The novel mouse connexin39 gene is expressed in developing striated muscle fibers

Gap-junctional channels consist of two hemichannels (connexons), which are contributed by each of two adjacent cells thereby allowing direct cell-to-cell communication. The connexon comprises six connexin (Cx) subunit proteins that cross the plasma membrane four times and consist of two extracellular domains, one cytoplasmic loop and N-terminal as well as C-terminal regions of variable length. The expression of connexin genes is regulated in a spatiotemporal pattern in different cell types. Different combinations of connexins in gap-junction channels further enlarge their regulatory possibilities (see Kumar and Gilula, 1996). Besides the well-documented function of gap junctions as intercellular channels exchanging electrical potentials, nutrients, ions, metabolites, waste products and second messengers, recent evidence supports the notion that hemichannels are also involved in intercellular signalling without being incorporated into gap junctions (Goodenough and Paul, 2003; Ebihara, 2003).

So far, 20 connexin genes have been discovered in the mouse and 21 in the human genome (Söhl and Willecke, 2003). Except for hCx25, mCx33 and hCx59, connexin genes seem to be orthologous between mouse and man according to their deduced nucleotide and protein sequence. However, a low degree of sequence similarity between two orthologues appears to correlate with higher divergence in their expression profiles (Söhl et al., 2003). Although most connexins are expressed during embryogenesis, developmental defects were only found in a few of the corresponding connexin-deficient mice. Redundancy in the expression of different connexins may have prevented the detection of essential functions of distinct connexins before and after implantation (Kidder and Winterhager, 2001). However, functional loss of three mouse connexins (i.e. Cx26, Cx43 and Cx45) results in independent developmental defects that lead to death between day 10 and 21 of embryonic development (Willecke et al., 2002).

The functional impact of gap-junctional coupling, if any, during development of striated muscle fibers as well as the expression pattern of different connexins in this tissue, is largely unknown. Ultrastructural analysis demonstrate gap junction-like structures between interacting myoblasts, and between myoblasts and myotubes in developing muscles of the rat (Kalderon et al., 1977; Ling et al., 1992) but the expression of only Cx40 and Cx43 has been reported in subcompartments of developing muscles in mouse and rat (Constantin and Cronier, 2000).

Here we characterize the mCx39 gene that shows, similar to connexin36, an interrupted coding region located on exon 1 and exon 2. Cx39 expression is upregulated during the final third of embryogenesis, remains present around birth but could not be detected in various adult tissues when tested by northern blot hybridization and immunoblot analysis. Antibodies directed to a C-terminal peptide of mCx39 yielded strong immunosignals in tissues where myogenesis occurred in the mouse embryo after the second embryonic week. HeLa cells transfected with the DNA encoding Cx39 served as a positive control for antibody specificity but showed no coupling after microinjection of various tracers. However, spreading of Alexa488 but not rhodamine dextran, microinjected into myotubes of the neonatal diaphragm, suggests functional coupling of mCx39 gap-junctional channels in this tissue, as expression of no other known connexin could be verified in these cells.

We used the HUSAR-derived subprogram GENESCAN (http://genius.embnet.dkfz-heidelberg.de) to search for suitable N-terminal sequences upstream of a novel mouse genomic sequence and its putative human orthologue that could lead after splicing to possible connexin coding regions. These were denoted mCx39 and hCx40.1 (Eiberger et al., 2001). In order to clone full-size coding regions of these putative connexins, large scale upstream primers (mCx39N-Term and hCx40.1N-Term, see Table 1) were designed, which spanned the whole coding sequence of exon 1 and annealed at the 5′ region of exon 2. In this case, complete coding regions of mCx39 and hCx40.1 were restored omitting the putative splice event. An additional XhoI restriction site was integrated in the 5′ end of the upstream primer. The corresponding downstream primers contained an EcoRI restriction site (see Table 1). The coding region of mCx39 was amplified by PCR from mouse genomic DNA (C57BL/6 strain), and the coding region of hCx40.1 was PCR-amplified from human genomic DNA (gift of S. Dhein, Herzzentrum Leipzig, Leipzig, Germany). Fragments were excised after gel electrophoresis, purified by using the Perfectprep® Gel Cleanup procedure for PCR-fragments (Eppendorf, Hamburg, Germany), subcloned into the pGEM-T_easy vector system suitable for cloning of PCR-fragments (Promega, Madison, WI, USA) and commercially sequenced by AGOWA, Berlin, Germany.

For stable transfection of HeLa cells, the mCx39 coding region was cut out of the pGEM-T_easy vector with XhoI and EcoRI and cloned into the XhoI/EcoRI restricted multiple cloning site of the expression vector pMJgreen (J. Degen, unpublished results). This vector contained an AsnI/StuI fragment (2570 bp) derived from the pEGFP-N1 vector (Clontech, Palo Alto, CA, USA), which was cloned into the NotI/ClaI treated pBEHpac18 vector (∼3700 bp) (Horst and Hasilik, 1991). The resulting expression plasmid was denoted pMJCx39.

