INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

CYSTIC FIBROSIS (CF) is the most common fatal genetic disease among Caucasians. CF is caused by mutations in the gene encoding the cystic fibrosis transmembrane conductance regulator (CFTR), which is a cAMP/protein kinase A-regulated Cl channel that is absent or defective in CF. CF patients most commonly succumb to respiratory complications that arise because of colonization of the respiratory tract by Pseudomonas aeruginosa and other opportunistic bacteria. One hypothesis to explain how a reduction of Cl permeability could lead to colonization of CF lungs by P. aeruginosa is that pH of the Golgi (pH_G), trans-Golgi, and trans-Golgi network, which are normally acidic relative to the cytosol, are alkaline in CF. CFTR, which could serve as the counterion conductance, may be required to prevent generation of large lumen-positive membrane voltages during pumping by the electrogenic H⁺ vacuolar (V)-ATPase (1, 2). The lack of CFTR would lead to an alkaline pH_G, which in turn would alter the activities of resident enzymes responsible for proper sialylation, sulfation, and fucosylation [due to sialyltransferase (ST), sulfotransferase, and fucosyltransferase, respectively] of secreted and surface membrane components. These alterations could lead to changes in the chemical properties of membrane and secreted glycoproteins and glycolipids, such as increased asialo-GM1, a hypothesized bacterial binding site on epithelia (5, 11, 25, 63) [see also Schroeder et al. (45) for conflicting opinion].

A number of groups have attempted to test the organelle pH hypothesis. Lukacs et al. (33) demonstrated that CFTR was functional in endosomes of Chinese hamster ovary (CHO) cells heterologously expressing CFTR, but they concluded that factors other than CFTR were the major determinants of endosomal pH. Dunn et al. (12) found that endocytic acidification was independent of CFTR when the CF pancreatic cell line (CFPAC) and CFTR-corrected CFPAC cells were compared. Seksek et al. (48) microinjected liposomes containing pH-sensitive fluid-phase dyes into the Golgi of fibroblasts and epithelial cells and found that pH_G was the same in both cell types and also in epithelial cells that normally do (Calu-3) and do not (Madin-Darby canine kidney, SK-MES-1) express CFTR (49).

There were three reasons for performing a rigorous test of the organelle pH hypothesis. First, human airway epithelial cells may be different from fibroblasts, lymphocytes, and non-airway epithelial cells from other species with regard to pH_G regulation. Second, none of the previous comparisons of pH_G was performed on cells that were genetically matched F508 (deletion of Phe-508 in CFTR) and wild-type CFTR-expressing (WT-CFTR) respiratory cell lines. Third, it was important to measure the pH of Golgi cisternae where ST and the other critical enzymes reside because pH regulatory mechanisms may differ in different organelles (54), and none of the other studies had definitively measured pH in the ST-containing region of the Golgi.

We developed a ratiometric, green fluorescent protein (GFP)-based pH sensor that was targeted to the Golgi with ST in genetically matched F508-CF and CFTR-corrected F508-CF human airway epithelial cells. Various pH-sensitive mutants of GFP (pK_a 5-7) have been targeted to the Golgi (as well as mitochondria and endoplasmic reticulum) (28, 32, 55), but artifacts can arise with the use of the single-wavelength intensity changes of these GFPs when the apparent or actual fluorophore concentration changes (e.g., due to bleaching or changes in path length) and pH does not. By fusing ultraviolet enhanced GFP (GFPuv) and enhanced yellow fluorescent protein (EYFP), we created a chimeric protein that has an excitation spectrum with pH-dependent (490 nm) and relatively pH-independent (440 nm) wavelengths. Measurements of pH_G in HeLa cells were used to confirm the method, and measurements in CF (F508 CFTR) tracheal (CFT1) and human nasal epithelial (JME) cells and in CFTR-expressing tracheal (CFT1-CFTR) and human bronchial epithelial (HBE) cells were used to determine the role of CFTR in the control of pH_G.

