I. INTRODUCTION

Top Previous Next References

The discovery of aquaporin membrane water channels by Agre and co-workers (6, 7, 191, 192) answered a long-standing biophysical question of how water crosses biological membranes specifically, and provided insight, at the molecular level, into the fundamental physiology of water balance and the pathophysiology of water balance disorders. Out of at least 10 aquaporin isoforms, at least 7 are known to be present in the kidney at distinct sites along the nephron and collecting duct. Aquaporin-1 (AQP1) is extremely abundant in the proximal tubule and descending thin limb where it appears to be the main site for proximal nephron water reabsorption. It is also present in the descending vasa recta. AQP2 is abundant in the collecting duct principal cells and is the chief target for the regulation of collecting duct water reabsorption by vasopressin. Acute regulation involves vasopressin-induced trafficking of AQP2 between an intracellular reservoir in vesicles and the apical plasma membrane. In addition, AQP2 is involved in chronic/adaptational control of body water balance, which is achieved through regulation of AQP2 expression. Importantly, multiple studies have now underscored a critical role of AQP2 in several inherited and acquired water balance disorders. This includes inherited forms of nephrogenic diabetes insipidus, acquired states of nephrogenic diabetes insipidus, and other diseases associated with urinary concentrating defects where AQP2 expression and targeting are affected. Conversely, AQP2 expression and targeting appear to be increased in some conditions with water retention such as pregnancy and congestive heart failure. AQP3 and AQP4 are basolateral water channels located in the kidney collecting duct and represent exit pathways for water reabsorbed via AQP2. Several additional aquaporins have been identified in kidney, including AQP6, which is expressed at lower abundance in the collecting duct intercalated cells. Four additional aquaporins (AQP7, -8, -9, and -10) are also expressed in kidney, but less is known about expression sites and their pathophysiological function. In this review we focus mainly on the role of collecting duct aquaporins in water balance regulation and in the pathophysiology of water balance disorders.

	II. DISCOVERY OF THE FIRST MOLECULAR WATER CHANNEL

Top Previous Next References

The molecular identity of membrane water channels long remained elusive until the pioneering discovery of AQP1 by Agre and colleagues around 1989-1991, and the "route" of this discovery has been covered in several reviews (6).

Previous attempts to purify water channel proteins from tissues, or to isolate water channel cDNAs by expression cloning in oocytes, were unsuccessful (see reviews in Refs. 4, 8, 15). The identification of the first water channel took place via a convoluted route. The Agre lab at that time worked on identifying the function of Rhesus polypeptides and in the process of isolating a 32-kDa bilayer-spanning polypeptide component of the red cell Rh blood group antigen (7, 205); a 28-kDa polypeptide was partially copurified (46) which displayed a number of biochemical characteristics (6). It was comprised of hydrophobic amino acids and exhibited an unusual detergent solubility allowing purification and biochemical characterization. The 28-kDa polypeptide was found to exist as an oligomeric protein with the physical characteristics of a tetramer. The NH₂-terminal amino acid sequence was identified (213) that subsequently allowed cDNA cloning (191). This protein was first known as "CHIP28" (channel-like integral protein of 28 kDa), but was later redubbed aquaporin-1 or AQP1 (8). The Human Genome Nomenclature Committee has accepted this nomenclature for all related proteins.

The Xenopus laevis oocyte expression system, largely developed by Wright et al. (254), was used by Preston et al. (191) and oocytes injected with cRNA for "CHIP28" (or AQP1), exhibited remarkably high osmotic water permeability causing the cells to swell rapidly and burst in hypotonic buffer. In contrast, control oocytes injected with water alone exhibited water permeability one order of magnitude lower. Thus this strongly suggested that AQP1 was a molecular water channel. Morever, the oocyte studies demonstrated that AQP1 behaves like the water channels in native cell membranes (192). The biophysical characteristics of AQP1 in oocytes revealed a series of the water channel specific characteristics. To establish the role of AQP1 as a molecular water channel further, and to rule out the remote possibility that AQP1 protein expression was leading to coexpression of a native molecular water channel, a series of studies were performed on purified AQP1 reconstituted in liposomes. Highly purified AQP1 protein from human red blood cells was reconstituted with pure phospholipid into proteoliposomes and were compared with liposomes without AQP1. The analyses were performed by rapid transfer to hyperosmolar buffer (267). The result confirmed that AQP1 is a molecular water channel. Together these studies indicated that AQP1 is both necessary and sufficient to explain the well-recognized membrane water permeability of the red blood cell and strongly suggests that AQP1 water channels are of fundamental importance for transmembrane or transcellular water transport in tissues where it is expressed. This laid the ground for the identification of a number of related channels by homology cloning and other means, which has led to the understanding that aquaporins play essential roles in transmembrane and transepithelial water transport in the kidney and elsewhere and are critical in renal regulation of body water balance and in many water balance disorders.

	III. AQUAPORIN STRUCTURE

Top Previous Next References

Since the discovery of aquaporins, major efforts have been aimed at elucidating their structural organization. Hydropathy analysis of the deduced amino acid sequence of AQP1 led to the prediction that the protein resides primarily within the lipid bilayer (191), consistent with the initial studies of AQP1 in red cell membranes (46). AQP1 contains an internal repeat with the NH₂- and the COOH-terminal halves being sequence related and each containing the signature motif Asn-Pro-Ala (NPA) (181, 252). This is consistent with earlier observations on the homologous major intrinsic protein from lens, (MIP, now referred to as AQP0). When evaluated by hydropathy analysis, six bilayer-spanning domains are apparent (Fig. 1); however, the apparent interhelical loops B and E also exhibit significant hydrophobicity. Critical to the topology is the location of loop C which connects the two halves of the molecule. Epitope mapping has been extensively used for analyzing the AQP1 structure by adding a peptide epitope at various sites that can be localized to intracellular or extracellular sites with antibodies and by selective proteolysis of intact membranes or inside-out membrane vesicles. Using such an approach, Preston et al. (194) demonstrated that loop C resides at the extracellular surface of the oocytes, confirming the obverse symmetry of the NH₂- and COOH-terminal halves of the molecule. Site-directed mutagenesis of AQP1 expressed in oocytes led to the observations that Cys-189 in the E loop is the site of mercurial inhibition (193), and this has been found by other workers (271). Thus loop E has been implicated as a structural component of the aqueous pathway. Moreover, it has been demonstrated that loops B and E were functionally essential for water permeability, leading to the "hour-glass" model (107) in which these domains overlap midway between the leaflets of the bilayer, creating a narrow aqueous pathway (Fig. 1).

