INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

THE EYE IS a water-transporting organ. The corneal endothelium expels water into the anterior chamber, thereby maintaining transparency and counteracting the swelling tendencies of passive water uptake through corneal epithelium and endothelium (11, 15). Movement of water from the ciliary stroma across the epithelium, within the trabecular meshwork, and across the endothelial wall into Schlemm's canal is essential for the secretion and reabsorption of aqueous humor (1, 32). In addition to the major drainage route provided by Schlemm's canal, some aqueous humor leaves the eye by the uveoscleral route (1). The iris has a high water permeability, which may facilitate rapid changes in shape during pupillary constriction, and the iris also plays an important role in changing the composition of the aqueous humor by diffusion across its surface. The retinal pigment epithelium transports water toward the choroid, thereby enhancing retinal adhesion (18, 36). Proper balance of ions and water between the cytoplasm and the extracellular space between the lens fiber cells helps maintain lens clarity. In addition, the functions of each cell type are dependent on regulation of cell volume.

The driving forces and the pathways for epithelial water transport are not completely understood (39). Molecular understandings were advanced by the identification of aquaporin water channel proteins in the plasma membranes of multiple epithelia, and at least six different mammalian aquaporins (AQP) are now recognized. AQP1 is highly expressed in several epithelia, including kidney proximal tubules and descending thin limbs and capillary endothelium (26); in the eye, AQP1 is present in corneal endothelium, iris, and ciliary and lens epithelia (13, 25, 34). AQP2 (8) is the vasopressin-regulated water channel and is restricted to the kidney collecting duct (23, 31). AQP3 is expressed in the kidney collecting duct and other organs including the conjunctival epithelium (5-7). AQP4 is predominantly expressed in the brain (14, 16), where it is localized in glial cell processes at multiple sites, including the cerebellum and supraoptic nuclei (24), as well as ependymal cells lining the ventricles (24). AQP4 is also distributed outside the central nervous system (7). The cDNA encoding AQP5 was isolated from salivary glands; in situ hybridization and Northern blotting revealed expression of the transcript in salivary glands, corneal epithelium, and lung (29). Major intrinsic protein of the lens (MIP) is designated AQP0 and is abundantly expressed in lens fiber cells (37). This protein has recently been shown to be a molecular water channel (21, 38). Recently, an extensive study using RT-PCR demonstrated AQP1, AQP3, and AQP4 mRNAs in multiple dissected rat ocular epithelia (27). AQP1 mRNA was found in almost all ocular tissues, including the retina, and AQP4 was found in the retina and ciliary body (27).

Although aquaporin transcripts and immunoreactivities have been detected in various ocular tissues, a comprehensive analysis of eye tissues for each member of the aquaporin family of proteins has not been reported at the protein level. Here we utilize immunoblotting, immunocytochemistry, and immunoelectron microscopy of intact human and rat tissues to define the cellular and subcellular distribution of aquaporins (AQP1-AQP5) in the eye.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Tissue preparation. Human eyes were provided by the Department of Eye Pathology, University of Copenhagen, and the Eye Department of Rigshospitalet and were fixed in 4% paraformaldehyde in 0.1 M sodium cacodylate buffer, pH 7.2. Male Wistar rats weighing 200-300 g (Møllegard Breeding Centre, Eiby, Denmark) were allowed free access to food and water. Rat eyes, lacrimal glands, sclera, and conjunctiva were fixed by cardiac perfusion with 4% paraformaldehyde in 0.1 M sodium cacodylate buffer, pH 7.2. Tissue blocks, including small blocks of human and rat eyes trimmed from the cornea, ciliary epithelia, iris, and retina, were further fixed by immersion for 2 h in 4% paraformaldehyde in 0.1 M sodium cacodylate buffer, pH 7.2, infiltrated for 30 min with 2.3 M sucrose containing 2% paraformaldehyde, mounted on holders, and rapidly frozen in liquid nitrogen. Tissues for immunoblotting were obtained after cardiac perfusion with PBS.

