Therapeutically Active RIG-I Agonist Induces Immunogenic Tumor Cell Killing in Breast Cancers Free

Breast cancer is the most frequently diagnosed cancer in women (1). Despite advances in early detection and treatment, breast cancer remains the second leading cause of cancer-related deaths for women. With an eye toward new treatment strategies, recent attention has focused on immune-checkpoint inhibitors (ICI), antibodies that block regulatory receptors that dampen adaptive immunity (e.g., PD-1, PD-L1, CTLA-4). ICI-mediated inhibition of checkpoint receptors releases regulatory restraints on adaptive immunity, permitting a proinflammatory lymphocytic response against tumor neoantigens, and resulting in robust and durable antitumor immune responses (2). ICI treatments have seen remarkable success in cases of melanoma and lung cancer (2–4). However, ICI response rates reported in breast cancer clinical trials have thus far been disappointing, achieving success in only a fraction of patients (5–7). The relatively diminished response to ICIs in breast cancer is not completely understood, but may relate to fewer tumor-infiltrating lymphocytes (TIL; refs. 8, 9), a decreased mutational burden (10, 11), limited or absent expression of antigen presentation machinery on tumor cells (12), or enhanced expression of counterregulatory factors in breast cancers as compared with what is seen in other cancer types (13). A fraction of the highly aggressive triple-negative breast cancer (TNBC) subtypes, which on average harbor a greater number of TILs and a greater mutational burden than the other clinical breast cancer subtypes, have shown greater response to ICI over those breast cancer subtypes expressing estrogen receptor (ER⁺) or harboring HER2 gene amplification (HER2⁺). Further, decreased TILs within the aggressive TNBC subtype predict poor outcome and decreased response to ICI. Interestingly, certain chemotherapeutic regimens increase TILs in breast cancers, which often correlates with improved response to treatment. Therefore, a new treatment paradigm may be needed in breast cancers to promote de novo inflammation to instigate antitumor immunity, or to enable efficacy of existing ICIs. It is possible that treatment strategies that increase TILs improve antigen presentation or increase inflammatory cytokines in the tumor microenvironment might improve immunogenicity of all breast cancer subtypes.

Pattern recognition receptors (PRR) of the innate immune system, which recognize conserved pathogen-associated molecular patterns (PAMP; e.g., viral nucleotide motifs), are gaining interest as a potential treatment strategy (14, 15). PRR activation by their viral nucleotide ligand induces proinflammatory transcription factors, including NF-κB, signal transduction and transcription, and interferon regulatory factors (IRF), which drive production of IFNs and other proinflammatory cytokines that orchestrate antimicrobial innate immune responses and stimulate adaptive immunity (14, 16). Certain PRRs are expressed in nearly every cell in the human body, including cancer cells, suggesting that some PRRs might be leveraged therapeutically as part of a cancer treatment strategy. This idea is being explored extensively in regard to the PRR known as stimulator of interferon genes (STING; refs. 17, 18). Synthetic STING ligands potently induce type I IFNs and support antitumor immunity across a variety of cancers, including breast, CLL, colon, and squamous cell carcinoma (19–23). However, there is increasing evidence that STING signaling is defective in many cancers, including breast some breast cancers, due to mutations, promoter methylation, and decreased expression of genes in the STING pathway (24, 25).

Retinoic acid-inducible gene I (RIG-I) is another PRR, playing a key role in recognizing RNA viruses. In contrast to the frequent STING pathway alterations seen in breast cancers, alterations in the RIG-I gene DDX58 have been infrequently reported, and DDX58 promoter methylation was not significantly higher in breast tumor versus normal breast tissue (24). RIG-I recognizes double-stranded viral RNAs (dsRNA) containing two or three 5′-phosphates (26–29). RIG-I activation by its ligand causes RIG-I translocation to mitochondria, where it interacts with its binding partner Mitochondrial antiviral signaling (MAVS) to activate signaling pathways that produce proinflammatory cytokines (30). Importantly, RIG-I activation also promotes the elimination of virally infected cells through apoptotic pathways (31–33). These attributes make RIG-I mimetics an attractive therapeutic approach in immune oncology.

Therapeutic efficacy of RIG-I mimetics has been seen in several cancer cell lines originating from a variety of tissues, although the impact of RIG-I activation in breast cancers is relatively understudied as compared with other cancers. Further, the use of RIG-I agonists as a cancer treatment requires a specific and potent RIG-I ligand that is functional in vivo, which has only recently been reported in a study using a minimal 5′-triphophosphorylated stem-loop RNA (SLR) sequence for intravenous delivery to mice (28). The stem-loop structure enhances structural stability of the complex, a key determinant of RIG-I ligand potency. Delivery of SLR sequences to mice in vivo activated in RIG-I signaling, IFN induction, and expression of genes required for potent antiviral immunity. However, the efficacy of RIG-I ligands, including SLRs, in animal models of cancer has not yet been tested.

