The Calmodulin-binding Domain of the Catalytic γ Subunit of Phosphorylase Kinase Interacts with Its Inhibitory α Subunit

Chemical cross-linking as a probe of conformation has consistently shown that activators, including Ca2+ ions, of the (αβγδ)4phosphorylase kinase holoenzyme (PhK) alter the interactions between its regulatory α and catalytic γ subunits. The γ subunit is also known to interact with the δ subunit, an endogenous molecule of calmodulin that mediates the activation of PhK by Ca2+ions. In this study, we have used two-hybrid screening and chemical cross-linking to dissect the regulatory quaternary interactions involving these subunits. The yeast two-hybrid system indicated that regions near the C termini of the γ (residues 343–386) and α (residues 1060–1237) subunits interact. The association of this region of α with γ was corroborated by the isolation of a cross-linked fragment of α containing residues 1015–1237 from an α−γ dimer that had been formed within the PhK holoenzyme by formaldehyde, a nearly zero-length cross-linker. Because the region of γ that we found to interact with α has previously been shown to contain a high affinity binding site for calmodulin (Dasgupta, M., Honeycutt, T., and Blumenthal, D. K. (1989) J. Biol. Chem. 264, 17156–17163), we tested the influence of Ca2+ on the conformation of the α subunit and found that the region of α that interacts with γ was, in fact, perturbed by Ca2+. The results herein support the existence of a Ca2+-sensitive communication network among the δ, γ, and α subunits, with the regulatory domain of γ being the primary mediator. The similarity of such a Ca2+-dependent network to the interactions among troponin C, troponin I, and actin is discussed in light of the known structural and functional similarities between troponin I and the γ subunit of PhK.

Phosphorylase kinase (PhK) 1 , a Ca 2ϩ -dependent enzyme involved in the regulation of glycogenolysis, is among the largest and most complex enzymes known. Structurally, PhK is composed of four copies each of four different subunits, (␣␤␥␦) 4 and has a mass of 1.3 ϫ 10 6 Da (for reviews see Refs. [1][2][3]. Of the four subunits, ␥ is catalytic, whereas the remaining three are regulatory: ␣ and ␤ exert quaternary constraint on the activity of ␥, and ␦ is an intrinsic molecule of calmodulin (CaM). To fully understand how PhK integrates diverse physiological signals to regulate glycogenolytic flux in skeletal muscle, it is first essential to understand how intrasubunit and intersubunit interactions within the hexadecameric holoenzyme change in response to effector ligands, and in so doing, control its catalytic activity. Despite the increased availability of structural information regarding PhK, interactions associated with activation and involving specific regions of individual subunits, in particular the ␣ and ␤ subunits, have largely remained uncharacterized. In this study, we have focused on delineating interacting regions between the large ␣ and catalytic ␥ subunits to advance our understanding of how structural perturbations correlate with activation of this complex holoenzyme. By using chemical cross-linkers as structural probes, alterations in the interactions between the regulatory ␣ and catalytic ␥ subunits of PhK consistently emerge as a common structural marker of enzyme activation by multiple effectors, including Ca 2ϩ ions (4 -7). Although there have been many studies on the activation of PhK by Ca 2ϩ , it has not been clear how the binding of Ca 2ϩ to the ␦ subunit (CaM) relays structural information to the remainder of the holoenzyme, especially to the ␣ subunit. It has been shown that Ca 2ϩ increases the accessibility of specific regions of the ␥ subunit (5), and the C-terminal region of the ␥ subunit has been shown by a variety of experimental approaches to contain the regulatory domain that binds ␦ (8 -11), thus conferring Ca 2ϩ -sensitivity to the holoenzyme. In fact, truncation of ␥ to eliminate this regulatory domain renders it constitutively active by itself and Ca 2ϩ /CaMindependent (9,11). Our findings reported herein suggest that the flow of structural information from ␦ to ␣ in the holoenzyme is directly mediated by the C-terminal regulatory domain of ␥. These are the first results to define a specific region of ␥ that interacts with any region of either the ␣ or ␤ subunits of PhK. Furthermore, the finding that the region of ␣-␥ interaction includes a portion of the CaM binding domain of ␥ provides a possible explanation for the previously observed changes in ␣-␥ interactions induced by Ca 2ϩ .
