Gap Junction Processing and Redistribution Revealed by Quantitative Optical Measurements of Connexin46 Epitopes in the Lens

Abstract

purpose. To map changes in the structure and function of fiber cell gap junctions that occur with lens differentiation.

methods. Equatorial lens sections were fluorescently labeled with antibodies to the gap junction protein connexin (Cx)46, the membrane marker wheat germ agglutinin, and the nuclear stain propidium iodide. Two-photon microscopy and digital image analysis were used to quantify label and cell morphology as a function of radial distance (r/a) across the lens. Loop- and tail-specific Cx46 antibodies were used to identify regions of posttranslational modification. Local fiber cell coupling was imaged in situ using two-photon flash photolysis of caged fluorescein.

results. Antibody labeling showed that the cytoplasmic tail of Cx46 was removed in two zones (r/a ∼ 0.9 and r/a ∼ 0.7). In addition, with increasing depth, the large radially aligned plaques of peripheral fiber cells became fragmented and dispersed around the cell membrane, and cells became more circular in cross section. Fluorescein transfer between peripheral fiber cells was highly anisotropic and occurred predominantly within a column of fiber cells, resulting in radially directed transport. In regions beyond the zone of nuclear loss, transport was more isotropic and occurred across columns of fiber cells.

conclusions. The cleavage of Cx46 is associated with a spatial redistribution of gap junction plaques. The distribution of gap junction plaques around the cell membrane can explain the observed directionality of intercellular solute transfer. The findings suggest that the processing and redistribution of gap junction proteins is central to controlling radial and circumferential solute gradients in different regions within the lens.

Gap-junction–mediated intercellular communication can have profound effects on cell development, differentiation, and growth. ^{1 2} Gap junction channels are formed by the connexin family of proteins, which comprises at least 20 distinct isoforms. ³ At the molecular level, gap junction channels consist of two hemichannels (or connexons) that are in turn composed of six connexin subunits. ⁴ The ability to form gap junctions depends on the isoforms present, ⁵ and gap junctions formed by different isoforms exhibit distinct channel permeabilities, gating properties, and regulation. ⁶ It follows that the connexin composition of a gap junction plaque not only determines the extent of intercellular communication, but also the types of messages that can be exchanged between cells. Thus, during differentiation and growth, changes in the pattern of connexin expression among cells can create selective communication pathways that dictate the ultimate fate of particular groups of cells. ⁷  

In the ocular lens, changes in expression level of three connexin isoforms after differentiation have been noted. ⁸ Other studies have shown that the size, morphology, and subcellular localization of gap junction plaques vary as fiber cells elongate. ^{9 10 11} Furthermore, it has been proposed that regional variations in gap junction plaque composition and density could contribute to the generation of a lens microcirculation system. ¹² This system facilitates nutrient delivery, metabolic waste removal, and cell volume regulation in the avascular lens. ^{13 14} The importance of gap junction channels to the maintenance of lens transparency has been highlighted by the discovery that mutations in the genes encoding either of the fiber cell connexins cause cataract in humans. ^{15 16} Furthermore, targeted ablation of either Cx46 ¹⁷ or Cx50 ¹⁸ in mice produces distinctive cataractous phenotypes, suggesting that each connexin isoform contributes specific properties to intercellular communication in the lens. 

The relative simplicity of the lens connexin expression pattern and the extensive literature on differentiation and development in the lens make it a useful system in which to assess the relationship between epithelial cell differentiation and connexin expression patterns. The lens consists of two cell types: a cuboidal epithelium that covers the anterior surface of the lens and the fiber cells that form the bulk of the lens. At the equator, the cuboidal cells divide, differentiate, and elongate to form the fiber cells, which have a characteristic hexagonal cross section. ¹⁹ During the course of fiber cell differentiation, cells lose their nuclei and other light-scattering cellular organelles. ²⁰ This process of fiber cell differentiation continues throughout life, with new fiber cells being laid down in the lens periphery over existing older fiber cells, which become internalized, creating an age or differentiation gradient across the lens. Whereas the cuboidal epithelial cells express Cx43, fiber cell differentiation is associated with the expression of Cx46 ²¹ and Cx50. ²² Biochemical experiments conducted on fiber cell membranes collected from either the outer cortex or inner core of the lens have shown that both Cx46 and Cx50 undergo age-dependent processing that removes the cytoplasmic tails of the connexin proteins. ^{10 23} Although it is known that the cleavage of Cx50 coincides with the loss of fiber cell nuclei and is performed by the protease calpain, ²⁴ neither the location of Cx46 cleavage in the lens nor the enzyme(s) that performs cleavage is known. 

