Novel Isozymes of cAMP-dependent Protein Kinase Exist in Human Cells Due to Formation of RIa-RIP Heterodimeric Complexes*

We report that a human neoplastic B cell line (Reh) contains CAMP-dependent protein kinase (cAK) type I (cAKI), but is practically devoid of cAK type I1 (cAKII). However, these cells contain a novel cAKI isozyme con-sisting of an RIa-RIP heterodimer in association with phosphotransferase activity (RIaRIPC,) eluting from DW-cellulose columns at a salt concentration charac- teristic of a cAKII.

upon binding of two CAMP molecules to each regulatory (R) subunit proceeds by a concerted reaction resulting in an R subunit dimer with four CAMP molecules bound and two free active catalytic (C) subunits that phosphorylate specific serine and threonine residues on target proteins and thus mediate the action of CAMP (Corbin et al., 1978;Deskeland, 1978).
Two major isozymes of eukaryotic cAK, types I and I1 (cAKI and cAKII, respectively), were initially described by their pattern of elution from DEAE-cellulose columns (Reimann et al., 1971). Furthermore, the cAKI and cAKII holoenzymes (eluting at concentrations between 25 and 50 mM and 150 and 200 mM NaC1, respectively) were shown to contain different regulatory subunits, termed RI and RII (Beebe and Corbin, 1986). However, more recent reports have revealed a greater heterogeneity of cAK subunits. At present, four regulatory subunits (RIa, RIP, RIIa, and RIIP) and three catalytic (Ca, CP, and Cy) subunits have been reported (for references, see Scott (1991)). Splice variants of two of the C subunits (Ca2 and CPZ) add to this complexity (Thomis et al., 1992;Wiemann et al., 1991).
Apart from the different biochemical and functional properties, several lines of evidence support specific roles for the different cAK isozymes. Most cells, with some exceptions (Lange-Carter et al., 1990), have cAKI and cAKII; however, the proportion of cAKI to cAKII and the level of expression of the different subunits vary greatly between cells and tissues (Corbin et al., 1975). Furthermore, expression of cAK subunits is altered as a consequence of differentiation (Liu, 1982;Schwartz and Rubin, 1983), hormonal regulation (Jahnsen et al., 1985; 0yen et al., 1988), and mitogenic signal^.^ In addition, specific subcellular localization of cAK isozymes further supports the notion that specific functions are assigned to each isozyme. Type I1 cAKs have been reported to localize to the Golgi apparatus and centrosomes (Nigg et al., 1985;Keryer et al., 1993), probably due to interaction with RIIa-or RIIP-specific binding proteins (for review, see Scott (1991) and Scott and Carr (1992)). The cAKI holoenzyme (RIa2C2) is primarily R. Solberg Scott, T. Jahnsen, and S. S. Taylor, submitted for publication. Jahnsen, and T. Lea, manuscript in preparation. soluble and cytoplasmic  and releases the C subunit to be translocated to the nucleus upon activation Adams et al., 1991;Fantozzi et al., 1992), as does the more recently described cAKI However, compartmentalized effects mediated through cAKI have also been reported (Moger, 1991;Lanotte et al., 1991;Skilhegg et al., 1992b). Furthermore, we have recently shown that cAKI (RIa2Cz) in human T cells localizes to and associates with the T cell antigen receptor-CD3 complex during T cell activation and capping: indicating that cAKI can also be specifically localized to certain subcellular structures and mediate isozymespecific effects.
Since CAMP through cAK inhibits cell replication in normal (Blomhoff et al., 1987;Skilhegg et al., 1992b) and neoplastic (Blomhoff et al., 1987) human lymphocytes and cAKI (RIazCP2) specifically mediates the inhibitory effects of CAMP on the T cell antigen receptor-CD3-mediated T cell replication (Skilhegg et al., 1992b), we embarked on a study characterizing isomeric forms of cAK in a neoplastic B cell line (Reh). Reh cell proliferation is also counteracted by CAMP (Blomhoff et al., 1987), and this cell type provides an excellent model for studies of cAMP-mediated growth control without the implications of cell differentiation. Surprisingly, our results demonstrated that Reh cells are practically devoid of cAKII isozymes and that the apparent isozyme eluting in the position of type I1 cAK from DEAEkellulose columns represents a cAKI (RIaRIPCz). RIa-RIP heterodimeric complexes could also be demonstrated in normal human T cells employing specific site-directed antibodies. Furthermore, purified recombinant RIa and RIP proteins dimerized in vitro.

