In vivo expression of aquaporin-4 by reactive microglia

Authors


Address correspondence and reprint requests to Josefina Cano, Departamento de Bioquímica, Bromatología, Toxicología y Medicina Legal, Facultad de Farmacia, Universidad de Sevilla, C/ Prof. García González s/n., 41012-Sevilla, Spain. E-mail: josefina@us.es

Abstract

Aquaporin-4 (AQP4) is the most abundant aquaporin in the brain and it is widely accepted that this AQP is solely expressed by astrocytes and ependymal cells. AQP4 is particularly enriched in plasma membranes of ependymal cells and astrocyte membrane domains facing blood vessels and pia. AQP4 has gained much attraction due to its involvement in the physiopathology of brain edema, a major cause of death in humans. Consequently, it is of paramount importance to ascertain the phenotypic nature of AQP4 mRNA-expressing cells in the CNS before attempting future clinical studies aimed at minimizing the development of brain edema. We have used intranigral injections of lipopolysaccharide (LPS), a potent immunostimulant that causes disruption of the blood brain barrier, vasogenic edema, loss of reactive astrocytes and activation of microglial cells. These LPS-induced features are ideal for testing the possibility that reactive microglial cells express AQP4 in the adult brain. We have studied AQP4 at the mRNA and protein level. We provide strong evidence that reactive microglial cells highly express AQP4 mRNA and protein in response to LPS injections.

Abbreviations used
AQP4

aquaporin-4

BBB

blood–brain barrier

CSF

cerebrospinal fluid

GFAP

glial-fibrillary acidic protein

IH

immunohistochemistry

ISH

in situ hybridization

LPS

lipopolysaccharide

MHC

major histocompatibility complex

SN

substantia nigra

The aquaporins (AQPs) are a family of water-selective channel proteins which are present in many kinds of tissues (Agre et al. 2002; Papadopoulos et al. 2002; Verkman 2002; Amiry-Moghaddam and Ottersen 2003). AQPs mediate the efficient movement of water across the membrane. Eleven isoforms of AQPs, AQP0–AQP10, have been identified in mammalian cells to date, from which AQP4 is the most abundant water channel in the CNS (Jung et al. 1994). The highly polarized location of AQP4 on the membrane of astrocytic foot processes opposed to the brain capillaries, pia, and ependymal epithelium, points to its critical function in water transport across the blood–brain barrier (BBB) and brain–cerebrospinal fluid interfaces (Nielsen et al. 1997; Verkman and Mitra 2000; Petrov and Rafols 2001; Venero et al. 2001).

Astrocytes at the BBB, a complex glio-vascular system that controls the CNS homeostasis, act to prevent the non-specific passage of hydrophilic solutes between blood and neuropile. The presence of AQP4 on astrocyte end-feet, together with the absence of AQP1 in the endothelium (for review see Badaut et al. 2002; Papadopoulos et al. 2002), may make water movements from the extracellular space easier with astrocytes than with the blood compartment through the BBB. The pattern of distribution of AQP4 within the perivascular space might be related to the control of the perivascular volume, a function that may be crucial for the maintenance of cerebral blood perfusion (Badaut et al. 2000). AQP4 may also play a role in physiopathologic conditions, as shown by the reduced edema formation observed after water intoxication and focal cerebral ischemia in AQP4-knockout mice (Manley et al. 2000). Several studies have also unveiled its crucial role in response to brain edema formation after osmotic, traumatic and ischemic insults (Manley et al. 2000; Vajda et al. 2000; Ke et al. 2001). Altogether, these experiments have attracted a high interest on this AQP. Therefore, it is of high interest to identify the mechanism of water balance involving the expression and localization of AQP4 in pathological situations in the CNS. Considering the role of AQP4 in the physiopathology of brain edema and that it is believed that this AQP is solely expressed by astrocytes and ependymal cells, it is very important to ascertain the phenotypic nature of AQP4 mRNA-expressing cells in the adult brain in response to injury. The present study is the first to examine the expression of AQP4 in microglia. We have combined AQP4 mRNA in situ hybridization (ISH) and immunohistochemistry (IH) for glial-fibrillary acidic protein (GFAP) and OX-6 as markers of astrocytes and reactive microglial cells, respectively. In addition, we have carried out parallel studies of colocalization by immunohistochemistry of AQP-4 + OX-6 and AQP4 + GFAP. These experiments were conducted by injecting lipopolysaccharide (LPS) into the substantia nigra (SN) to investigate if microglial cells express AQP4. The SN is especially susceptible to the LPS-induced neurotoxicity as compared with other brain areas, an effect that has been ascribed to the activation of microglia (Herrera et al. 2000; Kim et al. 2000). Intraparenchymal injections of LPS in different areas including cortex, hippocampus, striatum and dorsal raphe failed to induce neurodegeneration with minimal glial reaction in contrast to intranigral injections (Herrera et al. 2000; Kim et al. 2000). In keeping with this view, it was originally thought that LPS intracerebral administration induced minimal inflammatory reaction in rat brain (Montero-Menei et al. 1994). The region-specific susceptibility to LPS-induced degeneration is most likely attributable to the fact that SN is the structure with highest microglia concentration in the brain (Lawson et al. 1990). Consequently, the single intranigral injection of LPS is a very suitable model for studying the expression of AQP4 by reactive microglia. Under these conditions, we provide strong evidence that reactive migroglial cells express AQP 4 mRNA and protein. The significance of this data is discussed.

