Ion Pumps in Polarized Cells: Sorting and Regulation of the Na ,K - and H ,K -ATPases*

The physiologic function of an ion transport protein is determined, in part, by its subcellular localization and by the cellular mechanisms that modulate its activity. The Na (cid:1) ,K (cid:1) -ATPase and the H (cid:1) ,K (cid:1) -ATPases are closely related members of the P-type family of ion transporting ATPases. Despite their homology, these pumps are sorted to different domains in polarized epithelial cells, and their enzymatic activities are subject to distinct regulatory pathways. The molecular signals responsible for these properties have begun to be elucidated. It appears that a complex array of inter- and intramolecular interactions govern trafficking, distribution, and catalytic capacities of these proteins. The P-type ATPase family, whose members include the Na (cid:1) ,K (cid:1) -, H (cid:1) ,K (cid:1) -, and Ca 2 (cid:1) -ATPases, comprises a nearly ubiquitous collec-tion of ion pumps with catalytic activities that drive myriad phys- iologic processes (1). These pumps are responsible for such funda-mental manifestations of cellular homeostasis as the maintenance of osmotic balance and intracellular ionic composition. Further-more, almost every transport operation performed by the cells of epithelial tissues is coupled in some fashion to the action of a P-type ATPase (2). The enzymatic activities of these pumps consti- tute some of the principal means through which all animal cells convert the energy embodied in ATP into electrochemical gradients that can be exploited by all manner of metabolic pathways. The function of any ion transport can be appreciated at the backbone structure of the Ca 2 (cid:1) -ATPase (78). The yellow residues correspond to the NH 2 -terminal 85 residues of the H (cid:1) ,K (cid:1) -ATPase, which serves in all pump chimeras as an epitope tag that permits the chimeras to be distinguished from the host cell’s own population of sodium pumps. This portion of the H (cid:1) ,K (cid:1) -ATPase does not confer any apical sorting information (25). The turquoise residues correspond to the predicted fourth transmembrane domain of the H (cid:1) ,K (cid:1) -ATPase, which is sufficient to direct the apical accumulation of those Na (cid:1) ,K (cid:1) -ATPase/H (cid:1) ,K (cid:1) -ATPase chimeric molecules that include this H (cid:1) ,K (cid:1) pump domain. The red residues and the green residues correspond to the sequences that flank TM4 on its cytoplasmic and extracytoplasmic sides, respectively. Chimeras incorporating these H (cid:1) ,K (cid:1) -ATPase ectodomain or cytoplasmic domain residues alone in the background of the Na (cid:1) ,K (cid:1) -ATPase (cid:1) -subunit behave as basolateral proteins. In contrast, a chimera that incorporates both sets of flanking residues behaves as an apical protein. Thus, apical sorting information can be com-municated by TM4 or by its flanking domains, demonstrating that the apical sorting determinant is the product of transmembrane conformational interactions (25).

basolateral Na ϩ ,K ϩ -ATPase to the apical surface in transfected epithelial cells has been identified ( Fig. 1). This motif resides entirely within the fourth domain (TM4) of the ten predicted transmembrane domains of H ϩ ,K ϩ -ATPase ␣-subunit (25).