We tested the hypothesis that RIG-I–mediated activation of innate immune responses might be therapeutically efficacious in breast cancers, while increasing the inflammatory phenotype of breast cancers. We demonstrate here that RIG-I activation in breast cancer cells resulted in tumor cell–intrinsic tumor cell death due in part to activation of pyroptosis and induced expression of inflammatory cytokines, leukocyte-recruiting chemokines, and increased expression of major histocompatibility (MHC)-I components. Delivery of synthetic RIG-I ligands to breast tumors in vivo recapitulated these results, recruiting leukocytes to the tumor microenvironment and decreasing tumor growth and metastasis.

Oligoribonucleotides sequence OH-SLR20 (5′-GGACGUACGUUUCGACGUACGUCC) was synthesized on an automated MerMade synthesizer (BioAutomation) using standard phosphoramidite chemistry. The hydroxylated oligonucleotide was deprotected and gel purified as previously described (34). Triphosphorylated oligoribonucleotide SLR20 (5′ppp-GGACGUACGUUUCGACGUACGUCC) was synthesized as described (35), deprotected, and gel purified. The triphosphorylation state and purity were confirmed using mass spectrometry. The oligonucleotides were resuspended in RNA storage buffer (10 mmol/L MOPS pH 7, 1 mmol/L EDTA) and snap cooled to ensure hairpin formation, as previously described (28).

The human cell lines MCF-7, BT474, and murine cell line 4T1 were purchased from ATCC in 2015. All cells were maintained at low passage in DMEM with 10% fetal bovine serum and 1% antibiotics and antimyotics. Cell identity was verified by ATCC using genotyping with a Multiplex STR assay. All cell lines were screened monthly for Mycoplasma. Cells were used within 20 passages for each experiment.

SLR20 and OH-SLR20 were delivered to cells in serum-free Opti-MEM media at a final concentration of 0.25 μmol/L using lipofectamine 2000 (Invitrogen). Where indicated, cells were treated with staurosporine (Cell Signaling Technology) at a final concentration of 1 μmol/L in serum-free Opti-MEM. Cells expressing shRNA against RIG-I were generated by transduction with pLKO lentiviral particles (Sigma-Aldrich) harboring shRNA sequences against human or mouse RIG-I (DDX58) and selected with puromycin (2 μg/mL). Cells were treated with inhibitors for caspase-9 (Z-LEHD-FMK, BD Pharmingen), caspase-10 (Z-AEVD-FMK, Cayman Chemical), and caspase-1 (Ac-YVAD-CHO, Cayman Chemical) at a final concentration of 5 μmol/L.

Whole-cell lysate was harvested by homogenization of cells in ice-cold lysis buffer [50 mmol/L Tris pH 7.4, 100 mmol/L NaF, 120 mmol/L NaCl, 0.5% nonidet P-40, 100 μmol/L Na₃VO₄, 1 × protease inhibitor cocktail (Roche), 0.5 μM MG132 (Selleck Chem)]. Mitochondrial and cell membrane extracts were harvested from cells using the Cell Fractionation Kit (Cell Signaling Technologies) according to the manufacturer's instructions. Lysates (20 μg protein measured by BCA assay) were resolved on 4% to 12% polyacrylamide gels (Novex) and transferred to nitrocellulose membranes (iBlot), blocked in 3% gelatin in TBS-T (Tris-buffered saline, 0.1% Tween-20), incubated in primary antibodies from Cell Signaling Technologies: (RIG-I (D14G6, 1:1,000), MAVS (3993, 1:1,000), SOD2 (D3 × 8F, 1:1,000), p65 (D14E12, 1:1,000), P-p65 Y701 (D4A7, 1:1,000), STAT1 (D1K9Y, 1:1,000), P-STAT1 S536 (93H1, 1:1,000), PARP (9542, 1:1,000), caspase-1 (2225, 1:1,000), cleaved caspase-1 (D57A2, 1:1,000), gasdermin D (96458, 1:1,000), Rab11 (7100, 1:1,000); β-actin (Sigma-Aldrich, AC-15, 1:10,000); and E-cadherin (BD Transduction Laboratories, 610182, 1:1,000). Secondary antibodies were from PerkinElmer [goat anti-rabbit (1:5,000) and goat anti-mouse (1:10,000)]. Western blots were developed with ECL substrate (Thermo Fisher Scientific).

