Localization of gap junctions and tracer coupling in retinal m�ller cells

Localization of Gap Junctions and TracerCoupling in Retinal Muller Cells

ALEXANDER K. BALL1* AND JOHN S. MCREYNOLDS2

1Department of Biomedical Sciences, McMaster University,Hamilton, Ontario L8N 3Z5, Canada

2Department of Physiology, University of Michigan, Ann Arbor, Michigan 48109–0622

ABSTRACTPhysiological studies have demonstrated the existence of direct intercellular communica-

tion, presumably mediated by gap junctions, both between neurons and between glial cells inthe vertebrate retina. We localized gap junctions in the retinas of rat, goldfish, and mudpuppyby using antisera directed against proteins that make up the connexon channels in two tissuesfrom which connexins have been isolated: liver (connexin 32; CX32) and heart (connexin 43;CX43). Although the antiserum against CX32 stained liver gap junctions, it did not reveal anystaining in rat or goldfish retina. The antiserum against CX43 stained gap junctionsassociated with the intercalated disk in rat heart and also stained gap junctions betweenpigment epithelium cells in rat, goldfish, and mudpuppy retina. Anti-CX43 also stained gapjunctions between Muller cells in goldfish and mudpuppy retina but not in rat retina.Intracellular injections of the tracer Neurobiotin into Muller cells in the mudpuppy retinarevealed that these glial cells are extensively tracer coupled. Staining with the tracer formed asyncytium of thin processes surrounding every neuron from the outer limiting membrane tothe inner limiting membrane. Confocal microscopy demonstrated that the Muller cells were inclose apposition with one another at every level of the retina. However, CX43 immunoreactiv-ity was heaviest at the outer limiting membrane, where the apical processes of Muller cells arelocated. Some anti-CX43 staining was observed at the level of the outer nuclear layer and theinner plexiform layer but not in the ganglion cell layer or at the Muller cell end feet formingthe inner limiting membrane. J. Comp. Neurol. 393:48–57, 1998. r 1998 Wiley-Liss, Inc.

Indexing terms: connexin-32; connexin-43; immunohistochemistry; Neurobiotin; goldfish;

mudpuppy

Gap junctions represent an important mode of intercellu-lar communication in the retina, permitting molecules aslarge as 1 kDa to pass from cell to cell and allowingmetabolic and/or electrical coupling between cells. In bothmammalian and nonmammalian retinas, there are manyexamples of electrical coupling between neurons of thesame type. For example, the existence of electrical cou-pling between photoreceptors (Baylor et al., 1971; Detwilerand Hodgkin, 1979; Gold, 1979; Gold and Dowling, 1979)and between horizontal cells (Kaneko and Stuart, 1984;Piccolino et al., 1984; Lasater and Dowling, 1985; DeVriesand Schwartz, 1989; Dong and McReynolds, 1991) is wellestablished in most species studied. Gap junctions and/orelectrical coupling have also been reported between bipo-lar cells (Witkovsky and Stell, 1973; Wong-Riley, 1974;Saito and Kujiraoka, 1982, 1988; Van Haesendonck andMissoten, 1983; Kujiraoka and Saito, 1986; Marc et al.,1988; Cuenca et al., 1993; Umino et al., 1994) and betweenamacrine cells (Wong-Riley, 1974; Famiglietti and Kolb,1975; Naka and Christenson, 1981; Teranishi et al., 1984;

Vaney, 1991; Hampson et al., 1992; Hidaka et al., 1993;Mills and Massey, 1995). The main type of glial cell in theretina is the Muller cell (Neuman and Reichenbach, 1996);in mammalian retinas, Muller cells do not appear to becoupled to each other by gap junctions (Wolburg et al.,1990; Hollander et al., 1991; Robinson et al., 1993). Thesituation may be different in lower vertebrates, however(Uga and Smelser, 1973). In amphibians, gap junctionshave been observed between the apical processes of Mullercells in mudpuppy (Miller and Dowling, 1970) and newt(Tonus and Dickson, 1979). Muller cells have been shownto be electrically coupled in turtle retina (Conner et al.,

Grant sponsor: National Institutes of Health; Grant numbers: EY01653and EY07003; Grant sponsor: NSERC; Grant number: OGP0171190.

*Correspondence to: Alexander K. Ball, Department of Biomedical Sci-ences, McMaster University, 1200 Main Street West, Hamilton, OntarioL8N 3Z5, Canada. E-mail: [email protected]

Received 8 July 1997; Revised 3 November 1997; Accepted 4 November1997

THE JOURNAL OF COMPARATIVE NEUROLOGY 393:48–57 (1998)

r 1998 WILEY-LISS, INC.

1985) and both electrically coupled and dye-coupled inaxolotl retina (Mobbs et al., 1989).

The existence of gap junctions at specific locations hasalso been determined by transmission or freeze-fractureelectron microscopy and, more recently, by using antibod-ies directed against the connexin proteins, which form thegap junctions. Connexin-43 (CX43), originally found inheart muscle, and connexin-32 (CX32), originally found inliver, are both found in neurons and glia (Dermietzel andSpray, 1993). Despite the abundant anatomical and electro-physiological evidence for gap-junction coupling betweenneurons and between Muller cells in the retina, the specificconnexin proteins of which these gap junctions are consti-tuted have been identified only recently (Janssen-Bien-hold et al., 1996; Wagner et al., 1996; Giblin and Chris-tensen, 1997).

