Pre-embedding Staining for GAD 67 Versus Postembedding Staining for GABA as Markers for Central GABAergic Terminals

Abstract

Pre-embedding immunocytochemistry for the active form of glutamate decarboxylase (GAD₆₇) and postembedding staining for γ-aminobutyric acid (GABA) were compared as markers for central GABAergic terminals in the phrenic motor nucleus, in which phrenic motor neurons had been retrogradely labeled with cholera toxin B-horseradish peroxidase. Nerve terminals with or without GAD₆₇ immunoreactivity were identified in one ultrathin section. GABA was localized with immunogold in an adjacent section after etching and bleaching. GABA labeling density was assessed over 519 GAD₆₇-positive and GAD₆₇-negative nerve terminals in the phrenic motor nucleus. Frequency histograms showed that statistically higher densities of gold particles occurred over most GAD₆₇-positive terminals. However, some GAD₆₇-negative terminals also showed high densities of gold particles, and some GAD₆₇-positive terminals showed low densities. Preabsorption of the anti-GABA antibody with a GABA-protein conjugate, but not with other amino acid-protein conjugates, significantly reduced gold labeling over both GAD₆₇-positive and GAD₆₇-negative terminals. These results show that the presence of GAD₆₇ immunoreactivity correlates strongly with high densities of immunogold labeling for GABA in nerve terminals in the phrenic motor nucleus. Preabsorption controls indicate that authentic GABA was localized in the postembedding labeling procedure. Only a small proportion of intensely GABAimmunoreactive terminals lack GAD₆₇, suggesting that both GAD₆₇ and GABA are reliable markers of GABAergic nerve terminals.

Keywords

amino acid immunocytochemistry motor neuron spinal cord ultrastructure

GABAergic nerve terminals throughout the central nervous system have been identified using either preor postembedding immunogold methods to localize γ-aminobutyric acid (GABA), or pre-embedding immunocytochemistry to visualize glutamate decarboxylase (GAD), the enzyme required for synthesis of GABA from l-glutamate (McLaughlin et al. 1975; Gabbott et al. 1986; Babb et al. 1988; Zhan et al. 1989; Helfert et al. 1992; Tai and Goshgarian 1996). A general presumption is that all nerve terminals containing GAD₆₇, the active form of the enzyme, contain high levels of GABA, but this relationship has not been analyzed quantitatively. Maxwell et al. (1989) showed that, in the lateral cervical nucleus of the cat, GAD-immunoreactive terminals exhibited high densities of gold labeling for GABA but did not show quantitatively whether the density of GABA labeling was high over all GAD-positive terminals and low over all GAD-negative terminals. We have previously used tissue fixed with 2.5% glutaraldehyde to localize GAD₆₇ immunoreactivity and showed that it was suitable for immunogold localization of the amino acid neurotransmitter glutamate (Murphy et al. 1996).

Using similarly processed tissue in this study, we determined whether GAD₆₇-containing nerve terminals labeled by pre-embedding immunocytochemistry showed high densities of gold particles after postembedding immunogold staining for GABA in adjacent etched and bleached ultrathin sections.

Materials and Methods

Retrograde Tracing

The right hemidiaphragms of three male Wistar-Kyoto rats (250-350 g) were injected with cholera toxin B-subunit conjugated to horseradish peroxidase (CT-HRP; List Biological Laboratories, Campbell, CA) (Murphy et al. 1996).

Tissue Processing

After 2-3 days, the rats were anesthetized with sodium pentobarbital (80 mg/kg ip) and perfused with phosphate-buffered 2.5% glutaraldehyde. Cervical spinal cords were divided in two, postfixed for 60-90 min, and cut parasagitally at 50μm on a Vibratome. The sections were washed in 50% ethanol for 45 min (Llewellyn-Smith and Minson 1992). Retrogradely transported CT-HRP was visualized with a tetramethylbenzidine-tungstate (TMBT) reaction, which was subsequently stabilized with cobalt-diaminobenzidine (Llewellyn-Smith et al. 1993).