MATERIALS AND METHODS
Cell Cultures-A neoplastic B precursor cell line (Reh) derived from a patient with acute lymphoblastic leukemia was kindly provided by Dr. M. F. Greaves (Imperial Cancer Research Fund Laboratories, London). Cells were grown as suspension culture in RPMI 1640 medium supplemented with 10% heat-inactivated fetal calf serum (Sera-Lab Ltd., Sussex, United Kingdom), penicillin (100 IU/ml; Life Technologies, Inc., Paisley, Scotland), and streptomycin (100 pg/ml; Life Technologies, Inc.) at 37 "C in a humidified atmosphere of 5% CO,. The number of cells were determined in a Coulter Counter. Peripheral blood T lymphocytes were prepared as described in detail elsewhere (Skllhegg et al., 1992b).
Northern Blot Analysis-Extraction of total RNA was performed as described previously (Taskh et al., 1990). Reh cells (4 x lo7) were transferred to 50-ml tubes and pelleted by centrifugation at 500 x g for 15 min at 4 "C. The media were discarded, and cells were washed once in cold phosphate-buffered saline and repelleted. The cell pellets were then lysed directly by suspension in guanidium isothiocyanate, and total RNA was isolated by centrifugation through a cesium chloride gradient and purified by phenolkhloroform extractions. Total RNA (20 pg) from each sample was denatured in 50% (v/v) formamide, 6.0% formaldehyde at 50 "C for 15 min. T h e RNA samples were resolved on a 1.5% agarose gel containing 6.7% (v/v) formaldehyde in a 20 m M sodium phosphate buffer, pH 7.0. Quality of the RNA and migration of an RNA standard (0.3-9.5 kb; Life Technologies, Inc.) were assessed by ethidium bromide staining of the gel before transferring by capillary blotting technique to a nylon membrane (Biotrans, ICN, Irvine, CA). The membrane was baked (80 "C, 1 h), prehybridized, and hybridized as recommended by ICN. After prehybridization using single-stranded salmon sperm DNA to decrease nonspecific binding, the filters were hybridized at 42 "C in 50% formamide with a labeled cDNA probe (1 x lo6 cpm/ml). The membranes were washed four times in 2 x SSC, 0.1% SDS at room temperature for 5-10 min, followed by two washes at 50 "C in 0.1 x SSC, 0.1% SDS. Autoradiography was performed using Amersham Hyperiilm MP. Before rehybridizations, filters were washed in 50% formamide, 10 m~ sodium phosphate, pH 6.5, at 65 "C for 1 h.
Antibodies-A monoclonal antibody against human RIa (clone 4D7) was made by immunizing mice with purified human testis RI (SkBlhegg et al., 1992a) employing a protocol described in detail elsewhere (Mollnes et al., 1985). Mouse ascites was fractionated by protein Acoated Sepharose 4B chromatography, and the resulting IgG-purified monoclonal antibody (IgG,) was used at a dilution of 1:500 for radioimmunolabeling and 1:50 for immunoprecipitations.
Anti-human RIP antisera were made by immunizing rabbits with a hemocyanin-coupled synthetic peptide CNRQILARQKSNSQSDSHDE-NH2; Multiple Peptide Systems, San Diego, CA) corresponding to amino acids 61-79 of the human RIP sequence. The anti-RIP antisera were diluted 1:50-1:lOO for radioimmunolabeling and 1:15 for immunoprecipitations.
RII subunits (RIIa and RIIP) were recognized by highly specific antisera made against synthetic peptides from the amino-terminal end of the human RIIa and RIIp amino acid sequence^.^ RII antisera were diluted 1:lOOO for radioimmunolabeling and 1:50 for immunoprecipitations.
An anti-rat RIa antisemm produced against a synthetic peptide of rat RIa (a gift from Dr. H. S. Huitfeldt, The National Institute of Public Health, Oslo, Norway) was diluted 1:50 for immunoprecipitations (see Fig. 8).
A rabbit anti-C subunit antiserum raised against purified bovine heart C (kindly provided by Dr. S. M. Lohmann, Medisinische Universitatsklinik, Wurzburg, Germany) was used at a 1:200 dilution for radioimmunolabeling.
Preparation of Soluble Protein Extracts-For direct photoafinity labeling or immunoprecipitations, cells were homogenized for 2 x 15 s (Ultra Turrax, full speed) in buffer containing 10 m~ potassium phosphate, pH 6.8, 1 m M EDTA, 250 m M sucrose (PES buffer), 50 kallikrein inhibitor unitdml aprotinin (Sigma), and 10 pg/ml concentrations of each of the following protease inhibitors: chymostatin, leupeptin, antipain, and pepstatin A (Peninsula Laboratories, Inc., Belmont, CAI. For DEAE-cellulose chromatography, cells were homogenized in PES buffer with 0.1% Triton X-100 (Sigma) and centrifuged at 15,000 x g for 30 min at 4 "C, and the supernatants were used for fractionation.
