Cystic fibrosis transmembrane conductance regulator-associated ATP and adenosine 3'-phosphate 5'-phosphosulfate channels in endoplasmic reticulum and plasma membranes.

Cystic fibrosis (CF) is characterized by abnormal regulation of epithelial ion and fluid transport due to mutations in the CF transmembrane conductance regulator (CFTR), an apical membrane-localized Cl− channel, that usually prevent it from exiting the endoplasmic reticulum. Defective or absent CFTR in the epithelium is believed to disrupt fluid balance in human airways and thereby contribute to chronic respiratory inflammation. Patch-clamp of the plasma membrane and outer membrane of the nuclear envelope of nuclei isolated from CFTR-expressing Chinese hamster ovary cells revealed that CFTR is associated with a regulated ATP channel in both membrane compartments. CFTR expression was also shown to be associated with permeability to another adenine nucleotide, adenosine 3′-phosphate 5′-phosphosulfate, the universal sulfate donor in cells. These results may provide a link between the ion channel function of CFTR and abnormal glycoprotein processing observed in CF.

Cystic fibrosis (CF) 1 is a common genetic disease characterized by abnormal regulation of epithelial ion and fluid transport due to mutations in CFTR, an apical membrane-localized cAMP-regulated Cl Ϫ channel (1)(2)(3)(4)(5)(6)(7)(8). Hundreds of different mutations (1) can affect CFTR Cl Ϫ channel function by several mechanisms (8). However, most CF is caused by lack of CFTR in the plasma membrane due to mutations, including the most common one (⌬F508-CFTR) (8), that affect the ability of fully translated CFTR to exit from the endoplasmic reticulum (ER), where it is synthesized (9 -12). Defective or absent CFTR in the epithelium is believed to disrupt fluid balance in human airways and thereby compromise mucociliary clearance of inflammatory particles, including bacteria (13). The resulting chronic inflammation and tissue damage in the lung are the cause of most mortality in CF (14). Nevertheless, it remains unclear if the lack of plasma membrane Cl Ϫ channel activity can fully account for airway pathology in CF (13,15). Human airways are Na ϩ -absorbing tissues (16,17), and the role of CFTR in transepithelial Cl Ϫ transport is still undefined because Cl Ϫ seems to be at electrochemical equilibrium across the apical membrane (18), and NaCl absorption is enhanced in CF (17,19,20). Furthermore, other transport and biochemical abnormalities in the lung in CF may not be readily explained by the lack of plasma membrane CFTR Cl Ϫ channel activity, including elevated Na ϩ permeability through amiloride-sensitive channels (13,17,19,21) and abnormal regulation by cAMP of another airway Cl Ϫ channel, the outward rectifier (22)(23)(24)(25). In addition, altered glycoprotein processing is observed in respiratory cells from CF patients (26 -29). CF airway epithelial cells express higher levels of the receptor for Pseudomonas aeruginosa, which may be a consequence of its diminished sialylation (14,28), and secreted and cell-surface glycoconjugates are hypersulfated in the lung in vivo and in primary cultures of airway cells from CF patients (27, 30 -32). The evidence suggesting that abnormal protein sulfation is a fundamental biochemical defect in CF (31, 32) seems paradoxical in light of the identification of CFTR as a plasma membrane Cl Ϫ channel.
