Phosphorylation of CCAAT-enhancer binding protein by protein kinase C attenuates site-selective DNA binding.

Four DNA-recombinant proteins, corresponding to the DNA-binding domain of CCAAT/enhancer binding protein (C/EBP), were phosphorylated in vitro by protein kinase C (PKC). High-performance liquid chromatography-peptide mapping of 32P-labeled C/EBP indicated the presence of three major 32P-labeled peptides: S299 (P)RDK, AKKS277 (P)VDK, and GAAGLPGPGGS248 (P)LK. Phosphorylation of C/EBP by PKC or M-kinase resulted in an attenuation of binding to a 32P-labeled CCAAT oligodeoxynucleotide. Three other truncated forms of C/EBP, C/EBP87, C/EBP87S-C, and C/EBP60, were studied to define the sites of phosphorylation affecting DNA binding. Phosphorylation of the C/EBP87, containing sites Ser299 and Ser277, and C/EBP60, containing only site Ser299, by PKC also resulted in attenuation of DNA binding. In contrast, phosphorylation of C/EBP87S-C, which retained Ser277 but had a Cys in place of Ser299, had no effect on DNA binding. Ser299 could not be phosphorylated by PKC if the protein is already bound to specific DNA. Phosphorylation of intact C/EBP from liver nuclear extract by PKC or M-kinase occurred at Ser299 and Ser277 and at an additional site, as demonstrated by immunoprecipitation and peptide mapping.

phorbol esters (11,12), the mechanism of signal transduction from the phorbol esters to the transcriptional process remains elusive. Recently, evidence for a nuclear PKC (rat liver) was reported (13), suggesting a possible PKC-mediated phosphorylation of nuclear transcription factors. Although many transcription factors are potential PKC substrates, e.g. SP1, SREF,GCN4,, based on the presence of Thr or Ser residues flanked by basic residues, there is little evidence to date that they can be phosphorylated by PKC.
CCAAT/enhancer binding protein (C/EBP) is a member of the bZIP family of DNA-binding proteins. These proteins are characterized by a leucine repeat-dimerization interface immediately preceded by a region rich in basic amino acids that is believed to constitute the DNA contact surface (for reviews see . The presence of potential PKC phosphorylation sites in C/EBP, coupled with the colocalization of C/ EBP and PKC in rat liver nuclei raised the possibility that C/EBP may be a substrate of PKC. We report here the in vitro phosphorylation by PKC of proteins corresponding to the DNA-binding domain of C/EBP and holo-C/EBP. We have identified the sites of PKC-mediated phosphorylation of C/EBP and have studied the subsequent effect that modification exerts on sequence-specific DNA binding. Our results suggest that site-selective phosphorylation at SerZg9 of C/EBP results in attenuation of the DNA binding.
Methods-PKC was purified from rat brain by the procedure of Huang et al. (24). Catalytic fragment ("kinase) of trypsinized type I1 PKC was purified by chromatography on a Mono Q column as previously described (25). PKC activity was measured in 30 mM Tris-C1 (pH 7.5), 6 mM magnesium acetate, 10 pM [Y-~*P]ATP (28 Ci/ mmol), +400 p~ CaCl,, +40 pg/ml PS, +8 pg/ml DAG, k1.25-9 pg/ ml PKC, and variable concentrations of substrates. Reactions were carried out at 30 "C in final volumes of 20-40 p1 and incorporation of '"P was determined, as previously described, using 30% trichloroacetic acid containing 25 mM ATP as the stop solution and 20% trichloroacetic acid containing 0.2 M KC1 as the chromatographic solvent (26). In the case of the smaller C/EBP proteins (C/EBP87, C/EBP87S-C, and C/EBP6O), incorporation of 32Pi was determined by using P81 phosphocellulose paper (27). In the case of gel retardation and DNase I protection assays, phosphorylation reactions were stopped by dilution with an equal volume of 20 mM EDTA, pH 7.5, containing 20% glycerol. SDS-polyacrylamide gel electrophoresis was by the procedure of Laemmli (28).
