Transporters involved in regulation of intracellular pH in primary cultured rat brain endothelial cells

Abstract

Fluid secretion across the blood–brain barrier, critical for maintaining the correct fluid balance in the brain, entails net secretion of HCO₃⁻, which is brought about by the combined activities of ion transporters situated in brain microvessels. These same transporters will concomitantly influence intracellular pH (pH_i). To analyse the transporters that may be involved in the maintenance of pH_i and hence secretion of HCO₃⁻, we have loaded primary cultured endothelial cells derived from rat brain microvessels with the pH indicator BCECF and suspended them in standard NaCl solutions buffered with Hepes or Hepes plus 5% CO₂/HCO₃⁻. pH_i in the standard solutions showed a slow acidification over at least 30 min, the rate being less in the presence of HCO₃⁻ than in its absence. However, after accounting for the difference in buffering, the net rates of acid loading with and without HCO₃⁻ were similar. In the nominal absence of HCO₃⁻ the rate of acid loading was increased equally by removal of external Na⁺ or by inhibition of Na⁺/H⁺ exchange by ethylisopropylamiloride (EIPA). By contrast, in the presence of HCO₃⁻ the increase in the rate of acid loading when Na⁺ was removed was much larger and the rate was then also significantly greater than the rate observed in the absence of both Na⁺ and HCO₃⁻. Removal of Cl⁻ in the presence of HCO₃⁻ produced an alkalinization followed by a resumption of the slow acid gain. Removal of Na⁺ following removal of Cl⁻ increased the rate of acid gain. In the presence of HCO₃⁻ and initial presence of Na⁺ and Cl⁻, DIDS inhibited the changes in pH_i produced by removal of either Na⁺ or Cl⁻. These are the expected results if these cells possess an AE-like Cl⁻/HCO₃⁻ exchanger, a ‘channel-like’ permeability allowing slow influx of acid (or efflux of HCO₃⁻), a NBC-like Cl⁻-independent Na⁺−HCO₃⁻ cotransporter, and a NHE-like Na⁺/H⁺ exchanger. The in vitro rates of HCO₃⁻ loading via the Na⁺−HCO₃⁻ cotransporter could, if the transporter is located on the apical, blood-facing side of the cells, account for the net secretion of HCO₃⁻ into the brain.

Endothelial cells from brain microvessels form the structural and functional components of the blood–brain barrier. It is well known that this cell layer blocks the uncontrolled movements of many solutes and allows rapid movements of selected substances such as glucose, oxygen and carbon dioxide. It also secretes the principal ionic components of the interstitial fluid at a rate which is variously estimated to be 10–50% of the total rate of secretion into the extracellular fluid of the brain (Milhorat, 1987; Cserr & Patlak, 1992; Abbott, 2004; Redzic & Segal, 2004) and thus will include something of the order of 1–5 mmol day⁻¹ HCO₃⁻. It is known that secretion by the cells of the blood–brain barrier is driven by the sodium pump expelling Na⁺ on the abluminal side (Eisenberg & Suddith, 1979; Betz, 1983; Ennis et al. 1996); the transendothelial permeability to radiolabelled K⁺ is much greater than that to Na⁺ and Cl⁻ (Smith & Rapoport, 1986); Na⁺, K⁺ and Cl⁻ can enter the cells on the luminal surface via a Na⁺−K⁺−2Cl⁻ cotransporter (O'Donnell et al. 2004; Foroutan & O'Donnell, 2005); the endothelial cells have functional amiloride (or ethylisopropylamiloride; EIPA)-sensitive Na⁺/H⁺ exchangers (Betz, 1983; Vigne et al. 1991; Hsu et al. 1996; Sipos et al. 2005); the resting intracellular pH, pH_i, is decreased by removal of external Na⁺ (Hsu et al. 1996); there is evidence for a Na⁺- and Cl⁻-dependent HCO₃⁻ influx (Hsu et al. 1996) and there is a Na⁺- and HCO₃⁻-dependent component of the pH recovery process after an imposed acidification (Sipos et al. 2005). In addition there must be a transporter or channel to allow efflux of Cl⁻ on the abluminal side and transporters or channels to allow recycling of K⁺ on both the luminal and abluminal sides. However, nothing is known of the rates of transport for H⁺ or HCO₃⁻ and the identities of the transporters remain unknown.

The transporters discussed in this paper are found in many cells throughout the body where they are involved in pH regulation and secretion. There have been several studies of pH regulation in vascular endothelia that do not secrete (see, e.g. Faber et al. 1998; Sun et al. 1999) and a great many on HCO₃⁻ transport in secretory epithelia and a variety of cell lines (see, e.g. Reinertsen et al. 1988; Tonnessen et al. 1990; Romero et al. 2004 and references therein). Nevertheless because these tissues either subserve different functions or have different embryological origins the transport activities in brain microvascular endothelial cells cannot be inferred from studies on any of these systems.

Because the area of the blood brain barrier is so large, approximately 100 cm² g⁻¹ (see, e.g. Bradbury, 1979), the rate of secretion of fluid and bicarbonate across the blood–brain barrier expressed per unit area is small. Partly for this reason there is at present no in vitro system that will allow direct, quantitative measurement of the net rate of HCO₃⁻ secretion across the endothelial cells. However, the transporters responsible for the secretion will per force also influence the intracellular pH, pH_i, of the endothelial cells. Transporters implicated in regulation of pH_i in various cell types include the sodium–hydrogen exchangers and several members of the bicarbonate transporter superfamily. The Na⁺/H⁺ exchangers, the NHEs (gene family SLC9), extrude protons in exchange for extracellular Na⁺ (Wakabayashi et al. 1997; Putney et al. 2002). They either are active under most circumstances (e.g. NHE3 in kidney proximal tubules) or more commonly are activated by a fall in pH_i. Members of the NHE family can be inhibited by amiloride and more specifically by its derivative, ethylisopropylamiloride (EIPA; Vigne et al. 1983; Kleyman & Cragoe, 1988). The bicarbonate transporter superfamily (SLC4; Romero et al. 2004) contains Na⁺-independent Cl⁻/HCO₃⁻ exchangers, also termed anion exchangers (AE1, AE2, and AE3; Alper et al. 2001), Na⁺-driven Cl⁻/HCO₃⁻ exchangers (NDCBE; Boron, 2001; Grichtchenko et al. 2001), and Na⁺−HCO₃⁻ cotransporters (NBC; Romero & Boron, 1999; Soleimani & Burnham, 2001). It is currently unclear if NCBE is a Na⁺-driven Cl⁻/HCO₃⁻ exchanger or a Na⁺−HCO₃⁻ cotransporter (Wang et al.; Romero et al. 2004). AE activity is increased at alkaline pH_i (Reinertsen et al. 1988; Alper et al. 2001) which leads to HCO₃⁻ efflux from the cell; NBC, NDCBE or NCBE activity normally leads to HCO₃⁻ influx (with the important exception of the 1Na⁺−3HCO₃⁻ mode of activity of NBCe1 in kidney proximal tubules). Members of the multifunctional anion exchanger family (SLC26) of transporters (Mount & Romero, 2004) can also transport HCO₃⁻ and OH⁻. With the exception in some tissues but not others of the electroneutral Na⁺−HCO₃⁻ cotransporter, NBCn1 (SC4A7; Choi et al. 2000; Praetorius et al. 2004a; Bouzinova et al. 2005; Boedtkjer et al. 2006; Damkier et al. 2006), all of the known isoforms of the HCO₃⁻-dependent transporters can be inhibited by the stilbene derivative 4,4′-diisothiocyanostilbene-2,2′-disulphonic acid (DIDS) (Mount & Romero, 2004; Romero et al. 2004).

