Ion transporters in brain endothelial cells that contribute to formation of brain interstitial fluid

Ions and water transported across the endothelium lining the blood–brain barrier contribute to the fluid secreted into the brain and are important in maintaining appropriate volume and ionic composition of brain interstitial fluid. Changes in this secretion process may occur after stroke. The present study identifies at transcript and protein level ion transporters involved in the movement of key ions and examines how levels of certain of these alter following oxidative stress. Immunohistochemistry provides evidence for Cl−/HCO3− exchanger, AE2, and Na+, HCO3− cotransporters, NBCe1 and NBCn1, on brain microvessels. mRNA analysis by RT-PCR reveals expression of these transporters in cultured rat brain microvascular endothelial cells (both primary and immortalized GPNT cells) and also Na+/H+ exchangers, NHE1 (primary and immortalized) and NHE2 (primary cells only). Knock-down using siRNA in immortalized GPNT cells identifies AE2 as responsible for much of the Cl−/HCO3− exchange following extracellular chloride removal and NHE1 as the transporter that accounts for most of the Na+/H+ exchange following intracellular acidification. Transcript levels of both AE2 and NHE1 are increased following hypoxia/reoxygenation. Further work is now required to determine the localization of the bicarbonate transporters to luminal or abluminal membranes of the endothelial cells as well as to identify and localize additional transport mechanisms that must exist for K+ and Cl−.

of CSF, with the remainder ascribed to the choroid plexuses [1, 15-17, 21, 46, 57, 74]. 1 Secretion of fluid by the blood-brain barrier must also be involved in local brain tissue swelling, i.e. focal oedema. Tissue volume increases must reflect either increased secretion or maintained secretion with decreased net rate of removal of fluid from the region to the CSF [67]. If there were to be changes in production or outflow of CSF, this would involve changes in fluid volume in spaces outside the brain parenchyma and so be associated primarily with hydrocephalus. Thus, to understand normal homeostasis of brain interstitial fluid and its failures, e.g. after ischaemic stroke with or without reperfusion, it is essential to understand the secretion process at the blood-brain barrier.
The mechanism of secretion of Na + and Cl − across the endothelial cells that line the blood vessels forming the blood-brain barrier is known in broad outline (for a review, see [52]). Na + and Cl − enter the endothelial cells largely via a Na + , K + , 2 Cl − cotransporter, NKCC1, located in the luminal (blood facing) membrane [26,54,72], while the Na + leaves these cells largely via the Na + , K + -ATPase, the Na + pump, located in the abluminal (brain facing) membrane [5,6,25,26,48,65]. The exit route for Cl − is unknown: possibilities include anion channels and K + , Cl − cotransporters.
For HCO 3 − the major evidence for its mechanism of transport is the demonstration in cultured rat brain endothelial cells that under ambient conditions there is both Na + , HCO 3 − cotransport that alkalinizes the cells and Na + -independent Cl − /HCO 3 − exchange that acidifies the cells, these occurring at comparable rates for intracellular pH (pH i ) near 7.1 [49,76]. Thus, if Na + , HCO 3 − cotransport were to be primarily at the luminal, blood-facing, membrane and the Cl − /HCO 3 − exchange at the abluminal, brain-facing, membrane, then there would be net secretion of HCO 3 − into the brain. Combined activity of carbonic anhydrase [27,76] with Na + /H + exchange produces the same result as Na + , HCO 3 − cotransport, but at pH i above 7 the rate would be much less [49,76]. Only at lower pH i , e.g. 6.5, does Na + /H + exchange become dominant [49]. The molecular identities of the transporters likely to move HCO 3 − at the blood-brain barrier have not so far been determined. Brain endothelial cells are known to express appropriate DIDS-sensitive SLC4A transporters. The cells also express NHE1 and NHE2 whose properties are suitable for the Na + /H + exchange [41,68,76]. The present study identifies at transcript and protein level the ion transporters that could account for some of the activities observed. Immunohistochemistry provides clear evidence for AE2 and indications for NBCn1 and NBCe1 in the microvessels. Results from knock-down experiments using siRNA in cells of the GPNT brain endothelial cell line indicate that AE2 is responsible for the major part of the Cl − / HCO 3 − exchange observed when Clis removed from the external solution and that NHE1 accounts for most of the Na + /H + exchange observed following intracellular acidification. Both AE2 and NHE1 transcripts are shown to be upregulated following hypoxia/reoxygenation.

Immunohistochemistry
Paraffin-embedded tissue Brain and kidney were fixed by perfusion of halothane-anaesthetized animals via the abdominal aorta with 4 % paraformaldehyde in 0.1 M cacodylate buffer, pH 7.4, according to the protocol of Praetorius [58]. Each organ was then cut into two equal halves, dehydrated using increasing concentrations of ethanol, embedded in paraffin and cut into 2 μM sections using a rotary microtome (Leica, Heidelberg, Germany). Sections were incubated in xylene overnight before staining. Initial rehydration was carried out by exposing slides to decreasing concentrations of ethanol. Endogenous peroxidase activity was blocked by 30 min of incubation in 0.5 % H 2 O 2 in absolute methanol. Some sections were exposed to 0.1 % sodium dodecyl sulphate for 30 min prior to antigen retrieval that involved boiling in 1 mM Tris, pH 9, supplemented with 0.5 mM EGTA for 13 min. Non-specific binding was limited by exposure to 50 1 The mechanisms by which ions and water exchange between ISF and CSF have not been fully established. When viewed over distances less than a few hundred microns, movements of solutes and water within the ISF appear as diffusion (for reviews, see [21] and [73]). Viewed over larger distances, there is a superimposed, convective component, which can occur both towards and away from CSF and is primarily localized to the perivascular spaces surrounding blood vessels larger than capillaries [1, 15-17, 32, 33]. The uncertainty concerns the relative importance of convection and whether or not there is continuous circulation of fluid and solutes across the outer surfaces of the brain from CSF into ISF and back to CSF [1,32,33]. Such circulation may be very important for the delivery of intrathecally administered drugs to within diffusion distance of their targets (see e.g. [30]) but does not alter the need for a net efflux of ions and water to CSF to balance secretion across the blood-brain barrier.