Total RNA from various mouse tissues (C57BL/6 strain) was isolated using the TRIzol® -reagent according to instructions provided by the manufacturer (GibcoBRL, Eggenstein, Germany). Aliquots of 2 μg RNA were incubated with 1 μg oligo(dT)₁₅ primer (0.5 μg/μl; Promega) for 10 minutes at 70°C in a total volume of 12 μl and briefly chilled on ice. After primer annealing, polyA⁺ RNA was reverse transcribed by using 200 U of SuperScript™ II reverse transcriptase (Invitrogen, Karlsruhe, Germany). The RT buffer contained 50 mM Tris-HCl (pH 8.3), 75 mM KCl, 3 mM MgCl₂, 10 mM dithiothreitol (DTT), 40 U RNasin (Promega) and 500 μM of each dNTP (Promega). Reaction mixtures (20 μl) were incubated for 50 minutes at 42°C and then for 15 minutes at 70°C. Aliquots of the transcribed cDNA (1/20 of tissue and cell reaction mix, ∼0.1 ng) were amplified using the following combination of Cx39 specific primers: upstream primer mCx39KXhoI and downstream primer mCx39Eco (see Table 1). Primer combinations specific for β-actin were used (De Sousa et al., 1993). For amplification of Cx26, Cx30.3, Cx31, Cx31.1, Cx32, Cx37, Cx40, Cx43, Cx45, Cx46 and Cx50, primer pairs were used as reported (Davies et al., 1996). Cx30, Cx36 and Cx47 were amplified using primer combinations as reported (Dahl et al., 1996; Söhl et al., 1998; Teubner et al., 2001). Other primer combinations are listed in Table 1. Amplicon sizes were deduced from published sequences or from Table 1. Reaction mixtures (50 μl) contained 20 mM Tris-HCl (pH 8.4), 400 μM dNTPs, 2 mM MgCl₂, 50 mM KCl, 0.4 μM of each primer and 1 unit Taq DNA-polymerase (Promega). PCR was carried out for 35 cycles using a PTC-100 Thermal Cycler (MJ Research, Watertown, MA, USA) with the following program: first denaturing step at 94°C for 3 minutes, then a cyclic denaturing at 94°C for 1 minute, annealing at 65°C for 1 minute, elongation at 72°C for 2 minutes and a final elongation for 7 minutes. The annealing temperature was gradually reduced by 1°C during each successive cycle until the annealing temperature of 55°C was reached. After gel electrophoresis in a 0.8% agarose gel, DNA fragments were visualized with ethidium bromide (Sambrook and Russel, 2001). The integrity of the primer combination used to amplify Cx39 was verified by using the vector construct containing the restored Cx39 cDNA sequence (pMJCx39) as template. Fragments of interest were cloned and sequenced as mentioned above.

Total RNA (20 μg) was separated (Sambrook and Russel, 2001) and transferred onto HybondN nylon membrane (Amersham International, Amersham, Bucks, UK) by capillary diffusion in 20×SC and prehybridized at 42°C for 30 minutes in preheated ULTRAhyb™ hybridization solution (Ambion, Austin, Texas) according to the manufacturer. All membranes were probed with a 765 bp fragment of mouse Cx39 that was cut out of the pMJCx39 vector by PstI/EcoRI double digestion (see above). It coded for half of the presumptive cytoplasmic loop up to the stop codon (from nucleotides 1877 to 2642, see Fig. 1). The probe was [³²P] labelled, using the random primed method (multiprime labelling Kit, Amersham) to a specific activity of 0.7-1.0×10⁹ cpm/μg DNA and added to the ULTRAhyb™ hybridization solution at 1.4× 10⁶ cpm/ml. Hybridization at high stringency was carried out overnight at 42°C. Filters were first washed twice for 5 minutes in 2×SC/0.1% SDS at 42°C, then twice for 15 minutes in 0.1×SC/0.1% SDS at 42°C and exposed to XAR X-ray film (Eastman Kodak, Rochester, NY, USA) with an intensifying screen at –70°C. The amount of total RNA on the northern blot was roughly standardized relative to the intensities of ethidium bromide-stained 18S rRNA.

Two different peptides corresponding to part of the cytoplasmic loop (AVGGACRPQVPDL) or the C-terminus (TSGSAPHLRTKKSEWV) of mCx39 were commercially synthesized, coupled to keyhole limpet hemocyanin, and injected into rabbits following the doubleX procedure at Eurogentec (Seraing, Belgium) for immunization. Polyclonal serum was affinity purified with each peptide by using separate HiTrap affinity columns (Pharmacia Biotech, Freiburg, Germany) according to the guidelines of the manufacturer. After elution with 3 M KSCN in phosphate-buffered saline without Mg²⁺ (PBS^–; pH 7.4) and dialysis against PBS^–, containing 0.01% sodium azide, both antibody fractions were concentrated by ultrafiltration through Centricon microtubes 30 (Amicon, Beverly, MA, USA) and finally stored in PBS^- with 0.5% bovine serum albumin (BSA, Miles Diagnostics, Kankakee, IL, USA).