Sections were incubated in the following reagents: 10% nonimmune horse serum (NHS) for 30 min, rabbit anti-GAD₆₇ antiserum (Chemicon; Temecula, CA) 1:20,000 in 10% NHS in Tris-PBS for 48-72 hr, biotinylated donkey anti-rabbit IgG (Jackson ImmunoResearch, West Grove PA; 1:500 in 1% NHS-TPBS) for 16-24 hr, and finally ExtrAvidin-horseradish peroxidase (Sigma, St Louis, MO; 1:1500 in TPBS) for 4-16 hr. A nickel-intensified diaminobenzidine reaction revealed GAD-immunoreactive nerve fibers (Llewellyn-Smith et al. 1992b). The primary antibody to GAD is well characterized and recognizes only GAD₆₇ (Kaufman et al. 1991).

Stained sections were osmicated, exposed to uranyl acetate, dehydrated, and embedded flat in Durcupan (Sigma). At least two sets of paired serial sections were cut from one tissue block from each of the three animals. Serial ultrathin sections were collected on formvar-coated nickel single-slot grids. Grids were kept in pairs and in serial order. One grid of each pair was stained with Reynold's lead citrate to identify GAD₆₇-immunoreactive nerve terminals. GABA was localized by postembedding immunocytochemistry in the other.

Postembedding Immunocytochemistry

Sections were etched in 2% aqueous periodic acid for 7 min and then bleached for 7 min in 2% aqueous sodium metaperiodate [adapted from Somogyi and Soltész (1986)]. Between etching and bleaching, the sections were rinsed in three changes of distilled water, followed by immersion for 5 min. Sections were then processed for immunogold localization of GABA (Llewellyn-Smith et al. 1992a). After washing in Trisbuffered saline with Triton X-100 (TBST), pH 7.6, the grids were immersed overnight in rabbit anti-GABA antiserum (Arnel Products, Brooklyn, NY; 1:2500 or 1:5000 in TBST, pH 7.6). The primary antibody recognizes GABA but not a wide range of other amino acids (Hepler et al. 1986). The grids were washed in TBST, pH 7.6 then pH 8.2, before incubation for 1 hr in secondary antibody conjugated to 15-nm gold particles (Amersham, Arlington Heights, IL; 1:25 GAR Auroprobe EM in TBST, pH 8.2). After washing, the grids were stained with uranyl acetate and lead citrate.

Absorption Controls

Controls for the specificity of GABA immunogold labeling included omitting the primary or secondary antibody and replacing the primary antibody with normal rabbit serum. Overnight preabsorption of the primary antibody with a variety of amino acid-hemocyanin-glutaraldehyde conjugates was also carried out to test crossreactivity (Seguela et al. 1984). Hemocyanin conjugates of glutamate, glycine, and glutamine were prepared, resuspended in 500μl distilled water, and diluted to a final conjugate concentration of 1:100 with 1:2500 anti-GABA antibody in TBST, pH 7.6. A GABA-hemocyanin conjugate was similarly diluted with primary antibody to give final conjugate concentrations of 1:100 to 1:2500. The conjugate-antibody mixtures were stored overnight at 4C, as was 1:2500 anti-GABA without conjugate. Postembedding staining was done on two sections from runs of three serial sections, with the third section left unstained. Primary antibody preabsorbed with one of the four amino acid-hemocyanin conjugates (GABA, glutamate, glycine, or glutamine) was used on the first section and unabsorbed anti-GABA antibody on the second. After staining, the density of gold particles over 20 nerve terminals in the phrenic motor nucleus was measured for each treatment. The nerve terminals assessed for GABA immunoreactivity were then located in the third unstained section and assessed for GAD₆₇ immunoreactivity. In separate control experiments, the primary antibody was replaced with amino acid-glutaraldehyde-hemocyanin conjugate (1:2500).

