RNA4 segment of a severe isolate of Rice stripe virus, isolated from Liaoning province and designated as PJ isolate, was amplified, cloned and sequenced. RNA4 of PJ isolate had 2157 nucleotide in length. When compared with RNA4 of T and M isolates of Japan and CX isolate of Yunnan province of China that had been previously reported, we found that these four isolates could be divided into two groups. PJ, T and M isolates shared 97.0% and 97.0%-97.5% identities in vORF4 and vcOFR4 at the nucleotide level, respectively and formed one group. The sequences in 5' and 3' terminal non-encoding region were completely identical among these three isolates. In this group, PJ isolate was more closely related to T isolate than to M isolate. The length of intergenic region(IR4) of PJ isolate was as same as that of T isolate, and had 93.0% sequence identity. However, PJ IR4 had an insertion of 19 bp in length compared with isolate M and had only 85.0% sequence identity. CX isolate belonged to another group, which shared only 94.0% and 92.5%-93.5% sequences identities in vORF4 and vcORF4 at the nucleotide level, respectively, even though there were not significant difference between these two group at the amino acid level. There was an insertion of 84 bp in length in the IR4 of CX isolate compared with PJ isolate and the sequence identity between two group reached to 72.0%-75.0%. Even though no base variation occurred in 5' terminal non-coding region, there was two bases substitution in 3' terminal non-coding region. These results showed that the isolates were grouped according to their geographical location. Additionally, highly consensus in 5' and 3' non-encoding region suggested that these regions played an very important role in transcription and replication of viral genome. Finally, the molecular epidemiology and gene functions of Rice stripe virus were discussed in this paper.