ARTICLE

Top Abstract Article References

THE TARGETING OF POLARIZED membrane and secreted proteins in epithelia begins in the trans-Golgi network where apical or basolateral proteins are loaded into different vesicles. A number of signals have been discovered that appear to code for the mechanism responsible for this separation (19). However, a general hypothesis that applies for all apical or basolateral proteins has not emerged. Indeed, it appears that the same protein might be targeted to the apical membrane of one epithelial cell but to the basolateral membrane of another. For instance, the -subunit of the Na⁺-K⁺-ATPase is apically located in choroid plexus and retinal pigment epithelium but is basolateral in most other epithelia (28). Two other proteins in the retinal pigment epithelium also have flexible polarities (29, 39). The low-density lipoprotein (LDL) receptor transgene is apical in some kidney cells but basolateral in intestinal and liver cells (51). Furthermore, glycosyl-phosphatidylinositol-linked proteins are targeted to the apical membrane of most epithelia, yet in one cell line derived from the thyroid, such proteins are basolateral (74). We previously found that the kidney form of the anion exchanger AE1 (here, termed kAE1) is targeted to the apical membrane in -type of renal intercalated cell but to the basolateral membrane of another type of intercalated cell (-type) (64, 65). This flexibility (or plasticity) in the targeting of these proteins can be influenced by environmental cues. The retinal pigment epithelium targets the neural cell adhesion molecule, N-CAM, apically in situ, but when it is separated from the retina and cultured in vitro, it now targets this adhesion molecule to the basolateral domain (29). The intercalated cell, the subject of this review, reverses the targeting of kAE1 and the H⁺-ATPase in vitro in response to seeding density (65) and in vivo in response to acid feeding (59). These studies suggest that it might be possible to reorient the targeting machinery of a cell, a suggestion that is quite surprising given that epithelial cells are thought to be terminally differentiated. Recent studies in nonpolarized cells have shown the existence of "dormant" polarized pathways in that when apical or basolateral proteins were transfected into lymphocytes, osteoclasts, or macrophages, they did not share the same vesicles or routes of traffic (48, 73). However, the provision of external cues, for instance, binding to another cell or to a surface, directed some of these protein toward a new "polarized" surface.

The reorientation of epithelial cell proteins during this process of plasticity does not reverse the polarity of all proteins; for instance, the glucose transporter of intercalated cells remains in a basolateral location, and the peanut lectin binding protein continues to be apical regardless of the state of kAE1 or the H⁺-ATPase (65). Perhaps the targeting machinery of a cell has two separable pathways; one that is flexible and another that is fixed, but the biochemical mechanisms underlying these putative routes are unknown at present. That such plasticity of targeting exists in many cells is suggested by studies of sorting of the Na⁺-K⁺-ATPase in ATP depletion or ischemic acute renal failure (46), where it was found that the basolateral Na pump of the proximal tubule was retargeted to the apical membrane.

