| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
PHYSIOLOGICAL REVIEWS Vol. 79 No. 1 January 1999, pp. S77-S107
Copyright ©1999 by the American Physiological Society
Laboratory of Cardiac/Membrane Physiology, and Laboratory of Molecular and Cellular Neuroscience, The Rockefeller University, New York, New York
I. INTRODUCTION
A. Domain Organization of the CFTR Molecule
B. CFTR Forms an Ion Channel
C. Overview of Regulation of CFTR Channel Function
D. CFTR in Cardiac Myocytes
II. REGULATION OF CYSTIC FIBROSIS TRANSMEMBRANE CONDUCTANCE REGULATOR BY PROTEIN KINASES
A. PKA
B. PKC
C. cGMP-Dependent Protein Kinases
D. Ca2+/Calmodulin-Dependent Protein Kinases
E. Protein Tyrosine Kinases
III. REGULATION OF CYSTIC FIBROSIS TRANSMEMBRANE CONDUCTANCE REGULATOR BY PROTEIN PHOSPHATASES
A. Candidate Phosphatases for Regulating CFTR
B. Functional Effects of Exogenous Phosphatases
C. Findings With Phosphatase Inhibitors
D. Differential Dephosphorylation of Multiple Sites
IV. MECHANISMS OF OPENING AND CLOSING OF PHOSPHORYLATED CYSTIC FIBROSIS TRANSMEMBRANE CONDUCTANCE REGULATOR CHANNELS
A. Phosphorylation of CFTR Controls ATP Hydrolysis and Channel Gating
B. How Does Phosphorylation by PKA Modify CFTR Function?
C. Distinct Functions of the Two NBDs
D. Which NBD Opens, and Which Closes, a CFTR Channel?
V. WORKING MODEL OF CYSTIC FIBROSIS TRANSMEMBRANE CONDUCTANCE REGULATOR'S CATALYTIC AND GATING CYCLES AND INFLUENCE OF PHOSPHORYLATION
A. Catalytic Cycles of the Two NBDs
B. Modulation of CFTR Channel Gating by Incremental Phosphorylation
C. Lingering Uncertainties
VI. CONCLUDING REMARKS
REFERENCES
 |
ABSTRACT |
|---|
 Top References
|
|---|
Gadsby, David C., and Angus C. Nairn. Control of CTFR Channel Gating by Phosphorylation and Nucleotide Hydrolysis. Physiol. Rev. 79, Suppl.: S77-S107, 1999. 
The cystic fibrosis transmembrane conductance regulator (CFTR) Cl
channel is the protein product of the gene defective in cystic fibrosis, the most common lethal genetic disease among Caucasians. Unlike any other known ion channel, CFTR belongs to the ATP-binding cassette superfamily of transporters and, like all other family members, CFTR includes two cytoplasmic nucleotide-binding domains (NBDs), both of which bind and hydrolyze ATP. It appears that in a single open-close gating cycle, an individual CFTR channel hydrolyzes one ATP molecule at the NH2-terminal NBD to open the channel, and then binds and hydrolyzes a second ATP molecule at the COOH-terminal NBD to close the channel. This complex coordinated behavior of the two NBDs is orchestrated by multiple protein kinase A-dependent phosphorylation events, at least some of which occur within the third large cytoplasmic domain, called the regulatory domain. Two or more kinds of protein phosphatases selectively dephosphorylate distinct sites. Under appropriately controlled conditions of progressive phosphorylation or dephosphorylation, three functionally different phosphoforms of a single CFTR channel can be distinguished on the basis of channel opening and closing kinetics. Recording single CFTR channel currents affords an unprecedented opportunity to reproducibly examine, and manipulate, individual ATP hydrolysis cycles in a single molecule, in its natural environment, in real time.
A. Domain Organization of the CFTR Molecule
The cystic fibrosis transmembrane conductance regulator (CFTR) protein, the single polypeptide product of the gene defective in cystic fibrosis (CF) patients (157, 166, 204), is now known to form a Cl
B. CFTR Forms an Ion Channel
The CFTR is the only ABC transporter so far established to function as an ion channel [despite an earlier suggestion of channel-like activity of Pgp (70, 136), subsequently disavowed (77)]. However, there is growing evidence that CFTR is also able to modify the activity of other ion channels, such as outwardly rectifying Cl The possibility that CFTR subserves other roles notwithstanding, several lines of evidence have led to the incontrovertible conclusion that CFTR itself comprises a metabolically gated Cl One of the most fundamental questions, presently unresolved for any ABC transporter, concerns the quaternary structure of the functional unit. A thorough immunoprecipitation analysis of coexpressed CFTR mutants bearing different epitopes suggested strongly that CFTR functions as a monomer (127). Analyses of electron microscopic images of purified Pgp were similarly interpreted as suggesting that active Pgp is monomeric, even though the size of the particles was consistent with that of a Pgp dimer (168). However, recent analysis of particle sizes for a range of recombinant membrane proteins, using freeze-fracture electron microscopy of oocyte membranes, led to the conclusion that CFTR exists as a dimer (56). Moreover, preliminary electrophysiological analysis of concatenated CFTR dimers, comprising linked monomers with different gating characteristics, has now prompted Zerhusen and Ma (229) to propose that a single CFTR channel contains two CFTR polypeptides. Interestingly, complementation and reconstitution studies of the E. coli ABC transporter Ars, responsible for arsenical extrusion, had already led to a model in which the transporter functions as the equivalent of a dimer in which the NH2-terminal NBD of each monomer interacts with the COOH-terminal NBD of the other to form a total of only two catalytic sites (114). C. Overview of Regulation of CFTR Channel Function
Although CFTR seems to be the only ABC molecule that forms an ion channel, recent biochemical measurements have demonstrated that, as previously shown for other ABC transporters (e.g., Refs. 179, 180), purified intact CFTR (113), as well as an NH2-terminal CFTR NBD (NBD1; Ref. 108) or COOH-terminal CFTR NBD (NBD2; Ref. 160) peptide, expressed in fusion with the maltose-binding protein, is able to hydrolyze ATP. Moreover, ATP hydrolysis by intact CFTR was found to be enhanced after phosphorylation by PKA (113). These biochemical results provide a satisfying corollary to a substantial body of functional data that has established that opening and closing of CFTR channels is linked to ATP hydrolysis (reviewed in Ref. 68) and that, even in the presence of millimolar MgATP, CFTR channels will not open unless they are first phosphorylated by PKA (3, 140, 194). As reviewed in sections IV and V, the details of this complex interplay between phosphorylation of CFTR, ATP hydrolysis at CFTR's NBDs, and CFTR channel gating remain incompletely understood. Nevertheless, it seems likely that, during each open-close channel gating cycle, a single phosphorylated CFTR channel hydrolyzes one molecule of ATP at NBD1 (Fig. 1) to open, and then hydrolyzes a second ATP at NBD2 to permit the channel to close (14, 25, 74, 75, 91). However, it is also evident that activation of CFTR channels by PKA-mediated phosphorylation is not simply an all-or-nothing process, because biochemical measurements show that PKA readily phosphorylates the R domain of CFTR to a stoichiometry of at least 5 mol/mol (52, 152), and electrophysiological measurements have now identified at least three functionally distinct phosphorylated states of individual CFTR channels that appear to reflect the different actions of specific protein phosphatases on phosphorylated CFTR (49, 59, 88, 120, 226). Our goal in this review is to critically evaluate the experimental evidence that has led to this complex gating mechanism for CFTR channels in which phosphorylation and dephosphorylation of multiple sites in the R domain, and perhaps other domains of the protein, orchestrates the cycles of ATP binding and hydrolysis at the two NBDs that regulate opening and closing of the anion-selective pore. D. CFTR in Cardiac Myocytes
About a decade ago, mammalian cardiac ventricular myocytes were found to display a PKA-activated Cl A. PKA
1. Red herring: outwardly rectifying Cl
I. INTRODUCTION
Top
Next
References
channel subject to complex regulation (for reviews, see Refs. 68, 165, 216). However, this was far from obvious when the gene was first identified, because the predicted amino acid sequence of CFTR (166, 167) placed it squarely in the superfamily of ATP-binding cassette (ABC) transporters. Other well-known members of this large and still growing family include a host of bacterial periplasmic transporters, each selective for a particular substrate, the yeast a-mating factor exporter (STE6), and the mammalian P-glycoprotein (Pgp) linked to multidrug resistance. All known ABC transporters seem to require two nucleotide-binding domains (NBDs) and two transmembrane domains (TMDs) to function normally (82). Often, particularly in the case of the bacterial ABC transporters, separate genes encode these individual domains, or fused pairs of domains in various combinations (i.e., TMD-TMD, NBD-NBD, or TMD-NBD). An important implication of expressing several domains of a transporter as separate gene products is that the individual polypeptides must contain the necessary information to enable them to self-assemble into a fully functional transporter. It seems likely that most of those important interactions between domains have been preserved also in ABC transporters expressed from a single gene and that the nature of those interactions may provide clues as to how all of these molecules function. The CFTR is one example in which all of the functional domains are encoded by a single gene. Accordingly, the NH2- and COOH-terminal halves of CFTR both contain a TMD comprising six putative membrane-spanning 
-helices followed by an NBD (Fig. 1), and the two halves are linked by a cytoplasmic regulatory (R) domain (not found in other ABC transporters) that incorporates multiple sites for phosphorylation by cAMP-dependent protein kinase (PKA) and protein kinase C (PKC; Ref. 166). The original definition of the domain boundaries in CFTR was guided largely by the location of intron-exon junctions and by the extent of regions of sequence similarity with other ABC transporters (166). Although the proposed topology and overall organization of these domains in CFTR have proven remarkably resilient, very recent information from homology modeling (e.g., Ref. 6) and functional studies of CFTR (e.g., Ref. 29), and from the high-resolution structure of an NBD from the Escherichia coli ribose ABC transporter (7), makes it seem likely that NBD1 extends up to 60 residues into the R domain as initially defined (166). Presumably, if this new domain boundary is correct, the COOH-terminal limit of NBD2 in CFTR will need to be similarly extended.
