Cell, Vol. 67, 775-764,

November

15, 1991, Copyright

0 1991 by Cell Press

Nucleoside Triphosphates Are Required to Open the CFTR Chloride Channel Matthew P. Anderson,’ Herbert A. Berger,’ Devra P. Rich,” Richard J. Gregory,t Alan E. Smith,t and Michael J. Welsh’ *Howard Hughes Medical Institute Department of Internal Medicine Department of Physiology and Biophysics University of Iowa College of Medicine Iowa City, Iowa 52242 tGenzyme Corporation One Mountain Road Framingham, Massachusetts 01701

Summary The CFTR Cl- channel contains two predicted nucleotide-binding domains (NBDl and NBDP); therefore, we examined the effect of ATP on channel activity. Once phosphorylated by CAMP-dependent protein kinase (PKA), channels required cytosolic ATP to open. Activation occurred by a PKA-independent mechanism. ATPyS substituted for ATP in PKA phosphorylation, but it did not open the channel. Several hydrolyzable nucleotides (ATP > GTP > ITP = UTP > CTP) reversibly activated phosphorylated channels, but nonhydrolyxable analogs and Mg*+-free ATP did not. Studies of CFTR mutants indicated that ATP controls channel activity independent of the R domain and suggested that hydrolysis of ATP by NBDl may be sufficient for channel opening. The finding that nucleoside triphosphates regulate CFTR begins to explain why CF-associated mutations in the NBDs block Cl- channel function. Introduction Cystic fibrosis (CF), a common lethal genetic disease characterized by defective epithelial Cl- transport (Boat et al., 1989; Quinton, 199Oa), is caused by mutations in the gene encoding cystic fibrosis transmembrane conductance regulator (CFTR) (Riordan et al., 1989; Kerem et al., 1989; Rich et al., 1990; Drumm et al., 1990). Amino acid sequence analysis and comparison with other proteins suggested that CFTR consists of five domains (Riordan et al., 1989) (Figure 1): two membrane-spanning domains, each composed of six transmembrane segments; an R domain, which contains many potential consensus phosphorylation sites; and two nucleotide-binding domains (NBDs). Recent studies indicate that CFTR is a Cl- channel and have begun to provide information about the functional roles of the membrane-spanning and R domains. The membrane-spanning domains contain hydrophobic residues that are predicted to cross the lipid bilayer as c helices (Riordan et al., 1989). CFTR expression produces CAMP-activated Cl- channels in CF epithelial cells (Rich et al., 1990; Drumm et al., 1990) and in several other cell types that normally lack such channels (Anderson et al.,

