Neuronal electrical activity causes only modest changes in global intracellular pH (pHi). We have measured regional pHi differences in isolated patch-clamped neurones during depolarization, using confocal imaging of 8-hydroxypyrene-1,3,6-trisulfonic acid (HPTS) fluorescence. The pHi shifts in the soma were as expected; however, substantially larger shifts occurred in other regions. These regional differences were still observed in the presence of CO2-HCO3−, they decayed over many seconds and were associated with changes in calcium concentration. Lamellipodial HPTS fluorescence fell by 8.7 ± 1.3 % (n= 9; ∼0.1 pH unit acidification) following a 1 s depolarization to 0 mV; this was more than 4-fold greater than the relative shift seen in the soma. Depolarization to +40 mV for 1 s caused a 46.7 ± 7.0 % increase (n= 11; ∼0.4 pH unit alkalinization) in HPTS fluorescence in the lamellipodia, more than 6-fold that seen in the soma. Application of 5 % CO2-20 mm HCO3− did not significantly reduce the size of the +40 mV-evoked local pH shifts despite carbonic anhydrase activity. The pHi gradient between regions ∼50 μm apart, resulting from acid shifts, took 10.3 ± 3.1 s (n= 6) to decay by 50 %, whereas the pHi gradient resulting from alkaline shifts took only 3.7 ± 1.4 s (n= 12) to decay by 50 %. The regional rates of acidification and calcium recovery were closely related, suggesting that the acidic pH microdomains resulted from Ca2+-H+ pump activity. The alkaline pH microdomains were blocked by zinc and resulted from proton channel opening. It is likely that the microdomains result from transmembrane acid fluxes in areas with different surface area to volume ratios. Such neuronal pH microdomains are likely to have consequences for local receptor, channel and enzyme function in restricted regions.
Calcium signalling is known to play a pivotal role in controlling neuronal function. Calcium ions entering during electrical activity are extruded in exchange for protons (Ahmed & Connor, 1980; Schwiening et al. 1993; Trapp et al. 1996b). It is generally believed that proton buffering, diffusion and regulation reduce the size of the resultant pH changes such that they are of little physiological importance. However, indirect data suggest that pH transients dynamically limit neuronal dendritic calcium entry (Tombaugh, 1998) and affect synaptic function by modulating NMDA receptors (Traynelis & Cull-Candy, 1990).
In this study we have used confocal imaging of pH-dependent fluorescence during depolarization under voltage clamp to explore the possibility of transient cytoplasmic pH microdomains. The data presented here were recorded from isolated Helix aspersa neurones plated onto glass coverslips. Like other molluscan neurones, they showed thin lamellipodial processes ∼1 h following attachment to the glass (Schacher & Proshansky, 1983). Depolarization to 0 mV was used to trigger calcium entry and thereby evoke neuronal acidification as a result of calcium extrusion in exchange for protons (Schwiening et al. 1993). Larger depolarizations, to +40 mV, were also used to open voltage-gated proton channels and cause alkalinization (Thomas & Meech, 1982).
We are thus able to show directly, for the first time, large spatially localized intracellular pH (pHi) signals in neurones resulting from transient proton fluxes through proton channels and the calcium-hydrogen pump (Ca2+-H+ pump). The activity-dependent pH shifts are larger in regions of the cell with a high surface area to volume ratio. Spatial pH gradients have been reported in enterocytes and myocytes (Stewart et al. 1999; Spitzer et al. 2000), in the absence of CO2-HCO3−. In contrast the localized pH shifts we report here occur even in the presence of the mobile pH buffer CO2-HCO3−. It is likely that other neurones will show similar large acidic pH shifts in dendritic regions during electrical activity. These findings raise the possibility that protons may have a similar local intracellular signalling role to calcium. Localized pH signals, as a result of transmembrane proton fluxes, may act on many cellular and synaptic processes including local dynamic modulation of channel activity (Turrigiano et al. 1998; Schiller et al. 2000), growth cone turning (Zheng, 2000) and glutamate uptake (Zerangue & Kavanaugh, 1996).
Suboesophageal ganglia were removed from garden snails, Helix aspersa, cut into eighths and incubated in 1 mg ml−1 protease (Sigma Type XIV) in snail Ringer solution (composition (mm): NaCl 80, KCl 4, CaCl2 7, MgCl2 5 and Hepes 20; pH 7.5) at ∼30 °C for ∼1.5 h. The fragments were then washed in Ringer solution and neurones were isolated by passing the tissue through the polished tip of a plastic pipette (internal tip diameter ∼1 mm). The isolated neurones were placed in a superfusion chamber on the stage of a Zeiss Axiovert microscope and allowed to settle for at least 0.5 h.
