ABSTRACT
Gap junctions in the epithelium and superficial fiber cells from young mice were examined in lenses prepared by rapid-freezing, and processed for freeze-substitution and freeze-fracture electron microscopy. There appeared to be three structural types of gap junction: one type between epithelial cells and two types between fiber cells. Epithelial gap junctions seen by freeze-substitution were ∼20 nm thick and consistently associated with layers of dense material lying along both cytoplasmic surfaces. Fiber gap junctions, in contrast, were 1516 nm (type 1) or 17-18 nm thick (type 2), and had little associated cytoplasmic material. Type 1 fiber gap junctions were extensive in flat expanses of cell membrane and had a thin, discontinuous central lamina, whereas type 2 fiber gap junctions were associated with the ball- and-socket domains and exhibited a dense, continuous central lamina. Both types of fiber gap junction had a diffuse arrangement of junctional intramembrane particles, whereas particles and pits of epithelial gap junctions were in a tight, hexagonal configuration. The type 2 fiber gap junctions, however, had a larger particle size (∼9 nm) than the type 1 (∼7.5 nm). In addition, a large number of junctional particles typified the E-faces of both fiber types but not the epithelial type of gap junction. Gap junctions between fiber and epithelial cells had structural features of type 1 fiber gap junctions. These structural features suggest that the epithelial and the type 2 fiber gap junctions are of the common communicating type, whereas the type 1 fiber gap junctions may represent a new type of intercellular contact, not necessarily even derived from gap junction proteins, which serves an adhesive function in the lens.
INTRODUCTION
The lens is composed of numerous sheets of slender fiber cells covered by a monolayer of epithelium at its anterior surface. The lens permits incident light to pass through and helps to form a focused image on the retina through the mechanism of accommodation. The lens possesses several unique features serving these specific functions. It contains a high concentration of transparent crystallin proteins, which increase the refractive index. The lens has no blood supply and its mature fiber cells lose their organelles during the maturation process, which eliminates light scattering. Metabolic activities in the lens cortex and nucleus are low and proceed anaerobically and the exchanges of ions and small metabolites between lens cells depend on gap junctions (Rae, 1979; Goodenough et al., 1980; Mathias and Rae, 1989). Lens fibers also possess an elaborate interlocking system and adherens junctions for maintaining their structural stability, especially during the deformation that accompanies visual accommodation (Cohen, 1965; Kuwabara, 1975; Harding et al., 1976; Kuszak et al., 1980; Willekens and Vrensen, 1981; Lo, 1988).
Interlocking connections between lens fibers include ball- and-sockets, protrusions, mounds, spikes and tongue- and-grooved ridges (Dickson and Crock, 1972; Kuwabara, 1975; Harding et al., 1976; Kuszak et al., 1980; Willekens and Vrensen, 1981; Lo and Harding, 1984). Large numbers of ball- and-socket domains are found in the superficial cortex and, at reduced frequency, in the deeper cortex and lens nucleus in many species. They are typically distributed most abundantly along the wide sides of the hexagonal fiber cells (Harding et al., 1976; Kuszak et al., 1980; Willekens and Vrensen, 1981). Protrusions resemble ball- and-socket domains but are much smaller and are located primarily at the corners of fiber cells (Willekens and Vrensen, 1981). The tongue- and-grooved ridges are small membrane inter- digitations between deep cortical and nuclear fiber cells of the lens. They are associated with 12-nm, wavy thin junctions or square array junctions (Zampighi et al., 1982, 1989; Lo and Harding, 1984; Costello et al., 1985, 1989). The adherens junctions between lens cells are believed to play a role in binding cells together (Lo, 1988). Tight junctions and desmosomes are lacking between fiber cells (Maisel et al., 1981; Rafferty, 1985; Lo, 1988).
