Regulation of the Cystic Fibrosis Transmembrane Conductance Regulator Cl− Channel by Its R Domain*

Since the sequence of the CFTR Cl channel was discovered, the function of its R domain has been puzzling. CFTR is a member of the ATP-binding cassette (ABC) transporter family, and it contains the features characteristic of this family: two nucleotide binding domains (NBDs) and two membrane-spanning domains (MSDs) (1). Yet in addition, CFTR also contains the R domain, a unique sequence not found in other ABC transporters or any other proteins. The R domain serves as the major physiologic regulator of the CFTR Cl channel (2, 3). Upon elevation of cAMP levels, cAMP-dependent protein kinase (PKA) phosphorylates the R domain allowing ATP to open and close the channel. Yet how phosphorylation activates the channel is not well understood. Some models propose that the R domain prevents the channel from opening and that phosphorylation relieves this inhibition. Other models suggest that phosphorylation of the R domain stimulates activity. Here we briefly review current knowledge that may help explain the function of this interesting domain.

Since the sequence of the CFTR 1 Cl Ϫ channel was discovered, the function of its R domain has been puzzling. CFTR is a member of the ATP-binding cassette (ABC) transporter family, and it contains the features characteristic of this family: two nucleotide binding domains (NBDs) and two membrane-spanning domains (MSDs) (1). Yet in addition, CFTR also contains the R domain, a unique sequence not found in other ABC transporters or any other proteins. The R domain serves as the major physiologic regulator of the CFTR Cl Ϫ channel (2,3). Upon elevation of cAMP levels, cAMP-dependent protein kinase (PKA) phosphorylates the R domain allowing ATP to open and close the channel. Yet how phosphorylation activates the channel is not well understood. Some models propose that the R domain prevents the channel from opening and that phosphorylation relieves this inhibition. Other models suggest that phosphorylation of the R domain stimulates activity. Here we briefly review current knowledge that may help explain the function of this interesting domain.

Sequence and Boundaries of the R Domain
The R domain was originally defined as those residues encoded by exon 13 (aa 590 -830) ( Fig. 1) (1). Although the precise boundaries of the R domain remain uncertain, recent studies suggest that the N-terminal portion of exon 13 is actually part of NBD1, whereas the C-terminal portion constitutes the structural and functional "R domain." Supporting this conclusion, the N-terminal portion of exon 13 has sequence similarity with NBDs of other ABC transporters (4,5). More recently, crystal structures were solved for the NBDs of the ABC transporters HisP, MalK, and Rad50 (6 -8). In CFTR, residues analogous to those NBDs would extend to approximately aa 642. Experimental evidence for the boundaries came with the demonstration that deletion of aa 708 -835, but not deletions extending further in the N-terminal direction, produced a channel that was processed correctly, opened in the presence of ATP, and had conductive properties like those of wild-type CFTR (4,9). In addition, severing CFTR after residues 633 or 835 and coexpressing the two halves in Xenopus oocytes produced functional channels, whereas severing the channel between exons 12 and 13 abolished function (10,11). All these data suggest that the N-terminal R domain boundary begins between residues 634 and 708 and that the C-terminal boundary ends in the region of residue 835.
The most striking feature of the R domain is the presence of multiple consensus PKA phosphorylation sites that are highly conserved across species (Fig. 1). There is little other sequence similarity in the R domain, unlike the substantial sequence conservation throughout the rest of CFTR. We are not aware of any other conserved protein motifs in the R domain, although there is a relatively high percentage (28%) of charged residues. Sixteen of the 24 basic residues are in PKA consensus sites. Several of the acidic residues are clustered within 817-838 (12).
The functional consequences of phosphorylation have been evaluated by studying CFTR mutated at one or many of the phosphorylatable serines (13, 15, 18 -23). Serine mutations reduced but did not abolish phosphorylation-stimulated activity, and no single serine was required for stimulation. Even when all the consensus phosphoserines were mutated, a small amount of PKA-mediated Cl Ϫ current remained, suggesting a contribution from sites that are not normally phosphorylated in vivo such as Ser-753 (15,19). Nonetheless, the majority of PKA-dependent stimulation appears to result from phosphorylation of Ser-660, Ser-700, Ser-737, Ser-795, and Ser-813.
Although no one serine is essential, individual serines seem to stimulate channel activity to different degrees. For example, mutation of either Ser-660 or Ser-813 reduced the open state probability (P o ) of CFTR studied under several conditions (20,21,23). Thus, these two residues play key stimulatory roles. Phosphorylation of Ser-795 also stimulated, although to a lesser extent (21,23). Interestingly, the consequences of mutating Ser-737 differed depending upon how it was studied. In excised, cell-free patches, the S737A mutant reduced P o (20), but in Xenopus oocytes and epithelia, the S737A mutant increased cAMP-stimulated current (21,23). We speculate that these differences may be because of loss of kinases, phosphatases, or associated molecules after membrane excision.
Various phosphoserines also show functional interactions. For example, mutating either Ser-660 or Ser-813 alone reduced phosphorylation-stimulated activity. However, when either Ser-660 or Ser-813 was present in a variant in which the other phosphorylation sites were mutated, current did not increase; both Ser-660 and Ser-813 were essential for current to reach wild-type levels (23). Likewise, Ser-737 was unable to stimulate or inhibit unless other serines were also present (21,23). These data suggest either concomitant or sequential phosphorylation events may be required.