With an internal pH perhaps 0.3 less than that of the extracellular fluid and a membrane potential at least as negative as −30 mV, H⁺ will passively enter and HCO₃⁻ will passively leave rat brain endothelial cells via any channels or other available routes. The cells must therefore possess some means of extruding acid or equivalently of acquiring base. In cells such as brain endothelial cells, which are not called upon to generate large pH gradients, the normal source of energy for the acid extrusion is the Na⁺ concentration gradient generated by the Na⁺ pump. Brain endothelial cells must also mediate a net flux of HCO₃⁻ from blood to brain. We present evidence here for three transport activities seen at intracellular pH near the presumed resting value as indicated in Fig. 1 along with ‘channel-like’ permeability and the Na⁺ pump. Throughout this paper we follow the usual convention (see, e.g. Bevensee & Boron, 1998b) and refer to acid loaders and acid extruders even though much of the acid loading occurs via efflux of HCO₃⁻ and much of the acid extrusion occurs via influx of HCO₃⁻.

Figure 1. Putative transporter activities discussed in this paper.

CBex,sens: DIDS-sensitive, Na⁺-independent Cl⁻/HCO₃⁻ exchange like that mediated by the AE transporters; NBcot: DIDS-sensitive Na⁺,HCO₃⁻ cotransport like that mediated by the NBC transporters; NHex: EIPA-sensitive Na⁺/H⁺ exchange like that mediated by NHE1; NadrivenCBex: DIDS-sensitive, Na⁺-dependent, Cl⁻/HCO₃⁻ exchange like that mediated by NDCBE and possibly NCBE; Na⁺,K⁺-ATPase: the sodium pump; The ‘channel-like’ permeability, routes for HCO₃⁻ exit out of the cell (or for H⁺ entry into it). Arrows indicate the direction of transport in the ‘resting’ state of the cells. The data in this paper provide only indirect support for the presence of NadrivenCBex activity (see text).

Methods

Solutions and chemicals

Unless otherwise stated, all materials were obtained from Sigma (Poole, Dorset, UK). The compositions of the various buffered solutions are indicated in Table 1.

Table 1.

Composition of solutions used in the experiments

	Hepes buffers				NH4+ Hepes	HCO3− buffers
	Normal	Na+ free	Cl− free	NaCl free	Normal	Normal	Na+ free	Cl− free	NaCl free
NaCl	127	—	—	—	107	102	—	—
NaHCO3	—	—	—	—	—	25	—	25
KCl	4	2.9	—	—	4	4	2.9	—
MgSO4	0.6	0.6	0.6	0.6	0.6	0.6	0.6	0.6	0.6
CaCl2	0.3	0.3	—	—	0.3	0.3	0.3	—
Na2HPO4	1.1	0	1.1	—	1.1	1.1	—	1.1
KH2PO4	0.6	1.7	0.6	1.7	0.6	0.6	1.7	0.6	1.7
Hepes	10	10	10	10	10	10	10	10	10
Glucose	6	6	6	6	6	6	6	6	6
NaOH	—	—	—	—	—	5	—	—
NMDG	—	132	0	132	—	—	132	—	132
Na-gluconate	—	—	127	—	—	—	—	102
K-gluconate	—	—	4	2.9	—	—	—	4	2.9
Ca-gluconate	—	—	0.3	0.3	—	—	—	0.3	0.3
Gluconolactone	—	—	—	127	—	—	—	—	104
NH4Cl	—	—	—	—	20	—	—	—

All values are stated in mm. Solutions were adjusted to pH 7.4 at 37°C using either HCl, NaOH or in the case of NaCl free solutions, Tris (tris hydroxymethylaminomethane). Buffers containing added HCO₃⁻ were bubbled with 5% CO₂−95% O₂ until the pH was stable.

The standard solutions for pH calibration contained 110 mm potassium gluconate or 110 mm potassium aspartate, 20 mm KCl, 0.5 mm MgCl₂, 20 mm Hepes and 20 μm nigericin adjusted to various pH values using KOH. Calibration buffers used over a pH range of 6–6.9 and 7.9–8.4 contained 20 mm Pipes and 20 mm Tricine, respectively, in place of 20 mm Hepes

The acetoxymethyl ester of 2′,7′-bis(2-carboxyethyl)-5(6)-carboxyfluorescein (BCECF-AM) (Molecular Probes, Netherlands) at 0.5 mm, EIPA at 100 mm, DIDS at 125 mm and ethoxzolamide at 100 mm were prepared in anhydrous DMSO (41648, Fluka, Buchs, Germany). All except DIDS were aliquoted and frozen; DIDS solutions were prepared on the day of the experiment. Nigericin was stored as a 10 mm stock in ethanol. The final concentrations of DMSO or ethanol never exceeded 0.2%.

Animals

Wistar rats (200–350 g, Charles River, Ramsgate, UK) were killed using carbon dioxide asphyxiation in accordance with the Animals (Scientific Procedures) Act 1986.

Cell preparation

Brain microvessels were isolated from the cortical grey matter removed from rats using a method previously described (Abbott et al. 1992; Barrand et al. 1995) and were maintained as described by Seetharaman et al. (1998). Two to three days prior to experiments, cells were plated onto a small area of a glass coverslip coated first with poly d-lysine and then collagen. Medium was replaced 1–2 days before the experiment, and cells were used at passages 2–4.

Measurement of intracellular pH

Measurement of pH_i was performed using the pH-sensitive fluorescent dye, BCECF, and a fluorescence spectrophotometer (F-2000, Hitachi). The cuvette was maintained at 37°C and the solutions were stirred by a small magnetic flea throughout the experiments. Fluorescence was recorded at 5 s intervals at 526 nm with excitation alternately at 440 nm and 502 nm using 5 nm (or in some experiments 10 nm) excitation and emission slit widths.

Approximately (2–3) × 10⁴ cells were plated and grown on glass coverslips in a region slightly larger than the 2 mm × 7 mm rectangle illuminated by the excitation beam. For experiments with Hepes-buffered medium, the coverslip with adherent cells was rinsed and mounted in the cuvette containing 2.5 ml of experimental solution without BCECF-AM. After recording the ‘cells-only’ backgrounds at the two excitation wavelengths, the cuvette was exchanged for another containing medium and 0.5 μm BCECF-AM (0.1% DMSO) and the initial medium fluorescence was noted. The coverslip holder, coverslip and cells were then transferred to the new cuvette. The fluorescence at both wavelengths increased linearly with time during loading. At the end of loading, typically for 900 s, the coverslip holder was transferred to a fresh cuvette and the final fluorescence of the loading medium noted. Changes in experimental solutions were made by transfer of the coverslip between cuvettes.

The procedures for experiments with CO₂/HCO₃⁻-buffered medium included minor modifications to reduce loss of CO₂. Solution equilibrated with 5% CO₂–95% O₂ by bubbling were taken up into a syringe and dispensed into the cuvette through a short length of wide bore tubing. Care was taken not to produce bubbles or an aerosol. The cuvette was then closed off with a snug-fitting Teflon cap. The lack of change of pH_i during three successive transfers of cells between cuvettes is shown in a figure reported previously (Hladky et al. 2000) demonstrating that the layer of buffered solution adherent to the coverslip provided adequate protection for the cells during the transfer. Since there were only a small number of cells in the cuvette during an experiment and the buffer volume was 2.5 ml, the amount of CO₂ generated from cellular metabolism was too small to produce any detectable changes in pH_i. The ease with which CO₂/HCO₃⁻-buffered solutions can be handled in these experiments is a major advantage of the cuvette-based technique.