In a preliminary study ²⁵ we used two-photon microscopy (TPM) ²⁶ and advanced image analysis ^{27 28} to quantify the distribution of the uncleaved Cx46 gap junction protein. To clarify the relationship between the subcellular configuration of gap junctions, their possible modification by cleavage, and their functional properties, we comprehensively extended those studies. To locate the region in the lens where cleavage of Cx46 occurs, immunocytochemical mapping of the cleaved and uncleaved forms of Cx46 was performed and related to markers of fiber cell differentiation. Functional consequences of the structural changes in gap junctions through fiber cell differentiation were then probed in situ with a novel two-photon–activated dye transfer method. Our data show that Cx46 is cleaved at two distinct stages in fiber cell differentiation and that the subcellular distribution of gap junction plaques and associated local patterns of fiber cell coupling also change markedly as a function of this differentiation. The results are discussed with reference to additional markers of fiber cell differentiation ²⁹ and previous functional studies. ^{12 30}  

Materials and Methods

Antibodies and Reagents

Rabbit anti-Cx46 cytoplasmic tail (carboxyl terminus)–specific, affinity-purified antibodies were obtained from Alpha Diagnostic International (San Antonio, TX). Rabbit anti-Cx46 cytoplasmic-loop–specific antibodies were kindly provided by Nalin Kumar (Dept. of Ophthalmology and Visual Sciences, University of Illinois at Chicago, Chicago, IL). Goat anti-rabbit Alexa Fluor 488 secondary antibody, Alexa Fluor 350–conjugated wheat germ agglutinin (WGA), fluorescein bis-(5-carboxymethoxy-2-nitrobenzyl) ether [CMNB-caged fluorescein], and 200-nm-diameter subresolution fluorescent beads were obtained from Molecular Probes (Eugene, OR). Unless otherwise stated, all other chemicals were purchased from Sigma-Aldrich Australia/New Zealand (Sydney, Australia). Phosphate-buffered saline (PBS) was prepared from tablets (Sigma), yielding (in mM): phosphate buffer 10, KCl 2.7, and NaCl 137 (pH 7.4). 

Tissue Sectioning

Twenty-one-day-old Wistar rats were killed by CO₂ asphyxiation in accordance with protocols approved by the University of Auckland Animal Ethics Committee and with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Lenses were fixed and cryosectioned according to a protocol designed to obtain a balance between antibody signal and morphology. Briefly, lenses were dissected in PBS and fixed in 0.75% wt/vol paraformaldehyde in PBS for 24 hours at room temperature. Lenses were then washed 3 × 10 minutes in PBS followed by cryoprotection in 10% wt/vol sucrose in PBS for 1 hour at room temperature, followed by 20% sucrose in PBS and then 30% sucrose in PBS at 4°C overnight. Fixed lenses were stored up to 7 days before mounting in optimal cutting temperature (OCT) compound (Tissue-Tek; Sakura Finetek, Torrance, CA) at 4°C on prechilled chucks. Chucks were immersed in liquid nitrogen for 25 seconds to freeze the lens and then kept on dry ice. Ten-μm-thick sections were cut with disposable cryosectioning blades (S-35; Feather Safety Razor Co., Osaka, Japan) at −18°C on a cryostat (CM3050; Leica, Heidelberg, Germany) and placed onto slides coated with poly-l-lysine. Slides were then washed four times for 5 minutes each in PBS before labeling with Cx46 antibodies and WGA. 

Fluorescent Labeling

All antibody labeling was performed either at room temperature for the times indicated or at 4°C overnight in a humid box. Slides were treated for 1 hour with blocking solution (3% wt/vol bovine serum albumin, 3% vol/vol fetal calf serum in PBS), washed three times for 5 minutes each in PBS, and treated for 2 hours with anti-Cx46 antibody diluted 1:200 in blocking solution. Slides were washed three times for 5 minutes each in PBS and treated for 1.5 hours in the dark with anti-rabbit Alexa Fluor 488 antibody diluted 1:200 in blocking solution. After washing three times for 5 minutes each in PBS, slides were labeled for 1.5 hours in the dark with Alexa Fluor 350–conjugated WGA (40 μg/mL in PBS). Finally, after washing three times for 5 minutes each in PBS, slides were mounted in 10 μL antifade medium (AF1; Citifluor, London, UK) and stored at 4°C. The lectin WGA labels glycosylated membrane proteins and provides a reliable marker of fiber cell surface, ³¹ as well as a benchmark for comparison with other labels. To stain nuclei, slides were washed 5 minutes in 125 μM propidium iodide in PBS. 