Patch pipettes (tip diameter ∼1.5 μm) were pulled from 1.5 mm diameter unfilamented borosilicate glass (Clark Electromedical, UK) and filled with 110 mm CsCl containing either 500 μm 8-hydroxypyrene-1,3,6-trisulfonic acid (HPTS; Sigma) for pH measurements, or 100 μm Oregon Green BAPTA-1 (Oregon Green; Molecular Probes) for calcium measurements. A neurone with an axonal projection was selected and patch clamped in the whole-cell configuration after a seal of > 1 GΩ was achieved. The isolation procedure yielded neurones with a range of sizes and variable calcium current amplitudes. The largest neurones were avoided because of the delays in fluorescent dye loading and problems in achieving good voltage control. Cells exhibiting no inward calcium current were not analysed on depolarization to 0 mV. Electrophysiological recordings were made using an Axoclamp 2B (0.1 gain headstage; Axon Instruments) and data were sampled at 200 Hz using the software Spike2 (Cambridge Electronic Design, UK).
Single wavelength confocal measurements of fluorescence were made using a Zeiss LSM 510 (× 40 oil objective; NA, 1.3; excitation wavelength: HPTS, 458 nm; Oregon Green, 488 nm; emission wavelength: HPTS and Oregon Green, > 505 nm). Images were collected continuously or, when little change in the signals was expected, in either short bursts or at a lower frequency to minimize photobleaching. The discontinuous collection can be seen in some of the figures as breaks in the fluorescence traces. Pixel-based data analysis was performed using a customized Visual Basic program to extract regional data and construct background-subtracted relative fluorescence (F/F0) plots. F0 images were mean images from the control period, 2 s before depolarization.
pHi estimation and ΔpH calibration
Single-point estimations of pHi from F/F0 data were possible because HPTS fluorescence, excited at 458 nm, completely quenches at acidic pH (Willoughby et al. 1998) and the buffering power (β) of these cells is known. The data points in Fig. 1A show the 12-bit absolute fluorescence intensity of a calibration solution (110 mm CsCl, 500 μm HPTS and 5 mm Hepes) at 14 different known pH values. The continuous line, fitted by least squares to the data points, was plotted from the standard Grynkiewicz equation (Grynkiewicz et al. 1985) with values of 7.18 for the pK, 0 for the minimum fluorescence and 3150 for the maximum fluorescence of HPTS. We used two simple physicochemical equations to estimate pHi. Equation (1) is a derivative of the Grynkiewicz equation modified for use with HPTS when excited at 458 nm.
Using eqn (1), the pH shift (ΔpH) of an HPTS-containing solution can be calculated from the fluorescence shift (F/F0). In eqn (1), F0 is the absolute fluorescence yielded by the solution before the pH shift, at the starting pH (pHstart), and F is the absolute fluorescence following the pH shift. Figure 1B shows six lines plotted from eqn (1), each line having a different pHstart. For example, it can be seen from Fig. 1B that eqn (1) predicts that an HPTS-containing solution with a starting pH of 6.75, will show an 80 % increase in fluorescence when its pH is increased by 0.4 pH units (F/F0= 1.8, see data point marked by the filled square in Fig. 1B). The open circles in Fig. 1B show seven data points from Fig. 1A replotted with the pH 7.00 calibration solution as pHstart. Figure 1B shows that the agreement between the measured ΔpH values and those predicted by eqn (1), from the fluorescence shifts, is excellent. Equation (1) was used in Fig. 2 to calibrate pHi.
Equation (2) allows the pH shift (ΔpH) to be calculated on application of a weak base at a concentration C (m), of dissociation constant pKb, at a known extracellular pH (pHo), to a cell at pHi if the buffering power (β) is also known.
Linear and single exponential fits were performed by least squares. Means ±s.e.m. are shown, where n represents the number of experiments performed. Significance was determined using Student's unpaired t test.