Frozen tissue Whole rat brains were 'snap'-frozen in isopentane at −18°C on dry ice and stored at −80°C.
Twenty-micrometre sections cut on a cryostat were thawmounted onto Vectabond™ (Vector Laboratories) coated slides, fixed in methanol at −20°C for 2 min, and rehydrated in Tris-buffered saline (TBS, 50 mM Tris-HCl, pH 7.6, 150 mM NaCl) containing 0.2 % (v /v ) Tween20 (TBS/Tween), and non-specific binding sites were blocked with 50 μg ml −1 BSA in TBS/Tween before application of the first primary antibody in blocking solution for 1 h at room temperature. Control slides were exposed for the same length of time but to blocking solution alone. Slides were then washed well and treated for 1 h at room temperature with the secondary antibody mixture containing Alexa Fluor™ 488 goat highly cross-absorbed anti-mouse IgG (H+L) conjugate at 1:50 (40 μg/mL) and Alexa Fluor™ 594 goat highly cross-absorbed anti-rabbit IgG (H+L) conjugate at 1:150 (13.33 μg/mL) (Molecular Probes, Portland, OR, USA). Sections were then washed well and mounted in glycerol/PBS (9:1v /v ) containing an anti-fading agent (DABCO at 2.5 mg ml −1 ) and visualised by fluorescence microscopy. An SDS epitope unmasking procedure was carried out for detection of both AE1 and AE2 [11]. Without this treatment, SA6 antibody stains mainly AE1, whereas the antibody reacts with both AE1 and AE2 following epitope unmasking. Briefly, the procedure involves fixing the sections in a solution containing 4 g paraformaldehyde, 2.739 g L -lysine and 0.428 g sodium metaperiodate in 200 ml phosphate buffer, pH 7.4, for 30 min at room temperature or immersion-fixing overnight at 4°C. The sections were then washed and rehydrated in PBS containing 0.2 % (v /v ) Tween-20 and incubated with 10 mg ml −1 sodium dodecyl sulphate in the same buffer for 4 min before treatment with primary and secondary antibodies as described earlier. Previous studies using this method have allowed immunolocalization of AE2 in basolateral membranes of several rat kidney tubular epithelial cell types [2] and on the basolateral membrane of epithelial cells in the initial segment and intermediate zone of the rat epididymis [35].
Sections were visualised by fluorescence microscopy, with light being provided either by a mercury arc lamp (Zeiss Axioskop) or a scanning confocal laser (Biorad Micro-Radiance). With the Bio-Rad system, sequential images were obtained (Kalman average of eight to ten individual scans) and the two parent images were merged using Lasersharp software.
Antibodies The antibodies used with thin sections cut from wax-embedded tissue are listed in Table 1. For the work with thick sections cut from frozen tissue, the same antibodies were used for AE2, AQP4 and NBCe1. Other antibodies used with thick sections are listed in Table 2.
The GPNT cell line [62] is an immortalised rat brain endothelial cell line derived from the GP8 cell line [29]. Cells used in this study were a kind gift from Dr. I. Romero (Open University, Milton Keynes, UK). Cells were maintained on collagen-coated flasks in a 1:1 mix of Ham's F-10: MEM-α supplemented with 10 % (v/v) FBS, 2 mM L -glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin, 2 μg/ mL basic fibroblast growth factor (First Link) and 5 μg/mL puromycin.

Application of oxidative challenge
Cells grown to confluency in multiwell plates were exposed to oxidative challenge in one of two ways as described in [63], either by direct application of medium containing 200 μM H 2 O 2 to each well or by applying hypoxic conditions to cells for 6 h followed with reoxygenation as described [63]. For this latter method, cells were treated with medium that had been pre-bubbled with 95 % N 2 in CO 2 for 3 min and then placed in a hypoxia chamber (Bilups-Rothenburg, www.brincubator. com) flushed with a mixture of 95 % air/5 % CO 2 and 95 % N 2 /5 % CO 2 and maintained at 1 % O 2 for 6 h. Cells were reoxygenated by feeding with fresh, fully oxygenated medium and were replaced into the incubator for the times indicated before analysis.

