| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
MEMBRANE TRANSPORTERS, ION CHANNELS, AND PUMPS
Department of Cell and Developmental Biology, University of Colorado Health Sciences Center, Aurora, Colorado
Submitted 1 December 2004 ; accepted in final form 2 September 2005
 |
ABSTRACT |
|---|
 TOP ABSTRACT MATERIALS AND METHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
-32P]ATP into sealed Golgi vesicles that maintain in situ orientation was consistent with movement through the GOLACs because it exhibited little temperature dependence and was saturated with an apparent Km = 25 µM. Finally, after transport of [
-32P]ATP, a protease-protection assay demonstrated that proteins are phosphorylated within the Golgi lumen, and after SDS-PAGE, the proteins in the phosphorylated bands were identified by mass spectrometry. GOLAC conductances, [-32P]ATP transport, and protein phosphorylation have identical pharmacological profiles. We conclude that the GOLACs play dual roles in the Golgi complex, providing pathways for Cl– and ATP influx into the Golgi lumen. Golgi complex; Cl channel; mass spectrometry; phosphorylation
-cells (23). In addition, ATP in the Golgi lumen may provide the ATP present in exocytotic vesicles that mediate extracellular ATP secretion for activation of purinergic receptors (45). Thus Golgi luminal ATP is required for diverse functions. The mechanism of ATP transport into the Golgi lumen is not well established. ATP requires ion channels or transporters to mediate its flux across Golgi membranes because it is an anion at physiological pH, with a valence equal to –2 when complexed with Mg2+ or Ca2+ and equal to –3 and –4 as free ATP. The best-characterized ATP transporters are the mitochondrial (3) and bacterial ATP/ADP exchangers (43), which couple ADP export to ATP import. ATP movement through anion channels is less well characterized and often controversial. Several types of anion channels (26–29) have been suggested to provide pathways for ATP movement across membranes. These pathways include two ATP-binding cassette transporters that form anion channels, the multidrug resistance gene product (mdr1, Ref. 1) and the cystic fibrosis transmembrane conductance regulator (7, 26), although the cystic fibrosis transmembrane conductance regulator remains controversial (25, 36). In addition, the plasma membrane volume-activated osmolyte anion channel, which permits ATP release from cells after swelling (28), and a sarcoplasmic reticulum anion channel (17) conduct ATP. Furthermore, a patch-clamp study of heterologous expression of the ADP/ATP exchanger of Neurospora crassa has demonstrated large-conductance anion channel activity (6).
ATP transport into the Golgi lumen has been studied at the biochemical level by Hirschberg and colleagues (8, 9), who interpreted their data from the perspective of a classic transporter. On the basis of the evidence that large-conductance anion channels can mediate ATP flux, we explored the possibility that Golgi anion channels provide ATP to the Golgi lumen in addition to or in place of a classic ATP transporter. We previously characterized two Cl–-conducting ion channels in Golgi membranes, the Golgi anion channels GOLAC-1 and GOLAC-2, by electrophysiological methods (20, 39). We predicted that these channels are important in providing charge neutrality for the organelle's v-type H+-ATPase, which generates an acidic lumen (2, 22). Both channels have five subconductance levels (L1–L5) that are approximately one-fifth increments of the fully open level. The GOLACs are open >95% of the time in symmetrical 150 mM KCl, suggesting that they can function as constitutively open pathways for the flux of both small and large organic and inorganic anions across Golgi membranes. The channels are inhibited by 4,4'-diisothiocyano-2,2'-disulfonic acid stilbene (DIDS), as is the Golgi ATP transporter (9). The GOLACs discriminate poorly between anion species, conducting those with diameters <6.2 Å. Molecular modeling of the effective diameter of MgATP2– resulted in a value of 5.9–6.1 Å, which was slightly larger than the predicted diameter, 5.7–5.8 Å of ATP4– (7). Others have reported similar diameters (27, 29). Thus we reasoned that the ATP molecule, either as MgATP2– or ATP4–, has the potential to be permeable through the GOLACs.
Three different protocols were used to test the hypothesis that the GOLACs function as pathways for ATP entry to the Golgi lumen. First, electrophysiological data collected from single GOLAC channels incorporated into planar lipid bilayers showed that they conduct ATP. Second, biochemical transport assays using [-32P]ATP showed that transport into the Golgi lumen had little temperature dependence and was sensitive to channel blockers that inhibit the GOLACs, but was insensitive to five structurally distinct transport blockers that do not affect the GOLACs. Finally, proteomic techniques showed phosphorylated proteins in the lumen of the Golgi after transport of [-32P]ATP, and this phosphorylation had the same pharmacological sensitivity as the channel currents and ATP transport. Data from each of these three experimental approaches all supported the hypothesis that the Golgi anion channels function in ATP influx into the Golgi lumen.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
-32P]ATP (specific activity 6,000 Ci/mmol; product no. ARP101, American Radiolabeled Chemicals), [-32P]ATP (specific activity 6,000 Ci/mmol; Amersham), and [3H]-inulin (101 mCi/g, American Radiolabeled Chemicals). Preparation of an enriched Golgi fraction. An isolated stacked Golgi fraction (SGF1) from rat liver enriched 300–400 times for trans- and cis-Golgi markers and 400–700 times for medial Golgi markers over the postnuclear supernatant was prepared as previously described (38). For use in electrophysiological recordings of GOLACs, proteins in transit through the Golgi, including those destined for the plasma membrane, were first cleared before SGF1 isolation by in vivo treatment with the protein synthesis inhibitor cyclohexamide (CHX). This protocol allowed enrichment of and focus on anion channels endogenous to the Golgi (20, 39). ATP transport assays were performed on both CHX-treated and control (not CHX treated) SGF1, and no differences in transport properties were observed.
