Aquaporin water channels and endothelial cell function

Abstract

The aquaporins (AQP) are a family of homologous water channels expressed in many epithelial and endothelial cell types involved in fluid transport. AQP1 protein is strongly expressed in most microvascular endothelia outside of the brain, as well as in endothelial cells in cornea, intestinal lacteals, and other tissues. AQP4 is expressed in astroglial foot processes adjacent to endothelial cells in the central nervous system. Transgenic mice lacking aquaporins have been useful in defining their role in mammalian physiology. Mice lacking AQP1 manifest defective urinary concentrating ability, in part because of decreased water permeability in renal vasa recta microvessels. These mice also show a defect in dietary fat processing that may involve chylomicron absorption by intestinal lacteals, as well as defective active fluid transport across the corneal endothelium. AQP1 might also play a role in tumour angiogenesis and in renal microvessel structural adaptation. However, AQP1 in most endothelial tissues does not appear to have a physiological function despite its role in osmotically driven water transport. For example, mice lacking AQP1 have low alveolar-capillary water permeability but unimpaired lung fluid absorption, as well as unimpaired saliva and tear secretion, aqueous fluid outflow, and pleural and peritoneal fluid transport. In the central nervous system mice lacking AQP4 are partially protected from brain oedema in water intoxication and ischaemic models of brain injury. Therefore, although the role of aquaporins in epithelial fluid transport is in most cases well-understood, there remain many questions about the role of aquaporins in endothelial cell function. It is unclear why many leaky microvessels strongly express AQP1 without apparent functional significance. Improved understanding of aquaporin-endothelial biology may lead to novel therapies for human disease, such as pharmacological modulation of corneal fluid transport, renal fluid clearance and intestinal absorption.

Introduction

The aquaporins (AQP) are a family of homologous water-transporting proteins that are expressed in many epithelial, endothelial and other tissues (reviewed in Borgnia et al. 1999; Ishibashi et al. 2000; Verkman & Mitra, 2000). There are at least 10 aquaporins in mammals and many more in plants, amphibians and lower organisms. Functional measurements indicate that mammalian aquaporins 1, 2, 4, 5 and 8 are probably water-selective, whereas aquaporins 3, 7, 9 and 10 (called ‘aquaglyceroporins’) also transport glycerol and other small solutes. AQP6 has been proposed to function in kidney endosomes as a pH-sensitive chloride channel. Structural studies of AQP1 indicate a homotetrameric assembly in membranes in which each monomer contains six tilted helical segments forming a barrel surrounding a putative aqueous pathway. Rare human mutations exist for three of the aquaporins. Humans with mutations in the major intrinsic protein of lens fibre (AQP0), which may function as a weak water transporter, develop cataracts (Berry et al. 2000). Humans with mutations in the AQP1 (Colton blood group antigen) manifest a urinary concentrating defect (King et al. 2001) and humans with mutations in the vasopressin-regulated water channel AQP2 have autosomal hereditary nephrogenic diabetes insipidus (Deen et al. 1994).

Our laboratory has generated a series of single and double knockout mice lacking aquaporins (reviewed in Verkman, 2002), as well as a mouse model of a nephrogenic diabetes insipidus in which a human disease-causing mutation (AQP2-T126M) was introduced into the mouse genome by targeted gene replacement (Yang et al. 2001). A number of interesting phenotypes were found that mostly involve defective epithelial cell fluid transport. For example, deletion of AQP1 resulted in a profound urinary concentrating defect, in part because of reduced water transport in epithelia of kidney proximal tubule and thin descending limb of Henle (Ma et al. 1998; Schnermann et al. 1998; Chou et al. 1999). Deletion of AQP3 produced a urinary concentrating defect with different characteristics because of reduced water permeability in collecting duct epithelia (Ma et al. 2000c). AQP3 null mice also have an abnormally dry stratum corneum in skin as a result of defective epidermal cell barrier/transport function (Ma et al. 2002). AQP4 null mice had impaired hearing because of reduced water permeability (and possibly K⁺ recycling) in supportive epithelial cells adjacent to electrically excitable hair cells (Li & Verkman, 2001). Deletion of AQP5 impaired fluid secretion by salivary (Ma et al. 2000c) and airway submucosal (Song & Verkman, 2001) glands by reducing water permeability of acinar epithelial cells (Krane et al. 2001).

