Analysis of distribution patterns of gap junctions during development of embryonic chick facial primordia and brain

Mammalian embryonic facial development includes a sequence of events that results in the formation of outgrowths that enlarge and unite with each other to form the structures of the midface including the upper lip, nose and maxilla. The primordia that form these structures are the medial and lateral nasal processes, which arise as outgrowths on either side of the invaginating nasal placode, and the maxillary process, which emerges from the cranial end of the first pharyngeal arch. The fourth major primordium, the mandibular process, also arises from the first pharyngeal arch and forms the mandible, lower lip and associated structures (Streeter, 1948; Patten, 1968; Hamilton and Mossman, 1972; for avian development see Romanoff, 1960).

One of our major interests for the past several years has been the analysis of tissue interaction and spatial organization that result in the formation of these primordia, employing the chick embryo as a model system (Saber et al. 1989; Xu et al. 1990; Minkoff, 1990). Prior studies focused on the analysis of morphogenetic movement and cell cycle kinetics of facial mesenchyme (Minkoff and Kuntz, 1977, 1978; Minkoff, 19806, 1984; Minkoff and Martin, 1984; Patterson et al. 1984; Patterson and Minkoff, 1985; Bailey et al. 1988). More recent studies of epithelial-mesenchymal interaction in the facial primordia (Saber et al. 1989; Minkoff, 1990) indicated the presence of stage-dependent gradients of cell viability of mesenchyme in organ culture. Developmental compartments, most noticeable at early stages of development, appeared within facial mesenchyme as a result of epithelial-mesenchymal interaction. The means by which these gradients and compartments could have been established led to a consideration of cell communication and signaling mechanisms.

Among the mechanisms by which cells communicate with each other during development, a region of junctional specialization in the plasma membrane -the gap junction -allows movement of water-soluble molecules having a molecular weight of less .than approximately 1000 to pass freely from the cytoplasm of one cell to another. A population of cells coupled by gap junctions becomes, therefore, a syncytium with regard to passage of ions and small molecules. This integrative effect can result in coordinate metabolic activity and could influence the activities of cell populations during inductive tissue interactions, morphogenetic change and pattern specification. Since gap junctions mediate ionic and metabolic coupling, the distribution of these membrane specializations would indicate the potential for metabolic coupling within specific anatomical regions or among particular cell populations.

In earlier studies, we had mapped the distribution of gap junctions in the facial primordia employing transmission electron microscopy (TEM). These analyses contained inherent limitations since large numbers of sections were required and only small areas could be studied at any given time. Therefore, indirect immunofluorescence employing antibodies to gap junction proteins (Hertzberg, 1984; Hertzberg and Skibbens, 1984; Hertzberg et al. 1985; Yamamoto et al. 1990) was utilized for localization. This provided a significant advantage since the distribution of gap junctions could be seen over large areas and allowed observations to be made that would not have been possible employing TEM or other related techniques.

Our studies utilized antibodies to connexin 32, a gap junction protein in rodent fiver with an apparent mobility of approximately 27 ×10³M_r on SDS-PAGE (Henderson et al. 1979; Hertzberg and Gilula, 1979; Hertzberg, 1984; Paul, 1986). Immunocytological localization indicated that gap junctions were ubiquitously distributed throughout the epithelium and mesenchyme of all of the facial primordia examined. The distribution appeared uniform except in the developing nasal placode. In this structure, two sharply delineated regions were found. The superficial region contained a concentration of signal not found anywhere in the facial primordia or developing primary palate. Beneath this layer, the remaining deeper region of the placode was completely devoid of signal. This unusual pattern of distribution of gap junctions has been observed, thus far, in only one other tissue -the developing neuroepithelium.

Antibodies to connexin 43 (Beyer et al. VTST, Beyer et al. 1988), a gap junction protein initially isolated from heart tissue (Manjunath and Page, 1985), were also utilized as part of our ongoing efforts to map the distribution of gap junctions in developing chick facial primordia and nervous tissue. As in the case of connexin 32, a non-uniform distribution of immunoreactivity to connexin 43 was found, with an indication of only partial regional overlap of expression of these gap junction proteins.

