ABSTRACT: Since macrophage activation can now be studied at a global level using modern microarray and proteomic analyses, discovery of novel macrophage activation genes is inevitable and important for understanding HIV-associated dementia (HAD). We isolated two different types of primary human macrophages, microglia and monocyte-derived macrophages (MDM) from brain tissue and whole blood respectively. The microarray analysis of differentially regulated macrophage activation genes reported here supports our previous assertions that the mixed glia (MIX) cultured in starvation conditions (DMEM alone) are a non-activated, or quiescent, tissue culture model for studying macrophage activation in the brain. Transcript levels from these quiescent cultures provided a background level of gene expression, and allowed for the identification of upregulated macrophage activation genes in the MIX brain cultures upon treatment with an array of soluble activation factors: serum components, cytokines, and growth factors. We found that 914 genes in the MIX cultures and 734 genes in the MDM cultures had a greater than 2-fold increase in expression. We discovered 180 genes that were increased greater than 2 fold in both culture types. Microarray-specific statistical analyses were performed to complement fold change analysis: Significance Analysis of Microarrays (SAM), and Partek Pro. In the MIX cultures, we detected over a hundred fold increase in IL-1? and TIMP1 transcription; Caspase 9, S100A8 and 9, MMP-12, IL-8, MCP-1, MRC-1, and IL-6 were also upregulated. Activation of starved MDM cultures resulted in fewer upregulated genes compared to MIX cultures. Genes upregulated in both MIX and MDM included CCL2 (MCP1), CCL7, CXCL5, TNFSF14, kinases, and phosphatases. These microarray data may provide leads for identifying previously unknown neurotoxins, disease biomarkers, and pathways responsible for the neuronal apoptosis observed in HAD, and for the eventual identification of therapeutic targets and treatments. For each gene, we performed a per-gene normalization, whereby an expression ratio was calculated by dividing the raw data from the activated media samples by the raw signal for the same gene from the non-activated media samples (Fig. 1). The same microarray data were also analyzed again using the very common per-chip normalization method recommended by the manufacturer where raw signals for each chip were multiplied by a scaling factor such that the trimmed mean for each chip became an arbitrary number (150) (Affymetrix, 2001). The per-chip normalization method reduces person-to-person biological variability and potentially hides important changes in gene expression (Hill et al., 2001),(Bolstad et al., 2003). Therefore, we report our data using a per-gene normalization method, which yielded higher signal and low noise (uninteresting non-biological variability caused by chips, and/or experimental sample preparation), and given our experimental design, reduced person-to-person variability that these experiments cannot address directly. For later statistical analyses, the data were log transformed in order to obtain a normal distribution of the expression values of the >12,625 genes (Bolstad et al., 2003)