Electrophysiology. The electrophysiological properties of endogenous GOLACs were studied after incorporation into planar lipid bilayer membranes as previously described (20, 39). The Cl–-containing recording solution comprised (in mM) 150 KCl, 1 EGTA, 1.2 CaCl2, 1 MgCl2, and 10 MOPS (pH 7.0 with CsOH). Recording solutions containing ATP were made by replacing the 150 mM KCl with 75 mM K2ATP and CaCl2 and MgCl2 with either 1.2 or 5 mM Ca(OH)2, at pH 7. The chamber was washed with 10 vol (15 ml) to exchange the Cl– solution for the ATP-containing solution. Single-channel conductances were calculated using Ohm's law (G = i/V), where i is the maximal single-channel current and V is the ionic driving force, V = Vm – Vr, where Vm is the membrane potential and Vr is the potential where the current is zero and was corrected for liquid junction potentials (see below). V was applied for 20 s in 25-mV steps from –75 to +75 mV. Thus, conductances reported at ±75 mV were calculated for Vm = Vr ± 75 mV. Potentials are reported with respect to the trans-chamber (the ground).
The relative permeabilities (P) of ATP4–, MgATP2–, and Cl– under bi-ionic conditions, with Cl– in the trans chamber and either K2ATP or MgATP in the cis chamber, were determined using the Goldman-Hodgkin-Katz constant field equation (18) of the form
 | (1) |
represents the sum of all permeant ions, j, in the solutions, I is current, zj is valence of ion j, aj is the activity of the ion in the cis or trans-compartment, Pj is the permeability of the ion, and R, T, and F have the usual thermodynamic definitions. The valence of free ATP was assumed to be –4, and the valence of MgATP was –2 for these calculations (see below). Calculations were based on the change in Vr of single GOLAC channels after replacement of the 150 mM KCl solution in the cis- chamber with 75 mM K2ATP or MgATP. Vr was adjusted for errors due to liquid junction potentials. For replacement of 150 mM KCl with either 75 K2ATP or 75 MgATP, junction potentials were determined by measuring the voltage offset across a planar lipid bilayer that was "channel free" (i.e., no proteins added to the chamber) with the ATP-containing solution on the cis side and the KCl solution on the trans-side. These values were –14 mV for K2ATP:KCl and –11 mV for MgATP:KCl, and were notably larger than those predicted (–5.5 mV for K2ATP and –0.4 mV for MgATP) using JPCalc software (Dr. Peter Barry, University of New South Wales, Sydney, Australia) provided with the pCLAMP software. Osmolarity of all solutions was routinely monitored, and sucrose was added in some cases to ensure that changes in channel behavior were not due to an inherent osmosensitivity. However, neither GOLAC-1 nor GOLAC-2 were significantly affected by osmolarity in the range tested, 280 to 650 mosmol/l (data not shown).
Determination of ionic activities.
All calculations used ionic activities instead of concentrations to account for the interactions of ions in solution, which become increasingly important in high ionic strength solutions for ions with a valence >1. Ionic activity () is dependent on the total ionic strength (µ), which was calculated as µ = 0.5 zi2ci, where zi is the valence and ci is the concentration of the ion (i) in the solution. For solutions with µ < 0.5 M, the extended Debye-Hückel equation of the form
 | (2) |
, where B = 1.5 and is an empirical term that incorporates both the diffuse ionic cloud around the ions and the intermolecular distance between anion cation pairs, such as ATP4– and K+ (5), and |z + z–| represents the absolute value of the product of the respective valences of the cation-anion pair; for example,|z+ z–| for K2ATP is 4. MgATP or K2ATP will dissociate in our solution as mixed salts of free ATP4– + 2 K+ and Mg ATP2–. Furthermore, at pH 7, the free ATP will be present as either ATP4– or ATP3–, with the majority (
75%) as ATP4–. The valence of MgATP was taken as –2, and that of the free acid form of ATP as –4, for all calculations (26). To determine the relative activities of MgATP2– and ATP4– in solution, we used the Internet-based program MaxChelator (http://www.stanford.edu/cpatton/webmaxc/webmaxcE.htm) to first determine the concentrations of each ionic species in the K2ATP- and MgATP-containing solutions. These were then converted into ionic activities by using Eq. 2. In the 75 mM K2ATP and MgATP solutions, = 0.18 and 0.56, respectively. Concentration, free concentration (i.e., nonchelated), and corresponding activities used for all calculations are provided in the online data supplement (see Supplemental Table 1; http://ajp-cell.physiology.org/dgi/content/full/00585.2004/DC1). For simplicity, MgATP2– and CaATP2– were considered equivalent.