Proteins and Enzymes-Nonactivated PhK was purified from fasttwitch skeletal muscle of New Zealand White rabbits (15), dialyzed against 50 mM Hepes (pH 6.8), 10% sucrose, and 0.2 mM EDTA and either used immediately or stored frozen at Ϫ80°C. The full-length ␥ subunit was purified from isolated holoenzyme by the method of Paudel and Carlson (16). Purified yeast and bovine brain CaM were generously provided by Drs. Trisha M. Davis (University of Washington) and Shengli Huang (University of Missouri Kansas City), respectively. Glycogen phosphorylase-b (P-b) was isolated from rabbit skeletal muscle as described (17), and residual AMP was removed with activated charcoal. The concentrations of the isolated ␥ subunit and the two CaM isoforms were determined by the Bio-Rad protein assay with BSA as standard; P-b and PhK concentrations were determined spectrophotometrically by their respective absorbance indices (18,19). Anti-PhK ␣, ␤ and ␥ subunit-specific mAbs were those previously described (20,21). Anti-CaM was purchased from Signal Transduction Laboratories; all other detection conjugates were from Southern Biotechnology.
Two-hybrid Plasmid Construction-All PhK ␣ constructs were engineered as previously described (22). Rabbit skeletal muscle PhK ␥ cDNA was kindly provided by Dr. Donald J. Graves (Iowa State University) and used as the template for the preparation of all ␥ constructs. The cDNA fragments were ligated either to pLexA (CLONTECH), a 2-HIS3 plasmid, to generate a fusion protein consisting of the DNA BD (amino acids 1-202) of LexA fused to PhK ␥ or to pB42AD (CLON-TECH), a 2-TRP1 plasmid, to produce a B42 AD protein fused to PhK ␥. Constructs ␥65C, ␥110C, and ␥FL were directionally engineered into the EcoRI-BamHI restriction sites of pLexA by subcloning from previously prepared constructs in plasmid pGAD424. Subsequently, these three constructs were linearized from pLexA by digestion with EcoRI and XhoI restriction enzymes and ligated to pB42AD. Constructs ␥150C, ␥205C, ␥300C, and ␥342C were generated by PCR using primers to yield cDNAs flanked 5Ј and 3Ј with EcoRI and BamHI restriction sites, respectively (sense-strand, 5Ј-TGAATTCACCCGCGACGCGGCA-3Ј; antisense strand, (␥150) 5Ј-ATGGATCCTTACAGGTCCCGATG-CAC-3Ј, (␥205) 5Ј-ATGGATCCTTAGCCTGGGTGGTTGTC-3Ј, (␥300) 5Ј-ATGGATCCTTAGGGGCTGAAGTGGC-3Ј, (␥342) 5Ј-ATGGATCCT-TACAGAGGTCGGAGGGC-3Ј) and ligated to the EcoRI and BamHI sites of pLexA. To prepare the corresponding pB42AD ␥ constructs, pLexA␥150C, pLexA␥205C, pLexA␥300C, and pLexA␥342C were digested with EcoRI and XhoI, and the corresponding linear ␥ fragments were purified and subcloned into pB42AD. Fragments of ␥ corresponding to its regulatory tail were generated by PCR using the following primers and ligated into both pLexA and pB42AD at EcoRI sites: sense strands, (␥301N) 5Ј-TGAATTCCGGGGGAAGTTCAAGGT-3Ј and (␥343N) 5Ј-TGAATTCCGCCGCCTCATCGACG-3Ј; antisense strand, 5Ј-TGAATTCTTAGTAGTCATCCTCAGCCAG-3Ј. The orientation and proper sequence of all LexA and B42 ␥ fusion proteins were verified by dideoxy sequencing and/or restriction mapping. ␣-␥ Interactions by Two-hybrid Screening-To screen for ␣-␥ interactions, two series of C-terminal deletion mutants of the ␣ and ␥ subunits were assayed for interactions as LexA and B42 fusion proteins in all possible binary combinations. Yeast strain EGY48, possessing the pSH18 -34 lacZ reporter plasmid, was transformed by a modified lithium acetate procedure as previously described (23), and transformants were grown at 30°C on synthetic medium lacking histidine, tryptophan, and uracil (SDϪHisϪTrpϪUra) for 3 days. Protein expression of all ␣ and ␥ constructs was verified by Western analysis, with minor modification, as previously described (24) using either a DNA BD crossreactive LexA polyclonal Ab (kindly provided by Dr. Erica Golemis, Fox Chase Cancer Center) or an AD cross-reactive hemagglutinin mAb (Roche Molecular Biochemicals). Positive associations between ␣ and ␥ constructs were monitored by transcriptional activation of the LEU2 gene by growth on defined media lacking leucine (Leu Ϫ ) and of the lacZ reporter gene by liquid ␤-galactosidase assays using o-nitrophenyl galactopyranoside as substrate (22,25).