A recent study suggested that abnormal regulation of the outward rectifier in CF involves defective plasma membrane ATP permeability (33) and results from two studies by independent groups indicated that CFTR may either conduct ATP or be closely associated with a separate ATP conductance (33,34). However, these results are currently rather controversial because they have not been observed by all investigators (35)(36)(37). The possibility that CFTR is associated with ATP permeability has prompted us to consider whether other transport and protein-processing abnormalities in CF are related to this conductance. Specifically, in the present study we have considered whether protein processing alterations in CF might be related to an association of CFTR with ATP permeability in intracellular membranes. We recently demonstrated that CFTR as well as ⌬F508 CFTR function as Cl Ϫ channels when they are localized in the ER membrane (38). That wild-type CFTR functions in the ER as well in plasma membranes suggests that it also likely functions in intermediate compartments, including the Golgi, the site of most protein processing, including sialylation and sulfation. CF in most patients may therefore be associated with lack of Cl Ϫ channel activity not only in the plasma membrane but in the Golgi as well. If CFTR is associated with nucleotide conductances, altered nucleotide permeability in intracellular membranes may possibly contribute to cellular dysfunction in CF. cultured, and nuclei were isolated as described previously (38). Standard patch-clamp techniques were applied: (i) excised inside-out and cell-attached configurations for the plasma membrane, and (ii) excised patches (cytoplasmic side facing into the pipette) for the outer nuclear membrane (38). Stable (15-60 min) seals (10 -100 gigohms) were obtained using heat-polished electrodes (tip, Ͻ0.5 m in diameter) with 10 -50 megohms resistance. Ag/AgCl electrodes were used without salt bridges. When using low (ϳ5.2 and 0.7 mM) Cl Ϫ solutions (below) in the pipette, the electrodes were not backfilled with Cl Ϫ -containing solutions. These low Cl Ϫ conditions did not affect our ability to record currents, probably because the currents that the electrodes were required to pass in these experiments were small. Junction potentials were corrected in all experiments by an initial offset applied by the amplifier, with one exception relating to the third part of Fig. 4A, which has not been corrected for the junction potential that arose during the switch from the ATP to the Cl Ϫ solutions. The measured offset was ϳ6 mV, which does not significantly affect the interpretations. All experiments were performed at room temperature as described previously (38). Data were digitally filtered at 300 Hz unless indicated. Channel openings and closings were automatically detected using the Tac computer program, and amplitudes were determined by computer-assisted manual measurements of detected events. Data analyses were limited to records with low base-line drift. Kinetic analyses were restricted to single-channel patches. Results are expressed as mean Ϯ S.E. The following solutions were used: Solution A, high Cl Ϫ solution containing 120 mM N-methyl-D-glucamine chloride, 3 mM MgCl 2 , 0.1 mM CaCl 2 , 1.1 mM EGTA, 10 mM HEPES, and 5 mM glucose, pH 7.3 with HCl; Solution B, high ATP solution containing 100 mM Na 2 ATP, 5 mM KCl, 1 mM MgSO 4 , 0.1 mM CaCl 2 , 1.1 mM EGTA, and 10 mM HEPES, pH 7.3 or 7.1 with NaOH for plasma and nuclear membrane patches, respectively; Solution C, high PAPS solution containing 100 mM Li 4 PAPS, 5 mM KCl, 1 mM MgSO 4 , 0.1 mM CaCl 2 , 1.1 mM EGTA, and 10 mM HEPES, pH 7.3 with NaOH. Unless otherwise specified, PKA (180 nM) and MgATP (1 mM) were added to or included in the bath or pipette solutions for excised plasma membrane or nuclear membrane patches, respectively. Thus, MgATP 2concentration was always ϳ2 mM. In the ATP solutions, the extra 99 mM ATP is mostly ATP 4because it was added as Na ϩ or Tris ϩ salts. Therefore, the free Mg 2ϩ concentration was always Յ1 M. Thus, neither Mg 2ϩ nor MgATP 2were significant variables in our experiments. The bath was continuously perfused until PKA was added to it. The perfusion was halted during these periods because of the high cost of PKA and the large quantities required for perfusion. Cell-attached patches were stimulated by a mixture designed to elevate intracelluar levels of cAMP containing forskolin (5 M), 3-isobutyl-1-methylxanthine (0.1 mM), and the cAMP analog chlorophenylthio-cAMP (0.1 mM) applied directly to the bath to achieve the desired final concentrations. DIDS was added to the pipette or bath for plasma membrane or nuclear membrane patches, respectively.

