Osmotic swelling characteristics of glial cells in the murine hippocampus, cerebellum, and retina in situ

Authors


Address correspondence and reprint requests to Dr Petra G. Hirrlinger, Paul Flechsig Institute of Brain Research, University of Leipzig, Jahnallee 59, 04109 Leipzig, Germany.
E-mail: petra.hirrlinger@medizin.uni-leipzig.de

Abstract

Glial cells are proposed to play a major role in the ionic and osmotic homeostasis in the CNS. Swelling of glial cells contributes to the development of edema in neural tissue under pathological conditions such as trauma and ischemia. In this study, we compared the osmotic swelling characteristics of murine hippocampal astrocytes, cerebellar Bergmann glial cells, and retinal Müller glial cells in acutely isolated tissue slices in response to hypoosmotic stress and pharmacological blockade of Kir channels. Hypoosmotic challenge induced an immediate swelling of somata in the majority of Bergmann glial cells and hippocampal astrocytes investigated, whereas Müller cell bodies displayed a substantial delay in the onset of swelling and hippocampal astroglial processes remained unaffected. Blockade of Kir channels under isoosmotic conditions had no swelling-inducing effect in Müller cell somata but caused a swelling in brain astrocytic somata and processes. Blockade of Kir channels under hypoosmotic conditions induced an immediate and strong swelling in Müller cell somata, but had no cumulative effect to brain astroglial somata. No regulatory volume decrease could be observed in all cell types. The data suggest that Kir channels are differently implicated in cell volume homeostasis of retinal Müller cells and brain astrocytes and that Müller cells and brain astrocytes differ in their osmotic swelling properties.

Abbreviations used
aCSF

artificial CSF

EGFP

enhanced green fluorescent protein

GFAP

glial fibrillary acidic protein

PBS

phosphate-buffered saline

RVD

regulatory volume decrease

Neuronal activity is associated with rapid ion shifts between intra- and extracellular spaces which may cause osmotic imbalances in the neural tissue. A major functional role of glial cells in the CNS including the sensory retina is the control of osmotic and ionic homeostasis which is mediated by transglial ion and water transport (Orkand et al. 1966; Newman and Reichenbach 1996; Kofuji and Newman 2004; Nagelhus et al. 2004; Bringmann et al. 2006). Brain astrocytes and retinal glial (Müller) cells buffer local activity-dependent imbalances in the extracellular potassium concentration predominantly via passive currents through inwardly rectifying potassium (Kir) channels (Newman 1993; Newman and Reichenbach 1996; Kofuji and Newman 2004; Butt and Kalsi 2006). Glial cells express various subtypes of Kir channels (Kofuji et al. 2002; Raap et al. 2002; Ishii et al. 2003; Hibino et al. 2004; Olsen and Sontheimer 2004), particularly Kir4.1 (Kalsi et al. 2004; Butt and Kalsi 2006; Olsen et al. 2006) which has been proposed to be most important for the mediation of potassium buffering currents in retinal Müller cells (Kofuji et al. 2000). The co-localization of Kir4.1 and aquaporin-4 water channels in distinct membrane domains of Müller cells has led to the suggestion that transglial potassium transport is associated with a concomitant water flux through aquaporin water channels (Nagelhus et al. 1999). Aquaporin-4 is the predominant water channel in the CNS and is primarily expressed in brain astrocytes and retinal Müller cells (Simard and Nedergaard 2004).

Osmotic swelling of glial cells is a major cause of edema under various pathological conditions such as ischemia and trauma (Kimelberg 2005). It has been shown that deletion of aquaporin-4 reduces cerebral edema and swelling of pericapillary astrocytes in mice (Manley et al. 2000), suggesting that water transport through aquaporins has a key role in development of edema and glial cell swelling. In addition to an alteration in water transport, changes in transmembranous ion transport have been implicated in glial cell swelling. There are various different ion transport mechanisms proposed to be responsible for astrocytic swelling, such as an up-regulated influx of potassium and chloride ions under extracellular high-potassium conditions, activation of the sodium/proton exchanger leading to intracellular sodium accumulation, and glutamate-induced cellular swelling because of a transporter-mediated uptake of sodium and glutamate (Kimelberg 2005). In vitro in cultured astrocytes astrocytic swelling induced by hypoosmotic medium is typically followed by a regulatory volume decrease (RVD) that is partly mediated by activation of volume-sensitive potassium and anion channels; the release of osmolytes from the cells is associated with a water efflux out of the cells (Pasantes-Morales et al. 1994; Vitarella et al. 1994; Lang et al. 1998). In retinal Müller cells, transmembranous currents through Kir channels have been suggested to play a crucial role in preventing osmotic swelling (Pannicke et al. 2004). The volume homeostasis of Müller cells under hypoosmotic conditions has been especially associated with the expression of Kir4.1 channels, and is absent at the early postnatal stage and under pathological conditions when Müller cells lack prominent Kir channel-mediated membrane currents (Pannicke et al. 2004, 2005, 2006; Wurm et al. 2006a,b). However, it is unclear whether Kir channels are also involved in the regulation of cellular volume in brain astrocytes. To determine a possible role of Kir channels in cell volume regulation of astrocytes, cell size recordings on astrocytes in situ seem to be necessary as cultured astrocytes often lack a prominent expression of Kir channels (Benfenati et al. 2007). The aim of the present study was to compare the swelling characteristics of retinal and brain glial cells under normal and hypoosmotic conditions and to determine a possible role of Kir channels in cellular volume regulation. We recorded the alterations in the soma size of hippocampal protoplasmatic astrocytes, cerebellar Bergmann glial cells (together referred to as brain astrocytes), and retinal Müller cells as well as alterations of cell processes of hippocampal astrocytes in acutely isolated tissue slices. In a second part of the study, we investigated the expression pattern of Kir4.1 and aquaporin-4 in Müller cells and brain astrocytes. We found striking differences in swelling characteristics between astrocytes of different brain regions but also between cells of the same area, which might be related to Kir4.1 expression and function.

