Mitochondria-Rich, Proton-Secreting Epithelial Cells

Several vertebrate transporting epithelia contain a subset of epithelial cells that are highly specialized for proton or bicarbonate secretion (Al-Awqati, 1996; Brown, 1989; Madsen and Tisher, 1986; Schuster, 1993; Steinmetz, 1986; Whitear, 1975). The presence of these characteristic cells in the kidney collecting duct (Schachowa, 1876), the amphibian urinary bladder (Schiefferdecker, 1881) and the amphibian epidermis (Schulze, 1876) has been known for over a century, and they are now known to play a vital role in proton translocation and related epithelial ion-transport processes. These cells have several typical features, listed in Table 1, that can be used to group them together as a family of cells that were originally referred to as ‘mitochondria-rich’ (MR) cells because they have more mitochondria than the adjacent cells from the same epithelium. The most important feature of these cells that is critical for their function is an extremely high expression of a vacuolar-type proton-pumping ATPase (H⁺V-ATPase) both on intracellular vesicles and on specific domains of their plasma membrane (Brown et al. 1988a). This review will focus on some key features of MR cells from a variety of epithelia.

Intercalated cells were first described as a distinct population of cells in the kidney collecting duct by Schachowa (1876). They have several characteristic morphological features revealed by electron microscopy, including many mitochondria, abundant intracellular vesicles and apical microvilli that are often very well-developed (Kaissling, 1982; Madsen and Tisher, 1986) (Fig. 1A). These cells, which comprise about 30–40 % of the total epithelial cells in collecting ducts, contain high levels of carbonic anhydrase II (Brown et al. 1983; Dobyan et al. 1982; Holthofer et al. 1987; Lonnerholm et al. 1985; Rosen, 1972a; Spicer et al. 1979), a feature that first indicated a potential role in distal urinary acidification. An interesting peculiarity of intercalated cells in rat kidney is that the Na⁺/K⁺-ATPase is virtually undetectable by immunocytochemistry on the basolateral plasma membrane, in contrast to most other renal tubular cell types (Kashgarian et al. 1985). However, these cells do contain an H⁺/K⁺-ATPase that may be involved in K⁺ reabsorption and proton secretion (Wingo and Smolka, 1995).

At least two subtypes of intercalated cells are present in the kidney cortex, so-called A (or alpha)-cells and B (or beta)-cells. These subtypes are involved in proton secretion and bicarbonate secretion, respectively, and their functions have been characterized in a number of studies (Al-Awqati, 1996; Brown, 1989; Kaissling, 1982; Madsen et al. 1991; Schuster, 1993; Steinmetz, 1986). A-cells have an H⁺V-ATPase on their apical plasma membrane and on intracellular vesicles (Alper et al. 1989; Brown et al. 1988a), and they express the alternatively spliced kidney form of the band 3 Cl⁻/HCO₃⁻ exchanger (AE1) on their basolateral plasma membrane (Alper et al. 1989; Drenckhahn et al. 1985; Holthofer et al. 1987; Schuster et al. 1991) (Figs 2, 3). B-cells contain no detectable AE1, but show a remarkable variability in the cellular location of the H⁺V-ATPase (Alper et al. 1989; Bastani et al. 1991; Brown et al. 1988a; Brown and Sabolic, 1993). Cells that are traditionally identified as B-cells by the absence of AE1 staining can have apical, basolateral, bipolar or diffuse intracellular vesicular staining with anti-H⁺V-ATPase antibodies in rat kidney (Fig. 3). However, in mouse and rabbit, cells with unique and strong basolateral staining for H⁺V-ATPase are rarely found – most cells appear diffusely stained by immunofluorescence in these species (Breton et al. 1995; Schuster et al. 1991). Basolateral membrane staining can, however, be detected by immunogold staining in some rabbit B-cells (Verlander et al. 1994). Intercalated cells express the so-called ‘kidney’ isoform of the H⁺V-ATPase 56 kDa subunit. A distinct ‘brain’ isoform of the 56 kDa subunit is expressed in the proximal tubule (Nelson et al. 1992). Interestingly, other MR cells that have been examined by immunocytochemistry also express the ‘kidney’ isoform of this 56 kDa subunit (Brown et al. 1992; D. Brown, unpublished data), with the exception of the osteoclast, which expresses the ‘brain’ isoform (see below).

