You are here

  • Home
  • Archive
  • Volume 12, Issue 5
  • Colorectal cancer-specific IFNβ delivery overcomes dysfunctional dsRNA-mediated type I interferon signaling to increase the abscopal effect of radiotherapy

Article Text

Article menu
Basic tumor immunology
Original research
Colorectal cancer-specific IFNβ delivery overcomes dysfunctional dsRNA-mediated type I interferon signaling to increase the abscopal effect of radiotherapy
  1. Kevin Chih-Yang Huang1,2,3,
  2. Shu-Fen Chiang4,
  3. Hsin-Yu Chang2,5,
  4. Wei-Ze Hong5,
  5. Jhen-Yu Chen1,2,5,
  6. Pei-Chih Lee3,6,
  7. Ji-An Liang7,8,
  8. Tao-Wei Ke9,10,
  9. Shin-Lei Peng1,
  10. An‑Cheng Shiau1,5,7,
  11. Tsung-Wei Chen11,
  12. Pei-Chen Yang5,
  13. William Tzu-Liang Chen8,9,12 and
  14. K S Clifford Chao5,7,8
  1. 1Department of Biomedical Imaging and Radiological Science, China Medical University, Taichung, Taiwan
  2. 2Translation Research Core, China Medical University Hospital, Taichung, Taiwan
  3. 3Cancer Biology and Precision Therapeutics Center, China Medical University, Taichung, Taiwan
  4. 4Lab of Precision Medicine, Feng-Yuan Hospital Ministry of Health and Welfare, Taichung, Taiwan
  5. 5Proton Cancer, China Medical University Hospital, Taichung, Taiwan
  6. 6Graduate Institute of Biomedical Science, China Medical University, Taichung, Taiwan
  7. 7Department of Radiation Oncology, China Medical University Hospital, Taichung, Taiwan
  8. 8School of Medicine, China Medical University, Taichung, Taiwan
  9. 9Department of Colorectal Cancer, China Medical University Hospital, Taichung, Taiwan
  10. 10School of Chinese Medicine, China Medical University, Taichung, Taiwan
  11. 11Department of Pathology, Asia University, Taichung, Taiwan
  12. 12Department of Colorectal Surgery, China Medical University HsinChu Hospital, China Medical University Hospital, HsinChu, Taiwan
  1. Correspondence to Professor Kevin Chih-Yang Huang; flylerd0425{at}gmail.com; Dr K S Clifford Chao; d94032{at}mail.cmuh.org.tw

Abstract

Background Cancer-intrinsic type I interferon (IFN-I) production triggered by radiotherapy (RT) is mainly dependent on cytosolic double-stranded DNA (dsDNA)-mediated cGAS/STING signaling and increases cancer immunogenicity and enhances the antitumor immune response to increase therapeutic efficacy. However, cGAS/STING deficiency in colorectal cancer (CRC) may suppress the RT-induced antitumor immunity. Therefore, we aimed to evaluate the importance of the dsRNA-mediated antitumor immune response induced by RT in patients with CRC.

Methods Cytosolic dsRNA level and its sensors were evaluated via cell-based assays (co-culture assay, confocal microscopy, pharmacological inhibition and immunofluorescent staining) and in vivo experiments. Biopsies and surgical tissues from patients with CRC who received preoperative chemoradiotherapy (neoCRT) were collected for multiplex cytokine assays, immunohistochemical analysis and SNP genotyping. We also generated a cancer-specific adenovirus-associated virus (AAV)-IFNβ1 construct to evaluate its therapeutic efficacy in combination with RT, and the immune profiles were analyzed by flow cytometry and RNA-seq.

Results Our studies revealed that RT stimulates the autonomous release of dsRNA from cancer cells to activate TLR3-mediated IFN-I signatures to facilitate antitumor immune responses. Patients harboring a dysfunctional TLR3 variant had reduced serum levels of IFN-I-related cytokines and intratumoral CD8+ immune cells and shorter disease-free survival following neoCRT treatment. The engineered cancer-targeted construct AAV-IFNβ1 significantly improved the response to RT, leading to systematic eradication of distant tumors and prolonged survival in defective TLR3 preclinical models.