The present study was undertaken to determine whetherthe gap junctions that are known to exist in the retina arecomposed of either CX32 or CX43. Immunocytochemicalexperiments on retinas from rat, goldfish, and mudpuppydemonstrated no immunoreactivity to CX32. On the otherhand, immunoreactivity to CX43 was found between pig-ment epithelium cells in all three species but not betweenneurons in any retina. In goldfish and mudpuppy retina,but not in rat retina, CX43 immunoreactivity was alsofound between Muller cells. Intracellular injections of thesmall tracer Neurobiotin into Muller cells was used toestablish that these cells were coupled. Confocal micros-copy demonstrated that Neurobiotin spread readily be-tween Muller cells in mudpuppy retina. Even though thecoupled Muller cells were in close apposition with eachother at several levels in the retina, CX43-like immunore-activity was found in highest density nearest the outerlimiting membrane.

MATERIALS AND METHODS

Immunohistochemistry

Adult Wistar rats (200–220 g) were killed by exposure toCO2 in an enclosed chamber. The eyes were enucleated, thecornea and lens were removed, and the resulting eye cupswere placed in fixative consisting of 4% paraformaldehyde,0.1 M Sorensen’s phosphate buffer (0.1 M dibasic sodiumphosphate and 0.1 M monobasic potassium phosphate), pH7.3, 1% sucrose, and 0.2 mM CaCl2. The thoracic andabdominal cavities were opened, and small, 0.5 3 0.5 mmsamples of myocardium from the heart and parenchymafrom liver were excised and placed in the same fixative.Eye cups and heart samples were fixed for 2–4 hours atroom temperature and then washed for 2 hours in a rinsebuffer consisting of 0.1 M sodium phosphate buffer, pH 7.3,3% sucrose, and 0.2 mM CaCl2. Eye cups and heartsamples were cryoprotected in 20% sucrose in rinse bufferovernight, mounted in OCT compound (Tissue-Tek; SakuraFinetechnical USA Inc., Torrence, CA), and 10-µm sectionswere cut on a cryostat. Sections cut perpendicular to theretinal layers were picked up on APTEX-treated slides (no.A-3648; Sigma, St. Louis, MO) and dried for at least 2hours at room temperature.

Goldfish (Carassius auratus) 4–6 inches in length ormudpuppies (Necturus maculosus) 9–12 inches in lengthwere killed by decapitation, pithed, and the eyes wereenucleated and hemisected. The lens was removed, andthe resulting eye cups were placed in fixative, processed,and sectioned as described above.

Slides containing sections were washed in phosphate-buffered saline (PBS), pH 7.3, to remove OCT compound,and sections were incubated overnight in a humid cham-ber in rabbit anti-CX43 antibody (polyclonal; lots 18A/12and 16A/12; gift from Dr. Elliot Hertzberg) or in rabbitanti-CX32 antibody (monoclonal; lot 72F; gift from Dr.Elliot Hertzberg), 1:500–1:2,000 in rinse buffer containing1% normal goat serum (Sigma), 0.2% Triton X-100, and1% dimethylsulphoxide (DMSO). Sections of mudpuppyretina were also incubated in rabbit antiglycine antibody(Q45725R; BioDesign, Kennebunkport, ME; 1:500 inrinse buffer containing 1% normal goat serum, 0.2%Triton X-100, and 1% DMSO) as described for connexinantisera.

Slides were washed in PBS and incubated in secondarygoat anti-rabbit serum conjugated to Texas red (BioCan,Mississauga, Ontario, Canada; 1:100 in rinse buffer con-taining 0.2% Triton X-100 and 1% DMSO). Slides werethen washed in PBS and coverslipped in VectaShieldmounting medium (H-1000; Dimension Laboratories, Mis-

Fig. 1. Confocal photomicrographs of rat myocardium showingconnexin 43 (CX43)-like immunoreactivity. a: Low-power photomicro-graph showing staining associated with myocyte intercalated disks(arrows). b: Thin optical section at higher magnification reveals thatthe staining was located in punctate loci (arrowheads). Scale bars 5 50µm in a, 10 µm in b.

GAP JUNCTIONS IN RETINA 49

sissauga, Ontario, Canada). Slides were viewed with aZeiss LSM-10 confocal microscope (Thornwood, NY) byusing a 543-nm HeNe laser, and digitized photomicro-graphs were printed with a Mitsubishi CP-110U printer(Yokohama, Japan). Digitized photomicrographs wereprinted in negative (black on white) for clarity with thisprinting technique.

