Reply to Bonnefoix and Callanan: Significance of Akt/Wnt pathway in radiation resistance of TICs

Abstract

Whereas we appreciate the statistical concerns expressed about the use of the single-hit Poisson (SHPM) model to evaluate the data shown in tables 1–3 of ref. 1) and summarized in Fig. 1 here, it is unfortunate that these authors (2) did not bother to read or evaluate the previous studies upon which the identification of the LinnegCD29high/CD24high cells as tumor initiating cells (TICs) in the p53 model were based (3), nor did they take into consideration accompanying molecular and biochemical analyses shown in figs. 1–4 and the supplementary data (1). Furthermore, they did not evaluate the conclusions of our study in the context of other recent publications cited concerning targeting the Akt/Wnt pathway in cancer stem cells in both breast cancer (4) and gliomas (5). In the previous study by Zhang et al. (3), a much larger dataset involving 10 tumors and 426 limiting dilution transplantations showed very similar relative effect sizes to data in ref. 1 and was, perhaps due to the larger sample size, consistent with SHPM. As clearly discussed in ref. 1, the data shown in table 1 represent a much smaller number of tumors and transplants and are included to illustrate how the calculation of absolute TIC frequency can be influenced by cell isolation methods and FACS sorting methodology. However, the fold difference between the Linneg and TIC populations is still ≈65-fold. The biochemical and molecular analyses shown in figs. 1–4 of ref. 1 also clearly illustrate that the statistically significant decrease in PTEN expression in TICs is accompanied by an increased activity of the Akt and Wnt pathways. The use of a Wnt pathway reporter shows a highly significant overlap with TICs identified by FACS sorting using the cell surface markers. Furthermore, treatment with perifosine inhibits the selective activation of the Akt and Wnt pathways in the TIC population, whether assayed by FACS using the cell surface markers or the Wnt pathway reporter. Again a smaller number of tumors are used in the more labor-intensive and costly limiting dilution analyses of the Wnt(TOP-GFP) and perifosine-treated samples, giving low power to test that slopes are nonzero, although failure to reject does not mean the slope is zero. Bonnefoix and Callanan rightly point out that the data in the perifosine experiment are so sparse that the models may not converge. Despite this, the differences between the cell types are so large that Fisher's exact tests at individual doses are significant (Fig. 1). Most importantly, the results of our studies identify significant differences in the response to DNA damage in the TIC population as compared with the other tumor bulk subpopulations and show this difference was abrogated by perifosine treatment both in mammosphere cultures and following in vivo treatment. Thus, although we think it is valuable to emphasize the need for the application of appropriate statistical tests to specific datasets, especially because the SHPM is most often used for the evaluation of TIC frequency, Bonnefoix and Callanan (2) are remiss in dismissing the conclusions of our studies without having properly evaluated the supporting literature or accompanying data.