Decitabine activates type I interferon signaling to inhibit p53‐deficient myeloid malignant cells

Dear Editor, p53 is the most frequently mutated protein in cancers. We recently reported that arsenic trioxide (ATO) can efficiently restore tumour-suppressor function to structural mutant p53,1,2 spawning a series of p53-targeted ATO clinical trials registered on ClinicalTrials.gov. However, ATO, together with other p53-targeted agents, still lacks reports for statistically significant clinical efficacy at present. Targeting the molecular alterations caused by p53 mutations (for example, become addicted to a signaling), rather than directly targeting mutant p53 itself, is an alternative strategy to treat p53-mutated patients. We and others previously reported a high complete response rate of the decitabine (DAC) in treating p53mutated acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS).3–5 This observation is particularly encouraging considering the high hazard ratio of TP53 mutation for overall survival in myeloid malignancies (Figure S1A-D). However, all responding p53-mutated patients relapsed in a few months.4,5 Thus, understandingmechanisms underlying the high response rate and the inevitable relapse is critical. DAC and analogues were reported to trigger interferon (IFN) signaling in mouse embryonic fibroblasts (MEFs) and cancer cells.6–8 We confirmed the preferentially inhibition of cell growth and upregulation of IFN signaling genes byDAC inTrp53−/−MEFs (Figure S2A,B).Using isogenic myeloid malignant THP-1 cells (originally null p53) infected with wild-type (WT) p53, mutant p53 (R282W, a hotspot p53 mutation frequently detected in AML/MDS5) or vector (Figure S2C), we found that DAC also preferentially inhibited p53-deficient THP-1 lines (Figure S2D). In RNA-seq, 5 μMDAC significantly upregulated 216 and 332 genes in p53-R282W cells and p53 WT cells, respectively. The encoded proteins were highly enriched in an interaction network centering type I IFN signaling in p53-R282W cells, but not in p53 WT cells (Figure 1A,B). Consistently, IFN signaling was themost enriched gene ontology biolog-

Dear Editor, p53 is the most frequently mutated protein in cancers. We recently reported that arsenic trioxide (ATO) can efficiently restore tumour-suppressor function to structural mutant p53, 1,2 spawning a series of p53-targeted ATO clinical trials registered on ClinicalTrials.gov. However, ATO, together with other p53-targeted agents, still lacks reports for statistically significant clinical efficacy at present. Targeting the molecular alterations caused by p53 mutations (for example, become addicted to a signaling), rather than directly targeting mutant p53 itself, is an alternative strategy to treat p53-mutated patients.
We and others previously reported a high complete response rate of the decitabine (DAC) in treating p53mutated acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS). [3][4][5] This observation is particularly encouraging considering the high hazard ratio of TP53 mutation for overall survival in myeloid malignancies ( Figure S1A-D). However, all responding p53-mutated patients relapsed in a few months. 4,5 Thus, understanding mechanisms underlying the high response rate and the inevitable relapse is critical.
DAC and analogues were reported to trigger interferon (IFN) signaling in mouse embryonic fibroblasts (MEFs) and cancer cells. [6][7][8] We confirmed the preferentially inhibition of cell growth and upregulation of IFN signaling genes by DAC in Trp53 −/− MEFs ( Figure S2A,B). Using isogenic myeloid malignant THP-1 cells (originally null p53) infected with wild-type (WT) p53, mutant p53 (R282W, a hotspot p53 mutation frequently detected in AML/MDS 5 ) or vector ( Figure S2C), we found that DAC also preferentially inhibited p53-deficient THP-1 lines ( Figure S2D). In RNA-seq, 5 µM DAC significantly upregulated 216 and 332 genes in p53-R282W cells and p53 WT cells, respectively. The encoded proteins were highly enriched in an interaction network centering type I IFN signaling in p53-R282W cells, but not in p53 WT cells ( Figure 1A In the heatmap showing the relative expression levels of genes involved in type I IFN signaling, a batch of genes including MX1, ISG15, DDX58 and IFI27, but not IFN alpha protein-encoding genes (IFNA1, IFNA2, IFNA4, et al.), were efficiently upregulated following DAC treatment in p53-R282W cells ( Figure 1E). As expected, weaker upregulation was observed in p53 WT cells ( Figure 1E). Upregulation of type I IFN genes was also observed in the myeloid malignant U937 cells expressing p53-R282W upon DAC treatment ( Figure S2E). The RNA-seq results were confirmed by qRT-PCR ( Figure 1F).
