Abstract: Lefua costata is a primary freshwater fish confining to cold areas in north of the Huai River. They are the suited model for zoogeographic studies for their high environmental dependence. Here, based on D-loop mtDNA gene, the genetic diversity and phylogeography of L. costata were analyzed on the basis of 211 specimens from 18 sampling in 9 basins in north-eastern China. 55 haplotypes were identified, high haplotype diversity (h=0.9304) and nucleotide diversity (π=0.0087) were detected. Haplotype diversity and nucleotide diversity of population in Liaodong Peninsula (LD) were the highest (h=0.8920, π=0.0074), and the farther away from the Liaodong Peninsula, the lower haplotype diversity of the population. The analysis of molecular variance (AMOVA), pairwise comparisons of genetic distance (K2P) and genetic differentiation index (F_ST) among the populations showed that the genetic differentiation between the LD population and other populations were highly significant. Phylogenetic analyses based on ML and BI methods indicated all haplotypes from the 9 basins were not clustered separated but mixed with each other according to the geographical distribution. Network analysis showed that all haplotypes were divided into three branches: LD population was clustered in branch A; Huai-Huang River population (HH) and Hai-Luan River population (HL) were clustered in branch C; branch B contained all geographic populations. Neutral test and mismatch analysis showed that overall L. costata population experienced population expansion. According to the average evolution rate (1.00%Ma) and fossil calibration point, the divergence time of the populations of L. costata was 0.0477—2.3784 Ma. The earliest divergence time between LD population and other populations was about 2.3784 Ma, and it was speculated that the Liaodong Peninsula is one of the origin center of L. costata and little affected by Quaternary glaciation, other population had been experienced multiple gene exchange through the Huangbohai Plain estuary and the Songliao Ancient Great Lake.