RESULTS AND DISCUSSION
Stably transfected CHO cells were used to examine whether plasma membrane-localized as well as ER-localized CFTR are conductive to ATP. We first reconfirmed (39) the presence of CFTR Cl Ϫ channels in the plasma membrane of CFTR-expressing CHO cells. With Cl Ϫ solutions in the bath and pipette, elevation of intracellular cAMP levels activated CFTR Cl Ϫ channel activities in cell-attached patches (eight of nine patches) with familiar properties (39, 41, 42) (data not shown). In excised inside-out patches with the catalytic subunit of PKA and MgATP in the bath, 6.1 Ϯ 0.7 pS linear Cl Ϫ channels were observed (16 of 18 patches) with typical CFTR characteristics (Fig. 1A).
We used the inside-out excised patch configuration to determine whether membranes from these cells exhibited ATP conductances. With 100 mM ATP solutions in the bath and pipette, channel activities were observed with features similar to those of CFTR Cl Ϫ channels ( A 10-fold ATP gradient shifted the reversal potential from 0 mV to approximately Ϫ12 mV, indicating permeability to ATP (Fig. 1C). Of note, this reversal potential is close to that expected for permeation by a quadrivalent anion, i.e. ATP 4Ϫ (RT/4F ln C o /C i ϭ ϳ15 mV), the overwhelmingly dominant ionic species in these experiments. Cl Ϫ was eliminated as the current carrier because similar channels (4.6 Ϯ 0.2 pS) were present (10 of 11 patches; Fig. 3A) in Tris 2 ATP and Na 2 ATP solutions in which the Cl Ϫ concentration was reduced from the normal 5.2 mM (Figs. 1B and 2) to only 0.7 mM (Fig. 3). Na ϩ was also excluded as the current carrier because in inside-out patches under symmetrical conditions, 4.6 Ϯ 0.1 pS channels were present when Na 2 ATP was replaced with Tris 2 ATP (eight of nine patches; Fig. 3B), and no channels were observed when Kinetic analysis for three patches that contained single channels (data not shown) revealed P o ϭ 0.6, o1 ϭ 16 ms (77% of openings), o2 ϭ 1150 ms (23%), c1 ϭ 18 ms (56%), and c2 ϭ 455 ms (44%) at ϩ80 mV. B, ATP channels from inside-out patch in symmetrical ATP solution (Solution B) with PKA (180 nM) in the bath. In single-channel patches (n ϭ 4) at positive potentials (ϩ80, ϩ90, and ϩ100 mV), P o ϭ 0.4, o1 ϭ 17 ms (28%), o2 ϭ 247 ms (72%), c1 ϭ 7 ms (81%), and c2 ϭ 1350 ms (19%). This figure is representative of 13 experiments. C, I/V relationship for CFTR Cl -(triangles) and ATP (circles) channels under symmetrical (q) and asymmetrical (E) conditions. Slope conductance for symmetrical Cl Ϫ and ATP solutions was 6.1 Ϯ 0.7 (n ϭ 3 patches) and 4.5 Ϯ 0.3 pS (n ϭ 3 patches), respectively. In asymmetrical conditions the pipette solution was Solution B, whereas bath contained 10 mM Na 2 ATP, 5 mM KCl, 1 mM MgSO 4 , 0.1 mM CaCl 2 , 1.1 mM EGTA, 10 mM HEPES, and 200 mM glucose, pH ϭ 7.3 with NaOH (n ϭ 6 patches). Na 2 ATP was replaced with sodium gluconate (zero of four patches; Fig. 3C). Importantly, no ATP channels were observed in control cells (0 of 10 patches; Fig. 2D).
