Effect of extracellular acid–base disturbances on the intracellular pH of neurones cultured from rat medullary raphe or hippocampus

Abstract

Previous reports suggest that an important characteristic of chemosensitive neurones is an unusually large change of steady-state intracellular pH in response to a change in extracellular pH (ΔpH_i/ΔpH_o). To determine whether such a correlation exists between neurones from the medullary raphe (a chemosensitive brain region) and hippocampus (a non-chemosensitive region), we used BCECF to monitor pH_i in cultured neurones subjected to extracellular acid–base disturbances. In medullary raphe neurones, respiratory acidosis (5% → 9% CO₂) caused a rapid fall in pH_i (ΔpH_i ∼0.2) with no recovery and a large ΔpH_i/ΔpH_o of 0.71. Hippocampal neurones had a similar response, but with a slightly lower ΔpH_i/ΔpH_o (0.59). We further investigated a possible link between pH_i regulation and chemosensitivity by following the pH_i measurements on medullary raphe neurones with an immunocytochemistry for tryptophan hydroxylase (a marker of serotonergic neurones). We found that the ΔpH_i/ΔpH_o of 0.69 for serotonergic neurones (which are stimulated by acidosis) was not different from either the ΔpH_i/ΔpH_o of 0.75 for non-serotonergic neurones (most of which are not chemosensitive), or from the ΔpH_i/ΔpH_o of hippocampal neurones. For both respiratory alkalosis (5% → 3% CO₂) and metabolic alkalosis (22 mm → 35 mm HCO₃⁻), ΔpH_i/ΔpH_o was 0.42–0.53 for all groups of neurones studied. The only notable difference between medullary raphe and hippocampal neurones was in response to metabolic acidosis (22 mm → 14 mm HCO₃⁻), which caused a large pH_i decrease in ∼80% of medullary raphe neurones (ΔpH_i/ΔpH_o = 0.71), but relatively little pH_i decrease in 70% of the hippocampal neurones (ΔpH_i/ΔpH_o = 0.09). Our comparison of medullary raphe and hippocampal neurones indicates that, except in response to metabolic acidosis, the neurones from the chemosensitive region do not have a uniquely high ΔpH_i/ΔpH_o. Moreover, regardless of whether neurones were cultured from the chemosensitive or the non-chemosensitive region, pH_i did not recover during any of the acid–base stresses.

In mammals, an increase in arterial CO₂ partial pressure (P_CO2) is the most powerful stimulus for ventilation, acting through the peripheral and central chemoreceptors (for reviews, see Nattie, 1999; Richerson et al. 2001). The prevailing view is that changes in arterial P_CO2 produce their effects by inducing changes in the intracellular pH (pH_i) of chemoreceptor cells (Peers & Buckler, 1995; Wiemann et al. 1998; Filosa et al. 2002; Wang et al. 2002). The peripheral chemoreceptors are in the carotid and aortic bodies, which contain chemosensitive type I or glomus cells with neuronal properties. The identities of the central chemoreceptor neurones have not yet been unequivocally defined. Chemosensitive neurones are present in many brainstem nuclei that are linked to respiratory control, including the ventrolateral medulla (VLM), nucleus of the tractus solitarius (NTS), medullary raphe, locus coeruleus, and the hypothalamus (Richerson, 1998; Nattie, 1999). Indeed, in many of these regions, inducing local acidosis causes ventilation to increase (Nattie, 1999). It remains to be proven which of these chemosensitive neurones are responsible for the normal ventilatory response to small, physiological changes in pH/CO₂, but accmulating evidence suggests that serotonergic neurones within the medullary raphe nuclei are likely to play an important role (Richerson et al. 2001; Wang et al. 2001; Richerson, 2004).

Most types of cells, when subjected to a respiratory acid–base disturbance (a pH change produced by a change in P_CO2) or a metabolic acid–base disturbance (a pH change produced by a change in [HCO₃⁻] at a fixed P_CO2), exhibit a pH_i change that is 20–30% as large as the change in extracellular pH (pH_o) (Ellis & Thomas, 1976; Vaughan-Jones, 1986; Tolkovsky & Richards, 1987; Glunde et al. 2002). By contrast, the chemosensitive type I cells of the carotid body respond to the above-mentioned acid–base disturbances with unusually large changes in pH_i (Buckler et al. 1991), having a ΔpH_i/ΔpH_o ratio of 60–70% without recovery during sustained exposure. Similarly high ΔpH_i/ΔpH_o ratios have also been reported in subsets of neurones in the NTS and locus coeruleus (Richerson, 1998; Ritucci et al. 1998; Filosa et al. 2002), two regions that contain putative central chemoreceptor neurones, as well as in chemoreceptor neurones of the pulmonate terrestrial snail (Goldstein et al. 2000). Consult the review by Putnam (2001) for a discussion of pH_i regulation in neurones in chemosensitive brain regions, and the review by Chesler (2003) for a more general discussion of pH regulation in the brain.

The above observations have led to the concept that the absence of a pH_i recovery ensures that the primary stimulus – intracellular acidosis – continues to drive ventilation as long as the respiratory acidosis persists (i.e. the steady-state ΔpH_i/ΔpH_o is large). The implicit assumption is that, when subjected to sustained respiratory acidosis, the normal response of non-chemosensitive neurones is to return pH_i nearly to baseline levels in an attempt to stabilize the intracellular milieu (i.e. the steady-state ΔpH_i/ΔpH_o is small).

In determining whether putative central chemoreceptor neurones have a unique pH_i response to extracellular acid–base disturbances, it is important to compare this response to that of non-chemosensitive neurones. Such an analysis is the purpose of the present study, in which we have examined the pH_i response to acid–base disturbances in neurones from the medullary raphe versus the hippocampus. We chose to work with cultured neurones because (1) the preparation is ideal for monitoring pH_i while changing extracellular composition rapidly and effectively, and (2) cultured medullary raphe neurones retain the electrophysiological response to acidosis first identified in slices (Wang et al. 1998). Thus, working with cultured neurones would allow us to compare our pH_i data directly with electrophysiological data previously obtained on cultured medullary raphe and hippocampal neurones under the same conditions (Wang & Richerson, 2000). In culture, a subset (∼20%) of medullary raphe neurones responds to respiratory acidosis (a shift from 5% to 9% CO₂) by tripling their firing rate; these neurones are all serotonergic (Richerson, 1995; Wang et al. 2001). A second subset of medullary raphe neurones (∼20%) responds to respiratory acidosis in the opposite manner, decreasing firing rate to an average of 20–50% of control (Wang et al. 2001); none of these neurones is serotonergic. These two classes of medullary raphe neurones are excellent candidates for contributing to the physiological ventilatory response because of their extreme pH sensitivity. In addition, the chemosensitive serotonergic neurones are closely apposed to large arteries in the ventral medulla, where they could faithfully monitor arterial P_CO2 (Bradley et al. 2002). A third subset (∼60%) of medullary raphe neurones does not change firing rate in response to acidosis. Here we compared the pH_i responses of medullary raphe neurones with those of hippocampal neurones, none of which exhibit a large change in firing rate in response to acidosis (Wang & Richerson, 2000).

