Adenosine 3’S’-Monophosphate-regulated Phosphorylation of Erythrocyte Membrane Proteins SEPARATION OF MEMBRANE-ASSOCIATED CYCLIC ADENOSINE 3’:5’-MONOPHOSPHATE-DEPENDENT PROTEIN KINASE FROM ITS ENDOGENOUS SUBSTRATES*

An adenosine 3’:5’-monophosphate (cyclic AMP)-binding protein in the human erythrocyte plasma membrane was isotopically labeled using a photoaffinity analog of cyclic AMP, NO-(ethyl Z-diazomalo-nyl) cyclic [aH IAMP. The cyclic AMP-binding site is located in a polypeptide chain having a molecular weight of 48,000. Cyclic AMP-binding protein and cyclic AMP-dependent protein kinase were solubilized with 0.5Yo Triton X-100 in 56 mM sodium borate, pH 8, but SZP-labeled membrane phosphoproteins were retained in the Triton-insoluble fraction, suggesting that the membrane-associated binding protein is not a primary substrate for protein kinase. Triton-solubilized and membrane-associated protein kinase activities were stimulated 15- and 17-fold by cyclic AMP, suggesting that the degree of association between the catalytic and cyclic AMP-binding components was very similar in both preparations. Fractionation and characterization of membrane have shown that protein III and a co-migrating minor protein are substrates for protein kinase but membrane sialoglycoproteins are not phosphorylated.


S. RUBIN+
From the Departments of Neuroscience and Molecular Biology, Albert Einstein College of Medicine, Bronx, New York 10461 An adenosine 3':5'-monophosphate (cyclic AMP)-binding protein in the human erythrocyte plasma membrane was isotopically labeled using a photoaffinity analog of cyclic AMP, NO-(ethyl Z-diazomalonyl) cyclic [aH IAMP. The cyclic AMP-binding site is located in a polypeptide chain having a molecular weight of 48,000. Cyclic AMP-binding protein and cyclic AMP-dependent protein kinase were solubilized with 0.5Yo Triton X-100 in 56 mM sodium borate, pH 8, but SZP-labeled membrane phosphoproteins were retained in the Triton-insoluble fraction, suggesting that the membrane-associated binding protein is not a primary substrate for protein kinase.
Triton-solubilized and membrane-associated protein kinase activities were stimulated 15-and 17-fold by cyclic AMP, suggesting that the degree of association between the catalytic and cyclic AMP-binding components was very similar in both preparations. Fractionation and characterization of membrane phosphoproteins have shown that protein III and a co-migrating minor protein are substrates for protein kinase but membrane sialoglycoproteins are not phosphorylated.
Cyclic AMP'-dependent protein kinases are integral components of plasma membranes (l-8) as well as cardiac muscle sarcoplasmic reticulum (9,10). These kinases catalyze the transfer of the y-phosphate of ATP to serine and threonine hydroxyl groups in specific membrane polypeptides (l-10). The juxtaposition of cyclic AMP-dependent protein kinase and phosphate-acceptor proteins in plasma membranes could provide a system capable of directly modifying the properties of the cell surface in response to polypeptide hormones, biogenic amines, and other activators of adenylate cyclase. Such a system could operate entirely within the domain of the plasma membrane and remain independent of cytoplasmic and nuclear factors during the initial phase of cellular response (i.e. membrane modification) to effector binding. Many observations have implicated cyclic AMP and protein phosphorylation and dephosphorylation reactions in the regulation of membrane function and structure (11). For example, Greengard  and pharmacological evidence (12, 13) consistent with the proposal that synaptic transmission is modulated by the cyclic AMP-regulated phosphorylation of two polypeptides responsible for controlling cation permeability in postsynaptic plasma membranes; the phosphorylation of a single polypeptide in the plasma membrane of the turkey erythrocyte has been temporally correlated with isoproterenol-induced increases in intracellular cyclic AMP concentration and passive permeability to Na+ (14); and the phosphorylation of two small polypeptides at the external surface of rat adipocytes has been associated with the inhibition of .insulin-stimulated glucose transport in intact cells (6) and is controlled by cyclic AMP in purified adipocyte plasma membranes (7). Nevertheless, functional roles have not been unequivocally defined for any of the plasma membrane proteins which serve as phosphate-acceptors for intrinsic cyclic AMP-dependent protein kinases. Human erythrocyte plasma membranes provide an excellent system for extending the characterization of endogenous substrates because they can be isolated in highly purified form, their major polypeptide components are few in number, and many of their structural, enzymic, and antigenic features have been documented (15)(16)(17). Approximately 80% of the protein kinasel activity of the human erythrocyte is localized on the internal surface of the plasma membrane (2,18). Intrinsic protein kinase catalyzes the cyclic AMP-regulated phosphorylation of two membrane polypeptides that exhibit apparent molecular weights of 88,000 and 48,000 on electrophoresis in analytical Na dodecyl-SO,polyacrylamide gels (3)(4)(5) as well as the cyclic AMP-independent phosphorylation of a considerably larger protein having a molecular weight of 215,000 (4). These phosphoproteins correspond to stained protein bands previously designated III, IVc, and II,' respectively (Ref. 4 see Fig. 1).  Brunswick and Cooperman (23). Pronase was purchased from Calbiochem, casein and serum albumin were from Miles, and Triton X-100 was obtained from Rohm and Haas.