Does the Unphosphorylated R Domain Inhibit
Channel Activity? Additional clues to R domain function came from studies of R domain deletions. Deleting aa 708 -835 (⌬R) or 768 -830 ( Fig. 2) generated channels that were open in the presence of ATP but did not require phosphorylation, i.e. they were constitutively active (4,24,25). These data suggested that the unphosphorylated R domain inhibits activity and that phosphorylation or deletion eliminates the inhibition. This led to the proposal that the R domain might be an inhibitory particle, much like the N terminus of the Shaker K ϩ channel (26 to inhibit constitutive activity. For example, coexpressing two halves of CFTR (aa 1-835 and aa 837-1480) that retain the R domain generated a constitutively active channel ( Fig. 2) (11,27). Likewise, translocating aa 709 -835 to the C terminus of CFTR-⌬R produced constitutively active channels (28). Moreover, adding isolated unphosphorylated R domain proteins encompassing aa 645-834, 590 -858, or 708 -831 did not reduce constitutive activity of CFTR-⌬R (20,24,29).
To identify R domain sequences that might prevent constitutive activity, other deletions were studied. A smaller R domain deletion (aa 760 -835) also induced constitutive activity, whereas deletions of residues 708 -759, 784 -835, and 780 -830 did not (Fig. 2) (25,28). This focused attention on residues 760 -783; deletion of these residues generated a constitutively active channel (28). There are no obvious sequence motifs between aa 760 and 783, although the sequence 765 RRQSVL(N/D)LMT 774 is conserved across species. Ser-768 is located in this region, but it is not phosphorylated in vivo (13,14), and its mutation did not induce constitutive activity (21,28). Interestingly, adding residues 760 -783 as isolated peptides or translocating them to the C terminus failed to prevent constitutive activity (Fig. 2).
Thus, the R domain probably has an inhibitory function that requires a small portion in the middle of the domain. However, those residues alone are not sufficient to prevent constitutive activity, and to have an effect it appears that they must be retained in their normal location between NBD1 and MSD2.

Does the Phosphorylated R Domain Stimulate
Channel Activity? Direct evidence that the phosphorylated R domain is stimulatory came from studies that added isolated R domain proteins to CFTR-⌬R channels. Phosphorylated, but not unphosphorylated, proteins consisting of residues 645-834, 590 -858, and 708 -831 each stimulated Cl Ϫ current (Fig. 2) (20,24,29). In addition, translocating aa 709 -835, 709 -759, or 760 -835 to the C terminus restored PKA-stimulated activity to CFTR-⌬R (28). One study suggesting that the R domain is not stimulatory found that channels generated by coexpressing aa 3-633 and 837-1480 opened as often as wild type (Fig. 2) (11). However, the P o of those severed channels was only half that of wild-type channels. Another study found that deleting the negatively charged region at aa 817-838 generated a constitutively active channel whose activity was not increased by PKA (Fig. 2) (12). This finding suggested that this relatively conserved region was required for stimulation.
These data clearly indicate that the R domain can stimulate activity, although the mechanism is unknown. Interestingly, stimulation can come from multiple R domain regions, more than one phosphoserine can stimulate, and no one specific phosphoserine or unique sequence is required. The data also suggest that stimulatory regions differ from those preventing constitutive activity.