We know of no systematic studies of the effects of respiratory and metabolic acidosis/alkalosis on steady-state pH_i in either medullary raphe or hippocampal neurones. Working on organotypic cultures of rat medullary neurones, Wiemann and colleagues (Wiemann et al. 1998, 1999; Wiemann & Bingmann, 2001) found that respiratory acidosis generally causes a modest but sustained fall in pH_i. In the present study, we used the fluorescent pH-sensitive dye BCECF and a video-imaging system to monitor pH_i in primary neuronal/glial cultures subjected to extracellular respiratory acidosis and alkalosis as well as extracellular metabolic acidosis and alkalosis. In some experiments, we used immunocytochemistry to determine which neurones studied in the imaging experiments were serotonergic. We made the surprising observation that, in response to respiratory acid–base disturbances, as well as metabolic alkalosis, all medullary raphe and hippocampal neurones have similarly large ΔpH_i/ΔpH_o values, and neither group exhibits a significant pH_i recovery during any of these three stresses. Moreover, among medullary raphe neurones, the responses of serotonergic and non-serotonergic neurones to respiratory disturbances were indistinguishable. The major exception to this pattern of uniformity was the response to metabolic acidosis: about 80% of medullary raphe neurones underwent a large, sustained, and reversible pH_i decrease; the rest exhibited only a small pH_i decrease. Conversely, less than 30% of hippocampal neurones underwent the large, sustained and reversible response; the majority exhibited only a small pH_i decrease from which they rebounded after the removal of the insult. Based on our comparison of medullary raphe and hippocampal neurones, we conclude that a high ΔpH_i/ΔpH_o and a lack of pH_i recovery are not unique to chemosensitive neurones.

Methods

Cell culture

Experiments were performed on cell cultures prepared from the ventromedial medulla or hippocampus from neonatal (P0–P2) Sprague-Dawley rats. The methods for collecting tissues have been previously described (Wang et al. 1998; Wang & Richerson, 2000). The Yale University Animal Care and Use Committee approved all procedures. Briefly, rats were killed by decapitation and the medulla was rapidly removed in an aseptic environment. A region was microdissected that contained neurones and glia from the raphe pallidus, raphe magnus, a portion of the raphe obscurus and tissue immediately adjacent to these nuclei. Hippocampal neurones were obtained from an entire hippocampus removed from one hemisphere. The collected tissues were placed in oxygenated Ringer solution with the following composition (mm): NaCl, 130; KCl, 4; MgCl₂, 1; CaCl₂, 1.5; Hepes, 10; dextrose, 10; titrated to pH 7.3 with NaOH. Dissected tissue was digested with papain, triturated, and plated onto poly l-ornithine- and laminin-coated coverslips. Cells were fed with glial-conditioned medium (10% fetal bovine serum (FBS) in 60% modified Eagle's medium (MEM) + 40% Neurobasal medium with B27 supplement + penicillin–streptomycin; conditioned for 1 day by ventromedial medullary cultures). Basic fibroblast growth factor (bFGF; 0.1–1 ng ml⁻¹) and fibroblast growth factor 5 (FGF-5; 1–10 ng ml⁻¹) were also added to the culture medium to enhance survival. Cells were first fed on days 4–7 with a half-volume change of Neurobasal–B27, supplemented with cytosine β-d-arabino-furanoside hydrochloride (Ara-C) (3 µm) to inhibit glial growth. They were then fed every 1–2 weeks with a half-volume change of Neurobasal–B27. Cultures were maintained in an incubator at 37°C and 5% CO₂. This protocol leads to cultures that consist of a bed of glia, with neurones scattered on the surface. In the case of the hippocampal neurones, recordings were made after cells were cultured for at least 11 days, and up to 53 days (average, 30 days). Hippocampal neurones achieve strong synaptic coupling within 14 days of culture. In the case of the medullary raphe neurones, the neurones were cultured for at least 14 days, and up to 50 days (average, 31 days). Medullary raphe neurones achieve a mature electrophysiological response to acid–base disturbances within 21 days in culture (Wang & Richerson, 1999). However, the responses of the five medullary raphe neurones that we studied after fewer than 21 days in culture were indistinguishable from those of other medullary raphe neurones. Culture media, including FBS, bFGF, Neurobasal medium and B27 supplement were purchased from Life Technologies, Inc. (Rockville, MD, USA). MEM (no. 56419) was purchased from JRH Biosciences (Lenexa, KS, USA). All other salts and chemicals were purchased from Sigma Chemical Co. (St Louis, MO, USA).

Solutions

The compositions of the physiological solutions are given in Table 1. Osmolalities, measured using a vapour-pressure osmometer (Model 5100C, Wescor, Inc., Logan, UT, USA), were adjusted to 300 ± 3 mosmol kg⁻¹ by adding water or NaCl. All solutions were titrated to the indicated pH at 37°C. Solution 1 was gassed with 100% O₂ for 30–40 min. The other solutions were similarly gassed with a certified gas mixture (Airgas East, Cheshire, CT, USA) containing the indicated level of CO₂, balance O₂. Solutions were delivered at 7 ml min⁻¹ using a syringe pump (Model 55-2222, Harvard Apparatus, South Holliston, MA, USA) to drive 140 ml plastic syringes (Monoject 140 cc, Sherwood Medical Industries Ltd, Ballymoney, N. Ireland), which exhibit a negligible loss of CO₂ over 24 h. The output of the syringe was connected to an array of three-way valves mounted in series with Tygon tubing (R-3603, 1/16 inches i.d. × 1/8 inches o.d., Norton Performance Plastics Corp., Akron, OH, USA). The output of the valves was connected to stainless-steel tubing surrounded by a water jacket that warmed the solutions sufficiently to yield 37°C in the chamber. Using a P_CO2 electrode as previously described (Zhao et al. 2003), we found no significant loss of CO₂ from our perfusion system.

Table 1.

Physiological solutions

	1	2	3	4	5	6
Components	Hepes	Standard equilibrated CO2–HCO3−	Respiratory acidosis	Respiratory alkalosis	Metabolic acidosis	Metabolic alkalosis
NaCl	146	124	124	124	136	115
NaH2PO4	1.3	1.3	1.3	1.3	1.3	1.3
NaHCO3	0	22	22	22	13.9	34.9
KCl	3	3	3	3	3	3
MgCl2	2	2	2	2	2	2
CaCl2	2	2	2	2	2	2
Glucose	10	10	10	10	10	10
CO2 (%)	0	5	9	3	5	5
Hepes	30	0	0	0	0	0
pH	7.40	7.40	7.16	7.62	7.19	7.61

The concentrations are millimolar except for CO₂ (given as a percentage).

We introduced the pH calibration solutions (which contained nigericin for the BCECF calibration – see below) to the chamber using an auxiliary gravity-fed solution-delivery system that was completely separate from the primary solution delivery system used for the physiological solutions. This precaution avoided contamination of the plumbing fixtures by nigericin. After each experiment, we washed the chamber extensively with 70% ethanol in water to remove traces of nigericin (Bevensee et al. 1999).

Fluorescence measurements

Prior to the fluorescence measurements, we washed the coverslip with cells attached with the Hepes-buffered solution 1 (Table 1) to replace the culture medium, and then transferred the coverslip to an air incubator at 37°C for ∼20 min. After removing the coverslip from the incubator, we used vacuum grease to attach the coverslip to the bottom of a perfusion chamber, and secured the assembly on the stage of an inverted microscope (Olympus IX70) equipped for epifluorescence. We then loaded the neurones with the pH-sensitive dye 2′,7′-bis-2-carboxyethyl)-5(and-6)carboxyfluorescein (BCECF) by incubating the cells at room temperature for ∼15 min in our Hepes-buffered solution 1 containing 5 µm of the esterified form of BCECF (BCECF-AM, Molecular Probes Inc., Eugene, OR, USA). Unhydrolysed BCECF-AM was washed from the cells by flowing solution 1 (37°C) through the chamber for approximately 3–4 min (i.e. ∼21–28 ml of fluid) before starting the fluorescence measurements.