The localization, function and regulation of Cl⁻/HCO₃⁻ anion exchange proteins in intercalated cells are especially complex (Emmons and Kurtz, 1993). New data have challenged the simple picture in which A-cells have basolateral AE1 and B-cells have a functionally detectable apical anion exchanger that has not yet been identified at the protein level (Schuster and Stokes, 1987; Weiner et al. 1994). A third type of intercalated cell, the gamma-cell, can have both apical and basolateral anion exchangers, at least in rabbit cortical collecting ducts (Emmons and Kurtz, 1994). Studies in cultured cells have suggested that the apical exchanger in B-cells is also AE1 (van Adelsberg et al. 1993), but other work has shown only very low levels of AE1 expression in B-cells (Fejes-Toth et al. 1994). Furthermore, no reports have yet shown reversal of polarity of AE1 by immunocytochemistry (Hayashi et al. 1988), although it is possible that apically located AE1 might not be recognized by available antibodies, perhaps because of some conformational change in the protein (van Adelsberg et al. 1993).

As mentioned above, the immunocytochemical detection of apical anion exchangers has remained elusive until recently. It has now been shown that AE3 isoforms of the Cl⁻/HCO₃⁻ exchanger can be expressed both apically and basolaterally in intercalated cells (Alper et al. 1995), thus adding an additional degree of complexity to an already complicated picture.

The capacity of A-type intercalated cells to secrete protons is regulated under physiological conditions by shuttling H⁺V-ATPase molecules to and from the plasma membrane via a specialized intracellular acidic vesicular compartment (Bastani et al. 1991, 1994; Brown, 1989; Madsen et al. 1991; Schuster, 1993; Schwartz et al. 1985). Under conditions of systemic acidosis or basolateral perfusion of isolated collecting tubules with high-[CO₂] solutions, vesicle exocytosis results in the delivery of more pumps to the apical plasma membrane and an increase in H⁺ secretion. Upon reversal of the stimulus, pumps are removed from the membrane by endocytosis. The vesicles that are involved in this shuttling process are highly characteristic and have a well-defined ‘coat’ structure formed of the peripheral, cytoplasmic V₁ subunit complexes of H⁺V-ATPases (Brown et al. 1987a) (Fig. 4A,E). Protein-A–gold immunocytochemistry using antibodies against the H⁺V-ATPase revealed a dense labeling of the apical plasma membrane (Fig. 4B) and of the cytoplasmic side of the coated vesicles (Brown et al. 1987a). Large regions of the cytoplasmic surface of the apical membrane of toad bladder MR cells are decorated by stud-like projections that are seen as regular, hexagonal arrays by the rapid freezing and freeze-drying technique (Fig. 4C) (Brown et al. 1987a). Identical arrays of studs are also seen on phospholipid liposomes in which purified H⁺V-ATPase molecules have been reconstituted, providing strong evidence that this coating material is related to the H⁺V-ATPase (Brown et al. 1987a). Rod-shaped intramembranous particles (IMPs) are found on the plasma membrane of MR cells (Fig. 4D) and on intracellular vesicles by freeze–fracture electron microscopy (Humbert et al. 1975; Orci et al. 1981; Stetson et al. 1980). The endocytotic process involved in H⁺V-ATPase recycling does not involve clathrin (Brown and Orci, 1986; Brown et al. 1987b), and the mechanism underlying this internalization process is unknown.

Recent work has also failed to detect caveolin, a protein associated with another endocytotic pathway involving caveolae (Rothberg et al. 1992), in these endocytotic vesicles (S. Breton and D. Brown, unpublished results). In collecting duct principal cells, the ‘kidney’ isoform of the 56 kDa peripheral V₁ subunit has been detected in association with water-channel-containing endosomes, which do not contain other subunits of the proton pump and which do not acidify their lumen (Sabolic et al. 1992). It is possible, therefore, that the 56 kDa subunit has a function that can be independent of the H⁺V-ATPase and that is, in some as yet undefined way, involved in a novel endocytotic mechanism in intercalated cells, as well as in other cell types.