It was also expected that pH_G would be critically affected by the magnitude of H⁺ leaks. Because CFTR may control both Na⁺ conductance (29, 39, 50) and anion (Cl/HCO) exchange (31) in the plasma membranes of epithelial cells, it seemed possible that CFTR mutations might also affect pH_G by altering H⁺ leak pathways. We therefore compared H⁺ leaks in CFT1 and CFT1-CFTR cells. We also developed a mathematical model that used measurements of pH_G, cytosolic pH [pH_C, with the cytosolic dye 2',7'-bis(2-carboxyethyl)-5(6)-carboxyfluorescein (BCECF)], and buffer capacity of the Golgi to make the first in vivo estimates of Golgi membrane H⁺ permeability (P_H⁺, in cm/s) in CFT1 cells.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Materials

Construction of Plasmids

Bacterial expression vector for pHERP. To create a bacterial expression vector for pH-sensitive excitation ratiometric green fluorescent protein (pHERP), EYFP (S65G, S72A, T203Y, H231L) was PCR-amplified with a sense primer (5'-cccaagcttgatggtgagcaagggcgag-3') containing a HindIII restriction site (underlined) and an antisense primer (5'-gacgagctgtacaagggaggaggtctagag-3') that codes for a linker region. An XbaI restriction site (underlined) is present and eliminated the EYFP stop codon. Cloning the product into GFPuv (F99S, M153T, V163A; Clonetech) put EYFP upstream of GFPuv with an intervening linker region having the amino acid sequence GGGLEDPRVPVEK.

ST-pHERP mammalian expression vector. We PCR amplified the ST fragment, amino acids 1-70 containing the cytosolic, transmembrane, and truncated luminal domains, from human 2',6-sialyltransferase (courtesy of Dr. Brian Seed, Harvard Medical School and Massachusetts General Hospital). This portion of ST has been used to target chimeric molecules to the Golgi (60, 61). PCR amplification was performed with primers (5'-cgcgggaagcttgccaccatgattcacaccaacctg-3' and 5'-cgcgggcggatcctgggtgctgcttgagga-3') that allowed cloning of the PCR product into the pcDNA3 vector (Invitrogen, San Diego, CA) with 5' HindIII and 3' BamHI restriction sites. PCR amplification of pHERP was performed with a sense primer (5'-cgcgggagatctagaattcgtgagcaagggcgag-3') that eliminates EYFP's ATG and has a BglII site (underlined) and with an antisense SP6 primer (5'-gatttaggtgacactatag-3'). The EYFP-GFPuv PCR product was subcloned downstream of ST in pcDNA3 with BglII and ApaI. The final construct codes for the chimeric protein with amino acids 1-70 of ST, a 3-amino acid linker (LEF) between ST and EYFP, amino acids 2-239 of EYFP, a 13-amino acid linker (GGGLEDPRVPVEK), and amino acids 1-238 of GFPuv. GT-EGFP [enhanced GFP (EGFP) targeted to Golgi with galactosyltransferase (GT)] was provided by the laboratory of Roger Tsien (Howard Hughes Medical Institute and University of California, San Diego).

In Vitro Spectra of pHERP

Cell Culture

Transfections

Solutions

Fluorescence Ratio Imaging of pH_C and pH_G

Data were collected by electronically selecting regions of the image for quantitation. Cytosolic measurements were made from entire cells. When measurements were made on Golgi, only the brightest perinuclear regions were selected. Like FITC and BCECF, the fluorescence of pHERP excited at 490 nm increases with pH, whereas fluorescence at 440 nm is relatively insensitive to pH. Intensities were balanced with neutral density filters. Photobleaching was negligible during our experiments.

Methods describing calibration of cytosolic and organelle pH measurements have been reported previously (54).

Determination of Golgi Buffer Capacity

10

5

(1)

Calculation of H+ Permeability of Golgi Membranes

One experimental protocol was to add the H⁺ V-ATPase inhibitor bafilomycin to cells and model measured rates of alkalization of pH_G while pH_C remained constant (see Fig. 8A). The second approach was to acidify both cytosol and Golgi cells with an NH₄Cl pulse and then to model changes of pH_G while pH_C also changed (see Figs. 6 and 8B). Without loss of generality, the equation that describes transient changes in pH_G during the above-mentioned experimental manipulations is

(2)

where J_pump is the H⁺ flux of an individual V-ATPase, _pump is the density of H⁺ pumps in the Golgi, and J_leak is the passive flux of H⁺ across a unit area of Golgi membrane; the term in parentheses is the total flux of H⁺ across a unit area of membrane. These fluxes depend on the Golgi-cytosol [H⁺] gradient, the change in pH defined as pH_G  pH_C, and the membrane potential (_G). S and V are the Golgi surface area and volume, which were assumed to be 8 × 10⁶ cm² and 6 × 10¹² cm³, respectively (30). _G is the buffering capacity of the Golgi (determined as in Fig. 7).