Conclusion Our results support that increasing cancer-intrinsic IFNβ1 expression is an immunotherapeutic strategy that enhances the RT-induced antitumor immune response in locally patients with advanced CRC with dysfunctional TLR3.

  • Radiotherapy
  • Abscopal effect
  • Colorectal Cancer
  • Toll-like receptor 3

Data availability statement

Data are available on reasonable request. The original dataset is available on request from the corresponding author.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/jitc-2023-008515

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • Currently, cancer-intrinsic type I interferon (IFN-I) production triggered by radiotherapy (RT) is known to depend mainly on cytosolic double-stranded DNA (dsDNA)-mediated cGAS/STING signaling to increase cancer immunogenicity and enhance antitumor immunity. cGAS/STING deficiency may limit the RT-induced antitumor immune response and the therapeutic efficacy of RT.

    WHAT THIS STUDY ADDS

    • This study revealed that RT induces cytosolic dsRNA accumulation and release, and this dsRNA can be recognized by TLR3 to promote type I IFN signaling in colorectal cancer. In patients with defective TLR3, RT-induced antitumor immunity may be attenuated, leading to poor response and survival outcomes. Colorectal cancer-specific overexpression of IFNβ1 significantly overcame the attenuation of defective TLR3-mediated IFN-I production, thereby improving the response to RT, leading to systematic eradication of distant tumors and prolonged survival in preclinical models.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • This study elucidates the important role of TLR3 in regulating RT-induced dsRNA-mediated type I IFN signaling to affect antitumor immunity in colorectal cancer.

    Background

    Radiotherapy (RT) can initiate a tumor-specific immune response with systemic effects to eradicate local tumors as well as distant, non-irradiated metastatic tumors, improving the survival outcome of patients with cancer.1 2 RT has been found to remodel the tumor microenvironment (TME) by converting an immunologically “cold” tumor into a “hot” tumor through immunogenic cell death (ICD).3 These processes release damage-associated molecular patterns as immune adjuvants for the recruitment of immune cells and activation of tumor-specific adaptive immunity.4 5 The therapeutic efficacy of RT is dependent mainly on the secretion of type I interferon (IFN-I) by irradiated cancer cells, which are activated by the cytosolic double-stranded DNA (dsDNA) sensor cyclic GMP-AMP synthase (cGAS)/STING signaling.2 6 However, the downregulation of cGAS/STING in colorectal cancer (CRC) may limit the RT-induced antitumor immune response.7

    Apart from cGAS/STING-mediated IFN-I production, other dsRNA sensors, such as toll-like receptor 3 (TLR3), MDA5 and RIG-1, have also been reported to elicit IFN-I production.8 Recent studies have revealed the role of TLR3-mediated IFN-I signaling in antitumor immunity. For instance, the chemotherapeutic drug doxorubicin has been shown to induce TLR3 activation to trigger ICD in breast cancer cells,9 10 leading to autocrine/paracrine IFN-I signaling and subsequent secretion of CXCL10.9 High TLR3 and CXCL10 levels are prognostic factors and are predictive of favorable 5 year survival outcomes for patients with esophageal squamous cell carcinoma who receive adjuvant chemotherapy after surgery.11 Additionally, the TLR3 polymorphism (L412F/rs3775291) and a low IFN-I signature are considered unfavorable prognostic factors for patients with breast cancer undergoing chemotherapy and for patients with CRC.9 12 Moreover, several studies have indicated that DNA demethylating agents can induce a “viral mimicry state” to enhance antitumor immunity via cytosolic dsRNA accumulation.13–15 Targeting EZH2 can also activate a dsRNA-STING-ISG stress response to augment antigen presentation in the context of immune checkpoint blockade (ICB) treatment.16 The increased dsRNA content induced by RT can enhance the response to immunotherapy in poorly immunogenic cancers such as microsatellite stable CRC (MSS-CRC) and pancreatic adenocarcinoma (PDAC),14 15 suggesting that dsRNA may be another significant factor in the tumor response to RT.

