Albumin elicits calcium signals from astrocytes in brain slices from neonatal rat cortex

Abstract

1. Albumin causes calcium signals and mitosis in cultured astrocytes, but it has not been established whether astrocytes in intact brain also respond to albumin. The effect of albumin on intracellular calcium concentration ([Ca²⁺]_i) in single cells was therefore studied in acutely isolated cortical brain slices from the neonatal rat.

2. Physiological concentrations of albumin from plasma and from serum produced an increase in [Ca²⁺]_i in a subpopulation of cortical cells. Trains of transient elevations in [Ca²⁺]_i (Ca²⁺ spikes) were seen in 41 % of these cells.

3. The cells responding to albumin are identified as astrocytes because the neurone-specific agonist NMDA caused much smaller and slower responses in these cells. On the other hand NMDA-responsive cells, which are probably neurones, exhibited only small and slow responses to albumin. The residual responses of astrocytes to NMDA and neurones to albumin are likely to be due to crosstalk with adjacent neurones and astrocytes, respectively.

4. Methanol extraction of albumin removes a polar lipid and abolishes the ability of albumin to increase intracellular calcium.

5. Astrocyte calcium signalling caused by albumin may have important physiological consequences when the blood-brain barrier breaks down and allows albumin to enter the CNS.

Cultured astrocytes have recently been shown to generate repetitive calcium oscillations and to proliferate when exposed to albumin (Nadal, Fuentes, Pastor & McNaughton, 1995,1997). The existence of this response was something of a surprise, as astrocytes are normally isolated from blood proteins such as albumin by the tight blood-brain barrier. During pathological breakdown of the blood-brain barrier, however, astrocytes come into contact with albumin. In these circumstances albumin is likely to be a signal for astrocyte proliferation, leading to the formation of a glial scar at the injury site. Glial scars have a protective function but also prevent the regeneration of neural tissue, and elucidation of the processes leading to their formation may be of value in preventing unwanted scar formation after damage to the brain.

Calcium signals and mitosis in astrocytes are caused not by albumin itself but by an attached factor, which is firmly bound to albumin in aqueous solution but which can be removed by solvent extraction. On the basis of its solvent extraction profile the factor has been identified as a polar lipid (Nadal et al. 1995). Similar calcium signals are produced by serum albumin in a variety of other cell types (Tigyi, Dyer, Matute & Miledi, 1990; Tigyi & Miledi, 1992; Nadal et al. 1995; Fuentes, Nadal, Jacob & McNaughton, 1997), and the lipid factor responsible in these cases has been identified as lysophosphatidic acid, or LPA (Jalink, Van Corven & Moolenaar, 1990; Tigyi & Miledi, 1992; Moolenaar, 1994). The lipid factor active in astrocytes is unlikely to be LPA, however, because plasma albumin, which contains little bound LPA, and is therefore inactive in producing calcium signals in most cell types, produces vigorous calcium signals in astrocytes (Nadal et al. 1995).

Calcium signals in response to albumin have to date been studied only in cultured astrocytes, and an obvious question is whether these experiments have any relevance to astrocytes in vivo in the brain. In the present study we recorded [Ca²⁺]_i from freshly isolated brain slices, and we examined the effects of serum albumin, plasma albumin and lipid-free (i.e. methanol-extracted) albumin. Cells were identified as neuronal or non-neuronal on the basis of their response to the neuronal agonist NMDA, to which cultured astrocytes do not respond. Plasma and serum albumin were found to generate calcium signals in a subpopulation of non-neuronal cells, presumed to be astrocytes. As in cultured astrocytes, this response was abolished by methanol extraction. We conclude that astrocytes in the intact brain do respond to albumin, and that this response is therefore likely to be important as a signal for breakdown of the blood-brain barrier.

METHODS

Preparation

Brain slices of 300 μm thickness were obtained from the cerebral cortices of 1- to 5-day-old rat pups following standard procedures described elsewhere (Geijo-Barrientos & Pastore, 1995; de la Peña & Geijo-Barrientos, 1996). Rat pups were killed by cervical dislocation followed by decapitation and the brain was rapidly removed. Cortical slices were cut on a vibrating microtome (Vibraslice, Campden Instruments, Loughborough, UK) and were allowed to recover for 30-60 min in a medium containing (mM): 140 NaCl, 5 KCl, 1.2 KH₂PO₄, 1.3 MgSO₄, 26 NaHCO₃, 2.4 CaCl₂ and 10 glucose, continuously gassed with 5 % CO₂-95 % O₂. After the recovery period slices were loaded with calcium indicator by incubation in 10 μM of either fura-2 AM or fluo-3 AM (Molecular Probes), for at least 1 h at room temperature. Lipid-free albumin (2 mg ml⁻¹) was added to the incubation medium to improve dispersal of the calcium indicator. Slices were transferred to the stage of an upright microscope for calcium imaging (see below) and were kept in position with a nylon net.

