Endosomal pathways for water channel and proton pump recycling in kidney epithelial cells

The process of vesicle recycling is involved in establishing, maintaining and modulating the plasma membrane composition of virtually all eukaryotic cells. Because epithelial cells from the kidney must react rapidly to changes in their environment in order to maintain a constant milieu intérieur for the organism, they have developed highly specialized mechanisms that enable them to modify transepithelial transport processes in response to a variety of stimuli. Thus, specific plasma membrane components are inserted on demand by exocytosis of specific transport vesicles, and are subsequently removed from the cell surface by endocytosis of membrane segments in which particular proteins are concentrated. While such events occur in most, if not all, regions of the kidney tubule, the most remarkable examples of plasma membrane recycling in response to physiological stimuli occur in the collecting duct.

Urinary acidification and urinary concentration are finely regulated in this tubule segment, and occur independently in two distinct cell types, the intercalated cell and the principal cell. In the intercalated cell, hydrogen ion secretion is controlled in response to body acid-base status by the cycling of vesicles carrying a vacuolar type proton pumping ATPase to and from the plasma membrane (Brown, 1989; Brown et al., 1988a; Madsen and Tisher, 1986; Madsen et al., 1991; Schwartz and Al-Awqati, 1985; Schwartz et al., 1985). As will be described below, proton pumps can be inserted into opposite poles of subsets of intercalated cells, and this represents, therefore, an intriguing and unusual example of membrane protein targetting (Brown et al., 1988a; Schwartz et al., 1985). In principal cells, the antidiuretic hormone, vasopressin, induces the insertion of vesicles that contain proteinaceous water channels into the apical cell membrane, thus increasing the permeability to water of the epithelial cell layer (Abramov et al., 1987; Brown, 1989; Handler, 1988; Verkman, 1989).

The polarized insertion of membrane components involves at least two distinct steps once the proteins have been packaged into appropriate transporting vesicles at the level of the Golgi. First, vesicles must move through the cytosol towards their target membrane; and second, a recognition process must take place that results in fusion of the vesicle with the correct membrane domain. Movement through the cytosol involves interaction of vesicles with cytoskeletal elements, including microtubules, via microtubule motors (Schroer and Sheetz, 1991; Vale, 1987) and actin microfilaments, in part via myosin isoforms (Adams and Pollard, 1986; Johnston et al., 1991; Zot et al., 1992). Vesicle recognition and fusion steps are poorly understood, but probably involve a specific interaction between cyto-plasmically oriented protein domains that are characteristic of the transported vesicle, the membrane target, or both. A considerable amount of recent work has implicated GTP-binding proteins in vesicle trafficking and targetting, but well-defined roles for these molecules are still lacking (Balch, 1990; Goud and McCaffrey, 1991).

Whereas principal cells recycle water channels in non-acidic endosomes, the opposite is true of collecting duct intercalated cells. The endosomal pathway in this cell type is specialized to transport proton pumps to and from the plasma membrane, and these endosomes do not contain water channels (Lencer et al., 1990a). Thus, there is a clearcut division of labor between these two cell types in this segment of the urinary tubule.

During systemic acidosis in vivo, or an increase in pCO₂ at the basolateral pole of isolated tubules in vitro, specialized cytoplasmic vesicles that are enriched in proton pumps, fuse with the apical plasma membrane of intercalated cells (Dorup, 1986; Madsen and Tisher, 1983; Madsen et al., 1991; Schwartz and Al-Awqati, 1985). This process increases the number of proton pumps at the cell surface. Similar carbonic-anhydrase-rich, proton-secreting cells are also found in toad and turtle urinary bladder, and an identical exocytotic process also occurs in these tissues (Dixon et al., 1986; Gluck et al., 1982; Stetson and Steinmetz, 1983). The transporting vesicles are highly characteristic, having a cytoplasmic ‘coat’ that, by conventional electron microscopy, resembles clathrin. However, morphological and immunocytochemical evidence identified the coating material as the cytoplasmic domain of a vacuolar-type proton pumping ATPase (Brown et al., 1987; Brown and Orci, 1986). Specific antibodies against different subunits of the proton pumping ATPase from bovine kidney medulla were used to show that the vesicle-coating material contained defined subunits of the proton pump (Brown et al., 1987, 1988b). The ‘coat’ seen on the cytoplasmic side of the plasma membrane of intercalated cells also contained the same pump subunits.

