Self-association of the α Subunit of Phosphorylase Kinase as Determined by Two-hybrid Screening*

The structural organization of the (αβγδ)4 phosphorylase kinase complex has been studied using the yeast two-hybrid screen for the purpose of elucidating regions of α subunit interactions. By screening a rabbit skeletal muscle cDNA library with residues 1–1059 of the α subunit of phosphorylase kinase, we have isolated 16 interacting, independent, yet overlapping transcripts of the α subunit containing its C-terminal region. Domain mapping of binary interactions between α constructs revealed two regions involved in the self-association of the α subunit: residues 833–854, a previously unrecognized leucine zipper, and an unspecified region within residues 1015–1237. The cognate binding partner for the latter domain has been inferred to lie within the stretch from residues 864–1059. Indirect evidence from the literature suggests that the interacting domains contained within the latter two, overlapping regions may be further narrowed to the stretches from 1057 to 1237 and from 864 to 971. Cross-linking of the nonactivated holoenzyme withN-(γ-maleimidobutyroxy)sulfosuccin-imide ester produced intramolecularly cross-linked α-α dimers, consistent with portions of two α subunits in the holoenyzme being in sufficient proximity to associate. This is the first report to identify potential areas of contact between the α subunits of phosphorylase kinase. Additionally, issues regarding the general utility of two-hybrid screening as a method for studying homodimeric interactions are discussed.

Phosphorylase b kinase (PhK) 1 is among the largest and most complex of the protein kinases, with a mass of 1.3 ϫ 10 6 Da and a subunit stoichiometry of (␣␤␥␦) 4 (reviewed in Refs. 1 and 2). The catalytic ␥ subunit of this complex oligomer is allosterically controlled through alterations in quaternary structure initiated by the regulatory ␣, ␤, and ␦ subunits, the latter being an intrinsic molecule of calmodulin (3). Studies have shown that activation of PhK by multiple effectors occurs concomitantly with common conformational changes in the ␣ and ␤ subunits (4 -6); moreover, the activity of free ␥ subunit is inhibited by the ␣ and ␤ subunits (7,8). It has been proposed that the ␣ and ␤ subunits of PhK impose steric constraint upon ␥ that, when removed, leads to activation of the kinase (7)(8)(9)(10).
We have chosen to study the largest of the regulatory subunits, ␣, because relatively few studies have been devoted to understanding its role in the structure and function of PhK. Previous studies using electron microscopy and immunoelectron microscopy have suggested that the ␣ subunit is peripherally located in the holoenzyme (11,12), which is consistent with its high susceptibility to proteolysis by a wide variety of proteases (5). Based upon the probable exposed location of ␣ within the PhK hexadecamer, we hypothesized that this subunit may interact with other muscle proteins and investigated potential PhK ␣ interactions by screening a skeletal muscle cDNA library for potential binding partners of ␣ using the yeast two-hybrid system.
The two-hybrid system is based on the modularity of eukaryotic transcription factors possessing functionally separate DNA binding and activation domains that need not be covalently linked to activate transcription (reviewed in Ref. 13). This attribute has been exploited to directly evaluate potential interactions between proteins fused to either DNA binding domain (BD) or activation domain (AD) modules. For positive associations, coexpression of the chimeric proteins in yeast results in positioning of the transcriptional domains in sufficient proximity to direct the activation of a phenotypically detectable downstream reporter gene.
In this study, by screening residues 1-1059 of the ␣ subunit of PhK against a rabbit skeletal muscle cDNA library for possible binding partners, we identified many independent, overlapping C-terminal transcripts of the ␣ subunit of PhK, the majority of which possess a previously unidentified leucine zipper dimerization motif. Additional mapping of the domains giving rise to self-association of ␣ identified a second region, along with its implicit cognate binding partner. Although intramolecular, cross-linked ␣-␣ dimers have previously been reported to be formed during cross-linking of PhK under conditions in which it is activated (14,15), the data presented herein are the first to identify regions of potential ␣-␣ contact.
* This work was supported by National Institutes of Health Grant DK32953 (to G. M. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Transformation and Library Screening-Yeast library transformants were generated by a modified lithium acetate procedure (21) and grown at 30°C on the appropriate synthetic media. Yeast strain EGY48, already containing the pSH18 -34 lacZ reporter plasmid and the pLexA␣1059C bait plasmid, was transformed with the rabbit skeletal muscle cDNA library and grown on synthetic complete medium containing 2% glucose but lacking histidine, tryptophan, and uracil (SD ϪHisϪTrpϪUra). The transformation efficiency of the library was 10 4 colony-forming units/g of DNA with a total of 6.4 ϫ 10 6 library clones being amplified. All transformants were pooled and stored at Ϫ80°C. To induce the expression of the library proteins under the control of the Gal1 promoter, an aliquot of the amplified library was diluted and grown for 3 h in synthetic complete medium containing 2% galactose, 1% raffinose but lacking histidine, tryptophan, and uracil (SD Gal/Raf ϪHisϪTrpϪUra). A final concentration of 6 ϫ 10 7 transformants (10 cells/amplified clone) was plated onto SD Gal/Raf ϪHisϪTrpϪUraϪLeu medium in the presence of 80 g/ml 5-bromo-4-chloro-3-indolyl ␤-Dgalactopyranoside (X-gal) in order to simultaneously screen for galactosedependent lacZ expression and leucine prototrophy of colonies expressing cDNA-encoded proteins that interact with the ␣1059 bait protein.