Data Analysis

One pair of serial ultrathin sections per block was randomly selected for identification of GAD₆₇-immunoreactive varicosities in the phrenic motor nucleus. In each section postembedding-stained for GABA, nerve terminals in the vicinity of retrogradely labeled phrenic motor neurons were photographed randomly, numbered, located in the adjacent section on the unlabeled grid, and scored for GAD₆₇ immunoreactivity. Terminals that could not be identified in the untreated sections were not analyzed.

GABA immunoreactivity was assessed on micrographs printed to × 37,000-40,000. The areas of axon terminals and of mitochondria within them were measured with a digitizing tablet. Gold particles over axoplasmic matrix plus synaptic vesicles and over mitochondria were counted manually. GABA labeling density (gold particles/μm²) was then calculated for axoplasmic matrix plus synaptic vesicles (excluding mitochondria), for mitochondria, and for the whole terminal (axoplasmic matrix, synaptic vesicles, and mitochondria). Labeling over blank resin was determined from gold particle densities over the lumina of at least four blood vessels per sample and ranged from 0.5 to 1.6 gold particles/μm². Because this labeling was present at such a low level, we did not correct for it in any of our measurements.

Terminals identified as GAD₆₇-positive or GAD₆₇-negative were sorted into groups on the basis of their gold particle labeling density for GABA. Frequency distributions for the densities of immunogold labeling over synaptic vesicles and axoplasmic matrix were constructed for all analyzed terminals from each section because this tissue compartment appeared most likely to reflect the distribution of neurotransmitter GABA.

Statistics

All statistical analyses were carried out according to Sokal and Rohlf (1995) and Rohlf and Sokal (1995).

Nerve terminals from GABA-labeled sections were sorted into GAD₆₇-positive and GAD₆₇-negative categories. In the four sets of data collected, tests for homogeneity of variance between the GAD₆₇-positive and GAD₆₇-negative groups showed that the variances of these two groups were heterogeneous in almost all cases. Where this occurred, these data were log-transformed before unpaired t-tests were used to compare the density of labeling of the two groups of nerve terminals. The mean densities are given after back-transformation from their log values.

In the preabsorption control experiments, comparisons of the density of gold labeling were made over groups of GAD₆₇-positive and GAD₆₇-negative nerve terminals. The same terminals were identified in adjacent sections. One section of each pair was stained for GABA and the other with anti-GABA antibody that had been preabsorbed overnight with one of the four amino acid-hemocyanin conjugates. In these cases, the density of labeling over GAD₆₇-positive or GAD₆₇-negative nerve terminals over the area of synaptic vesicles and axoplasmic matrix was compared with a paired t-test. In several cases, the variances of the densities of gold labeling were heterogeneous, so data were log-transformed before the statistical test was performed.

Results

Four pairs of adjacent ultrathin sections through the phrenic motor nucleus were examined from the three rats that had received injections of CT-HRP into the diaphragm. One section of each pair was untreated so that nerve terminals could be classified as GAD₆₇-positive or GAD₆₇-negative. The other section was stained for GABA after being etched to remove resin and bleached to remove DAB reaction product. An initial sample of 697 GAD₆₇-immunoreactive terminals in the phrenic motor nucleus was photographed; 25% of these (178) made synapses and direct contacts with retrogradely labeled phrenic motor neurons. Five hundred and nineteen of the 697 terminals could be identified in both GABA- and GAD₆₇-labeled sections and were analyzed further for their density of gold labeling and their GAD₆₇ content.