Using the previously described model of the V-ATPase (19), we found computed values for J_pump to be relatively constant over moderate luminal pH ranges but sensitive to changes in membrane potential (calculated as shown in Eq. 4), consistent with previous current-voltage data (7). In the present work, calculated J_pump changed by only 1-10% over the time course of any experiment.

Although the true nature of the proton leak, J_leak, is not known, we modeled this transport as simple, passive diffusion (23)

(3)

where i denotes the ionic species, J_i is the ionic flux density, P_i is the permeability of the membrane to ion i, [C_i]_C and [C_i]_G are the concentrations of the ion in the cytoplasm and Golgi, respectively, z_i is the valance of the ion, and U = _GF/(RT), where U is reduced voltage, _G is the Golgi membrane potential, and F, R, and T have their usual meanings.

Because _G will affect both J_pump and J_leak, we included this effect by writing an explicit form for _G in terms of the excess charge inside the Golgi, the membrane of which was treated as a parallel plate capacitor, and by assuming that the dominant counterions were K⁺ and Cl

(4)

where C_m is the total capacitance of the membrane, [K⁺]_G and [Cl]_G are the molar concentrations of K⁺ and Cl in the Golgi, the integral term represents the total amount of H⁺ in the Golgi lumen (buffered plus free), and B (a constant) is the molar concentration of charged species that are trapped in the Golgi. When the concentration of trapped protein, B, is balanced by the net sum of all ionic species in the lumen, _G is zero.

When K⁺ and Cl permeabilities were assumed to be 10⁵ cm/s (22), _G was calculated (from Eqs. 2 and 3) to be <10 mV. This result was consistent with recent experiments showing that the Golgi and trans-Golgi network had relatively large conductances to both K⁺ and Cl and that _G was likely to be an unimportant determinant of either J_pump or J_leak (43, 61). Schapiro and Grinstein (43) arrived at this conclusion by finding that [K⁺]_G, measured to be 107 mM using a null point method, was similar to the cytosolic [K⁺] ([K⁺]_C).

Equation 2 together with equations describing the passive fluxes of K⁺ and Cl (from Eq. 3) forms a set of three ordinary differential equations that are coupled by the algebraic constraint of Eq. 4 and that uniquely determine the time course of changes of ionic concentrations in the Golgi given an initial set of conditions. We assumed [K⁺]_C = 130 mM and [Cl]_C = 20 mM throughout each experimental run, and initial [K⁺]_G and [Cl]_G were chosen to keep the Golgi close to electroneutral (consistent with _G < 10 mV).

pH_C and transient changes of pH_G (similar to those shown in Fig. 8, A or B) were measured in separate experiments, and then the model was used (Eq. 3) to predict the changes in Golgi acidification. This was done by varying P_H⁺, _pump, and the concentration of fixed negative charge in the lumen (B) until a best fit to pH_G was obtained. When results were modeled from experiments using V-ATPase inhibitors (similar to Fig. 8A), _pump was set to zero and pH_C was held constant to match experimental conditions. The Na⁺-free, acid-loaded Golgi experiments (see Fig. 8B) were more complex because pH_C and pH_G both varied. However, this procedure was advantageous because it allowed us to acidify the Golgi and keep pH_C relatively constant (see Figs. 6 and 8B). For these experiments, we reported the average of six predicted P_H⁺ values, for each pH_G experiment, where each pH_G run was fit against a different pH_C run (see Possible errors in predicting P_H⁺). All searches were performed with a Nelder-Mead algorithm, and the ordinary differential equations were solved with a stiff method in both Matlab and Berkeley Madonna software (38).

Although Eqs. 2-4 yielded estimates of P_H⁺, B, and the number of active H⁺ V-ATPases (NOP = _pump · S), we have reported only predicted values for P_H⁺. The model yielded average values of B  140 mM and NOP  2,000. Within a particular data set, these two parameters had counteracting effects: a decrease in _pump together with an increase in B sometimes resulted in very similar pH fits. This made it very difficult for the search algorithm to find a unique, best fit, and the same data set with different initial search conditions would yield very different values for B and _pump. In contrast, estimates of P_H⁺ were more robust to the initial conditions of the search algorithm, and the same best value for P_H⁺ was usually found regardless of the values of B and _pump. As mentioned above, J_pump changed little for any given run. Thus the parameter P_H⁺ had the strongest influence over the shape of the predicted pH curves.