1991b; Kartner et al., 1991; Berger et al., 1991), suggesting that CFTR forms a transmembrane Cl--conducting pore (Figure 1). More direct evidence that the membrane spanning domains may form a channel came from the finding that mutation of amino acids within the proposed membrane-spanning sequences alters the anion selectivity of Cl- channels generated by expression of CFTR (Anderson et al., 1991a). Phosphorylation of the R domain appears to regulate the Cl- channel (Figure 1). This conclusion is supported by several observations. The first is that addition of the catalytic subunit of PKA to the cytosolic surface of excised, cell-free patches of membrane activates the CFTR CIchannel (Berger et al., 1991). Second, four serines (residues 880, 737, 795, and 813) in the R domain are substrates for CAMP-dependent protein kinase (PKA) in vitro and are phosphorylated in vivo when cellular levels of CAMP increase (Cheng et al., 1991). Third, mutation of all four of those serines to alanine prevents CAMP-dependent activation of CFTR Cl channels (Cheng et al., 1991). Finally, expression of CFTR in which most of the R domain has been deleted generates Cl- channels that are constitutively active without an increase in CAMP (Rich et al., 1991). Despite this progress in understanding the domain structure of CFTR, the function of the NBDs remains an enigma. Understanding their function is particularly important, because the majority of missense mutations that cause CF are found within the NBDs (Cutting et al., 1990; Kerem et al., 1990). Many NBD missense mutations result in abnormal processing of CFTR, suggesting that the protein may never reach the plasma membrane (Cheng et al., 1990; Gregory et al., 1991). However, some NBD mutants that are processed normally still fail to generate CAMPdependent Cl- channel activity (Gregory et al., 1991). Although these results do not tell us how the NBDs function, they clearly indicate that the NBDs are important for normal function of the CFTR Cl- channel. The NBDs were originally defined (Riordan et al., 1989) by their sequence similarity with two nucleotide-binding sequences (motifs A and B; Walker et al., 1982) and their sequence similarity to the traffic ATPase or the ABC transporter family of proteins, which includes MDR, STEG, and periplasmic permeases (Riordan et al., 1989; Ames et al., 1990; Hyde et al., 1990). Several members of this family utilize the energy released during ATP hydrolysis to actively transport substrate across the cell membrane (Ames et al., 1990; Hyde et al., 1990). This hydrolysis occurs in the NBDs. Because CFTR functions as a Cl- channel rather than as a Cl- pump (i.e., it passively conducts CIdown its electrochemical gradient), we considered the possibility that the CFfR NBDs may regulate the channel, rather than support active transport. We therefore asked, does ATP regulate the CFTR Cl- channel? The approach to this question is complicated because, as shown in Figure 1, ATP may be involved in channel regulation at more than one site. At a minimum, ATP is required for PKA-

Cdl 776

A. 100

1

ATP JlJ&

ATP

a 8 E50 2 f 0 0lJ!!lLl 0

1000

2000

Time(s) B.

Figure

1. Model Showing

the Proposed

Domains

of CFTR

MSD refers to membrane-spanning domain, NBD refers to nucleotidebinding domain, and PKA refers to CAMP-dependent protein kinase. Membrane is cross-hatched area. See text for details. Time(s)

dependent phosphorylation (Cheng et al., 1991; Berger et al., 1991). We were interested in learning whether ATP might function at a second point, independent of PKA. Thus, to test for additional effects of ATP, we first phosphorylated the channel and then assessed the effect of ATP. We found that in addition to its role in phosphorylation, ATP directly regulates the channel. Results ATP Regulates the CFTR Cl- Channel We measured macroscopic currents in excised, cell-free patchesof membrane that contained large numbersof low conductance Cl- channels, to increase the signal-to-noise ratio and to reduce the variability that is observed when only one or a few channels are present in the patch. We previously described the single channel properties of the CFTR Cl- channel in excised, inside-out membrane patches and showed that the single-channel and macroscopic currents had the same properties as Cl- channel currents measured with the whole-cell patch or in the apical membrane(Bergeret al., 1991; Anderson et al., 1991a; Anderson and Welsh, 1991). Addition of ATP (1 mM) to the cytosolic surface of excised patches had no effect, but the combination of ATP plus the catalytic subunit of PKA activated Cl- channels (Figure 2A) (Berger et al., 1991). When we eliminated PKA, but kept ATP on the cytosolic surface, the current did not reverse. Only when ATP was removed did current return to prestimulation levels (Figure 2A). Moreover, the channel was reactivated by readding ATP. These results indicate that ATP regulates the channel, but only when it has first been phosphorylated by PKA. They also suggest that PKA-dependent phosphorylation is relatively irreversible in the excised, cell-free patch. This result contrasts with our previous findings with whole-cell (Anderson et al., 1991 b) and cell-attached patches (Berger et al., 1991) that current from CFTR Cl- channels decreased when CAMP agonists were removed. That reversibility most likely resulted from dephosphorylation of CFTR by a cytosolic

Time(s)