Problems were experienced with dye penetration into the slice. Cells close to the cut surface, and therefore likely to be damaged, were well loaded with dye but typically failed to respond to either albumin or NMDA. Poorer dye loading was observed in cells 15-30 μm below the cut surface, but at this depth the probability of recording calcium signals was higher. These deeper cells were therefore routinely used, and all results in this study were obtained from cells at this depth.

In common with other groups we were unable to obtain adequate dye loading in tissue slices from adult cortex. All experiments were therefore carried out on slices from neonatal animals. Stability of the recording was also a problem, as even a small vertical movement of the slice caused changes in the fluorescence signal because of movement out of the focal plane. Results presented in this paper are from the small minority of preparations in which a stable baseline was observed both before and after application of test solutions.

Calcium imaging

In initial studies brain slices were imaged using conventional epifluorescence microscopy and ratiometric calcium measurement using fura-2. Single cells within the slice could not be visualized using this technique, because the high fluorescence from cells near the cut surface swamps the signal from undamaged cells deeper in the slice. For later experiments, therefore, slices were loaded with fluo-3 and imaged in a confocal microscope, which because of its ability to image an optical section eliminates background fluorescence from cells out of the focal plane and allows visualization of cells from within the slice. For ratiometric calcium imaging, fura-2 loaded slices were excited at 340 and 380 nm through a water immersion objective mounted on an upright Zeiss Axioskop. The fluorescence emitted at 510 nm was recorded by a Hamamatsu C2400 intensifier-Dage 72 video camera. Images were digitized, stored, ratioed and spatially analysed using a MCID M4 System (Imaging Research Inc., St Catherine's, Ontario, Canada). The ratio of fluorescence at 340 nm to that at 380 nm (F₃₄₀/F₃₈₀) was plotted as the change in ratio (ΔR) expressed as a percentage of the basal fluorescence ratio (R) observed in the absence of stimulus.

For confocal microscopy intracellular calcium in fluo-3 loaded slices was imaged under a BioRad MRC-600 confocal microscope using a Zeiss ×40 Plan Neofluar water immersion lens, numerical aperture 0.9. Images were collected at 2 s intervals and were stored on disk for later analysis. Fluorescence signals from individual cells in a given field of view were measured as a function of time by using the BioRad TCSM software package. Results are plotted as the change in fluorescence intensity (ΔF) expressed as a percentage of the basal fluorescence intensity (F) observed in the absence of stimulus.

In order to quantify the fluorescence signal from fura-2 or fluo-3 in terms of [Ca²⁺]_i it is necessary to determine maximum and minimum values of fluorescence ratio (for fura-2) or fluorescence (for fluo-3) by artificially altering [Ca²⁺]_i to saturating and zero levels. Application of the solutions required to alter [Ca²⁺]_i, however, always caused slight movement of the brain slice, probably because of cell swelling or shrinkage, and consequently caused an unacceptable change in recorded fluorescence because of movement of cells from the focal plane. Records are therefore presented as percentage change from the baseline level, allowing a qualitative measure of changes in [Ca²⁺]_i.

Preparation and source of albumin

Bovine plasma albumin (BPA, Sigma A9306) was used throughout except in Fig. 1A in which human plasma albumin (HPA, Pentex cat. no. 823022, from ICN Biomedicals Ltd, Thame, UK) and human serum albumin (HSA, Sigma cat. no. A 1653) were used. Lipid-free albumin (LFA) was produced by methanol extraction as described by Nadal et al. (1995). Commercial fatty acid-free albumin (Sigma A6003 and A3803) had the same effect as LFA prepared by this method.

Figure 1. Fluorescence signals generated in brain slices by albumin.

A, increases in fluorescence induced by different albumins, bovine plasma albumin (BPA), human plasma albumin (HPA) and human serum albumin (HSA), in slices loaded with the Ca²⁺ sensitive dye fura-2 and imaged using conventional epifluorescence calcium imaging (see Methods). Records are expressed as the increment of fluorescence ratio, ΔR, divided by the prestimulus value of R in order to normalize records (where R = F₃₄₀/F₃₈₀). All albumin concentrations were 20 mg ml⁻¹. The line at the top indicates time of application of albumin. Results shown are typical of observations in 5 brain slices. B, effect of BPA (20 mg ml⁻¹) in a single cell within a brain slice, loaded with fluo-3 and imaged using a confocal microscope. Record expressed as the increment of fluorescence, ΔF, divided by the prestimulus value of F. In 41 % of the cells responding to BPA an oscillatory increase in [Ca²⁺]_i was observed (22 cells out of 78 responding to albumin in 8 different slices). C, effect of BPA, LFA (obtained by methanol extraction of BPA) and a second application of BPA (all 20 mg ml⁻¹). Results obtained using fura-2 and conventional epifluorescence imaging.