Single wavelength HPTS calibration
We chose the pyrene-based dye HPTS, rather than a fluorescein-based dye, to assess pHi changes since it does not inhibit the Ca2+-H+ pump (Willoughby et al. 1998). As this is the first time, to our knowledge, that confocal imaging of HPTS fluorescence has been used to assess regional pH changes we sought to verify that it behaves consistently in different cellular regions. To do this we induced stable and identical pH displacements across the entire isolated snail neurone. Firstly, we used the outwardly rectifying proton channels to drive pHi to thermodynamic equilibrium (Thomas & Meech, 1982) throughout the cell. Secondly, we applied a weak acid or base extracellularly to displace pHi to a new steady-state value. Figure 2 illustrates the first approach. Figure 2Aa shows a transmitted light image of a neurone and patch pipette. In Fig. 2Ab, a side view of a rendered z-series confocal reconstruction, from the HPTS fluorescence, of the same patch-clamped neurone is shown. Figure 2Ab shows that both the soma and the axon foot had made attachments to the glass coverslip (indicated by the vertical yellow line). By focusing the confocal slice (1 μm thick) just above the coverslip we obtained the greyscale image shown in Fig. 2Ac. The image also has marked on it two coloured regions of interest from which mean fluorescence intensity was calculated. This fluorescence is shown plotted against time in Fig. 2B, as absolute intensity and as a relative fluorescence shift (F/F0).
HPTS fluorescence, when excited at 458 nm, is known to increase on alkalinization, and decrease on acidification (Wolfbeis et al. 1983; Willoughby et al. 1998). When clamped at −60 mV the mean fluorescence intensity in the axonal lamellipodia (shown in red) was low when compared with that of the cell body (blue) and showed only a small absolute rise when the holding potential was changed to +40 mV. However, the F/F0 trace shows that the size of the steady-state relative fluorescence increase was almost identical in the two regions. This would be expected if the pH-sensitive properties of HPTS fluorescence were independent of cellular location and total fluorescence intensity. This observation was confirmed in other experiments (data not shown) in which superfusion of the weak base trimethylamine (3 mm) induced similar F/F0 shifts in regions of differing absolute fluorescence intensity. It is therefore unlikely that the fluorescence properties of HPTS are markedly changed at the different cytoplasmic locations.
We have thus been able to use F/F0 as a reporter of pH throughout the cytosol. Stable pHi displacements, by prolonged depolarization or the application of weak acids or bases, enable F/F0 to be calibrated. Since the total proton conductance is large and inactivates slowly (Mahaut-Smith, 1989), it can be used to drive pHi to thermodynamic equilibrium, and thereby calibrate a single point on the F/F0 trace. We used such a calibration technique in the experiment shown in Fig. 2 where the steady-state pHi, during sustained depolarization to +40 mV, was calculated to be 8.2 when pHo was 7.5. Furthermore, since 458 nm-excited HPTS fluorescence is quenched completely at acidic pH (Schwiening & Willoughby, 2000), a relative fluorescence shift from a given pH can be calibrated if one assumes no change in apparent dye concentration (see eqn (1)). Thus we calculated a starting pH of ∼7.2. Although sustained depolarization resulted in a uniform displacement of pHi, the fluorescence shift in the lamellipodia (red region) occurred more rapidly than that seen below the soma (blue region). Similar results were obtained in a total of five experiments. In Fig. 2B the difference in the rate of alkalinization demonstrates a pH gradient within the cytosol. The maximal pH gradient was ∼0.3 pH units and occurred ∼5 s after the start of depolarization.