After colonies were grown at 30°C for 3 days, putative primary lacZ ϩ / Leu ϩ colonies were replica-plated onto master SD ϪHisϪTrpϪUra plates and stored at 4°C for secondary analysis. Primary positive colonies were restreaked onto SD ϪHisϪTrpϪUra plates to allow segregation of multiple AD library plasmids. Additionally, duplicate library clones were sorted by AluI restriction analysis of polymerase chain reaction-amplified inserts. Positive AD library plasmid DNA was isolated from a single colony, rescued in KC8 cells, and reintroduced into the original selection strain and into a second strain expressing the nonspecific Drosophila Bicoid-LexA fusion protein to test for the specificity of the interaction. Specific interactor cDNAs, those that interacted with the ␣1059 LexA fusion but not the Bicoid-LexA protein, were identified via dideoxy sequencing.
␣-␣ Interactions by Two-hybrid Screening-To screen for binary ␣-␣ interactions, a series of C-terminal deletions of the ␣ subunit were assayed for interactions as LexA and B42 fusion proteins in all possible combinations. Yeast strain EGY48, possessing the pSH18 -34 lacZ reporter plasmid, was transformed as described previously, and transformants were grown at 30°C on SD ϪHisϪTrpϪUra medium for 3 days. Positive associations between ␣ constructs were monitored by transcriptional activation of the LEU2 gene by growth on defined media lacking leucine (ϪLeu) and activation of the lacZ reporter gene, resulting in expression of ␤-galactosidase activity.
Quantification of Reporter Gene Activity via ␤-Galactosidase Activity-␤-Galactosidase activity of all positive ␣-␣ interactors, as determined by the previously described in vivo plate assays, was subsequently quantified by liquid o-nitrophenyl-␤-D-galactopyranoside spectroscopic assays (22). Briefly, yeast were grown on SD ϪHisϪTrpϪUra plates for 3-5 days at 30°C. Several colonies from each plate were scraped into 5 ml of liquid SD ϪHisϪTrpϪUra medium and grown overnight at 30°C with shaking at 250 rpm. From the liquid culture, 1 ml was removed and diluted into 5 ml of induction medium (SD Gal/Raf ϪHisϪTrpϪUraϪLeu) and grown at 30°C to mid-log phase (A 600 of 0.5-0.7). Optical density measurements at 600 nm were taken in order to normalize ␤-galactosidase activity to yeast cell density. A 1-ml aliquot of each yeast culture was briefly centrifuged to pellet the cells and resuspended in 1 ml of Z-buffer (60 mM Na 2 HPO 4 , 60 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , 38 mM ␤-mercaptoethanol, pH 7.0). After the addition of 50 l of chloroform and 50 l of 0.1% SDS, the cell suspension was vortexed vigorously for 10 s. Samples were preincubated for 5 min at 30°C, and ␤-galactosidase assays were initiated by the addition of 200 l of o-nitrophenyl-␤-D-galactopyranoside (4 mg/ml). Reactions were terminated by the addition of 400 l of 1 M Na 2 CO 3 . Reaction mixtures were centrifuged for 10 min at 13,000 ϫ g, and the A 420 was measured for each supernatant.
Cross-linking of PhK-PhK was isolated from fast twitch skeletal muscle of New Zealand White rabbits as described previously (23). The mAbs against the ␣, ␤, and ␥ subunits of PhK were produced in mice using holoenzyme as antigen and have been previously characterized (12,24); the anti-CaM mAb was from Signal Transduction Laboratories. All other detection conjugates were purchased from Southern Biotechnology. Prior to cross-linking, PhK was preincubated for 2 min at 30°C, and the cross-linking reactions were initiated with sulfo-GMBS (Pierce) and incubated at 30°C for 10 min. The final concentrations in the standard cross-linking reaction were as follows: 3.07-3.21 M ␣␤␥␦ protomer of PhK, 50 mM Hepes (pH 6.8), 1.0 mM EDTA, and 10 M cross-linker. Reactions were quenched by dilution with an equal volume of SDS buffer (0.125 M Tris (pH 6.8), 20% glycerol, 5% ␤-mercaptoethanol, 4% SDS). After heating for 5 min at 90°C, all samples (30 g/lane) were run on SDS-polyacrylamide gels (6%) (25), stained with Coomassie Blue, and destained in 40% methanol plus 10% acetic acid (2 h) and 7% acetic acid plus 4% methanol (15 h). In order to identify the cross-linked species, apparent molecular masses of the cross-linked conjugates were determined by comparison with the migration of protein standards (the ␤ (125-kDa) and ␣ (138-kDa) subunits of PhK, myosin (200 kDa), and cross-linked ␣␤ dimer from PhK (264 kDa)). The subunit composition of each cross-linked species was verified by Western blotting with subunitspecific mAbs following SDS-polyacrylamide gel electrophoresis on 4% bisacrylamide gels (cross-linking ratio 80:1). Samples were transferred by a semidry technique to nitrocellulose and blotted as described previously (12,24).