Protein homogenates from different mouse tissues (C57BL/6 strain) and protein extracts from cultured HeLa cells were obtained by pulverizing frozen tissue in liquid nitrogen and homogenizing by sonification in 1×Complete® (Roche, Mannheim, Germany). Protein concentrations were determined after using the protein determination kit with bichinchoninic acid (BCA, Sigma), according to the instructions provided. Aliquots of 100 μg protein per lane in Laemmli sample buffer (Traub et al., 1994) containing mercaptoethanol were separated by 12% SDS-PAGE and transferred to nitrocellulose membranes (Sambrook and Russel, 2001). Membranes were blocked with blocking solution (8.5 mM Tris-HCl, 1.65 mM Tris-base, 50 mM NaCl and 0.1% Tween) containing 5% milk powder for 1 hour and incubated for 2 hours at room temperature with a 1:100 dilution of polyclonal Cx39 antibodies. After washing (twice for 5 minutes and twice for 10 minutes) in washing buffer, membranes were incubated for 1 hour at room temperature with a 1:20,000 dilution of horseradish peroxidase-conjugated goat anti-rabbit (Dianova) and washed again twice for 5 minutes and twice for 10 minutes in washing buffer. For detection of the complexed antibodies, we used the SuperSignal® West Pico Chemiluminescent detection kit (Pierce, Rockford, IL, USA). Standardization of immunoblots was performed by using a 1:500 dilution of mouse monoclonal β-actin antibodies (Sigma) and a 1:5000 dilution of horseradish peroxidase-conjugated goat anti-mouse secondary antibodies (Dianova). Bands were visualized by chemiluminescence as described for northern blot analysis.

Cryosections (10 μm) or whole diaphragms of mouse embryos at different developmental stages and neonatal mice were fixed in absolute ethanol at –20°C for 10 minutes and washed twice in PBS^– for 5 minutes. RNase 2000 (Roche) diluted 1:100 in PBS^– was applied for 15 minutes. Then the sections were washed twice with PBS^–, incubated with 40 μM propidium iodide for 15 minutes and preincubated for 45 minutes in blocking reagent (PBS^– containing 4% BSA and 0.1% Triton X-100). For detection of Cx39, sections were incubated for 2 hours with a 1:100 dilution of affinity-purified polyclonal Cx39 antibodies at room temperature. After three washes in PBS^– for 5 minutes, tissue samples were stained for 45 minutes with a 1:400 dilution of Alexa488-conjugated goat anti-rabbit secondary antibodies (MoBiTec, Göttingen, Germany). Cx40 antibodies (Alpha Diagnostics, USA), Cx43 antibodies (Traub et al., 1994), Cx26 antibodies (DPC-Biermann), Cx32 antibodies (Zymed) and Cx37 antibodies (Traub et al., 1998) were used in this study at dilutions recommended for immunofluorescence analysis of tissues. After incubation, samples were washed three times in PBS^– for 5 minutes and mounted in fluorescence mounting solution (DAKO, Hamburg, Germany). Fluorescent signals were recorded with a Zeiss laser-scanning microscope (LSM 510) equipped with a ×40 objective and appropriate filters.

Cryosections (20 μm) of diaphragms from neonatal mice were fixed in 0.2 % glutaraldehyde for 5 minutes. Staining was performed using the substrate 5-bromo-4-chloro-3-indolyl-β-galactoside (X-gal) according to Krüger et al. (Krüger et al., 2000).

Human cervix carcinoma HeLa cells (ATCC CCL1; American Type Culture Collection, Rockville, MD, USA) were cultured in Dulbecco's modified Eagle's medium (DMEM) containing low glucose, 2 mM glutamine, 10% fetal calf serum, 100 U/ml penicillin and 100 μg/ml streptomycin (all from Life Technologies). HeLa transfectants were maintained in standard medium containing puromycin (0.5 mg/ml; Sigma, Deisenhofen, Germany) and cultivated in a moist atmosphere containing 10% CO₂ at 37°C. HeLa cells were transfected with 20 μg of the linearized pMJgreen plasmid containing mouse Cx39 cDNA (pMJCx39) using the calcium phosphate transfection protocol (Elfgang et al., 1995). Briefly, after exposure to DNA for 24-48 hours, puromycin was added to the medium. Clones were picked after 2 weeks and grown under selective conditions. Iontophoresis of Lucifer yellow, propidium iodide, ethidium bromide, DAPI and neurobiotin was performed as described (Elfgang et al., 1995).

After sacrificing the neonatal mice, the whole diaphragm was dissected and transferred into Krebs-Henseleit-buffer with 10 mM glucose, warmed to 37°C and oxygenated with carbogen. Microinjection of the dye Alexa488 (570.48 Da; 5% Alexa Fluor 488 hydrazide in H₂O; Molecular Probes, Leiden, The Netherlands) or rhodamine dextran (10 kDa, 1% in H₂O, Sigma Deisenhofen, Germany) was performed by iontophoresis into a single myotube under visual inspection. The dye was injected with 2 nA for 2 minutes (Romualdi et al., 2002). Sharp glass electrodes were filled with 100 mM KCl. We measured the membrane potential before and after injection; injections at less than –30 mV of membrane potential were excluded from the evaluation. The intercellular dye spreading was followed by visual inspection under UV light and phase-contrast optics. Micrographs were taken on the Laser-scanning microscope.

