Connexin31 cannot functionally replace connexin43 during cardiac morphogenesis in mice

In the gastrulating mouse embryo, the gap junction protein connexin43 is expressed exclusively in cells derived from the inner cell mass, whereas connexin31 is expressed in cells of the trophoblast lineage. Since connexin43 and connexin31 do not form heterotypic gap junction channels in exogenous expression systems, such as HeLa cells and Xenopus oocytes, previous studies have suggested that the incompatibility of these two connexins could contribute to the separation of connexin43-expressing and connexin31-expressing compartments between embryo and extraembryonic tissues at gastrulation, respectively. Thus, we have generated connexin43 knock-in connexin31 mice, in which the coding region of the connexin43 gene was replaced by that of connexin31. Interbreeding of heterozygous connexin43 knock-in connexin31 mice resulted in homozygous connexin43 knock-in connexin31 mice, but none of them survived to adulthood. As these mice were born at the expected Mendelian frequency, we conclude that the reported incompatibility of connexin43 and connexin31 to form heterotypic gap junction channels does not interfere with normal embryonic development. Neonatal homozygous connexin43 knock-in connexin31 hearts showed malformation in the subpulmonary outlet of the right ventricle, similar to general connexin43-deficient mice. Electrocardiograms of neonatal hearts in homozygous connexin43 knock-in connexin31 mice revealed significantly low voltage of the QRS complex. This is in contrast to previous results from our laboratory which showed that replacement of connexin43 by connexin40 resulted in morphologically and functionally normal hearts. We conclude that connexin31 cannot functionally replace connexin43 during cardiac morphogenesis.


Introduction
Connexin (Cx) proteins are subunits of gap junction channels that permit diffusional exchange of ions, secondary messenger molecules and metabolites up to 1000 Da and can serve for transmission of electrical signals between two contacting cells. Each gap junction channel consists of two hemichannels, the connexons, which are contributed by two neighboring cells. Each connexon is composed of six connexin protein subunits (Nicholson, 2003). In the mouse genome, 20 connexin genes (Cx) have been identified (Söhl and Willecke, 2003). They show cell type-and differentiation-dependent expression patterns, whereby most cell types express more than one connexin isoform. Connexins can assemble into homomeric or heteromeric hemichannels. The hemichannels can dock to hemichannels of the same or different connexin composition in the plasma membrane of apposed cells to form homotypic or heterotypic channels, respectively (Willecke et al., 2002). Gap junction channels composed of different connexin isoforms differ from each other in unitary conductance and permeability to small molecules (Harris, 2001). This raised the question 'to what extent do the different connexin isoforms limit or support functional specialization of different cell types?'.
Of the 20 known mouse connexin genes, connexin43 (Cx43) is most abundantly expressed in different cell types. In mouse heart, Connexin43 protein (Cx43) is found in the working myocardium and Purkinje fibers (van Veen et al., 2001). Cx43deficient mice died shortly after birth due to obstruction of the right ventricular outflow tract of the heart (Reaume et al., 1995). It has been described that the lethality of Cx43-deficient mice could be rescued by Cx32 or Cx40, indicating that Cx43, Cx40 and Cx32 share at least some common functions. However, Cx43 knock-in Cx32 (Cx43 32/32 ) and Cx43 knock-in Cx40 (Cx43 40/40 ) mice differed functionally and morphologically from each other and from wild-type mice (Plum et al., 2000). Cx31 is not expressed in the heart and Cx31 deficiency in mice caused transient placental dysmorphogenesis (Plum et al., 2001). In transfected HeLa cells or in RNA-injected Xenopus oocytes, Cx31 hemichannels did not form functional heterotypic channels with any other connexin, including Cx43 hemichannels, when analyzed by transfer of microinjected tracer molecules (Elfgang et al.,

In the gastrulating mouse embryo, the gap junction protein connexin43 is expressed exclusively in cells derived from the inner cell mass, whereas connexin31 is expressed in cells of the trophoblast lineage. Since connexin43 and connexin31 do not form heterotypic gap junction channels in exogenous expression systems, such as HeLa cells and