The majority of the polypeptides that have been shown to play a role in membrane protein sorting, such as cytoskeletal elements, adapters, and COP proteins, are soluble and reside in the cytosol (27). The exit of newly synthesized Na ϩ ,K ϩ -ATPase from the Golgi complex, for example, appears to require an intact Golgi cytoskeleton containing ␤I spectrin and ankyrin Gly-119. Disruption of this matrix results in the retention of a specific subset of membrane proteins, including the sodium pump, within a subcompartment of the Golgi complex (28). It is somewhat surprising, therefore, that information capable of directing the apical sorting of a pump chimera can be entirely buried within the plane of the membrane and thus inaccessible to interpretation by this molecular machinery. This same problem is confronted by proteins anchored to the membrane via covalent attachment to glycosylphosphatidylinositol (GPI). 1 In many epithelial cell types, GPI-linked proteins accumulate predominantly at the apical plasmalemma (29). The signal responsible for this anisotropic distribution appears to be attributable to the GPI anchor itself, which is able to associate in the plane of the membrane with glycosphingolipid (GSL) and cholesterol molecules to form "lipid rafts" (30). These lipid rafts, which coalesce in subdomains of the Golgi complex, appear to give rise to apically directed transport vesicles (31). Thus GPI-linked, as well as at least some transmembrane proteins, are apparently sorted to the apical surface by virtue of their capacity to partition into GSL-rich rafts during their biosynthetic passage through the Golgi (32). Presum-ably, the targeting proteins that endow these nascent vesicles with their apical specificity are incorporated into GSL-rich rafts through a similar partitioning mechanism. It has recently been suggested that the vectorial basolateral targeting of the Na ϩ ,K ϩ -ATPase (33,34) depends upon its exclusion from GSL-rich raft domains during its passage through the Golgi complex (35,36).
Numerous proteins that are capable of partitioning into GSLrich rafts resist detergent solubilization in 1% Triton X-100 at 4°C (37). This is not the case, however, for the H ϩ ,K ϩ -ATPase. The native H ϩ ,K ϩ -ATPase and all of the chimeras that incorporate its TM4 segment are fully soluble under these conditions, as is the Na ϩ ,K ϩ -ATPase (25). These data indicate that the H ϩ ,K ϩ -ATPase TM4 does not appear to exert its influence on sorting through interactions with lipid molecules but may instead mediate intramembranous protein-protein associations or alter the global conformation of the pump protein to create a motif that is accessible to interactions with cytoplasmic polypeptides. Support for the global conformation hypothesis derives from the sorting behaviors of the non-gastric H ϩ ,K ϩ -ATPase proteins. These heterodimeric pumps are close relatives of both the Na ϩ ,K ϩ -and gastric H ϩ ,K ϩ -ATPases and appear to play a role in potassium reabsorption in the colon and the kidney (38). Although there is physiologic evidence to suggest that (as their name implies) these pumps transport protons in exchange for potassium, recent functional expression studies indicate that sodium may in fact be the preferred counterion in potassium transport (39,40). The nongastric H ϩ ,K ϩ -ATPase ␣-subunits share ϳ60% protein sequence identity with both their Na ϩ ,K ϩ -and gastric H ϩ ,K ϩ -ATPase ␣-subunit counterparts (41). Alignment of the protein sequences of the ␣-subunits from these three pump families reveals that the TM4 segments of the non-gastric pumps are highly homologous to that of the sodium pump and markedly divergent from that of the gastric H ϩ ,K ϩ -ATPase. If the TM4 motif were necessary to account for the apical sorting of the gastric H ϩ ,K ϩ -ATPase, then the non-gastric H ϩ ,K ϩ -ATPases would be expected to behave as basolateral proteins. Transport studies performed on intact epithelial tissues, however, indicate that these pumps are functionally present in the apical plasma membrane (42). Similarly, when exogenously expressed by transfection in Madin-Darby canine kidney cells the human non-gastric H ϩ ,K ϩ -ATPase, ATP1AL1, also accumulates at the apical surface (43). Thus, although the gastric H ϩ ,K ϩ -ATPase TM4 transmembrane motif is clearly sufficient to specify apical sorting, other pump determinants appear to participate in this targeting process.
The generation of further gastric H ϩ ,K ϩ -ATPase/Na ϩ ,K ϩ -ATPase chimeras provides additional support for the contention that the TM4 sequence does not act alone. In fact, the discontinuous sequence domains that flank the fourth transmembrane domain can cooperate to recapitulate the apical sorting signal of the fourth transmembrane domain ( Fig. 1) (25). Thus, the apical sorting determinant appears to be the product of a conversation between the fourth transmembrane domain and the motifs that abut it. Interestingly, a similar conformational interaction dramatically influences the cation selectivities of these pump chimeras (44,45). The precise mechanisms through which these sequences cooperate remain to be determined.

