In vivo selection strategies that convey a survival advantage to genetically modified cells carrying mutant forms of MGMT (P140K) have the potential to improve autologous and allogeneic stem cell gene therapy and transplantation. Previously, we have shown efficient in vivo selection and marrow protection in a clinically relevant canine model. Here we describe retrovirus integration site (RIS) mapping in the canine genome after chemotherapy with O6BG and BCNU or temozolomide (TMZ). This study includes 4 dogs that received chemotherapy (2 allogeneic and 2 autologous transplants) and an untreated control dog. Animals received CD34-enriched marrow cells transduced with an RD114-pseudotype retroviral vector expressing a bicistronic message containing P140K and GFP while the control dog retroviral vector encoded only GFP. After stable engraftment 4 dogs were treated with O6BG and either BCNU or TMZ. Initial granulocyte marking (10–16%) was increased to >98% with dose-escalating chemotherapy and stabilized at 66–97% with stable increases in all cell lineages analyzed. For LAM-PCR DNA was extracted from peripheral blood leukocytes after various rounds of chemotherapy. The DNA samples analyzed for the control animal corresponded chronologically with that of the chemotherapy treatment animals. Criteria for a ‘true’ RIS are a BLAT (http://genome.ucsc.edu/) score of 35 or greater, >90% identity to the canine genome (actual identity >99%) and the second ranked score <98% of the first score. To map the RIS relative to genes in the canine genome we utilized alignment of human RefSeq genes to the canine genome (UCSC BLAT: xenoRefFlat). It has been previously described that gammaretrovirus integrate preferentially around promoter and enhancer regions and that is recapitulated in our data. Analysis of a quadrant <5kb around transcription start sites contains 13.3% and 25.5% of the total RIS sites analyzed for the chemotherapy and control animals respectively and when the quadrant is increased to (+/−)50kb that respectively comprises 60.0% and 75.0% of the RIS. We cataloged oncogenes <50kb from RIS to determine if chemotherapy has selected for potentially pre-malignant clones. Interestingly, this does not seem to be the case and in fact the only oncogene identified <50kb from a RIS in a chemotherapy animal (LASP1) was only isolated from a small subset of samples before chemotherapy was administered. In the control animal we identified three oncogenes (TAL1, LYL1 and TFG) within 50kb of a RIS. This suggests that chemotherapy has not selected for clones in or near oncogenes beyond that which already occur for gammaretroviruses. Initial RIS analysis of CFUs has shown that the average copy number is 1.52 and 1.46 for chemotherapy and control animals respectively, suggesting multiple integration events are not required for chemo-protection. We are in the process of carrying out a more thorough analysis of RIS before and after chemotherapy and will track the contribution to hematopoiesis of specific clones using real time-PCR. Our studies suggest that after aggressive in vivo selection of chemo-protected cells multiple clones contribute to hematopoiesis and no side effects or malignancies were observed. These studies and future studies in the nonmyeloablative setting are required to assess the long-term safety of chemo-protective gene therapy.