Electrophoresis and immunoblotting. Tissue from rat eyes or cerebellum were minced finely and homogenized in dissecting buffer (0.3 M sucrose, 25 mM imidazole, 1 mM EDTA, pH 7.2, with 8.5 mM leupeptin and 1 mM phenylmethylsulfonyl fluoride) with five strokes of a motor-driven Potter-Elvehjem homogenizer at 1,250 rpm. The homogenates were centrifuged in a Beckman L8M centrifuge at 4,000 g for 15 min at 4°C, and the supernatants were centrifuged at 200,000 g for 1 h. The high-speed pellets were resuspended in dissecting buffer and assayed for protein concentrations by the method of Lowry.

Immunofluorescence and immunoperoxidase cytochemistry. The procedures have been described previously (23, 26). For light microscopy, paraffin sections of whole human eyes (2-4 µm thick prepared using standard techniques) or cryosections of different ocular epithelia obtained with a Reichert-Jung cryoultramicrotome (0.85 µm thick) were used. Cryosections or paraffin sections were placed on gelatin-coated glass slides. After preincubation with PBS containing 1% BSA or 0.1% skim milk and 0.05 M glycine, the sections were incubated with anti-aquaporin antibodies. The use of the affinity-purified antibodies against aquaporins for immunocytochemistry has previously been described (referenced above for each antibody): anti-AQP1 (0.1-0.2 µg IgG/ml), anti-AQP2 (0.1-0.8 µg/ml), anti-AQP3 (0.5 µg/ml), anti-AQP4 (1-5 µg/ml), and anti-AQP5 (5-20 µg/ml). The labeling was visualized by use of peroxidase-conjugated secondary antibody (P-448, Dako; 1:100). Sections were counterstained with Meier reagent (an unspecific nuclear stain). For fluorescence microscopy, the sections were incubated with anti-AQP1 antibodies, washed, incubated with goat anti-rabbit conjugated to FITC (Z-205, Dako; 1:40), and finally washed and mounted with Glycergel, as previously described (5, 35). The specimens were studied in a Leitz Laborlux S fluorescence microscope.

Immunoelectron microscopy. Tissue blocks were subjected to freeze substitution and embedded in Lowicryl HM-20 by use of procedures previously described (24). The frozen samples were freeze substituted in a Reichert auto freeze-substitution unit (Reichert, Vienna, Austria). Samples were sequentially equilibrated over 3 days in 0.5% uranyl acetate in methanol at temperatures gradually increasing from 80 to 70°C, rinsed in pure methanol for 24 h at 70 to 45°C, and infiltrated at 45°C with Lowicryl HM-20 and methanol at 1:1, 2:1, and finally pure HM-20 (1 day in each solution) before ultraviolet polymerization in pure HM-20 for 2 days at 45°C and 2 days at 0°C. Immunolabeling was performed on ultrathin sections (40-60 nm) incubated with affinity-purified aquaporin antibodies (described above) and visualized with goat-anti-rabbit IgG conjugated to 10-nm colloidal gold particles (GAR.EM10, BioCell Research Laboratories, Cardiff, UK). Sections were stained for 10 min with uranyl acetate and examined in an electron microscope (model CM100, Philips).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Immunoblotting of AQP1, AQP4, and AQP5 in membranes from the rat eye. Specificity of our antibodies for rat ocular tissues was confirmed by immunoblotting isolated membranes. AQP1 immunolabeling is abundant in membranes from the cornea and ciliary body. At this exposure no signals were obtained in membrane fractions from the retina (Fig. 1A), although higher exposures revealed a weak signal (not shown). Strong AQP4 immunolabeling is apparent in membranes from the retina, with the 29-kDa band migrating at the same level as in control membranes from rat cerebellum (Fig. 1B), a site with abundant AQP4 expression (7, 16, 24). A weak AQP4 signal is also observed in membranes from the ciliary body. AQP5 immunolabeling is strong in membranes prepared from the cornea (Fig. 1C), whereas none appeared in membranes from the ciliary body or retina. Controls prepared with nonimmune IgG revealed no labeling (Fig. 1D). On the basis of these immunoblots, immunocytochemical and immunoelectron-microscopic analyses were performed on the three layers of rat and human eye: fibrous tunica, uvea, and retina.