Xenopus oocytes, previous studies have suggested that the incompatibility of these two connexins could contribute to the separation of connexin43-expressing and connexin31expressing compartments between embryo and extraembryonic tissues at gastrulation, respectively. Thus, we have generated connexin43 knock-in connexin31 mice, in which the coding region of the connexin43 gene was replaced by that of connexin31. Interbreeding of heterozygous connexin43 knock-in connexin31 mice resulted in homozygous connexin43 knock-in connexin31 mice, but none of them survived to adulthood. As these mice were born at the expected Mendelian frequency, we conclude that the reported incompatibility of connexin43 and connexin31 to form heterotypic gap junction channels does not interfere with normal embryonic development. Neonatal homozygous connexin43 knock-in connexin31 hearts showed malformation in the subpulmonary outlet of the right ventricle, similar to general connexin43-deficient mice. Electrocardiograms of neonatal hearts in homozygous connexin43 knock-in connexin31 mice revealed significantly low voltage of the QRS complex. This is in contrast to previous results from our laboratory which showed that replacement of connexin43 by connexin40 resulted in morphologically and functionally normal hearts. We conclude that connexin31 cannot functionally replace connexin43 during cardiac morphogenesis. 1995) or by electrical conductance (Yeager and Nicholson, 2000). We refer to this as the incompatibility of Cx43 and Cx31 hemichannels. During embryonic development, both Cx43 and Cx31 are present in the compacted morula and throughout the blastocyst. After implantation, Cx43 protein is detected exclusively in cells derived from the inner cell mass, whereas Cx31 protein is localized in cells of the trophectoderm lineage (Dahl et al., 1996;Reuss et al., 1996). It has been hypothesized that the incompatibility of these two connexins might contribute to the separation of Cx43-expressing and Cx31expressing compartments between embryo and extraembryonic tissues at gastrulation (Dahl et al., 1996). Using Cx43 knockin Cx31 mice (Cx43 31/31 ), in which the coding region of the Cx43 gene was replaced by that of Cx31, we have now investigated whether the incompatibility between these two isoforms is necessary for embryonic development and whether the loss of Cx43 can be compensated for by Cx31.

Generation of floxCx43KI31neo mice
To circumvent possible embryonic lethality, we first generated a mouse line that carries a 'floxed' Cx43 coding region, i.e. flanked by loxP recognition sites for the Cre recombinase (Nagy, 2000). The Cx31 coding region was inserted behind the second loxP site (Fig. 1A). After crossing the floxed mice with PGK-Cre mice that expressed the Cre recombinase under control of the phosphoglycerate kinase (PGK) promoter, the Cx43 coding region was removed by Cre-mediated deletion and Cx31 was expressed under control of the Cx43 promoter. This approach allowed us not only to study the general replacement of Cx43 by Cx31, but will also be useful in the future for a cell-type-specific replacement by crossing with other mouse lines that express Cre under control of cell-typespecific promoters.
To check whether expression of Cx43 and Cx31 in the Lane 1: 100 bp marker; lane 2: a 2.1 kb amplicon indicates that homologous recombination has occurred in the floxCx43KI31neo allele using primer 1 (Cx31Ki2) and 2 (Cx43-3Ј-RO3); lane 3: using primer 3 (Cx43/31fw) and primer 4 (Cx43/31rev), the wild-type allele yielded a 326 bp DNA fragment, whereas floxCx43KI31neo allele gave rise to a 375 bp fragment due to an additional loxP site; lane 4: after Cre activity, a 736 bp amplicon was generated using primer 3 (Cx43/31fw) and primer 5 (Cx31Ki3rev); lane 5-7: a multiplex PCR was established using primer 6 (Cx43-3Ј-HO2), primer 7 (Cx31Ki3) and primer 8 (Cx43-3Ј-RO4). A 381 bp fragment indicates the wild-type allele and a 615 bp fragment the knock-in allele. Lane 6: the Cx43 +/31 genotype. Cx43 knock-in Cx31 mice floxCx43KI31neo mice (here abbreviated as Cx43 flox/flox ) was influenced by the loxP sites, the PGK-neomycin cassette or the silent Cx31 gene, Cx43 +/flox mice were crossed with each other. Cx43 flox/flox animals were born at the expected Mendelian frequency. We performed western blot analysis of heart and brain lysates using anti-Cx43 and anti-Cx31 antibodies. Compared to wild-type animals, the protein level of Cx43 in the heart was slightly increased in Cx43 +/flox or Cx43 flox/flox mice (data not shown). By contrast, no significant change of Cx43 expression was found in the brain. As in wild-type mice, Cx31 was not expressed in brain or heart of Cx43 +/flox and Cx43 flox/flox mice (data not shown). Therefore, we concluded that floxCx43KI31neo mice resemble wild-type mice with regard to Cx43 or Cx31 expression.