Fertile White Leghorn eggs were incubated at 37°C in a humidified atmosphere to yield normal embryos ranging between stages 22 and 28 (Hamburger and Hamilton, 1951; HH). This corresponds to 3 to 6 days of incubation and encompasses the formation of the primary palate in the chick embryo. Embryo heads were embedded in Tissue-Tek OCT compound (Miles Scientific, Naperville, IL) and rapidly frozen by immersion in liquid nitrogen. 10/mt cryostat sections were mounted on gelatinized slides. The slides were stored at −20°C until stained (refer to Fig. 1 for orientation and plane of sectioning).

Tissue sections were incubated for 20 min with 10% rabbit serum in phosphate-buffered saline (R-PBS) to block nonspecific binding of antibodies. After overnight incubation at 4°C with a 1:400 dilution of 2.0/tgml^-1 affinity-purified sheep antibodies to the 27 ×10³M_r rat liver gap junction protein in R-PBS, the slides were washed in PBS and incubated for one hour at room temperature with a 1:60 dilution in R-PBS of pre-adsorbed fluorescein-conjugated rabbit anti-sheep IgG (ICN, Lisle, IL). The fluoresceinated IgG was pre-adsorbed for Ih at room temperature with a homogenate of chick embryo tissue, centrifuged and the supernatant fraction preadsorbed a second time. Serum was substituted for gap junction antibody in negative control slides. The slides were washed and mounted with glycerol containing 10 % PBS and 1 % p-phenylenediamine (PPD) to reduce photo-bleaching (Johnson and de C. Nogueira Araujo, 1981). In preliminary experiments, a high background due to the fluorescein-conjugated anti-sheep IgG secondary antibody was found with a sparse, but reproducible, signal discernible. The absorption protocol described above was, therefore, developed to lower this background. However, even with this approach and with the use of PPD, a labile signal was observed which was subject to relatively rapid photo-bleaching.

Rabbit antibodies to a peptide corresponding to amino acids 346–360 of the 43×læ M_r ‘heart’ sequence (Beyer et al. 1987) were employed in a similar fashion (Yamamoto et al. 1990). Tissue sections were blocked with 3% bovine albumin in PBS (BSA/PBS). After overnight incubation with a 1:1000 dilution of the primary antibody in BSA/PBS, slides were washed in PBS and incubated for one hour at room temperature with a 1:60 dilution of fluorescein-conjugated goat anti-rabbit IgG (Sigma). Slides were washed and mounted as described above.

Specimens were then examined with a Nikon Optiphot microscope equipped for epifluorescence. Photographs were prepared using Tri-X Pan ASA 400 film. The number of sections examined by developmental stage is displayed in Table 1.

Affinity-purified sheep antibodies to connexin 32 (Hertzberg, 1984; Saez et al. 1990) and a rabbit antibody to connexin 43 (Yamamoto et al. 1990) were used to identify immunoreactive proteins on Western blots. Tissue samples for Western blots were homogenized in 1 mM NaHCO₃ to which phenylmethylsulfonyl fluoride was added to 2 mM immediately before use from a 0.1M stock solution in isopropanol. 50 μg protein samples were resolved by electrophoresis on 10% SDS-polyacrylamide gels, electrophoretically transferred to nitrocellulose and placed in a block solution overnight prior to incubation with primary antibodies (Hertzberg and Skibbens, 1984; Hertzberg et al. 1988). Subsequent to incubation with primary antibody, blots probed with sheep antibody were incubated with rabbit anti-sheep IgG followed by ¹²⁵I-Protein A. Blots probed with rabbit antisera were incubated with ¹²⁵I-Protein A directly, and all autoradiographs exposed at −80°C using an intensifying screen. For samples in which preliminary experiments indicated an absence or low level of immunoreactivity, membrane protein-containing fractions enriched for gap junction proteins were prepared by extracting the homogenate in 20 HIM NaOH and analyzing the material in the pellet obtained by centrifugation at 48000g for 20 min (Hertzberg and Skibbens, 1984).

