Glial Versus Neuronal Uptake of Glutamate

Results are reported from a preliminary autoradiographic study of uptake of glutamate, glutamine and taurine into well-defined cerebellar cell cultures. The uptake of glutamate into granule neurones was minimal, although there was a substantial uptake into glia. The uptake of glutamine and taurine showed no specific distribution. It is possible that granule neurones in vitro are too immature to express sites for glutamate uptake, however sites for neuronal uptake of GABA are present from the first day of culture. Therefore these observations add some further weight to the view that high affinity glutamate uptake is exclusively located in glia. If glial uptake also predominates in vivo then the reductions in glutamate uptake which follow the loss of certain neuronal populations, thought to use glutamate as their transmitter, may be explained in terms of a glial-neuronal interaction. Thus the capacity of glial glutamate uptake would be dependent either on the level of release of glutamate from adjacent nerve terminals or on specific diffusible or contact-dependent factors.

Several amino acids have been implicated as neurotransmitters in the central nervous system on the basis of electrophysiological and neurochemical data (Curtis & Johnston, 1974). Termination of their synaptic activity is generally proposed to be via high-affinity uptake into surrounding glia and the presynaptic neurones, as has been found for monoamine transmitters (Iversen, 1974). Thus, gamma-aminobutyric acid (GABA) is well established as an inhibitory transmitter in spinal cord and brain and a specific high-affinity GABA uptake into glia and GABAergic neurones has been demonstrated (e.g. Kelly & Dick, 1976). Glycine has been found to act as an inhibitory transmitter in the spinal cord (Curtis et al. 1968) but not in cortical areas (Kelly & Kmjevic, 1968) and high-affinity glycine uptake is similarly restricted in distribution (Johnston & Iversen, 1971). Glutamate and aspartate have been proposed as neurotransmitters, due primarily to their excitatory action when iontophoresed onto many types of CNS neurones (Curtis & Johnston, 1974). Support for a transmitter role for these acidic amino acids came initially from the demonstration of a specific sodium-dependent high-affinity uptake system in brain slices and synaptosomes (Balear & Johnston, 1972; Logan & Snyder, 1972).

In the last decade, a great deal of information on the uptake and metabolism of transmitter amino acids has emerged from several lines of investigation. These include biochemical studies of metabolic compartmentation in brain tissue, autoradiographic studies of uptake localization, and studies of single cell types maintained in culture. However, some areas of uncertainty remain, in particular with respect to the role of glutamate. Several studies have now shown that lesions of excitatory pathways in the brain lead to decreases in glutamate content, high-affinity glutamate uptake and stimulus-induced glutamate release. In particular, substantial decreases in glutamate content and high-affinity glutamate uptake have been correlated with the loss of cerebellar granule cells, whether caused by virus infection (Young et al. 1974), X-irradiation (Valcana, Hudson & Timiras, 1972), or neurological mutation (McBride, Aprison & Kusano, 1976). Transection of the lateral olfactory tract caused a reduction in glutamate and aspartate content in guinea-pig and rat pre-pyriform olfactory cortex (Harvey et al. 1975 ; Collins, 1979). Lesions of the entorhinal input to the dentate gyrus result in a decrease of glutamate uptake and of Ca²⁺-dependent, K⁺-induced release (Nadler et al. 1976). Glutamate uptake, and its loss after lesions of excitatory inputs to the hippocampus were shown in autoradiographic studies to be localized in the known target areas of these projections (Storm-Mathisen, 1977).

Aspartate is taken up by a high-affinity transport system which appears to be the same as that for glutamate and therefore cannot be distinguished in uptake studies. In addition, aspartate release is often found to be associated with glutamate release (e.g. Baughman & Gilbert, 1980). However, differential neural sensitivity to the excitatory actions of aspartate or glutamate has been found in some regions of the central nervous system (Watkins, 1978) and at least three different types of receptor are now postulated (Watkins, 1980).