A multiplex PCR (Fig. 1A) was established for genotyping using one common reverse primer from the 3Ј untranslated region of Cx43 (primer 8), one forward primer from the coding region of Cx43 (primer 6) and the other from the coding region of Cx31 (primer 7). A 381 bp amplicon was derived from the wild-type allele and a 615 bp amplicon was delivered from the knock-in allele (Fig. 1E). In the Southern blot hybridization, the Cx43 external probe recognized a 8 kb wild-type fragment and a 2.9 kb knock-in fragment (Fig. 1B). The Cx43 internal probe detected the 8 kb wild-type fragment only in wild-type but not in homozygous mutant DNA (Fig. 1C). The Cx31 internal probe labeled the 12 kb wild-type fragment in all three genotypes and the additional 5.3 kb knock-in fragment in heterozygous and homozygous mutants (Fig. 1D). Western blot analyses of postnatal day 7 (P7) heart and brain lysates demonstrated that Cx43 protein was absent in the Cx43 31/31 mice and Cx31 protein was translated in Cx43 +/31 and Cx43 31/31 mice in a gene dosagedependent manner (Fig. 2). Immunofluorescence analysis of P0 ventricular myocardium confirmed the results of the western blot analysis: no gap junctional plaques due to Cx43 protein were detected in the Cx43 31/31 mice, whereas signals of the Cx31 protein were found in Cx43 +/31 and Cx43 31/31 mutants ( Fig. 3). In addition, we performed double immunofluorescence analysis of Cx43 and Cx31 on the Cx43 +/31 heart tissue ( Fig. 4). The working cardiomyocytes of the adult ventricle were extensively interconnected at the intercalated disk ( Fig. 4D,E). Cx43-and Cx31-containing plaques were  Cryosections of Cx43 +/+ (A,D), Cx43 +/31 (B,E) and Cx43 31/31 (C,F) hearts at P0 were immunolabeled with anti-Cx43 (A-C, red), anti-Cx31 (D-F, green) and counterstained with propidium iodide (shown as blue nuclear staining). Cx43 was expressed in Cx43 +/+ and Cx43 +/31 ventricles, but not in Cx43 31/31 ventricles. Cx31 was not detected in Cx43 +/+ hearts, but in Cx43 +/31 and Cx43 31/31 ventricles. Bars, 20 m. predominantly co-localized (Fig. 4F). By contrast, P0 ventricular cardiomyocytes did not show intercalated disks (Fig.  4A,B) and the co-localization of Cx43 and Cx31 was low (Fig.  4C). Immunofluorescence analysis of Cx43 31/31 heart showed regular expression of Cx40 in the atria and vessels. Upregulation of Cx40 was not found in Cx43 31/31 heart (data not shown). Western blot analysis showed that the total amount of Cx40 was not changed in the P0 heart lysate of Cx43 31/31 mice (data not shown). By immunofluorescence, Cx45 was detected in the conduction system, but not in the working cardiomyocytes of either Cx43 +/+ or Cx43 31/31 mice (data not shown). In the western blot analyses, Cx45 was not found to be upregulated in the Cx43 31/31 heart lysate (data not shown).
Cx43 31/31 mice die postnatally because of a dysmorphogenesis of the right ventricular outflow tract To obtain Cx43 31/31 mice, Cx43 +/31 mice were intercrossed. At birth, we found the expected ratio of homozygous, heterozygous and wild-type mice (Fig. 5). This indicated that the embryos developed to term, although Cx31 was presumably expressed both in the extraembryonic tissue and the embryo proper (instead of Cx43). However, no Cx43 31/31 mice survived to adulthood. 50% of the Cx43 31/31 animals survived at P0, 30% at P1, none of them lived longer than 10 days (Fig. 5). At P0, the average weight of the wild-type mice was 1.4±0.15 g and the weight of Cx43 31/31 mice was 1.2±0.14 g. Several days later, the living Cx43 31/31 animals could be identified by their lower weight. They did not grow as fast as their Cx43 +/+ and Cx43 +/31 littermates. There was no obvious difference in the phenotype between Cx43 +/+ and Cx43 +/31 mice.
