Topology of the M, 27,000 Liver Gap Junction Protein CYTOPLASMIC LOCALIZATION OF AMINO- AND CARBOXYL TERMINI AND A HYDROPHILIC DOMAIN WHICH IS PROTEASE-HYPERSENSITIVE*

Hydropathy analysis of the M, 27,000 rat liver gap junction protein sequence deduced from a cDNA clone tological localization of monoclonal and polyclonal antibody binding by indirect immunofluorescence due to the lower background obtained when utilizing commercially available fluorescein-conjugated second- ary antibodies to rat and mouse immunoglobulins. (PA)-derived peptides which were sequenced in two independent experiments. The proteinase A (PA) cleavage site after the Leu at position 9 was found in the 10-kDa fragment, as were two chymotryptic (CT, and CT,) cleavage sites (sequenced as a mixture of 10-kDa peptides), before amino acids at positions 8 and 11. No cleavage site near the amino terminus of the protein was detected for S. aureus V8 protease (VB), thermolysin (TH), or proteinase K (PK). The proteinase A (PA) cleavage site responsible for generating the 18-kDa peptide was after the leucine at position 108. Cleavage sites leading to the formation of the 17-kDa peptide were found after amino acids 109 and 119 for S. aureus (V8) (sequenced as a mixture of 17-kDa peptides), 123 for chymotrypsin (CT), and 116 for thermolysin (TH). Sequence analysis of the undigested 27-kDa protein gave a single sequence as described earlier by others (24). No inconsistencies were found with the predicted protein sequence and our sequence analysis of the various peptides.

(10-13) and plays a central role in this unique form of direct intercellular communication (14)(15)(16). Structural studies utilizing x-ray diffraction and electron microscopic image-processing techniques (1-4), while providing some structural details, have not yet provided adequate resolution of this membrane structure to visualize the disposition of the peptide backbone or other atomic features. In principle, recent publication of the deduced primary structure of the rat liver gap junction protein (11,17) and a hydropathicity analysis (18) based upon the sequence data (11) have provided an additional, independent source of structural information.
By any consideration of the properties or structure of gap junctions, this protein should contain one hydrophobic domain (contacting the hydrocarbon region of the bilayer) and three hydrophilic domains, the transmembrane pore and the regions of the protein present at the cytoplasmic surface and facing the extracellular space at the gap. As for other membrane proteins, biochemical and immunological approaches may be used to distinguish the various domains of the gap junction protein. Unlike the situation for other membrane proteins, which are accessible to hydrophilic or bulky reagents on either side of the membrane bilayer, the close juxtaposition of extracellular domains of gap junctions along the narrow "gap" (-1.5 nm) between cells renders these regions inaccessible to macromolecular probes such as antibodies or proteases (19,20). Hence, treatment of isolated, intact gap junctions with these probes involves only the binding and cleavage sites available on what was the cytoplasmic surface of the gap junction in uiuo. Indeed, gap junctions retain most of their morphological characteristics even after extensive treatment with proteases (19,20). Moreover, the punctate pattern of antibody binding, which localizes to cell peripheries, as observed by indirect immunofluorescence (7,(9)(10)(11) has been shown by immunoelectron microscopy to represent binding at the cytoplasmic surface of gap junctions (10)(11)(12)(13)15,19).
In the series of experiments using proteases and antibodies on isolated gap junctions reported here, we have capitalized upon the protectiveness of the narrow, intercellular gap (19,20) in order to identify those segments of the protein protruding from the cytoplasmic surface. The data support the model suggested by hydropathicity plots (11) in which the M, 27,000 gap junction protein spans the plasma membrane bilayer four times. They directly demonstrate, as was recently proposed (19), that both the amino and carboxyl termini of the M, 27,000 rat liver gap junction protein are located at or near the cytoplasmic surface of the gap junction, and identify a region within the molecule which is extremely sensitive to proteolytic cleavage.

EXPERIMENTAL PROCEDURES
Gap junctions were isolated subsequent to treatment of rat liver plasma membranes with 20 mM NaOH as described earlier (7).