View larger version (54K): [in this window] [in a new window]   Fig. 1.   Immunoblots of membrane fractions isolated from various rat ocular tissues (20 µg membrane protein/lane). A: immunoblot reacted with affinity-purified anti-aquaporin 1 (anti-AQP1). Antibody selectively recognizes 28-kDa nonglycosylated and 35- to 50-kDa glycosylated moieties of AQP1 in membrane fractions of cornea (cor) and ciliary body (cil). B: immunoblot reacted with affinity-purified anti-AQP4; 29-kDa band and higher-molecular-weight band are seen in membranes from cerebellum (cer; control tissue) and in retina (ret). C: immunoblot reacted with affinity-purified anti-AQP5; 28-kDa AQP5 band is seen exclusively in membrane fractions from cornea. D: immunolabeling of control, where nonimmune IgG used as primary antibody revealed no labeling.

Immunolocalization of AQP1, AQP3, and AQP5 in fibrous tunica and external tissues. Corneal keratocytes react strongly with anti-AQP1 (Fig. 2, a and b), whereas no AQP1 was observed in the corneal epithelium in the rat eye (Fig. 2a), and controls confirmed the lack of nonspecific labeling (Fig. 2a, inset). The apical and basolateral membranes of the corneal endothelium are also strongly labeled with anti-AQP1 in the rat (Fig. 2b) and human (Fig. 2b, inset) eye. The corneal epithelium exhibited only modest labeling with anti-AQP3 (Fig. 2c) and none with anti-AQP4 (Fig. 2d), whereas significant AQP5 immunolabeling is present (Fig. 2e). When probed with anti-AQP3, anti-AQP4, or anti-AQP5, corneal keratocytes failed to label (Fig. 2, c-e).

View larger version (114K): [in this window] [in a new window]   Fig. 2.   Immunocytochemical localization of aquaporins in fibrous tunica (a-i) of rat eye and rat lacrimal gland (j). Original magnification ×430. a-e are from cornea, f and g from conjunctiva, h from limbal area, and i from sclera. In a, arrows indicate AQP1-immunolabeled keratocytes. Corneal epithelium (EP) did not display AQP1 immunolabeling. Inset: immunolabeling control. b: AQP1 immunolabeling of keratocytes (arrows) and corneal endothelium (EN; arrowheads). Inset: immunofluorescence microscopy showing AQP1 immunolabeling of keratocytes and endothelium of human cornea. In c, plasma membranes of corneal epithelial cells exhibit weak AQP3 immunolabeling (arrows), with no labeling of corneal stroma (). In d, corneal epithelium exhibits no AQP4 immunolabeling. , Stroma. In e, plasma membranes of corneal epithelium exhibit abundant AQP5 immunolabeling (arrows). , Stroma. Inset: AQP5 immunolabeling of corneal epithelium at low magnification. In f, sections of conjunctival epithelium exhibit extensive AQP3 immunolabeling, with no labeling of underlying stroma (CS). In g, no expression of AQP4 is detected in conjunctival epithelium or stroma. In h, limbal area contains abundant AQP1 in scleral cells (S), ciliary epithelia, epithelia of iris (I), and aqueous outflow apparatus; endothelium of trabecular meshwork (T), including uveal meshwork (U); and endothelium in Schlemm's canal (). In contrast, there is no AQP1 immunostaining of stroma of ciliary body (CB) or conjunctival epithelium. In i, scleral fibroblasts exhibit significant AQP1 immunolabeling (arrows). In j, lacrimal gland acini exhibit AQP5 immunolabeling of apical plasma membrane (arrows), but not basolateral membranes (arrowheads).