Since Cx43-deficient mice died shortly after birth due to a dysmorphogenesis of the right ventricular outflow tract (Reaume et al., 1995), we investigated neonatal Cx43 31/31 hearts histologically (Fig. 6). We examined six Cx43 +/+ (three P0, two P4 and one P5), six Cx43 +/31 (two P0, two P4 and two P5) and ten Cx43 31/31 (eight P0, one P4 and one P5) hearts. The pulmonary outlet of all ten Cx43 31/31 heart was prominent with irregular hypertrophic trabeculation forming abnormal pouchlike spaces, surrounding the narrow but still open substantial lumen. The morphology of Cx43 +/31 hearts was basically normal. We conclude that the morphological defect of Cx43 31/31 hearts was similar to Cx43-deficient hearts and Cx31 could not replace the functions of Cx43 during cardiac morphogenesis.
No differences in development and placental architecture were found in Cx43 31/31 conceptuses Cx43 is expressed in spongiotrophoblast and giant cells in the mature placenta. In order to test whether or not replacement of Cx43 by Cx31 leads to impairment of trophoblast cell lineage differentiation, we examined placentas for phenotypic changes. Placental morphology was investigated at embryonic day (E) 10.5 and E17.5. No obvious differences in placental architecture, correlated to the developmental stage, were observed between Cx43 +/+ , Cx43 +/31 and Cx43 31/31 mice from the same litter (Fig. 8). In mature placentas, all different trophoblast cell types analyzed were normally developed, for example the labyrinthine trophoblast, the spongiotrophoblast, interspersed by nests of glycogen cells, and trophoblast giant cells located at the feto-maternal border.

Discussion
We prepared the floxed Cx43KI31neo construct, since we anticipated that Cx43 31/31 mice might die during development at the gastrulation stage. At this time, Cx43 is expressed in the embryonic compartment, whereas Cx31 was found exclusively in the surrounding extraembryonic compartment (Dahl et al., 1996). Since Cx31 and Cx43 did not form heterotypic gap junction channels in transfected HeLa cells (Elfgang et al., 1995) and in RNA-injected Xenopus oocytes (Yeager and Nicholson, 2000), we hypothesized that this incompatibility might be important for the further development of the gastrulating mouse embryo. Our results presented here clearly show that the incompatibility of Cx43 and Cx31 hemichannels cannot be essential for mouse embryonic development, because Cx43 +/31 and Cx43 31/31 mice were born at the expected Mendelian frequencies after interbreeding of Cx43 +/31 mice Fig. 6. Transverse sections through P4 hearts of Cx43 +/+ (A), Cx43 +/31 (B) and Cx43 31/31 mice (C). The heart of the Cx43 31/31 mouse showed irregular hypertrophic trabeculae with abnormal pouches (asterisks) in the subpulmonary outlet of the right ventricle. The heart of the Cx43 +/31 mouse was almost normal. aa: ascending aorta; la, left atrium; ra, right atrium; rv, right ventricle; pv, pulmonary valve. Bar, 500 m. Fig. 7. Representative signal-averaged ECG recordings performed at P0. Cx43 +/+ (A) and Cx43 +/31 mice (B) show normal ECG patterns. Note significant low voltage and prolonged P wave as well as QRS duration in Cx43 31/31 mice (C). (Fig. 5). There might be several reasons why the separation between embryonic and placental lineage might not have any effect in our experimental system. For example, the separation between the embryo and the surrounding extraembryonic tissues at gastrulation might have other explanations, such as discrete tissue masses that do not establish gap-junctional communication across an extracellular matrix barrier or, alternatively, the abolition of the separation does not impair the development of the two cell lineages.
Cx43 31/31 newborns showed a significantly reduced weight compared to their Cx43 +/+ and Cx43 +/31 littermates. The weight of embryos or newborns has not been analyzed so far for Cx43 -/mice (Reaume et al., 1995), but recently Kibschull et al. (Kibschull et al., 2005) showed for the Cx31/Cx43 double knockout mouse that at 17.5 dpc the Cx31 +/-/Cx43 -/embryos were also significantly smaller compared to their Cx31 +/-/Cx43 +/and Cx31 +/-/Cx43 +/+ littermates. Neither the weight nor the morphology of the Cx31/Cx43 double knockout placentas appeared to be affected by Cx43 expression. In accordance to this data, the Cx43 31/31 placentas also did not show any morphological changes compared to wild-type littermates. Therefore, we assume that the weight of the Cx43 31/31 embryos is unlikely to have been influenced by placental development or function. The reduced weight of the Cx43 31/31 newborns is probably the result of the loss of Cx43 in the embryo itself, as has been discussed for the Cx31/Cx43 double knockout mouse (Kibschull et al., 2005). Since the replacement of Cx43 by Cx31 did not lead to any obvious impairment in embryogenesis and placental development, we focused our characterization of Cx43 31/31 mice on the cardiac phenotype.