The three putative transmitter amino acids - glutamate, aspartate and GABA - comprise, with the physiologically inactive glutamine, 70 % of the free amino acid in the adult brain. Although there may be considerable variation in the levels of each component, the total concentration of this amino acid system remains constant under a variety of conditions (including neuronal division and degeneration) which suggests the existence of a homeostatic mechanism (Margolis, Heller & Moore, 1968). Glutamate is formed by reductive amination of α-ketoglutarate (via glutamate dehydrogenase) and glutamate and aspartate can interchange with TCA cycle intermediates via a transaminase. Glutamine is produced by further amination of glutamate (glutamine synthetase) and GABA is formed by decarboxylation of glutamate (glutamic acid decarboxylase, GAD).

A compartmentation of glutamate metabolism in the brain was suggested some time ago, in an attempt to explain anomalies in precursor/product relations (fieri & Clarke, 1969; Balázs & Cremer, 1973; Berl, Clarke & Schnieider, 1975). For instance, radioactive glutamate administered intracranially labels brain glutamine to a greater extent than it mixes with the main pool of glutamate resulting in a far higher specific activity in the product than in its immediate precursor. Similarly, several TCA cycle inhibitors, such as fluoroacetate, were found to inhibit the labelling of glutamine from a variety of precursors far more effectively than that of glutamate, an intermediate in glutamine synthesis (Clarke, Nicklas & Berl, 1970). However, the labelling of GABA, which also derives from glutamate, was hardly affected by fluoroacetate. These results have been explained in terms of a separate TCA cycle, with its own (small) pool of glutamate, which is particularly active in glutamine synthesis and makes little contribution to brain glutamate or GABA pools. This ‘small’compartment of glutamate metabolism is preferentially labelled by acetate and glutamate, whereas glucose labels both compartments equally well.

Several lines of evidence have pointed to a neuronal location for the large compartment which synthesizes GABA and glutamate and to a glial site for the small compartment. For instance, a tetrodotoxin-sensitive increase in neuronal respiration can be caused by veratrine alkaloids, which act on neuronal ion channels. Selective inhibitors of glutamate metabolism in the small compartment, such as fluoroacetate, do not prevent the veratrine-stimulated increase in neuronal respiration (Benjamin & Quastel, 1972). Malonate, on the other hand, prevents the veratrine stimulated increase in respiration and the stimulated release of glutamate, aspartate and GABA, but not the release of glutamine.

The high-affinity uptake of extracellular glutamate seems from biochemical studies to be linked with the small (glial) compartment of glutamate metabolism. Thus, when glutamate was applied to brain slices or intracranially, it labelled mainly glutamine. Exogenous glutamate has been found to exchange poorly with endogenous glutamate, i.e. with the large (neuronal) compartment of glutamate metabolism, and to be a poor precursor for GABA (Benjamin & Quastel, 1972). In contrast, glutamine is generally a better precursor for neuronal glutamate metabolism and GABA synthesis than glutamate itself (e.g. Berl, Lajtha & Waelsch, 1961). Indeed, glutamine may be superior to glucose as a precursor of neuronal stimulus-evoked glutamate release (Hamberger et al. 1979).

These biochemical observations were drawn together in a proposal linking neurones and glia which would account not only for the regulation of the levels of the physioiogically active amino acids, but also for the removal of tissue ammonia (Benjamin & Quastel, 1975). This hypothesis stemmed from the observation under a variety of experimental conditions of a striking constancy in total nitrogen content of the suni of tissue ammonia, glutamate, glutamine, GABA and aspartate. Inactivation of synaptically-released glutamate, and maintenance of the low (< 10 μM) extracellular glutamate concentration is proposed to be via uptake into surrounding glial cells. Addition of ammonia to glutamate by the glial enzyme glutamine synthetase would yield glutamine which can be released to the extracellular space. The high extracellular level of glutamine (∼ 500 μM) is available for neuronal re-uptake and metabolism to glutamate, GABA and the other components of the large compartment. Similarly, GABA released by neurones and taken up by glia could be returned to the large glutamate pool via glutamine (Van den Berg & Garfinkel, 1971). This simple arrangement is consistent with the biochemical data on compartmentation and accounts for most of the apparently anomalous aspects of glutamate metabolism. The cellular locations of the biochemical compartments have been inferred, but cannot be established directly by precursor/product experiments on brain slices. For confirmation of the glutamate-glutamine cycle hypothesis, it is essential to examine the uptake kinetics of glutamate, GABA and glutamine into neurones and glial cells and the activities of the relevant enzymes in each cell type.