View larger version (127K): [in this window] [in a new window]   Fig. 3.   Immunoelectron micrograph of AQP1 in a scleral stroma cell. Fibroblast (F) shows abundant AQP1 immunolabeling over plasma membranes (arrows). C, closely packed fibrils of collagen. Original magnification ×80,000.

Immunolocalization of AQP1 and AQP4 in the uvea. Immunocytochemistry of the pars plicata of the human ciliary body reveals strong AQP1 immunolabeling of the internal nonpigmented epithelium, including basal infoldings and lateral plasma membranes (Fig. 4a), whereas controls showed no specific labeling (Fig. 4b). Although it is clear that there is no labeling of the basolateral portion of the ciliary pigmented epithelial cells, melanized pigment granules in the apical portion of the ciliary pigmented epithelial cells prevent unambiguous interpretation of immunoperoxidase labeling of the apical membranes. To resolve this issue, immunofluorescence microscopy was performed revealing anti-AQP1 labeling of the apical nonpigmented epithelial plasma membranes (Fig. 4, c and d). Although autofluorescence signals were emitted by small granules in the nonpigmented epithelial cells, no AQP1 immunolabeling was seen in the cytoplasm (Fig. 4, b and c). In contrast to the strong immunolabeling of the internal nonpigmented epithelium, anti-AQP1 labeling was not observed in the heavily pigmented external epithelium (CPE, Fig. 4, a, c, and d). The ciliary stroma was unlabeled, but the endothelium of capillaries and postcapillary venules in the stroma exhibit strong AQP1 immunolabeling (Fig. 4d). Immunoelectron microscopy of ciliary nonpigmented epithelium in the pars plicata confirmed that AQP1 immunolabeling is abundant in basolateral plasma membranes (Fig. 5a) as well as in apical plasma membranes facing the unlabeled external ciliary epithelium (Fig. 5b).

View larger version (133K): [in this window] [in a new window]   Fig. 4.   Immunocytochemical localization of AQP1 in uveal tract. a-d are from human ciliary body, e from human iris, and f from rat choroid. Original magnification ×430. In a, paraffin section of ciliary body shows AQP1 immunolabeling of basolateral membranes (arrows) of ciliary nonpigmented epithelium (NPE), but no labeling of external ciliary pigmented epithelium (CPE). In b, parallel to a, immunolabeling control shows no labeling. Brown spots represent melanin granules. In c, immunofluorescence microscopy of paraffin section demonstrates AQP1 immunolabeling of basolateral and apical NPE plasma membranes (arrows) facing unlabeled CPE (arrowhead). In d, immunofluorescence labeling of a semithin cryosection shows labeling of plasma membranes of NPE cells (arrows). Endothelial cells of vessels (V) show significant labeling. Autofluorescence signals appear over small granules within NPE cells. In e, a section from iris reveals significant AQP1 immunolabeling (arrows) of posterior pigmented epithelium (PE) as well as anterior epithelium (AE). In f, sections of choroid show AQP1 immunolabeling of choriocapillary endothelium (arrows) with no staining of retinal pigment epithelium (RPE) or rod-and-cone layer (RCL). Inset: immunolabeling control, where nonimmune IgG was used as primary antibody.

View larger version (127K): [in this window] [in a new window]   Fig. 5.   Immunoelectron micrograph of AQP1 in pars plicata of rat ciliary NPE. In a, extensive AQP1 immunolabeling appears over basolateral plasma membranes (arrows). ILM, inner limiting membrane; M, mitochondrion; N, nucleus. In b, extensive AQP1 immunolabeling of apical plasma membranes extends along cytoplasmic processes (P) projecting from apical surface into intercellular spaces (ICS). Original magnification ×85,000.