Two-hybrid Library Screening-A rabbit skeletal muscle cDNA library (22) was used to screen for interactors of PhK␥ by using ␥300C and ␥FL constructs (described above) as bait. Yeast library transformants containing either of the DNA-BD bait plasmids and the AD cDNA library were generated as previously described (22). Transformation efficiencies of the library for both ␥300C and ␥FL were 10 4 colonyforming units/g DNA with 9.5 ϫ 10 6 and 5.4 ϫ 10 6 total library clones eventually being amplified, respectively. All transformants were pooled and stored at Ϫ80°C. Library screens were performed on SD/Gal/Raf (ϪHis, ϪTrp, ϪUra, ϪLeu) synthetic media in the presence of 80 g/ml 5-bromo-4-chloro-3-indoyl-␤-D-galactopyranoside in order to simulta-neously screen for galactose-dependent LacZ expressing cDNA-encoded proteins that interact with either of the two bait proteins. After colonies were grown at 30°C for 3-7 days, putative primary LacZϩ/Leuϩ colonies were restreaked onto SD ϪHis, ϪTrp, ϪUra and subjected to secondary and tertiary analyses as previously described (22). Specific interactor cDNAs were ultimately identified by dideoxy sequencing.
Enzymatic Assays of the ␥/CaM Complex-The activity of the renatured ␥/CaM complex was determined by following the incorporation of 32 P into P-b at pH 7.0 using the filter paper assay of Reimann et al. (28). Prior to assaying, each renaturation sample was diluted 10-fold with buffer containing 100 mM Hepes (pH 7.0), 0.2 mM DTT, 0.5 mM CaCl 2 , 0.1 mM EDTA, and 1 mg/ml BSA. To initiate the reaction, one volume of this diluted solution was added to three volumes of reaction mixture. Cross-linking of PhK by Formaldehyde-PhK was cross-linked by formaldehyde (5 mM), prepared by the hydrolysis of paraformaldehyde (29), either alone or in the presence of Ca 2ϩ . The final concentrations in the standard cross-linking reaction were: PhK (0.43 M), Hepes (35 mM, pH 6.8), Ca 2ϩ (1.25 mM), and EDTA (1 mM). Following cross-linking, reactions were quenched by an equal volume of SDS buffer (0.125 M Tris (pH 6.8), 20% glycerol, 5% ␤-mercaptoethanol, 4% SDS). Cross-linked conjugates were resolved on SDS-PAGE gradient gels (4 -20%) and characterized by their apparent mass and cross-reactivity against subunit-specific mAbs (20,21).
Partial Proteolysis of Native, Ca 2ϩ -activated, and Cross-linked PhK-PhK, either in the absence or presence of 1.25 mM Ca 2ϩ , was digested with chymotrypsin under conditions that promoted selective cleavage of the ␣ subunit. The standard proteolysis reaction, containing 0.43 M PhK, 0.12 g/ml chymotrypsin, 39 mM Hepes (pH 6.8), and 1 mM EDTA, was carried out for 10 min at 30°C and subsequently quenched by 2ϫ SDS buffer with brief mixing and heating at 80°C. Samples were resolved by SDS-PAGE on a 4 -20% gradient gel, stained with Coomassie blue, and characterized by their apparent mass. The extent of ␣ subunit digestion, which was linear over the time period used, and of the corresponding formation of proteolytic fragments were determined by optical integrated densitometry of the appropriate protein bands.
To localize the region of ␣ cross-linked to ␥ by formaldehyde, PhK was cross-linked as described above, but with the reaction quenched by addition of 1 M Tris (final concentration of 100 mM) and subsequently digested by chymotrypsin (0.5 g/ml). The subunit composition of crosslinked species was determined by Western blotting and peptide sequencing. For sequencing, samples were electrophoretically transferred to polyvinylidene difluoride membranes in 10 mM 3-(cyclohexylamino)propanesulfonic acid (pH 11)/10% ethanol and stained with amido black for visualization. Bands of interest were excised and submitted for amino acid analysis and N-terminal sequencing to the Harvard Microchemistry Facility.