Conditions for treatment of gap junctions with various proteases were optimized using 1-2 pg of isolated gap junctions/reaction (-20 pl final volume). Incubations were carried out overnight at 37 "C using a range of protease concentrations. Inhibition of proteinase A (Sigma) for control experiments was achieved by evaporating 3 p1 of pepstatin (Boehringer Mannheim) (1 mg/ml in methanol) to dryness in a microfuge tube, then adding proteinase A (100 pl, 0.1 mg/ml), and incubating at room temperature for 4 h prior to its addition to gap junctions.
Subsequent to incubation of gap junctions with protease, the reaction mixture was usually solubilized in SDSl-containing buffer for polyacrylamide gel electrophoresis (7). In the case of proteases which retain activity in SDS, notably proteinase K and pronase, the reaction mixture was first diluted with 1 mM NaHC03, and centrifuged in a Beckman Microfuge (13,000 X g, 30 min, 4 "C) prior to solubilization.
The peptides resolved on the SDS-polyacrylamide gels were visualized by silver staining (21). In order to optimize resolution of the lower molecular weight peptides, resolving gels were cast containing 22% acrylamide from a stock solution of 60% acrylamide, 0.4% bisacrylamide (22). Apparent molecular weights of peptides were estimated using a molecular weight marker kit (MW-SDS-17) consisting of myoglobin-derived peptides of known mass (Sigma).
For microsequence analysis, proteolytic treatments were scaled up to approximately 5-10 nmol of isolated gap junctions in a final volume of 100-200 pl. The protease-treated gap junctions were diluted and washed once in 1 mM NaHC03 using a Beckman microfuge (13,000 X g, 30 min, 4 'C) prior to solubilization and electrophoresis in order to remove the bulk of the proteases, many of which are still active in SDS. Staining of the gel for a minimum time in Coomassie Brilliant Blue R, followed by destaining and electroelution of peptides for microsequencing was essentially as described (23). Eluted and dialyzed samples were lyophilized, resuspended by sonication in 50 pl of water, precipitated with 1 ml of 95% ethanol overnight at -20 "C, lyophilized again, and resuspended in 50-100 pl of 50 mM NaHC03, with trifluoroacetic acid added to 25%, as necessary, to achieve solubilization. The samples were then sequenced using an Applied Biosystems model 470A sequencer; phenylthiohydantoin amino acids were identified by HPLC on the Applied Biosystems model 120A phenylthiohydantoin-analyzer. The amount of peptide for which sequence data were obtained varied from 50-350 pmol.
Hybridomas were isolated which secrete a monoclonal antibody to a peptide corresponding to amino acids 224-234 (RSNPPSRKGSG(C)) in the sequence of the 27-kDa rat liver gap junction protein encoded by a cDNA clone (11). The peptide was synthesized on an Applied Biosystems 430A Peptide Synthesizer using tBoc chemistry. A cysteine was added at the carboxyl terminus of the peptide to facilitate subsequent coupling to carrier protein for immunization. The peptide was analyzed by reverse-phase HPLC, and quantitated, and the composition verified by amino acid analysis.
The synthetic peptide was conjugated to BSA for immunization. 2 mg of succinimidyl 4-(N-maleimidomethyl)cyclohexane-l-carboxylate (Pierce Chemical Co.) dissolved in 100 pl of dimethylformamide was added in 10-pl aliquots, with stirring, to a solution containing 10 mg of BSA (1 ml, 10 mM NaPi, pH 7.4). After incubating with continuous inversion for 1.5 h at room temperature, the mixture was centrifuged in a microfuge to remove insoluble material. The supernatant fraction was then chromatographed on Sephadex G-25, and developed with 50 mM NH4HC03. Material eluting in the void volume, containing activated BSA, was pooled and brought to dryness using a Speed Vac (Savant). The activated BSA was dissolved in 0.1 M NaPi, 1 mM EDTA, pH 7.0 immediately before addition of peptide.