R Domain Phosphorylation Enhances ATP Interaction
with the NBDs Phosphorylation raises P o by increasing the rate at which the channel opens (20,22,24). Phosphorylation also increases the apparent affinity of the NBDs for ATP; this effect differs for different phosphoserines (20,22). In general the more phosphorylation, the greater the activity. This is likely to be of physiologic importance, as graded phosphorylation could generate graded apical membrane Cl Ϫ channel activity allowing precise control of transepithelial Cl Ϫ transport.

The R Domain Is Predominantly Unstructured in Solution
To better understand the functional and structural features of this unique domain, we expressed and purified an R domain protein comprising aa 708 -831 (named R8) (29). Analytical ultracentrifugation showed that soluble R8 was a monomer, and limited proteolysis indicated that most of R8 was accessible to proteases. Phosphorylation did not change these properties, suggesting that it did not induce global conformational changes. Importantly, phosphorylated R8 stimulated channel activity, suggesting that if the R domain has a specific structure, it is retained in this protein.
Circular dichroism (CD) suggested that R8 had little well ordered secondary structure; helical content was about 5% with the remaining protein being random coil (29). The results were the same irrespective of whether spectra were obtained in high pH, low salt, buffers used for patch-clamp studies, trimethylamine oxide (which promotes protein folding), or R8 incorporated in phospholipid micelles. Phosphorylated and nonphosphorylated R8 had similar CD spectra, suggesting that phosphorylation did not induce large structural changes. CD spectra of larger R domain proteins including all or parts of NBD1 revealed significant secondary structure; aa 595-831 had 10% ␣-helix and 30% ␤-sheet (30), whereas aa 404 -830 had 19% ␣-helix and 43% ␤-sheet (16). The secondary structure in those proteins may arise from the NBD1 portion and the random coil from the R domain.
Although other ABC transporters do not contain a region homologous to the R domain, P-glycoprotein contains a central region connecting NBD1 and MSD2 that includes two PKA motifs. Deleting this linker region (aa 653-686) disrupted drug transport and ATP hydrolysis (31). Replacing the linker with an unrelated sequence predicted to form a flexible structure restored both drug transport and ATP hydrolysis. Interestingly, when linker residues 644 -689 from P-glycoprotein replaced aa 780 -830 in CFTR, stimulated channel activity was restored to wild-type levels (25). These observations suggest that a defined amino acid sequence is not critical, but PKA consensus motifs, a certain degree of flexibility, and possibly an optimal length are key in determining channel activity.
Several other observations support the conclusion that the functional R domain is predominantly random coil. First, the entire R domain was neither required to stimulate activity nor to prevent constitutive activity; portions were sufficient (Fig. 2). If the R domain had a well ordered tertiary structure, such alterations would likely be disruptive. Second, the R domain sequence is not well conserved across species (Fig. 1). Most highly structured domains show substantial sequence conservation; examples are the NBDs. Moreover, many of the conserved residues in such proteins have hydrophobic side chains buried within the protein core where they help maintain structure (32). In contrast, the conserved R domain sequences are consensus PKA phosphorylation motifs that are expected to be solvent-exposed where they serve functional rather than structural roles. Third, of the more than 972 mutations and variations reported in the CFTR gene, 2 only 9 missense variations lie between aa 708 and 831; two are at conserved PKA motifs (Fig. 1). Note, however, that not all of these reported sequence variations are actually known to cause cystic fibrosis because clinical information is limited. Of these 9, for the few variants that have been studied, their biosynthesis does not differ from wild-type (34), consistent with the idea that an amino acid change in random coil would be unlikely to disrupt structure. Moreover, these variants show relatively small functional differences (34).