The structure of the membrane coat was elucidated using the rapid-freeze, deep-etch technique, in which high-resolution microscopic images of proteins can be obtained. In proton-secreting cells from the toad bladder, the membranecoating material had a structure that was identical to that of immunoaffinity-purified proton pumps, incorporated into phospholipid liposomes, thus confirming its identity as part of the proton pumping ATPase responsible for distal urinary acidification (Brown et al., 1987). Part of the underside of a plasma membrane domain rich in proton pumps is shown in Fig. 1. The pumps are tightly packed into this membrane region, at a density of about 14,000 per pm². Vesicles inside the cell with a similar structure were also observed using the deep-etching procedure (Brown et al.,1987)..

In the cortical collecting duct (Schwartz et al., 1985) and in the turtle urinary bladder (Stetson and Steinmetz, 1985), morphological variants of carbonic anhydrase-rich cells have been described (Fig. 2), which are believed to be proton-secreting (A-cells) or bicarbonate-secreting (B-cells). Both types co-exist in kidney cortical collecting ducts, in accord with the ability of this tubule segment to secrete either net acid or net base under different physiological conditions of acidosis or alkalosis. Using antiproton pump antibodies, all medullary intercalated cells were found to have proton pumps associated with their apical plasma membrane, whereas cortical intercalated cells had three patterns of labeling (Fig. 2A): apical, basolateral, and diffuse cytoplasmic or bipolar (Brown et al., 1988a). This finding provided direct support for the idea that different intercalated cells in the cortex are responsible for either proton (apical pumps) or bicarbonate (basolateral pumps) secretion into the tubule lumen. The cells with a diffuse staining might be an intermediate or transitional cell type.

Provocative data by Schwartz et al. (1985) suggested that A and B cells may be interconvertible, and may change their functional polarity by inserting proton pumps into either apical or basolateral plasma membranes, depending on prevailing acid-base conditions of the animal. However, others have argued that A and B cells do not interconvert during alterations in acid-base status (Verlander et al., 1988; Schuster, 1988) and Schwartz has also concluded that a direct interconversion of A- and B-cells is unlikely to occur at least in an in vitro model of acidification (Koichi et al., 1992). While this question has not yet been clearly resolved, several observations are pertinent. The A cells have a basolateral band-3-like chloride-bicarbonate exchanger, AE1 (Alper et al., 1989; Drenckhahn et al., 1985; Schuster et al., 1986), whereas no band-3-like chloride-bicarbonate exchanger has yet been located by immunocytochemistry in type B cells, either in the apical or the basolateral plasma membrane. Double-staining studies show that all B-type intercalated cells with basolateral or diffuse proton pump staining lack detectable AE1, while the vast majority of cells with apical proton pumps show strong basolateral AE1 staining (Fig. 2). However, we consistently found a small (1%) population of cortical intercalated cells with apical proton pumps but no basolateral AE1 (Alper et al., 1989). Whether this cell type is a distinct population or a transitional type of intercalated cell is unknown. The absence of AEl-like antigenicity on the apical surface of all intercalated cells demonstrates that if a simple exchange of transporting molecules from apical to basolateral plasma membrane and vice versa occurs in these cells during adaptation to acid-base loads, then the apical anion exchanger must be modified in some way that renders it undetectable by a range of anti-band 3 antibodies (Schuster, 1988). In aci-dotic animals, functionally detectable apical anion exchangers appear to be internalized by B-intercalated cells (Satlin and Schwartz, 1989), but whether or not proton pumps are subsequently inserted into the apical membrane of the same cells is still unclear. By immunofluorescence microscopy, a decrease in the number of basolaterally stained intercalated cells is found in the cortex of rats with metabolic acidosis and the number of cells with apical proton pumps increases, both after acute (6 hour) gavage with NH4CI (Brown, Saboli ć and Gluck, unpublished) and after chronic (14-day) NH4CI treatment (Bastani et al., 1991).

In contrast, immunocytochemically-detectable AE1, as determined by quantitative laser confocal microscopy, significantly diminishes in basolateral plasma membranes of A-type intercalated cells in the cortex after just 6 hours gavage with HCO?,“, which induces metabolic alkalosis (Alper et al., 1991). Whether this decrease reflects turnover and degradation of AE1, or a phenomenon of epitope masking is unclear at present. Nevertheless, this decrease would be an expected adaptation to alkalosis, because it would presumably reduce the rate of net acid secretion by the cortical collecting duct by decreasing the H⁺ secreting capacity of A intercalated cells.