Calculation of cell associated fluorescence ratios

The initial, ‘cells-only’ readings were taken for use as the background fluorescence values (i.e. those not resulting from cell associated BCECF). pH_i was calculated as described below from the ratio (measured fluorescence, excitation 502 nm – background, excitation 502 nm)/(measured fluorescence, excitation 440 nm – background, excitation 440 nm). To confirm the validity of using the ‘cells only’ readings for the background corrections, pH_i was also calculated from the ratio of the slope of the cell-associated fluorescence versus time curve measured at 502 nm during loading to that at 440 nm. Cell-associated fluorescence during loading was in turn calculated as the difference between the fluorescence measured in the presence of the cells and the fluorescence of the medium, which were each shown to increase linearly with time during loading. In the few experiments for which there were discrepancies between the values of pH_i determined just before the end of loading or just after loading, the data were discarded. All pH_i values reported in this paper are calculated from raw fluorescence values that are at least twice (usually more than three times) the background (at both excitation wavelengths).

Calibration of BCECF and the fluorimeter

The pH calibration curve used to convert background-corrected fluorescence ratios to values of pH depends on the properties of the fluorescent indicator, the light source and the detection system. An initial calibration curve was constructed using values of pH_i imposed with the nigericin high-K⁺ method (Thomas et al. 1979; Bevensee & Boron, 1998a) (see Fig. 2A) using 140 mm K⁺ and 20 mm Cl⁻ and fitted using the standard equation for titration between the acidic and basic forms of the indicator:

where R_acid, and R_base are the fluorescence ratios of the two forms. The fitted parameters are R_acid = 1.06, R_base = 11.7 and pK′_D = 7.16. There were no significant differences between measurements with gluconate⁻- or aspartate⁻-based solutions. These may be compared with the values obtained for the indicator in free solution (Fig. 2A), 0.8, 12.0 and 7.00. The use of BCECF as an indicator for pH requires that the fluorescence ratio is not sensitive to other factors that may change within the cell. Thus it is reassuring that the two calibrations are so close. To investigate the discrepancy further, calibration measurements were obtained using solutions with 140 mm Cl⁻ (and no gluconate⁻). With BCECF in cell-free solution the fluorescence ratios are not affected by this change (data not shown). With BCECF loaded into cells the fluorescence ratios with the high Cl⁻ solutions at each of pH 6.9, 7.4 and 7.9 were 0.8 ± 0.1 (P < 0.01, n = 4) greater than when measured with the low Cl⁻ solutions (corresponding to a shift of ∼0.2 pH units). This effect of altering Cl⁻ outside on the fluorescence of BCECF inside suggests that intracellular ion concentrations are changed during the calibration procedure, which may invalidate the assumption made that pH_i = pH_o.

Figure 2. Calibration of BCECF and measurement of intracellular buffering.

A, pH imposed upon cells loaded with BCECF (□, continuous line) using the nigericin–high K⁺ method or pH in a solution of BCECF (○, dashed line) plotted against the background corrected fluorescence ratio that was produced. The calibration curve used in the rest of this paper is indistinguishable on this graph from the solution curve. B, example of null-point offset data for addition of a 6: 1 ratio of butyric acid to trimethylamine (20 mm total). pH, as calculated using the cells curve in A, is plotted against the change in pH that was observed. The straight line is fitted by linear regression with intercept 7.15 which is 0.14 units higher than the null-point pH, 7.01, for the 6: 1 ratio. C, chord estimates of buffering versus pH_i either just after loading cells with base or just before a load of acid or base is removed. The chord estimate is the amount of base added divided by the change in pH. ×, addition of NH₄Cl; ▵, removal of NH₄Cl; O, removal of CO₂/HCO₃⁻; –, the theoretical predictions for a single intracellular buffer with pK = 6.44 and maximum capacity 74 mm.

The null-point method (Eisner et al. 1989) avoids the need to make the assumptions (Chaillet & Boron, 1985; Boyarsky et al. 1988) underlying the nigericin–high-K⁺ method. The null-point method is based on determining the ratio of the total concentrations of a weak acid, A_T, and a weak base, B_T, which when added simultaneously produce no change in pH_i. The value of pH_i at this null point can then be calculated as:

Combinations of butyric acid and trimethylamine (combined total concentration 20 mm) were applied to cells in the standard Hepes buffer at a variety of different starting values of pH_nk calculated from the nigericin–high-K⁺ calibration. As shown in Fig. 2B for a 6: 1 ratio, the changes in pH became less positive or more negative at larger initial pH_nk. The null-point can be estimated as the intercept on the pH axis by linear regression. In addition (data not shown) the changes in pH_nk became less positive or more negative as the ratio A_T/B_T was increased. The data for ratios from 4: 1 to 16: 1 have been fitted assuming that the change in pH is proportional to the difference between the actual pH_i and the null-point pH, and that the offset, OF, between the nigericin-high-K⁺ estimate and the actual pH is constant, which leads to

The best fit value of the nonlinear least squares fit of the predicted changes to those observed was obtained for an offset of 0.15 ± 0.04 (P < 0.05 against any proposed value of the offset outside this range). The calibration curve used to calculate pH_i was therefore shifted by this amount relative to pH_nk, yielding R_acid = 1, R_base = 12 and pK′_D = 7.04.

It was confirmed that in free solution the fluorescence ratios were not affected by Cl⁻, NH₄⁺, or EIPA over the ranges of concentration encountered in this study. DIDS in solution produced fluorescence somewhat lower than the typical background values. As only changes in pH in the continuous presence or absence of DIDS are interpreted in this paper and the pH calibration curve is approximately linear over the entire region of interest, no correction was made. It was also confirmed that the fluorescence ratios were independent of dye concentration over a range producing emission intensities spanning that recorded in the experiments with cells. The ratio differed by less than 2% even for emission intensities 5 × higher. There has been a previous report (Hegyi et al. 2004) demonstrating that the calibration curve for BCECF and a microscope based optical system can vary with fluorescence intensity. The experiments reported here differ from those in that a fluorimeter has been used instead of a microscope, the detectors are different, and the dye concentrations are lower.

Calculation of intracellular buffering

The non-bicarbonate buffering in the endothelial cells was estimated from the immediate change in pH_i following withdrawal of CO₂/HCO₃⁻ or after the addition or removal of NH₃/NH₄⁺. It was not useful to make stepwise changes because pH_i changed too rapidly after each step. The intrinsic buffering was calculated as:

where the base added was obtained from the composition of the external solution and the measured values of pH_i. To calculate the internal bicarbonate concentration immediately before the removal of CO₂/HCO₃⁻ using the Henderson-Hasselbach equation:

it is necessary to know the combination of constants:

This value was determined as 6.11 ± 0.01 (n = 6) under the conditions used to prepare the experimental solutions by measuring the pH produced by additions of 135 mm NaHCO₃−135 mm NaCl at 37°C with constant bubbling with 5% CO₂−95% O₂.

The data (see Fig. 2C) were fitted using the integrated, single-buffer expression for the ratio of the base added to the change in pH_i:

with pK = −log(K_d/1 m) = 6.44 and A = 74 mm. Because in these cells changes in [CO₂] rapidly produce changes in [HCO₃⁻]_i (see, e.g. Fig. 3), the total buffering has been calculated as (see, e.g. Roos & Boron, 1981):

Figure 3. Changes in pH_i following transfer between Hepes and CO₂/HCO₃⁻+ Hepes-buffered solutions in the presence (grey) or absence (black) of 10 μm ethoxzolamide.

At the first break in the trace, the solution was replaced with one containing ethoxzolamide or a solvent control. If added at this point, ethoxzolamide was also present in all subsequent solutions. At the second break the cells were transferred to Hepes-buffered solution and at the third back to HCO₃⁻. The rapid responses immediately following the solution changes require rapid intracellular conversion of HCO₃⁻ to CO₂. Ethoxzolamide blunts these initial responses. The changes in pH_i plotted are the calculated pH_i values minus the average of five points just before the change from CO₂/HCO₃⁻ to Hepes (7.36 absence and 7.41 presence of ethoxzolamide).