Preparation of ␣ Peptides for Epitope Mapping-In order to further define the epitope of the anti-␣ mAb (mAb157), several recombinant histidine-tagged peptides were cloned and expressed (Table IV). To construct the clone for peptide P1, cDNA encoding ␣ subunit amino acids 1097-1237 was prepared by polymerase chain reaction amplification with the following synthetic primers using ␣D1 as a template: 5Ј-ACTGCAGCATGCACTTTCTGGTG-3Ј and 5Ј-AAAGCTTCATTG-CATGGCACAG-3Ј. The product was cloned into the PstI and HindIII sites of plasmid expression vector pQE-31 (Qiagen). The clone for peptide P2, residues 1097-1185, was prepared by linearizing the clone for P1 with restriction enzymes HindIII and EcoRV, filling in the HindIII overhang with the Klenow fragment of DNA polymerase, and religating the cDNA. Clones for P5 (residues 1193-1237), P3 (residues 1132-1237), and P4 (residues 1132-1194) were generated by digesting the construct for P1 with appropriate restriction enzymes (P5, BamHI and HindIII; P3, KpnI and HindIII; and P4, KpnI and BamHI, respectively) and ligating their purified restriction fragments into pQE-31. A fusion protein of the amino-terminal domain of DHFR and ␣ amino acids 1132-1237 was generated by ligating the 318-base pair KpnI/HindIII fragment into expression plasmid pQE-41 (Qiagen). To generate the clone encoding a fusion protein of amino-terminal DHFR and ␣ residues 1188 -1202 (P6), complementary synthetic oligomers 5Ј-GATCTATCA-TGTTGTTGGCAAAGGATCCTGCATCTGGCATCTGTACTCTTCTG-3Ј and 5Ј-AGCTCAGAAGAGTACAGATGCCAG ATGCAGGATCCTTTGC-CAACATGATA-3Ј were hybridized and ligated into the pQE-40 expression plasmid (Qiagen). The DNA sequence and correct reading frame were verified for all clones by dideoxy chain termination sequencing.
Bacterial expression of the clones was induced by the addition of 2 mM isopropyl-␤-D-thiogalactopyranoside, and the expressed, histidinetagged fusion proteins were purified over Talon affinity resin (CLON-TECH) according to the manufacturer's guidelines.
Immunoassays-An ELISA was used to evaluate the binding of mAb 157 to each expressed His-tagged ␣ peptide, PhK holoenzyme, and DHFR. Briefly, each test antigen was diluted with plating buffer (50 mM Hepes, 0.5 mM EGTA, pH 6.8) to 5 g/ml, and 0.1 ml was added to the wells of Immulon I microtiter plates (Dynatech). Unbound protein was removed by washing the plates with ELISA buffer (plating buffer plus 1% (w/v) bovine serum albumin) followed by blocking of nonspecific binding sites with plating buffer plus 5% bovine serum albumin (w/v). Each well was then incubated with monoclonal anti-␣ antibody (mAb 157), diluted 50-fold in ELISA buffer, or with ELISA buffer alone. Wells were washed three times with ELISA buffer and incubated with alkaline phosphatase-labeled goat anti-mouse IgG (diluted 500-fold in ELISA buffer). After three additional washes, the hydrolysis of pnitrophenylphosphate (1 mg/ml in 0.9 M diethanolamine, 10 mM MgCl 2 , pH 9.8) was detected by absorbance at 405 nm.

RESULTS
The ␣ Subunit of PhK Interacts with Itself in the Yeast Two-hybrid System-A yeast two-hybrid system was used to investigate potential protein interactions of the ␣ subunit of PhK. Given that the ␣ subunit is farnesylated at Cys-1234 (26), potentially directing it to membranes, we chose to use only amino acids 1-1059 of ␣ fused to the LexA DNA BD protein (␣1059C-BD) as a molecular bait for screening a rabbit skeletal muscle cDNA library, in order to improve the chances for nuclear import of the bait. The site of C-terminal truncation at residue 1059 was chosen so as not to disrupt predicted secondary structure domains. Complete characterization of our bait protein prior to library screening demonstrated that it was correctly expressed, as determined by Western blotting with a polyclonal antibody directed to the LexA protein; that it possessed no intrinsic transcriptional activation of the LEU2 and lacZ reporter genes independently or in the presence of an empty activation domain vector; and that it was properly transported to the nucleus, as measured by a ␤-galactosidase repression assay (27).