Materials and methods

Animals and lesions

Adult female Wistar rats (200–250 g) born in our laboratory were used in this study. They were housed in a room maintained under constant temperature (22 ± 2°C) and humidity conditions (60%) with a 12/12 h light/dark cycle. Food and water were available ad libitum. Animals were cared for and treated in accord with the Guidelines of the European Union Council (86/609/EU), following the Spanish regulations (BOE 67/8509–12, 1988) for the use of laboratory animals and approved by the Scientific Committee of the University of Sevilla.

Rats were anesthetized with 400 mg/kg choral hydrate and positioned in a stereotaxic apparatus (Kopf Instruments, Tuyunga, CA, USA) to conform with the brain atlas of Paxinos and Watson (1986). The LPS was injected into the left SN. The injection needle was lowered through a drill hole 5.5 mm posterior, 1.5 mm lateral, and 8.3 mm ventral to the bregma. The right side was used as a control. LPS (from Escherichia coli, serotype O26:B6; Sigma, St Louis, MO, USA) was dissolved (1 mg/mL) in a solution of 1% Monastral blue in phosphate-buffered saline (PBS), and 2 µL were injected into the SN. Control injections were of 1% Monastral blue in PBS. All animals were analyzed at 48 h postinjection. Perfused and non-perfused animals were used for either immunohistochemical analysis (n = 6) or in situ hybridization analysis (n = 6). In some cases, immunohistochemical analysis for GFAP and OX-6 were performed on frozen sections from non-perfused animals. For western blot analysis of AQP4, mesencephalon was rapidly removed, frozen in liquid nitrogen and stored at − 80°C until analysis (n = 6).

Immunohistological evaluation of OX-6, GFAP and vimentin

Two days after the intranigral injection of LPS, animals were perfused through the heart under deep anesthesia (chloral hydrate) with 100 mL of PBS containing 10 U/mL heparin followed by 150–200 mL of 4% paraformaldehyde in phosphate buffer, pH 7.4. Brains were removed and then immersed in sucrose in PBS, pH 7.4, first in 10% sucrose for 24 h and then in 30% sucrose until sunk (2–5 days). Tissues were then frozen in isopentane at −15°C and 25-µm sections were cut on a cryostat and mounted in gelatin-coated slices. Primary antibodies used were mouse-derived OX-6 (Serotec, Oxford, UK; 1 : 200), antiglial fibrillary acidic protein (anti-GFAP; Chemicon, Temecula, CA, USA; 1 : 300) and antivimentin (anti-Vimentin; DakoCytomation, Glostrup, Denmark; 1 : 200). All incubations and washes were in Tris-buffered saline (TBS), pH 7.4, unless otherwise noted. All work was done at room temperature. Sections were first washed in PBS and then treated with 0.3% hydrogen peroxide in methanol for 15 min, washed again, and incubated in a solution containing TBS and 1% horse serum for 60 min in a humid chamber. Slices were drained and further incubated with the primary antibody in TBS containing 1% horse serum and 0.25% Triton-X-100 for 24 h. Sections were then incubated for 2 h with biotinylated horse anti-mouse IgG (Vector, 1 : 200) followed by a second 1-h incubation with ImmunoPure® Standard Ultra-Sensitive ABC Staining Kit (Pierce, Rockford, IL, USA; 1 : 100). The antibody was diluted in TBS containing 0.25% Triton-X-100, and its addition was preceded by three 10-min rinses in TBS. The peroxidase was visualized with a standard diaminobenzidine/hydrogen peroxidase chromogen reaction for 5 min.