How Does an R Domain Composed Primarily of Random
Coil Regulate Activity? If the R domain is predominantly random coil and if phosphorylation does not increase overall structure, how does the R domain regulate the channel? There are several possibilities. (a) Phosphorylation might induce a conformational change but only in small discrete regions. (b) Introduction of negative charge by phosphorylation may be sufficient to initiate activity; consistent with this, replacement of serines with aspartates stimulates activity (18). (c) Phosphorylation may not directly change conformation but may be required before the R domain can interact with other sites within CFTR. (d) The R domain may adopt a more ordered structure only upon contact with the rest of CFTR. The last two possibilities are reminiscent of the interaction of the kinase-inducible domain (KID) of the cAMP-responsive element-binding protein (CREB) with CREB-binding protein (CBP). Several structural parameters indicate that KID is unstructured, and serine phosphorylation does not increase structure (35)(36)(37)(38)(39). However, just as phosphorylation allows the R domain to activate the Cl Ϫ channel, phosphorylation allows KID to bind and activate CBP. This binding induces a conformational change in KID, which is limited to the phosphoserine and a few neighboring amino acids.
How does a phosphorylated random coil R domain stimulate activity? Fig. 3 shows some speculative models (29). The phosphorylated R domain might interact with a single (B) or multiple sites (C and D) on the rest of CFTR. Interactions may be specific (C) or nonspecific (D). The number of phosphoserines and interaction sites is arbitrary and for illustrative purposes only. In this model, we emphasize the advantages of an unstructured but phosphorylated R domain for stimulation. However, the model does not show how an unphosphorylated R domain prevents constitutive activity. 2 CF Genetic Analysis Consortium, unpublished data. Our current level of understanding makes it difficult to include both stimulatory and inhibitory aspects in the same model.
Model A shows the unphosphorylated R domain as a random coil. Without phosphorylation, it does not stimulate activity. Model B shows a phosphorylated R domain. Because the R domain is unstructured, any of several different phosphoserines could interact with the single site to stimulate activity. Model B could explain the ability of multiple different phosphoserines to stimulate, possibly with quantitatively different effects on activity. Model C shows multiple phosphoserines interacting with unique sites in CFTR. This model could account for increasing activity with increasing phosphorylation; if there were 3, 4, or 5 phosphoserines, there would be 3, 4, or 5 matching interaction sites. This model could also account for quantitatively different effects generated by phosphorylation of different serines.
Model D shows multiple phosphoserines interacting with multiple binding sites. Here we show only two, suggesting the possibility of at least one interaction site at each NBD. However, more interaction sites may exist, as evidenced by the observation that regions within the R domain and the N terminus interact (40). Model D could account for the apparent promiscuous relationship between phosphorylation sites and stimulation of activity as depicted by the arrows. In such a model, a relatively unstructured R domain would allow multiple different phosphoserines to stimulate, no one phosphoserine would be required, different phosphoserines could have quantitatively different effects, and the more serines phosphorylated, the greater the activity.

Advantages of Intrinsically Unstructured Domains
There is increasing recognition that many protein domains and full-length proteins are intrinsically unstructured (41,42). Examples include the cell cycle kinase inhibitor p21 waf1/Cip1/Sdi1 (43), the FlgM flagellar protein of Salmonella typhimurium (44), and the fibronectin receptor-binding protein (45). Interestingly, a regulatory sequence in another channel, the fast inactivation domain of the Shaker B K ϩ channel, also acts in a structure-independent manner (33). The advantages of unfolded proteins for signaling or regulation are also becoming more apparent (41,42). In CFTR, an R domain composed predominantly of random coil would retain the flexibility that permits facile interaction with multiple regions within the rest of CFTR and allow prompt, discrete, and variable reactions to phosphorylation of different serines.
Despite extensive study on the R domain, a number of critical questions remain. Why is spacing of PKA motifs conserved within random coil? Why are certain regions within the R domain necessary but not sufficient for inhibition, and why can't those residues be translocated within the molecule and retain function? Finally, what is the mechanism by which the R domain stimulates and inhibits the channel?