RT-PCR analysis and reference gene selection
Total cellular RNA was isolated using TriPure Isolation Reagent (Roche, Derbyshire, UK) and cDNA was generated by reverse transcription using random hexamers as described previously [63]. Fragments specific to the rat genes of interest were then amplified by PCR using SensiMix (Quantace Ltd, Watford, UK) containing SYBR green and the appropriate primer pairs (see Table 3). Reactions were run on Rotorgene 3000 (Corbett Research, Sydney, NSW, Australia). Thermal cycling conditions included initial denaturation for 10 min at 95°C followed by 45 cycles of 95°C for 30 s, 55°C for 30 s and 72°C for 30 s, with terminal elongation being performed at 72°C for 2 min with a melt curve analysis from 72 to 99°C. Data were analysed using the comparative quantitation option in Rotor-gene 3000 software (Corbett Research). As identified by and described in Robertson et al. [63], the reference gene most suitable for H 2 O 2 exposure experiments was beta-actin and for hypoxia/reoxygenation it was hypoxanthine-guanine phosphoribosyltransferase.

Luciferase reporter gene assay
To detect the activation of NHE1 promoter, a construct (kindly donated by Dr. Katarzyna Kania, Polish Academy of Sciences, Lodz, Poland) containing a NHE1-specific gene promoter sequence inserted into the pGL4. 10[luc2] vector was transfected by electroporation using the A m a x a e n d o t h e l i a l t r a n s f e c t i o n k i t ( L o n z a , Basel, Switzerland). A vector encoding a secretable form of alkaline phosphatase (pCMV-SEAP) under the control of a cytomegalovirus-derived promoter was co-transfected as described in Robertson et al. [63]. Following electroporation, cells were divided between six and nine wells of 24-well plates and maintained in medium for 24 h before treatment with hypoxia/reoxygenation as described earlier.
After the requisite treatment times, the medium was harvested for alkaline phosphatase activity and the cells were analysed for luciferase activity, with values being normalised to the alkaline phosphatase activity as described previously [63].
Use of siRNA targeted against AE2 or NHE1 siRNA transfection mixes were prepared as follows: 50 nM of the siRNA duplexes (Qiagen) (sequences shown below) was diluted in 100 μl Opti-MEM with gentle mixing, 3 μl of INTERFERin TM (Polyplus transfection), then added to the solutions and allowed to incubate for 10 min. To apply to GPNT cells already grown to~50 % confluency in 24 well plates, the growth medium was removed from each well and 0.5 ml fresh medium was added. A total of 100 μl of the siRNA mix was then added and the plate was incubated at 37°C. For RNA, cells were harvested after 24 h, and for fluorimetry or immunocytochemistry cells were trypsinised after 48 h and plated on glass coverslips.

Functional studies
Na + /H + and Cl − /HCO 3 − exchange activities were inferred from measurements of intracellular pH, pH i . This was measured as described in detail by Taylor et al. [76] using the fluorescent indicator 2′,7′-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein (BCECF) introduced into cells via exposure to membrane-permeant BCECF-AM. Fluorescence was excited alternately at a pH-insensitive wavelength 440 nm and a pH-sensitive wavelength 502 nm, and the pH was calculated from the ratio of the background-corrected emissions measured at 530 nm. A calibration curve for converting the ratio to pH was determined for rat brain endothelial cells using a combination of nigericin-high K + and weak acid/weak base null point methods [76]. Na + / H + exchange was assessed using an ammonium prepulse as described previously [49,76]. Briefly, cells were exposed to 20 mM NH 4 Cl, replacing the same amount of NaCl in the external Hepes-buffered solution (nominally HCO 3 − free) for 200 s followed by return to a fresh aliquot of the original solution. During the exposure, pH i initially shifts rapidly to a more alkaline value corresponding to diffusion into the cells of NH 3 and its combination there with H + to form NH 4 + . pH i then partly recovers, which, because the NH 3 concentration is then constant, implies a further increase in NH 4 + .
When the cells are returned to NH 4 Cl-free solution, all the NH 4 + exits as NH 3 , leaving behind H + that greatly reduces pH i . In the nominal absence of HCO 3 − , recovery is virtually abolished by either replacement of external Na + with NMDG + or application Jaulmes et al., 2006 [34] of 5 μM 5-(N-ethyl-N-isopropyl)amiloride (EIPA) [49,76]. The NHE1 inhibitor, cariporide (HOE642), was made up within 3 h of use as a 100-mM stock in distilled water. Cl − / HCO 3 − exchange was assessed from the increase in pH i when external Cl − was removed. This removal leads to a DIDS-inhibitable exchange of internal Cl − for external HCO 3 − [76].

Detection of transporters in rat brain
Immunohistochemical studies to detect the location of ion transporters in rat brain were initially carried out using a panel of antibodies selective for Na, K-ATPase, AE2, AQP4, AQP1, NHE1, NHE3, NBCe1 and NBCn1 as shown in Table 1. These were kindly supplied by Dr. Jeppe Praetorius.
The techniques followed those published by Praetorius and collaborators and involved 2-μM sections cut from waxembedded rat brain. These were able to detect the transporters concerned in choroid plexus or kidney, with examples shown for the first three in Fig. 1. However, with the exception of AQP4, which is known to be expressed at high levels in the astrocyte end-feet that surround the microvessels [50], none of the other antibodies produced detectable specific staining around the microvessels in the brain slices. An antibody against CD31, PECAM, which is known to be present in brain microvessels, also failed to produce detectable staining in the thin sections. It was concluded from this that though these techniques were appropriate for localizing transporters that are highly expressed, they were inadequate for detecting transporters that are expressed at lower levels as may be the case in or on the microvessel walls. Hence, these results indicated that the ion transporters of interest must be expressed in brain and the microvessels at levels lower than those found in the specialized transporting epithelia of choroid plexus and kidney.