Neurobiotin tracer coupling

Mudpuppy eye cups were prepared as described above.The vitreous was removed with filter paper, and the eyecup was placed in a chamber and superfused with Ringer’ssolution consisting of 110 mM NaCl, 2.5 mM KCl, 1.8 mMCaCl2, 11 mM glucose, and 5 mM HEPES buffer, adjusted

Fig. 2. Photomicrographs showing CX43-like immunoreactivity ingoldfish retina. a: Toluidine blue-stained section of goldfish retinashowing retinal layers observed in vertical sections. Cone innersegments (cis) are located immediately above the outer limitingmembrane (olm) formed by the apical processes of Muller cells. ris, rodinner segments; opl, outer plexiform layer; ipl, inner plexiform layer.b: Confocal photomicrographs of goldfish neural retina stained withantisera directed against gap junction connexin CX43. CX43-like

immunoreactivity was most intense at the olm (large arrowhead).Sparse labeling was also observed between the olm and the opl (smallarrowheads). c: Fluorescence photomicrograph of oblique sections ofthe goldfish retina reveals staining at the olm between ris and cis.Staining at the apical processes of Muller cells surrounded photorecep-tor cells (large arrowheads) passing through the olm. Punctatestaining at the apical processes of Muller cells was frequently paired(small arrowheads). Scale bars 5 25 µm in a and b.

50 A.K. BALL AND J.S. MCREYNOLDS

to pH 7.8, with 10 N NaOH. Intracellular recordings weremade by using electrodes pulled on a Brown-Flamingpuller (Sutter Instrument Co., Novato, CA) and by usingconventional electronics. The electrodes were filled with0.15 M Neurobiotin (SP-1120; Vector Laboratories, Burlin-game, CA; FW 322.8) in 0.01 M potassium acetate or 0.01M KCl and had resistances of 500–1,000 MV. Muller cellswere tentatively identified by their very negative restingpotentials (260 to 280 mV) and their lack of any responseto light. Although Miller and Dowling (1970) recordeddepolarizing responses to light from identified Muller cellsin mudpuppy, in the present study, no light responses wereseen in the cells, which were subsequently identified asMuller cells. The absence of light responses in Muller cellsin our recordings may have been due to the relatively dimlight stimuli, which were used to keep the retinas darkadapted for other reasons. Tracer was injected into thecells by passing 10–15 nA, 200-msec depolarizing pulses at3.3 Hz for 2–5 minutes. The current was monitoredcontinuously during the injection.

After injection of the tracer, eye cups were placed in thefixative described above. The retinas were then removedfrom the eye cups and kept flat under a Teflon gauzeduring fixation for 1–2 hours. Then, they were washed inrinse buffer for 2 hours and placed in Streptavidin-Texasred (016-070-084; BioCan; 1:500 in rinse buffer containing0.2% Triton X-100 and 1% DMSO) overnight on a rotator.Retinas were again washed in rinse buffer for 20 minutesand flatmounted in VectaShield for subsequent examina-tion by conventional fluorescence microscopy and confocalmicroscopy, as described above.

Animals were treated according to the guidelines of theCanadian Council on Animal Care and the Handbook forthe Use of Animals in Biomedical Research published bythe Association for Research in Vision and Ophthalmology(1993).

RESULTS

Immunohistochemistry

Antisera directed against CX32 stained gap junctionsbetween rat liver hepatocytes, but no staining was ob-served in either rat, goldfish, or mudpuppy retina at anydilution used. Antisera directed against CX43 stained gapjunctions between cardiac muscle fibers (Fig. 1) at alldilutions used. Staining was typical of gap-junction label-ing previously described for rodent cardiac tissue (Luqueet al., 1994). Low-power confocal photomicrographs of ratmyocardium showed extensive CX43-like immunoreactiv-ity associated with intercalated disks (Fig. 1a). At highermagnification, thin optical sections revealed that thestaining was located in punctate loci measuring 1.5–2.0µm in diameter (Fig. 1b).

Similar punctate staining with CX43 antisera was seenin the retinas of rat, goldfish, and mudpuppy. Figures 2and 3 show results from goldfish retinas. Toluidine blue-stained, frozen sections (Fig. 2a) were used to identifyretinal layers and to facilitate the location of immunoreac-tive staining in antibody-treated retinas. Cone inner seg-ments were visible immediately distal to the outer limitingmembrane formed by the apical processes of Muller cells(Fig. 2a). The rod inner segments were located between theouter limiting membrane and the outer plexiform layer(Fig. 2a). Confocal photomicrographs (Fig. 2b) showed thatCX43-like immunoreactivity was most intense at the outerlimiting membrane (Fig. 2b, large arrowhead). Additional

Fig. 3. High-magnification confocal photomicrographs showingCX43-like immunoreactivity in goldfish pigment epithelium and neu-ral retina. a: In the retinal pigment epithelium (rpe), immunoreactiv-ity was localized to the irregular junctions between cells (arrows). n,Pigment epithelium cell nucleus; ros, rod outer segments (in thesubretinal space). b: In the neural retina, staining was limited to theouter retina. Vertical sections revealed punctate staining at the outerlimiting membrane (olm) and between the olm and the outer plexiformlayer (opl; same magnification as Fig. 3c). c: Composite overlay of tenhalf-micrometer confocal sections demonstrates the high density ofstained loci between the olm and the opl. Scale bars 5 25 µm in a,10 µm in c.


staining was observed in the region between the outerlimiting membrane and the outer plexiform layer. Stainingwas never observed at any other locations in the neuralretina. Fluorescence photomicrographs of oblique sectionsof the retina (Fig. 2c) revealed staining at the outerlimiting membrane between rod inner segments and coneinner segments. Staining at the apical processes of Mullercells (Fig. 2c, large arrowheads) surrounded photoreceptorcells passing through the outer limiting membrane. Punc-tate staining at the apical processes was frequently paired(Fig. 2c, small arrowheads), suggestive of multiple gap-junction connections between the thin sheets of Mullercells surrounding photoreceptors at their apical ends.