DEAE-cellulose Chromatography-DEAE-cellulose columns (0.9 x 10 cm, 6.4 ml; DE52, Whatman) were equilibrated in 10 m M potassium phosphate buffer, pH 6.8, 1 m M EDTA (PE buffer). Cell extracts were applied in a total volume of 2 ml. Each column was then washed with 6 bed volumes of PE buffer followed by gradient elution (0-400 m M NaCl in PE buffer, total volume of 50 ml) of retained protein. Fractions of 1 ml were collected and immediately assayed for phosphotransferase activity and specific PHICAMP binding. NaCl concentrations in eluted fractions were determined by conductivity measurements (Hoelzle and Chelius, Neu-Isenburg, Germany). Fractions of interest were concentrated by ultrafiltration on Centricon-30 columns (4500 x g, 1 h, 4 "C; Amicon Corp.) for further analysis either by radioimmunolabeling or by photoaffinity labeling and immunoprecipitation.
cAMPBinding Activity-Determination of specific L3H]cAMP binding of soluble R subunits was carried out according to Cobb and Corbin (1988) in a solution containing [2,8-3HlcAMP (3 p~; specific activity of 15.7 Ci/mmol; Du Pont-New England Nuclear). Molar amounts of R subunits were calculated assuming two CAMP-binding sitedmonomer.
Phosphotransferase Activity of cAMP-dependent Protein Kinase -Catalytic activity of cAK was assayed by phosphorylating a cAK-B. F. Landmark, B. S. Skllhegg, K. Tasken, T. Jahnsen, and V.
Hansson, unpublished data. specific substrate (phosphate acceptor peptide, Peninsula Laboratories, Inc.) (Kemp et al., 1977) using [Y-~~PIATP (specific activity of 5000 CUmmol, 10 mCUml; Amersham Corp.) in an assay mixture described by Roskoski (1983). Calculation of the molar concentration of the C subunit was based on the specific activity of homogenous bovine heart C subunit (15 pmollmidmg). Phosphotransferase activity was measured both in the presence and absence of CAMP (5 PM). Kinase activity in the absence of CAMP (data not shown) was generally below 5% of the activity in the presence of CAMP.
Heterologous Expression of RIa and R I p A n NcoI-blunt end restriction fragment corresponding to nucleotides 103-1474 of the human RIa cDNA (Sandberg et al., 1987) was inserted into the vector pGEX-KG (Guan and Dixon, 1991), kindly provided by Dr. J. E. Dixon (Department of Biochemistry and Walther Cancer Institute, Purdue University, West Lafayette, IN). The resulting construct was sequenced by the dideoxy chain termination method (Sanger et al., 1977) to verify correct insertion of the fragment. Human RIa was expressed as a fusion protein with glutathione S-transferase in the Escherichia coli strain BL21/DE3 (Studier and Moffatt, 1986) and purified from bacterial lysate by absorption to glutathione-agarose beads as described in detail elsewhere (Smith and Johnson, 1988; Guan and Dixon, 1991L2 Subsequently, purified fusion protein was digested with thrombin, and glutathione Stransferase was absorbed on glutathione-agarose beads to yield soluble RIa including an amino-terminal extension of 15 amino acid residues from the glycine kinker segment of pGEX-KG. Heterologous expression and purification of recombinant human RIP2 were performed similarly to those of RIa and yielded a protein with an identical amino-terminal extension. Purified recombinant bovine RIa (Uon et al., 1991) had an amino-terminal extension of 10 amino acids from the lac2 gene of the plasmid pUC 7 (Saraswat et al ., 1986).
