This review outlines recent advances related to the molecular mechanisms and pathways of aquaporin-2 (AQP2) water channel trafficking. AQP2 is a fascinating protein, whose sorting signals can be interpreted by different cell types to achieve apical or basolateral membrane insertion, in both regulated and constitutive trafficking pathways. In addition to the well-known cAMP-mediated, stimulatory effect of vasopressin on AQP2 membrane insertion, other signaling and trafficking events can also lead to AQP2 membrane accumulation via cAMP-independent mechanisms. These include1) elevation of cGMP, mediated by sodium nitroprusside (a nitric oxide donor), atrial natriuretic factor, andl-arginine (via nitric oxide synthase); 2) disruption of the actin cytoskeleton; and 3) inhibition of the clathrin-mediated endocytotic arm of the AQP2 recycling pathway by dominant-negative dynamin expression and by membrane cholesterol depletion. Recent data also indicate that AQP2 recycles constitutively in epithelial cells, it can be inserted into different membrane domains in different cell types both in vitro and in vivo, and these pathways can be modulated by factors including hypertonicity. The roles of accessory proteins, including small GTPases and solubleN-ethylmaleimide-sensitive factor attachment protein receptor proteins in AQP2 membrane insertion, are also being uncovered. Understanding cAMP-independent mechanisms for membrane insertion of AQP2 is especially relevant to the therapeutic bypassing of the mutated, dysfunctional vasopressin receptor in patients with X-linked nephrogenic diabetes insipidus.
vasopressin stimulation results in the accumulation of aquaporin-2 (AQP2) on the plasma membrane of principal cells in the kidney via a membrane trafficking mechanism that involves the recycling of AQP2 between intracellular vesicles and the cell surface (14, 44, 58). The basic paradigm underlying these cell biological events, which are a crucial part of the urinary concentrating mechanism, was worked out many years ago using amphibian bladder and epidermis as experimental models (4, 73). The so-called “shuttle hypothesis” of vasopressin action was confirmed by direct observations of water channel (aquaporin) recycling when antibodies against the vasopressin-regulated AQP2 became available (18, 42, 43, 56). However, despite several years of investigation in the postaquaporin era, many basic features of vasopressin-regulated AQP2 membrane insertion and retrieval remain unclear. This brief review will summarize selected recent advances that have been made in this active field of research. In some cases, new data have resolved questions and confirmed hypotheses that were first raised over 20 years ago. Other results, however, have forced a reassessment of older paradigms and indicate that the cell-surface expression of AQP2 can be induced by mechanisms other than the traditional vasopressin receptor (V2R) and cAMP signal transduction cascade.
cAMP-INDEPENDENT CELL-SURFACE EXPRESSION OF AQP2
The cGMP Pathway
Activation of PKA by cAMP leads to phosphorylation of AQP2 on a serine residue at position 256 on the cytoplasmic COOH terminus. With the use of a serine 256-to-alanine (S256A) point mutation of AQP2, expressed in LLC-PK1 cells, it was clearly shown that phosphorylation of the serine 256 residue by PKA is required for vasopressin-induced translocation of AQP2 from vesicles to the plasma membrane (17, 28). Vasopressin also stimulates serine 256 phosphorylation of native AQP2 in collecting duct principal cells in situ (46, 77). However, studies of alternative (or parallel) trafficking pathways that do not seem to involve cAMP or PKA are being uncovered. Hormones and drugs that increase cGMP levels (sodium nitroprusside, l-arginine, atrial natriuretic peptide) as well as permeable cGMP analogs also induce AQP2 trafficking in cultured cells and in collecting ducts in vitro (Fig. 1,A and B) (1). PKG can phosphorylate the COOH terminus of AQP2 in vitro, and cGMP-stimulated membrane accumulation of AQP2 does not occur in cells expressing the S256A mutation of AQP2. Figure2 shows the potential interactions of the cAMP-dependent and cGMP-dependent signaling pathways on AQP2 membrane insertion. It is not yet clear whether PKG directly phosphorylates AQP2 in vivo or whether the cGMP/PKG effect is ultimately mediated by activation of PKA. The physiological relevance of this alternative signaling pathway is undetermined, but interestingly the effect is most pronounced in the collecting ducts of the outer medulla. The cGMP effect is barely detectable in the cortical and inner medullary collecting ducts. It is possible that this cGMP-mediated AQP2 insertion may at least partially explain the apparently vasopressin-independent urinary concentration that occurs in some animal models and in humans under certain conditions. These issues were discussed in more detail in a previous report (1).
