The Effects of Calcium Site Occupancy and Reagent Length on Reactivity of Calmodulin Lysyl Residues with Heterobifunctional Aryl Azides MAPPING INTERACTION DOMAINS WITH SPECIFIC CALMODULIN PHOTOPROBE DERIVATIVES*

The relationship of structural and functional moie-ties on calmodulin is important in all venues of cell activity. In this study, we investigate the effect of lysine modification on calmodulin function. Azidosalicylate reagents containing different “linker arm” lengths, between the photoactive terminus and an amine-reactive N-hydroxysuccinimidyl ester moiety were used to modify calmodulin lysines at three different positions in a calcium-dependent manner. The short cross-linker, (ASNE-2 (where ASNE represents azidosalicylate N-hydroxysuccinimidyl ester), modifies Lys-75, whereas the longer reagent, ASNE-6, modifies lysines 21, 75, and 94. The modification of these different lysines is shown to be calcium-depend-ent. At 1-100 PM levels of calcium, only Lys-94 is modified, suggesting that modification of this residue is directed by both the binding of calcium to calcium- binding loops I11 and IV and the hydrophobic pocket exposed between these two loops as a result of calcium binding. At higher calcium concentrations (>200 p ~ ) , where sites I and I1 become filled, modification of Lys- 21 or Lys-75 also was observed. All the modified cal-modulins were able to stimulate 3’,5’-cyclic-nucleotide phosphodiesterase fully although the Kact for the Lys-75

three-dimensional x-ray crystal structure of the calcium-replete molecule indicates that CaM exists as a rigid protein with two globular lobes, each containing two calcium-binding sites, at either end of a long central helix (Babu et al., 1985(Babu et al., , 1988. However, substantial data now exist suggesting that this rigid structure does not represent the active conformation of Ca:'/CaM in solution. Persechini and Kretsinger (1988a) demonstrated that the two halves of a recombinant CaM that contain cysteinyl residues inserted at positions 3 and 146, 37 A apart in the crystal structure, could be cross-linked with the thiol-directed reagent bismaleimidohexane, which could span at most 19 A. Furthermore, this cross-linked derivative retained the ability to activate rabbit skeletal muscle myosin light chain kinase. O'Neil et al. (1989) demonstrated that a synthetic peptide corresponding to the calmodulin-binding domain of myosin light chain kinase but containing photoactive p-benzoylphenylalanine derivatives at either end would photolabel methionyl residues in the NH,-and COOH-terminal hydrophobic pockets of CaM, an observation that further supported the foided conformation of Cai'/CaM in solution. Persechini and Kretsinger (1988b) have proposed a model for this interaction with myosin light chain kinase based on extensive molecular modeling and use of the folded CaM structure that suggests that flexibility in the central helix of CaM, particularly in the backbone folding in the region around residue 80, allows for CaM to bind to and activate its target enzymes.
Lysine 75 modification and the functional consequences are particularly interesting because of the position of its side chain at the mouth of a hydrophobic "pocket" composed of both aliphatic side chains and a well-ordered set of phenylalanyl side chains. Previous studies indicated that hydrophobic interaction is required for modification at Lys-75 (Mann and Vanaman, 1988). NMR studies of the binding of pheno-thiazines to CaM indicated that the drugs may intercalate into the stacked aromatic side chains that line each pocket (Anderson et al., 1983;Dalgarno et al., 1984). Interestingly, differential effects on the degree of 3',5'-cyclic-nucleotide phosphodiesterase activation have been demonstrated for aliphatic versus aromatic adducts of Lys-75. The norchlorpromazine Lys-75 derivative of CaM is unable to activate 3',5'cyclic-nucleotide phosphodiesterase (Newton et al., 1983), whereas the cyclohexyl adduct still gives maximal activation, but shows decreased binding affinity (Mann and Vanaman, 1988).
Despite the numerous studies to date, the structural and steric considerations that direct the modification of specific lysyl residues in CaM and the relationship of Ca2+ site occupancy to lysine reactivity remain to be defined. This study, using a set of related aryl azide N-hydroxysuccinimidyl ester compounds of different lengths, demonstrates that both factors are important in dictating lysyl reactivity, which can be accounted for based on the known three-dimensional crystal structure of CaM.
The resulting photoactive CaM derivatives are used to further examine the interaction of CaM with two target enzymes, 3',5'-cyclic-nucleotide phosphodiesterase and erythrocyte membrane Ca2'-ATPase. Photocross-linking studies with the latter enzyme also demonstrate that the length and site of attachment of the photoprobe as well as its photochemical properties are important in determining the efficiency with which such CaM derivatives cross-link to target proteins.

