To clarify the carcinogenic factors--whether it is the kind of carcinogen or their length of exposure--that determine whether colorectal cancer develops from an adenoma or develops de novo in the absence of an adenoma, we histopathologically analyzed a total of 229 rat colon tumors induced by administration of 1,2-dimethyl-hydrazine (DMH) or N-methyl-N'-nitro-N-nitrosoguanidine (MNNG) for three or 15 weeks. In the three-week-exposure groups, 71% of DMH-induced carcinomas and 82% of MNNG-induced carcinomas coexisted with low-grade dysplasia (adenomatous remnant). However, in the 15-week-exposure groups, lowgrade dysplasia was observed in only 10% of DMH-induced and 27% of MNNG-induced carcinomas. Even in the tumors smaller than 20 mm3, it was observed in only 10% of DMH-induced and 32% of MNNG-induced carcinomas. Furthermore, carcinomas without low-grade dysplasia predominated from the initial period of tumor occurrence. Next, we investigated association of K-ras and APC gene mutations with these carcinogenesis patterns in 80 tumors. K-ras mutations were not detected in any tumors induced by three weeks of exposure. However, in the 15-week-exposure groups, this mutation was observed in 57% of DMH-induced tumors and 13% of MNNG-induced tumors. APC mutations in the region homologous to the human mutation cluster region were observed in only 6% of tumors. Thus, our results suggest that the carcinogenesis patterns in rat colon are dependent on the length of exposure to carcinogen and that K-ras mutations were partly involved in a subset of them.