The Actin Cytoskeleton and Membrane Accumulation of AQP2
A role of the actin cytoskeleton in vasopressin-stimulated water channel insertion was also proposed based on work in toad bladder (15, 50). Subsequently, studies of collecting ducts confirmed these observations (23). However, an unexpected recent finding was that AQP2 translocation can be achieved simply by modulating actin polymerization, in the complete absence of hormonal stimulation. The potential role of actin in AQP2 membrane insertion was directly examined in transfected inner medullary collecting duct (IMCD) cells in culture. Exposure of these cells toClostridium toxin B caused actin depolymerization and accumulation of AQP2 in the plasma membrane (67). This toxin inhibits RhoGTPases that are involved in regulating the polymerization state of the actin cytoskeleton (52). AQP2 translocation was also seen in cells treated with Y-27632, the downstream Rho kinase inhibitor (32). This occurred in the absence of any detectable elevation of intracellular cAMP. Conversely, expression of constitutively active RhoA in these cells induced stress fiber formation, indicating actin polymerization, and inhibited the normal AQP2 translocation response to forskolin. These data provide strong evidence for a major regulatory role of the actin cytoskeleton in the vasopressin-induced recycling of AQP2 between intracellular vesicles and the cell surface. Whether the endocytotic or exocytotic pathways are affected has not yet been determined directly, however (see below). Interestingly, principal cells of both rat and mouse kidney contain high levels of the actin-severing protein adseverin, which may be involved in regulating the actin cytoskeleton in these cells in vivo (34).
APICAL VS. BASOLATERAL AQP2 EXPRESSION IN EPITHELIAL CELLS
Since the original studies on AQP2 localization in kidney collecting ducts, it has been clear that AQP2 can be expressed not only on the apical plasma membrane but also at the basolateral surface of collecting duct principal cells (2, 10, 42). Furthermore, in our original study describing trafficking of AQP2 transfected into LLC-PK1 cells (29), AQP2 was inserted into the basolateral plasma membrane after vasopressin stimulation. In contrast, subsequent studies reported the apical delivery of AQP2 in Madin-Darby canine kidney (MDCK) cells and rabbit collecting duct cells after vasopressin stimulation (12, 68). Remarkably, however, a predominantly basolateral AQP2 expression in response to acute vasopressin stimulation occurred in primary cultures of IMCD cells (9, 37). Thus, while regulated trafficking of AQP2 occurs in a variety of transfected cell lines, the polarity signals on AQP2 (as well as those on some other proteins) seem to be interpreted in different ways by a variety of cell types (21, 51). AQP2 in the vas deferens of the male reproductive tract is inserted constitutively into the apical plasma membrane of epithelial cells in this tissue, further underlining the cell specificity of the targeting process (61).
What could contribute to this variable polarity of AQP2 expression? Recent data indicate that interstitial osmolality may be at least partially responsible for the basolateral targeting of AQP2 in the IMCD (72). In MDCK cells adapted to hypertonic culture medium, basolateral rather than apical targeting of AQP2 is induced by forskolin. Furthermore, kidney slices taken from vasopressin-deficient Brattleboro rats show apical insertion of AQP2 in the inner medulla when exposed to vasopressin. In contrast, slices from normal rats (with a hypertonic interstitium in vitro) show a marked basolateral insertion of AQP2 when challenged acutely with vasopressin in vitro (Fig. 1,C and D) (72). Finally, inner medullary principal cells in kidney slices from Brattleboro rats pretreated with vasopressin for 11 days in vivo also show basolateral AQP2 insertion when exposed to acute vasopressin stimulation in vitro. While hypertonicity cannot be the only factor involved in this change in polarity of AQP2 insertion (because cortical connecting segments also show basolateral AQP2 insertion, and LLC-PK1 cells with basolateral AQP2 expression are grown in isotonic culture medium), this observation provides a starting point for the examination of gene expression patterns that are involved in determining the polarity of insertion of AQP2 in epithelial cells.