View larger version (86K): [in this window] [in a new window]   Fig. 1. A: schematic representation of the structural organization of aquaporin-1 (AQP1) monomers in the membrane (top and bottom). Aquaporins have six membrane-spanning regions, both intracellular NH2 and COOH termini, and internal tandem repeats that, presumably, are due to an ancient gene duplication (top). The topology is consistent with an obverse symmetry for the two similar NH2- and COOH-terminal halves (bottom). The tandem repeat structure with two asparagine-proline-alanine (NPA) sequences has been proposed to form tight turn structures that interact in the membrane to form the pathway for translocation of water across the plasma membrane. Of the five loops in AQP1, the B and E loops dip into the lipid bilayer, and it has been proposed that they form "hemichannels" that connect between the leaflets to form a single aqueous pathway within a symmetric structure that resembles an "hourglass." B: AQP1 is a multisubunit oligomer that is organized as a tetrameric assembly of four identical polypeptide subunits with a large glycan attached to only one.

Reconstitution of highly purified red cell AQP1 into membranes forms crystals with very uniform lattices that retain full osmotic water permeability (248). High-resolution electron microscopic evaluation of such membranes subjected to negative staining allowed the production of a three-dimensional reconstruction (247). Image projections revealed the presence of multiple bilayer-spanning domains, and atomic force microscopy further defined the orientation and extramembranous dimensions of AQP1 (249). Electron crystallographic analysis of cryopreserved specimens has been undertaken which allowed determination of the three-dimensional structure of AQP1 at 6-Å resolution (21, 134, 246). These images confirm the tetrameric organization, within which each subunit is comprised of six tilted, bilayer-spanning -helices that form a right-handed bundle surrounding a central density, an arrangement which is strikingly similar to the original proposed hour-glass arrangement of the B and E loops (107). This approach eventually provided the AQP1 structure at 3.8-Å resolution. By modeling the AQP1 sequence within the electron density, the three-dimensional atomic structural model of AQP1 has emerged (162a). This model first provided a molecular answer to the long-standing problem, How can water channels avoid passage of protons (H₃O⁺)? As predicted, loops B and E are associated by Van der Waals interactions between the two NPA motifs. Free hydrogen bonding occurs in the column of water within the pore, except at the very center where a single water molecule transiently reorients to bond with the two asparagines residues of the NPA motif. This results in minimum resistance to the flow of water, thus permitting kidneys to perform their important physiological roles of reabsorbing water while excreting acid (see below).

The structural organization of other aquaporins such as bacterial aquaporin-Z and plant aquaporins have also been studied, providing further information about the general structure of aquaporins. They display marked similarities to AQP1. This work has been reviewed in detail previously (200). Virtually simultaneous to the studies of AQP1, the atomic structure of GlpF, the glycerol-transporting homolog of Escherichia coli, was solved at 2.2-Å resolution by X-ray crystallographic analysis (78a). The structure of GlpF is virtually identical to AQP1; however, the aqueous pore is slightly wider, and key hydrophic residues provide a "slippery slide" for the carbon backbone of glycerol. Although the physiological need for glycerol transport is not understood in some tissues, it may confer important properties to other tissues (e.g., glycerol release from adipocytes in response to starvation).

As discussed above, the crystallographic studies have confirmed the tetrameric organization of AQP1 in the membrane (Fig. 1). Studies by Brown and colleagues (235, 239, 258) using Chinese hamster ovary (CHO) cells transfected with AQP1 to -5 have indicated that AQP2, -3, and -5 may also form tetramers in the membrane.

Not all aquaporins appear to assemble in the plasma membrane as tetramers. Recently, several studies revealed that AQP4 forms larger oligomeric structures in the plasma membrane. It is well established that freeze-fracture analyses of glial cells and other cells, later found to express high amounts of AQP4, have a high density of intramembrane particle square arrays also called orthogonal arrays (clusters of intramembrane particles in a special systematic/geometric organization). The subsequent demonstration that square arrays are absent in cells from transgenic knockout mice lacking AQP4 protein (240) supports the view that AQP4 may form these square arrays. Recently, the presence of AQP4 within these square arrays was established directly using freeze fracture immunogold labeling by Rash et al. (198). It is not yet clear if these square arrays are themselves made up of tetramers.

Similarly, structures (referred to as linear parallel grooves) exist in the "intramembrane particle arrays" described in toad bladder (which is a functional homolog of the mammalian collecting duct) and thought to represent the sites of antidiuretic hormone (ADH)-induced water flow in this tissue (22, 110). Although suspected, it remains to be proven that these represent members of the aquaporin family.

	IV. BIOPHYSICS OF AQUAPORIN FUNCTION

Top Previous Next References

Expression of AQP1 in X. laevis oocytes by Preston et al. (192) demonstrated that AQP1-expressing oocytes exhibited remarkably high osmotic water permeability (P_f ~200 × 10⁴ cm/s), causing the cells to swell rapidly and explode in hypotonic buffer. The osmotically induced swelling of oocytes expressing AQP1 occurs with a low activation energy and is reversibly inhibited by HgCl₂ or other mercurials. Only inward water flow (swelling) was examined, but it was predicted that the direction of water flow through AQP1 is determined by the orientation of the osmotic gradient. Consistent with this, it was later demonstrated that AQP1-expressing oocytes swell in hyposmolar buffers but shrink in hyperosmolar buffers (160).

Highly purified AQP1 protein from human red blood cells was reconstituted with pure phospholipid into proteoliposomes and were compared with liposomes without AQP1. The analyses were performed by rapid transfer to hyperosmolar buffer (267). The unit water permeability (conductance per monomeric AQP1) was extremely high (P_f ~3 × 10⁹ water molecules subunits/s). As demonstrated in oocytes, AQP1 reconstituted in liposomes also displayed sensitivity to HgCl₂, which dramatically and reversibly reduced the osmotic water permeability. The AQP1-containing liposomes also exhibited a low activation energy for water permeation. This was confirmed in other studies (234). Zeidel et al. (268) also demonstrated that AQP1 proteoliposomes are not permeable to various small solutes or protons, thus revealing that AQP1 is water selective. Together, these studies indicated that AQP1 is both necessary and sufficient to explain the well-recognized membrane water permeability of the red blood cell and strongly suggests that AQP1 water channels are of fundamental importance for transmembrane or transcellular water transport in tissues where it is expressed.

Over the past 4 years a series of studies have explored the issues of selectivity and polytransport function of aquaporins. This has led to a division of aquaporins (4) into a group that transports water relatively selectively (the "orthodox" set or "aquaporins") and a group of water channels that also conduct glycerol and other small solutes in addition to water (the "cocktail" set or aquaglyceroporins). This appears to represent an ancient phylogenetic divergence between glycerol transporters and pure water channels (185). Recently, it has become clear that transport properties are even more diverse, since AQP6 has been demonstrated to conduct anions as well (263), and it has also been demonstrated that aquaporins can be regulated by gating, as discussed below.

AQP1 is generally believed to be a constitutively active, water-selective pore. Nonetheless, some observations contradict this. A small degree of permeation by glycerol has been seen in AQP1 oocytes, which may represent opening or presence of a leak pathway (1), but the biological significance of this remains unclear (160). Yool et al. (266) reported that forskolin treatment of oocytes expressing AQP1 induces a cation current, indicating the presence of cation permeation. However, several other research groups were unable to reproduce this effect (5).