Erythrocyte Membmnes-Packed erythrocytes were obtained from the Jacobi Hospital blood bank, Bronx, New York. Erythrocyte plasma membranes were prepared according to the method of Dodge et al. (24).

Protein
Kinase Assay-Assays of protein kinase activity, using protamine or casein as exogenous substrates, were carried out as described by Rubin et al. (2).  Fig. 1). Three membrane polypeptides were phosphate acceptors for endogenous protein kinase (Fig. 2). These phosphoproteins exhibit the same relative mobilities as proteins II, III, and IVc in Na dodecyl-SO,-polyacrylamide gels ( Fig. 2), but are now designated Pl, P2, and P3, respectively, to indicate that coincidence of electrophoretic mobilities is not a sufficient criterion for determining the identity of radioactively labeled and stained polypeptides (see below). Phosphoproteins P2 and P3 are also not readily distinguished from sialoglycoproteins PAS-l and PAS-2 in the standard gel system (Fig. 2).
Pl is the principal phosphate acceptor, but its phosphorylation was unaffected by cyclic AMP (Fig. 2). In contrast, the levels of phosphorylation of P2 and P3 were stimulated 2-and &fold by 3 PM cyclic AMP in the experiment shown in Fig. 2. Cyclic AMP had no effect on lipid phosphorylation. [%H]AMP), was employed to label specifically cyclic AMP-binding proteins of the erythrocyte membrane. Cyclic AMP and its analog were tightly bound by the membrane. Table I presents the binding  parameters  obtained from reciprocal plots of bound cyclic nucleotide uersus free cyclic nucleotide.
Erythrocyte membranes exhibited an affinity constant (K,) for EtN,Mal-cyclic AMP that was one-half as great as the K. for cyclic AMP and also contained approximately the same number of binding sites for both cyclic AMP and its photoaffinity analog ( Table I). Preincubation of membranes at 0" with saturating concentrations of either SH-labeled cyclic nucleotide followed by lo-fold dilution and washing with buffer at O-2" resulted in the retention of 70 to 75% of the bound cyclic nucleotide by the membranes (Table II). High retention of bound cyclic nucleotides under these conditions is due to the low rate of cyclic nucleotide dissociation at O-2" (2). Membranes pretreated with nonradioactive cyclic AMP or EtN,Mal-cyclic AMP as described above (see Table II) and subsequently incubated with an excess of either JH-labeled cyclic nucleotide bound only 26 to 34% of the maximum amount of the labeled compounds (Table II). These observations and the similarity in the number of binding sites and K, values (Table I) suggest that EtN,Mal-cyclic AMP is bound at the same binding site as cyclic AMP.
In the absence of ultraviolet illumination the binding of EtNJ'vIal-cyclic AMP is noncovalent as indicated by its dissociation from the membrane at 37" and the release of all bound LH when the "membrane .EtN,Mal-cyclic [*HIAMP complex" is denatured by trichloroacetic acid (Table I). EtN,Mal-cyclic AMP also mimicked cyclic AMP in stimu-  (A,.,,,, = 254 nm, see "Experimental Procedures").
Photolysis of the 2-diazo group generates a highly reactive carbene intermediate which rapidly forms a covalent bond at the cyclic AMP binding site (33,34). Electrophoretic analyses of the photolytically labeled membranes showed that the incorporation of *H into specific membrane proteins reached a maximum after 4 min irradiation (Fig. 3). As expected, covalently bound *H remained associated with the proteins of photoaffinity labeled membranes under a variety of conditions which readily dissociate noncovalently bound ligands (Table III) (Table  III). Preincubation with 5'-AMP, adenosine, or cyclic GMP had no effect on *H incorporation.