Comparison of rates of acid gain in different solutions

Following a 200 s period in one dye free test solution, cells were exposed to a second test solution for 200 s. As indicated in Table 3 paired comparisons are either between these two periods, or between the second periods for coverslips from the same preparation measured on the same day.

Table 3.

Rates of acid gain following ion substitutions or additions of DIDS or EIPA

		Rate of gain of acid μm s−1
	Paired comparison of 1st and 2nd intervals for a single cover slip	1st interval	2nd interval	n	P
1	NaCl−Hepes: NaCl−Hepes	7.5 ± 1.8	8.0 ± 1.4	10	0.78
2	NaCl−HCO3: NaCl−HCO3	12.1 ± 1.9	15.0 ± 2.3	9	0.11
3	NaCl−Hepes: NMDGCl−Hepes	7.8 ± 1.1	10.1 ± 1.0	15	< 0.02
4	NaCl−Hepes: NaCl−Hepes + 5 μm EIPA	6.8 ± 1.6	11.0 ± 1.0	10	< 0.01
5	NaCl−HCO3: NMDGCl−HCO3	8.1 ± 1.7	33.5 ± 3.8	27	< 0.001
6	NaCl−HCO3+ 250 μm DIDS: NMDGCl−HCO3+ 250 μm DIDS1	9.1 ± 5.7	4.2 ± 8.3	6	0.25
7	NaCl−HCO3: early Na-gluconate−HCO3	9.7 ± 1.8	−20.0 ± 5.0	14	< 0.001
8	NaCl−HCO3: late Na-gluconate−HCO32	9.7 ± 1.8	9.5 ± 2.02	14	0.75
9	NMDGCl−HCO3: early NMDG-gluconate−HCO3	19.6 ± 2.8	−9.1 ± 1.5	5	< 0.005
10	Na-gluconate−HCO3: NMDG-gluconate−HCO33	12.4 ± 0.8	29.6 ± 3.8	10	< 0.001

¹DIDS was first applied at the start of loading and was then present throughout.

²The acidwards drift after removal of Cl⁻ was preceded by an alkaline shift in pH of 0.16 ± 0.03 units (see row 7 and Fig. 5). The rate reported is the slope seen after the transient.

³Gluconate⁻ replaced Cl⁻ during loading with BCECF and throughout the experiment. ⁴Initial rates of change, i.e. at the onset of the transient alkalinization.

Acidification of cells by the NH₄Cl pulse protocol

Following a 200 s period in dye-free solution, cells were exposed to 20 mm NH₄Cl (replacement of Na⁺ by NH₄⁺). The presence of the NH₄Cl results in a rapid initial intracellular alkalinization followed by a slow acidification. After 200 s the cells were transferred to a NH₄Cl-free Hepes-buffered solution and the subsequent changes in pH_i measured. In all cases the cells undergo a marked intracellular acidification followed by a slower alkalinization (Boron & De Weer, 1976; Roos & Boron, 1981; Thomas, 1984). The recovery process was fitted as a single exponential decay towards a constant value using the solver in Excel.

RT-PCR analysis

Total RNA was isolated from freshly dissected rat brain and cultured rat brain endothelial cells using TriPure™ Isolation Reagent (Roche Diagnostics, Burgess Hill, W. Sussex, UK) according to manufacturer's instructions. All samples were checked for integrity and relative content of RNA by electrophoretic separation through agarose and volumes adjusted accordingly before proceeding to RT-PCR analysis. cDNA synthesis was performed with Bioscript (Bioline Ltd, London) according to the manufacturer's instructions. Samples were stored at −20°C. PCR was performed with a Rotor-Gene 3000TM (Corbett Research, Concord, NSW, AUS) using SensiMix DNA Kit (Quantace Ltd, Watford, UK) (12.5 μl enzyme mix, 0.5 μl Sybr green I stock, 0.75 μl 50 mm MgCl₂) together with 500 nm primers and 8.75 μl template in 25 μl final volume. Initial denaturation for 10 min at 95°C was followed by 45 cycles of amplification (95°C for 15 s and 60°C for 60 s) and melting curve analysis between 60 and 99°C in 1°C, 5 s steps.

Primer sequences are listed in Table 2. The results were analysed with the Rotor-Gene software v6 (Corbett Research, Concord, NSW, AUS) using the comparative quantification feature to compare expression of the mRNA of interest with that of NHE1. Sizes of the PCR products were confirmed using agarose gel separation.

Table 2.

Primers for PCR

Gene	Primer pairs sequence	Product size	Accession number
β2-Microglobulin	5′-CCGTGATCTTCTGGTGC-3′	250	Y00441
	5′-TTCAGTGTGAGCCAGGAT-3′
NHE1	5′-ATGTGGCTGGGAAACAAGAC-3′	218	NM012652
	5′-AGTGGTCCTCATCACCCAA-3′
AE1	5′-GCTTCTGGCTTATCCTGCTG-3′	166	NM_01651
	5′-AGCGGGTAGTCCTGGAAAAT-3′
AE2	5′-GAGCCCTTCTGCTGAAACAC-3′	190	NM_017048
	5′-CGCTCTCTATCCACCTCCAG-3′
AE3	5′-AAAAGCCTGAAACTGCTGGA-3′	217	NM_017049
	5′-GCCCTAGCTCGTGATAGTCG-3′
NBCe1	5′-TAGCGACAACGACGATTCTG-3′	227	NM_053424
	5′-TTGAGGCACGACTTTCACTG-3′
NBCe2	5′-GTTCGTCTTGCCCAGTCTTC-3′	209	NM_212512
	5′-TCATCAACTCCTGCGATCAG-3′
NCBE	5′-GGCATGTACCTCTCCGAAAA-3′	183	NM_178092
	5′-CATCGAGCCAGCTAAGTTCC-3′
NBCn1	5′-AAACATCACCACCGGAGAAG-3′	227	NM_058211
	5′-TCTTCGAACTTCAGCCACCT-3′
NDCBE	5′-ACGGGGATTGCCTACTCTCT-3′	183	NM_199497

Sequences were designed using Primer 3 (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi).

Statistical analysis

Data are expressed as mean values ± s.e.m. with n indicating the number of experiments. P values are calculated using a paired or unpaired (unequal variance) two-tailed Student's t test as appropriate. The parameters of the buffer curve and the offset of the pH calibration curve were fitted by ordinary least squares minimization. The significance of the improvement in fit allowed by introducing a variable parameter was assessed by an F test on the reduction in the residual sum of squares compared to the mean squared error per remaining degree of freedom (the variance ratio test). Analysis of variance for 429 measurements for cells in Hepes which were derived from 13 primary cultures showed that the variation of measurements between cells derived from a single primary culture (root mean squared deviation 0.09 pH units) was significantly less than the variation between the mean values for each culture (root mean squared deviation 0.2 pH units; P < 10⁻⁶). Thus the ranges of values reported in this paper reflect primarily variations in the properties of the cells rather than inaccuracy in the measurements. To minimize the effect of this variation, most measurements are paired with controls measured on the same day and, when possible, on the same cells (see Tables 3 and 4).

Table 4.

