Characterization of the mitogen-activated protein kinase phosphorylation sites on the connexin-43 gap junction protein.

We have previously demonstrated that epidermal growth factor induced a rapid, transient decrease in gap junctional communication and increase in serine phosphorylation on the connexin-43 gap junction protein in T51B rat liver epithelial cells. The kinase(s) responsible for phosphorylation and specific serine targets in connexin-43 have not been identified. There are three consensus mitogen-activated protein (MAP) kinase serine phosphorylation sequences in the carboxyl-terminal tail of connexin-43 and purified MAP kinase phosphorylated connexin-43 in vitro on tryptic peptides that comigrated with a subset of peptides from connexin-43 phosphorylated in vivo in cells treated with epidermal growth factor. These data suggested that MAP kinase may phosphorylate connexin-43 directly in vivo. We have utilized a glutathione S-transferase fusion protein containing the cytoplasmic tail of connexin-43 to characterize MAP kinase phosphorylation. Site-directed mutagenesis, phosphotryptic peptide analysis, and peptide sequencing have confirmed that MAP kinase can phosphorylate connexin-43 at Ser255, Ser279, and Ser282, which correspond to the consensus sites recognized earlier. Characterization of MAP kinase-mediated phosphorylation of connexin-43 has defined potential targets for phosphorylation in vivo following activation of the epidermal growth factor receptor and has provided the basis for studies of the effects of phosphorylation, at specific molecular sites, on the regulation of gap junctional communication.

Gap junctions are aqueous membrane channels that permit the exchange of small (Ͻ1000 Da) regulatory ions, molecules, and metabolites between cells. Gap junctional communication (GJC) 1 allows for synchrony in events such as contraction in the uterus and myocardium and is believed to play an important role in regulating growth and differentiation (reviewed in [1][2][3]. Gap junctions form between hexameric structures of connexin molecules (connexons) that interact with connexons in neighboring cells to form membrane pores (4,5). Connexins are a conserved family of proteins with four membrane-spanning regions and with cytoplasmic amino and carboxyl termini, yielding one intracellular and two extracellular loops.
GJC is known to be regulated by posttranslational phosphorylation on connexin-43 (Cx43). Musil et al. (6) have demonstrated that a basal level of posttranslational phosphorylation on serine residues in Cx43 may be essential for functional assembly and activation of gap junctions. Up-regulation of GJC has been associated with increased levels of cyclic AMP and increased serine phosphorylation on Cx43 (7,8). GJC and serine phosphorylation on Cx43 were also up-regulated in communication-deficient S180 mouse cells following transfection with liver cell adhesion molecule DNA (6). In contrast, disruption of GJC has been associated with increased tyrosine and/or serine phosphorylation on Cx43 (9 -12). Down-regulation of GJC was associated with increased serine phosphorylation in cells expressing the ras oncogene (10) and in cells stimulated with epidermal growth factor (EGF; Refs. 13 and 14) and with increased tyrosine phosphorylation in src (15) and fps (16) transformed cells. Studies from this laboratory have demonstrated that Cx43 can serve as a direct substrate for the pp60 src tyrosine kinase (17), however, it is not known whether the fps kinase directly phosphorylates Cx43 or activates a downstream tyrosine kinase responsible for Cx43 phosphorylation (16). The possible increased serine phosphorylation on Cx43 in cells transformed by the src and fps oncogenes is presumably mediated by the activation of downstream serine/threonine kinases through signal transduction events.
A rapid and transient decrease in GJC, which correlated with an increase in serine phosphorylation on Cx43, was observed in T51B rat liver epithelial cells following stimulation of the epidermal growth factor receptor (EGFR; Refs. 13 and 14). The data suggested that the receptor tyrosine kinase activated a downstream serine/threonine kinase(s) that phosphorylated Cx43. The identity of this kinase(s) and the specific Cx43 sites phosphorylated have not been characterized. EGF stimulates pathways that lead to the activation of protein kinase C (PKC; Refs. 18 and 19). However, down-regulation of 12-O-tetradecanoylphorbol 13-acetate-sensitive forms of PKC did not prevent EGF-induced disruption of GJC or Cx43 phosphorylation (14). EGF also activates mitogen-activated protein kinase (MAPK) by signal transduction events that begin with ligand activation of the EGFR and proceed through a sequence of protein-protein interactions coupled to a protein kinase cascade (20 -22). MAPK was activated in EGF-treated T51B rat liver epithelial cells with kinetics that supported a potential role for MAPK in signal transduction events leading to Cx43 phosphorylation or perhaps in directly phosphorylating Cx43 (14). The cytoplasmic, carboxyl-terminal tail of Cx43 possesses three putative consensus MAPK phosphorylation sequences PX 1-2 (S/T)P (23)(24)(25), underlined in Fig. 1. Furthermore, activated MAPK may phosphorylate Cx43 directly in vivo, since MAPK phosphorylated Cx43 in vitro on phosphotryptic peptides that comigrated with a subset of EGF-responsive phosphotryptic peptides obtained from Cx43 phosphorylated in vivo in EGF-stimulated cells (14).