Of the approximately 6 ϫ 10 7 library transformants screened, more than 300 primary positive colonies were obtained that showed galactose-dependent leucine prototrophy and lacZ expression. Of these initial positives, 80 were selected for further secondary and tertiary analyses, ultimately resulting in 23 library cDNAs being identified by dideoxy sequencing. Sequencing results identified 16 of the 23 clones as independent, overlapping transcripts of the ␣ subunit of PhK containing its carboxyl terminus but varying in amino-terminal start sites (Table I). Moreover, 12 of the sequences contained a region of ␣ having a previously unrecognized leucine zipper motif from residues 833-854 ( Fig. 1). Leucine zippers form parallel coiledcoil ␣-helices via the interactions of hydrophobic side chains, and while initially identified as dimerization domains in transcription factors, they have subsequently been shown to exist in other classes of proteins (28).
Because four ␣ clones (␣969N, ␣976N, ␣1013N, and ␣1015N) that were identified in our screen contained only the extreme C terminus of ␣ and did not possess any portion of the leucine zipper, we investigated the role of the region distal to the leucine zipper by assaying three representative library AD fusion proteins (␣591N, ␣730N, and ␣1015N) against a series of ␣ C-terminal deletion mutants fused to the DNA binding domain (Fig. 2). To further evaluate the role of the leucine zipper in ␣-␣ associations, we also made an AD fusion protein from clone ␣730N (␣730 -863) lacking the C terminus of ␣ but maintaining the leucine zipper. None of the three AD library clones screened interacted with the BD construct containing the first 305 residues of PhK ␣ (Table II), suggesting that the amino and carboxyl termini between two ␣ subunits do not interact with each other. However, the addition of amino acids 306 -863 of ␣ to the BD construct (i.e. ␣863C) produced strong interactions with library clones ␣591N and ␣730N (6.2 and 22.0 Miller units, respectively) but not with clones ␣ 1015N, lacking the leucine zipper, or ␣ 730 -863, lacking the C terminus. Upon the addition of residues 864 -1059 of ␣ , interactions were observed between this construct, ␣1059C, and all four library clones, although the ␤-galactosidase activity between ␣1059C and ␣730 -863 was 76 -89% less than the activity measured with the other three constructs. Interactions between ␣1059C and clones ␣591N, ␣730N, and ␣1015N had the highest ␤-galactosidase activity of any constructs and may contain the regions of highest affinity between ␣ subunits, given that the measured strength of protein associations discriminated in vivo by twohybrid analyses has been reported to correlate with the corresponding binding affinities determined in vitro (16).
The sum of the results from Tables I and II is consistent with the involvement of at least three domains, two identified and one implicit, in the self-association of PhK's ␣ subunit. The first of these is the newly identified leucine zipper between residues 833 and 854, which would be expected to interact with the same region of another ␣ subunit. The second association domain directly identified is contained within the stretch from 1015 to 1237, present in all library clones detected by ␣1059C (Table I).
The final and implicit association domain, that recognized by the domain within residues 1015-1237, has been only indirectly identified. That ␣1015N interacts with ␣1059C, but not with ␣863C (Table II), suggests that the cognate partner for the second domain lies somewhere within residues 864 -1059, a unique region not found in the otherwise 80% homologous ␤ subunit of PhK (29) that lies just C-terminal to the leucine zipper. It should be noted that the broad regions containing the second domain and its implicit cognate partner (1015-1237 and 864 -1059) overlap between residues 1015 and 1059; the significance of this overlapping region is addressed under "Discussion." Domain Mapping of Potential ␣-␣ Interactions-To further define the regions of potential ␣-␣ interactions, all of the Cterminal deletion mutants of the ␣ subunit depicted in Fig. 2 were assayed for interactions in every possible binary combination, and the results are shown in Table III. As was previously observed with the library ␣ clones, no interactions occurred when constructs containing only the N terminus of ␣ were screened, residues 1-305 in both domains or 1-470 in . b We were unable to sequence through the C terminus of the construct; however, the size of the transcripts, as determined by agarose gel electrophoresis, indicates the library insert extends through the termination codon and contains part of the 3Ј-untranslated region.
only the activation domain. We were unable to screen the reciprocal ␣470C-BD construct due to its inability to express properly in the yeast cells.