Several in vitro preparations enriched in glial or neuronal cells or cell fragments have been used to investigate, and quantify the cellular locations of aspects of glutamate metabolism.

Studies on glial material began with the demonstration of high-affinity uptake of glutamate and GABA into a bulk preparation of glial cells from adult rat brain (Henn & Hamberger, 1971). Human gliomas show high-affinity uptake of glutamate and GABA in slices (Snodgrass & Iversen, 1974) and when cultured (Stewart et al. 1976). Several studies have demonstrated sodium-dependent transport of glutamate and GABA at high-affinity into tumour cell lines of glial origin (Hutchison et al. 1974; Schrier & Thompson, 1974; Balear, Borg & Mandel, 1977). Primary cultures of normal astrocytes also exhibit high-affinity uptake of GABA (Schousboe, Hertz & Svenneby, 1977) and glutamate (Hertz et al. 1978). This preparation has virtually the same homogeneity as clonal cell lines and in addition seems more faithfully to reflect in vivo kinetic properties, particularly when quantifying the capacity (V_max) of the uptake system (Schousboe, 1977). From these uptake data and the astrocytic activities of glutamine synthetase, glutamate dehydrogenase, glutamate oxaloacetate transferase and other enzymes involved in metabolism of glutamate, Hertz (1977) calculated that the glial capacity for uptake and metabolism of glutamate was sufficient to account for the known flux between neuronal and glial metabolism. In contrast, glial uptake of GABA and its metabolism via GABA trans-aminase, although significant, could only account for a small part of the total GABA degradation in vivo, implying a requirement for neuronal GABA uptake and breakdown. The neuronal-glial cycle of glutamate and glutamine proposed by Benjamin & Quastel (1975) is supported by these results. However, the fact that significant levels of low-affinity glutamine uptake and of glutaminase have been found in cultured astrocytes (Schousboe et al. 1979a) may indicate that some portion of glutamine breakdown can occur in glia.

Most quantitative studies of glutamate and GABA uptake into neuronal elements have involved the use of synaptosomal preparations. High-affinity uptake of GABA, glutamate and glycine has been found in synaptosomes from spinal cord and brain (Iversen & Johnston, 1971 ;Logan & Snyder, 1972; Bennett, Logan & Snyder, 1973). Separation by density-gradient centrifugation of synaptosome populations accumulating different amino acids has been claimed (Arregui et al. 1972), although there is a contrary report (Balear & Johnston, 1973). Contamination of synaptosomal preparations by a proportion of glial particles must be borne in mind when assessing these results (Cotman, Herschman & Taylor, 1971;Schon & Kelly, 1975 ; Henn, Anderson & Rustad, 1976). However, localization of glutamate uptake in a small proportion of nerve terminals in a synaptosome preparation has been observed (Beart, 1976). A five times greater activity for glutamate uptake in synaptosomes compared with bulk-isolated glia has been reported (Weiler, Nystrom & Hamberger, 1979) although, as the authors point out, membrane damage in the glial fraction may cause losses in apparent uptake capacity.

In order to obtain comparable data for neurones to those found for glia grown in primary culture, it is necessary to study a preparation of neurones grown in the absence of glia. Neuronal primary culture systems have been reported by several groups (e.g. Dichter, 1978 ; Pettman, Louis & Sensenbrenner, 1979; Currie & Dutton, 1980) which contain a small and quantifiable proportion of glia. However, none have been used for such a study, although the work of Lasher (1975) on GABA uptake demonstrated the usefulness of this approach.