Live cells were incubated in MitoTracker Red (Invitrogen) for 45 minutes to stain mitochondria then 100% methanol fixed, blocked in TBS-T 3% gelatin and stained with rabbit anti–RIG-I (1:100, Cell Signaling Technology) and goat anti-rabbit Alexa Fluor 488 (1:1,000, Invitrogen). For apoptotic analysis, live cells were stained with Annexin V, AlexaFluor-488 conjugate (1:500, Invitrogen) for 4 hours before imaging. For pyroptotic studies, live cells were stained in propidium iodide (PI; Sigma-Aldrich, 1:1,000) for 1 hour before imaging.

Amphiphilic diblock copolymer composed of a 10.3 kDa dimethylaminoethyl methacrylate (DMAEMA) first block and a 31.0 kDa, 35% DMAEMA, 39% butyl methacrylate (BMA), and 26% propylacrylic acid (PAA) second block were synthesized as previously described (36). Dry amphiphilic diblock polymer was dissolved into ethanol at 50 mg/mL, rapidly diluted into phosphate buffer (pH 7.0, 100 mmol/L) to 10 mg/mL, concentrated, and buffer was exchanged into PBS (Gibco) using 3 kDa molecular weight cutoff centrifugal filtration columns (Ambion, Millipore) and sterile filtered. Polymer concentration was measured by absorbance at 310 nm (Synergy H1 microplate reader, BioTek). Concentrated polymer solution was rapidly mixed with SLR20 (or OH-SLR20) at a charge ratio of 5:1 (N:P) for 30 minutes and diluted into PBS (pH 7.4, Gibco) to 20 μg of SLR and 400 μg of polymer in 50 μL total volume.

All studies were performed in accordance with Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC) guidelines and were approved by the Institutional Animal Care and Use Committee at Vanderbilt University. All mice were housed in pathogen-free conditions. Left inguinal mammary fat pads of wild-type (WT) female Balb/c mice or athymic (nu/nu) Balb/c mice (Jackson Labs) were injected with 10⁶ 4T1 cells. Mice were randomized into treatment groups when tumors reached 50 to 100 mm³. Intratumoral injection of nanoparticle in 50 μL of saline (or saline without nanoparticle) was performed at 48-hour intervals for a total of 3 treatments, or at 72- to 96-hour intervals for a total of 4 treatments. Intraperitoneal injection of InVivoMab αPD-L1 (B7-H1) and control IgG2b κ from Bio X Cell were delivered at 25 mg/kg in sterile saline twice weekly for 10 days. Mice were monitored daily, and tumor volume was measured with calipers twice weekly for up to 25 days.

Lungs and tumors were formalin-fixed and paraffin-embedded, and sections (5 μm) were stained with hematoxylin and eosin. In situ TUNEL analysis was performed on paraffin-embedded sections using the ApopTag kit (Millipore). IHC was performed using the following antibodies: RIG-I (Invitrogen, PA5-20276, 1:400), P-STAT1 (Cell Signaling Technology, 9167, 1:100), Ki67 (Biocare Medical, CFM325B, 1:100), CD45 (Abcam, ab10558, 1:5,000), F4/80 (Bio-Rad, MCA497GA, 1:100), CD4 (eBioscience, 14-0195-82, 1:1,000), CD8 (eBioscience, 14-0195-82, 1:100), TRAIL (GeneTex, 6TX11700, 1:800). Immunodetection was performed using the Vectastain kit (Vector Laboratories) according to the manufacturer's instructions.

Total RNA was extracted using NucleoSpin RNA (Machery-Nagel), reverse transcribed (iScript cDNA Synthesis; Bio-Rad), and using for qPCR with iTaq Universal SYBR Green (Bio-Rad) on a Bio-Rad CFX96 thermocycler. Gene expression is normalized to 36B4. The following primers were obtained from Integrated DNA Technologies: IFNB1 [forward 5′-TGCTCTCCTGTTGTGCTTCTCC; reverse 5′-GTTCATCCTGTCCTTGAGGCAGT]; Ifnb1 [forward 5′-CAGCTCCAAGAAAGGACGAAC; reverse 5′-GGCAGTGTAACTCTTCTGCAT]; HLA-A [forward 5′-GCGGCTACTACAACCAGAGC; reverse 5′-GATGTAATCCTTGCCGTCGT]; TNF [forward 5′-CCTCTCTCTAATCAGCCCTCTG; reverse 5′-GAGGACCTGGGAGTAGATGAG]; Tnf [forward 5′-CCCTCACACTCAGATCATCTTCT; reverse 5′-GCTACGACGTGGGCTACAG]; TNFSF10 [forward 5′-TGCGTGCTGATCGTGATCTTC; reverse 5′-GCTCGTTGGTAAAGTACACGTA]; Tnfsf10 [forward 5′-ATGGTGATTTGCATAGTGCTCC; reverse 5′-GCAAGCAGGGTCTGTTCAAGA]. Other genes were analyzed via PCR array (RT² Profiler Array, Qiagen).