To enhance the sensitivity of detection, immunohistochemistry using fluorescent secondary antibodies was applied to thicker sections cut from frozen blocks. The advantage of thicker sections is that a field of view will contain more transporters, allowing easier detection; the disadvantage is that spatial discrimination and resolution allowing structural features to be displayed are inevitably lost. It was however possible to locate blood vessels in the frozen sections using an anti-PECAM antibody. This is illustrated in Fig. 1f for anti-AQP4 and anti-PECAM antibodies raised in different species and hence detectable with different secondary antibodies. The fluorescence corresponding to the location of the AQP4 antibody (shown in red) is seen to surround that for the PECAM antibody (shown in green).
This approach was applied to detect AE2, NBCe1 and NBCn1 in frozen rat brain slices as shown in Fig. 2. To reveal the presence of AE2, it was necessary to use an unmasking procedure [11] before the application of the primary antibody as described in "Methods" and the figure legend. With unmasking, AE2 was seen to be distributed along the walls of the microvessels (Fig. 2c, d). The antibody used to detect AE2 also binds to AE1 within the red blood cells, which can be seen clearly with or without unmasking (Fig. 2a-d). With NBCn1, there is binding along the course of the vessels and also to other components in the brain, while with NBCe1 the same appears to occur but less clearly. Background staining with the secondary antibodies alone was minor, i.e. less than the diffuse staining seen in Fig. 2a, b. Immunohistochemistry provides evidence for the presence of AE2, NBCn1 and possibly NBCe1 on the microvasculature. No evidence was found using these methods for NBCe2, NCBE (NBCn2) or NDCBE protein along the microvessels consistent with either no expression or low expression comparable to that seen on the surrounding cells.
Evidence for ion transporter expression in cultured rat brain endothelial cells In order to explore the possible functional significance of the ion transporters detected in the brain vascular endothelium, it was necessary to undertake experiments on primary cultured rat brain endothelial cells (RBECs) and on immortalized rat brain endothelial cells (GPNT cells). To ensure that these cells expressed the ion transporters detected in vivo, mRNA analyses were firstly carried out.
As previously reported [76] for RBECs, compared to that for NHE1, a relatively large amount of mRNA for AE2, ca. sixfold less for AE3 and ca 100-fold less for AE1 was found. Of the NBCs, NBCe1 and NBCn1 were most highly expressed, NCBE and NDCBE somewhat less and NBCe2 ca. 80-fold less. In a more recent study where AEs were compared in primary cultured RBECs and in cells of the immortalized GPNT line, there was found to be less AE3 (relative to NHE1 in the same cell type) but somewhat more AE2 than seen previously (see Fig. 3).
Expression of NHE 2-5 relative to NHE1 was also examined. Results are shown in Fig. 3 for RBECs and GPNT. NHE2 mRNA is approximately 50-fold less than that of NHE1 in the RBECs and at least 500-fold less prominent relative to NHE1 in the GPNT cell line. Primers were validated using whole rat brain and whole rat kidney. They detected NHE 2, 4 and 5 at more than 10 % of the level for NHE1 in whole rat brain and detected NHE3 at the same level as NHE1 in whole rat kidney.
Experiments were also performed to determine whether oxidative challenge might alter the expression of AE2 and NHE1 and the activity of the NHE1 promoter. For this purpose, GPNT cells were used, these being more readily amenable to transfection of the reporter gene construct. Oxidative challenge was imposed by direct application of H 2 O 2 or by exposing the cells to 6 h of hypoxia, 1 % O 2 and 5 % CO 2 in reduced serum medium, followed by 16 h of normoxia, 5 % CO 2 in air. As shown in Fig. 4, both types of challenge significantly increased the expression of mRNA for AE2 and NHE1. Furthermore, using a luciferase reporter gene construct, it was also shown that hypoxia/reoxygenation increased the activity of the NHE1 promoter.
Evidence for ion transporter activity in cultured rat brain endothelial cells Having established the presence of mRNA for the above ion transporters in cultured rat brain endothelial cells, studies were undertaken to look for the functional activities of these transporters. Procedures for detecting the activity of various types of ion transporter in these cells were by monitoring changes in pH i following various challenges which have been presented earlier (Taylor, 2006;Nicola et al., 2008). Na + /H + exchange can be assayed from the rate of recovery of pH i after an acid challenge imposed by transient exposure to NH 4 Cl for cells in nominally bicarbonate-free medium. Cl − /HCO 3 − exchange can be assayed with cells in bicarbonate-buffered medium by the rate of alkalinization observed when external Cl − is removed. Na + /H + exchange in RBECs has been shown to be inhibited by EIPA with an IC 50 near 0.5 μM (Taylor, 2006). Inhibition of this exchange by cariporide in the GPNT cells is shown in Fig. 5 where the IC 50 Fig. 1 Immunohistochemical detection of Na, K-ATPase, AE2, AQP4 and PECAM in rat brain. a, c and e Images obtained from 2-μm wax embedded sections of brain stained for Na, K-ATPase, AE2 and AQP4 using the techniques previously employed to detect these proteins in choroid plexus and kidney. The staining in c shows AE1 in red blood cells within a larger blood vessel. b, d Images that illustrate positive controls for Na, K-ATPase and AE2 in choroid plexus obtained in parallel experiments conducted at the same time with the same materials. Similarly, negative results for the microvessels and positive results for appropriate controls were obtained with antibodies for NHE1, NHE2, NHE3, AQP1, NBCe1 and NBCn1 (not shown). Only for AQP4 (shown in e) was there a clear specific staining surrounding small blood vessels in the thin brain sections. No staining was obtained in the thin sections with antibody for PECAM (also known as CD31). By contrast, in a~20-μM frozen section from brain (f) stained for both PECAM (green) and AQP4 (red), a fluorescent signal for each is clearly evident following the course of the blood vessels. Scale bars: black 10 μM, white 50 μM cells as can be seen in the controls for the experiments reported in Fig. 7.