Materials
Calmodulin was isolated from bovine testes according to the procedure of Jamieson and Vanaman (1979). Activator-deficient bovine brain 3',5'-cyclic-nucleotide phosphodiesterase was purchased from Sigma. Erythrocyte plasma membrane Ca*+-ATPase was isolated from fresh pig blood according to the method of Niggli et al. (1987). ~-l-Tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin was purchased from Worthington. HPLC-grade acetonitrile was purchased from Burdick & Jackson Laboratories Inc. Aqueous buffers were prepared with Milli-Q water and filtered through 0.22-pm Millipore filters before use. Reverse-phase C, columns and C4 bulk resin were purchased from Altex, and phenyl pBondapak columns were from Waters. C4 columns (4 X 250 mm) were packed at 5000 p.s.i. in 100% Fisher HPLC-grade methanol using a Shandon HPLC column packer. Mono Q columns were purchased from Pharmacia LKB Biotechnology Inc. The ASNE reagents (shown in Fig. 1) were synthesized and characterized as described by Imai et al. (1990). Enzy-0 n=2 n=4 n=6 FIG. 1. Chemical structure of ASNE reagents. The n refers to the number of methylene groups in the linker arm region of the reagent (ie. succinimidyl N-[2-[(4-azidosalicyloyl)oxy]ethyl]subermate ( n = 6) is represented by the abbreviation ASNE-6). mobeads were purchased from Bio-Rad. Carrier-free Na'"1 (17 Ci/ mg iodine) was purchased from ICN. All other chemicals were purchased from Sigma.

Methods'
Reaction of ASNE Compounds with CaM-Immediately before use, the ASNE reagents were dissolved in acetonitrile at a concentration of 1 mg/ml. For most studies, 2 (ASNE-6) or 5 (ASNE-2) molar eq of reagent were added to reaction mixtures containing 30 mM HEPES (pH 7.4). 2 mM CaCL (or 2 mM EGTA), and 10 p~ CaM. The conditions were altered to prepare large quantities of ASNE-6 monoadducts: those mixtures contained 50 p~ ASNE-6 (5 molar eq) plus 1.38 mM EGTA and 1.36 mM CaC1' to obtain a free Ca*+ concentration of 2.1 X 1O"j M (based on studies of the Ca'+ dependence of modification of specific lysyl residues described under "Results"). Reactions were allowed to proceed at 25 "C for 2 h, at which time they were quenched by the addition of lysine to a final concentration of 10 mM. The extent of reaction was monitored using reverse-phase HPLC as described by Mann and Vanaman (1988), which separates unmodified CaM from CaM derivatives because of the addition of the hydrophobic aryl azide group. The ASNE reagents exhibit a UV absorbance maximum at 310 nm due to the aryl azide, and the ratios of 310:230 nm absorbances reflect the extent of modification of CaM derivatives.
Assessment of Calcium Dependence of Modification of Specific Lysyl Residues by ASNE-6 Reagent-Reactions were performed in 30 mM HEPES (pH 7.4) and 2 mM EGTA with increasing CaC12 concentrations, fixed using CaClJEGTA buffers as described by Goldstein (1979), in a total volume of 250 p1. The entire reaction was subjected to reverse-phase C4 HPLC. The native CaM and adduct peaks were integrated, and the percent of total CaM present in each peak was calculated based on the sum of the areas of the native, monoadduct, and polyadduct peaks detected at 230 nm.
Purification of Monoadducts-CaM-ASNE monoadducts were isolated on a large scale using a combination of fast protein liquid chromatography and HPLC techniques (see Miniprint). Large-scale reaction mixtures containing 6 mg of CaM were resolved on an anionexchange Mono Q fast protein liquid chromatography column. Further analysis by reverse-phase Ct (ASNE-2) and C4 (ASNE-6) HPLC showed that the monoadducts of the ASNE-2 reaction and the low calcium ASNE-6 reaction could be purified by application to the Mono Q column. One of the monoadducts from the high calcium ASNE-6 reaction was pure after Mono Q chromatography. The other monoadduct required further purification on C4 HPLC with multiple runs. These purified monoadducts were desalted and characterized.