Under microscopic observation (using a × 40 oil-immersion objective, NA 1.35, with a 1.5 × magnification-selector knob), we selected a field that contained one to four pyramidal neurones with distinct processes, homogeneous cytoplasm, and a three-dimensional appearance when we focused through the soma using differential-interference contrast optics. The fluorescence excitation light source was a 75 W xenon arc lamp; two excitations wavelengths of 440 nm and 490 nm were obtained by mounting two excitations filters (440 ± 5 nm and 495 ± 5 nm, Omega Optical Inc., Brattleboro, VT, USA) on a filter wheel (Ludl Electronic Products Ltd, Hawthorne, NY, USA) in the excitation light path. We also used an appropriate neutral-density filter (Omega Optical Inc.), mounted on a second wheel, to equalize as nearly as possible the emitted light and avoid over-illumination of the cells. The excitation light was directed to the cells via a long-pass dichroic mirror (DM 510, Omega Optical Inc.) and the objective mentioned previously. The emitted light was collected by the same objective and, via a band-pass filter (530 ± 35 nm, Omega Optical Inc.), was directed to an intensified CCD camera (Model 350F, Video Scope International LTD, Dulles, VA, USA). A typical data-acquisition cycle consisted of an illumination period of 370 ms at 440 nm (yielding the emitted light intensity I₄₄₀) and then 370 ms at 490 nm (yielding I₄₉₀). During the collection time we averaged four successive video frames using an image-processing board (DT3155, Data Translation, Marlboro, MA, USA). The rate of acquisition varied from once every 2.5 s to once every 20 s; a shutter on the filter wheel protected the cells from the light between data-acquisition cycles. The data acquisition was controlled by software developed in our laboratory using the Optimas (Media Cybernetics, Inc., Silver Spring, MD, USA) platform. The software allowed us to draw manually an area of interest (AOI) that surrounded only the soma of the cells. The pixel intensity of the AOI at I₄₉₀ (background subtracted) was divided by the pixel intensity of the AOI at I₄₄₀ (background subtracted). The fluorescence-excitation ratio I₄₉₀/I₄₄₀ is a sensitive indicator of pH_i, with relatively little contamination by changes in parameters such as dye concentration.

We computed the pH_i values by using the high-K⁺–nigericin technique (Thomas et al. 1979), as modified in our laboratory to exploit a one-point calibration at pH_i 7.00 (Boyarsky et al. 1988). Briefly, at the end of each experiment we applied a solution containing a H⁺ ionophore (nigericin) in a high-K⁺ solution buffered at pH 7.00, which clamps pH_i to 7.00. The entire data set of I₄₉₀/I₄₄₀ ratios from a single cell was then normalized to the I₄₉₀/I₄₄₀ for that cell obtained at pH_i 7.00. We used the following equation to calculate pH_i:

In a separate series of experiments, we exposed neurones to high-K⁺–nigericin solutions at 10 different pH values ranging from 5.8 to 8.5 (Boyarsky et al. 1988), and thereby obtained the pK (negative log of the molar dissociation constant) and the b values. We used a non-linear least-squares method to fit the parameters in the above equation, which forces the best-fit curve to pass through unity at pH_i = 7.00, to the calibration data. The best-fit values were pK = 7.32 ± 0.01 (s.d.) and b = 2.65 ± 0.02 (s.d.), for the medullary raphe neurones, and pK = 7.34 ± 0.004 (s.d.) and b = 2.56 ± 0.007 (s.d.) for the hippocampal neurones.

Immunocytochemistry

Recently, Wang et al. (2001) showed that all medullary raphe neurones stimulated by respiratory acidosis are serotonergic. In one series of experiments, we compared the pH_i responses of serotonergic and non-serotonergic raphe neurones. For these experiments, we cultured the neurones on coverslips etched with a grid (CELLocate®, Eppendorf, Westbury, NY, USA) for later identification by immunostaining. Before beginning to monitor pH_i, we used the CCD camera to capture images of the neurones and coverslip grid in the field. At the end of the experiment, cells were fixed by immersion in 30% ethanol (EtOH)–1% acetic acid, followed by 60% EtOH–1% acetic acid, and finally 95% EtOH–1% acetic acid for 30 min each. Neurones were then incubated at 4°C for 24–48 h with a mouse monoclonal antibody (Sigma, St Louis, MO, USA; 1/2000) directed against the enzyme tryptophan hydroxylase (TPH). The presence of TPH was visualized using the peroxidase method, using diaminobenzidine as the chromogen, as previously described (Wang et al. 2001).

Data analysis

Excessive illumination of intracellular dyes can lead to the formation of free radicals that harm the cell, leading to increased membrane leakiness and cell death. Therefore, as previously reported, we used the fractional rate of loss of intracellular BCECF (−k₄₄₀), computed from the change in I₄₄₀ over time, as an index of cell viability (Bevensee et al. 1995). For a full discussion of the −k₄₄₀ technique, including the fluctuations in −k₄₄₀, see Bevensee et al. (1995). In the present study, we rejected neurones with a −k₄₄₀ value > 5% min⁻¹. Figure 1 shows the time courses of I₄₉₀/I₄₄₀ and −k₄₄₀ for a single medullary raphe neurone. Initially, with the neurone in a Hepes buffer, I₄₉₀/I₄₄₀ is fairly stable and −k₄₄₀ is near zero, indicating a very low rate of dye loss and thus a healthy neurone. Introducing 5% CO₂–22 mm HCO₃⁻ causes a rise in I₄₉₀/I₄₄₀ (consistent with a rise in pH_i). After the subsequent switch to 9% CO₂–22 mm HCO₃⁻, I₄₉₀/I₄₄₀ at first falls (consistent with the expected fall in pH_i) before beginning to rise, at first slowly, and then far more rapidly. This apparent rise in pH_i is, in fact, an artifact caused by the loss of the cell membrane integrity, which is reflected by the precipitous increase of −k₄₄₀. Without this criterion of cell viability, a pH_i recovery that occurs during respiratory acidosis could be interpreted as a compensatory acid-extrusion mechanism in response to the fall in pH_i. We found that the chances of observing such a I₄₉₀/I₄₄₀ pattern (which presumably reflects photodynamic damage to the neurone) increased with the number of data-acquisition cycles and/or the frequency of data acquisition. By limiting the frequency of data-acquisition in lengthy experiments, we were able to maintain values of −k₄₄₀ that were relatively low.

Figure 1. Time course of −k₄₄₀ in a medullary raphe neurone that loses its membrane integrity.

A, record of I₄₉₀/I₄₄₀. At the indicated time, we switched the extracellular fluid from a Hepes buffer (solution 1 in Table 1) to 5% CO₂–22 mm HCO₃⁻ (solution 2 in Table 1), and then to 9% CO₂–22 mm HCO₃⁻ (solution 3 in Table 1). Increasing values of I₄₉₀/I₄₄₀ correspond to increasing pH_i. We did not convert the I₄₉₀/I₄₄₀ values to pH_i values because the cell perished before we could perform the calibration, which normally would have occurred at the end of the experiment. B, record of −k₄₄₀, the rate constant describing the decrease in I₄₄₀ (an index of intracellular dye concentration). The −k₄₄₀ record is for the same neurone as in A. Note that −k₄₄₀ remained below our standard limit of 5% min⁻¹ until late in the experiment, when −k₄₄₀ began to rise dramatically, reflecting a sudden loss of dye, which in turn was due to a leaky cell membrane.

Statistics

Data are reported as mean ± s.e.m., followed by the number of cells (n), and the number of coverslips (N). The s.e.m. values were computed on the basis of n − 1. Means were compared using, as indicated, Student's paired and unpaired t tests (two tails). P < 0.05 was considered significant. In the regression analyses, the slopes and the intercept of linear fits were compared using Student's t test (Kleinbaum et al. 1988). The test statistic was the difference between the slopes (or x-intercepts) divided by the standard deviation of the estimated difference between the slopes (or x-intercepts).