The addition of amino acids 471-650 to generate the BD construct ␣650C independently activated transcription of both the LEU2 and lacZ reporter genes (data not shown); thus, this construct was not usable. The AD construct ␣650C induced high levels of ␤-galactosidase activity (15.7, 16.3, and 19. 8 Miller units) when screened against the DNA BD constructs ␣863C, ␣1059C, and ␣FL, respectively; however, for several reasons, we think that these results are artifactual. First, further elongating ␣650C to form ␣863C eliminates interactions. In fact, ␣863C fused to either the DNA BD or AD does not interact with any of the ␣ C-terminal deletion mutants, except ␣650C, suggesting that the interactions of ␣650C-AD may be artifacts resulting from a normally buried domain that is exposed in this truncation. Second, that ␣650C-BD independently activated transcription raises concerns regarding the behavior of its AD counterpart. Third, the results obtained with ␣650C-AD are completely inconsistent with the interacting library clones identified by ␣1059C (Table I), none of which lacked the C-terminal region of ␣ and most of which began after residue 650.
The constructs ␣1059C and ␣FL interact with each other as both DNA BD and AD fusion proteins (7.3 and 5.3 Miller units, respectively) and, to a lesser extent, as isologous pairs, i.e. ␣1059C-␣1059C and ␣FL-␣FL (Table III). While the expression of all fusion proteins has been confirmed by immunoblot anal-ysis (data not shown), the observed minor variations in transcriptional activation between the same pair of ␣ constructs expressed in opposite domains is probably due to dissimilar levels of nuclear transport or protein expression (16).
Homodimeric Interactions of the ␣ Subunit within the PhK Holoenzyme-Because the yeast two-hybrid system identifies only potential dimeric contacts, it was necessary to demonstrate the relevance of the observed ␣-␣ interactions within the context of the oligomeric (␣␤␥␦) 4 holoenzyme. Although proximal, but unidentified, regions of two ␣ subunits have been detected by covalent cross-linking, this was under conditions where the enzyme is activated (14,15). In fact, in the case of cross-linking by transglutaminase, ␣-␣ dimers were found to be formed only with activated enzyme (15). Potential ␣-␣ interactions have not been previously observed with the nonactivated conformer of the PhK holoenzyme. We selected chemical crosslinking at the nonactivating pH of 6.8 as the method to probe for these interactions, because the unambiguous identification of cross-linked conjugates is now possible using PhK subunitspecific mAbs (30). From screening a variety of bifunctional reagents, the heterobifunctional cross-linker sulfo-GMBS was observed to form an ␣-␣ dimer (268.5 kDa, 3.0% error) (Fig. 3) with nonactivated PhK, in addition to ␣-␤ dimers and several uncharacterized high molecular weight oligomers. The crosslinking was determined to be intramolecular, i.e. within a PhK hexadecamer, because the cross-linked protein coeluted with native, hexadecameric PhK on size exclusion HPLC (data not shown), a method that we routinely use to determine the molecularity of cross-linking (15,31). These cross-linking results indicate that portions of at least pairs of ␣ subunits are indeed proximal within the PhK holoenzyme.
Defining the Epitope of Anti-␣ mAb 157-A mAb against a

FIG. 2. PhK ␣ subunit deletion mutants.
A, C-terminal deletion mutants. Constructs were designed based upon secondary structure predictions and 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 the last C-terminal residue expressed beginning from amino acid 1, e.g. ␣305C is a fusion protein of amino acids 1-305 of the ␣ subunit. Functionally significant domains are represented by various patterns as follows: gray, region missing in ␣Ј (48); striped, region missing in ␤ (29); and black and white, leucine zipper. B, PhK library constructs. Regions of PhK ␣ were identified by yeast two-hybrid screening using a construct of ␣ containing residues 1-1059 as bait as described under "Experimental Procedures." Constructs are named as subunit, followed by the first N-terminal residue expressed ending with the final amino acid of ␣ (position 1237), e.g. ␣591N is a fusion protein of amino acids 591-1237 of the ␣ subunit. Sequence domains are represented by the various patterns as described for A.
region within the C terminus of the ␣ subunit was previously shown by immunoelectron microscopy to bind near the tips of the lobes of PhK (12), which is arranged as a dimer of perpendicular (␣␤␥␦) 2 dimers, with each octameric dimer composing one long (22.2-nm) lobe (32). The epitope for this mAb (mAb 157, which was used in generating Fig. 3) was initially only broadly defined; by cross-reactivity against proteolytic fragments of ␣, it was shown to lie within the C-terminal residues 1015-1237 and, based on indirect evidence, was proposed to lie within either residues 1097-1133 or 1152-1233 (12). Because the anti-␣ mAb binds at a nearly terminal position on the long lobes and because our two-hybrid analyses identified one domain of ␣ subunit self-association as being in the broad region between residues 1015 and 1237 that also contains the epitope for that mAb, we thought it incumbent to narrow the epitope for the anti-␣ mAb in order to determine whether this region of ␣ self-association could possibly contribute to homodimeric interactions, given the location of the ␣ subunits in the PhK holoenzyme.