It should be noted that in the limit of  = 0 mV, Eqs. 2-4 can be replaced by an intuitive and simple expression for the P_H⁺ of the Golgi in terms of the instantaneous rate of change of the Golgi, the H⁺ gradient, and a few physical parameters

(5)

Thus P_H⁺ could also be determined from initial rates of alkalinization after treatment with bafilomycin to block the H⁺ pumps without appealing to the more complicated model. For the bafilomycin-treated Golgi, this method predicted P_H⁺ values that were within factors of 2-4 of the values determined using the full model.

Possible Errors in Predicting P_H+

Surface-to-volume ratio. We arbitrarily chose the value of 1.33 × 10⁶ cm¹ for the surface-to-volume ratio (S/V) of the Golgi obtained from rat kidney cells (30). Golgi S/V in terminal tubule and acinar cells of the rat submandibular gland (51) is an order of magnitude smaller. If the latter estimates were correct, our P_H⁺ values would be 10-fold larger; thus our calculations provide a lower estimate. This S/V parameter is most likely the largest source of error in our study.

Buffering. If the true buffering capacity is 20% larger or smaller than the value measured, which is consistent with our errors, then the true P_H⁺ will be 20% larger or smaller than predicted.

H+ gradients. In many of the bafilomycin-induced alkalization experiments, pH_C was assumed to be 7.55 (the average value). Varying the assumed pH_C by ±0.2 pH units for a typical run had a ±25% effect on the estimated P_H⁺. For the experiments where the Golgi has been acidified, each pH_G run (see examples in Figs. 6 and 8B) was fit against six pH_C runs that represented the entire set of measured cytoplasmic recovery experiments. Most of these reported P_H⁺ values had SD of 2 × 10⁴, although one had a SD of 2 × 10³.

Membrane potential. The movement of counterions influences H⁺ movement through effects on _G as shown in Eq. 4. When the Golgi membrane was assumed to have K⁺ and Cl permeabilities >5 × 10⁹ cm/s, _G was calculated (from Eqs. 2 and 3) to be <10 mV. When permeabilities to K⁺ and Cl were <5 × 10⁹ cm/s, K⁺ and Cl movements became quite slow, and _G began to affect pH_G. The present P_H⁺ predictions remain unchanged for K⁺ and Cl permeabilities >5 × 10⁹ cm/s. Below this value, movements of counterions resulted in transient changes of _G that affected calculations of P_H⁺. When _G was arbitrarily varied from +7 mV to +50 mV and held constant over the time course of the simulation, the predicted P_H⁺ values varied from +50% to 50% of the original value determined with our model of _G. Therefore, the P_H⁺ values determined with Eqs. 2 and 3 were relatively insensitive to _G.

Obtaining Initial Alkalization Rates and P_H+ From the Literature

Statistics

As an objective measure of the quality of our fits, the conventional root mean square (RMS) value was computed. When a particular run did not have an RMS value <0.175 pH units, it was dropped from the analysis. The average RMS value for all runs fit by the model was 0.06 ± 0.03 pH units (mean ± SD).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

pH Sensitivity and Golgi Targeting of pHERP

View larger version (19K): [in this window] [in a new window]   Fig. 1.   Construction of sialyltransferase (ST)-pH-sensitive excitation ratiometric green fluorescent protein (pHERP). A: in vitro excitation spectrum of green fluorescent protein (GFP) constructs measured at 520 nm. Spectra for enhanced yellow fluorescent protein (EYFP) and ultraviolet enhanced GFP (GFPuv) were obtained at pH 7.5. EYFP and GFPuv were fused together, separated by a 13-amino acid linker to create pHERP. The spectrum of pHERP is extremely sensitive to pH at ~490 nm and relatively insensitive at 440 nm, allowing for ratiometric imaging (ex, excitation). B: schematic of ST-pHERP. Amino acids 1-70 of ST were used to target pHERP to the Golgi. N, NH2 terminal; C, COOH terminal. C: representative in vivo calibration of ST-pHERP (performed at the end of each experiment). Golgi pH was allowed to equilibrate with solutions of various pH values in the presence of 10 µM each of nigericin and monensin.