Materials and methods

Preparation of tissue slices

All experiments were performed in accordance with the European Communities Council Directive 86/609/EEC, and were approved by the local authorities. Adult (7 weeks to 12 months) C57/BL6 mice were used, as well as adult transgenic heterozygous TgN(hGFAP-EGFP) mice in which astrocytes were labeled by enhanced green fluorescent protein (EGFP) under the control of the human glial fibrillary acidic protein (GFAP) promoter (Nolte et al. 2001). Unfortunately, these FVB/N-derived transgenic mice carry a mutation in a phosphodiesterase gene leading to retinal degeneration (Gimenez and Montoliu 2001) and in addition, the expression level of EGFP in Müller cells was far too low to image these cells properly. Therefore, we used C57/BL6 wildtype mice in which Müller cells were labeled with the vital dye Mitotracker Orange (Uckermann et al. 2004). Animals were killed by cervical dislocation. The brains of TgN(hGFAP-EGFP) mice were dissected and placed in ice-cold, carbogen (95% oxygen/5% carbon dioxide)-saturated artificial CSF (aCSF) (normosmotic control aCSF; in mM: 125 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2, 1.25 NaH2PO4, 25 NaHCO3, and 25 d-glucose, pH 7.4). Frontal or sagittal brain slices (thickness 300 μm) were cut using a vibratome (HM650V, Microm Int., Walldorf, Germany) and stored in aCSF at room temperature (20–23°C) for at least 20 min before experiments were started. The brain tissue slices were transferred to recording chambers and kept submerged in control aCSF by a metal grid with nylon threads for mechanical stabilization. Acutely isolated retinas of C57/BL6 mice were mounted onto a nitrocellulose filter, cut into slices (thickness 1 mm) using a custom made cutter equipped with a razor blade. The filter stripes with the retinal slices were placed into a recording chamber with plastic holders and loaded with Mitotracker Orange (10 μM in aCSF; Molecular Probes, Eugene, OR, USA). It has been shown that this vital dye is taken up selectively by Müller glial cells in the retina whereas neurons, astrocytes, and microglial cells remain unstained (Uckermann et al. 2004).

Experimental conditions

All experiments were performed at room temperature (20–23°C). The recording chambers were mounted on the stage of the microscope, and slices were continuously perfused with control aCSF at a flow rate of 2 mL/min. The hypoosmotic aCSF contained (in mM) 60 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2, 1.25 NaH2PO4, 25 NaHCO3, and 25 d-glucose, pH 7.4. Osmolarities of the solutions were checked before experiments were started (control aCSF: 322 ± 1.8 mosm/L, n = 17; hypoosmotic aCSF: 206 ± 3.4 mosm/L, n = 13). To block Kir channels, barium chloride (1 mM) was added to both solutions; slices were pre-incubated in barium-containing control aCSF for 10 min. For each condition tested, recordings were made in slices of three to five mice.

Image recording

Somata and processes of EGFP-labeled hippocampal protoplasmatic astrocytes and Bergmann glial cells were recorded with an upright confocal two-photon laser scanning microscope (Zeiss LSM 510NLO; Axioskop FS2M, Zeiss, Oberkochen, Germany; equipped with Chameleon laser, Coherent, Dieburg, Germany) and an Achroplan IR 40×/0.8 water immersion objective (Zeiss). EGFP was excited at 890 nm with the lowest excitation intensity possible; emission was recorded with a 500–550 nm band-pass filter and non-descanned detection. Three-dimensional image stacks with xy frame sizes of 256 × 256 pixel (cell somata) or 512 × 512 pixel (covering the whole cell dimensions for analysis of cell processes; pixel size of 0.5–2 μm) at intervals of 1 μm in the z direction were recorded in a tissue depth of ≥ 40 μm. The acquisition time for one image stack was ≤ 90 s for cell somata.

The Mitotracker Orange-stained Müller cell bodies in retinal slices were recorded with an upright confocal laser scanning microscope (LSM 510 Meta; Zeiss) and an Achroplan 63×/0.9 water immersion objective. Mitotracker Orange was excited at 543 nm with a He–Ne laser, and emission was recorded with a 560 nm long-pass filter. During the experiments, the vital dye-stained cell somata in the inner nuclear layer were recorded at the plane of their largest extension. To assure that the maximum soma areas were precisely recorded, the focal plane was continuously adjusted in the course of the experiments.