In this current study, the epitope for mAb 157 was further defined by evaluating its binding to expressed fusion peptide antigens containing various portions of the C-terminal 140 residues of the ␣ subunit (Table IV). The mAb 157 bound strongly to an immobilized peptide (P1, 1097-1237) containing both of the candidate epitopes listed above, producing ELISA readings of 55% that with the holoenzyme, confirming that the epitope is located within the last 140 residues of ␣. Furthermore, a shorter peptide (P3, 1132-1237) containing only the second candidate epitope also interacted with the mAb, albeit at a lower level of cross-reactivity in ELISAs (5% that of the holoenzyme). These peptides were also expressed as N-terminal fusion proteins with DHFR, allowing analysis of crossreactivity by Western blots, and bands of approximately equal intensity were observed in the immunoblots of the two peptides (data not shown). In contrast, a peptide containing the entire first candidate epitope but missing half of the second (P2, 1097-1185) showed no cross-reactivity with the mAb in either ELISAs (Table IV) or immunoblots (data not shown). Thus, the epitope for the anti-␣ mAb lies somewhere within amino acids 1132-1237 (P3), presumably within the second (1152-1233) of the two originally proposed candidate epitopes. When the cross-reactive peptide P3 was divided in the middle of the second candidate epitope at residue 1194 or 1193, neither resultant peptide (P4 and P5, respectively) cross-reacted with the a Construct ␣730 -863 was derived by removing a XhoI-XhoI fragment (amino acids 864 -1237) from clone ␣730N, thus removing the C terminus while leaving the leucine zipper intact.

TABLE III
Domain mapping of ␣-␣ interactions ␤-Galactosidase activity from yeast lysates was determined using o-nitrophenyl-␤-D-galactopyranoside as substrate as described under "Experimental Procedures." Activity is expressed as Miller units according to the following formula: A 420 ϫ 1000/(volume)(time)(A 600 ). Data represent the mean Ϯ S.E. of three or six independent assays performed in duplicate. Since all measured vector control values were between 0.0 and 1.8, a positive interaction is determined as Ն2.0 Miller units.

Vector encoding B42AD
Vector encoding LexA Vector

FIG. 3. Identification of cross-linked ␣-␣ dimer by immunoblot.
PhK was cross-linked by sulfo-GMBS prior to immunoblotting with anti-␣ and anti-␤ mAbs as described under "Experimental Procedures." For each blot, lane 1 shows uncross-linked PhK, and lane 2 shows cross-linked PhK. The ␣-␣ dimer is indicated by the arrow. In the control lanes, there are small amounts of ␣-␤ dimers, which are frequently observed after storage of the enzyme. Cross-linked samples were also subjected to immunoanalysis using mAbs against the other two PhK subunits, ␥ and ␦, and no cross-reactivity was observed for the cross-linked complexes displayed (data not shown).

TABLE IV
The epitope of anti-␣ mAb 157 An ELISA was used to compare the relative binding of mAb 157 to six immobilized peptides corresponding to different regions of the ␣ subunit. Peptides P1-P5 were tested as purified His-tagged fusions, whereas P6 was an N-terminal fusion with DHFR. Each sample was assayed in triplicate. Relative binding refers to the ratio of the mean absorbance at 405 nm for each test antigen to the mean absorbance at 405 nm for PhK holoenzyme, as described under "Experimental Procedures." Control values of samples in the absence of primary ␣ mAb were subtracted. anti-␣ mAb (Table IV), suggesting that residues surrounding 1193-1194 contribute to the epitope. A short tetradecameric peptide covering a limited span on either side of these residues (P6, 1188 -1202) and tested as a fusion protein with DHFR did not, however, bind the mAb in either ELISAs (Table IV) or Western blots (data not shown). Taken together, the immunoassay results demonstrate that the epitope recognized by mAb 157 is located within residues 1132-1237, considerably removed from the identified regions of possible ␣-␣ interaction and, additionally, suggest that the epitope, perhaps a discontinuous one, may involve amino acids very near the C terminus of ␣ in a region near residues 1193-1194.

DISCUSSION
By using the genetic two-hybrid screen, we have identified two regions involved in the self-association of the regulatory ␣ subunit of PhK: residues 833-854, a previously unrecognized leucine zipper, and an unspecified region within residues 1015-1237. In addition, the cognate binding partner for the latter domain has been inferred to lie somewhere within the stretch from 864 to 1059. That homodimeric interactions between ␣ subunits may occur in the nonactivated, native holoenzyme was shown by the formation of ␣-␣ dimers through chemical cross-linking. This is the first report to demonstrate ␣-␣ formation in nonactivated PhK and to define potential areas of contact between two ␣ subunits. It is noteworthy that after multiple, stringent specificity tests, 70% of all the positive clones identified in our library screen contained overlaps of the identical region of the ␣ subunit of PhK (Table I). This high percentage of overlapping ␣ clones isolated from a skeletal muscle library implies specificity in the interactions of the domains identified.