Results

Steady-state pH_i in absence and presence of CO₂–HCO₃⁻ buffer

At the beginning of most experiments, we bathed the neurones in a Hepes-buffered solution (solution 1 in Table 1). In the absence of CO₂–HCO₃⁻, the mean steady-state pH_i was 7.17 ± 0.02 (n = 183, N = 87) for medullary raphe neurones and 7.03 ± 0.03 (n = 49, N = 29) for hippocampal neurones. The difference between the two mean steady-state pH_i values was statistically significant (P < 0.0007, unpaired two-tail t test). In a study of freshly dissociated hippocampal neurones, Bevensee et al. (1996) found that the distribution of pH_i values was unimodal for neurones isolated from ‘immature’ rats, but bimodal for neurones isolated from ‘mature’ rats. As shown in Fig. 2A, the frequency distributions of the two populations of neurones in the present study were unimodal.

Figure 2. Initial steady-state pH_i values.

A, distribution of steady-state pH_i values in a Hepes-buffered solution. The histogram (bin width = 0.1pH units) represents the steady-state pH_i values of 183 medullary raphe neurones (blue columns) and of 49 hippocampal neurones (red columns) measured during an exposure to solution 1 (Table 1). B, change in steady-state pH_i produced upon switching the bath from a Hepes- (solution 1) to a 5% CO₂–22 mm HCO₃⁻-buffered solution (y-axis) versus the initial pH_i in Hepes (x-axis). The blue (medullary raphe neurones) and red (hippocampal neurones) lines represent linear regression fits. A parallelism analysis according to Kleinbaum et al. (1988) shows that the difference between the two slopes is statistically significant (P < 0.01).

Switching the extracellular solution from one buffered with Hepes to one buffered with 5% CO₂–22 mm HCO₃⁻ (solution 2), at a fixed pH_o of 7.40, caused an immediate fall in pH_i, followed by a variable pH_i recovery (i.e. alkalinization) to a final steady-state value (not shown). For both types of neurones, the general trend was for the ΔpH_i (i.e. difference between final and initial values) to fall with increasing initial pH_i (Fig. 2B). The dependence of ΔpH_i on initial pH_i was steeper for the hippocampal neurones. The difference between the slopes of the two linear fits was statistically significant (P < 0.01) for the two populations of neurones, but the difference between the x-intercepts was not (P = 0.13).

The relationship between initial pH_i versus ΔpH_i for hippocampal neurones (Fig. 2B, blue symbols) is nearly identical to previous data from freshly dissociated rat hippocampal neurones (Schwiening & Boron, 1994; Bevensee et al. 1996; Brett et al. 2002) consistent with the idea that pH_i regulation is similar in cultured and freshly dissociated neurones.

Respiratory acidosis

In the following experiments, we lowered extracellular pH (pH_o) from 7.40 to ∼7.17 by increasing the P_CO2 from 5% to 9% CO₂, all at a constant [HCO₃⁻] of 22 mm.

Medullary raphe neurones

The blue record in the upper panel of Fig. 3A shows the effect of respiratory acidosis on the pH_i of a single medullary raphe neurone. We initially superfused the medullary raphe neurones with a Hepes-buffered solution (solution 1), and then switched to the solution containing 5% CO₂–22 mm HCO₃⁻ (solution 2) at a fixed pH_o of 7.40. As is typical for a neurone with a relatively low initial pH_i, this switch caused steady-state pH_i to increase. In this particular example, the transient CO₂-induced acidification is not apparent because of the low data-acquisition rate (one sample every 20 s). After ∼5 min, we switched to the solution equilibrated with 9% CO₂–22 mm HCO₃⁻ (solution 3 in Table 1), and found that pH_i rapidly fell, presumably reflecting the diffusion of CO₂ into the cell and the subsequent formation of HCO₃⁻ and H⁺. Even though we maintained the exposure to 9% CO₂ for 10 min, pH_i at most recovered only very slightly during the plateau phase in this neurone. The blue record Fig. 3B shows the normalized pH_i time course for a series of such experiments.

Figure 3. Effect of respiratory acidosis/alkalosis on medullary raphe and hippocampal neurones.

A–C, respiratory acidosis. A, representative pH_i records obtained from a single medullary raphe neurone (blue) and from a single hippocampal neurone (red), subjected to respiratory acidosis. These data are from two separate experiments with an identical sequence of solution changes, but slightly different timings. At the beginning of each experiment, the cells were bathed in a Hepes-buffered solution (solution 1 in Table 1). At the indicated time, the neurones were sequentially exposed to 5% CO₂–22 mm HCO₃⁻ (solution 2), 9% CO₂–22 mm HCO₃⁻ (solution 3), and then back to 5% CO₂–22 mm HCO₃⁻. The two records are temporally aligned at the switch from 5% to 9% CO₂ (vertical dashed line). The record from the medullary raphe neurone ends with a one-point nigericin calibration at pH_o 7.00 (see Methods). The lower portion of A shows the rate of dye loss (−k₄₄₀) for the same two neurones. B, time course of the mean change in pH_i (ΔpH_i) for 46 medullary raphe neurones (blue) on 14 coverslips, and for 20 hippocampal neurones (red) on 8 coverslips, induced by switching the bath from 5% to 9% CO₂. For each neurone, we obtained the ΔpH_i values by computing the mean steady-state pH_i during the last minute of the initial exposure to 5% CO₂–22 mm HCO₃⁻ (brace labelled ‘a’), and then subtracting from this mean the actual pH_i at each point during the exposure to 9% CO₂. For the blue record, each point represents 46 medullary raphe neurones at the beginning, gradually decreasing to 24 neurones at the end; for the red record, the comparable numbers are 20 and 9 hippocampal neurones. The vertical bars represent s.e.m. values. C is similar to B, except that it represents mean ΔpH_i for TPH⁺ (black) and TPH⁻ (pink) medullary raphe neurones. Each black point represents 4–10 TPH⁺ neurones and each pink point represents 45–64 TPH⁻ neurones. D–F, respiratory alkalosis. The plots are similar to those in A–C, except that the exposure to 9% CO₂ in A–C is replaced in D–F by an exposure to 3% CO₂–22 mm HCO₃⁻ (solution 4 in Table 1). Each point in E represents 8–31 medullary raphe neurones, or 17 hippocampal neurones; each point in F represents 6–15 TPH⁻ neurones and TPH⁺ 4–5 neurones.

As summarized in Table 2, raising P_CO2 from 5% to 9% caused the mean pH_i to fall from 7.06 to 6.89 during the first 2–3 min. (For each neurone, we computed the mean pH_i value three times over a period of 1 min, i.e. 4 samples, 20 s apart. The first period ended just before switching to 9% CO₂. The second period began 2–3 min after the solution change, after pH_i had stabilized. The third period began 1 min prior to switching back to 5% CO₂.) At the end of the 10 to 12 min exposure to 9% CO₂, the mean pH_i was also 6.89. Thus, pH_i did not recover from the acidosis. The mean steady-state ΔpH_i/ΔpH_o was 0.71 (Table 2).

Table 2.

Steady-state pH_i and ΔpH_i/ΔpH_o in medullary raphe and hippocampal neurones during respiratory acidosis and alkalosis

As shown by the blue record in the upper panel of Fig. 3A, switching the bath solution from 9% CO₂ back to 5% CO₂–22 mm HCO₃⁻ caused pH_i to return to a value only slightly higher than the previous value in 5% CO₂. The mean pH_i after the return to 5% CO₂ was 7.07 (Table 2), which is not significantly different from the previous value of 7.06. The blue record in the upper panel of Fig. 3A also shows a typical one-point nigericin calibration at pH_i 7.00 (see Methods). The blue record in the lower panel of Fig. 3A shows that the rate of dye loss during the experiment did not exceed 5% min⁻¹, indicating that the neurone was healthy (Bevensee et al. 1995).