Two groups have followed up the possible glutamatergic nature of granule cells by investigating the uptake of labelled amino acid into separated cerebellar perikarya. Campbell & Shank (1978) found that a population separated from 10-day-old mice containing predominantly granule cells showed high-affinity uptake for [¹⁴C]glutamate with a V_max of about one-third that of a glial-enriched (although heterogeneous) population. The other kinetic parameters were identical and included an unusual two-component high-affinity uptake system. A parallel tissue culture study showed the granule fraction, although remarkably homogeneous, to include 5 % astrocytes and other cells, while the glial-enriched fraction was much more heterogeneous and contained an unstated proportion of astrocytes (Campbell & Williams, 1978). It is difficult, on the basis of these workers ‘results, to exclude the possibility that the glutamate uptake shown by the granule cell population may be a high level of uptake into a small minority of glial cells, or subcellular glial fragments. East, Dutton & Currie (1980) showed the distribution of glutamate uptake into separated fractions of cerebellar perikarya isolated from 10-to 11-day-old rats to coincide with that of β-Lalanine (see below), the highest levels occurring in the glial-enriched fractions. There was no uptake of glutamate or β-alanine by fractions of granule neurones. Although it is possible that the mildly trypsinized perikarya used by Campbell & Shank (1978) and East et al. (1980) had lost the sites responsible for neuronal glutamate uptake, the latter found GABA uptake to be retained by fractions enriched in Purkinje cells. Granule cells isolated from animals of this age could also be too immature to express the ability to take up glutamate.

Information on the cellular location of amino acid uptake has also been provided by autoradiographic studies. Uptake of labelled GABA in brain tissue has been investigated by autoradiography after incubation of slices (Iversen & Bloom, 1972), ventricular injection (Hokfelt & Ljungdahl, 1970, 1972; Schon & Iversen, 1972) and direct microinjection (Kelly & Dick, 1976). The contribution of neuronal and glial components of GABA uptake were clarified by the finding that their different substrate specificities could be distinguished by the use of GABA analogues (Kelly & Dick, 1976). Thus, glial uptake of [³H] GABA was inhibited by β-alanine, while labelled β-alanine was selectively accumulated by glia (Schon & Kelly, 1975 ; Kelly & Dick, 1976; Weitsch-Dick, Jessell & Kelly, 1978). Similarly, diaminobutyric acid (DABA) showed specificity for the high-affinity GABA transport sites of known and putative GABAergic neurones. Aminocyclohexanecarboxylic acid (ACHC) has also been found to be an effective competitive inhibitor of neuronal GABA uptake (Bowery, Jones & Neal, 1976) and a wide range of further analogues of varying selectivity have been reported (e.g. Schousboe et al. 1979b).

In cell culture, GABA is taken up by astrocytes and some types of neurone which can plausibly be identified as cells known to be GABAergic in vivo. Thus, uptake into cerebellar interneurones (Lasher, 1974; Messer, 1977), cortical inhibitory neurones (White, Snodgrass & Dichter, 1980) and olfactory bulb granule neurones (Currie & Dutton, 1980) has been shown in dissociated cell cultures. Neuronal and glial uptake have been distinguished using competitive analogues (ACHC and β-alanine) in cultures of cerebellum and olfactory bulb (Currie & Dutton, 1980) (see Fig. 2).

The presence of neuronal GABA uptake in GABAergic cells accords well with the calculation of Hertz (1977) that glial GABA uptake, although substantial, is not sufficient to account for more than a fraction of the neuronal release. There is no evidence that non-GABAergic cells which receive GABA innervation possess GABA uptake sites; the granule cells of the cerebellum, for instance, receive GABAergic input from Golgi cells, but show no GABA uptake (Kelly & Dick, 1976; Currie & Dutton, 1980). The evidence on GABA uptake from studies vivo and in vitro using kinetic and autoradiographic approaches is thus broadly in agreement. In the case of the major putative excitatory amino acid transmitter, glutamate, the situation is more problematic.

It might be thought that the most direct way to determine whether high-affinity glutamate uptake is located in neurones or glia would be via autoradiography of uptake into intact tissue at the light, and particularly the electron, microscopic level. However, results have not been unequivocal.