RESULTS
The C Terminus of PhK ␣ Interacts with the Regulatory Domain of ␥ in the Yeast Two-hybrid System-Our laboratory has previously demonstrated the formation of ␣-␥ complexes through chemical cross-linking of the PhK holoenzyme (4 -7); however, relatively long cross-linkers were used in those studies, and as a result, it was not unequivocally established that the ␣ and ␥ subunits within the hexadecameric holoenzyme actually interact, as opposed to simply being proximal. To determine whether the observed cross-linking of ␣ with ␥ results from the actual association of these two subunits, and if so, to define the regions involved in that interaction, we screened a series of C-terminal truncations of ␣ and ␥ (Fig. 1) against one another in the yeast two-hybrid system. To avoid potential disruption of secondary structural elements, the truncated mutants were designed based upon the known crystal structure of the catalytic domain of the ␥ subunit (30) and the predicted secondary structure of the ␣ subunit. When all binary combinations of the ␣ and ␥ constructs were assayed against each other, no interactions were observed for any C-terminal truncation of either ␣ or ␥; a significant interaction did occur, however, when full-length constructs of both subunits were expressed (Table I). High levels of ␤-galactosidase activity were induced by the interaction of ␥FL and ␣FL as either BD or AD fusions, but with the pair BD-␣FL/AD-␥FL demonstrating ␤-galactosidase activity 3.4-fold greater than that observed with the reciprocal combination of constructs (133 versus 39. 3 Miller units). Such domain effects are not uncommon in twohybrid screens and have been observed with many proteins, including the transcriptional regulators Myc and Max (31).
The fact that none of the C-terminal deletions, but only the full-length ␣ and ␥ constructs, interacted suggested that regions near the C termini of both subunits (residues 1060 -1237 of ␣ and 343-386 of ␥) are those that interact to form the ␣-␥ complex. To test this hypothesis, two constructs comprising either the entire C-terminal tail of ␥, amino acids 301-386 (␥301N), or the shorter region therein implicated by deletion analysis to interact with ␣, amino acids 343-386 (␥343N), were engineered as both DNA BD and AD fusion constructs and screened against all ␣ subunit constructs (Fig. 1). Significant interactions were observed only between the ␥343N construct and the full-length ␣ subunit, as either DNA BD or AD fusion proteins (14.0 and 23.6 Miller units, Table I) agreeing with the implied region of association between ␣ and ␥ from the truncation analyses. The reason why the longer ␥ construct (␥301N) did not interact with any ␣ construct is unclear. It should be noted, however, that in interpreting two-hybrid data, each BD and AD fusion protein is unique and must be evaluated as such; for instance, it is possible that the longer construct, but not the shorter, resulted in a chimeric protein structure in which the interaction of LexA with its targets was sterically blocked. Similar to the results obtained with ␥301N, when a construct expressing only the C terminus of ␣ (residues 1015-1237 (␣1015N)) was screened as both a DNA BD and AD fusion protein against all of the ␥ constructs shown in Fig. 1, no interactions were observed with any ␥ mutant regardless of the transcriptional domain (data not shown). However, as is described below, a slightly shorter ␣1031N construct was found to interact with ␥FL.
Additional evidence that the C-terminal CaM-binding regulatory domain of the ␥ subunit interacts with the C-terminal region of the regulatory ␣ subunit came from screening a rabbit skeletal muscle cDNA library (22) using ␥FL versus ␥300C BD fusions as bait. Of ϳ6 ϫ 10 7 library transformants screened with ␥FL, 165 primary LacZϩ/Leuϩ colonies were obtained. Of these initial positives, 63 were selected for further analysis, and 21 library cDNAs were ultimately sequenced. The sequencing results showed every single one of these 21 clones to be overlapping transcripts of the ␣ subunit of PhK containing its entire C terminus, but varying in N-terminal start sites (Table  II). Screening the very same cDNA library using ␥300C as bait resulted in 137 primary LacZϩ/Leuϩ library cDNAs, with 23 eventually being sequenced. All but two of these 23 were independent clones, but not a single one of them corresponded to the ␣ subunit of PhK. Thus, elimination of the terminal 86 residues of ␥ eliminated the ␣ subunit as a target in the cDNA library, further demonstrating that in the two-hybrid system the C-terminal regulatory region of the ␥ subunit of PhK interacts with the C-terminal region of its regulatory ␣ subunit.