In parallel with the cellular redistribution of H⁺V-ATPases during acidosis and alkalosis, it has been shown that the expression of AE1 protein and/or mRNA in A-type intercalated cells is increased during acidosis and decreased during alkalosis (Alper et al. 1991; Da Silva et al. 1991; Fejes-Toth et al. 1994). These changes would modulate proton and bicarbonate secretion in parallel with the observed rearrangement of H⁺V-ATPases under these conditions.

It is still controversial whether these intercalated cell phenotypes represent distinct cellular subtypes or whether polarized AE1 expression and H⁺V-ATPase localization can both be reversed within the same cell, under different physiological conditions (Al-Awqati, 1996; Koichi et al. 1992; Madsen et al. 1991; Schuster, 1993). Results from a number of studies are inconclusive, because of different possible interpretations of the data, and a complete discussion of this complex issue is not the aim of this review. However, cells with basolateral AE1 always have apical or subapical H⁺V-ATPases, but it is clear that AE1-negative cells, classified therefore as B-cells, can also have apical H⁺V-ATPases (Fig. 3C) (Alper et al. 1989). H⁺V-ATPase staining of intercalated cells cannot, therefore, be used alone to distinguish A- and B-cells in the renal cortex. Thus, transcytosis of H⁺V-ATPases from apical to basolateral membranes, or vice versa, appears to be a feature of AE1-negative cells, but may not occur in ‘typical’ A-cells. For example, basolaterally located H⁺V-ATPases have never been found in the inner stripe or inner medulla of the collecting duct, where only ‘typical’ A-cells are found. However, it should be borne in mind that, as mentioned above, the intensity of basolateral AE1 staining of A-intercalated cells in the cortex is dependent on acid–base status and further confounds the precise identification of A- and B-cells in this region of the kidney.

Recent data from our laboratory have also shown that A-cells and B-cells have different intracellular trafficking pathways for the lysosomal glycoprotein LGP120 that may parallel the ability of B-cells (but not that of A-cells) to insert H⁺V-ATPases into apical and basolateral membrane domains (Breton and Brown, 1995). B-cells may insert newly synthesized LGP120 first into their basolateral plasma membrane, before re-targeting the protein to lysosomes. In contrast, LGP120 appears to be directly targeted to lysosomes in A-cells, with no detectable accumulation at the cell surface. Finally, using FITC–dextran as a marker of fluid-phase endocytosis in the kidney, we have found that not only typical A-cells with basolateral AE1 have a high rate of apical endocytotic activity (Fig. 5), but that AE1-negative cells (B-cells) in the collecting duct and connecting tubule often also show an extremely high rate of apical endocytosis (S. Breton, P. King and D. Brown, unpublished results). Many previous studies, particularly in isolated perfused tubules and cell culture (Koichi et al. 1992; Schwartz et al. 1985; van Adelsberg et al. 1994), have utilized apical endocytosis as a means of classifying A-cells (endocytosis-positive) and B-cells (endocytosis-negative). Clearly, this is an oversimplification, because both cell types can display active apical endocytosis in vivo. In addition, remodeling of bicarbonate-secreting B-cells, probably by endocytosis of apical membrane components including an anion exchanger and a protein binding peanut lectin, has been shown to occur after an acid load in isolated perfused rabbit collecting tubules (Satlin and Schwartz, 1989). Recently, an extracellular matrix protein has been proposed as a regulator of intercalated cell phenotype on the basis of its ability to reverse the functional polarity of intercalated cells in culture (van Adelsberg et al. 1994).

Intercalated cells are found in collecting ducts and connecting segments prior to birth. In the rat kidney, they are most common in the developing papilla and rarer in the outer medulla and cortex (Kim et al. 1994; Matsumoto et al. 1993). Cells with apical H⁺V-ATPases and cells with basolateral H⁺V-ATPases appear in collecting ducts and connecting segments simultaneously during development. In the 3 weeks after birth, the situation reverses and the intercalated cell population increases in the cortex and outer medulla, but decreases considerably in the inner medulla (Brown et al. 1983; Kim et al. 1994; Madsen et al. 1988). It is not known to what extent selective cell death is involved in the disappearance of these cells in the inner medulla, although apoptotic figures can be found in some B-intercalated cells in this region after birth, while A-cells appear to be deleted by extrusion from the epithelium and shedding into the lumen (Kim et al. 1996). Alternatively, it has been repeatedly suggested that the different cell types in the collecting duct are developmentally related, and that principal and intercalated cells may interconvert (Evan et al. 1980; Hancox and Komender, 1963; Ordonez and Spargo, 1976; Richet and Hagege, 1975). Although there is no direct evidence in vivo showing such a relationship, studies in vitro using immuno-isolated A- and B-intercalated cells and principal cells indicate that antigens characteristic of one cell type can appear in cultures that originally did not contain this cell type. Thus, purified cultures of B-intercalated cells eventually give rise to cells that stain positively with monoclonal antibodies directed against A-cell and principal cell antigens (Fejes-Toth and Naray-Fejes-Toth, 1992, 1993).

