THE JOURNAL OF COMPARATIVE NEUROLOGY 307:296-310 (1991)

An Electron Microscopic, Immunogold Analysis of Glutamate and Glutamine in Terminals of Rat Spinocerebellar Fibers ZHONGQI JI, JAN-ERIK AAS, JON LAAKE, FRED WALBERG, AND OLE PETTER OTTERSEN Department of Anatomy, University of Oslo, Blindern, N-0317 Oslo 3, Norway

ABSTRACT A semiquantitative, electron microscopic immunocytochemical procedure based on the use of colloidal gold particles as markers was employed to analyze the subcellular distribution of glutamate and glutamine, a major glutamate precursor, in a subpopulation of spinocerebellar mossy fiber terminals. These terminals were identified by anterograde transport of a horseradish peroxidase-wheat germ agglutinin conjugate, injected in the thoracic spinal cord. Gold particles signalling glutamate-like immunoreactivity were enriched over clusters of synaptic vesicles relative to organelle-free cytoplasmic matrix, and there was a strong positive correlation between gold particle and synaptic vesicle densities (correlation coefficient 0.94). Gold particles indicating glutamine-like immunoreactivity showed a much weaker correlation with vesicle density (correlation coefficient 0.36) and were about equally concentrated over cytoplasmic matrix as over clusters of synaptic vesicles. Compared with the mossy fibers, the putative GABAergic Golgi cell terminals exhibited a lower level of glutamate-like immunoreactivity, which was very weakly correlated with the vesicle density (correlation coefficient 0.27). The level of glutamine-like immunoreactivity in the Golgi cell terminals was similar to that in mossy fibers, but much lower than that in glial cells. The anterogradely labelled mossy fiber terminals were not enriched in immunoreactivities for aspartate or GABA. These results suggest that the level and subcellular distribution of glutamate in presumed glutamatergic terminals differs from that in terminals in which glutamate only serves metabolic or precursor roles, and that these differences can be exploited in immunocytochemical studies aimed at identifying glutamate-using neurons. In contrast, glutamine immunocytochemistry does not seem to be generally useful in this regard. Key words: synapticvesicles, cerebellum, neurotransmission, quantitation

The interest in glutamate as a CNS transmitter has increased substantially during the past few years following the characterization of its receptors and the realization that receptor-mediated effects of glutamate may play crucial roles in diverse processes such as long-term potentiation, spread of epileptic activity, and neuronal degeneration in ischemia and other pathological states (Lodge, '88). An understanding of glutamate's role in these conditions requires not only a detailed insight in the properties of the receptors, but also knowledge about the cellular and subcellular localization of glutamate and the factors that govern glutamate turnover and release. For glutamate the paradoxical situation exists that only a minor proportion of it is assumed to be releasable and relevant for the receptors: the major proportion is believed to subserve a variety of metabolic functions restricted to intracellular compart-

o 1991 WILEY-LISS, INC.

ments (Fonnum, '84).It is possible, however, that metabolic glutamate may gain access to the extracellular space in certain nonphysiological conditions (Nicholls, '89; Torp et al., '90). These considerations underline the importance of being able to distinguish between transmitter glutamate and metabolic glutamate. Synaptosome preparations offer valuable clues in this regard (Nicholls, '891, but in the ideal situation it should be feasible to make this distinction in preparations with intact anatomical integrity, To this end, Accepted January 9,1991. Zhongqi J i is on leave of absence from the Department of Anatomy, Capital Institute of Medicine,You An Men Street, Beijing, China. Address correspondence to Ole P. Ottersen, Department of Anatomy, University of Oslo, P.O. Box 1105Blindern, N-0317 OSL03, Norway.