Although it is known that GJC can be regulated by posttranslational phosphorylation on connexin, little is known about the molecular sites in the connexin molecule that are critical for regulating intercellular communication through gap junctions. The increased phosphorylation on Cx43 at specific serine sites that follows activation of the EGFR may be the direct cause of the observed functional disruption of GJC. Thus, EGF-induced phosphorylation provides an excellent system for the identification of sites in Cx43 critical to GJC and for the characterization of the signal transduction events leading from the EGF receptor to the activation of downstream serine/threonine kinases that mediate Cx43 phosphorylation. The studies presented here were carried out to further delineate the role of MAPK in Cx43 phosphorylation by identifying specific MAPK serine phosphorylation sites on Cx43. We utilized a GST (glutathione S-transferase) fusion protein containing the cytoplasmic, carboxyl-terminal tail of Cx43 (GST-Cx43-CT; Ref. 17) as a MAPK substrate and have identified the sites of phosphorylation by deletion and site-directed mutagenesis, phosphotryptic peptide mapping, and peptide sequence analysis. The results of these studies support the concept that MAPK phosphorylates Cx43 in vivo in response to a signal transduced by the activated EGFR and define the phosphorylation sites in Cx43 that may be directly related to the functional disruption of GJC observed in EGF-treated cells.
Recombinant, Baculovirus-expressed Cx43-Full-length Cx43 was expressed in Sf9 insect cells infected with a recombinant baculovirus containing the rat Cx43 cDNA (17). Cx43 was isolated from Sf9 cell homogenates by detergent extraction of the membrane pellet and was immunoaffinity purified by incubation with a monoclonal antibody to Cx43, chemically coupled to protein G-Sepharose, as described (17).
GST-Cx43-CT Fusion Proteins-A GST fusion protein containing the cytoplasmic, carboxyl-terminal tail of Cx43 (amino acids 236 -382, see Fig. 1) was generated by polymerase chain reaction amplification of nucleotides 907-1356 of the rat Cx43 cDNA and cloned into the BamHI/ EcoRI sites of the pGEX-KG expression vector as described (17). This construct is referred to as wild type (wt) GST-Cx43-CT. Deletion mutants were prepared by recombinant polymerase chain reaction (26) using mutagenic oligonucleotide primers designed to delete nucleotides encoding amino acids 253-256 (PLSP) or amino acids 274 -284 (PTA-PLSPMSPP) of Cx43 as described. 2 Underlined amino acids represent the consensus MAPK phosphorylation sequences (serine sites 255, 279, and 282). Deletion mutants were cloned into the pGEX 2TK expression vector (Pharmacia Biotech, Inc.). Serine site mutants of GST-Cx43-CT were prepared from double-stranded DNA (Chameleon site-directed mutagenesis kit, Stratagene) with oligonucleotide primers (Ransom Hill Biosciences, Inc.) designed to alter a serine residue to alanine: primer 5Ј-CTGGCCCACTGGCCCCATCAAAAGAC-3Ј was used to alter the AGC codon for Ser 255 to GCC, a codon for alanine (mutant S255A), and primer 5Ј-GCTCCACTCGCGCCTATGGCTCCTCCTGG-3Ј was used to alter the codons for Ser 279 (TCG) and Ser 282 (TCT) to codons for alanine (GCG and GCT, respectively; double mutant S279A,S282A). Selection for mutants was provided by the simultaneous use of a selection primer (27) designed to eliminate a unique restriction site in the parental DNA (AlwNI at position 2617 or MluI at position 3647 in the pGEX-KG expression plasmid). The fidelity of all GST-Cx43-CT mutants was confirmed by DNA sequencing (28).
GST fusion proteins were expressed in Escherichia coli (DH5␣) by induction with 0.1 mM isopropyl-␤-D-thiogalactopyranoside for 3 h at 37°C. Cells were lysed by brief sonication on ice in PBS with 4 mM EDTA, 1 mM benzamidine, and 0.2 mM phenylmethylsulfonyl fluoride and then solubilized with 1% Triton X-100. GST fusion proteins were affinity purified from clarified cell lysates by a 2-h incubation at 4°C with glutathione-Sepharose 4B-agarose beads (Sigma) followed by extensive washes with PBS (17).