Another more specific method for identification of transporters is provided by transfection of siRNA to reduce the levels of transporter expressed. This technique was applied to both RBECs and GPNT cells but was found to be more effective in the immortalized GPNT cells. Using either of two different siRNA constructs in the GPNT cells, it was possible to reduce the level of mRNA for AE2 by about with excitation only at 540 nm detects only one or another of the transporters. With excitation at both 488 and 540 nM and emission detection at 530±30 nm (shown as green) and >600 nm (shown as red) PECAM is also detected. The sa6 antibody used in a-d recognises AE1 present in red blood cells, but after unmasking it also detects AE2 which is seen to outline the vessel with little labelling elsewhere. In e and f, anti-NBCn1 appears to label the blood vessels. In g and h, there is a suggestion that anti-NBCe1 labelling also follows the blood vessels. In these two frames, the red channel signal has been multiplied by 1.  Fig. 3 Expression of NHEs and AEs at mRNA level in cultured rat brain endothelial cells. Expression is shown relative to that for NHE1 in the same cell type. Black bars for RBEC primary cultured cells; open bars are for the GPNT immortalised cell line. Data are shown as mean ± s.e.m. n =5 or 6 except for AE1 RBEC n =1, AE1 GPNT n =1, AE3 RBEC n =2 and AE3 GPNT n =3. In the GPNT cells, there was no detectable expression of NHE2 in any experiment (NHE4 was not tested), while for AE1 there was no detectable expression in either RBEC or GPNT cells in two experiments. In GPNT cells, the expression of NHE1 relative to GAPDH was 0.017±0.001, n =6 Promoter activity was measured using luciferase reporter gene assay as described in "Methods". For H/R, expression levels were first normalized to those for HPRT under the same conditions, and the ratios were then compared between exposed cells and unexposed controls. For H 2 O 2 exposure normalization was to expression of β-actin. These reference genes have been shown previously to vary little with hypoxia/ reoxygenation and peroxide exposure, respectively [63]. All ratios shown (mean ± s.e.m.) are significantly different from 1. From left to right: n =8, p <0.01; n =5, p <0.05; n =4, p <0.01; n =4, p <0.05; n =6, p <0.05; ttests for the logs of the ratios compared to zero 50 %, and using one siRNA construct, similar results were obtained for NHE1 (see Fig. 6). That the reduction was not due to the transfection process itself was demonstrated from the lack of effect of transfection with the "AllStars" construct.
(Successful transfection of the fluorescent "AllStars" construct was confirmed by flow cytometry; data not shown.) Functional assays were undertaken in the transfected GPNT cells with altered NHE1 and AE2 expression, and the results are shown in Fig. 7. Example traces are given in Fig. 7a-f and results are summarized in Fig. 7g, h. The two siRNAs targeted at AE2 each produced a significant reduction in Cl − /HCO 3 − exchange (Fig. 7g), while the siRNA targeted against NHE1 significantly reduced Na + /H + exchange.

Detection of transporters and their functions
To identify ion transporters on brain microvessels, immunohistochemical staining using thin wax-embedded brain  Fig. 5 Inhibitory effect of cariporide on the rate of recovery of intracellular pH (pH i ) after an ammonium pulse in GPNT cells. pH i was measured using BCECF as described in "Methods". Data are shown as mean ± s.e.m., n =4. The square symbol to the right indicates the rate of change after complete replacement of external Na + by NMDG + , which should eliminate all transport via NHEs and is therefore taken as an estimate of the limiting value for pH i changes in the presence of large cariporide concentrations. The fitted curve corresponds to a IC 50   -and Cl − -containing medium and after 1,150 s were transferred to medium in which all Cl − was replaced by gluconate which stimulates HCO 3 − influx in exchange for Cl − efflux. The initial rate was calculated from the slope of a straight line fitted by eye to the 100 s of the trace after Cl − removal. NHE function shown in b, d, f and h was assessed using the initial rate of increase in intracellular pH in nominally HCO 3 − -free medium during recovery after a 200-s exposure to 20 mM NH 4 Cl. The initial rate was calculated from an exponential curve fitted to the data by least squares minimization (example shown in f). a-f Examples of traces observed. a and b show rates observed in cells not transfected with siRNA, c and d are for cells transfected with "All Stars" siRNA that should have no specific effect on transporter levels, and e and f are for cells transfected with siRNA-1 specific for AE2 or siRNA-3 specific for NHE1, respectively. Ratios of the rates observed for the transfected cells to those for the corresponding paired controls are shown for AE2 in g and for NHE1 in h. Data for the "All Stars" transfected cells are not significantly different from control, while those for siRNA-1,2,3 are significantly less (n =3, p <0.05, 0.05 and 0.01, respectively; t -test against 0 for the logarithm of the ratio) sections was used initially. Such a method has been a key tool in identifying the molecular species responsible for ion transport in specialized epithelia such as kidney and choroid plexus where high-resolution optical images clearly indicate basolateral or apical locations for many of the transporters of interest (see references in Table 1 and [18]). However, these same procedures when used on brain sections (see Fig. 1) were unable to detect any staining on microvessels, the reasons almost certainly being that the lining endothelial cells are very thin, with membranes lacking much folding and that the transporters they express are at lower densities than in specialized epithelia.