View larger version (148K): [in this window] [in a new window]   Fig. 2.   ST-pHERP morphology. Human nasal epithelial (JME) cells were transiently transfected with ST-pHERP and mounted on the stage of the microscope. Fluorescence image overlaid on the bright-field image shows that pHERP exhibit a perinuclear staining pattern typical of Golgi.

pH_G Measured With pHERP

View larger version (13K): [in this window] [in a new window]   Fig. 3.   ST-pHERP reports Golgi pH (pHG) in HeLa cells. A: a pulse of 30 mM NH4Cl alkalinized the Golgi due to NH3 entry. The slower NH entry led to a slight acidification, which was apparent when the NH4Cl was washed away. Golgi pH rapidly recovered from the acidification. B: histogram of steady-state pHG measured with ST-pHERP at the beginning of each experiment.

ST-pHERP consists of ST fused to the relatively pH-insensitive GFPuv and pH-sensitive EYFP. EYFP has recently been reported also to be sensitive to [Cl], with decreases in [Cl] causing increases in fluorescence (57). Although the changes in EYFP fluorescence due to changes in [Cl] are smaller than those due to pH (32, 57), we were concerned that changes in ST-pHERP fluorescence ratio in vivo would be larger than predicted from calibrations in which pH was varied but [Cl] was held constant. We therefore compared measurements of pH_G made with ST-pHERP with those obtained with GT-EGFP (13, 28, 32). EGFP is insensitive to [Cl] but sensitive to pH, with a pK_a of 6.4 (32, 57). When expressed in both CFT1 and CFT1-CFTR cells, GT-EGFP and ST-pHERP yielded similar results when pH_G was perturbed. When cells expressing either ST-pHERP (see Fig. 6) or GT-EGFP (data not shown) cells were pulsed with NH₄Cl followed by removal and then Na⁺-free treatment, pH_G acidified and recovered partially. Full recovery was completed only when Na⁺ was added back to the cells. These results support the conclusion that ST-pHERP was accurately reporting pH_G. The fact that ST-pHERP was more easily calibrated in terms of pH_G made it preferable to GT-EGFP.

CFTR and Steady-State pH_G

The activity of CFTR is increased by cellular cAMP concentration, and the effects of cAMP on pH_G have been controversial (32, 48). We therefore tested the effects of cAMP in CFT1 and CFT1-CFTR cells. In both cell types, there was little change in pH_G in response to cAMP whether experiments were performed in the presence of 25 mM HCO/5% CO₂ or in nominally HCO-free solutions (Fig. 4).

View larger version (15K): [in this window] [in a new window]   Fig. 4.   Increasing cAMP concentration does not alter pHG. Left: experiments were conducted in the presence of 5% CO2 and 25 mM HCO. Intracellular cAMP concentration was increased by application of 10 µM forskolin (arrow). Right: experiments were conducted in HEPES-buffered nominally HCO-free conditions. cAMP concentration was increased by application of 10 µM forskolin and 500 µM dibutyryl-cAMP (arrow). Traces are from Golgi of single cells. CFTR, cystic fibrosis transmembrane conductance regulator; CFT1(CFTR), CFTR-expressing tracheal cells.

Golgi H+ "Leak" and Potential Effect of CFTR

2

2

Golgi-to-Cytosol [H+] Gradient

7

7

Golgi H+ Permeability and Buffer Capacity

3

View larger version (26K): [in this window] [in a new window]   Fig. 5.   The Golgi H+ leak is large. A: a representative trace of cytosolic pH (pHC). Cells were exposed to a pulse of 30 mM NH4Cl for 5 min, and Na+ was then replaced with N-methyl-D-glucamine (NMDG) 2.5 min into the NH4Cl pulse. When NH was removed, the cells acidified and remained acidic. Reapplication of Na+ allowed pHC to recover to control levels. B: a representative trace of pHG during an experiment similar to that performed on the cytosol in A, although the cell was first exposed to bafilomycin (250 nM), a specific inhibitor of the H+-ATPase. C: cytosolic and Golgi pH alkalization curves from A and B are superimposed and the time scale is magnified. Slope at pH 7 was taken for a number of similar experiments. D: model of the likely transporters and channels that are regulating pH during the experiment.