Prior to addition of peptide to the activated BSA for conjugation, 10 pmol of peptide were suspended with sonication in 0.3 ml 0.1 M NaZB407, pH 9.1, and reduced by addition of 50 pl of freshly prepared 0.1 M NaBH4. After incubation for 30 min at room temperature, excess borohydride was destroyed by acidification to pH 2.5 with 1 N HCl. The pH of the solution was then neutralized to pH 6.5-7.0 by addition of 1 N NaOH and added to the activated BSA. The solution was briefly bubbled with nitrogen to remove dissolved oxygen and incubated overnight at room temperature with continuous inversion. After removal of insoluble material by centrifugation in a microfuge, unconjugated peptide and other low molecular weight materials were removed by chromatography on Sephadex G-25 as described above. Material eluting in the void volume was pooled and brought to dryness * The abbreviations used are: SDS, sodium dodecyl sulfate; BSA, bovine serum albumin; HPLC, high pressure liquid chromatography; HEPES, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid.
as before and dissolved in water for use.
Immunization of mice, screening of sera for immunoreaction with the synthetic peptide, fusion of the splenocytes, and selection of antipeptide antibody secreting hybridomas were carried out by standard techniques modified as described here. 8-week-old female BALB-C mice were injected intraperitoneally with an emulsion of 200 pg of conjugatedpeptide 224-234 and an equal volume of complete Freund's adjuvant. The mice received intraperitoneally boosts of 50-200 pg of conjugated peptide in incomplete Freund's adjuvant biweekly for a total of nine boosts. Sera were screened by enzyme-linked immunosorbent assay for the presence of immunoglobulins binding peptide 224-234 (below) and on Western blots of isolated gap junction protein. At 2-day intervals prior to the fusion, the mice received 1-3 intraperitoneal injections with 100 pg each of conjugated peptide without adjuvant.
Beginning on day 12, hybridoma supernatant fractions were screened by enzyme-linked immunosorbent assay, Western blots, and by indirect immunofluorescence on frozen unfixed rabbit liver sections (below). Positive wells were expanded and diluted over a feeder layer of 5.0 X IO6 mouse thymocytes/well in humidified 96-well plates. Two consecutive dilutions of two positive, original fusion wells yielded two stable hybridoma cell lines, 8-26C and 8-92B, from single colony wells.
Growth of hybridomas as an ascites tumor was achieved in pristane-primed (0.5 ml intraperitoneal) 6-week-old male BALB/c mice. 10-14 days after administration of pristane, mice received injections of 9.0-18.0 X IO6 cells in 2 ml of hybridoma medium. Ascites fluid was collected 1 week following injection of the hybridomas. Both of the hybridomas reported here readily grow as an ascites and were found to be K chain-containing IgG,s by use of a subtyping kit (Bio-Rad). Results presented here were obtained using suitably diluted ascites fluid. Data indistinguishable from those described here were obtained using culture supernatant fractions of these hybridomas.
Culture supernatant fractions were assayed for the presence of immunoglobulins binding to peptide 224-234 by enzyme-linked immunosorbent assay. Briefly, wells in ProBind assay plates (Falcon) were coated overnight with 500 ng/well of peptide dissolved in 100 pl of 50 mM Na2CO3, pH 9.6. After washing and incubation with culture supernatant fractions, the presence of antibody was detected using alkaline phosphatase-conjugated goat anti-mouse immunoglobulins (Sigma) using the supplier's suggested protocol.
A rat hybridoma, designated R5.21C, producing monoclonal antibody specific for the 27-kDa liver gap junction protein (12) was kindly provided by Dr. Daniel Goodenough, Harvard Medical School. This monoclonal was found to be of the IgGl subclass based upon binding of only a rabbit anti-rat IgGl (Zymed) when screened on Western blots of purified rat liver gap junctions.
Immunochemical analysis of binding of monoclonal antibodies and affinity purified rabbit antibodies to the 27-kDa rat liver gap junction protein (16) on Western blots was carried out as described earlier (7,9,16). Autoradiography subsequent to incubation with '*'I-Protein A was used to visualize the binding of primary antibody. Frozen sections of rabbit liver, instead of rat liver (7, 16), were used for immunocy-tological localization of monoclonal and polyclonal antibody binding by indirect immunofluorescence due to the lower background obtained when utilizing commercially available fluorescein-conjugated secondary antibodies to rat and mouse immunoglobulins.