View larger version (58K):
[in a new window]
FIG. 1.
Proposed topological model of cystic fibrosis transmembrane conductance regulator (CFTR) showing cytosolic NH2 (N) and COOH (C) termini, nucleotide binding domains (NBD1, NBD2), and regulatory (R) domain, predicted membrane-spanning -helices (M1-M12), and glycosylation sites in M7-M8 extracellular loop (166). Thickened section of M2-M3 cytoplasmic loop represents 30 amino acids encoded by exon 5, believed spliced out of cardiac CFTR isoform (86). Lengths of extracellular and intracellular loops are in rough proportion to number of residues they contain. NBD1 and NBD2 are drawn with fold of p21-ras (149) to emphasize proposed broad functional, and possibly even structural, homology between CFTR's NBDs and catalytic sites of G proteins (cf. Refs. 26, 68, 129). NBDs are drawn in close proximity to each other and in contact with R domain to emphasize likelihood that two NBDs directly interact with each other and with R domain. [From Gadsby and Nairn (69).]
channels (53, 66) and amiloride-sensitive Na+ channels (190-192), but it remains unclear whether such modulation is direct, resulting from association of CFTR with other channels (110), or from a chain of such protein-protein interactions perhaps beginning at CFTR's COOH-terminal PDZ (postsynaptic density protein, disc-large, ZO-1)-domain binding motif (76, 106, 158, 185, 213,), or is indirect, as suggested for CFTR's apparent influence on outward rectifier Cl channels (175). Interestingly, it has recently become clear that another member of the ABC family, the sulfonylurea receptor (SUR; Ref. 96), forms an oligomeric complex with inward rectifier K+ channel subunits (the stoichiometry is 4 SUR per tetrameric K+ channel) and regulates their function (36, 187). This regulation appears to depend on ATP hydrolysis at the NBDs of SUR but occurs by a mechanism that is presently unknown (e.g., Ref. 8). pore. The most straightforward evidence is the demonstration that recombinant CFTR expressed in Sf9 cells can be purified after solubilization with detergent, and then renatured and incorporated into lipid bilayers where, upon activation by PKA catalytic subunit plus ATP, it gives rise to typical small (~10 pS in symmetrical, physiological Cl) ohmic-conductance Cl channels (15). Additional strong evidence is that certain mutations introduced into putative transmembrane -helices M5 or M6 (Fig. 1) modify the characteristics of anion interaction with the permeation pathway (reviewed in Ref. 44). For example, charge-reversing mutations at K335 or R347 (both in M6) were found to alter both anion binding within the pore and permeation through it (4, 125, 195), and the charge-neutralizing mutation R347H rendered single-channel conductance switchable between wild-type and mutant values simply by changing cytoplasmic pH back and forth between 5.5 and 8.7 (195). Further evidence comes from experiments that probed the ability of residues in M6 to interact with water-soluble reagents that could modify pore function. The mutation S341A, for instance, altered the apparent affinity (and its voltage dependence) for block of open CFTR channels by diphenylamine-2-carboxylate (133), and a cysteine-scanning method has demonstrated the accessibility to hydrophilic cysteine-modifying reagents of residues in M6 (35). The reactive cysteines appeared to lie on one face of an -helix which therefore may line the pore, and comparisons of reaction rates for anionic and cationic reagents suggested that the anion selectivity filter likely resides toward the cytoplasmic end of a wide water-filled pore (35). current with an approximately linear whole cell current-voltage relationship in symmetrical ~150 mM Cl solutions (10, 78, 132). Unitary current recordings in cell-attached (54, 55) and excised (55, 140) patches subsequently confirmed that these cardiac Cl channels possess all of the hallmark characteristics used to identify epithelial CFTR Cl channels. Thus, in excised inside-out patches, after the required phosphorylation by PKA catalytic subunit, cardiac CFTR Cl channels close promptly upon ATP withdrawal but can be reopened by ATP or GTP, but not by ADP or 5'-adenylylimidodiphosphate (AMP-PNP); their unitary conductance is ohmic and ~12 pS in roughly symmetrical ~150 mM Cl solutions, their open probability (Po , the average fraction of time that a channel spends open) is approximately voltage independent, and their rates of opening and closing are very low (140). Northern blot analyses confirmed the presence of CFTR mRNA in myocytes from regions of the heart and species in which PKA-regulated Cl currents can be recorded, but not in tissue from regions with no CFTR-like currents (112, 140). Sequencing of the amplified transcript from rabbit ventricle indicated that cardiac CFTR is an alternatively spliced isoform lacking the 30 residues at the COOH-terminal end of the first cytoplasmic loop (Fig. 1, thickened segment) that are encoded by exon 5 (86). The only functional difference between cardiac and epithelial CFTR channels so far documented is that upon activation by PKA, whether in intact cells or excised patches, the Po of cardiac CFTR channels often attains a level of 0.7-0.8 (e.g., Refs. 14, 54, 55, 91, 140), whereas epithelial CFTR channels have rarely been reported to display a Po above 0.4-0.5 (e.g., Refs. 211, 223, 224, but cf. Ref. 79). However, it seems most unlikely that the higher Po of cardiac CFTR channels can be attributed to lack of exon 5 because, on the contrary, experimental deletion of exon 5 from CFTR cRNA appears to interfere with trafficking of the channels in mammalian cells, and when the channels are incorporated into bilayers, they display a marked reduction (2- to 3-fold) of Po compared with wild-type CFTR channels (225). Although the electrocardiological role of cardiac CFTR channels remains unclear (but see Refs. 196, 198), and their very existence in human heart is controversial (148, 169, 214), studies of cardiac CFTR channels have nevertheless afforded crucial insights into the gating mechanisms of epithelial CFTR channels, as we discuss here.