Immunoprecipitations-Preparation of Staphylococcus aureus cells (SACS) expressing protein A (Pansorbin cells, Calbiochem) for immunoprecipitations was performed as follows. One volume of SAC buffer (50 mM Tris-HCI, pH 7.5, 150 m M NaCI, 1 mM EDTA, 0.5% Triton X-100) containing 1% BSA, 0.5% SDS, and 2.5% Triton X-100 was mixed with 1 volume of SACs; sonicated for 3 x 30 s (sonicator model W-10, Ultrasonics, Plainview, N Y ) ; and pelleted by centrifugation (5000 x g, 5 rnin). Pellets were resuspended by sonication (3 x 30 s) in 1 volume of SAC buffer containing 1% BSA. One additional wash was performed with 1 volume of SAC buffer with no addition, and finally, SACS were resuspended by sonication in 0.9 volume of SAC buffer. Samples (labeled R subunits or cell extracts) were adjusted to a volume of 200 pl in buffer containing phosphate-buffered saline, 2 m M EGTA, 2 m M EDTA, 2.5% Triton X-100, 50 kallikrein inhibitor unitdm1 aprotinin, and 10 pg/ml concentrations of each of the antiproteases chymostatin, leupeptin, antipain, and pepstatin A. SACS (250 pl) were added to the samples, which were incubated by shaking for 20 min a t 20 "C and thereafter centrifuged (15,000 x g, 3 mid. Supernatants were transferred to new tubes, and antibody was added. Samples were then incubated by shaking for 90 min a t 20 "C, after which 75 pl of SACs was added, and the incubations were continued for another 60 min. Antigen-antibody complexes bound to protein A were then pelleted (15,000 x g, 3 mid, and pellets were washed twice in SAC buffer with additional 0.5% Triton X-100. Washed pellets were resuspended in 80 pl of SDS sample buffer and boiled for 5 min, and SDS-PAGE was performed according to Laemmli (1970) employing 4.5% stacking gels and 7.5% separating gels. Coomassie Blue-stained, k e d , and dried gels were subjected to autoradiography using Hyperfilm MP and SuperRapid intensifying screens (Kodak Eastman, Rochester, N Y ) .
Reduction and Alkylation--To disrupt disulfide bridges and to prevent reassociation of R subunits, 8-a~ido-[~~P]cAMP photoafinity-labeled crude cell extracts (400 pg) were treated with the reducing agent dithioerythritol(1 mM; Sigma) for 30 min a t 20 "C followed by treatment with the alkylating agent N-ethylmaleimide (NEM) (20 mM; Sigma) for 30 min at 20 "C. Alkylation with NEM also neutralized the reductive potential of the remaining dithioerythritol and prevented the destruction of antibodies in the subsequent immunoprecipitations.
In Vitro Dimerization of Purified Recombinant RI Subunits-Purified recombinant RI subunits (50 ng) were labeled with 8-a~ido-[~~P]cAMP. Subsequently, a 5-10-fold molar excess of unlabeled subunit was added, and samples were incubated in buffer containing 20 m M potassium phosphate, pH 7.4,2 mM EDTA, 1 mM MgC12, 5% glycerol, 0.01% BSA, and 100 m M P-mercaptoethanol for 30 min a t 20 "C to reduce disulfide bridges and to produce monomeric R subunits. Samples were dialyzed against the same buffer without p-mercaptoethanol and BSA for 16 h at 4 "C to remove P-mercaptoethanol and to optimize redimerization. Following dialysis, samples were divided equally in two, and immunopre-cipitations with antibodies against both RIa and RIP were performed. Concentrations of antibodies and SACs were doubled in immunoprecipitations (compared to the protocol described above) to accommodate the high concentrations of unlabeled RI subunit present.

Subunits of &in a Human Lymphoid B Cell Line (Rehl-To
determine which of the cAK subunits were expressed at the mRNA level in Reh cells, total RNA was extracted from cells in culture, and Northern analysis was performed. Fig. 1 shows the resulting filter after hybridization with probes against human cAK R and C subunits. The subunits RIa, RIP, RIIP, Ca, and CP were detected. The levels of RIa and Ca mRNAs were comparably higher (exposure of 36 h) than the levels of RIP, RIIP, and Cp mRNAs (exposures of 12, 11, and 5 days, respectively). The RIIa and Cy mRNAs could not be detected, even after very long exposures (21 days).
Levels of immunoreactive cAK subunits in Reh cells were next assessed by Western analysis (Fig. 2). Immunoreactive C (antiserum does not distinguish Ca and CP), RIa, and RIP subunits were readily detected and are shown in comparison to purified standards (25 ng) (Fig. 2). In contrast, no RII subunits were detected even after prolonged exposures of the blots shown in Fig. 2.
To investigate further the possible presence of RII subunits in this cell line, extracts from Reh cells were fractionated on DEAE-cellulose columns eluted with a linear gradient of NaCl (0-400 mM). The resulting elution profile (Fig. 3-4) shows two peaks of specific CAMP binding and phosphotransferase activ- The mobilities of the ribosomal RNAs (18 S and 28 S) have also been indicated.

FIG. 2. Immunoreactive cAK subunits expressed in Reh cells.