The Intercalated Cells of the Collecting Tubule

The Apical Anion Exchanger of the -Intercalated Cell

Plasticity of polarity can be rigorously examined only in a clonal cell line. To produce such a cell line, we first isolated a purified peanut lectin binding population of intercalated cells from rabbit kidney cortex and transfected them with the temperature-sensitive SV40 T antigen. Under the appropriate conditions (see below), the clonal cells grow to form an epithelial monolayer capable of secreting HCO₃ into the apical medium in a Cl-dependent manner (20). Measurement of their intracellular pH demonstrated apical but not basolateral Cl/HCO₃ exchange (64). These cells reproduced all the other known characteristics of -cells in situ, including the lack of staining with antibodies against kAE1. To study the polarized location of kAE1 biochemically, the current standard method would be to biotinylate the apical surface by an impermeant reagent and then purify the biotinylated proteins on avidin beads (55). However, when the apical surface was biotinylated, no proteins were precipitated by avidin beads, likely because the thick coat of glycolipids and glycocalyx present in the apical membranes prevented access of the reagents to any apical proteins. We then used the earlier approach, the cell-ripping method in which apical membranes are fused to a glass slide and the cells are sheared (65). In this method, it is actually the apical cytoplasm as well as the apical membrane that remains on the slide, and to remove all bound vesicles from the cytoplasmic surface of the apical membranes, we treated the slide with 0.1 M Na₂CO₃, pH 11. Electron micrographs of these clonal cells had previously shown that this region of the cell is sparsely populated by vesicles (20). Using this method, we found that these membranes contained kAE1 (64, 65). Fejes-Toth et al. (22) isolated intercalated cells from rabbit kidney using their -cell-specific monoclonal antibody and found by Western blots of whole cell lysates that no AE1 was detected. Perhaps this was due to its low abundance in these cells, and perhaps we were able to detect it because of our use of purified apical membranes. It had been demonstrated by Fejes-Toth et al. (22) and by us that these cells express mRNA for kAE1, but its abundance was low. That we were able to find kAE1 by biochemical but not immunocytochemical methods suggests to us that the antigen is "latent," as has been recently found for AE2 in the kidney (3, 9).

In addition to lack of staining by antibodies, kinetic analyses of apical Cl/HCO₃ exchange have led others to suggest that this transporter is not kAE1. Apical exchange is insensitive to the inhibitor of AE1, DIDS, and has a more alkaline pH activation profile (57). We recently showed that when the erythroid AE1 is reconstituted in apical type of lipids (i.e., enriched in glycosphingolipids and gangliosides), it was no longer sensitive to DIDS and had different pH activation profiles (66). These results in the aggregate demonstrate that the apical anion exchanger of -cells is kAE1 and that the absent signal on immunostaining is either due to low abundance or "latency." Brown and Alper and their colleagues (3, 9) have shown that kidney sections do not stain with antibodies to AE2 despite the fact that this anion exchanger is expressed in various segments of the kidney tubule. However, when the sections were treated with the harsh detergent SDS for a short period, the antigenicity of AE2 was uncovered. Similarly, the intercalated cells did not stain for the Na⁺-K⁺-ATPase except when treated with SDS (3, 9). These important results should inject a note of humility into conclusions about the presence or absence of AE1 in a cell based only on immunocytochemical methods. But, perhaps because seeing is believing, these doubts regarding the molecular identity of the exchanger will remain until the problem of the antigenic accessibility is solved or another Cl/HCO₃ exchanger that has these characteristics is discovered.

Recent interesting results by Cox et al. (15, 16) have shown that the chicken kidney also expresses alternately spliced variants of AE1; one such variant, AE1-3, is similar to the mammalian splice variant, kAE1. When transfected into MDCK cells, it is sorted to the apical membrane, whereas another variant, AE1-4, which encodes a protein that is longer by 63 amino acids, is sorted to the basolateral membrane (1). We had found that kAE1 is expressed in the clonal cell line; however, neither we nor others have examined for the presence of erythroid AE1 or other potential splice variants in these cells. It had been found that rat kidney expresses erythroid AE1; hence, the new avian studies suggest that a reexamination of this issue might provide new and interesting information.

Plasticity of Polarity in Intercalated Cells During the Response to Metabolic Acidosis

In vitro Plasticity in a -Intercalated Clonal Cell Line

When these clonal cells were plated at subconfluent density, they grew to confluence forming monolayers that secreted HCO₃. When plated at superconfluent density, they secreted acid and had vigorous apical endocytosis, apical H⁺-ATPase, and basolateral kAE1 (65). Hence, at least in vitro and in immortalized clonal cells, the plasticity of epithelial polarity was demonstrated. The reversal of polarity was dependent only on the seeding density, since 1 wk after plating, both low-density and high-density cells had a similar number of cells (based on their DNA content). Both low- and high-density cells were "real" epithelia; they had strictly polarized distribution of the four membrane proteins tested, kAE1, H⁺-ATPase, peanut lectin binding protein, and the glucose transporter, GLUT-1. Both had tight junctions, appropriate adherent junctions mediated by E-cadherin, basolateral actin cytoskeleton, and were able to have steady-state transepithelial transport of H⁺ and HCO₃ (S. Vijayakumar and Q. Al-Awqati, unpublished results). Given that the two cell forms originated from the same clone, we conclude that low- and high-density cells are two phenotypes of the same progenitor cell.