Exclusively glial uptake has been found in several studies, including uptake into Müller glial cells in retina (Ehinger & Falck, 1971), Bergmann glial cells in cerebellum (Hokfelt & Ljungdahl, 1972) and satellite glial cells in sensory ganglia (Schon & Kelly, 1974). An EM autoradiographic study of glutamate uptake into brain after microinjection or topical application found silver grains to be located over glial cells (McLennan, 1976). Unfortunately, there have been no glutamate analogues described which distinguish between neuronal and glial uptake, a facility which clarified the distribution of GABA uptake.

The most extensive autoradiographic investigation of glutamate uptake is that of Storm-Mathisen and his co-workers (Storm-Mathisen, 1977; Storm-Mathisen & Iversen, 1979) who have adopted a 10 min incubation of thin (200 μm) slices of rat hippocampus with labelled glutamate. Under these conditions, the majority of label in some regions was associated with axons and boutons in a ‘hypothetical grain analysis’ of EM autoradiographs. There was minimal labelling of cell bodies or dendrites. The label was concentrated over areas in which glutamatergic projections terminate, e.g. the inner zone of the area dentata molecular layer which is a target of pyramidal axons from zones CA3 and CA4 (Storm-Mathisen & Iversen, 1979). Particularly important was the large decrease in labelling found some days after transection of the contralateral and ipsilateral projections of these pyramidal cells (Storm-Mathisen, 1977). Similar observations of loss of glutamate uptake on denervation have been made in several systems, including lesions of corticostriatal paths (McGeer et al. 1977; Divac, Fonnum & Storm-Mathisen, 1977), corticofugal fibres (Lund, Karlsen & Fonnum, 1978) and hippocampal afferents to septum and hypothalamus (Storm-Mathisen & Woxen Opsahl, 1978). This association of substantial loss of glutamate uptake with removal of excitatory synaptic input represents the most compelling body of evidence in favour of a predominantly neuronal location of glutamate uptake.

Similar evidence for the cerebellum has been provided by the observation of a large decrease in glutamate uptake and glutamate content associated with viral-induced loss of granule neurones (Young et al. 1974). However, interpretation of these cerebellar results is complicated by the finding that loss of glutamate content in agranular mutant mice extended to the deep cerebellar nuclei which contain no granule neurones (Roffler-Tarlov & Sidman, 1978). It has recently been reported that cerebellar uptake of labelled D-aspartate, a non-metabolisable substrate for the glutamate uptake system, was confined to glia in a light and EM autoradiographic study (Gordon et al. 1980).

The pattern of GABA uptake in vitro was found to reflect faithfully in vivo behaviour. It may be possible, therefore, to resolve the site of high-affinity glutamate uptake in dissociated cell cultures where cells can be readily distinguished, and where the cellular composition of the culture can be manipulated. Accordingly, a study has been undertaken of glutamate uptake into cerebellar cell cultures which are enriched in granule neurones. These small excitatory interneurones are generally accepted as glutamatergic on the basis of the considerable (circumstantial) evidence reviewed earlier.

The system of defined cell cultures of cerebellum used here has been described previously (Currie, 1980; Currie & Dutton, 1980). Taking advantage of the limited and well-established range of cell populations in the cerebellum, it has been possible to define the composition of such cultures in terms of known cell types (Currie, 1980) using reliable criteria such as GABA uptake autoradiography (Currie & Dutton, 1980) And immunological markers (Raff et al. 1979). Cells are prepared from one-week old rat cerebellum by a controlled trypsinisation procedure and seeded at 2 × 10⁶/cm² of 13 mm glass cover slips coated with poly-L-lysine. Addition of fluorodeoxyuridine after, in this case, 48 h allowed a substantial population of astrocytes to grow up, but prevented overgrowth which adversely affects neuronal survival. Cultures such as those shown here contain 75–80% granule neurones with 15&3–20% astrocytes and 5–8% GABAergic neurones, probably stellate and basket cells (Currie & Dutton, 1980); fibroblasts, oligodendrocytes and macrophages are present at less than 1 %.

