Targeting nucleic acid sensors in tumor cells to reprogram biogenesis and RNA cargo of extracellular vesicles for T cell-mediated cancer immunotherapy

Abstract

Tumor-derived extracellular vesicles (EVs) have been associated with immune evasion and tumor progression. We show that the RNA-sensing receptor RIG-I within tumor cells governs biogenesis and immunomodulatory function of EVs. Cancer-intrinsic RIG-I activation releases EVs, which mediate dendritic cell maturation and T cell antitumor immunity, synergizing with immune checkpoint blockade. Intact RIG-I, autocrine interferon signaling, and the GTPase Rab27a in tumor cells are required for biogenesis of immunostimulatory EVs. Active intrinsic RIG-I signaling governs composition of the tumor EV RNA cargo including small non-coding stimulatory RNAs. High transcriptional activity of EV pathway genes and RIG-I in melanoma samples associate with prolonged patient survival and beneficial response to immunotherapy. EVs generated from human melanoma after RIG-I stimulation induce potent antigen-specific T cell responses. We thus define a molecular pathway that can be targeted in tumors to favorably alter EV immunomodulatory function. We propose "reprogramming" of tumor EVs as a personalized strategy for T cell-mediated cancer immunotherapy.

Keywords: RIG-I; RNA; STING; cancer immunotherapy; cancer resistance; extracellular vesicles; innate immunity; nucleic acid receptors; personalized therapy.

Graphical abstract

Figure 1

Tumor cell-intrinsic RIG-I signaling mediates the release of immunogenic EVs B16.OVA melanoma cells were transfected with a RIG-I ligand (3p-RNA, RIG-I-EVs), interferon-stimulating DNA (cGAS-EVs), or synthetic RNA (synRNA-EVs). Extracellular vesicles (EVs) were enriched from the culture supernatant of treated and untreated (ctrl-EVs) cells. Precipitation of 3pRNA-liposomes in the absence of tumor cells was performed as negative control (mock-EVs). (A) IFN-β release (ELISA) by dendritic cells (DCs) exposed to tumor cell-derived EV samples. Some DCs were directly transfected with in vitro transcribed 3pRNA as positive control. (B–D) IFN-β release from DCs exposed to RIG-I-EV preparations: (B) prepared from different melanoma clones that lack specific downstream signaling components of nucleic acid receptor pathways; (C) enriched from various murine tumor cell lines: mammary (4T1), pancreatic (Panc02), colon (C26) carcinoma; and (D) enriched by precipitation or size-exclusion chromatography. All data are presented as mean values ± SEM of at least quadruplicate technical replicates per group and are representative of two independent experiments. Asterisks without brackets indicate statistical comparison with Ctrl-EV-treated cells. (E) Presence of the antigen OVA (by western blot) in melanoma cell EV samples. (F) Treatment model. C57BL6/j mice were repeatedly injected subcutaneously with tumor EV samples prepared from cultures of wild-type, RIG-I-deficient (RIG-I^−/−), or IRF3/7-deficient (IRF3/7^−/−) B16.OVA melanoma cells. (G) IFN-γ release by CD8⁺ T cells from draining lymph nodes after ex vivo OVA restimulation (flow cytometry). Data are presented as mean values ± SEM of n = 4–5 individual mice per group and were pooled from two independent experiments. Unstim, unstimulated. See also Figures S1–S3.

Figure 2

EVs released from RIG-I-activated tumor cells induce potent cytotoxic antitumor immunity and synergize with checkpoint inhibitors (A) B16.OVA-bearing mice were injected peritumorally with melanoma-cell-derived EV samples. (B and C) (B) Tumor growth and (C) overall survival. (D) Tumor growth after rechallenge with a second, contralateral injection of viable B16.OVA cells in mice with initially complete tumor regression in response to EV treatment. (E and F) Abundance of activated tumor-infiltrating (E) DCs and (F) CD8⁺ T cells (n = 8–12 mice per group). (G and H) Tumor growth in mice additionally injected with (G) anti-CD8a- or anti-NK1.1-depleting antibodies, or (H) anti-PD-1/CTLA-4 checkpoint inhibitors. (I) Panc02 pancreatic carcinoma growth in mice injected peritumorally with Panc02-cell-derived EV samples. (J) IFN-β release from DCs exposed to RIG-I-induced EVs from either culture of B16.OVA melanoma cells or ex vivo cell suspensions of freshly isolated bulk tumors (mean ± SEM of n = 8 biological replicates pooled from at least three independent experiments). All data are presented as mean tumor growth ± SEM and are pooled from or representative of at least two independent experiments. See also Figures S3 and S4.