In Vitro MAPK Assays-MAPK assays were carried out with GST-Cx43-CT fusion proteins immobilized on glutathione-Sepharose 4B beads (10 l). The beads were washed three times with cold kinase buffer (20 mM HEPES, pH 7.4, 10 mM MgCl 2 , and 1 mM dithiothreitol) and then incubated for 15 min at 30°C in kinase buffer with the addition of purified MAPK (0.1-1 l) and 10 Ci of [␥-32 P]ATP (Amersham Corp., 6000 Ci/mmol) per reaction. Full-length Cx43, isolated from baculovirus-infected Sf9 cells, was phosphorylated by MAPK in a 20-l reaction mixture containing 0.2 g of Cx43. Kinase reactions were terminated by the addition of 2 ϫ SDS sample buffer and heating the sample tubes in boiling water. To phosphorylate preparative amounts of GST-Cx43-CT, 500 l of beads was incubated in the same kinase buffer with 250 Ci of [␥-32 P]ATP and 1.5 l of MAPK, and the addition of 0.16 mol of unlabeled ATP. Two MAPK preparations were used in these studies: (i) activated, recombinant human MAPK produced in Sf9 cells coinfected with Raf-1 and v-Ras (29), purified by HPLC (30) and used at 1 l/reaction or 100 ng/reaction (kindly provided by Dr. Rikiro Fukunaga, The Salk Institute, La Jolla, CA); and (ii) recombinant His-tagged rat MAPK (31) produced in bacteria and activated in vitro with an activated, His-tagged human MAPK kinase (MEK) preparation (32), used at 0.1 l or 50 ng/reaction (generously provided by Drs. Natalie Ahn, University of Colorado, Boulder, CO and Melanie Cobb, University of Texas, Dallas, TX). The histidine tag on the bacterially expressed recombinant MAPK and MEK allowed for rapid and convenient purification of these kinases without significant loss of enzymatic activity. The His-tagged, activated rat MAPK preparation was used for all the in vitro phosphorylations in the figures presented in this paper. The deletion mutants of GST-Cx43-CT were phosphorylated with the activated human MAPK preparation from recombinant baculovirus-infected Sf9 cells.
Two-dimensional Phosphopeptide Mapping-Phosphotryptic peptide maps were prepared as described by Boyle et al. (33) and Kanemitsu and Lau (14). Briefly, MAPK-phosphorylated GST-Cx43-CT fusion proteins were resolved by SDS-PAGE and visualized by autoradiography. The protein bands were excised from the wet gel and extracted with a buffer containing 50 mM NH 4 HCO 3 , 0.1% SDS, and 5% ␤-mercaptoethanol. SDS was removed from the samples by precipitating the proteins with cold trichloroacetic acid in the presence of carrier protein (RNase A, 10 -20 g). Precipitated proteins were washed with ice-cold acetone, oxidized with performic acid at 0°C, lyophilized to dryness, and then digested overnight at 37°C with TPCK-treated trypsin (Worthington). Phosphotryptic peptides were resolved on thin-layer cellulose chromatography (TLC) plates (EM Scientific) by electrophoresis for 65 min at 1000 V in pH 1.9 buffer (2.5% formic acid (88%), 7.8% glacial acetic acid, v/v), followed by ascending liquid chromatography in the second dimension in isobutyric acid buffer (62.5% isobutyric acid, 1.9% n-butanol, 4.8% pyridine, 2.9% glacial acetic acid, v/v). Phosphorylated peptides were visualized by autoradiography at Ϫ70°C with the aid of an intensifying screen. MAPK-phosphorylated wt and deletion mutants of GST-Cx43-CT produced the same phosphorylation patterns when the kinase reactions were carried out in solution with proteins eluted from the glutathione-Sepharose beads (10 g of fusion protein/reaction), indicating that phosphorylation was not hindered when the fusion proteins were immobilized on beads.
Protein Sequencing and Edman Degradation-Tryptic digests of preparative amounts of MAPK-phosphorylated wt GST-Cx43-CT were prepared as described, except that the carrier protein (RNase A) was omitted during trichloroacetic acid precipitation and a sequencing grade of TPCK-treated trypsin was used for digestions (Promega). Tryptic digests were fractionated by HPLC on a Bio-Rad Microsorb CB reverse phase column (5 m, 300 Å) using a water/acetonitrile/trifluoroacetic acid gradient that changed from 85.9/14/0.1 to 75.9/24/0.1 over 30 min. Fractions were collected every minute, and the radioactive fractions were characterized by their migration on two-dimensional phosphotryptic maps. The fraction that migrated with peptide b was identified and submitted for protein sequence analysis (University of Minnesota, St. Paul, MN) and Edman degradation (Ref. 34; University of Virginia, Charlottesville, VA).
Phosphoamino Acid Analysis-Phosphoamino acid analysis was carried out as described (35) on tryptic digests of MAPK-phosphorylated GST fusion proteins.

The in Vitro MAPK Phosphorylation Sites in Cx43
Are in the Cytoplasmic Carboxyl-terminal Tail-An earlier study demonstrated that numerous tryptic peptides were nominally phosphorylated in Cx43 isolated from control-treated T51B rat liver epithelial cells and that phosphorylation on these same peptides was markedly enhanced (ϳ3-fold) in cells treated with EGF (14). This study also demonstrated that a membrane preparation of full-length Cx43, isolated from recombinant baculovirus-infected Sf9 cells, was phosphorylated by purified MAPK in vitro on four tryptic peptides that comigrated with a subset of the EGF-responsive tryptic peptides obtained from Cx43 phosphorylated in vivo in EGF-treated cells (14). Three putative consensus MAPK phosphorylation sequences are present in the cytoplasmic, carboxyl-terminal tail of Cx43 (underlined in Fig. 1). We utilized a GST fusion protein, containing the carboxyl-terminal tail of Cx43 (Val 236 -Ile 382 ), as a substrate to identify MAPK phosphorylation sites. Activated MAPK was prepared in vitro by phosphorylation with constitutively active MAPK kinase (MEK; Refs. 31 and 32) or purified by HPLC from recombinant baculovirus-infected Sf9 cells expressing activated human MAPK (29,30). The two MAPK preparations yielded similar phosphorylation patterns for wt GST-Cx43-CT and for serine site mutants of GST-Cx43-CT (data not shown).