297

NEUROACTIVE AMINO ACIDS IN MOSSY FIBERS immunocytochemistry would seem to be the method of choice. Antibodies that selectively recognize fixed glutamate are now available (Storm-Mathisen et al., '83; Ottersen and Storm-Mathisen, '84; Hepler et al., '88; Chagnaud et al., '89; Liu et al., '89). Under identical conditions with respect to fixation and accessibility, such antibodies naturally reveal both transmitter and metabolic glutamate with the same ease, since different functions are not betrayed by differences in molecular structure. However, differences in immunolabelling intensity may be exploited in attempts to separate transmitter and metabolic glutamate, assuming higher glutamate concentrations in the former pool (Somogyi et al., '86; Ottersen, '89; Watson, '88; Montero and Wenthold, '89; Bramham et al., '90; Maxwell et al., '90). This distinction is useful for the interpretation of nerve terminal immunolabelling, but presumably of less relevance with regard to the immunolabelling of cell bodies, which are distant from the site of transmitter release and whose glutamate contents may predominantly reflect the level of metabolic glutamate (Ottersen and Storm-Mathisen, '84). However, even in nerve terminals a demonstrable enrichment of glutamate is unreliable as the sole criterion for the identification of a transmitter pool, as pronounced variations in the level of metabolic glutamate possibly occur also in this compartment. In the present study we explored whether the subcellular distribution of glutamate and of its major precursor, glutamine (Bradford et al., '78; Hertz and Schousboe, '88), could serve as additional guides for distinguishing between the different pools of glutamate. The subcellular distribution of fixed glutamate can now be studied with great accuracy and in a semiquantitative manner following the adaptation of the postembedding immunogold procedure (Somogyi et al., '86; Ottersen, '87). We have employed this procedure and used mossy fiber boutons of the cerebellum as a model of a presumed glutamatergic terminal. Mossy fiber boutons are enriched in glutamate-like immunoreactivity (Somogyi et al., '86; Liu et al., '89; Ottersen, '89), which can be depleted in a calcium-dependent manner on depolarization with high K' (Ottersen, '89; Ottersen and Laake, '901, as expected if the immunolabelling reflects a transmitter pool. Electrophysiological and pharmacological data are in line with the immunocytochemical results inasmuch as they indicate that the mossy-fiber-induced excitation is mediated by a compound acting on excitatory amino acid receptors, mainly of the non-N-methyl-D-aspartate (NMDA) type (Garthwaite and Brodbelt, '89, '90). Further, the mossy fibers are convenient objects for the present type of study due to their extraordinarily large size and distinct compartmentation. A possible pitfall, however, lies in the fact that the mossy fiber system appears to be biochemically heterogeneous. In addition to glutamate, y-aminobutyric acid (GABA) and several peptides including enkephalin, substance P, and corticotropin-releasing factor have been reported to occur in this system (Korte et al., '80; Schulman et al., '81; King et al., '87; Cummings et al., '88; Hamori and Takacs, '89). The biochemical heterogeneity reflects the multiple origins of the mossy fibers (Dietrichs and Walberg, '79; Bloedel and Courville, '81; Tolbert, '82; Walberg, '82; Brodal, '87; Grant and Xu, '88). On this background, it was deemed necessary to single out a subpopulation of mossy fibers for the present study. To this end we identified spinocerebellar mossy fibers from a restricted part of the spinal cord, by use of the

anterograde tracer horseradish peroxidase-wheat germ agglutinin conjugate (HRP-WGA). The distributions of glutamate-like and glutamine-like immunoreactivities in these terminals were compared with those of other amino acids thought to be involved in synaptic functions.

MATERIALS AND METHODS Experimental material Male Wistar rats (n = 15; 200-300 g body weight) were deeply anaesthetized with pentobarbital (50 mgkg i.p.) and an aqueous solution of HRP-WGA (Sigma V1, 2%, total amount 0.2-0.4 p1) was subsequently injected through a Hamilton microsyringe into the spinal cord at two different sites (separated rostrocaudally by 1-2 mm) at the level of Th7-Th8. After a survival time of 68-72 hours, the rats were anaesthetized as above and perfused for 10-15 seconds through the ascending aorta with 2% dextran (MW 70,000) in 0.1 M sodium phosphate buffer (pH 7.4,4"C)and then for 15 minutes with a mixture of 1% paraformaldehyde and 2.5% glutaraldehyde (G) in the same buffer (room temperature). The cerebellum and the injected segments of the spinal cord were immediately removed and sectioned in the transverse plane on a Vibratome (50 pm) and a freezing microtome (60 pm), respectively. After treatment with tetra-methyl-benzidine/HzOz(Henry et al., '85; Broman et al., '901, the injection site and anterogradely labelled fibers were identified in a dissection microscope and suitable specimens isolated and treated overnight with 1% OsO, in sodium phosphate buffer, pH 5.5. After rapid dehydration in graded ethanols and acetone (total duration 25-30 minutes), the blocks were flat embedded in Durcupan (ACM, Fluka). Anterogradely labelled terminals were reidentified in semithin sections before series of ultrathin sections were obtained for electron microscopy.