Considering that a CaM-binding regulatory region of ␥ (8) overlaps the region implicated above to bind ␣ and that ␣-␥-␦ complexes can be isolated following partial dissociation of the rabbit muscle (␣␤␥␦) 4 holoenzyme (32) raised the question of whether, during our two-hybrid screening, endogenous yeast CaM, despite its structural and functional divergence from mammalian CaM (33,34), may nevertheless interact with ␥, giving rise to formation of ternary (␣-␥-CaM), as opposed to binary (␣-␥), complexes. To evaluate the feasibility of this notion, we determined whether yeast CaM could bind and subsequently activate PhK ␥. When the isolated ␥ subunit of PhK was renatured in the presence of either purified yeast CaM or bovine brain CaM, the yeast CaM stimulated the P-b conversion activity of ␥ to a similar extent as that observed with the bovine brain isoform. The concentrations for half-maximal ac- Constructs were engineered as either transcriptional DNA BD LexA or AD BD42 fusion proteins as described under "Experimental Procedures." Amino acid numbers are given above the first construct. Constructs are named as subunit followed by either the last C-terminal residue expressed beginning from residue 1 (e.g. ␣305C is a fusion protein of amino acids 1-305 of the ␣ subunit) or by the first N-terminal residue expressed ending with the final amino acid of ␣ (1237) (e.g. ␣1015N is a fusion protein of amino acids 1015-1237 of ␣). Functionally significant domains are represented by the various patterns as follows: gray, region missing in ␣Ј (50); hatched, region missing in ␤ (51); and black and white, leucine zipper. B, PhK ␥ deletion mutants. Constructs were engineered as described under "Experimental Procedures" and labeled as described in A. Sequence domains are represented by the various patterns as follows: checkered, unique N-terminal region; gray, hinge region; hatched, C-terminal regulatory domain; and black, CaM binding domains. tivation by the yeast and bovine brain CaMs were very similar (K a ϭ 0.42 Ϯ 0.18 M, n ϭ 5 versus K a ϭ 0.30 Ϯ 0.10 M, n ϭ 3, respectively), the maximal activation by the yeast CaM was only 15% less than by the mammalian CaM (Fig. 2). These results indicate that in yeast, free CaM is capable of binding to ␥ with high affinity; thus, it is possible that the observed ␣-␥ interactions identified in the two-hybrid system may actually involve yeast CaM also and correspond to interactions occurring within the ␣-␥-␦ trimer, assuming of course that ␣ and CaM do not exclusively compete for binding to the C-terminal region of ␥.
Formaldehyde Cross-links the C Terminus of the ␣ Subunit to the ␥ Subunit within the PhK Holoenzyme-To further examine the contact regions between the ␣ and ␥ subunits, we sought a very short cross-linker that would enable us to substantiate the observed two-hybrid ␣-␥ interactions within the context of the (␣␤␥␦) 4 holoenzyme and, additionally, to correlate observed conformational changes between the two subunits with the activation of PhK. Formaldehyde was chosen as the crosslinking agent because its reaction with nucleophiles results in the insertion of but a single methylene bridge (ϳ2.9 Å, Ref. 35) between reactive side chains. When PhK was cross-linked by formaldehyde, one major conjugate with an apparent mass of 178 kDa (based on sequences an ␣-␥ dimer has a theoretical mass of 183 kDa) was formed in large amounts (Fig. 3A). The cross-linking was determined to be intramolecular, i.e. within a PhK hexadecamer because the cross-linked protein coeluted with native, hexadecameric PhK on size exclusion high pressure liquid chromatography (data not shown), a method that we routinely use to determine the molecularity of cross-linking (36,37). The formation of this 178-kDa conjugate, as followed by change in optical density, increased with time along with a corresponding loss in density of both the ␣ and ␥ subunits. Furthermore, the cross-linked species cross-reacted with only anti-␣ and anti-␥ subunit-specific mAbs by Western analysis; no cross-reactivity was observed with the anti-␤ or anti-CaM (␦) mAbs (Fig. 4B). Taken together, these data indicate the formation of an ␣-␥ dimer within the PhK holoenzyme by the very short cross-linker formaldehyde.