RESULTS
Conditions for protease treatment of intact gap junctions were optimized with respect to buffer, pH, and protease concentration. Digestion of intact gap junctions with increasing concentrations of proteinase A (Fig. 1) led to degradation of the 27-kDa gap junction protein and the formation of two major peptides with apparent molecular masses of 10 and 18 kDa. An apparent intermediate digestion product, migrating slightly slower than the 18-kDa product on this gel, was transiently observed, suggesting the occurrence of a cleavage prior to formation of the 10-and 18-kDa products. Sensitivity to digestion by proteinase A is also seen for the minor 21-kDa protein, characteristic of rat liver gap junction preparations, which has a mobility on this gel system slightly greater than that of the 27-kDa protein.
Prolonged treatment of intact gap junctions with many different proteases again generated two characteristic fragments which, by SDS-polyacrylamide gel electrophoretic analysis, had apparent molecular masses of 10 and 17-18 kDa, respectively. Specifically, the peptides generated by treatment with trypsin, proteinase K, chymotrypsin, Staphylococcus aureus V8 protease, elastase, and subtilisin had apparent molecular masses of 10 and 17 kDa on SDS-polyacrylamide gel electrophoresis (Fig. 2, left panel). On occasion, the 17-kDa band was found to be diffuse (e.g. trypsin-treated gap junctions, lune A ) . The 10-and 17-kDa bands were also obtained after treatment of gap junctions with other proteases, including papain, clostripain, thermolysin, and pepsin (not shown). Only treatment with proteinase A (lune G) generated a band with an apparent molecular weight that was higher than the slower-migrating band of the routinely observed doublet (18 as compared to 17 kDa).

I) and subtilisin (lune J).
In many cases, a peptide migrating more rapidly than the 27-kDa native protein band, but more slowly than the 17-18-kDa digestion product, was also observed. A 23-kDa band was readily detected in the samples treated with trypsin, proteinase K, endoproteinase lys-C, and subtilisin (lunes A , B, I, and J, respectively) but was also observed as a minor component in samples treated with the other proteases. This band is likely to be an intermediate, since increasing incubation time or concentration of protease led to its loss, accompanied by an increase in the lower molecular weight peptides, as demonstrated above (Fig. 1) for proteinase A.
The ability of antibodies to the 27-kDa rat liver gap junction protein (9, 16) to bind to the 10-and 17-18-kDa proteolytic products was characterized using Western blots (Fig. 2, right  panel). The data indicate that these polyclonal antibodies bind to the 17-18-kDa peptides, with no detectable binding to the 10-kDa fragment.
With several of the proteases, treatments were carried out on a preparative scale to generate peptides which could be electroeluted from the gels for microsequencing. The sequence data obtained for these peptides, summarized in Fig. 3, demonstrated that the faster migrating (10 kDa) band in each case contained the amino-terminal portion of the protein molecule. For proteinase K, S. aureus V8 protease and thermolysin, the only sequence observed in this faster-migrating band was that beginning with the amino-terminal methionine. These findings are consistent with the data of Zimmer et al. (19) for the 10-kDa band they obtained after treating rat liver gap junctions with either trypsin or endoproteinase Lys-C (19). With proteinase A, however, we observed a single, internal sequence, beginning with Leu at position 10. Moreover, in the case of chymotrypsin, microsequence analysis of the 10-kDa fragment indicated the presence of three distinct peptides. While about 50% of the sequenced material began with Met at position 1 (fragment denoted as CTl in Fig..3), the remainder was nearly equally divided between peptides  (11) based upon analysis of a cDNA clone. The putative membrane spanning sequences of amino acids (relatively hydrophobic stretches, including residues with hydropathy values greater than 0.5, when averaged over 19 amino acids, are assumed to span the bilayer) are underlined. Intact gap junctions were treated with protease, the resulting 10 and 17-18 kDa resolved on SDS-polyacrylamide gels and electroeluted for microsequence analysis as indicated under "Experimental Procedures." Cleavage sites are indicated by vertical arrows separating two amino acids (t), with the protease cleaving at the site abbreviated above the vertical arrow.