Immunogold procedure The procedure was based on that of Somogyi and Hodgson ('85) and has been described in detail previously (Ottersen, '87, '89). Following etching in 1% HIO, (7 minutes) and 9% NaIO, (15 minutes) in distilled water, the ultrathin sections were preincubated in 1%human serum albumin in Tris-phosphate-buffered saline (TPBS; 20 minutes), and then incubated in one of the following antiserum preparations [antiserum code no., dilution (in TPBS), additive, incubation time]: 1)glutamate 03, 1:500, aspartate-G (300 pM) and glutamine-G (300 pM), 2 hours; 2) glutamine 34, 1500, aspartate-G (100 pM), 2 hours; 3) aspartate 18, 1:30, glutamate-G (30 pM) and taurine-G (30 pM), overnight; 4) GABA 25, 150, glutamate-G (300 p M ) and p-alanine-G (300 pM), overnight. The amino acid-glutaraldehyde complexes (prepared as described by Ottersen et al., '86) were added to the antisera to remove traces of crossreactivity. After a brief rinse in polyethyleneglycol (50 mg/100 ml in 0.05 M TrisiHCl buffer, pH 7.4), the sections were exposed for 2 hours to goat anti-rabbit TgG coupled to colloidal gold particles (diameter 15 nm; Janssen), diluted 1:20 in the polyethyleneglycol solution. The sections were stained with 1%uranyl acetate (20 minutes) and 1%lead citrate (2 minutes) and examined in a Philips EM400 or CMlO electron microscope,

Specificity controls All antisera used in the present report have been characterized previously (Storm-Mathisenet al.,'83; Ottersen and

298 Storm-Mathisen, '84, '85; Laake et al., '86; Zhang et al., '90) and shown to react selectively with the appropriate amino acid conjugate. The antisera distinguish perfectly between glutamate, aspartate, glutamine, and GABA, and display no significant crossreactivity with the about 40 naturally occurring molecules that have been tested (Ottersen et al., '86). In addition, the glutamate and aspartate antisera have been screened against a small molecular extract of rat brain after separation by thin layer chromatography. In this test, each antiserum labelled a single spot that had co-migrated with authentic glutamate and aspartate, respectively (Zhang et al., '90). As a current quality control, each tissue section that was used in the present study was incubated together with a test section (Figs. 2 and 7 ; prepared according to Ottersen, '87), containing fixation complexesof six different amino acids. These amino acids are the ones that are most likely to interfere with the immunocytochemicalreaction, inasmuch as they are structurally interrelated and occur in particularly high concentrations in the mammalian brain. The present strategy, entailing a simultaneous and identical processing of tissue sections and test antigens, ensures that the test data adequately reflect the quality of the immunocytochemical reaction (Ottersen, '89). Blocking experiments were performed by preadsorbing the antisera with glutaraldehyde complexes of the respective amino acids (300 FM). As an additional control, some experiments were carried out using a preimmune serum instead of the specific antiserum.

Z. JI ET AL. be seen in fortunate cases (Fig. 1). Labelled structures resembling granule cells were never detected, suggesting an absence of transsynaptic HRP-WGAtransport.

Specificity The model sections that had been incubated together with the tissue sections confirmed that the antisera reacted selectively with the appropriate antigen under the conditions of the immunocytochemicalprocedure (Figs. 2 and 7; Table 1).Preabsorption of the sera with glutaraldehyde complexes of the homologous amino acids abolished labelling of tissue and model sections, as did substitution of the primary antiserum with a preimmune serum.