II. REGULATION OF CYSTIC FIBROSIS TRANSMEMBRANE CONDUCTANCE REGULATOR BY PROTEIN KINASES
Top
Previous
Next
References
channels
channels could be activated by purified protein kinases (PKA or PKC) applied directly to the cytoplasmic surface, but not when the cells came from CF patients (89, 115, 116, 172). The Cl channels in question, when symmetrically exposed to ~150 mM Cl solutions on either side of the membrane, had single-channel conductances of some 30-50 pS, depending on the precise point along the curved, outwardly rectifying, single-channel current-voltage relationship at which the conductance was measured. Much smaller conductance (~15 pS) Cl channels with linear current-voltage relationships had also been observed in an earlier study of airway cells (64), but those channels were overlooked in the experiments examining kinase action on excised patches. When the CF gene was subsequently identified and sequenced and the large cytoplasmic R domain of its protein product, human epithelial CFTR, was predicted to contain multiple sites for phosphorylation by PKA and PKC (166), this seemed to afford a straightforward and reasonable explanation for the observed lack of kinase-mediated regulation of the 30- to 50-pS, outwardly rectifying, Cl channels in the membranes of cells from CF patients. However, this hope was soon dashed by the observation that, no matter what cell type was chosen as the host for expressing CFTR cRNA, the Cl channels that resulted from CFTR expression were small ohmic conductance channels, not outwardly rectifying channels (3, 20, 43, 50, 94, 104, 116). As already mentioned, the defective regulation of outwardly rectifying Cl channels in CF airway cells is nowadays attributed to (and constitutes the initial evidence for) some kind of modulatory interaction, whose mechanism remains to be established, between CFTR and other channels (53, 66, 175; cf. Refs. 191, 192).
2. Biochemical analysis of phosphorylation by PKA
A) DIBASIC VERSUS MONOBASIC CONSENSUS MOTIFS. It is now abundantly clear that CFTR itself constitutes a small ohmic Cl channel in which five serines (Ser-660, -700, -737, -795, and -813) appear to be readily phosphorylated when PKA is activated in cells that express either native or recombinant CFTR (34, 152). It is also clear, however, that CFTR harbors a total of at least 10 serine residues (Fig. 2) that can be phosphorylated by PKA in vitro, and it seems likely that most of these may, under appropriate conditions, also be phosphorylated by PKA in vivo. Thus a variety of biochemical methods, including direct amino acid sequencing, site-directed mutagenesis combined with two-dimensional peptide mapping, and mass spectrometry (34, 141, 144, 152, 178, 199), have established that Ser-660, -700, -712, -737, -753, -768, -795, and -813 may be phosphorylated, in vitro, by PKA applied to full-length CFTR, before or after immunoprecipitation, and to recombinant R-domain or NBD1-R-domain peptides. In addition, phosphorylation of Ser-422 in NBD1-R-domain peptide (144) and of Ser-670 in R-domain peptide (141) have been demonstrated, although neither has yet been found to be phosphorylated in intact CFTR, either in vitro or in vivo. The reason for this failure is unclear, since functional consequences have been observed upon mutation of Ser-422 in a CFTR mutant already containing nine other serine-alanine mutations (31), or when Ser-670 is mutated in wild-type CFTR (220).
|
As indicated in Table 1, only 2 (Ser-670 and Ser-753) of these 10 phosphorylated serines so far identified in human CFTR lie within a monobasic R-X-S sequence (X represents any amino acid), the most common (twice as frequent as the classical dibasic) consensus sequence for PKA substrates (150). Curiously, in all other species examined, an additional Arg residue confers a dibasic R-R-X-S consensus motif on the Ser-670 site. The other eight phosphorylated serines occur in classical dibasic PKA consensus sites, with an amino acid sequence R-R/K-X-S/T. Other than Ser-422, which lies just NH2 terminal to NBD1 (Fig. 2), these sites so far shown to be phosphorylated by PKA are all contained within the R domain.
|
B) VARIATION IN PHOSPHORYLATION KINETICS. That R-domain peptide is nevertheless phosphorylated only to a stoichiometry of ~5-6 mol/mol after a limited exposure to PKA (52, 152) likely reflects variation in the kinetics of phosphorylation of the different sites. In which case, phosphorylation with higher levels of PKA and for longer periods ought to eventually result in stoichiometric phosphorylation of all the above PKA sites. Consistent with this interpretation, when each site was incorporated into a short synthetic peptide, a >10-fold variability was found in the catalytic efficiency of PKA toward the sites, presumably attributable to intrinsic differences in the ease with which PKA can bind to and phosphorylate individual serines (152). In intact CFTR, the kinetics of phosphorylation of particular serines are likely to be further influenced by local secondary, tertiary, and possibly quaternary structure, and, in cells, also by the proximity, activity, and selectivity of specific phosphatases. For example, the fact that phosphorylation of Ser-768 has not yet been demonstrated biochemically in intact cells, although it is the best in vitro PKA site, suggests that this site is also a very good substrate for one or more protein phosphatases, and this serves to emphasize the important role played by protein dephosphorylation in cellular regulation of CFTR function. A further consideration is that the pattern of phosphorylation might be ordered, because phosphorylation of one site might alter the structure of the protein and thereby influence the rate of phosphorylation (or dephosphorylation) at another site (e.g., Refs. 23, 30, 42, 141). A conformational change of the R domain upon phosphorylation by PKA was demonstrated by analysis of circular dichroism (CD) spectra (51, 52), and several distinct conformations of variably phosphorylated R-domain peptide are suggested by the multiple bands revealed by SDS-PAGE (52, 141, 152). The largest individual mobility shift appears to attend phosphorylation of Ser-737 (23, 109, 141). Moreover, a strong interdomain interaction influencing R-domain phosphorylation is suggested by the observation that, in NBD1-R-domain peptide, occupancy of NBD1 by nucleotide seemed to impair phosphorylation of the R domain (148). Along the same lines, coimmunoprecipitation experiments suggested that phosphorylation of recombinant R domain strengthened its interaction with recombinant NBD1 peptide (73). It remains to be seen whether the same interactions occur within intact CFTR and inside cells.
3. Mutational analysis of phosphorylation sites
A major goal of current research is to understand just how phosphorylation of multiple R-domain serines by PKA effectively regulates CFTR channel function, and this encompasses a number of obvious questions. Does phosphorylation of a given site have a specific function? Or is overall phosphorylation titer the functionally important parameter? If each site does contribute to a particular function, which site is linked with what function? Site-directed mutagenesis of R-domain serines, individually or in groups, provides a means to address these questions. The essential underlying assumption is that if CFTR function is altered by mutation of a serine residue (e.g., to alanine), then that site must normally become phosphorylated. For this reason, two significant caveats must be borne in mind. First, the altered function could be merely a consequence of a change in protein structure resulting from the mutation, rather than a specific response to loss of a phosphorylation site. In this regard, Dulhanty et al. (51) found that the CD spectrum of R-domain peptide bearing Ser-Ala mutations at nine sites differed substantially from that of dephosphorylated (or phosphorylated; Ref. 53) wild-type R domain, suggesting that the mutations (at least 9 of them together) did indeed alter protein structure. Second, unless molecular function can be examined at the microscopic level and/or over a wide range of conditions, there might be little chance of discriminating among effects of the same single mutation introduced into different consensus sites one at a time. It was presumably the latter difficulty that was responsible for the initial failure (34, 163) to detect effects on grossly measured cAMP-stimulated anion permeability of individually mutating to alanine any one of the five readily phosphorylated serines (Ser-660, -700, -737, -795, or -813), or of mutating Ser-686, -712, or -768, or even of simultaneously mutating two or three of them in several combinations. However, simultaneous mutation of four of the principal serines (660737795, and 813) to alanine (to yield the 4SA mutant) clearly reduced both cAMP-stimulated iodide efflux and CFTR channel Po , but only by approximately twofold (31, 163). Additional mutation of up to all eight dibasic R-domain serines (8SA: the 4SA mutant with Ser-686, -700, -712, and -768 also mutated to alanine), or all nine dibasic R-domain sites (9SA: the 8SA mutant with Thr-788 also mutated to alanine), further reduced PKA-activated channel function only slightly, whether assayed by iodide efflux or by channel Po (31, 163). The same assays indicated small, but measurable, functional decrements attending incorporation of additional mutations at Ser-422 to yield a 10SA mutant (31), at Ser-753 to give the 11SA mutant (178), and at four remaining R-domain serines and threonines (S670, T690, T787, and S790) mutated together to yield a (11+4)SA mutant (177). When five additional serines and threonines (S728, S742, S756, T757, and T760) were simultaneously mutated to alanine in the 11SA background to yield a 16SA mutant, there was no observable further decrease (relative to 11SA) in either phosphorylation level or iodide efflux, indicating that those five are not sites for phosphorylation by PKA (178). Corresponding levels of phosphorylation of these mutants were determined either by labeling intact cells with 32P and immunoprecipitating CFTR (31, 163) or by immunoprecipitation followed by phosphorylation in vitro (163, 178). Compared with wild-type CFTR, cAMP-stimulated phosphorylation was substantially reduced in the 4SA mutant, further reduced in the 8SA mutant, greatly diminished in the 10SA mutant, and barely detectable in the 11SA, (11+4)SA, or 16SA mutants (Table 1; Refs. 31, 163, 177, 178).4. Do functionally significant PKA phosphorylation sites exist outside the R domain?