Gap junctions between lens fiber cells of various species have been demonstrated by morphological and physiological studies (Leeson, 1971; Benedetti et al., 1974; Kuszak et al., 1978; Rae, 1979; Goodenough, 1979; Lo and Harding, 1986; Itoi et al., 1991). Morphological studies show that fiber gap junctions are unusually large both in size and in number, and that their structural characteristics and distribution are also unique. For example, the typical 2-4 nm intercellular gap has rarely been seen in the fiber gap junctions of the intact lens (Kuwabara, 1975; Goodenough, 1979; Lo and Harding, 1984; Lo and Kuck, 1987; Costello et al., 1989; Itoi et al., 1991). Also, gap-junctional plaques are distributed mainly in a single row along the middle of the narrow sides of hexagonal fiber cells, but have a random distribution on their wide sides (Lo and Harding, 1986; Gruijters et al., 1987a). In addition, fiber gap junctions have been found frequently in the ball- and-socket domains in all species studied (Kuwabara, 1975; Okinami, 1978; Goodenough, 1979; Lo and Harding, 1986; Lo and Kuck, 1987). Electrophysiological studies suggest that fiber cells are electrically coupled throughout the lens (Duncan, 1970; Rae, 1979; Mathias et al., 1991).
Biochemical and immunocytochemical studies have shown that several proteins are associated with fiber gap junctions. Candidates include MP46, MP56, MP70 and MP20 (Kistler et al., 1985, 1988; Galvan et al., 1989; Louis et al., 1989; Paul et al., 1991; Rup et al., 1993). Recently, both Cx46 and Cx50 (MP70) have been shown to be members of a connexin family of fiber gap junction proteins in the lens (Paul et al., 1991; White et al., 1992). These two connexins are not found in the lens epithelium from which the fiber cells are derived. On the other hand, connexin43 has been proven to be a gap junction protein in the lens epithelium, cardiac cells and other tissues, but is not found in the fiber gap junctions (Beyer et al., 1989; Musil et al., 1990).
In this study, we investigate gap junctions in the young mouse lens using a rapid-freezing technique in order to avoid any structural alteration during chemical fixation. Freshly isolated mouse lenses were slam-frozen with a liquid helium freezing apparatus, and then processed by freeze-substitution and freeze-fracture electron microscopy. These new techniques reveal that there is one epithelial type and two fiber types (type 1 and type 2) of gap junctions based on their distinct structural characteristics. The type 1 fiber gap junctions are associated mostly with the flat cell membranes, whereas the type 2 fiber gap junctions are specifically associated with ball- and-socket domains. Certain aspects of type 1 fiber gap junctions are atypical of gap junction structure because they exhibit a thinner pentalam- inar profile with a discontinuous, thin central lamina and a smaller particle size than the type 2 fiber gap junction and the epithelial gap junction. We suggest that the epithelial type and the type 2 fiber gap junctions are real communicating junctions, whereas the type 1 junction serves an adhesive function in the lens.
MATERIALS AND METHODS
Mice (CD-1, Charles River), 10-15 days old, were decapitated and the intact lenses were quickly removed from freshly enucleated eyes. Lenses were oriented so that the anterior surface (including the epithelium and fibers) or the posterior surface (including the fibers only) could be quickly (within 60 s of decapitation) slam frozen on a copper block cooled by liquid helium. No fluids were added to the lens surface as the transfer and freezing were done rapidly enough to preclude drying.
Freeze-substitution
Frozen lenses mounted on the aluminum substrata were immersed in acetone containing 5% osmium tetroxide at −80°C for 12-24 h and then warmed up gradually over 12 h to 4°C. The tissue was rinsed in acetone, and stained in saturated uranyl acetate in acetone for 1 h followed by 0.1% hafnium chloride in acetone for 1 h at room temperature. Blocks were rinsed in methanol and embedded in Araldite (Polysciences, Inc.). Thin sections (80 nm) were stained with 5% uranyl acetate followed by Reynold’s lead citrate, and examined with a JEOL 1200EX or JEOL 200CX electron microscope.
Freeze-fracture
Cryofractures of rapidly frozen lenses were made in a modified Balzers 301 freeze-fracture unit, at a stage temperature of -160°C in a vacuum of approximately 2×10−7 Torr (1 Torr = 133.3 Pa). The lens tissue was fractured by scraping a steel knife across a frozen surface to an appropriate depth to explore either the epithelium or fiber cells underlying the lens capsule. The fractured surface was immediately replicated with a platinum-iridium-tantalum alloy followed by carbon film. The replicas, obtained by unidirectional or rotary shadowing, were cleaned with household bleach and examined with the electron microscope.