High-magnification confocal microscopy revealed thelocation of CX43 staining in the distal goldfish retina (Fig.3a–c). Individual punctae of immunoreactivity were ob-served at the junction between pigment epithelial cells(Fig. 3a, arrows). These loci measured 0.5–1.5 µm indiameter. Similar staining was observed at the outerlimiting membrane (Fig. 3b) and between cells at the outernuclear layer, where rod inner segments were located (Fig.3b). A composite overlay of ten 0.5-µm-thick confocaloptical sections, 1 µm apart, demonstrated the density ofthe punctate staining located between the outer limitingmembrane and the outer plexiform layer (Fig. 3c).

In the rat retina, CX43-like immunoreactivity was ob-served between pigment epithelial cells but not in theneural retina (Fig. 4a). In oblique sections (Fig. 4b), thestaining in the pigment epithelium was observed as punc-tate staining surrounding each pigment epithelial cell.

In the mudpuppy retina, CX43-like immunoreactivitywas observed between pigment epithelial cells as well as

between cells in the neural retina. Confocal microscopyrevealed that the punctate immunoreactivity was concen-trated most heavily at the outer limiting membrane (Fig.5a, large arrowhead). Immunoreactivity was also locatedbetween the outer limiting membrane and the outerplexiform layer but to a lesser extent than at the outerlimiting membrane. This immunoreactivity was observedalong the columns of darkly stained Muller cells extendinglongitudinally through the retina from the outer plexiformlayer to the inner plexiform layer (Fig. 5a, small arrow-heads). A composite overlay of ten 0.5-µm-thick confocaloptical sections, 1 µm apart, clearly demonstrated the highdensity of immunoreactivity located at the outer limitingmembrane (Fig. 5b). Staining was never observed proxi-mal to the inner plexiform layer. High-magnification confo-cal photomicrographs of immunoreactive staining at theouter limiting membrane showed a dense aggregation ofpunctate deposits in the space between photoreceptors(Fig. 5c, arrowheads). Individual loci of staining measur-ing 1–2 µm were visible between photoreceptors and at theouter plexiform layer (Fig. 5c, arrows).

Tracer coupling of Muller cells

Injections of Neurobiotin into goldfish or rat retinal cellsrarely resulted in extensive tracer spread between Mullercells. This is consistent with the observations of Zahs andNewman (1997), who found that Neurobiotin did notspread between rat retinal Muller cells. By comparison,injections of Neurobiotin into mudpuppy retinal cellsusually resulted in Muller cell labeling, even when neu-rons were the intended target of study. On the occasions

Fig. 4. CX43-like immunoreactivity in rat retina. a: Confocalphotomicrograph of a vertical section showing staining between cellsin the pigment epithelium (large arrowhead) but not in the neuralretina. b: In an oblique section through the pigment epithelium layer,

immunoreactivity is visible surrounding pigment epithelial cells (ar-rowheads). n, retinal pigment epithelium cell (rpe) nucleus (arrow);olm, outer limiting membrane; opl, outer plexiform layer; ipl, innerplexiform layer.


when more than one Muller cell was labeled after injectingrat or goldfish retina, we assumed that this was due todamage and leakage of Neurobiotin from the micropipette,because the pattern of staining was unlike the extensiveand symmetrically radial labeling pattern observed inmudpuppy. To study this phenomenon further, we injectedeight mudpuppy Muller cells, each in a different retina,and the dye was allowed to diffuse for from 5 minutes to 1hour prior to fixation. In all cases, Neurobiotin diffused to

other Muller cells over 250 µm from the injection site. Suchextensive labeling was observed in confocal photomicro-graphs of flatmounted retinas (Fig. 6a). The phi-Z capabil-ity of the confocal microscope was used to optically sectionthrough the flatmount and to reconstruct a vertical sectionthrough the tracer-coupled regions (Fig. 6b). These recon-structions demonstrated that the coupling was exclusive toMuller cells. The apical processes of Muller cells (Fig. 6b,small arrow), cell bodies (Fig. 6b, large arrowhead), andend feet (Fig. 6b, small arrowheads) were revealed byusing this technique. Small processes of the tracer-filledcells filled the inner plexiform layer and outlined thesomata of amacrine cells (Fig. 6b) and ganglion cells (Fig.6b). The morphology of the these reconstructed, tracer-filled cells was identical to that of mudpuppy Muller cellsspecifically stained with antisera directed against glycine(Fig. 6c), which labels both glycinergic amacrine cells (Fig.6c) and glia (Fig. 6c, large arrowhead). Confocal micros-copy was used to characterize the Muller cell syncytium atthe outer limiting membrane, inner nuclear layer, innerplexiform layer, and optic nerve fiber layer. Equivalentlevels are indicated on the vertical section through theretina (Fig. 6c, levels a–d).