|
-32P]ATP to cold K2ATP was kept constant at 0.11 µCi per 100 µM K2ATP and was added at the appropriate final activity of ATP as indicated. This ATP-containing solution also contained 0.11 µCi of [3H]inulin to allow subtraction of nontransported radioactivity associated with the outside of sealed Golgi vesicles as previously reported (9). On average, 0.3 ± 0.1 pmol [3H]inulin/mg protein was associated with the outside of the vesicles (i.e., <0.01% of the added inulin remained after the vesicles were washed). With the exception of experiments to determine the temperature sensitivity of ATP transport, all experiments were performed at 25°C for 3 min and stopped by dilution of the total reaction (100 µl) into 10 ml of 150 mM KCl, 10 mM Tris·HCl (pH 7.5) and 1 mM MgCl2. All calculations and plots used ionic activities of free ATP, which were calculated as described above. The diluted reaction mixture was filtered through 0.45 µm nitrocellulose filters and washed three times with 10 ml of the KCl buffer used to stop the reaction. The filters were immersed in 5 ml of scintillation fluid, and radioactivity was counted using a scintillation counter (model LS 1801, Beckman Instruments). Background was determined by adding equal amounts of the radiolabeled ATP/inulin solution to the filters in the absence of SGF1, and further treated in parallel to the transport reactions. Background radioactivity in counts per minute (CPM) was always <1% of the transported amount, and this number was subtracted from the transported CPM. Total [-32P]ATP transported in units of pmol/mg protein/3 min was calculated as the following: CPM membrane/(specific activity of [-32P]ATP per pmol)(mg SGF1)(time in min). In some experiments, 1 ml of the SGF1 fraction was diluted twice in 150 mM KH2PO4/K2HPO4, 10 mM MOPS, and 100 µM Ca[OH]2 at pH 7.2, which is the same buffer solution minus sucrose that was used to isolate the SGF1 fraction (38), centrifuged at 13,000 g for 15 min at 4°C in a Beckman desktop centrifuge, and then resuspended in 1 ml of 250 mM sucrose, 10 mM MOPS, and 100 µM Ca(OH)2. This step was performed to remove anions, especially chloride, present in the isolation buffer.
Anion channel blockers were incubated with SGF1 at the appropriate final concentration for 10 min to 1 h on ice before the sample was warmed to 25°C for transport assays. Anion channel blockers were added either to the anion-free solution before the addition of 100 µM ATP, including [-32P]ATP and [3H]inulin as described above, and 150 mM KCl or directly to the SGF1 fraction in isolation buffer. With the exception of 9-AC, all blockers were similarly effective under either condition (isolation buffer or anion-free buffer). Block by 9-AC required removal of small anions. It has been reported that anion channel block by 9-AC is both voltage and chloride dependent (11, 21), and we observed that a net anion flux was required for 9-AC to be an effective blocker in both electrophysiological (i.e., applied voltage was required) and ATP transport experiments (Cl– flux-dependent block).
Quantification of protein phosphorylation by densitometry.
Protein phosphorylation was determined by autoradiography after transport of [-32P]ATP. The assay consisted of 100 µg of SGF1 resuspended in the 150 mM KCl solution used for electrophysiology recordings, and was incubated for 5 min at 30°C with 5–10 µCi of [-32P]ATP and enough cold K2ATP to bring the final ATP ionic activity to 11.4 µM. The free concentration of ATP4– is 3.8 mM under these conditions, which is in the range of the measured cytosolic concentration of ATP (15). After the initial incubation with ATP, the sample was divided into three equal portions: the first was the control, the second contained 200 µg/ml proteinase K, and the third contained 200 µg/ml proteinase K plus 1–2% Triton X-100 (TX-100) and was incubated on ice for 1 h. The reactions were stopped by the addition of an SDS sample buffer containing 2 mM PMSF to yield a final protein concentration of 1 µg/µl. SGF1 vesicle orientation was confirmed after protease protection by immunoblotting with monoclonal antibodies that recognize epitopes on the luminal domains of Golgi resident transmembrane proteins.
PAGE was carried out using 10 µg of [-32P]ATP-treated CHX-SGF1 per lane using a 10% acrylamide gel with a 4% stacker and labeled bands were visualized by autoradiography after exposure of dried gels to a phosphorimaging screen (Kodak) for 12–24 h. Autoradiograms were obtained at a pixel size of 50–100 µm with the use of an imaging system (Typhoon 8200; Amersham), stored as digital images, and quantified using National Institutes of Health Image J software, as follows. Background 32P intensity was determined for a small region of each lane that did not contain radiolabeled bands. The background was subtracted from the total density of the lane, determined for a window drawn around that lane. Thus each lane had its background subtracted. The fraction of protease-protected 32P-labeled proteins was calculated from the following equation: fractional 32P density = (B – C)/(A – C), where A is the mean intensity of the control lane (without protease), B is the mean intensity of the protease-treated lane, and C is the intensity of the protease + detergent lane. The term C was included to remove the contribution of signal due to incomplete action of the protease. Adjacent lanes were compared in groups of 3 (i.e., control or blocker treated, +protease, and protease + detergent).