Truncated C/EBP proteins (3 pM), phosphorylated with PKC or "kinase using [Y-~'P]ATP, were precipitated with 2 volumes of 30% trichloroacetic acid containing 25 mM ATP. Pellets were washed with 20% trichloroacetic acid, resuspended in 0.2 N NaOH, reprecipitated with 20% trichloroacetic acid, and washed with ice-chilled ether. Protein pellets were then successively cleaved with CNBr and endoproteinase Lys-C as follows. CNBr cleavage (100-fold molar excess over total Met) was carried out in 70% formic acid for 16-24 h at room temperature (22-pl final volume). The reaction was terminated by evaporating the formic acid with an air stream in a fume hood; the pellet was mixed with 22 pl of formic acid and evaporation was repeated residual CNBr was then sublimated (70 "C, 5 rnin). CNBrcleaved protein was digested with endoproteinase Lys-C (l/lO, w/w) in 100 mM NH4HC03, pH 9.0, for 4 h a t room temperature (22-pl final volume) and directly analyzed by peptide mapping. HPLCpeptide mapping was carried out employing a Vydac 218TP54 (C-18, 5 pm, 300 A, 25 X 0.46 cm) column and elution was by a 0.25%/min gradient from solvent A (0.1% trifluoroacetic acid) to solvent B (0.1% trifluoroacetic acid in CH3CN) a t a flow rate of 1 ml/min. Fractions (1 ml) were collected and counted in a liquid scintillation counter. '"P-Labeled peptide peaks were pooled, dried in a Speed Vac, and their amino acid sequences were determined using a gas-phase protein sequencer (Applied Biosystems Model 470A). Void volume radioactive peaks from the C-18 HPLC chromatography were further purified by drying the pooled peaks in a Speed Vac; pellets were resuspended in 10 mM NH,HCO,, pH 7.5, and applied to a Sep-Pak NHz cartridge. Elution of 32P-labeled peptides was by sequential step-wise elution with 0.5 ml each of 10,25,50,100,150,200,250,300,400, and 500 mM NH4HCO3, pH 7.5. A single broad radioactive peak in the 25-400 mM NH4HC03 concentration range was divided into two separate pools (50-150 and 200-300 mM NH,HCOJ, corresponding to the leading and tailing edge, for amino acid sequence analysis. Palindromic oligodeoxynucleotides (28 and 20 residues in length) corresponding to an avid C/EBP binding site, 5"AAGCTGCA-GATTGCGCAATCTGCAGCTT-3' and 5"TGCAGATTGCGCAA-TCTGCA-3' (20), were synthesized using an Applied Biosystems 380B DNA synthesizer. The correct sizes were confirmed and the purity of the synthesized oligodeoxynucleotides were >95% as determined by analysis on a DNA-sequencing gel. The oligodeoxynucleotides were annealed in TEK buffer (10 mM Tris-C1, pH 7.9, containing 1 mM EDTA, and 50 mM KCI) and then labeled with 32P using T, polynucleotide kinase. Oligodeoxynucleotides (140 ng) were labeled in the presence of 50 mM Tris-C1 (pH 7.5), 10 mM MgCI,, 5 mM DTT, 50 pg/ml bovine serum albumin, 20 pCi of [Y-~'P]ATP (>5000 Ci/ mmol), and 10 units of T, polynucleotide kinase in a final volume of 20 yl for 30 min at 37 "C. Radiolabeled oligodeoxynucleotides were purified by applying the reaction mixture and 20 p1 of the reactiontube rinse (TEK buffer) to a NICK column equilibrated in TEK buffer. The NICK column was washed with 0.4 ml of TEK buffer, and the "2P-labeled oligodeoxynucleotide was then eluted with 0.5 ml of the same buffer.