MAPK-phosphorylated full-length Cx43 and wt GST-Cx43-CT were excised from SDS-PAGE gels (see Fig. 2), eluted, and subjected to tryptic digestion as described under "Materials and Methods." Multiple phosphopeptides were observed on two-dimensional tryptic analysis of Cx43 (Fig. 3A) and wt GST-Cx43-CT (Fig. 3B). Peptides a-d are labeled according to the corresponding EGF-responsive phosphopeptides of MAPKphosphorylated full-length Cx43 in a previous study (14). Phosphotryptic peptides of wt GST-Cx43-CT comigrated (Fig. 3D) with a subset of the tryptic peptides of Cx43 phosphorylated in vivo when T51B rat cells were treated with EGF ( Fig. 3C) and also comigrated with phosphopeptides obtained from fulllength Cx43 phosphorylated by MAPK in vitro (data not shown). Control reactions, performed in the absence of MAPK, failed to phosphorylate full-length Cx43 or wt GST-Cx43-CT, indicating that endogenous kinases, capable of phosphorylating these substrates, were not present in these preparations (Fig. 2, lanes 2 and 5). The GST portion of the fusion protein was not significantly phosphorylated by MAPK (ϳ0.13% and 0.35% of wt GST-Cx43-CT in two experiments, see Fig. 2, lane 1) and autophosphorylation was not detected in the kinase alone control reaction (data not shown). These data demonstrated that the major MAPK target sites in Cx43 are located in the cytoplasmic, carboxyl-terminal tail (Val 236 -Ile 382 ) and that GST-Cx43-CT is a suitable substrate to characterize in vitro MAPK phosphorylation of Cx43.

Identification of Tryptic Peptides of GST-Cx43-CT Containing MAPK Phosphorylation
Sites-MAPK is a proline-directed serine/threonine kinase (23)(24)(25). To identify tryptic peptides in Cx43 phosphorylated by MAPK, we utilized GST-Cx43-CT mutants with deletions of the proline-rich consensus MAPK phosphorylation sequences (Fig. 1, enclosed amino acids). The deletion of Pro 253 -Pro 256 (⌬253-256) or Pro 274 -Pro 284 (⌬274 -284) was expected to alter the mobility of the peptides containing these amino acids (predicted tryptic peptides Ser 244 -Lys 258 and Tyr 265 -Lys 287 , respectively). These mutant peptides are smaller and have a different amino acid composition than those in wt GST-Cx43-CT and, thus, will migrate differently on twodimensional analysis. The phosphopeptide map of the Cx43 ⌬253-256 mutant contained phosphopeptides b and c; however, phosphopeptides d and e were absent (data not shown; refer to wt GST-Cx43-CT for phosphopeptide labeling, Fig. 3B). This indicated that the Ser 244 -Lys 258 tryptic peptide contained a site phosphorylated by MAPK in wt GST-Cx43-CT. Furthermore, it was clear that phosphopeptides d and e are related, since both disappeared when the Ser 255 consensus sequence These studies with Cx43 deletion mutants indicated that MAPK phosphorylated Cx43 on two tryptic peptides that contain the consensus MAPK phosphorylation sequences (peptides Ser 244 -Lys 258 and Tyr 265 -Lys 287 ).
We focused on the identification of peptide b because it is a major phosphopeptide on two-dimensional tryptic maps of MAPK-phosphorylated full-length Cx43 (Fig. 3A) and wt GST-Cx43-CT (Fig. 3B) and is also phosphorylated in Cx43 labeled in vivo in EGF-treated cells (Fig. 3C). Peptide b was isolated by HPLC from tryptic digests of MAPK-phosphorylated wt GST-Cx43-CT and submitted for amino acid sequence analysis.
Clear sequence data that corresponded to the predicted tryptic peptide beginning at Tyr 265 (see Fig. 1) was obtained through at least the first 15 amino acids of this peptide. The identity of peptide b is consistent with the absence of this phosphopeptide in tryptic maps of the Cx43 ⌬274 -284 deletion mutant and with MAPK phosphorylation on the synthetic peptide that corresponded to Cys 271 -Lys 287 . Peptide b contains two tandem consensus MAPK phosphorylation sequences (underlined in Fig. 1; Ser 279 and Ser 282 ).