The striking activation of IFN signaling by DAC in p53deficient THP-1 cells can be achieved by upregulation of viral double strand RNAs (dsRNAs) [6][7][8] or direct DNA demethylation of promoters of IFN signaling genes (Figure 2A). In qRT-PCR, we observed upregulation of the representative dsRNAs MER57B1 and ERV3 ( Figure 2B). In methylation-specific PCR (MSP), the promoter of IFNB1, but not IRF7 or IFNA1, was demethylated upon DAC treatment ( Figure 2C). Thus, both mechanisms contribute to the observed activation of IFN signaling ( Figure 2A).
We further explored two potential mechanisms underlying DAC resistance ( Figure 2D). The expanding of preexisting p53-mutated subclones in DAC relapsed patients supported the second mechanism. 5 We thus prepared DAC-resistant THP-1 cells by long-term DAC treatment ( Figures S3A,B and 2E). RNA-seq results revealed that the DAC-mediated upregulations of dsRNAs and type I IFN signaling genes were largely compromised in DACresistant cells ( Figure 2F-H). The RNA-seq results were confirmed by qRT-PCR ( Figure 2I).
We next tried to generate a mouse model of p53-deficient AML for in vivo studies. The p53-deficient mice (Trp53 −/− ) did not develop AML over their lifespan ( Figure 3A). In contrast, the mice transplanted with Trp53 −/− ; Kras G12D fetal liver cells developed lethal neoplasia on day 20 in   Figure 3B,C). Immature blast cells were clearly detected in the peripheral blood (PB) and bone marrow (BM) at day 7 upon injection of Trp53 −/− ; Kras G12D fetal liver cells ( Figure 3D). Hepatomegaly, splenomegaly, anaemia and reduced platelet counts were also observed ( Figure S4A,B). Staining of the isolated BM, PB and spleen (SP) using myeloid ( Figure   S4C), lymphoid ( Figure S4D), T-cell ( Figure S4D), erythrocyte ( Figure 3E), hematopoietic stem cell and hematopoietic progenitor cell lineage markers ( Figure S4E) suggested that an erythrocyte-abnormal AML mouse model was successfully generated ( Figure 3B, lower panel).
The established AML mice were then intraperitoneally injected with 0.2 mg kg −1 DAC or PBS for 5 consecutive Error bars represent mean ± SD (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001) days each week ( Figure 3F). Mice in the DAC-treated group survived significantly longer than the untreated group, with median survival increasing from 14.5 to 17.5 days (Figure 3F, p = 0.0009). In a parallel experiment, the spleens at day 15 in the DAC-treated group were generally smaller ( Figure 3G), and the percentages of AML cells in the spleens were significantly lower ( Figure 3H). In the BM and SP, the mRNA levels of IFN signaling genes were significantly upregulated in the DAC group ( Figure 3I), which were consistent with the finding in Figure 2I.
In our AML/MDS Sample Bank at the Shanghai Institute of Hematology, five BM and PB samples were recorded as p53-mutated at prognosis ( Figure 4A), with the samples from patients #4 and #5 receiving DAC treatment but experiencing disease progression at the time of sample collection. We confirmed these p53 mutants were functionally deficient by the classical luciferase reporter assay 9 (Figure 4B). The cryopreserved primary BM mononuclear cells (BMMC) or PBMCs were thawed and cultured for 3 days with or without DAC. The current culture system did not greatly cause PBMCs death ( Figure S5A,B). Upon DAC treatment for 3 days, the mRNA levels of the representative IFN signaling genes were consistently upregulated in the cultured primary cells from patients #1, #2 and #3, with IFI27 efficiently upregulated ( Figure 4C). By contrast, we did not observe significant upregulation of these IFN signaling genes in the samples derived from patients #4 and #5 with DAC relapse ( Figure 4C). We failed to perform additional experiments on this set of primary cells such as RNA-seq because of the limited cell amounts.
In summary, we report the activation of IFN signaling in DAC-treated p53-deficient myeloid malignant cells and the failure of DAC in activating IFN signaling in resistant cells.