We recently identified two nucleotide sequence tags (one of mouse and one of human origin), encoding three spaced cysteine motifs in two putative extracellular connexin domains: [C(X₆)C(X₃)C], Cx39 residues 54-65 and [C(X₄)C(X₅)C], Cx39 residues 153-164 (see Fig. 1), which were identical to all other connexin genes except Cx31 (Eiberger et al., 2001). In both tags, the N-terminal coding sequence was not found adjacent to the cysteine motifs. Thus, we used the HUSAR-based GENESCAN program to search for possible N-terminal coding sequences in the 5′ upstream regions of both genomic sequences. One of them, encoding 21 amino acid residues, was predicted to be spliced in-frame to the partial mouse connexin coding region, about 1.5 kb further downstream, thus resulting in an uninterrupted reading frame of 1095 bases (stop codon inclusive) coding for 364 amino acid residues (39.995 Daltons) (see Fig. 1). Accordingly, a possible N-terminal coding sequence was found 1.95 kb upstream of the human connexin coding region leading to an open reading frame (ORF) of 1113 bases (stop codon inclusive) covering 370 amino acids (40.139 Daltons) after splicing. Thus, we denoted these novel connexins mCx39 and hCx40.1, respectively, in accordance with their theoretical molecular masses (Eiberger et al., 2001). We did not group these novel connexin genes into one of the previously described α-, β- or γ-groups (Gimlich et al., 1990) owing to their low sequence similarity to any of these subgroups. Instead, structures of mCx39 and hCx40.1 resembled those of murine and human Cx36 (Condorelli et al., 1998; Söhl et al., 1998; Belluardo et al., 1999) as depicted at the bottom of Fig. 1. The putative splice donor and acceptor sites are in accordance with canonical motifs described for eukaryotic genes (Shapiro and Senapathy, 1987). In order to determine the extent of the 3′ untranslated region (3′-UTR) of mCx39, the sequence downstream of the translational stop codon (TGA) was screened using the HUSAR-device `Termination' for canonical AAUAAA poly(A)⁺ recognition motifs, which were discovered about 1.3 kb and 1.8 kb further downstream (Fig. 1; position 3995-4000 and 4470-4475). Thus, the size of exon 2 might be 2.5 or 3.0 kb, respectively. This largely matches the size of already cloned Cx39 cDNA sequences, denoted to be full-sized (GenBank accession numbers AK034663 and AK053016).

The orthologous counterpart of mouse Cx39 was calculated to be human Cx40.1. Although the percentage of nucleotide and protein sequence similarities (65% identical nucleotides; 56% identical amino acid residues) are the lowest among orthologous connexin pairs (Söhl et al., 2003), both sequences were readily aligned when compared to a pool of 19 mouse and 20 human connexins (Eiberger et al., 2001). hCx40.1 was weakly expressed in some adult human tissues tested but there was no indication of mCx39 expression in the corresponding adult mouse tissues (Söhl et al., 2003).

There are several observations that are consistent with the low percentage of sequence similarity and thus the low extent of orthology between both connexins: mouse Cx39 and human Cx40.1 do not only differ in their total number of amino acid residues (mCx39, 364; hCx40.1, 370) but also in sequence segments of putative cytoplasmic domains. A deletion of 22 residues in the second part of the cytoplasmic loop of mCx39 compared to hCx40.1, led to an unusual short cytoplasmic loop of about 21 amino acids. In contrast, hCx40.1 contains two minor and one larger deletion of 17 residues in the C-terminal part (Fig. 2). The overall sequence similarity in different aligned domains is very low, although it is significantly higher in domains like the extracellular loop or the N-terminus. However, although the C-terminal domain shows the lowest amino acid similarity between mCx39 and hCx40.1, the final 12 residues present a conserved cluster out of which 10 residues are identical (APHLR-X-X-KSEWV) (Fig. 2). This might serve as a conserved binding motif for proteins influencing the localization or function of both connexins within their species. Finally, the GC content of both reading frames is relatively high but differs in its total value (mCx39, 56%; hCx40.1, 68%).

Using the HUSAR-derived subprogram PROSITE, we identified various potential consensus motifs for post-translational modifications from which putative phosphorylation sites were directly compared between mCx39 and hCx40.1 (Fig. 2). In general, both connexins largely differ in their total number, position and kinase subtype of phosphorylation sites. Mouse Cx39 contains ten sites whereas hCx40.1 has 12. Only one conserved putative casein kinase II (CKII)-site is evident in both connexins at residues 290 (mCx39)/301 (hCx40.1).