Scanning electron microscopy
Freshly isolated lenses of young mice (CD-1, 2-3 weeks old) were fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.3, at room temperature for 24-48 h. Lenses were postfixed in 1% aqueous OsO4 for 1-2 h at room temperature, dehydrated in a graded ethanol series and dried in a critical-point dryer. Lenses were oriented and mounted on the specimen stub, and the lens cortex was carefully fractured with a sharp razor blade or needle to expose the fiber cell membrane surface, which was subsequently coated with gold in a sputter coater. Micrographs were taken with a JEOL 820 scanning electron microscope at 10 kV.
RESULTS
Observations of gap junctions were confined to the epithelium and a few layers of underlying superficial fiber cells, where ice crystal damage remained minimal. One epithelial type and two fiber types of gap junctions were consistently recognized, even in the same section or the same replica. Two fiber types of gap junctions were also observed in the same section or the same replica prepared from the posterior surface of the lens. Structural details characteristic of each type of gap junction served as the basis for this classification.
Gap junctions between superficial fiber cells
Two types of gap junction between superficial fiber cells were recognized after freeze-substitution. Type 1 gap junctions were 15-16 nm thick, with a discontinuous, thin central lamina. They were typically long (up to ∼10 μm) and were usually associated with the flat regions of the cell membrane (Fig. 1A,B). The type 2 fiber gap junctions were distinctly thicker, 17-18 nm across, and always exhibited a continuous, heavy central lamina sometimes bisected by an intercellular gap. The type 2 fiber gap junction was specifically associated with the ball- and-socket domains, as shown by their ring shape in cross-sections (Fig. 2A,B). They were much shorter than the type 1 junctions, with lengths ranging from 0.5 to 1.5 μm. An important difference between the two types of fiber gap junction was in the overall thickness of the junctional profile and its intercellular space. The type 2 junction had a wider intercellular space, which stained more densely during freeze-substitution, giving it the appearance of a continuous, thick central lamina. In contrast to the epithelial type of gap junction, both fiber types of gap junctions had little fuzzy material on their cytoplasmic surfaces.
Freeze-fracture electron microscopy showed distributions of particles characteristic of gap junctions (Goodenough, 1979). Our results were consistent over three sessions of freezing and were always studied in areas of the replica free of freezing and drying artifacts. The gap-junctional particles and the complementary pits of both type 1 and type 2 fiber gap junctions were in a loose, random configuration (Figs 1C, 2C). Also, in contrast to the epithelial gap junctions, many of the junctional particles (>25%) remained on the E-face in type 1 and type 2 junctions (Figs 1C, 2C). However, the type 1 gap junction had distinctly smaller intramembrane particles (∼7.5 nm in diameter) than the type 2 fiber gap junction (∼9 nm; Figs 1C, 2C). Viewed in a cartographic stereo viewer (equipped with a height measurement device), the particles in the type 2 junction appeared to be more variable in height and to include particles higher than those in the type 1 junction.
The finding of a specific association between the type 2 fiber junction and the ball- and-socket domains prompted us to examine the distribution of this domain by scanning electron microscopy. Many ball- and-socket configurations were found on the wide sides of hexagonal cortical fiber cells (Fig. 2D). Their range of sizes was consistent with the sizes of type 2 fiber gap junctions seen by freeze-substitution and freeze-fracture.
Gap junctions between epithelial cells
Freeze-substitution revealed that epithelial gap junctions were usually small (∼0.3 μm) in diameter and were typically found between lateral membranes near the apical surface of the epithelium (Fig. 3A). The epithelial gap junctions were ∼20 nm thick and had distinct layers of fuzzy material (∼1.5 nm thick) consistently associated with the junctional membranes at their cytoplasmic surfaces (Fig. 3B). This junctional profile had a continuous central lamina.
Freeze-fracture showed that these junctions have an intramembrane structure typical of gap junctions in other locations; arrays of particles on their P-faces and complementary pits on their E-faces arranged in a tight, hexagonal arrangement (Fig. 3C). These gap junction particles were 8.5-9 nm in diameter and very few remained on the E-face of junctional membrane.