In the case of LLC-PK1 cells, it was shown that the polarity of expression of a transfected LDL receptor is apical, rather than basolateral, as in MDCK cells (48). This change in polarity was ascribed to the lack of a key sorting adaptor, μ1b, which interacts with some tyrosine-based sorting motifs and is not expressed in LLC-PK1 cells. Restoration of the missing μ1b adaptor into LLC-PK1 cells by cotransfection restored the basolateral polarity of the LDL receptor in these cells. However, the absence of this adaptor does not lead to a generic disruption of apical/basolateral polarity in these cells, and we have shown that AQP2 is still inserted basolaterally in LLC-PK1 cells that express the μ1b protein (Sun T, Bouley R, and Brown D, unpublished observations).
MEMBRANE ACCUMULATION OF AQP2 BY INHIBITION OF CLATHRIN-MEDIATED ENDOCYTOSIS
While the precise pathway leading to AQP exocytosis has not been identified, recent data have confirmed a role for clathrin-coated pits in the endocytotic pathway (64, 65). A role for clathrin in water channel internalization was first proposed almost two decades ago, based on indirect evidence (5, 6, 63). New data using label-fracture on AQP2-transfected LLC-PK1 cells and antibodies against an externally oriented epitope of AQP2 in collecting duct principal cells have shown that AQP2 accumulates in clathrin-coated plasma membrane domains during vasopressin stimulation and washout (Fig. 3). Furthermore, when clathrin-mediated endocytosis was inhibited by the expression of a dominant-negative form of the protein dynamin in LLC-PK1 cells, AQP2 accumulated on the plasma membrane and was depleted from cytoplasmic vesicles (Fig.4) (65). Dynamin is a GTPase that is involved in the formation and pinching off of clathrin-coated pits to form clathrin-coated vesicles (24, 41, 59). The dominant-negative form has a single point mutation, K44A, that renders the protein GTPase deficient and arrests clathrin-mediated endocytosis (8, 49, 59). However, dynamin mutants also inhibit endocytosis via caveolae (47), and the electron microscopic detection of AQP2 in clathrin-coated pits described above was, therefore, essential to prove the involvement of clathrin in AQP2 endocytosis.
Why does inhibition of clathrin-mediated endocytosis result in the plasma membrane accumulation of AQP2? Studies in transfected LLC-PK1 cells have shown that AQP2 can recycle constitutively between the plasma membrane and intracellular vesicles, even in the absence of increased intracellular cAMP (22). However, accumulation of AQP2 at the cell surface does not usually occur under baseline conditions presumably because the relative rates of exo- and endocytosis are in equilibrium. The amount of AQP2 on the plasma membrane could, therefore, be increased either by increasing the rate of exocytosis, by inhibiting endocytosis, or both. It has been shown, for example, that the insulin-regulated glucose transporter, GLUT4, also recycles constitutively but the amount at the cell surface is increased by insulin treatment of target cells (7, 26,49).
How rapidly is AQP2 accumulated at the cell surface after inhibition of endocytosis? While expression of dominant-negative dynamin in cells results in increased cell-surface AQP2 expression, several hours are required before the effect takes place in these transfection experiments. More recently, an extensive accumulation of AQP2 at the cell surface has been accomplished within 30 min after inhibition of endocytosis by treatment with methyl-β-cyclodextrin (64). This drug depletes membranes of cholesterol and results in a rapid inhibition of endocytosis (53). Therefore, at least in LLC-PK1 cells, AQP2 can accumulate at the cell surface with a time course similar to that seen after vasopressin stimulation, simply by blocking endocytosis. This implies that constitutive AQP2 recycling between intracellular vesicles and the cell surface is rapid in this system. It must now be determined whether a similar constitutive process occurs in the collecting duct in situ. When interpreting the results of manipulations that result in a net increase in AQP2 cell-surface expression, one must therefore bear in mind the contribution of an inhibition of endocytosis to the overall effect.