To increase the sensitivity of detection, thick frozen sections were used, though this led to some loss of spatial resolution. The positions of the microvessels within the sections were determined by labelling with an antibody to a known endothelial cell marker, PECAM (CD31), together with that to the transporter of interest. This dual-labelling procedure provided clear evidence for AE2 protein along the microvessels (Fig. 2). There was also labelling along microvessels for NBCn1 and NBCe1. For the other transporters investigated (NBCe2, NCBE, NDCBE and NHE1), it was not possible to distinguish staining of microvessels from that of surrounding parenchymal cells. While the methods used here provide evidence for the presence of AE2, NBCn1 and NBCe1 on brain microvessels, they are incapable of localizing these transporters to specific membranes. It is interesting to note that O'Donnell and coworkers [41,53] had to use immunoelectron microscopy in slices from rat brains to locate NHE1 and NHE2 on luminal and abluminal endothelial cell membranes with 65-75 % and 75-80 %, respectively, being found on the luminal side.
Functional activities of the transporters were investigated using brain endothelial cells in culture. Previous studies on cultured rat brain endothelial cells (RBECs) have detected Na + , HCO 3 − cotransport, Cl − /HCO 3 − exchange and Na + /H + exchange [49,68,76]. All of these activities were found also in the GPNT-immortalized rat brain endothelial cells, which were used here for transfection studies (see later discussion). There was substantial expression at mRNA level of NBCn1 and NBCe1, the same Na + , HCO 3 − cotransporters detected at protein level in the brain slices. AE2 was also seen prominently both at mRNA level in the cultured cells and as protein in the slices and could account for the Cl − /HCO 3 − exchange observed in the cultured cells. NHE1 was seen prominently as mRNA in both the RBEC and GPNT cells, while mRNA for NHE2 was present at about 50-fold lower levels than NHE1 in the RBECs. With RBECs [75], as Lam et al. [41] reported for bovine cortical microvascular endothelial cells, both NHE1 and NHE2 were seen in western blots. As shown previously, the Na + /H + exchange observed in RBECs was inhibitable by EIPA [68,76,79]. However, this does not distinguish between the two NHE forms, so for a more precise identification, EIPA was replaced by cariporide in the investigations with GPNT cells (see Fig. 5). IC 50 values for cariporide acting on NHE1 and NHE2 have been reported by others to be 0.03-3.4 μM and 4.3-62 μM, respectively [44]. The IC 50 value found here for GPNT cells of 1.1 μM falls clearly within the lower range, consistent with the mRNA observations of NHE1 but not NHE2 in these cells. The value observed for bovine cortical microvascular endothelial cells, 7 μM [41], may indicate both NHE1 and NHE2 to be important in those particular cells. The use of siRNA to depress levels of mRNA and function in cells provided another means of identifying the molecular species responsible for the activities observed. The effects of two siRNA sequences targeted against AE2 and one against NHE1 were assessed in the GPNT cells. Each of the sequences directed against AE2 significantly depressed both AE2 mRNA expression and the alkalinization resulting from Cl − /HCO 3 − exchange following the removal of external chloride from HCO 3 − -buffered medium. Similarly, the sequence directed against NHE1 significantly depressed both NHE1 mRNA expression and the alkalinization following an imposed acid load seen with nominally HCO 3 − -free medium. These data provide evidence that AE2 and NHE1 not only are present but also account at least in part for the activities in GPNT cells as described earlier. However, it remains possible that in RBEC cells where NHE2 is detectable, NHE2 may also contribute. Experiments were performed to determine whether oxidative challenge might alter not only the expression of AE2 and NHE1 but also the activity of the NHE1 promoter. For this purpose, GPNT cells were used, these being more readily amenable to the necessary transfection. Either direct application of H 2 O 2 or exposure of the cells to 6 h of hypoxia, 1 % O 2 , followed by 16 h of normoxia significantly increased the expression of mRNA for AE2 and NHE1. Furthermore, using a luciferase reporter gene construct, it was also shown that the activity of the NHE1 promoter was increased by hypoxia/ reoxygenation. These observations accord with those of Lam et al. [41] who found increased Na + /H + exchange activity following exposure to hypoxia for 30 min in cortical microvascular endothelial cells. If such changes were to occur in vivo, the greater activities of NHE1 and AE2 would be expected to increase the net transport of Na + and HCO 3 − into the brain. This could have implications for ion transport/fluid secretion at the blood-brain barrier following ischaemia (see final section).