Two-Photon Microscopy

Excitation for TPM was provided by a mode-locked Ti:sapphire laser (Coherent, Santa Clara, CA) that was coupled to a modified confocal microscope ³² (LSM 410; Carl Zeiss Meditec, Jena, Germany). The microscope provided either TPM and confocal imaging modes or a confocal scanning mode with stationary spot two-photon excitation. The Ti:sapphire laser was tuned to a center wavelength of approximately 755 nm to excite both labels simultaneously. Slides were imaged using a 40× 1.2-numeric-aperture, water-immersion objective. Excitation and emission filters were custom-made to the authors’ specifications. The main dichroic filter separating illumination and emission wavelengths from each other was a short-pass filter with an edge centered at 650 nm (Chroma Technology, Brattleboro, VT). Emitted light was collected from both fluorescent labels simultaneously, with all emission pinholes fully open (because optical sectioning in TPM does not require a confocal pinhole). The emitted light was split into two channels by a long-pass filter centered at 495 nm (495DRLP; Omega Optical, Brattleboro, VT). Emission band-pass filters centered at 450 nm (450DF65, bandwidth 65 nm; Omega Optical) and 535 nm (HQ535/50, bandwidth 50 nm; Chroma Technology) were used to collect the signals from Alexa Fluor 350 and Alexa Fluor 488, respectively. Image stacks were collected at a sampling resolution of approximately 200 nm in plane and 300 nm along the optical axis. To extend the total field of view, we acquired a number of 1024 × 1024 × 30-pixel image stacks, so that a total region of 750 × 385 μm was covered. Images were acquired at a gray-scale resolution of 8 bits. However, the effective bit resolution in Figure 3 is more than 18 bits. The reason for that is twofold: (1) D/D _max, where D is density, was calculated as a (floating-point) average over more than 1000 pixels. This increases the bit resolution by approximately 10 bits; and (2) the photomultiplier tube (PMT) gain was varied when acquiring individual tiles, increasing it to fit the dynamic 8-bit range at deeper locations where signals were weaker. The intensities of image stacks were adjusted for these changes in gain (with all data in 32-bit floating point) when images were combined into the montage. This was achieved by referring to a calibration curve that tabulated detector gain versus nominal software gain settings, which had been recorded for each stack. Scanned images were written directly to hard disc for off-line analysis and archived on recordable CDs. Point-spread functions (PSFs) for deconvolution were determined by imaging 200-nm-diameter subresolution fluorescent beads. The measured PSF had a full width at half maximum of 400 nm in plane and 900 nm axially. 

Flash Photolysis Experiments

Lenses were dissected into PBS at 37°C and transferred to a plate for cutting, where they were bathed in an intracellular medium containing (in mM) MgCl₂ 1, EGTA 0.5, NaCl 10, Na₂ATP 2, KCl 20, K-gluconate 120, and HEPES 10 (pH 7.3; 300 mOsm/kg). Lenses were cut in half through the equator with a fresh scalpel blade. Lens hemispheres were inspected for structural integrity by dissecting microscope and transferred to a perfusion chamber sealed at the bottom with a coverslip so that the cut surface of the lens faced the coverslip. The chamber contained 1 mM CMNB-caged fluorescein and 100 μg/mL WGA in intracellular medium. The chamber was mounted on the confocal/two-photon microscope and the caged fluorescein released by controlling the beam intensity from the Ti:sapphire laser with a Pockels cell. The stationary spot two-photon flash photolysis (TPFP) was restricted to the focal volume of the microscope (∼400 × 400 nm in the x–y plane; ∼900 nm along the optical axis ³³ ). The time course of the diffusion of the released fluorescein was measured by scanning the 488-nm line of the Ar⁺ laser associated with the confocal microscope across the field of view simultaneously. After collection of time course data, the Ti:sapphire laser beam path was switched to scanning mode, and TPM was used to image the cell membranes labeled by WGA. 

Image Processing

Image processing and analysis were performed by using custom routines written in the IDL programming language (Research Systems, Boulder, CO) running on a computer workstation (Silicon Graphics, Mountain View, CA). Distributions of WGA and Cx46 signals were isosurface-rendered using the OpenDX data visualization package (http://www.opendx.org; Visualization and Imagery Solutions, Inc., Missoula, MT). Digital deconvolution of image stacks was performed by using a maximum-likelihood algorithm with the measured PSF. ³⁴ The image-processing strategy for analyzing fiber cell gap junction distribution is illustrated in Figure 1 . Extended-focus images of Cx46 (Fig. 1A) and WGA (Fig. 1B) staining were generated by maximum projection of image stacks over a depth of 2 μm. Bi-level masks that distinguish between foreground (signal) and background (no signal) were calculated for each signal by an adaptive thresholding algorithm. Adaptive thresholding used the mean and SD of fluorescence in a 32 × 32-pixel region around a central point to provide a local threshold. The resultant bi-level images of Cx46 and WGA were merged to generate a new bi-level mask that traced the cell outlines (Fig. 1C) . This combined bi-level mask was “thinned” ³⁵ to produce a membrane “skeleton” in which the thickness of membranes was reduced to 1 pixel (Fig. 1D) . The membrane skeleton was used to calculate “masked” fluorescence images in which the fluorescence signal was only retained at pixel locations belonging to the skeleton (Fig. 1E) . To determine plaque width and number, the masked Cx46 image was converted to a bi-level image (Fig. 1F) . The membrane signal was also used to obtain measurements of local average cell size (cross-sectional area), average membrane length per cell cross section and membrane density per unit area. The size, number, and fragmentation of plaques were quantified by further analysis of the bi-level Cx46 mask (Fig. 1F) . Plaque width was also quantified by measuring the size of regions that did not contain WGA staining. Standard connected-component labeling techniques ³⁵ were used to count the number and determine the size distribution of plaques as a function of radial position. The fragmentation of plaques was characterized by counting the number of plaques per cell cross-section. 

