Neuroendocrine tumors of lung, including typical carcinoid (TC), atypical carcinoid (AC), large cell neuroendocrine carcinoma (LCNEC), and small cell lung carcinoma (SCLC) constitute a spectrum of malignancies in which the pathologist at times has difficulty in discerning tumor subtype and aggressiveness in a reproducible fashion. Therefore, 59 primary neuroendocrine lung tumors including 10 TCs, 26 ACs, 15 LCNECs, and 8 SCLCs were selected from cases collected from 1976 to 1988 and immunostained for p53 protein. All of these tumors were also genotyped for specific point mutational damage affecting p53 (exons 5, 7, and 8; with ACs additionally sequenced for p53 exon 6); 13 tumors for K-ras-2 (exon 1); and 31 tumors for c-raf-1 (exon 15) growth-regulatory genes. Genotyping was performed on topographically selected, minute tumor samples removed from unstained formalin-fixed, paraffin-embedded tissue sections (topographic genotyping) using polymerase chain reaction and direct sequencing. The distribution of p53 immunohistochemical staining had four patterns: negative in TCs, one-half of ACs, 3 of 15 LCNECs, and 1 of 8 SCLCs; less than 10% but more than five tumor cells per 10 high power fields (focal) in a subset (7 of 26) of aggressive ACs; 10 to 49% of tumor cells (patchy) in a subset (6 of 26) of ACs with a higher grade of aggressiveness; and 50 to 100% of tumor cells (diffuse), exclusively seen in LCNECs (12 of 15) and SCLCs (7 of 8). Three patterns of immunohistochemical staining intensity of p53 protein were seen: negative, weak or mild, and moderate to marked. SCLCs and LCNECs accounted for cases of moderate to marked staining and were the only ones to have mutations in p53 exons 5, 7, or 8. No mutations were found in AC and TC, showing absent to weak staining and no staining, respectively. The difference in distribution and staining intensities between LCNEC and SCLC compared with AC and TC was statistically significant (P < 0.001). Patients having AC with patchy p53 immunostaining usually had survival limited to 3 years, whereas those having AC with focal p53 immunostaining subsequently developed metastatic or recurrence of AC disease (P < 0.05). The absence of point mutations in cases with patchy or focal immunostaining suggests increased expression of wild-type p53 tumor suppressor protein likely in response to growth deregulation in a more aggressive subtype of AC. A novel hypothesis is presented in regard to these findings. K-ras-2 and c-raf-1 gene sequence analysis showed no evidence of point mutational change in any of the tumors studied. The TC and AC categories are therefore genetically distinct from the higher grade neuroendocrine SCLC and LCNEC. Immunohistochemistry for p53 on AC lung tumors may be helpful to delineate cases at higher risk for aggressive behavior. Additionally, although LCNEC is categorized as a non-small-cell carcinoma, it is more akin genetically and immunohistochemically to SCLC.