PATHWAYS OF INTRACELLULAR AQP2 RECYCLING
The development of transfected cell models has greatly facilitated the dissection of intracellular AQP2 trafficking pathways. AQP2 is internalized by a clathrin-mediated mechanism (65), and internalized fluid phase markers can be detected in subapical vesicles, as well as in tubular structures in proximity to the Golgi region in both cultured cells and collecting duct principal cells in situ (6). Several rounds of exo- and endocytosis of AQP2 could be followed in LLC-PK1 cells despite the complete inhibition of de novo AQP2 synthesis (27), but details of the pathway(s) followed by this recycling AQP2 were vague. The nature of this pathway was addressed in cultured cells using manipulations that blocked transport at intermediate steps within the cell. Lowering the temperature to 20°C or incubating cells with the H+-ATPase inhibitor bafilomycin caused an accumulation of AQP2 in a perinuclear compartment that partially overlapped with clathrin immunostaining but did not stain for Golgi markers such as giantin or β-coat protein. The 20°C block prevents exit of proteins from the trans-Golgi (40), and clathrin-coated vesicles are enriched in this cellular compartment (20). In addition, recycling AQP2 is at least partially colocalized with internalized transferrin in recycling endosomes in LLC-PK1 cells (76). Motifs in the sixth transmembrane domain of AQP2, including a dileucine motif, are involved in regulated trafficking of this water channel (75). Domain-swap experiments, however, show that while the cytoplasmic COOH terminus of AQP2 is necessary for regulated insertion of AQP2, it is not sufficient, implying that other domains of the protein play a role in this process (13).
AQP2 EXOCYTOSIS AND VESICLE FUSION
While intracellular transport processes (via microtubules, actin, or simple random motion of vesicles) can deliver AQP2-containing vesicles close to the appropriate membrane domain, regulated vesicle fusion with target membranes involves the concerted interaction of many other accessory proteins. In common with membrane fusion events in many cell types, the final step in the exocytotic fusion of AQP2-containing vesicles with the plasma membrane involves solubleN-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) protein complexes (19) as well as intracellular calcium mobilization (9). The role of so-called SNARE proteins in vesicle fusion events was first appreciated in synaptic vesicle fusion events. Since the original “SNARE hypothesis” was proposed (60), the precise details of how proteins on target membranes and donor membranes interact with soluble cytosolic proteins to regulate vesicle fusion have been subjected to intense investigation (54, 55, 74). In collecting duct principal cells, vesicle-associated membrane polypeptide (VAMP)-2 is present on AQP2-containing vesicles (45), and the target (t)-SNAREs syntaxin 3, synataxin 4 (35), and soluble N-ethylmaleimide-sensitive factor attachment protein (SNAP)-23 (25) are also expressed. Treatment of collecting duct cells in culture with tetanus toxin, which cleaves the SNARE protein VAMP-2/synaptobrevin, abolishes vasopressin-induced AQP2 translocation to the plasma membrane (19). However, the precise SNARE proteins that are involved in this process in vivo remain unclear. The t-SNAREs syntaxin 3 and syntaxin 4 were initially reported to have basolateral and apical localizations, respectively, in adult rat kidney (3, 35,36). However, a recent study using different antibodies has reported exactly the opposite polarity of these two SNARE proteins in principal cells (33). Both sets of antibodies were raised against seemingly unique peptide sequences, and thus the reason for this major discrepancy is unclear. A further problem to be resolved is that heavy syntaxin 3 staining of the proximal tubule brush border was described in the most recent study (33), yet previous work did not detect syntaxin 3 mRNA by RT-PCR in microdissected proximal tubules (36). This process becomes even more complex when one considers that AQP2 can be inserted apically and/or basolaterally in different cell types and under different physiological conditions (see apical vs. basolateral aqp2 expression in epithelial cells). Does this also require a reorientation of the vesicle fusion machinery, or can the existing polarized fusion machinery adapt to catalyze fusion of AQP2-containing vesicles with either the apical or basolateral plasma membrane?