Results

Mapping Cx46 Plaque Density and Fiber Cell Differentiation

Because fiber cells do not degrade but are retained throughout life, older fiber cells become internalized and the size of the lens increases with age. Thus, sections taken through a lens represent a “snapshot” of lens development and differentiation. Sections taken through the equator contain regular rows of fiber cells cut in cross section, whereas axial sections cut through the poles of the lens contain fiber cells cut in a longitudinal direction (Fig. 2) . In an axial section, epithelial cell nuclei can be seen to extend over the anterior surface of the lens. The differentiation of these cells into fiber cells occurs in the “bow region” at the equator of the lens where the fiber cell nuclei are concentrated. Fiber cell nuclei then disperse and become smaller with depth into the lens. This region corresponds to the zone where fiber cells lose other cellular organelles as well as nuclei. ²⁹  

Using fluorescently tagged antibodies and WGA, changes in Cx46 density were measured with increasing distance from the lens periphery and correlated to fiber cell differentiation (Fig. 3) . The distribution of Cx46 was probed with two antibodies, one of which was specific to the tail region of Cx46, whereas the other was targeted to an epitope located in the intracellular loop of Cx46. Biochemical experiments have established that lens gap junction connexin proteins undergo a differentiation-dependent processing that removes a portion of the cytoplasmic tail peptide residues. ^{10 23} To clarify spatial properties of this processing, we used the cytoplasmic tail antibody to detect uncleaved Cx46, whereas the antibody directed against the cytoplasmic loop was used to detect both cleaved and uncleaved forms of Cx46. Labeling obtained with the tail-specific Cx46 antibody was most intense at the lens periphery (where fiber cells are youngest), whereas lower levels of signal were present deeper in the lens (Fig. 3Aa) . In contrast, the signal from the loop-specific Cx46 antibody remained largely constant across the whole region that was imaged (Fig. 3Ba) . Quantitative analysis of the membrane labeling density of the two antibodies was performed (see the Methods section) to measure the radial (and therefore age) dependence of these changes. The Cx46 tail-specific labeling intensity appeared to form three discrete regions separated by two narrow transition zones (Fig. 3Ab) . In contrast, the Cx46 loop-specific signal density was approximately constant (Fig. 3Bb) across the lens, suggesting that the density of the membrane-embedded portion of Cx46 changed little. It is notable that the transition zones observed for the cytoplasmic tail antibody corresponded to areas where nuclei started to disperse and then disappeared (Fig. 3C) . To facilitate the comparison between antibody labeling and the changes in fiber cell nuclei, we normalized radial distance (r) from the center of the lens radius (a). Using this measure (r/a) we found that the initial dispersion of nuclei at r/a ∼ 0.9 correlated with the first decrease in Cx46 signal intensity, whereas the loss of nuclei at r/a ∼ 0.7 coincided with the secondary loss of signal intensity. Thus, it appears that two distinct cleavage events occur during fiber cell differentiation that remove the cytoplasmic tail of Cx46 while leaving the remainder of the protein embedded in the membrane. 

Plaque Remodeling

In addition to the changes in Cx46 signal density observed with the tail-specific antibody, radial changes in plaque size and distribution were visible. The image shown in Figure 4A is a dual-labeled (Cx46 and WGA) image consisting of overlapping high-resolution image stacks from which high-resolution two- and three-dimensional images from defined areas of interest were extracted. Near the periphery, the fiber cell shape was quite uniform in cross section, with large gap junction plaques in the broad sides of the cells and much smaller plaques located in the narrow sides (Fig. 4B) . In the region where cell nuclei disperse (r/a ∼ 0.8), the cell shape was less uniform and broad-side plaques were smaller and more fragmented (Fig. 4C) . At r/a ∼ 0.7 fiber cells had lost their distinctive hexagonal shape and had adopted a more circular cross section (Fig. 4D) . At this depth, small gap junction plaques appeared to be randomly dispersed around the entire cell periphery. 

To examine whether the observed changes in fiber cell morphology and subcellular distribution of gap junction plaques were related, quantitative morphometric analysis of these image data was performed. Semiautomated measurements of fiber cell cross-sectional area and perimeter from the combined Cx46 and WGA signals were used to calculate an ellipticity index (E = [4π × cross-sectional area/(cell perimeter)²]^1/2) to describe how fiber cell morphology changed with radial depth into the lens. The ellipticity index serves as a measure of “circularity” in the cell boundary, where values of 0 and 1 imply a line and a circle, respectively. ³⁶ Figure 5A shows that the ellipticity index increased smoothly from the lens periphery inward, reflecting a gradual change from the classic hexagonal peripheral fiber cell profile to a more circular profile in the center of the lens. It follows that the sudden changes in gap junction processing that we have observed do not correspond to a sudden change in cell surface morphology per se. 