Determination of Modification Sites-Purified CaM-ASNE monoadducts were subjected to trypsin digestion as described by Vanaman (1983). Modified CaM (100-500 pg) in 100 mM (pH 8.0) was digested with L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin (1:20 trypsin/CaM) for 2 h (twice) at 37 "C. The resulting mixture of peptides was applied to a phenyl pBondapak reverse-phase HPLC column for peptide isolation (Vanaman, 1983). The peptides were separated at 25 "C using a l%/min gradient of Buffer A (10 mM NaH,PO4 (pH 6.5), 2 mM EGTA) and Buffer B (5 mM NaH2P04 (pH 6.5), 1 mM EGTA, 50% acetonitrile) from 0 to 50% Buffer B. Modified peptides were identified on the basis of UV absorbance at 310 nm. Fractions containing the modified peptides were desalted over the same column by elution with a gradient (l%/min) formed with 0.1% (v/v) trifluoroacetic acid and increasing amounts (0-50%) of 88% (v/v) acetonitrile, 0.05% (v/v) trifluoroacetic acid. In each case, a single peak absorbing at both 230 and 310 nm was obtained. This peptide-containing fraction was subjected to automated sequence analysis on an Applied Biosystems Model 477A Peptide Sequencer at the University of Kentucky Macromolecular Structure Analysis Facility.
Functional Characterization of CaM-ASNE Monoadducts-CaM-ASNE monoadducts were tested for their ability to activate bovine brain 3',5'-cyclic-nucleotide phosphodiesterase and the porcine erythrocyte plasma membrane CaZ+-ATPase. The 3',5'-cyclic-nucleotide phosphodiesterase assays were performed as described by Wallace et al. (1983), except that CAMP hydrolysis was measured by release of free phosphate (Lanzetta et al., 1979) from 5'-AMP. Ca*+-ATPase assays were done according to the procedure of Niggli et al. (1987), with released phosphate also detected using Lanzetta reagent (Lanzetta et al., 1979). Concentrations of stock solutions of CaM and CaM-ASNE monoadducts were determined by amino acid analysis following acid hydrolysis as previously described (Vanaman, 1983).
Cross-linking CaM-ASNE Monoadducts to ATPase-CaM-ASNE monoadducts were radioiodinated by the Enzymobead method according to the procedure supplied by Bio-Rad. One hundred pCi of '''1 (17 Ci/mg iodine) were used for each iodination. Iodinated monoadducts were desalted before use. Specific activities of each of these monoadducts were determined to be as follows: Lys-21 CaM, 1.1 mCi/ mmol; Lys-75 CaM, 1.65 mCi/mmol; and Lys-94 CaM, 1.2 mCi/ mmol. Cross-linking was performed as described by Imai et al. (1990), except that purified monoadducts were used. Each reaction mixture contained 1.5 p~ Ca2+-ATPase, 6 p~ radiolabeled CaM-ASNE monoadduct, 30 mM HEPES (pH 7.4), 130 mM NaCl, 2 mM MgCl', 0.05% (v/v) Triton X-100, 5% (v/v) glycerol, and 0.5 mg/ml phosphatidylcholine. Cross-linking reactions performed in the presence of calcium contained 100 p~ CaC12. Reactions without calcium contained 10 mM EGTA. Native CaM competition reactions contained 12 p~ (2-fold molar excess) unmodified CaM. Photolysis was done using a handheld ultraviolet light (4600 milliwatts/cm2, Mineralite model UVS-11) with the glass face removed. Reaction mixtures were photolyzed on ice at a distance of 4 cm for 1 min, suspended in an equal volume of protein-solubilizing buffer (65 mM Tris-HC1 (pH 8.8), 2% (w/v) SDS, 10% (v/v) glycerol, 5% (v/v) 2-mercaptoethanol, and 0.001% (w/v) bromphenol blue), and boiled for 5 min. Reaction products were resolved on 7.5% SDS-polyacrylamide gel (Laemmli, 1970). The gel was dried, autoradiographed for 16 h, rehydrated, and stained with silver (Morrissey, 1981). Fig. 2 shows reverse-phase CB HPLC analysis of the products obtained when CaM was reacted with ASNE-2 and ASNE-6 in the presence (upper panels) or absence (lower panels) of calcium. Reactions were carried out at pH 7.4 for 2 h as described under "Methods." The solid line shows 230 nm absorbance, and the dashed line shows 310 nm absorbance. The asterisk in each trace indicates the elution position of unmodified CaM, determined from a prior analysis and detected as a peak at 230 nm with little 310 nm absorbance. Products eluting prior to 10 min were shown to be hydrolyzed ASNE or adducts of ASNE with free lysine used to quench excess reagent (data not shown). The products eluting later than unmodified CaM and having both 230 and 310 nm absorbance were CaM-ASNE adducts. The major products (indicated by arrows in upper panels were determined to be monoadducts by virtue of the ratio of the UV absorbance at 310 versus 230 nm and by subsequent sequence analysis (see below). The other products eluting after the monoadducts were determined to be polyadducts on the basis of their 310:230 nm absorbance ratios and peptide analysis on purified derivatives (data not shown).