Journal of Cell Science 119 (4) Durations and intervals are given as group means ± s.e.m.; *P<0.05; **P<0.01; ***P<0.001; n.d., evaluation not possible because voltage was too low.  Our results show that Cx31 cannot replace the essential role of Cx43 during cardiac morphogenesis. Mice lacking Cx43 die shortly after birth from an obstruction of the right ventricular outflow tract (Reaume et al., 1995). In Cx43 31/31 mice, the right ventricular outflow tract is still open, but its lumen is so narrow that the supply of blood is likely to be limited. As a consequence, all of them died within ten days (Fig. 5). By contrast, Cx43 40/40 and Cx43 32/32 mice can survive to adults, although 40% of them died during the first 3 weeks after birth and most of them were infertile (Plum et al., 2000). Thus, Cx40 and Cx32 can replace Cx43 during cardiac development, but Cx31 cannot. Our results fit into the general scheme, that the different connexin isoforms can replace each other in certain tissues but fulfill unique essential functions in other tissues (Plum et al., 2000;White et al., 2002;Alcolea et al., 2004).
The reasons, at the molecular level, for the malformation of hearts during development of general Cx43-deficient mice are still a matter of debate. The hearts of mice with cardiacrestricted inactivation of Cx43 by use of the Cre/loxP system are structurally normal (Gutstein et al., 2001). Studies of transgenic mice exhibiting gain or loss of Cx43 suggest that the migration rate of cardiac neural crest cells is increased with overexpression of Cx43 and decreased with deficiency of Cx43 (Lo et al., 1999). Additional studies have shown involvement of another extracardiac cell population, the proepicardial cells in these cardiac anomalies (Li et al., 2002;Wei et al., 2004). Cx43 31/31 mice show similar cardiac malformations as the general Cx43-deficient mice. The hearts of Cx43 32/32 mice also suggest a similar morphologic defect, but to a much lesser extent. This underlines that the special properties of Cx43 lead to normal morphogenesis of the heart.
Functionally, replacement of Cx43 by Cx31 in the heart leads to low voltage of the QRS complex, and prolonged P and QRS duration in the electrocardiogram. Recently, Danik et al. (Danik et al., 2004) generated a heart-specific Cx43-deficient subline (O-CKO mice) in which cardiac Cx43 expression decreased progressively throughout the ventricular myocardium. From 25 to 45 days, the number of immunofluorescent clusters in the O-CKO hearts decreased from 67% of control to 34% and the percentage area from 59% to 30%. The QRS amplitude of O-CKO mice showed an agedependent decrease, which closely paralleled the loss of Cx43 expression. In order to determine the mechanism of diminished QRS amplitude, the authors later used voltage-sensitive dyes and high resolution optical mapping techniques and found that reduced intercellular coupling leads to paradoxical propagation across Purkinje-ventricular junctions and aberrant myocardial activation (Morley et al., 2005). The diminution of the QRS amplitude seems to be related to the pattern of ventricular excitation. Compared with O-CKO mice, in Cx43 +/31 and Cx43 31/31 hearts Cx43 is replaced by Cx31. Double immunofluorescence using anti-Cx43 and anti-Cx31 showed in the ventricle of Cx43 +/31 mice at P0 that Cx43-and Cx31containing plaques were not co-localized, whereas in the adult Cx +/31 ventricle, Cx43 and Cx31 were predominantly colocalized in intercalated disks (Fig. 4). Since the ECG of Cx43 +/31 mice was normal, we assume that the presence of Cx31 plaques does not lead to an increase in impulse propagation across the Purkinje-ventricular junction. The reduction of the QRS amplitude in the Cx43 31/31 heart is possibly due to deletion of Cx43 and the inability of Cx31 to function like Cx43 in the myocardium. Although Cx43 is the predominant connexin expressed by ventricular and atrial cardiomyocytes in wild-type mice, Cx45 is also expressed at very low levels (Severs et al., 2004). The total amount of Cx45 in the Cx43 31/31 heart was not upregulated, as determined by western blot analysis (data not shown).