Cells (1 × 10⁶) were seeded. After 24 hours, cells were transfected with SLR20 or OH-SLR20 as described above. Cell culture media were removed 32 hours after transfection, filtered with a 0.2 micron strainer, and immediately added to blocked membranes from Human Cytokine Antibody Array C1000 (RayBiotech), and processed according to the manufacturer's instructions. Chemiluminescent cytokine arrays were imaged digitally using Amersham Imager 600 (GE Healthcare).

Experimental groups were compared with controls using Student unpaired, two-tailed t test. Multiple groups were compared across a single condition using one-way analysis of variance (ANOVA). P < 0.05 was used to define significant differences from the null hypothesis. qPCR array data sets were compared using multiple t tests with an FDR cutoff of 0.05.

Animals were housed under pathogen-free conditions, and experiments were performed in accordance with AAALAC guidelines and with Vanderbilt University Institutional Animal Care and Use Committee approval.

To assess the potential applicability of a RIG-I agonist in breast cancer, we examined RIG-I/DDX58 expression in a clinical invasive breast cancer data set curated by The Cancer Genome Atlas (TCGA; ref. 37). We found genomic DDX58 deletion in only 1 of 817 tumors and mRNA downregulation in only 8 of 817 tumors (Fig. 1A), suggesting that loss of RIG-I expression is a rare event. Similar results were produced upon the analysis of 2509 breast tumors from the METABRIC invasive breast cancer data set (Supplementary Fig. S1A; ref. 38). Whole-exome sequencing data identified 3 nonrecurrent missense mutations within DDX58, and no recurrent, truncating, or in-frame mutations (Supplementary Fig. S1B), suggesting that RIG-I/DDX58 is rarely lost or mutated in breast cancers.

Western analysis confirmed RIG-I expression in two human breast cancer cell lines, MCF7 (ER⁺), and BT474 (HER2 amplified; Fig. 1B), but not in HER2-amplified, ER⁺ MDA-MB-361 cells (Supplementary Fig. S2A). To determine if RIG-I signaling pathways are functional in breast cancer cells, we used a previously described synthetic minimal RIG-I agonist composed of a double-stranded, triphosphorylated 20-base pair stem-loop RNA, which was then modified with a 5′ triphosphate sequence (SLR20; ref. 39). Previous studies demonstrated that SLRs containing the 5′ ppp motif, but not those lacking the motif, activate type I IFN production via RIG-I/MAVS signaling. We transfected SLR20 (and the nonphosphorylated, but otherwise identical sequence, OH-SLR20) into MCF7 cells and measured RIG-I expression and distribution by immunofluorescence. RIG-I expression was robustly increased in cells transfected with the RIG-I ligand SLR20 as compared with the control ligand OH-SLR20 (Fig. 1C). Counterstaining of mitochondria demonstrated mitochondrial localization of RIG-I in many cells following SLR20 treatment. Further, analysis of mitochondrial cell fractions by Western analysis confirmed mitochondrial RIG-I localization in MCF7 and BT474 cells transfected with SLR20, but not OH-SLR20 (Fig. 1D). Western analysis confirmed RIG-I upregulation following transfection with SLR20 in MCF7, BT474, and mouse 4T1 cells, a mammary tumor line used as a model of aggressive, metastatic, and poorly immunogenic TNBC (Fig. 1E). Importantly, SLR20 increased phosphorylation of the proinflammatory transcription factors p65 (an NF-κB subunit) and STAT1 in MCF7, BT474, and 4T1 cells. Importantly, SLR20 did not affect P-p65 in MDA-MB-361 cells, which lack RIG-I expression (Supplementary Fig. S2B). These data support use of these breast cancer cell lines, and the SLR20 agonist, to model the therapeutic impact of RIG-I signaling in breast cancer.