Our recent studies on mice that are deficient in carbonic anhydrase type II show that, at birth, both intercalated and principal cells are present in medullary collecting ducts, but that in the adult mouse, intercalated cells are virtually absent from all regions of the medulla (Breton et al. 1995). These results suggest either that intercalated cells are selectively depleted from these collecting ducts or that they convert during postnatal differentiation into principal cells. The total number of cells in the collecting ducts of these mice is unchanged compared with normal littermates, indicating either a conversion of intercalated cells into principal cells or a concomitant increase in principal cell proliferation to replace intercalated cells lost by an as yet undetermined mechanism. Our finding of some ‘hybrid’ cells that contain both intercalated (AE1) and principal (aquaporin 2) cell markers (S. Breton and D. Brown, unpublished data) suggests that earlier work, which provided circumstantial evidence for a potential interconversion of these two functionally quite distinct cell types, merits further investigation.

Finally, cells with rod-shaped IMPs, which are present in all MR cell membranes and may represent the ‘V₂’ membrane region of H⁺V-ATPases (see below for discussion of rod-shaped IMPs), are present even in the rabbit mesonephric kidney (Schiller and Tiedemann, 1981), and more recent data have confirmed the presence of A-type intercalated cells in mesonephric rat and rabbit kidney (Matsumoto et al. 1993, 1994), indicating a complex developmental relationship among the different collecting duct cell types.

Special epithelial cell subtypes were first described in amphibian epidermis by Rudneff (1865) and in amphibian urinary bladder by Schiefferdecker (1881). While the nature of these cells was unknown, the camera lucida illustrations prepared by these authors provided a highly accurate reflection of the morphology and distribution of MR cells in these tissues. Similar cells are also found in the turtle urinary bladder, and this system proved to be a powerful model in which to examine transepithelial proton secretion in many pioneering studies (Steinmetz, 1986).

The turtle bladder is an epithelium with a high capacity to translocate protons. It has been widely used as a model system to examine different aspects of proton secretion, including the cellular biology of the specialized cell type that is primarily responsible for acid–base transport, the MR cell or the CAII-rich cell, as it is commonly referred to in this tissue (Schwartz et al. 1972, 1982; Steinmetz, 1985). A series of functional studies on the turtle bladder provided critical insights into mechanisms of acid–base handling by the renal collecting duct, and these are discussed in reviews by Steinmetz (1985, 1986). The MR cells in the turtle bladder have many of the morphological features of renal intercalated cells, but so far immunocytochemical localization of H⁺V-ATPases in reptilian bladder has not been performed. However, using the characteristic membrane-associated studs (originally named portasomes) as a morphological marker of the V₁ peripheral region of H⁺V-ATPases (Harvey, 1980), it was shown that α- and β-type MR cells having an opposite polarity of the H⁺V-ATPase could be identified (Stetson and Steinmetz, 1985). On the basis of this observation, a model for proton and bicarbonate transport across transporting epithelia, activated by a proton-pumping V-ATPase, was developed (Stetson et al. 1985) and the bicarbonate-secreting cells were identified as MR cells (Fritsche and Schwartz, 1985). Bicarbonate secretion is linked to the activity of an H⁺V-ATPase, presumably located in the basolateral membrane of β-type MR cells (Fritsche et al. 1989). In addition, vibrating probe studies provided direct evidence that MR cells were the source of the proton current in turtle bladder (Durham and Nagel, 1986). The turtle bladder was also the first tissue in which regulation of apical proton secretion by exo- and endocytosis of acidic vesicles containing H⁺V-ATPases was demonstrated (Cannon et al. 1985; Dixon et al. 1986; Gluck et al. 1982; Stetson and Steinmetz, 1983).