Yasui et al. (263) recently demonstrated that AQP6 expressed in oocytes displays unique biophysical properties. It has a low, but significant, basal water permeability. Treatment of AQP6 oocytes with mercurials dramatically increases the osmotic water permeability but interestingly also induces membrane conductance for other small anions. Thus AQP6 can function as an anion channel. Moreover, it was demonstrated that reducing the pH to <5.5 resulted in a marked increase in anion conductance revealing that changes in pH may provide a physiological mediator of these effects. The physiological role of anion conductance remains to be established, but it is suspected that this may be involved in vesicle acidification, since AQP6 is expressed only on intracellular vesicles.

The "aquaporin of the toad bladder" may allow the passage of protons. Gluck and Al-Awqati (86) showed an increase in proton permeability that occurred in parallel with the ADH-induced increase in water permeability of this tissue. This proton permeability was blocked by methohexital, an inhibitor of water permeability, but not by inhibitors of urea or sodium permeation. The authors concluded that the protons were crossing through the water channels, but direct proof of this remains elusive.

Recently, Yool and colleagues (10) provided evidence that an additional water channel (viz AQP1) can also provide a pathway for anions. Using two-electrode voltage-clamp analyses, the authors showed activation of an ionic conductance in AQP1-expressing oocytes after direct injection of cGMP. Current activation was not observed in control (water-injected) oocytes or in AQP5-expressing oocytes with osmotic water permeabilities equivalent to those seen with AQP1. Patch-clamp recordings revealed large-conductance channels in excised patches from AQP1-expressing oocytes after the application of cGMP to the internal side. These results indicate that AQP1 channels have the capacity to participate in ionic signaling after the activation of cGMP second messenger pathways. These observations were qualitatively confirmed using purified AQP1 reconstituted into planar bilayers (208a); however, the stoichiometry (only one ion channel per 10⁷ AQP1 molecules) and permeation through a pathway distinct from the aqueous pore raise serious doubts about possible physiological significance.

Recently, the possible permeation of small gases through aquaporins has been investigated using CO₂. The rates of pH change are ~40% higher in oocytes expressing AQP1 (166) than in control oocytes. As described in a follow-up paper (41), this observation is consistent with several interpretations. The authors found that expressing the AQP1 in Xenopus oocytes increases the CO₂ permeability of oocytes in an expression-dependent fashion, whereas expressing the K⁺ channel ROMK1 has no effect. The mercurial p-chloromercuribenzenesulfonate (PCMBS), which inhibits the water conductance through AQP1, also blocks the AQP1-dependent increase in CO₂ permeability. The mercury-insensitive C189S mutant of AQP1 increases the CO₂ permeability of the oocyte to the same extent as does the wild-type channel, but the C189S-dependent increase in CO₂ permeability is unaffected by treatment with PCMBS. These results therefore suggest that CO₂ passes through the same pore in AQP1 as water does. It should be mentioned that experiments on the CO₂ permeability of the erythrocytes and lung cells from AQP1 null mice and wild-type mice revealed no differences, arguing against a major CO₂ permeability through AQP1 (260). The potential physiological relevance of AQP1 permeation by gases warrants more study (199). Thus, although the evidence that AQP1 functions as a water channel is incontrovertible, the possibility of yet undiscovered transport functions cannot be excluded.

Whereas some previously cloned aquaporins are only permeable to water, some homologs are permeated by water, glycerol, and other small solutes. AQP3 was noted to be genetically closer to the E. coli glycerol transport protein GlpF. The structural explanation for how AQP3 may permit transport of water and glycerol is debated (60, 99, 160); however, the atomic structure of GlpF now provides a clear explanation (78a). Multiple other aquaglyceroporins are now being identified by cDNA cloning and computer search of expressed sequence tagged (EST) cDNA libraries. A cDNA encoding AQP7 was isolated from rat testis (100, 218) and AQP9 from liver (97, 122, 123, 225, 226). These homologs are also thought to function as glycerol transporters as well as water channels.

Early, but indirect, evidence for the gating of water channels came from studies looking at the effect of pH on toad bladder water permeability. Lowering the pH both inhibited the increase in water permeability induced by ADH and reversed the increase after it had been established. These changes were also seen after stimulation with cAMP instead of ADH, indicating that the effect of pH was at a step distal to the activation of adenylate cyclase. This correlated with the prevention of the appearance of particle aggregates thought to represent the sites of water flow in the apical plasma membrane, or their disappearance in an established response. However, at low temperatures, when removal of the aggregates from the plasma membrane was slowed, lowering the pH led to a fall in water permeability without a parallel fall in aggregate area, suggesting that the channels could be closed. Furthermore, raising the pH led to a rapid restoration of water flow, consistent with the channels reopening (23, 182-184). Thus pH appeared to affect both the trafficking of the water channels and the permeability of the channels themselves. These effects could be caused by lowering either mucosal or serosal pH, but only if a buffer system, such as bicarbonate/CO₂, which could modify cytosolic pH, was used. This implies that it is the cytosolic pH that is responsible for the altered channel permeability, whereas extracellular pH has no effect.

A breakthrough discovery came in a recent report by Agre and associates on AQP6 (264), a water channel residing in intracellular vesicles in kidney collecting duct intercalated cells (263). When expressed in X. laevis oocytes, AQP6 exhibits low basal water permeability (P_f), but (surprisingly) when treated with mercurials, known inhibitors of many aquaporin water channels, the osmotic water permeability of AQP6 oocytes rapidly rises ~10-fold (and is accompanied by ion conductance, as described above). Because AQP6 is present in intercalated cells, which are engaged in renal acid/base regulation, it was tested if pH may be a physiological parameter that may alter the conductance. At pH <5.5, water permeability and anion conductance are rapidly and reversibly activated in AQP6-expressing oocytes. Site-directed mutation of lysine to glutamate at position 72 in the cytoplasmic mouth of the pore changes the cation/anion selectivity but leaves low pH activation intact. These studies demonstrated unprecedented biophysical properties of an aquaporin, viz regulation by gating, and in addition provided evidence for anion channel function.

Evidence has also appeared that additional aquaporins, such as AQP3, are also regulated by gating in response to pH changes (270).

	V. AQUAPORINS IN KIDNEY

Top Previous Next References

Absorption of water out of the renal tubule depends on the osmotic driving force for water reabsorption and the equilibration of water across the tubular epithelium (119). The driving force is established, at least in part, by active Na⁺ transport. Moreover, the generation of a hypertonic medullary interstitium results as a consequence of countercurrent multiplication. This requires active transport and low water permeability in some kidney tubule segments, whereas in other segments there is a need for high water permeability (either constitutive or regulated). A series of studies over the past 10 years has made it clear that osmotic water transport across the tubule epithelium is chiefly dependent on aquaporin water channels.