This review is focused on the roles of aquaporins in endothelial cell function. AQP1 is expressed strongly throughout microvascular endothelial beds outside of the brain, such in microvessels in kidney, lung and airways, secretory glands, skeletal muscle, pleura and peritoneum (Nielsen et al. 1993; Folkesson et al. 1994; Hasegawa et al. 1994; Effros et al. 1997; Nielsen et al. 1997; Devuyst et al. 1998; Gresz et al. 2001; Hurley et al. 2001). AQP1 is also expressed in endothelial cells in cornea, trebecular meshwork in canal of Schlemm, and central lacteals in small intestine (Nielsen et al. 1993; Hasegawa et al. 1994; Wen et al. 2001). Aquaporins have not been identified in endothelial cells in brain, though AQP4 is strongly expressed in astrocyte foot processes that comprise the blood–brain barrier in close contact with endothelial cells. We systematically review available data for involvement of aquaporins in endothelial cell physiology in various organ systems.

Renal vasa recta

At least six aquaporin water channels are expressed in kidney, four of which have been shown to be important for renal function (Fig. 1A) (reviewed in Yamamoto & Sasaki, 1998; Verkman, 2000; Kwon et al. 2001). AQP1 is expressed in proximal tubule and thin descending limb of Henle epithelia, and in the microvascular endothelium of outer medullary descending vasa recta (OMDVR). AQP2 is expressed in collecting duct principal cells and undergoes vasopressin-regulated trafficking between an intracellular vesicular compartment and the cell apical plasma membrane. AQP3 and AQP4 are coexpressed at the basolateral membrane of collecting duct epithelial cells. The formation of a concentrated urine requires the trapping of NaCl and urea in the renal medulla, as a well as high water permeability across the collecting duct epithelium. In portions of OMDVR, water efflux occurs despite the existence of Starling forces (hydrostatic and oncotic forces) that favour volume influx (Michel, 1995; Edwards et al. 2000), suggesting that water efflux involves a water-only (AQP1) pathway in which NaCl and urea osmotic gradients drive water movement.

Fig. 1.

AQP1 facilitates osmotic water transport across renal vasa recta endothelium. (A) Location of the four principal aquaporin water channels in kidney tubules and microvasculature. (B) Light micrographs of OMDVR microdissected from wildtype (+/+) and AQP1 null (−/−) mice. Unperfused OMDVR from wildtype mice have diameter −10 μm. (C) Osmotic water permeability (P_f) measured in isolated microperfused OMDVR. P_f was measured in response to 200 mOsm gradients of NaCl or raffinose. Data from Pallone et al. (2000).

To investigate the role of AQP1 in vasa recta function, OMDVR were microdissected from vascular bundles and perfused in vitro (Pallone et al. 2000). Interestingly, OMDVR from AQP1 null mice were remarkably larger in diameter than those from wildtype mice (Fig. 1B) and were more easily freed from vascular bundles. It was proposed that deletion of AQP1 might lead to remodelling of the microvessel wall as a means of long-term adaptation. Osmotic water permeability in response to a 200-mm NaCl gradient (bath > lumen) was reduced by more than 10-fold in AQP1 null mice (Fig. 1C), and by ∼2-fold when driven by a raffinose gradient. These data, together with p-chloromercuribenzenesulphonate inhibition measurements, suggested that most NaCl-driven water transport occurs by a transcellular route through AQP1, whereas raffinose-driven water transport also involves a parallel, AQP1-independent, mercurial-insensitive pathway. In a mathematical model of the medullary microcirculation, deletion of AQP1 resulted in diminished concentrating ability due to enhancement of medullary blood flow, partially accounting for the observed urinary concentrating defect. These data support the conclusion that AQP1 in OMDVR is an important component of the urinary concentrating mechanism. Mechanistically, NaCl and urea gradients drive water transport from the OMDVR lumen to the medullary interstitium across AQP1, thereby shunting blood flow from descending to ascending vasa recta in the outer medulla. The reduction in blood flow to the deep inner medullary portions of the vasa recta is predicted to enhance the efficiency of countercurrent exchange by reducing solute washout.