In parallel with observations on the turtle bladder, the amphibian bladder was also found to be an acidifying epithelium (Ludens and Fanestil, 1972). Cells bearing the hallmark features of MR cells are found in variable numbers in the urinary bladder of many amphibian species (Frazier, 1978; Keller, 1963; Orci et al. 1975; Peachy and Rasmussen, 1961; Rosen, 1972b; Wade, 1976) (Fig. 1B). A key observation that provided the basis for many studies on the cellular origin of the proton flux in this tissue was that toads (Bufo marinus) from the Dominican Republic and from Columbia had considerably different capacities to acidify their urine (Rosen et al. 1974). Morphological examination revealed that the poor acidification ability of Dominican toads correlated with a paucity of MR cells in the bladder epithelium, whereas Columbian toads, which could acidify vigorously, had considerably greater numbers of MR cells. Thus, the amphibian bladder provided another system in which acidification was related to the presence of specialized MR cells. The morphology of these MR cells was specifically modified during stimulation of acid secretion (Frazier, 1978), and MR cells appear to be specific targets for aldosterone binding in toad bladder (Sapirstein and Scott, 1975). The cytoplasmic ‘studs’ that coat the inner surface of MR cell membranes, as well as many intracellular vesicles, were first identified as cytoplasmic subunits of an H⁺V-ATPase by rapid-freeze, deep-etch microscopy studies on toad bladder membranes (Brown et al. 1987a).

The amphibian epidermis is another acid-secreting tissue that contains MR cells (Machen and Erlij, 1975). In the skin, they are often referred to as flask cells, because of their typical elongated shape, with a wide base tapering to a narrow neck. Epidermal MR cells are rich in carbonic anhydrase (Rosen and Friedley, 1973), have elongated IMPs seen by freeze–fracture (Brown et al. 1978) and show characteristic morphological features of the MR cell family (Whitear, 1975). Recent work has demonstrated a band-3-like anion exchanger in MR cells of Bufo marinus skin (Devuyst et al. 1993), but it was located apically. In contrast, we have found that a band-3-like anion exchanger is basolaterally located in MR cells, while proton pumps are apically located (S. Breton and D. Brown, unpublished data; Fig. 6). In addition to their role in transepithelial proton movement, it has also been shown that the H⁺V-ATPase in these cells energizes transepithelial Na⁺ transport in this tissue (Harvey, 1992), although the MR cells themselves are not the principal site of amiloride-sensitive transepithelial Na⁺ transport (Dorge et al. 1990; Ehrenfeld et al. 1976). However, several studies have shown that epidermal flask cells are the pathway for the high transepithelial chloride conductance in this tissue (Foskett and Ussing, 1986; Katz and Scheffey, 1986; Larsen et al. 1987). This conductance is regulated by β-adrenergic agonists via cyclic AMP (Willumsen et al. 1992). The functional state of epidermal MR cells is modulated by aldosterone (Devuyst et al. 1991a) and prostaglandins (Yorio et al. 1991).

The number of flask cells varies considerably among different regions of skin in the same animal and also among different species. In several species, the number of epidermal flask cells can be increased by maintaining the animals in distilled water and decreased by maintaining them in saline solutions (Budtz et al. 1995; Devuyst et al. 1991b; Ilic and Brown, 1980; Katz and Gabbay, 1988). Many flask cells are shed during each molting cycle and replaced by differentiation of epithelial cells, although the lineage of the new cells is not clear (Muto et al. 1990).