When GAD₆₇-positive and GAD₆₇-negative terminals were assessed for GABA gold labeling in the adjacent section, GAD₆₇-positive nerve terminals generally appeared to have higher levels of GABA labeling than GAD₆₇-negative nerve terminals (Figures 1, 2A, and 2B). Frequency distributions for gold labeling density over synaptic vesicles and axoplasmic matrix were constructed first for all nerve terminals examined in each section (Figure 3A) and then for nerve terminals that had been sorted into GAD₆₇-positive and GAD₆₇-negative classes (Figure 3B). These histograms confirmed that GAD₆₇-positive nerve terminals generally showed higher densities of gold labeling for GABA than GAD₆₇-negative terminals. The presence of two distinct peaks in the overall frequency distribution (Figure 3A) suggested a population of nerve terminals with a low GABA labeling density and a population with a high GABA labeling. However, when frequency distributions were constructed after sorting terminals into GAD₆₇-positive and GAD₆₇-negative classes, the two populations were not as well separated as the overall frequency distribution suggested, i.e., there was some degree of overlap between the labeling densities of GAD₆₇-negative and GAD₆₇-positive nerve terminals (Figure 3B). There were nerve terminals that lacked GAD₆₇-immunoreactivity in the plane of section but showed a high density of GABA labeling. There were also GAD₆₇-positive nerve terminals that displayed low levels of GABA labeling. Nevertheless, in all four pairs of sections from the three animals, unpaired to-tests showed that the density of gold labeling was significantly higher over GAD₆₇-positive nerve terminals than over GAD₆₇-negative nerve terminals (Table 1). This was true regardless of whether the density of gold labeling for GABA was measured over synaptic vesicles and axoplasmic matrix, over mitochondria alone, or over the entire nerve terminal.

Control experiments confirmed that the heavy labeling over GAD₆₇-immunoreactive terminals was likely to be authentic GABA immunoreactivity. Overnight preabsorption of the primary antibody with increasing concentrations of GABA-hemocyanin resulted in a graded decrease in labeling density over both GAD₆₇-positive and GAD₆₇-negative nerve terminals (Figures 2C and 2D), with the highest concentration (1:100) virtually abolishing gold labeling. This reduction in gold particle density was statistically significant (paired t-test p < 0.001; Table 2). In contrast, incubation of the anti-GABA antiserum with glutamate-, glycine-, or glutamine-hemocyanin conjugates did not significantly alter gold labeling density (Table 2), although the density of labeling over both GAD-positive and GAD-negative nerve terminals showed a decrease with all conjugates tested. When the GABA-glutaraldehyde-hemocyanin conjugate diluted 1:2500 was substituted for primary antibody or when primary antibody was omitted, no gold labeling was observed. Substitution of normal rabbit serum for the primary antibody resulted in very sparse labeling, with densities similar to that seen in sections incubated with GABA-hemocyanin-preabsorbed primary antibody.

Figure 1

After postembedding immunogold localization of GABA, GAD-positive nerve terminals that synapse on phrenic motor neurons show a higher density of gold particles than GAD-negative nerve terminals. (A) Three nerve terminals (1-3, box) appose the cell body of a phrenic motor neuron. They are shown at higher magnification in B. Asterisks, TMBT crystals from retrograde transport of CT-HRP. Bar = 2 μm. (B) Terminals 1 and 2 are GAD-positive and 3 is GAD-negative. (C) Adjacent section stained for GABA. After etching and bleaching, the two GAD-positive nerve terminals (1, 2) show a higher density of immunogold labeling for GABA (48 gold particles/μm² and 78 particles/ μm², respectively) than the GAD-negative nerve terminal (3; 30 particles/μm²). All three terminals make synapses with the phrenic motor neuron (arrowheads). Bar = 500 nm.

Discussion

Our correlation of the densities of GABA labeling with the occurrence of GAD₆₇ immunoreactivity in adjacent sections of nerve terminals in the phrenic motor nucleus and of terminals synapsing on retrogradely labeled phrenic motor neurons has demonstrated that GAD₆₇-positive nerve terminals have a statistically higher density of GABA gold labeling than GAD₆₇-negative terminals. Although some GAD₆₇-positive terminals showed low levels of GABA labeling and some GAD₆₇-negative terminals showed high levels, the overall correspondence between the presence of immunoreactivity for GABA and GAD₆₇suggests that both are useful markers of GABAergic terminals.