Image processing

The time-dependent drift of brain slices was corrected using custom written scripts in Matlab (Mathworks, Munich, Germany). The soma and process area of brain astrocytes was estimated by analyzing the area covered by cellular structures in maximum intensity projections of the confocal z stack images. The projections were filtered with anisotropic diffusion (20 iterations) using NIH ImageJ to reduce the background noise. The area of cell somata was calculated by counting the number of pixels over a distinct threshold (ImageJ; http://rsb.info.nih.gov/ij/; area calculator plugIn) keeping the threshold constant for all time points of one experiment. For cell process analysis, an additional background subtraction after filtering was performed and the area containing the cell soma was cleared. The area of cell processes was calculated by counting pixels over threshold (ImageJ particle analysis tool). To determine the volume changes of Müller cell somata, the cross-sectional area of the somata was measured using the image analysis software of the laser scanning microscope.

Data analysis

The values are expressed as mean ± SEM, in percentage of control data measured before normo- or hypotonic challenge (t = 0 min set as 100%). Two groups of cells were defined: cells with a soma area ≤ 100% at t = 32 min were defined as non-swelling cells, whereas all cells displaying an increase in soma area > 100% were defined as swelling cells and included into the statistical analysis. In case of conditions without cell swelling (control for all three cell types as well as 1 mM barium chloride for Müller cells) all analyzed cells were used for statistics. Graphs were prepared using SigmaPlot 10.0 (SPSS Inc., Chicago, IL, USA). Statistical analysis was performed using the Prism program (Graphpad Software, San Diego, CA, USA); significance was determined by anova (Kruskal–Wallis test) followed by Dunn’s comparisons for multiple groups.

Immunohistochemistry

Immunohistochemical stainings were performed with the tissues of three animals per condition; at least three slices per animal were stained. The animals were deeply anesthetized with diethylether and cardioperfused with phosphate-buffered saline (PBS) followed by 4% (w/v) p-formaldehyde in PBS. The brains of transgenic mice, and the eyes of wildtype mice, were removed and post-fixed for 24 h in the same fixative. Isolated retinal tissues were embedded in PBS containing 3% agarose (w/v). The brains and embedded retinas were cut into 40-μm thick slices using a vibratome. The tissues were permeabilized with 0.4% Triton X-100 in PBS for 30 min, blocked in 4% goat serum/0.2% Triton X-100 in PBS for 30 min, and incubated overnight with the first antibodies in 1% goat serum/0.05% Triton X-100 in PBS. After washing with PBS, the secondary antibodies were applied for 2 h in 1.5% goat serum in PBS. The slices were mounted with Shandon Immu-Mount (Thermo Fisher Scientific, Waltham, MA, USA). The following antibodies were used: rabbit anti-Kir4.1 (1 : 100; Sigma-Aldrich, Taufkirchen, Germany), rabbit anti-aquaporin-4 (1 : 100; Sigma-Aldrich), mouse anti-glutamine synthetase (MAB302; 1 : 250; Chemicon, Temecula, CA, USA), Cy2-conjugated goat anti-mouse IgG (1 : 100), and Cy3-conjugated goat anti-rabbit IgG (1 : 1000; Dianova, Hamburg, Germany). Images were obtained using a confocal laser scanning microscope and stored and processed with the Zeiss LSM software and Adobe Photoshop (Adobe Systems, Munich, Germany). Three-dimensional image stacks, covering the entire cell dimensions, were recorded and analyzed for the presence/absence of unequivocal contacts to blood vessels. A given cell was considered to establish such contacts if we found typical astroglial endfeet aligned around or along vascular structures; these endfeet could be clearly identified in the EGFP-labeled cells.

Results

To determine the volume responses of retinal and brain glial cells to hypoosmotic challenge, we recorded the size of the cell bodies of hippocampal astrocytes, Bergmann glial cells, and retinal Müller cells in acutely isolated tissue slices as well as astrocytic processes in the hippocampus. Brain astrocytes were visualized by transgenically expressed EGFP, and Müller cells were stained with the vital dye Mitotracker Orange. In brain astrocytes, the soma and process size was estimated as the area covered by somata and processes, respectively, yielded in maximum intensity projections of confocal z stack images [Figs 1(c,d), 2(b,c), and 3(c,d)]. In Müller cells, the cross-sectional area of somata was measured (Fig. 4c and d). To examine a possible involvement of Kir channels in the glial volume regulation, recordings were made in the absence and presence of the Kir channel blocker barium chloride (1 mM). In all three glial cell types examined we found that most but not all cells analyzed showed swelling of their cell bodies in response to osmotic stress and/or pharmacological blockade of Kir channels (cf. Table 1).

Figure 1.

