Although some studies have shown that the cell penetrating peptide (CPP) TAT can enter a variety of cell lines with high efficiency, others have observed little or no transduction in vivo or in vitro under conditions mimicking the in vivo environment. The mechanisms underlying TAT-mediated transduction have been investigated in cell lines, but not in primary brain cells. In this study we demonstrate that transduction of a green fluorescent protein (GFP)-TAT fusion protein is dependent on glycosaminoglycan (GAG) expression in both the PC12 cell line and primary astrocytes. GFP-TAT transduced PC12 cells and did so with even higher efficiency following NGF differentiation. In cultures of primary brain cells, TAT significantly enhanced GFP delivery into astrocytes grown under different conditions: (1) monocultures grown in serum-containing medium; (2) monocultures grown in serum-free medium; (3) cocultures with neurons in serum-free medium. The efficiency of GFP-TAT transduction was significantly higher in the monocultures than in the cocultures. The GFP-TAT construct did not significantly enter neurons. Experimental modulation of GAG content correlated with alterations in TAT transduction in PC12 cells and astrocyte monocultures grown in the presence of serum. In addition, this correlation was predictive of TAT-mediated transduction in astrocyte monocultures grown in serum free medium and in coculture. We conclude that culture conditions affect cellular GAG expression, which in turn dictates TAT-mediated transduction efficiency, extending previous results from cell lines to primary cells. These results highlight the cell-type and phenotype-dependence of TAT-mediated transduction, and underscore the necessity of controlling the phenotype of the target cell in future protein engineering efforts aimed at creating more efficacious CPPs. Biotechnol. Bioeng. 2009; 104: 10–19 © 2009 Wiley Periodicals, Inc.