Reh cells (2.5 x lo6) corresponding to 100 pg of protein were boiled directly in SDS sample buffer, resolved by SDS-PAGE on 7.5% gels, and blotted onto nitrocellulose. Western analyses were performed employing antibodies directed against the C subunits, RIa, RIP, RIIa, and RIIP (see "Materials and Methods" for details). Immunoreactive proteins were visualized by 1251-protein A and autoradiography. Standards were run in parallel lanes (Std. ) and represent purified bovine heart C subunit (25 ng) (anti-C), purified human testis RIa subunits were calculated. The conductivity was measured in every second fraction, and the concentrations of NaCl were calculated (0). B , the DEAE fractions indicated were concentrated by ultrafiltration, subjected to SDS-PAGE, and blotted onto nitrocellulose, and Western blots were prepared employing antibodies directed against RIa and RIP (both antibodies together). Standards were purified human testis RIa (50 ng) and recombinant human RIP (50 ng). NaC1. After concentration of DEAE fractions corresponding to peak I (5-fold) and peak I1 (eO-fold), proteins were separated by SDS-PAGE, and radioimmunolabeling was performed on parallel blots employing antibodies directed against RIa and RIP (Fig. 3B, both antibodies together) or antibodies against RIIa or RIIP (data not shown). RIa was the only R subunit present in peak I (Fig. 3B, DEAE fraction 9). Furthermore, a reappearance of immunoreactive RIa was seen in peak I1 together with RIP ( Fig. 3B, DEAE fractions 19-29). Prolonged exposures of parallel immunoblots with RIIa or RIIP antiserum indicated that RII subunits were below the level of detection both in peaks I and 11. The reappearance of RIa together with RIP in peak I1 from DEAE-cellulose columns suggested the existence of a cAKI holoenzyme (RIaRIpC2) resulting from the binding of two C subunits to an RIa-RIP heterodimer in Reh cells. The To precipitate antigen-antibody complexes, SACS expressing protein A were added, and antigen-antibody complexes bound to protein A were pelleted by centrifugation (see "Materials and Methods"). SAC pellets were washed and subsequently boiled in SDS sample buffer, and immunoprecipitated protein samples were subjected to SDS-PAGE and autoradiography.
(150-200 m M NaCl) is clearly different from that usually described for cAKI (RIaZC2; 25-50 m M NaCI), but similar to that found for ClWI.
Specificity of RIaIRIP Antibodies in Immunoprecipitations " T o examine the possible presence of RI heterodimeric complexes more closely, an immunoprecipitation assay employing antibodies against RIa and RIP was established. To assure specificity in immunoprecipitations, human RIa was expressed in E. coli (see "Materials and Methods" for details), purified, and labeled with 8-a~ido-[~~P]cAMP together with purified recombinant human RIP.2 Immunoprecipitations were performed with antibodies against both purified recombinant RI subunits (Fig. 4). As seen in Fig. 4, human RIa was immunoprecipitated only with the anti-human RIa monoclonal antibody. Similarly, human RIP was immunoprecipitated only with the anti-RIP antiserum, and no cross-reactivity between the antibodies could be observed. Furthermore, no cross-reactivity with either of these antibodies could be observed toward in vitro translated, [35Slmethionine-labeled RIIa and RIIP pro-teins6 Fig. 5A (left lane) shows direct photoaffinity labeling of crude Reh cell extracts followed by SDS-PAGE, revealing two CAMP-binding proteins (49 and 55 kDa). The major band represents a 49-kDa protein that co-migrates with RIa, whereas the minor band represents a larger protein (55 kDa) with migration identical to that of RIP. The labeling of both proteins was counteracted by addition of a 100-fold excess of unlabeled CAMP as a competitor (-and + lanes). Fig. 5 further shows the specific immunoprecipitation of 8-a~ido-[~~P]cAMP-labeIed RIa with the RIa antibody (anti-RIa lane) giving a strong signal representing a protein with an apparent molecular mass of 49 ma. When weak exposures were examined closely, a faint band of higher molecular mass could also repeatedly be observed to be immunoprecipitated with the same antibody, but was easily overshadowed by the strong signal from the RIa protein. In contrast, immunoprecipitations with the antiserum against RIP (anti-RIP) revealed two bands of equal intensity (Fig. 5A,  right lane). The upper band co-migrated with labeled recombinant RIP, whereas the lower band co-migrated with labeled recombinant RIa. Furthermore, when photoafinity-labeled cell extracts were treated with the reducing agent dithioerythritol (1 mM, 30 min, 20 "C) followed by the alkylating agent NEM (20 mM, 30 min, 20 "C), only the larger protein could be immunoprecipitated with the RIP antibody (Fig. 5B). After reduction- alkylation, labeled extracts were also simultaneously immunoprecipitated with the RIa antibody, showing that reductionalkylation did not alter the immunoreactivity of RIa (data not shown). Photoaffinity-labeled R subunits in concentrated DEAE fractions corresponding to peak I (25-50 m M NaCl) and peak I1 (150-200 m NaCl) were immunoprecipitated (Fig. 6). Direct electrophoresis and autoradiography of the labeled fractions from both peaks revealed a number of nonspecific bands that were present both in the absence (-lanes) and presence (+ lanes) of excess unlabeled CAMP. In contrast, immunoprecipitation followed by electrophoresis revealed only RIa in peak I, and no radioactive protein was seen after immunoprecipitation with the RIP antibody (Fig. 6, upper panel). However, in peak 11, separation by SDS-PAGE of immunoprecipitates employing the anti-RIP antiserum revealed two equally strong bands labeled with 8-a~ido-[~~P]cAMP (Fig. 6, lower  panel). Close examination of the immunoprecipitates from peak I1 using the anti-RIa antibody also shows the presence of two differently sized proteins, although the lower band is much stronger (due to free RIa homodimer that is trailing into peak 11).