The absence of "H incorporation in experiments utilizing either heat-denatured membranes or prephotolyzed EtN,Mal-cyclic AMP indicates that both the native cyclic AMP-binding protein and the diazo form of the cyclic AMP analog are required for photoaffinity labeling (Table III). having a minimum molecular weight of 48,000 (estimated from the semilog plot in Fig. 1). In preliminary investigations Guthrow et al. (34) suggested that the erythrocyte membrane cyclic AMP-binding protein might be phosphorylated.
The possibility that the cyclic AMP-binding protein served as a substrate for membrane-associated, cyclic AMP-dependent protein kinase was examined in the following manner: membranes were labeled with SH using the photoaffinity technique or with sT via endogenous protein kinase in the presence of [y-*2P]ATP or with both radioisotopes by sequential photoaffinity labeling and phosphorylation, and then were subjected to electrophoretic analysis on polyacrylamide gels. Both mixtures of $*P-and $H-labeled membranes and preparations of double-labeled membranes gave the same patterns of radioactive and Coomassie blue-stained polypeptides. A typical electrophoretogram is shown in Fig. 4. The *H-labeled cyclic AMP-binding protein and phosphoprotein P3 exhibited very similar relative mobilities, suggesting that the binding protein might be phosphorylated by the catalytic subunit of cyclic AMP-dependent protein kinase. Reservations about this interpretation of the results arose when a series of five independent experiments showed that a high degree of overlap between these two radioactive proteins was reproducibly observed (Fig. 4), but the peaks of $H and s2P were never completely coincident.
The SH-labeled cyclic AMP-binding protein consistently migrated more slowly (average R, = 0.509) than P3 (R, = 0.521) and consequently appeared to be a slightly larger polypeptide (Fig. 4). Thus, the relationship between the binding protein and P3 could not be unambiguously established by application of the standard Na dodecyl-SO, polyacrylamide gel electrophoresis system. This ambiquity was eliminated by coupling Na dodecylS0, polyacrylamide electrophoresis to the selective extraction of specific membrane proteins by the nonionic detergent Triton X-100.

Solubilization
and Separation of Membraneassociated Protein Kinase from Substrate Phosphoproteins-Cyclic AMP-dependent protein kinase was selectively solubilized by extraction of the membranes with 0.5% Triton X-100 in low ionic strength borate buffer (56 mM sodium borate, pH 8.0, p = 0.008, see Ref. 31). In addition to kinase activity 0.5% Triton X-100 eluted Coomassie blue-staining proteins III and VI as well as a portion of IVa (Fig. 5). The selectivity of the extraction procedure is illustrated by the data presented in Table IV (also see Fig. 5). Over 90% of cyclic AMP-stimulated protein kinase activity was released into the detergent-soluble fraction as compared to approximately 50% of the total membrane protein and 1% or less of the polypeptides comprising "spectrin" (I, II, and V). It is also interesting to note that the enrichment of cyclic AMP-dependent protein kinase in the soluble fraction was paralleled by a significant enrichment of cyclic AMP-independent casein phosphokinase activity (Table IV) in the insoluble pellet, suggesting the occurrence of at least two distinct phosphotransferases in erythrocyte membranes.
The insoluble membrane residue contained 5% of cyclic AMP-dependent protein kinase activity with respect to exogenous substrates and was unable to catalyze the phosphorylation of endogenous membrane proteins.
Solubilized protein kinase had kinetic properties and metal ion requirements similar to those of the membrane-bound enzyme * and appeared to contain a normal complement of catalytic and cyclic AMP-binding activities based on (a) the recovery of 87% of the total kinase units, and (b) a 15-fold stimulation of activity by an optimal concentration of cyclic AMP as compared to a 17-fold enhancement in the unextracted membranes ( Table V). The elution of the cyclic AMP-binding protein was documented and quantitated independently by direct measurement of cyclic AMP-binding activity (see "Experimental Procedures"). More than 90% of the cyclic AMP-binding activity was solubilized in 0.5% Triton X-100 (Table  IV). When photoaffinity-labeled membranes were extracted, 72% of the incorporated 3H was eluted into the 0.5% Triton X-100/56 mM borate supernatant fraction and the 15 to 20% decrease in extractability paralleled a 10 to 20% decrease in the solubility of membrane proteins in 0.5% Triton X-100 subsequent to ultraviolet irradiation.6 Decreased protein solubility may arise from ultraviolet irradiation-induced protein aggregation (35). Solubilization of catalytic and cyclic AMP-binding activities was unaffected by incubation of membranes with cyclic AMP or by phosphorylation of endogenous substrates (Table IV). In contrast to the behavior of cyclic AMP-dependent protein kinase "P-labeled phosphoproteins Pl and P3 remained associated with the Triton-insoluble membrane pellet while the radioactivity in P2 was distributed *between the soluble (33%) and particulate (67%) fractions (Fig. 5). The latter result suggests the possibility of additional heterogeneity among the intrinsic substrates for protein kinase. The retention of *2Plabeled P3 in the Triton X-IOO-insoluble pellet (Fig. 5) and the solubilization of the cyclic AMP-binding protein with 0.5% Triton X-100 (Table IV  were undertaken to effect the separation of these components and determine which polypeptides might be phosphate acceptors. As a first approach the relative mobilities of BT-phosphoproteins, Coomassie blue-stained proteins and Schiff-stained sialoglycoproteins were determined as functions of increasing acrylamide concentration and the data were plotted according to the method of Ferguson (Ref. 21 and Fig. 6). This method takes advantage of the fact that heavily glycosylated, PAS-positive erythrocyte membrane proteins bind considerably less Na dodecyl-SO, than other membrane proteins (36) and exhibit abnormally low free electrophoretic mobilities.