A current model of ion transport at the blood-brain barrier The ion transporters identified in the present study may be inserted into the current model of blood-brain barrier transport as shown in Fig. 8. The primary route for Na + entails entry from the blood via NKCC1 [26,54,72]; however some must also enter in conjunction with HCO 3 − transport. Net exit of Na + into the brain occurs primarily via Na + , K + -ATPases [5,6,25,26,48,65], but possibly some Na + could exit by cotransport with HCO 3 − . HCO 3 − can enter the cell from the blood directly by one or both of the NBCs identified here, i.e. NBCe1 (in its 2 HCO 3 − to 1 Na + mode) or NBCn1. It can also enter indirectly by diffusion of CO 2 , its hydration to carbonic acid catalyzed by carbonic anhydrase, dissociation of that to form H + and HCO 3 − and transport of the H + to the blood via NHE1 and possibly NHE2. Under unchallenged conditions where endothelial cell pH i is >7, Na + /H + exchange and thus CO 2 hydration are relatively unimportant [49], which can explain the lack of inhibition by acetazolamide (an inhibitor of carbonic anhydrase) of blood-brain barrier secretion observed in vivo (see page 739 in [21]). When pH i is reduced, as occurs under ischaemic conditions, Na + /H + exchange then becomes important [49], and as shown here NHE1 mediates at least part of this activity. Net exit of HCO 3 − into the brain may occur via Cl − /HCO 3 − exchange mediated by AE2 and possibly via NBCe1 (in 1 Na + /3 HCO 3 − mode as in renal proximal tubules [64,70,83] and as shown for NBCe2 in choroid plexus [18]). Cl − enters the endothelial cells from the blood via NKCC1 and, if internal Cl − is less than about 95 mM, will also enter the cells via AE2. It may leave into the brain by as yet unidentified Cl − channels or K + , Cl − cotransporters. A volume-activated Cl − channel has been found in these cells [80], which, though unlikely to be the route for Cl − exit in the absence of cell swelling, may be important in Cl − transport during ischaemia. More Cl − enters cells across the luminal membrane via NKCC1 than is transported across to the brain, so at least some must exit on the luminal side. Thus, there must be an as yet unidentified route for Cl − to leave cells on the blood side and again the suggested routes are channels and K + , Cl − cotransporters.
As indicated in Fig. 8, K + can enter via NKCC1 on the luminal side and via the Na + pump on the abluminal side [26], in both cases in much larger amounts than are transported across the endothelium. Thus, there must be routes for K + recycling across both membranes [47], which could be K + channels, some of which are known to be present, 2 or possibly K + , Cl − cotransporters. The presence of passive transport routes for K + in both membranes is consistent with observations that the K + permeability of the blood-brain barrier in vivo is ca. tenfold higher than the permeability to Na + or Cl − [22,69]. 3 The importance of blood-brain barrier ion transport Under normal conditions, ion transport and relatively high water permeability across the blood-brain barrier are important in helping to determine the ionic composition and volume of the ISF. The net flow of fluid produced also aids in the removal of high molecular weight blood-brain barrier impermeant materials from the brain. [17].
However, following ischaemia, uptake and release of ions by cells within the brain are also important in determining the composition and volume of the ISF. Within a few minutes of onset of ischaemia, there is generation of non-volatile acid and consequent decrease in both intracellular and ISF pH, net cellular uptake of Na + and Cl − and release of K + , cell swelling and decreased ISF volume [24,38,43,67,73]. Nevertheless, for the total volume to increase, i.e. for there to be oedema, the ISF must be replenished by a rate of fluid secretion across the blood-brain barrier that exceeds the rate of fluid loss from the brain [4]. The rate of loss will be reduced from its preischaemic value by the decreased size of interstitial spaces [67], and furthermore the rate of secretion is increased [7,14].
While continued or increased transport across the bloodbrain barrier during ischaemia may be beneficial in countering changes in interstitial K + concentration and pH, secretion of Na + and Cl − will contribute to oedema. Thus, in targeting the blood-brain barrier to reduce the effects of ischaemia, it would be desirable to inhibit Na + and Cl − secretion while not impairing or even increasing the transport of K + from the brain and the secretion of HCO 3 − into the brain. Of the transport activities now known to exist in the endothelial cells, 2 There is evidence for several types of K + channel in brain endothelial cells, though some is contradictory. Thus, Hoyer et al. [31] found inwardly rectifying K + channels on the abluminal membrane, a location well suited to recycling K + that entered the cells via Na + , K + -ATPase; Millar et al. [47] found both mRNA and functional fingerprints for Kv1 and Kir2 channels but not Ca 2+ -activated channels; and van Renterghem et al. [78] and Chen et al. [13] found Ca 2+ -activated channels. The differences may be the result of different experimental conditions for the patch-clamp experiments used to identify the channels. Regardless, the available evidence indicates that there are several possible routes for K + efflux from brain endothelial cells. 3 Despite widespread acceptance of the view that net movement of K + across the blood-brain barrier is from brain to blood, see e.g. [52], the direction of this net movement has not been established and may vary. K + concentration is less in ISF than in plasma (see [21] for a review); this has been taken as evidence that K + must be actively transported from brain to blood across the blood-brain barrier. However, if K + is removed from the brain via ISF outflow, then at steady state K + concentration within ISF is determined by the ratio of the net rates at which K + and volume are added to the brain compartment by all other routes to replace materials removed via the outflow. Thus, to achieve a lower ISF K + concentration, there is no need that K + be transported back to the blood by an active process. A more subtle argument can be based on the data of Bito and Davson [8] who found that the K + concentration of CSF in the cortical subarachnoid spaces, 2.65 ±0.1 mequiv/kg H 2 O, is lower than that of CSF in the cisterna magna, 2.98±0.06 mequiv/kg H 2 O (data for dogs, with similar results for cats). They took this observation to imply that K + was being removed from the CSF by diffusion into the cortical parenchyma as the CSF flowed from the cisterna magna to the subarachnoid spaces. This removal could be sustained at steady state if there were net K + transport out of the ISF across the blood-brain barrier. However, again there are alternative explanations. Thus, the CSF K + concentration could also be reduced if fluid with a lower K + concentration were added to the CSF during passage from cisternae to subarachnoid space.