The binding of truncated C/EBP protein to 32P-laheled DNA substrate was carried out in the presence of 0.28 ng of 32P-labeled oligodeoxynucleotide, 20 mM HEPES (pH 7.9), 10% glycerol, 50-500 mM KCl, 0.1 mg/ml bovine serum albumin, 1 mM DTT, and varying concentrations of truncated C/EBP protein, in a final volume of 20 p1 for 30 min at room temperature. In the case of the C/EBP87S-C protein, 25 mM DTT was employed. Reaction mixtures (15 p l ) were then immediately loaded on a 5% native polyacrylamide (29:1, acrylamide:bisacrylamide) gel, containing 50 mM Tris, 50 mM boric acid, and 1 mM EDTA, pH 8.3 (TBE). Electrophoresis was carried out with TBE buffer as electrolyte for approximately 1 h at 15 mA per mini-gel (5 X 5 X 0.15 cm) until bromphenol blue migrated to within 2 cm of the bottom of the gel. Gels were then sealed in plastic wrap and autoradiographed for detection of free and retarded 32P-labeled DNA.
DNase I protection assays were performed with a 130-bp (EglII/ KpnI) (32P-end labeled) DNA fragment containing high affinity C/ EBP binding sites as describedpreviously (32). Briefly, the 32P-laheled DNA fragment was incubated with truncated C/EBP proteins (10 ng) for 10 min and digested with a 500-fold dilution of 1 mg/ml DNase I for 1 min. Reactions were stopped with buffer containing 1% SDS and 0.1 mg/ml proteinase K and analyzed on a sequencing gel.
Rat liver nuclear extract (5 mg of protein), prepared by the method of Gorski et al. (33) in the presence of 10 pg/ml aprotinin, 1 pg/ml leupeptin, 1 pg/ml pepstatin, and 1 pg/ml chymostatin, was heattreated (80 "C, 5 min) and precipitatedprotein (>go% of total protein) was removed by centrifugation (32). The supernatant was concentrated to 40 pl using a Centricon 10 and was phosphorylated with Mkinase in the presence of 30 mM Tris-C1 (pH 7.6), 6 mM magnesium acetate, 100 p~ [ Y -~~P I A T P (230 pci), in a final volume of 88 p1 (37 "C, 15 rnin). The reaction was stopped by the addition of 0.1 volume of 100 mM EDTA (pH 7.5) and 32P-labeled C/EBP was immunoprecipitated using anti-serum generated against a synthetic 14-amino acid peptide or holo-C/EBP expressed in and purified from E. coli as described (31). Immunoprecipitates were resuspended in 2fold SDS sample buffer, boiled for 5 min, and resolved on a 14% SDS-polyacrylamide gel. Proteins in the gel were then electroblotted (30 W, 2 h) to Immobilon membrane (34). Membrane corresponding to 32P-labeled C/EBP was cut out, incubated with 1 mg bovine serum albumin in 100 pl of 20 mM Tris-C1 (pH 8.0), 137 mM NaCl (TBS, 1 h, at room temperature) and successively washed with TBS containing 0.05% Tween 20 (TBST, 5 min X 2) and TBS (5 min X 2). The protein on the membrane strip was then digested with 150 mM CNBr in 70% formic acid (150 p l , 16-24 h, at room temperature). The reaction was terminated by evaporating the formic acid/CNBr with an air stream in a fume hood, peptide fragments were washed with 70% formic acid, and the evaporation was repeated. Residual CNBr was sublimated (70 "C, 5 min). CNBr-generatedpeptides were further cleaved with 1 pg of endoproteinase Lys-C in 100 mM NH,HC03 (pH 9.0) in a final volume of 100 pl (4 h, at room temperature). Peptides were eluted from the membrane by addition of an equal volume of CH3CN, followed by several washes (100 pl, 20 min) of 50% CH3CN in 200 mM NH4HCO3 (pH 9.0). The combined eluates were dried on a Speed Vac concentrator and washed, and peptides were resuspended in 0.1% trifluoroacetic acid and analyzed on a C-18 HPLC column as described above. In the case of Ca2+-and lipid-dependent phosphorylation of intact C/EBP, PKCII was employed using the above procedure through the Immobilon blotting step.