Commonly cited limitations of two-hybrid analyses to identify protein-protein interactions include such possibilities as the improper folding of truncation mutants or the instability of expressed proteins within yeast cells; however, additional complications arise when probing homodimeric interactions. One such complication introduced when studying homodimeric interactions in the two-hybrid system is that of stable, but nonproductive, associations. For example, within the yeast, ␣FL can potentially dimerize as three different pairs of fusion proteins, ␣FL-AD/␣FL-AD, ␣FL-BD/␣FL-BD, and ␣FL-BD/␣FL-AD, of which only the last possesses the potential to activate transcription of downstream reporter genes. Thus, as little as one-third of the potential transcriptional activation might be observed for ␣FL-␣FL interactions. So, our finding that two ␣FL chimeras produce transcriptional levels lower than the interactions of ␣FL with ␣1059C assayed in either domain (Table III) could be explained by the nonproductive dimerization of ␣FL; even if the ␣FL-␣1059C interactions were actually weaker than ␣FL-␣FL, they potentially could still result in a higher activity being measured, depending on the relative affinities of the different binary combinations. An alternative explanation for the lower values with ␣FL is that a fraction of these constructs could be directed to membranes, unlike ␣1059C, which lacks the site for farnesylation (26).
Another complication when probing homodimeric interactions in the two-hybrid system is the meaning of the observed interactions; this is especially true in the case of an oligomer that itself may further associate. For example, the stable, native (␣␤␥␦) 4 PhK hexadecamer is capable of further associating through uncharacterized interactions to form even larger complexes (10,11,33,34). Consequently, a positive, nonartifactual, homodimeric interaction between truncated ␣ constructs in the two-hybrid system could potentially represent, disregarding interacting leucine zippers, any of three different possibilities with respect to the native protein: (a) intrasubunit interactions that would normally stabilize that subunit's tertiary structure; (b) intersubunit, intramolecular interactions stabilizing the (␣␤␥␦) 4 parent molecule's quaternary structure; or (c) intersubunit, intermolecular interactions between (␣␤␥␦) 4 molecules. Given these alternative possibilities, the most general interpretation for interacting constructs of the ␣ subunit not containing the leucine zipper is that they contain regions of selfassociation that may or may not represent domains of dimerization that stabilize the quaternary structure of the holoenzyme.
For the observed two-hybrid homodimeric interactions of ␣ to possibly represent ␣-␣ dimers within the (␣␤␥␦) 4 holoenzyme (i.e. possibility (b) in the paragraph above), it is necessary to establish that two ␣ subunits are, in fact, sufficiently close to associate intramolecularly, which was achieved by chemical cross-linking, and that the identified regions of potential association are consistent with what is known concerning the quaternary structure of PhK. As determined by electron microscopy and immunoelectron microscopy, the overall structure of PhK is a pseudotetrahedron of four ␣␤␥␦ protomers that form two lobes, each an (␣␤␥␦) 2 dimer, arranged in D 2 symmetry (32). In the hexadecamer, an epitope of the ␣ subunit within residues 1015-1237 has been localized to near the distal tips of each lobe by immunoelectron microscopy (12). That the potential ␣-␣ contact regions identified in this study, the leucine zipper and domains C-terminal to residue 864, are not far removed in primary structure from the broadly defined, previously reported anti-␣ mAb epitope at the lobe tips raised concern as to whether these regions could contribute to intramolecular ␣-␣ interactions. Because the average distance between tips of a single lobe of the holoenzyme is 22 nm (32), it would be unlikely that the more distal regions of potential contact between ␣ subunits would be in sufficient proximity to touch if the epitope on ␣ localized to the lobe tips were actually proximal to residue 1059 (12). Our refinement of the epitope for the anti-␣ mAb to residues 1132-1237 moves the portion of ␣ located at the lobe tips at least another 113 residues closer to the extreme C terminus of ␣ (Table IV), thus adding considerably more distance between the lobe tips and the potential regions of ␣-␣ association.
We have identified in the ␣ subunit, but not in the homologous ␤ subunit of PhK, a leucine zipper dimerization motif, generally defined as a row of 4 -7 heptad leucine repeats (Fig.  1). Although leucine zippers were first discovered in transcription factors, subsequently they have been shown to mediate dimerization in many non-DNA-binding proteins (28,(35)(36)(37). High resolution NMR and x-ray diffraction studies have shown that zippers interact as parallel coiled-coil ␣-helices, with the dimer interface being formed from the hydrophobic side chains of the a and d positions of one helix packing side by side against the aЈ and dЈ positions, respectively, of the other helix (38). Generally, a conserved Asn residue is buried in the hydrophobic dimer interface and is thought to contribute, via destabilization of the coiled-coil, to the selectivity for hetero-versus homodimerization of monomers (Ref. 38 and references therein). The leucine zipper of PhK aligns well with other known zippers, although it does not contain the conserved Asn but instead has a Cys residue at that position (Fig. 1). Stabilization of leucine zippers is often further mediated via electrostatic interactions between the e-gЈ and eЈ-g residues, respectively (38), and in the PhK leucine zipper, there are several charged residues in these positions available for salt bridge formation (Fig. 1). It should be pointed out that not all binary combinations of constructs having two leucine zippers showed interactions in the two-hybrid system (Tables II and III); however, in all cases but one, at least one of the two constructs being tested terminated at residue 863, only nine residues removed from the end of the zipper.