It is well established that the cytosolic buffering capacity varies with pH (58, 59), which is a function of the pK_a and concentration of titratable groups (41). Previous measurements of Golgi buffer capacity have reported only one value (42, 61) or claimed that _G was constant between pH 6 and 7 (13). We felt that it was necessary to measure _G over a wide pH range to calculate P_H⁺ of the Golgi at the various pH values observed in our experiments. Values for _G were obtained by first treating the cells with bafilomycin (250 nM) and then titrating in various amounts of weak base (NH₃) or weak acid (HOAc). We measured _G at pH < 7 by following the bafilomycin treatment with an acidification step (NH₄Cl prepulse followed by incubation in Na⁺-free solution) and then adding either NH₄Cl or HOAc to the Na⁺-free solutions. A typical experiment in which _G was measured is shown in Fig. 7A. Data summarizing measurements of _G as a function of pH are summarized in Fig. 7B. Buffer capacities were grouped into 0.2-pH unit buffering domains for simplicity. At pH 6.9 ± 0.02 (mean ± SE), _G = 17.2 ± 4 mM per pH unit (mean ± SE), in good agreement with previous results of 10-40 mM per pH unit (13, 42, 61). At higher and lower pH values, _G varied, and the variation was well fit by a single exponential (Fig. 7B), which allowed us to extrapolate to acidic values that were attained in some measurements of P_H⁺.

By using the experimentally determined _G and a model of pH_G (Eqs. 2-4), we were able to estimate the P_H⁺ of bafilomycin-treated cells when pH > 6.4 (Fig. 8, A and C). Additionally, the model allowed estimates of P_H⁺ under various conditions where organelle and pH_C varied over a considerable range: recovery of both pH_G and pH_C were measured after acidification (using an NH prepulse) and treatment with Na⁺-free solution (see example in Fig. 8, B and C). These data provided P_H⁺ when pH_G < 6.4. These calculations showed that the log of Golgi P_H⁺ was inversely correlated with the pH_G (Fig. 8C). The data were well fit by a linear equation, and the slope of this line was significantly different from the slope of 0 (P < <0.01).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

ST-pHERP for Measuring pH_G

GFPs have many advantages as organelle pH sensors: any organelle can be targeted, a variety of mutants with a wide range of pK_a values (5.5-7.0) is available (28, 32, 55), and changes of pH can be measured with little background signal, no diffusive loss of fluorophore, minimal bleaching, and no requirement for exogenous dyes that require hydrolysis and might lead to cytotoxicity. Despite these many advantages, calibrating the single-wavelength intensity changes of these GFPs is problematic, because artifacts can arise when apparent fluorophore concentration changes and pH does not. Miesenbock et al. (34) overcame this problem by developing a GFP with a spectrum that shifted with pH, thereby providing an excitation ratiometric pH indicator. Our development of pHERP allows an alternative approach that has many of the same advantages as pHlorin for measuring pH_G.

CFTR, Counterion Conductance, and Membrane Potential in Determining pH_G

Al-Awqati and colleagues (1, 2) suggested that the Golgi Cl permeability was able to limit H⁺ flux by dissipating the membrane potential. Subsequent data in the literature indicated that conductances to both K⁺ and Cl in the Golgi and trans-Golgi network are so high that membrane voltage is small and unimportant in determining organelle pH (10, 43, 61). This is consistent with observations that neither ouabain nor its membrane-permeant analog acetyltrophanthidin have any effect on pH_G in HeLa cells, which have no CFTR (Wu M and Machen T, unpublished observations), implying that the Na⁺-K⁺-ATPase activity was low or nonexistent. Our model confirmed and extended the conclusions regarding the small Golgi membrane potential and also allowed us to estimate the number of counterion channels necessary to achieve this goal. Our model showed that K⁺ and Cl permeability could be decreased to as low as 5 × 10⁹ cm/s before membrane potential became a factor in fitting the pH transients (see METHODS and Fig. 8). For a Golgi with a surface area of 8×10⁶ cm² (30), these very low permeabilities corresponded to K⁺ and Cl conductances <20 pS for the entire Golgi, which could be achieved with between one and two K⁺ or Cl channels. Calculation of K⁺ conductance was performed by using the Goldman-Hodgkin-Katz (GHK) current equation (23) after compensating for the assumed surface area of the Golgi. Conductance was calculated as the slope of the current with respect to voltage at 0 mV, assuming [K⁺]_G = 110 mM (43) and [K⁺]_C =130 mM. In this circumstance, it is likely that eliminating CFTR has no effect on pH_G, because other K⁺ and Cl channels can provide sufficient counterion conductance to ensure that membrane potential in the Golgi is small and therefore has little effect on H⁺ pumps or leaks. Two other studies have provided evidence for the presence of Cl channels in the Golgi: Nordeen et al. (35) found an anion channel in enriched Golgi fractions, and Schwappach et al. (46) found that Gef1p (the only yeast ClC Cl channel) localizes to the Golgi. Additionally, elimination of Gef1p from yeast does result in altered cation homeostasis (17) but does not effect Kar2p secretion or glycosylation of invertase, both dependent on an acidic Golgi (46).