In inside-out plasma membrane patches with Cl Ϫ and ATP in the pipette and bath solutions, respectively, currents at ϩ70 mV had slope conductances of 6 -8 pS, whereas at Ϫ70 mV they were 4 -5 pS (6 of 6 patches; Fig. 4), consistent with the currents being carried by Cl Ϫ and ATP, respectively. Replacement of bath ATP by perfusion with 120 mM NaCl solution restored 6 -8 pS activities at Ϫ70 mV (Fig. 4A). Under reversed bi-ionic conditions in inside-out patches (i.e. pipette, ATP; bath, Cl Ϫ ), 6 -8 pS Cl Ϫ currents were observed at Ϫ70 mV, whereas 4 -5 pS ATP currents were present at ϩ70 mV (8 of 9 patches; data not shown).
We previously demonstrated by patch-clamp of the outer membrane of nuclei isolated from CFTR-expressing CHO cells that ER-localized CFTR was functional as a Cl Ϫ channel, with properties similar to those exhibited when it resided in the plasma membrane (38). Using a similar approach, we evaluated the ATP conductance of the ER, using nuclei isolated from the same CHO cells. With PKA and MgATP in the pipette (because the cytosolic face of the membrane patch faces into the pipette in this configuration), PKA-dependent channels with linear I/V relation and 4.6 Ϯ 0.5 pS conductance were recorded in symmetrical ATP solutions (33 of 35 patches; Fig. 5). Other properties of the ATP channels in the nuclear/ER membrane were also similar to those observed for wild-type CFTR in the plasma membrane (Fig. 5).
Our studies suggest that CFTR is associated with a regulated ATP channel in the plasma membrane and ER, and by extension in membranes in intermediate compartments including Golgi. The single ATP channel conductance, I/V relation, PKA sensitivity, pharmacological sensitivities, and dependence on expression of CFTR we have observed in the present study are similar to those reported previously (33,34) in different cell types, suggesting that CFTR is closely associated with or is in fact an ATP conductance. We do not think that the Cl Ϫ conduction pathway in CFTR is the ATP conduction pathway, although the resemblance of the biophysical, kinetic, and pharamacological properties of the ATP conductance to those of the CFTR Cl Ϫ conductance is striking and suggests that gating of the ATP permeability is strongly tied to CFTR gating. The reasons why some investigators have observed nucleotide conductances associated with CFTR (33, 34) although others have not (35)(36)(37) are not addressed by our study, and our results provide no insights into this issue. We would point out, however, that we have also observed CFTR Cl Ϫ channels that were not associated with ATP conductances in a different CHO cell clone, 2 suggesting that coupling of the two permeabilities may be cell-type dependent. The clear channel activities we observe will enable us in future studies to define variables that might regulate the manifestation of the ATP conductances and therefore resolve these conflicting observations.
The presence of a CFTR-associated ATP conductance in intracellular membranes raises the possibility that altered ATP permeability in intracellular compartments in CF may contribute to disease pathogenesis. ATP-translocase activity has been detected in ER and Golgi membranes (44), and ATP binding to ER proteins, ATP-dependence of the structure or activities of ER-resident proteins, and lumenal ATP requirement for protein translocation in the ER (45)(46)(47) suggest that intralumenal ATP content is physiologically regulated. ATP binding by ERlocalized molecular chaperones, including BiP and calnexin, is necessary for ATPase activity, structural changes, and oligomerization required for chaperone functions (45)(46)(47). These observations may have significance for CF because CFTR and ⌬F508-CFTR both associate with calnexin in the ER membrane (40), and the trafficking defect of ⌬F508-CFTR is likely a chaperone-assisted process (12). If ATP permeability through ER-or Golgi-localized CFTR affected intralumenal ATP concentrations, mutant CFTRs could conceivably modify lumenal ATP concentrations and consequently affect protein processing. However, measurements of lumenal ATP concentrations will be required to test this hypothesis.