Figure

2. ATP Regulates

the CFTR Cl- Channel

ATP (1 mM, Mg2+ salt), the catalytic subunit of PKA (75 nM) (Promega, Madison, Wisconsin), and PKI (3uM, either a gift of Ft. A. Maurer, University of Iowa or from Sigma Chemical Co., St. Louis, Missouri) were present in the internal solution (bath) during the times indicated by the bars. Representative examples from 3T3 fibroblasts stably expressing CFTR ([A] and [Cl) (n = 36) or HeLa cells transiently expressingCFTR(B)(n = 4). PKIdidnotinhibittheATPdependentCIfchannel activity (C). In 3T3 fibroblasts stably expressing CFTR, baseline current was 3 f 1 pA (mean f SEM) and remained at 3 2 1 pA after 1 mM MgATP, but increased to 136 2 29 pA in the presence of PKA and ATP (n = 14). After removing PKA but retaining ATP in the bath, current remained at 82 f 5% (n = 9) of the value in the presence of ATP plus PKA. In HeLa cells, current was 4 f 1 pA (n = 7) under basal conditions, 3 f 1 pA, 169 * 21 s after addition of 1 mM MgATP, and 193 f 64 pA after addition of PKA (n = 7, average time to peak, 121 f 33 s). Current remained elevated at 146 f 72 pA (n = 6) after PKA was removed but PKA remained. PKI (3 uM) inhibited the effect of PKA by 86% f 6% (n = 6) of the subsequent response to PKA in the absence of PKI. When PKI was added to phosphorylated channels in the presence of ATP, current was 65% f 6% control after 251700 s (n = 6).

phosphatase that is lost when membrane patches are excised from the cell. Similar results were obtained in HeLa cells transiently expressing CFTR (Figure 28). ATP alone did not open channels, but once channels were phosphorylated by PKA, ATP reversibly activated current multiple times (four of four patches). However, the total current often decreased over time in both HeLa cells and NIH 3T3 fibroblasts (e.g., Figure 28; see also Figures 38 and 48 below). A similar phenomenon, often referred to as “rundown,” has been observed with several other ion channels (Horn and

Nucleotide

Regulation

of CFTR

777

Table

0

700

1400

2100

1. Effect

of Nucleotides

Nucleotide

Percent

ATP AMP-PNP ATP-I-S AMP-PCP AMP-CPP Me*-free ADP CAMP

100.0

ATP

7.5 6.5 4.0 75.1 12.1 10.2 0.1

on CFTR ATP f f f f f f f

2.8 3.1 1.4 4.5 4.0 4.0 1.4

Cl- Channel f

SEM

Activity n

5 4 5 3 4 5 4

Results are for each nucleotide expressed as the percent of current observed with MgATP in the same patch. All nucleotides are present at 1 mM concentration. n is number of patches studied. For Mg-free ATP, 1 mM EDTA replaced EGTA, no MgN was added, and 1 mM NeATP was used.

0-c 0

1000

2000

Time(s)

150-

$ loot 3

$ -:“ OJ

Figure

3. Effect

of Nucleotides

lb

2’0 [ATP) mM

on CFTR

Clt Channel

I

30

Activity

(A) Effect of nonhydrolyzable ATP analogs and other nucleotides on channel activity. MgATP, AMP-PNP, AMP-PCP, ADP, and CAMP were present at 1 mM during the times indicated by bars. The concentration of catalytic subunit of PKA was 75 nM. Calyculin A (100 nM, L. C. Services Corp., Woburn, Massachusetts) was present throughout; see below. Data are from experiments with 3T3 fibroblasts expressing CFTR. (B) Effect of hydrolyzable nucleotide triphosphates on channel activity. Example is patch from a 3T3 fibroblast expressing CFTR. CTP, ITP, ATP, GTP, and UTP were present at 1 mM during indicated times. At start of record, channels had already been activated using 75 nM catalytic subunit of PKA and 1 mM MgATP. (C) Effect of ATP concentration on channel activation. Data points are mean * SEM of at least three values at each concentration. Data are from eight membrane patches. Because each patch contains a different number of CFTR Clt channels, current was normalized to the activation obtained with 1 mM ATP. Pipette and bath solutions were as described in methods except 30 mM MgClr was used in the bath.