The water permeability of the kidney collecting duct epithelium is regulated by vasopressin-induced recycling of water channels between an intracellular vesicular compartment and the plasma membrane of principal cells (Fig. 3); an analogous mechanism exists in other vasopressin-sensitive epithelia such as the toad urinary bladder and the amphibian epidermis (Abramov et al., 1987; Brown, 1989,1991a,b; Handler, 1988; Harris et al., 1991b; Hays, 1983; Hays et al., 1987; Verkman, 1989). The apically derived endosomes in principal cells are generated via clathrin-mediated endo-cytosis (Brown and Orci, 1983; Brown et al., 1988; Strange et al., 1988) and they appear to function primarily to recycle membrane components, including water channels, back to the apical membrane during hormonal stimulation (Fig. 3). These vesicles do not, however, acidify their interior. This was shown using FITC-dextran as an in vivo marker of endocytosis and acidification, coupled with co-localiza-tion of a lysosomal glycoprotein LGP 120; most of the internalized fluorescent probe did not move into lysosomes and remained in an apical, non-acidic compartment (Lencer et al., 1990b).

These specialized endosomes were purified by differential and density-gradient centrifugation from kidney papilla, and some of their transport characteristics were measured, along with their protein composition (Sabolic et al., 1992b). Fluorescence quenching measurements showed that the isolated vesicles maintained a high, HgCh-sensitive water permeability, consistent with the presence of vasopressinsensitive water channels. They did not, however, exhibit ATP-dependent luminal acidification, nor any A-ethyl-maleimide-sensitive ATPase activity, properties that are characteristic of most acidic endosomal compartments (Fig. 4). Western blotting with specific antibodies showed that the 31 kDa and 70 kDa cytoplasmically oriented subunits of the vacuolar proton pump were not present in these apical endosomes from the papilla, whereas they were readily detectable in endosomes prepared in parallel from the cortex. In contrast, the 56 kDa subunit of the proton pump was present in papillary endosomes, and was localized at the apical pole of principal cells by immunocytochemistry.

In addition, an antibody that recognizes the 16 kDa transmembrane subunit of oat tonoplast ATPase (Lai et al., 1988) cross-reacted with a distinct 16 kDa band in cortical endosomes, but no 16 kDa band was detectable in endosomes from the papilla. Therefore, early endosomes derived from the apical plasma membrane of collecting duct principal cells fail to acidify because they lack functionally important subunits of a vacuolar-type proton pumping ATPase, including the 16 kDa transmembrane domain that serves as the proton-conducting channel, and the 70 kDa cytoplasmic subunit that contains the ATPase catalytic site. Interestingly, endosomes that are involved in water channel internalization and recycling in the toad urinary bladder also fail to acidify in response to ATP, and they also lack the 31 and 70 kDa subunits of the proton pump, while retaining the 56 kDa subunit (Harris et al., 1991a).

The specialized nature of apical endosomes in vasopressin-sensitive cells presumably reflects the superficial vesicle-shuttling mechanism that characterizes their physiological response to hormonal stimulation. However, other cell types also use a similar process to recycle functionally important cell surface molecules between cytoplasmic vesicles and the plasma membrane in response to a variety of stimuli. For example, the insulin-sensitive glucose transporter Glut-4 has a similar recycling mechanism (Suzuki and Kono, 1980), as does the H⁺,K⁺-ATPase in gastric parietal cells (Forte et al., 1977). In the mammalian urinary bladder, sodium channels are inserted into the apical membrane of surface epithelial cells as a result of stretch-induced exocytosis (Lewis and DeMoura, 1984). Whether these and other vesicles involved in specialized recycling processes also lack the capacity to acidify remains to be determined.

The identification of early endosomes devoid of proton pump subunits suggests that the vesicles that initially pinch off from the apical plasma membrane of collecting duct principal cells also do not acidify. Because, in the case of water channel endocytosis, this is a clathrin-mediated process (Brown and Orci, 1983; Brown et al., 1988; Strange et al., 1988), our results imply that endocytotic clathrin-coated vesicles lack the capability to acidify. However, when total cellular clathrin-coated vesicles are isolated from various sources, including the kidney, proton pump subunits can be detected in these vesicles, and ATP-dependent acidification can be measured in the total vesicle population (Forgac et al., 1983; Stone, 1988; Xie and Stone, 1986). The most likely explanation for these results is that the coated vesicles that are specifically involved in (apical) endocytosis are not acidic vesicles, but that other clathrin-coated vesicles within the cell, that subserve other transport functions, do contain membrane-associated proton pumps, and are capable of generating an internal acidic pH. In support of this, staining of fibroblasts with the morphological pH marker, DAMP, has demonstrated intracellular heterogeneity of coated vesicle labeling (Anderson et al., 1984). In addition, it has been reported that clathrin-coated vesicles involved in endocytosis in other cell types are nonacidifying vesicles (Fuchs et al., 1987).