The NCBI Annotation Project (http://www.ncbi.nlm.nih.gov/Homology/) has localized both connexin genes on their corresponding mouse and human chromosomes. Mouse Cx39 (on contig NT_039674.1) was found on chromosome 18, A1 ∼9.3 cM, between a putative cyclin-box carrying protein upstream and the mouse orthologue for the Drosophila-derived frizzled homolog 8 (Fzd8) downstream. Human Cx40.1 (on contig NT_008705.14) is localized on chromosome 10p11.21 between the cyclin fold protein1 (CFP1) upstream and the human orthologue for the Drosophila-derived frizzled homolog 8 (FZD8) downstream. Applying the NCBI-supported program HomoloGene (http://www.ncbi.nlm.nih.gov/Homology/) resulted in hCx40.1 and mCx39 as calculated orthologues showing 77.5% identity based on genomic sequences. The human/mouse homology map aligned the chromosomal regions around both human and mouse orthologues of the Drosophila-derived frizzled homolog 8. Thus, hCx40.1 and mCx39 are localized in a syntenous genomic region. Up to now, no other connexin genes are mapped on mouse chromosome 18 and human chromosome 10 according to the Human Genome Browser (www.ensembl.org/Homo_sapiens/familyview.html).

The translational start point and the Kozak-consensus motif (Kozak, 1989) located immediately upstream influencing translational initiation, were directly compared between mCx39 (–12 ttttcttcaagcATGg +4) and hCx40.1 (–12 cattctggaagcATGg +4). We calculated theoretical values for translational efficacy (Cx39, 4.14 and Cx40.1, 4.47) according to Iida and Masuda (Iida and Masuda, 1996), and found both values to be within a similar range, probably because 12 out of 16 bases within this motif are identical. Total values concerning translational efficacy range from 6.7421 to 0.3219 (Iida and Masuda, 1996). To validate translational initiation at other putative ATG start codons of mCx39, we also calculated their theoretical translational efficacy. The ATG at position 1614-1616 in Fig. 1 is embedded into a Kozak-consensus motif partially coinciding with the splice acceptor site of exon 2. Without splicing, its efficacy is about 3.54 but after splicing, it slightly increases to 3.82. Owing to these relatively low values, translational initiation from this ATG seems unlikely. Furthermore, efficacy of a second ATG in exon 1 (position 139-141, Fig. 1) is about 5.63 and thus more convenient for translational initiation compared to 4.14 of the ATG at position 79-81. However, the scanning hypothesis contradicts the use of this second ATG further downstream within exon 1 (Kozak, 1989).

After functional cloning, the complete mCx39 and hCx40.1 coding DNAs from either mouse genomic DNA (C57BL/6 strain) or human genomic DNA, respectively, were sequenced. We identified four base-pair exchanges compared to the already published mCx39 sequence (GenBank accession number AJ414562) (Eiberger et al., 2001), which was taken from the Celera mouse database assembled after genomic sequencing of different mouse 129Sv-derived sub-strains. Only one of these nucleotide exchanges resulted in an amino acid exchange (position 345, Gly to Cys) (Fig. 1 marked in bold italics).

To determine the transcriptional pattern of mCx39, we hybridized total RNA from various tissues to a probe specific for Cx39. In the whole mouse embryo, we found two prominent signals at about 2.7 and 3.2 kb, highly expressed at ED (embryonic day) 13.5 and ED16.5 and still present in newborn animals (P0). Neither signals were seen at ED11.5 in the embryo, nor in embryonic yolk sac, placenta or in utero (Fig. 3A). Analysis of neo- and postnatal tissues revealed that Cx39 mRNA is only slightly expressed in eye, chest and head at P0 as well as in eye and chest of P2 mice but could not be found in any of the other tissues tested (Fig. 3B). The following adult mouse tissues were found negative for Cx39 expression after northern blot hybridization (not shown): whole brain, heart, stomach, kidney, liver, spleen, skin, skeletal muscle, lung, testis, uterus, intestine, ovary, pancreas, thymus, tongue, cortex, cerebellum, hippocampus, epididymis, prostate, bladder, colon, oesophagus, salivary gland, lachrymal gland, lens and retina. Furthermore, none of these adult tissues yielded amplicons specific for Cx39 after RT-PCR analysis (not shown).

Instead, an amplicon of about 1.1 kb, representing the spliced form of mCx39, was amplified and sequenced using cDNA fractions from the whole embryo at ED11.5, 13.5, 16.5 and newborn (not shown) and from ED16.5 brain, eye, chest, skin, intestine, heart, lung and liver (Fig. 4A). Additionally, we detected RT-PCR amplicons of mCx39 in newborn (P0) brain, eye, chest, heart and skin but not in lung, liver and intestine (Fig. 4B). As total RNA harvested from chest contains a mixed population of RNA from bone and the attached intercostal muscle, three different developing muscles were prepared and analyzed separately (diaphragm, intercostal muscle and hind-limb muscle) revealing amplicons for Cx39 (Fig. 4B). In contrast, different postnatal (P2 to P7) organs of the mouse were all found to be negative (not shown). Thus, transcription of mCx39 seems to be confined to a temporal window starting after ED11.5 and ending after birth.

To exclude the transcription of other known members of the connexin gene family, a detailed RT-PCR analysis was performed with cDNA from P0 mouse diaphragm (Fig. 4C). Amplicons of the appropriate size were found for Cx26, Cx32, Cx37, Cx39, Cx40 and Cx45 whereas the absence of amplicons for Cx29, Cx30, Cx30.2, Cx30.3, Cx31, Cx31.1, Cx33, Cx36, Cx43, Cx46, Cx47, Cx50 and Cx57 suggests that these connexins are not expressed in the neonatal mouse diaphragm.