The localization of connexin 32-containing gap junctions was accomplished employing indirect immunofluorescence localization with antibodies raised against the 27×10³M_r rat liver gap junction protein. Based upon double diffusion analysis in which precipitin arcs were observed, the antibodies were directed against multiple determinants on the junction polypeptide. Specificity for the 27×10³M_r gap junction polypeptide has been demonstrated by immunoreplica analysis of liver plasma membranes and isolated gap junctions. By indirect immunofluorescence staining, the antibodies appear to bind to punctate regions of the hepatocyte plasma membrane in a manner consistent with the distribution of gap junctions determined by electron microscopy (Hertzberg, 1984, 1985). Cross-reacting 27×10³M_r polypeptides were detected in liver from mammalian, fish and avian species, including chick, by immunoreplica analyses and were localized to punctate regions of the plasma membrane by indirect immunofluorescence (Hertzberg and Skibbens, 1984; Der-mietzel et al. 1984). Polyclonal antibodies to peptides in the 43×10³M_r ‘heart’ gap junction protein (peptide 346-360 of the heart sequence conjugated to BSA) were also employed for Western blot analysis and immunolocalization (Yamamoto et al. 1990).

Nasal placodes were dissected from stage 25 chick embryos and immediately placed in 3 % buffered glutaraldehyde at 4°C for 24h.

The tissues were rinsed in cold Millonig’s phosphate buffer at pH 7.4 and post-fixed in 2% phosphate-buffered osmium tetroxide for 60 min at 4°C. The tissue was then rinsed with Millonig’s phosphate buffer and dehydrated with three successive changes of ethanol for 15 min at each ethanol concentration (50%−70% −95% −100%), 3 changes of propylene oxide, and a final change of 1 part propylene oxide to 1 part epoxy resin uncorked overnight on a rotator. The supernatant fraction was decanted and the tissue was infiltrated with epoxy resin in a tightly corked vial for 1 h on a rotator and then transferred to capsules with epoxy mixture and allowed to polymerize overnight at 65°C. Thin sections were prepared and viewed on a JEOL-JEM-IOOCXII electron microscope.

Immunofluorescent localization employing antibodies to the 27×10³M_r gap junction protein revealed a ubiquitous distribution of immunoreactivity in both the mesenchyme and epithelium in all of the facial primordia as well as adjacent regions such as the roof of the stomodeum at all stages examined. Distinct punctate signal was seen throughout the mesenchyme and epithelium. There was a uniform distribution of signal with no indication of concentration in specific regions. This uniform and ubiquitous distribution can be seen in representative sections from the maxillary and lateral nasal processes and roof of the stomodeum at stage 22 (Fig. 2), the medial nasal and mandibular processes and roof of the stomodeum at stage 25 (Fig. 3), and from the maxillary process and frontonasal region at stage 28 (Fig. 4). TEM photographs (Fig. 5) confirmed the presence of gap junctions in both epithelium and mesenchyme. In a prior TEM study, gap junctions were found throughout mesenchyme not only between cell bodies but between cell bodies and cell processes and between cell processes (unpublished data).

The only exception to this ubiquitous distribution of gap junctions was a unique non-random pattern found in the epithelium of the nasal placode (Fig. 6). In this region, a distribution characterized by two zones was found; a distinct heavy concentration of signal in the superficial layer of cells adjacent to the exterior surface and a region devoid of detectable signal in the interior cell layer of the nasal placode adjacent to the mesenchyme. This pattern was seen during all stages of placode invagination that were examined. For example, in Fig. 6A this distinctive pattern is seen at the base of the nasal placode at stage 22 where signal is found only in the region adjacent to the nasal groove. The deeper cell layers of placodal epithelium are completely devoid of signal. In contrast, the subjacent mesenchyme displays the characteristic uniform distribution of signal found throughout the primordia of the developing face. The corresponding pattern was seen in nasal placode epithelium at stages 25 (Fig. 6B) and 28 (Fig. 6C).

