In order to investigate the histogenesis and facilitate the diagnosis of desmoplastic small round cell tumor (DSRCT), 39 cases were studied by immunohistochemical methods using a large battery of antibodies directed against a wide variety of epithelial, mesenchymal, and neural-associated proteins. Sixteen of these tumors were also studied by electron microscopy. Thirty-seven of 39 cases reacted for cytokeratin using a "cocktail" of 3 monoclonal antibodies (CAM 5.2/AE1/AE3), 39/39 for desmin, 24/25 for epithelial membrane antigen, 22/27 for vimentin, 18/25 for neuron-specific enolase, 10/15 for CD57 (Leu-7), 3/19 for synaptophysin, 1/22 for chromogranin, 3/19 for muscle-specific actin, 3/16 for α-smooth-muscle actin, 11/16 for CD15 (Leu-M1), 5/12 for CA-125, 6/17 for CD99, 9/10 for MOC-31, 2/6 for NB84, 5/7 for Ber-EP4, and 8/9 for the Wilms tumor (WT1) protein. No staining was obtained in any of the cases tested for cytokeratin 5/6 or 20, neurofilament proteins, glial fibrillary acidic protein, peripherin, CA19-9, thrombomodulin, alphafetoprotein, carcinoembryonic antigen, TAG-72 (B72.3), placental alkaline phosphatase, S-100 protein, HMB-45, myoglobin, or for the two myogenic regulatory proteins myogenin and MyoD1. A frequent ultrastructural finding was the presence of juxtanuclear aggregates of intermediate filaments, but microfilaments with densities or Z-band-like material suggestive of either smooth or skeletal muscle differentiation were not seen in any case. Dendritic-like processes containing microtubules and dense core granules were seen in four tumors and all of these tumors reacted for at least one of the neural markers investigated. Although ultrastructural and immunohistochemical studies confirmed previous observations that DSRCTs present epithelial, mesenchymal, and neural phenotypes, a great variation was found in the frequency of expression of the different markers used to demonstrate each line of cell differentiation. The absence of expression of cytokeratin 5/6 and thrombomodulin together with positive staining for CD15, MOC-31, and Ber-EP4 argues against the possible mesothelial origin that has been suggested for this tumor. Additionally since none of the tumors reacted for myogenin or MyoD1, desmin expression in DSRCT cannot be regarded as evidence of skeletal muscle differentiation. Although the histogenesis of DSRCT remains unknown, it is believed that this tumor originates from a progenitor cell with potential for multiphenotypic differentiation.