Alveolar microvasculature

The lung and airways participate in fluid movement during airway hydration, reabsorption of alveolar fluid in the neonatal period and formation/resolution of pulmonary oedema resulting from heart failure or lung injury. The barriers to water movement between the air space and capillary compartments consist of an epithelium, interstitium and endothelium (Fig. 2A). The alveolar epithelium is composed mainly of type I cells which express AQP5 at their apical membrane. The microvascular endothelium expresses AQP1. AQP4 is expressed at the basolateral membrane of surface epithelial cells in large and small airways, and AQP3 in airway epithelial cells in nasopharynx and trachea. The fluid transport-related functions in the peripheral lung include alveolar fluid clearance, gas exchange, and adaptation to acute and subacute lung injury. As described below, the principal finding was that although AQP1 and AQP5 provide the major pathways for osmotically driven water transport across the serial endothelial and epithelial barriers, they are not required for physiologically important lung functions (reviewed in Borok & Verkman, 2002).

Fig. 2.

AQP1 facilitates osmotic water transport across alveolar microvascular endothelium. (A) Schematic of barriers in airways and lung showing sites of aquaporin expression. (B) Osmotic water permeability across the air space–capillary barrier measured in perfused mouse lung in which the air space was filled with isosmolar saline containing a membrane-impermeant fluorescent indicator. Fluorescence changes as water moves into or out of the air spaces in response to changes in pulmonary artery perfusate osmolality. (C) Data shown for wildtype and AQP1 null mice. (D) Measurement of lung microvascular endothelial water permeability. The air space was filled with an inert perfluorocarbon, and the pulmonary artery was perfused with solutions of indicated osmolalities containing FITC-dextran. In response to an increase in perfusate osmolality, water moves into capillaries, resulting in fluorophore dilution and a prompt decrease in pleural surface fluorescence signal. Fluorescence returns to its initial level as osmotic equilibrium is established between capillary fluid and interstitium. (E) Gravimetric measurement of lung water permeability. The pulmonary artery was perfused with solutions of specified osmolality and lung weight was measured continuously by a gravimetric transducer. Data from Bai et al. (1999), Ma et al. (2000a) and Song et al. (2000b).

Osmotic water transport between the air space and capillary compartments was measured by a pleural surface fluorescence method as shown schematically in Fig. 2(B) (Carter et al. 1996). The air space was filled with fluid containing a membrane-impermeant fluorophore, and the pulmonary artery was perfused with solutions of specified osmolalities. Because of the finite penetration depth of the excitation light, the surface fluorescence signal is directly proportional to the air space fluorophore concentration. In response to an osmotic gradient, water flows between the air space and the perfusate compartments, resulting in a change in fluorophore concentration that is detected continuously by measurement of pleural surface fluorescence. Airspace-capillary osmotic water permeability was slowed ∼10-fold by deletion of AQP1 (Fig. 2C) or AQP5, and by > 30-fold by deletion of AQP1 and AQP5 together in double knockout mice (Bai et al. 1999; Ma et al. 2000a).

A modified pleural surface fluorescence strategy was developed to measure microvascular endothelial water permeability in which the air space was filled with an inert, water-insoluble perfluorocarbon to restrict lung water to two compartments, the interstitium and capillaries, so that the microvascular endothelium becomes the single rate–limiting permeability barrier (Carter et al. 1998). The pulmonary artery was perfused with solutions of specified osmolalities containing identical concentrations of a fluorescent volume marker. In response to a change in perfusate osmolality, water is osmotically driven into or out of the capillaries, resulting in fluorophore dilution or concentration. This produces a prompt change (decrease for fluorophore dilution; Fig. 2D) in pleural surface fluorescence, whose magnitude increases with increased water permeability or decreased pulmonary arterial flow. The prompt deflection is followed by a slower return of fluorescence signal to the original level as interstitial and capillary osmolalities equilibrate. Using a three-compartment model to compute capillary P_f from the fluorescence data, microvascular P_f was very high (∼0.03 cm s⁻¹), inhibited by mercury compounds, and greatly decreased by AQP1 deletion (Fig. 2D, bottom). Another useful approach to measure both osmotically induced water transport and hydrostatic filtration in mouse lungs is gravimetry, in which lung weight is measured continuously in response to changes in pulmonary artery perfusate osmolality or pressure (Song et al. 2000b). When the air space compartment is filled with isosmolar saline, changes in perfusate osmolality drive water movement across endothelial and epithelial barriers, producing changes in the water content of the interstitial and air space compartments observed as changes in lung weight. Figure 2(E) shows gravimetric data in which decreasing the perfusate osmolality from 300 to 200 mOsm produced reversible increases in lung weight as water was driven into the air space. AQP1 deletion produced remarkable slowing of the lung weight increase (Fig. 2E, bottom). The gravimetric method has also been used to measure hydrostatically driven fluid uptake in the lung (Song et al. 2000b).

