Blast Crisis CML Cells Require IFN-γ Conditioning For Effective GVL Whereas Chronic Phase CML Cells Do Not: An Explanation For Chronic Phase CML GVL-Sensitivity

While the GVL effect is potent against chronic phase (CP) CML, it is less effective against blast crisis (BC) CML and AML. To better understand this difference we have been studying GVL against mouse models of CP-CML (mCP-CML) and BC-CML (mBC-CML) induced by infecting bone marrow (BM) with retrovirus encoding bcr-abl alone (mCP-CML) or also with a retrovirus encoding the NUP98-HOXA9 fusion (mBC-CML). We previously reported MHCII^-/- mBC-CML and mCP-CML to be completely resistant to CD4-mediated GVL, but the leukemia stem cells (LSCs) of each expressed little if any MHCII. We hypothesized that MHCII might be upregulated in an alloimmune environment. Consistent with this, in mice transplanted with allogeneic T cells, mBC-CML and mCP-CML LSCs upregulated MHCII as well as MHCI and ICAM-1. To test the hypothesis that IFN-γ upregulated LSC MHC post BMT, we created IFN-γR^-/- mBC- and mCP-CML. IFN-γR^-/- mBC-CML LSCs did not upregulate MHCII, MHCI or ICAM-1 in vivo, whereas they were upregulated on IFN-γR^-/- mCP-CML LSCs. Consistent with this, IFN-γR^-/- mBC-CML was completely resistant to GVL (C3H.SW→B6) mediated by CD4 or CD8 cells whereas GVL against IFN-γR^-/- mCP-CML was equivalent to that against wt mCP-CML. Therefore independence from IFN-γ conditioning is a key property that at least in part explains the sensitivity of mCP-CML to GVL. In contrast, type I IFN receptor-deficient mCP-CML and mBC-CML upregulated MHC and were fully GVL-sensitive. That IFN-γR^-/- mBC-CML was resistant to CD8-mediated GVL was surprising given that mBC-CML LSCs are constitutively MHCI⁺. These data suggest that IFN-γ does more than simply upregulate MHCI. To further explore this we performed gene expression analysis. Illumina arrays revealed that wt mBC-CML and not IFN-γR^-/- LSCs from alloimmune mice have higher mRNA levels of death pathway genes Ripk1, Fas, Caspases, 1,2,8, Diablo, Apaf1 and Mdm2 and antigen presentation genes Ciita, Tap1, and proteasome subunit Psmb9, in addition to the canonical IFN-γ-induced gene IRF1. In contrast, c-myc expression was reduced. We next focused on the effects of IFN-γ on mBC-CML LSCs. MHC upregulation did not require that T cells make TCR:MHC interactions as CD8 cells induced upregulation of MHCII in MHCI^-/- mBC-CML and CD4 cells induced MHCI upregulation on MHCII^-/- mBC-CML. MHCII⁺ mBC-CML were bona fide LSCs as they transferred mBC-CML, but the resultant leukemias were MHCII^-, indicating that MHCII upregulation was dynamic. In vitro IFN-γ led to STAT1 phosphorylation within 15 minutes whereas MHCII upregulation took 24-48 hours. Upregulation could be initiated with an IFN-g pulse as short as 30’ but was maximal with a 24 hour stimulation. Donor T cells were the key source of IFN-γ as IFN-γ^-/- T cells failed to induce MHCII upregulation and did not mediate GVL, despite engraftment and expansion. Given that cognate interactions were not required for IFN-γ-induced MHCII upregulation, we tested whether GVL could be augmented by IFN-γ-producing T cells that could not directly kill LSCs. To do so, irradiated B6 mice were reconstituted with BALB/c BM, B6 MHCI^- mBC-CML without or with BALB/c T cells as follows: a) wt CD4; b) IFN-γ^-/- CD4; c) wt CD4 + wt CD8; or d) IFN-γ^-/- CD4 + wt CD8. Despite not being able to interact with MHCI^- mBC-CML, wt CD8 rescued LSC MHCII upregulation in recipients of IFN-γ^-/- CD4 cells and improved early GVL in recipients of both wt and IFN-γ^-/- CD4 cells as measured by the number of mBC-CML cells in spleen, blood and BM on day ⁺14. However, the effect was short-lived as later post transplant mBC-CML LSCs lost MHCII expression, likely due to contraction of IFN-γ-producing T cells, and mice succumbed to mBC-CML. These studies suggest that a requirement for IFN-γ stimulation is a key feature that distinguishes GVL against BC-CML and AML from GVL against CP-CML. This could explain why in the clinic a lower intensity alloimmune response is sufficient to contain mCP-CML whereas such responses are generally insufficient against BC-CML and AML. Our work also suggests that T cells that do not recognize leukemia cells can augment GVL. This hypothesis is consistent with a recent report that early CMV reactivation is associated with a lower risk for AML relapse independent from aGVHD (M.L. Green et al,Blood, 2013). These data provide a strong rationale to develop methods for safely delivering IFN-γ to AML and BC-CML cells post alloSCT.