Therefore, sialoglycoproteins migrate aberrantly relative to other membrane proteins in Na dodecylS0, gels and yield unique Ferguson plots (see Fig. 6 and Refs. 22 and 36). Phosphoproteins P2 and P3 displayed the same retardation coefficients, K,, (defined as the slope of the Ferguson plot) as stained proteins III and IVc, respectively, and were clearly differentiated from PAS-l and PAS-2 (Fig. 6). Fig. 6 also shows (2/l) as described by Hamaguchi and Cleve (37). The aqueous layer contained all the sialoglycoproteins but no ST-labeled material; the insoluble pellet at the organicaqueous interface comprised the remainder of the membrane proteins, including protein kinase substrates, Pl, P2, and P3 (Fig. 7). Thus all the major phosphate acceptors are readily separable from the sialoglycoproteins.
Finally, Fig. 5 shows that only 33% of s*P incorporated into P2 may be associated with III while 67% was retained in a of radioactivity were carried out as described in "Experimental Procedures." greater than the difference in mobilities between the cyclic AMP-binding protein and P3. ' The cyclic AMP-binding protein was clearly shown to be separate and distinct from phosphoprotein P3 by the selective solubilization of (a) cyclic AMP-dependent protein kinase activity, (b) independently determined cyclic AMP-binding activity, and (c) the SH-labeled cyclic AMP-binding protein by 0.5% Triton X-100 in 56 mM borate buffer, and by the retention of SzP-labeled P3 in the Triton-insoluble pellet (Tables IV and  V and Fig. 5). Thus, the cyclic AMP-binding protein does not appear to be a major phosphate-acceptor in the erythrocyte membrane and definition of the functional role of P3 will require further investigation.
The current studies have not completely excluded the possibility that the binding protein is phosphorylated to a limited extent since phosphorylation of 10% of the binding protein would have resulted in the incorporation of less than 50 cpm under the assay conditions employed in these studies (using [y-3*P]ATP at a specific activity of 600 cpm/pmol, 60 pg of protein/gel).
Studies are now in progress to assess the possible phosphorylation of the cyclic AMP-binding protein in preparations of Triton-solubilized protein kinase. Several additional characteristics distinguish the phosphorylation of P3 from the autophosphorylation of regulatory subunits of cytoplasmic protein kinases: determination of the maximal incorporation of 3zP, into phosphoprotein P3 and estimation of the concentration of cyclic AMP-binding protein in saturation binding experiments would suggest incorporation of 4 or 5 mol of phosphate/m01 of binding protein'; and phosphorylation of P3 is relatively slow and highly dependent on cyclic AMP (Refs. 3 and 4 and Fig. 2). In contrast, the cyclic AMP-binding subunits of cytoplasmic protein kinases are rapidly phosphorylated in a cyclic AMP-independent fashion and a maximum of 1 mol of P, is incorporated per mol of binding protein subunit (19,20).
Solubilization of the cyclic AMP-binding protein in Triton X-100 was paralleled by the extraction of cyclic AMP-dependent protein kinase activity (Table IV). The protein kinase activity ratio, defined as protein kinase activity in the absence of cyclic AMP divided by kinase activity in the presence of cyclic AMP, is a reliable indicator of the degree of association of the cyclic AMP-binding and catalytic subunits of protein kinases (38). Data presented in Table V yield activity ratios (reciprocals of the stimulation factor) of 0.06 and 0.07 for the membrane associated and Triton-solubilized protein kinase preparations, suggesting that the subunits are highly associated in both forms of the enzyme. Little information is currently available regarding either physicochemical properties of membrane-associated protein kinases or their relationship to cytoplasmic protein kinases. The selectivity, high yield, and conservation of subunit association obtained with the Triton-borate extraction procedure may facilitate the purification and characterization of a membrane-associated protein kinase.