Localization of AQP4 in astroglial and microglial cells by fluorescence immunohistochemistry

Animals were perfused and sections were prepared as described above. For double-labeling of GFAP or OX-6 with AQP4, sections were rehydrated in PBS for 10 min, and then blocked with PBS containing 1% normal horse serum for 1 h. The blocking solution was replaced with either GFAP (1 : 200), OX-6 (1 : 200) diluted in the blocking solution containing 0.25% Triton-X-100, and the slides were then incubated overnight at 4°C. The slides were washed three times in PBS and subsequently incubated with a horse anti-mouse secondary antibody conjugated to fluorescein (Vector, 1 : 300) for 1 h at room temperature in the dark. The secondary antibody was diluted in PBS containing 0.25% Triton-X-100. Sections were washed three times for 10 min and then blocked with PBS containing 1% normal goat serum for 1 h. The slides were washed three times in PBS and then incubated overnight at 4°C with anti-AQP4 (Chemicon, 1 : 250) diluted in PBS containing 1% normal goat serum and 025% Triton-X-100. Sections were incubated with goat anti-rabbit secondary antibody conjugated to fluorescent Cy3 (Chemicon, 1 : 400) for 1 h at room temperature in the dark and its addition was preceded by three 10-min rinses in PBS. As a control, it was performed another set of experiments where the sections were only incubated with one antibody (GFAP, OX-6 or AQ4), and then visualized with both fluorescence filters. No signal was detected with GFAP or OX-6 alone when Cy3 filter was used (photomicrograph not shown). The same was true with AQ4 when fluorescein filter was used. Immunofluorescence was visualized using an Olympus BX61 microscope. Images were acquired using a digital camera (Olympus DP70) and processed using the associated software package to the camera (Olympus DPController and Olympus DPManager).

Dual localization of microglial cells by lectin staining and vimentin fluorescence immunohistochemistry

First, it was performed the vimentin fluorescence IHC as described above for OX-6 or GFAP and photographed (anti-vimentin; DakoCytomation; 1 : 200). Lectin histochemistry was then performed. Sections were soaked in PBSC (PBS containing 0.3% Triton X-100 and 0.1 mm each CaCl2, MgCl2, and MnCl2) at 4°C for 1 h. Sections were then incubated in 10 µg/mL horseradish peroxidase (HRP)-labeled lectin from Griffonia simplicifolia (GSA I-B4–HRP; Sigma, St. Louis, MO, USA) in PBSC overnight at 4°C. Sections were then briefly rinsed with PBS and lectin binding sites visualized using 3,3′-diaminobenzidine–H2O2 and photographed again.

Sodium dodecyl sulfate–polyacrylamide gel electrophoresis and western blotting of AQP4

Dissected substantia nigra was homogenized and processed for sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and western blotting as described in detail before (Garcia-Rodriguez et al. 2003). Loading conditions were 20 µg protein. Membranes were incubated overnight with anti-AQP4 antibody (1 : 1000, polyclonal antibody from Chemicon International). All experiments were repeated in triplicate.

The products were analyzed by densitometry using AnalySYS® software.

Preparation of riboprobes

pBluescript SK plasmid containing the cDNA sequence for AQP4 (Jung et al. 1994) was purchased from the American Type Culture Collection (Rockville, MD, USA). Antisense riboprobe contains a 394-nucleotide fragment from the 5′ end of the rat AQP4 gene, which corresponds to nucleotides 1201–1595 of the rat AQP4 sequence. It was generated after SspI digestion by using the T7 RNA polymerase. AQP4 sense riboprobe size was 160 nucleotides and was generated after EspI digestion by using the T3 RNA polymerase. Sense and antisense riboprobes used for ISH were transcribed in the presence of [35S]UTP (1300 Ci/mmol; Amersham International) according to a protocol provided by the RNA polymerase supplier (Bethesda Research Laboratories, Bethesda, MD, USA).