To localize regions of the ␣ subunit cross-linked to the ␥ subunit, we relied on the highly selective proteolysis of the ␣ subunit within the holoenzyme by chymotrypsin, which degrades that subunit essentially to completion without significant hydrolysis of the ␤, ␥, or ␦ subunits (38). Partial digestion of nonactivated PhK generated major fragments of ␣ having apparent masses of 78, 60, 58, 30, and 24 kDa; two fragments of ␣ (58 and 24 kDa, Fig. 4A) have previously been shown to cross-react with an anti-␣ mAb whose epitope is known to be near the C terminus of that subunit between residues 1132-1237 (7,20). When cross-linked enzyme was digested with chymotrypsin, two new bands were observed, ␥-␣ 1 (102 kDa) and ␥-␣ 2 (68 kDa), which cross-reacted with both the anti-␣ and anti-␥ mAbs (Fig. 4B). Based upon mass, these new bands corresponded to the entire ␥ subunit cross-linked to the 58-and 24-kDa C-terminal fragments of ␣, respectively. Examination of the proteolytic digestion pattern of cross-linked PhK demonstrated corresponding losses in density of both the 58-and 24-kDa fragments of ␣ in the anti-␣ blot, as well as a significant loss in density of the ␥ subunit in the anti-␥ blot with respect to uncross-linked proteolysed holoenzyme.  2. Stimulation of isolated ␥ subunit catalytic activity by CaM. PhK ␥ was renatured in the presence of either yeast (q) or bovine brain (E) CaM and assayed for enzymatic activity as described under "Experimental Procedures." Data points represent the average of triplicate assays with the bars indicating the S.D. Kinetic parameters given in the text were determined by a linear regression analysis of at least three to five separate experiments.
Because ␥-␣ 2 was the better resolved of the two cross-linked ␣-␥ complexes and was present in greater quantities than ␥-␣ 1 , it was used for subsequent sequence determination. N-terminal analysis of this conjugate resulted in two sequences being identified, TRDAALPG and RRLSISTE, which correspond exclusively to the N termini of ␥ and the 24-kDa fragment of ␣, respectively. The latter is known to result from cleavage at Phe-1014 and correspond to residues 1015-1237 of ␣ (20), which make up the C-terminal one-sixth of the ␣ subunit.
Effects of Ca 2ϩ on the Conformational Stability of the ␣ and ␥ Subunits-Ca 2ϩ is the most fundamental activator of PhK, regulating its activity through the ␦ subunit, an endogenous molecule of CaM. Because the C terminus of ␥, known to bind to ␦ (8 -11), was identified in this study to also associate with ␣, we investigated the effect of Ca 2ϩ on ␣-␥ dimer formation by formaldehyde. The extent of ␣-␥ dimer produced by cross-linking the holoenzyme in the presence of Ca 2ϩ increased by 2-fold over nonactivated PhK (Fig. 3). This enhancement of the formation of ␣-␥ by Ca 2ϩ corroborates the observed two-hybrid interaction of ␣ with the CaM-binding C terminus of ␥ and is consistent with other cross-linking studies from our laboratory correlating Ca 2ϩ -mediated activation of the PhK holoenzyme with perturbations in the interaction between its ␣ and ␥ subunits (5-7).
Because the ␦-mediated effects of Ca 2ϩ resulting in the activation of ␥ are concomitantly transmitted to the C-terminal region of ␣, we further examined the Ca 2ϩ -induced perturbations in this region of ␣ by its selective cleavage with chymotrypsin. When the PhK holoenzyme was partially digested in the presence of Ca 2ϩ , the pattern of cleavage of the ␣ subunit did not change; however, there was a preferential 3-fold increase 2 in the rate of formation of the 24-kDa immunoreactive fragment of ␣ previously shown to result from cleavage at residue 1014. These cross-linking and partial proteolysis results indicate that the binding of Ca 2ϩ ions to the ␦ subunit of PhK causes a distinct conformational change in the C-terminal region of its ␣ subunit. DISCUSSION Whereas the catalytic ␥ subunit of PhK is undoubtedly involved in complex interactions with all three of its regulatory subunits, no specific contact region between ␥ and either of the large inhibitory ␣ and ␤ subunits has been determined previously. In this study, we have identified a region within the stretch of residues 343-386 at the C terminus of the ␥ subunit that interacts with a region near the C terminus of the ␣ subunit. This region of ␥ is of particular importance because it contains one of the two distinct, noncontiguous CaM-binding domains present in the C-terminal regulatory region of the ␥ subunit. In a thorough study utilizing a series of 18 overlapping 25-mer peptides corresponding to the C-terminal 110-residue sequence (amino acids 277-386) of the ␥ subunit, Dasgupta et al. (8) identified two domains of ␥ with nanomolar affinity for CaM: domain N (residues 287-331) and domain C (residues 332-371). Within domain C, the peptide with the highest affinity for CaM corresponded to residues 342-366, which but for one residue are contained entirely within the region of ␥ shown herein to interact with ␣.
