Lack of an appropriate animal model for dengue virus (DEN), which causes dengue fever and dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS), has impeded characterization of the mechanisms underlying the disease pathogenesis. The cardinal feature of DHF/DSS, the severe form of DEN infection, is increased vascular permeability. To develop a murine model that is more relevant to DHF/DSS, a novel DEN strain, D2S10, was generated by alternately passaging a non-mouse-adapted DEN strain between mosquito cells and mice, thereby mimicking the natural transmission cycle of the virus between mosquitoes and humans. After infection with D2S10, mice lacking interferon receptors died early without manifesting signs of paralysis, carried infectious virus in both non-neuronal and neuronal tissues, and exhibited signs of increased vascular permeability. In contrast, mice infected with the parental DEN strain developed paralysis at late times after infection, contained detectable levels of virus only in the central nervous system, and displayed normal vascular permeability. In the mice infected with D2S10, but not the parental DEN strain, significant levels of serum tumor necrosis factor alpha (TNF-alpha) were produced, and the neutralization of TNF-alpha activity prevented early death of D2S10-infected mice. Sequence analysis comparing D2S10 to its parental strain implicated a conserved region of amino acid residues in the envelope protein as a possible source for the D2S10 phenotype. These results demonstrate that D2S10 causes a more relevant disease in mice and that TNF-alpha may be one of several key mediators of severe DEN-induced disease in mice. This report represents a significant advance in animal models for severe DEN disease, and it begins to provide mechanistic insights into DEN-induced disease in vivo.