AXL targeting restores PD-1 blockade sensitivity of STK11/LKB1 mutant NSCLC through expansion of TCF1+ CD8 T cells

Abstract

Mutations in STK11/LKB1 in non-small cell lung cancer (NSCLC) are associated with poor patient responses to immune checkpoint blockade (ICB), and introduction of a Stk11/Lkb1 (L) mutation into murine lung adenocarcinomas driven by mutant Kras and Trp53 loss (KP) resulted in an ICB refractory syngeneic KPL tumor. Mechanistically this occurred because KPL mutant NSCLCs lacked TCF1-expressing CD8 T cells, a phenotype recapitulated in human STK11/LKB1 mutant NSCLCs. Systemic inhibition of Axl results in increased type I interferon secretion from dendritic cells that expanded tumor-associated TCF1⁺PD-1⁺CD8 T cells, restoring therapeutic response to PD-1 ICB in KPL tumors. This was observed in syngeneic immunocompetent mouse models and in humanized mice bearing STK11/LKB1 mutant NSCLC human tumor xenografts. NSCLC-affected individuals with identified STK11/LKB1 mutations receiving bemcentinib and pembrolizumab demonstrated objective clinical response to combination therapy. We conclude that AXL is a critical targetable driver of immune suppression in STK11/LKB1 mutant NSCLC.

Trial registration: ClinicalTrials.gov NCT03184571.

Keywords: Axl; NSCLC; STK11/LKB1 mutation; TCF1 CD8 T cells; immunotherapy.

Graphical abstract

Figure 1

Stk11/Lkb1 mutant NSCLC lacks anti-PD-1-responsive T cells (A) C57BL/6J mice (n = 5) were injected with 1 × 10⁶ KP (left) or KPL (right) tumor cells and treated with anti-PD-1 (10 mg/kg, day 7, 10, 14). Tumor volume was measured every 3 days. ∗∗∗p < 0.005. (B) UMAP of sub-clustering tumor-infiltrating CD8 T cells in KP and KPL tumors. T cell clusters are denoted by color. (C) Composition of CD8 T cells in each cluster from KP and KPL. (D) The expression level of Tcf7 in CD8 T cells of KP and KPL detected from scRNA-seq are compared and visualized through feature plot (left) and violin plot (right). (E) Abundance of TCF1⁺PD-1⁺ cells among gated CD8 tumor-infiltrating lymphocytes (TILs) (per mm³ of tumor) on day 14 post tumor cell injection. ∗∗p < 0.01. (F) Visualization and localization of TCF1⁺ (orange) expressing CD8⁺ (green) T cells in KP (left) and KPL (right) tumor microenvironment through immunohistochemistry. Scale bar, 100 μm. (G) Quantification of immunohistochemistry staining of CD8, Granzyme B (GzB), and CD45RO in STK11/LKB1 wild-type and mutant individuals. ∗p < 0.05; ∗∗p < 0.01. (H) Correlation of TCF7 expression, tumor purity and CD8 T cell infiltration in TCGA lung adenocarcinoma-affected individuals (left). The expression of TCF7 negatively correlates with tumor purity, suggesting that the main source of TCF7 expression detected is stromal and immune cells (correlation = −0.367, p = 3.14e-17). The correlation of CD8 T cell infiltration and STK11 mutation in lung adenocarcinoma individualss modified from TIMER deconvolution (right).