In this study, GAD₆₇-positive nerve terminals showed statistically higher GABA labeling over synaptic vesicles and axoplasmic matrix, over mitochondria, and over the whole nerve terminal, than was seen over GAD₆₇-negative nerve terminals. This difference between GAD₆₇-positive and GAD₆₇-negative nerve terminals was apparent in frequency distribution histograms based on gold particle density. Nevertheless, all of the frequency distributions for GABA labeling densities over the GAD₆₇-negative and GAD₆₇-positive groups of terminals showed some degree of overlap. Some GAD₆₇-positive terminals showed low levels of GABA labeling and some GAD₆₇-negative terminals showed levels of GABA labeling that were as high as those seen in GAD₆₇-positive terminals. GAD₆₇ terminals were marked by peroxidase reaction product, the presence of which unequivocally identifies them as possessing the active form of the enzyme necessary for GABA synthesis. However, the absence of GAD₆₇ immunoreactivity from a nerve terminal, particularly from a single ultrathin section through a nerve terminal, does not prove that the terminal cannot synthesize GABA. Because the primary antibody used here recognizes GAD₆₇ and not GAD₆₅ (Kaufman et al. 1991), we may have underestimated the true proportion of GABAergic input. Terminals that lacked GAD₆₇ mmunoreactivity but showed high levels of GABA labeling might contain GAD₆₅ and therefore be capable of producing GABA under the appropriate intracellular conditions (Kugler 1993).

Only one previous study has directly examined GABA immunogold labeling over GAD-positive and GAD-negative nerve terminals (Maxwell et al. 1989). However, it did not show quantitatively whether all GAD-positive nerve terminals were heavily labeled with GABA and all GAD-negative terminals were lightly labeled. In synaptosomes, GABA concentrations and GAD activity are generally well correlated (Löscher et al. 1985; Santiapillai et al. 1989), but these data cannot be compared with our present results because they do not provide information about individual nerve terminals. In the opossum piriform cortex, the presence of GABA immunoreactivity correlated with [³H]-GABA uptake in nerve terminals (Haberley et al. 1987), but the relationship between the presence of GAD immunoreactivity and [³H]-GABA uptake was not as consistent, possibly because the antiserum used to localize GAD did not recognize both forms of the enzyme (Feldblum et al. 1993). We believe that, in this study, the impact of any possible underestimation of the number of GAD₆₇-containing nerve terminals is small because we saw relatively few GAD₆₇-negative nerve terminals with high densities of gold labeling. For example, beyond the midpoint of the frequency distribution in Figure 3B, 10 of 11 (92%) terminals most heavily labeled for GABA were GAD₆₇-positive and only one was GAD₆₇-negative. Across the other sections examined, GAD₆₇-positive varicosities comprised 62%, 80%, and 83% of all the varicosities that exceeded the midpoint of the range of gold particle density. It is possible that the small number of GAD₆₇-negative terminals that display high levels of GABA labeling are a consequence of inadequate penetration of antibodies in the pre-embedding immunocytochemical procedure. More problematic are the nerve terminals at the lower ends of the frequency distribution, those that were GAD₆₇-positive but showed low levels of GABA gold labeling equivalent to the levels for some GAD₆₇-negative nerve terminals. If these terminals contain the active form of GAD, why should they have levels of GABA immunoreactivity similar to nerve terminals that lack GAD₆₇? Possibly the presence of some GABA in GAD₆₇-negative nerve terminals is due to production of GABA by other enzymatic reactions that do not require GAD₆₇, such as the conversion of putrescine to GABA through aldehyde dehydrogenase (Kugler 1993), or by uptake of GABA via GABA transporters (Iversen and Kelly 1975).