 Osmotic soma swelling of protoplasmatic astrocytes in frontal hippocampal slices. (a) Time-dependent changes in the soma area. Frontal brain slices were superfused with normosmotic or hypoosmotic aCSF, in the absence and presence of barium chloride (1 mM). The decrease in the control value is due to photobleaching of EGFP. The gray line highlights the 32 min time point analyzed in (b–d). (b) Mean (± SEM) soma areas which were measured after 32 min of perfusion with the respective solutions. The mean values were calculated from cells that displayed soma swelling and from all analyzed cells after control perfusion. Cell numbers are given in the bars. Significant differences versus control: ***p < 0.001. (c and d) Maximum intensity projections of confocal z stack images display examples of EGFP-expressing astrocytic somata before (left) and after (right) perfusion with normosmotic (c) and hypoosmotic aCSF (d), respectively. Scale bars: 5 μm.

Figure 2.

 Osmotic process swelling of protoplasmatic astrocytes in frontal hippocampal slices. Frontal brain slices were superfused with normosmotic or hypoosmotic aCSF, in the absence and presence of barium chloride (1 mM). (a) Mean (± SEM) relative process areas which were measured after 32 min of perfusion with the respective solutions. The cell somata were not included in calculation. Cell numbers are given in the bars. Significant differences versus control: **p < 0.01. (b and c) Maximum intensity projections of confocal z stack images display examples of EGFP-expressing astrocytes before (left) and after (right) perfusion with normosmotic (b) and hypoosmotic barium-containing aCSF (c), respectively. Scale bars: 10 μm.

Figure 3.

 Osmotic swelling of cerebellar Bergmann glial somata in sagittal brain slices. (a) Time-dependent changes in the soma area. Sagittal cerebellar slices were superfused with normosmotic or hypoosmotic aCSF, in the absence and presence of barium chloride (1 mM). The decrease in the control value is due to photobleaching of EGFP. The gray line highlights the 32 min time point analyzed in (b–d). (b) Mean (± SEM) soma areas which were measured after 32 min of perfusion with the respective solutions. The mean values were calculated from cells that displayed soma swelling and from all analyzed cells after control perfusion. Cell numbers are given in the bars. Significant differences versus control: *p < 0.05. (c and d) Examples of maximum intensity projections of EGFP-expressing somata of Bergmann glial cells before (left) and after (right) perfusion with normosmotic (c) and hypoosmotic aCSF (d), respectively. Scale bars: 5 μm.

Figure 4.

 Osmotic swelling of Müller cell somata in retinal slices. (a) Time-dependent changes in the soma area. Retinal slices were stained with the vital dye Mitotracker Orange and superfused with normosmotic or hypoosmotic aCSF, in the absence and presence of barium chloride (1 mM). The gray line highlights the 32 min time point analyzed in (b–d). (b) Mean (± SEM) soma areas which were measured after 32 min of perfusion with the respective solutions. The mean values were calculated from cells that displayed soma swelling after perfusion with hypoosmotic and from all analyzed cells after perfusion with normosmotic solutions. Cell numbers are given in the bars. Significant differences versus control: *p < 0.05; ***p < 0.001. (c and d) Examples of dye-filled Müller cell bodies recorded before (left) and after (right) perfusion with normosmotic (c) and hypoosmotic aCSF (d), respectively. Scale bars: 5 μm.

Table 1.   Volume responses of somata of hippocampal astrocytes, Bergmann glial cells, and retinal Müller cells to hypoosmolarity (∼65% of control osmolarity), blockade of Kir channels by barium chloride (1 mM), and under combination of both treatments
Glial cell typeCondition
HypoosmolarityBariumHypoosmolarity and barium
  1. The numbers of cells that showed soma swelling within 32 min of stimulation are given, as well as the numbers of all cells investigated (in parenthesis). Cells were defined as swelling cells if the soma area was increased to > 100% after 32 min.

Hippocampal astrocytes8 (10)10 (12)7 (8)
Bergmann glial cells8 (16)18 (20)22 (22)
Retinal Müller cells 8 (9)0 (9)8 (8)

Swelling of hippocampal astrocytes

Enhanced green fluorescent protein-labeled protoplasmatic astrocytes were recorded in frontal slices of the hippocampus. Because of the photobleaching of EGFP, we recorded only astrocytes that displayed bright EGFP fluorescence. Perfusion of the slices with normosmotic aCSF (up to 47 min) did not cause swelling of astrocytic somata (Fig. 1a and c) or processes (Fig. 2a and b). A slight apparent decrease in the soma (Fig. 1a) and process size (Fig. 2b) can be attributed to photobleaching of EGFP (after 32 min; somata: 93.8 ± 2.5%; n = 9; Fig. 1b and processes: 96.5 ± 1.8%; n = 8; Fig. 2a). Perfusion of hippocampal slices with hypoosmotic solution (∼65% of control osmolarity) showed cells with different swelling behaviors, it resulted in a strong time-dependent increase in soma size in 8 out of 10 cells investigated (Table 1). In 2 of 10 cells the soma area was ≤ 100%. When analyzing the group of swelling cells, this increase was nearly linear within the first 37 min of hypoosmotic stimulation (Fig. 1a). After 32 min of perfusion, the mean soma area of swelling cells increased to 139.9 ± 8.5% (n = 8; p < 0.001; Fig. 1b). In contrast to cell somata, glial processes and endfeet were not affected after 32 min hypotonic challenge (99.9 ± 5.3%; n = 8; Fig. 2a). Perfusion of the slices with barium-containing normosmotic aCSF caused a non-significant increase in soma size (Fig. 1a and b) of the majority of cells investigated (83%; Table 1) to 114.8 ± 2.2% (n = 10), but a significant increase in the area of processes (116.6 ± 2.2%, n = 8; p < 0.01; Fig. 2a) after 32 min. Addition of barium to the hypoosmotic aCSF caused a strong increase in the soma area of seven out of eight cells to 136.4 ± 5.6% (p < 0.001; Fig. 1b). This value is not different to soma swelling observed under hypoosmotic conditions in the absence of barium. A significant increase was also observed in process area after 32 min superfusion with barium-containing hypoosmotic aCSF (129.2 ± 10.7%, n = 8; p < 0.01; Fig. 2a and c). No RVD was observed within the perfusion period of 47 min under any condition. The data indicate that the majority of hippocampal protoplasmatic astrocytes display soma swelling, but no swelling of their processes under hypoosmotic conditions, and that a blockade of Kir channels has a slight but statistically non-significant swelling-inducing effect on the size of astrocytic somata but a significant effect on the process area.