    In this study, we observed a significant increase in tumor IFNβ following RT in cGAS/STING-deficient CRC, indicating that cytosolic dsRNA accumulation may activate IFN-I signaling to promote antitumor immunity. We found that inhibition of TLR3 activation using genetic or pharmacological methods significantly reduced RT-induced IFNβ and CXCL10 production in cGAS/STING-deficient CRC, resulting in a decreased density of CD8+ tumor-infiltrating lymphocytes (TILs) in vivo. These findings suggest that RT-induced dsRNA accumulation promotes IFN-I signaling for antitumor immunity via TLR3. Additionally, we found that patients with advanced CRC bearing a TLR3 variant (L412F/rs3775291) exhibited poor therapeutic response to neoCRT and worsened survival outcomes. CRC-specific IFNβ1 overexpression mediated by adenovirus-associated virus (AAV) transduction significantly augmented the therapeutic response to RT by promoting high infiltration of dendritic cells, activating IFNγ+CD8+ T cells, and activating IFNγ+ NK cells. Furthermore, CRC-specific IFNβ1 overexpression combined with RT significantly delayed non-irradiated (abscopal) tumors, indicating that the immunomodulation of tumor IFNβ potentiates the response to RT by promoting the antitumor immune response. These findings suggest that the TLR3 variant can serve as an independent biomarker to stratify patients with CRC according to the potential benefit of neoCRT. Moreover, IFNβ-based immunotherapy may be an alternative therapeutic strategy to overcome the impact of defective TLR3 and promote the antitumor response to RT in CRC.

    Methods

    Cell culture, treatment, western blot analysis

    The CRC cell lines SW480 (cGAS-deficient), HCT116 (cGAS-deficient), HT29 and CT26 were obtained from the American Type Culture Collection (Virginia, USA).17 The cells were authenticated, cultured and maintained in RPMI1640 medium supplemented with 10% fetal bovine serum (Thermo Fisher Scientific, California, USA), 2 mM glutamine, 100 U/mL penicillin, 100 mg/mL streptomycin, and 1 mM sodium pyruvate at 37°C in a humidified, 5% CO2 atmosphere in limited passage (<10 passage). For radiation experiments, SW480 and HCT116 cells were irradiated and harvested at the indicated times. The cell lysates and medium were then analyzed by immunoblotting, and the results were quantified. HCT116 cells were treated with 2.5 µg/mL poly (I:C) (Sigma, Missouri, USA) or 10 µM selective TLR3 inhibitor CU-CPT4a (Tocris, Pennsylvania, USA) for 24 hours. All reagents not otherwise mentioned were purchased from Sigma. National RNAi Core Facility at Academia Sinica in Taiwan for providing shRNA, Crispr-Cas9 reagents and related services to generate knockdown and knockout cell lines.

    The total lysates (30 µg) were separated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE; 6%–12% resolving gel) and were electrophoretically transferred onto a PVDF membrane (GE Healthcare, Amersham, UK). The membranes were blocked with 5% non-fat milk and probed with specific antibodies overnight at 4°C. Then, horseradish peroxidase (HRP)-conjugated secondary antibodies (1:10,000, GE Healthcare) were applied to the membranes, followed by detection using the Immobilon Western Chemiluminescent HRP Substrate (Millipore, Massachusetts, USA) and analysis by an ImageQuant LAS 4000 biomolecular imager (GE Healthcare). The digital images were processed with Adobe Photoshop V.7.0 (Adobe Systems, California, USA). Each blot was stripped using Restore western blot Stripping Buffer (Pierce, Iowa, USA) and incubated with the other antibodies. The results were assessed using ImageJ software (NIH, Maryland, USA). The following antibodies were used: cGAS (ab224144, Abcam, Cambridge, UK), STING (#13647, Cell Signaling Tech., California, USA), GAPDH (Ab9485, Abcam), RIG1 (A0550, Abclonal, Massachusetts, USA), TLR3 (ab62566, Abcam), MDA5 (A2203, Abclonal), MAVS (A5764, Abclonal), beta-actin (IR2-7, iREAL Biotech.), p-IRF3S388 (AP1333 and AP0263, Abclonal, IRF3 (#41075, Cell Signaling Tech.), p-STAT1T701 (bs-1657R, Bioss antibodies and AP0054, Abclonal) and STAT1 (sc-464, Santa Cruz, California, USA).