In-gel digestion of phosphoproteins and identification by mass spectrometry. [32P]-Labeled bands from Coomassie-stained, dried gels were identified by autoradiography, excised, in-gel digested, and analyzed by tandem mass spectrometry (MS/MS) as previously described (44). MS/MS spectra were analyzed by SEQUEST software using a rat, mouse, and human database. Proteins were identified using DTASelect (37).
|
RESULTS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
|
|
Analysis of the blocking effect of K2ATP on the channel substrates indicated that block of L2 through L5 saturated at 23–37% inhibition (Table 1), with apparent Km values ranging from 16 to 115 µM activities of ATP4–. Both channels had similar sensitivity to ATP4–, and no significant differences in the Km of sublevels L2–L5 were detected. The channels were rarely found in the L0 and L1 states, and therefore, those states could not be analyzed. The increased occurrence of rapid flickering could not be quantified by measuring the mean open times for each subconductance state because transitions <1.5 ms must be excluded from the analysis because of restrictions imposed by the low-pass filter and because at higher activities of ATP4– flicker was faster than the minimal detectable event duration (i.e., Fig. 1A). Block by ATP also occurred if ATP was added to the trans-chamber of the bilayer (data not shown), which is equivalent to the lumenal side of the membrane (see below).
The data shown in Fig. 1 are consistent with a high-affinity, open channel block of GOLAC chloride currents by ATP4– (16). However, ATP is a chelator of divalent metals such as Ca2+, and the [Ca2+] decreased with addition of increasing amounts of K2ATP from 200 µM in the control condition to 5 µM in the solution containing 4.8 mM ATP4– activity. Thus, to rule out the possibility of an indirect action of ATP4– via Ca2+ chelation, we recorded channel activity under control conditions and with nominally Ca2+-free (zero added Ca2+ + 1 mM EGTA) solutions. Similar to our previous report (20), single-channel conductances and gating of both GOLAC-1 and GOLAC-2 were independent of free Ca2+ (data not shown), suggesting that Cl– current block by increasing ATP4– is a consequence of direct competition between the anions.
We next tested whether channel inhibition by ATP was dependent on the charge of the ATP by using cation-ATP complexes that have different valences (i.e., 2K+ + ATP4– or MgATP2–). MgATP was added to the cis-chamber at increasing ionic activities. Cl– currents were blocked by MgATP2– in a manner similar to block by K2ATP, but the threshold concentration for block required substantially greater ionic activity compared with that for K2ATP. For GOLAC-2, inhibition of L4 by ATP4– occurred at micromolar K2ATP activities (Table 1). Ideally, we would have liked to compare the Km of inhibition for K2ATP and MgATP. However, this was not possible because block by MgATP did not occur until the activity reached >30 mM and the generation of a dose-response curve would have required unacceptably high ionic strength solutions. However, this difference suggests that free ATP4– rather than the chelated form, produced the block because only a small fraction of MgATP will fully dissociate into ATP4–.
The reversibility of the effects of ATP on the channels (the wash in Fig. 1, A–C), the rapid flicker of the blocking events, and the apparent channel block by ATP4–, as opposed to MgATP2–, all argue against block of GOLAC Cl– currents by a mechanism involving ATP hydrolysis. This was confirmed by addition of the nonhydrolyzable ATP analog, ATP--S (0.5 mM ionic activity; 3 mM concentration) to the cis-chamber (Fig. 2), which reduced the single-channel conductance of GOLAC-2 to 62 ± 2% (n = 3) of the control value (at Vm = –75 mV). GOLAC-1 conductance was reduced to 80 ± 8% (n = 3; data not shown). ATP--S also produced rapid, flickery blocking events (Fig. 2A) similar to those observed with ATP (Fig. 1). In summary, the data in Figs. 1 and 2 suggest that Cl– currents were blocked by a direct, saturable binding of ATP4– to both GOLACs, possibly at the same site as Cl–.
|
Two types of information are obtained from these experiments: 1) the relative permeabilities of ATP4– and MgATP2– to each other and to Cl–, which were calculated using the shift in reversal potential and ionic activities of each anion under the bi-ionic conditions (see Supplementary Table 1), and 2) the conductance of the channels for ATP4– and MgATP2–, which is determined under symmetrical conditions and is an electrical measure of the movement of the ATP anions through the pore. If ATP is permeable and conducted by the GOLACs, single-channel currents will be observed under both bi-ionic and symmetrical ATP conditions. The relative permeabilities of ATP4– and Ca/MgATP2– to Cl– and to each other were calculated using the Goldman-Hodgkin-Katz constant field equation (see MATERIALS AND METHODS). Vr was corrected for errors due to liquid junction potentials using experimentally measured junction potentials (see MATERIALS AND METHODS). For GOLAC-1, the mean Vr was –25.5 ± 3.1 mV (n = 5) with K2ATP cis and KCl trans and was –43.4 ± 4.2 mV (n = 5) with MgATP cis and KCl trans. These values were significantly different (P < 0.05; Student's unpaired t-test). For GOLAC-2, Vr with K2ATP cis and KCl trans (–36.3 ± 1.6; n = 6) was significantly different (P < 0.05) from that in cis MgATP and trans- KCl (–52.9 ± 3.6; n = 5). The Goldman-Hodgkin-Katz constant field equation was used to calculate relative permeabilities of ATP4–, MgATP2–, and Cl– (Table 2). Note that both channels have the relative permeability sequence of Cl– > ATP4– > MgATP2–, although for GOLAC-2, both ATP anions had similar permeability, which may be related to the almost twofold larger conductance of GOLAC-2 vs. GOLAC-1.
|
|
The conductance changes suggesting that the pores are shared by the different anions was evident only for GOLAC-2 and were most dramatic with cis-K2ATP and trans- Cl–. Under these conditions, the ATP4– conductance increased to 506 ± 31 pS, which is larger than the conductance of Cl– or ATP4– under all ionic conditions tested (Table 3). The Cl– conductance for cis to trans movement under the opposite bi-ionic conditions (Cl– cis and ATP trans) was about one-half of that observed for ATP movement in the same direction (Fig. 3; Table 3). The conductance data in Fig. 3 and Table 3 clearly indicate that Cl– and ATP interact when moving in opposite directions through the pore of GOLAC-2. In some instances, currents are augmented and in others they are attenuated. This interaction, together with the data in Fig. 1, provides strong evidence that the Cl– and ATP currents utilize the same channel.