RIa-RIP Heterodimeric Complexes in Crude Reh Cell Homogenates-
In Vitro Dimerization of RI Proteins-Purified recombinant human RIa and RIP were labeled separately with 8-azido-[32P]cAMP; a 10-fold molar excess of the opposite unlabeled subunit was added; and the mixtures were treated with p-mercaptoethanol (100 mM, 30 min, 20 "C) to produce monomeric R subunits. Mixtures were then dialyzed to remove p-mercaptoethanol and to allow dimerization. Control experiments without addition of excess unlabeled subunit were performed in parallel. All reactions were analyzed by separate immunoprecipitations with the RIa and RIP antibodies. Fig. 7 (middle  panel) shows experiments in which the human RIa protein was used in combination with the human RIP protein. Control experiments (lanes 5-8) revealed no cross-reactivity of the antibodies. When antibodies directed against the labeled subunits were applied, strong bands could be detected (lanes 5 and 71, whereas no signals were observed when antibodies against the

FIG. 6. Immunoprecipitation of photoafRnity-labeled R subunits in peaks I and I1 from DEAE-cellulose chromatography.
Cell extracts from Reh cells were fractionated by DEAE-cellulose chromatography, and fractions corresponding to peaks I and I1 were pooled and concentrated by ultrafiltration. Small aliquots were used for direct photoaffinity labeling with 8-azido-["P]cAMP in the absence (-) or presence (+) of excess cAMP (100 p~) . The rest was labeled in the absence of CAMP, and equal amounts were used for immunoprecipitations with antibodies against RIa (anti-Rla) and RIP (anti-RIP). Imraphy together with directly labeled samples (-and + lanes) and 10 ng munoprecipitates were analyzed by SDS-PAGE followed by autoradiog- opposite subunits were used (lanes 6 and 8). When dimerization was performed with both recombinant subunits present, we demonstrated that radioactive RIa incubated with unlabeled RIP is precipitated by the anti-RIa antibody (lane 1). However, the RIP antibody failed in coimmunoprecipitating labeled RIa (lane 2). Furthermore, radioactive labeled RIP incubated with unlabeled RIa was readily precipitated by the RIP antibody (lane 3 ) . In this case, coimmunoprecipitation of labeled RIP by the anti-RIa antibody was also observed (lane 4 ) . Heterologous expression and purification of human RIa yielded a protein with an amino-terminal extension of 15 amino acid residues from the vector pGEX-KG. Such an extension could possibly interfere with the binding of the anti-RIP antibody to a heterodimeric complex of human RIP and human RIa since the peptide used to generate this antibody (amino acids 61-79) is in close proximity to the part of the RI molecule assumed to be responsible for dimerization (amino acids 1 4 5 ) (Scott, 1991). Therefore, purified recombinant bovine RIa that had a shorter amino-terminal extension (10 amino acid residues) was labeled with 8-azid0-[~~P]cAMP, and dimerization was allowed to proceed as described above (Fig. 7, lower panel).
This combination allowed the coimmunoprecipitation of labeled RIa with the RIP antibody when incubated in the presence of excess unlabeled RIP (lane 21, whereas no signal was detected in the control experiment in which no RIP had been added (lane 6). In the inverse experiment (labeled RIP, unlabeled RIa, and anti-RIa antibody), we were unable to precipitate the labeled RIP subunit with our RIa antibody due to reduced cross-species reactivity between bovine RIa and the mouse monoclonal antibody against human RIa. However, when labeled bovine RIa and labeled human RIP were allowed to dimerize in the presence or absence of a &fold excess of the opposite unlabeled subunit and the monoclonal antibody against human RIa was substituted with a polyclonal anti-rat RIa antiserum raised in rabbits against a synthetic peptide of rat RIa (Fig. 8), the formation of heterodimeric complexes could be detected with both antibodies (lunes 2 and 4 ) . In this case, the RIa antibody was able to precipitate the labeled RIP protein and vice versa. In the control experiments, immunological cross-reactivity was not observed (lunes 6 and 8).