Marty, 1988). Rundown was variable, but was often most noticeable during the initial exchanges of the cytosolic bath. This rundown also occurred in the presence of phosphatase inhibitors and in CFTR lacking the R domain (see below). These results indicate that ATP opens CFTR Cl- channels in two cell types using two different expression systems. Because the vaccinia expression system suppresses host protein synthesis, these results cannot be explained by production of some other protein, synthesized as a result of CFTR expression. ATP does not reversibly stimulate the channel through the activity of residual PKA. PKI, aspecific peptide inhibitor

of PKA (Cheng et al., 1988) blocked activation of Cl- channels by PKA (Figure 2C). However, once channels were phosphorylated by PKA, PKI did not alter the current regardless of whether it was added after ATP (Figure 2C) or before ATP (for an example, see Figure 58). These results indicate that ATP is required during phosphorylation of the CFTR Cl- channel by PKA. But more importantly, they establish that channel opening requires ATP in a second distinct mechanism that is independent of PKA. The remainder of this work focuses on the PKAindependent mechanism of channel regulation. Hydrolyzable Analogs of AfP Are Required to Open the PKA-Phosphorylated CFTR Cl- Channel The experiments described above raised the question of whether ATP hydrolysis is required to open phosphorylated channels. To address this question, we tested the effect of ATP analogs. Adenosine 5’-(8,y-methylene) triphosphate (AMP-PCP) contains a methyl group in place of oxygen between the 8 and y phosphate; this substitution inhibits hydrolysis (Moos et al., 1980). Its structural analog, adenosine 5’-(a$-methylene) triphosphate (AMPCPP), contains the methyl group between the a and 8 phosphate; this analog is hydrolyzed by some ATPases (Krug et al., 1973). We found that AMP-CPP, but not AMPPCP, substituted for ATP in regulating the channel (Table l), suggesting that hydrolysis of the terminal phosphate was important. Another nonhydrolyzable ATP analog, adenosine 5’~(8,y-imino) triphosphate (AMP-PNP) (Yount et al., 1971), as well as ADP and CAMP, failed to activate channels (Figure 3A and Table 1). MgZ+, a cofactor required in ATP hydrolysis reactions, was also required for nucleotide regulation of the CFTR Cl- channel (Table 1). We found that several other nucleoside triphosphates also opened PKA-phosphorylated CFTR Cl- channels. At a concentration of 1 mM, the nucleotide selectivity sequence was (Figure 38 and Tables 1 and 2): ATP > AMPCPP > GTP > ITP = UTP > CTP. This broad nucleotide specificity contrasts with the high specificity for ATP observed for a number of kinases (PKA, Lemaire et al., 1974; and IGFI receptor, Sasaki et al., 1985) and the Na+/K+ ATPase (Glynn and Hoffman, 1971). ATP increased Cl-channel activity in a dose-dependent

Cdl 770

Table 2. Effect of Hydrolyzable Cl- Channel Activity Nucleotide

Percent

ATP GTP ITP UTP CTP

100.0 60.8 48.5 42.2 25.2

Nucleotide ATP f

f f f f

7.1 6.6 4.4 5.6

Triphosphates SEM

on CFTR

A. ATP

n

12 6 6 5

PKA

PKA

0 0

Results for each nucleotide are expressed as the percent of current observed with ATP in the same patch. All nucleotides are present at 1 mM concentration. “n” is number of patches studied.