RESULTS
C/EBP145, containing the carboxyl-terminal 145 residues of C/EBP (Fig. l), was phosphorylated by PKC in a manner dependent on time, Ca2+, lipid activators, enzyme concentration, and isozyme type (Fig. 2). The rate and extent of phosphorylation was dependent on PKC isozyme type. Types I1 and I11 PKC phosphorylated at similar rates and to a similar extent, whereas type I PKC exhibited an -5-fold lower rate and an -3-fold lower extent of phosphorylation. A maximum of 1 mol of 32Pi/mol protein monomer could be incorporated into C/EBP145. Under these assay conditions autophosphorylation of PKC was relatively insignificant as compared to phosphorylation of C/EBP. Several quantitative methods for measuring phosphorylation, instant thin-layer chromatography (26), phosphocellulose paper (27), and SDSpolyacrylamide gel electrophoresis (28) yielded similar measurements of phosphate incorporation into C/EBP by PKC. The K,,, values for the C/EBP145, using type I1 PKC, were -10-20 pg/ml (data not shown) which is similar to that for histone-IIIS on a molar basis (-1 p~) (35). Additionally, C/ EBP145 and histone 1 1 1 s are comparable substrates for PKC as they exhibit similar V,,, values, as well as requirements for Ca2+ and PS/DAG as activating factors (Fig. 2B). Phos-

00-
FIG. 1. Amino acid sequence and schematic representation of the C/EBP145, CIEBP87, C/EBP87'-", and C/EBPGO truncated C/ERP proteins. Truncated C/EBP proteins corresponding to the C-terminal half of the protein contained the encoded residues and NYX'-G.'"' (C/EBP6O) with leader sequences of GKA-and MGS-(for the latter three), respectively. A , amino acid sequence of C/ EBP145. Sequence of isolated R'P-labeled peptides (stippling), phosphorylated Ser residues (capsules), basic regions A and B (cigar shapes), and the leucine zipper residues (boxed) are as indicated. B, schematic representation of truncated C/EBP proteins. Phosphorylation sites (Ser"", Ser'", and SerJ9' ) ( S ) , the basic DNA binding regions A and B ( B R and stippling), and leucine zipper (LZ and stippling) are as indicated. C/EBP87S'C has a Cys residue in place of phoamino acid analysis of the "P-labeled C/EBP145 indicated that only Ser residues were modified (data not shown). Revered-phase HPLC-peptide mapping carried out on C/ EBP145, '"P-labeled by PKC (type 11) and cleaved sequentially with CNBr and endoproteinase Lys-C, indicated the presence of three major "'P-labeled peptides ( T R = 4', 17', and 60') and several minor ones. Amino acid sequencing of abundant "'P-labeled peaks yielded the following: T R = 4', S'99 (P)RDK; T,? = 17', AKKS'" (P)VDK; and 60', GAAG-LPGPGGV8 (P)LK. Extended endoproteinase Lys-C cleavage of '"P-labeled C/EBP145 resulted in loss of the T R = 17' peak and appearance of SZi7 (P)VDK eluting at 7'; some variability in cleavage at one or both of the lysines in this heptapeptide was observed. Resistance to cleavage by endoproteinase Lys-C a t a lysine residue in the presence of an adjacent basic residue or a phospho-acceptor residue has been reported (36,37). In the case of CAMP-dependent protein kinase phosphorylation of Leu-Arg-Arg-Ala-Ser-Leu-Gly (kemptide) a t a Ser residue with an upstream basic residue, phosphorylation results in resistance to cleavage by trypsin at that basic residue (38).
The effect of phosphorylation of truncated C/EBP proteins on DNA binding was studied by using a '"P-labeled synthetic oligodeoxynucleotide probe containing a high affinity C/EBP binding site. C/EBP145 phosphorylated by PKCII in the ard assay conditions. Incubations were carried out until 1 mol of mol of protein was incorporated as determined in a parallel experiment in the presence of (-y-:"P]ATP. Nonphosphorylated ( A ) and phosphorylated ( B ) C/EBP145 were incubated with CCAAT probe (0.28 ng of "'P-labeled 20-mer) in molar ratios (C/EBP145/DNA probe) of 0 (lanes 1, 4, 7, and I O ) , 6.9 (lanes 2, 5, 8, and 11 ), and 69  (lanes 3, 6, 9, and 12). KC1 concentrations of the binding reactions were 50 ( a ) , 100 ( b ) , 250 ( c ) , or 500 ( d ) mM. Binding of the labeled probe to C/EBP145 was analyzed by electrophoresis on a 5% native polyacrylamide gel. Free and retarded '"P-labeled CCAAT probe were detected by autoradiography.