pH changes during brief depolarizations
Since sustained large depolarizations are not seen physiologically in neurones, we decided to investigate whether such pH gradients might be visible using briefer, more physiological, depolarizations. We chose to depolarize for 1 s under voltage clamp to three different potentials: 0, +20 and +40 mV. Depolarization to all three potentials results in calcium influx which leads to an intracellular acid load when the calcium is extruded via the Ca2+-H+ pump (Schwiening et al. 1993; Trapp et al. 1996b). However, depolarization to +20 or +40 mV can also activate proton channels resulting in acid efflux if pHi is low enough to result in an outward electrochemical gradient for protons. Therefore, depolarization can lead to an acidification, an alkalinization or a mixture of both. Figure 3 shows an example of such pHi changes in an isolated snail neurone as a result of depolarizations for 1 s. Figure 3Aa shows the neurone that was patch clamped in the whole-cell configuration and dialysed with HPTS. The confocal image (Fig. 3Ab) shows the fluorescence of the dye in an optical slice just above the coverslip. Three regions of interest were selected: the cell body (shown in green), the axon foot (shown in blue) and the lamellipodia (shown in red). The lamellipodia had relatively low fluorescence when compared with the other regions. This low fluorescence intensity results from the thinness of the lamellipodia (< 1.2 μm) when compared with the confocal optical slice (2.4 μm), rather than from a lower dye concentration, or a persistent pHi gradient. To normalize for differences in fluorescence within the optical slice, the relative HPTS fluorescence (F/F0) was plotted (Fig. 3B, lower trace). The first depolarization to 0 mV for 1 s caused a fall in pH, as a result of Ca2+-H+ pump activity. On depolarization to +20 mV the lamellipodia, and to a lesser extent the axon foot, showed biphasic fluorescence changes. The initial upward (alkaline) shift occurred within a few milliseconds of depolarization, whilst the later fall in fluorescence (acidification) directly followed repolarization. Depolarization to +40 mV resulted in a large and rapid increase in fluorescence in the lamellipodia, with progressively smaller shifts in the axonal and somatic regions. The final depolarization in Fig. 3 shows that the proton channel blocker zinc (Mahaut-Smith, 1989) abolished the rapid increase in fluorescence on stepping to +40 mV and unmasked an underlying fall in pH. All neurones exposed to zinc (n= 5) showed inhibition of the alkaline shift. In the lamellipodia, the inhibition was 100 ± 10 % (n= 3; 50 μm zinc) and 64 % (n= 2; 20 μm zinc). The direction of these pH changes (Fig. 3) is as expected from the acid-equivalent fluxes known to exist in snail neurones (Thomas & Meech, 1982; Schwiening et al. 1993). Proton efflux through voltage-gated, outwardly rectifying proton channels causes alkaline shifts whilst the acid shifts occur as a result of calcium extrusion via the Ca2+-H+ pump. The relative magnitudes of the pHi changes may be related to the surface area to volume ratio of the regions, assuming a uniform H+ permeability of the plasma membrane. We calculated that the lamellipodia had a total exposed surface area of ∼750 μm2 and a volume of ∼450 μm3, assuming a uniform thickness of 1 μm. The cell body had a total surface area of ∼4100 μm2 and a volume of ∼24 000 μm3, assuming it to be a smooth sphere. Thus the lamellipodia in Fig. 3 had a surface area to volume ratio that was ∼10-fold greater than that of the cell body.
Figure 4 shows mean regional data for the pH transients evoked by depolarization. The peak alkalinizations (open bars in Fig. 4), on depolarization to +40 mV for 1 s, are expressed as the percentage shift in HPTS fluorescence. The fluorescence increases were graded, with the change in the lamellipodia (equivalent to ∼0.4 pH units) being greater than that in the axon (∼0.15 pH units) which in turn is greater than that in the soma (∼0.08 pH units). Although the depolarization-evoked alkaline shifts were large, the pH gradient between the cell body and the end of the axon or lamellipodia collapsed relatively rapidly. The time to half-collapse of the gradient was 3.7 ± 1.4 s, calculated from regions separated by 40.8 ± 7.4 μm (n= 12). The depolarization-induced fluorescence decreases (shaded bars in Fig. 4) in the lamellipodia (equivalent to ∼-0.1 pH units) were also significantly larger (P < 0.01) than those seen in the cell body (∼-0.02 pH units). The pH gradients caused by these acid shifts collapsed significantly more slowly (P < 0.05) than those caused by alkalinization, with a time to half-collapse of 10.3 ± 3.1 s over a distance of 51.7 ± 9.3 μm (n= 6). The slow decay of this pH gradient is probably due to the relatively prolonged activity of the Ca2+-H+ pump.
Calcium changes during depolarization
Since acidifications result from calcium influx we decided to investigate the relationship between the pH microdomains and calcium by performing parallel experiments with the calcium-sensitive indicator Oregon Green. An example of such an experiment is shown in Fig. 5. The calcium transients evoked by depolarization were larger, reached a peak more rapidly, and then recovered faster in the lamellipodia (red region) than in the cell body (green) or axon (blue). The calcium transients, in any particular region, were qualitatively similar on depolarization to 0, +20 or +40 mV; however, they did appear to increase in size with larger depolarizations. Figure 6 shows mean data for the peak calcium change in the three different regions on depolarization to 0 and +40 mV for 1 s. The changes in calcium concentration show a similar trend to that seen for the changes in pH (Fig. 4).