Figure 3

Therapeutic in situ activation of the RIG-I pathway modulates tumor EV generation for systemic antitumor immunity (A) RNA-seq of bulk tumor after a single intratumoral injection of 3pRNA in B16.OVA melanoma-bearing mice. Heatmap shows Z-transformed expression of genes associated with tumor EV biogenesis and cargo loading. (B and C) (B) Rab27a mRNA copy numbers of two different Rab27a-deficient (Rab27a^−/−) clones and (C) corresponding protein expression (western blot). (D and E) EVs were enriched from Rab27a^−/− B16 cell cultures. (D) Particle quantification within EV preparations by NTA. (E) IFN-I induction in DCs exposed to EVs from Rab27a^−/− B16.OVA cells. (F) Mice bilaterally bearing either wild-type or Rab27a^−/− melanomas were injected with 3pRNA into right-sided tumors. (G) Mean and individual frequency of H-2Kb-SIINFEKL tetramer⁺ CD8⁺ T cells in peripheral blood of n = 10 mice per group pooled from two independent experiments. (H) Mean volume of left-sided tumors ± SEM of n = 15 individual mice per group pooled from three independent experiments. All in vitro data show mean ± SEM of at least quadruplicate technical replicates, representative of at least two independent experiments.

Figure 4

Immunogenicity of RIG-I-induced isEVs is mediated via host cytosolic nucleic acid receptor signaling and IFN-I activity in myeloid antigen-presenting cells (A) Mice were treated with anti-IFNaR1 antibodies prior to immunization with B16.OVA melanoma cell EV samples. IFN-γ release by CD8⁺ T cells (flow cytometry). (B and C) IFN-γ release by CD8⁺ T cells upon tumor EV immunization in mice with genetic deficiency (B) for IFNaR1 in DCs (CD11c-Cre Ifnar^fl/fl) or macrophages (LysM-Cre Ifnar^fl/fl), and (C) globally for MAVS (Mavs^−/−) or STING (Sting^gt/gt). Data represent the mean value ± SEM of individual mice per group pooled from at least two independent experiments. (D and E) IFN-I production in (D) Mavs^−/− or Sting^gt/gt DCs and (E) TLR-3-deficient (Tlr3^−/−) or Myd88-deficient (Myd88^−/−) DCs exposed in vitro to melanoma EV samples. All in vitro data show mean ± SEM of at least triplicate technical replicates, representative of at least two independent experiments. Asterisks without brackets indicate statistical comparison to vehicle-treated control cells. See also Figure S4.

Figure 5

Tumor-intrinsic RIG-I pathway activity mediates shuttling of immunostimulatory RNA within EVs (A) Transfer of EV RNA to DCs after exposure to B16 melanoma EVs with fluorescently labeled RNA cargo (flow cytometry). (B) IFN-β release by DCs upon transfection with tumor EV extracted RNA or in vitro transcribed 3pRNA. Some EV RNA was treated with alkaline phosphatase prior to DC transfection (ELISA, mean value ± SEM of at least triplicate technical replicates, representative of at least two independent experiments). (C–G) RNA-seq of small RNA content extracted from B16.OVA melanoma cell EVs (n = 3 biological replicates per group). (C) Principal component analysis of the RNA content of Ctrl-EVs vs. RIG-I-EVs. (D) Absolute (upper panel) and relative abundance (lower panel) of the indicated RNA biotypes. (E) Relative abundance of the indicated short non-coding RNA. (F) Differential gene expression volcano plot of RNA content in RIG-I- vs. Ctrl-EVs. Significantly differentially expressed small non-coding RNAs are color-coded by their respective biotypes. (G) Cumulative abundance of U1 and U2 splicesosomal RNA reads in the RNA content of EVs, normalized to the overall library size. See also Figures S5 and S6; Tables S1 and S2.

Figure 6

Tumor-intrinsic RIG-I activation in human melanoma mediates shuttling of immunostimulatory RNA within EVs and associates with beneficial clinical response (A and B) Production of CXCL10 and IFN-I in human peripheral blood mononuclear cells (PBMCs) upon exposure to (A) human melanoma cell (D04mel)-derived RIG-I-EVs vs. Ctrl-EVs or (B) RNA purified from melanoma EV samples. (C) IFN-I response in monocytic THP-1 reporter cells after exposure to EV samples derived from human melanoma cell lines with deficiency in RIG-I, cGAS, STING, or IFNAR1. (D) IFN-I response in THP-1 cells with overexpression (STING^+/+) or gene-deficiency for STING (STING^−/−) upon exposure to different human melanoma EV preparations. (E) Melanoma antigen-specific T cell response against Melan A (upper panel) or tyrosinase (lower panel) induced by melanoma EV samples in co-culture with autologous HLA-matched DCs (IFN-γ ELISpot, n = 2 technical replicates per group). (F–G) Transcriptional activity of RIG-I-encoding DDX58 and the Tumor EV pathway gene set in tumor samples from individual patients undergoing immune checkpoint inhibition. The dotted lines give mean transcriptional activity of the indicated genes. Gene expression and treatment response in (F) 20 melanoma patients treated with anti-CTLA-4 and (G) 41 melanoma patients treated with anti-PD-1. All bar graphs show mean ± SEM of at least triplicate technical replicates, representative of at least two independent experiments. nd, not determined. See also Figures S7 and S8; Table S3.