A distinct change in the distribution of signal could be seen in the region where the placodal epithelium merges with the epithelium of the medial and lateral nasal processes (Fig. 7). In this transitional zone, the uniform distribution of signal in the epithelium of the nasal primordia becomes restricted to the superficial layers of the epithelium; in the thickened, invaginated region of the placode, the characteristic two zones are seen.

This pattern of distribution of signal was also observed in developing neuroepithelium of the brain (Fig. 8A and B). Here, as in the nasal placode, two zones were delineated. An extremely heavy concentration of signal was found in the neural epithelium lining the ventricles of the brain. The deeper layers were devoid of signal and presented a pattern identical to that of the placodal epithelium (compare Fig. 8 with Fig. 6 and 7). Separation of signal was also observed in the neural retina (Fig. 8C). In this instance, however, signal was concentrated in the layer adjacent to the pigmented retina. Because of the inversion of cell layers during optic cup formation, the region in which signal was concentrated in the neural retina corresponds to the cell layer lining the ventricle of the brain.

It should be noted that no physical barriers exist between the two regions in either the nasal placode or the brain. In the nasal placode the cells are contiguous with no morphological boundaries such as a basement membrane separating the cell layers (Fig. 9).

Preliminary studies employing immunolocalization with antibodies to connexin 43 also revealed nonrandom patterns of distribution. In contrast to the relatively photo-labile signal obtained with the anti-connexin 32 antibodies, a robust, stable signal was obtained with our antibodies to connexin 43. Initial studies with these antibodies indicate concentration of signal in the cell layers of the brain adjacent to the ventricle (Fig. 10). Distinct patterns of distribution also appear to be present in the retina. These patterns, in both the brain and retina, however, are somewhat discordant from those found with antibodies to the 27×10³M_r protein (compare Fig. 10 with Fig. 8). In addition, initial observations of distribution patterns in mesenchyme with antibodies to the 43×10³M_r protein indicate regional concentrations of signal, most noticeably in the lateral and medial nasal processes (data not shown). These patterns appear markedly different from those found with antibodies to the 27×10³M_r protein.

Identification of immunoreactive proteins recognized by these antibodies immunocytologically (above) was undertaken by Western blot analysis (Fig. 11). Pooled brain, retina and placode homogenates were analyzed using the antibodies to connexin 32 (Fig. 11A) and connexin 43 (Fig. 11B).

In most experiments with antibodies to connexin 32, it was difficult to resolve, reproducibly, antibody binding to proteins in chick samples even when membrane proteins were concentrated by extraction of soluble and peripheral membrane proteins with alkali. In some experiments, however, antibody binding was observed to proteins in the alkali residue of the chick material which co-migrated with 27×10³M_r liver gap junction protein and its 47xlO³M_r dimer (Fig. 11 A). Unlike the situation with the anti-connexin 32 antibodies, protein in the chick homogenate co-migrating with heart connexin 43 was generally detectable (not shown), as it was in alkali-residues (Fig. 11B). The difference between detectability of connexin 43 and connexin 32 immunoreactivity on Western blots appears to correlate with the difficulties encountered in obtaining clear, but labile, signal immunocytologically with the connexin 32 antibody and is likely to reflect the relative abundance of these proteins in the samples analyzed. However, it is also possible that tissue manipulations prior to SDS-PAGE lead to differences in recovery of these connexins. Alternatively, greater differences may exist in the primary structure of chick connexin 32 relative to that of rat as compared to the sequence of connexin 43 containing the peptide to which the anti-connexin 43 antibody was raised.

Localization of structures immunoreactive with an antibody to the 27×0³M_r liver gap junction protein (Hertzberg, 1984; Hertzberg and Skibbens, 1984), also referred to as connexin 32 (Paul, 1986) has revealed a ubiquitous distribution of immunoreactivity in all regions of the primary palate including the facial primordia. The distribution appeared to be uniform without region-specific concentration of signal in either mesenchyme or epithelium with the exception of one structure -the nasal placode. In the placodal epithelium, a unique, non-random pattern of distribution of signal was found.