In of the measurements of ATP conductance described above, we used high concentrations (75 mM) of total ATP, where the activity of ATP4– in these solutions was 47 mM (See Supplementary Table 1 data supplement). Do the channels conduct ATP at more physiological concentrations similar to those used to block Cl– currents (Fig. 1)? To address this question, we first performed experiments with low concentrations of KCl that are typical of intracellular Cl concentrations. In symmetrical 10 mM KCl (7 mM activity), the gating of the channels (both GOLAC-1 and GOLAC-2) was drastically reduced, with only a few openings in the 20-s recording period, with the probability of being open changing from >95% to 2% (compare Fig. 4, A and B). We did not test the possibility that the reduced open probability was due to the unphysiologically low [K+] (10 mM). Although the cytoplasmic concentration of ATP has been reported in the range of 1–5 mM (15), the activity level is not known. However, at 10 mM total K2ATP (activity = 3 mM), we detected channel openings with a similar frequency and magnitude as those observed in 10 mM KCl (compare Fig. 4, B and C). As expected, increasing the activity of ATP4– also increased the single-channel conductance (Fig. 4D). In summary, the data presented above indicate that the GOLACs are ATP-permeable anion channels and that both Cl and ATP currents can be recorded at anion activities close to the physiological range.
|
|
|
-32P]ATP and [3H]inulin and the accumulation of 32P was determined by scintillation counting. ATP with an [-32P] label was used to avoid transfer of the radiolabel during phosphorylation reactions, which would bias the measurement of transport rates in favor of enhanced influx. [3H]Inulin was included because, as previously reported (8, 9), this molecule is not transported and allows for subtraction of radioactivity associated with the volume of the extravesicular space. Two additional controls were used to address the possibility that binding of [3H]inulin to the outside of vesicles differs from [-32P]ATP binding: 1) vesicles were first solubilized with 1% TX-100 before incubation with the radioactive solution, and 2) influx was blocked by DIDS, and as reported by Capasso et al. (9), this was taken as the amount of [-32P]ATP binding to the outside of Golgi vesicles.
If the ATP-permeable GOLACs are the pathway for ATP flux into the Golgi lumen, it follows that the time constant (
) of [-32P]ATP transport would increase by a factor of 1.7 with a 20°C temperature change. This is because conduction by ion channels is well predicted by the Q10 for diffusion (1.3; see Ref. 16). However, if a "classic" transport mechanism (i.e., an exchanger) is involved, at 4°C is expected to be 9x greater than that at 25°C because the Q10 for transporters, such as the mitochondrial ATP/ADP exchanger, is 3 (16, 24).
We determined at 4 and 25°C using a free ATP4– of 2 µM, which included 0.11 µCi [-32P]ATP. The low activity of ATP used allowed rates to be calculated because at higher concentrations saturation occurred before the first time point (10 s). Figure 5B plots [-32P]ATP accumulation vs. time at 4° and 25°C. Fit of the experimental data at 25°C with a single exponential gave a = 0.48 ± 0.07 min with a maximum transport capacity of 690 ± 27 pmol/mg of protein. These values were then used to compare at 4°C to that predicted by diffusional or enzymatic processes. At 4°C, [-32P]ATP inside vesicles was measured as = 0.57 ± 0.16 min, and saturated at 775 ± 52 pmol/mg. For a purely diffusional process, is predicted to equal 0.8 min (Fig. 5B, dashed line) and 4.7 min for an enzymatic process (Fig. 5B, dotted line). The important point from the data in Fig. 5B is that the predicted shift in influx is best described by a diffusional process and is consistent with the GOLACs mediating ATP influx into the Golgi.
The concentration dependence of [-32P]ATP influx at 25°C was determined and is plotted in Fig. 5C, which allowed comparison of our data to that of Hirschberg's group. Transport had an apparent Km of 25.3 ± 6.2 µM ATP activity and Vmax = 1,858 ± 202 pmol/mg of protein for a 3-min incubation time. As further controls, we determined that influx of [-32P]ATP was linear between 10 and 1,000 µg total SGF1 protein (we used 100 µg of SGF1 for all assays) and was blocked by a 1,000-fold excess of cold ATP. Accumulation by the Golgi preparation was prevented by preincubation of the SGF1 with 1% TX-100 (data not shown). DIDS inhibition (Fig. 6A) permitted us to correct for nonspecific binding to the outside surface of the Golgi vesicles, as previously described by Capasso et al. (9). This binding was 10–15% of the amount of ATP accumulated. In summary, the collective data demonstrate that the GOLACs are highly permeable to ATP and that ATP influx into the Golgi lumen has the temperature dependence expected for facilitated diffusion, rather than that for an enzymatic process.
|
-32P]ATP influx have the same pharmacology.