A recent study examining SLR delivery in vivo confirmed rapid induction of type I IFNs following delivery of a 10-bp SLR sequence (SLR10; ref. 28). However, the impact of RIG-I activation in the complex breast tumor microenvironment has not been explored. We used a nanoparticle-based platform previously optimized for oligonucleotide delivery in vivo for intratumoral (i.t.) treatment of breast tumors with SLR20 (Fig. 2A). These pH-responsive nanoparticles (NP) were composed of amphiphilic diblock copolymers formulated with a hydrophobic core-forming block that is endosomolytic and drives micellar assembly, and a polycationic corona for electrostatic complexation with oligonucleotide (i.e., SLR20), as described previously (36). This formulation has been shown to maximize cytoplasmic delivery of oligonucleotides, an ideal scenario for cytoplasmic RIG-I activation by SLR20. NPs were delivered i.t. to 4T1 mammary tumors grown in WT Balb/c female mice when tumors reached 50 to 100 mm³. As an additional control, a third group of tumor-bearing mice were treated by i.t. injection of saline, the vehicle in which NPs were delivered. A total of 3 treatments were administered (days 0, 2, and 4; Fig. 2B). IHC of tumors collected at day 5 (24 hours after final treatment) revealed RIG-I protein upregulation in 4T1 tumors treated with SLR20 NPs over saline-treated or OH-SLR20 NP-treated tumors (Fig. 2C–D and Supplementary Fig. S2C). Further, tumors treated with SLR20 NPs, but not OH-SLR20 NPs, exhibited a 3-fold increase in phosphorylation of STAT1 (Fig. 2C and D), confirming RIG-I signaling in SLR20-treated tumors in vivo.

We used a slightly modified treatment scheme to assess the therapeutic impact of SLR20-NP on tumor growth. Tumors were treated on days 0, 3, 6, and 9 with i.t. delivery of SLR20 NPs, at which point treatment stopped and tumor volume was monitored through day 25 (Fig. 2E). Tumors treated with SLR20 NPs did not increase in volume during the treatment window (days 0–9), while tumors treated with saline or with OH-SLR20 NPs increased nearly 4-fold (Fig. 2F). Once treatment was complete, tumors treated with SLR20 resumed volumetric increase, but still grew at a diminished rate as compared with tumors treated with OH-SLR20 NPs or with saline. Lungs harvested from mice on day 25 revealed a decreased number of lung metastases in the SLR20-treated mice as compared with control groups (Fig. 2G). Treatment extended through treatment day 25 (Supplementary Fig. S3A) resulted in sustained tumor growth inhibition in response to SLR20 (Supplementary Fig. S3B).

We investigated potential mechanisms responsible for decreased tumor growth and metastasis following treatment with SLR20 NPs, first measuring Ki67-positive cells by IHC as a marker of cell proliferation (Fig. 3A; Supplementary Fig. S4A). Assessing 4T1 tumors collected on treatment day 5, we found a decreased percentage of Ki67⁺ tumor cells in samples treated with SLR20 NPs as compared with samples treated with OH-SLR20-NPs or with saline (Fig. 3B). Conversely, tumor cell death, measured by terminal dUTP nick-end labeling (TUNEL) analysis (Fig. 3A), was increased 5-fold in samples treated with SLR20 NPs (Fig. 3B). RIG-I signaling is capable of inducing programmed cell death in many cell types, including some cancer cell types (40), although this possibility remains unclear in breast cancers. Therefore, we transfected MCF7, BT474, and 4T1 cells in culture with SLR20, assessing cells 12 hours after transfection for PARP cleavage, a molecular marker of cell death. Cleaved PARP was increased in cells transfected with SLR20 versus OH-SLR20 (Fig. 3C). Annexin V-FITC staining was used to enumerate apoptotic cells, revealing increased Annexin V⁺ cells following SLR20 treatment in MCF7, BT474, and 4T1 cells (Fig. 3D), but not in MDA-MB-361 cells, which lack RIG-I expression (Supplementary Fig. S4B). Importantly, knockdown of RIG-I in MCF7, BT474, and 4T1 cells using shRNA sequences against RIG-I (Fig. 3E) abrogated the increased Annexin V staining in response to SLR20 (Fig. 3F), while SLR20 remained capable of inducing Annexin V staining in cells expressing nontargeting shRNA sequences. These findings demonstrate that SLR20 activates RIG-I signaling in breast cancer cells, inducing cell death in a tumor cell–intrinsic manner.

Because RIG-I signaling is reported to induce apoptosis through several distinct pathways, including intrinsic apoptosis, extrinsic apoptosis, and pyroptosis pathways across a variety of cell types (33), it is unclear by which pathway RIG-I induces cell death in breast cancers. We investigated this using an apoptosis expression array, assessing expression changes in 84 genes associated with the intrinsic and extrinsic apoptosis pathways. RNA harvested from BT474 cells collected 16 hours after transfection harbored changes in 18 of the 84 genes assessed. Genes arranged in order of expression fold change revealed that genes regulating intrinsic apoptosis (e.g., BAD, BAX, CASP9) were downregulated, while expression of genes regulating the extrinsic apoptosis pathway (TNFSF10, FAS, CASP10, CASP8) were upregulated (Fig. 4A), suggesting that the extrinsic apoptosis pathway might be activated in response to RIG-I signaling in breast cancer cells. We confirmed that SLR20 induced expression of the extrinsic apoptotic factor TNFSF10 in MCF7, BT474, and 4T1 cells (Fig. 4B). Additionally, 4T1 tumors treated in vivo with SLR20 NPs were assessed by IHC for expression of the Tnfsf10 gene product, TRAIL. Although TRAIL was expressed at only low levels in 4T1 tumors treated with saline or with OH-SLR20 NPs (Fig. 4C), TRAIL protein levels were markedly upregulated in samples treated with SLR20 NPs.