View larger version (73K): [in this window] [in a new window]   Fig. 6.   Immunocytochemical localization of AQP4 in cryostat sections of immersion-fixed rat ciliary body (from rat strain with pigment in ciliary pigmented epithelium). Original magnification ×190. In a, cryostat section of ciliary body shows AQP4 immunolabeling of basolateral and apical membranes (large arrows) of ciliary NPE, but no labeling of CPE (small arrows). Large open arrow, initiation of retina. b: Parallel view using interference microscopy. c: Immunolabeling control performed without primary antibody on a cryostat section from ciliary body. Background labeling (produced by secondary antibody) is seen in deeper portions of ciliary body, whereas no staining is found in ciliary epithelium, confirming selective AQP4 labeling demonstrated in a. Arrows as described in a.

Immunolocalization of AQP4 in the retina.¹ Immunoperoxidase cytochemistry in paraffin sections revealed no anti-AQP1 labeling of cells in the layers in the neuronal part of rat retina (layers 2-10) or in the retinal pigment epithelium (Fig. 7a, layer 1) but distinct labeling of scleral cells. However, use of thin cryosections revealed very low AQP1 labeling in the outer nuclear layer (Fig. 7a, inset), presumably of glial elements. The labeling was weak, and it was not possible with use of standard techniques to obtain significant labeling at the electron-microscopic level (not shown). In contrast, immunocytochemistry revealed very abundant AQP4 labeling of Müller cells and astrocytes (Fig. 7, b-d) extending from the outer limiting membrane (Fig. 7c, layer 3) throughout the retina, as demonstrated by immunoperoxidase (Fig. 7, layers 3-7) and by immunoelectron microscopy (Figs. 8 and 9). Also, glial processes in layers 8 and 9 exhibit similar anti-AQP4 labeling (not shown). Higher magnification reveals strong AQP4 immunolabeling of Müller cell processes surrounding capillaries (Fig. 7d, arrows) and weaker but significant labeling of glial processes surrounding neurons (Fig. 7, b and d, arrowheads). Very prominent labeling is seen in the inner plexiform layer (layer 7); significant immunolabeling is also seen in the inner nuclear layer (layer 6), where Müller cell bodies reside. All AQP4 immunolabeling controls were negative (Fig. 7c, inset). In addition, anti-AQP4 failed to label retinal pigment epithelium or the rod-and-cone layer (Fig. 7, b and c, layers 1 and 2), and no anti-AQP3 or anti-AQP5 labeling of the retina was observed (not shown). Immunoelectron microcopy reveals abundant AQP4 immunolabeling by plasma membranes of Müller cell processes in the inner limiting membrane (lower magnification, Fig. 8a; higher magnification, Fig. 8, b and c). Also, perivascular Müller cell processes exhibit extensive anti-AQP4 labeling, whereas retinal capillary endothelium and neurons showed no labeling (Fig. 9). Thin cryosections of human retina labeled with affinity-purified anti-human AQP4 revealed a similar labeling pattern, with specific labeling of astroglial cells (data not shown).

View larger version (138K): [in this window] [in a new window]   Fig. 7.   Immunocytochemical localization of aquaporins in retina of rat eye (a-d) and human lens (e). Original magnification ×430 unless otherwise noted. Retinal layers are as follows: retinal pigment epithelium (1), rod-and-cone layer (2), outer limiting membrane (3), outer nuclear layer (4), outer plexiform layer (5), inner nuclear layer (6), inner plexiform layer (7), ganglion cell layer (8), nerve fiber layer (9), and inner limiting membrane (10). a: Paraffin section of rat retina. AQP1 immunolabeling is present in choroid (CH) and in sclera (S) but weak or absent in layers of retina. Inset: AQP1 immunolabeling using thin cryosections. In b, anti-AQP4 shows extensive labeling of Müller cells (large arrows and arrowheads). C, capillary. AQP4-labeled astrocytes (arrowhead) are abundant in inner plexiform layer and ganglion cell layer. AQP4 immunostaining of Müller cells commences in outer limiting membrane at apexes and continues internally with staining in outer nuclear layer and outer plexiform layer. c: Distinct AQP4 immunolabeling of Müller cells (arrows) in outer limiting membrane, outer nuclear layer, and outer plexiform layer. Inset: immunolabeling control. In d, a higher magnification of part of b shows distinct AQP4 immunolabeling of Müller cells in outer plexiform layer, consistent with known existence of many branching processes of Müller cells in this layer. Capillaries (C) are seen in layers 5-7 surrounded by heavily AQP4-stained perivascular Müller cell processes (small arrows). Large arrows, Müller cells; arrowheads, astrocytes. Note striking light-microscopic difference between these 2 types of retinal glia. Original magnification ×860. e: Cryosection (0.85 µm thick) of human lens. AQP1 labeling is abundant in apical and basolateral membranes of lens epithelium (EP). No AQP1 labeling is seen in lens capsule (CA) or lens fiber cells (LC).