DEAE-cellulose
Columns- Fig. 9 shows specific [3H]cAMP binding and phosphotransferase activity eluting from DEAEcellulose columns a t different concentrations of NaCl. The elution profile from Reh cells (Fig. 9A) is compared to the elution profile from human peripheral blood T lymphocytes (Fig. 9B), which are known to contain cAKII (Skllhegg et ul., 1992b). Fig. 9C represents the elution profile from a mixture of Reh cells and T cells. Peak I1 elutes at 150-200 m M NaCl in all three experiments, which were performed sequentially (Fig. 9, A-C). Immunoprecipitations of pooled, concentrated, and 8-azido-  (lanes 1, 2, 5, and 6) or RIP (lanes 3, 4, 7, and 8 ) subunit was photoafinity-labeled with 8-azid0-[~~PJcAIvlF'. Subsequently, a 10-fold molar excess of the opposite unlabeled subunit (RIP, lanes 1 and 2; RIa, lanes 3 and 4 ) was added. Controls (lanes 5-8) received no addition of unlabeled subunit. Reactions were incubated in buffer containing 20 m M potassium phosphate, pH 7.4, 2 mM EDTA, 1 m M MgC12, 5% glycerol, 0.01% BSA, and 100 m M P-mercaptoethanol for 30 min at 20 "C and dialyzed against the same buffer without p-mercaptoethanol and BSA for 16 h a t 4 "C. Following dialysis, all samples (corresponding to lanes 1 and 2, lanes 3 and 4, lanes 5 and 6, and lanes 7 and 8) were split in two, and antibodies against either RIa (aRIa,lanes I ,4 ,5 ,and 8 ) or RIP (aRIP, lanes 2,3, 6, and 7) were added for immunoprecipitations. Immunoprecipitated subunits were separated by SDS-PAGE and subjected to autoradiography. Upper panel, experimental regimes; middle panel, combination of human recombinant RIa and human recombinant RIP; lower panel, combination of bovine recombinant RIa and human recombinant RIP. Co-migration of cAKI (RIcrRIpC2) from Reh cells with cAKII (RIIa2C2) from human T cells on DEAE-cellulose columns could also be demonstrated in Fig. 9C, where immunoprecipitation revealed the presence of RIa-RIP heterodimeric complexes (lunes 1 and 2 ) as well as RIIa (lune 3 ) .

DISCUSSION
In this study, we have examined the expression of cAK subunits in a neoplastic B cell line (Reh) and demonstrate that a cAK isozyme eluting as cAKII from DEAE-cellulose columns represents RIa-RIP heterodimeric complexes binding catalytic activity.
Reh cells reveal high level expression of RIa and Ca mRNAs. Lower levels of RIP and CP mRNAs are also present as well as low levels of RIIP mRNA. Western analysis of whole cell lysates readily demonstrates immunoreactive C subunit and RIa and small amounts of RIP, whereas neither RIIa nor RIIP could be detected. Thus, in spite of small amounts of RIIP mRNA present, immunoreactive RIIP is, on Western analysis, below the level of detection.
Our initial fractionations of Reh cell extracts on DEAE-cellulose columns indicated the presence of both cAKI and cAKII in addition to some free RI subunits eluting in between these two peaks. However, Western analysis of the various fractions gives no indication of cAKII in any of the peaks. Whereas the first peak contains immunoreactive RIa, the second peak consists only of immunoreactive RIa and RIP, which show an increase through the area where cAKII is expected to elute. The reappearance of RIa together with RIP and associated phosphotransferase activity raised the possibility of RIa-RIP heterodimeric complexes associated with catalytic activity. Employing monoclonal or antipeptide antibodies entirely specific for the regulatory subunits in question, we examined whether immunoprecipitation of RIa would coprecipitate RIP or vice versa. This was performed in cell lysates in which the regulatory subunits were labeled with 8-a~ido-[~~P]cAMP. Because of the large excess of RIa compared to RIP, it was difficult to see a distinct band representing RIP in the anti-RIa immunoprecipitates. However, the anti-RIP antibody readily precipitates both RIP and RIa subunits in stoichiometric amounts.