At the level of the apical processes (Fig. 7a), stainedprocesses were located at the outer limiting membrane,separating photoreceptors (Fig. 7a, asterisk). Thin sheetsof cytoplasm labelled with Neurobiotin (Fig. 7a, arrows)extended from the apical processes of Muller cells, separat-ing photoreceptor inner segments and isolating them fromtheir neighbors. At the level of the Muller cell bodies (Fig.7b, arrowheads), stained cells were found among thesomata of bipolar and amacrine cells (Fig. 7b, asterisk).Approximately 40% of somata in the middle of the innernuclear layer were labelled with Neurobiotin, indicating ahigh glia/neuron ratio in this retina. Thin sheets of cyto-plasm extended from Muller cell bodies, outlining theneurons found at this level (Fig. 7b, arrows). At the level ofsublamina a of the inner plexiform layer (Fig. 7c), Neuro-biotin filled an array of cytoplasmic columns, each measur-ing 6–8 µm in diameter (Fig. 7c, arrowheads). Thesecolumns extended through the extent of the inner plexi-form layer. Very small processes radiated from the col-umns, extending laterally in the plexiform layer, andisolated the neuronal elements of the neuropil (Fig. 7c,asterisk). At the level of the optic nerve fiber layer and theinner limiting membrane (Fig. 7d), the columns of theinner plexiform layer expanded to form the end feet (Fig.7d, arrowheads) observed in vertical sections. Like the

Fig. 5. Confocal photomicrographs showing CX43-like immunore-activity in mudpuppy retina. a: In low-magnification confocal photomi-crographs of vertical sections showing the entire thickness of themudpuppy retina, immunoreactive staining was concentrated at theouter limiting membrane (olm; large arrowhead). Staining was alsoobserved between the olm and the outer plexiform layer (opl; mediumarrowhead) and along columns of darkly stained Muller cells extend-ing longitudinally through the retina (small arrowheads). b: Compos-ite overlay of ten half-micrometer confocal sections at the samemagnification as in a demonstrating that the highest density of CX43immunoreactivity is located at the olm. Staining was never observedproximal to the inner plexiform layer (ipl). c: Higher magnificationconfocal photomicrograph showing punctate immunoreactivity be-tween Muller cells at the olm (arrowheads). The dense staining at theapical processes of Muller cells was interrupted by photoreceptorsextending through the olm. Individual loci of immunoreactive stainingwere located along the cytoplasm of Muller cells between photorecep-tors (arrows). cn, Cone nucleus. Scale bars 5 50 µm in a, 25 µm in c.


other horizontal optical sections, thin sheets of cytoplasm(Fig. 7d, small arrows) extended from the end feet toseparate ganglion cell and displaced amacrine cell somata(Fig. 7d, asterisk). The axons of ganglion cells, whichformed fascicles of optic nerve fibers (Fig. 7d, large arrowmarked onf), were also outlined by the Neurobiotin tracer.These observations demonstrate that Muller cells were inclose apposition through thin sheets of cytoplasm at every

level of the retina but that gap junctions, revealed byCX43-like immunoreactivity, are located in the outer retina,mostly at the outer limiting membrane.

DISCUSSION

The antibodies we used against CX32 did not stain anycells in the retina or pigment epithelium of any of the threespecies studied, whereas antibodies to CX43 stained pig-ment epithelial cells in all three species and stainedMuller cells in goldfish and mudpuppy but not in rat. Bothtypes of antisera stained cells in nonretinal tissues knownto contain these types of connexins.

Although the CX43-like immunoreactivity suggests thatgap junctions between pigment epithelial cells in all threespecies use CX43, none of the gap junctions betweenneurons appeared to use this form of connexin. Among gapjunctions between glial cells in the central nervous system,CX32 is found in oligodendrocytes, and CX43 is found inastrocytes (Dermietzel and Spray, 1993). The finding ofCX43-like immunoreactivity in Muller cells in mudpuppyand goldfish retinas is consistent with their being consid-ered a type of astrocyte (the vascularized retinas ofmammals also contain other astrocytes, which, togetherwith Muller cells, form a larger class termed macroglia;Hollander et al., 1991). In cat, gap junctions have beenfound on astrocytes but not on Muller cells, except whenthey were grown in culture (Wolburg et al., 1990; Hol-lander et al., 1991). In rabbit, the tracer molecules Luciferyellow and biocytin injected into astrocytes spread toMuller cells, but there was no spread in the oppositedirection from Muller cells to astrocytes and no spreadbetween Muller cells (Robinson et al., 1993). Similarresults were seen after Neurobiotin was injected into ratretinal glial cells (Zahs and Newman, 1997). In rat retina,the spread of calcium waves through Muller cells appearsto be via spread through the astrocyte syncytium, withsecondary spread from astrocytes into Muller cells but notdirectly between Muller cells (Newman and Zahs, 1997).