View larger version (128K): [in this window] [in a new window]   Fig. 8.   Immunoelectron micrographs of AQP4 in Müller cells in inner rat retina. a: 2 adjacent Müller cells, their basement membrane (the inner limiting membrane), and vitreous. Original magnification ×80,000. b and c: higher magnifications of a. Original magnification ×110,000. In a-c, extensive AQP4 immunogold labeling (arrows) of Müller cell (M) plasma membranes is shown. Basement membrane (B) is closely related to fibrils in vitreous (V).

View larger version (128K): [in this window] [in a new window]   Fig. 9.   Immunoelectron micrographs showing AQP4 immunolabeling of retinal perivascular glia (perivascular Müller cells). a: Thin glial process with dense AQP4 staining of plasma membrane. Original magnification ×80,000. b: Thicker portion of a glial cell process with prominent AQP4 staining. Original magnification ×80,000. c: Higher magnification showing association of immunogold particles with plasma membrane of glial process. Original magnification ×100,000. G, perivascular glial cell; P, pericyte; E, endothelial cell; L, capillary lumen.

Immunolocalization of AQP1 in the lens. Anterior lens epithelium was strongly labeled by anti-AQP1. Apical and basolateral plasma membrane domains are immunolabeled, but lens fiber cells and lens capsule did not label (Fig. 7e). No labeling of the lens was produced with the antibodies specific for AQP3, AQP4, or AQP5 (not shown).

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

In this study we report the cellular and subcellular localization of multiple members of the aquaporin family of membrane water channels in ocular tissues (Fig. 10). We provide novel information about the distinct expression patterns for each aquaporin, including the presence of AQP4 in Müller cells and astrocytes with a polarized distribution, the expression of AQP4 in ciliary nonpigmented epithelium, the presence of AQP5 and AQP3 in corneal epithelium, the presence of AQP3 in basolateral membranes of conjunctival epithelium, and the presence of AQP5 in apical domains of lacrimal gland epithelium; moreover, AQP1 was shown to be abundant in scleral cells in addition to the known broad expression in multiple ocular tissues. Furthermore, we defined the absence or low expression of aquaporin proteins in certain tissues. The studies were performed in rat and human tissues at light- and electron-microscopic levels. Together with previous data, we conclude that each of these aquaporins is expressed at distinct sites in the eye, predicting specific roles at each site.

View larger version (29K): [in this window] [in a new window]   Fig. 10.   Sites of aquaporin distribution in eye. MIP, major intrinsic protein of lens.

AQP3 was not found in intraocular epithelia but is very abundant in basolateral plasma membranes of conjunctival epithelium, and AQP5 is abundant in the apical domain of lacrimal acinus cells and in corneal epithelial cells. The participation of AQP3 and AQP5 in tear formation and external moistening is suspected. Naturally occurring mutations in genes encoding AQP3 and AQP5 are not known. AQP1 is present in corneal endothelium and corneal keratocytes, consistent with a role in maintenance of corneal hydration, which is critical for corneal transparency and refraction.

