Interferon-ε loss is elusive 9p21 link to immune-cold tumors, resistant to immune-checkpoint therapy and endogenous CXCL9/10 induction

Introduction: Copy-number (CN) loss of chromosome 9p, or parts thereof, impair immune response and confer ICT resistance by direct elimination of immune-regulatory genes on this arm, notably IFNγ genes at 9p24.1, and type-I interferon (IFN-I) genes at 9p21.3. We recently found that the primary 9p-loss human-tumor immune readout, however, is indirect (CXCL9/10 depletion at 4q21.1), and in mice, uncovered little-studied IFN-I interferon-ε (IFNϵ) deletion as the pivotal 9p21.3 link to TME immune-cell suppression. The central role of CXCL9 and/or CXCL10 in TME, has generated intense interest in cellular sources and regulation of these chemokines. We developed a focal gene-deletion strategy, termed MACHETE, to study the contribution of individual IFN-I genes to TME immune-cell populations in murine models. In this report, MACHETE-engineered deletions of Cdkn2a/b alone, MTAP vs Cdkn2a/b with progressively increasing numbers of IFN-I genes, ΔS and ΔL, at mouse chr4C4 syntenic to human chr9p21.3, were used to assess IFN-I contribution of to cxcl9 and cxcl10 expression levels.

Methods: This research perspective updates and explicates the rapidly emerging body of clinical 9p CN alteration (CNA)/ICT data (13 reports, 36 cohorts, 3.5 years), and executes clinical and experimental 9p, IFNϵ and CXCL9/10 studies of this novel genomic ICT-resistance mechanism. We analyzed 9p, 9p21.3 and 9p24.1 influence on IFN-I gene-expression and using CIBERSORT, Kassandra, MCPcounter, xCell immune-cell deconvolution probed CD8 T-cells, dendritic cells (DCs), macrophages, neutrophils; subtypes, molecules, and sub-cluster mechanisms in HPV^- HNSC (TCGA, n=343; CPTAC (n=105) and 32 cell lines, and pancreatic ductal adenocarcinoma (PDAC) (177 TCGA, 44 lines). We also include pan (34)-tumor analysis, focused on 4 highly aneuploid tumors-HPV^- HNSC, NSCLC (non-squamous [NS) and squamous [LUSC]), and PDAC-and mouse-model PDAC and NS NSCLC studies. To identify CXCL9/10-CXCR3 axis sources and regulation, we analyzed 9p21.3, IFN-I deletion size and depth in human tumors and cell lines, and scRNA-sequencing of mouse models, for cell type, subtype and subcluster expression of CXCL9 and CXCL10. The latter metrics included numbers of Cxcl9/10⁺ immune cells, percentages of Cxcl9/10 -expressing cells, per-cell expression levels of each CXCL gene, and total cell Cxcl9 and cxcl10⁺ fractions..

Results: In HPV^- HNSC, IFNϵ was the most highly expressed IFN-I gene in the TME, the only IFN-I gene detectable in cell lines; suppressed (with IFNA1, IFNA13 and IFNK) in 9p21.3 (but not 9p24.1) loss tumors, adjusting for SCNA level, and in mediation analysis, IFNϵ loss was a statistically significant direct 9p21 link to effector-cell suppression (of CD8, T-cells, myeloid DCs and neutrophils), and was profoundly tissue specific. GSEA-pathway analysis of IFNϵ identified NFκB and inflammatory response as the top two IFNϵ-loss depleted pathways in TCGA and cell lines. 9p21.3 shallow and deep (and ΔS and ΔL) deletions were associated with progressive CXCL9/10-CXCR3 axis suppression in multiple multivariable models. Ifnϵ was the primary cell-intrinsic IFN-I signal to Cxcl9/10 in PDAC, confirmed in KPL-3M lung scRNA-seq data. In support of a causal link between IFNϵ and TME, CD8 T-cells and myeloid DCs, and CXCL9/10, this analysis revealed higher numbers of Cxcl9/10⁺ DCs, macrophages and neutrophils in IFN-intact ΔS (vs ΔL). We found higher percentages of CXCL9- and CXCL10-expressing DCs, macrophages and neutrophils in IFN-I WT ΔS (vs ΔL) tumors, and Cxcl9/10⁺ per-cell expression levels in macrophages (P=6.2e^-4 for CXCL9: P=3.3e^-6 for CXCL10). M1 was the main macrophages subtype driving the difference in CXCL9 between ΔS and ΔL tumors (P=2.1e^-3), and CCL5+ in CXCL10 (P=0.018). DC subclustering revealed that both cDC1 and cDC2 produced CXCL9, while only cDC2 produced CXCL10, and the difference between ΔS and ΔL was mainly driven by cDC2. Although no difference was observed in overall per-cell expression level of each CXCL gene in ΔS vs ΔL tumors, total DC CXCL9+ and CXCL10+ fractions were higher in ΔL.

Conclusion: We identify IFNϵ loss as the elusive 9p21 link to human immune-cold, CXCL9/10-CXCR3 axis-depleted tumors. Extending mouse-model studies of IFN-I on TME immune-cell levels, we found that IFNϵ loss is the primary cell-intrinsic 9p21 immune signal to DC and macrophage subtype and subcluster expression of CXCL9 and CXCL10, the major sources of these chemokines. Larger deletions to 9p24 further restrict CXCL9/10 induction via loss of IFN-γ-pathway gene, JAK2. 9p-loss tumors with these distinct IFN defects operative in the TME, lack the capacity of endogenous CXCL9/10 induction in an immune-desert, ICT-resistant state. These data, the extensive 9p loss/ICT resistance body of evidence, and early NSCLC DC-chemokine vaccine trials, have led to a DC vaccine engineered with a CXCL9/10 payload, designed to bypass the specific, severe chemokine deficit in 9p loss tumors.