Glutamate-like immunoreactivity

The distribution of glutamate immunoreactivity within the granule cell layer of the rat and cat cerebellum has been described in detail in previous reports (Somogyi et al., '86; Ottersen, '87, '89; Liu et al., '891, and the present findings are in agreement with those. The different constituents of the cerebellar glomeruli, i.e., the mossy fiber terminals, granule cell dendritic digits, and Golgi cell terminals, showed different immunolabelling intensities, with the highest gold particle density in the former, and the lowest particle density in the latter component (Fig. 3). Cell bodies and processes of glial cells were weakly labelled. The general pattern of immunolabelling in the HRP-reacted sections did not differ from that in control sections incubated in the same drops of immunoreagents. Computer analysis The ultrathin sections of HRP-reacted material revealed Electron micrographs (55,500 final magnification) were that the structures that were identified as HRP-positive in analysed on a digitising tablet coupled to a Cromemco semithin sections were mossy fiber terminals containing CS300 computer. The gold particles over the different aggregates of HRP reaction product (Figs. 3 and 9). The tissue profiles and subcellular compartments were counted amount and appearance of the HRP reaction product were and the particle densities were determined by means of a similar in immunolabelled and nonimmunolabelled seccomputer program (MORFOREL; Blackstad et al., '90). tions, indicating that it had not been significantly quenched This program allowed simple statistical evaluations; for by the immunocytochemicalprocedure. The visual impresanalysis of variance, the data were transferred to a commer- sion was that the subcellular distribution of gold particles cial package (SPSS/PC+). The areas occupied by HRP in HRP-labelled mossy fibers mimicked that in the HRPnegative ones, but that very few gold particles occurred over reaction products were excluded from analysis. the areas occupied by the HRP reaction product. The For establishing the correlation between gold particle and vesicle densities, the number of particles and vesicles HRP-positive mossy fiber terminals did not exhibit morphowere counted within a set of squares (each covering an area logical features that could serve to distinguish them from corresponding to 0.04 or 0.025 ym2 in the electron micro- the HRP-negative ones. Quantitative analysis (Figs. 4 and 5) revealed that algraphs), drawn on a transparent overlay. Squares were accepted for analysis only if their total area fell within though both HRP-positive and HRP-negative mossy fiber cytoplasmic matrix or vesicle-containingareas of the profile terminals were strongly labelled, the former group of in question. Thus, squares encroaching on mitochondria or terminals was associated with slightly lower particle dension neighbouring profiles were disregarded. Matching pairs ties than the latter. Golgi cell boutons consistently disof data were subjected to regression analysis, using a played lower particle densities than the mossy fiber boutons (Fig. 4). computer program (Teoregres). Analysis of the subcellular distribution of glutamate immunoreactivity in HRP-positive mossy fibers showed that the particle density was significantly higher over RESULTS clusters of synaptic vesicles than over organelle-free cytoSections through the injection site revealed that the plasmic matrix (Fig. 5).The highest particle concentrations HRP-WGA involved almost the entire area of the ipsilateral were found over mitochondria. HRP-negative and HRPhalf of the spinal cord, and a large part of the contralateral positive mossy fibers displayed similar patterns of immunogray matter (Fig. 1).The anterogradely labelled fibers were labelling (Fig. 5 ) . The values for vesicle immunolabelling given in Fig. 5 distributed in the granule cell layer of cerebellar cortex as previously described (Yaginuma and Matsushita, '87, '89; were based on countings over the entire vesicle-containing Matsushita, '88). Both in Vibratome and in semithin sec- domains, since the resolution of the procedure (lateral tions, the HRP reaction product identified structures with resolution about 25 nm) does not allow conclusions to be the size and irregular outline typical of mossy fiber termi- drawn with regard to individual vesicles. This approach nals (Palay and Chan-Palay, '74). Preterminal axons could leads to an underestimation of the level of immunoreactiv-

NEUROACTIVE AMINO ACIDS IN MOSSY FIBERS

299

Fig. 1. Anterogradely labelled mossy fibers (arrowheads) in a Vibratome (A) and semithin (B)section ofvermis following an injection of HRP-WGA in the thoracic spinal cord ( C ) . Arrows in A indicate preterminal axon. G , granule cell layer; W, white matter. Asterisks in B,

granule cell bodies. D, dorsal; V, ventral. Hatched area in C shows the diffusion area of the injected tracer. The micropipette track is indicated (arrow).Scale bars = 20 pm in A and 10 pm in B.