AQP1 is present in the nonpigmented ciliary and iris epithelium, suggesting a role in aqueous humor secretion. AQP1 is also present in the endothelium covering Schlemm's canal, scleral stromal cells, and capillary endothelium, tissues that are involved in aqueous humor drainage. AQP1 in the anterior epithelium of the lens suggests participation in maintaining optical refraction of the lens. Although humans with knockout mutations in AQP1 do not show a severe clinical phenotype, their extreme rarity suggests that they may bear a compensating mutation (28). In contrast, MIP is known to be an aquaporin (AQP0), and naturally occurring mouse mutations result in congenital cataracts (33).

Abundant expression of AQP4 is noted in glial elements in the retina, including Müller cells, with extensive expression in processes facing the capillary endothelium or the inner limiting membrane facing the vitreous body. Thus the AQP4 expression pattern in the retina is very similar to that in the brain (24). Although the role of AQP4 in glial cells in the retina and brain remains undefined, it may be speculated that AQP4 participates in volume regulation in response to potassium siphoning during synaptic transmission. We have no clue from genetics, since AQP4 mutants are not yet known. Very low levels of AQP1 labeling were also found in nuclear layers of the retina. These observations are consistent with a previous report by Patil et al. (27), who found significant levels of AQP4 and AQP1 mRNA using RT-PCR of dissected retina. The cells expressing low levels of AQP1 are presumably glial cells, since AQP1 has consistently been found to be absent from neurons elsewhere (13, 24, 25).

Aquaporins in the cornea. The abundant expression of AQP1 in corneal endothelium (Fig. 2b) indicates a role of AQP1 in the water transport necessary to prevent swelling and thereby maintain transparency. The corneal endothelium (but not the corneal epithelium) is well established to be essential for maintaining corneal transparency (4, 15, 20). Active mechanisms are thought to involve Na⁺-K⁺-ATPase and a bicarbonate-dependent Mg²⁺-ATPase (reviewed in Ref. 15). Inhibition of Na⁺-K⁺-ATPase activity with ouabain is known to produce marked swelling of the cornea (9), suggesting that the enzyme is responsible for an outward transport of salt from the cornea into the aqueous humor. Nevertheless, the Na⁺-K⁺-ATPase has been localized to the basolateral plasma membranes facing the stroma, not the anterior chamber (19). If this is the predominant ouabain-inhibitable ATPase, then the pumping direction appears to be the reverse of the orientation found in the kidney proximal tubule. Thus it remains to be established whether the functional inhibition with ouabain is caused by this basolateral Na⁺-K⁺-ATPase or by another not yet defined ouabain-inhibitable transporter. The water transport secondary to the gradient established by active ouabain-sensitive salt transport is likely to take place via AQP1, which is abundant in apical and basolateral plasma membranes.