Isotopic in situ hybridization histochemistry

ISH on brain sections was carried out following a modification of a procedure described in detail elsewhere (Pasinetti et al. 1989). Thaw-mounted 12-µm sections were postfixed for 30 min in 4% paraformaldehyde, followed by three 10-min washes in phosphate-buffered saline (pH 7.4). Sections were treated for 1 min in 0.1 m triethanolamine, followed by 10 min in acetic anhydride/0.1 m triethanolamine in order to decrease non-specific binding. Following a 1 min wash in two times standard saline citrate (SSC), sections were dehydrated in a series of increasing concentrations of ethanol and then air-dried. The sections were hybridized for 3 h at 50°C with the [35S] cRNA, rinsed in four times SSC/20 mm dithiothreitol, then four times SSC alone. Sections were subjected to 30 min of RNase digestion at 37°C (20 µg/mL RNase A in 0.5 m NaCl, 0.01 m Tris-HCl, 0.001 m EDTA; pH 8.0), washed for 2 h in two times SSC at 25°C followed by 0.1 × SSC at 60°C for 1 h, dehydrated in a series of ethanols, air-dried and processed for emulsion autoradiography. Autoradiograms were generated by apposing the labeled tissue to βmax Hyperfilm for 2 weeks. After looking the dry film autoradiography, slides were dipped in Amersham LM1 emulsion (diluted 1 : 1 with water) and exposed in the dark for 10–15 days at 4°C. Slides were then developed in D-19 (Kodak) at 15°C for 2.5 min fixed for 10 min in fixer (Kodak), and counterstained with cresyl violet (Sigma-Aldrich, SA).

Quantification of isotopic in situ hybridization

Quantification of AQP4 mRNA specific hybridization was performed at the cellular level by measuring grain densities over individual cell bodies using a Image Analysis Program (analySIS®, Soft Imaging System, Germany). The system was calibrated by measuring increasing silver grain densities. The background level usually ranged between two and four grains per cell. These low background levels along with the robust lesion-induced increase in AQP4 mRNA expression ensure the reliability of the quantification.

Combined AQP4 mRNA in situ hybridization and GFAP/OX-6 immunohistochemistry

Thaw-mounted 12-µm sections were postfixed for 30 min in 4% paraformaldehyde, followed by three washes of 10 min each in PBS, pH 7.4. Sections were processed for GFAP or OX-6 IH as described above. After visualization of peroxidase, section were washed in PBS, treated with 0.1 m triethanolamine and processed for ISH as previously described.

Results

Effect of LPS injection on glial population

Figure 1(a,b) shows two representative sections from sham-injected animals in terms of GFAP (Fig.1a) and OX-6 (Fig. 1b) immunoreactivity. The effect of a single intranigral injection of saline had minimal effect on the astroglial population (Fig. 1a) and a slight up-regulation in the OX-6 staining confined to the injection track (Fig. 1b). In confirmation of earlier findings (Herrera et al. 2000; Tomás-Camardiel et al. 2004), the injection of LPS in the SN produced the disappearance of the astrocytes, observed as a lack of GFAP immunoreactivity (Fig. 1c). It is associated with a BBB breakdown (Tomás-Camardiel et al. 2004). Surrounding the area lacking GFAP-immunopositive cells, we found reactive astrocytes with more abundant and larger processes than resting astrocytes oriented towards the depleted area (Fig. 1c). Figure 1(e), which is a high magnification of 1C, shows in contrast the typical morphology of resting astrocytes.

Figure 1.

Effect of intranigral LPS injection on glia. (a,b) A single intranigral injection of saline induced minor changes over the astocytic population in the ventral mesencephalon (a), along with activation of microglia as seen by OX-6 immunoreactivity in an area confined to the needle track (b). In contrast, a single intranigral injection of LPS caused the disappearance of astrocytes, observed as an area lacking GFAP immunoreactivity (c). The area devoid of GFAP immunostaining is completely filled with activated microglia immunostained with OX-6 antibody (d) as seen in a consecutive section to (a). (e) High magnification of the asterisk-marked zone in (c), where the typical morphology of resting astrocytes is patent. (f) High magnification of the asterisk-marked zone in (d) showing the characteristic morphology of the activated microglial cells present in the core. Scale bars: (a,b) 800 µm; (c,d) 400 µm; and (e,f) 100 µm (b,d).