Modification of CaM by ASNE Reagents-
In the absence of calcium (Fig. 2, lowerpanels), <5% of the calmodulin was modified by either reagent. As shown in Fig.  2 (upper left panel), reaction of CaM with ASNE-2 (5 molar eq) in the presence of 2 mM CaClz yielded one major new peak of material on HPLC analysis. Further characterization (see below) showed this peak to contain a single CaM monoadduct modified at Lys-75. Reaction with ASNE-6 (2 molar eq) under similar conditions yielded two major peaks on HPLC analysis (Fig. 2, upper right panel). Further separation on C4 columns and subsequent analyses showed that three distinct monoadducts of CaM were present in these two peaks. Fig. 3 shows a typical separation of CaM and the three CaM-ASNE-6 monoadducts using reverse-phase C4 HPLC. Separation of these monoadducts was determined to be optimal at 18 "C (data not shown), the temperature used in the separation shown here. The peak eluting at 13.1 min (indicated by arrow) showed little 310 nm absorbance and was unmodified CaM. The peaks eluting at 14.2, 14.6, and 15.2 min, respectively, were shown to be unique CaM-ASNE-6 monoadducts modified at Lys-21, Lys-75, or Lys-94 as discussed below. For ease of discussion, these monoadducts will be referred to as 1, 2, or 3 based on their order of elution from C4 columns. The relative abundance of these three monoadducts was dependent on the concentration of calcium present during the reaction as discussed below.
Calcium Dependence of Modification by ASNE-6"Previous analysis of the reaction of CaM and ASNE-6 showed the production of three major monoadducts (see Fig. 3). The generation of these monoadducts was shown to be dependent on the calcium concentration in the reaction mixture as illustrated in Fig. 4. CaM modification reactions were performed as described before, except that the calcium concentration was varied over a range of lo-' to M by the use of a calcium/EGTA buffering system (Goldstein, 1979). The reactions were performed in triplicate and were analyzed by C4 HPLC at 18 "C. Previous analysis had determined the elution position of unmodified CaM as well as each of the monoadducts and diadducts. Using the peak areas obtained FIG. 2. Reverse-phase Cs HPLC separation of reaction products of calmodulin and reagents ASNE-2 and ASNE-6. The separation was achieved using a l%/min gradient of Buffer B (100% acetonitrile) and Buffer A (10 mM NaH2P0, (pH 6.0), 2 mM EGTA, and 5% acetonitrile) from 20 to 40% Buffer B. The arrows indicate the elution positions of the CaM monoadducts, whereas the asterisks denote unmodified CaM. The left a i s (-) is the UV absorbance scale at 230 nm, whereas the right axis (---) is the UV absorbance scale at 310 nm. The presence of calcium (+Ca+') or EGTA (-Ca+') in the reaction is indicated in the upper right corner of each panel. . Separation of CaM and three CaM-ASNE-6 monoadducts using reverse-phase C4 HPLC. Separation was achieved using the same buffers and conditions as in the CB separations (see Fig. 2), except that the column was cooled to 18 "C. As before, the left axis (-) indicates the UV absorbance scale at 230 nm, whereas the right axis (---) indicates the UV absorbance scale a t 310 nm. The arrow indicates the elution position of unmodified CaM, whereas the numbers above the peaks denote the three CaM-ASNE-6 monoadducts.  Lys-75 (peak 3), and Lys-94 (peak 2) modifications, respectively. Free calcium concentration is depicted by pCa (-log of the free calcium). Points were determined as the average of the analysis of three separate reactions. Error bars show standard deviation. by integration, the amount of total CaM recovered was calculated (values were within 10% in each of three separate analyses). The individual peak areas were used to determine the fraction of total CaM that each monoadduct comprised. These values are plotted as a function of pCa in Fig. 4. The disappearance of native CaM is indicated (0). At a pCa value of 8.2, monoadduct 2 began to appear (+). This was the only monoadduct formed until the pCa reached 4.2, where monoadducts 1 and 3 began to appear (0 and 0, respectively). Diadducts also were formed efficiently beginning at a pCa of 4.2, which accounts for the parallel decline in the amount of both monoadduct 2 and unmodified CaM at the higher Ca2+ concentrations.