RESULTS

Figure 1A shows the effects of three different albumins, bovine plasma albumin (BPA), human plasma albumin (HPA) and human serum albumin (HSA), on intracellular calcium levels measured with fura-2 in a brain slice from the neonatal rat. These records were obtained using conventional epifluorescence microscopy, where individual cells within the slice cannot be clearly distinguished because of the fluorescence signal from cells out of focus above and below the focal plane. The records therefore show the average fluorescence ratio from a number of cells in an area of 200 μm × 200 μm. All three albumins caused a similar maintained increase in [Ca²⁺]_i. Astrocytes in culture generate trains of calcium spikes in response to both serum and plasma albumin (Nadal et al. 1995), while neurones are unresponsive to albumin (Nuñez & Garcia-Sancho, 1996). It is therefore reasonable to suppose that the calcium increase shown in Fig. 1A is generated by astrocytes, with the oscillatory changes which occur in single cells being averaged out over the many cells in the field.

The existence of a calcium signal in response to the plasma albumins (see Fig. 1A) distinguishes cultured astrocytes and the cortical slice preparations described in the present study from other cells tested to date, namely Xenopus oocytes, PC12 cells, endothelial cells and fibroblasts, all of which respond to serum but not plasma albumin (Tigyi et al. 1990; Tigyi & Miledi, 1992; Moolenaar, 1994; Nadal et al. 1995; Fuentes et al. 1997). In common with all other cell types tested to date, however, the calcium signal in brain slices was abolished by prior methanol extraction of the albumin to produce lipid-free albumin (LFA; Fig. 1C), showing that the response depends on a polar lipid attached to albumin.

Confocal microscopy was employed in subsequent experiments in order to obtain records from individual undamaged cells within the brain slice. Figure 1B shows a typical record from such a cell located 15-30 μm below the slice surface. The response pattern of an initial spike followed by trains of oscillations of period 12-30 s is similar to the calcium signals observed in cultured astrocytes (Nadal et al. 1997). The proportion of cells generating maintained trains of oscillations after the initial spike was 41 % (out of 78 cells in 8 slices which responded to BPA with an initial calcium spike), similar to the proportion observed in cultured astrocyte preparations (Nadal et al. 1997).

In order to establish the cell type responsible for these albumin-induced responses we took advantage of the absence of any increase in [Ca²⁺]_i in astrocytes in response to the glutamate receptor agonist N-methyl-D-aspartate, NMDA (Cornell-Bell, Finkbeiner, Cooper & Smith, 1990; McNaughton, Lagnado, Sokolovsky, Hunt & McNaughton, 1990; Jensen & Chiu, 1990; Dani, Chernjavsky & Smith, 1992; Porter & McCarthy, 1996; but see Muller, Grosche, Ohlemeyer & Kettenmann, 1993 and Steinhauser, Jabs & Kettenmann, 1994 who recorded NMDA-activated currents from other glial cell types). Most cortical neurones, on the other hand, exhibit a strong increase in [Ca²⁺]_i in response to NMDA, but neurones do not respond to albumin (Nuñez & Garcia-Sancho, 1996).

In brain slice preparations a maintained increase in [Ca²⁺]_i is observed in some cells in response to NMDA (see Fig. 2, upper panel; NMDA applied in the absence of Mg²⁺ and in the presence of glycine to potentiate its action in elevating [Ca²⁺]_i). Treatment with plasma albumin generated responses in a non-overlapping subset of cells (Fig. 2, lower panel; arrows in both upper and lower panels mark the positions of albumin-responsive cells). Figure 3 shows records of fluorescence as a function of time from four of the cells responding to NMDA or albumin in the experiment of Fig. 2. These results are typical of thirty-five cells in the four slices from which stable recordings were obtained during application of both NMDA and albumin. All NMDA-responsive cells exhibited a maintained increase in [Ca²⁺]_i, as shown in the top record of Fig. 3. Cells responding to BPA exhibited a more variable pattern of responses (lower three records in Fig. 3), with some cells exhibiting a maintained increase in [Ca²⁺]_i, and others generating repetitive spikes. A qualitatively similar distribution of responses to albumin is seen in cultured astrocytes (Nadal et al. 1997). All cells could be clearly separated into NMDA- and albumin-responsive classes, on the basis that the increase in [Ca²⁺]_i was substantially larger in response to NMDA or albumin than to the other. Small responses to NMDA were however often observed in albumin-responsive cells and vice versa (Fig. 3) but these responses are probably due to secondary glutamate release from other cells rather than to a primary response to the agonist (see Discussion).

