Inhibition of Translation Initiation and Heat Shock Factor 1 (HSF1): A Novel Strategy Targeting FLT3 Double Mutant and Non-FLT3 Mutant AML Progenitor/Stem Cells

The transcription factor Heat shock factor 1 (HSF1) is a key sensor of proteotoxic stress and plays a major role in cancer biology. In various tumors, downregulation of HSF1 inhibits growth and induces cell death, and its activation was reported to be an adverse prognostic factor in breast and lung cancers. We have reported that inhibition of translation represses DNA-binding of HSF1 in cancers; large scale drug screening identified the eIF4a antagonist rohinitib (RHT) as a potent inhibitor of the HSF1 activation state (Santagata et al, Science 2013). Review of public databases provided us with additional rationale for targeting HSF1 and eIF4a in acute myeloid leukemia (AML): 1) mRNA levels of HSPA8, a primary HSF1 target, are correlated with poor prognosis in AML; 2) eIF4a mRNA levels are highest in AML among 12 cancer types; and 3) Gene set enrichment analysis using microarray dataset of functionally-defined leukemia stem cells (LSCs) and non-LSCs (Nature, Ng et al), based on intra-patient comparisons, reveals that a gene set for translation initiation is highly enriched in LSCs (NES 2.73, FDR = 0.000, p= 0.000).

RHT induced marked apoptogenic effects in seven human AML cell lines in culture at low nanomolar concentrations (LD50s; 9.5 to 99.5 nM, IC50s; 4.7 to 8.8 nM, based on AnnexinV/PI-positivity at 72hr). RHT inhibited HSF1 transcriptional activity, as shown by 70% reduction in HSPA8 mRNA levels, and 50% reduction in a luciferase reporter driven by the HSPA1A or HSPA6 promoters. We then tested primary samples from 17 AML patients and bone marrow (BM) samples from eight healthy donors. RHT potently induced apoptosis in AML cells, while relatively sparing normal BM cells (%apoptosis: 65.4% vs 6.67%, P < 0.0001). Importantly, a similarly significant difference in sensitivity was also observed between AML and normal stem/progenitor cell populations (CD45+CD34+CD38-). We further found that the activity of RHT was significantly higher in FLT3-ITD than in FLT3-wt AMLs. Comparing isogenic versions of the cell line Ba/F3, RHT more potently killed FLT3-ITD cells than wild-type (wt) cells (LD50s; 65.3 vs 20.1 nM). Immunoblot analysis showed higher activation-associated phospho-HSF1 (Serine 326) in FLT3-ITD Ba/F3 than in FLT3-wt cells, and growth inhibition by HSF1 knockdown was FLT3-ITD-dependent in those lines, indicating greater dependence of FLT3-ITD cells on HSF1. In an AML xenograft model with GFP-luciferase labeled MOLM-13 cells, RHT significantly reduced tumor burden as measured by bioluminescence imaging and reduction in GFP+ leukemic cells peripheral blood (PB) and BM by day 16. Median survival of the treatment group was significantly longer (24 vs 18 days, p < 0.0001).

We also found that RHT was synergistic when combined with FLT3 inhibitors of Type I (sorafenib, quizartinib, midostaurin) or Type II (gilteritinib, and E-6201). Immunoblot analysis showed that oncogenic FLT3 activation is associated with increased phosphorylation of HSF1 (S326), suggesting that decreased phosphorylation of the S326 site associated with FLT3 inhibition and diminished DNA-binding by HSF1 driven by eIF4a inhibition combine to promote synergistic lethality. In support of this, HSF1 knockdown sensitized MOLM-13 and FLT3-ITD+ Ba/F3 cells to FLT3 inhibitors. Importantly, similar combinatorial synergistic effects with type II FLT3 inhibitors were observed in cells with dual mutations of FLT3-ITD and FLT3-D835Y/H. This observation has significant clinical relevance because the development of secondary point mutations is a major resistance mechanism against FLT3 inhibitors.

Finally, to test the hypothesis that eIF4a inhibition impairs LSC functions, we treated primary patient-derived xenograft AML PDX cells with RHT ex vivo, then injected identical numbers of live cells into recipient NSG mice. Subsequent leukemia engraftment was delayed by one week vs. controls, and human CD45+ cells in PB and BM at 4 weeks post transplantation were significantly reduced (n = 9, 3.30% vs 0.26% in PB, P < 0.0001). This reduced engraftment capacity of RHT-treated PDX AML-LSCs indicates a toxic effect of eIF4a inhibition on AML-LSCs.

Collectively, these preclinical findings strongly support the further development of eIF4a inhibitors for the treatment of AML, especially for high-risk FLT3-mutated AML, with additional potential as LSC-targeted therapy.