Determination of Sites of Modification-Using the calcium concentrations indicated in the previous experiment, largescale preparations of each CaM-ASNE-6 monoadduct as well as of the CaM-ASNE-2 monoadduct were prepared as described in the Miniprint. The pure monoadducts were subjected to trypsin digestion, and the resulting peptides were purified by phenyl pBondapak HPLC monitored at 230 and 310 nm. The resulting peptide maps are shown in Fig. 5 . For ease of presentation, only the 310 nm traces are shown for the modified CaMs (Fig. 5, C-F). The 230 and 310 nm traces obtained with the trypsin digest of the unmodified protein are shown in Fig. 5 ( A and B , respectively). The 230 nm traces obtained with digests of the monoadducts looked similar to Fig. 5A, except that a new peak having both 230 and 310 nm absorbance was present in each case. Unmodified CaM peptides showed very little 310 nm absorbance (Fig. 5 B ) . Fig. 5 (C-E) shows the HPLC analyses of the trypsin digests of the CaM-ASNE monoadducts using detection at 310 nm. The ASNE-2 monoadduct digest is shown in Fig. 5C, where one major 310 nm absorbing peak is seen eluting at 37.2 min. The ASNE-6 monoadducts are shown in Fig. 5 (D-F). One major 310 nm absorbing peak was seen for each monoadduct analyzed.
The fraction from each separation containing the 310 nm by guest on March 24, 2020 http://www.jbc.org/ Downloaded from absorbing material was pooled and reapplied to the phenyl pBondapak column to remove salt. Only one peak was seen for each fraction with the acidic buffer system used for elution during this desalting step (data not shown). The peak fraction was collected and submitted to sequence analysis as described under "Methods." The sequences were identified as to their position in the known CaM sequence (Watterson et al., 1980) as shown in Table I. In each case, a cycle corresponding to a lysyl residue showed no identifiable phenylthiohydantoinderivative on HPLC analysis, but did produce an unidentified peak. In the case of the ASNE-2 monoadduct, this peak occurred at 19.5 min, whereas in the sequence of the ASNE-6-containing peptides, the peak appeared at -27 min (see Miniprint). The CaM-ASNE-2 monoadduct is modified at Lys-75, as would be expected from previous modification studies (Mann and Vanaman, 1988). CaM-ASNE-6 monoadduct 3 is also modified at Lys-75. CaM-ASNE-6 monoadduct 1 is modified at Lys-21, whereas the modification that takes place at subsaturating calcium concentrations occurs at Lys-94, yielding monoadduct 2. Functional Studies Using CaM-ASNE Monoadducts-Since the modifications were in different regions of the CaM molecule, functional studies with these adducts were undertaken to determine whether modification altered binding to and activation of two enzymes that appear to interact with different regions of the CaM molecule, bovine brain 3',5'-cyclicnucleotide phosphodiesterase and porcine erythrocyte membrane Ca2+-ATPase. Figs. 6 and 7 show the results of assays using native CaM and the characterized CaM-ASNE monoadducts. The CaM-ASNE monoadducts activated the Ca2+-ATPase in a manner indistinguishable from native CaM (see Fig. 6). Previous studies with the CaM-ASNE-2 monoadduct using the human erythrocyte Ca2+-ATPase showed that the modification had no effect on its ability to activate the enzyme fully (Imai et al., 1990). In the case of bovine brain 3',5'cyclic-nucleotide phosphodiesterase, modifications at Lys-75 and Lys-21 affected the ability of CaM to activate the enzyme. An 8-10-fold increase in the concentration of both the CaM-ASNE-2 (0) and CaM-ASNE-6 (0) Lys-75 monoadducts was required to stimulate 3',5'-cyclic-nucleotide phosphodiesterase fully (Fig. 7). The ASNE-6 Lys-21 monoadduct (V) required a 50-fold higher concentration to stimulate the enzyme fully. Modification at Lys-94 (+) did not affect 3',5'-cyclicnucleotide phosphodiesterase activation.