However, despite the dependence of lung fluid osmotic water permeability on AQP1, there was no effect of AQP1 (or AQP5) deletion on alveolar fluid absorption, even when fluid absorption was maximally stimulated by beta agonists and type II cell up-regulation (Bai et al. 1999; Ma et al. 2000a). Also, the rapid reabsorption of fluid from the air space just after birth was not impaired by aquaporin deletion, and aquaporin deletion did not affect the response of the adult lung in experimental models of lung injury, including acid-induced epithelial injury, thiourea-induced endothelial injury and hyperoxic subacute lung injury (Song et al. 2000a). The remarkably slower rate of alveolar fluid absorption compared to proximal tubule fluid absorption and saliva secretion was proposed to explain the lack of effect of AQP1 and AQP5 deletion on alveolar fluid clearance. Also, AQP1 deletion in mice did not impair airway humidification (Song et al. 2001) or lung carbon dioxide transport (Yang et al. 2000; Fang et al. 2002), which had been proposed to involve AQP1 (Nakhoul et al. 1998). Together these results show that although AQP1 in alveolar microvascular endothelia facilitates osmotic water transport, its deletion does not impair lung fluid or gas transport physiology.

Secretory gland function

The salivary gland expresses multiple aquaporins including AQP1 in microvascular endothelia, AQP4 in ductal epithelia and AQP5 at the apical membrane of serous acinar cells. Saliva secretion involves active salt transport into the acinar lumen across epithelial cells, which drives osmotic water transport across serial AQP1-containing endothelial and AQP5-containing epithelial barriers. The role of aquaporins in pilocarpine-stimulated saliva secretion was studied by saliva collections in knockout mice (Ma et al. 1999). Saliva secretion was remarkably reduced in AQP5 null mice, but unimpaired in AQP1 or AQP4 null mice. Interestingly, the saliva in AQP5 knockout mice was hypertonic and hyperosmolar, as expected for unimpaired salt pumping across a relatively water-impermeable luminal barrier. The lack of effect of AQP1 deletion indicates that osmosis across the endothelial barrier is not a rate-limiting determinant of fluid secretion. Similarly, deletion of AQP1 in mice does not affect tear secretion by lacrimal glands (Moore et al. 2000) which occurs by a mechanism similar to that in salivary gland.

Pleural and peritoneal barriers

Fluid is continuously secreted into and cleared from the pleural space. Pleural fluid can accumulate in congestive heart failure, lung infection, lung tumour and the acute respiratory distress syndrome. Fluid entry into the pleural space involves filtration across microvascular endothelia near the pleural surface, and movement across a mesothelial barrier lining the pleural space, whereas fluid clearance occurs primarily by lymphatic drainage (Staub et al. 1985). AQP1 is expressed in microvascular endothelia near the visceral and parietal pleura and in mesothelial cells in visceral pleura (Fig. 3A) (Song et al. 2000c). The role of AQP1 in pleural barrier osmosis, fluid absorption and fluid accumulation was studied using AQP1 null mice. Osmotic water permeability was measured in anaesthetized, mechanically ventilated mice from the kinetics of pleural fluid osmolality after instillation of hypertonic or hypotonic fluid into the pleural space. Osmotic equilibration of pleural fluid was rapid in wildtype mice (50% equilibration in < 2 min), and slowed by ∼4-fold in AQP1 null mice (Fig. 3B). However, the clearance of isosmolar saline instilled in the pleural space (∼4 mL kg⁻¹ h⁻¹) was not affected by AQP1 deletion (Fig. 3C), nor was the accumulation of pleural fluid (∼0.035 mL h⁻¹) in a fluid overload model produced by intraperitoneal saline administration and renal artery ligation. Therefore, although AQP1 facilitates rapid osmotic equilibration across the pleural surface, AQP1 did not appear to play a role in physiologically important mechanisms of pleural fluid accumulation or clearance.

Fig. 3.

Involvement of AQP1 in water transport across the pleural barrier. (A) Immunofluorescence localization of AQP1 protein in pleural microvascular endothelia (arrows) from wildtype and AQP1 null mice. Arrowheads indicate pleural surface. Scale bar = 150 μm. (B) Reduced osmotically induced water transport across the pleural barrier in AQP1 null mice. After anaesthesia and mechanical ventilation, the pleural space was infused with 0.25 mL of a hyperosmolar (500 mOsm) solution and pleural fluid osmolality was measured at indicated times. (C) Isosmolar fluid absorption from the pleural space. The pleural space was infused with 0.25 mL of an isosmolar solution containing 1% albumin and pleural fluid volume was measured at indicated times. Adapted from Song et al. (2000c).