In order to reduce the risk of crossreactivity, we raised polyclonal antibodies against two different epitopes of Cx39, which show no similarities to other connexins (boxed in Fig. 1). Antibody fractions directed to the cytoplasmic loop or the ultimate C-terminus were used to probe immunoblots of lysates from HeLa cells, stably transfected with the Cx39 coding region (HeLa-Cx39) in order to verify their specificity. One prominent protein band migrating at about 40 kDa was detected in lysates of Cx39-transfected HeLa cells after incubation with the antibody fraction directed to the C-terminus missing in lysates of wild-type HeLa cells. Furthermore, a protein fraction of about 80 kDa was very often detected, possibly indicating a dimerization of Cx39, since this signal was also absent from lysates of wild-type cells (Fig. 5A). The antibody fraction directed to the cytoplasmic loop yielded no detectable signals and was therefore omitted from further studies. In tissue lysates, a band migrating at around 40 kDa was detected in total mouse embryo (ED16.5, Fig. 5A), in embryonic brain, skin, intestine, liver, lung and diaphragm, but not in heart or bone (Fig. 5B). In newborn brain, heart, eye, liver, diaphragm, intercostal muscle and hind limb muscle, a 40 kDa signal was found which was absent from bone, skin, lung and intestine (Fig. 5C). In all tissue lysates from three-month-old animals, no protein was detected by Cx39 antibodies (Fig. 5D). Ponceau staining of membranes (not shown) and reincubation of blots with actin antibodies indicated that similar amounts of protein were loaded (Fig. 5B-D). Thus, Cx39 protein expression was detected only in ED16.5 and neonatal (P0) tissues but was absent from all the other adult tissues tested. These results correspond well to the transcriptional pattern of mCx39.

We studied the expression of Cx39 protein in the mouse embryo by immunocytochemistry. After staining sagittal cryosections (ED16.5 and P0), immunoreactive signals were detected in different parts of the body, in structures around the developing eye, between cells of the pericard, in the tongue and in the region around the occipital bone of the head (not shown) and most prominently in the zone between the lung and liver harbouring the preliminary diaphragm (Fig. 6A-D). Signals around the eye were located outside the sclera, thus ruling out mCx39 expression in the developing retina or lens. The diaphragm transverses the whole embryo from dorsal to ventral corresponding to the pattern of immunofluorescent signals. A punctate staining pattern was also seen in the embryonic and neonatal intercostal muscle (Fig. 6E,F) and observed at a higher magnification in the area of the embryonic hind limb muscle (Fig. 6G). Punctuated patterns of Cx39 immunofluorescence were generally detected in distinct regions that developed into either the functional diaphragm or intraperitoneal striated muscles. This staining was not visible after omitting the primary antibodies, therefore excluding cross-reactivity of the secondary antibodies (Fig. 6B,D). To verify that Cx39 can be excluded from adult diaphragm, we studied diaphragms from three-month-old mice. In accordance with the results from immunoblots of adult tissues (Fig. 5D), no Cx39 immunoreactivity was found (not shown). In contrast, a whole-mount micrograph of a P0 diaphragm stained with anti-Cx39 antibodies shows abundant mCx39 expression throughout this tissue (see Fig. 9A).

Very recently, we found transient expression of mCx39 protein by immunofluorescence analysis of regenerating adult skeletal muscle 2-10 days after injection of BaCl₂ into the tibialis anterior muscle of 6-week-old mice. These results suggest that the mCx39 protein, in addition to its expression in striated muscles during development, may also contribute to muscle regeneration in adult mice.

Coupling-deficient HeLa wild-type cells (Elfgang et al., 1995) were stably transfected with a vector directing Cx39 protein expression. Cx39 transfectants showed punctate staining on plasma membranes of adjacent cells only with the antibody fraction directed to the C-terminus of Cx39 (Fig. 6H). Staining was absent when primary antibodies were omitted (not shown). In contrast, immunofluorescence studies revealed no punctate staining of wild-type cells after incubation with Cx39 antibodies and subsequent secondary antibodies (not shown). However, microinjections of dyes or tracers like Lucifer yellow, propidium iodide, DAPI, ethidium bromide and neurobiotin into Cx39-HeLa transfectants yielded no transfer to neighboring cells, similar to that observed in the wild type (not shown). Microinjections in Cx43-HeLa transfectants as controls revealed the expected permeation of the injected tracers to higher orders of surrounding cells (see Elfgang et al., 1995). Thus, although Cx39 appeared to be located properly in the plasma membrane of Cx39-HeLa transfectants, no tracer transfer was seen after microinjection. However, this does not totally exclude the exchange of small molecules or ions through Cx39 gap junctions between Cx39-transfected HeLa cells.