To obtain further support that the GOLACs are ATP pathways in Golgi membranes, anion channel blockers were tested for inhibition of GOLAC single-channel currents and [-32P]ATP influx. The concentration dependence of inhibition of these two processes should be similar. It is well known that anion channel pharmacology is highly nonspecific because different anion channel families are blocked by similar drugs, and there is a marked lack of high-affinity specific blockers. Therefore, we compared the pharmacological profile of inhibition of GOLACs' ATP and Cl– conductance in the lipid bilayer to predict each anion channel blocker's effect (or lack thereof) on [-32P]ATP influx. Six structurally distinct anion channel blockers, DIDS, 9-AC, IAA-94, NPPB, niflumic acid, and glibenclamide, in addition to the mitochondrial ADP/ATP exchanger inhibitor ATR, were tested.
We previously reported that both GOLACs are blocked by DIDS, with a Kd of 200 µM (Fig. 6A, inset; Refs. 20, 39). DIDS completely inhibited [-32P]ATP influx with a Kd = 158 ± 23 µM (Fig. 6A), which is similar to the values determined for direct channel block. The anion channel blocker, 9-AC, also inhibited both channels with a Kd for inhibition for GOLAC-1 of 1.8 ± 0.3 mM (n = 6) and for GOLAC-2 of 1.5 ± 0.4 mM (n = 6; Fig. 6B, inset). Block by 9-AC was voltage dependent and resulted in increased channel flicker (Fig. 6C), consistent with an open channel-blocking mechanism, similar to what has been described for the ClC-1 chloride channel (11). As shown in Fig. 6B, 9-AC (0.1–10 mM) inhibited [-32P]ATP influx with a Kd = 1.8 ± 0.7 mM (n = 4), which is similar to the values for direct channel inhibition. Anion channel blockers used at concentrations known to completely block other anion channels (composition, in mM: 0.1 glibenclamide, 0.1 IAA-94, 0.1 niflumic acid, and 0.1 NPPB) failed to affect the single-channel currents of GOLACs (data not shown) and, as predicted, also did not inhibit [-32P]ATP influx (Fig. 7A). The inhibitor of the mitochondrial ATP/ADP exchanger, ATR, at 5–50 µM, was also ineffective, demonstrating that the observed ATP influx was not due to contamination of the SGF1 preparation with mitochondria (see also Ref. 9).
|
|
-type H+-ATPase (13). To test this, the SGF1 fraction was incubated with 0.5 mM DIDS to block the GOLACs and with 1 µM valinomycin, a K+ ionophore, to maintain a zero Vm. DIDS remained effective at blocking [-32P]ATP transport when a zero Vm was maintained using valinomycin (Fig. 7B). Thus, the GOLACs and [-32P]ATP movement into sealed Golgi vesicles have the same pharmacology, and the block of [-32P]ATP accumulation is not caused by the development of a membrane potential.
Vesicular ATP is used in the phosphorylation of lumenal Golgi proteins.
To demonstrate the physiological relevance of the accumulated ATP, we assayed for phosphorylated (i.e., 32P labeled) proteins via SDS-PAGE and autoradiography. Proteins in radiolabeled bands were identified using mass spectrometry. Incubation of the SGF1 vesicles with [-32P]ATP resulted in protein phosphorylation (Fig. 8A). A protease-protection assay similar to that in Fig. 5A was used to determine the sidedness of the phosphorylated proteins. The data show that both protease-sensitive and -insensitive 32P labeling occurred (Fig. 8A). Quantification of the 32P signal using densitometry indicated that 1/3 of the label was associated with protease-protected proteins and were thus located inside sealed Golgi vesicles (Fig. 8B). Solubilization of the Golgi fraction with 1% TX-100 in the presence of protease almost completely eliminated the signal from 32P-labeled, protease-protected proteins (Fig. 8A, lane 3, and 8B, lane 4), confirming that protein phosphorylation occurred on both sides of the Golgi membrane. TX-100 alone had no effect on phosphorylation (Fig. 8B, lane 2). Phosphorylation of proteins in the SGF1 fraction was inhibited by DIDS (Fig. 9A) and 9-AC (Fig. 9B). As a further correlation, anion channel blockers that failed to affect the GOLAC currents and [-32P]ATP movement into the SGF1 vesicles (Figs. 6 and 7) were also ineffective at preventing phosphorylation of lumenal Golgi proteins (Fig. 9C).
Four labeled bands were excised from the gels, in-gel digested and prepared for analysis using MS/MS. Three of these were protease protected (Bands 1–3; Fig. 8A) and one was protease sensitive (Band 4; Fig. 8A). Proteins identified by at least five nonoverlapping peptides or from fewer peptides but found in at least three separate experiments are provided in the Supplementary Table 2. In each band more than one protein was identified, and for each, a known phosphorylated Golgi protein was present. A limitation of these data is that because the bands were from one-dimensional gels, multiple identified proteins were present in the 32P-containing bands. Of the identified proteins, 65% were known Golgi proteins such as sialyltransferase 4A, nucleobindin, and mannosidase-1 and -, 15% were known endoplasmic reticulum proteins, and the remaining 20% were either ubiquitous cellular proteins (i.e., HSP70) or secretory proteins. Only two proteins were found in the protease-sensitive band 4, clathrin heavy chain and MG-160. Clathrin heavy chain is known to be phosphorylated (14). MG-160 is unlikely to be the phosphorylated protein, as it is protease insensitive (see Fig. 5) and band 4 was protease sensitive (Fig. 5A). These data show that the ATP accumulated by the Golgi vesicles is used to phosphorylate lumenal proteins.