In addition to SNARE proteins, other accessory proteins known to be involved in the AQP2 shuttling and/or fusion process are heterotrimeric GTP binding proteins of the Gi family (70). Treatment of cultured rabbit cortical collecting duct cells with pertussis toxin, or exposure of permeabilized cells to synthetic peptides derived from the COOH terminus of the Gαi-3-subunit, abolished cAMP-induced AQP2 trafficking to the cell surface. However, it has also been proposed that proteins of the Gi family inhibit AQP2 accumulation at the cell surface via calcium receptor signaling upon an increase in luminal calcium levels (57). Furthermore, earlier work demonstrated that Gαi-3 exerts a negative regulatory effect on the secretion of heparan sulfate proteoglycans from cultured cells and that secretion is actually increased after pertussis toxin treatment, which inhibits Gi protein activity (62). Thus the precise step at which this G protein exerts its effect on AQP2 trafficking remains to be determined.
ROLE OF PHOSPHORYLATION IN AQP2 TRAFFICKING
Several putative phosphorylation sites for kinases are present in the AQP2 sequence. These include PKA and PKG, PKC, and casein kinase II sites. Most work has focused on the role of PKA-induced phosphorylation in the vasopressin-induced signaling cascade, as described above. Some PKA anchoring proteins (AKAPs) are enriched in AQP2-immunopurified vesicles from IMCD cells. Inhibition of forskolin-induced AQP2 translocation with a peptide that prevents PKA-AKAP interaction demonstrated that, besides its enzymatic activity, tethering of PKA to subcellular compartments is essential for AQP2 translocation (30,31). While available data consistently show that serine 256 phosphorylation is required for the vasopressin-induced cell-surface accumulation of AQP2 (17, 28), dephosphorylation of AQP2 is probably not necessary for its subsequent internalization. Prostaglandin E2 stimulates removal of AQP2 from the surface of principal cells when added after AVP treatment but does not alter the phosphorylated state of AQP2 (77). In support of this, it was shown in cell cultures that PKC-mediated endocytosis of AQP2 is also independent of the phosphorylation state of this water channel at residue serine 256 (71).
A major unresolved issue is how phosphorylation of AQP2 at serine 256 increases the cell-surface expression of this water channel. Phosphorylation could modify the interaction of AQP2 cargo vesicles with the cytoskeleton, via microtubule and/or microtubule motors or accessory cross-linking proteins. Phosphorylation-dependent protein-protein interactions may be necessary to augment AQP2 trafficking to the cell surface, although, as previously discussed, constitutive membrane insertion occurs rapidly without vasopressin stimulation or an elevation of intracellular cAMP. Alternatively, phosphorylation could inhibit the endocytotic step of AQP2 recycling, leading to accumulation at the cell surface. However, some data (discussed above) show that serine 256-phosphorylated AQP2 can still be internalized. Preventing dephosphorylation of AQP2 with the phosphatase inhibitor okadeic acid has the expected effect of increasing cell-surface accumulation of AQP2 in cultured cells, but, surprisingly, the same effect of okadeic acid was observed in the presence of the PKA inhibitor H-89. The authors concluded that okadeic acid stimulates the membrane translocation of AQP2 in a phosphorylation-independent manner (69). As expected, an AQP2 mutant containing a serine 256-to-aspartic acid amino acid switch was constitutively expressed at the cell surface, presumably because the negatively charged aspartic acid residue mimics the phosphorylated serine 256 residue of the wild-type protein (71). As yet, no effect of phosphorylation on the interaction of accessory proteins with AQP2 has been reported. This is an area toward which future efforts will certainly be directed.