The relationship between QRS duration and uncoupling due to Cx43-deletion is controversial in the literature. A modest increase in QRS duration was detected only at 25 days in the O-CKO mice. At 35 and 45 days of age, however, there was no significant difference in the QRS duration between the CKO and control mice (Danik et al., 2004). Our data are consistent with the results of Eckardt et al. (Eckardt et al., 2004) who showed in inducible Cx43 Cre-Er(T)/fl mice that ventricular activation was significantly delayed after deletion of Cx43, as indicated by significant prolongation of the QRS complex. Prolongation of the P wave was found in Cx40-deficient mice (Kirschhoff et al., 1998;Simon et al., 1998). Gap junction channels among atrial myocytes contain abundant Cx40, colocalized with Cx43 (Vozzi et al., 1999;Severs et al., 2001). An increase in P wave has not been shown in conditional Cx43deficient mice (Danik et al., 2004;Eckardt et al., 2004). The Cx43 31/31 mice exhibited significant prolongation of the P wave. This might be due to the different level of atrial Cx43 in different mouse lines. Atrial expression of Cx43 in the O-CKO hearts at 45 days of age was not changed in immunoblot analyses (Danik et al., 2004). By contrast, no Cx43 protein was detected in the Cx43 31/31 heart, either by western blot (Fig. 2) or by immunofluorescence (Fig. 3) analyses. Direct compensation of Cx43 deletion by upregulation of Cx40 was not found in Cx43 31/31 heart, similar to O-CKO mice (Danik et al., 2004) and inducible Cx43 Cre-Er(T)/fl mice (Eckhardt et al., 2004).
Quantitative echocardiography of neonatal mice on P0 could not be performed in this study because of the small size of the hearts. On P5, echocardiography revealed an impaired systolic function in surviving Cx43 31/31 . In these animals, mean Doppler-derived systolic velocities in aorta and pulmonary trunk showed no significant differences between the genotypes. Assessment of Doppler-derived flow velocities within the stenotic outflow tract was not possible at this age. Slightly lower velocities measured in the putative post-stenotic pulmonary trunk of Cx43 31/31 mice could be the consequence of right-ventricular outflow tract stenosis.
In recent years, several heart-specific conditional Cx43deficient mice have been generated (Gutstein et al., 2001;Eckardt et al., 2004;van Rijen et al., 2004;Danik et al., 2004;Gros et al., 2004). In summary, 80-95% decrease of Cx43 expression resulted in delayed ventricular activation, arrhythmia and sudden death. Cx43 31/31 mice died before P10. At this young age, telemetric ECGs could not be measured. Thus, we do not know the ECG shortly before death. Since we can induce cardiac-specific replacement of Cx43 with Cx31 in adult mice by breeding floxCx43KI31neo mice with Cx43 Cre-ER(T)/+ mice, we can study the consequence of this replacement also in adult hearts in the future. Furthermore, the floxCx43KI31 mice that were generated in this work could be bred with mice that express the Cre recombinase cell type specifically, in order to study the functional consequence of replacement of Cx43 by Cx31 in other cell types. We assume that the functional diversity of different connexin isoforms in different cell types is probably due to specific regulatory features of these proteins, such as permeability characteristics, posttranslational modifications, specific binding proteins or a combination of these differences.

Generation of mice
A fragment of genomic 129/Sv mouse DNA spanning about 12 kb of the Cx43 locus was isolated as previously described (Plum et al., 2000) from a recombinant lamda phage library (Stratagene, La Jolla, CA). The Cx43KI31 construct (Fig. 1A) was embedded in a pBluescriptII SK + vector backbone (Stratagene). A 5.3 kb SacI-HincII fragment and a 1.6 kb ClaI-AvaI fragment were used as 5Ј and 3Ј homologous regions, respectively. The first loxP site was inserted into the HincII site, 188 bp upstream of the Cx43 start codon. A 160 bp DNA fragment harboring a SV40 poly(A) signal was inserted into the ClaI site 26 bp downstream of the Cx43 stop codon, followed by a 1.95 kb phosphoglycerate kinase (PGK) promoter-neomycin selection marker flanked by two frt sites and the second loxP site (Meyers et al., 1998). Between the second loxP site and the 3Ј homologous region of Cx43, several DNA fragments were inserted: a 188 bp fragment upstream of the Cx43 start codon including the endogenous splice acceptor site, the 813 bp coding region of Cx31 and the following 80 bp downstream of the Cx31 stop codon. In order to generate a NcoI site, a cytosine residue was inserted two nucleotides upstream of the Cx43 start codon. The final targeting vector pCx43KI31 was analyzed by restriction mapping and partial sequencing. The functions of loxP sites and frt sites were verified by transformation of Cre or Flp recombinase-expressing E. coli bacteria (Buchholz et al., 1996).