Site-directed Mutagenesis of the Consensus MAPK Sites-To identify the MAPK phosphorylation targets in wt GST-Cx43-CT, we prepared mutants with one or more consensus MAPK phosphorylation site(s) altered from serine to alanine. Twodimensional tryptic phosphopeptide maps of phosphorylated serine site mutants are shown in Fig. 4. Eliminating the Ser 255 consensus site (mutant S255A, panel B) produced a tryptic map with phosphopeptides d and e missing (compare with wt GST-Cx43-CT in panel A), confirming that Ser 255 is a MAPK phosphorylation site in wt GST-Cx43-CT. Eliminating either the Ser 279 or the Ser 282 consensus site (mutants S279A or S282A) resulted in phosphopeptide maps indistinguishable from that of wt GST-Cx43-CT (data not shown). This is consistent with the possibility that MAPK phosphorylates peptide b at either or both of these sites in wt GST-Cx43-CT, since the Tyr 265 -Lys 287 peptide is phosphorylated when either consensus MAPK site is eliminated. Since peptides b and c migrated diagonally in relation to each other (see panels A and B), they may represent phosphoisomers (33) or different degrees of phosphorylation on the same peptide (peptide b may represent doubly phosphorylated peptide, and peptide c represent singly phosphorylated peptide). The addition of a negatively charged phosphate group retards peptide mobility in both dimensions on two-dimensional analysis. However, peptide b (presumably doubly phosphorylated) was present in the tryptic maps of the S279A and S282A single site mutants, suggesting that at least one additional site on the Tyr 265 -Lys 287 peptide may be phosphorylated. It is not clear from these data whether this other site is a preferred site or an alternate MAPK phosphorylation site that is phosphorylated in the absence of one of the primary consensus sites.
To clarify this issue, two-dimensional phosphotryptic maps of GST-Cx43-CT double or triple mutants with consensus serine sites altered to alanine were prepared. Eliminating both consensus MAPK sites in peptide b (mutant S279A,S282A, Fig.  4C) left the Ser 255 site intact and peptides d and e were phosphorylated. However, surprisingly, phosphopeptides b and c were both present in this tryptic map, suggesting the presence of two additional phosphorylation sites in the Tyr 265 -Lys 287 peptide that are phosphorylated by MAPK in the absence of the MAPK consensus sites. Results from the phosphorylation of the Cx43 triple mutant S255A,S279A,S282A (all three consensus MAPK serine sites altered to alanine, Fig. 4D) are consistent with these data. Peptides d and e were not phosphorylated; however, peptide b was phosphorylated, indicating the presence of two phosphorylation sites on peptide b, in addition to the two consensus MAPK sites. Eliminating both the Ser 255 and Ser 279 sites (mutant S255A,S279A) produced a peptide map consistent with the data presented previously; peptides d and e were absent (elimination of Ser 255 ) and peptides b and c were present (consistent with phosphorylation of peptide b at Ser 282 and at an additional site; data not shown).
Identification of the Preferred Sites of Phosphorylation by MAPK on Peptide b-The data obtained from two-dimensional phosphotryptic peptide maps of the serine site mutants suggested that MAPK phosphorylates GST-Cx43-CT at the consensus MAPK sites in the Tyr 265 -Lys 287 peptide, but may also phosphorylate this peptide at alternate sites when the consensus sites are absent. To unequivocally identify the preferred sites of MAPK phosphorylation in peptide b, we submitted peptide b (Fig. 5B), purified by HPLC from tryptic digests of wt GST-Cx43-CT (Fig. 5A), to Edman degradation. Each cycle was monitored for the release of 32 P by Cerenkov counting. Three major peaks of radioactivity were identified at cycles 5, 15, and 18 (Fig. 5C). The peaks at positions 15 and 18 correspond to Ser 279 and Ser 282 in the Tyr 265 -Lys 287 peptide (peptide b), indicating that the consensus MAPK sites are the preferred phosphorylation sites in wt GST-Cx43-CT. The Ser 282 site appears to be phosphorylated to a lesser extent than the Ser 279 site, even accounting for reduced yields with repetitive cycles (ϳ85% recovery at each cycle). Possible explanations for this include: 1) increased ␤-elimination of phosphate at Ser 282 relative to Ser 279 ; 2) some endogenous phosphorylation on Ser 282 in wt GST-Cx43-CT, decreasing the MAPK-mediated phosphorylation at Ser 282 in in vitro kinase reactions; and 3) small amounts of peptide c in the fraction submitted for Edman degradation may contribute unequally to the phosphorylation data (peptide c may be phosphorylated primarily at the Ser 279 site).
The peak at cycle 5, Asn 269 in the Tyr 265 -Lys 287 peptide, presumably is due to a contaminating phosphopeptide. Although no evidence of a contaminating peptide (other than small amounts of peptide c) was obtained on two-dimensional peptide maps of the sample submitted to Edman degradation (Fig. 5B), it was possible that a contaminating peptide was undetected because it comigrated with peptide b. It is unlikely that the contaminating peptide originated from the carboxyl terminus of Cx43 because none of its predicted tryptic peptides contain a potential serine or threonine phosphorylation site at position 5 (see Fig. 1). GST was not significantly phosphorylated by MAPK (0.13% and 0.35% of wt GST-Cx43-CT in the absence of other MAPK substrates); thus, its level of phosphorylation is too low to account for the radioactivity in cycle 5. Furthermore, predicted tryptic peptides of GST do not contain phosphorylatable residues at position 5 or recognized MAPK phosphorylation consensus sequences. A more plausible explanation for the 32 P-radioactivity obtained in cycle 5 is partial hydrolysis of the Tyr 265 -Lys 287 peptide (peptide b) that occurred during the Edman degradation procedure. Hydrolysis of the Thr 274 -Pro 275 bond would produce a shortened peptide and yield phosphorylation peaks at cycles 5 and 8. This possibility is consistent with the data (Fig. 5C) and with the lack of evidence for other phosphopeptide contaminants in the sample analyzed (Fig. 5B). A small increase in radioactivity in cycle 8 (reflecting the same difference in radioactivity compared with cycle 5 as seen between cycles 18 and 15) is consistent with this hypothesis. Although the radioactivity at cycles 5 and 15 was determined to be equivalent, when corrected for repetitive losses at each cycle, the radioactivity at cycle 15 was actually 4 times that in cycle 5, suggesting that ϳ20% of the peptide may have been hydrolyzed.