The detailed RT-PCR analysis (Fig. 4C) of all known connexins in the mouse yielded amplicons for Cx26, Cx32, Cx37, Cx39, Cx40 and Cx45. By immunofluorescence analysis, however, presence of Cx26 and Cx32 could be confined to the developing liver rather than to the diaphragm (Fig. 7A,B). Thus, positive RT-PCR signals for Cx26 and Cx32 might reflect a contamination of residual liver specimen harvested during diaphragm preparation. Cx37, which is expressed in endothelial tissue (Traub et al., 1998), was excluded from myotubes by immunofluorescence analysis (Fig. 7C). Neither Cx40 nor Cx43 (Fig. 7D,E), previously shown in developing skeletal muscles (Dahl et al., 1995; Reinecke et al., 2000), could be detected in myotubes. Immunofluorescence signals for Cx40 and Cx43 suggest that these connexins are expressed in the vascular endothelium (Fig. 7D) and in limiting membranes of the diaphragm (Fig. 7E). Based on expression analysis of the LacZ reporter gene in Cx45(LacZ⁺) mice (Krüger et al., 2000), we conclude that Cx45 is expressed in blood vessels (Fig. 7F) but not in myotubes of neonatal diaphragm.

During development of striated muscle fibers in the rat, gap junctions have widely been demonstrated between secondary myotubes or between secondary and primary myotubes from ED18 up to ED 21 (Ling et al., 1992). These authors detected gap-junctional channels at ED16 in the rat diaphragm, a developmental stage at which only primary myotubes are prominent in this tissue.

We investigated whether mCx39 is also expressed at the corresponding time point in mouse diaphragm. After immunofluorescence analysis of cryosections from mouse ED14.5 intercostal muscle and diaphragm, we found specific staining for mCx39 even at this very early developmental stage (Fig. 8A,B; no signals could be detected in the absence of primary antibodies (Fig. 8C).

As it has been shown that in the mouse diaphragm secondary myoblasts progressively enter muscle differentiation to secondary myotubes after ED15.5 (Wigmore and Dunglison, 1998), mCx39 gap-junctional channels may already be present at ED14.5 between primary myotubes.

Dye injections into single cells of dissected neonatal diaphragm (P0) were used to assess functional coupling of mCx39 gap-junction channels. Before and after injecting into a single myotube, we measured a membrane potential of about –30 mV and thus confirmed its physiological integrity. Appropriate cells were then filled with Alexa488 for two minutes. The dye spread into the neighboring myotubes up to the second order (Fig. 9B). Dye transfer between myotubes in the mouse diaphragm at P0 implied functional coupling of gap-junction channels in this tissue. In contrast, rhodamine dextran (10 kDa molecular mass) injected into a single myotube failed to spread into the neighbouring cells (Fig. 9C). Owing to its size, this dye does not permeate through gap-junction channels but via cytoplasmic bridges. Thus, our results indicate the absence of such cytoplasmic bridges, instead coupling in neonatal diaphragm is probably due to functional expression of Cx39 in developing striated muscle fibers of this tissue.

The extent of sequence similarity between mCx39 and hCx40.1 is the lowest among all known orthologous connexins between men and mouse (Söhl et al., 2003). The coding region of the mCx39 and hCx40.1 genes is interrupted by an intron, which has been so far only been described for the orthologous connexin pair mCx36 and hCx36 (Condorelli et al., 1998; Söhl et al., 1998; Belluardo et al., 1999). mCx39 and hCx40.1 show an unexpectedly high GC content of similar magnitude to mCx30.2 and hCx31.9 (Nielsen and Kumar, 2003), share a syntenous chromosomal region, contain a common motif at their C-terminal ends (possibly indicating similar processing and attachment), and have nearly the same theoretical translational efficacy.

In contrast, mCx39 and hCx40.1 differ largely in their transcriptional pattern, with respect to the predicted phosphorylation sites from which only one out of twelve is similar in both proteins. Mouse Cx39 is transcribed only from ED11.5 up to birth, whereas hCx40.1 mRNA could be detected in various adult human tissues (Söhl et al., 2003) and was also found in cDNA libraries generated from human embryonic stages. However, their major differences occur in sequence similarities and theoretical phosphorylation sites. These two features imply that both genes might have unequally adopted or lost distinct functions during evolution. However, according to its expression profile, hCx40.1 must have adopted additional tasks to accomplish during adulthood in humans (Söhl et al., 2003). In conclusion, gross regulation of both proteins (mCx39 and hCx40.1), such as translation and intracellular transport/assembly, seem to be similar, whereas fine regulation with respect to spatiotemporal expression, phosphorylation and possibly permeability is likely to be different.