Experimentally determined sequences of electroeluted peptides are indicated by horizontal arrows (-+), with all peptides sequenced through at least 10 cycles of the microsequenator in order to assure correct assignment of cleavage sites. All peptide sequences were derived in a single analysis, except for the amino terminus 10-kDa peptide generated by proteinase K (PK) and the two proteinase A (PA)-derived peptides which were sequenced in two independent experiments. The proteinase A (PA) cleavage site after the Leu at position 9 was found in the 10-kDa fragment, as were two chymotryptic (CT, and CT,) cleavage sites (sequenced as a mixture of 10-kDa peptides), before amino acids at positions 8 and 11. No cleavage site near the amino terminus of the protein was detected for S. aureus V8 protease (VB), thermolysin (TH), or proteinase K (PK). The proteinase A (PA) cleavage site responsible for generating the 18-kDa peptide was after the leucine at position 108. Cleavage sites leading to the formation of the 17-kDa peptide were found after amino acids 109 and 119 for S. aureus (V8) (sequenced as a mixture of 17-kDa peptides), 123 for chymotrypsin (CT), and 116 for thermolysin (TH). Sequence analysis of the undigested 27-kDa protein gave a single sequence as described earlier by others (24). No inconsistencies were found with the predicted protein sequence and our sequence analysis of the various peptides.
beginning with Thr at position 8 (CT2) and Ser at position 11 (CTd. The data obtained using proteinase A and chymotrypsin provide evidence that the amino terminus of the protein is accessible to these proteases and is found, therefore, at or near the cytoplasmic surface of the gap junction. A control experiment was conducted with proteinase A to determine whether this accessibility was an artifact due to the presence of lipolytic or detergent-like activities in the protease preparation. Isolated gap junctions were incubated for 18 h with proteinase A that had been inhibited by pretreatment with pepstatin. If other activities in the proteinase A preparation were altering the membrane, subsequent addition of the highly active, relatively nonspecific protease, proteinase K, would be expected to cleave bonds at or near Leu 10 (as for proteinase A). However, this treatment did not in any way alter cleavage of the amino-terminal region of the protein by proteinase K, with the amino-terminal methionine-containing sequence being the only one detected in the lower (10 kDa) band.
Microsequencing of the slower-migrating peptides (apparent M, 17,000 for S. uureus V8 protease, thermolysin, and chymotrypsin, and M, 18,000 for proteinase A), indicated the presence of a protease-sensitive region from amino acids 109-123 (observed cleavage sites indicated in Fig. 3).
The binding of several antibodies to peptides arising from proteolytic treatment of intact gap junctions was analyzed using Western blots (Fig. 4). As also shown above (Fig. 2), the affinity purified rabbit polyclonal antibodies to the 27-kDa gap junction protein bound to this protein and its 47-kDa dimer (Fig. 4, PcAb, lune c ) as well as to both the 18-kDa carboxyl terminus peptide generated by proteinase A (lune a) and the 17-kDa carboxyl terminus peptide generated by thermolysin (lune b). The same pattern of antibody binding was generated using two monoclonal antibodies to peptide 224-234 (McAbl, 26C and McAbz, 92B) and a rat monoclonal antibody specific for the 27-kDa liver gap junction protein shown to bind at the cytoplasmic surface of the gap junction by immunoelectron microscopy (12) (McAbs). Binding to lower molecular weight peptides was not observed.
Immunocytological characterization of these antibodies by indirect immunofluorescence on frozen sections of rabbit liver (Fig. 5) indicated that they all bind with a punctate pattern surrounding individual hepatocytes. Focusing through the section indicated that antibody binding was not confined to the surface of the sections. As earlier (7,9), no antibody binding was observed in the absence of primary antibody or with preimmune sera.