The uniqueness of this pattern was striking. In only one other region of the craniofacial complex was such a pattern observed. The developing neuroepithelium of the brain and retina displayed compartmentalization of signal. For example, neuroepithelium lining the ventricles of the brain contained a concentration of signal as dense as that found in the superficial layers of the placodal epithelium. In a similar fashion, the deeper layers of the neuroepithelium were devoid of signal. It is noteworthy that prior studies, employing freeze fracture analysis (e.g. Revel and Brown, 1976; Schoen-wolf and Kelley, 1980), suggested the presence of juxtaluminal concentrations of gap’ junctions during development of the neural tube in the chick embryo. More recently Dermietzel et al. (1989) demonstrated the presence and regional distribution of gap junction proteins (connexin 26 and 43) in the developing mammalian brain. A pattern of immunoreactivity was observed that revealed separation of signal between the ventricular and subventricular zones similar in appearance to what was observed in the present study of the embryonic avian brain.

The reasons for the presence of this unique pattern are, at present, obscure. In both instances, the cell layers containing dense concentration of signal are zones of cell proliferation where all mitotic activity within the epithelium occurs. Communication by means of gap junctions could serve to synchronize those cell populations for co-ordinate proliferative activity. This, however, does not explain why the deeper regions of the cell layers of either structure are devoid of signal although it is possible that a different protein,, not recognized by our antibodies, is expressed in gap junctions of these regions. In the case of the developing brain, the deeper layer -the mantle layer -contains differentiating neurons that may require the absence of cell communication by means of gap junctions as part of the program by which axonal specificity is established. In this regard, it should be noted that in regions such as the developing retina of the chick embryo, associations have been found between the disappearance of gap junctions and the progressive differentiation of neural epithelium (Fujisawa et al. 1976).

This, however, would not be the case with placodal epithelium. Although olfactory neurons arise from the base of the nasal placode and send axonal projections to the olfactory centers of the brain, the major cell type in the nasorespiratory tract is pseudostratified ciliated columnar epithelium. In addition, other cell types are present whose lineage is, presumably, from placodal epithelium. It is difficult, therefore, to envision the necessity for an impedence of cell communication among cell populations with such diverse lineages.

Two recent studies that, however, may be of potential relevance to the pattern of distribution in placodal epithelium concern the origin of neurons of the central nervous system that produce luteinizing hormone -releasing hormone (LHRH) and which control the release of gonadotropic hormones from the anterior pituitary. These reports (Schwantzel-Fukuda and Pfaff, 1989; Wray et al. 1989) provide evidence that these cells originate in the olfactory epithelium and migrate into, and then through, the forebrain. Their path of migration is associated with the nervus terminalis medial to the axons of the olfactory nerves (Schwantzel-Fukuda et al. 1989). Many of the immunopositive-stained LHRH cells that migrated out of the olfactory epithelium formed what were described as ‘tracks’. These cells differed morphologically from those within the olfactory epithelium in that the former now became bipolar or unipolar and contained processes that appeared to contact each other (Wray et al. 1989). Since these ‘tracks’ extended from the olfactory epithelium into the telencephalon and rostral diencephalon, it is not unreasonable to consider a role for gap junctions in the coordination of migratory activity of this unique cell population.

An alternative explanation for the presence of this unusual pattern of gap junction distribution, and, by implication, of cell communication, may be related to morphogenetic changes that occur in these structures. The nasal placode invaginates to form a pit, then a groove and finally a tube, forming the nasal chamber and associated structures. The developing brain also undergoes numerous alterations in form associated with flexure of the neural tube, enlargement of hemispheres, constrictions, etc. The uniqueness of the pattern of gap junction distribution could be associated with the coordination of cell movement required for these extensive morphological changes that accompany the development of these structures.