To quantify the distribution of plaques across the cell surface as a function of radial distance, plaque width and the number of plaques per cell cross section were measured. Three independent methods were used to measure plaque width (Fig. 5B) . We considered that the WGA, which reveals the presence of glycosylated membrane proteins, should provide a “negative stain” for gap junction plaques, because the plaques should (1) effectively exclude other surface membrane proteins by a volume fraction effect, and (2) restrict the extracellular space within the plaque to inhibit WGA entry and labeling (of any glycosylated protein that may be present). This view was supported by the reciprocal staining patterns seen with Cx46 antibodies and WGA (Fig. 1) . Thus, by using both Cx46 and WGA labeling, we could measure gap junction morphologic changes across the lens and correlate them with changes in cell morphology. The validity of the Cx46 signals as a general measure of gap junction plaque size was then verified by analyzing the negative staining pattern in the WGA signal. The results of the WGA analysis were in good agreement with signals from both Cx46 antibodies which suggests that gap junction plaque morphology has been adequately captured with our methods. Regardless of method, we observed a rapid decline in average plaque width from a peak of approximately 2.0 μm at the lens periphery to approximately 0.9 μm, which then remained nearly constant with increasing depth into the lens (Fig. 5B) . To reconcile this measurement with the relative constancy of Cx46 loop antibody labeling across the lens (Fig. 3) , we also measured the number of plaques per cell cross section (Fig. 5C) . At the periphery, the number of plaques was low, reflecting the concentration of large plaques in the broad sides of the cells. The number of plaques then increased with depth, consistent with the visual impression that plaques appear to “fragment” over the region where nuclei disperse and disappear (Fig. 4) . Plaque width decreased more rapidly than plaque number increased in the first 200 μm (r/a 0.9), but both measures became more constant in the deeper lens (r/a 0.7). 

From these data, we suggest that the subcellular remodeling of plaques is associated with two independent processes: an initial decrease in the size of plaques confined to the broad sides of the fiber cells, followed by fragmentation (and dispersal) of the smaller plaques around the entire cell perimeter. A comparison of Figures 5A and 5C shows that the changes in ellipticity index and the number of plaques per cross section paralleled each other from r/a of 1.0 to ∼0.7. This shows that the loss of the hexagonal profile is closely associated with plaque dispersal, as indicated by the increase in the number of plaques per cross section. It also appears that, deeper in the lens (r/a < 0.7), fiber cells continue to become more circular in cross section although no further change in gap junction plaque dispersion takes place (Fig. 5C) . 

The process of plaque remodeling was explored further by examining frequency distributions of plaque widths at various distances into the lens (Fig. 6) . In the peripheral location (Fig. 6A) the distribution of plaque widths could be decomposed into two component distributions: an exponentially declining distribution comprising mainly small plaques (dashed line), and an approximately Gaussian distribution of larger plaques that reflected the presence of broad-side plaques in the periphery (dotted line). Deeper within the lens, the distribution of plaque widths was describable by the sum of an exponential and a Gaussian distribution (Fig. 6B) . However, the mean of the Gaussian component had decreased, indicating a reduction in the size of the broad-side plaques. Deeper still, the distribution of plaque widths was purely exponential, consistent with loss of a hexagonal morphology and large broad-side plaques (Fig. 6C) . Because the minimum nominal plaque width reported by our image analysis was equal to one pixel (or 0.2 μm), it is unclear whether the increasing dominance of the exponential component with depth into the lens represents an increase in the frequency of a different small-plaque class or simply the remodeling of large plaques into smaller plaques. However, the latter possibility seems more likely because the decreasing mean of the Gaussian component and increase in plaque number with little change in Cx46 protein would be consistent with larger plaques “breaking up.” In summary, these analyses suggest that during fiber cell differentiation, the large broad-side plaques found in the lens periphery are fragmented into (or replaced by) numerous smaller plaques or clusters of channels that become uniformly distributed around the membranes of deeper, older fiber cells. 