DISCUSSION
The existence of four transmembrane segments in the 27-kDa rat liver gap junction protein was initially suggested by Paul (11) untreated (lanes c). One replicate was stained with silver (Ag) and the remainder transferred to nitrocellulose and screened for antibody binding. The antibodies used were polyclonal, affinity purified rabbit antibody (PcAb); the two monoclonal antibodies to peptide 224-234 (26C, McAb, and 92B, McAb,); and a rat monoclonal antibody (12) to the liver gap junction protein (McAb,). encoded by a cDNA clone. Identification of protease-sensitive sites in the primary structure of the M , 27,000 rat liver gap junction protein has been used, in the present study and by others (19), to elucidate several of the basic topological features of the gap junction model. In the latter study (19), a single protease cleavage site was identified between the second and third putative transmembrane segments, localizing that hydrophilic portion of the protein to the cytoplasmic surface of the gap junction. It was inferred from other proteasedigestion experiments that the carboxyl-terminal hydrophilic region of the protein is likewise cytoplasmically disposed (19,24). However, while the amino terminus of the protein was also placed at the cytoplasmic surface, the experiments designed to test this assignment were unable to demonstrate the presence of predicted sites of either protease action or sitespecific antibody binding (19). The data obtained in our study directly demonstrate cytoplasmic localization of the amino terminus of the 27-kDa rat liver gap junction protein, confirm the cytoplasmic localization of a hydrophilic domain of the protein connecting the putative second and third transmembrane segments, and demonstrate that some part of the hydrophilic carboxyl terminus of the protein, including amino acids 224-234, is also at the cytoplasmic surface of the gap junction.
Two major fragments (10 and 17-18 kDa, respectively) have been characterized in this study as the peptides generated by treatment of intact gap junctions with a wide variety of different proteases. Our findings are consistent with earlier reports that treatment of isolated gap junctions with trypsin yields two fragments (19,24) and with the SDS-polyacrylamide gel profile of endoproteinase Lys-C generated fragments of rat liver gap junctions (19).
In the study of Zimmer et al. (19), more extensive degradation of the gap junction protein to fragments of approximately 6 kDa in size was reported with either pronase or proteinase K. We have made similar observations with pro- nase and have attributed them to an inability to remove the protease from the gap junction membranes prior to solubilization in SDS; this enzyme retains proteolytic activity even in the presence of SDS. Our recovery of larger fragments (10 and 17 kDa) relative to those obtained by Zimmer et al. (19) following digestion with proteinase K, also known for retention of activity in SDS, may simply reflect our greater success in removing the enzyme from the gap junctions during washing steps prior to solubilization in SDS. We have noted that failure to remove or inhibit protease prior to addition of SDS can introduce new cleavage sites with specific proteases, such as a cleavage after Glu-169 by S. aureus V8 protease (not shown).
Microsequencing of the 10-kDa peptides demonstrated that they are derived from the amino terminus of the protein, and that they are relatively resistant to proteolysis near their amino-terminal ends. No cleavage was detected within this part of the protein in our study using proteinase K, S. aureus V8 protease, or thermolysin, and by others using trypsin, endoproteinase Lys-C, or endoproteinase Arg-C (19). We have, however, mapped three cleavage sites between Tyr-7 and Ser-11, one site for proteinase A and two partial sites for chymotrypsin, indicating that this region is exposed at the cytoplasmic surface of the gap junction. Microsequencing of the 17-18-kDa peptides, which begin in the region of the protein including amino acids 109-124, indicates that these peptides represent the carboxyl-terminal portion of the protein.
The fact that the 13-amino terminal residues are primarily hydrophobic in nature suggests that they may, in part, be buried in the lipid bilayer. Hence, while protease specificity could be a factor, partial occlusion of the amino terminus by the lipid bilayer may account for both the observed paucity of proteolytic cleavage sites and the reported inability of antibodies to a synthetic peptide, corresponding to amino acids 7-21, to bind to gap junctions by immunocytological criteria (19).
The location of the amino terminus at or near the cytoplasmic surface would explain the lack of detectable N-linked glycosylation of the 27-kDa rat liver gap junction protein which, based upon the presence of the sequence Asn-Trp-Thr, would be expected to occur at Asn-2 (26). The enzymatic machinery necessary for this type of processing is found in the lumen of the endoplasmic reticulum and Golgi apparatus, exposure to which occurs only at the extracellular domains of membrane and secretory proteins (27).