The second association domain, which occurs somewhere between residues 1015 and 1237, and its cognate binding partner, which we infer to lie at an unspecified region between residues 864 and 1059, overlap in the stretch between residues 1015 and 1059; however, for various reasons we think that this overlapping stretch is an unlikely sequence to contain an association domain. First, Wullrich et al. (39) have found that amino acids 1012-1024 and 1025-1041 are deleted, as a result of differential mRNA splicing, independently or in combination in various tissues of the rabbit. Second, the stretch from 1012 to 1056 is a region of very low homology between the rabbit muscle and liver ␣ subunits (40). Third, at least seven phosphorylation sites occur between residues 972 and 1030 (20). The sum of this evidence strongly suggests that the overlapping region contains regulatory domains, as opposed to an association domain. Thus, the second association domain probably occurs somewhere C-terminal to residue 1056, and its cognate binding partner most likely lies N-terminal to residue 972, the first of the known phosphorylation sites. These additional indirect data would then place the second pair of candidate domains of association somewhere within residues 1057-1237 and 864 -971.
Besides the regions of ␣ implicated in its self-association that were identified in this study, low resolution data are also available concerning regions of this subunit that interact with other subunits within the (␣␤␥␦) 4 PhK holoenzyme. Using various cross-linkers as structural probes of the hexadecamer, the ␣ subunit forms cross-linked complexes with the catalytic ␥ subunit, and these conjugates are substantially increased upon activation of the enzyme (30,41). One region of ␣ cross-linked to the ␥ subunit has been delineated to lie C-terminal to residue 1015 (41), which overlaps the region containing the second association domain identified herein. As noted above, this general region of ␣ contains multiple phosphorylation sites, and given that phosphorylation of PhK leads to increased activity of its catalytic ␥ subunit (42), it is reasonable to hypothesize that both ␣-␥ and ␣-␣ contacts are altered during the activation of PhK. Such a notion is supported by previously reported differences in ␣-␣ cross-linking under activating versus nonactivating conditions (15). With respect to the ␤ subunit of PhK, cross-linking has also shown that the ␤ and ␣ subunits abut each other at a region somewhere within the N-terminal half of ␣ (31). The remaining subunit of PhK, ␦, an endogenous molecule of calmodulin (3), apparently binds predominantly to the ␥ subunit within the holoenzyme (43,44); however, exogenous calmodulin, termed ␦Ј, also binds to PhK, at least in part to its ␣ subunit (45,46). Although not modified in the holoenzyme by derivatized calmodulin (46), a calmodulin-binding domain of ␣ from residues 660 -677 (47) that is missing in the cardiac and slow muscle ␣Ј isoform (48) has been suggested to be surfaceexposed and possibly involved in the binding of ␦Ј (47). From our results, this region is not involved in the self-association of ␣ , especially given the identification in our two-hybrid screen of an ␣ clone lacking residues 654 -712, as in the ␣Ј isozyme (Table I) (48).
The regions of self-association of the ␣ subunit identified in this study may be relevant to the inherited glycogen storage disease type IX, which results from mutations in PhK (reviewed in Ref. 49). The ␣ subunit of PhK is encoded by two separate genes, PHKA1 (muscle) and PHKA2 (liver) (50,51). Many mutations leading to liver-specific glycogenoses have been mapped to the ␣ gene PHKA2 and are biochemically differentiated as either XLG I or XLG II, based upon phenotype. XLG I is characterized by reduced PhK activity in liver, erythrocytes and leukocytes (52), while XLG II is characterized by variable hepatic PhK activity, with normal or elevated activity in erythrocytes and leukocytes (53). It has been postulated that the two phenotypes may arise from differences in mutations that result either in regulatory changes in PhK (XLG II) or in a disruption of the structural integrity of the holoenzyme (XLG I) (54). When one examines the location of the various identified mutations on the ␣ gene, a distinct clustering to either the N-terminal or C-terminal regions of ␣ is observed, with no mutations found to occur between residues 400 and 765 (55). The 18 mutations in the N-terminal region lead to a roughly equal distribution of the XLG I and XLG II phenotypes, with eight of the former and 10 of the latter. In contrast, 11 of the 13 mutations clustered in the C-terminal region of ␣ lead to the XLG I phenotype, which is suggested to be due to a structural instability of PhK (54,55), consistent with the findings herein of several regions of self-association C-terminal to residue 833 of the ␣ subunit.