Phosphate activated glutaminase is a key enzyme in glutamate synthesis. Here we have employed a quantitative and high-resolution immunogold procedure to analyse the cellular and subcellular expression of this enzyme in the cerebellar cortex. Three main issues were addressed. First, is phosphate activated glutaminase exclusively or predominantly a mitochondrial enzyme, as biochemical data suggest? Second, to what extent is the mitochondrial content of glutaminase dependent on cell type and transmitter identity? Third, can individual neurons maintain a subcellular segregation of mitochondria with different glutaminase content? An attempt was also made to disclose the intramitochondrial localization of glutaminase, and to correlate the content of this enzyme with that of glutamate and glutamine in the same mitochondria (by use of triple labelling). Antisera to the N- and C-termini of glutaminase revealed strong labelling of the putatively glutamatergic mossy fibre terminals. The vast majority of gold particles (approximately 96%) was associated with the mitochondria. Equally high labelling intensities were found in mitochondria of perikarya and dendrites in the pontine nuclei, a major source of mossy fibres. The level of glutaminase immunoreactivity in parallel and climbing fibres (which like the mossy fibres are thought to use glutamate as transmitter) was only about 20% of that in mossy fibres, and similar to that in non-glutamatergic neurons (Purkinje and Golgi cells). Glial cell mitochondria were devoid of specific glutaminase labelling and revealed a much lower glutamate:glutamine ratio than did the mitochondria of mossy fibres. As to the submitochondrial localization of glutaminase, immunogold particles were often found to be aligned with the cristae, suggesting an association of the enzyme with the inner mitochondrial membrane. However, the existence of a glutaminase pool in the mitochondrial matrix could not be excluded. The outer mitochondrial membrane was unlabelled. The present study provides quantitative evidence for a substantial heterogeneity in the mitochondrial content of glutaminase. This heterogeneity applies not only to neurons with different transmitter signatures, but also to different categories of glutamatergic pathways. In terms of the routes involved, the synthesis of transmitter glutamate may be less uniform than previously expected.