Abstract 3730: A quick and cost effective 12-cell line panel assay to predict drug activity in human tumor xenograft models

The procedure to identify and develop an anti-cancer drug first involves testing drug candidates in cell lines followed by human tumor xenograft models, usually selected based upon the histotype of the cell lines in which the drug showed optimal activity. Many drugs fail at this stage, as activity in cell lines does not often correlate with activity in xenograft models. This is not surprising, as we have previously shown that gene expression in xenograft models does not necessarily correlate with the cell line from which it was derived. In an attempt to improve the success rate of drugs tested in xenograft models, we have developed a fast and cost effective 12-panel human tumor cell line assay that represents the genetic diversity of all our xenograft models and several different cancer histotypes. Affymetrix genomic analysis was performed on 100 human tumor xenograft and cell line models. The genomic profiles obtained underwent Unsupervised Hierarchical Cluster Analysis to group models with similar genetic profiles. This analysis resulted in 12 distinct clusters; a representative cell line was chosen from each cluster. Stocks of each representative cell line were frozen and tested to ensure exponential growth immediately upon thawing, resulting in no waiting time for drug testing. It follows that if a candidate drug shows activity in one or more of these representative cell lines, other cell lines and/or xenograft models in the same cluster can also be tested. As the cell lines and xenograft models within the same cluster will have a similar genetic profile, the chances of success should thus be increased. To test the effectiveness of this approach, we used our database to further develop an internal compound. SRI-20900 had been tested previously in the CCRF-CEM and CAKI-1 xenograft models. The compound showed no activity in CCRF-CEM cells, but excellent activity in CAKI-1 cells. These models were in completely different clusters. So, based on these data, we tested the compound in the SKOV-3 and IGROV-1 xenograft models, as these clustered closely to the CAKI-1 model. The compound showed excellent activity in both SKOV-3 and IGROV-1 models. Although these data provide proof of principle, further work needs to be done by testing targeted compounds in the 12-cell line panel, followed by testing in xenograft models within the same cluster as the cell lines that show optimal activity. In addition, it would follow that a xenograft model within the same cluster