HM-1 embryonic stem cells were transfected with 300 g DNA of the targeting vector pCx43KI31, linearized upstream of the 5Ј homologous region by NotI and selected by 350 g G418 (Sigma, St Louis, MO) per ml of medium as described previously (Magin et al., 1992;Degen et al., 2004). Resistant clones were analyzed for homologous recombination by PCR and Southern blot hybridization. Correctly recombined ES clones were injected into C57BL/6 blastocysts to generate chimeras, as described by Nagy et al. (Nagy et al., 2003). For germline transmission of the mutated allele, the chimeras were crossed to C57BL/6 mice. Male heterozygous offspring, designated as floxCx43KI31neo, were further crossed to female PGK-Cre mice (Lallemand et al., 1998) to generate the final allele Cx43KI31. All analyses were carried out on a mixed 129/Ola/C57BL/6 genetic background using littermates as controls.
Mice were kept under standard housing conditions with a fixed 12 hours light:12 hours dark cycle, and food as well as water ad libitum. All experiments were carried out in accordance with the German law for animal welfare and with permission from the local state authorities.

Genotyping of ES cells and mouse tissues
G418 resistant ES clones were prepared and processed for two PCR analyses (Fig.  1A,E). In the first PCR, recombination at the 3Ј homologous site was examined using a Cx31-coding-region-specific upstream primer (primer 1: 5Ј-GTG AGA TCT GCT ACC TCA TCT TCC-3Ј) and a 3Ј downstream primer external to the targeting vector (primer 2: 5Ј-GCA GCA ACA TAC CAT TGC ACA GAG-3Ј). A 2.1 kb amplicon was indicative of homologous recombination. In the second PCR, the existence of the 5Ј located loxP site was verified using the primers: primer 3 (5Ј-GCA CTT GGT AGG TAG AGC CTG TCA GGT C-3Ј) and primer 4 (5Ј´-CTC CAG TCA CCC ATG GTC TGG GCA CCT C-3Ј). The wild-type allele yielded a 326 bp DNA fragment, whereas the mutated allele gave a 375 bp fragment because of the first loxP site.
For genomic Southern blot hybridization, DNA from PCR-positive clones was digested by XbaI. A 550 bp AvaI/AvaI fragment outside the 3Ј homologous region of Cx43 served as external probe. The Cx43 internal probe was a 762 bp HincII/ClaI fragment of the coding region. In addition, a Cx31 internal probe (544 bp, PstI/PstI) was used. After XbaI digestion and hybridization to the Cx43 external probe, the Cx43 wild-type allele yielded a 10 kb fragment compared to a 6.4 kb fragment of mutated allele. The Cx43 internal probe detected a 10 kb fragment of the wild-type allele and a 4 kb fragment of the mutated allele. The Cx31 internal probe hybridized to a 5.3 kb fragment of wild-type and a 6.4 kb fragment of the recombined allele.
After action of the Cre recombinase, the deletion of the Cx43 coding region and PGK-neo cassette was analyzed by using primer 3 (5Ј-GCA CTT GGT AGG TAG AGC CTG TCA GGT C-3Ј) and primer 5 (5Ј-CTA AAC AAG TAG GTC CAC CAC AGG-3Ј). The PCR yielded a 736 bp amplicon covering the junction between the intron of Cx43 and the coding region of Cx31. For detection of the PGK-Cre transgene, the primers up-pgk1 (5Ј-GCT GTT CTC CTC TTC CTC ATC TCC-3Ј) and IntCre_rev (5Ј-TCC ATG AGT GAA CGA ACC TGG TCG-3Ј) were used, generating a 500 bp amplicon.
For routine genotype analysis, genomic DNAs from tail tips were used for PCR, including three primers: primer 6 (coding region of Cx43; 5Ј-CGC AAT TAC AAC AAG CAA GCC AGC-3Ј), primer 8 (3Ј untranslated region of Cx43; 5Ј-CGC CTC ATT ACT GAG GTT GTT GAG-3Ј) and primer 7 (coding region of Cx31; 5Ј-CCT GTG GTG GAC CTA CTT GTT TAG-3Ј) that generated a 381 bp amplicon of wildtype and a 615 bp amplicon of the knock-in allele.

