Preferential In Vitro and In Vivo Induction of Endogenous Retrovirus Transcription from a Monoallelic Chromosome 7q in Decitabine (DAC) Treated AML Blasts

Background: Hypomethylating agents (HMAs) show an encouraging but not yet well-understood activity in AML/MDS patients (pts) with adverse cytogenetics, such as -7/7q-, often embedded in a complex, monosomal karyotype (Lübbert et al., Haematologica 2012). Integrative methylome/transcriptome studies have provided evidence for aberrant hypermethylation/silencing on monoallelic gene loci, including tumor suppressor genes (TSGs, Kotini et al., Nat. Biotechnol. 2015). Recently, an alternative mechanism of HMA action was described (Roulois et al., Chiappinelli et. al, Cell 2015): induction of silenced endogenous retroviral (ERV) dsRNA transcription, resulting in activation of an interferon-mediated, antitumor immune response ("viral mimicry"). To unravel in vivo HMA activity in different (cyto)genetic backgrounds, we hypothesized that in AML haploinsufficient for 7q, transcriptional repression of monoallelic genes in this region may be preferentially reversed by HMA treatment.

Materials and Methods: For identification of genes induced by HMAs, two AML cell lines were selected, ELF-153 (ELF) and UCSD-AML1 (AML1): copy number variations were analyzed by Genome-Wide Human SNP Array 6.0 (Affymetrix), and revealed no monosomies but loss of 7p; in AML1 loss of one entire chromosome 7 was the only numerical aberration. Cells were treated with DAC at equitoxic and -effective concentrations for 4 days, harvested on day 5. Total RNA was depleted of ribosomal RNA (Ribo-Zero rRNA Removal Kit, Illumina) and cDNA libraries (strand-specific, 50 bp) were sequenced with ≥33 million paired-end reads, read quality checked with FastQC (≥95% >Q30). Alignment to the reference genome hg19 was performed with TopHat2, read-counting with HTseq, differential expression testing with DESeq2 (protein-coding transcripts only). Selected target genes were validated by RT-qPCR. Peripheral blood mononuclear cells (PBMC) were collected from 9 AML pts (5 pts with 7q-, 4 cytogenetically normal [CN] pts) treated with DAC (i.v. 20 mg/m² for 5 days) within the DECIDER trial (NCT00867672). Leukemic blasts before treatment and at day 8 were isolated using automatic magnetic sorting of cells labeled with anti-human CD34 and CD117 MACS microbeads.

Results: DAC treatment of AML1 and ELF cells resulted in massive transcriptome changes in both cell lines, with 1,704 genes induced in AML1 (total of 3,583) and 1,206 in ELF (total of 1,867; adjusted FDR <0.01). Comparing DAC-induced expression changes only of genes on 7q, we identified 43 genes significantly differentially expressed only in AML1 cells (up: 15), 24 only in ELF (up: 21), and 19 in both (up: 12). The 15 genes on 7q that were selectively upregulated in AML1 included 3 with TSG features: zyxin (ZYX; up 6.0-fold), homeodomain interacting protein kinase 2 (HIPK2; up 3.8-fold), and high mobility group box transcription factor 1 (HBP1, up 3.1-fold). However, the most heavily upregulated transcripts in AML1 were the endogenous retrovirus 3 group member 1 (ERV3-1 ; up 86.4-fold in AML1 vs 14.3-fold in ELF) and the neighboring zinc finger 117 (ZNF117; up 119.4-fold in AML1vs. 6.5-fold in ELF), both transcribed from the ERV3-1 promoter. ERV3-1 induction was accompanied by induction of the interferon type I response genes retinoic acid inducible gene I (RIG-I) and interferon regulatory factor 7 (IRF7) in both cell lines. Therefore, we next interrogated induction of these genes by DAC in vivo, utilizing serially sorted, matched primary AML blasts (purified before and 3 days after DAC administration). Stronger induction of ERV3-1 and RIG-I mRNA was seen in the 7q- pts compared to the CN pts: By qRT-PCR, median ERV3-1 and RIG-I expression was induced 1.5- and 1.4-fold, respectively, in the pts with 7q- (range: 1.1-2.1 and 1.0-3.2), 1.1- and 0.7-fold, respectively, in the CN pts (range: 0.8-1.7 and 0.3-0.9).

Conclusions: We successfully developed an unbiased RNA-seq approach of AML cell lines either mono- or bi-allelic for 7q, demonstrating that DAC treatment preferential upregulates several monoallelic TSGs, and massively activates the ERV3-1 gene (implicated in "viral mimicry"). Induction of ERV3-1 and the dsRNA sensor RIG-I was validated in vivo, and preferentially seen in 7q- AML purified primary blasts. Thus, under clinically established treatment conditions, both the ERV and an interferon response gene can be activated by DAC, supporting a combination with immunotherapy.