Relationship between calcium and pH changes
To further investigate the relationship between the calcium and pH microdomains single exponentials were fitted to the falling phase of the calcium transients and the onset of the pH responses following depolarization to 0 mV for 1 s. In Fig. 7A, the time regions fitted with the exponentials are illustrated. The pH data, representing the rate of acid influx via the Ca2+-H+ pump, were fitted from the start of the depolarization. However, the calcium data were fitted from the end of the depolarization, following the closure of the voltage-gated calcium channels. Figure 7B shows the rate of acidification (reciprocal of the exponential time constant) plotted against the rate of fall of calcium. The rates of acidification were remarkably similar to the rates of calcium recovery, in any given region, supporting the suggested link between the two ionic species. Also, the rates of acidification and calcium recovery were ∼5-fold greater in the lamellipodia than in the soma. The similarity between the calcium recovery and acidification rates together with the similar regional pattern of peak pH and calcium transients (Fig. 4 and Fig. 6) support the hypothesis that a simple physical phenomenon, such as surface area to volume ratio, underlies the generation of the spatially heterogeneous ionic concentration changes.
The effect of CO2-HCO3− on pHi gradients
The generation of such calcium microdomains is well known, and relies on the slow diffusion of calcium through the cytosol. However, protons are thought to diffuse so rapidly, especially in the presence of CO2-HCO3− and the enzyme carbonic anhydrase, that such gradients in pH could not be sustained for physiologically important time periods. Stewart et al. (1999) were unable to show pH gradients, in enterocytes, in the physiological buffer CO2-HCO3−. Superfusion of snail neurones with CO2-HCO3−-buffered Ringer solution has four main effects that could shape the intracellular pH gradients seen here. Firstly, extracellular HCO3− activates pH regulation (Boron & De Weer, 1976; Thomas, 1977). Secondly, the open buffer system of the weak acid CO2 greatly increases the H+ buffering power (Roos & Boron, 1981). Thirdly, it increases the H+ mobility, and finally, by acidifying the cell, it shifts the proton equilibrium potential in a negative direction. This final effect would increase the size of the alkaline shift seen on activation of proton channels at +40 mV (Thomas & Meech, 1982). We have attempted to correct for this by reducing pHo by 0.25 pH units. To achieve this, the HCO3− concentration in the 5 % CO2 Ringer solution used was decreased from 40 to 20 mm, to produce a pHo of 7.25 rather than the normal pH of 7.50. This should compensate for the shift in the proton equilibrium potential caused by intracellular acidification on the addition of CO2 and thus minimize the magnification of the alkaline shifts seen on depolarization. The effects of 5 % CO2-20 mm HCO3− on the spatial and temporal nature of the pH gradients in isolated neurones were investigated, and are shown in Fig. 8. A transmitted light image of the patch-clamped cell is shown in Fig. 8Aa. In Fig. 8Ab, an enlarged confocal fluorescence image of the axonal lamellipodia (area delineated by the rectangle in Fig. 8Aa) is shown with two regions of interest indicated (blue and red). Fluorescence from the yellow ‘ladder’ region (Fig. 8Ab) was used to construct the linescan plots shown in the lower panels of Fig. 8C and D. The fall in fluorescence (F/F0) on addition of CO2-HCO3− (Fig. 8B), and the subsequent overshoot on its removal, are characteristic of the pHi changes induced by the weak acid CO2 in snail neurones (Thomas, 1977). Depolarization to +40 mV, in both the absence and presence of 5 % CO2, produced large relative fluorescence shifts (Fig. 8C and D) in the red region (lamellipodia) compared with the blue region (axonal foot). Figure 8C shows the spatiotemporal characteristics of the pH gradient induced by depolarization to +40 mV, in the absence of added CO2-HCO3−, for multiple regions of interest along the yellow ladder region. The pH gradient was visible for more than 7 s after the end of the depolarization. The gradient appeared to be steep at the end of the axon foot where the lamellipodia began. Little gradient was visible along the length of the axon, although the alkaline shift in the axon foot was slower and smaller than that seen in the part of the axon closer to the cell body. The magnitudes of these shifts are consistent with the apparent surface area to volume ratios of the regions. In the presence of CO2-HCO3− the pH gradient could still be seen, although it dissipated faster. In all eight cells exposed to 5 % CO2-20 mm HCO3− pH gradients were still visible in the presence of CO2. In the absence of added CO2-HCO3−, the mean peak relative shifts in fluorescence, on depolarization to +40 mV, were 16 ± 4 % in the axon and 42 ± 7 % in the lamellipodia (n= 8). Addition of 5 % CO2-20 mm HCO3− had no significant effect on these shifts (axon, 12 ± 3 %; lamellipodia, 46 ± 6 %; n= 8).