Swelling of Bergmann glial cells

Virtually all Bergmann glial cells of the transgenic mice used displayed labeling with EGFP. Perfusion of sagittal cerebellar brain slices with control aCSF resulted in a time-dependent slight decrease in the relative soma area of Bergmann glial cells (Fig. 3a and b) which is attributed to the photobleaching of EGFP. The decrease was more pronounced as in hippocampal astrocytes, likely because of the smaller size of the Bergmann glial cell bodies (after 32 min: 89.8 ± 1.5%; n = 21). No swelling of somata was observed up to 47 min of normosmotic perfusion. Perfusion of the slices with hypoosmotic aCSF resulted in an increase in the soma size of 8 out of 16 Bergmann glial cells investigated (Table 1). The soma swelling was nearly linear during the time period investigated (Fig. 3a). After 32 min of perfusion, the mean soma area of swelling cells increased to 132.9 ± 13.6% of control (n = 8; p < 0.05; Fig. 3b). Perfusion of slices with barium-containing control aCSF induced a statistically significant swelling of Bergmann glial somata in 18 out of 20 cells investigated (Table 1) to 115.3 ± 2.8% (p < 0.05; Fig. 2b). A blockade of Kir channels by barium under hypoosmotic conditions caused soma swelling in all cells investigated (n = 22; Table 1). The amplitude of swelling (133.1 ± 7.8%) was in the same range as observed during perfusion with hypoosmotic aCSF without barium (Fig. 3b). As in hippocampal astrocytes, no RVD was observed in Bergmann glial cells. The data indicate that a blockade of Kir channels with barium, as well as hypoosmolarity cause soma swelling in many Bergmann glial cells. An additional blockade of Kir channels under hypoosmotic conditions had no further effect on cell size after swelling when compared with hypoosmolarity without barium, however, it increased the incidence of swelling cells to 100% (Table 1).

Swelling of retinal Müller cells

Perfusion of retinal slices with control aCSF did not alter the size of Müller cell somata (Fig. 4a and c). Similarly, barium-containing aCSF had no effect on the soma size during the time period investigated; after 32 min of perfusion, the mean soma area of the investigated cells was 98.3 ± 1.7% (n = 9) which was not different to control (99.8 ± 2.1%; n = 9; Fig. 4b). In both conditions, all investigated cells were included into the statistical analysis. Perfusion with the hypoosmotic aCSF up to 20 min did not result in a significant increase in the size of Müller cell bodies; thereafter, the somata of eight out of nine cells investigated (Table 1) displayed swelling. After 32 min of perfusion, the soma area increased to 119.1 ± 3.5% of control (n = 8; p < 0.05; Fig. 4b). Perfusion with the barium-containing hypoosmotic aCSF led to an immediate strong swelling of Müller cell bodies (Fig. 4a). After 32 min of perfusion, the mean soma area was increased to 131.9 ± 3.1% of control (n = 8; p < 0.001; Fig. 4b). The data indicate that hypoosmotic conditions cause swelling of Müller cell bodies after a delay of several minutes. A blockade of Kir channels under control conditions does not alter the size of Müller cell bodies, however, it abolishes the lag period during hypoosmotic challenge. The same results were obtained in mice with mixed FVB/N/C57/BL6 background (50%/50%) generated by crossing TgN(hGFAP-EGFP) mice to the C57/BL6 wildtype mice, suggesting that the different swelling behavior is not because of the genetic background (data not shown).