It has previously been shown that RIa homodimer is established through interchain disulfide bonds between cysteines 16 and 37, giving an antiparallel alignment of the two strands (Zick and Taylor, 1982;Bubis et al., 1987). Similarly positioned cysteines are also present in the human RIP protein (Solberg et al., 1991). The fact that reduction combined with alkylation prevented the coprecipitation of the RIa subunit with the RIP antibody clearly demonstrates that RIa-RIP heterodimer formation involves disulfide bounds. In addition, phosphorylation/ dephosphorylation with the catalytic subunit of cAK or alkaline phosphatase does not alter the mobilities of the precipitated bands on SDS-PAGE (data not shown), excluding the possibility that the two bands observed represent a phospho and a dephospho form of the RIP subunit or that the 8-azido-[32P]cAMP-labeled bands observed in the immunoprecipitates could represent RII subunits. The demonstration that significant amounts of RIa can be precipitated by the RIP antibody in peak I1 (but not in peak I) from the DEAE-cellulose columns further proves that the RIa-RIP heterodimeric complexes elute as cAKII. Furthermore, mouse RIP (free .dimer) expressed in transfected cells as well as from mouse brain elutes at a higher salt concentration than RIa, but clearly in front of cAKII (Clegg et al., 1988).
To demonstrate heterodimer formation directly, we used recombinant RIa and RIP both in a homologous (human) and a heterologous (bovinehuman) system. We succeeded in showing that heterodimeric complexes are formed in both systems. Radioactive RIP can readily be coprecipitated with the RIa antibody in the homologous system (human), but the RIa subunit is not precipitated with the RIP antibody. In contrast, the opposite was found with the bovine RIdhuman RIP system. In this case, the RIP antibody clearly coprecipitates radioactive RIa, but not the other way around. The reason why the anti-RIP antibody is unable to precipitate RIa is probably related to the fact that the human RIa subunit used in these in vitro dimerization studies contains an amino-terminal extension of 15 amino acids. This may cause a steric hindrance for the RIP antibody, which is raised against a peptide sequence (amino acids 61-79) very close to the dimerization zone in human RIP.
The fact that bovine RIa, with a shorter amino-terminal extension, can be precipitated with the RIP antibody tends to support this conclusion. The finding that the monoclonal antibody against human RIa is unable to precipitate the RIP subunit in the bovine RIdhuman RIP system probably reflects reduced cross-species immunoreactivity. This conclusion is supported by our observations that a peptide antibody against rat RIa easily precipitates the RIP subunit in similar experiments.
In our studies on in vitro dimerization of recombinant RIa and RIP subunits, we demonstrate that efficient dimerization occurs at a ratio of labeled to unlabeled subunit of 1:5-1:lO. Efficient dimerization was also observed at ratios of unlabeled to labeled subunit close to unity (data not shown). This indicates that heterodimerization actually proceeds fairly easily and not only in cases in which one of the subunits is in great excess. The fact that almost all of the RIP subunits in Reh cells appear to be present as RIa-RIP heterodimeric complexes further indicates that RIa-RIP heterodimerization proceeds quite easily.
The precipitation of RIa by the RIP antibody and vice versa is not due to immunological cross-reactivity, but to coprecipitation of RIa due to RIa-RIP heterodimer formation. The strict 1:l stoichiometry after 8-a~ido-[~~P]cAMP labeling and immunoprecipitation, complete lack of immunological cross-reactivity in immunoprecipitation controls, as well as the fact that immunoprecipitation of RIa by the RIP antibody only and exclusively could be seen in DEAE fractions from the peak I1 area and never in the peak I area, in spite of the fact that far greater levels of RIa are present, strongly support this conclusion. In addition, coimmunoprecipitation of labeled RIa with the RIP antibody is possible only when unlabeled RIP that can react with the antibody is present. In control experiments in which no RIP is present, no cross-reactivity is observed, although large amounts of RIa is present.
The functional consequences of RIa-RIP heterodimer formation are unknown. However, the fact that such dimers are associated with kinase activity clearly indicates that they are functional and activated by CAMP. RII subunits are shown to interact with anchoring proteins localizing phosphotransferase activity in close proximity to relevant substrates (Scott, 1991;Scott and Carr, 1992). We have recently demonstrated, both by immunocytochemistry and immunoprecipitation, that the RIa subunit binds to and redistributes with the T cell antigen receptor-CD3 complex during T cell activation and ~a p p i n g .~ Any functional proteins binding to the RIP subunit have so far not been demonstrated. Formation of RIa-RIP heterodimers certainly adds to the complexity of cAK isozymes and represents a new isozyme (RIaRIPC2) present both in Reh cells and in normal T cells as well as testicular tissue (data not shown). This adds to the diversification of possible CAMP-mediated effects in cells containing this isozyme.