Stauro.

stauro

100

800

ATP Does Not Regulate the CPTR Cl- Channel via Reversible Phosphorylation One possible mechanism by which ATP might regulate CFTR Cl-channels is through phosphorylation by a kinase other than PKA. Such regulation would require that a kinase must be present consistently within all excised membrane patches. The kinase reaction must show an unusually high EC% for ATP and must demonstrate broad nucleotide specificity. The very consistent and rapid reversal of the nucleotide effect (reversal occurred within 76 of: 14s,n = 14, and was limited by the bath exchange rate) would require that a protein phosphatase be present in all excised membrane patches. Such a phosphatase must have specificity: dephosphorylating the nucleotide-regulated site, but not the PKA-phosphorylated sites. Despite these considerations, several additional studies were performed to further exclude this possibility. To determine whether a kinase mediates the effect of nucleoside triphosphates, we used a nonspecific protein kinase inhibitor, staurosporine (Tamaoki et al., 1966) at a high concentration (5 PM). Staurosporine is known to inhibit the following kinases with IC, vslues of GTP (79%) > TTP (26%) > CTP (11%) (Ames et al., 1989). The multidrug resistance protein also requires hydrolyzable ATP: AMP-PNP would not substitute for ATP in supporting drug transport (Horio et al., 1988); both ATP and GTP inhibited labeling with 8-azido-a-ATP (Cornwell et al., 1987); and the purified protein hydrolyzed both ATP and

GTP (Hamada and Tsuruo, 1988). These results contrast with data from the ATP-regulated K” channel in which nonhydrolyzable analogs of ATP substitute for ATP in channel inhibition (Ashcroft and Ashcroft, 1990). One property that distinguishes classes of ion motive ATPases is their sensitivity to vanadate (Forgac, 1989; Pedersen and Carafoli, 1987) a transition state analog of P04-. Those ATPases that form acovalent phosphorylated intermediate, such as the plasma membrane Na+/K+ ATPase, Ca2+ ATPase, and H+ ATPase, are inhibited by low concentrations of orthovanadate (10 f&l). In contrast, the F,F,,H+ ATPases and vacuolar H+ ATPases, in which phosphorylated intermediate states have not been found, are insensitive to vanadate. Our data showing that vanadate (500 f&l, Figure 4C) did not inhibit CFTR Cl- channel currents contrasts with reports on MDR and the histidine periplasmic transport system: vanadate inhibited substrate transport by 50% at 10 uM in MDR (Horio et al., 1988) and by 70% at 500uM in the histidine transport system (Ames et al., 1989). The differential sensitivity to vanadate suggests that nucleotide regulation of CFTR may occur through a different mechanism than in MDR and the histidine transport system. The observation that CFTR forms a Cl- channel that passively transports Cl- down its electrochemical gradient indicates that the energy of ATP hydrolysis is not directly involved in the movement of Cl- through the pore. How then does ATP regulate the channel? One possibility is that the energetically favorable conformation of the channel is the inactivated (or closed) state and that energy input is required for the transition to and maintenance of the activated (or open) state. That energy could come from ATP hydrolysis. By analogy, energy (in the form of a change in membrane voltage) is required for the conformational change that activates the family of voltage-gated channels. What might be the physiologic role of ATP regulation of CFTR? One possibility is that nucleotide regulation provides a way of matching the rate of Cl- secretion with the availability of cellular ATP. In a Cl-secreting epithelium, the transepithelial flux of electrolytes can be substantial. For example, at a secretory rate of 3 BEq x cm+ x hr-l (the rate obsenred in the Cl--secreting T84 cell line; Mandel et al., 1986) the entire intracellular Cl- content turns over every 1.6 min, and the intracellular Na+ content turns overevery0.8min(Schultz, 198l).Theresultingmetabolic demands are significant, and it is imperative for maintenance of cell composition and volume that transepithelial transport not exceed the metabolic energy necessary to power it. Because apical membrane Cl- channels are a key site in the regulation of transepithelial Cl- transport, they might be an advantageous point at which to couple cellular ATP levels to transport. A decrease in cellular ATP would close apical Cl- channels, thereby decreasing the rate of Cl- secretion and the metabolic demand on the cells (Quinton, 1990b). The obsenration that the concentrations of ATP that regulate CFTR are in the physiologic range suggests the plausibility of this regulatory mechanism. In contrast, ATP levels could not readily regulate transport via either PKA-dependent phosphorylation or the basolat-

era1 Na+/K+ ATPase, because the concentrations of ATP required for their half-maximal activation, 3 uM (Flockhart et al., 1984) and 0.2-2 uM (Taylor and Green, 1989) respectively, would be too low for effective coupling of transport to metabolic energy. In summary, our data show that nucleoside triphosphates regulate the CFTR Cl- channel. This regulation most likely occurs via the NBDs and may require hydrolysis because only hydrolyzable nucleotides activated the channel. These data begin to explain why CF-associated missense mutations in the NBDs block Cl- channel function, the hallmark abnormality in CF epithelia. Experimental