Gap junctions at the interface between the epithelium and fiber cells
Gap junctions were readily found at the interfaces between the epithelium and fiber cells (Fig. 4A), and they exhibited structural features similar to the type 1 fiber gap junctions (Fig. 4B). Gap junctions in the central and the peripheral regions of the epithelium were similar in size. Because cell membranes at the epithelial-fiber interface have an orientation similar to the tortuous lateral membranes near the apical surface of the epithelium as well as to the underlying fiber cell membranes, we were unable definitively to identify these gap junctions in freeze-fracture.
DISCUSSION
By the structural criteria used here, there are three types of gap junction near the surfaces of the young mouse lens: two types of fiber gap junctions and one epithelial type.
The epithelial gap junctions closely resemble cardiac gap junctions (Manjunath et al., 1984, 1987; Green and Severs, 1984; Shibata et al., 1985; Deleze and Herve, 1986). They have a similar thickness, distinct layers of fuzzy material along their cytoplasmic surfaces, and the same 43 kDa con- nexin (Beyer et al., 1989; Musil et al., 1990). It is therefore reasonable to conclude that the epithelial gap junctions are of the communicating type. However, this study shows that the particle arrangement of epithelial gap junctions in the mouse lens differs from those seen in the rapid-frozen chick lens (Miller and Goodenough, 1985). The gap junction particles of the mouse lens show a tight, hexagonal packing whereas those of the chick lens exhibit a loose arrangement; we attribute this structural heterogeneity to a species difference. Resolution of this issue would require further comparative investigation, which is outside the scope of this paper, and does not affect our conclusion that there are three types of junction. A similar difference in the particle arrangement of lens epithelial gap junctions among several other species has also been observed in the glutaraldehyde-fixed lenses by Lo and Harding (1986). They found that epithelial gap junctions in the frog lens always showed a loose particle packing whereas most epithelial gap junctions in other species (e.g. human, rabbit and rat) had a tight, crystalline configuration.
The structural characteristics of both types of fiber gap junction differ significantly from the epithelial gap junctions. Type 1 fiber gap junctions show a very thin profile (∼15 nm) and small particle size (∼7.5 nm) in contrast to the type 2 fiber gap junctions, which have a thickness (∼18 nm) and particle size (∼9 nm) similar to gap junctions in liver and other tissues as viewed with rapid-freezing electron microscopy (Raviola et al., 1980; Hirokawa and Heuser, 1982; Shibata et al., 1985; Hanna et al., 1985). The type 2 junctions also can display a real 2-4 nm intercellular gap, which is always lacking in the type 1 gap junctions. In contrast, the type 1 junctions consistently have an appearance atypical of gap junctions in that the center of the pentalaminar profile is very thin and discontinuous. There are also characteristic differences in the distribution of the type 1 and 2 fiber junctions. The type 1 fiber gap junction is associated with flat expanses of cell membrane, whereas the type 2 is associated with ball- and-socket domains.
The intramembrane particles in type 1 junctions can be induced by high calcium, low pH and anoxia to form orthogonal or rhomboic arrays (Peracchia and Peracchia, 1980a,b; Bernardini and Peracchia, 1981). This pattern of array differs from the tight, hexagonal pattern commonly seen in a typical gap junction of other cell types under the same treatments (Peracchia, 1980; Raviola et al., 1980). These observations and the present ones suggest that the type 2 junction may simply be a variant of the communicating type of gap junction found in many tissues, whereas the type 1 junctions may be either a new type of gap junction or a junction of another type, presumably not involved in intercellular communication.
The significance of the existence of two types of fiber gap junctions in the lens remains to be determined. There are two gap junction proteins (connexins 28 and 21) in the liver, although they appear to be co-localized in a single type of hepatic gap junction (Nicholson et al., 1987). Also, two or three gap junction proteins can be expressed in cardiac and brain tissues during development (Dermietzel et al., 1989; Beyer, 1990). The presence of multiple gap junction proteins could provide cell-to-cell communication of substances with different properties (Beyer, 1990). Indeed, the ionic and dye coupling between lens fiber cells could depend on the two different types of gap junctions (Rae, 1979; Goodenough et al., 1980; Rae and Kuszak, 1983; Mathias et al., 1991).