A recent exciting development in the field of water channel physiology has been the isolation, cloning and functional expression of a 28 kDa integral membrane protein known as CHIP28. First isolated from erythrocytes by Agre and colleagues (Denker et al., 1988), this member of a family of channel proteins has a kidney homolog (Zhang et al., 1993), and is expressed at high levels in proximal tubule and descending limbs of Henle in both apical and basolateral plasma membranes (Nielsen et al., 1993; Sabolic et al., 1992a) (Fig. 5). However, because these membrane domains are constitutively permeable to water, and because CHIP28 cannot be detected in papillary principal cells (although it cross-reacts weakly with principal cells in the cortex) CHIP28 is probably not the vasopressin-sensitive water channel that is recycled in the non-acidic endosomes described above. However, a partial homolog (referred to as WCH1) with approximately 50% sequence identity to porter Glut-1 (Thorens et al., 1990), was examined, no redistribution could be detected. Thus, by double staining the same cell, proton pumps were scattered throughout the cytoplasm as described earlier, but the other basolateral markers were still located in a distinct basolateral pattern (Fig. 6A,B). This result indicates that the effect of microtubule disruption is likely to be protein-specific, rather than membrane domain-specific. It also indicates that proton pumps must be segregated into specific endocytotic vesicles during the recycling process, and that other membrane proteins, including band 3 and Glut-1, are not internalized in parallel in the same vesicles.

Support for the contention that the microtubule involvement in membrane polarity is protein-specific comes from studies on the proximal tubule. In this nephron segment, the apical membrane is highly active endocytotically, and much of the apical membrane at the base of the microvilli is clathrin-coated (Brown and Orci, 1986; Rodman et al., 1984). Our CHIP28, has been identified in the papilla (Sasaka et al., 1992), and antibodies against this protein localize to collecting duct principal cells. This protein is, therefore, a candidate for the vasopressin-regulated water channel. The availability of full-length cDNAs for these water channels will permit the future analysis of structural elements that may determine non-polarized (CHIP28) versus polarized (WCH1, vasopressin-sensitive channel) targetting of these physiologically relevant proteins.

Many studies have shown that microtubules can support the movement of organelles in vitro, and are probably involved in organelle trafficking in vivo. Drugs that disrupt microtubules interfere with secretory events in many cell types (Busson-Mabillot et al., 1982; Orci et al., 1973; Patzelt et al., 1977; Stetson and Steinmetz, 1983; Taylor, 1977) and they result in a redistribution of intracellular organelles including lysosomes, RER and the Golgi apparatus (Dabora and Sheetz, 1988; Heuser, 1989; Kreis, 1989; Matteoni and Kreis, 1987; Moskalewski et al., 1973). Despite these major cellular alterations induced by microtubule disruption, the role of microtubules in determining cell polarity remains controversial (Achler et al., 1989; Gutmann et al., 1989; Hasegawa et al., 1987; Ojakian and Schwimmer, 1988; Pavelka et al., 1983; Rindler et al., 1987; Salas et al., 1986). In particular, some studies using virally infected cells as well as endogenous proteins have shown that microtubule disruption causes apical proteins to appear on both apical and basolateral plasma membranes. In contrast, other work, including our own studies on an apical membrane glycoprotein, gp330, in the proximal tubule, has shown that membrane proteins are not misdirected to an inappropriate membrane domain after microtubule disruption in these cells. In addition, it is generally believed that microtubules are not involved in the polarized insertion of basolateral membrane proteins.

Microtubules are also implicated in the physiological responses of intercalated cells and principal cells. Studies using both amphibian epithelia and mammalian kidney collecting duct have shown that microtubule depolymerizing agents significantly reduce vasopressin-induced transepithelial water flow (DeSousa et al., 1974; Dratwa et al., 1984; Kachadorian et al., 1979; Valenti et al., 1988), and that H⁺ secretion is also markedly inhibited by colchicine in turtle bladder (Stetson and Steinmetz, 1983). In both cases, the effect of microtubule disruption seems to occur by reducing the likelihood that vesicles carrying water channels or proton pumps will move to and fuse with the apical plasma membrane. However, because neither water flow nor H⁺ secretion are totally inhibited (maximum inhibition is around 60-70%), it appears that some transporting vesicles are still able to fuse with the appropriate target membrane even in the presumed absence of microtubules. This may occur as a result of random vesicle movement within the cell resulting in some residual ‘collisions’ of these vesicles with the apical membrane, thus allowing the recognition and fusion event to occur normally.