Fig. 6. Photomicrographs showing the spread of Neurobiotin be-tween Muller cells in mudpuppy retina. Neurobiotin was injected intoa single Muller cell. a: Low-power confocal photomicrograph of aretinal flatmount shows spread of tracer to other Muller cells over 250µm from the injection site. The level of focus is at the outer limitingmembrane (olm). b: The phi-Z scanning capability of the confocalmicroscope was used to reconstruct a vertical section through theretina to reveal that the spread of Neurobiotin occurred through theMuller cell syncitium. In the top half of b, the level of focus was nearthe middle of the inner nuclear layer (DZ 5 78.95 µm). The optical Zsection was taken at the position of the horizontal line (top half). Thedimensions of the X, Y, and Z planes are equal (Z:XY 5 1). Theresulting vertical section from the phi-Z scan is shown in the lower halfof b. The same cells visible in the flatmount (upper half) are visible inthe vertical section (lower half, arrowheads). Apical processes ofMuller cells (small arrow), cell bodies (large arrowhead), and end feetof Muller cells (small arrowheads) are visible. Neuronal elements,including amacrine cell (ac) and ganglion cell (g) somata, are un-stained. c: Confocal photomicrograph of mudpuppy retina showingglycine immunoreactivity in Muller cells (large arrowhead) and ama-crine cells (ac; open arrow). The vertical Z scan through the Neurobio-tin-filled cell network (bottom) was comparable to Muller cell labelingwith glycine antisera. Apical processes of Muller cells (arrow) at theolm were labeled as well as their end feet in the inner retina (mediumarrowheads) surrounding ganglion cell somata (g). Small arrowheads(a–d) indicate approximate levels where subsequent confocal photomi-crographs of Neurobiotin spread in flatmounts were made (see Fig. 7).Scale bars 5 100 µm in a, 50 µm in b.


The demonstration that tracer spread occurs from astro-cyte to Muller cell but not in the opposite directionsuggests the presence of an atypical rectifying gap junctionin rat retina (Zahs and Newman, 1997). Our resultsindicate that these are not composed of CX32 or CX43, orare composed of some other form of these connexins.Although Muller cells are unlikely to be involved inbuffering ions to any great extent in the rat retina, thespread of Neurobiotin tracer between mudpuppy Mullercells and their CX43 immunoreactivity suggests that thismay be an important role for Muller cells in this species.

The localization of CX43 immunoreactivity mainly atthe apical ends of the Muller cells in goldfish and mud-puppy retinas was similar, but the spread of Neurobiotinbetween mudpuppy Muller cells appears to be more exten-sive than between goldfish Muller cells. The reason for

such a difference is unknown, but it may reflect differencesin the light- and dark-adaptive coupling state in the twospecies.

A striking finding was that neither CX32 nor CX43,which are the major types of connexin found making upgap junctions in the brain, was found in any neurons in theretina, even though many of these cells have been shown tobe coupled by gap junctions. These results suggest that oneor more types of connexin other than CX32 or CX43, whichthe antibodies we used were directed against, make upmost of the gap junctions between retinal neurons. Prelimi-nary reports showing that variants of CX43 can be isolatedfrom fish retina (Janssen-Bienhold et al., 1996; Lu et al.,1996; Giblin and Christensen, 1997) suggest that multipleforms of CX43 exist. It is therefore possible that suchvariants of CX43 and CX32 exist between neurons in the

Fig. 7. Confocal optical horizontal sections of Neurobiotin-filledMuller cells at four different levels in flatmounted mudpuppy retinaindicated by a–d in Figure 6c. Potential cellular contact betweenMuller cells is seen at each level of the retina. a: At the level of theouter limiting membrane (olm), stained apical processes of Mullercells (arrowheads) were in close apposition through thin sheets ofcytoplasm (arrows) separating unlabeled photoreceptors (asterisk).b: At the level of the Muller cell bodies, stained cells were found amongthe somata of bipolar and amacrine cells (asterisk). Muller cell bodieswere in close apposition through thin sheets of cytoplasm thatseparated neurons (arrows). c: At the level of sublamina a of the inner

plexiform layer (ipl), Neurobiotin stained arrays of cytoplasmic col-umns that projected through the ipl (arrowheads). Small processes(arrows) radiated from these columns, isolating the neuronal elementsof the neuropil (asterisks). d: At the level of the inner limitingmembrane and the optic nerve fiber layer, the columns seen in the iplexpanded to form end feet (arrowheads). Thin sheets of cytoplasmextended from the end feet (small arrows) to separate the somata ofganglion cells and displaced amacrine cells (asterisk). The axons ofganglion cells forming fascicles of optic nerve fiber (onf; large arrows)were also outlined by Neurobiotin. Scale bars 5 50 µm.

55A.K. BALL AND J.S. MCREYNOLDS

goldfish and mudpuppy retina and could explain whythese connnexins were not revealed by our immunostain-ing.

When coupled Muller cells in mudpuppy retina werefilled with Neurobiotin, the tracer showed that the coupledcells were in close apposition with each other at severallevels in the retina. Confocal microscopy demonstratedthat the cells were in close apposition in the outer nuclearlayer, in the inner nuclear layer, and at the inner limitingmembrane. However, because CX43-like immunoreactiv-ity was found only in the outer retina, mainly at the level ofthe outer limiting membrane, this suggests that eitherthere are no gap junctions between Muller cells at otherlevels, where they were in close apposition with each other,or that gap junctions between Muller cells at other levels ofthe retina are composed of some other type of connexin or avariant of CX43. In another urodele amphibian, the axo-lotl, the spread of current between end feet of Muller cellswas measured (Mobbs et al., 1988). The calculated spaceconstant for the Muller cell network was 15 µM when itwas assumed that the gap junctions between Muller cellswere mainly at the level of the end feet. The present datasuggest that the gap junctions between Muller cells are atthe other end of the cells, near the outer limiting mem-brane. The physiological significance of the clustering ofgap junctions at the apical end of the Muller cells andwhether this would have any effect on the calculated spaceconstant remain unclear.