Granule neurones in cell culture are small, round or ovoid, phase-bright, bipolar cells which have a tendency to aggregate after some days in culture. As shown in Fig. 1, they bind tetanus toxin, a neuronal marker (Dimpfel, Huang & Habermann, 1977; Mirsky et al. 1978), but do not show neurone-specific GABA uptake (Fig. 2). The minor population(s) of tetanus-positive neurones which show GABA uptake, inhibitable by ACHC but not by β-alanine (Fig. 2), are probably stellate and basket neurones (Currie & Dutton, 1980). These neurones are usually larger than granule cells, and show no tendency to aggregate with each other or with granule neurones. Specific GABA uptake sites are present in these cells from the earliest times in culture : Fig. 2(c) shows that GABA uptake can be detected after only 3 h in culture.

When a similar procedure was used to study [³H]glutamate uptake (Figs. 3 a, b) high levels of labelling were found over the underlying background cells, but not over the granule neurones. In a parallel culture (Figs. 3 c, d) the flattened cells could be labelled by an antibody to glial fibrillary acidic protein (anti-GFAP) and therefore may be identified as astrocytes (Antanitus, Choi & Lapham, 1975). They also show glial-specific GABA uptake (Fig. 2,b) (Currie & Dutton, 1980). The absence of [³H] glutamate uptake into granule neurones is shown again in Figs. 4(a), (b), where unlabelled neuronal processes can clearly be seen running over labelled astrocytes. This pattern of uptake is similar to that shown by β-alanine and GABA inhibited by ACHC (Currie & Dutton, 1980); both are taken up at high-affinity by the glial GABA transport system, but not by the neurones. Neuronal uptake of glutamine is predicted by the glutamate/glutamine cycle hypothesis; when cultures were incubated with [³H] glutamine and processed for autoradiography in parallel (Fig. 4 c), grains were found over both granule neurones and glia. Studies of labelled glutamate uptake by autoradiography must be tempered by the knowledge that glutamate is rapidly metabolized into other components, particularly glutamine, which may pass between cells during the incubation period. Such an effect might account for misleading distributions of label after incubations of intact tissue. However cells in monolayer culture are accessible to a large volume of medium which will by dilution tend to prevent significant intercellular transfer of label.

From these results, it must be concluded that the granule cell does not show high-affinity glutamate uptake as it develops in vitro. It is possible that such uptake sites develop at a later stage of maturation, and that these studies must be extended to longer times in vitro. However, neuronal GABA uptake is present from earliest times in culture. Dissociated cerebral cortex neurones have been reported to show a similar lack of high-affinity glutamate uptake after three weeks in culture (White et al. 1980).

There is persuasive, although not yet conclusive, evidence that glutamate is an excitatory transmitter in several brain pathways, and that it is inactivated by high-affinity uptake into adjoining cells. Knowledge of the cellular location of glutamate uptake sites is thus central to an understanding of a major excitatory transmitter system. The existence of neuronal uptake which is restricted to glutamatergic neurones would have the additional methodological advantage that localisation of glutamate uptake could be used to identify such neurones, as has already been possible for GABA, glycine and the monoamine systems.

However, several lines of evidence suggest that glutamate uptake is restricted to glia. Biochemical studies have established that extracellularly applied glutamate is almost exclusively taken up into the small compartment of glutamate metabolism and converted to glutamine via the enzyme glutamine synthetase. A recent histochemical study indicates that this enzyme is to be found only in glia (Norenberg & Martinez-Hernandez, 1979); it has since been used as a marker for astrocytes (White, Dutton & Norenberg, 1981). Externally applied glutamate does not mix with the large intra cellularly synthesized pool of glutamate from which GABA is produced. Extracellular glutamate is a very poor precursor for glutamate released by Ca²⁺-dependent stimulation; better precursors are glucose and, particularly, glutamine (Hamberger et al. 1979)

The biochemically-defined compartments of glutamate metabolism involve, of course, considerable simplification. There can, for instance, be no unitary pool: GABA synthesis is associated with the large (neuronal) pool, but only a fraction of the neurones release GABA, and are rich in the enzyme glutamate decarboxylase. However, putative glutamate-releasing neurones from a substantial proportion of the total neurones and any proposal regarding glutamate uptake into these cells must be reconcilable with the established biochemical data. There are few studies of the possible heterogeneity of glia between different brain areas.