Immunoreactivities for Kir4.1 and aquaporin-4

Both Kir4.1 and aquaporin-4 have been implicated in potassium and water transport through glial cells (Nagelhus et al. 1999, 2004; Kofuji and Newman 2004; Bringmann et al. 2006; Butt and Kalsi 2006) and in glial cell swelling (Manley et al. 2000; Pannicke et al. 2004). Therefore, we compared the immunolocalization of Kir4.1 and aquaporin-4 in glial cells of the retina, hippocampus and cerebellum of adult mice. In hippocampus, only bright EGFP fluorescent protoplasmatic astrocytes (Fig. 5a and d) were investigated as these cells were used for analysis of swelling (Fig. 1). We found that 88 out of 124 hippocampal astrocytes investigated (71%) showed a strong immunostaining for Kir4.1 (Fig. 5b and c; arrows) while 36 cells (29%) displayed a weak labeling of Kir4.1 (Fig. 5b and c; arrowhead). In all cells, Kir4.1 immunostaining was enriched in the cell processes when compared with the soma. A similar staining pattern was found for aquaporin-4. In a total of 75 cells investigated, 45 cells (60%) showed a strong labeling for aquaporin-4 (Fig. 5e and f; arrows), and 30 cells (40%) displayed a weak staining for aquaporin-4 (Fig. 5e and f; arrowhead). It is known that Kir4.1 and aquaporin-4 are enriched in perivascular glial membranes in the retina (Nagelhus et al. 1998, 1999) and spinal cord (Dibaj et al. 2007). Therefore, we investigated whether the Kir4.1 and aquaporin-4 expressing hippocampal astrocytes have contacts to blood vessels in three dimensional image stacks. A clear contact to capillaries (revealed by the identification of perivascular endfeet in three-dimensional image stacks spanning the whole cell dimensions) could be detected in 92% of the astrocytes that displayed a strong labeling of Kir4.1, and in 96% of the cells that were strongly labeled with the anti-aquaporin-4 antibody. In contrast, unequivocal contacts to blood vessels were found in only 50% of the cells that showed a weak labeling for Kir4.1 and aquaporin-4.

Figure 5.

 Localization of Kir4.1 and aquaporin-4 proteins in murine hippocampal astrocytes. Frontal brain slices of transgenic TgN(hGFAP-EGFP) mice were immunolabeled for Kir4.1 (a–c) and aquaporin-4 (d–f). Strong Kir4.1 immunoreactivity (b) was detected in EGFP expressing protoplasmatic astrocytes (a) contacting a capillary (arrows). Weak Kir4.1 immunoreactivity was found in green fluorescent astrocytes without a detectable contact to blood vessels (arrowhead). Strong immunolabeling for aquaporin-4 (e) was found in EGFP expressing astrocytes (d) that had contacts to blood vessels (arrows) whereas astrocytes without a detectable contact to capillaries showed a weak expression of aquaporin-4 (arrowheads). hc, hippocampus.

In cerebellum and retina, both Bergmann glial cells (Fig. 6) and retinal Müller cells (Fig. 7) displayed immunoreactivity for Kir4.1 and aquaporin-4. However, we were unable to quantify the cells that were labeled with the antibodies against Kir4.1 and aquaporin-4 as individual glial cells displayed a dense network of fine side processes and could not be separated. In Bergmann glial cells, aquaporin-4 was localized in processes surrounding capillaries (Fig. 6e and f; arrowheads) and in endfeet abutting the pia mater (Fig. 6e and f). The aquaporin-4 labeling of the granule cell layer is most likely referred to a staining of astroglial processes. In contrast, Kir4.1 staining was more pronounced in glial processes contacting the pia mater (Fig. 6b and c, arrows). In retinal Müller cells, a strong staining of both channel proteins could be detected in the inner plexiform and ganglion cell layers (Fig. 7b, c, e, and f), likely reflecting an expression of the proteins by Müller cell processes and endfeet and by retinal astrocytes. In addition, Kir4.1 immunoreactivity was found in the inner nuclear layer, around vessels (Fig. 7b and c). Aquaporin-4 immunoreactivity was also detected in the outer plexiform layer, and around capillaries (Fig. 7e and f).

Figure 6.

 Localization of Kir4.1 and aquaporin-4 proteins in Bergmann glial cells. Sagittal brain slices of transgenic TgN(hGFAP-EGFP) mice were immunolabeled for Kir4.1 (a–c) and aquaporin-4 (d–f), respectively. Strong Kir4.1 (b) and aquaporin-4 (e) immunoreactivities were detected in virtually all EGFP expressing Bergmann glial cells (a and d). A prominent expression of Kir4.1 was found in endfeet contacting the pia mater (b and c, arrows), and of aquaporin-4 additionally around blood vessels (e and f, arrowheads). cb, cerebellum.

Figure 7.

 Localization of Kir4.1 and aquaporin-4 proteins in retinal Müller cells. Retinal slices of C57/BL6 mice were co-immunolabeled for the Müller cell marker protein glutamine synthetase (a and d) and Kir4.1 (b) or aquaporin-4 (e), respectively. Strong Kir4.1 (b) and aquaporin-4 (e) immunoreactivities were detected in virtually all glutamine synthetase positve Müller glial cells (a and d). Kir4.1 was localized predominantely in perivascular (a–c, arrowheads) membranes and in vitreal endfeet. A strong labeling for aquaporin-4 was also found around the vessels (d–f, arrowheads). GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; ONL, outer nuclear layer; OPL outer plexiform layer.