The majority of the phosphorylation sites identified so far occur within the R domain. The exception is Ser-422. Significantly, clear phosphorylation of Ser-422 has been demonstrated only in recombinant NBD1-R-domain peptide, and not in full-length CFTR (144). On the other hand, the clear-cut decrement in channel function seen upon adding the S422A mutation to the 9SA mutant (31) implies that Ser-422 does get phosphorylated in whole CFTR (at least, in 9SA mutant CFTR). It seems valid to question the significance of a small functional consequence of one additional point mutation made in a background in which many of the supposedly major phosphorylation sites have already been eliminated. On the other hand, it is conceivable that phosphorylation of minor, normally kinetically disfavored, sites may contribute to the regulation of wild-type CFTR channels when cellular PKA activity is sufficiently high, or when phosphatases are relatively inactive. But, it could also be that any residual phosphorylation detected in heavily mutated CFTR channels is somehow made possible by those mutations, and would not ever occur in wild-type CFTR regardless of kinase and phosphatase activity. The same kind of criticism can be leveled at the observation that PKA can still enhance activity of mutant 16SA channels that lack nearly all plausible PKA phosphorylation sites in the R domain (178). An alternative viewpoint is that the PKA-dependent activation of those heavily mutated CFTR channels (yielding Po around one-quarter of that for wild-type CFTR) is large enough to suggest that it is imprudent to ignore the role of phosphorylation sites outside the R domain. We have previously pointed out that the lack of influence of PKA on single-channel currents in mutant CFTR missing R-domain residues 708-835 (CFTR
R), with (163) or without (122, but cf. Ref. 164) additional mutation of Ser-660 to alanine (CFTRR-S660A), does not constitute proof that all the phosphorylation sites reside in the R domain (68). It is equally possible that phosphorylation sites outside the R domain require an intact R domain for transduction of their signal. Having said that, we note that Seibert et al. (178), from comparison of autoradiographs of cyanogen bromide (CnBr) digests of R-domain peptide and of intact CFTR, favor the interpretation that all detectable phosphorylation in intact CFTR occurs within the R domain, although this once again raises the question of phosphorylation of Ser-422. Among other possibilities for as yet undetected phosphorylation sites outside the R domain is a cluster of Arg, Ser, and Lys residues near the COOH terminus of CFTR (residues 1453-1457 of human epithelial CFTR). The sequence R-X-S-S-K/R at that location is conserved across species, and recent preliminary studies have shown that a synthetic peptide including these amino acids can be phosphorylated by PKA (K. Chan and A. C. Nairn, unpublished results). The safest conclusion is that all the functionally significant PKA phosphorylation sites in CFTR have not yet been identified.
5. Roles for individual PKA phosphorylation sites
The inability to discern specific functional roles for individual PKA phosphorylation sites, whether assumed to be of major (e.g., Ser-660, -700, -737, -795, and -813) or minor (e.g., Ser-422 and -753) importance, coupled with the apparent progressive decline of channel activity as the number of Ser-Ala mutations was increased, led to the idea of functional redundancy, or degeneracy, among the phosphorylation sites (31, 163). According to this idea, only the number of phosphorylated serines, and not their specific location, governs activation of CFTR (31, 34, 163). A suggested mechanism possibly underlying such an effect is a finely graded electrostatic movement of the R domain, without any specific change in its conformation, in rough proportion to the accumulated negative charge on the phosphoserines, that movement somehow leading to a progressive facilitation of ion flux through the channel pore (163, 217). One kind of observation used to support this suggestion was the finding that insertion of six to eight negative charges in place of R-domain serines (6SD-8SD: Ref. 163; 8SE: Ref. 31) yielded channels that did not require phosphorylation to be opened by ATP (although they had substantially diminished Po). On the other hand, the finding that incorporation of only four or five negative charges left channels apparently fully dependent on phosphorylation (but also with low maximal Po) implied at least some kind of threshold phenomenon (31, 163). Several additional arguments make the simplest, accumulated charge mechanism unlikely. As already mentioned, phosphorylation of the R domain, judged on the basis of mobility shifts of recombinant R domain on SDS-PAGE (23, 52, 141, 152), is associated with reproducible, discrete conformational changes and thus seems incremental rather than continuous. Also, strong phosphorylation by PKA alters the CD spectrum of R-domain peptide, implying changes in its secondary and tertiary structure (52). Importantly, the CD spectrum of mutant 8SE CFTR R domain differs markedly from that of phosphorylated wild-type CFTR R domain, as well as from that of dephosphorylated R domain or mutant 8SA R domain (51, 52). Perhaps most persuasive, recent results demonstrate that Ser-737 and Ser-768 both exert an inhibitory (rather than stimulatory) influence on CFTR channel activity under conditions of submaximal phosphorylation (42, 218, 220). That work examined dose-response curves for activation of CFTR channels expressed in Xenopus oocytes by IBMX [at submillimolar concentrations, most likely predominantly inhibiting cAMP phosphodiesterase (50), but at millimolar levels probably also directly stimulating CFTR channels (2, 81)] in the presence of forskolin (to activate PKA). The results showed that individual substitution of Ser-660, -670, -700, -795, or -813 with alanine increased the IBMX concentration required (K0.5) for half-maximal activation of CFTR, as expected for "stimulatory" phosphorylation sites. In general accordance with phosphorylation levels observed in intact cells, mutation of Ser-813 had the largest effect, while mutation of Ser-660 or Ser-795 had lesser effects, mutation of Ser-670 or Ser-700 had small effects, and mutation of Ser-712 or Ser-686 (a PKC site, see sect. IIB) had no measurable consequence. Significantly, mutation of Ser-737 (smaller effect) or Ser-768 (larger effect) decreased the K0.5 for IBMX, demonstrating that these are "inhibitory" sites. Recent measurements of Po of S768A CFTR channels in excised patches exposed directly to PKA catalytic subunit suggest that phosphorylation of Ser-768 somehow impedes phosphorylation of one or more serines in stimulatory sites (42). Similarly, biochemical assays have shown that phosphorylation of the other inhibitory site, Ser-737, in an R domain peptide slows phosphorylation of Ser-660, -700, -712, and -795 (109), at least three of which occur in demonstrably stimulatory sites. These findings emphasize the quantitatively and qualitatively distinct contributions made by individual phosphorylation sites, and they argue that activation of CFTR channels by phosphorylation with PKA cannot readily be explained in terms of a simple build-up of negative charges (68). On the contrary, the CFTR activating influence of phosphorylation of the prominent stimulatory site Ser-813 seems to be effectively canceled by phosphorylation of the major inhibitory site Ser-768, since the double mutant S768A/S813A had a K0.5 for IBMX comparable to that of wild-type CFTR channels (220). This recent mutagenesis work, examining one or two residues at a time, has begun to address the question of which PKA phosphorylation sites are involved in what aspect of CFTR function, although investigation of the underlying molecular mechanisms has yet to come. The finding that several phosphoforms of CFTR can be distinguished on the basis of channel function (49, 88) strongly suggests that individual phosphorylation sites do indeed play distinct, reproducible roles in regulating channel gating, presumably by somehow controlling ATP binding and hydrolysis by the NBDs (see, e.g., Refs. 128, 224).B. PKC
1. PKC phosphorylation sites
Although phosphorylation of CFTR by PKA is still believed to be the principal pathway for acute regulation of channel gating, it has become increasingly clear that CFTR can also be phosphorylated, and consequently regulated, by other protein kinases. Thus PKC, with or without Ca2+, phosphorylates the R domain of CFTR in vitro to a stoichiometry of ~2 mol/mol (21, 52, 152), predominantly on Ser-686 and -790 (Fig. 2), although some sites phosphorylated by PKA, like Ser-660 and -700, are also slowly phosphorylated by PKC (152). Interestingly, no change in CD spectrum (52) or in mobility on SDS-PAGE (52, 152) occurred when R-domain peptide was phosphorylated by PKC alone at moderately low levels of PKC activity. Activation of PKC by phorbol ester in intact cells resulted in phosphorylation of Ser-686 and of several PKA sites, but not of Ser-790 (152). However, it is not clear whether, in those cells, PKC directly phosphorylated PKA sites on CFTR, or simply phosphorylated specific PKC sites (e.g., Ser-686) which then somehow facilitated phosphorylation of other sites by low basal activity of PKA. In vitro, prephosphorylation of CFTR by PKC did seem to enhance subsequent phosphorylation by PKA (31, but see discussion in Ref. 99).2. PKC influence on CFTR channel function
Protein kinase C applied directly, and in the absence of exogenous PKA, to recombinant CFTR in excised patches has been consistently reported to activate some of the channels, although weakly in many cases (21, 31, 59, 130, 194). In intact cells, stimulation of PKC with phorbol ester can activate CFTR-mediated Cl efflux (46) as well as CFTR channels in cell-attached patches (11, 40). However, during whole cell current recording, phorbol ester alone failed to elicit CFTR current in cardiac myocytes (134) or in pancreatic duct cells (222). Nevertheless, PKC stimulation or application has invariably been found to potentiate subsequent CFTR channel activation by PKA, markedly in cell-attached or excised patches (11, 194) and moderately in intact cardiac myocytes (134) and pancreatic duct cells (222). Most surprisingly, this marked potentiation was still seen in 10SA mutant channels (31) that lack the Ser-686, -660, and -700 (but not Ser-790) sites identified as being phosphorylated by PKC in vitro. A possible implication is that PKC phosphorylation of Ser-790 enhances the functional consequence, in terms of channel activation, of PKA phosphorylation at, say, the monobasic sites Ser-753 or -670 (both spared in the 10SA mutant) or at sites not yet identified.