    For T cell migration assay, cancer cells were seeded in the lower well of transwell (6 µm pore size) for 24 hours, and then CSFE-labeled SupT1 (5 µM) cells were seeded in the upper wells for 6 hours. The migrated T cells were examined by flow cytometry.

    RNA extraction, qRT-PCR and RNA analysis

    HCT116 and SW480 cells were treated with 2.5 µg/mL poly (I:C) (Sigma) or 10 µM selective TLR3 inhibitor CU-CPT4a for 24 hours. Total RNA was extracted using TRIzol reagent (Zymo Research, California, USA) according to the manufacturer’s instructions. Briefly, 1 µg of total RNA was incubated with 0.5 µg of oligo dT (MD.Bio., Taipei, Taiwan) at 70°C for 15 min. Then, the RNA was mixed with buffer containing 0.25 mM dNTPs (MD. Bio.), 20 U of RNasin I Plus RNase Inhibitor (Promega, Wisconsin, USA) and 20 U of M-MLV Reverse Transcriptase (Promega) and incubated at 42°C for 90 min for cDNA synthesis. This mixture was then used for specific cDNA amplification in a GeneAmp PCR system 2400 (Perkin Elmer, Waltham, Massachusetts, USA).18

    For mRNA quantification, real-time PCR was performed using a standard Rotor-Gene Q SYBR Green I Master protocol on a Rotor-Gene Q SYB system (Qiagen, QIAGEN Taiwan, Taipei, Taiwan). The 25 µL PCR included 2 µL cDNA, 12.5 µL 2×SyberGreen PCR Mix, 0.5 µL 10 µM forward primer, 0.5 L 10 µM reverse primer and 9.5 µL ddH2O. All reactions were run in triplicate. The cycle number at which the reaction crossed the threshold cycle (Ct) was determined for each gene, and the relative amount of each gene to GAPDH was described using the Equation 2-ΔΔCt, where ΔCt = (Ctinterested gene – CtGAPDH) and ΔΔCT = ΔCT(a target sample)−ΔCT(a reference sample)

    For dsRNA analysis, the total RNA (50 µg) of each sample and the dsRNA marker (Abnova, California, USA) were electrophoresed on a 7.5% non-denaturing polyacrylamide gel at 100 V. The gel was run at 18 W for 2–3 hours.19

    Indirect immunofluorescence and confocal microscopy

    Cells were fixed with 4% paraformaldehyde for 15 min at room temperature and permeabilized with 0.1% Triton X-100 for 15 min at room temperature. Fixed and permeabilized cells were blocked with 2.5% BSA for 15 min and then incubated overnight at 4°C with diluted MabJ2 in PBS/1% BSA (1:500 dilution of the 0.2 mg/mL antibody), which is a specific antibody for dsRNA (clone J2, MABE1134, Merck Sigma, California, USA). Then, the cells were washed and stained with Alexa 546 rabbit anti-mouse IgG secondary antibodies (Thermo Fisher Scientific). Images were captured using a Leica SP2 Confocal Spectral Microscope. The images were processed using Adobe Photoshop.