Discussion

To visualize brain glial cells, we took advantage of transgenic mice with fluorescently labeled astrocytes (Nolte et al. 2001). These mice were already used in many studies where they showed no apparent differences to wildtype mice on cellular and physiological levels, and are well characterized (Grass et al. 2004; Hirrlinger et al. 2004; Szöke et al. 2006). In the hippocampus, only bright fluorescent cells were used for imaging (Matthias et al. 2003; Wallraff et al. 2004; Jabs et al. 2005). For imaging of Müller cells we used C57/BL6 wildtype mice labeled with the vital dye Mitotracker Orange (Uckermann et al. 2004). In mice, Kir4.1 protein is expressed in Müller glial cells and spinal cord astrocytes starting at postnatal day 14 (Kofuji et al. 2002; Dibaj et al. 2007). We therefore used adult mice of ≥ 7 weeks of age to ensure that mature glial cells with full expression of both proteins were investigated.

The present study shows that hypoosmotic conditions induce soma swelling in the majority of cells of all three glial cell populations investigated in situ. However, our study also revealed significant differences in the osmotic swelling characteristics and the involvement of Kir channels in the soma volume homeostasis between brain astrocytes (hippocampal astrocytes, cerebellar Bergmann glial cells) and retinal Müller cells, viz. (i) hypoosmotic challenge caused an immediate and strong swelling of the cell bodies of the majority of brain astrocytes investigated, whereas Müller cell bodies displayed a substantial delay in the onset of swelling; (ii) blockade of Kir channels under normosmotic conditions had no effect on the size of Müller cell somata, but induced soma swelling in the majority of brain astrocytes investigated (with a statistically non-significant swelling in hippocampal astrocytes and a significant swelling in Bergmann glial cells); and (iii) blockade of Kir channels under hypoosmotic conditions resulted in an immediate and strong soma swelling in Müller cells, but did not augment soma swelling of brain astrocytes in comparison to hypoosmolarity alone.

We conclude that retinal Müller cells are capable of efficient cell volume homeostasis under physiological conditions mediated by Kir-independent mechanisms, because a blockade of Kir channels had no effect on soma size under normosmotic conditions. In brain astrocytes, however, Kir channels are apparently involved in cell volume regulation under physiological, normosmotic conditions because the soma size of these cells increased after blockade of Kir channels under normosmotic conditions. However, a blockade of Kir channels under hypoosmolarity did not augment soma swelling in comparison to hypoosmolarity alone. This might suggest a less important role of Kir-channels in cell volume regulation during hypoosmolarity in brain astrocytes. Alternatively, astrocytic Kir channels might still be active under conditions of hypoosmolarity, but their regulatory capacity may be overwhelmed by the strong cell swelling-inducing force of hypoosmolarity. Müller cells, by contrast, need to involve Kir channels only upon exposure to hypoosmotic environment, because a blockade of Kir channels under hypoosmotic conditions results in an immediate and strong soma swelling. The reason for this difference between retinal and brain astroglial cells is unclear. Possibly, the expression of different Kir channel subunits and/or different levels of gap junctional coupling between glial cells may contribute to these differences. It might also be that the storage levels and/or the capacity of transport pathways of alternative osmolytes – such as taurine – differ between Müller cells and brain astrocytes, either absolutely or in relation to the functional demands of their respective surrounding neuronal elements. It should also be kept in mind that the major ‘sink’ of excess water in vivo, the bloodstream in the vasculature, was lost under the conditions of our experiments. This might be less of a problem for the Müller cells in retinal slices whose endfeet still face a large watery compartment than for the astrocytes in the depth of the brain slices. Even the subpial endfeet of cerebellar Bergmann glial cells are mainly exposed to their counterparts in the adherent folium, with only a minimal cleft between the two compartments (cf. Fig. 6b).

Similarly to previous investigations (Pannicke et al. 2004), we found that Müller cell somata showed a substantial delay in the beginning of swelling under hypoosmotic conditions. The effective soma volume regulation of Müller cells in the first 20 min of osmotic stress has been suggested to be mediated predominantly by a release of potassium through weakly rectifying Kir channels associated with water efflux from the cells (Pannicke et al. 2004; Wurm et al. 2006a), as a blockade of Kir channels under hypoosmotic conditions results in an immediate swelling of the cell soma. Apparently the homeostatic mechanisms of Müller cells are exhausted after ∼20 min of osmotic stimulation.

Interestingly, there were also cells in all three glial cell populations that displayed no soma swelling under hypoosmotic stress or in the presence of barium. The reason for this glial heterogeneity in cellular swelling is unclear, but might be because of a different expression of Kir channels. We were unable to find morphological differences between cells with and without soma swelling. Very recently, comparable findings were published for cortical astrocytes defining two groups of cells that showed a high and a low volume response to hypoosmotic challenge (Chvatal et al. 2007a).