Regulation of Ion Pump Function
Several factors have been identified that regulate Na ϩ ,K ϩ -ATPase activity. Most of the effects have been studied in the kidney, where the physiologic regulation of Na ϩ ,K ϩ -ATPase function has been well established. Acute regulation of the pumpmediated catalysis may be accomplished by either directly modulating the activity of the enzyme or by changing its localization. The cAMP-dependent protein kinase (PKA), protein kinase C (PKC), and tyrosine phosphorylation of the ␣-subunit of the Na ϩ ,K ϩ -ATPase all appear to be potential participants in the short-term regulation of Na ϩ ,K ϩ -ATPase activity. Steroid and thyroid hormones are responsible for more long-term regulation of pump capacity by exerting their effects at the transcriptional level (46). The evidence supporting each of these signaling pathways and their relevance to the modulation of pump function are discussed in great depth in an excellent recent review (47).
It is worth noting here, however, that the short-term control of The positions of residues exchanged in four different Na ϩ ,K ϩ /H ϩ ,K ϩ -ATPase ␣-subunit chimeras are shown in this space-filling color representation superimposed upon the backbone structure of the Ca 2ϩ -ATPase (78). The yellow residues correspond to the NH 2 -terminal 85 residues of the H ϩ ,K ϩ -ATPase, which serves in all pump chimeras as an epitope tag that permits the chimeras to be distinguished from the host cell's own population of sodium pumps. This portion of the H ϩ ,K ϩ -ATPase does not confer any apical sorting information (25). The turquoise residues correspond to the predicted fourth transmembrane domain of the H ϩ ,K ϩ -ATPase, which is sufficient to direct the apical accumulation of those Na ϩ ,K ϩ -ATPase/H ϩ ,K ϩ -ATPase chimeric molecules that include this H ϩ ,K ϩ pump domain. The red residues and the green residues correspond to the sequences that flank TM4 on its cytoplasmic and extracytoplasmic sides, respectively. Chimeras incorporating these H ϩ ,K ϩ -ATPase ectodomain or cytoplasmic domain residues alone in the background of the Na ϩ ,K ϩ -ATPase ␣-subunit behave as basolateral proteins. In contrast, a chimera that incorporates both sets of flanking residues behaves as an apical protein. Thus, apical sorting information can be communicated by TM4 or by its flanking domains, demonstrating that the apical sorting determinant is the product of transmembrane conformational interactions (25).

Minireview: Ion Pump
Sorting 29618 sodium pump function may involve regulated membrane trafficking. Dopamine stimulation of rat renal proximal tubules leads to endocytosis of Na ϩ ,K ϩ -ATPase from the basolateral plasmalemma through a mechanism that requires the involvement of class I(A) phosphoinositide 3-kinase (PI3K) (48 -52). PKC phosphorylation of a serine residue in the NH 2 -terminal tail of the Na ϩ ,K ϩ -ATPase ␣-subunit apparently permits PI3K to bind to an adjacent ␣-subunit proline-rich sequence via an SH3 domain in the p85 subunit of PI3K (52). This binding event may then recruit clathrin adapter complexes (51), leading to the internalization of the Na ϩ ,K ϩ -ATPase and the down-regulation of transport function.
The gastric H ϩ ,K ϩ -ATPase-driven secretion of acid into the lumen of the stomach is regulated at least as tightly as sodium pump function. In the resting state, the majority of H ϩ ,K ϩ -ATPase is restricted to the tubulovesicular elements (TVEs), a system of membranes that resides beneath the plasma membrane. Upon stimulation of the parietal cell by secretagogues there is a transient rise in the intracellular levels of cAMP, inositol trisphosphate, diacylglycerol, and Ca 2ϩ . Some combination of these second messengers induces the TVEs to fuse with the apical plasma membrane, thus allowing the H ϩ ,K ϩ -ATPase to secrete protons directly into the lumen of the gastric gland (53)(54)(55).