TABLE 1. Selectivity of Antisera Under the Conditions of the Immunocytochemical Procedure’

were prcpdr1.d and tmbedded a s previously descrihed ! Ottrrsrn. ‘67r a i d iiicul)ati:d in t h r same drops 01 immunimagcnts ni the tissue section Son(.. cirniplexeJ nrade by reacting hram macromolecules with glutaraldehyde in the ah. nrv nt‘uny free amiiiu acid Thc data wrrv ohtoined from inodel *t.ctiiin. similar to thuse in Figs. 2 and 7 Values represent inean numher uf yuld particles $m’ o w r 11 conjugate cluiiips after suhtractiny the gold particle drnsiry over cmpty rrsin r0.3 paiticlrj +m’ in left c h i n n , 1 1 in right column . ‘The alinoluir deiiaity v d u r s are not comparnhle nith thow owbr tissue pnihlra ’ince thr test cunju#dtes were prepared at high aminu acid ronrentratiuns correspondma t o 201) inM 111 the br.un 12 ..imilarly high srlectivity witi obtained for the other aiiiiwra iiwd in the preaent repurt !data not shown GABA, y;iminohutyiracid

difference in labelling intensity between “vesicle clusters” and mitochondria (see legend to Fig. 5). The approach on which Fig. 5 is based does not give a complete picture of the relationship between gold particle concentration and vesicle packing density, since the latter varies considerably within the vesicle-containing domains. To elucidate this relationship, we counted the number of gold particles and vesicles in several discrete areas within each terminal in order to establish a regression line (Fig. 6). In the HRP-positive mossy fibers, there was a strong linear correlation between gold particle and vesicle densities, with a correlation coefficient of 0.94 (Table 2). The correlation coefficient for the HRP-negative mossy fiber terminals was lower (0.74). The distribution of glutamate immunoreactivity in the presumed GABAergic Golgi cell terminals showed a very weak correlation with vesicle density (Table 2). The determination coefficient for the latter terminals was 0.07, indicating that differences in vesicle density only explain 7% of the total variation in gold particle density.

ity in the vesicles, since these are “diluted” by the less strongly labelled cytoplasmic matrix. Indeed, correcting for this dilution effect will strongly reduce or eliminate the

The distribution of glutamine-like immunoreactivity in the mossy fiber terminals differed considerably from that of glutamate immunoreactivity, inasmuch as the gold particle

Amino acid coniugate Glutamine Aspartate None Glycine Taurine Glutamate GABA “I’rst conjugates

36.2 f 8.0 ( 6 ) 15.3 i 4.0 (5) 11.9 f 3.2 (5) 1.2 f 0.8 (7) 1.6 ? 1.1 (5) 738.0 f 55.4 ( 6 ) 7.0 3.5 (5)

*

*

1,333.5 131.2 ( 6 ) 11.3 f 3.2 ( 6 ) 0.0 f 0.5 ( 6 ) 15.5 5.2 ( 6 ) 0.2 ? 1.3 (7) 24.8 f 5.8 (7) 19.8 f 2.9 (6)

*

~ A i i i i i i o ucid-glurllrllldeh?de.~r~ii~nacriimiilecuh~ voniplexesi

Glutamine-like immunoreactivity

300

Z. JI ET AL.

Fig. 2. Ultrathin test sandwich (Ottersen, '87) incubated in the same drop of glutamate antiserum as the tissue sections. Test antigens appear as electron-densebodies surrounded by resin. The test antigens were prepared by reacting a n amino acid with glutaraldehydei formaldehyde in the presence of an extensively dialyzed brain protein extract. They should thus be structurally similar to the amino acid

fixation complexes that are formed in the tissue during perfusion fixation. Spacer sections (2 pm) are from rat hippocampus. Arrows indicate bodies enlarged in right panel. The glutamate conjugates are selectively labelled (see Table 1 for quantitative analysis). GLN, glutamine; ASP, aspartate; GLY, glycine; TAU, taurine; GLU, glutamate; GABA, y-aminobutyric acid.

density was about equally high over organelle-free cytoplasmic matrix as over vesicular domains (Figs. 5, 7, 8). Mitochondria were more strongly labelled than the other intraterminal compartments, but this difference was more modest than in the case of glutamate (Fig. 5 ) . Horseradish-

peroxidase-positive and HRP-negative terminals showed similar labelling patterns (Figs. 5 , 7,8) and a similarly weak correlation between gold particle and vesicle density (correlation coefficient 0.36; Table 2). Astrocytes and Bergmann glia were enriched in glutamine-like immunoreactivity, and

NEUROACTIVE AMINO ACIDS IN MOSSY FIBERS

Fig. 3. Glutamate (GLU) immunolabelling of identified spinocerebellar mossy fiber boutons. The HRP reaction products are indicated by arrowheads. Note strong labelling over mitochondria (m) as well as over vesicle-containing areas. Asterisks, granule cell dendritic digits; Go, Golgi cell terminals; Mf, mossy fiber terminals; Ax, myelinated axon. Scale bar = 0.4 bm.