These data suggest that RIG-I might activate the extrinsic apoptosis pathway in breast cancer cells but do not rule out that RIG-I signaling might also induce breast cancer cells to undergo pyroptosis, an inflammatory type of programmed cell death that requires activation of caspase-1 and oligomerization of gasdermin D on the cell membrane (41). Western analysis of MCF7 and BT474 cells transfected with SLR20 revealed potent activation of caspase-1 (Fig. 4D) and localization of gasdermin D to cell membranes (Fig. 4E). These findings were confirmed in 4T1 cells (Supplementary Fig. S4C). Upregulation of CASP1 and CASP4 (encoding another mediator of pyroptosis, caspase-4) was seen in BT474 cells transfected with SLR20 (Fig. 4F). Importantly, Casp1 levels were increased in 4T1 tumors treated in vivo with SLR20 NPs, but not in tumors treated with OH-SLR20 NPs (Fig. 4G), suggesting the RIG-I–mediated activation of pyroptotic signaling pathway may be maintained even within the complex tumor microenvironment.

Next, we used a selective inhibitor of caspase-10, AEVD-FMK, to block the extrinsic apoptotic pathway, resulting in a moderate, but significant, diminution of Annexin V⁺ cells following treatment with SLR20 (Fig. 4H). In contrast, the caspase-9 inhibitor Z-LEHD-FMK, which blocks activation of the intrinsic apoptotic pathway, had little impact on Annexin V staining in cells transfected with SLR20. We also tested an inhibitor of caspase-1, Z-YVAD-FMK, to block the pyroptosis pathway in SLR20-transfected cells, resulting in partial inhibition of Annexin V staining. These results were confirmed in MCF7 cells (Supplementary Fig. S4D). Interestingly, the combination of the caspase-10 inhibitor with the caspase-1 inhibitor produced a greater reduction in Annexin V staining in BT474 cells as compared with either inhibitor used alone (Fig. 4H), consistent with the idea that these two inhibitors operate through distinct pathways in BT474 cells, and suggesting that RIG-I signaling in breast cancer cells may use both the intrinsic apoptosis pathway and pyroptosis to potently induce programmed cell death. Because pyroptosis produces pores in the plasma membrane (41), making them permeable to PI, we stained MCF7 and BT474 cells with PI at 12 hours after transfection with OH-SLR20 or SLR20, finding a robust increase in PI⁺ staining when cells were transfected with SLR20 (Fig. 4I). However, PI staining was completely abolished in MCF7 and BT474 cells pretreated with the caspase-1 inhibitor, confirming that pyroptosis is induced by SLR20 in breast cancer cells.

In contrast to the intrinsic apoptosis pathway, which is considered an immunologically silent form of programmed cell death, pyroptosis is thought to be an immunogenic form of cell death that may recruit inflammatory leukocytes to the site of a viral infection through cytokine modulation, while increasing immunogenicity of the infected cell through increased expression of the major histocompatibility complex (MHC)-I, the antigen presentation machinery expressed on most nucleated cells. Consistent with this idea, both MCF7 and BT474 breast cancer cells transfected with SLR20 showed upregulation of HLAB (Fig. 5A), encoding a key MHC-I component. The gene B2M, encoding another key MHC-I component, β2 microglobulin, was similarly upregulated in BT474 cells (Fig. 4A).