The data from our control experiments indicate that we have localized authentic GABA immunoreactivity in both GAD₆₇-positive and GAD₆₇-negative nerve terminals. First, there was no gold labeling over any tissue structures in the absence of primary antibody, and gold labeling was minimal in experiments in which the primary antibody was replaced with normal rabbit serum. We also showed concentrationdependent reductions in the density of gold labeling when the primary antibody to GABA was preabsorbed with increasing concentrations of a GABA-protein conjugate. This anti-GABA antibody has been used in our previous studies on amino acid content of synaptic inputs to rat sympathetic preganglionic neurons, and to bulbospinal neurons of the rostral ventrolateral medulla, in which we showed that terminals that contained high densities of immunogold labeling for GABA contained low levels of GLU labeling, and vice versa (Llewellyn-Smith et al. 1992a,1995). Furthermore, the labeling density produced by this antiserum was not significantly affected by preabsorption with glutamate-, glycine-, or glutamine-protein conjugates.

Figure 2

Control experiment showing effect of overnight preabsorption of the anti-GABA antibody with a GABA-glutaraldehyde-hemocyanin conjugate. (A) A GAD-positive nerve terminal (asterisk) is adjacent to a GAD-negative nerve terminal. (B) The GAD-positive nerve terminal (asterisk) contains a high density of gold labeling (72 gold particles/μm²) compared to the adjacent GAD-negative terminal (18 particles/μm²). (C) Section through the same nerve terminals as in A and B. The section was incubated with anti-GABA preabsorbed with a 1:500 dilution of GABA-glutaraldehyde-hemocyanin conjugate. There are a few randomly scattered gold particles over the GAD-positive (asterisk) and GAD-negative nerve terminals. (D) Serial section to C incubated with anti-GABA preabsorbed with a 1:2500 dilution of conjugate. Although there is a higher density of gold particles over this section than in C, the level of labeling is still much lower than in sections processed with the untreated primary antibody (compare with B). Bar = 500 nm.

Figure 3

Frequency distributions of gold particle densities over synaptic vesicles and axoplasmic matrix for all nerve terminals (A) and over the same nerve terminals after sorting them into GAD-negative and GAD-positive classes (B). Data from one section. Note some overlap between the labeling densities of GAD-negative and GAD-positive nerve terminals.

Table 1

Mean densities of GABA immunogold labeling (gold particles/μm²) over GAD-positive and GAD-negative terminals in the phrenic motor nucleus a

Section number		1	2	3	4
Dilution of anti-GABA		(1:5000)	(1:2500)	(1:2500)	(1:2500)
Gold labeling over synaptic vesicles and axoplasmic matrix	GAD-negative terminals	32.2 (69)	27.5 (78)	23.0 (108)	27.6 103
	GAD-positive terminals	61.4 (43)	45.5 (31)	41.2 (48)	62.5 (39)
		∗∗∗	∗∗∗	∗∗∗	∗∗∗ b
Gold labeling over mitochondria	GAD-negative terminals	123.3 (53)	85.7 (61)	84.3 (77)	85.5 (80)
GAD-positive terminals	151.0 (33)	124.2 (25)	125.6 (35)	180.7 (29)
		∗	∗∗ b	∗∗∗	∗∗∗ b
Gold labeling over whole nerve terminal	GAD-negative terminals	41.0 (69)	34.6 (78)	31.5 (108)	35.1 (103)
GAD-positive terminals	78.3 (43)	59.7 (31)	57.0 (48)	88.9 (39)
	∗∗∗	∗∗∗	∗∗∗	∗∗∗

Labeling densities were determined for the area occupied by synaptic vesicles and axoplasmic matrix, by mitochondria, and by the whole terminal. The number of terminals measured in each group is shown in parentheses. Gold labeling densities were determined from four sections, sections 1 and 2 coming from the same animal. Unpaired t-tests were used to compare gold particle densities for GAD-negative and GAD-positive terminals.

∗

p <0.05

∗∗

p <0.01

∗∗∗

p <0.001.

Log transformation not required before statistical analysis.