Rates of acid gain following ion substitutions or additions of the inhibitors EIPA or DIDS. Comparisons between paired or grouped cover slips

		Rate of gain of acid (μm s−1)
		1	2	n	P
Initial rates during 1st recording intervals in the buffer solution indicated
A	(1) NaCl−Hepes, (2) NaCl−HCO3	7.0 ± 0.7	8.8 ± 1.1	34	0.21
Comparison of 2nd recording intervals
B	Two successive periods of either;	12.5 ± 2.7	15.3 ± 2.4	6	0.43
	(1) NaCl−HCO3
	(2) NaCl−HCO3+ 250 μm DIDS
C	(1) NaCl−Hepes → NMDGCl−Hepes			15	0.52
	(2) NaCl−Hepes → NaCl−Hepes + 5 μm EIPA	10.1 ± 1.0	11.0 ± 1.0	10	unpaired
D	(1) NaCl−Hepes → NMDGCl−Hepes	9.6 ± 1.1	33.6 ± 5.1	16	< 0.001
	(2) NaCl−HCO3→ NMDGCl−HCO3			18	unpaired

Results

Initial pH_i and rate of change of pH_i in the presence or absence of HCO₃⁻

For cells loaded with the indicator BCECF in solutions buffered either with Hepes alone or with Hepes plus CO₂/HCO₃⁻, the pH_i after loading decreased gradually at a rate that remained uniform for more than 400 s. The value immediately after loading, called here the initial pH_i, was 7.098 ± 0.004 (n = 473) for cells in Hepes, which is close to but significantly less than the equivalent value for cells loaded in the presence of CO₂/HCO₃⁻, 7.146 ± 0.007 (n = 115, P < 0.001).

Because the rate of decrease of pH_i was less in the presence of HCO₃⁻ than in its (nominal) absence ((1.8 ± 0.2) × 10⁻⁴ s⁻¹ and (2.8 ± 0.3) × 10⁻⁴ s⁻¹, n = 34, P < 0.02, paired), the difference between the values of pH_i for cells in Hepes- or HCO₃⁻-buffered solutions increased during the experiments. Thus it appears that cells are better able to maintain pH_i in HCO₃⁻-buffered solution. To test this suggestion further, experiments were performed in which the solutions were interchanged. Following a change from HCO₃⁻+ Hepes to Hepes (see Fig. 3), pH_i rapidly increased; this corresponds to an efflux of CO₂ from the cells and conversion of intracellular HCO₃⁻ and H⁺ to water and CO₂. Subsequently, pH_i decreased and after 400 s was always lower than before the change (δ = 0.14 ± 0.02, P = 0.04, 7 out of 7). Similarly following a change from Hepes to Hepes + HCO₃⁻ solutions, pH_i initially decreased but after 300 s was higher (in 6 out of 7 experiments) than before the change (δ = 0.07 ± 0.02, n = 7, P = 0.04).

Rapid change in pH_i when CO₂ is added or withdrawn requires carbonic anhydrase within the cells to catalyse the interconversion of CO₂ and HCO₃⁻. It may also be aided by carbonic anhydrase exposed to the stirred external solution as there is evidence that a carbonic anhydrase isoform (CAIV) is present on the external surface of these cells (Ghandour et al. 1992). The importance of carbonic anhydrase was confirmed by the observation that the membrane-permeant carbonic anhydrase inhibitor, ethoxzolamide (6-ethoxy-2-benzothiazolesulphonamide) (Maren, 1963; Cousin et al. 1975) clearly blunts the response (see Fig. 3). The increase in pH_i measured 20 s after CO₂/HCO₃⁻ removal was reduced from 0.63 ± 0.6 to 0.17 ± 0.4 by 2 μm ethoxzolamide and to 0.18 ± 0.18 by 20 μm (n = 5, P < 0.002 at both concentrations). Similar effects were observed with acetazolamide (data not shown).

Rate of cellular acid gain and loss in the presence or absence of HCO₃⁻

The resting state pH_i of these cells (strictly the slow baseline rate of acid gain for cells in the experimental solutions) is presumed to be governed by a balance between the rates of acid loading and acid extrusion (see Discussion). As shown above, the rate of change in pH_i was smaller for cells in the presence than in the absence of HCO₃⁻. The cells are, however, more strongly buffered in the CO₂/HCO₃⁻-buffered solutions (at pH_i 7.1 β_total = 48 mm in the presence of HCO₃⁻ compared to 25 mm in its nominal absence) and thus the rate of gain of acid calculated from the rate of change in pH_i was nearly the same with the two buffer solutions (see Table 3 rows 1 and 2 and Table 4 row A). This implies that the extra mechanisms of transport brought into play by the presence of HCO₃⁻ produce cancelling effects with little overall change in the rate of acid loading.

For cells that are not required to generate large differences in pH, the commonly encountered means for acid extrusion, which are indicated in Fig. 1, entail either exchange of intracellular H⁺ for external Na⁺ or cotransport of HCO₃⁻ and Na⁺ into the cell. The most commonly encountered means for acid loading are Cl⁻/HCO₃⁻ exchange and channel-like permeability to H⁺ or HCO₃⁻. Therefore experiments involving removal of Na⁺ or Cl⁻ were undertaken in the absence or presence of HCO₃⁻ (see Tables 3 and 4) as described below.

Evidence for the presence of a Na⁺/H⁺ exchanger

Experiments to demonstrate the presence of a NHE-like Na⁺/H⁺ exchanger were conducted in the nominal absence of HCO₃⁻. Replacement of Na⁺ with NMDG⁺ (Table 3 row 3) (see Fig. 4) and addition of 5 μm EIPA (Table 3 row 4) produced equal, small increases (Table 4 row C) in the rate of acid gain. Because EIPA is a selective inhibitor of the NHE transporters this suggests that a transporter of the NHE family is the major mechanism of Na⁺-dependent acid extrusion in the nominal absence of HCO₃⁻. The extrusion partially offsets a somewhat larger rate of acid loading.

Figure 4. Effect on pH_i of replacing Na⁺ with NMDG⁺.

The black trace shows the response with HCO₃⁻-buffered solution, the grey trace with Hepes-buffered solution. The loading solutions were replaced with dye-free solutions at 800 s, these were replaced with Na⁺-free solutions at 1000 s, which were in turn replaced by the normal Na⁺-containing solutions at 1600 s. The dotted line indicates the time course for pH_i in cells in the HCO₃⁻-buffered solution if the Na⁺ had not been removed. Note that because the cell interior is more strongly buffered in the presence of HCO₃⁻, these traces show that in the absence of Na⁺ the rate of acid gain is much greater in the presence of HCO₃⁻ than in its absence. For cells in Hepes-buffered solutions, the removal of Na⁺ hardly affects the time course. The symbols show means ± s.e.m.; n = 3 for Hepes and 4 for HCO₃⁻ (see Table 2 for statistics).

NHE activity usually increases markedly as pH_i decreases below 7 (Aronson, 1985; Orlowski, 1993). To confirm this property of the Na⁺-dependent, HCO₃⁻-independent acid extrusion seen here, the NH₄Cl prepulse technique was used to acidify the cells (see Fig. 5). After the pulse, pH_i fell rapidly to 6.36 ± 0.01 (n = 72) followed by a slower recovery. The initial rate of this recovery in Hepes-buffered solutions corresponded to a rate of H⁺ loss of 0.27 ± 0.02 mm s⁻¹ (n = 72), which is about 35× faster than observed for pH_i near 7.1. The rate of recovery when Na⁺ was replaced by NMDG⁺ was reduced by 80 ± 4%, significantly different from both 0 (P < 0.001) and 100% (P < 0.01) (Fig. 6). The rate of acid removal was significantly reduced by EIPA in the range 0.5–5 μm (Fig. 6) with an apparent IC₅₀ of ∼0.5 μm (least squares fit including the data for NMDG⁺ as an estimate of the non-inhibitable recovery). DIDS at 250 μm, which should not inhibit NHE- mediated transport, did not produce any significant inhibition (P = 0.38 paired, n = 8).

Figure 5. Example of an NH₄⁺ pulse experiment.