3. Constitutive PKC phosphorylation?
The recent conclusion (99) that constitutive phosphorylation by cellular PKC does not by itself activate CFTR channels, but is a prerequisite for their subsequent activation by PKA, is beginning to shed new light on these apparently synergistic interactions between PKC and PKA sites. After pretreatment of cells with PKC inhibitors, recombinant CFTR channels in transfected Chinese hamster ovary (CHO) or baby hamster kidney (BHK) cells (99, 117), and native CFTR channels in Calu-3 cells (117) or in cardiac myocytes (134), were almost completely, or largely, refractory to PKA stimulation by elevation of cellular cAMP. These permissive PKC sites appeared to be dephosphorylated moderately slowly by membrane-bound phosphatases upon patch excision (at least, in CHO cells), because the ability of PKA catalytic subunit to reactivate rundown channels waned steadily over an ~10-min period, although that ability could be restored by exposing the patch to PKC (99). Further analysis suggested that phosphorylation by PKC did not alter the number of active channels in a patch, but enhanced the ability of PKA to increase channel Po , primarily by reducing mean closed time (99).4. Mechanism of PKC action
This new work strengthens the suggested modulatory role(s) for PKC in CFTR channel activation, although it is still not clear whether either form of modulation (i.e., stimulatory or permissive) reflects direct phosphorylation of CFTR by PKC, or whether such phosphorylation is essential for subsequent channel activation by PKA. It might be thought that evidence against an absolute need for PKC phosphorylation was provided by the finding that addition of only PKA and MgATP (i.e., without PKC) activated recombinant CFTR channels after their extensive purification from insect cells, reconstitution in pure phospholipids, and incorporation into a lipid bilayer (15). In fact, the same CFTR preparation was recently shown to remain partially phosphorylated [at sites sensitive to protein phosphatase (PP) 2A; see sect. IVA] throughout the purification procedure (113); although whether that basal phosphorylation occurred at PKA or PKC sites was not determined. In retrospect, a useful test might be to fully dephosphorylate the purified CFTR preparation with PP2A (perhaps together with PP2C) and then test whether PKA alone could rephosphorylate and reactivate it. Along these lines, CFTR channels in patches excised from NIH-3T3 or Calu-3 cells could be activated by either PKC alone or PKA alone, then deactivated by exposure to
-phosphatase (-PP), and finally reactivated by application of the "other" kinase, i.e., PKA or PKC, respectively. Regardless of the sequence of phosphorylation, the kinetics of CFTR channel opening and closing were faster after phosphorylation by PKC than after phosphorylation by PKA (59). If -PP could be shown to fully dephosphorylate CFTR channels under the conditions of those recordings, then the simplest interpretation of the findings would be that CFTR channels can be activated by phosphorylation either by PKC alone or by PKA alone. However, as already mentioned, PKC has previously been shown to phosphorylate some R-domain sites that are also phosphorylated by PKA (152). Also, it remains conceivable that basal activity of cellular PKC and PKA, before patch excision, was sufficient to phosphorylate key sites on CFTR which, although not themselves capable of activating CFTR channels, are not only obligatory for further phosphorylation, and hence activation, of the channels by PKA or PKC, but are also resistant to -PP.
5. PKC targets other than CFTR?
Of course, we still cannot be certain whether any (or all) of these effects of PKC on CFTR channels in patches or intact cells reflect PKC phosphorylation of CFTR itself. Thus phosphorylation of Ser-686 by PKC (152) might not occur in all cell types and, even when it does occur, might not have any discernible functional consequence. Indeed, although not examined at the single-channel level, the sensitivity of mutant S686A CFTR channels in oocytes to activation by IBMX in the presence of forskolin was no different from that of wild-type CFTR (220). So it remains possible that the described effects of PKC on CFTR channel activity might be indirect, reflecting phosphorylation of other targets, such as the cytoskeleton, which could lead, via membrane retrieval (222) or insertion (11), to changes in the density of CFTR channels incorporated in the cell surface. It has recently been demonstrated that membrane-targeted syntaxin 1A and Munc-18 (two proteins involved in vesicle fusion) reciprocally regulate CFTR channel function (syntaxin 1A inhibiting and Munc-18 relieving that inhibition; Ref. 142), apparently by interacting with the NH2-terminal cytoplasmic tail of CFTR (143). New findings also show that CFTR's COOH terminus binds to PDZ domains in an anchoring protein, EBP50 [ERM(ezrin-radixin-moesin)-binding phosphoprotein 50], which in turn physically interacts with ezrin, an actin-binding protein that also anchors PKA (76, 106, 158, 185, 213). These recent demonstrations suggest other possible targets for phosphorylation by PKC and, hence, pathways for PKC-mediated modulation of CFTR channel activity.6. PKC isoforms
Once the specific targets of phosphorylation by PKC are identified, we will need to learn which PKC isoforms are involved. Only 4 of the 11 isoforms of PKC so far identified (146) depend on Ca2+ for activity, and available evidence suggests that CFTR can be phosphorylated by PKC in both a Ca2+-dependent as well as a Ca2+-independent manner (21, 152). The first clear evidence of involvement of a specific isoform of PKC has come from the recent finding that 48-h pretreatment of Calu-3 cells with antisense oligonucleotide to PKC-
(but not to PKC-
or PKC-
) prevented the normal PKA-mediated increase in 36Cl efflux via CFTR, without affecting activity of PKA itself (117). The clear implication is that the constitutive permissive phosphorylation that facilitates activation of CFTR by PKA is carried out by Ca2+-independent PKC-. It will now be important to learn whether PKC- can directly phosphorylate CFTR in vitro and, if so, at what site(s). Also, detailed analyses of single CFTR channel responses to individual PKC isoforms will be required to learn the full range of possible mechanisms by which PKC isoforms, whether Ca2+ dependent or Ca2+ independent, and whether basally active or acutely stimulated, might regulate CFTR function in cells. Despite our present meager understanding of the underlying mechanisms, CFTR channel regulation by PKC is evidently important enough to warrant further investigation in this kind of detail.