Hippocampal neurones

In order to test the hypothesis that the absence of pH_i recovery and a large ΔpH_i/ΔpH_o is a hallmark for chemosensitivity, we repeated the protocol described above with neurones from a non-chemosensitive area of the brain – the hippocampus. As shown by the red records in Fig. 3A and B, and summarized in Table 2, the response of hippocampal neurones to respiratory acidosis was very similar to that of medullary raphe neurones. (In this neurone, the data-acquisition rate during the switch from the Hepes to the CO₂–HCO₃⁻ solution was one sample every 2.5 s, fast enough to allow us to observe the transient CO₂-induced acidification.) In particular, the pH_i recovery during the exposure to 9% CO₂ was negligible over a period of nearly 10 min. Moreover, the ΔpH_i/ΔpH_o was very high, although somewhat smaller than for the medullary raphe neurones (0.59 versus 0.71, P < 0.01).

Medullary raphe neurones: comparison of TPH⁺ and TPH⁻ neurones

Not all medullary raphe neurones are chemosensitive. Respiratory acidosis increases the firing rate of ∼20% of cultured medullary raphe neurones, decreases the firing rate of another ∼20%, and has no effect on ∼60% (Wang et al. 2001, 2002). A question that arises is whether the pH_i of chemosensitive and insensitive medullary raphe neurones respond differently to respiratory acidosis. Wang et al. (2001) showed that the subpopulation of medullary raphe neurones that respond to respiratory acidosis with increased firing are all immunopositive for tryptophan hydroxylase (TPH) – a marker for serotonergic neurones. Therefore, in experiments entirely separate from those described above, we determined whether the pH_i response of TPH-positive (TPH⁺) medullary raphe neurones is different from that of their TPH-negative (TPH⁻) neighbours. Before commencing the pH_i measurements, we determined the coordinates of individual neurones on a grid coverslip. We then subjected the neurones to the respiratory acidosis protocol in Fig. 3A. At the end of the experiment, we processed the coverslips for immunostaining, and used an antibody against TPH (see Methods) to categorize the neurones as either TPH⁺ or TPH⁻.

As summarized in Fig. 3C, the mean ΔpH_i time courses were similar for TPH⁺ (black record) and TPH⁻ (pink record) medullary raphe neurones. As summarized in Table 2, both populations had similarly high ΔpH_i/ΔpH_o values. The mean pH_i 2–3 min after the switch to 9% CO₂ was 7.14 versus 7.04 and the mean pH_i 10 min later was 7.12 versus 7.04 (for TPH positive versus negative, respectively). Moreover, after the first 2–3 min of exposure to 9% CO₂, neither population exhibited a statistically significant pH_i drift (P = 0.7 and P = 0.4, respectively).

Respiratory alkalosis

A second approach that we used was to increase pH_o by lowering P_CO2 from 5% to 3% at a constant [HCO₃⁻] of 22 mm.

Medullary raphe neurones

The blue record in the upper panel of Fig. 3D shows the effect of respiratory alkalosis on the pH_i of a single medullary raphe neurone. As in the experiment discussed earlier (Fig. 3A), we began by switching the extracellular solution from one buffered by Hepes to one buffered by 5% CO₂–22 mm HCO₃⁻. As is typical for a neurone with a relatively high pH_i, this switch caused steady-state pH_i to decrease. After letting the cell stabilize for ∼5 min, we changed to the solution equilibrated with 3% CO₂–22 mm HCO₃⁻ (solution 4 in Table 1), which caused pH_i to rise by nearly 0.20 and then – in this neurone – drift lower by ∼0.07. The blue record in Fig. 3E summarizes normalized data from a series of similar experiments, and shows that respiratory alkalosis caused pH_i to increase rapidly and then stabilize, with no recovery (i.e. no pH_i decrease).

As summarized in Table 2, lowering P_CO2 from 5% to 3% caused the mean pH_i to increase from 7.14 to 7.25 within the first 2–3 min, and then stabilize (pH_i= 7.24). The mean steady-state ΔpH_i/ΔpH_o was 0.51 (Table 2).

As shown by the blue record in Fig. 3D, switching the solution from 3% CO₂ back to 5% CO₂ caused pH_i to return to a value close to the previous baseline before the exposure to 3% CO₂. The blue record in the lower panel of Fig. 3D shows that, while the neurone was exposed to CO₂–HCO₃⁻, the rate of dye loss did not exceed 5% min⁻¹.

Hippocampal neurones

We next applied the same protocol to hippocampal neurones. As shown by the red records in Fig. 3D and E, and summarized in Table 2, the response of hippocampal neurones to respiratory alkalosis was very similar to that of the medullary raphe neurones. In particular, the mean pH_i showed no evidence of recovery during the period of respiratory alkalosis, and the mean ΔpH_i/ΔpH_o value for the hippocampal neurones (0.53) was not different from that of the medullary raphe neurones (0.51).

Medullary raphe neurones: comparison of TPH⁺ and TPH⁻ neurones

Recall that Fig. 3C summarized the effect of respiratory acidosis on TPH⁺ and TPH⁻ neurones; Fig. 3F does the same for respiratory alkalosis. In the experiments summarized in Fig. 3F, we used the same protocol described in Fig. 3D, but cultured the neurones (a different set of neurones from those used in Fig. 3D and E) on grid coverslips. As shown in Fig. 3F, the time course of the change in pH_i was very similar for TPH⁺ neurones (black symbols) and TPH⁻ neurones (pink symbols): a sustained alkalinization during respiratory alkalosis without significant pH_i recovery (i.e. acidification). The ΔpH_i/ΔpH_o values of 0.42 and 0.44 (Table 2) were not significantly different from the value of 0.51 obtained for medullary raphe neurones in Fig. 3D and E.

Metabolic acidosis

A third approach that we used to study the effect of altering pH_o on pH_i was to keep P_CO2 constant at 5% CO₂ while lowering [HCO₃⁻] from 22 mm to 13.9 mm, producing a metabolic acidosis with a pH_o of ∼7.19.

Medullary raphe neurones

The blue record in the upper panel of Fig. 4A shows the effect of metabolic acidosis on the pH_i of a single medullary raphe neurone. At the beginning of the experiment, we superfused the neurone with a Hepes-buffered solution (solution 1) and then switched to a solution containing 5% CO₂–22 mm HCO₃⁻ (solution 2). For this neurone, with a relatively high starting pH_i in Hepes, the switch to 5% CO₂ induced a rapid fall in pH_i followed by stabilization at a lower steady-state value. Changing the bath solution to one containing 5% CO₂–13.9 mm HCO₃⁻ (solution 4 in Table 1) now caused pH_i to fall rapidly and then to drift to slightly lower values over the ensuing 10 min of exposure to metabolic acidosis (metabolic acidosis induced a significant acid drift during the exposure; P < 0.05). We observed similar responses to metabolic acidosis in 16 of 20 medullary raphe neurones, as shown by the blue record for the normalized data in Fig. 4B. However, in the other four medullary raphe neurones, the metabolic acidosis caused only a slight fall in pH_i (black records in upper panel of Fig. 4A and in Fig. 4B). On two occasions, we observed both the blue and black pattern among neurones on the same coverslip.

Figure 4. Effect of metabolic acidosis/alkalosis on medullary raphe and hippocampal neurones.