We assessed leukocyte recruitment to 4T1 mammary tumors grown in immune-competent Balb/c mice following treatment with SLR20 NPs. IHC analysis for CD45, a pan-leukocyte marker, revealed substantially increased CD45⁺ cells in tumors treated with SLR20 NPs versus saline or OH-SLR20 NPs (Fig. 5B and Supplementary Fig. S5). Further, IHC analysis using antibodies against F4/80 (a mature macrophage marker), CD4 (a marker of helper T lymphocytes), and CD8, a marker of cytotoxic T lymphocytes (CTL), natural killer T cells (NK-T) and inflammatory dendritic cell (DC) populations, were each increased in tumors treated with SLR20 NPs as compared with saline OH-SLR20 NP-treated tumors (Fig. 5B and C). These data suggest that RIG-I activation results in active recruitment of leukocytes to the TME, consistent with a more immunogenic tumor microenvironment. Consistent with this notion, we found that SLR20 delivery to 4T1 tumors grown in immunocompromised athymic Balb/c (nu/nu) mice displayed more rapid resurgence of tumor growth once the SLR20 treatment was discontinued (Supplementary Fig. S6). This use of the RIG-I ligand to generate a more immunogenic tumor microenvironment was tested more directly using SLR20 in combination with the ICI, αPD-L1. Tumor-bearing WT Balb/c mice were randomized into groups to receive treatment with SLR20, OH-SLR20, or saline and were randomized further into groups receiving αPD-L1 or an isotype-matched control IgG (Fig. 5D). Tumors were treated twice weekly through treatment day 10 and monitored through treatment day 18. We found that tumors treated with SLR20 alone grew at a slower rate than tumors treated with OH-SLR20 or with saline (Fig. 5E). Tumor growth was inhibited by treatment with αPD-L1 alone. However, the combination of SLR20 with αPD-L1 decreased tumor growth to a greater extent than either agent alone, and to a greater extent than αPD-L1 in combination with the control OH-SLR20 NP. These findings are consistent with the idea that SLR20 increases the immunogenicity of the tumor microenvironment in this model of breast cancer.

Like many PRRs, RIG-I induces expression of inflammatory cytokines required for lymphocyte recruitment (42). Therefore, we measured expression of IFNB1 in MCF7, BT474, and 4T1 cells following transfection with SLR20, revealing IFNB1 upregulation (Fig. 6A). Notably, Ifnb1 upregulation was also seen in 4T1 tumors treated in vivo with SLR20 NPs (Fig. 6B). SLR20-mediated upregulation of IFN1B was impaired in MCF7 and BT474 cells expressing RIG-I-directed shRNA sequences (Fig. 6C; Supplementary Fig. S7). This suggests that RIG-I signaling in breast cancer cells might be capable of activating in trans the antigen presenting cells, such as macrophages, within the tumor microenvironment. We tested this hypothesis by harvesting cultured media from 4T1 cells transfected with SLR20 or OH-SLR20, and adding the cultured media to macrophage-derived Raw264.7 cells. After 30 minutes of exposure to cultured media harvested from 4T1 cells treated with SLR20, we found phosphorylation of the proinflammatory transcription factor STAT1 (Supplementary Fig. S8A). Induction of TNF gene expression following transfection with SLR20 also was seen in MCF7, BT474, and 4T1 (Fig. 6D), but not in cells expressing RIG-I shRNA sequences (Supplementary Fig. S8B). To confirm that these gene-expression changes were seen at the protein level, we assessed cultured media harvested from MCF7 cells 48 hours after transfection with SLR20 by cytokine array analysis. Although this array did not carry IFNβ, we observed increased protein levels of TNFα and TNFβ in the cultured media harvested from SLR20 transfected MCF7 cells (Fig. 6E). Additionally, MCF7 cells transfected with SLR20 harbored increased protein expression of several IFNβ-inducible chemokines known to recruit T lymphocytes, including chemokine (C-C) motif ligand (CCL)-3, CCL5, CCL13, C-X-C motif chemokine ligand 11 (CXCL11), lymphotactin/C-motif ligand (XCL1) and interleukin (IL)-8.

Although RIG-I–dependent anticancer immunity has been reported in several cancers, including pancreatic cancer, hepatocellular carcinoma (40), leukemias (43), and melanomas (31), little is known regarding RIG-I signaling in breast cancers. We used a synthetic agonist to activate the innate immune effector RIG-I in breast cancer cells in culture and in vivo, resulting in decreased tumor growth, decreased metastasis, increased TILs, and induction of tumor cell death via pyroptosis, an immunogenic form of cell death. These results suggest that RIG-I signaling remains intact in breast cancer cells and can be exploited to increase tumor cell death and, perhaps, tumor immunogenicity. Similar results have been reported from preclinical tumor models and clinical trials in cancers assessing the STING-mediated DNA/viral-sensing pathway (19, 20, 23), although STING signaling reportedly may be defective in a variety of cancers (24, 25). Nonetheless, these findings suggest that the developing field of RIG-I mimetics and the wider field of innate anticancer immunity may yield novel treatment strategies for breast cancers, which historically have not benefited to the same extent as other cancers from recent breakthroughs in immuno-oncology.