Southern blot analysis
For Southern blot hybridization of wild-type (Cx43 +/+ ), heterozygous (Cx43 +/31 ) and homozygous (Cx43 31/31 ) Cx43 knock-in Cx31 mice, genomic DNA was prepared from neonatal kidneys and digested with PstI for the Cx43 external and internal probe, and HindIII for the Cx31 internal probe. After digestion, genomic DNA was fractionated on 0.7% agarose gels and transferred onto nylon membrane (Hybond TM -N+; Amersham Biosciences, UK). The Cx43 external probe (487 bp outside the 3Ј homologous region), Cx43 internal probe (654 bp of Cx43 coding region) and Cx31 internal probe (508 bp of Cx31 coding region) were synthesized using a PCR DIG Synthesis Kit (Roche diagnostics GmbH, Penzberg, Germany). Hybridization, washing and detection were performed using DIG Prime DNA Labeling and Detection Starter Kit II (Roche Diagnostics GmbH, Penzberg, Germany) according to instructions provided by the manufacturer.

Western blot analysis
Mouse hearts and brains were dissected, immediately frozen on dry ice and stored at -70°C until further use. Homogenized tissue was taken up in protein lysis buffer (60 mM Tris HCl pH 7.4 and 3% sodium dodecyl sulphate), supplemented with protease inhibitor 'Complete' (Roche, Mannheim, Germany) and sonicated three times for 10 seconds on ice. Protein concentration was determined using the bicinchoninic acid protein assay (Sigma, Taufkirchen, Germany). 50 g protein were separated on 12.5% SDS-polyacrylamide gels and electroblotted on nitrocellulose membranes (Hybond ECL Biosciences, Bucks, UK) for 45 minutes at 100 V. Membranes were blocked for 1 hour with WBWP buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.1% Tween 20) and 5% milk powder (w/v) for anti-Cx43 and with Roti-Block (Roth, Karlsruhe, Germany) for anti-Cx31. Afterwards the membranes were incubated with rabbit anti-Cx43 (1:4000) (Traub et al., 1994) and rabbit anti-Cx31 (1:1000; Zymed, San Francisco, CA) overnight at 4°C. Anti-Cx43 was diluted in WBWP buffer and anti-Cx31 was diluted in Roti-Block. After incubation, the membranes were washed for 30 minutes in WBWP buffer (anti-Cx43) or in phosphate-buffered saline (PBS) plus 0.1% Tween 20 (anti-Cx31), followed by incubation with anti-rabbit horseradish peroxidase-conjugated secondary antibodies (1: 30000; Dianova, Hamburg, Germany) for 1 hour at room temperature. Membranes were washed as described before and incubated with an ECL chemiluminescence detection system (Amersham Biosciences, Freiburg, Germany). In order to prove that equal amounts of protein were loaded in all lanes, the immunoblots were stripped using Restore TM Western Blot Stripping (Pierce, Rockford, IL) and incubated with monoclonal anti-␣-tubulin (1:10.000; Sigma, Taufkirchen, Germany) for 1 hour at room temperature. After washing, the membranes were incubated with anti-mouse horseradish peroxidase-conjugated secondary antibodies (1: 30.000; Dianova, Hamburg, Germany), followed by ECL detection. Monoclonal anti-Cx45 (Chemicon, Hampshire, UK) and polyclonal anti-Cx40 (Alpha Diagnostic, San Antonio, TX) were used as described for anti-Cx43.

Histological analysis
For morphological analyses, hearts from postnatal days (P)0, P4 and P5 mice and placentas from E10.5 and E17.5 conceptuses were fixed in 4% paraformaldehyde in PBS for at least 2 days at 4°C, dehydrated and embedded in paraffin. Sections were stained with Hematoxylin and Eosin (HE) and mounted in Entellan (Merck, Darmstadt, Germany).

ECG recording of neonatal mice
ECG recordings were obtained on P0 from 81 neonates of both genders weighing between 1 and 1.4 g (1.3±0.2 g). Mice were fixed on a heating pad warmed to maintain body temperature. A surface 6-lead ECG was acquired digitally by means of a multichannel amplifier (PowerLab TM System, ADInstruments, Milford, MA). ECG channels were amplified, filtered between 10 and 100 Hz and sampled at a rate of 1 kHz per channel. Heart rate, P-wave duration, PR interval, QRS duration, QRS amplitude and QT interval were determined off-line as described recently