301

302

Z. JI ET AL.

.

160 140

N

120

E 1 \

100

m

-.U aJ

80

c L

CI

a

-

60

U

0

40

\ \

20

0

I

1

I

I

I

I

I

I

I

I

I

I

1

2

3

L

5

6

7

8

9

10

11

12

I ~

I

. l

l

13- 14- 1 5 16 p r o f i l e No

Fig. 4. Diagram showing the levels of glutamate-like immunoreactivity in HRP-negative mossy fiber boutons (thick line), HRP-positive mossy fiber boutons (thin line), and Golgi cell boutons (interrupted line). The intraterminal areas that were occupied by the HRP reaction products and the overlying gold particles were excluded from analysis. Each data point represents three terminals, except in the case of the HRP-positive mossy fibers, each of which is represented by a single dot. Each number along the abscissa represents a separate set of profiles, recorded from a restricted area of the granule cell layer in an individual tissue section. (To avoid bias, the HRP-negative terminals and Golgi cell terminals that were chosen for analysis were always those that were located in the closest proximity to the matching HRP-positive terminal.) Note that the relative immunolabelling density of the three profile

types is highly consistent. However, the absolute particle density in the different sets of profiles varies somewhat, probably as a result of inadvertent fluctuations in incubation parameters from section to section. The differences in labelling intensity between the three different types of profiles are highly significant statistically (P < 0.001; t-test for nonindependent samples). The gold particle density over HRP-positive terminals was 15.1% lower than that over the HRPnegative ones. The average vesicle density (total number of vesicles/ total nerve terminal area minus area occupied by HRP reaction products) in the 16 HRP-positive and 48 HRP-negative mossy fibers represented in this figure was 105.0 and 102.5 per +m2, respectively (difference not statistically significant).

constituted the most strongly labelled elements in the cerebellar cortex (Fig. 8; Table 3). Golgi cell terminals and parallel fiber terminals displayed gold particle densities similar to those over mossy fiber boutons (Table 3).

DISCUSSION Glutamate: transmitter pool versus metabolic pool

Aspartate-likeimmunoreactivity In sections treated with the aspartate antiserum, the HRP-positive mossy fibers were associated with a gold particle density similar to that over empty resin (Fig. 9B). This was not due to method failure, since: 1) adjacent sections through the same mossy fiber terminals, processed in the same immunocytochemical experiment, exhibited glutamate-like immunoreactivity as expected (Fig. 9A); and 2) model sections incubated in the same drop of aspartate antiserum as the tissue section revealed strong labelling of the aspartate-glutaraldehyde-brainprotein conjugate (inset, Fig. 9B).

GABA-like immunoreactivity None of the mossy fibers that were anterogradely labelled from the spinal cord was enriched in GABA-like immunoreactivity. Scattered gold particles could be found over some of their mitochondria (Fig. lo), but in no case was the particle density over mossy fiber boutons comparable to that over Golgi cell terminals.

The high concentration and even distribution of glutamate in the brain were two factors that contributed to the long delay between the discovery of the powerful excitatory effect of glutamate (Curtis and Watkins, '60) and the general acceptance of glutamate as a transmitter. These factors have also limited the usefulness of glutamate immunocytochemistry as a tool for identifying glutamatergic neurons. The major problem has been t o separate immunolabelling reflecting metabolic glutamate from immunolabelling representing the transmitter pool. The solution of this problem seems now to be under way, due to recent methodological advances based on the use of the postembedding immunogold technique (review: Ottersen, '89). This technique offers a high spatial resolution and lends itself easily to quantification. By virtue of these properties, two distinctive features of the transmitter pool, namely its enrichment in the appropriate classes of nerve terminals and its sensitivity to depolarizing stimuli, are now amenable to immunocytochemical analysis. In the present report we exploit further the high resolution of the postembedding immunogold technique to investigate whether glutamatergic terminals are characterized

NEUROACTIVE AMINO ACIDS I N MOSSY F I B E R S GLU

303

GLU whole terminal ,-%*

vesicle clusters

T

mltochondria

E

cytosol

lZ0I

.