Cells were initially in Hepes-buffered solution. They were transferred into solution containing 20 mm NH₄⁺ for the indicated period and subsequently Hepes-buffered solution again. A single exponential fit to the recovery phase is shown as the smooth curve. The fitted curve is used as a computational aid for calculating the initial value and slope of the recovery curve.

Figure 6. Inhibition by EIPA or by removal of Na⁺ of the rate of acid extrusion following acidification by the ammonium pulse technique.

The rate was measured in the nominal absence of HCO₃⁻. The control rate of extrusion in these experiments was 0.33 ± 0.03 mm s⁻¹ (n = 21). Bars with numbers indicate paired t tests with the number of measurements and the P values.

Evidence for the presence of a NBC-like Na⁺−HCO₃⁻ cotransport

In the presence of HCO₃⁻, inhibition (or reversal) of Na⁺-dependent transport by replacement of Na⁺ with NMDG⁺ produced a larger increase in the rate of acid gain (Table 3 row 5) (see Fig. 4) than was seen in the absence of HCO₃⁻ (Table 3 row 3) suggesting the presence of a Na⁺- and HCO₃⁻-dependent acid extruder. Removal of Na⁺ after prior removal of Cl⁻ (see below) still produced a clear increase in the rate of acid gain (Table 3 row 10), implying that most, possibly all, of the Na⁺-dependent HCO₃⁻ transport is Cl⁻ independent. DIDS is an inhibitor of many forms of Cl⁻ and HCO₃⁻ transport including acid loaders (such as the AEs) and acid extruders (such as the NBCs). Addition of 250 μm DIDS blocked the increase in rate of acid gain seen when Na⁺ was removed (compare Table 3 row 6 with row 5). These are the properties expected if Na⁺-dependent HCO₃⁻ transport into the cells is mediated by a transporter like NBCe1, NBCe2 or NBCn1 (Romero et al. 2004; Boedtkjer et al. 2006; Damkier et al. 2006).

Evidence for the presence of an AE-like Na⁺-independent Cl⁻/HCO₃⁻ exchange

In the absence of Na⁺, a higher rate of acid gain was seen in the presence of HCO₃⁻ than in its absence (Table 4 row D). This implies the presence of a Na⁺-independent, HCO₃⁻-dependent acid loader. Removal of external Cl⁻ in the presence of HCO₃⁻ (see Fig. 7A) produced an alkaline shift in pH_i (Table 3 row 7) followed by a return to a net rate of acid gain (Table 3 row 8). This can be interpreted as a transient reversal of a Cl⁻/HCO₃⁻ exchanger followed by its inactivity when internal Cl⁻ is depleted. Removal of Cl⁻ in the absence of Na⁺ (Table 3 row 9) produced a change in the rate of acid gain similar to that seen in the presence of Na⁺ (Table 3 row 7).

Figure 7. Effect of Cl⁻ removal on pH_i.

Time course in the presence of HCO₃⁻ with a change from NaCl to sodium gluconate at time 0 in the absence, upper trace, or presence, lower trace, of 250 μm DIDS. In the absence of DIDS an alkaline shift is followed by a return to a gradual gain of acid. DIDS abolishes the alkaline shift. Representative traces (see text and Table 3 for initial slopes and statistics).

In the presence of HCO₃⁻, addition of 250 μm DIDS, which inhibits AE transporters, produced little change in the net rate of acid gain when Na⁺ was present (Table 4 row B) but blocked or even reversed the increase in the rate of acid gain seen when Na⁺ was removed (compare Table 3 row 6 with row 5). As already noted, these observations imply that the acid extrusion mechanism inhibited by removal of Na⁺ was already inhibited by the DIDS. This in turn implies that DIDS must also have inhibited a balancing acid loader so that its net effect in the presence of Na⁺ was nearly neutral. Confirming this inference DIDS greatly slowed the alkaline shift when Cl⁻ was removed in the presence of HCO₃⁻ (see Fig. 7A), inhibiting 86 ± 8% (P < 0.002, n = 4) of the initial change in slope.

Evidence for the expression of HCO₃⁻ transporters in brain microvascular endothelial cells

Expression at the mRNA level of NBCe1, NBCe2, NBCn1, NCBE, AE1, AE2 and AE3 calculated relative to that for NHE1 was investigated using real-time PCR (see Fig. 8). A product with a well-defined melting point and of the expected size was found with primers specific for each of these transporters. Transcripts of NHE1, AE2 and NBCn1 were comparably abundant with that for NBCe1 present at ca 1/3rd this level, while those for AE1, AE3, NBCe2, NCBE and NDCBE were detected but at lower levels. For comparison data are also reported for mRNA isolated from chloroid plexus (isolated as in Redzic et al. 2005) and whole kidney, which was used as a positive control. While the relative levels of expression of mRNA do not imply the same relative levels of expression for protein, these results do provide suggested molecular candidates for the transporter functions, i.e. AE2 for Cl⁻/HCO₃⁻ exchange and either NBCn1 or NBCe1 for the Na⁺−HCO₃⁻ cotransporter.

Figure 8. Transcript levels of bicarbonate transporters calculated relative to NHE1 in RNA samples from rat brain microvascular endothelial cells, choroid plexus and whole kidney.

Results are shown as means ± s.e.m.n = 5–10 for RBEC, 2 for choroid plexus and 5 for kidney (except 2 for NDCBE).

Discussion

This study addresses the means by which brain endothelial cells are able to regulate their internal pH and secrete HCO₃⁻. These cells produce large quantities of acid from the metabolism of glucose. However, this metabolic challenge is dealt with by efflux of the acid products of metabolism, e.g. by diffusion outwards of CO₂ and the facilitated, coupled transport of lactate⁻ and H⁺ via the monocarboxylate transporters (Halestrap & Price, 1999). Thus the rate of net gain of acid within the cells is normally the difference between the rate of acid loading from outside, some of which can be passive, and the rate of acid extrusion, some of which must be active. Maintaining an average net rate of acid gain within the cells of zero can require careful balance of relatively large fluxes. In cells such as brain endothelial cells that are not called upon to generate large pH gradients, the normal source of energy for active acid extrusion is the Na⁺ concentration gradient generated by the Na⁺ pump.

Rat brain endothelial cells maintained in primary culture and then minimally disturbed were found to have a basal pH_i estimated to be 7.15 ± 0.01. This value is close to the values observed in previous studies, 7.18 ± 0.02 (Hsu et al. 1996) and 6.9 ± 0.08 (Sipos et al. 2005). The slow acidification of these cells under our experimental conditions suggests that some effect of transfer of the cells from growth medium to the experimental solutions, loading with BCECF and illumination has either reduced an acid extrusion system or increased an acid-loading system (which may be a channel-like permeability) to disturb the balance which must obtain in the growing cells. This slow acidification, like the normal balance, is the net result of competing processes. The results presented here provide evidence for activity in rat brain endothelial cells of several different ion transporters (see Fig. 1). These include a Na⁺/H⁺ exchanger, a DIDS-sensitive Na⁺–HCO₃⁻ cotransporter that continues to function in the absence of Cl⁻ and a DIDS-sensitive Na⁺-independent Cl⁻–HCO₃⁻ exchanger. This study reports the first evidence in brain microvascular endothelial cells for Na⁺-independent Cl⁻/HCO₃⁻ exchange and Cl⁻-independent Na⁺-HCO₃⁻ cotransport and the first determinations of the rates of Na⁺/H⁺ exchange and Na⁺−HCO₃⁻ cotransport for internal pH near the resting value. The results also provide the first demonstration of carbonic anhydrase activity in these cells. The transport rates observed near pH 7.1 are modest, which may explain why they have not been reported previously. But, nevertheless, they are sufficiently large that if the transporters are appropriately disposed on the two sides of the cells they could account for the net secretion of HCO₃⁻ across the blood–brain barrier.