A similar strategy was used to study the role of AQP1 in fluid movement into and out of the peritoneal cavity. The peritoneal cavity is lined by a membranous barrier that provides a large surface for fluid movement between peritoneal capillaries and the peritoneal cavity. Ascites can accumulate in conditions associated with decreased serum oncotic pressure, increased portal venous pressure, or peritoneal cavity inflammation/infection. The large peritoneal surface is exploited in peritoneal dialysis, where water, electrolytes, urea and uraemia-causing toxins are extracted from blood by repeated infusion and removal of dialysate solutions into the peritoneal cavity. Expression studies and measurements of mercurial inhibition of water transport have suggested the involvement of AQP1 in osmosis across the peritoneal barrier (Carlsson et al. 1996). AQP1 has been localized to capillary endothelia and mesangium near the peritoneal luminal surface. Osmotically induced water transport was compared in wildtype and AQP1 null mice (Yang et al. 1999). Hyperosmolar saline (saline +300 mm sucrose) was infused rapidly into the peritoneal cavity via a catheter and serial fluid samples were withdrawn to quantify the time course of osmotic equilibration using albumin as a volume marker. The albumin dilution data showed a ∼2.5-fold decreased water flux in the AQP1 null mice. However, the transport of radiolabelled urea was not affected by AQP1 deletion, which is consistent with the function of AQP1 as a water-selective channel. Therefore AQP1 has an important role in osmosis during peritoneal dialysis. However, fluid absorption after intraperitoneal saline infusion was not affected by AQP1 deletion, suggesting that AQP1 is not an important determinant of peritoneal fluid movement under physiological conditions.

Intestinal lacteals

AQP1 is expressed at sites in the proximal gastrointestinal tract that play a role in dietary fat processing including cholangiocytes in liver (bile production), pancreatic microvascular endothelium (pancreatic fluid production), gallbladder microvascular endothelium (bile storage) and intestinal lacteal endothelium (chylomicron absorption, Fig. 4A). The hypothesis was tested that AQP1 facilitates dietary fat processing (Ma et al. 2000b). It was found that young AQP1 null mice fed a high- (50%) fat diet failed to gain weight, whereas litter-matched wildtype mice gained weight rapidly (Fig. 4B). The young mice thrived when returned to a normal low-fat diet. The AQP1 null mice on a high-fat diet developed steatorrhea and had reduced serum triglyceride concentration. The null mice had elevated concentrations of pancreatic enzymes in their small intestine and stool (Fig. 4C), normal pH in duodenal fluid, and normal bile/pancreatic fluid production, suggesting a defect in absorption rather than digestion. The mechanisms by which AQP1 in lacteal endothelia might facilitate chylomicron absorption remain to be determined.

Fig. 4.

AQP1 in central lacteal endothelium in small intestine facilitates fat absorption. (A) Immunofluorescence localization of AQP1 in duodenum showing strong AQP1 immunostaining in central lacteals of wildtype mice. Scale bar = 50 μm. (B) Weight loss and appearance of mice on a 50% fat diet. (left) Weight curves of wildtype, heterozygous, and AQP1 null mice on a high-fat diet. Initial mouse weight was 10–12 g. Mice were switched to normal diet on day 8. (right) Photograph of mice on day 4 of a high-fat diet showing oily appearing fur and smaller size of AQP1 null mice. (C) Lipase activity in faeces and small intestinal lumen. Samples were collected from weight matched mice (10–12 g) before and after 3 days on high-fat diet. Data taken from Ma et al. (2000b).

Blood–brain barrier

AQP1 and AQP4 are expressed strongly in the brain – AQP1 in the ventricular-facing surface of choroid plexus (Fig. 5A), and AQP4 in astrocyte foot processes near blood vessels (Fig. 5B,C) and in ependymal and pial surfaces in contact with cerebrospinal fluid (Frigeri et al. 1995; Nielsen et al. 1997; Rash et al. 1998). AQP4 is assembled in membranes in regular square arrays called orthogonal arrays of particles (Fig. 5D), as shown by the absence of orthogonal arrays in AQP4 null mice (Verbavatz et al. 1997) and confirmed by immunogold-label freeze-fracture electron microscopy (Rash et al. 1998). So far aquaporins have not been identified on endothelial cells in the cerebral microvasculature (Kobayashi et al. 2001).