Aquaporins in the ciliary epithelium. As demonstrated in Figs. 4 and 5, AQP1 is abundantly expressed in the internal nonpigmented epithelium (Fig. 4) of ciliary tissue. In contrast, no labeling of the heavily pigmented external epithelium was observed (Fig. 4, a, c, and d). Immunoelectron microscopy confirmed the abundant labeling of apical and basolateral plasma membranes of the nonpigmented epithelium (Fig. 5). A weak expression of AQP4 was also noted (Fig. 6), indicating that AQP4 may also participate in conjunction with AQP1. The ciliary body has a major function in production of aqueous humor (32). The two layers of epithelia are apposed apex to apex, and the pigmented epithelium represents the forward continuation of the retinal pigment epithelium, whereas the nonpigmented epithelium represents a forward continuation of the neural retina. Although evidence suggests that the nonpigmented epithelium (which also expresses AQP1) is largely responsible for aqueous humor production, the double-layered epithelium may very well act as a functional unit (for a comprehensive review see Ref. 32). The ciliary epithelium is also the site for enzymes that are known to participate intimately in the rate of aqueous humor production, including Na⁺-K⁺-ATPase, adenylyl cyclase (in regulating aqueous humor production), and carbonic anhydrase (32). Notably the nonpigmented epithelium is the site for the major fraction of Na⁺-K⁺-ATPase activity, as demonstrated biochemically after separation of bovine pigmented and nonpigmented epithelia in density gradients (30). Immunocytochemistry has also shown abundant labeling of ₁-, ₂-, and ₃-subunits of the Na⁺-K⁺-ATPase in the basolateral plasma membranes of ciliary nonpigmented epithelial cells, whereas only the ₁-subunit was found in the pigmented ciliary epithelial cells (10). Importantly, there appears to be a close inverse correlation between aqueous humor production rates and ouabain inhibition of Na⁺-K⁺-ATPase activity (2). Thus a major driving force for Na⁺ transport, and hence water transport, is likely to be established mainly in the nonpigmented epithelium. The presence of AQP1 in the nonpigmented epithelium is therefore consistent with a distinct role of AQP1 in aqueous humor production. The expression pattern shares similarities with that in the kidney proximal tubule, with abundant AQP1 in apical and basolateral plasma membranes and abundant Na⁺-K⁺-ATPase in basolateral plasma membranes (31). Both epithelia are responsible for near-isosmotic water reabsorption, with release of water at the basolateral aspect of the cells. Whether AQP1 is essential for aqueous humor production or just provides an energetically favorable pathway for water transport is unknown and awaits experiments in mice with disruption of the Aqp1 gene. Very recently, Aqp1 knockout mice were produced, which in response to thirsting demonstrated a critical role of AQP1 in the kidney for urinary concentration (A. S. Verkman et al., unpublished observations). Potential defects in other organs including the eye await further studies.

Lack of AQP1, AQP3, AQP4, and AQP5 expression in the retinal pigment epithelium. The retinal pigment epithelium is responsible for bulk water transport out of the eye into the choroidal blood circulation. This epithelium does not express any of the mammalian aquaporins cloned so far. Although a noncharacterized aquaporin may be expressed at this site, other routes for water transport are likely to exist. Recently, Loo et al. (17) demonstrated that cotransporters, such as the sodium-coupled glucose transporter, may carry water along with their primary substrates. Thus, in addition to aquaporin-mediated water transport, other routes may exist in the retinal pigment epithelium. Indeed, extensive physiological and biophysical experiments have documented the presence of several different cotransporters in the retinal pigment epithelium. These include identification of cotransport of H⁺, lactate, and water in a membrane protein positioned in the apical plasma membrane of the retinal pigment epithelium of bullfrog eye (40) and human eye (12). Although direct evidence for the importance of this system is lacking, it appears likely that these cotransporters may contribute to transport of water across distinct ocular epithelia, which is under investigation in these laboratories.

Aquaporin expression in retinal glial cells. We report abundant expression of AQP4 in the processes of glial cells including Müller cells. As recently reviewed (22), Müller cells represent the principal glial cell of the retina and accomplish multiple functions. Müller cells express voltage-gated channels and neurotransmitter receptors, and they modulate neuronal activity by regulating the extracellular content of neuroactive substances including K⁺, glutamate, GABA, and H⁺. In the human retina the glial cells are divided into Müller cells, astrocytes, and microglia. However, in the rat eye, glial elements are comprised only of Müller cells and astrocytes, with astrocytes present exclusively in the innermost layers of the retina. Retinal blood vessels are entirely surrounded by Müller cell end-feet, except in the inner portion of the retina, where astrocyte end-feet also line the capillaries. The end-feet in contact with the capillary endothelium can release K⁺, acid equivalents, and water (22). As shown in the present study, AQP4 is extensively expressed in the glial processes facing retinal capillaries. This would be consistent with a role of AQP4 in water release to the capillaries, thereby contributing to maintenance of the extracellular osmolality during neuronal activity. AQP4 may also be hypothetically involved in K⁺ siphoning, a role that has recently been discussed in relation to the extensive polarized expression of AQP4 in glial processes in the cerebellum and elsewhere in the brain (24).