A and B, metabolic acidosis. These panels are very similar to A and B in Fig. 3, except that the exposure to 9% CO₂–22 mm HCO₃⁻ in Fig. 3A and B is replaced in Fig. 4A and B by exposure to 5% CO₂–13.9 mm HCO₃⁻ (solution 5 in Table 1). A, records for 4 neurones: majority-type medullary raphe neurones (blue), minority-type medullary raphe neurones (black), majority-type hippocampal neurones (red), and minority-type hippocampal neurones (green). Each blue point in B represents 13–16 neurones for the majority-type medullary raphe neurones, each black point represents 4 minority-type medullary raphe neurones, each red point represents 5–10 majority-type hippocampal neurones, and each green point represents 4 minority-type hippocampal neurones. C and D, metabolic alkalosis. The plots are similar to those in A and B, except that the exposure to 13.9 mm HCO₃⁻ in A and B is replaced in C and D by an exposure to 5% CO₂–34.9 mm HCO₃⁻ (solution 6 in Table 1). Each blue point in D represents 8–18 medullary raphe neurones, and each red point in D represents 4–10 hippocampal neurones.

Table 3 summarizes the data from both the majority and minority groups of medullary raphe neurones. For the majority group of neurones, the mean ΔpH_i/ΔpH_o was 0.71 (Table 3), which is similar to the values that we observed for both medullary raphe and hippocampal neurones subjected to respiratory acidosis (Table 2). On the other hand, for the minority group of medullary raphe neurones, the mean ΔpH_i/ΔpH_o was only 0.09 (Table 3), indicating that these neurones were almost completely resistant to the acidosis.

Table 3.

Steady-state pH_i and ΔpH_i/ΔpH_o in medullary raphe and hippocampal neurones during respiratory acidosis and alkalosis

After the 10 min exposure to 5% CO₂–13.9 mm HCO₃⁻, we switched back to 5% CO₂–22 mm HCO₃⁻, which caused pH_i to return towards the previous baseline (blue and black records in the upper panel of Fig. 4A). The blue and black records of the lower panel of Fig. 4A show that −k₄₄₀ was less than 5% min⁻¹ for the entire experiment.

Hippocampal neurones

When we applied the metabolic acidosis protocol to hippocampal neurones, we again observed that the pH_i response followed two distinct patterns, but with the proportions reversed. The red record in the upper panel of Fig. 4A is an example of the majority response (10 of 14 neurones), which is similar to the minority response of the medullary raphe neurones. The green record in the upper panel of Fig. 4A is an example of the minority response (4 of 14 neurones), which was similar to the majority response of the medullary raphe neurones except that we found no evidence of a slow acid drift (the slopes are not different from zero; P = 0.39). Both responses are summarized in Fig. 4B. For the majority of hippocampal neurones, ΔpH_i/ΔpH_o was 0.09 (Table 3), which is even smaller than the values typically reported for non-chemosensitive cells (Ellis & Thomas, 1976; Aickin, 1984; Vaughan-Jones, 1986; Tolkovsky & Richards, 1987). For the minority hippocampal neurones, the ΔpH_i/ΔpH_o was 0.63 (Table 3), which was not significantly different (P = 0.6) from the values we observed for the majority medullary raphe neurones subjected to metabolic acidosis (ΔpH_i/ΔpH_o = 0.71 in Table 3), or from the values we observed for both medullary raphe and hippocampal neurones subjected to respiratory acidosis (Table 2).

As shown by the red record in Fig. 4A, for a hippocampal neurone with a majority-type response to metabolic acidosis, switching back to 5% CO₂–22 mm HCO₃⁻ caused pH_i to overshoot slightly the pre-acidosis value. As summarized in Table 3, the final pH_i value in 5% CO₂–22 mm HCO₃⁻ was 7.20, compared to the initial value of 7.13 in 5% CO₂–22 mm HCO₃⁻. This was the only example in the present study in which the ‘Pre’ and ‘Post’ pH_i values were significantly different. The overshoot suggests that these neurones responded to metabolic acidosis by rapidly increasing net acid extrusion (explaining the limited fall in pH_i), and that this stimulation of net acid extruders remained at least partially in effect after return to control conditions, thereby explaining the overshoot (Roos & Boron, 1981).

Metabolic alkalosis

The fourth and final approach that we used to study the effect of altering pH_o on pH_i was to keep the P_CO2 constant at 5% CO₂ while raising [HCO₃⁻] from 22 mm to 34.9 mm, producing a metabolic alkalosis with a pH_o of ∼7.61.

Medullary raphe neurones

The blue record in the upper panel of Fig. 4C shows the effect of metabolic alkalosis on the pH_i of a single medullary raphe neurone. As in the previous experiments, this experiment started in Hepes buffer (solution 1). Switching the bath solution to 5% CO₂–22 mm HCO₃⁻ (solution 2) caused a rapid fall of pH_i followed by a slow, partial recovery. Changing the bath solution to one containing 5% CO₂–34.9 mm HCO₃⁻ (solution 6 in Table 1) caused pH_i to increase rapidly by ∼0.15. The blue record in Fig. 4D, which summarizes normalized data from a series of similar experiments, shows that metabolic alkalosis caused pH_i to increase rapidly; it also suggests that pH_i may have continued to drift upwards. Indeed, an analysis of the pH_i record for each neurone during the plateau phase of the metabolic alkalosis shows that the mean dpH_i/dt (7.80 × 10⁻⁵ ± 2.23 × 10⁻⁵ pH units s⁻¹) was significantly greater than zero (P < 0.004).

As summarized in Table 3, raising [HCO₃⁻]_o from 22 mm to 34.9 mm caused the mean pH_i to increase from 7.07 to 7.16 within the first 2–3 min of the stress. Afterwards, the mean pH_i did not significantly change, even though, as noted above, the mean dpH_i/dt value was significant. The mean ΔpH_i/ΔpH_o of 0.45 was not significantly different (P = 0.14) from the value of 0.51 calculated for respiratory alkalosis in medullary raphe neurones (Table 2).

As shown by the blue record in the upper panel of Fig. 4C, switching the bath solution back to 5% CO₂–22 mm HCO₃⁻ caused pH_i to return towards its previous baseline. The blue record of −k₄₄₀ in the lower panel of Fig. 4C shows that the rate of dye loss never exceeded 5% min⁻¹.

Hippocampal neurones

As shown by the red records in Fig. 4C and D, and as summarized in Table 3, the response of hippocampal neurones to metabolic alkalosis was very similar to that of medullary raphe neurones. In particular, the mean pH_i drift during the plateau phase of metabolic alkalosis was 6.72 × 10⁻⁵ ± 3.54 × 10⁻⁵ pH units s⁻¹, which is not different from the mean value for medullary raphe neurones (P = 0.8). Moreover, the ΔpH_i/ΔpH_o for the hippocampal neurones (0.49) was virtually identical to that for medullary raphe neurones (0.45).

Discussion

In this study we compared, in medullary raphe neurones (∼40% of which are strongly chemosensitive) versus hippocampal neurones (only a small percentage of which are weakly chemosensitive; Wang et al. 2000), the effects on pH_i of prolonged exposure to one of the four fundamental acid–base disturbances (i.e. respiratory acidosis, respiratory alkalosis, metabolic acidosis, and metabolic alkalosis). Generally, these acid–base disturbances had rather similar effects on these two groups of neurones, the lone exception being metabolic acidosis. In addition, for medullary raphe neurones, we compared the effects of respiratory acidosis or alkalosis in TPH⁺ neurones (all of which are expected to be stimulated by acidosis) versus TPH⁻ neurones (3/4 of which are expected to be non-chemosensitive), and found no significant differences.