An alternate possibility is that the type 1 fiber gap junction is not a communicating gap junction but serves for adherence of fiber cells to one another. There are several reasons to consider this possibility. First is the unusually large size and number of type 1 fiber junctions and the fact that they exhibit structural features atypical of gap junctions in other tissues. It is not clear whether there is a need for such extensive junctions because it is not clear whether the cortex and nucleus of the lens are metabolically active. The type 2 fiber gap junctions would provide an adequate cell-cell communication for metabolically active cortical fiber cells in the more superficial regions while the type 1 junctions are suitably deployed to maintain the narrow intercellular spaces over large areas required for the normal function of the lens (Mathias and Rae, 1985). The type 1 gap junctions could, therefore, assist the small-sized adherens junctions in providing structural stability of the lens during accommodation even if it were a species of gap junction (Lo, 1988). Another example of the changing uses of proteins during evolution of the lens is that several major lens crystallins are the products of genes for specific enzymes (Piatigorsky and Wistow, 1989; Wistow et al., 1990; Rao and Zigler, 1991). Despite the fact that these lens crystallins show enzymatic activities, they only play a structural role in the refractive properties of the lens. Perhaps the type 1 fiber gap junctions retain some structural features of gap junctions and have an adhesive, but not a communicating, function.
It should be noted that type 2 fiber gap junctions are specifically associated with the ball- and-socket domains. The ball- and-sockets are located mostly on the wide sides of fiber cells and are larger in size but much smaller in number than the protrusions found at corners (Willekens and Vrensen, 1981). Also, they are more frequent in the superficial cortex than in the deeper cortex and nucleus. Why type 2 fiber gap junctions are limited to the ball- and-socket domains is also not known. Perhaps the type 2 fiber gap junctions can exchange ions and metabolites more efficiently in the ball- and-socket domains, because these domains protrude deeply into neighboring fiber cells.
There are several membrane proteins associated with fiber junctions (e.g. MP26, MP46, MP56, MP70 and MP20) in the lens (Bok et al., 1982; Fitzgerald et al., 1983; Nicholson et al., 1983; Kistler et al., 1985; Sas et al., 1985; Galvan et al., 1989; Louis et al., 1989; Paul et al., 1991; Rup et al., 1993). Both Cx46 and MP70 (Cx50) have recently been shown to be members of a connexin family of fiber gap junction proteins in the lens (Paul et al., 1991; White et al., 1992). Antibodies to MP70 and Cx46 become localized in flat fiber gap junctions at both light and electron microscope levels (Kistler et al., 1985; Gruijters et al., 1987b; Paul et al., 1991). The same antibodies have also been localized in the ball- and-socket domains as determined by fluorescence labeling (Kistler et al., 1986; Gruijters, 1989). However, our re-examination of the photographs in these studies indicates that these ball- and-sockets should be classified as protrusions, according to the classification of Willekens and Vrensen (1981) and that used in the present study, because they are all located at the corners of fiber cells. The MP26 protein appears to be present in both square-array junctions and non-junctional membranes (Paul and Goodenough, 1983; Zampighi et al., 1989). Squarearray junctions are structurally quite different from fiber gap junctions. Square-array junctions have a 12-13 nm wide pentalaminar profile and are associated with tongue- and-groove or ridge interdigitations of deep cortical and nuclear fiber cells (Lo and Harding, 1984; Costello et al., 1985, 1989; Zampighi et al., 1989). Although MP26 can form channels in single bilayers (Ehring et al., 1990), it is not related to other members of the connexin family of gap junction proteins (Hertzberg et al., 1982; Nicholson et al., 1983; Gorin et al., 1984). In addition, MP20 has also been shown to be a possible structural protein of fiber gap junctions (Galvan et al., 1989; Louis et al., 1989). Immunolocalization studies reveal that MP20 antibody is labeled on the flat fiber gap junctions (Galvan et al., 1989). Molecular cloning studies indicate that MP20 is not related either to fiber connexins or to MP26 (Gutekunst et al., 1990; Kumar et al., 1993). Future studies on the isolation of two types of fiber gap junctions should provide further insights into the chemical and functional properties of the controversial fiber gap junctions in the lens.
ACKNOWLEDGEMENTS
The authors thank Kasia Hammer and John P. Chludzinski of NINDS/NIH, and Adell Mills of Morehouse School of Medicine for their excellent technical assistance. This study was supported in part by NIH grant R01 EY05314 (WKL).