Carbonic anhydrase activity
One possible explanation for the relatively small effect of CO2-HCO3− on the pH gradients would be an absence of carbonic anhydrase activity (Stewart et al. 1999) in these isolated cells. The absence of this enzyme would reduce the buffering ability of CO2-HCO3−. Figure 9 shows an experiment in which the carbonic anhydrase inhibitor acetazolamide was used to test for the presence of the enzyme. Figure 9A shows a neurone and the regions for which data were plotted. The simultaneous application of CO2-HCO3− and acetazolamide (Fig. 9B) caused the expected fall in fluorescence, due to intracellular acidification. The mean initial linear rate of acidification caused by CO2-HCO3− in the presence of acetazolamide (-0.87 ± 0.22 % HPTS fluorescence s−1; n= 5) was significantly (P < 0.01) slower than that without acetazolamide (-1.9 ± 0.3 % HPTS fluorescence s−1; n= 11). This indicates the presence of carbonic anhydrase activity. Comparison of the spatiotemporal gradients induced by depolarization to +40 mV revealed a small increase in the pH gradient in the presence of CO2 and acetazolamide compared with Hepes-buffered Ringer solution alone (Fig. 9C and D). There may be two reasons for this. Firstly, there is a slightly larger alkaline shift in the lamellipodia. Secondly, there is a smaller and slower alkalinization at the soma. Six experiments were performed to investigate the effects of CO2-HCO3−-buffered Ringer solution containing acetazolamide on depolarization-evoked F/F0 shifts. The fluorescence shifts at the soma showed a trend towards smaller values that was not statistically significant. The effects seen in other regions were clearly variable. In two experiments the shifts in the axonal regions were larger, in two experiments they were similar and in the remaining two they were depressed compared with the shifts seen in the absence of CO2-HCO3−. In the axonal region, mean data for the peak shifts induced by depolarization to +40 mV from all six cells were 18 ± 7 % before addition of CO2-HCO3− and 15 ± 6 % in the presence of 5 % CO2-20 mm HCO3− with acetazolamide.
In the absence of acetazolamide, carbonic anhydrase appeared to be active in catalysing the conversion of CO2 to HCO3−; however, this reaction did not significantly alter the size of the rapid pH transients induced by depolarization to +40 mV. This result is not completely unexpected since others have reported a relatively weak effect of CO2-HCO3− in buffering rapid pH shifts (Amos et al. 1996). Thus, Fig. 8 and Fig. 9 show that spatially heterogeneous pH signals can exist in isolated neurones in the presence of the physiological buffer system CO2-HCO3− catalysed by carbonic anhydrase.
The depolarization-induced pH shifts we report here are larger in flattened plasma membrane-restricted regions than in the soma, even in the presence of CO2-HCO3−. Furthermore the pH gradients, which result from these pH differences, collapse over many seconds rather than milliseconds. This is an unexpected finding and is the first time that such regional pH differences have been shown in neurones.
It is unlikely that these apparent pH gradients are an artefact of our pH-sensitive dye (see Fig. 1 and Fig. 2). The pH-sensitive single wavelength fluorescence properties of HPTS, during excitation with 458 nm laser light, are consistent with those previously reported when the dye is used ratiometrically (Willoughby et al. 1998). However, the calibration of the fluorescence is more problematic. We have derived a simple equation (eqn (1)) that describes the pH shifts in terms of relative fluorescence change from a known starting pH. This equation accurately predicts in vitro data (Fig. 1B) and we have used it to calibrate some of the pH shifts reported here. These calibrations rely either on pHi being driven to thermodynamic equilibrium through proton channels (Fig. 2) or on published values for global intracellular buffering power (Fig. 8 and Fig. 9). We have also estimated the size of other depolarization-induced pHi shifts using eqn (1) by assuming an intracellular pH value of 7.1. If pHi were more alkaline, the predicted pHi shifts would be larger (Fig. 1B). It is unlikely that HPTS will have influenced the size and rate of collapse of the pH transients. We used 500 μm HPTS, which should increase the total cellular H+ buffering power by no more than 3 % (when pH equals the pK of HPTS, assuming an intrinsic buffering power of 12 mm). Since 15 % of the hydrogen ion buffering power is believed to be diffusable (al-Baldawi & Abercrombie, 1992), it is unlikely to have affected the proton diffusion coefficient dramatically.