presence of Ca2+ and PS/DAG bound the oligodeoxynucleotide probe with decreased affinity relative to the nonphosphorylated starting material (Fig. 3). DNA binding by both the non-and phosphorylated forms of C/EBP145 was dependent on the protein concentration. Increasing the KC1 concentrations from 50 to 500 mM resulted in decreased binding of C/EBP145 to the "P-labeled 20-mer probe, indicating that ionic interactions play a significant role in the interaction of C/EBP145 with probe. Gel retardation experiments in the presence of 25 pg/ml poly(d1-dC) yielded similar results as in its absence which is consistent with a specific interaction between the CCAAT oligodeoxynucleotide and C/ EBP145.
In order to determine the effect of each phosphorylation site on DNA binding, several smaller C/EBP fragments, which contained the leucine zipper and adjacent upstream basic region, were produced. These proteins contained either two (C/EBP87, Ser2''9 and Ser2") or one (C/EBP87"-C, Ser2" only, and C/EBPGO, Ser299 only) phosphorylation site(s). C/ EBP87, C/EBP87"", and C/EBP6O were phosphorylated by PKC to near stoichiometric levels of 2.0, 0.75, and 0.7 mol of :''Pi/mol of protein monomer, respectively (Fig. 4). In all cases, modification proceeded in a Ca2+-and lipid-dependent manner. Peptide mapping (isoelectric focusing and reversed-phase HPLC) of :'2P-labeled C/EBP87, C/EBP87"-", and C/EBP6O indicated the presence of two, one, and one "'P-labeled peptides, respectively, each corresponding to the expected Ser residue(s). Phosphorylation of C/EBP87 and C/EBP6O to near stoichiometric levels with PKC resulted in an attenuation of DNA binding (Fig. 5), similar to that obtained with C/EBP145 (Fig. 3). In contrast, phosphorylation of C/ EBP87"'" showed no apparent effect on site-specific DNA binding (Fig. 5). Analogous results were obtained using a DNase I protection assay (Fig. 6). Thus, negative effects on protein/nucleic acid interaction were inferred to result from phosphorylation of Ser2". The maximum incorporation of 1 mol of '"PJmol of C/EBP145 monomer was distributed among the three major characterized sites, Ser299, Ser2", and Ser'", in the approximate ratio of 0.5:0.35:0.15, respectively, i.e. -0.50 mol of Pi/mol of SerZg9 site/mol of protein monomer (see Fig. 9B). It is not yet certain whether the E. coli-produced C/EBP145 is folded in the same manner as its smaller M , counterparts such that each polypeptide chain can be phosphorylated by PKC. The dramatic weakening of DNA binding by maximally phosphorylated C/EBP145 could be due to  phosphorylation of a single polypeptide chain within the C/ EBP dimer to a level of -1.0 mol of Pi/mol of SerJg9 site, or alternatively, only 50% of the recombinant protein is in the active conformation a t SerzgS to be recognized by PKC.
The observation that the phosphorylation of Ser'99 weakens binding of C/EBP to its DNA substrate raised the possibility that this portion of the protein might be involved in intimate association with the DNA. If so, prior interaction of the protein with the DNA substrate might mask the Ser"' site from subsequent modification by PKC. To test this hypothesis, C/EBP6O was exposed to the specific oligodeoxynucleotide probe and analyzed by phosphorylation with PKC (Fig. 7A ). Increasing mole equivalents of the DNA substrate probe per mol of C/EBP6O resulted in progressively greater inhibition of phosphorylation by PKC, with >75-80% inhibition being obtained a t 0.93 mol eq of DNA probe dimer per mol active C/EBP6O dimer. To test whether the observed effect was due to site-specific interaction of the protein with its DNA substrate, we repeated the experiment using a nonspecific oligodeoxynucleotide probe (Fig. 7B). In this case, no significant effect on PKC-catalyzed phosphorylation was obtained with up to 1.49 mol eq of nonspecific DNA per mol of C/EBPGO. Hence C/EBPGO, when bound in a site-specific manner to its DNA substrate, could not serve as a substrate for PKC.