Figure 2

Bemcentinib (BGB324) sensitizes KPL tumors to anti-PD-1 (A) C57BL/6J mice (n = 5) were injected with 1 × 10⁶ KPL tumor cells and treated with BGB324 (50 mg/kg, twice daily), anti-PD-1 (10 mg/kg, day 7, 10, 14), or the combination starting on day 7 post tumor cell injection. Control animals were treated with control IgG (10 mg/kg) and vehicle. Tumor volume was measured every 3 days. ∗∗∗∗p < 0.001. (B) Abundance of TCF1⁺PD-1⁺ cells among gated CD8⁺ T cells (per mm³ of tumor) on day 7 post therapy initiation (day 14 post tumor cell injection). ∗∗∗p < 0.005; ∗∗∗∗p < 0.001. (C) Treatment enriched distribution of stem, clonal expanded, and exhausted effector CD8⁺ T cells (see STAR Methods). The ratio of observed cell numbers to random expectation estimated by the R_o/e index through the chi-square test. +++ (R_o/e ≥ 3, p < 0.05) represents highly enriched, ++ (1.2 ≤ R_o/e < 3, p < 0.05) represents enriched, + (0.8 ≤ R_o/e < 1.2, p < 0.05) represents weakly enriched, − (0 < R_o/e < 0.8, p < 0.05) represents not significant or reduced. (D) Shared clonotypes of TCR between clusters of Cd8 T cells detected by single-cell TCR sequencing (left). Visualization of heatmap into a network plot is shown (right). See STAR Methods. (E) RNA velocity analysis of gene expression in Cd8 T cells from different treatment groups. Arrows indicate potential dynamic paths of differentiation. (F) C57BL/6J mice (n = 5) were injected with 1 × 10⁶ KPL-OVA tumor cells and treated with BGB324 (50 mg/kg, twice daily), anti-PD-1 (10 mg/kg, day 7, 10, 14), or the combination starting on day 7 post tumor cell injection. Control animals were treated with control IgG (10 mg/kg) and vehicle. Abundance of OVA antigen-specific cells among CD8⁺ T cells (per gram of tumor) in each treatment group analyzed on day 14 after tumor cell injection. ∗∗p < 0.01; ∗∗∗∗p < 0.001 (G) Same experimental schema as in (F). Statistical results for IFN-γ-producing cells (per mm³ of tumor) from each treatment group. Splenocytes were isolated and re-stimulated with E7 peptide (negative control) or OVA peptide for 48 h. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.005. (H) C57BL/6J background conditional knockout mice (n = 5) or control littermates (n = 3 in control, n = 4 in combination treatment) were injected with 1 × 10⁶ KPL tumor cells and treated with BGB324 (50 mg/kg, twice daily), anti-PD-1 (10 mg/kg, day 7, 10, 14), or the combination starting on day 7 post tumor cell injection. Control animals were treated with control IgG (10 mg/kg) and vehicle. Tumor volume was measured every 3 days. ∗∗∗∗p < 0.001.

Figure 3

Inhibition of Axl on dendritic cells sensitizes KPL tumors to PD-1 blockade (A) Axl^−/− mice (n = 5) were injected with 1 × 10⁶ KPL tumor cells and treated with anti-PD-1 (10 mg/kg, days 7, 10, 14). Tumor volume was measured every 3 days. ∗∗∗∗p < 0.001. (B) Abundance of TCF1⁺PD-1⁺ cells among gated CD8⁺ T cells (per mm³ of tumor) on day 7 post therapy initiation (day 14 post tumor injection). ∗p < 0.05. (C) UMAP of sub-clustered tumor-infiltrating myeloid cells in KPL tumors. (D) Ridge plot of Axl expression on myeloid cells in KPL tumor myeloid cell clusters. (E) Ridge plot of Axl expression on myeloid cells in KP and KPL tumors. (F) C57BL/6J mice (n = 5) were injected with 1 × 10⁶ KPL tumor cells and treated with macrophage depleting reagent (Liposome) or control liposomes (Ctrl) (see STAR Methods). Mice were treated with BGB324 (50 mg/kg, twice daily) and anti-PD-1 (10 mg/kg, day 7, 10, 14) (Combined), or corresponding IgG and vehicle (Veh) starting on day 7 post tumor cell injection. Tumor volume was measured every 3 days. ∗∗∗∗p < 0.001. (G) C57BL/6J mice (n = 5) and Batf3^−/− mice (n = 4 for control group; n = 5 for treatment group) were injected with 1 × 10⁶ KPL tumor cells. Treatment strategy same as (F). ∗∗∗∗p < 0.001.

