Prostate cancer that has progressed to metastatic disease remains largely untreatable. Nearly 90% of patients with advanced prostate cancer develop skeletal metastases, resulting in a substantial reduction in the quality of life and a drastic worsening of patient prognosis. The mechanisms involved in prostate cancer cell dissemination, however, remain poorly understood. We previously reported the identification of a highly tumorigenic E-cadherin positive prostate tumor stem cell subpopulation that expressed the embryonic stem cell markers SOX2 and OCT3/4. We herein demonstrate that this subpopulation is also highly invasive and, importantly, is capable of altering its E-cadherin expression during the process of invasion. The non-tumorigenic E-cadherin negative subpopulation which minimally expresses SOX2 or OCT3/4 was found to be poorly invasive. In addition, targeted knockdown of SOX2 or OCT3/4 markedly suppressed the invasion of prostate cancer cells. Taken together, these findings indicate that the expression of SOX2 or OCT3/4 is required for invasive cell capacity, but the ability to modulate E-cadherin is the key permissive factor enabling cancer stem cell invasion in vitro. We therefore propose a model in which the post-epithelial to mesenchymal transition phenotype progresses to a frank, aggressive, and invasive phenotype by a process requiring the acquisition of E-cadherin plasticity. Considering the clinical significance of the metastatic complications of prostate adenocarcinoma, the identification of factors that promote the dissemination of the malignant prostate phenotype is essential to establish effective therapies to combat this disease in future.