7. PKC influence on CFTR expression
In addition to possible modulation via direct phosphorylation of CFTR, PKC appears to regulate CFTR expression and degradation. Treatment of epithelial cells with phorbol ester for several hours leads to a significant reduction in transcription of CFTR mRNA (13, 45, 200), and Ca2+ ionophores such as A-23187 or ionomycin have a similar effect (13). These effects are thought to be consequences of long-term stimulation of PKC and are mediated by phorbol ester-sensitive elements in the CFTR promoter. Prolonged exposure of epithelial cells to phorbol ester has also been found to accelerate degradation of CFTR protein (24). However, because long-term treatment of cells with phorbol ester also causes rapid downregulation of sensitive PKC isoforms, the mechanisms of these effects of phorbol esters on CFTR remain unclear.C. cGMP-Dependent Protein Kinases
1. cGMP-dependent protein kinase isoform specificity
In vitro, both of the major cGMP-dependent protein kinase (PKG) isoforms, PKGI and PKGII, efficiently phosphorylate the R domain of CFTR (21, 63, 152) to high stoichiometry (>5 mol/mol) at sites that largely overlap those phosphorylated by PKA (63, 152). Surprisingly, however, although PKGII was found to activate CFTR channels in patches from NIH-3T3 or rat intestinal IEC-CF7 cells about as robustly as PKA did (63), the same channels could not be activated by PKGI (21, 63). The solution to this seeming paradox is that the apparent specificity of PKGII in regulating CFTR is attributable to myristoylation-dependent membrane targeting of PKGII, which gives it a crucial kinetic advantage due to its colocalization with CFTR in the plasma membrane (208-210). Thus expressed membrane-associated PKGII, but not expressed soluble PKGI, phosphorylated and activated CFTR channels when IEC-CF7 cells were stimulated with atrial natriuretic peptide, which activates guanylyl cyclase (210). Correspondingly, a chimeric PKGI incorporating the NH2-terminal membrane-anchoring domain of PKGII was able to strongly activate CFTR in stimulated IEC-CF7 cells, but activation by PKGII bearing mutations at the NH2-terminal myristoylation site was severely impaired (209).2. Tissue-specific responses to cGMP
In intestinal epithelium, stimulation of guanylyl cyclase and resulting generation of cGMP has been suggested as the mechanism by which the hormone guanylin, and heat-stable enterotoxins secreted by pathogenic strains of E. coli, activate CFTR channels (32). In support of the proposal that PKGII selectively mediates this activation of CFTR, "knockout" mice deficient in PKGII are resistant to the influence of E. coli enterotoxin on intestinal secretion (151). Related to this, Quinton (156) has suggested that the strong activation of CFTR in intestinal epithelia, and consequent stimulation of fluid secretion, by heat-stable enterotoxins acting via cGMP (also by heat-labile toxins, like cholera toxin, that act via cAMP) provides a mechanism by which individuals heterozygous for a lethal CF mutation might have gained a genetic advantage over people harboring no mutation. Cystic fibrosis heterozygotes are likely to have been afforded some protection against life-threatening enterotoxin-induced diarrhea (32, 65), and this might explain the relatively high frequency of theF508 mutation (1 in 25 Caucasians is a carrier; Ref. 204). Unlike intestinal epithelium, airway epithelium lacks the PKGII isoform (63), and cGMP does not appear to activate CFTR channels in permeabilized human airway epithelial monolayers (21).
3. More than one mechanism of cGMP action?
A rise in cellular cGMP concentration in cells expressing PKGII will likely activate CFTR channels via PKG-mediated phosphorylation (126, 209, 210). But even in cells devoid of PKGII, CFTR channels may be activated if cGMP concentrations reach sufficiently high levels to promiscuously "cross-activate" PKA (32, 61, 197). And even a moderate increase of cGMP concentration might lead to activation (or enhanced activity) of CFTR channels, also via PKA-mediated phosphorylation, in cells in which type III cGMP-inhibited phosphodiesterase (which destroys cAMP) is highly expressed (105, 147, 210). Finally, a direct activation of CFTR channels expressed in Xenopus oocytes by intracellular cGMP, via a PKG-independent pathway, has recently been suggested (193). The third cytoplasmic loop of CFTR includes a domain resembling the cyclic nucleotide-binding site(s) in proteins related to the catabolite-gene activator protein, and mutations within that domain impaired the ability of cGMP to activate CFTR channels, although their activation by cAMP seemed unaltered (193).D. Ca2+/Calmodulin-Dependent Protein Kinases
Although purified CFTR R domain can be phosphorylated in vitro by calmodulin (CaM) kinase I (152), though not by CaM kinase II (21, 152), or CaM kinase III, or casein kinase II (152), there has been no investigation yet of possible functional consequences, either in vitro or in vivo. The highest concentrations of CaM kinase I are found in neurons of the brain (153), cells that were originally believed not to express CFTR. However, the presence of CFTR mRNA and protein has recently been demonstrated using reverse transcriptase PCR, in situ hybridization, and immunocytochemistry in human hypothalamus (138) and in several regions of rat brain, including cerebral cortex (101), hypothalamus, thalamus, and amygdaloid nuclei (137). Moreover, CaM kinase I (154) and CFTR (85) are both expressed in the choroid plexus, allowing the possibility of CFTR regulation by CaM kinase I-mediated phosphorylation at least in that tissue. Outside the brain, however, a recent study found little evidence for coexpression of CaM kinase I and CFTR in nonneuronal tissues, including the intestine (131).
E. Protein Tyrosine Kinases
1. Tyrosine phosphorylation of CFTR
Much recent work has examined the striking enhancement of CFTR channel activity caused by the tyrosine kinase inhibitor genistein (94, 95, 162, 176, 226), although only within the past year has phosphorylation of tyrosyl residues in CFTR been detected (100). Coexpression of CFTR with v-Src, which encodes oncogenic, constitutively active, tyrosine kinase, caused in vivo tyrosine phosphorylation of CFTR as detected by antiphosphotyrosine antibody, and immunoprecipitated CFTR was phosphorylated, in vitro, by the tyrosine kinase p60c-Src (100). However, concomitant functional analyses suggested that this direct tyrosine phosphorylation of CFTR increases channel activity, in contrast to the implication of the findings with the inhibitor genistein which, at first glance, would be more consistent with a tyrosine kinase-mediated reduction of CFTR channel current. Thus application of Src to excised patches increased CFTR current, while restoring fast flickery gating, in PKA-activated CFTR channels (60), and Src was even able to strongly activate CFTR channels in the presence of PKA inhibitor (100). The latter finding is strikingly corroborated by undiminished activation by Src of 15SA CFTR channels (with mutations in 15 dibasic and monobasic PKA consensus sites) despite a roughly sixfold reduction in activation of the same channels by PKA (100).2. Influence of the tyrosine kinase inhibitor genistein
A) GENISTEIN EFFECTS. If tyrosine phosphorylation of CFTR enhances channel activity, and if tyrosine kinases were constitutively active in cells expressing CFTR, then exposing such cells to a tyrosine kinase inhibitor, like genistein, would be expected to allow endogenous tyrosine phosphatases to dephosphorylate CFTR and hence diminish channel activity. On the contrary, kinase-inhibiting concentrations of genistein have been demonstrated to increase, not decrease, CFTR channel activity in all cells tested, including NIH-3T3, IEC-CF7, HT-29/B6, HEK-293, Calu-3, and T84 cell lines, Hi-5 insect cells, and Xenopus oocytes (92, 94, 95, 162, 176, 215, 226). In these various cell types, genistein enhanced iodide efflux, macroscopic CFTR Cl current, and time-averaged single CFTR channel current (94, 95, 162, 176, 226), as well as phosphorylation of CFTR (92, 162). It is now clear that these effects of genistein are not mediated by inhibition of tyrosine kinase and that they are absolutely dependent on prior phosphorylation of CFTR by PKA. Evidence against tyrosine kinase inhibition as the mechanism of genistein action includes the findings that its effects 1) were not mimicked by the tyrosine kinase inhibitors erbstatin, tyrphostin A23, tyrphostin A51, tyrphostin B42, or herbimycin A, in T84 and HT-29/B6 cells (94), nor by AG126, tyrphostin 25, or herbimycin, in NIH-3T3 cells (47), nor by tyrphostin 47 in Xenopus oocytes (215); 2) were not abolished or prevented by the protein tyrosine phosphatase inhibitors pervanadate or orthovanadate in patches excised from T84 or HT-29 colonocytes (47), Xenopus oocytes (215), or NIH-3T3 cells (213a); and 3) the effects were similar in direction to those resulting from direct exposure of CFTR to the active tyrosine kinase Src (100). Evidence that genistein acts on CFTR channels only after their phosphorylation by PKA includes the observations that 1) genistein had no effect in permeabilized HT-29/B6 monolayers in the absence of cAMP (94), 2) genistein enhanced CFTR channel activity in resting Hi-5 insect and NIH-3T3 cells in which the channels were already active under basal conditions (i.e., even before stimulation of PKA) but had no effect on resting cells lacking basal CFTR channel activity (62, 226), and 3) genistein's effects on CFTR phosphorylation and iodide efflux in NIH-3T3 cells were inhibited by the PKA inhibitor H-89 (162).