To date, it does not appear that the H ϩ ,K ϩ -ATPase itself undergoes any covalent modifications during the activation of acid secretion. Instead, the membrane fusion events regulate the ability of the pump to actively secrete protons. Upon the withdrawal of stimulation, the enzyme, along with large portions of the plasma membrane, is re-internalized to create the TVEs. The ␤-subunit of the gastric H ϩ ,K ϩ -ATPase contains a tyrosine-based endocytosis motif, which is necessary for the re-internalization of the holoenzyme (56). Transgenic mice expressing H ϩ ,K ϩ -ATPase ␤-subunit in which the critical tyrosine residue is mutated to an alanine fail to re-internalize the enzyme, leading to hypersecretion of acid and chronic gastritis. A similar endocytosis-based mechanism for H ϩ ,K ϩ pump regulation appears to function in the kidney as well, because the mice expressing the mutated H ϩ ,K ϩ -ATPase ␤-subunit also hyper-reabsorb potassium from their urine (57).

Pump Interacting Proteins
All of these regulatory phenomena, as well as the sorting processes discussed above, must be mediated by specific protein-protein interactions. It is becoming increasingly clear that ion transport proteins do not exist as isolated individuals in the membranes of living cells. Instead, they appear to participate in an extremely wide array of interactions that help to determine their localization, life span, and susceptibility to control by signals from second messenger systems (58). These protein scaffolds are capable of simultaneously incorporating a mixture of transport proteins, as well as a generous selection of kinases, phosphatases, adapter proteins, and other such arbiters of subcellular signaling and traffic control. It would appear that the cell does not communicate with its complement of transport proteins through the three-dimensional diffusion of regulatory molecules but rather through the organization of a two-dimensional solid state signaling matrix that permits extremely precise integration and spatial constraint. In light of their central role in the maintenance of cellular and organismic homeostasis, it seems highly unlikely that P-type ion pumps would be exempt from participating in these sorts of protein assemblies. Little is known, however, about the specific repertoires of polypeptides that comprise pump-associated scaffolds.
Biochemical studies have identified membrane and peripheral membrane proteins that may interact directly or indirectly with the Na ϩ ,K ϩ -ATPase. The small polypeptide with a single membrane-spanning segment, known as the ␥-subunit, is assembled with the Na ϩ ,K ϩ -ATPase in a subset of renal epithelial cell types and appears to influence pump kinetics (47,59). A sodium pumpassociated protein related to both the ␥-subunit and to phospholemman was recently isolated from the shark rectal gland (60). This protein appears to be a substrate for phosphorylation by both protein kinases A and C. Phosphorylation may destabilize this protein's apparently inhibitory association with the Na ϩ ,K ϩ -ATPase and thus serve to indirectly regulate pump function.
A large body of data establishes the existence of interactions between both the Na ϩ ,K ϩ -ATPase and the gastric H ϩ ,K ϩ -ATPase ␣-subunits and the cytoskeletal protein ankyrin (61)(62)(63)(64)(65). Through their associations with ankyrin, both of these pumps are attached to the meshwork of structural proteins, including spectrin and actin, that constitute the subcortical cytoskeleton. At least in the case of the sodium pump, this interaction is thought to be an important factor in maintaining the anisotropic distribution of the pump in polarized cells (35,66). In addition to its role in stabilizing pump localization, the ankyrin interaction may also exert a stimulatory effect on pump catalysis by altering the rates of conformational transitions (67). Ankyrin appears to bind to two discontinuous sequence domains present in loops 2-3 and 4 -5 of the ␣-subunit cytoplasmic domain (68 -70). The NH 2 -terminal tail of the H ϩ ,K ϩ -ATPase may also serve as a substrate for ankyrin association (65).