There are three sites in the Tyr 265 -Lys 287 peptide that are potential candidates for alternate MAPK phosphorylation sites: Ser 272 , Ser 273 , and Thr 276 . Phosphoserine was the only radiolabeled phosphoamino acid detected in MAPK-phosphorylated wt GST-Cx43-CT and in Cx43 phosphorylated in vivo in EGF-treated cells (14), consistent with phosphorylation at the consensus MAPK serine sites. Phosphoserine was also the only radiolabeled phosphoamino acid detected in the MAPK-phosphorylated GST-Cx43-CT mutants with altered consensus MAPK sites in the Tyr 265 -Lys 287 peptide (double mutant S279A,S282A and triple mutant S255A,S279A,S282A; data not shown). Since threonine was not phosphorylated in these mutants, the alternate sites of MAPK phosphorylation on peptide b must be Ser 272 and Ser 273 . All of the data presented are consistent with the migration of peptide b as the doubly phosphorylated Tyr 265 -Lys 287 peptide (phosphorylated at Ser 279 and Ser 282 in wt GST-Cx43-CT) and peptide c migrating as the singly phosphorylated form of the Tyr 265 -Lys 287 peptide. DISCUSSION Earlier studies from this laboratory demonstrated an increase in serine phosphorylation on Cx43 in response to signals transduced by the activated EGF receptor (13,14). GJC was transiently disrupted in T51B rat liver epithelial cells following EGF treatment and coincided with increased phosphorylation on Cx43. Treating stimulated cells with okadaic acid, a serine/ threonine phosphatase inhibitor, prevented both the dephosphorylation of Cx43 and the restoration of GJC. These studies suggested that phosphorylation on Cx43 at specific serine sites was directly related to the disruption of GJC. The MAPK signaling cascade is activated by EGF and leads to the activation of downstream protein kinases, such as MEK and MAPK (20 -22). MAPK was activated in EGF-treated T51B cells, and putative consensus MAPK phosphorylation sequences are present in the cytoplasmic, carboxyl-terminal tail of Cx43. Furthermore, purified MAPK phosphorylated recombinant Cx43 in vitro on serine residues in tryptic peptides that comigrated with EGF-responsive peptides obtained from Cx43 phosphorylated in vivo in the EGF-stimulated cells (14).
In this study, we have utilized a GST fusion protein of the carboxyl-terminal tail of Cx43 as a substrate and demonstrated that the in vitro MAPK phosphorylation sites in Cx43 are located in the carboxyl-terminal tail. Tryptic peptides of MAPK-phosphorylated wt GST-Cx43-CT (Fig. 3B) comigrated (Fig. 3D) with a subset of the tryptic peptides obtained from Cx43 phosphorylated in vivo in EGF-treated cells (Fig. 3C) and produced the same phosphorylation pattern as that obtained for full-length Cx43 phosphorylated by MAPK in vitro (Fig. 3A). These data provided additional support for the potential of MAPK to mediate phosphorylation on Cx43 in EGF-treated cells. Using a combination of tryptic peptide analysis of MAPKphosphorylated GST-Cx43-CT mutants, phosphoamino acid analysis, and sequence analysis of isolated peptides, we have identified three primary MAPK phosphorylation sites in Cx43 at Ser 255 , Ser 279 , and Ser 282 . Tryptic analysis of MAPK-phosphorylated deletion mutants indicated that the major phosphopeptides contained the consensus MAPK sequences. Phosphopeptides d and e were absent in tryptic maps of the MAPKphosphorylated ⌬253-256 deletion mutant and the S255A serine site mutant (Fig. 4B). Thus, Ser 255 is the MAPK target site in the Ser 244 -Lys 258 peptide in wt GST-Cx43-CT. Peptides d and e are likely to represent different tryptic digestion products containing the Ser 244 -Lys 258 peptide, due to inefficient cleavage at Lys-Asp bonds (Lys 258 -Asp 259 ). Two phosphopeptides migrated as peptide d, and two or three phosphopeptides migrated as peptide e (see Fig. 3B or 4A). However, with more complete digestion, using sequencing grade TPCK-treated trypsin in the absence of the RNase A carrier protein, the peptides migrating at positions d and e resolved to single phosphopeptides (see Fig. 5A). Peptide e probably represents the Ser 244 -Lys 258 peptide and peptide d the larger, less hydrophobic Ser 244 -Lys 264 peptide. It is important to note that phosphopeptides d and e are present in tryptic maps of Cx43 obtained from EGF-treated cells (Fig. 3C) and were shown to be EGF-responsive peptides in the earlier study (14), suggesting that Ser 255 may be an in vivo target for phosphorylation on Cx43 in cells stimulated with EGF.