Figure 2. Effect of NMDA and albumin on individual cells in a brain slice.

A, effect of NMDA (100 μM, applied in 0 Mg²⁺+ 10 μM glycine). From left to right: basal level of fluorescence, 10 s after and 30 s after NMDA application. Arrows indicate position of cells responding to BPA (see B). B, effect of BPA (20 mg ml⁻¹) on the same field of cells. Brain slice loaded with fluo-3 (see Methods) and imaged with a confocal microscope. Images shown in false colour, with red representing high and blue representing low fluorescence intensity. Scale bar at bottom right is 50 μm long.

Figure 3. Calcium signals in response to NMDA and albumin.

Records of fluorescence intensity versus time from the experiment in Fig. 2. Top trace shows an example of a cell responding to NMDA. Three example traces of cells responded to BPA are shown below, the upper one illustrating a maintained increase in [Ca²⁺]_i, and the lower two repetitive spike-like increases in [Ca²⁺]_i.

In thirty-five cells, identified by a clearly visible cell body with a fluorescence intensity above the level of the surrounding area, fourteen (40 %) responded to NMDA, and are therefore probably neurones; eleven (31 %) responded to BPA, with either oscillations or a maintained increase, and are therefore probably astrocytes; and ten (29 %) responded to neither NMDA nor BPA. Cells exhibiting substantial increases in [Ca²⁺]_i in response to both NMDA and BPA were not observed, though as mentioned above small responses to albumin were often seen in NMDA-responsive cells and vice versa (see Discussion). These results suggest that distinct populations of NMDA-responsive and BPA-responsive cells, tentatively identified as neurones and astrocytes, respectively, exist in the cortex.

DISCUSSION

In previous studies we have shown that cultured astrocytes produce continuous trains of calcium spikes in response to plasma albumin, and that plasma albumin is an effective mitogen for astrocytes (Nadal et al. 1995, 1997). Doubts about the physiological relevance of these studies remained, however, because astrocytes in culture are known to diverge from the in vivo phenotype, for instance by expression of ion channels which are not observed in freshly isolated cells (Barres, Chun & Corey, 1990).

In the present study we have re-examined calcium signals in single cells in freshly isolated brain slices from neonatal rats. It has not been possible to identify histologically the cells from which we recorded the calcium signals, because the tissue shrinks during fixation and our attempts to identify cells, for instance by marking for glial fibrillary acidic protein (GFAP), an astrocyte marker, were for this reason unsuccessful. We sought instead to identify neurones from their response to NMDA, a specific glutamate receptor agonist. Cells were clearly divisible into NMDA- and albumin-responsive classes, suggesting that neurones expressing receptors for NMDA do not respond to albumin, while astrocytes or other glial cells respond to albumin but not to NMDA.

Other properties of the calcium signals elicited by albumin in brain slices resembled those of comparable signals recorded in cultured astrocytes. About 40 % of cells produced sustained trains of oscillations in response to albumin, similar to the proportion in culture (Nadal et al. 1997). The period of oscillations, 12-30 s, was also similar to that observed in culture (Nadal et al. 1995). Serum and plasma albumin both elicited similar responses in brain slices, behaviour which resembles that of cultured astrocytes, but is different from all other albumin-responsive cell types identified to date, in which calcium signals are seen in response to serum but not plasma albumin.

Recent work has identified crosstalk between neurones and glia, both in vitro (Purpura, Basarky, Liu, Eftinija & Haydon, 1994; Nedergaard, 1994) and in cortical slices (Pasti, Volterra, Pozzan & Carmignoto, 1997). Neuronal stimulation has been shown to cause increases in [Ca²⁺]_i in astrocytes which are blocked by metabotropic glutamate receptor antagonists, while the neuronal response to astrocytic stimulation is largely blocked by ionotropic glutamate receptor antagonists (Pasti et al. 1997). The small responses to albumin observed in the present study in NMDA-responsive cells, and to NMDA in albumin-responsive cells, are therefore probably due to glutamate released from astrocytes and neurones, respectively.

The results in this study show that responses of astrocytes in intact brain to albumin are similar to those of cultured astrocytes. Cultured astrocytes are therefore, as far as the response to albumin is concerned, a good model for astrocytes in vivo, and it is reasonable to propose that albumin will act as a mitogen as well as a generator of calcium signals in vivo. Albumin released into the brain space, after damage sufficient to cause rupture of the blood-brain barrier, is therefore likely to be an important trigger of mitosis leading to the formation of a glial scar.