Since oligonucleotide could potentially inhibit PKC activity by somehow competing for ATP or otherwise by poisoning the catalytic activity of PKC, control peptides, GAP43 (1-20), glycogen synthase(l-12), and myosin light chain(l1-23) peptide, were incubated with specific C/EBP DNA substrate EBP8ys (' as indicated. Samples were then digested with DNase I and analyzed on a sequencing gel. Maximally phosphorylated C/EBP proteins were prepared in the presence of PKCIII and nonphosphorylated in the ahsence, using standard assay conditions. followed by phosphorylation with PKC ( Fig. 7C and data not shown). All of the control peptide substrates showed no inhibition of phosphorylation by PKC with up to 1.38 mol eq of DNA probe/mol of peptide, indicating that the specific DNA substrate for C/EBP had no effect on PKC-catalyzed phosphorylation of these PKC substrates.
Since many protein kinases have overlapping substrate specificities, PKA and PKG, which have similar substrate specificities as PKC, were tested for their ability to phosphorylate C/EBP145. PKA (bovine heart) and its catalytic subunit could incorporate up to 0.1 mol of "Pi per mol of C/ EBP145; increasing the enzyme concentration or reaction time resulted in increased autophosphorylation levels but no further incorporation into the C/EBP145 (data not shown). PKG phosphorylated the C/EBP145 to a considerably lower extent than PKA, as demonstrated by SDS-polyacrylamide gel analysis. Phosphoamino acid analysis of C/EBP145 that had been '"P-labeled with PKA indicated that only Ser residues were phosphorylated. Reverse-phase HPLC peptide mapping of C/EBP145 that had been '"P-labeled with PKA, electroblotted to Immobilon membrane, cut out, and successively cleaved with CNBr and endoproteinase Lys-C, indicated insignificant levels of phosphorylation of the peptide eluting at the position of the SerZ4'-containing peptide. Other '"P-labeled material eluted a t 4' and 17', similar to "'P-labeled Ser""" and Ser'" containing peptides obtained with PKC, suggesting that PKA can phosphorylate Ser'2"9 and Serli7 but not Ser2"' in C/EBP145. Phosphorylation of the C/EBP145 with PKA had a slight attenuating effect on its ability to bind to the specific DNA substrate relative to nonphosphorylated protein (data not shown).
Although PKA and PKG could phosphorylate C/EBP145, the low stoichiometries of phosphorylation of this protein casts doubt upon a physiological role for these two protein kinases in regulating C/EBP function.
Phosphorylation of holo-C/EBP by PKC was tested by employing liver nuclear extract. Liver nuclear extract, that had been heat-treated and concentrated, was phosphorylated by PKCII in the presence of [y-"'P]ATP and either EGTA or (PDBu), as demonstrated by immunoprecipitation of "Plabeled C/EBP (42 kDa) (Fig. 8). A significant stimulation in the phosphorylation of C/EBP was apparent in the presence of Ca2+, PS, and PDBu as compared to basal phosphorylation. We found that making use of PKM, which does not require lipid cofactors, resulted in a better recovery of the immunoprecipitated :'2P-labeled C/EBP than with PKC. Phosphoamino acid analysis of the "'P-labeled holo-C/EBP, that was immunoprecipitated, analyzed on a 14% SDS-polyacrylamide gel, and transferred to Immobilon membrane, indicated that Ser was the only modified residue. HPLC-peptide mapping of '"P-labeled holo-C/EBP, that had been immunoprecipitated, cut out from the Immobilon membrane, and successively cleaved with CNBr and endoproteinase Lys-C, yielded a similar '"P-labeled peptide map as the truncated C/EBP proteins, with the exception of an additional major ""P-labeled peptide at 12' and a weak phosphorylation a t SerZ4' (60') (Fig. 9A). The immunoprecipitated ""P-labeled 30-kDa protein is putatively a fragment of C/EBP and the ratio of '"P-labeled 42-kDa C/EBP to 30-kDa material varied somewhat from preparation to preparation (39). Hence it appears that an additional PKC phosphorylation site is present in holo-C/EBP in addition to Ser""', Ser"' , and Ser"' sites. Because of the relatively low abundance of PKC and C/EBP in liver, we have not yet successfully demonstrated i n uiuo phosphorylation of C/EBP by PKC. In order to obtain specific immunoprecipitation of the in uitro labeled C/EBP it was necessary t o concentrate the heat-treated nuclear extract for phosphorylation by exogenously added PKM or PKC.