At least seven aquaporins are expressed in the kidney (Table 1). AQP1 (46, 191, 192) is present in the proximal tubule and in the descending thin limb (Fig. 2) but is absent in highly water-impermeable segments of ascending thin limb, thick ascending limb, and distal tubule (175). Moreover, it is absent from the collecting duct, making it clear that AQP1 does not participate in the regulation by vasopressin of renal water excretion (175). Subsequently, additional aquaporins were identified by homology cloning approaches. These include AQP2 (82), which is abundant in kidney collecting duct principal cells (Fig. 2), and this water channel has been found to be the chief target by which vasopressin regulates kidney collecting duct water permeability and hence renal regulation of body water balance (121, 169, 171). Two additional aquaporins are expressed in the collecting duct principal cells, AQP3 and AQP4 (59, 60, 73, 99, 137, 223, 224, 261). Both of these channels are present in the basolateral plasma membranes and are likely to represent exit pathways for water entering the cells through AQP2 (Fig. 2). AQP5 is not present in the kidney as determined by survey screening methods (immunohistochemistry). AQP6 is present in collecting duct intercalated cells type A (264). At the subcellular level, AQP6 is unique compared with the early aquaporins (AQP1 to -5) in that it is exclusively present in an intracellular location with no expression in the plasma membrane. The physiological role of AQP6 is undefined. However, as discussed above, its location and anion permeability have suggested a role in the acidification of vesicles. Several further new aquaporins have recently been identified, presumably by homology cloning approaches or as a consequence of the human genome project. AQP7 appears to be expressed in the brush border of the proximal tubule, but little is known about its role in the kidney. AQP8 through AQP12 have also been identified (see sect. VE for references), and some of these homologs appear by Northern blot or RT-PCR survey to be expressed in kidney although at low abundance. Their functional role remains to be determined.

View larger version (25K): [in this window] [in a new window]   Fig. 2. Diagrammatic representation of the localization of different aquaporins in the nephron and collecting duct system. AQP1 (blue) is present in the proximal tubule and descending thin limb. AQP2 (green) is abundant in the apical and subapical part of collecting duct principal cells, whereas AQP3 (red) and AQP4 (purple) are both present in the basolateral plasma membrane of collecting duct principal cells. AQP7 (orange) is confined to the apical brush border of straight proximal tubules. ADH, antidiuretic hormone.

The archetypal member of the aquaporin family, AQP1 (46, 191, 213), is highly abundant in the proximal tubule (Figs. 3 and 4) and descending thin limb (Figs. 3 and 4) (175, 176), and it constitutes almost 1% of total membrane protein in kidney cortex (175). AQP1 is absent in other nephron segments and absent from the collecting duct (175) and is thus exclusively expressed in segments of the kidney nephron that are constitutively, and highly, water permeable and is not involved in the vasopressin regulation of kidney water transport. Immunoelectron microscopic analysis has documented that AQP1 is abundant in both apical and basolateral plasma membranes in proximal tubules and descending thin limbs (Fig. 4), consistent with a role for AQP1 in the movement of water across both surfaces of the cells (175).

View larger version (126K): [in this window] [in a new window]   Fig. 3. Immunoperoxidase labeling of AQP1 in thin cryosections from kidney cortex (A) and inner stripe of the outer medulla (B). A: extensive labeling is seen in both of the apical plasma membrane domains and in the basolateral plasma membrane domains (arrows) in proximal tubule cells. Arrows point to basolateral plasma membrane domains, and arrowheads point to apical plasma membranes. Inset: apical and basolateral AQP1 labeling of multiple cross-sectioned proximal tubules. B: in descending thin limbs, both apical and basolateral plasma membrane domains are labeled. Arrows point to basolateral plasma membrane domains, and arrowheads point to apical plasma membranes. Asterisks indicate vascular structures, and C indicates collecting duct. Magnification: ×1,000.

View larger version (150K): [in this window] [in a new window]   Fig. 4. Immunoelectron microscopical localization of AQP1 in segment 3 proximal tubule cell (A) and descending thin limb cell (B) (cryosubstituted and low-temperature Lowicryl HM20 embedded tissue). A: AQP1 is extremely abundant in the apical plasma membrane of the brush border (BB). M, mitochondrion. B: in descending thin limb (DTL) cell, very strong labeling is seen in both the apical and basolateral plasma membrane (BM). L, lumen. Magnification: ×54,000.

Immunocytochemistry (174, 175, 204) revealed that AQP1 immunolabeling is especially abundant in the segment 3 proximal tubule, although high levels are also present in segments 1 and 2. This is the case both in rat kidney (see above) and in human kidney (159). Importantly, there is an axial heterogeneity and segment-to-segment difference in the expression levels of AQP1 in the descending thin limb segments (174, 175). As described below, this heterogeneity parallels closely the water permeability characteristics in the different segments, as established using isolated perfused descending thin limbs (28, 29, 32). AQP1 is extraordinarily abundant in type II descending thin limb epithelium (174, 175), which is also known to display an extremely high osmotic water permeability (32). Type II descending thin limb epithelium (present in the inner stripe of the outer medulla from long looped nephrons) continues into type III descending thin limb epithelium, which has a much simpler ultrastructure and has a lower osmotic water permeability (albeit still very high in absolute terms) than the type II epithelium. The labeling density of AQP1 in this segment (DTL III) is lower than in type II, consistent with the lower water permeability. Finally, type I epithelium, found in descending thin limbs from short loops, has a very simplified epithelium and a much lower osmotic water permeability than type II and type III (28, 94). Consistent with this, the AQP1 labeling is less intense than in type II and type III. The short loop descending limb itself appears heterogeneous with greater labeling in the early part than in the late part (242). Thus the abundance of AQP1 parallels the differences in the osmotic water permeability, consistent with the hypothesis that AQP1 is providing the main route for water movement across the tubule wall.

The absolute abundance of AQP1 in individual microdissected renal tubule segments of rat has been measured with a fluorescence-based enzyme-linked immunosorbant assay (ELISA) standardized using AQP1 purified from red blood cells (144). In general, these measurements confirmed the very high abundance of AQP1 in proximal tubules and descending limbs previously inferred from immunocytochemical observations. In proximal tubule segments, AQP1 levels (in molecules/mm tubule length × 10⁹) averaged the following: S-1, 6.5; S-2, 6.0; and S-3, 12.8. In descending limb segments, the values in the same units were as follows: type I, 7.8; type II, 52.1; and type III, 25.9.

The biophysical properties, the overall extraordinary abundance of AQP1 in the proximal tubule and descending thin limb, combined with it expression levels in different segments of the descending thin limb, strongly supported the view that AQP1 is essential for proximal nephron water handling and urinary concentration (119, 171). The critical role of AQP1 in urinary concentration was recently confirmed in humans lacking AQP1 and in transgenic knockout mice lacking AQP1.

Humans have been identified who totally lack the AQP1 protein (195). The human AQP1 gene was localized to chromosome 7p14, and the Co blood group antigens were previously linked to 7p, suggesting a molecular relationship. It was established that the Co blood group antigen results from an Ala/Val polymorphism of AQP1 (214). Only six individuals have been shown to lack Co, and most of these Co null individuals are women who developed anti-Co during pregnancy. Three of these Co null individuals were found to have mutations in the AQP1 gene (195). Recent studies have indicated that these Co null individuals have a urinary concentrating defect in response to vasopressin or water deprivation (116a).