The factor that induced the conversion was found to be an extracellular matrix (ECM) protein that was deposited on the filters by high-density cells. When cells were plated at low density on filters that were conditioned by high-density cells, they converted their polarity and phenotype. Using apical endocytosis as an assay, we purified a 230-kDa protein capable of inducing the change in polarity (65). We termed this protein "hensin," which is Japanese for "change in body." It is a glycosylated protein that is synthesized as a secreted protein and traffics to the basolateral medium. Hensin is expressed highly in the intestine, pancreas, and liver but also in the brain (64). In the kidney, it is present only in the collecting ducts, including intercalated and principal cells as well as the inner medullary collecting duct.

Hensin: a New Multidomain Protein

Studies in the past two years have identified other proteins that contain these three domains in different combinations. CRP-ductin was identified as a cDNA sequence highly expressed in intestinal crypts (11). It exists in two alternately spliced forms, one of which contains a transmembrane domain. Ebnerin was accidentally cloned when a rat taste bud library was screened with probes coding for the amiloride-sensitive sodium channel. Ebnerin (we think it is a partial sequence) appears to be present in von Ebner's salivary gland rather than the taste bud cell, but is not a sodium channel homolog (37). DMBT1 is a gene frequently deleted in malignant brain tumors and was identified by genomic cloning of that region of chromosome 10q25-26 (47). The similarity of the domain structure of these proteins is striking. Unfortunately, comparison of the available sequences is hampered by the fact that each of the four has been cloned from a different species. Hence, it is unclear at present whether they represent alternately spliced forms of a single gene or multiple genes forming a new family.

Signal Transduction by Hensin: a New Pathway?

Studies of signal transduction by fibronectin had suggested a working model for hensin; the fibronectin receptor (₅₁ integrin) needs to be activated before it can cause fibronectin to form dimers and fibrils that are higher order multimers (66, 72). Activation of these integrin receptors dramatically increases their affinity for fibronectin; this higher affinity allows the protein to change its conformation in such a way that fibrillar fibronectin forms, which by itself is capable of activating signaling through the integrins (60). Perhaps a similar mechanism is responsible for hensin precipitation. In addition (or alternatively), hensin could localize to the ECM because it binds to another protein. We recently discovered that hensin binds to a novel small-molecular-weight protein of 27 kDa that is localized to the ECM (C. Hikita and Q. Al-Awqati, unpublished observations). Its role awaits the identification of the full-length sequence.

The conversion of the -phenotype to an -phenotype is associated with a large number of changes, i.e., retargeting of kAE1 and the H⁺-ATPase, induction of apical endocytosis, and rearrangement of the apical membrane from microvilli to microplicae and several other cytoskeletal rearrangements. To begin to understand these events, we need a target downstream from the signal induced by hensin. Since kAE1 will be retargeted to the basolateral membrane, we attempted to identify proteins that bind to it.