The rat epididymis and vas deferens contain a distinct population of epithelial cells that express high levels of H⁺V-ATPase on their luminal plasma membranes, as well as in intracellular vesicles (Brown et al. 1992) (Fig. 7). These cells have many morphological features of MR cells, including the presence of rod-shaped IMPs on their apical plasma membrane (Brown and Montesano, 1980). An earlier histochemical study showed a distinct population of epithelial cells in the epididymis that contained large amounts of carbonic anhydrase (Cohen et al. 1976), an indication that these cells might be involved in proton secretion. It was then shown that carbonic-anhydrase-rich cells appear gradually in the epithelium after birth and are present at adult levels after 4–5 weeks (Membre et al. 1985). We have shown that H⁺V-ATPases and CAII are co-localized in the same cell type (Breton et al. 1996). These cells are the so-called ‘apical’ cells or ‘narrow’ cells in the head (caput) portion of the epididymis, and the ‘light’ cells or ‘clear’ cells in the body (corpus) and tail (cauda) (Reid and Cleland, 1957; Sun and Flickinger, 1979). The morphology of apical and clear cells is quite distinct, although both share an extremely high level of apical endocytotic activity, similar to that of renal intercalated cells (Al-Awqati, 1996; Brown et al. 1987b; Moore and Bedford, 1979). Furthermore, aldosterone binds avidly to epididymal clear cells (Hinton and Keefer, 1985), and clear cells are selectively ablated from parts of the epididymis by ethane dimethane sulfonate (Klinefelter et al. 1994; Liu et al. 1993) in parallel with a decrease in fertility. The presence of CAII in the epididymis prompted studies to determine whether the CA inhibitor acetazolamide would inhibit luminal acidification. Conflicting data were obtained. In one study, acetazolamide did partially inhibit acidification (Au and Wong, 1980), whereas subsequent work showed no effect on epididymal pH (Caflisch, 1992), possibly due to a failure to deliver an appropriate dose of the inhibitor to the CAII-rich cells.

In collaboration with Dr Peter Smith at the Woods Hole Marine Biology Laboratory, MA, the proton-selective, non-invasive vibrating probe has been used to detect a bafilomycin-sensitive proton secretion in the lumen of the vas deferens (Breton et al. 1996). Bafilomycin is a specific inhibitor of the H⁺V-ATPase. Up to 80 % of the transepithelial proton flux in the vas deferens was inhibited by bafilomycin, implicating H⁺V-ATPase-driven proton transport as the major source of luminal acidification. Because the ‘apical’ and ‘clear’ cells contain large numbers of H⁺V-ATPase molecules on their apical plasma membranes, it is likely that they are analogous in function to renal intercalated cells and that they are primarily responsible for luminal proton secretion into the excurrent duct system. However, at least in rat epididymis, no basolateral anion exchanger has yet been identified, and the cells do not stain with anti-AE antibodies that strongly label the basolateral plasma membrane of renal intercalated cells (Breton et al. 1996; Brown et al. 1992).

The luminal fluid along much of the male reproductive tract is maintained at an acidic pH in many species (Carr et al. 1985; Levine and Marsh, 1971; Rodriguez et al. 1990). It has been proposed that low pH is required for sperm maturation and that it is involved in maintaining sperm in an immotile state during their passage through the epididymis and vas deferens, probably in conjunction with factors including specific proteins, weak acids and other ions (Hinton and Palladino, 1995). Previous work on the perfused epididymis implicated an Na⁺/H⁺ exchanger in the process of acidification (Au and Wong, 1980), but on the basis of our observations, it is likely that the H⁺V-ATPase-rich MR cells described above play a key role in the maintenance and modulation of luminal pH in this tissue.

Osteoclasts reabsorb bone by creating a specialized extracellular acidic space between the bone-facing domain of their plasma membrane and the bone surface. The osteoclast membrane in this region is highly ruffled, and its cytoplasmic surface is decorated with typical H⁺V-ATPase studs. Immunocytochemistry and functional assays have shown that this membrane contains a high level of H⁺V-ATPase (Blair et al. 1989; Vaananen et al. 1990), while the non-absorptive surface of the osteoclast contains a band-3-like anion exchanger (Baron, 1989; Blair et al. 1993). In addition, like other MR cells, osteoclasts express abundant cytosolic carbonic anhydrase (Baron, 1989). The H⁺V-ATPase is, however, distinct from the intercalated cell enzyme in that the 56 kDa subunit expressed in osteoclasts is the ‘brain’ isoform (Bartkiewicz et al. 1995; van Hille et al. 1994) that is mainly expressed in the kidney in proximal tubules and not in intercalated cells (Nelson et al. 1992).