B) GENISTEIN SITE OF ACTION. The finding that genistein increased PKA-mediated incorporation of 32P into CFTR in NIH-3T3 cells (92, 162) prompted the suggestion that genistein might inhibit a serine/threonine phosphatase that dephosphorylates PKA sites in CFTR (92, 94, 162, 226). Because genistein's effect was additive with that of calyculin A (an inhibitor of PP1 and PP2A), a protein phosphatase distinct from PP1 or PP2A was considered a likely target of genistein, the obvious candidate being PP2C (226; cf. Ref. 162). However, the most recent studies on CFTR channels in excised patches argue strongly that genistein, in fact, interacts directly with CFTR itself. The clearest evidence is that genistein (applied in the absence of PKA) can still rapidly, and reversibly, enhance the activity of phosphorylated CFTR channels even when further PKA phosphorylation is prevented by a maximally effective concentration of peptide inhibitor (213a, 215), or after the channels have been made resistant to dephosphorylation by thiophosphorylating them, by briefly replacing ATP with adenosine 5'-O-(3-thiotriphosphate) (ATP
S) in the presence of PKA (62). Single-channel analysis indicates that low concentrations of genistein increase CFTR current by stabilizing the channel open state, suggesting a possible interaction with NBD2 (62, 213a, 226). At higher (>50 µM) concentrations of genistein, however, activation gives way to inhibition, indicating that genistein also binds to a lower affinity site at which it slows channel opening (213a; cf. Ref. 94). It is possible that, at high concentrations, genistein (or at least its anionic form) can also bind weakly to a site within the pore of CFTR where it may act as an open-channel blocker (111).
C) GENISTEIN MECHANISM. If genistein acts directly on CFTR channels to prolong channel opening even when no further phosphorylation is possible, how can we explain the observed genistein-induced increase in CFTR phosphorylation? The simplest explanation is that genistein acts at NBD2 to stabilize the open conformation of CFTR and that the open channel is a poorer substrate for dephosphorylation by PP2C than the closed channel (67, 92, 213a, 226). Such an action would enhance, in a PKA-dependent manner, time-averaged single-channel CFTR currents, macroscopic CFTR currents and fluxes, and the steady-state phosphorylation level of a subset of serines. An influence of channel gating on phosphorylation and/or dephosphorylation need not be surprising. Because complex interactions linking R-domain phosphorylation with NBD function are believed to underlie the regulation of CFTR channel gating, simple thermodynamic constraints require that NBD function affects phosphorylation and dephosphorylation of the R domain (84, 97, 188). Indeed, biochemical evidence for such an influence of NBD conformation on kinetics of R-domain phosphorylation by PKA has recently been obtained (145).
| |
III. REGULATION OF CYSTIC FIBROSIS TRANSMEMBRANE CONDUCTANCE REGULATOR BY PROTEIN PHOSPHATASES |
|---|
|
TopPreviousNextReferences
|
|---|
A. Candidate Phosphatases for Regulating CFTR
As already noted, the steady-state level of phosphorylation of a protein in vivo depends on the relative rates of phosphorylation and dephosphorylation of all of its phosphorylatable sites. In all cell types examined to date, CFTR Cl current activated by stimulation of PKA declines rapidly upon withdrawal of the PKA agonist, indicating that highly active endogenous protein phosphatases promptly dephosphorylate and inactivate CFTR (67, 68, 165, 216). The four main types of serine/threonine protein phosphatases in eukaryotic cells are encoded by members of two gene families. The protein phosphatases PP1, PP2A, and PP2B belong to the large PPP family which ensures substrate specificity in vivo by employing a variety of regulatory subunits (also known as B subunits) to target and modulate phosphatase activity, and PP2C belongs to the PPM family of Mg2+-dependent phosphatases that lack regulatory subunits (37, 38, 184). Both PP1 and PP2A are ubiquitous components of intracellular signaling pathways, and they are both potently inhibited by naturally occurring toxins like okadaic acid, microcystin, and calyculin A. Protein phosphatase 2B, also known as calcineurin, requires Ca2+ and calmodulin for its activity and is inhibited by the immunosuppressants cyclosporin or FK-506 bound to their respective partners, cyclophilin or FKBP-12. Protein phosphatase 2C, for which no good inhibitor is known, requires millimolar Mg2+ (or Mn2+) for its activity (123). In in vitro biochemical tests using PKA-phosphorylated CFTR, incubation with purified PP2A has been shown to cause substantial (18) or nearly complete (21) dephosphorylation, whereas PP1 and PP2B were both far less effective (21). More recent studies have shown that, like PP2A, PP2C can almost completely dephosphorylate PKA-phosphorylated intact CFTR as well as R-domain peptide (201). PKA-phosphorylated R-domain peptide was similarly found to be strongly, but incompletely, dephosphorylated by either purified PP2A alone or recombinant PP2C
alone, although, together, PP2A plus PP2C caused complete dephosphorylation, whereas neither PP1 nor PP2B was measurably effective (141). However, little is known yet about the specificity with which these different phosphatases attack individual phosphoserines on phosphorylated CFTR.
B. Functional Effects of Exogenous Phosphatases
Purified PP2A greatly reduced the Po (from 0.31 to 0.02) of PKA-phosphorylated wild-type CFTR channels reconstituted in lipid bilayers but had no effect on the constitutive activity of CFTRR (122). In patches excised from NIH-3T3 cells, direct application of purified PP2A also substantially deactivated PKA-phosphorylated epithelial CFTR channels, but PP1 and PP2B were both reported to have essentially no efffect (21). Surprisingly, however, in a more recent study, PP2B was shown to reproducibly deactivate PKA-phosphorylated CFTR channels, not only in patches excised from NIH-3T3 cells but also in patches from Calu-3 cells, and PP2B inhibitors strongly potentiated PKA-dependent activation of CFTR channels in the same NIH-3T3 cells (59). The reasons for these apparently discrepant findings with PP2B in NIH-3T3 cells remain unclear, although differences in enzyme source, and hence activity, might have contributed (59). Experiments testing all four phosphatase types on patches excised from BHK cells confirmed that PP1 is without effect, PP2B weakly inactivates, and PP2A and PP2C both strongly but incompletely deactivate PKA-phosphorylated CFTR channels (120). Because the more rapid deactivation induced by exogenous PP2C most closely resembled the pattern of channel rundown observed (1 in 4 patches showed rundown) following patch excision into PKA-free (and phosphatase-free) solution, the authors concluded that endogenous membrane-attached PP2C normally causes that rundown. Kinetic analysis confirmed that the characteristic changes in CFTR channel gating caused by purified exogenous PP2C mimicked those attending the spontaneous rundown (120). Protein kinase A-phosphorylated CFTR channels in excised patches from HeLa cells were also shown to be rapidly deactivated, although incompletely (<20% of patch current remaining), by recombinant PP2C (201). In the same study, coexpression of PP2C with CFTR channels in Fisher rat thyroid cells, grown as an epithelial monolayer, reduced the magnitude of short-circuit current activated by PKA agonists and accelerated its deactivation on agonist withdrawal, in comparison with monolayers expressing CFTR alone (201).
