To investigate RNA-dependent RNA polymerase (RdRp) further, mutational analysis of the N-terminal domain of the NS5B protein of Classical swine fever virus was performed. Results show that the N-terminal domain (positions 1-300) of the protein might be divided artificially into four different regions, N1-N4. The N1 region (positions 1-61) contained neither conserved lysine nor conserved arginine residues. NS5B protein with deletion of the N1 region has the capacity for elongative RNA synthesis, but not for de novo RNA synthesis on natural templates. All substitutions of the conserved lysines and arginines in the N2 region (positions 63-216) destroyed RdRp activity completely. Substitutions of the conserved arginines in the N3 region (positions 217-280) seriously reduced RdRp activity. However, all substitutions of the conserved lysines in this region enhanced RNA synthesis and made the mutants synthesize RNA on any template. Substitutions of the conserved arginines in the N4 region (positions 281-300) reduced elongative synthesis and destroyed de novo RNA synthesis. In contrast, substitutions of lysines in this region did not affect RdRp activity significantly. These data indicate that the N3 region might be related to the enzymic specificity for templates, and the conserved lysines and arginines in different regions have different effects on RdRp activity. In combination with the published crystal structure of bovine viral diarrhea virus NS5B, these results define the important role of the N-terminal domain of NS5B for template recognition and de novo RNA synthesis.