|
DISCUSSION |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
-32P]ATP has physiological relevance because protease-protected Golgi proteins were phosphorylated. Taken together, these data provide evidence for an anion channel-mediated pathway for ATP influx into the Golgi lumen. Golgi anion channels conduct ATP. In the presence of symmetrical 150 mM KCl, Cl– currents were inhibited by addition of low micromolar activities of K2ATP. The features of inhibition by K2ATP were characteristic of an open-channel blocking mechanism, and are consistent with ATP4– entering the pore and occluding Cl– conduction. These hallmark characteristics include: 1) voltage dependence of block, where currents are inhibited only at potentials at which ATP4– is driven into the pore and the stronger the driving force becomes, the greater the block, 2) the presence of the blocking anion induced a rapid flickering behavior and an apparent reduction in the single channel conductance, 3) chelating two of the negative charges of ATP, by using (Mg)ATP2– instead of (K2)ATP4– produced a weaker blocking effect, and 4) ATP4– blocks from both sides of the channel. These data indicate that ATP enters the pore of the channels and inhibits Cl– currents.
To distinguish between a blocking mechanism where ATP simply occludes the pore vs. co-conductance of ATP4– and Cl–, we replaced the Cl– with either K2ATP or MgATP and measured both ionic permeabilities and single-channel conductances. Both channels had ATP and Cl conductances that were similar in magnitude (Table 3). Both channels showed the following permeability sequence: Cl– > ATP4– 
MgATP2– with ATP and MgATP being 0.1 to 0.3 times more permeable than Cl. The approximately equal conductances of ATP4– and Cl– but reduced ATP permeability compared with Cl– may be a consequence of the fact that the valences are very different (flux of an ATP4– anion is equivalent to the flux of four Cl–) and could also be explained if Cl– experiences more resistance (i.e., stronger binding or higher friction) in the pore than ATP, reducing the relative rate of Cl– movement. Other ATP-permeable channels have been reported to have relative ATP to Cl– permeabilities on the order of 0.09 to 0.4 (10, 26, 28), which is similar to our values for the GOLACs. We found that the permeability of the GOLACs to free ATP4– was only about twofold greater than that of MgATP2–. This was unexpected because of the difference in charge and shape of the anions but has been reported for another anion channel (10, 29).
Could block by ATP occur by an allosteric mechanism through binding to an external domain of the channel? The voltage dependence of block (Fig. 1, B and C) implies that the binding site is within the membrane. Addition of ATP to the trans-side of the bilayer also produced block (data not shown), making ATP binding within the pore the simplest explanation for the block of Cl– currents. A parallel can be drawn between the action of ATP on the GOLACs and the action of calcium upon calcium channels. In the case of voltage-gated calcium channels, in the absence of calcium, the channels conduct monovalent cations such as sodium. Addition of micromolar concentrations of calcium blocks sodium current due to a high affinity calcium binding site in the pore, similar to the block of chloride currents by micromolar ATP (Fig. 1). At higher concentrations of calcium, the channel conducts calcium at high rates (31).
It might be argued that two distinct channels are present in the bilayer, one that conducts Cl– and another that conducts ATP. This scenario seems unlikely for the following reasons. First, when recordings were made with Cl– on one side and ATP on the other side of the bilayer, we did not observe new conductance levels. For this to occur, the ATP-conducting channel would have to initially be quiescent, and its opening would be synchronized precisely to the concomitant closure of the Cl– channel. This is especially unlikely in the asymmetrical conditions where both permeant anions would be present and each putative channel would be observed. Second, ATP and Cl– on opposite sides of the bilayer influence each other's conductance through the GOLACs, suggesting that the two anions move in opposite directions through the same pore. Third, ATP4– at low ionic activity directly reduced Cl– conductance of both GOLACs in a voltage-dependent manner, indicating competition for conduction. A similar argument has been used for voltage-dependent anion channels to explain conductance of multiple anion species (27). Fourth, the GOLACs show similar substate behavior for both ATP and Cl– (our unpublished observations). We conclude that GOLACs conduct both ATP and Cl–.
Like many anion channels, GOLACs have some permeability to cations (12, 16, 20, 39). Therefore, we considered the possibility that the currents attributed to movement of ATP were in fact cation currents. We present three arguments that the currents measured under symmetrical ATP conditions were not due to movement of cations. First, in the presence of impermeant anions (for example, with symmetrical 150 mM K+ gluconate), GOLAC single channel currents are abolished (20, 39). This is consistent with work by Franciolini and Nonner (12), who suggested a model for the pore of a neuronal Cl– channel that requires simultaneous binding of both anions and cations, where the cations do not transit through the pore independently of anions. Second, measurements of the permeability of the GOLACs to different cation-Cl salts (using 10-fold concentration differences between the cis and trans chambers) showed only slight permeability differences between different monovalent cations or divalent cations, and little effect of these cations on Cl– currents (our unpublished results). Third, the negative shift in Vr under bi-ionic conditions is inconsistent with an increase in cation permeability because the predicted reversal potential for cations was unchanged for both symmetrical and bi-ionic conditions. These data strengthen our conclusion that both GOLACs conduct ATP.