DISCUSSION
The DNA-binding domain of C/EBP can be phosphorylated by PKC in a manner dependent on Ca", lipid activators, and isozyme type. Both type I1 and I11 PKC were observed to be more active than type I enzyme in the phosphorylation of the truncated C/EBP proteins. Since type I1 PKC is localized in rat liver nuclei (13), it is possible that this particular PKC isozyme might phosphorylate C/EBP i n uiuo. Three major PKC phosphorylation sites, SerZsg9, Ser"" , and SerZ4', were identified in C/EBP145. Intact C/EBP from liver nuclear extract was phosphorylated by PKCII in a Ca2+-and lipiddependent manner, and this protein was modified by Mkinase at Ser2"' and Ser"', to a lower extent at Ser'14*, as well as at an additional unidentified site (Fig. 9). Although the former are sites of modification by PKC i n uitro, they are not typical PKC consensus sites. A consensus PKC phosphoryl- and C/EBP145 phosphorylated by PKC. A, liver nuclear extract (4.8 mg of protein), derived from 7.5 g of adult liver, was sequentially heat-treated, concentrated, and phosphorylated with "kinase and 230 pCi of [y-'"PIATP. After stopping the reaction with EDTA, '"Plabeled C/EBP was immunoprecipitated with antiserum against a 14amino acid peptide and analyzed on a 14% SDS-polyacrylamide gel. After transfer to Immobilon membrane (inset), radiolabeled C/EBP (42 kDa) was cut out and sequentially digested with CNBr and endoproteinase Lys-C and analyzed using reversed-phase HPLC.
Solvents A and B were aqueous 0.1% trifluoroacetic acid and CHsCN containing 0.1% trifluoroacetic acid, respectively. Radiolabeled peaks a t 4', 17', and 60' correspond to peptides containing Ser""', Ser'", and Ser"', respectively. H, C/EBP145 (3 p~) was phosphorylated with "kinase and [y-"'PIATP. The reaction was stopped with trichloroacetic acid, and "P-labeled C/EBP145 was washed and sequentially cleaved with CNBr and endoproteinase Lys-C. The "'Plabeled peptides were analyzed using reversed-phase HPLC as described above. Peaks a t 4', 17', and 60' correspond to peptides containing SerZg9, Ser"', and Ser'", respectively, and the determined sequences are as indicated. The ratio of '"P-labeled holo-C/EBP to immunoreactive 30-kDa material varied somewhat among different preparations. See "Experimental Procedures" for further details. ation site is immediately flanked by basic residues with at least an adjacent nonbasic residue between the Ser/Thr acceptor and the adjacent basic residues. However, despite being an imperfect match to the consensus PKC substrate sequences, these phosphorylation sites in C/EBP serve as unusually good substrates.
A model for the interaction of the DNA binding domain of C/EBP with its specific recognition sequence has recently been proposed (20). This "scissors-grip'' model hypothesizes that a dimerized set of a helices (the leucine zipper) bifurcates in such a manner as to project positively charged N helices in opposite directions into the major groove of DNA. Phosphorylation of Ser'"" in C/EBP resulted in significant attenuation in DNA binding affinity. According to the scissorsgrip model, SeI-"99 is positioned within the major groove of the DNA. One might therefore predict the observed effect. Since SerZy9 is thought to come in close contact with the negatively charged DNA substrate, it is not surprising that modification of this residue by phosphorylation would interfere with DNA binding. The added negative charge resulting from phosphorylation might interfere with DNA binding by ionic charge repulsion. Alternatively phosphate modification might interfere with specific interactions made between SerZg9 and the base pairs in the DNA recognition site. Since the scissorsgrip model predicts that SerZg9 will be buried within the major groove of DNA when C/EBP is bound to its DNA substrate, it might have been anticipated that DNA binding would inhibit phosphorylation of SerZg9 by any protein kinase. Consistent with this interpretation, we have observed that specific C/EBP DNA substrates block the capacity of PKC to phosphorylate SerZg9 of C/EBP.