OX-6 IH was performed in consecutive sections of those immunostained against GFAP (Fig. 1d,f). OX-6 is directed against a monomorphic determinant of the rat major histocompatibility complex (MHC) class II antigens, expressed by activated microglia but not for resting cells. Two days after the intranigral injection of LPS, microglial cells were activated in the injured area, just filling the area devoid of GFAP immunostaining (Fig. 1d,f). It could be argued that a loss of GFAP immunoreactivity could be due to a down-regulation of this intermediate filament or alternatively to a loss of immunogenity for GFAP without resulting in astroglial death. Consequently, we performed vimentin IH in adjacent sections to those used for GFAP IH (Fig. 2). We failed to detect an area devoid of vimentin immunoreactivity adjacent to the neurotoxin administration as seen for GFAP (see Fig. 2a,b). However, with no exception, all vimentin-labeled cells present in the GFAP-depleted area exhibited a pseudopodic to globular-shaped morphology, typical morphological features of reactive microglia (Fig. 2d). These unexpected morphological features of vimentin-immunolabeled cells were only evident in the area that precisely correlated with that devoid of GFAP immunoreactivity (Fig. 2c,d). Otherwise, vimentin-immunolabeled cells fulfilled with typical morphological features of astrocytes (Fig. 2e,f). To undoubtedly confirm the phenotypic nature of the vimentin immunopositive cells, we performed a dual localization of microglial cells by lectin histochemistry and vimentin fluorescence immunohistochemistry. By doing that, the microglial phenotype was demonstrated in most vimentin-immunopositive cells in the core of the lesion (Fig. 2g,h).

Figure 2.

Effect of intranigral LPS on vimentin immunoreactivity in comparison with GFAP immunoreactivity. (a,b) Low-magnification bright field photographs of the ipsilateral ventral mesencephalon immunostained with anti-GFAP (a) or anti-vimentin (b) in response to LPS. (c–f) High-magnification bright field photographs of the zones from the boxes depicted in (a) (c,e) and (b) (d,f). Each box and corresponding photograph represents a different lesion profile: boxes (c) and (d), core of the lesion; boxes (e) and (f), unlesioned ventral mesencephalon away from the core of the lesion. It can be observed that within the core of the lesion there is complete absence of GFAP-immunoreactive astrocytes (c) but high density of vimentin-immunolabeled cells showing typical morphological features of reactive microglia; pseudopodic to globular-shape morphology (arrow) (d). In an area away from the core of the lesion, both GFAP (e) and vimentin (f) exclusively label cells showing typical morphological features of astrocytes (arrows). Dual localization of microglial cells by lectin staining and vimentin fluorescence immunohistochemistry demonstrated the microglial phenotype in most of vimentin-immunopositive cells in the core of the lesion. Representative examples of double labeling are marked by arrows in (g) and (h). Scale bar: (a,b), 1 mm; (c–f), 50 µm; (g,h), 35 µm.

AQP4 mRNA expression after an intranigral injection of LPS

In the ventral mesencephalon of control animals, the presence of AQP4 mRNA-expressing cells was scarce and these cells were expressing very low levels and were distributed homogeneously with no regional differences between SN pars compacta and SN pars reticulata (Fig. 3a). Sham-operated animals exhibited a different pattern of AQP4 mRNA expression as compared with control intact animals (Fig. 3b). A single saline injection in SN induced the expression of AQP4 mRNA in a subset of cells distributed throughout the whole ipsilateral ventral mesencephalon (Fig. 3b). We have previously demonstrated that a single intraparenchymal injection of saline is enough to induce the expression of AQP4 mRNA, an effect that was associated to disruption of the BBB (Vizuete et al. 1999). The intranigral injection of 2 µg of LPS was accompanied by an overall and robust induction of AQP4 mRNA in the whole ipsilateral ventral mesencephalon, which was especially evident around the injection track (Fig. 3c; dash-lined area). In a high-power bright magnification of the area where the highest AQP4 mRNA expression is detected, it can be observed that it precisely corresponded with the GFAP-depleted area produced around the LPS-injection site, as described above (Fig. 3d). As we were interested in the cellular expression of AQP4 mRNA in the GFAP-depleted area, we quantified AQP4 mRNA levels in this particular area (Fig. 3e,f) and compared with those seen in the GFAP-positive area (Fig. 3g). Grain count analysis demonstrated that AQP4 mRNA in the GFAP-depleted area had 78% more grains than those found in the GFAP-positive area (Fig. 3h).