The AQP1-deficient mice were polyuric (140) and were unable to concentrate urine to more than ~700 mosmol/kgH₂O even in response to water deprivation, during which they become rapidly dehydrated; plasma osmolalities increased dramatically up to 400-500 mosmol/kgH₂O. Thus AQP1 is required for the formation of a concentrated urine. It is presumed that lack of AQP1 undermines the countercurrent multiplication process, which depends on the rapid equilibration of water across the descending thin limb of Henle's loop. Subsequent studies have demonstrated that the osmotic water permeability of perfused proximal tubules isolated from AQP1 knockout mice were only one-fifth of the permeabilities in proximal tubules dissected from kidneys of normal mice (210). Recently, Chou et al. (30) also demonstrated that the osmotic water permeability of descending thin limb (dissected from kidneys of AQP1-deficient animals) is reduced by 90%. These studies in AQP1-deficient mice and the subsequent studies in AQP1-deficient humans not only indicate a major importance of AQP1 for water reabsorption in the proximal nephron, but also provide strong evidence that the major pathway for water reabsorption in the proximal tubule and descending thin limbs is transcellular (via AQP1) and not paracellular. Additional support for a critical role of AQP1 for osmotic equilibration across the tubular epithelium came from free-flow micropuncture experiments in AQP1-deficient mice by Schnermann and colleagues (231). They determined that the osmolality difference between the plasma and the kidney proximal tubule luminal fluid is excessively increased in the AQP1-deficient mice compared with control mice, strongly supporting the view that AQP1 is important for efficient water transport across the tubular epithelium (231).

In addition to its presence in the proximal tubule and descending thin limb, AQP1 is also expressed in the descending vasa recta of rat kidney but not in the ascending vasa recta (174). Functional studies were performed using isolated perfused descending vasa recta briefly fixed with glutaraldehyde. This fixation was performed to eliminate the cellular toxicity of mercurials, and it has been shown in collecting duct (124), toad urinary bladders (241), and peritoneum (19) that fixation at least partially maintains the osmotic water permeability. The fixed descending vasa recta displayed a marked sensitivity for mercurials on the transendothelial water permeability, confirming that AQP1 at this site may play a significant role for water transport. This was later confirmed by Pallone et al. (180) using isolated perfused descending vasa recta from wild-type mice and from mice lacking AQP1. Water equilibration across the vasa recta may also play a critical role in avoiding the disruption of the osmotic gradient established by the countercurrent exchanger. It should also be mentioned that the demonstrated role of aquaporins, including AQP1 for peritoneal transport of water in experimental peritoneal dialysis (19), was also later confirmed in AQP1-deficient mice (259).

AQP2 (82) is abundant in the apical plasma membrane and apical vesicles in the collecting duct principal cells (169) (Figs. 5 and 6) and at lower abundance in connecting tubules (117, 135). In addition to its presence in the apical plasma membrane and intracellular vesicles (Fig. 6), some AQP2 immunostaining has also been found to be associated with the basolateral plasma membrane, especially in the inner medullary collecting duct principal cell (151, 169) (Fig. 5C).

View larger version (48K): [in this window] [in a new window]   Fig. 5. Immunofluorescence microscopy of AQP2 in cortical (A), outer medullary (B), and inner medullary (C and D) collecting duct and of AQP3 (E) and AQP4 (F) in inner medullary collecting duct. AQP2 is very abundant in the apical plasma membrane domains as well as in subapical domains (arrows in A, B, and D), whereas intercalated cells are unlabeled (arrowheads in A and B). In the inner medullary collecting duct, AQP2 is also present in the basolateral part of the cell. AQP3 is abundant in both basal and lateral plasma membranes, whereas AQP4 is predominantly expressed in the basal plasma membrane and less prominently in the lateral plasma membranes. Magnification: ×1,100 (A, B, D-F) and ×550 (C). [From Nielsen et al. (171).]

View larger version (113K): [in this window] [in a new window]   Fig. 6. Immunoelectron microscopical labeling of AQP2 in ultrathin cryosection of inner medullary collecting duct principal cell. AQP2 is abundant both in the apical plasma membrane (arrows) and in small subapical vesicles (arrowheads). Magnification: ×70,000.

AQP2 is the primary target for vasopressin regulation of collecting duct water permeability. This conclusion was solidly established in studies showing 1) the cellular and subcellular distribution (82, 169), 2) a direct correlation between AQP2 expression and collecting duct water permeability in rats (50), 3) a direct correlation between the osmotic water permeability and AQP2 levels in the apical plasma membrane of collecting duct principal cells in isolated perfused collecting ducts (168) and in whole animal experiments (only of the onset phase, Refs. 151, 203, 256), and 4) in studies demonstrating that humans with mutations in the AQP2 gene (45) or rats with 95% reduction in AQP2 expression (149) have profound nephrogenic diabetes insipidus. AQP2 regulation is described in detail below.

AQP3 and AQP4, which are expressed in cells in a wide range of organs (see, for example, Refs. 73, 74, 88, 158, 165, 170, 173, 228), are also present in the collecting duct principal cells (59, 60, 73, 99, 137, 223, 224, 261) where they are abundant in the basolateral plasma membranes (Fig. 5, D and E) and represent potential exit pathways from the cell for water entering via AQP2. There is some heterogeneity in the segmental and subcellular localization of these two aquaporins in kidney collecting duct. AQP3 is very abundant in connecting tubule as well as cortical, outer medullary, and inner medullary collecting duct (59, 223). In contrast, AQP4 is mainly abundant in the inner medulla, although some expression is also noted in the more proximal segments (223). At the subcellular level, AQP3 is abundant in both the basal and in the lateral plasma membranes of collecting duct principal cells. In contrast, AQP4 is mainly present in the basal plasma membrane of collecting duct principal cells in rat. In mice, AQP4 was also present in the collecting duct principal cell, but interestingly also in basolateral membranes of proximal tubule S3 segments (233). This was corroborated by freeze-fracture electron microscopy, revealing orthogonal arrays of intramembrane particles (OAPs; known to represent AQP4) on the basolateral membranes of the S3 segment in normal mice but lack of AQP4 immunostaining and OAPs in collecting duct and proximal tubule in AQP4 knockout mice. Neither AQP3 nor AQP4 is found in significant amounts in cytoplasmic vesicles, and there is no evidence for a rapid increase in labeling of the plasma membrane in response to vasopressin.

A number of studies have also focused on the issue of whether the expression of AQP3 and AQP4 are regulated and whether this correlates with changes in renal water balance regulation. In brief, it has been demonstrated that the expression of AQP3 is regulated by changes in vasopressin levels (224). Brattleboro rats, which lack endogenous vasopressin, have lower expression levels of AQP3 than do Long Evans rats (the parent strain of the Brattleboros, which have normal vasopressin expression). Long-term vasopressin treatment of Brattleboro rats (using implantable osmotic minipumps to deliver vasopressin for 5 days) results in a significant increase in AQP3 expression, in parallel with an increase in AQP2 expression. Thirsting of normal rats is also associated with increased AQP3 expression (224). In contrast, AQP4 appears not be regulated or at least not as markedly as AQP2 and AQP3. Long-term vasopressin infusion did not affect AQP4 expression.