However, the loss of these diffusable buffers through the patch pipette might enhance regional pH differences. The positioning of the patch electrode along the axon rather than the cell body had little effect on the observed pH shifts. Nevertheless we generally observed larger pH gradients as the dialysis progressed. Alternative explanations for this could be that there was: (1) a rise in the dye concentration improving the resolution of ΔpH; (2) greater space clamp as CsCl diffused into the cell; or (3) acidification of pHi, as judged by the proton equilibrium potential.
To quantify the size of the local pHi shifts we averaged pixel-based data in three cytoplasmic regions. The regions were selected because they were morphologically distinct and showed different pH and calcium concentration dynamics. The lamellipodia were characterized by the flattened extensions of plasma membrane continuous with the end of the axon (axon foot). In many cells there was no obvious boundary between these two regions. It is likely that subdivision of our classification of lamellipodia will yield larger and more rapid pH shifts. We did not investigate pH differences within the soma or at the junction of the axon and soma, although in some experiments pH gradients were visible in these locations.
The decay of the local pH shifts appears to be dependent upon the diffusional restrictions to other cellular regions with different surface area to volume ratios. This is illustrated in the lower panels of Fig. 8C and D where a steep pHi gradient can be seen, over a distance of less than 5 μm, at the junction of the axon end and the lamellipodia. The complex three-dimensional cellular morphology makes it difficult to arrive at a meaningful statistical analysis of the spatiotemporal characteristics of the pH shifts. For this reason we have reported some of our observations with figures and measurements which illustrate the common findings.
The relatively weak effect of CO2-HCO3− on the depolarization-induced regional pH shifts, despite the presence of the enzyme carbonic anhydrase, supports the hypothesis that these shifts may occur in vivo. The reason why pH gradients are visible in neurones, even in the presence of CO2, is that they have a complex morphology. There are large transmembrane acid fluxes that have little effect on pHi in the soma since the cytosolic volume is large. However, in regions with high surface area to volume ratios, pHi changes can still occur. The effect of mobile pH buffers, in collapsing the pH differences between regions, is restricted by the relative lack of volume. The neuronal soma, where observation of pHi gradients is the most difficult, is morphologically similar to the enterocyte, where pHi gradients have not been observed in the presence of CO2-HCO3−(Stewart et al. 1999).
The pH gradients dissipate more rapidly following alkalinization when compared with acidification. A likely explanation for this difference is that alkalinization, caused by the rapid efflux of protons through proton channels, is followed by an influx of protons via the Ca2+-H+ pump. This influx of protons tends to neutralize the alkaline pH shifts locally, thereby collapsing the gradient more rapidly than diffusion of protons through the cytosol alone. Acidification resulting from calcium extrusion via the Ca2+-H+ pump is not balanced by an alkaline flux. Hence the collapse of these acidic pH gradients is slow, depending largely upon cytosolic diffusion. Snail neurones are apparently unusual amongst nerve cells in that they show proton channel activity. It is unlikely that other neurones will show these large, proton channel-dependent alkaline shifts. However, it is already well known that there are activity-dependent acidic pH shifts in neuronal cell bodies (Trapp et al. 1996a,b) and it appears possible that there will be even larger acid shifts in dendritic regions where there is substantial Ca2+-H+ pump activity and a high surface area to volume ratio. The correlation between the dynamics of calcium extrusion and acidification in the various microdomains (Fig. 7) provides strong evidence for the Ca2+-H+ pump acting as a local modulator of pH. Such microdomains with a sustained pH difference, resulting from calcium efflux, are potential short-term modulators of local function.
In conclusion, substantial differences in pH may exist in neuronal cytosolic domains as a result of electrical activity, even in the presence of physiological buffers. The role of the large pH transients is currently unknown, but it is likely that some processes previously ascribed to calcium or the passage of time may be due to these local pH signals.
We thank the MRC for financial support and the MRC cellular calcium homeostasis co-operative (G9900182) for critical comments and assistance.