To access cell-type specific expression of mCx39, immunofluorescence studies using antibodies to its C-terminus were performed which revealed mCx39 protein in embryonic structures related to myogenesis. Immunofluorescent signals within the developing diaphragm, intercostal striated muscle and tongue as well as around the eye and occipital bone suggested the presence of Cx39 between myoblasts or myotubes. Interestingly, there is a detailed electron microscopy study (Ling et al., 1992), focusing on histogenesis of rat diaphragm that clearly demonstrates that myotubes (ED16) are closely associated with each other forming parallel sheets or palisades communicating by gap junctions. These gap junctions were detected before and during myogenesis connecting only primary myotubes and primary to secondary myotubes, which occur later at ED18 intercalating the primary ones. Around birth, separation of primary and secondary myotubes is almost complete, thus leading to the disruption and downregulation of gap-junctional coupling, although residual gap junctions on the surface of young striated muscle fibers remain expressed neonatally (Ling et al., 1992). This temporally confined occurrence of gap junctions corresponds well to the expression pattern of mCx39, at least in mouse diaphragm. From ED14.5 of mouse development, exclusively primary myotubes within the diaphragm were stained by Cx39 antibodies. During mouse development, two waves of myogenesis can be characterized. Between ED11.5 and ED15.5 the primary myotubes are formed and after ED15.5 up to birth the differentiation of secondary myoblasts to secondary myotubes occurs (Wigmore and Dunglison, 1998). This implies that Cx39 might be the molecular correlate for homologous gap-junction coupling between primary myotubes as well as for heterologous coupling between primary and secondary myotubes, persisting up to birth. After birth, mCx39 expression rapidly declined in neonatal pups and was completely absent in adults when assessed by transcript and immunofluorescence analyses. Thus, the mCx39 gene may code for a protein subunit of gap junctions occurring during myogenesis of the late mouse embryo.

More than 20 years ago, ultrastructural analyses of developing striated skeletal muscles in rat and chicken revealed the presence of gap junctions between myoblasts and between myoblasts and myotubes (Kalderon et al., 1977). These were shown to provide functional intercellular coupling in a reversible manner after blocking differentiation and fusion of these myotubes with octanol, β-glycyrrhetinic acid (Proulx et al., 1997) or heptanol (Constantin et al., 1997). A more recent study (Reinecke et al., 2000) confirmed that Cx43 is highly expressed in neonatal rat skeletal myoblasts and strongly downregulated after differentiation into myotubes. During fetal development, neither Cx43 transcript nor Cx43 protein could be detected in mouse embryonic back and intercostal muscles. Instead, Cx40 expression was first found at ED13 and later on in myotubes and myoblasts between ED14.5 and ED16.5 by in-situ hybridization. In the ED16.5 mouse embryo, however, Cx40 expression was restricted to the outermost cells within bundles of striated back muscles or intercostal muscles (Dahl et al., 1995). Our immunofluorescence data concerning Cx40 and Cx43 are partially consistent with these previous findings, as we found Cx40 in blood vessels of ED16.5 (Fig. 7D) and Cx43 in the outer boundaries of the diaphragm (Fig. 7E). Gap-junction coupling was suggested to mediate communication between primary and secondary myotubes that might allow focal activation at ED16.5 triggering a wave of activation that spreads across the entire muscle. This is probably essential for early coordination of respiratory movements, when phrenic innervation is not complete (Dennis et al., 1981). Cx39, but also Cx40, Cx43 and Cx45 might contribute to this function.

The highly specific expression of Cx39 during striated muscle development suggests that it may fulfil a unique role in this process. Targeted deletion of the Cx39 coding DNA in the mouse genome should allow the clarification of this role. It will be interesting to compare the phenotype of Cx39-deficient mice with human congenital diaphragmatic hernia (Hilfiker et al., 1998).

Although Cx39 HeLa transfectants express Cx39 protein within the plasma membrane of adjacent cells, microinjected dyes and tracers failed to diffuse into their nearest neighbours. This is not without precedent, however, as it was shown that mCx29 (Altevogt et al., 2002) or mCx33 (Chang et al., 1996) convey no functional coupling in transfected mammalian cells or in Xenopus oocytes. In the case of mCx39, distinct post-translational modifications or co-factors of this connexin, which is only expressed in myotubes, might be required for functional gap junction channels.

In order to circumvent the block of functional coupling between mCx39-transfected HeLa cells, we injected the dye Alexa488 into native myotubes of freshly prepared mouse neonatal diaphragm. Interestingly, the dye readily spread into neighbouring cells up to the second order. Since no other connexins could be detected in these cells, it is likely that mCx39 gap junction channels are functional within their normal cellular environment. This might be due to appropriate post-translational modifications in myotubes of the diaphragm such as phosphorylation or the presence of suitable co-factors or proteins that bind to the cytoplasmic loop or C-terminal region of mCx39 in order to open the pore.

Very recently, the expression pattern of cCx39 during chick development was published (Nicotra et al., 2004). Similar molecular masses and therefore the abbreviated names of mouse Cx39 and chicken Cx39 suggest that these connexins could be the corresponding orthologous genes in both species. However, these proteins differ largely with respect to their sequence similarity: after aligning cCx39 (GenBank accession number AY206667) to mCx39 (AJ414562), coding sequences show only 42% nucleotide sequence identity. Moreover, a detailed HUSAR-derived BLASTN database search within all connexin sequences selected from rodents, yielded cCx39 with the highest nucleotide sequence similarity (∼77%) to mouse or rat connexin37. Thus, cCx39 appears to be the orthologue of mCx37 and is not related to mCx39.

We thank Joana Fischer for technical assistance. This work was supported by grants of the German Research Association (Wi 270-251/2) and Fonds of the Chemical Industry to K.W.