Two other primary MAPK phosphorylation sites were identified in the tryptic peptide corresponding to Tyr 265 -Lys 287 . The identity of this peptide was determined by amino acid sequence analysis of peptide b isolated from tryptic digests of MAPK-phosphorylated wt GST-Cx43-CT. Edman degradation confirmed that the preferred MAPK phosphorylation sites on this peptide are the consensus MAPK sites, Ser 279 and Ser 282 (see Fig. 5C), and that peptide b is the doubly phosphorylated form of the Tyr 265 -Lys 287 peptide. When one of the consensus sites in the Tyr 265 -Lys 287 peptide is altered by the conservative substitution of alanine for serine, MAPK phosphorylates this peptide at an alternate site and doubly phosphorylated peptide b is still present in the tryptic maps of the single site mutants, S279A and S282A. The phosphotryptic map of the S279A,S282A double mutant confirmed that phosphorylation can occur at two alternate sites on the Tyr 265 -Lys 287 peptide in the absence of both consensus MAPK sites (Fig. 4C). However, the extent of phosphorylation on peptides b and c appears to be reduced relative to phosphorylation of these peptides in wt GST-Cx43-CT (see Fig. 4A) and the consensus Ser 255 phosphorylation site appears to be favored relative to phosphorylation at the alternate sites in the Tyr 265 -Lys 287 peptide (compare phosphorylation of the S279A,S282A double mutant in Fig. 4C with phosphorylation of wt GST-Cx43-CT in Fig. 4A).
Other explanations for the presence of phosphopeptides b and c in the S279A,S282A double mutant, such as comigration of a contaminating peptide or phosphorylation by a contaminating kinase, are unlikely. GST was not a substrate for MAPK, and no candidate serine MAPK sites are present in the GST-Cx43-CT fusion protein except for the described alternate sites, Ser 272 and Ser 273 , located adjacent to a COOH-terminal proline. Peptides b and c migrated as dimers on some twodimensional maps in the chromatographic dimension (see Figs. 4A and 5A). This was dependent on the extent of migration relative to the solvent front with increasing migration resulting in a loss of the dimeric form. Such differences in migration may be due to oxidation differences or to hydrophobicity differences due to phosphorylation at different sites on the peptide (33). No evidence of a contaminating phosphopeptide was found in the amino acid sequence analysis of peptide b isolated from wt GST-Cx43-CT. Also, contaminating kinases were not present in wt GST-Cx43-CT (see Fig. 2, lane 5), and two very different MAPK preparations yielded the same phosphorylation patterns for the site-directed mutants, even with reduced amounts of enzyme (3-10-fold), arguing against the presence of a common contaminating serine kinase in the MAPK preparations.
The singly phosphorylated Tyr 265 -Lys 287 peptide appears to be a better substrate for MAPK than unphosphorylated peptide, because the doubly phosphorylated form is apparent in tryptic maps of all of the site-directed mutants (see Fig. 4). Peptide b was observed on tryptic maps when either preparation of MAPK was used and with decreased amounts of enzyme, suggesting a preference for double phosphorylation, at tandem sites, on the Tyr 265 -Lys 287 peptide. Peptide b was also present in tryptic maps of Cx43 phosphorylated in vivo in EGF-treated cells (Fig. 3C), consistent with double phosphorylation on the Tyr 265 -Lys 287 peptide in vivo.
Since phosphoserine was the only radiolabeled phosphoamino acid detected in the MAPK-phosphorylated S279A,S282A double mutant, the alternate phosphorylation sites in the Tyr 265 -Lys 287 peptide are Ser 272 and Ser 273 . These sites are present in the Cx43 ⌬274 -284 deletion mutant, but were not phosphorylated (peptides b and c were missing rather than shifted in position). Presumably, MAPK did not phosphorylate these serine sites, since the adjacent proline residue (Pro 274 ) was deleted. A proline residue COOH-terminal to serine/threonine was determined to be critical for recognition by MAPK and a NH 2 -terminal proline, 1-2 amino acids away, constituted an optimal phosphorylation sequence in studies with peptide substrates (23,25). Although alternate MAPK phosphorylation sites are present in the Tyr 265 -Lys 287 peptide, phosphorylation appears to be tightly controlled with only two sites phosphorylated in wt GST-Cx43-CT and in Cx43 phosphorylated in vivo. Peptides b and c migrated diagonally in relation to each other on all tryptic maps, consistent with phosphoisomers, whereas peptide a migrated diagonally below peptide b on some tryptic maps (see Figs. 3A and 5A) but not on others (see Figs. 3B and 4D). The identity of peptide a is not known; however, it does not appear to be a phosphoisomer of peptide b or to be a major phosphopeptide of Cx43 phosphorylated in vivo (Fig. 3C).