Cross-linking Studies with CaM-ASNE Monoadducts and
ATPase-We have previously reported (Imai et al., 1990) that the CaM-ASNE-2 Lys-75 monoadduct gives no detectable cross-linking to the Ca2+-ATPase, but a mixture of the CaM-ASNE-6 monoadducts shows -8% cross-linking to this enzyme. Here we report that cross-linking with the separated CaM-ASNE-6 Lys-21, Lys-75, and Lys-94 monoadducts occurs with substantially different efficiencies. The monoadducts were purified, radiolabeled with lZ5I, and subjected to cross-linking. Fig. 8 shows SDS-polyacrylamide gel electrophoresis analysis of products of photocross-linking with each of the CaM-ASNE-6 monoadducts and the purified erythrocyte Ca2'-ATPase. The upper panel shows the silver-stained 7.5% polyacrylamide gel, whereas the lower panel shows the autoradiograph of that gel prior to silver staining. Lane I shows the untreated enzyme. Lanes 2-5 are CaM-ASNE-6 Lys-21 cross-linking, lanes 6-9 are CaM-ASNE-6 Lys-75 cross-linking, and lunes 10-13 are CaM-ASNE-6 Lys-94 crosslinking. Samples in lunes 2, 6, and 10 were photolyzed in the presence of calcium. Lanes 3, 7, and 11 contained reaction mixtures that were not irradiated. Lanes 4,8, and 12 contained samples that were photolyzed in the absence of calcium. Samples in lunes 5, 9, and 13 were photolyzed in the presence of 2 eq of unlabeled unmodified CaM. The porcine erythrocyte Ca2+-ATPase runs as a doublet at 121 and 138 kDa. Cross-linked species occurred at 153 kDa. Efficient cross-linking was seen only in the case of the Lys-75 derivative. Densitometric scanning of the silver-stained gel indicated -40% cross-linking efficiency with the Lys-75 derivative (Fig. 8, lane 6). The cross-linking was calcium-dependent (lune 8 ) and efficiently inhibited (80%) by a 2-fold excess of unlabeled CaM (lane 9). Limited cross-linking (-5%) also was observed with the Lys-94 derivative (lune 10). The Lys-21 derivative showed only a trace amount of crosslinking (lune 2 ) . The differences in apparent cross-linking efficiency cannot be due to differences in the specific radioactivity of the different CaM-ASNE-6 derivatives as the extent of iodination was comparable in all cases (see "Methods"). In addition, the percent of cross-linking was estimated by densitometry of the silver-stained gel bands, not the autoradiograph.

DISCUSSION
The preceding sections describe the use of heterobifunctional cross-linking reagents to study the structure of CaM and its mechanism of target activation. These reagents mod-   ified calmodulin in a calcium-dependent manner. Characterization of the resulting modified CaM monoadducts yielded important structural information. The ASNE-2 reagent preferentially modified Lys-75 in the presence of calcium, whereas ASNE-6 modified Lys-75 as well as Lys-21 and Lys-94. Lys-21 was described as extremely unreactive by Giedroc et al. (1985) in trace labeling studies with acetic anhydride, whereas Lys-94 was the next most reactive lysine after Lys-75. Other studies have shown that Lys-94 is preferentially modified by hydrophilic reagents (Mann and Vanaman, 1989). The Ca2+dependent modification of these 3 residues by ASNE-6 prompted further study into the factors directing lysine modification in CaM.

Lys-Met-Lys
Previous studies indicated that hydrophobic and/or aromatic compounds interact with the hydrophobic pockets in the amino-and carboxyl-terminal globular domains of calmodulin (LaPorte et al., 1980;Weiss et al., 1982;Newton et al., 1985;Mann and Vanaman, 1988). T o determine if the ASNE reagents were also interacting with the hydrophobic pockets, competition studies were performed with the calmodulin antagonist calmidazolium. One eq of calmidazolium was  6 7 8 9 1 0 1 1 1 2 1 3   FIG. 8. Silver-stained gel (upper panel) and autoradiogram (lowerpanel) showing products of cross-linking between porcine erythrocyte Ca"+-ATPase and purified CaM-ASNE-6 monoadducts. The positions of molecular weight markers are shown to the left of the gel and autoradiograph. All lanes contained purified Ca"-ATPase. The variables in reaction mixtures are depicted below the autoradiograph. CaM 21, CaM 75, and CaM 94 stand for the radiolabeled CaM-ASNE-6 Lys-21 monoadduct, CaM-ASNE-6 Lys-75 monoadduct, and CaM-ASNE-6 Lys-94 monoadduct, respectively. Ca2+ means that 100 j t~ CaC12 was present in the reaction. CaM indicates that 2 eq of unmodified CaM were added to the reaction mixture prior to photolysis. hu indicates whether or not the reaction was irradiated.
shown to inhibit the reaction of CaM with 5 molar eq of ASNE-2 (Mann, 1987). The reaction of ASNE-6 and CaM was also inhibited by calmidazolium, but the concentration of calmidazolium required to completely inhibit the reaction exceeded the solubility of calmidazolium in an aqueous solution (data not shown). In addition, this calmidazolium competition appeared to be biphasic, indicative of two binding events. The amino-terminal modifications were abolished at calmidazolium concentrations comparable to those seen with ASNE-2. The Lys-94 modification was not affected until higher calmidazolium concentrations were reached, which suggested that the modification of these 3 lysines was controlled by two separate phenomena.