Figure 3.

Intranigral LPS highly induces the expression of AQP4 mRNA in a GFAP-depleted area in the ventral mesencephalon. (a–c) Dark-field low-magnification photographs of AQP4 mRNA expression in the ventral mesencephalon from an intact animal (a), saline-injected animal (sham; b) and LPS-injected animal (c). A saline injection is enough to slightly induce the expression of AQP4 mRNA in the ventral mesencephalon (b versus a), which is not, however, comparable to that seen after LPS injection (c versus b). Note that LPS injection produces a large induction of AQP4 mRNA expression in the core of the lesion (dash-lined area). (d) Consecutive section to (c), immunostained with GFAP antibody. The same dash-lined area from (c) is devoid of GFAP staining. (e) High-power bright-field photograph of the zone from (c) where the highest AQP4 mRNA expression is observed. Arrow shows the presence of the inert tracer Monastral blue. Around this injection site, many cells expressed AQP4 mRNA (black grains). (f) and (g) are high-power bright field photographs illustrating cellular expression of AQP4 mRNA in the GFAP-depleted area (f) and in a GFAP-positive area located within the substantia nigra pars reticulata (g) following LPS injection. Cellular levels of AQP4 mRNA were performed in two well defined locations within the substantia nigra response to intranigral LPS; GFAP + and GFAP – areas (h). Cellular levels of AQP4 mRNA were significantly higher in the GFAP – area (h). Results represent the mean ± SEM of six animals and are expressed as number of grains per cell. (i) Analyses of AQP4 protein in SN by western blotting. A single intranigral injection of LPS produced a 9.5 fold increase of AQP4 protein expression as compared with the controls. Results represent the mean ± SEM of six animals and are expressed as percentage of controls. Statistical analysis (Student's test): *p < 0.001. Scale bars: (a–d), 200 µm; (e) 50 µm; (f,g), 10 µm.

Western blotting of AQP4

The expression of AQP4 protein in SN 48 h after the LPS injection was analyzed by western blotting. The intranigral LPS injection highly increased the presence of AQP4 protein (960% of control animals, Fig. 3i).

Phenotypic characterization of AQP4 mRNA expressing cells

We have performed dual AQP4 mRNA ISH and GFAP or OX-6 IH (Fig. 4). The colocalization of GFAP immunoreactivity and AQP4 mRNA-expressing cells has been previously reported (Vizuete et al. 1999). By using this approach, AQP4 mRNA-expressing astrocytes were found 2 days after the intranigral injection of 2 µg of LPS (Fig. 4a). However, these astroglial cells were always away from the core of the lesion (Fig. 4b). In fact, the inert tracer Monastral blue was not present, not even close, to those astrocytes represented in Fig. 4(a). Lying close to the core (in the presence of the inert tracer), cells expressing AQP4 mRNA were found, but its phenotype did not coincide with the astrocytic one, as these cells did not express GFAP immunostaining (Fig. 4b). As the GFAP-depleted area, which surrounds the injection site is filled with OX-6 immunopositive cells (Fig. 1), we decided to combine OX-6 IH and AQP4 mRNA ISH. Strikingly, we found that activated microglial cells were able to express AQP4 mRNA (Fig. 4c). Figure 4(c) shows the presence of OX-6 immunolabeled cells expressing AQP4 mRNA (arrows), confirming for the first time the expression of AQP4 mRNA in activated microglial cells. In a high-magnification photograph (Fig. 4d), these microglial AQP4 mRNA-expressing cells can be observed in detail.

Figure 4.

