In the present study, we analyzed human renal allografts using immunohistochemical techniques to determine the site, identity, and frequency of (a) cytotoxic and apoptotic cells, as identified by staining for GMP-17 (TIA-1), a component of cytotoxic granules; and (b) DNA fragmentation in situ, as detected by the TUNEL method. In acute cellular rejection (n = 15), GMP-17+ mononuclear cells accounted for 29% +/- 12% of the infiltrating cells in the interstitium (341 +/- 164/mm2) and were significantly more concentrated in tubulitis lesions, where they amounted to 65% +/- 14% of the mononuclear cells (96 +/- 61/mm2) (p < 0.01 versus interstitium). GMP-17+ mononuclear cells were also found in sites of endothelialitis. An estimated 80% of the GMP-17+ lymphocytes expressed CD8, and 10% to 20% expressed either CD4 or the macrophage marker CD14. The latter finding led us to analyze normal peripheral blood monocytes by flow cytometry, all of which were found to contain GMP-17. NK cells and neutrophils, which are known to express GMP-17, were detected only rarely in allografts. Specimens with cyclosporine A toxicity (n = 7) or acute tubular necrosis (n = 13) showed fewer GMP-17+ cells in the interstitium (22 +/- 46/mm2 and 62 +/- 50/mm2, respectively) and tubules (2 +/- 6/mm2 and 10 +/- 10/mm2, respectively) (all p < 0.01 versus rejection). These differences were due largely to less intense mononuclear cell infiltration. In cyclosporine A toxicity, however, the percentages of GMP-17+ mononuclear cells within tubules and the interstitium were significantly lower than in rejection (p = 0.02), whereas in acute tubular necrosis significantly lower percentages were found in the tubules (p = 0.04) but not in the interstitium. Native kidneys with end-stage diabetic nephropathy (n = 5) had very low proportions of GMP-17+ cells in interstitial infiltrates (7% +/- 6%) and in tubules (11% +/- 15%), although the infiltrates were focally intense (517 +/- 355/mm2). TUNEL+ cells were found in acute cellular rejection, predominantly in areas with intense mononuclear infiltrates and also within lesions of tubulitis and endothelialitis. Although some TUNEL+ cells were intrinsic renal cells, most appeared to be infiltrating mononuclear cells, and we were able to detect CD3 in some. In areas of intense cellular infiltration, the percentages of TUNEL+ cells (range, 0.5% to 4.2%) were comparable to those seen in the rat thymus, indicating a high level of apoptosis. Overall, in the allograft samples, the numbers of GMP-17+ cells and TUNEL+ cells were significantly correlated (r = 0.79; p < 0.01). These data provide new evidence that T cell (and possibly macrophage)-mediated cytotoxicity plays an important role in acute renal allograft rejection, particularly in the case of tubular injury, and furthermore suggest that apoptosis may be a mechanism not only for graft cell destruction, but also for elimination of activated T cells in the infiltrate.