5-aza-2′,2′-difluorodeoxycytidine (NUC013) has been shown to be significantly safer and more effective than decitabine in xenograft models of human leukemia and colon cancer. However, it suffers from a similar short half-life as other DNA methyltransferase inhibitors with a 5-azacytosine base, which is problematic for nucleosides that primarily target tumor cells in S phase. Because of the relative instability of 5-azanucleosides, a prodrug approach was developed to improve the pharmacology of NUC013. NUC013 was conjugated with trimethylsilanol (TMS) at the 3′ and 5′ position of the sugar, rendering the molecule hydrophobic and producing 3′,5′-di-trimethylsilyl-2′,2′-difluoro-5-azadeoxycytidine (NUC041). NUC041 was designed to be formulated in a hydrophobic vehicle, protecting it from deamination and hydrolysis. In contact with blood, the TMS moieties are readily hydrolyzed to release NUC013. The half-life of NUC013 administered intravenously in mice is 20.1 min, while that of NUC013 derived from intramuscular NUC041 formulated in a pegylated-phospholipid depot is 3.4 h. In a NCI-H460 xenograft of non-small cell lung cancer, NUC013 was shown to significantly inhibit tumor growth and improve survival. Treatment with NUC041 also led to significant tumor growth inhibition. However, NUC041-treated mice had significantly more tumors ulcerate than either NUC013 treated mice or saline control mice, and such ulceration occurred at significantly lower tumor volumes. In these nude mice, tumor regression was likely mediated by the derepression of the tumor suppressor gene p53 and resultant activation of natural killer (NK) cells.