Alterations of immune response of Non-Small Cell Lung Cancer with Azacytidine

Abstract

Innovative therapies are needed for advanced Non-Small Cell Lung Cancer (NSCLC). We have undertaken a genomics based, hypothesis driving, approach to query an emerging potential that epigenetic therapy may sensitize to immune checkpoint therapy targeting PD-L1/PD-1 interaction. NSCLC cell lines were treated with the DNA hypomethylating agent azacytidine (AZA - Vidaza) and genes and pathways altered were mapped by genome-wide expression and DNA methylation analyses. AZA-induced pathways were analyzed in The Cancer Genome Atlas (TCGA) project by mapping the derived gene signatures in hundreds of lung adeno (LUAD) and squamous cell carcinoma (LUSC) samples. AZA up-regulates genes and pathways related to both innate and adaptive immunity and genes related to immune evasion in a several NSCLC lines. DNA hypermethylation and low expression of IRF7, an interferon transcription factor, tracks with this signature particularly in LUSC. In concert with these events, AZA up-regulates PD-L1 transcripts and protein, a key ligand-mediator of immune tolerance. Analysis of TCGA samples demonstrates that a significant proportion of primary NSCLC have low expression of AZA-induced immune genes, including PD-L1. We hypothesize that epigenetic therapy combined with blockade of immune checkpoints - in particular the PD-1/PD-L1 pathway - may augment response of NSCLC by shifting the balance between immune activation and immune inhibition, particularly in a subset of NSCLC with low expression of these pathways. Our studies define a biomarker strategy for response in a recently initiated trial to examine the potential of epigenetic therapy to sensitize patients with NSCLC to PD-1 immune checkpoint blockade.

Figure 1. Azacytidine alters gene expression in NSCLC cell lines for multiple immune related pathways

(A) Top panel: Gene Set Enrichment Analysis (GSEA) for pathways up-regulated by azacytidine. Normalized enrichment scores are plotted as a heat map. Bottom panel: boxplot showing degree of demethylation in each cell line, as measured by the difference in beta values between the AZA and mock-treated cells immediately after drug withdrawal and 7 days later. (B) FACS analysis shows increased level of cell surface PD-L1 after AZA treatment by day 10 in NSCLC lines H838 and H1299. (C) to (J) AZA-mediated expression changes at day 10 in key genes from pathways outlined in (A). Y axis = Ratio of expression values (log2) of AZA -treated vs. mock-treated cells; X-axis = gene names.

Figure 2. Genetic knock out of DNA Methyltransferases mimics the effects of azacytidine mediated immune pathway up-regulation

Gene expression alterations when comparing wild-type HCT116 colon cancer cells to their isogenic DNMT1 and 3B knockout counterpart (DKO). The gene expression differences are given as the log2 ratio of expression in DKO over wild-type HCT116 (Y-axis) and the gene panels, A-H correspond to panels C-J in Fig. 1 for the NSCLC cell lines treated with AZA.

Figure 3. Identification of azacytidine up-regulated transcription factors and interferon signaling related genes, and their clustering of primary Non-Small Cell Lung Cancer in TCGA

(A) Identification of genes in Non-Small Cell Lung Cancer cell lines with low basal expression with high basal promoter region DNA methylation which are demethylated and re-expressed with AZA treatment. The red box encompasses genes meeting these criteria which are described specifically in methods. Among these, IRF7, a key immune-related transcription factor, was up-regulated in multiple cell lines. (B) Pathways up-regulated in NSCLC cell lines in response to AZA are enriched for IRF7 targets as determined by PScan analysis (-log₁₀ of p-values) and gene set enrichment analysis. (C) Heat map of RNA-Seq expression levels in primary lung cancers from TCGA database for genes 4-fold or more induced by AZA in the LUSC cell line H2170, the cell line with the greatest degree of IRF7 up-regulation. Top bar: red indicates LUAD and orange indicates LUSC samples. Genes used in the heat map are listed in supplemental table 4. (D) Bar panels show expression of PD-L1 and IRF7 in five quantile intervals (red for lower and green for higher expression). Heat map immediately below IRF7 expression bar shows corresponding Infinium platform DNA-methylation levels (Z-scores, red for more and green for less methylated) across the promoter region. Positions relative to transcription start site are shown to the right. CpG-island probes are labeled in green. Sample order in bar plots and methylation heat map is maintained from the main heat map.

Figure 4. Relationship of azacytidine-induced, immune-related pathways to primary lung tumors grouped by expression of IRF7-associated genes

TCGA samples are ordered by unsupervised clustering based on genes highly up-regulated in H2170, which are enriched for IRF7-targets, represented in the topmost heat map. Order of samples is maintained in all lower heat maps. PD-L1 and IRF7 expression are depicted in the top bar panels as in figure 3D. Supplemental Table 5 table shows the overlaps of genes from each pathway represented in the heat maps. That the observed clustering pattern is not due to chance or batch effect is demonstrated using random sets of 25 genes shown in the bottom two panels.