Although Ser277 lies in a region of net positive charge (Fig.  l ) , mutagenesis studies replacing these positively charged residues with uncharged amino acids resulted in only a 2-4fold decrease in DNA binding affinity (40). Similar sitedirected mutagenesis experiments in which the basic residues in the BR-A and BR-B regions (Fig. 1) were mutagenized indicated that these residues are critical for specific interaction with DNA. The lack of an effect on DNA binding when only SerZ7' is phosphorylated (Fig. 5, C/EBP87'-' ) is consistent with the notion that the basic region surrounding Ser277 is not of critical importance for DNA binding. We likewise infer that the detrimental effects of phosphorylation at SerZg9 indicate that it and its surrounding amino acid residues come in close contact with DNA. The observation that protein kinase C exerts a negative effect on the function of C/EBP is consistent with our current understanding of the biology of both C/EBP and PKC. Umek et al. (41) have shown that functional expression of C/EBP in a rapidly dividing cell line results in cessation of proliferation, an experimental result consistent with the observation that C/EBP is found only in differentiated tissues. In addition, the expression of PKC in several tissues is more prominent after differentiation (42). The coexistence of PKC and C/EBP in the nucleus of differentiated tissues suggests that PKC may play a role in controlling C/EBP function in these tissues.
According to our observations activation of PKC should inhibit C/EBP. Phorbol esters tend to induce cell proliferation, whereas C/EBP inhibits cell proliferation. It is possible that PKC may reverse the growth-suppressive effects of C/ EBP under conditions, for example, of liver regeneration. Under such conditions, Fos and Jun are induced (43). Although these two proteins are bZip proteins like C/EBP, perhaps they would not be negatively regulated by PKC because they have Cys at the position within the basic region corresponding to SerZg9 in C/EBP. We conclude by emphasizing the fact that our observations have resulted from assays carried out exclusively in uitro. Until in uiuo evidence is obtained, the relationship between PKC and C/EBP is entirely speculative. Despite this significant shortcoming, we do recognize that C/EBP is a very good substrate for PKC and that the effects of phosphorylation at SerZg9 are consistent with the scissors-grip model. Thus, although the ultimate connection between PKC and C/EBP has yet to be made, useful information has resulted from these studies.
Based upon sequence alignment comparisons of 11 bZIP proteins, and conservation of a Ser at position 299 (Fig. lo), we predict that GCN4, v-Jun, CPC-1, HBP-1, TGA-1, and Opaque 2 are also PKC substrates and that c-Jun, Fos, YAP-1, and Cys-3 are not. CREB has been shown to be a PKC substrate (44), however, rat brain and human placental CREB lack Ser at positions corresponding to SerZg9, Ser277, and SerZ4' in C/EBP (45, 46). The amino acid sequences of the 11 transcription factors indicate limited conservation of SerZg9,

O I P M U P R l O A L K E E P O l W E M~~~l f P l~E~E R l~R K R~N R l M S~K R K l~R l A R l E E~l l~O N S E L A
301 v -i m

301
C -j m

W P A P S A G L Y S R A G W I (~M~G C R A O S~~R G~O~S P E E E E K R R I R R E R N~C R Y R R R E L~D~~M E~D O L E D E K S A~O 280
fOS a a a

S R~E I E N E~R R l G l R D G E D S E~K K K G S K l S K K~l D P E l K O K R l A O N R M O R A f R E R K E R~L E K~S l E S l~N E
1D7 1 W -1 Ser residues of the C/EBP proteins phosphorylated by PKC (*), the basic DNA binding regions (BR-A, BR-B), the leucine zipper residues (a), and conserved phosphorylation site Ser residues (stippling) are as indicated (Refs.: 31,(47)(48)(49)17,50,46,51,18,16,52,and 20,respectively). and poor conservation of residues corresponding to SerZ4' and Ser277. Thus, modification of these residues could explain, in part, C/EBP functional specificity. The ultimate test of these ideas requires their evaluation in intact cells.