Previous investigations (3-5) did not discriminate among the possible phosphorylations of PAS-l, protein III, or minor components since all these proteins displayed the same mobilities on standard Na dodecyl-SO,-polyacrylamide gels (Fig. 2). The Ferguson plot (Fig. 6) and Fig. 7

DISCUSSION
Cyclic AMP-binding sites in the erythrocyte plasma membrane were covalently labeled using a highly specific photoaffinity analog of cyclic AMP, EtN,Mal-cyclic [SH ]AMP (Tables  I to III and Fig. 3). A unique polypeptide having a molecular weight of 48,006 was identified as the cyclic AMP-binding protein by determining the distribution of covalently bound *H among membrane proteins subsequent to labeling and electrophoresis in Na dodecyl-SO, (Fig. 4). The occurrence of a single binding protein and the observations that (a) membrane-associated and Triton-solubilized forms of protein kinase are highly stimulated by cyclic AMP (Table V), (b) detergent-solubilized catalytic and cyclic AMPbinding activities copurify on gel filtration, ion exchange chromatography, and sedimentation in sucrose gradients;' and (c) protein kinases are the only known receptors for cyclic AMP in mammalian cells (11) strongly suggest that the binding protein is a regulatory subunit of protein kinase. The molecular weight of the binding protein is also similar to the subunit molecular weights reported for the regulatory (cyclic AMPbinding) proteins of several cytoplasmic, cyclic AMP-dependent protein kinases (11).
The existence of a single cyclic AMP-binding protein does not preclude the presence of multiple species of cyclic AMPdependent protein kinase in the membrane. Such kinases could arise from the association of several distinct catalytic subunits with a single binding protein, or from the partition of the binding protein between monomeric and oligomeric forms.
Identification of this binding protein demonstrates the sensitivity and analytical power of the photoaffinity method since the cyclic AMP-binding protein accounts for only 0.07% of the total membrane protein, assuming one binding site per polypeptide chain, M, = 48,000, and 14 pmol of binding sites per mg of membrane protein (Table I). Furthermore, labeling of the cyclic AMP-binding protein represents a compromise between covalent labeling and irradiation-induced inactivation of binding activity, so that 8% of the available binding sites are labeled and 16% of the binding activity is lost after 4-min irradiation.
Further addition of fresh analog and repeated photolysis is ineffective because of binding site destruction, increased background labeling, and decreased solubility of membrane proteins in Na dodecyl-SO, after prolonged exposure to ultraviolet irradiation. Nevertheless, the labeling conditions employed in the present experiments (see "Experimental Procedures") led to incorporation of SH into a unique membrane protein and gave a 3H incorporated to SH background ratio of 10 in the peak gel slice (Fig. 4). The latter result represents a significant improvement over the ratio of 1.5 reported by Guthrow et al. (34).
In preliminary experiments Guthrow et al. (34) labeled erythrocyte membranes with EtN,Mal-cyclic [3H]AMP and proposed that the binding protein and P3 were identical, based solely on the migration of peaks of SH and aaP in separate experiments (3,34). It is unlikely that a limited number of separate, single isotope experiments would disclose the slight incongruity between the mobilities of the SH-labeled binding protein and SZP-labeled P3 observed in double isotope experiments (Fig. 4) Fairbanks and Avruch (39). They noted a cyclic AMP-dependent phosphorylation of a polypeptide slightly smaller than Pl when incubations were performed at low ionic strength, but found that this reaction was masked by the high level of cyclic AMP-dependent phosphorylation of Pl under conditions similar to those employed in the current studies.
Cyclic AMP-dependent phosphorylation of proteins with molecular weights corresponding to P2 or P3, or both, has been observed in membrane fractions derived from mammalian brain, smooth muscle, and toad bladder epithelium (40-42). It will be of considerable interest to determine whether the larger peak of 8zP represents single or multiple polypeptide species and whether smaller protein is separable from the cyclic AMP-binding protein.
In general, the approaches described here (photoaffinity labeling, separation of kinase and substrates, and subfractionation of phosphoproteins by detergent extraction) may be of considerable future use in establishing or eliminating possible functional roles of phosphoproteins in membranes.