Parameters that describe how cells govern pH_i

Before discussing our data, it is useful to define three terms that describe different aspects of how cells govern pH_i. (1) Buffering power describes the ability of the cytosol to resist pH changes as weak acids or bases, biochemical reactions and organelles reversibly consume or release H⁺. Buffers can minimize a change in pH_i, and can reduce rates of pH_i change. However, they cannot prevent pH_i from changing, nor can they cause pH_i to return to its initial levels. Finally, changes in buffering power have no effect on steady-state pH_i. The present study does not address the issue of buffering power. (2) pH_i homeostasis or regulation describes the dynamic ability of a cell to restore pH_i to its initial value after the imposition of an acute acid or alkali load under conditions that do not alter the extracellular acid–base status. The pH_i recovery from acute acid or alkali loads reflects the action of acid–base transport pathways (transporters and channels) located in the plasma membrane. The present study does not directly address pH_i homeostasis. (3) ΔpH_i/ΔpH_o describes the change in steady-state pH_i in response to an imposed change in pH_o. Steady-state pH_i can change only in response to changes in the fundamental properties of acid–base transport/production pathways, as is the case in the present study. For example, decreases in pH_o and/or [HCO₃⁻]_o can directly reduce the activity of acid extruders across a wide range of pH_i values. For a more thorough discussion of these concepts, refer to the review by Roos & Boron (1981).

Steady-state pH_i in Hepes versus CO₂–HCO₃⁻ buffers

Both medullary raphe neurones and hippocampal neurones exhibited a broad range of initial pH_i values while incubated in a Hepes buffer (Fig. 2A). For both groups of cells, this broad pH_i range could reflect the presence of more than one type of neurone and/or different functional states of the same neurone type. However, we do not believe that the broad distributions of pH_i values represent different states of cell viability because all of the neurones met established criteria (Bevensee et al. 1995) for retaining the fluorescent dye BCECF.

Both medullary raphe neurones and hippocampal neurones had similar responses to the switch from a Hepes to a 5% CO₂–22 mm HCO₃⁻ buffer: an initial acidification due to the CO₂ diffusion into the cell, followed by a pH_i recovery, the extent of which depended on the initial pH_i value in the Hepes buffer (as described in Fig. 2B). (Examples of such pH_i recoveries can be seen in the initial portions of the pH_i records, i.e. the transition from the Hepes solution to the CO₂–HCO₃⁻ solution, in Figs 3A and D and 4A and C). Moreover, in both cases, the extent of the recovery tended to increase at lower values of initial pH_i (Fig. 2B), a behaviour previously reported for hippocampal neurones freshly isolated from rats (Schwiening & Boron, 1994; Bevensee et al. 1996; Brett et al. 2002). The pH_i dependence of this recovery was about 50% greater for the hippocampal versus the medullary raphe neurones, as shown by the difference in slopes of the linear fits in Fig. 2B.

Respiratory acid–base disturbances

A striking observation in the present study – which contrasts with the results of previous studies (Buckler et al. 1991; Lyall et al. 1997; Ritucci et al. 1997; Goldstein et al. 2000) – is that neurones isolated from a chemosensitive brain region (i.e. the medullary raphe) and neurones isolated from a non-chemosensitive brain region (i.e. the hippocampus) both respond to respiratory acidosis with unusually high ΔpH_i/ΔpH_o values. In other words, there was not a marked difference in ΔpH_i/ΔpH_o in the neurones from the chemosensitive region compared to the neurones from the non-chemosensitive region. Although, respiratory acidosis elicited a slightly greater ΔpH_i/ΔpH_o in medullary raphe than in hippocampal neurones (Table 2), in both cases, the ratios were large. Moreover, respiratory alkalosis elicited virtually identical responses in the two groups of neurones. Even more striking, we found that TPH⁺ medullary raphe neurones – all of which respond to respiratory acidosis by markedly increasing their firing rates (Wang et al. 2001, 2002) – exhibited ΔpH_i/ΔpH_o values that, for both respiratory acidosis and alkalosis, were indistinguishable from those of TPH⁻ medullary raphe neurones. These observations are particularly relevant because, in the short term, central chemoreceptors are far more sensitive to respiratory than to metabolic acid–base disturbances.

One consistent difference between medullary raphe neurones and hippocampal neurones was in their differential response to respiratory acidosis versus respiratory alkalosis. For all medullary raphe neurones, respiratory acidosis elicited a moderately greater ΔpH_i/ΔpH_o than did respiratory alkalosis, indicating that the neurones are more effective at minimizing upward changes in pH_i. On the other hand, for hippocampal neurones, the ΔpH_i/ΔpH_o values are indistinguishable for respiratory acidosis versus alkalosis. In conclusion, although we sometimes observed modest differences in ΔpH_i/ΔpH_o values between medullary raphe and hippocampal neurones, our results suggest that, at least during respiratory acid–base disturbances, the kinetics of the acid–base transporters, e.g. the dependencies on pH_i, pH_o, [HCO₃⁻]_i and [HCO₃⁻]_o, are not strikingly different between the chemosensitive and non-chemosensitive neurones that we studied.

Metabolic acid–base disturbances

Metabolic acidosis

In contrast to our observations with respiratory acid–base disturbances, where we observed no major differences between medullary raphe and hippocampal neurones, metabolic acidosis elicited distinctly different responses in medullary raphe versus hippocampal neurones. This is the same general conclusion reached by Ritucci et al. (1998) in their study on NTS or VLM neurones (from chemosensitive regions) versus inferior olive or hypoglossal neurones (from non-chemosensitive regions), except that the details are quite different. Ritucci et al. (1998) found that, in neurones from chemosensitive regions, metabolic acidosis universally caused a large and sustained pH_i decrease, whereas in neurones from non-chemosensitive regions, metabolic acidosis universally caused an initially large pH_i decrease, followed by a partial pH_i recovery. In contrast, we found that, in both medullary raphe and hippocampal neurones, metabolic acidosis unmasked two functional populations: one with a large and sustained pH_i decrease (ΔpH_i/ΔpH_o indistinguishable from the values observed during respiratory acidosis), and another with only a very small pH_i decrease. What distinguished the neurones in the two regions in our study was that metabolic acidosis caused large and sustained pH_i decreases in ∼80% of the medullary raphe neurones, whereas it caused minimal pH_i decreases in ∼70% of the hippocampal neurones. It is unclear whether this difference in pH_i response is related to the property of chemosensitivity. We emphasize that underlying these striking differences in steady-state pH_i must have been fundamental differences in the kinetics of acid-extrusion versus acid-loading processes; by themselves, changes in intracellular buffering power cannot produce changes in steady-state pH_i (Bevensee et al. 2000).

The most striking observation, is that most hippocampal neurones were almost totally resistant to acidification in response to metabolic acidosis. To our knowledge, such behaviour has never before been reported except for gastric gland cells exposed to acid at the side of the apical membrane, which is virtually impermeable to H⁺ (Waisbren et al. 1994a), CO₂, HCO₃⁻, NH₃ and NH₄⁺ (Waisbren et al. 1994b). It is unlikely that the hippocampal neurones in the present study are similarly impermeable to H⁺ and HCO₃⁻. We suggest that the majority population of hippocampal neurones somehow sensed the metabolic acidosis and compensated almost instantly by increasing acid extrusion and/or decreasing acid loading, thereby minimizing the fall in pH_i. On the other hand, these same hippocampal neurones were apparently not able to compensate for respiratory acidosis, which produced a large and sustained pH_i decrease. How is it that most hippocampal neurones were able to resist metabolic acidosis to pH_o 7.19, but not to resist respiratory acidosis to the nearly identical pH_o of 7.16. Did the hippocampal neurones, when subjected to metabolic acidosis, sense some altered parameter in the Henderson-Hasselbalch equation that they did not sense when subjected to respiratory acidosis?

(1) [CO₂]_o and [CO₂]_i. The imposition of metabolic acidosis had no effect on either [CO₂]_i or [CO₂]_o, which remained at ∼5%. In respiratory acidosis, [CO₂]_o and [CO₂]_i rose to ∼9%. If anything, the increase in [CO₂]_o or [CO₂]_i, independent of pH or [HCO₃⁻], should have stimulated acid extrusion, as has been observed in the renal proximal tubule (Zhou et al. 1999; Zhao et al. 2003). Yet, pH_i failed to recover during respiratory acidosis.