ATP movement into Golgi lumen is mediated by conduction through GOLACs.
The single-channel data indicated that the GOLACs are ATP-permeable anion channels, and this prompted us to test whether the channels form an important pathway for ATP movement into the Golgi lumen. Several converging lines of evidence support this hypothesis, including the direct block of single GOLAC currents by ATP and the same pharmacological profile of the channels for both [-32P]ATP influx and phosphorylation of lumenal proteins. The pharmacological profile of the GOLACs (block by DIDS and 9-AC and not by the other inhibitors tested) is distinct from the 11 anion channels (excluding the GOLACs) that are known to be expressed in the liver (19), the tissue source of the SGF1 fraction. However, both transporters and ion channels can be inhibited by DIDS and 9-AC, and thus blocking with these molecules on its own provides only circumstantial evidence and does not exclude the possibility that both a classic transporter and a channel are functioning in parallel. This limitation will only be resolved once the GOLACs have been cloned, expressed and studied in isolation from other proteins.
ATP transport into the Golgi lumen has been studied at the biochemical level by Hirschberg and colleagues (8, 9). They reported that ATP transport into the Golgi was blocked by DIDS, that ATP import was coupled to the exchange of AMP, that ATP was concentrated 30 fold in the lumen, and that transport was temperature dependent. These authors concluded that the ATP uptake mechanism of the Golgi was due to the activity of a classic transporter. Our conclusion that ATP enters the Golgi via the GOLACs rather than a transporter is in contrast to that of Hirschberg and colleagues (8, 9), but our data for the apparent Km and Vmax of [-32P]ATP influx are consistent with their data and suggests that we are investigating the same process. Could additional organellar ATP transporters be working in parallel with the Golgi proteins during these influx assays? Besides the mitochondrial ATP/ADP exchanger, which we blocked with carboxyatractyloside, there are reports of other ATP transporters in rough endoplasmic reticulum (33) and peroxisomes (40). However, there are no specific inhibitors of these transporters, and they should be present only as minor contaminants of our Golgi fraction (38).
Three aspects of our data argue for the presence of simply a channel as the mechanism of ATP uptake by the Golgi. First, the GOLACs conduct ATP at concentrations close to the concentration found in cells (Fig. 4), suggesting that the channels are functional under normal cell conditions. Second, in the reports by Hirschberg and colleagues, ATP transport was associated with a concentration in the Golgi lumen of both ATP (30 fold) and the by-products of ATP metabolism (up to 400 fold), including AMP, ADP, and Pi. ATP channels would be predicted to dissipate a high concentration inside the lumen, which is consistent with our data that ATP moves in both directions through the channels. Thus it is worth considering alternative explanations for the concentration of ATP and ATP by-products in the lumen. The concentration of [-32P]ATP Hirschberg observed could have been generated independently of a transport mechanism if ATP interacted with ATP binding proteins within the Golgi lumen or equivalently, if the radiolabeled phosphate were transferred to a protein. Both Capasso et al. (9) and our current study have demonstrated phosphorylation of lumenal Golgi proteins. Another consideration is that the development of a concentration gradient of ATP metabolic products (ADP, AMP, and Pi) in the Golgi lumen to a greater extent than ATP is counterintuitive. Elevated concentrations of AMP and ADP in the Golgi lumen inhibit ATP-requiring reactions. Because we show phosphorylation of lumenal Golgi proteins, it is unlikely that high concentrations of ADP and AMP accumulate in the vesicle lumen. Finally, Capasso et al. (9) stated that ATP transport was temperature dependent, but experimental data or Q10 values were not provided. In contrast, the data provided herein show that radiolabeled ATP entry into the Golgi was not significantly affected by a 20°C temperature change, which strongly suggests that facilitated diffusion through the aqueous pores of ion channels is the mechanism of ATP influx into the Golgi. If ATP influx were mediated by an energy-dependent process, or via a classic ATP/ADP exchanger (antiporter), such as that found in mitochondria, cold would markedly inhibit influx (16, 24). The lack of temperature dependence of ATP accumulation (Fig. 5B) is strong evidence that a channel mediates ATP influx.
The direction of ATP flux through the channel is determined by the electrochemical gradient for ATP. Because ATP is an anion at physiological pH and because there is a small or zero Vm between the cytoplasm and Golgi lumen (22, 32, 34), the GOLACs' ATP conductance depends solely upon the ATP concentration gradient. ATP concentration is 3–5 mM in the cytoplasm and is unknown in the Golgi lumen. (Note that ATP activity in the cytosol is probably in the range of 0.5 to 1 mM ATP because of the physiological ionic strength). If the ATP concentration in the Golgi lumen is lower than in the cytoplasm, the GOLACs could constitutively provide ATP to the Golgi lumen.
In conclusion, our data demonstrate that anion channels play dual roles in the Golgi complex; they provide Cl– ions to permit lumenal acidification (20, 39) and function as ATP channels. The Cl– flux through the GOLACs is required for charge neutralization of the acidic Golgi lumen that results from the activity of the organelle's -type H+-ATPase. An acidic lumen is important to maintain optimal activity of many Golgi enzymes (22). The promiscuous nature of anion conductance by the GOLACs is consistent with poor selectivity of most anion channels. We propose that the broad selectivity of the GOLACs endows these channels with multiple functions, i