Figure 4

Bemcentinib induced type I interferon secretion is required for TCF1⁺PD-1⁺ CD8 cell expansion (A) BMDCs were co-cultured with irradiated KPL cells (40 Gy) and treated with BGB324 (40 nM) for 24 h. Conditioned medium was collected for IFN-β ELISA. ∗∗p < 0.01. (B) C57BL/6J mice (n = 5) were injected with 1 × 10⁶ KPL tumor cells and treated with a control IgG (IgG) or interferon alpha receptor blocking antibody (anti-IFN-αR, see STAR Methods). Mice were treated with BGB324 (50 mg/kg, twice daily) and anti-PD-1 (10 mg/kg, day 7, 10, 14) (combined), or corresponding IgG and vehicle (Ctrl) starting on day 7 post tumor cell injection. Tumor volume was measured every 3 days. ∗∗∗∗p < 0.001. (C) Abundance of TCF1⁺PD-1⁺ cells among gated CD8⁺ T cells (per mm³ of tumor) at day 7 post therapy initiation. Experimental schema is same as (B). ∗∗∗∗p < 0.001. (D) Mean fluorescent intensity (MFI) for TCF1 in CD8⁺ OT-1 cells. BMDCs were co-cultured with isolated CD8⁺ T cells stimulated with ovalbumin (see STAR Methods). Each dot represents one biological replicate. ∗∗∗∗p < 0.001. (E) MFI of TCF1 in CD8⁺ OT-1 cells. Wild-type BMDCs or Axl^−/− BMDCs were co-cultured with isolated CD8⁺ T cells stimulated with ovalbumin for 48h, respectively. IFN-αR blocking strategy is as in (B). ∗∗p < 0.01. (F) Rag1^−/− mice (n = 5) were transferred with CD8⁺ T cells isolated from wild-type or Ifnar1^−/− mice. After 7 days of being transferred, mice were injected with 1 × 10⁶ KPL tumor cells and treated with BGB324 (50 mg/kg, twice daily) and anti-PD-1 (10 mg/kg, day 7, 10, 14) (combined), or corresponding IgG and vehicle (Ctrl). Tumor volume was measured every 3 days. ∗∗∗∗p < 0.001.

Figure 5

Bemcentinib sensitizes human STK11/LKB1 mutant NSCLC tumors in humanized mice (A) Flow cytometric analysis of Axl expression on A549 and H2122 tumor cells in vitro. MFI is shown. (B) The effect of BGB324 on cell viability (IC₅₀) of A549 and H2122 cells. (C) AXL immunohistochemistry in A549 and H2122 xenografts grown in humanized mice. A549 scale bar, 250 μm; H2122 scale bar, 100 μm. (D) Humanization and treatment strategy for tumor-bearing humanized mice. (E) Humanized NSG-SGM3 mice were injected with 1.5 × 10⁶ of A549 cells (n = 5, right flank) and treated with BGB324 (50 mg/kg, twice daily), pembrolizumab (10 mg/kg, day 7, 10, 14), or the combination. Control animals were treated with control IgG (10 mg/kg) and vehicle (Ctrl). Treatment was withdrawn after 20 days. ∗∗∗p < 0.005. (F) Humanized NSG-SGM3 mice were injected with 1.5 × 10⁶ of H2122 cells (n = 2 for control group and 3 for treatment group, bilateral). Treatment schema is as in (D). Treatment was withdrawn after 20 days. ∗∗∗∗p < 0.001. (G and H) Abundance of TCF1⁺PD-1⁺ cells among gated CD8⁺ T cells (per mm³ of tumor) on day 7 post therapy initiation in A549 (G) and H2122 (H) xenografts. ∗∗p < 0.01; ∗∗∗∗p < 0.001.

Figure 6

Clinical evidence of the importance of AXL in STK11/LKB1 mutant NSCLC individualss (A) STK11 mutant NSCLC individuals treated with bemcentinib and pembrolizumab combination therapy. All three individuals showed partial response or stable disease, including subjects 2 and 3 who had already progressed on anti-PD-1/PD-L1 therapy. (B–D) Representative AXL IHC of (B) chemo-refractory individual (scale bar, 100 μm; inset 50 μm) and (C and D) two anti-PD-1/PD-L1 refractory individuals from (A). Magnified regions show examples of AXL⁺ tumor (left) and immune (right) cells in (B) and AXL⁺ immune cells in (C and D). Scale bar, 100 μm, inset 20 μm in (C); scale bar, 100 μm, inset 50 μm in (D). (E) Representative AXL IHC staining (left) and the quantification (right) of a tissue microarray consisting of tumors from 62 NSCLC-affected individuals. For the quantification, none represents non-detectable expression of AXL, medium represents 1%–5% percentage of positive staining, and high represents >5% of positive staining. The p value was calculated through chi-square analysis. Scale bar, 300 μm, inset 200 μm. ∗p < 0.05.