if they were to be distributed as shown in Fig. 8, the most appropriate target for inhibition would be NKCC1. Inhibiting this transporter would reduce the movements of Na + , K + and Cl − towards the brain without directly impairing secretion of HCO 3 − or removal of H + . Indeed marked reduction in oedema and infarct size following middle cerebral artery occlusion in the rat has been seen with bumetanide, an inhibitor of NKCC1 [52,54]. The NHE1 and NHE2 inhibitor, cariporide, has also been shown to reduce oedema and infarct size following middle cerebral artery occlusion [53]. If its action is on secretion across the blood-brain barrier, then it must be inhibiting, directly or indirectly, both Na + and Cl − movements towards the brain. Such inhibition would occur if both the Na + /H + exchanger, NHE1 or NHE2, and Cl − /HCO 3 − exchanger, AE2, were located in the luminal membrane of the endothelial cells as suggested by O'Donnell [52]. Direct NHE inhibition would lead to a decrease in intracellular pH and HCO 3 − concentration, which would in turn reduce Cl − /HCO 3 − exchange, with the net effect of reducing NaCl entry across the luminal membrane. However, while placing AE2 in the luminal membrane is consistent with the data for Cl − /HCO 3 − exchange in cultured brain endothelial cells [49,76], it is not consistent with the proposed role for AE2 in normal secretion across the blood-brain barrier: abluminal to luminal membranes during ischaemia; cariporide may not be having its effects via inhibition of NHE in the endothelium but rather by inhibition of NHE in cells within the brain parenchyma. This last option is also pertinent to interpretation of the in vivo effects of bumetanide as discussed earlier. NKCC1 and the NHEs are known to be present on neurons and astrocytes within the brain parenchyma. If inhibitors of these transporters reach the parenchyma, they will produce marked reductions in the responses of the parenchymal cells to ischaemia with less cellular uptake of Na + and Cl − , less release of K + , reduced cell swelling [12,36,56,71,81,82] and presumably also less contraction of the interstitial spaces. These inhibitor effects within the parenchyma may have consequences quite independently of any direct effects of the inhibitors on the endothelial cells that counter changes in the rates of blood-brain barrier secretion and ISF loss to CSF seen blood endothelial cell interstitial fluid Fig. 8 Schematic diagram of the transporters putatively involved in movement of ions across the blood-brain barrier. Transporters drawn as solid lines are already identified and with known locations, those shown as dashed lines are already identified but with suggested locations and those drawn as dotted lines indicate possible activities in the membranes indicated. Some route/s in addition to those shown solid or dashed must exist in each membrane for each of K + and Cl − . Transporter names are defined in the tables. Transported species are indicated at the source end of each arrow. NHE is shown in both membranes [41,53] without the inhibitors (see e.g. [66,67]). Whether or not the inhibitors reach sites within the parenchyma in sufficient concentrations is at present unknown, 5 and it is prudent to regard the reductions in the extent of oedema and infarct volume as results of effects on both the endothelial cells and parenchymal cells. The relative importance of these sites of action may be different for bumetanide and cariporide.
In summary, results of the current study identify AE2, NBCn1 and NBCe1 on microvessels in the brain. Experiments on endothelial cells in culture demonstrate that these transporters can make a substantial contribution to movement of ions into and out of the endothelium. Further work is now required to determine the localization of these bicarbonate transporters to luminal or abluminal membranes of the endothelial cells as noted by O'Donnell [52] as well as to identify and localize additional transport mechanisms that must exist for K + and Cl − . 5 Bumetanide is a carboxylic weak acid that will be present primarily in anionic form at pH 7.4, while cariporide, pKa 6.28 [44], is a weak base present primarily in neutral form. Both are absorbed well from oral doses, and thus they succeed in crossing the epithelial cells of the gut wall. For bumetanide, in the rat, concentrations have been measured simultaneously in blood and brain following single doses. The earlier of two studies reports bumetanide concentration in brain to be increasing over an hour [10], while the later finds that the concentrations in both serum and brain increase to a peak at half an hour and then fall by more than 75 % over the next half hour [42]. Both find that bumetanide enters brain tissue but to different extents and with different kinetics. In the rat, approximately 98 % of bumetanide in plasma is bound to plasma proteins [39]. Thus, total concentrations in the brain as small as 2 % of those in plasma might reflect either slow penetration across the blood-brain barrier or a lack of binding in brain tissue or both. Low concentrations with rapid removal from the brain [42] would imply little binding and rapid penetration. No data for penetration of cariporide into cells or across the blood-brain barrier appear to have been published [23]. However, it is a small molecule, MW 284, with a neutral form that is known to cross cell membranes in the gut. As a weak base, it might even be a candidate for weak base trapping in ischaemic regions of low pH within the brain parenchyma.