The four predominant cellular phosphatases, PP1, PP2A, PP2B, and PP2C, are not the only phosphatases capable of dephosphorylating, and deactivating, CFTR. For instance, it has been shown that exogenous alkaline phosphatase can dephosphorylate PKA-phosphorylated CFTR protein (18) and can deactivate CFTR channel currents in patches excised from CHO cells (18, 194), pancreatic cells (17), NIH-3T3 cells (21), and BHK cells (120). In addition, a variety of suggested inhibitors of alkaline phosphatase, including IBMX, theophylline, levamisole, and p-bromotetramisole, were reported to slow deactivation of CFTR channels upon patch excision (17, 18). However, there are several reasons for believing that a physiological contribution of alkaline phosphatase to CFTR regulation is most unlikely. First, and foremost, although alkaline phosphatase is localized to the apical membrane of polarized pancreatic cells, it is oriented with its catalytic domain in the extracellular milieu (17), i.e., on the opposite side of the membrane from its proposed target, the R domain. Second, such high concentrations of the various inhibitors were required (e.g., bromotetramisole was used at concentrations several orders of magnitude higher than needed to inhibit alkaline phosphatase in standard assays; Ref. 120) that their specificity must be questioned. Third, deactivation of CFTR channels by alkaline phosphatase itself occurred only at concentrations eight orders of magnitude greater than those at which PP2A and PP2C deactivated the same channels in the same patches (120), a result underscored by the known ability of alkaline phosphatase to nonspecifically dephosphorylate many protein and nonprotein substrates. Similarly, -PP, the viral nonspecific serine/threonine/tyrosine phosphatase, has been shown to deactivate CFTR channels previously activated in NIH-3T3 cell patches by p60c-Src (60), PKA, or PKC (59).
C. Findings With Phosphatase Inhibitors
Evidently, test applications of exogenous protein phosphatases (or catalytic subunits of phosphatases) can tell us what these enzymes are capable of, but not necessarily what happens either in cells that naturally express native CFTR or in transfected cells expressing recombinant CFTR. Information on which cellular phosphatases actually regulate CFTR comes from use of selective inhibitors. Okadaic acid or microcystin enhanced forskolin-activated CFTR Cl current in cardiac myocytes and slowed, and made incomplete, its deactivation on washout of forskolin (Fig. 3; Ref. 88). Okadaic acid also prevented deactivation of CFTR current in isolated sweat duct, although no enhancement of cAMP-activated Cl current was noted (161). In NIH-3T3 cells expressing CFTR, calyculin A enhanced CFTR-dependent iodide efflux in the absence of any treatment to stimulate PKA activity, and this effect was paralleled by increased phosphorylation of CFTR measured biochemically (162). In insect cells transfected with CFTR, calyculin A had no effect by itself but increased dramatically (>17-fold) forskolin-stimulated CFTR Cl current, which did not fully deactivate when forskolin was withdrawn in the maintained presence of calyculin A (226). A phosphorylated peptide inhibitor of PP1 was found to be ineffective in a preliminary test using cardiac myocytes, suggesting that, at least in those cells, PP2A was the target of microcystin and okadaic acid (90), although it should be noted that those agents also inhibit the novel PP2A-like protein phosphatases PP4 and PP5 (e.g., Ref. 33). The conclusion that PP1 plays no role, but PP2A plays an important role, in deactivating native CFTR channels in two of their natural environments, sweat duct epithelial cells and cardiac myocytes, corresponds well with the results of tests with exogenous phosphatases described above. However, okadaic acid did not alter the amplitude or deactivation rate of PKA-mediated short-circuit current in monolayers of human airway cells or T84 cells (201), and calyculin A similarly failed to affect PKA-mediated short-circuit current in T84 cell monolayers (120), implying a less dominant role for PP2A in airway and intestinal epithelial cells.
|
An obligatory role for PP2B in deactivation of CFTR channels in cardiac myocytes can be ruled out, because channels deactivated completely on washout of forskolin (Fig. 3, A and B, c-a) despite the fact that the intracellular solutions employed for the whole cell current recordings lacked Ca2+ and included 10 mM EGTA (88, 91). The same conclusion may be drawn from the observations that the PP2B inhibitor FK-506 neither augmented the forskolin-induced activity of epithelial CFTR channels expressed in insect cells (226) nor affected the size or deactivation rate of CFTR current in human airway, or T84, epithelia (201). In contrast, the PP2B inhibitors cyclosporin A and deltamethrin were found to strongly enhance PKA-mediated activation of CFTR current in NIH-3T3 cells, but not in Calu-3 cells or in HT-29 epithelial monolayers, despite the fact that exogenous PP2B could deactivate CFTR channels in patches excised from both NIH-3T3 cells and Calu-3 cells (59). The latter findings emphasize the danger of drawing inferences about regulatory events in intact cells by extrapolating results obtained with exogenous components, since the exogenous component, even if shown to be expressed in the cell under study, might not colocalize with the target of interest (cf. Ref. 59). The same study also demonstrated that exogenous PP2B could deactivate not only PKA-activated but also PKC-activated CFTR channels in patches from NIH-3T3 cells (59). However, the likely synergism between PKA and PKC phosphorylation of CFTR channels raises the questions of whether the PP2B deactivation of PKC-activated channels in fact reflects dephosphorylation of essential, basally phosphorylated, PKA sites and/or whether the increase in PKA-activated current by PP2B inhibition in fact reflects interference with ongoing dephosphorylation of permissive PKC sites?
D. Differential Dephosphorylation of Multiple Sites
As mentioned in the previous section, selective inhibition of PP1/PP2A in cardiac myocytes with maximally effective concentrations of okadaic acid or microcystin not only augmented CFTR current activated by forskolin and slowed deactivation following forskolin removal, as expected for inhibition of a functionally important phosphatase, but it also rendered deactivation incomplete (Fig. 3, A and C, f-d; Ref. 88). Analogous results have recently been obtained with epithelial CFTR channels expressed in insect cells (226). Because up to 50% of the CFTR current persisted indefinitely in the continued presence of those phosphatase inhibitors (although not with an inhibitor of PP1), and the persistent current was insensitive to PKA inhibition with PKI (indicating that it did not depend on continuing phosphorylation by PKA), we were able to conclude that full deactivation of native cardiac CFTR requires that certain phosphoserine residues be dephosphorylated by PP2A (88). However, the fact that partial deactivation still occurs when PP2A is fully inhibited means that some phosphatase other than PP2A can dephosphorylate additional sites on CFTR. The obvious insensitivity of that phosphatase either to the presence of PP1/PP2A inhibitors or to the absence of Ca2+ (and, hence, of PP2B activity) implies that this partial deactivation can be attributed to PP2C (68, 88, 90, 91, 226). The subsequent finding (91) that a single CFTR channel could initially open to a low Po state characterized by brief open times, but then, during continued exposure to PKA, could switch to a higher Po state with longer openings (Fig. 4), suggested an explanation for the persistent CFTR current seen following deactivation of PKA when PP2A was inhibited. The persistent current (Fig. 3, A and C, f-d) was suggested to reflect the activity of partially phosphorylated CFTR channels, phosphorylated exclusively at sites that require PP2A for their dephosphorylation and that support only limited channel activity, restricted to brief openings and, hence, characterized by a low Po . The larger current recorded in the presence of forskolin (Fig. 3, A and C, e-d) was suggested to reflect activity of the same population of CFTR channels, but additionally phosphorylated at sites which were susceptible to dephosphorylation by PP2C, and which supported longer channel openings and, hence, a higher Po (91). It is not yet clear whether, under p
) sites. Failure to detect incorporation of 32P into Ser-768 in cells is an important issue not yet resolved, but could reflect stable basal phosphorylation at that site (e.g., Ref. 