Functional Consequences of Plaque Remodeling

It has been suggested that the alignment of the large, broad-side, gap junction plaques in the lens periphery contributes to the outwardly directed solute flux that can be measured at the lens equator. ¹³ To investigate the functional consequences of the changing membrane location(s) of plaques seen in the present study, we performed TPFP ³³ on lenses loaded with the nonfluorescent compound CMNB-caged fluorescein to assess local gap junction coupling patterns. By applying TPFP to a point in the cytoplasm of a single fiber cell, a microscopic source of fluorescein was created within the selected cell (see the Methods section). The subsequent diffusion of the fluorescein away from this point of release to neighboring fiber cells was monitored with time by conventional confocal microscopy. By applying this approach to different areas within the lens, the local patterns of cell-to-cell coupling could be related to the local distribution of gap junction plaques (Fig. 7) . In peripheral fiber cells where large broad-side plaques predominate, dye diffusion was highly anisotropic and occurred primarily in a radial direction (within a given fiber cell column) with minimal coupling to cells in the circumferential direction (Figs. 7A 7C) . In contrast, at locations beyond the zone of nuclear loss (r/a < 0.7, where plaques were distributed more evenly around the fiber cell membrane) the pattern of fluorescein diffusion was approximately isotropic (Figs. 7B 7D) . Thus, the local pattern of intercellular coupling appeared to change from a radial direction in the lens periphery to a more uniform pattern in the deeper fiber cells. The observed correlation between the shift in local diffusion patterns and the developmental dispersal of plaques would be consistent with the changes in intercellular communication in the lens observed by others ¹² and is, we suggest, the consequence of differentiation-dependent remodeling of gap junction plaques. 

Discussion

In this study we quantified Cx46 expression at high spatial resolution across equatorial sections of the rat lens. Our analysis shows that Cx46 processing and gap junction plaque remodeling vary with position across the tissue. By using two-photon imaging combined with antibodies targeted against different parts of the Cx46 protein we showed that the terminal fragment of Cx46 is cleaved in two distinct locations within the lens. Analysis of gap junction plaque morphology showed that, with increasing depth into the lens, gap junction plaques become smaller and more dispersed around the cell periphery. The observed structural remodeling of gap junctions correlated with a change in the dye-coupling pattern of fiber cells, which was visualized by two-photon–excited, high-resolution flash photolysis of caged fluorescein. 

Previous biochemical studies have shown that both Cx46 ¹⁰ and Cx50 ^{23 24} are cleaved in the lens, the latter by the protease calpain. ²⁴ In the present study, the analysis of Cx46 signal density profiles obtained using cytoplasmic loop and tail antibodies revealed that Cx46 cleavage occurs at two distinct stages during fiber cell differentiation. The initial cleavage of Cx46 occurs at r/a ∼ 0.9 and coincides with the dispersal of fiber cell nuclei, whereas the second occurs at r/a ∼ 0.7 and is associated with the complete loss of fiber cell nuclei. Thus, although both Cx46 and Cx50 undergo a cleavage event at r/a ∼ 0.7 coincident with a loss of fiber cell nuclei, the major site of Cx46 cleavage occurs at r/a ∼ 0.9. Whether this earlier cleavage event is also mediated by calpain or another unknown protease is an open question. Of interest, it has recently been reported that in the chick lens the differentiation-dependent cleavage of Cx45.6 is mediated by caspase-3. ³⁶ Thus, in the rat lens it is possible that multiple proteases may be involved. In principle, the loss of labeling of the tail-specific antibody could also be due to epitope masking rather than cleavage. This possibility appears unlikely, because biochemical experiments confirmed that Cx46 cleavage occurs, and the concomitant loss of plaque localization with the loss of signal is compatible with a loss of anchoring due to tail cleavage (although it is not impossible that masking of the tail region by another protein could produce a similar effect). 

The functional significance of the two Cx46 cleavage zones is uncertain. One possibility is that Cx46 processing is related to changes in the pH sensitivity of fiber cell gap junctions. The reliance of fiber cells on anaerobic metabolism generates a radial pH gradient in the lens. ³⁷ Gap junction channels are normally uncoupled by a reduction in intracellular pH. Thus, the acidic environment observed in the lens core (pH ∼ 6.5) would be expected to disrupt gap-junction–mediated communication between the periphery and the center of the lens. However, impedance measurements conducted on whole lenses have shown that despite the presence of a pH gradient, all fiber cells in the lens core are electrically coupled, and furthermore that fiber cell gap junction channels abruptly lost their sensitivity to changes in pH at r/a 0.75. ¹² This observed loss of the pH sensitivity has been attributed to the cleavage of the fiber cell connexins, ³⁰ which removes the cytoplasmic tail portion of the pH gate. Comparison of the pH sensitivities of Cx46 and Cx50 expressed in HeLa cells ³⁸ showed that although both connexins have a similar pK_A, Cx50 was much more sensitive to changes in pH (Hill coefficient = 6.8) than Cx46 (Hill coefficient = 2.2). Furthermore, whereas cleavage dramatically reduced Cx50 pH sensitivity, ³⁰ it had only a mild effect on the pH sensitivity of Cx46 channels, with a small shift in pK_A from 6.8 to 6.56. ³⁸ The dominant effect of Cx50 on the pH sensitivity of fiber cell gap junctions is supported by experiments in Cx50 knockout mice, which appear to express pH-insensitive (Cx46) gap junction channels. ³⁹ From this discussion, it appears that most Cx46 cleavage occurs in a region separate from where pH sensitivity is lost and that, in any case, Cx46 cleavage would have little effect on pH sensitivity. Thus, the loss of pH sensitivity observed in the previous impedance measurements is not primarily due to Cx46 cleavage, but more likely to the cleavage of Cx50. It is possible that the secondary loss of Cx46 tail signal at r/a ∼ 0.7 that we observed is a consequence of the processing directed toward Cx50 to reduce pH sensitivity. In this case, the residual uncleaved Cx46 at r/a ∼ 0.7 would be modified as a “bystander” to Cx50 tail processing and would have no real impact on pH regulation, per se. 