To obtain further information about the physiological role of AQP3 and AQP4, gene knockout mice have been produced (31, 138, 139, 240). Although such studies are complicated by potential interfering compensatory mechanisms during fetal and postnatal development, the use of these gene knockout mice by a variety of research groups has been informative. Transgenic knockout mice lacking AQP4 showed a mild urinary concentrating defect (139), and studies using isolated perfused collecting ducts from inner medulla (IMCDs) from AQP4-lacking mice revealed a fourfold reduction in the vasopressin-stimulated osmotic water permeability (31). This indicates that AQP4 is responsible for a substantial majority of the basolateral membrane water movement in IMCDs under maximal vasopressin stimulation. The lower abundance of AQP4, together with higher abundance of AQP3 in cortical and outer medullary collecting duct, raises the possibility that AQP3 may play a more significant role at these more proximal segments of the collecting duct. Recently, AQP3 knockout mice were produced, and they revealed a marked urinary concentrating defect with very severe polyuria (138). After deamino-8-D-arginine vasopressin (dDAVP) administration or water deprivation, the AQP3-deficient mice were able to concentrate their urine partially to ~30% of that in wild-type mice. Osmotic water permeability of dissected, nonperfused cortical collecting duct measured by an optical method was reduced in IMCDs from AQP3 knockout mice. One complicating factor is that in the AQP3 knockout mice AQP2 expression in the cortical collecting duct is also reduced extensively, raising the possibility that part of the polyuria in these AQP3 knockout mice may be caused by the reduced expression of AQP2.

Three further aquaporin cDNAs have been identified in kidney. One of these is AQP6. Initially a cDNA from rat kidney (WCH3) and the human homolog (hKID) was identified, and when expressing these in X. laevis oocytes they increased the osmotic water permeability slightly, but unsuccessful attempts to raise antibodies led the authors to conclude that the protein is not immunogenic (136, 141). The International Human Genome Nomenclature Committee adopted the name aquaporin (symbol AQP) as the referencing system (8). AQP6 was designated for rat WCH3 and human hKID (3), although the cellular locations were not established at the time. While searching for new aquaporins in kidney by PCR, a rat cDNA clone was isolated encoding a water channel protein (AQP6), which is closely related to rat WCH3 and human hKID. Production of antibodies against AQP6 allowed the definition of the cellular and subcellular localization of AQP6 in rat kidney. AQP6 was expressed in collecting duct intercalated cells in cortical, outer medullary, and inner medullary collecting duct (264). The presence in collecting duct intercalated cells in the outer medulla and inner medulla allowed the conclusion that AQP6 was present in type A intercalated cells and in the type A-like intercalated cell found in the inner medullary collecting duct. A major surprise was that AQP6 was almost exclusively found associated with intracellular vesicles in the cells described above, with an almost complete absence of AQP6 immunogold labeling in the plasma membranes. Thus AQP6 appears, at least in rat kidney, to be an exclusively intracellular water (and ion) channel.

AQP7 is highly abundant in spermatocytes (96, 100, 218), but it may also be present in other tissues. Preliminary studies using antibodies to rat or mouse AQP7 have indicated that AQP7 is expressed in the proximal tubule brush border especially in the segment 3 proximal tubule (Table 1). The cellular and subcellular localization of AQP8 (98, 142) has recently been established and AQP8 is present mainly in intracellular domains of proximal tubule and collecting duct cells (63a) in addition to being present in many other organs (63a, 83a). AQP9 (97, 122, 123, 225, 226) is not expressed in kidney but is abundant in many other tissues. Some of them are expected to be present in the kidney based on RT-PCR analysis, but the exact presence and abundance needs to be clarified using immunolabeling methods.

	VI. RENAL AQUAPORINS AND REGULATION OF BODY WATER BALANCE

Top Previous Next References

Much of the early work on vasopressin action was done in amphibian skin or bladder, which are functional analogs of the kidney collecting duct. Using freeze-fracture electron microscopy, Chevalier et al. (22) identified large clusters of particles, arranged in characteristic arrays (linear parallel grooves), which appeared in the apical plasma membrane of ADH-responsive cells during hormonal stimulation. The number of these so-called "particle aggregates" in the membrane was subsequently shown to correlate with the increase in water permeability of the epithelium under most circumstances (93, 109). In summary, these and subsequent studies in amphibian tissues revealed 1) that membrane turnover was dramatically increased in response to antidiuretic hormone, 2) cytoskeletal inhibitors markedly inhibited the water permeability response to vasopressin, 3) the intramembrane particle clusters or aggregates were found in intracellular structures in the absence of vasopressin stimulation and could be found in so-called fusion structures after vasopressin stimulation, and 4) membrane capacitance measurements revealed an increase in apical plasma membrane area in response to vasopressin stimulation. This led Wade et al. (244) to propose the "membrane shuttle hypothesis," which proposed that water channels were stored in vesicles and inserted exocytically into the apical plasma membrane in response to vasopressin. However, it proved remarkably difficult to produce definitive evidence for this in the absence of a molecular definition of the water channels or in the absence of good water channel blockers or probes/antibodies to the channels.

The identification of the aquaporins and subsequently AQP2, later shown to be the predominant vasopressin-regulated water channel (for recent review, see Ref. 171), made it possible to prepare antibodies and investigate the effects of vasopressin in mammalian collecting ducts directly. As shown in Figure 5, AQP2 is present in the apical and subapical parts of collecting duct principal cells, and immunoelectron microscopy (Fig. 6) showed that AQP2 is very abundant both in the apical plasma membrane and in small subapical vesicles (169).

Water flow out of the collecting duct is determined by the apical plasma membrane of the collecting duct cells (Fig. 7), which provides the rate-limiting barrier to water movement (71). Regulation of the osmotic water permeability of this membrane could, in principle, occur via one or both of two distinct mechanisms: either 1) the permeability of existing channels could be altered, probably by chemical modification (e.g., phosphorylation of the channel), or 2) the number of water channels present in the membrane could be altered by vasopressin or by insertion or removal of channels from an intracellular store, as proposed by the membrane shuttle hypothesis. The presence of AQP2 in small vesicles favored the latter hypothesis, and several in vitro and in vivo studies have now underscored the importance of regulated trafficking of AQP2. In vitro studies using isolated perfused tubules allowed a direct analysis of both the on-set and off-set responses to vasopressin. In this study it was demonstrated that changes in AQP2 labeling density of the apical plasma membrane correlated closely with the water permeability in the same tubules, while there were reciprocal changes in the intracellular labeling for AQP2 (Fig. 8) (168). In vivo studies using normal rats or vasopressin-deficient Brattleboro rats also showed a marked increase in apical plasma membrane labeling of AQP2 in response to vasopressin or dDAVP treatment (151, 203, 256). The off-set response has been examined in vivo using acute treatment of rats with vasopressin V₂-receptor antagonist (35, 91) or acute water loading (to reduce endogenous vasopressin levels, Ref. 206). These treatments (both reducing vasopressin action) resulted in a prominent internalization of AQP2 from the apical plasma membrane to small intracellular vesicles further underscoring the role of AQP2 trafficking in the regulation of collecting duct water permeability.