The calcium dependence of the reaction was determined to investigate further the reactions of ASNE-6 with CaM. At subsaturating calcium concentrations, only the CaM-ASNE-6 Lys-94 monoadduct was formed. At saturating calcium concentrations, Lys-21 and Lys-75 were also modified. This calcium dependence provides further evidence that the carboxyl-and amino-terminal pairs of Cat+-binding sites have high and low Cat+ affinities, respectively, as determined by direct ligand binding (Watterson et al., 1976;Crouch and Klee, 1980;Klee, 1988), NMR (Klevit, 1983;Linse et al., 1991), and specific Ca"'-site mutagenesis (Beckingham, 1991). These results also provide definitive evidence for separate conformational changes accompanying Ca" binding in the by guest on March 24, 2020 http://www.jbc.org/ Downloaded from carboxyl-and amino-terminal globular domains, resulting in the sequential exposure of each respective hydrophobic pocket.
As part of this study, we examined the diadducts formed in the reaction of ASNE-6 with CaM that were favored in the presence of increased ASNE concentrations. Preliminary examination of diadducts formed in the reactions of CaM with ASNE-6 in the presence first of low, then high calcium concentrations yielded derivatives modified either at . No diadduct modified at both Lys-75 and Lys-21 was detected, indicating that the reaction of the reagents with CaM was specific for each hydrophobic pocket. Information on reaction rates at each of these lysines has yet to be determined at any calcium concentration, but 2.5-fold less (on a molar basis) ASNE-6 reagent is required to modify 50% of the CaM compared with ASNE-2 in the presence of saturating calcium. Whether this is due to the availability of multiple modification sites for ASNE-6 or an increased binding affinity of the reagent has yet to be determined.
Molecular modeling was employed to address further the specificity of lysyl modification. In the fully !taggered conformation, the ASNE-2 adduct measures 16.1 A from the Ca of lysine to the termin+ azide nitrogen, whereas that with ASNE-6 measures 21 A. This difference in length appears to be the key factor that determines which CaM lysyl residues are modified by the two reagents. Fig. 9 shows the 2.2-A crystal structure of calmodulin (Babu et al., 1988). Lys-21 resides at the "top" of calcium-binding loop I in Fig. 9. Lys-75 sits at the "base" of the amino-terminal hydrophobic pocket. Lys-94 in calcium-binding loop I11 is in an analogous position to Lys-21 in calcium-binding loop 1.
The distances between the lysyl residues and the hydrophobic pockets were measured using the solvent surface depiction of the crystal structure of calmodulin. As shown in Fig. 10  (upper panel), the amino-terminal hydrophobic pocket is a crescent-shaped cleft defined by the stacked aromatic rings of the phenylalanyl residues. Measurements made from the camino group of Lys-21 over the lip of the hydrophobic pocket and into its center was -18 A. This would allow the ASNE-6 (but not the ASNE-2) reagent to reach from the pocket to the Lys-21 t-amino group. Lys-75, on the other hand, is positioned at the base of the pocket and is accessible to all the reagents.
The carboxyl-terminal hydrophobic pocket (Fig. 10, lower panel) is cup-shaped and lined with the aromatic rings of the tyrosyl and phenylalanyl residues. Measurements indicated that the distance from Lys;94 c-NH, to the center of the aromatic cluster is -18-19 A. Once again, only the ASNE-6 reagent is long enough to span the distance from this hydrophobic pocket to Lys-94. The importance of these distance  relationships is highlightet by the fact that the ASNE-4 reagent, which is only 2.5 A shorter than ASNE-6, modifies primarily Lys-75 in a Ca2+-dependent manner (Mann, 1987).
The availability of photoactive derivatives of CaM modified at specific sites in the molecule with reagents of different lengths provides substantial opportunity for studying CaMtarget protein interaction. The results of photocross-linking studies with erythrocyte membrane Ca2+-ATPase presented in Fig. 8 show that the ASNE-6 Lys-75 monoadduct is capable of efficient (40%) cross-linking, whereas the corresponding CaM-ASNE-2 derivative gives little cross-linking despite the fact that both Lys-75 monoadducts activate Ca2'-ATPase activity with indistinguishable dose-response curves. The actual cross-linking efficiency with CaM-ASNE-6 actually may be higher than 40% as the Ca2+-ATPase used in the crosslinking reaction is present in micelles, presumably in two orientations, with the CaM-binding region either exposed to the buffer or on the inside of the micelle.
The ASNE-6 Lys-21 and Lys-94 adducts gave very little Ca2+-dependent cross-linking to the Ca2'-ATPase. In preliminary studies (Imai et al., 1990), a mixture of the CaM-ASNE-6 monoadducts were used in cross-linking experiments, yielding cross-linking efficiency of only 8% (Imai et al., 1990). The results presented in the present work clearly demonstrate the need to work with purified and characterized modified derivatives if the cross-linking efficiencies are to be assessed.