Interestingly, unlike in somata, in processes of hippocampal astrocytes hypoosmotic challenge had no swelling-inducing effect. Kir channels may have a specific role in volume regulation of hippocampal astrocytic processes under physiological and pathophysiological conditions: Like in somata, pharmacological blockade of Kir channels led to a swelling of processes under normosmotic conditions. This process swelling further increased when hypoosmotic conditions were applied during blockade of Kir channels. This is in line with recently published data obtained in spinal cord astrocytes in situ (Dibaj et al. 2007) and indicates a regulatory role of Kir channels in process swelling. These differences between the swelling characteristic of soma and processes might be because of a polarized distribution of Kir4.1 and aquaporin4 protein, with an enrichment in endfoot structures embracing capillaries (Fig. 5c and e; Nagelhus et al. 2004). In order to find a possible correlation between soma swelling and channel protein expression, we investigated the localization of Kir4.1 and aquaporin-4 proteins in the three glial cell populations. All three glial cell populations displayed immunoreactivities for both proteins, particularly pronounced in their blood vessel-contacting processes. Previous studies have shown that the Kir4.1 protein is expressed in about one half of brain astrocytes depending on the brain region (Takumi et al. 1995; Poopalasundaram et al. 2000; Higashi et al. 2001). In the spinal cord, the expression levels of Kir4.1 in astrocytes differ depending on the spinal region, which was found to be associated with different astrocytic potassium clearance rates (Olsen et al. 2007). We found that apparently all Müller and Bergmann glial cells showed immunoreactivity for Kir4.1 and aquaporin-4. Bright EGFP fluorescent hippocampal astrocytes represent the major population of protoplasmatic hippocampal astrocytes (Matthias et al. 2003). They are extensively coupled via gap junctions, show a passive membrane current pattern, and express glutamate transporters but not glutamate receptors (GluT cells; Jabs et al. 2005; Matthias et al. 2003; Wallraff et al. 2004). Aside of these protoplasmatic astrocytes, weak fluorescent cells in the hippocampus with a different protein expression and electrophysiological pattern appear to be a new type of glial cells, called GluR cells (Matthias et al. 2003). These weakly fluorescent cells were not included in our study. However, we found that even the bright fluorescent astrocytes are heterogenous with respect to the intensity of immunolabeling for Kir4.1 and aquaporin-4 proteins. Interestingly, strong Kir4.1 and aquaporin-4 immunoreactivity was localized to > 90% of astrocytes that had visible contacts to blood vessels, and to only 50% of astrocytes without unequivocal vessel contacts (analyzed in three-dimensional image stacks spanning the entire cells). It is known that Kir4.1 and aquaporin-4 proteins are prominently colocalized in vitreal/subpial and perivascular endfeet of glial cells, suggesting that both channels act in concert in mediating ion and water homeostasis of the CNS tissue (Nagelhus et al. 2004). However, whether there is a functional interaction between both proteins remains ambiguous, because recent studies using aquaporin-4 knockout mice found striking evidence against this hypothesis (Ruiz-Ederra et al. 2007; Zhang and Verkman 2008). An elevated expression level of both proteins in distinct hippocampal astrocytes may be related to their contact to blood vessels. It remains to be determined whether different expression levels of Kir4.1 and aquaporin-4 are associated with different osmotic swelling properties of the cells, and whether a contact to blood vessels may determine, at least in part, the expression level of the channel proteins in hippocampal astrocytes.

We further observed that retinal Müller cells and brain astrocytes in situ fail to display any RVD in response to hypoosmotic swelling within a time period of 47 min. RVD is a characteristic response of cultured astrocytes in vitro to hypoosmolarity-induced cellular swelling (Kimelberg 2005). It cannot be excluded that the occurrence/failure of RVD is temperature-dependent (our experiments were performed at 20–23°C whereas cell cultures are usually studied at 37°C); however, it appears to be more likely that the presence (in our experiments) or absence (in culture) of the normal cellular environment – such as the extracellular matrix and neighboring neurons – is crucial. Indeed, recently published data investigating the swelling characteristics of cortical and spinal cord astrocytes in situ using the same transgenic TgN(hGFAP-EGFP) mouse line as used in our study (Chvatal et al. 2007a,b; Neprasova et al. 2007) are in accordance with our results: an RVD was found only in poorly swelling cortical astrocytes but neither in strongly swelling cortical nor in spinal cord astrocytes (Chvatal et al. 2007a; Neprasova et al. 2007). Again, the reasons for this heterogeneity of astrocytes are unclear, it may be related to the brain area, and highlights the diversity of astrocytes even in the EGFP labeled population of the transgenic mouse line.

In conclusion, we show that hypoosmolarity is a stimulus for swelling of astroglial cell bodies, but not processes, in situ, and that Kir channels are involved in astrocytic volume homeostasis, either only under control conditions (in hippocampal astrocytes and cerebellar Bergmann glial cells) or, at least, upon exposure to hypoosmotic environments (in retinal Müller cells). Additional osmoregulatory mechanisms, independent of Kir channels, may be active under normosmotic conditions, at least in Müller cells.

Acknowledgements

The authors like to thank J. Krenzlin and C. Schulze for expert technical assistance and F. Kirchhoff for the generous supply of TgN(hGFAP-EGFP) mice. This work was supported by the Interdisciplinary Center of Clinical Research (IZKF) Leipzig at the Faculty of Medicine of the University of Leipzig (projects N05, C35) and Deutsche Forschungsgemeinschaft (Grant numbers GRK 1097/1, RE 849/10-2, and RE 849/12-1).

Ancillary