Table 2

Mean density of GABA immunogold labeling (gold particles/μm²) with or without overnight preabsorption of the primary antibody a

Comparison	GAD-positive nerve terminals	GAD-negative nerve terminals
GABA only	30.41	11.36
preabsorbed with GABA-HC	1.71	1.43
	∗∗∗ b	∗∗∗ b
GABA only	26.10	6.00
preabsorbed with Glu-HC	20.28	4.89
		n.s.	n.s.
GABA only	20.33	9.19
preabsorbed with glycine-HC	14.24	2.47
		n.s. b	n.s.
GABA only	22.55	7.07
preabsorbed with glutamine-HC	24.82	5.08
		n.s.	n.s.

GABA labeling density was determined over 10 GAD-positive nerve terminals and 10 GAD-negative terminals in each of four GABA-labeled sections, and over the same 20 terminals in adjacent sections incubated with GABA antiserum that had been preabsorbed overnight with one of four amino acid-hemocyanin (HC) conjugates. Data were analyzed with paired t-test.

∗

p <0.05

∗∗

p <0.01

∗∗∗

p <0.001; n.s. not significantly different.

Analysis performed on log-transformed data.

In this study, the presence of peroxidase reaction product in GAD₆₇-labeled nerve terminals did not appear to cause any significant quenching of immunogold labeling for GABA, as has been reported in other studies (e.g., Ji et al. 1991) because there was statistically higher GABA immunogold labeling over peroxidase-containing (i.e., GAD₆₇-positive) nerve terminals. Moreover, our quantitative results support the general observation of Maxwell et al. (1989) that GAD-positive nerve terminals were more heavily labeled for GABA than GAD-negative terminals.

In summary, our postembedding immunogold studies have shown that, in the phrenic motor nucleus, GAD₆₇-immunoreactive nerve terminals, including many terminals that synapse with phrenic motor neurons, are more heavily stained for GABA than terminals that lack GAD₆₇ immunoreactivity. Our results indicate that both immunoreactivity for GABA and immunoreactivity for GAD₆₇ are good markers for central GABAergic terminals.

Footnotes

Acknowledgments

Supported by grants from the National Health and Medical Research Council of Australia, the National Heart Foundation of Australia, and the National Sudden Infant Death Syndrome Research Council of Australia. SMM was supported by a Post-doctoral Research Fellowship from the National SIDS Research Council.

We would like to thank Carolyn Martin for expert technical assistance.

References

Babb

Pretorius

Kupfer

Brown

(1988) Distribution of glutamate decarboxylase-immunoreactive neurons and synapses in the rat and monkey hippocampus: light and electron microscopy. J Comp Neurol 278:121–138.

Feldblum

Erlander

Tobin

(1993) Different distributions of GAD₆₅ and GAD₆₇ mRNAs suggest that the two glutamate decarboxylases play distinctive functional roles. J Neurosci Res 34:689–706.

Gabbott

PLA

Somogyi

Stewart

Hamori

(1986) GABA-immunoreactive neurons in the rat cerebellum: a light and electron microscope study. J Comp Neurol 251:474–490.

Haberley

Hansen

Feig

Presto

(1987) Distribution and ultrastructure of neurons in opossum piriform cortex displaying immunoreactivity to GABA and GAD and high-affinity tritiated GABA uptake. J Comp Neurol 266:269–290.

Helfert

Juiz

Bledsoe

Bonneau

Wenthold

Altschuler

(1992) Patterns of glutamate, glycine, and GABA immunolabeling in four synaptic terminal classes in the lateral superior olive of the guinea pig. J Comp Neurol 32:305–325.

Hepler

Petrusz

Rustioni

(1986) Antisera to GABA, glutamate and aspartate: characterization by immunoabsorption and immunocytochemistry. J Histochem Cytochem 34:110.

Iversen

Kelly

(1975) Uptake and metabolism of g-aminobutyric acid by neurones and glial cells. Biochem Pharmacol 24:933–938.

Aas

J-E

Laake

Walberg

Ottersen

(1991) An electron microscopic, immunogold analysis of glutamate and glutamine in terminals of rat spinocerebellar fibres. J Comp Neurol 307:296–310.