The fact that the same region within the regulatory domain of the ␥ subunit potentially interacts with both the ␦ (CaM) and ␣ subunits provides a plausible mechanism to explain how Ca 2ϩ induces tertiary and quaternary structural changes that are associated with activation of the PhK holoenzyme. It has been previously suggested that activation of PhK by various effectors occurs via a hierarchy of tiered conformational changes, largely reflecting differing states of release of quaternary constraint imposed by the regulatory subunits upon the catalytic ␥ subunit, with the most fundamental change being that induced by Ca 2ϩ (discussed in detail in Ref. 5). By crosslinking analysis, it is known that structural perturbations involving the ␣ and ␥ subunits occur upon activation of PhK by the binding of Ca 2ϩ ions to the ␦ subunit (5-7). Therefore, it is reasonable to hypothesize that the binding of the C-terminal regulatory region of ␥ to both ␣ and ␦ may be modulated in a Ca 2ϩ -dependent manner; that Ca 2ϩ affects the conformation of the very region of ␣ found to bind ␥ (Fig. 4) supports this hypothesis. The evidence indicates that perturbation of the ␦-␥ interactions caused by the binding of Ca 2ϩ to ␦ occurs concomitant with perturbation of the ␥-␣ interactions, resulting in a ␦-␥-␣ communication network. Recently, three-dimensional structures of the holoenzyme obtained by image reconstruction of PhK particles observed by electron microscopy Ϯ Ca 2ϩ (39) revealed that Ca 2ϩ does indeed induce distinct conformational changes over a region of PhK previously shown to be occupied by portions of its ␣, ␥, and ␦ subunits (20,21,37).
Given that purified yeast CaM activates the isolated ␥ subunit (Fig. 2), it is possible that endogenous CaM may be binding during our two-hybrid analyses to those ␥ constructs that have a CaM-binding domain, and as a result, the positive interaction we observe between full-length ␣ and ␥ may actually represent a trimeric ␣-␥-␦ instead of a dimeric ␣-␥ complex. Typically, yeast CaM is a poor activator of mammalian target enzymes 2 Ca 2ϩ causes a modest increase (ϳ20%) in the activity of chymotryp-sin (49). To control for this activation, the amount of the 24-kDa fragment generated from the ␣ subunit was determined by densitometry and expressed as a percentage of the total ␣ subunit consumed. During standard digestions, the percent of the 24-kDa fragment formed per total ␣ consumed for Ca 2ϩ -activated PhK was 15.4% versus 4.9% for non-activated PhK. including protein kinases (34,40), due to its significant divergence both structurally and functionally from mammalian CaMs; it shares only 60% identity in primary structure (33) and does not bind Ca 2ϩ at site IV (41,42). It is not surprising, however, that yeast CaM is capable of binding to and activating PhK ␥ because the ␥ subunit seems to be more sensitive to mutations in the second and third domains of CaM than in its 1st and fourth (43); moreover, the quite dissimilar TnC also activates ␥, albeit to a lesser extent than CaM (44).