View larger version (18K): [in this window] [in a new window]   Fig. 7. Changes in the osmotic water permeability in isolated perfused inner medullary collecting ducts in the presence of arginine vasopressin (AVP) for 40 min (middle portion, AVP) or before and after AVP stimulation. Osmotic water permeability is significantly increased with AVP stimulation.

View larger version (87K): [in this window] [in a new window]   Fig. 8. Immunoelectron microscopic labeling of AQP2 in isolated perfused inner medullary collecting ducts perfused in the presence of AVP for 40 min (top panel, AVP) or in the absence of AVP stimulation (middle panel, Pre-AVP). Bottom panels display the AQP2 labeling density of AQP2 in isolated perfused tubules perfused in absence of AVP stimulation, with AVP stimulation or after AVP stimulation (washout). AQP2 labeling density increased significantly in response to AVP treatment (bottom left panel), and this is parallelled by a similar increase in the osmotic water permeability (bottom right panel). Magnification: ×70,000.

With respect to the alternative or additional mode of regulation viz. regulation of conductance, two studies have attempted to address this issue. Kuwahara et al. (126) demonstrated that protein kinase A (PKA)-induced phosphorylation of AQP2 in Xenopus oocytes was associated with only a very small (~30%) increase in water permeability (126). Because vasopressin causes much greater changes in collecting duct water permeability (3- to 10-fold increase), this demonstrated that changes in water conductance of AQP2 by PKA-mediated phosphorylation could not explain more than a small fraction of the hormonal effect on the collecting duct. Consistent with this, Lande et al. (130) found that treatment with PKA caused no change in the water permeability of AQP2-bearing vesicles harvested from kidney inner medulla (130). Thus the major changes in the subcellular distribution of AQP2 in response to vasopressin or vasopressin receptor antagonist treatment strongly support the view that collecting duct water permeability, and hence body water balance, is acutely regulated primarily by vasopressin-regulated trafficking of AQP2.

Several groups have now successfully reconstituted the AQP2 delivery system using cultured cells transfected with either wild-type AQP2 or AQP2 tagged with a marker protein or a fluorescent protein (43, 114-116, 229). Using such cultured cells stably transfected with AQP2, the authors have shown shuttling of AQP2 from vesicles to the plasma membrane, albeit in some cases to the basolateral membrane, as well as retrieval and subsequent trafficking back to the surface upon repeated stimulation. This recycling of AQP2 also occurs in LLC-PK₁ cells in the continued presence of cycloheximide, preventing de novo AQP2 synthesis. Whether repeated trafficking and recycling also occurs in the native tissue remains to be established.

The signal transduction pathways have been described thoroughly in previous reviews (e.g., Ref. 121). cAMP levels in collecting duct principal cells are increased by binding of vasopressin to V₂ receptors (62, 125). The synthesis of cAMP by adenylate cyclase is stimulated by a V₂ receptor-coupled heterotrimeric GTP-binding protein, G_s. G_s interconverts between an inactive GDP-bound form and an active GTP-bound form, and the vasopressin-V₂ receptor complex catalyzes the exchange of GTP for bound GDP on the -subunit of G_s. This causes release of the -subunit, G_s-GTP, which subsequently binds to adenylate cyclase, thereby increasing cAMP production (Fig. 9A). PKA is a multimeric protein that is activated by cAMP and consists in its inactive state of two catalytic subunits and two regulatory subunits. When cAMP binds to the regulatory subunits, these dissociate from the catalytic subunits, resulting in activation of the kinase activity of the catalytic subunits.

View larger version (52K): [in this window] [in a new window]   Fig. 9. Signaling cascades and molecular apparatus involved in vasopressin regulation of AQP2 trafficking. See text for details. A: signaling cascades involving vasopressin V2 receptors (V2R), adenylyl cyclase (AC), cAMP, and protein kinase A (PKA). B: protein kinase A phosphorylates AQP2 monomers in Ser-256 of AQP2. Phosphorylation of 3 or 4 monomers of the homotetramers is necessary for trafficking to the apical plasma membrane. C: other kinases are likely also to play a role in regulation of AQP2 trafficking. D: cytoskeletal elements are involved in AQP2 trafficking and include microtubule-based motor proteins (dynein and dynactin) as well as actin filament binding proteins such as myosin-1. E: the docking mechanism may involve vesicle targeting receptors that have been found in the collecting duct cells associated with membrane domains housing AQP2. This includes VAMP2 and syntaxin-4. F: regulation of intracellular calcium is involved in AQP2 trafficking and appears to involve a calcium sensing receptor.

Early studies demonstrated that PKA induces phosphorylation of various membrane proteins in bovine kidney (52) and that vasopressin treatment of saponin-permeabilized outer medullary collecting duct segments induced phosphorylation of at least two 45- and 66-kDa proteins (83). It has also been shown that inhibition of PKA activity with H-89 inhibits the vasopressin-induced increase in water permeability in isolated perfused rabbit collecting ducts (215). AQP2 contains a consensus site for PKA phosphorylation (RRQS) in the cytoplasmic COOH terminus at serine-256 (82). Recent studies using ³²P labeling or using an antibody specific for phosphorylated AQP2 (see below) showed a very rapid phosphorylation of AQP2 (within 1 min) in response to vasopressin treatment of slices of the kidney papilla (178). This agrees well with the time course of vasopressin-stimulated water permeability of kidney collecting ducts (245). As described above, PKA-induced phosphorylation of AQP2 apparently does not change the water conductance of AQP2 significantly. Importantly, it was recently demonstrated that vasopressin or forskolin treatment failed to induce translocation of AQP2 when serine-256 was substituted by an alanine (S256A) in contrast to a significant regulated trafficking of wild-type AQP2 in LLC-PK₁ cells (115). A parallel study by Sasaki and colleagues (81) also demonstrated the lack of cAMP-mediated exocytosis of mutated (S256A) AQP2 transfected into LLC-PK₁ cells. Thus these studies indicate a specific role of PKA-induced phosphorylation of AQP2 for regulated trafficking. To explore this further, an antibody was designed that exclusively recognizes AQP2, which is phosphorylated in the PKA consensus site (serine-256). In normal rats, phosphorylated AQP2 is present in both intracellular vesicles and in apical plasma membranes, whereas in Brattleboro rats, phosphorylated AQP2 is mainly located in intracellular vesicles as shown by immunocytochemistry and immunoblotting using membrane fractionation (37). Surprisingly, there was no significant increase in the total amount of phosphorylated AQP2 after vasopressin treatment in normal rats (37). However, recent evidence that at least three of the subunits in each tetramer need to be phosphorylated to be transported to the plasma membrane (Fig. 9B) (111) may explain this; if there is a low level of phosphorylation and dephosphorylation occurring continuously, such that many tetramers are close to the threshold for insertion, only a very modest increase in phosphorylation may be needed to induce substantial transfer to the plasma membrane. Moreover, dDAVP treatment of Brattleboro rats caused a marked redistribution of phosphorylated AQP2 to the apical plasma membrane, which is in agreement with an important role of PKA phosphorylation in this trafficking (37). Conversely, treatment with V₂-receptor antagonist induced a marked decrease in expression of phosphorylated AQP2 (36) likely to be due to reduced PKA activity and/or increased dephosphorylation of AQP2, e.g., by increased phosphatase activity.