It is possible that Cx46 cleavage serves to initiate plaque remodeling. At the lens periphery, large gap junction plaques of up to 6 μm in width are found on the broad sides of hexagonal fiber cells, with smaller plaques being located on the narrow sides. The well-defined geometry of opposite broad-side and narrow-side plaques is initially maintained as the processing of Cx46 proceeds. Eventually, plaque alignment is lost, and plaques disperse as they decrease in size. In parallel to the dispersion of plaques, we have shown that a progressive rounding of fiber cell cross sections occurs with increasing depth into the lens. Changes in fiber cell morphology are known to be associated with the age-dependent processing of cytoskeletal elements and/or membrane anchors. ^{40 41 42 43} Thus, it is possible that cleavage of the cytoplasmic tail of Cx46 removes the ability of gap junction plaques to interact with elements of the cytoskeletal network that constrain the plaques to specific membrane domains in the lens periphery. For example, the age-dependent proteolysis of the membrane attachment proteins spectrin and ankyrin in the chick lens has been proposed to ease constraints on membranes, thereby allowing a repositioning of integral membrane channels. ⁴¹ Furthermore, the scaffolding protein zonula occludens (ZO)-1, which has been linked to the function of adherens junctions and gap junctions, ⁴⁴ has recently been shown to bind to Cx46 and Cx50. ⁴⁵ This finding raises the possibility that ZO-1 may mediate interactions between gap junctions and the cytoskeleton of fiber cells and that untethered cleaved Cx46 channels would be free to redistribute themselves more evenly throughout the fiber cell membrane. 

The delivery of nutrients and removal of waste from a large avascular tissue such as the lens would be problematic if it relied solely on passive diffusion. The observation that ion fluxes are directed into the lens at the poles and out of the lens at the equator ^{46 47} prompted Mathias et al. ¹³ to propose that the lens relies on an internal microcirculation system. The patterns of local cell–cell coupling that we have described may help explain how this circulation can provide nutrient supply and waste removal to fiber cells in the lens core. The dominant structures at the poles are the sutures where fiber cells from adjacent hemispheres meet. ⁴⁸ It is important to note that sutures are radially oriented, and, if extracellular metabolite penetration into the deeper lens occurs predominantly through these structures, it would lead to gradients of metabolites in a circumferential direction that reflects suture geometry. To dissipate such gradients, circumferential transport between cells must be facilitated. In contrast, toward the lens equator (once all circumferential gradients are dissipated), all intercellular transport would be purely radial, and it is in this direction that gap junctions are best placed to facilitate movement of metabolites. Our demonstration of isotropic diffusion deep within the lens and radial diffusion toward the periphery (observations that are in good agreement with the measured distribution of gap junction protein) are consistent with this model for lens circulation. The potential problem of how the lens might dissipate circumferential gradients deeper within the lens can now be explained by the reorganization of gap junction plaques in the cell membrane to facilitate more homogeneous transport, whereas the outward radial flow at the periphery (where only radial gradients occur) is facilitated by concentrating the gap junction plaques in the fiber cell broad sides. In connection with the above points, it is interesting to note that as mature fiber cells have lost nuclei, they may have very limited ability to synthesize new membrane proteins and as a result have to rely on modification of existing proteins to match protein expression to function. 

In conclusion, our high-resolution quantitative mapping of the remodeling of gap junctions and processing of Cx46 has revealed isoform-specific changes in the subcellular location and processing of gap junction proteins that occur at precise times during fiber cell differentiation. In peripheral fiber cells, the full-length connexins form large gap junction plaques on discrete broad-side domains, which results in strong radial cell–cell coupling, consistent with the known efflux at the equator powered by transporters. ¹² Cx46 processing near the periphery may be directed toward conversion of radially directed intercellular transport to isotropic transport by permitting the dispersal of gap junction plaques around the fiber cell membranes. Still deeper in the lens, the pH sensitivity of gap junction plaques is reduced by further connexin processing to keep channels open in the acidic lens core and, while this process should be primarily directed against more pH-sensitive Cx50 isoforms, some further Cx46 processing also occurs. Thus, the structure of gap junctions is modified by precise connexin processing and gap junction plaque remodeling to create functional specializations in subregions of the organ, which allow the maintenance of lens circulation, homeostasis, and transparency. 

 Supported by the University of Auckland Research Committee; the Marsden Fund of New Zealand; and the Wellcome Trust, United Kingdom.