The data shown herein are the first (to our knowledge) showing that RIG-I signaling provides a therapeutic benefit in breast cancer cells, per se, and in a mouse model of breast cancer in vivo. Interestingly, a previous report identified RIG-I/DDX58 as belonging to an antimetastatic gene signature in breast cancer cells (44); however, the significance of this interesting finding remains unclear. Our findings here are consistent with a previous report using poly(I:C), a double-stranded RNA that binds to a RIG-I like receptor known as MDA5, as well as Toll-like receptor 3 (TLR3), to demonstrate that viral-sensing RNA helicases can activate proinflammatory signaling pathways that can be exploited therapeutically in breast cancers (45). We have built upon these important early studies by investigating models of each of the three major clinical subtypes of breast cancer, including poorly immunogenic ER⁺ breast cancers, and highly aggressive TNBCs (1). We have also studied the impact of RIG-I signaling in the context of an immune-competent mouse model, finding that RIG-I signaling not only induces tumor cell death and cytokine modulation, but also increases tumor infiltration by leukocytes, a significant finding given that increased TILs predict a better outcome for patients with breast cancer and correlates with increased response to ICIs (5, 8, 9).

Although proinflammatory cytokines, such as those induced by RIG-I signaling, support antitumor immunity, these are likely to have multifaceted effects on antitumor immunity, depending on their expression dose and duration. For example, type I IFNs promote DC maturation and T-cell priming during an antitumor immune response, but prolonged IFN exposure induces regulatory factors that restrain inflammation and antitumor immunity, such as PD-L1, IDO-1, and others. This has been shown in the context of therapeutic STING signaling in tumors, causing sustained induction of type I IFNs, which ultimately recruited immune-suppressive myeloid-derived suppressor cells (MDSC) to the tumor microenvironment (46). Further, studies demonstrating that RIG-I may respond to endogenous “unshielded” long noncoding RNAs (lnRNA) or genomically incorporated retroviral sequences, derived from neighboring tumor-associated fibroblasts and delivered to tumor cells through exosomal transport, may actually increase tumor cell growth, treatment resistance and malignant progression (47), despite production of inflammatory cytokines. These observations support further investigation into the longer-term consequences of RIG-I signaling in tumors. Future studies also need to consider the potential risk for unrestrained inflammation, because RIG-I is expressed in virtually all cell types, and RIG-I activation induces feed-forward signaling to amplify RIG-I and IFN-responsive genes. Recently described conditional RIG-I agonists, in which the 5′-triphosphorylated terminus of the RNA duplex remains shielded until release by predetermined molecular cues, may help enrich delivery of RIG-I agonist to the target tissue (48).

In summary, we demonstrate that RIG-I signaling induces immunogenic tumor cell death and upregulates expression of MHC-I components, proinflammatory cytokines, and chemokines in ER⁺ breast cancer cells, HER2⁺ breast cancer cells, and TNBC cells, resulting in inhibition of tumor growth and increased TILs in vivo. These findings suggest that RIG-I activation using a synthetic agonist activates innate immunity in breast cancer cells increases immunogenicity of breast cancers and may be a feasible treatment approach for treatment of breast cancers, including those with lower mutational burden that are considered poor candidates for immunotherapy.

No potential conflicts of interest were disclosed.

Conception and design: D.L. Elion, J.T. Wilson, R.S. Cook

Development of methodology: D.L. Elion, M.E. Jacobson, V. Sanchez, J.T. Wilson, R.S. Cook

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.L. Elion, E. Jacobson, D.J. Hicks, P.I. Gonzalez-Ericsson, O. Fedorova

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.L. Elion, B. Rahman, R.S. Cook

Writing, review, and/or revision of the manuscript: D.L. Elion, A.M. Pyle, J.T. Wilson, R.S. Cook

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.L. Elion, D.J. Hicks, B. Rahman, P.I. Gonzalez-Ericsson

Study supervision: D.L. Elion, D.J. Hicks, J.T. Wilson, R.S. Cook

We would like to acknowledge the shared resources at Vanderbilt University, Vanderbilt University Medical Center, and the Vanderbilt-Ingram Cancer Center that contributed to the studies reported herein, including the VICC Breast SPORE pathology service under the direction of Dr. Melinda Sanders, the VUMC Translational Pathology Shared Resource under the direction of Dr. Kellye Boyd, the Vanderbilt Digital Histology Shared Resource (DHSR) under the direction of Joseph Roland for access to a slide scanner and assistance with histology quantification. This work was supported by Specialized Program of Research Excellence (SPORE) grant NIH P50 CA098131 (VICC; to R.S. Cook, D.J. Hicks, V. Sanchez, and P.I. Gonzalez-Ericsson), Cancer Center Support grant NIH P30 CA68485 (VICC; to R.S. Cook, D.J. Hicks, V. Sanchez, and P.I. Gonzalez-Ericsson), CTSA UL1TR000445 (R. S. Cook) from the National Center for Advancing Translational Sciences, W81XWH-161-0063 (J.T. Wilson and R.S. Cook) from the Congressionally Directed Medical Research Program, and CBET-1554623 (J.T. Wilson) from the National Science Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.