It was proposed that CFTR may normally provide a Cl Ϫ conductance in Golgi membranes to facilitate lumenal acidification by a proton pump and that lack of such activity in CF, by compromising Golgi pH, might account for altered glycoprotein processing (26). Nevertheless, a direct test of this model failed to provide support for it (48). Nor can altered protein sulfation 2 M. Sugita and J. K. Foskett, unpublished results. in CF airway epithelial cells be accounted for by changes in plasma membrane sulfate transport or cell content of inorganic sulfate (31,49). Intracellular sulfation reactions utilize sulfate that has been "activated" by its incorporation into PAPS (50). PAPS is synthesized in the cytoplasm by a reaction involving ATP and sulfate, catalyzed by a bifunctional ATP-sulfurylaseadenosine 5Ј-phosphosulfate kinase, and transported into the Golgi lumen by a Golgi-specific saturable process that is temperature-dependent, independent of pH gradients or ATP hydrolysis, and inhibitable by stilbenes (51,52). Once transported into the Golgi lumen, PAPS donates sulfate to sulfotransferases for protein sulfation. PAPS is an intermediate for synthesis of all naturally occurring sulfated compounds, including sulfated glycoproteins, fibronectins, sulfatides, and glycosaminoglycans (51,52). We have considered whether PAPS, as an adenine nucleotide, might also be permeable through CFTR.
With high PAPS on both sides of excised plasma membrane patches from CFTR-expressing CHO cells, channels were observed with essentially identical biophysical and pharmacological properties as the ATP channels observed in the plasma membrane and ER (Fig. 6). The channels had a linear I/V relation, had a 4.7 Ϯ 0.1-pS conductance (Fig. 6C), and were DIDS-insensitive (eight of eight patches; Fig. 7B). They were observed in the presence of PKA (26 of 28 patches; Fig. 7A) but were not observed in its absence (0 of 7 patches; Fig. 7A). The channels were PAPS-selective because a 10-fold PAPS gradient (100 mM PAPS ext /10 mM PAPS cyt ) shifted the reversal potential from 0 mV to approximately Ϫ15 mV (6 of 7 patches; Fig. 6, B and C) (the expected Nernstean potential is ϳϪ15 mV because PAPS, like the ATP in these experiments, has a valence of Ϫ4). Li 1ϩ was excluded as a current carrier because no channel activities were observed when currents were recorded in insideout patches in symmetrical Li 2 SO 4 solution (Solution C with 150 mM Li 2 SO 4 replacing Li 4 PAPS) (0 of 13 patches; data not shown). In control cells, no activities were observed in insideout patches in Solution C in the presence or absence of PKA (zero of seven patches; data not shown).
By demonstrating that PAPS permeability is associated with CFTR, our results raise the possibility that the concentration of PAPS in the Golgi lumen may be regulated in part by CFTR. In a simple model, activity of CFTR in this intracellular compartment would constitute a PAPS "leak" in parallel with the PAPS pump and would therefore tend to lower the PAPS concentration in the Golgi lumen. Lack of CFTR in this compartment in CF would shift the balance of these activities in favor of the pump, resulting in a higher Golgi lumenal PAPS concentration. Because the PAPS concentration in the Golgi lumen is likely rate-limiting for sulfation reactions (50,53,54), this model predicts hypersulfation of proteins in CF, consistent with observations. The physiological consequences of hypersulfation may include altered visco-elastic properties of airway secre-tions and diminished glycoprotein sialylation because sialylation and sulfation reactions are competitive on some glycoprotein residues (55,56). Importantly, because decreased sialylation has been implicated in the adherence of Pseudomonas to airway epithelial cells (14,28,57), the observed PAPS permeability of CFTR may therefore contribute to the heretofore unexplained propensity of Pseudomonas to colonize the lungs of CF patients (57). This PAPS hypothesis therefore links CFTR to glycoprotein processing and possibly to mucus viscosity and bacterial adherence. Because bacterial infection accounts for up to 90% of morbidity and mortality in patients with CF (58), detailed investigations of the PAPS hypothesis are now warranted to determine its relevance for understanding pathophysiology observed in CF.