NEPHROGENIC DIABETES INSIPIDUS AND AQP2 TRAFFICKING
Most cases of hereditary nephrogenic diabetes insipidus result from mutations in V2R, although autosomal forms in which AQP2 is defective have provided a great deal of important information about the pathophysiology of the disease, and about AQP2 trafficking and oligomerization (11, 14, 38). In the absence of a functional V2R at the cell surface in nephrogenic diabetes insipidus patients, it is critical to develop strategies to bypass the V2R-dependent signaling cascade that normally results in an accumulation of AQP2 on principal cell plasma membranes. Many of the observations related to the cell biology of AQP2 trafficking outlined above could be relevant in this regard. For example, it is now known that AQP2 is trafficked in response to an elevation of cGMP, in addition to cAMP, in some regions of the collecting duct (Figs. 1 and2). Modulation of the actin cytoskeleton alone can cause cell-surface accumulation of AQP2. Finally, AQP2 can move to the cell surface in a constitutive pathway, so that inhibition of endocytosis in this constitutive pathway can be envisaged as a means of accumulating AQP2 on the plasma membrane (Fig. 4).
In addition, many of the V2R and AQP2 mutations described in nephrogenic diabetes insipidus patients result in a misfolded protein that is blocked at some point [often the rough endoplasmic reticulum (RER)] inside the cell. Some of these proteins would be functional if they could reach the cell surface. It has been shown that some chemical chaperones (e.g., glycerol, DMSO) can help misfolded AQP2 to reach its final cell-surface destination (39, 66), but these agents may be difficult or impossible to apply in vivo. In an interesting recent development, it has been shown that relatively brief treatment of cultured cells and transgenic mice with the Ca-ATPase inhibitor thapsigargin can increase the cell-surface expression of ΔF508 CFTR (16). This effect is probably a result of the depletion of calcium stores from the RER and the subsequent inactivation of calcium-dependent chaperones that would otherwise bind to the misfolded ΔF508 CFTR protein and direct it into degradative pathways. While this approach is not, of course, without potential caveats related to the long-term depletion of calcium stores, the authors propose that it could be applicable to a variety of genetic diseases resulting from the defective trafficking and RER retention of misfolded proteins.
As outlined above, considerable progress has been made in understanding several aspects of AQP2 trafficking over the past few years. In large part, this has resulted from the development of model in vitro systems in which regulated apical and basolateral AQP2 trafficking has been reconstituted. The trafficking of AQP2 between intracellular vesicles and the cell surface involves a complex series of membrane fission and fusion steps at various points along the recycling pathway and involves passage through a trans-Golgi and/or recycling endosome-like compartment. AQP2 vesicles also interact with the actin cytoskeleton and the microtubule network of the cell during the trafficking process, and calcium mobilization in addition to cyclic nucleotide elevation in the cytosol contributes to the exocytotic insertion of AQP2. These events are regulated by hormone-induced protein phosphorylation in ways that remain undetermined and involve the interaction of GTP binding proteins. Additional features of AQP2 trafficking that have emerged recently are that AQP2 can recycle in a constitutive pathway, in addition to a vasopressin-regulated pathway, and that the polarity of membrane insertion of AQP2 can be different (apical vs. basolateral) in a variety of cell types in both cultured cells and different regions of the collecting duct. While cell culture systems will continue to play a major role in dissecting the cell biology of vasopressin action, it is also important to continue to relate these in vitro findings to whole organ and whole animal physiology. Thus alternative experimental systems from isolated collecting ducts, tissue slices, perfused kidneys, and transgenic mouse models are expected to play an increasing role in aquaporin research over the next few years. Genomic and proteomic approaches promise to reveal a wealth of new information about the molecular mechanisms of AQP2 trafficking in the collecting duct and about the many as yet undiscovered binding partners and accessory proteins that regulate the vesicular transport and membrane insertion of this protein. From both the physiological and cell biological points of view, AQP2 is a fascinating protein that will continue to occupy the efforts of many laboratories for years to come.
I thank all of my colleagues who have made contributions to these studies, as well as those who work in the aquaporin field, for making this such a stimulating and rewarding area of research.
The work from our laboratory that is described in this review has received continuous support from the National Institutes of Health, the National Kidney Foundation, and the American Heart Association.
Address for reprint requests and other correspondence: D. Brown, Renal Unit/Program in Membrane Biology, Massachusetts General Hospital East, 149 13th St., Charlestown, MA 02129 (E-mail:).
- Copyright © 2003 the American Physiological Society