Co-localization of AQP4 mRNA and GFAP/OX-6 2 days after the intranigral injection of LPS. Sections were processed for GFAP or OX-6 IH followed by AQP4 mRNA ISH, emulsion autoradiography and counterstained with cresyl violet. (a) GFAP-immunopositive astrocytes (arrows) expressed AQP4 mRNA (black grains). Nevertheless, these astrocytes were not close to the injection track, because the inert tracer is not present. (b) Close to the core, observed by the presence of the inert tracer Monastral blue, AQP4 mRNA-expressing cells were found. However, these cells lacked GFAP immunoreactivity. (c) Combined OX-6 IH and AQP4 mRNA ISH show the presence of activated microglial cells (brown) expressing AQP4 mRNA (arrows). These microglial cells were located near the injection track. (d) Highmagnification photograph of (c). Scale bar: 15 µm (a,b,d) and 25 µm (c).

Immunohistochemical characterization of AQP4 protein expression after LPS

To undoubtedly ascertain the microglial expression of AQP4 under the conditions of brain inflammation, double IH experiments were performed. AQP4 protein expression was analyzed together with either the astrocyte marker, GFAP, or the reactive microglial marker, OX-6. Double fluorescence IH for AQP4 and GFAP demonstrated high expression of AQP4 expression in resting and reactive astrocytes and in astrocytic domains facing blood vessels and pia (Fig. 5). An intense immunofluorescence of AQP4 immunoreactivity was seen in the substantia nigra ipsilateral to the LPS injection, which was particularly intense in the area devoid of GFAP staining (Fig. 5). Double fluorescence IH for AQP4 and OX-6 confirmed our observation that reactive microglial cells express AQP4. Figure 6 illustrates the massive up-regulation of AQP4 protein expression within reactive microglia in the same area shown to express high levels of AQP4 mRNA expression and absence of GFAP-immunoreactive astrocytes.

Figure 5.

AQP4 localization in reactive astrocytes in the ipsilateral substantia nigra following LPS administration. Staining of astrocytes was performed using anti-GFAP, labeled green (a,c). Anti-AQP4 was labeled red (b,d). Photographs were taken at low- (a,b) and high-magnification (c,d). Scale bar: (a,b) 50 µm; (c,d) 10 µm.

Figure 6.

AQP4 localization in reactive microglia in the ipsilateral ventral mesencephalon following LPS administration. Staining of reactive microglia was performed using anti-OX-6, labeled green (a,c). Anti-AQP4 was labeled red (b,d). Photographs were taken at low (a,b) and high magnification (c,d) and correspond to the core of the lesion characterized by absence of GFAP immunoreactivity and high density of reactive microglial cells. Scale bar: (a,b) 50 µm; (c,d) 5 µm.

Discussion

We provide, for the first time, direct evidence demonstrating the expression of AQP4 mRNA and protein within reactive microglial cells in vivo. It has been long established that AQP4 is constitutively expressed by astrocytes (reviewed by Venero et al. 2001; Badaut et al. 2002; Amiry-Moghaddam and Ottersen 2003). Immunoelectron microscopy analysis have consistently demonstrated that AQP4 was restricted to glial cells with morphological features typical of astrocytes and to subpopulations of ependymal cells (Nielsen et al. 1997; Rash et al. 1998) and have failed to detect AQP4 within neurons (Nielsen et al. 1997; Nagelhus et al. 1998; Wen et al. 1999). Recent data indicates that endothelial cells are endowed with AQP4 mRNA (Kobayashi et al. 2001) and express low levels of AQP4 protein (Amiry-Moghaddam and Ottersen 2003). However, data dealing with AQP4, the most abundant aquaporin in the brain, in microglial cells is lacking. We have used intranigral injection of LPS, which mimics the BBB disruption in meningoencephalitis (Quagliarello et al. 1986) and provokes severe vasogenic brain edema. We have previously demonstrated that LPS injection produces an area of disappearance of astrocytes probably due to the opening of the BBB along with the massive microglia/macrophage reaction inside the GFAP-depleted area (Castano et al. 1998; Herrera et al. 2000; Castano et al. 2002; Tomás-Camardiel et al. 2004). The loss of astrocytes was confirmed by using two different astrocytic markers; GFAP and vimentin. In confirmation of earlier findings, an area devoid of GFAP-astrocytes was evident; this was densely packed with activated OX-6-positive cells of macrophage morphology (Herrera et al. 2000; Tomás-Camardiel et al. 2004; present data). Vimentin-labeled cells in the area devoid of GFAP staining did not exhibit any morphological features typical of reactive astrocytes including thick processes and hypertrophied somata. With no exception, all vimentin-labelled in the GFAP-depleted area exhibited a pseudopodic to globular-shaped morphology, typical morphological features of reactive microglia (Raivich et al. 1999; Streit et al. 1999). This view was confirmed by performing dual labeling of microglia and vimentin immunoreactivity. Interestingly, reactive microglia has been shown to express vimentin in response to different injury stimulus, and particularly under conditions of inflammation (Graeber et al. 1988; Abd-el-Basset and Fedoroff 1995; Mor-Vaknin et al. 2003). Taken together, we conclude that following intraparenchymal LPS administration there is a loss of reactive astrocytes in a restricted area filled with reactive microglia. These LPS-induced features are therefore very suitable for investigating the presence or not of AQP4 within reactive microglia. We first found a massive up-regulation of AQP4 mRNA within the GFAP-depleted area. Dual GFAP and AQP4 mRNA ISH demonstrated the presence of reactive astrocytes expressing AQP4 mRNA, but always in an area away from the core of the lesion, which was characterized by the presence of a high density of reactive microglial cells. Double fluorescence IH for AQP4 and GFAP were in line with the ISH data. Consequently, to ascertain the phenotypic nature of the AQP4 mRNA and protein-expressing cells in the GFAP-depleted area, we performed dual OX-6 IH and AQP4 mRNA ISH and double fluorescence IH for AQP4 and OX-6. By doing that, we demonstrated that all cells expressing AQP4 mRNA and protein in the GFAP-depleted area were positive to OX-6 immunoreactivity in response to LPS injection. These results demonstrate that AQP4 is expressed by reactive microglial cells. Up-regulation of AQP4 mRNA and protein expression within microglial cells may represent a molecular adaptation to maintain ion water homeostasis in the injured brain.