The tracer-filled Muller cells reveal a network of glialprocesses, demonstrating that the neurons in mudpuppyretina are completely surrounded by the processes of thesecells. Lateral extensions from the Muller cells also ap-peared to surround, and perhaps compartmentalize, neu-ral processes in the inner plexiform layer. Indeed, onefunction of Muller cells may be to insulate the somata ofganglion cells and decrease the extracellular conductancearound them. Alternatively, the lateral processes of Mullercells in the plexiform layers may serve to increase theirsurface area to enhance siphoning of K1 and the uptake ofneurotransmitters, such as glutamate. It has been sug-gested that, for uniform illumination, changes in extracel-lular K1 produced by neuronal activity will be uniformacross the retina (although they will vary with the depth ofthe retina), and no spatial buffering of K1 ions of one cell tothe other will occur. For localized illumination, howeverthere will be a lateral gradient of [K1]o across the retina,and spatial buffering will occur through the electricalsynapses between Muller cells (Mobbs et al., 1988).

In summary, we have demonstrated extensive tracercoupling between Muller cells that appears to be mediatedby a CX43-like connexin. The gap junctions formed bythese connexins are located mainly in the distal retina.Given the morphological and electrophysiological abun-dance of evidence for the existence of gap-junction connec-tions between retinal cell types, it is surprising that noCX43- or CX32-like immunoreactivity in neurons was seenwith the antibodies used in our experiments.

Localization of different connexins to glia and differentneuronal subtypes may reveal a pattern of specificityrelated to functional differences, such as the modulation ofgap-junction resistance between retinal cells. Furtherstudies using different antibodies will be necessary toidentify the connexins that make up the known gapjunctions between the retinal neurons and to correlate

different connexin types with functional differences in gapjunctions.

ACKNOWLEDGMENTS

The authors thank Dr. Elliot Hertzberg for the kind giftof the connexin antisera used in this study and Drs. BobGarfield and Wm. Baldridge for their help with using theantisera. This work was supported by NIH grants EY01653and EY07003 to J.S.M. and by NSERC grant OGP0171190to A.K.B.

LITERATURE CITED

Baylor, D.A., M.G.F. Fuortes, and P.M. O’Bryan (1971) Receptive field ofcones in the retina of the turtle. J. Physiol. (London) 214:265–294.

Conner, J.D., P.B. Detwiler, and P.V. Sarthy (1985) Ionic and electrophysi-ological properties of retinal Muller (glial) cells of the turtle. J. Physiol.(London) 362:79–82.

Cuenca, N., E. Fernandez, M. Garcia, and J. DeJuan (1993) Dendrites ofrod dominant ON-bipolar cells are coupled by gap junctions in carpretina. Neurosci. Lett. 162:34–38.

Dermietzel, R. and D. Spray (1993) Gap junctions in the brain: Where, whattype, how many and why? Trends Neurosci. 16:186–192.

Detwiler, P.B. and A.L. Hodgkin (1979) Electrical coupling between cones inturtle retina. J. Physiol. (London) 291:75–100.

DeVries, S. and E.A. Schwartz (1989) Modulation of an electrical synapsebetween solitary pairs of catfish horizontal cells by dopamine andsecond messengers. J. Physiol. (London) 414:351–375.

Dong, C.-J. and J.S. McReynolds (1991) The relationship between light,dopamine release and horizontal cell coupling in the mudpuppy retina.J. Physiol. (London) 440:291–309.

Famiglietti, E.V. and H. Kolb (1975) A bistratified amacrine cell andsynaptic circuitry in the inner plexiform layer of the retina. Brain Res.84:293–300.

Giblin, L.J. and B.N. Christensen (1997) Connexin 43 immunoreactivity inthe catfish retina. Brain Res. 755:146–150.

Gold, G.H. (1979) Photoreceptor coupling in the retina of the toad Bufomarinus. II. Physiology. J. Neurophysiol. 42:311–328.

Gold, G.H. and J.E. Dowling (1979) Photoreceptor coupling in the retina ofthe toad Bufo marinus. I. Anatomy. J. Neurophysiol. 42:292–310.

Hampson, E.C.G.M., D.I. Vaney, and R. Weiler (1992) Dopaminergicmodulation of gap junction permeability between amacrine cells inmammalian retina. J. Neurosci. 12:4911–4922.

Hidaka, S., M. Maehara, O. Umino, Y. Lu, and Y. Hashimoto (1993) Lateralgap junction connections between retinal amacrine cells summatingsustained response. Neuroreport 5:29–32.

Hollander, H., F. Makarov, Z. Dreher, D. van Driel, T. Chan-Ling, and J.Stone (1991) Structure of the macroglia of the retina: Sharing anddivision of labour between astrocytes and Muller cells. J. Comp. Neurol.313:587–603.