It might be argued that since our experiments were performed on single cells, they are not comparable to those performed on confluent monolayers. We attempted to make pH_G measurements on confluent cells grown on filters, but the background fluorescence contributed by the filter exceeded the specific fluorescence of pHERP when excited at 440 and 490 nm under acidic conditions. Therefore, it was impossible to perform experiments when cells were grown on filters. Because of technical difficulties, we were able to obtain only one pH_G measurement from a CFT1-CFTR cell in a confluent patch. Results from this experiment clustered with most of the other CFT1-CFTR data (Fig. 8C). We also note that CFTR transits through the Golgi and becomes functional in the plasma membrane of a variety of cells grown on glass either as single cells or in confluent patches (14, 15). We therefore believe, but have not proven, that pH_G experiments performed on isolated cells grown on glass reflect those of confluent cells.

Roles of H+ Pump, pH-Dependent Leak, and pH_C in Determining pH_G

As summarized in Fig. 8C, P_H⁺ in Golgi of CFT1 cells was ~7.5 × 10⁴ cm/s when pH_G was 6.5, and there was little difference to CFTR-corrected CFT1 cells. Similar values for Golgi P_H⁺ of HeLa, Vero, and CHO cells were calculated using our model (Eqs. 2-4), and bafilomycin-induced alkalinization data were obtained from the literature (Table 2). These P_H⁺ values are large compared with typical permeability values for Na⁺, K⁺, and Cl: 10¹² cm/s for a membrane without channels (37) and up to 10⁵ cm/s for a membrane with channels (22, 24). However, our estimates of P_H⁺ compare favorably with P_H⁺ in lipid bilayers and liposomes: 10⁷-10² cm/s (8, 21, 37, 40). Similar to the data presented for P_H⁺ of the Golgi, artificial lipid bilayers also have pH-dependent P_H⁺ (8, 21). It therefore seems possible that P_H⁺ of the Golgi might be solely due to simple H⁺ diffusion through membranes, and variations in P_H⁺ could therefore be due to differences in fatty acids (which act as weak acid shuttles) and/or lipids (which can form water "wires") (8, 20, 21) in the Golgi of different cells. In addition, H⁺ channels may also provide leak pathways for H⁺ in the Golgi (43). However, estimates of P_H⁺ > 5 × 10¹ cm/s for plasma membranes of lung alveolar cells with H⁺ channels [using GHK, reported pH gradients, and measured current densities (6, 9)] was 100 times larger than the largest P_H⁺ observed in the present experiments. Finally, Na⁺/H⁺ exchange in the Golgi might contribute to the H⁺ permeability. However, Schapiro and Grinstein (43) showed that NHEs play no role in the efflux of H⁺ from the Golgi. Also, we have found that treatment of CFT1 cells with hexamethylene amiloride (membrane-permeant analog of amiloride) had no effect on steady-state pH_G (data not shown).

A consequence of the Golgi having a relatively large, pH-dependent P_H⁺ is that pH_G will be a complicated function of pH_C, P_H⁺, and H⁺ pumping. The role of pH_C in determining pH_G is shown by experiments in which cells were treated with NH₃/NH and both cytosol and Golgi were first alkalinized and then secondarily acidified (Figs. 3 and 6). Because this secondary acidification also occurred in the presence of bafilomycin (Fig. 5) and thus was not due to the H⁺ V-ATPase (as proposed in Refs. 27 and 42), it may have been due to H⁺ (or NH) accumulation across the plasma membrane into the cytosol, which then equilibrated with the Golgi. When extracellular Na⁺ was added back to acidified cells, the Golgi alkalinized at a rate that was only 2.5-fold slower than that of the cytosol, likely due to a delayed leak of H⁺ from the Golgi into the cytosol, which was being alkalinized due to active H⁺ extrusion by NHE in the plasma membrane.