Kaufman

Houser

Tobin

(1991) Two forms of the g-aminobutyric acid synthetic enzyme glutamate decarboxylase have distinct intraneuronal distributions and cofactor interactions. J Neurochem 56:720–723.

10.

Kugler

(1993) Enzymes involved in glutamatergic and GABAergic neurotransmission. Int Rev Cytol 147:285–336.

11.

Llewellyn-Smith

Minson

(1992) Complete penetration of antibodies into vibratome sections after glutaraldehyde fixation and ethanol treatment: light and electron microscopy for neuropeptides. J Histochem Cytochem 40:1741–1749.

12.

Llewellyn-Smith

Minson

Pilowsky

Arnolda

Chalmers

(1995) The one hundred percent hypothesis: glutamate or GABA in synapses on sympathetic preganglionic neurons. Clin Exp Hypertens 17:323–333.

13.

Llewellyn-Smith

Phend

Minson

Pilowsky

Chalmers

(1992a) Glutamate-immunoreactive synapses on retrogradelylabelled sympathetic preganglionic neurons in rat thoracic spinal cord. Brain Res 581:67–80.

14.

Llewellyn-Smith

Pilowsky

Minson

(1992b) Retrograde tracers for light and electron microscopy. In Bolam

, ed. Experimental Neuroanatomy: A Practical Approach. Oxford, Oxford University Press, 31–59.

15.

Llewellyn-Smith

Pilowsky

Minson

(1993) The tungstatestabilised tetramethylbenzidine reaction for light and electron microscopic immunocytochemistry and for revealing biocytinfilled neurons. J Neurosci Methods 46:27–40.

16.

Löscher

Böhme

Müller

Pagliusi

(1985) Improved method for isolating synaptosomes from 11 regions of one rat brain: electron microscopic and biochemical characterization and use in the study of drug effects on nerve terminal g-aminobutyric acid in vivo. J Neurochem 45:879–889.

17.

Maxwell

Christie

Somogyi

(1989) Synaptic connections of GABA-containing boutons in the lateral cervical nucleus of the cat: an ultrastructural study employing pre- and postembedding immunocytochemical methods. Neuroscience 33:169–184.

18.

McLaughlin

Barker

Saito

J-Y

(1975) Immunocytochemical localization of glutamate decarboxylase in the rat spinal cord. J Comp Neurol 164:305–322.

19.

Murphy

Pilowsky

Llewellyn-Smith

(1996) Vesicle shape and amino acids in synaptic inputs to phrenic motoneurons: do all inputs contain glutamate or GABA?. J Comp Neurol 373:200–219.

20.

Rohlf

Sokal

(1995) Statistical Tables. 3rd ed. New York, WH Freeman.

21.

Santiapillai

Gray

Phillips

Richardson

(1989) Isolation of nerve terminals from crustacean muscle. J Neurochem 53:1527–1535.

22.

Seguela

Geffard

Buijs

Le Moal

(1984) Antibodies against g-aminobutyric acid: specificity studies and immunocytochemical results. Proc Natl Acad Sci USA 81:3888–3892.

23.

Sokal

Rohlf

(1995) Biometry. 3rd ed. New York, WH Freeman.

24.

Somogyi

Soltész

(1986) Immunogold demonstration of GABA in synaptic terminals of intracellularly recorded, horseradish peroxidase-filled basket cells and clutch cells in the cat's visual cortex. Neuroscience 19:1051–1065.

25.

Tai

Goshgarian

(1996) Ultrastructural quantitative analysis of glutamatergic and GABAergic synaptic terminals in the phrenic nucleus after spinal cord injury. J Comp Neurol 372:343–355.

26.

Zhan

W-Z

Ellenberger

Feldman

(1989) Monoaminergic and GABA-ergic terminations in phrenic nucleus of rat identified by immunohistochemical labeling. Neuroscience 31:105–113.