The nature of the reactive species generated on photolysis of the aryl azide is also extremely important in governing cross-linking efficiencies between donor and target structures. Previous studies (Crocker et al., 1990) using the CaM Lys-75 monoadduct of succinimidyl 2-(4-azido-2,3,5,6-tetrafluorophenyl)thiazole-4-carboxylate demonstrated that this perfluorinated aryl azide derivative photocross-linked to the membrane Ca2+-ATPase with an efficiency approaching that of the ASNE-6 Lys-75 derivative despite the fact t h!t it is the shortest of the reagents thus far employed (9.8 A from lysine Ca to the azido group). Whereas conformational and steric considerations cannot be excluded, one logical explanation for these apparently disparate results is the nature of the intermediates formed on photolysis of these reagents.
The nitrene generated on photolysis of aryl azides has been shown to generate a ring-expanded dehydroazepine as the putative cross-linking species (Torres et al., 1986;Leyva et al., 1986;Shields et al., 1987;Schuster and Liang, 1987). The dehydroazepine is an electrophilic alkylating reagent that does not have the ability to insert into C-H and C-C bonds as does a highly reactive singlet nitrene. The perfluorinated aryl azide was selected specifically to avoid ring expansion and thereby provide access to a reactive singlet nitrene (Crocker et al., 1990). It is interesting to speculate that the shorter perfluorinated aryl azide might photoinsert into hydrophobic side chains in the CaM-ATPase molecule, in close proximity to the Lys-75 residue of CaM in the complex, whereas the dehydroazepine generated on photolysis of the ASNE compounds must react with nucleophilic side chains at a distance spanned only by the ASNE-6 derivative. In this regard, James et al. (1988) have identified a putative CaM-binding domain in the human erythrocyte membrane Ca2'-ATPase by photocross-linking using CaM modified with the Denny-Jaffe reagent (Jaffe et al., 1980). The lysine adduct of this Teagent would be expected to span a maximum distance of 27 A based on computer modeling. The amino acid sequence of the region of the Ca2+-ATPase labeled on photolysis of the complex with this modified CaM derivative contained a number of potential nucleophiles for alkylation by the expected dehydroazepine intermediate formed on photolysis. Unfortunately, the authors did not characterize the lysyl residue(s) in CaM at which the Denny-Jaffe reagent was incorporated. Clearly, further work is required to fully exploit the use of these various reagents in delineating the interaction of calmodulin with the plasma membrane Ca*+-ATPase and other CaM target enzymes.
The ASNE-CaM monoadducts also provided information on the mechanism of activation of 3',5'-cyclic-nucleotide phosphodiesterase. Previous studies had shown that modifications at Lys-75 reduced the efficiency of CaM activation of the enzyme (Newton et al., 1985;Mann and Vanaman, 1988). Modification at Lys-21 also reduced the efficiency of activation of the enzyme. This suggests that Ca2+-binding loop I is important in binding and/or activation of 3',5'-cyclic-nucleotide phosphodiesterase. Previous studies have indicated that modification of Cys-27 (Mann and Vanaman, 1989) in plant CaM or mutations in Ca2+-binding loop I (George et al., 1990) did not affect 3',5'-cyclic-nucleotide phosphodiesterase activation, suggesting that modification at Lys-21 by ASNE-6 does not affect enzyme binding and activation simply because of its position in the first domain. It is possible that the effect results from the presence of the aryl azide portion of the reagent in the amino-terminal hydrophobic pocket. However, a n alternative explanation arises from the observation that modification at Lys-21 may affect calcium binding in that loop. The CaM-ASNE-6 Lys-21 monoadduct is more susceptible to proteolysis than is either unmodified CaM or the other monoadducts (data not shown). This suggests that the Lys-21 monoadduct may not be able to bind calcium as effectively as native CaM. Mackall and Klee (1991) have reported that CaE+/CaM is more susceptible to proteolysis than either apoCaM or Ca:'/CaM. Preliminary experiments showed that addition of 10 mM calcium improves the ability of CaM Lys-21 to activate the 3',5'-cyclic-nucleotide phosphodiesterase (data not shown).
These studies illustrate, once again, that CaM may interact with different target enzymes in different ways. The model of Persechini and Kretsinger (1988b) involving a flexible central helix that wraps around target-binding sites is consistent with present knowledge. The flexibility of the central helix allows calmodulin to contact diverse target CaM-binding sites and to bind to them using contact regions most suited to each site. Each enzyme may have subtly different requirements for binding and activation. Studies of the type presented here can yield important information concerning CaM-binding site structures and specific information about the residues in the target enzymes that lie in close proximity to specifically modified CaM residues.