View larger version (12K): [in this window] [in a new window]   Fig. 6.   Golgi recovery from an acid load is partially dependent on extracellular Na+. A: the same representative pHC trace from Fig. 5. B: representative trace of pHG during the same protocol performed on the cell in A. C: model of the likely transporter and channels regulating pH during the experiment.

The role of H⁺ pumps in setting pH_G was suggested by the fact that HeLa cells had a more acidic Golgi than respiratory epithelial cells (Table 1). This was not solely due to the more acidic pH_C in HeLa cells because the Golgi-to-cytosol [H⁺] gradient was larger in HeLa (3.6 × 10⁷ M) than in respiratory epithelial cells (0.75-1.7 × 10⁷ M) (Table 1). Given that P_H⁺ of HeLa and other cells was similar in the pH 6.4-6.6 range (Fig. 8C), the most likely explanation is a difference in H⁺-pump activity between HeLa cells and CFT1 cells. This explanation would also apply to the difference between pH_G of CFT1 and CFT1-CFTR cells, although why or how various pump activities would arise in different cells is not clear.

View larger version (13K): [in this window] [in a new window]   Fig. 7.   Golgi buffer capacity (G) is pH dependent. A: representative trace of 1 cell for which Golgi buffer capacity was obtained. The cell was first acid-loaded with an NH4Cl prepulse and then exposed to varying amounts of NH4Cl under Na+-free/high-K+ conditions. B: summary of all Golgi buffer capacity experiments performed with NH4Cl and KHOAc. Buffer capacities were grouped by 0.2-pH unit increments. Data are means ± SE. The smooth line is an exponential fit through the data. (G = 10.4 × 104 e1.3 pH units).

View larger version (12K): [in this window] [in a new window]   Fig. 8.   PH+ is pH dependent. A: representative trace showing that when cells were treated with bafilomycin, pHG rapidly alkalinized. Data from the average of 6 cells are represented by points. The smooth line is the best fit of the data obtained from Eqs. 2-4 with the pumps turned off. B: superimposition of pHG (Golgi) and pHC (cytosol) traces after compartments were acid loaded (as in Fig. 6, A and B). Notice that on this time scale pHC was relatively constant, while pHG recovered. Golgi pH (data points) was fit using Eqs. 2-4 (smooth line). At ~7 min, the average gradient returned to 2.0 × 107 M. C: log PH+ plotted against pHG. Filled diamonds represent data from bafilomycin-treated CFT1 cells as described in Fig. 7A. Asterisks represent data from bafilomycin-treated CFT1-CFTR cells (1 CFT1-CFTR in the cluster of 5 was a cell from a confluent patch). Open diamonds represent acid-loaded cells that were not treated with bafilomycin, as in Fig. 7B. Open circles (a-d) represent values extrapolated from the literature (see Table 2). The smooth line is a linear regression through the CFT1 data where log PH+ = 0.53 pH  6.50. The slope of the line is significantly different from a line with a slope of zero (P < 0.01). Data from the literature and CFT1-CFTR cells were not included in the linear regression.

A potential regulatory role for the H⁺ pump has been considered previously by Kim et al. (27), who found that pH_G was constant in the face of altered pH_C. In contrast, the present data show that when pH_C was acidified from an average of 7.5 to 6.5, pH_G acidified on average from 6.5 to 6.2. These averages predict that the Golgi-to-cytosol [H⁺] gradient is constant in the face of an apparently decreasing P_H⁺ and constant H⁺ pump. It will be necessary to measure pH_C and pH_G in the same cells to determine whether H⁺ pump and leak may be regulated. It also will be important to determine the roles of pH_C, H⁺ pump, and pH-dependent H⁺ leak in generating the wide range of pH values found in other acidic organelles [e.g., endosomes (pH 6.0-6.5), lysosomes (pH 4-5), and secretory granules (pH 4-5)] in different cells.