If our two-hybrid interactions do, in fact, involve not only ␣ and ␥, but also CaM, then additional issues arise concerning Ca 2ϩ dependence and the interactions of the ␦ subunit of PhK. Because the three stable species of PhK (the hexadecameric (␣␤␥␦) 4 holoenzyme, the ␣-␥-␦ trimer, and the ␥-␦ dimer) all differ in the extent to which their activity is dependent on Ca 2ϩ ions, it has been suggested that increasing enzyme complexity, i.e. progressing from ␥-␦ to ␣-␥-␦ to holoenzyme, results in a progressive increase in the Ca 2ϩ -requirement for catalytic activation of the complex (32). Because the ␥-␦ complex readily dissociates in 8 M urea, whereas ␣-␥-␦ does not, it has been further suggested that extended interactions exist among these three subunits in which ␦, besides binding ␥, is stabilized from dissociation by the presence of ␣ (45). The sum of these findings suggests that ␣ may contribute, at least indirectly, to regulating the effects of Ca 2ϩ on catalysis. Our current finding that the C terminus of ␣, a regulatory region that undergoes structural perturbations in response to a variety of activating stimuli including Ca 2ϩ (7), interacts with ␥ within a region where it binds ␦ supports the existence of an intricate, Ca 2ϩ -sensitive communication network among the ␣, ␥, and ␦ subunits, an idea also strongly supported by the previously mentioned structures of PhK from electron microscopy (39). Because Ca 2ϩtriggered communication can flow from the ␦ to the ␣ subunit, then it is reasonable to expect that some manifestation of the reverse process ought to be present, as well. This expectation may be born out by early reports that activation of the PhK holoenzyme by phosphorylation of its ␣ and ␤ subunits increases affinity of the ␦ subunit for Ca 2ϩ ions (46,47).
Ca 2ϩ -dependent structural changes, similar to those described above for PhK, have been observed in the regulation of skeletal muscle troponin, specifically between the inhibitory region of TnI and its protein targets, actin and TnC (a CaM homologue). In reconstituted thin filaments, the inhibitory region of TnI, which shares remarkable sequence similarity with the C-terminal regulatory domain of the ␥ subunit of PhK (8,44), interacts preferentially with actin in the absence of Ca 2ϩ but with TnC in the presence of Ca 2ϩ (48). The possibility that the inhibitory domain of TnI and the homologous region within the ␥ subunit of PhK share a similar mechanism in their interactions with their respective protein targets (TnC and actin versus the ␦ and ␣ subunits) is supported by the following striking structural and functional similarities. (a) The regions of greatest sequence identity between ␥ and TnI (amino acids 301-325 and 103-115, respectively) contain in the case of TnI those very residues most critical for its specific interactions with actin and TnC (discussed in Ref. 44) and in the case of ␥, one of its two adjacent CaM-binding domains (8). (b) TnC activates PhK ␥ (44). (c) Actin inhibits ␥-CaM and ␥-TnC complexes (44). (d) ␥ inhibits actomyosin ATPase (44). The ␥ subunit of PhK is thought to have evolved from the fusion of a protein kinase protogene with a progenitor of exon VII of the TnI gene (44), which encodes its inhibitory domain. The probable evolutionary link between TnI and PhK ␥, the homologies that exist between them, and our current findings suggest that in skeletal muscle the troponin complex and the PhK holoenzyme may share related Ca 2ϩ -dependent alterations in quaternary structure, all leading to the simultaneous stimulation by Ca 2ϩ ions of contraction and energy production, respectively. FIG. 4. Partial proteolysis of native and cross-linked PhK by chymotrypsin. A, native or cross-linked PhK was partially digested with chymotrypsin, resolved by SDS-PAGE and stained for protein with Coomassie. Samples are as follows: 1, native PhK; 2, native PhK digested with chymotrypsin; 3, formaldehyde cross-linked PhK; 4, formaldehyde cross-linked PhK digested with chymotrypsin. The 58-and 24-kDa C-terminal fragments of ␣ are indicated by arrows. B, parallel samples were transferred to nitrocellulose and probed with anti-␣, anti-␤, anti-␥, and anti-CaM mAbs as described under "Experimental Procedures." Sample order is identical to that of A. The two new bands corresponding to ␥-␣ 1 (102 kDa) and ␥-␣ 2 (68 kDa) formed by partial digestion of cross-linked PhK are labeled. A high molecular weight, anti-␤ cross-reactive band is observed for native PhK digested by chymotrypsin and is most likely derived from an ␣-␤ dimer in which the ␤ subunit is cross-linked to an N-terminal proteolytic fragment of ␣ (36); a small amount of cross-linking of the holoenzyme by exposure to ultraviolet light, which occurs during typical handling and storage, has previously been observed (52).