Once the amino terminus is fixed at the cytoplasmic surface, one may infer localization at the cytoplasmic surface for the carboxyl terminus of the protein (see below) as well as for a hydrophilic loop connecting the second and third transmembrane spans. Data described here for four proteases (Fig. 3) have demonstrated a series of five protease-sensitive sites from amino acids 109-123, within the predicted span of a cytoplasmically disposed loop connecting transmembrane segments 2 and 3 (amino acids 94-132). Cleavage by endoproteinase Lys-C or trypsin after  is consistent with these findings.
It is noteworthy that the fragment obtained with proteinase A (cleavage between residues 108 and 109) appears to be significantly larger than that generated by S. aureus V8 protease (which cleaves between residues 109 and 110) based upon their mobilities on SDS polyacrylamide gels (Fig. 2). While these two peptides begin within one amino acid of each other, the overall size difference implies an additional cleavage site(s) close to the carboxyl terminus of the protein. The carboxyl terminus of the protein must, therefore, also be located on the cytoplasmic surface of the gap junction, a conclusion reached earlier by others (19,24) using a similar rationale. The presence of cleavage sites close to the carboxyl terminus of the protein could also account for production of the 23-kDa peptide (Fig. 2) as well as the proteinase Agenerated peptide migrating somewhat more slowly than the 18-kDa product (Fig. 1) as intermediates in formation of the 10-and/or  Our polyclonal antibodies bind to all of the 17-18-kDa COOH-terminal generated peptides in this study (Fig. 2), an observation similar to that described by other investigators (19).
Further evidence for cytoplasmic localization of the carboxyl terminus comes from characterization of monoclonal antibodies to a peptide corresponding to amino acids 224-234 in the predicted sequence of the protein. As in the case of protease, accessibility of bulky immunoglobulins and fluorescein-conjugated second antibodies is restricted to the cytoplasmic surface of the gap junction. As with the polyclonal antibodies, these monoclonal antibodies bind to the 17-18-kDa carboxyl terminus-derived proteolytic peptide on Western blots (Fig. 4). The patterns revealed by indirect immunofluorescence on frozen sections of rabbit liver (Fig. 5) are indistinguishable from those analyzed here and earlier with a number of polyclonal antibodies to the 27-kDa rat liver gap junction protein (7,(9)(10)(11)(12)(13)19). Hence, binding of these sitespecific monoclonal antibodies demonstrates that at least this part of the hydrophilic carboxyl terminus is cytoplasmically disposed and suggests that the polyclonal antibodies also bind to this portion of the protein. Given the relative paucity of hydrophobic amino acids in the remainder of the protein (12 of 59 amino acids in the sequence from residue 235 to the predicted carboxyl-terminal cysteine 283) and their distribution, it is unlikely that this part of protein could insert itself in a lipid bilayer. It is likely, based upon the data presented in Figs. 4 and 5, that the rat monoclonal antibody developed by others (12) also binds to the carboxyl terminus region of the gap junction protein.
Phosphorylation of the 27-kDa rat liver gap junction protein has been observed (29, 30) and appears to be involved in channel regulation (29). The constraints thus far imposed upon our topological model of the gap junction protein suggest the cytoplasmic localization of a phosphorylatable serine (residue 233), which is preceded by two basic amino acids (residues 230 and 231) as is required for phosphorylation by the CAMP-dependent protein kinase (31). The presence of additional serine and tyrosine residues which are potential substrates for other protein kinases has been noted (17) in this cytoplasmically disposed carboxyl terminus of the protein.
Collectively, our data provide strong support for the model suggested by hydropathicity plots (11) and provide several constraints upon models being developed based upon crystallographic and image processing techniques. Additional experiments will be necessary to further delineate the topological features of the gap junction protein, especially the assignment of residues to the extracellular domain of the protein. The data described in this report do, however, reinforce the emerging model of a gap junction protein spanning the bilayer four times with both the amino and carboxyl termini localized at the cytoplasmic surface. gelides, Michael Bennett, and David Spray for their suggestions 15. Young, J. D., Cohn, Z. A., and Gilula, N. B. (1987) Cell 48, 733concerning the manuscript, and to Kathe M. Hertzberg for her help 743