In the absence of Na⁺, the presence of CO₂/HCO₃⁻ increases the net rate of acid loading in the cells, implying the existence of a Na⁺-independent HCO₃⁻ exit mechanism. This mechanism is inhibited by DIDS. In the presence of HCO₃⁻, removal of external Cl⁻ produces a transient alkalinization as if it were the reversal of a Cl⁻/HCO₃⁻ exchanger. In this explanation the reversal is transient because it is limited by the supply of Cl⁻ within the cells.

In the presence of HCO₃⁻ removal of Na⁺ produces a sustained increase in the rate of acid gain which in many experiments was preceded by a short period of more rapid acid gain (see Fig. 4). The behaviour is plausibly explained as transient reversal of the Na⁺-dependent acid extrusion process followed by its inactivity. The sustained rate of acid gain would then result from the combination of the ‘channel-like’ permeability and the continued function of the Cl⁻/HCO₃⁻ exchanger. Indeed the activity of the exchanger may be increased, as the abolition of Cl⁻ entry via the Na⁺−K⁺−2Cl⁻ cotransporter, NKCC1 (O'Donnell et al. 2004; Foroutan & O'Donnell, 2005), will lead to decreased intracellular Cl⁻ concentration and hence a more favourable Cl⁻ gradient for Cl⁻/HCO₃⁻ exchange.

The transient net acid extrusion when Cl⁻ is removed in the presence of HCO₃⁻ (see Fig. 7A) is simply and plausibly explained as reversal of Cl⁻/HCO₃⁻ exchange. However, if the only effect of Cl⁻ removal is on the anion exchanger, the rate of the sustained acid gain after the transient should be slower than the rate observed in the presence of Cl⁻. This has not been observed (Table 3, row 9). This is evidence that Cl⁻ removal has an additional effect. The data could be explained if Cl⁻ removal also inhibits a component of acid extrusion like that which would be mediated by a Na⁺ driven Cl⁻/HCO₃⁻ exchanger, e.g. NDCBE or possibly NCBE. The present data are consistent with this type of explanation but without quantitative modelling taking into account factors like changes in membrane potential they cannot be said to require it. Comparison of the increase in the rate of acid gain produced by Na⁺ removal in the absence of Cl⁻ (Table 3 row 10) with that seen in the presence of Cl⁻ (Table 3 row 5) suggests that the major component of Na⁺-coupled HCO₃⁻ transport is Cl⁻ independent.

The properties of the acid extrusion and the acid loading in the presence of HCO₃⁻ correspond to those that could be mediated by known transporters. Na⁺ and HCO₃⁻ dependence, DIDS sensitivity and independence of Cl⁻ are properties of many NBC-like Na−HCO₃⁻ cotransporters (including NBCe1, NBCe2 and, possibly, NBCn1) which would function as an acid extruder in these cells. The mRNAs for NBCe1 and NBCn1 are clearly present in the cells. Cl⁻ and HCO₃⁻ dependence, Na⁺ independence and DIDS sensitivity are properties of the Cl⁻/HCO₃⁻ exchangers of the AE family and some members of the anion exchanger (SLC26) superfamily (Mount & Romero, 2004). There is clear expression of mRNA for AE2 in these cells.

In the nominal absence of HCO₃⁻ the principal acid extruder appears to be an NHE because the mechanism is inhibited by Na⁺ removal or by EIPA, a selective inhibitor of the NHEs, and it becomes much more prominent as pH_i is reduced. In addition to mRNA for NHE1 these cells also express mRNA for NHE2, 3 and 4 (Sipos et al. 2005) and NHE5 (Taylor et al. 2002). In the absence of Na⁺ there is still a significant rate of recovery from an acid load imposed by an NH₄Cl prepulse that reduces pH_i to ∼6.3. Sipos et al. (2005) reported no recovery in the absence of Na⁺, but the acid loads they imposed reduced pH_i only to 6.5. At the lower pH_i of 6.3 the electrochemical gradient for H⁺ is likely to be outwards and that for HCO₃⁻ inwards and the recovery could be mediated by a ‘channel-like’ permeability or reversal of the Cl⁻/HCO₃⁻ exchanger.

Molecular candidates for the transporters mediating the functions observed can be suggested by the pattern of expression levels of mRNA observed using real-time PCR. A comparison between mRNA derived from brain microvascular endothelial cells, choroid plexus epithelial cells and extracts from whole kidney (used as a positive control) is shown in Fig. 8. For each tissue or cell type the expression levels have been normalized to the expression for NHE1 as that transporter is expected to be highly expressed in all. The patterns of expression are very different. The striking feature of the results for the kidney is the prominent expression of NBCe1, which is hardly surprising as basolateral NBCe1 and apical NHE3 in proximal tubule cells mediate the bulk of renal HCO₃⁻ reabsorption. In choroid plexus there is high expression of NCBE, NBCn1 and NBCe2. This is consistent with the results from PCR and immunohistochemistry reported previously (Praetorius et al. 2004b; Bouzinova et al. 2005; Damkier et al. 2006). There is also clear expression of NHE1 consistent with previous binding studies (Kalaria et al. 1998). However, it is should be noted that NHE proteins have not been detected in choroid plexus using immunohistochemistry (Praetorius et al. 2004b). In the brain endothelial cells the prevalent mRNA species are those for the reference Na⁺/H⁺ exchanger NHE1, the Cl⁻-independent Na⁺−HCO₃⁻ cotransporters, NBCn1 and NBCe1 and the Na⁺-independent Cl⁻/HCO₃⁻ exchanger, AE2. mRNAs for NCBE, which may be a Na⁺-dependent Cl⁻/HCO₃⁻ exchanger, NDBCE, AE1, AE3 and NBCe2 were detected but at substantially lower abundance. Conventional PCR studies (amplification followed by visualization on gels) using RNA isolated from brain microvascular endothelial cells have previously shown expression for NHE1 and 5 (Taylor et al. 2002) and NHE1, 2, 3 and 4 (Sipos et al. 2005).

Table 5 presents a comparison of the transport activites observed in vascular endothelia from blood vessels outside the central nervous system. These have in common with brain microvascular endothelial cells the need to regulate intracellular pH but differ in that they do not mediate a net secretion of fluid. Estimates of the rates of transport exist only for endothelial cells from human umbilical vein. In that tissue the dominant mode of HCO₃⁻ entry at resting pH is Cl⁻ dependent while in the brain microvascular cells it is Cl⁻ independent. Damkier et al. (2006) have recently shown that NBCn1 is present in endothelial cells from many locations in the vasculature outside the CNS, which suggests an important role for Cl⁻-independent Na⁺−HCO₃⁻ cotransport in peripheral endothelia as well.

Table 5.