The functional polarity necessary for vectorial proton secretion is generated in this non-epithelial cell by its adherence to the surface of bone (Baron, 1989). Interestingly, the bone-facing membrane surface also contains lysosomal membrane markers (reminiscent of the cell surface staining for LGP120 that is characteristic of B-type intercalated cells in the kidney), and soluble lysosomal enzymes are secreted from this surface (Baron et al. 1990). The lacuna formed between the osteoclast membrane and the bone surface can be considered in several respects to be an extracellular lysosome that is specialized for bone reabsorption. Intracellular vesicles also contain H⁺V-ATPase molecules in the osteoclast, and modulation of the surface density of proton pumps probably occurs by vesicle recycling in a fashion similar to that in other MR cells, including the kidney intercalated cell (Baron et al. 1990). Thus, as described in a review (Gluck, 1992), the osteoclast can be considered as a ‘unicellular proton-transporting epithelium’. Indeed, osteoclasts are derived from phagocytes, which also have H⁺V-ATPases on their plasma membranes (Grinstein et al. 1992).

Specialized transporting epithelia from many insects contain cells that closely resemble MR cells. For instance, ion-secreting cells of Malpighian tubules and sensory sensilla contain abundant mitochondria. In the lepidopteran midgut, the ‘goblet cell’, which also contains numerous mitochondria, has long been a model for active K⁺ transport, since K⁺ is the principal extracellular and intracellular cation in this tissue (Harvey et al. 1983). The apical surface of goblet cells in lepidopteran midgut is involuted to form a deep well surrounded by a highly specialized membrane with many microvilli (Anderson and Harvey, 1966). In this respect, goblet cells are similar to the flask cells of amphibian kidney collecting ducts (Brown, 1978). The microvillar membrane is decorated with cytoplasmic studs that resemble the F₁ region of mitochondrial F₁F₂-ATPase, and these studs were referred to as ‘portasomes’ because of their proposed involvement in transmembrane transport processes, particularly K⁺ excretion (Harvey, 1980; Harvey et al. 1983). It is now known that they are part of the peripheral ‘V₁’ region of an H⁺V-ATPase complex that is abundant in this membrane domain (Russell et al. 1990; Wieczorek et al. 1989, 1990). In the midgut, this high concentration of H⁺V-ATPases has a well-defined role in K⁺ secretion. The electrical gradient formed by active proton pumping is used to energize the secretion of K⁺, which is found at high levels in the diet of leaf-eating insects, via an H⁺/K⁺ antiporter that is located in the same membrane domain (Wieczorek et al. 1991; Lepier et al. 1994).

Similar portasomes are found in cells of the insect salivary gland, Malpighian tubules and all sensory sensilla (Harvey et al. 1983). In cockroach salivary gland, an H⁺V-ATPase has been localized in apical microvillar membrane domains in the distal duct cells (Just and Walz, 1994).

In addition to the examples discussed above, application of freeze–fracture microscopy or (immuno)cytochemical procedures to a variety of tissues has revealed cells that have some characteristics of MR cells, although the complete battery of diagnostic assays has not yet been performed for these cells. Listed in no particular order, these cells include the following types.

1. A characteristic mitochondria-rich pneumocyte in turtle lung, distinct from type I and type II pneumocytes, which has rod-shaped IMPs on its plasma membrane (Bartels and Welsch, 1984).

2. The supporting cells in rodent olfactory mucosa, which have rod-shaped IMPs (Kerjaschki and Horandner, 1976). However, the supporting cells do not have high levels of carbonic anhydrase. Instead, the adjacent olfactory sensory cells in this epithelium are rich in carbonic anhydrase (Brown et al. 1984b). This is an unusual example of these two features of the MR cell family being present in different cells within the same epithelium. Furthermore, some taste bud cells in rats are also rich in carbonic anhydrase, but other features that would link them to the MR cell family have not yet been studied (Brown et al. 1984a). The detection of high levels of carbonic anhydrase in these cells supported an earlier proposal that this enzyme was involved in the sense of taste (Hansson, 1960).

3. In the intestine of several species, including humans, scattered cells containing abundant carbonic anhydrase have been reported (Lonnerholm, 1977, 1983; Lonnerholm and Wistrand, 1983), and some cells with linear IMP arrays that resemble rod-shaped IMPs have been found (Neutra, 1979). About 1 % of the epithelial cells lining the hen coprodeum also contain large numbers of rod-shaped IMPs on their plasma membrane, but this number can be increased to 20 % by feeding animals a Na⁺-deficient diet (Eldrup et al. 1980). This observation prompted the authors to speculate that the rod-shaped IMPs were Na⁺ channels.