Fig. 5.

Aquaporins and the blood–brain barrier. (A) Immunofluorescence showing AQP1 protein expression in the apical surface of choroid plexus cells. (B) Diagram of a brain microvessel in cross-section. Endothelial cells (red) are joined by tight junctions (black) and surrounded by astrocyte foot processes (blue). AQP4 water channels (yellow) are expressed at the astrocyte foot processes. (C) Human brain showing AQP4 immunoreactivity (brown) around microvessels. Scale bar = 10 μm. (D) Freeze-fracture electron micrograph (E-face) of rat brain showing AQP4 square arrays in an astrocyte foot process. Inset: arrays in P-face micrograph. (E) Survival of wildtype vs. AQP4 knockout mice after acute water intoxication produced by intraperitoneal water infusion. (F) (top) Brain sections of mice at 24 h after ischaemic stroke produced by permanent middle cerebral artery occlusion. Note midline shift and marked oedema in brain from wildtype mice. (bottom) Averaged hemispheric enlargement expressed as a percentage determined by image analysis of brain sections. Adapted from Manley et al. (2000) and Papadopoulos et al. (2002).

AQP1 has been proposed to facilitate cerebrospinal fluid formation, though direct evidence for this is not yet available. AQP4 has been proposed to be involved in cerebral oedema and in regulating interactions of astroglia with neurones and endothelia. As reviewed recently (Papadopoulos et al. 2002) a number of indirect lines of evidence support a role of AQP4 in brain oedema, including up-regulation of AQP4 expression in several experimental models of brain injury and in brain tumours. AQP4 appears in chick embryos at the time of development of the blood–brain barrier (Nico et al. 2001). Studies in AQP4 null mice provide direct evidence for the involvement of AQP4 in brain swelling. Initial evaluation of AQP4 null mice showed no overt neurological abnormalities or defects in osmoregulation (Ma et al. 1997). Brains from AQP4 null mice show reduced osmotic water permeability as measured in isolated membrane vesicles (Ma et al. 1997) and brain slices (Solenov et al. 2001). We tested the hypothesis that AQP4 plays a role in the generation of brain oedema in response to two established neurological insults – acute water intoxication produced by intraperitoneal water infusion (resulting in hyponatremia and vasogenic brain oedema) and ischaemic stroke produced by permanent middle cerebral artery occlusion (resulting in a combination of vasogenic and cellular oedema) (Manley et al. 2000). The survival of AQP4 null mice after water intoxication was greatly improved (Fig. 5E), which corresponded to significantly reduced brain swelling particularly in astrocytic foot processes. In addition, at 24 h after brain ischaemia produced by permanent middle cerebral artery occlusion, there was improved clinical outcome and much less brain swelling (Fig. 5F).

Recently, AQP1 was localized to neurones in superficial layers of dorsal horn in spinal cord, which contains C-fibres involved in pain sensation. Osmotically induced spinal cord swelling was reduced in AQP1 null mice in dorsal horn (Solenov et al. 2001), and markedly impaired pain sensation was demonstrated in response to thermal (tail flick test) and chemical (capsaicin injection) stimuli (Oshio et al. 2001). Together with data showing impaired hearing (Li & Verkman, 2001) and retinal signal transduction (Li et al. 2002) in AQP4 null mice, these results support a role for aquaporins in neural signal transduction, possibly in rapid water and/or K⁺ recycling.

Eye physiology

The intra-ocular pressure (IOP) is an important index of eye physiology, which when chronically elevated can cause glaucoma. The principal determinants of IOP are the rates of aqueous fluid production and drainage. Aqueous fluid production involves fluid secretion by the ciliary epithelium, and aqueous fluid drainage (outflow) involves pressure-driven bulk fluid flow in the canal of Schlemm as well as fluid movement through the sclera by seepage across the ciliary muscle and supraciliary space. AQP1 is expressed strongly in non-pigmented ciliary epithelium and trabecular meshwork endothelium. We recently studied the role of AQP1 in IOP and aqueous fluid dynamics (Zhang et al. 2002). IOP and aqueous fluid outflow were measured using micropipettes, and aqueous fluid production by in vivo confocal microscopy after transcorneal iontophoretic introduction of fluorescein. IOP in wildtype mice was 16.0 mmHg, aqueous fluid volume was 7.2 μL, aqueous fluid production was 3.6 µL h⁻¹, and aqueous fluid outflow was 0.36 µL h⁻¹ mmHg⁻¹ (outflow linearly related to IOP). IOP was significantly decreased by up to 1.8 mmHg and fluid production by up to 0.9 µL h⁻¹ in age/litter-matched mice lacking AQP1. However, AQP1 deletion did not significantly affect aqueous outflow despite its strong expression in trabecular meshwork endothelium. The reasons for AQP1 expression in trabecular meshwork endothelium remain unclear, particularly since a pressure-driven bulk flow mechanism for aqueous fluid outflow would not be expected to involve transcellular water-only transport.