(2) pH_i. The imposition of metabolic acidosis had no immediate effect on pH_i (mean value, 7.13; Table 3) because [CO₂]_o ≅ [CO₂]_i was constant. The imposition of respiratory acidosis caused pH_i to fall rapidly from a mean value of 7.10 in 5% CO₂ to 6.94 in 9% CO₂ (Table 2). If anything, this lower pH_i should have stimulated acid extruders and promoted a pH_i recovery, which did not occur.

(3) [HCO₃⁻]_i. The imposition of metabolic acidosis should have had no immediate effect on [HCO₃⁻]_i because [CO₂]_o ≅ [CO₂]_i was fixed and pH_i was initially stable. The imposition of respiratory acidosis caused the calculated mean [HCO₃⁻]_i to rise from 11 mm in 5% CO₂ (mean pH_i = 7.10, Table 2) to 13.7 mm early in the exposure to 9% CO₂ (mean pH_i = 6.94). It is true that an increase in [HCO₃⁻]_i might have inhibited HCO₃⁻ uptake (i.e. acid extrusion) and/or stimulated HCO₃⁻ efflux (i.e. acid loading), thereby opposing a pH_i recovery, consistent with the sustained pH_i decrease we observed during respiratory acidosis (mean pH_i = 6.96 late during the exposure to 9% CO₂). However, assuming a Michaelis-Menten-type dependence of HCO₃⁻ transporters on [HCO₃⁻]_i, and recognizing that Na⁺–H⁺ exchangers are insensitive to changes in [HCO₃⁻]_i, it is unlikely that a mere 25% increase in [HCO₃⁻]_i could have prevented pH_i from rising to the steady-state levels that we routinely observed for hippocampal neurones during metabolic acidosis (mean pH_i = 7.11, Table 3).

(4) [HCO₃⁻]_o. The imposition of respiratory acidosis had no immediate effect on [HCO₃⁻]_o. However, the imposition of metabolic acidosis was associated with a decrease in [HCO₃⁻]_o from 22 mm to 13.9 mm. This decrease in [HCO₃⁻]_o, by itself, should have lowered HCO₃⁻-dependent acid extrusion and/or raised HCO₃⁻-dependent acid loading, thereby promoting in a gradual decline in pH_i, consistent with the large fall in pH_i that we observed with most medullary raphe neurones. The hippocampal neurones could have resisted this tendency toward acidification only if they were able to sense the decrease in [HCO₃⁻]_o and respond by increasing the activity of acid extruders and/or by decreasing the activity of acid loaders.

Thus, we suggest that during metabolic acidosis the majority population of hippocampal neurones and the minority population of medullary raphe neurones somehow sense the fall in [HCO₃⁻]_o, and respond by enhancing their ability to resist the tendency to acidify. On the other hand, the minority population of hippocampal neurones and the majority population of medullary raphe neurones apparently do not have such a robust response to a decrease in [HCO₃⁻]_o.

What might be the relevance of the majority-type responses to metabolic acidosis in medullary raphe versus hippocampal neurones? During chronic metabolic acidosis, [HCO₃⁻] in the cerebrospinal fluid decreases, though to a smaller extent than [HCO₃⁻] in arterial blood (for review, see Dempsey & Forster, 1982; Nattie, 1998). Moreover, patients with metabolic acidosis (e.g. diabetic ketoacidosis), consistently exhibit a significant hyperventilation (Dempsey & Forster, 1982), which in its extreme form is referred to as Kussmaul's respiration. A majority-type response in the medullary raphe would allow chemosensitive neurones to respond to chronic metabolic acidosis by lowering pH_i, which in turn would allow them to increase ventilation. A majority-type response in the hippocampus would stabilize pH_i and thus help the neurones perform their normal tasks even in the face of metabolic acidosis.

Metabolic alkalosis

We found that the response to metabolic alkalosis, unlike the response to metabolic acidosis, was roughly the same for all neurones we studied, both medullary raphe and hippocampal. In both cases, pH_i at first rose rapidly and then tended to continue to drift upward more slowly. Also in both, the ΔpH_i/ΔpH_o values were similar to those for medullary raphe and hippocampal neurones exposed to respiratory alkalosis. These results are not very different from those of Ritucci et al. (1998), except for the lack of the slow, alkalinizing drift in their study.

Does a high ΔpH_i/ΔpH_o represent a good marker of CO₂/H⁺ chemosensitivity?

The concept that a high ΔpH_i/ΔpH_o is unique to chemoreceptors originated in the study of Buckler et al. (1991) on type I cells from the carotid body. They showed that a brief (∼5 min) respiratory or metabolic acid–base disturbance (ΔpH_o) elicited a rapid shift in pH_i (ΔpH_i) from which the cell did not recover. It is important to recognize that, in the study of Buckler et al. (1991) the cells – after an acute acid load imposed by an ammonium prepulse (Boron & De Weer, 1976) – were fully capable of returning pH_i to normal when the extracellular acid–base was also normal. The failure of pH_i to recover following a decrease in pH_o– the conditio sine qua non for a high ΔpH_i/ΔpH_o – probably reflected an underlying inhibition of acid extrusion and/or stimulation of acid loading. The concept that a high ΔpH_i/ΔpH_o is unique to chemoreceptors gained weight with the work by Ritucci et al. (1997), who made similar observations on neurones from chemosensitive areas in the brain, as opposed to neurones from non-chemosensitive areas.

A steep relationship between pH_i and pH_o might very well be necessary for chemoreception if the sensed parameter is pH_i, as is generally believed (Wiemann et al. 1998; Filosa et al. 2002; Wang et al. 2002). However, the results from the present study temper this notion. We found that the responses of TPH⁺ versus TPH⁻ neurones to respiratory acid–base disturbances were indistinguishable. The same is true with the response to respiratory alkalosis for mixed medullary raphe versus hippocampal neurones. The only statistically significant difference for respiratory acid–base disturbances was the response to respiratory acidosis for mixed medullary raphe versus hippocampal neurones, and this difference was small. In summary, it may be true that chemosensitive neurones exhibit a large ΔpH_i/ΔpH_o in response to respiratory acid–base disturbances, but at least one variety of non-chemosensitive neurones – hippocampal neurones – is capable of exhibiting equally large ΔpH_i/ΔpH_o responses. Thus, it could be incorrect to assume that a high ΔpH_i/ΔpH_o is a specialization of a neurone that is a chemoreceptor. Our results suggest that chemosensitivity, for example, the firing rate response to respiratory acidosis, is conferred not by the unique pH_i physiology of the neurone, but by the unique electrophysiological response of the neurone to a pH_i change.

Another interesting example in which a high ΔpH_i/ΔpH_o does not predict pH chemosensitivity is the response of the renal proximal tubule to respiratory acidosis. It has been known for nearly half a century that the kidney responds to respiratory acidosis by increasing its rate of HCO₃⁻ reabsorption (Brazeau & Gilman, 1953; Dorman et al. 1954), the renal counterpart of hyperventilation. Investigators predicted that a decrease in pH_i mediated the stimulation of HCO₃⁻ reabsorption. However, the use of out-of-equilibrium CO₂–HCO₃⁻ solutions (Zhao et al. 1995), which made it possible to change [CO₂]_o, [HCO₃⁻]_o, or pH_o independently, has led to the observation that HCO₃⁻ reabsorption by the proximal tubule cell is completely insensitive to even large changes in pH_i or pH_o (Zhou et al. 2001). Instead, the tubule cell responds to changes in [CO₂] and [HCO₃⁻] at or near the basolateral membrane.