4. Specialized ‘flask’ cells in the amphibian kidney collecting duct have rod-shaped IMPs (Brown, 1978) and are rich in carbonic anhydrase (Jonas, 1981). These are likely to be the amphibian homolog of the mammalian kidney intercalated cell.

5. Rod-shaped IMPs have been detected in endometrial cells lining the uterus of rats after treatment with progesterone (Murphy et al. 1979) and after implantation (Murphy et al. 1982).

6. Proton pumps and cytoplasmic studs are present on membranes of cells of the chorioallantoic membrane villus cavity cells in chick embryos (Narbaitz et al. 1995). The acidification activity of these cells is required to eliminate protons produced during metabolism of the embryo and to release calcium from the shell for use in embryonic growth.

In all cases so far examined, cells that have membrane-associated rod-shaped IMPs by freeze–fracture electron microscopy also have an active plasma-membrane-associated proton-pumping ATPase, and it has been postulated that these IMPs are the membrane-embedded ‘V₂’ regions of (V₁V₂) H⁺V-ATPases. However, the reverse situation is not always true. Some cells that have immunocytochemically detectable H⁺V-ATPases on their plasma membrane have no detectable rod-shaped IMPs by freeze–fracture electron microscopy. Two examples are found in the kidney. The apical plasma membrane of the proximal tubule in rat has a high level of functional and antigenically detectable proton pumps, yet has no typical rod-shaped IMPs (Brown et al. 1988b; Orci et al. 1981). The same applies to the initial portion of the thin descending limb of Henle (long loops) in rat kidney (Orci et al. 1981). In contrast, the thick ascending limb of Henle contains proton pumps (Brown et al. 1988b) and has readily detectable rod-shaped IMPs that are located in small clusters throughout the apical plasma membrane (D. Brown and S. Breton, unpublished data). In the epididymis, apical cells in the head have typical rod-shaped IMPs (Brown and Montesano, 1980), yet clear cells in the body and tail do not (Brown and Montesano, 1981), despite expressing high levels of H⁺V-ATPase (Brown et al. 1992).

In addition, freeze–fracture studies on bovine medullary H⁺V-ATPases reconstituted into pure phospholipid vesicles did not reveal the presence of ‘convincing’ rod-shaped IMPs (Brown et al. 1987a). Thus, the identity of the rod-shaped IMP remains enigmatic. Although circumstantial evidence correlates the presence and number of these structures with the activity of MR cells (Stetson and Steinmetz, 1983; Stetson et al. 1980), it remains to be proved that they are indeed the morphological correlate of the V₂ transmembrane complex of the H⁺-translocating V₁V₂-ATPase. Alternatively, they could be a membrane protein that is often located in the same membrane domain as the H⁺V-ATPase or they could reflect an organizational state of the enzyme, perhaps in conjunction with specific lipid membrane components, that is variable among different cell types and under different physiological conditions.

In this review, we have brought together a considerable amount of scattered information related to a class of highly specialized cells collectively known as mitochondria-rich (MR) cells. These cells are found in many epithelia in which proton and/or bicarbonate secretion is functionally important. A variety of physiological stimuli modify proton transport and related ion-transport processes in these tissues, and MR cells are intimately involved in the regulatory response at several levels. Because of the highly polarized, but variable, expression of transport proteins in these cells, they make an intriguing model in which to examine the establishment and regulation of functional epithelial cell polarity.

Future studies on these cells will have several goals, including dissecting intracellular pathways and signaling molecules that are responsible for the polarized targeting of proton pumps and anion exchangers in these cells; elucidating factors involved in the developmental expression of MR cells and their characteristic battery of transporting proteins; determining the relationship between MR cells and the surrounding epithelial cells in adult tissues; and last, but not least, providing definitive evidence concerning the potential interconversion of MR cell subtypes, especially in the kidney collecting duct.

We are grateful to the many colleagues who have contributed to these studies. In particular, we thank Drs Seth Alper (Beth Israel Hospital, Boston, MA), Steve Gluck (Jewish Hospital, St Louis, MO) and Ivan Sabolic (Institute of Occupational Health, Zagreb, Croatia) for stimulating collaboration over the past several years. This work has been supported by an NIH grant DK42596 to D.B. for the past 5 years and by a grant to S.B. from the National Sciences and Engineering Research Council of Canada.