A critical issue in understanding Huntington's disease (HD) pathogenesis is how the ubiquitously expressed mutant huntingtin (mhtt) with an expanded polyglutamine repeat can cause selective toxicity of striatal and cortical neurons. Two potential cellular models may contribute to such specificity: expression of mhtt in these vulnerable neurons alone may be sufficient to result in their dysfunction and/or degeneration (cell-autonomous model); or mhtt in other cell types can elicit pathological cell-cell interactions to cause the vulnerable neurons to become dysfunctional and be at risk for degeneration (cell-cell interaction model). To distinguish between these two models, we have selectively expressed a neuropathogenic fragment of mhtt-exon1 in striatal medium spiny neurons (MSNs) by crossing a conditional mouse model of HD with a striatal-specific Cre mouse line. In this striatal model of HD, we observed progressive and cell-autonomous nuclear accumulation of mhtt aggregates in MSNs. Surprisingly, unlike the mouse model expressing mhtt-exon1 in all the neurons in the brain, the striatal model lacks significant locomotor deficits and striatal neuropathology including gliosis and dark degenerating neurons. Electrophysiological findings from acutely dissociated MSNs revealed a cell-autonomous deficit in N-methyl-d-aspartate (NMDA) receptor sensitivity to Mg2+, a deficit also present in other mouse models of HD. In conclusion, this study provides the first in vivo genetic evidence that pathological cell-cell interactions are necessary for striatal pathogenesis in a conditional mouse model of HD, and suggests a ''two-hit'' hypothesis in which both cell-autonomous toxicity and pathological cell-cell interactions are critical to HD pathogenesis.