The results shown here support the conclusion that high-affinity glutamate uptake is solely a glial property. The lack of glutamate uptake shown by the supposedly glutamatergic granule neurones is in complete contrast with the neuronal-specific uptake of GABA into GABAergic neurones from the first hours in culture. The major site of glutamate uptake in cerebellar cultures is the population of astrocytes which can be labelled with an antibody to glial fibrillary acidic protein. Glutamate uptake into freshly isolated cerebellar perikarya was also concentrated in glial fractions, and no transport into granule neurones occurred (East et al. 1980). Autoradiographic studies in vivo have shown glutamate (or D-aspartate) uptake to be located in cerebellar glia, particularly the characteristic Bergmann glial cells, but not in neurones (Hokfelt & Ljungdahl, 1972; Gordon et al. 1980). The restriction of glutamate uptake to glia is thus supported by several lines of evidence; however certain studies have indicated a contrary conclusion.

An autoradiographic investigation of exogenous glutamate uptake into hippocampal slices has indicated a primarily neuronal uptake site (Storm-Mathisen & Iversen, 1979), the largest proportion of labelling occurring in nerve terminals, with no labelling of the parent cell bodies as commonly found for neuronal GABA uptake. would be useful to extend the culture studies to hippocampus, and to longer times in vitro, as it is possible that one-week-old cultures of cerebellar granule neurones have not developed sufficient synapses to show detectable uptake. A nerve terminal uptake site would be in accordance with the basic compartmentation evidence and glutamate/glutamine cycle hypothesis if terminal-accumulated glutamate were exclusively re-released without mixing with the larger main neuronal glutamate pool, or entering metabolism via the transaminase. However, this scheme cannot easily be reconciled with the observation that external glutamate is a very poor precursor for glutamate released from hippocampal and other tissues by Ca^2,+-dependent stimulation (Hamberger et al. 1979).

The main body of evidence against an exclusive uptake of glutamate by glia comes from denervation studies in several brain areas. The great reductions in glutamate uptake capacity found after transection of certain projections (Storm-Mathisen, 1977; McGeer et al. 1977; Lund Karlsen & Fonnum, 1978) or selective loss of a neuronal population (Young et al. 1974) are probably accompanied by a proliferation of glial cells.

One hypothesis which might account for these denervation effects is that the presence of glutamate-releasing nerve terminals stimulates the uptake of glutamate into glia. An uptake system of high capacity is necessary to remove all extracellular glutamate released by neurones (Hertz, 1977). It is plausible that a regulatory mechanism ensures that the uptake sites for glutamate are concentrated in the glia in close proximity to glutamatergic terminals. The presence of high-affinity glycine uptake in the spinal cord and not in the brain could be an example of such control (Johnston & Iversen, 1971). The capacity of glial glutamate uptake (i.e. the number of sites) could be controlled directly by the level of extracellular glutamate, by specific factors released from glutamatergic neurones, or by a contact-dependent interaction between nerve terminals and adjacent glia. The interesting observation that astrocytes from different brain areas differ in their glutamate uptake capacity (Schousboe & Divac, 1979) suggests that glial cells are not homogeneous in this respect. Similar interactions of neurones and glia have been proposed previously (Henn, 1976). Studies in defined cell cultures offer good possibilities for resolving such cell-cell interactions.

The initial observations on glutamate uptake in culture were made by DNC in collaboration with Dr G. R. Dutton. Gifts of materials are gratefully acknowledged from: Dr B. Anderton and Dr J. Kahn (anti-GFAP), Dr N. G. Bowery (ACHC), Dr R. Mirsky (anti-tetanus) and Dr R. O. Thompson (tetanus toxin). This work was partially supported by an MRC grant to the authors.