and mRNA, respectively. This is a crucial level of understanding of brain organization, since it is estimated that over half of the 100,000 genes in the mammalian genome are expressed exclusively or predominantly within the brain, however, many of them are not expressed ubiquitously or in a uniform pattern. In fact, the selective and nonuniform expression of genes is a critical element in each brain region, cell type, or neural circuit's particular neurochemical phenotype, i.e., the set of proteins that are required for that region, cell class, or circuit to perform its unique role in brain function. The regional and cellular delineation of gene expression patterns is thus important as a reflection of function, but it is also increasingly critical as a background for analysis of genetic manipulations, particularly in mouse transgenic models.
While in situ hybridization and several biochemical approaches have been applied in a quantitative fashion on the regional level to reveal relative mRNA levels in one brain region compared with another, the analysis of mRNA or protein levels on a quantitative cellular level has presented special problems with respect to obtaining quantitative data. Just as we need quantitative databases that define neuronal structure, we need quantitative data on gene expression patterns, otherwise we will not be able to measure the effect of a genetic manipulation in the context of neural circuits. This will require precise measurements of protein concentration and/or mRNA levels in specific regions, identified circuits, cell classes, neuronal compartments, and at the level of the synapse. There is little doubt that the emerging gene chip technology will augment efforts to obtain regional data on gene expression patterns, and the single-cell mRNA amplification approaches pioneered by Eberwine and colleagues have already been useful to obtain mRNA expression data on the single-cell level (Eberwine et al., 1992; Kacharmina et al., 1999). In addition, quantitative immunocytochemical analyses of protein levels on a cellular level haVe been fruiTful, although the measurements generally are interpreted as relative protein levels rather than absolute molar measurements (Gazzaley et al., 1996a, 1996b). These quantitative cellular measurements are crucial if we are to obtain data at the desired level of resolution (see earlier discussion) that can be linked to individual cell classes and circuits. The analyses of putative age-related or experimentally induced shifts in glutamate receptors (GluRs) offer an excellent example of this approach, particularly in the hippocampus, where information flow through the trisynaptic circuit is highly ordered anatomically and mediated through glutamate receptors, with the NMDA receptor in particular strongly implicated in memory and age-related changes in memory (Caramanos and Shapiro, 1994; Barnes, 1994; Barnes et al., 1997a, 1997b).
Both the available data and the missing data on changes in NMDA receptor distribution with aging offer an excellent example of the importance of