In erythrocytes, the cytoplasmic domain of AE1 connects the cortical actin cytoskeleton through its binding to ankyrin, which in turns binds the actin-binding protein spectrin. Mapping of the ankyrin binding sites showed that at least one of these domains is located in the extreme NH₂ terminus of AE1 (43). The NH₂-terminal three exons of AE1 are spliced out in the intercalated cell (7, 33, 68), resulting in the loss of ankyrin binding, at least in vitro (17, 69). Remarkably, ankyrin and fodrin colocalize with kAE1 in the basolateral membrane of -intercalated cells (18). However, recent studies have shown that the ankyrin of intercalated cell is a product of the Ank3 gene, whereas the red cell forms (Ank1 and Ank2) are absent in the kidney (52). Because the Ank repeat regions of all these ankyrins are homologous, it is likely that kAE1 and Ank3 do not interact; however, no direct studies have been published. To investigate whether other proteins might bind to the cytoplasmic domain of kAE1, we used the yeast two-hybrid system and identified a protein (which we term kanadaptin, for kidney anion exchanger adaptor protein) that binds to the cytoplasmic domain of kAE1 but does not bind to AE1 (10). In the kidney, kanadaptin is expressed only in the collecting tubules in both intercalated and principal cells. In the -intercalated cells, it labels vesicles that carry kAE1, but does not colocalize with the exchanger when it is retained in the basolateral membrane. Its sequence is new to the database and contains a proline-rich region that is predicted to allow binding to src-homology 3 (SH3)-domain-containing proteins. This suggests that this adaptor protein might participate in signaling.

We tested a small library of SH3-containing proteins in the yeast two-hybrid system and found that most such domains did not interact with kanadaptin. Specifically, spectrin, which contains SH3 domains, did not interact with it. The only protein (of the 10 tested) that bound to it was the oncogene vav, which contains, in addition to its SH3 domain, a GEF domain capable of activating GDP:GTP exchange of the small-molecular-weight G proteins, rho and rac (32). Rho has been implicated in targeting vesicles to the polarized bud in yeast, and rac functions in binding to the cytoskeleton (34, 50). However, we do not know at present whether vav is expressed in the intercalated cell (it is not present in total kidney mRNA). Finding proteins that interact with kanadaptin should allow us to begin to understand the mode of regulation of trafficking of the kAE1 vesicles.

Acid-Base Balance and In Vitro Plasticity

Seeding Density and Terminal Differentiation

During nephrogenesis, early epithelial cells (proto-epithelial cells) of the primitive nephron express cell junctions (tight, adherent, and focal adhesions), as well as differentiated ECM proteins. However, they lack many of the specialized structures that develop later as the cells mature to become the 14 recognizably different renal epithelial cells. Activation of the differentiation programs results in induction of new cytoskeletal proteins resulting in changes of cell shape (columnarization) and reorganization of the apical cytoplasm to form brush border microvilli. Terminal differentiation in other organs such as the intestine is also associated with columnerization of the cell and induction of microvillar proteins (25, 38). Remarkably, the -to- conversion in vitro recapitulates both of these processes.

It had been suggested that some intercalated cells in situ are not well-differentiated, since they occasionally exhibit nonpolarized staining patterns of kAE1 and the H⁺-ATPase in situ. In addition, the observation that -cells could lead to the generation of -type (23, 59) as well as principal cells (23) suggested that they might represent an earlier phenotype in the spectrum of differentiation. We found that low-density intercalated cells were flat and had a surface area almost three times that of high-density cells. The height of low-density cells was almost half that of high-density cells. Furthermore, previous studies using scanning electron microscopy had shown that the apical surface of -intercalated cells in situ had only a few microvilli, whereas that of -cells had thick ridges termed microplicae (30, 36, 61). These results had suggested that the apical cytoplasm and cytoskeleton might be different in the two cell types, a notion strengthened by our finding that -cells had no apical endocytosis, whereas -cells exhibited vigorous apical endocytosis (59, 65). We recently examined the composition of the apical cytoskeleton in the clonal intercalated cell line and found that the low-density cells had none of the critical components of microvillar structure; apical actin, cytokeratin 19, or villin. On the other hand, high-density cells had high expression of apical actin, cytokeratin 19, and villin. Similar to the studies in situ, we also found by scanning electron microscopy that the apical surface of low-density cells was simplified with only a few microvilli, whereas high-density cells had very thick microplicae (S. Vijayakumar and Q. Al-Awqati, unpublished observations).

Future Directions