100

w

a,

801

GLN

GLN

601

*

T

**

L

0 -0 a

-

T

HRP +

HRP-

Fig. 5. Subcellular distribution of glutamate (GLU)-like and glutamine (GLN)-like immunoreactivity in horseradish peroxidase (HRP)positive (HRP+) and HRP-negative (HRP-) mossy fiber terminals. “Cytosol” represents areas of cytoplasmic matrix (larger than 0.1 p,m? devoid of vesicles or other organelles. Values for “vesicle clusters” were obtained by counting particles over vesicles plus intervening cytoplasmic matrix. Assuming that the vesicles constitute one-third of the volume in the vesicle clusters, the level of glutamate immunoreactivity

HRP+

HRP-

over vesicles would correspond to 141 particles/pm* (calculated for the HRP+ terminals). The absolute values presented here are not comparable to those shown in the previous figure and Table 3, which were from different experiments. The values are based on observations from 20 terminals. Standard deviation is indicated. Asterisks: significantly different from vesicle clusters (*, P < 0.05; **, P < 0.01; ***, P < 0.001). Star: significantly different from mitochondria ( P < 0.001; Student’s t-test).

TABLE 2. Correlation Between Immunolabelling Intensity (in terms of gold uarticlesium*)and Packing Densitv of SvnaDtic Vesicles’ Amino acidtype of profile Glutamate in mossy fiber terminals of spinal ongm (HRP+) HRPgeneral population (HRP+ plus HRP-) Glutamine in mossy fiber terminals of spinal ongm (HRP+) HRPgeneral population (HRP+ plus HRP-) Glutamate in Gals cell termmals

Correlation coefficient (r)

Determination coefficient (r2)

Y-intercept

Slope

n

t -

0.94*** 0.74*** 0.79*** 0.36** 0.36** 0.36***

0.88 0.55 0.63 0.13 0.13 0.13 0.07

-0.2 0.15 0.03 -0.49 -0.04 -0.19 -0.01

0.42 0.35 0.37 0.17 0.12 0.14 0.17

26 101 127 62

6 9 15 4 6

0.27*

72 134 55

10

13

‘The data were accumulated as explained in Materials and Methods. The values for glutamate are derived from the data presented in Fig. 6. All parameters were determined by use of a computer program for regression analysis (Teoregres).n, number of observations; t, number of nerve terminals. The values for the “general population” of mossy fibers are somewhat biased towards HRP-positive (HRP+ 1 terminals since these are overrepresented in the sample. HRP-, HRP-negative mossy fiher terminals. Asterisks: values of r significantly different from zero.

*P= 0.05. **P< 0.01. ***P< 0.001.

by a distinctive subcellular distribution of glutamate. For this purpose we identified a small subpopulation of cerebellar mossy fibers, by injecting an anterograde tracer in a restricted portion of the spinal cord. Our finding that the anterogradely labelled mossy fibers were enriched in glutamate-like immunoreactivity (although slightly less so than was the majority of the mossy fibers; see below) is in line with previous studies of mossy fibers in general (Somogyiet al., ’86;Ottersen, ’87, ’89;Liu et al., ’89) and supports the notion that these fibers are glutamatergic. Within the identified mossy fiber terminals, a high level of glutamate-likeimmunoreactivity was found in mitochondria. This fits with the mitochondria1 localization of phosphate-activated glutaminase, the major glutamate synthesizing enzyme (Palaiologos et al., ’89). However, allowance

must also be made for the possibility that the mitochondria offer particularly favourable fixation conditions for glutamate (see below), and that this may have contributed to the strong labelling in this compartment. Of more direct relevance with regard to the distinction between metabolic and transmitter glutamate was the finding that gold particles were enriched over synaptic vesicle-containing areas relative to organelle-free cytoplasmic matrix. Similar results were obtained for mossy fiber terminals in the hippocampus (Ottersen et al., ’90) and a class of nerve terminals in the locust (Watson, ’88). To gain a better impression of how the immunolabelling intensity is related to the packing density of synaptic vesicles, the correlation between these two parameters was established by regression analysis. Our finding that the gold particle

Z. JI ET AL.

304

:/

I

.

,

: : :/ X

I

..::,:;..