PGE₂ inhibits vasopressin-induced water permeability by reducing cAMP levels (for detailed review, please see Ref. 121). In preliminary studies, Zelenina et al. (269) investigated the effect of PGE₂ on PKA phosphorylation of AQP2 in kidney papilla, and the results suggest that the action of prostaglandins is associated with retrieval of AQP2 from the plasma membrane, but that this appears to be independent of AQP2 phosphorylation by PKA.

Phosphorylation of AQP2 by other kinases, e.g., protein kinase C or casein kinase II, may potentially participate in regulation of AQP2 trafficking (Fig. 9C). Phosphorylation of other cytoplasmic or vesicular regulatory proteins may also be involved. These issues remain to be investigated directly.

Since the fundamentals of the shuttle hypothesis have been confirmed, interest has turned to the cellular mechanisms mediating the vasopressin-induced transfer of AQP2 to the apical plasma membrane. The shuttle hypothesis has a number of features whose molecular basis remains poorly understood. First, AQP2 is delivered in a relatively rapid and coordinated fashion, and vesicles move from a distribution throughout the cell to the apical region of the cell in response to vasopressin stimulation. Furthermore, AQP2 is delivered specifically to the apical plasma membrane. Finally, AQP2-bearing vesicles fuse with the apical plasma membrane in response to vasopressin, but not to a significant degree in the absence of stimulation (e.g., in vasopressin-deficient Brattleboro rats where <5% of total AQP2 is present in the apical plasma membrane; Refs. 87, 203, 256). Thus there must be some kind of a "clamp" preventing fusion in the unstimulated state and/or a "trigger" when activation occurs.

The coordinated delivery of AQP2-bearing vesicles to the apical part of the cell appears to depend on the translocation of the vesicles along the cytoskeletal elements. In particular, the microtubular network has been implicated in this process, since chemical disruption of microtubules inhibits the increase in permeability both in the toad bladder and in the mammalian collecting duct (189, 190). Because microtubule-disruptive agents inhibit the development of the hydrosmotic response to vasopressin, but have no effect on the maintenance of an established response, and because they have been reported to slow the development of the response without affecting the final permeability in toad bladders (230), it has been deduced that microtubules appear to be involved in the coordinated delivery of water channels, without being involved in the actual insertion process, or in recycling of water channels. Presumably, the processes in the kidney collecting duct are similar.

Microtubules are polar structures, arising from microtubule organizing centers (MTOCs), at which their minus ends are anchored, and with the plus ends growing away "into" the cell. In fibroblastic cells, there is a single MTOC in the perinuclear region, and the plus ends project to the periphery of the cell. However, there is increasing evidence that in polarized epithelia microtubules arise from multiple MTOCs in the apical region, with their plus ends projecting down toward the basolateral membrane (2). If this is the case in collecting duct cells, and there is some evidence that it is (153), then a minus end-directed motor protein such as dynein would be expected to be involved in the movement of vesicles toward the apical plasma membrane. Recently, it has been shown that dynein is present in the kidney of several mammalian species (for references, see Ref. 152) and that both dynein and dynactin, a protein complex believed to mediate the interaction of dynein with vesicles, associate with AQP2-bearing vesicles (152). Furthermore, both vanadate, a rather nonspecific inhibitor of ATPases, and erythro-9-(2-hydroxy-3-nonyl)adenine (EHNA), a relatively specific inhibitor of dynein, inhibit the antidiuretic response in toad bladder (47, 147). Thus it seems likely that dynein may drive the microtubule-dependent delivery of AQP2-bearing vesicles toward the apical plasma membrane (Fig. 9D).

The apical part of the collecting duct principal cells contains a prominent terminal web made up of actin filaments. These also appear to be involved in the hydrosmotic response, since disruption of microfilaments with cytochalasins inhibits the response in the toad bladder (108, 186, 241). Cytochalasins can also inhibit an established response, and even the offset of the response (162). From this it has been concluded that microfilaments are probably involved in the final movement of vesicles through the terminal web, their fusion with the plasma membrane, and the subsequent endocytic retrieval of the water channels (187). Interestingly, vasopressin itself causes actin depolymerization (51), suggesting that reorganization of the terminal web is an important part of the cellular response to vasopressin, a conclusion reached on morphological grounds by DiBona (49).

The problem of delivering vesicles to a particular domain and allowing them to fuse when, and only when, a signal arrives is conceptually very similar to the situation in the neuronal synapse. It therefore seemed possible that a molecular apparatus similar to the SNAP/SNARE system described there (201) might be present in the collecting duct principal cells. This hypothesis postulates that there are specific proteins on the vesicles (vSNAREs) and the target plasma membrane (tSNAREs) that interact with components of a fusion complex to induce fusion of the vesicles only with the required target membrane. The process is thought to be regulated by other protein components that sense the signal for fusion (i.e., increased calcium in the synapse). Several groups have now shown that vSNAREs such as VAMP-2 are present in the collecting duct principal cells and colocalize with AQP2 in the same vesicles (72, 101, 172). tSNAREs are also present. Syntaxin 4, but not syntaxins 2 or 3, is present in the apical plasma membrane of collecting duct principal cells (145, 146). Another putative tSNARE, SNAP23, has also been found in collecting duct principal cells both in the apical plasma membrane and in AQP2-bearing vesicles (95). Some soluble components of the fusion complex, including NEM-sensitive factor (NSF) and -soluble NSF-associated protein (SNAP), have also been identified in these cells. Thus it seems likely that the exocytic insertion of AQP2 is indeed controlled by a set of proteins similar to those involved in synaptic transmission (Fig. 9E), although considerable work remains to be done in isolating and characterizing the components, their regulation, and prime physiological function.

In addition to increasing cAMP levels in collecting duct principal cells, vasopressin acting through the V₂ receptor has also been demonstrated to transiently increase intracellular Ca²⁺ (20, 56, 143, 216). The increase occurs in the absence of activation of the phosphoinositide signaling pathway (33) and has recently been demonstrated to be due to activation of ryanodine-sensitive calcium release channels in the collecting duct cells (27). Buffering intracellular calcium with BAPTA or inhibition of calmodulin completely blocked the water permeability response to vasopressin in isolated perfused inner medullary collecting ducts, suggesting a critical role for calcium at some step in the process of AQP2 vesicle trafficking (34).