In a prior study, performed several years ago (Minkoff, 1983), the distribution of gap junctions in mesenchyme of the maxillary process and the roof of the stomodeum was mapped employing transmission electron microscopy. A non-random distribution was found in this preliminary study which indicated that the concentration of gap junctions was greater in mesenchyme subjacent to maxillary process epithelium. Further, it was observed that the distribution of gap junctions in these regions (i.e. the maxillary process and roof of the stomodeum) tended to parallel previously reported differences in rates of cell proliferation (Minkoff and Kuntz, 1978; Minkoff, 1980a). These differences in gap junction distribution, however, were not supported by analysis employing immunolocalization with antibody to the 27×lO³M_r gap junction protein. The latter analysis demonstrated a random distribution in the mesenchyme of both the maxillary process and roof of the stomodeum.

These contradictions may be a reflection of the necessity to examine very small regions of the maxillary process and roof of the stomodeum with TEM and then to combine these data for analysis. The difficulties encountered in obtaining greater amounts of data and increased sample size led to the use of alternative techniques employed in the present study. Quantitative analysis of the numbers of gap junctions and cell bodies per unit area, however, employing immunofluorescent localization with antibodies to the 27 ×10³M_r protein indicated a random distribution even when normalized for cell number (data not shown). The resolution of this issue will require additional studies, now in progress, with antibodies to other gap junction proteins. Preliminary studies of 43×10³M_r gap junction distribution, as noted above, revealed concentrations of signal in the nasal primordia and show promise of clarifying this issue.

The pattern of immunoreactivity described for the 27×10³M_r protein may well represent but a subset of gap junctions recognized by these antibodies. Distinct proteins homologous to the 27×lO³M_r liver gap junction protein (Beyer et al. 1988) have been identified in fiver (21×10³M_r on SDS-polyacrylamide gels) (Nicholson et al. 1987; Hertzberg et al. 1988), heart (connexin 43) (Manjunath et al. 1987; Beyer et al. 1987) and lens epithelial cells (MP70) (Kistler et al. 1988). It has also been shown that individual gap junctions can have more than one gap junction protein (e.g. 27 and 21×10³M_r in hepatocytes (Nicholson et al. 1987)). Furthermore, heterologous coupling has been demonstrated in Xenopus oocytes expressing different connexin mRNAs leading to channels with asymmetric gating properties (Swenson et al. 1989; Werner et al. 1989). Hence, the possibility remains of communication compartments in the tissues examined beyond those delineated here. In addition, discrete regulatory mechanisms within regions of tissue reflecting gating properties of individual gap junction proteins (connexins) may also arise in these regions during development.

For these reasons, continuing studies of the 43×10³M_r gap junction protein, as well as other gap junction proteins, are indicated especially in view of the difficulty encountered in obtaining a signal with antibodies to the 27×lO³M_r rat liver gap junction protein on Western blot analysis. Analysis of differential expression of gap junction proteins by Dermietzel et al. (1989) may be particularly significant in this regard. Connexins 26 and 43 were found in embryonic mammalian brain tissue while connexin 32 was reported to be virtually absent. Only after birth was connexin 32 expressed to a significant degree. This coincided with decreased expression of connexin 26 in the postnatal mammalian brain. Connexin 43 expression was observed in both embryonic and mature brain tissue. As noted by Dermietzel et al. (1989) discrepancies between this and other studies (Dudek et al. 1988; Nagy et al. 1988; Shiosaka et al. 1989) may be a result of differences in the composition of polyclonal sera.

The polyclonal antibody to the 27×10³M_r gap junction protein employed in the studies reported here may have recognized a broader spectrum of antigenic sites, potentially including those on other connexins (or related proteins) present in an abundance too low to visualize routinely on Western blots. However, our ability to demonstrate antibody binding to a protein in chick samples co-migrating with liver connexin 32 in some, but not all, Western blot experiments does suggest its presence, albeit at low abundance, in these samples.

This study was supported by NIH grants DE-07674 (R.M.), GM3O667 (E.L.H.) and a Research Career Development Award (HD00713 to E.L.H.). E.L.H. is the recipient of an Irma T. Hirschi Career Scientist Award. The authors wish to express their appreciation to Mr Udayan Parikh for assistance with the transmission electron microscopy, to Mr Richard Coprina for assistance with the Western Blots, and to Ms Barbara Barton for her expert secretarial assistance in the preparation of this manuscript.