The Ser 273 alternate phosphorylation site in Cx43 has an adjacent COOH-terminal proline as a minimal MAPK recognition signal. The Ser 272 site (COOH-terminal proline 2 amino acids away) does not appear to conform to a recognized MAPK phosphorylation site. MAPK phosphorylated a synthetic peptide with an alanine residue inserted between the phosphorylation site and the COOH-terminal proline 85-fold less effectively than a peptide with the COOH-terminal proline residue adjacent to the phosphorylation site (23). Amino acids flanking Ser 272 in Cx43 may permit a more favorable geometry for MAPK phosphorylation at a site 1 amino acid away from the COOH-terminal proline or the presence of a phosphate group on Ser 273 may favor subsequent phosphorylation at the adjacent Ser 272 site. It is important to remember that the phosphorylation at Ser 272 and Ser 273 observed in these studies is induced by the elimination of the primary consensus MAPK phosphorylation sites and probably does not reflect events occurring in vivo. However, the potential for phosphorylation to occur at these secondary sites must be considered in DNA transfection experiments designed to examine the effects of these mutations on connexin phosphorylation and function.
Ser 262 (in the Ser 244 -Lys 264 peptide) also has an adjacent COOH-terminal proline, but is not readily phosphorylated by MAPK in vitro. However, this site may be minimally phosphorylated in the absence of consensus MAPK phosphorylation sites, since the phosphopeptide marked as d in the MAPKphosphorylated S255A,S279A,S282A triple mutant (Fig. 4D) comigrated with peptide d of wt GST-Cx43-CT (data not shown). Ser 262 is flanked by proline and glycine residues and may not be readily accessible to MAPK in GST-Cx43-CT.
Although there are numerous reports of posttranslational modifications on connexin molecules associated with alterations in GJC, little is known about the specific molecular sites in the molecule essential for the regulation of GJC. One study in Xenopus oocytes, transfected with Cx43 mRNA, linked phosphorylation at Tyr 265 in Cx43 with altered GJC (12). Oocytes coexpressing pp60 v-src lost the ability to communicate, whereas GJC was not disrupted in oocytes that coexpressed pp60 v-src with a mutant form of Cx43 that could not be phosphorylated at position 265 (Y265F). This study concluded that phosphorylation at the Tyr 265 site on Cx43 was sufficient for directly disrupting GJC. However, these results should be interpreted cautiously, since events demonstrated in an oocyte system may not reflect completely the signaling events occurring in mammalian cells, where more than one tyrosine residue in Cx43 may be phosphorylated, directly or indirectly by pp60 v-src (15,17). Moreover, pp60 v-src associates with the Shc adaptor protein, resulting in the activation of the Ras/Raf signal transduction pathway and leading to the activation of MAPK (36).
A study by Britz-Cunningham et al. (37) demonstrated that a serine to proline mutation at position 364 of Cx43 (S364P) was associated with congenital heart defects in children. Cx43 is the main connexin expressed in heart tissue (38), and right ventricular cardiac malformation was the primary cause of neonatal death in mice lacking the Cx43 gene (39). The cytoplasmic, carboxyl-terminal tail of Cx43 contains consensus phosphorylation sequences for several protein kinases (40) that may regulate GJC, such as PKC, MAPK, and glycogen synthase kinase 3. L929 cells transfected with the S364P mutant of Cx43 did not exhibit the enhanced GJC observed in L929 cells transfected with wt Cx43 and differed in their responses to microinjected cAMP-dependent protein kinase and PKC (37). Additional mutations in serine or threonine residues in the carboxyl-terminal tail of Cx43 were noted in children with visceroatrial heterotaxia syndromes; however, the functional significance of these mutations has not been characterized (37). Nevertheless, it is clear that alterations in specific phosphorylated residues in the carboxyl-terminal tail of Cx43 can affect the regulation of GJC and the normal development of the heart.
The data presented in this study provide strong support for MAPK's role in mediating EGF-induced phosphorylation on Cx43 and identify the specific phosphorylated serine sites in the cytoplasmic, carboxyl-terminal tail of the protein that may be functionally related to the disruption of GJC. Importantly, the tryptic peptides containing the phosphorylated Ser 255 and Ser 279 /Ser 282 sites comigrated with major phosphopeptides of Cx43 phosphorylated in vivo, suggesting that these peptides are phosphorylated in response to activation of the EGF receptor in intact cells. Although alternate signaling pathways, leading to activation of other kinases such as JNK (41,42) or involving the Rac and/or Rho GTPases (43)(44)(45), have not been excluded from acting on Cx43 by these studies, we have demonstrated that Cx43 is a substrate for MAPK. Thus, Cx43 joins a growing number of MAPK substrates that include ribosomal S6 kinase (46), c-Myc (24), c-Jun (24,47), myelin basic protein (48), and the transcription factor p62 TCF (49). Cx43 is also a substrate for the pp60 src tyrosine kinase (17), the p130 gag-fps tyrosine kinase (either directly or indirectly; Ref. 16), and the PKC serine kinase. 3 Identification of the protein kinases responsible for phosphorylating Cx43 at specific sites is important to a better understanding of how GJC may be regulated in normal development and differentiation, synchronous contraction, and cell growth.