The mechanism of translational initiation by the 5' noncoding region (5'NCR) of hepatitis C virus (HCV) genome was analyzed. Using an in vitro translation system with artificial RNA containing a modified 5' NCR of HCV under the various KCl conditions, nucleotides (nt.) 62 to 341 of the HCV 5'NCR were not functional as an internal ribosome entry site (IRES). However, the full-length 5'NCR (nt. 1 to 341) produced an efficient internal initiation. To identify the essential region of the HCV-IRES, various mutants were produced in which stem-loops, predicted by secondary structure analysis of the HCV 5'NCR, were deleted. These constructs were analyzed by in vitro translation. Comparison of translation efficiency among these mutants suggested that the alpha- or both alpha- and beta-branches of domain II are essential for efficient translation. Moreover, the formation of correct secondary structure of IRES seems to be stabilized by the presence of domain I in 5'NCR. Furthermore, the uncapped 5'NCR of HCV promotes translation more efficiently than capped truncated 5'NCR constructs. Our results strongly suggested that complete 5'NCR containing all stem-loop structures is necessary for initiation by HCV-IRES.