Recovery from neuronal activation requires rapid clearance of potassium ions (K+) and restoration of osmotic equilibrium. It is well known that glial cells play an important role in regulating the homeostasia to ensure an appropriate neuronal environment. The astrocytes may help clear excess K+ around active neurons (reviewed by Newman et al. 1995; Ransom and Orkand 1996; Amedee et al. 1997). Within this context, AQP4 seems to play an essential role (Nagelhus et al. 1998; Amiry-Moghaddam and Ottersen 2003). The most compelling evidence about this role has been obtained in brains of alpha-syntrophin-null mice, which show a remarkable deficiency of AQP4 protein in skeletal muscle and brain (Frigeri et al. 2001; Vajda et al. 2002; Amiry-Moghaddam et al. 2003). With all these precedents, it has been suggested that water flux through perivascular AQP4 in the astrocyte is needed to sustain efficient removal of K+ after neuronal activation (Amiry-Moghaddam et al. 2003). These findings are consistent with the idea that K+ channels and AQP4 in the astrocytes cooperate in the siphoning of K+. The loss of astrocytes and proliferation of activated microglia could suggest that activated microglia is implicated in the clearance of K+ and restoration of osmotic equilibrium in absence of astrocytes. We have found AQP4 mRNA and protein expression in activated microglia. Inward rectifying K+ channels exists predominantly in the resting microglia, whilst both inward and outward rectifying K+ channels exist in the microglia activated by LPS (Kettenmann et al. 1990; Norenberg et al. 1992; Jou et al. 1998). It has been reported that several K+ channels are expressed in cultured rat microglial activation (Khanna et al. 2001). In addition, LPS-induced activation differentially regulates voltage-dependent K+ channels expression in macrophages (Vicente et al. 2003).

Microglial activation is accompanied by dramatically altered gene expression (Streit et al. 1999) and the functional significance of these events has only just begun to be understood. There is reason to believe that the microglial activation which accompanies acute CNS injuries is a beneficial process reflecting the CNS's attempt to cope with a drastically altered microenvironment, like those generated upon BBB disruption. Reactive microglia is tightly linked with rapid removal of debris and repair processes, processes associated with osmotic gradients, hence the theoretical importance of AQP4. In fact, increased AQP4 expression in reactive microglia may be related to its physiological role in the maintenance of K+ homeostasis. The robust up-regulation of AQP4 by reactive microglia in response to LPS seems to be in line with this issue.

Acknowledgement

This work was supported by a grant from Spanish Ministerio de Ciencia y Tecnología BFI 2001–3600.

Ancillary