Comparison of transport properties for vascular endothelia

				Brain microvascular
Preparation	Aortic Umbilical vein endothelial cells	Pulmonary endothelial cells	endothelial cells endothelial cells	Other studies	This study
Basal pH with HCO3−	7.26	∼7.2		6.9, 7.18**	7.15
Basal pH without HCO3−	7.21	∼7.2, 7.08*	7.02	6.83	7.1
Intrinsic buffer capacity near 7.1	15 mm		69 mm		25 mm
Near pH 7.1	∼17 μm s−1	Yes		Yes*, yes**	∼4 μm s−1
Na+/H+ exchange
Na+-dependent HCO3− influx	yes	Yes			∼25 μm s−1
Cl− dependent	∼17 μm s−1				suggestion
Cl− independent					∼17 μm s−1
Na+-independent Cl−/HCO3− exchange		Yes	See note*†		∼28 μm s−1 when reversed by Cl− removal
after NH4Cl prepulse#.	Yes*	Yes*	Yes	Yes, yes**	∼270 μm s−1 at pH 6.36
Na+/H+ exchange	> 50 μm s−1 at pH 6.9
Na+-dependent HCO3− influx	Yes	Yes*		Yes
Cl− dependent	> 34 μm s−1 at pH 6.9	Yes*		Yes data not shown
Net transport	None	None	None	secretion: apical (blood) to basolateral (brain)
References	Sun et al. (1999)	Ziegelstein et al. (1992)	Cutaia & Parks (1996)	Sipos et al. (2005)	This study
	Escobales et al. (1990)*	Ziegelstein et al. (1998)	Cutaia et al. (1998)	Vigne et al. (1991)*
		Wittstein et al. 2000,Faber, 1998*	Nozik-Grayck et al. (2003)*	Hsu et al. (1996)**

All HCO₃⁻ transport indicated has been shown to be either DIDS or H₂DIDS sensitive, all Na⁺/H⁺ exchange has been shown to be sensitive to either EIPA or DMA, (#) or other form of imposed acidification.

^†Nozik-Grayck et al. (2003) demonstrated ·O_2•⁻/HCO₃⁻ exchange which was presumed to be mediated by the Na⁺-independent Cl⁻/HCO₃⁻ exchanger, AE2.

The HCO₃⁻ and H⁺ transport demonstrated here is consistent with the same repertory of transporters that have been seen in many epithelia and in immortalized cell lines (for reviews see Reinertsen et al. 1988; Bonanno, 2003; Mount & Romero, 2004; Orlowski & Grinstein, 2004; Romero et al. 2004). Table 6 compares the results obtained with brain microvascular endothelial cells with those for choroid plexus and corneal endothelium (corneal posterior epithelium), both of which secrete HCO₃⁻. The former is included because it is the alternative source of fluid secretion into the brain, the latter because it illustrates some of the differences between secretory tissues that can be observed. Each of these three presents a different pattern of transporters. This serves to emphasize that the properties of each cannot be inferred but must be determined directly by experiments on the cell type in question.

Table 6.

Comparison of brain microvascular endothelial cell transport and transporters with those of choroid plexus epithelium and corneal endothelium (corneal posterior epithelium). Results from this study are in bold

Transport activity	Brain microvascular endothelial cells	Choroid plexus	Corneal endothelium
Basal pH with HCO3−	7.151, 6.92	7.38	7.151, 7.372
Na+/H+ exchanger	Functional at pH 7.11	NHE1 (mRNA)1	Functional1,2
	Functional at acid pH1,2
	NHE1 (mRNA)1
	NHE1 binding assay3
	NHE1-4 (mRNA)2
Na+-dependent,	Functional at pH 7.1 (presumed apical)1	NBCe2(ap)2	Functional1,3,4
Cl− independent	NBCe1 (mRNA)1	NBCn1 (bl)3	bl5
HCO3− influx	NBCn1 (mRNA)1		NBCe15,6,7
Na+ dependent,	Transport at acid pH2	NCBE# (bl)3
Cl− dependent	NCBE# (mRNA)1
HCO3− influx	NDBCE (mRNA)1
Na+-independent Cl−/HCO3− exchange	Functional1	AE2 (bl)4,5,6	Functional1,4
	AE2 (mRNA)1		ProbablyAE2 (bl)5
Na+,K+-ATPase	bl4	ap7,8	bl8
Na+−K+−2Cl− cotransport	ap5	ap6	bl5,9
Net transport	ap (blood) to bl (brain interstitial fluid)	bl (blood) to ap (CSF)	bl (corneal stroma) to ap (aqueous humour)
Secretory function	Secretion of brain	Secretion of CSF	Maintenance of osmotic pressure difference to oppose stromal swelling pressure
	Interstitial fluid
Embryological origin	Mesoderm6	Neural tube	Neural crest10
References	1. This study	1. Kalaria et al. (1998)	1. Jentsch et al. (1988)
	2. Sipos 2005	2. Bouzinova et al. (2005);	2. Bonanno & Giasson (1992a)
	3. Kalaria et al. (1998)	3. Praetorius et al. (2004b)	3. Jentsch (1984)
	4. Betz et al. (1980); Betz (1983)	4. Lindsey et al. (1990)	4. Bonanno & Giasson (1992b)
	5. O'Donnell et al. (2004)	5. Alper et al. (1994)	5. Bonanno (2003)
	6. Gage et al. (2005)	6. Wu et al. (1998)	6. Li et al. (2005)
		7. Mazusawa et al. (1984)	7. Usui et al. (1999)
		8. Ernst et al. (1986)	8. Guggenheim & Hodson (1994)
		Reviews: Brown et al. (2004) and Praetorius & Nelson (2006)	9. Jelamskii et al. (2000)
			10. Gage et al. (2005)

ap, apical; bl, basolateral. #The Cl⁻ dependence of NCBE remains controversial; see Wang et al. (2000), Choi et al. 2002 and Romero et al. (2004). The assignment adopted here follows that of Bouzinova et al. (2005) and Praetorius & Nielsen 2006.

It is possible from the net rate of acid extrusion via Na⁺− HCO₃⁻ cotransport reported here to calculate a net influx of HCO₃⁻ per unit area of the endothelial cells. The HCO₃⁻ would leave the cells via either the Cl⁻/HCO₃⁻ exchanger or a ‘channel-like’ permeability to HCO₃⁻. If one assumes a 0.5 μm thickness for the cells, the estimate is ∼25 μm s⁻¹× 0.5 μm = 1.25 pmol cm⁻² s⁻¹. This flux is small compared to those seen in transporting epithelia but nevertheless large enough to account for current estimates of HCO₃⁻ secretion across the blood–brain barrier. Assuming that the surface area of the capillaries in brain tissue is ∼100 cm² g⁻¹ (see, e.g. Bradbury, 1979) and for a 1000 g brain, this secretion rate is ∼10 mmol day⁻¹. Thus the transporters described here could account for current estimates, 1–5 mmol day⁻¹, of net secretion of HCO₃⁻ into the brain (Milhorat, 1987; Cserr & Patlak, 1992; Abbott, 2004; Redzic & Segal, 2004).

In summary when cultured rat brain endothelial cells are transferred from growth medium to a HCO₃⁻-buffered solution, the modest requirement for acid efflux appears to be largely but not completely met by Na⁺−HCO₃⁻ cotransport. Following intracellular acidification, as might occur in vivo during hypoxia, both Na⁺/H⁺ exchange and Na⁺−HCO₃⁻ cotransport (with or without exchange for Cl⁻) (Sipos et al. 2005) is likely to be important in the responses of the cells. HCO₃⁻ secretion by the endothelial cells requires a means of base entry into the cells (or equivalently acid extrusion) on one surface of the cell and of base extrusion (acid loading) on the other. This study provides evidence for at least two mechanisms for acid extrusion, an NBC-like Na⁺−HCO₃⁻ cotransporter and an NHE-like Na⁺/H⁺ exchanger, and two mechanisms for acid loading, channel-like permeability and an exchanger. The most important acid extruder in HCO₃⁻-buffered solutions is a Na⁺−HCO₃⁻ cotransporter (possibly NBCn1 or NBCe1). The transporters identified functionally in vitro can account for secretion of HCO₃⁻ provided the Na⁺-linked influx of HCO₃⁻ occurs primarily across the luminal membrane while the Na⁺-independent efflux of HCO₃⁻ occurs primarily across the abluminal surface. Multiple molecular candidates for the transporters that may fulfil these roles have been identified from real-time PCR measurements of mRNA. Determining which are present at the level of protein and how they are localized within the cells will be the subject of future studies.