In cornea, AQP1 is expressed in endothelial cells and AQP5 in epithelial cells. We compared corneal thickness, water permeability and response to experimental swelling in wildtype mice and transgenic null mice lacking AQP1 or AQP5 (Thiagarajah & Verkman, 2002). Corneal thickness was remarkably reduced in AQP1 null mice (101 µm) and increased in AQP5 null mice (144 µm) compared to wildtype mice (123 µm). After exposure of the corneal endothelial surface to hypotonic saline by anterior chamber perfusion, the rate of corneal swelling (7.1 µ min⁻¹ in wildtype mice) was reduced by AQP1 deletion (1.6 µm min⁻¹), indicating that AQP1 provides the principal route for osmotically driven water transport across the corneal endothelium. Although baseline corneal transparency was not impaired by AQP1 deletion, the recovery of corneal transparency and thickness after hypotonic swelling (10 min exposure of corneal surface to 100 mOsm solution) was remarkably delayed in AQP1 null mice, with ∼75% recovery at 7 min in wildtype mice compared to 5% recovery in AQP1 null mice. The impaired recovery of corneal transparency in AQP1 null mice provides evidence for the involvement of AQP1 in active extrusion of fluid from the corneal stroma across the corneal endothelium.

Tumour microvasculature

Expression of AQP1 protein has been reported in endothelial cells (but not in tumour cells) in rat models of glioblastoma and mammary carcinoma (Endo et al. 1999). A microarray study of differential gene expression also found increased AQP1 transcript expression in human glioblastoma, which was confirmed by AQP1 immunohistochemistry (Markert et al. 2001). AQP1 expression in microvessels of neoplastic brain was proposed to contribute to increased blood–brain barrier water permeability in aggressive brain tumours (Papadopoulos et al. 2001). In preliminary experiments, the possible involvement of AQP1 in tumour growth and angiogenesis was studied in AQP1 null mice (Ma and Verkman, unpublished results). After subcutaneous implantation of Lewis lung carcinoma cells, average tumour growth rate was ∼2-fold slowed in AQP1 null mice, and the tumours showed less vascularity. Experiments are needed to investigate whether AQP1 deletion in tumour microvessels impairs angiogenesis directly, since the slowed tumour growth in AQP1 null mice may be secondary to systemic differences in haemodynamics or nutritional status.

Conclusions

The phenotype studies suggest that aquaporins can have physiological importance for rapid water transport in response to osmotic gradients created by continuous flow (as in kidney) or active salt pumping (as in saliva secretion). In addition aquaporins appear to be involved in electrically excitable tissues, perhaps by dissipating osmotic gradients created during rapid K⁺ recycling. AQP1 is widely expressed in microvascular endothelia outside of the brain; however, only in renal vasa recta, where transendothelial osmotic transport driven by gradients of small solutes occurs, is AQP1 of clear physiological significance. The reasons for the expression of aquaporins in microvascular endothelia without apparent functional significance remain unclear. The possible involvement of AQP1 in microvessel growth and remodelling in tumours and kidney requires further study. In non-microvascular endothelia, AQP1 in intestinal lacteals appears to be important for intestinal absorption. AQP1 in corneal endothelium appears to be important for restoration of corneal water content and transparency following oedema. The tight endothelium in brain, where water channels have not yet been found, is in close contact with AQP4-expressing astrocytic foot processes. The reduced brain swelling in experimental models of brain injury suggests that AQP4 is an important determinant of brain water homeostasis. The phenotype studies suggest that pharmacological modulation of aquaporin function may provide new therapies in human disease, such as diuretics and regulators of intra-ocular pressure and swelling in brain and cornea (Verkman, 2001).