 Submitted for publication February 10, 2003; revised April 13 and June 3, 2003; accepted June 6, 2003.

 Disclosure: M.D. Jacobs, None; C. Soeller, None; A. Sisley, None; M.B. Cannell, None; P.J. Donaldson, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Summary of image-processing methods. The images shown illustrate a selected region (60 μm²) of the lens to which the image-processing methods for the analysis of gap junction and membrane features have been applied. Extended-focus images of (A) raw Cx46 signal and (B) raw WGA signal. (C) A bi-level mask formed by merging Cx46 and WGA distributions obtained after adaptive thresholding. (D) A membrane skeleton produced by thinning of image (C). This membrane skeleton was used to produce images (E) and (F). (E) Masked image of the Cx46 signal. (F) Masked bi-level image of the Cx46 signal used to extract data on the size and density of plaques. Scale bar, 5 μm.

Figure 2.

 

Lens morphology. Schematic diagram of lens morphology showing equatorial and axial sectioning planes and the dispersion of fiber cell nuclei. Directional axes indicated by arrows aid in orientation in large-scale expression studies (r, radial; c, circumferential; z, z-axial).

Figure 3.

 

Quantitative mapping of Cx46 signal density versus fiber cell differentiation. (A) Composite image of an equatorial radius of lens assembled from a set of four overlapping image stacks, showing (a) Cx46 tail-specific labeling and (b) the extracted normalized Cx46 tail-specific signal density. (B) Composite image showing (a) Cx46 loop-specific labeling and (b) the extracted normalized Cx46 loop-specific signal density. (C) Image of an axial section through a lens showing cell nuclei in the anterior epithelium and peripheral fiber cells. Fiber cell nuclei are first tightly aligned and then disperse before undergoing degradation. The scale represents radial distance (r) from the center of the lens plotted as a proportion of lens radius (a), so that r/a = 1 at the lens periphery and r/a = 0 at the lens center. Parameters extracted from the analysis of the images shown in (A) and (B) are plotted against distance and r/a. Zones where fiber cell nuclei are aligned or dispersed are represented by solid and dashed horizontal lines, respectively. Similar results were seen in sections from two other lenses for each label. Scale bars, 50 μm.

Figure 4.

 

Subcellular gap junction plaque morphology. (A) Overview of an equatorial radius of a lens assembled from a set of four overlapping image stacks, showing the r/a (relative radial distance) scale. Cx46 tail-specific (red) and WGA (green) signals are shown overlaid. (B–D) Representative images of deconvolved two-dimensional optical slices (left) and three dimensional isosurface projections (right), showing Cx46 (red) and WGA (green) labeling obtained at r/a of 0.95 (B), 0.8 (C), and 0.7 (D) acquired from the composite image shown in (A; arrowheads on x-axis). Arrows in (B–D) represent 1-μm scale bars and indicate the directional axes in the lens that were defined previously (Fig. 2) . For display purposes, the signal intensities have been adjusted and should not be taken as indicative of actual signal levels.

Figure 5.

 

Quantification of gap junction plaque remodeling. Physical measurements of fiber cell morphology and plaque structure were extracted from overview data shown in Figures 3 and 4 and plotted against radial distance. (A) Fiber cell ellipticity. (B) Plaque widths calculated from either the Cx46 tail-specific (•) or loop-specific (○) antibody signals and from the size of WGA-negative membrane regions (▴). (C) Plaque fragmentation. The number of discrete plaques labeled with Cx46 tail-specific (•) or loop-specific (○) antibody per cell over a 2-μm cell section. Solid and dashed lines: zones of aligned or dispersed bow region nuclei, respectively. Each data point in (A) was obtained by semiautomated analysis of 10 cells. For data points in (B) and (C) more than 100 cells were automatically analyzed.

Figure 6.

 

Size distributions of gap junction plaques. Frequency distributions of plaque widths were assessed at three depths into the lens corresponding to the three zones of Cx46 signal intensity shown in Figure 3 . (A) At r/a = 0.95, the distribution of plaque widths can be described by a sum of exponential (dashed line) and Gaussian (dotted line) distributions. (B) At r/a = 0.8, the mean of the Gaussian distribution decreased. (C) At r/a = 0.65, only the exponential distribution remained.

Figure 7.

 

Functional consequences of the redistribution of gap junction plaques. The images show the diffusion of the gap junction tracer fluorescein in the lens periphery (A, C) and inner lens (B, D; r/a < 0.7) after its release from a nonfluorescent cage by stationary spot TPFP. (A, B) A series of images showing the time course of fluorescein diffusion in the two areas of the lens, Δt = 10 seconds between images. Right: range of fluorescence intensity. (C, D) Fluorescein steady state diffusion patterns with superimposed fiber cell morphology that was visualized by WGA labeling of cell membranes. Similar photolysis data sets were recorded in three other lenses. Scale bars, 5 μm.