Procedures

Cells and CFTR Expresslon Systems We used two different cell types and CFTR expression systems. NIH 3T3 fibroblasts infected with a retrovirus expressing human CFTR were prepared and maintained as previously described (Anderson et al., t99lb). HeLa cells were used for the transient expression of CFTR, CFTRAR. and CFTR-KIPBOM. CFTRAR was constructed as previously described (Rich et al., 1991) and contains a deletion of bases 2254 through 2637 of the CFTR cDNA corresponding to amino acids 706-635 (Rich et al., 1991). CFTR-K125OM was constructed as previously described (Gregory et al., 1991) and contains methionine at position 1250 instead of lysine. CFTR-K1250M was expressed by transfection of the plasmid pTM-CFTR-K1250M and infection with a vaccinia virus expressing the T7 RNA polymerase (Elroy-Stein et al., 1969). We expressed CFTR or CFTRAR using a double infection with a recombinant virus expressing the bacteriophage T7 RNA polymerase (Elroy-Stein et al., 1969) and a second recombinant virus containing either CFTR or CFTRAR sequences preceded by a T7 promoter. Recombinant vaccinia virus containing the coding sequence for CFTR or CFTRAR was constructed by homologous recombination of wild-type vacciniavirus (WR strain) with either pTMCFTR4 (Gregoryet al., 1990) or pTM-CFfRAR (Rich et al., 1991) (which contain flanking sequences of thevacciniavirus thymidine kinasegene; Moss et al., 1990). Homologous recombination was initiated by infection of HeLa cells (5.4 x 10’ cells per cm’) with 0.1 MOI of wild-type vaccinia virus. After infection, cells were transfected with pTMCFfR4 or pTM-CFTRAR (5 pg/ Iv cells) using lipofectin (41 ratio of lipid to DNA) as previously described (Rich et al., 1990). Growth media was added to the cells 6 hr after transfection. Cells were harvested in modified Eagle’s medium after 2-3 days, and virus was released by three cycles of freezing and thawing. Recombinant viruses were selected by plaque assay of cell lysates on TK- 143 cells in the presence of 25 kg/ml 5-bromo-2’deoxyuridine (Mackett et al., 1965). Recombinant viruses were plaque purified twice and screened for incorporation of CFTR DNA using the polymerase chain reaction (PCR) (Innis and Gelfand, 1990). Large stocks of positive viral recombinants were subsequently prepared in HeLa cells, and viral titer was determined as described (Mackett et al., 1965). Electrophyslology We used the excised, inside-out patch clamp technique as previously described (Hamill et al., 1961; Berger et al., 1991). A List EPC7 (Medical Electronic, Darmstadt, Federal Republicof Germany) amplifierwas used for current amplification and voltage clamping, and a laboratory computer system (Indec Systems, Inc., Sunnyvale, California) was used for data acquisition and analysis. Current was filtered at 1 kHz and sampled every 500 ps. The pipette (external) solution contained (mM): 140 N-methyl-o-glucamine, 2 MgC&, 5 CaCl*, 100 L-aspartic acid, and 10 HEPES @H 7.3 with HCI) (Cl- concentration, 49 mM). The bath (internal) solution contained (mM): 140 N-methyl-o-glucamine. 3 MgCl*, 1 CsEGTA, and 10 HEPES (pH 7.3 with HCI) (Cl- concentration, 147 mM). The estimated free Ca- concentration in the internal solution was