One possibility that might explain differential targeting of proton pumps to opposite membrane domains in intercalated cells is that the apical and basolateral forms of the proton ATPase may not be identical in subunit composition; this difference could confer domain specificity on the targeting process. There is increasing evidence that different isoforms of some proton pump subunits are selectively amplified and expressed in different cell types, as well as in different membrane domains within the same cell (Nelson et al., 1992; Puopolo et aL, 1992). In particular, isoforms of the 56 kDa subunit show distinct patterns of cellular location in the kidney, and one isoform (the so-called ‘kidney’ isoform) is amplified in intercalated cells, but is not detectable in proximal tubule epithelial cells (Nelson et al., 1992). So far, however, no isoform that is restricted to either the apical or basolateral membrane in intercalated cells has been identified. Another possibility is that the microtubular network is somehow different in cells with apical or basolateral proton pumps, and that vesicles containing the pumps move selectively to only one pole of the cell. As discussed above, it has been proposed that microtubule tracks are important for the delivery of apical proteins, but that basolateral proteins are routed to the base of the cell by a default pathway. In this way, microtubule depolymerization might be expected to result in basolateral, rather than apical insertion of proton pumps. This prediction assumes that vesicles carrying proton pumps are competent to fuse with either apical or basolateral plasma membranes, and that the specificity of the process lies exclusively in the delivery of the vesicles to the appropriate membrane domain. However, this is clearly not the case with another apical membrane protein, gp330, which is present in vesicles that do not fuse with the wrong (i.e. the basolateral) plasma membrane, even when they are apparently close enough for fusion to occur (Gutmann et al.,1989).

To examine the possible involvement of microtubules in the polarized insertion of proton pumps, studies were performed on colchicine-treated rats (Brown et al., 1991). In these animals, proton pumps were no longer polarized to one pole of the cell, but they were concentrated in numerous vesicles scattered throughout the cytoplasm after 4-6 hours of drug treatment. A similar scattered distribution was seen in A and B intercalated cells, as well as in epithelial cells of the proximal tubule (Fig. 6). One especially important result was that A-type cells in the inner stripe of the outer medulla did not concentrate proton pumps in their basolateral plasma membranes. These results indicate that microtubule depolymerization alone is not sufficient to account for the final differences in the membrane distribution of proton pumps seen in A and B intercalated cells. As is also the case for gp330-containing vesicles in the proximal tubule, an additional level of control must govern the final fusion process between these re-routed vesicles and their target membrane region.

The effects of colchicine on the distribution of basolateral proton pumps in intercalated cells did not, however, extend to all basolateral membrane proteins. When the localization of two other exclusively basolateral proteins, the band 3 anion exchanger and a facilitated glucose trans previous studies show that proteins (gp330 and H⁺-ATPase) that are present either in the clathrin-coated membrane domains at the apical membrane, or in the membrane immediately adjacent to these regions at the base of the microvilli, are redistributed in colchicine-treated animals (Fig. 7A). However, some membrane proteins including dipeptidylpeptidase IV (DPP IV) and carbonic anhydrase type IV (CAIV) are restricted to the brush border microvilli (Brown and Waneck, 1992), and are not normally concentrated in the endocytically active zone at the microvillar base. After colchicine treatment, these proteins remain in their original microvillar location, are not redistributed with the other proteins (Fig. 7B). Interestingly, both CA IV and DPP IV are proteins that can be attached to the membrane via a glycosyl phosphatidylinositol anchor (Hooper et al., 1990; Waheed et al., 1992).

Based on results from many groups, the final level of control of vesicle fusion with target membranes could be related to the presence of specific GTP-binding proteins on different populations of vesicles (Balch, 1990; Goud and McCaffrey, 1991), or to the presence of other peripheral and integral membrane proteins that are necessary for specific fusion process to occur (Clary and Rothman, 1990; Clary et al., 1990; Hooper et al., 1990; Weidman et al., 1989). Apical, basolateral and endosomal membranes of many kidney epithelial cells have significant differences with respect to their content of GTP-binding proteins (Stow et aL, 1991). The structure of individual membrane proteins may, therefore, contain information ensuring that the protein is packaged into vesicles with the correct address, while other constituents of these vesicles may be responsible for controlling the final series of interactions with target membrane domains. The role of microtubules in this process would be, therefore, to concentrate vesicles containing either newly synthesized or recycling proteins at the appropriate membrane domain, and thereby increase the chances for a subsequent specific fusion event to occur.