Janssen-Bienhold, U., R. Dermietzel, and R. Weiler (1996) Distribution ofconnexin-43 immunoreactivity in the retinas of different species. Soc.Neurosci. Abstr. 22:882.

Kaneko, A. and A.E. Stuart (1984) Coupling between horizontal cells in thecarp retina revealed by diffusion of Lucifer yellow. Neurosci. Lett.47:1–7.

Kujiraoka, T. and T. Saito (1986) Electrical coupling between bipolar cells incarp retina. Proc. Natl. Acad. Sci. USA 83:4063–4066.

Lasater, E.M. and J.E. Dowling (1985) Dopamine decreases conductance ofthe electrical junctions between cultured horizontal cells. Proc. Natl.Acad. Sci. USA 82:3025–3029.

Lu, C., D.G. McMahon, and D. Speck (1996) Partial sequence of a gapjunction from the hybrid striped bass retina. Soc. Neurosci. Abstr.22:883.

Luque, E.A., R.D. Veenstra, E.C. Beyer, and L.F. Lemanski (1994) Localiza-tion and distribution of gap junctions in normal and cardiomyopathichamster heart. J. Morphol. 222:203–213.

Marc, R.E., W.-L.S. Liu, and J.F. Muller (1988) Gap junctions in the innerplexiform layer of the goldfish retina. Vision Res. 28:9–24.

Miller, R.F. and J.E. Dowling (1970) Intracellular responses of the Muller(glial) cells of the mudpuppy retina: Their relation to b-wave of theelectroretinogram. J. Neurophysiol. 33:323–341.


Mills, S.L. and S.C. Massey (1995) Differential properties of two gapjunctional pathways made by AII amacrine cells. Nature 377:734–737.

Mobbs, P., H. Brew, and D. Attwell (1988) A quantitative analysis of glialcell coupling in the retina of the axolotl (Ambystoma mexicanum). BrainRes. 460:235–245.

Naka, K.-I. and B.N. Christensen (1981) Direct electrical connectionsbetween transient amacrine cells in the catfish retina. Science 274:462–464.

Newman, E. and A. Reichenbach (1996) The Muller cell: A functionalelement of the retina. TINS 19:307–311.

Newman, E. and K.R. Zahs (1997) Calcium waves in retinal glial cells.Science 275:844–847.

Piccolino, M., J. Neyton, and H.M. Gerschenfeld (1984) Decrease of gapjunction permeability induced by dopamine and cyclic adenosine 38:58-monophosphate in horizontal cells of turtle retina. J. Neurosci. 4:2477–2488.

Robinson, S.R., E.C.G.M. Hampson, M.N. Munro, and D.I. Vaney (1993)Unidirectional coupling of gap junctions between neuroglia. Science262:1072–1074.

Saito, T. and T. Kujiraoka (1982) Physiological and morphological identifica-tion of two types of on center bipolar cells in the carp retina. J. Comp.Neurol. 205:161–170.

Saito, T. and T. Kujiraoka (1988) Characteristics of bipolar-bipolar couplingin the carp retina. J. Gen. Physiol. 91:275–287.

Teranishi, T., K. Negishi, and S. Kato (1984) Dye coupling betweenamacrine cells in carp retina. Neurosci. Lett. 51:73–78.

Tonus, J.G. and D.H. Dixon (1979) Neuro-glial relationships at the externallimiting membrane of the newt retina. Exp. Eye Res. 28:93–110.

Uga, S. and G.K. Smelzer (1973) Comparative study of the fine structure ofretinal Muller cells in various vertebrates. Invest. Ophthalmol. 12:434–448.

Umino, O., M. Maehara, S. Hidaka, S. Kita, and Y. Hashimoto (1994) Thenetwork properties of bipolar-bipolar cell coupling in the retina ofteleost fishes. Visual Neurosci. 11:533–548.

Van Haesendonck, E. and L. Missotten (1983) Interbipolar contacts in thedorsal inner plexiform layer in the retina of Callionymus lyra L. J.Ultrastruct. Res. 83:303-311.

Vaney, D.I. (1991) Many diverse types of retinal neurons show tracercoupling when injected with biocytin of Neurobiotin. Neurosci. Lett.125:187–190.

Wagner, T.L.E., D.G. McMahon, and D.K. Vaughan (1996) Cloning of a gapjunction gene from the retina of the giant Danio. Soc. Neurosci. Abstr.22:883.

Witkovsky, P. and W.K. Stell (1973) Retinal structure in the smooth dogfishMustelus canis: Electron microscopy of serially sectioned bipolar cellsynaptic terminals. J. Comp. Neurol. 150:147–168.

Wolburg, H., W. Reichelt, J.-U. Stolzenburg, W. Richter, and A. Reichenbach(1990) Rabbit retinal Muller cells in cell culture show gap and tightjunctions which they do not express in situ. Neurosci. Lett. 111:58–63.

Wong-Riley, M.T.T. (1974) Synaptic organization of the interplexiform layerin the retina of tiger salamander. J. Neurocytol. 3:1–33.

Zahs, E.R. and E.A. Newman (1997) Asymmetric gap junctional couplingbetween glial cells in the rat retina. Glia 20:10–22.


Documents

Localization of gap junctions and tracer coupling in retinal m�ller cells