It has yet to be determined whether the pump-ankyrin interaction is subject to any form of dynamic regulation. In the case of the Na ϩ ,K ϩ -ATPase, this complex is thought to play a fairly static role in stabilizing the sodium pump in the plasma membranes of epithelial cells (35,66). In the gastric parietal cell, the cycling of the H ϩ ,K ϩ -ATPase between the apical cell surface and the TVEs necessitates that its association with ankyrin be transient. It is interesting to note in this context that disruption of the fodrin cytoskeleton, through genetic means in Drosophila (71) or via expression of dominant negative constructs in tissue culture cells (72), appears to result in an analogous redistribution of the majority of the Na ϩ ,K ϩ -ATPase to intracellular compartments.
Adducin is another spectrin binding protein that can interact directly with the sodium pump ␣-subunit. Adducin appears to serve as a link between spectrin and actin filaments. It can bind calmodulin and is a substrate for PKC and Rho-associated kinase-mediated phosphorylation (73). Adducin stimulates Na ϩ ,K ϩ -ATPase activity in vitro by accelerating the rate-limiting potassium deocclusion step in the catalytic cycle (67). The ␣-subunit residues that comprise the adducin binding site have yet to be identified. Polymorphisms in the sequence of an adducin polypeptide are correlated with hypertension, both in humans and in the Milan hypertensive strain of rats (74). Both rat and human adducins bearing these polymorphisms bind to the ␣-subunit with higher affinity than do their wild type counterparts (67). These observations suggest the interesting possibility that adducin association may be an important determinant in the regulation of renal Na ϩ ,K ϩ -ATPase activity under normal circumstances. Furthermore, in the context of certain adducin mutations, adducin-mediated overstimulation of renal pump function may lead to an expansion of extracellular fluid volume and thus to hypertension.

Future Directions
A more thorough understanding of the cell biologic regulation of ion pumps will require a more complete picture of the protein complexes in which they participate. Although the roster of pumpinteracting proteins enumerated above is enticing, it is almost certainly far from complete. Systematic searches for Na ϩ ,K ϩ -or H ϩ ,K ϩ -ATPase ␣-subunit-associated proteins have not yet been completed. Similarly, no potential partners for either pump's ␤-subunits have been identified to date. To understand the mechanisms that determine pump distribution and control the level of pump catalysis it will be necessary to undertake a thorough inventory of these binding partners.
Many studies of the regulation and trafficking of ion pumps make use of pump constructs expressed by transfection in tissue culture cell lines. The most meaningful systems in which to evaluate the importance of particular molecular signals and proteinprotein interactions in the determination of pump properties, however, are those that faithfully recapitulate the pump's normal physiologic setting. Recent work in the creation of genetic model systems has facilitated the study of the pumps in relation to physiologic systems. The genes encoding several pump subunit proteins have been disrupted by homologous recombination, and knock-out mice have been generated in which the activities of these pumps have been genetically abrogated (75)(76)(77). These model systems allow the role of each pump protein to be assessed in a biologically meaningful setting. They illuminate the pathways that converge to modulate pump activity and reveal the adaptations that occur to compensate for disruptions in pump function.

Minireview: Ion Pump Sorting 29619
Ideally, pump subunit constructs mutagenized to alter signals relevant to trafficking or regulation should be expressed in their native cell types, in their native tissues, and in the context of a physiologically intact organism. Such an approach has been utilized in the transgenic mouse studies that demonstrated the requirement for a tyrosine-based signal in the cessation of gastric acid secretion (56). To be interpretable, however, experiments of this sort must either investigate the effects of pump mutations that produce dominant phenotypes or else be performed in an expression system whose own endogenous complement of normal pump molecules has been silenced. The recent generation of knock-out mouse strains, in which the expression of various pump subunit polypeptides has been disrupted, provides precisely these sorts of expression systems. By observing both the cell biologic and physiologic phenotypic consequences of expressing mutated pump proteins in pump subunit knock-out mice it should be possible to determine whether and how a particular signal or interaction operates to control pump function under normal physiologic circumstances. Such studies will hopefully illuminate the manner in which perturbations of these interactions might reproduce such clinically significant pathologies as gastric ulcer disease or hypertension.