    Bioinformatic analysis

    Raw messenger RNA (mRNA) expression profiles and clinical features of GSE46862 and GSE15781 were downloaded from the GEO database (http://www.ncbi.nlm.nih.gov/geo/). Gene set enrichment analysis (GSEA) was performed on the The Cancer Genome Atlas (TCGA) dataset using a software package downloaded from www.broadinstitute.org/gsea.20 The number of permutations was set to 1000. The relationship between CXCL10 mRNA and survival outcome in the TCGA-COAD dataset was determined using the online GEPIA platform (http://gepia.cancer-pku.cn/).21

    Treatment with established tumors in animals

    BALB/c mice (female, 4 weeks old) were maintained in specific pathogen-free conditions in a temperature-controlled environment with 12 hours light, 12 hours dark cycles and received food and water ad libitum according to the institutional guidelines. Four-week-old female BALB/c mice were obtained from BioLASCO Taiwan Co. (Taipei, Taiwan). CT26 cells (3×105 cells/mouse) were suspended in 50% 100 µL Matrigel matrix and injected subcutaneously into the right leg of each mouse. On day 10, when tumors were 100 mm3, animals were randomly assigned to three groups receiving or not receiving local RT (5 Gy×2 fractions), in combination or not in combination with 0.4 mg/mouse TLR3/dsRNA complex inhibitor CU-CPT4a (1 hour before and after RT by intraperitoneal injection).

    The tumor volume was measured every 3 days and was calculated according to the formula (width2×length)/2. The mice were sacrificed at the termination of the experiments, and tumor tissues from representative mice were collected for lysis, subjected to immunoblotting analysis and stained by immunohistochemistry (IHC). A tumor size of 3000 mm3 was considered the endpoint of survival time.

    Therapeutic studies of recombinant IFNβ protein in a mouse CRC model

    CT26 cells (3×105 cells/mouse) were subcutaneously injected into the right leg of BALB/c mice on day 0 and received local RT (5 Gy×2 fractions) in combination with 0.4 mg/mouse TLR3/dsRNA complex inhibitor CU-CPT4a (1 hour before and after irradiation by intraperitoneal injection) on days 7, 10, and 13. On days 8, 11 and 14, 10,000 U of mouse IFNβ (BioLegend, California, USA) was intratumorally administered. The antitumor effect was determined by measuring the tumor volume every 3 days. Mice were sacrificed when tumors reached 3000 mm3.

    Patient characteristics, clinical staging, treatment, and pathological evaluation

    211 patients with locally advanced rectal cancer were treated at our hospital from 2006 to 2014. Among these patients, 171 received neoCRT followed by surgery. Patients with biopsy-proven locally advanced rectal cancer (cT3-4 or cN+by endorectal ultrasonography (EUS), CT, or MRI) who were treated with preoperative chemoradiotherapy followed by radical resection at China Medical University Hospital comprised the study cohort. Tumors were staged based on the American Joint Committee on Cancer staging system. EUS, MRI or CT was used to assess the pretreatment clinical stage, and pretreatment biopsies were reviewed by pathologists, as previously described.22 The tumor regression grade (TRG) system was used according to Dworak’s criteria.23

    Patients were treated with chemoradiotherapy with a median RT dose of 50.4 Gy in 28 fractions and concurrent fluoropyrimidine-based chemotherapy (mainly single-agent capecitabine, daily 500 mg/m2/two times per day, orally administered 7 days/week during a conventional RT period of approximately 6 weeks). Patients were assessed for their clinical response 6–8 weeks after the completion of neoCRT according to rigorous criteria of clinical, endoscopic, and radiologic findings. The three criteria for complete clinical response (cCR) were (a) the absence of a residual ulceration, mass, or mucosal irregularity on clinical/endoscopic assessment; (b) whitening of the mucosa and the presence of neovasculature; and (c) radiologic imaging, such as CT, EUS, or MRI, without evidence of extrarectal residual disease.

    Construction of the tissue microarray

    Tissue microarrays (TMA) were constructed from 133 pair-matched pre-neoCRT biopsies and post-neoCRT surgical tissue from patients with rectal cancer, and other specimens were not available (materials were not suitable for IHC or SNP genotyping, online supplemental table S1).24 Areas of tumor cells were marked on H&E-stained slides. The corresponding area on the matching paraffin block (donor block) was then identified and marked. We used the AutoTiss 10C system (EverBio Technology, Taipei, Taiwan) to remove the tissue core from these areas of the donor blocks into the recipient block in a precise, arrayed fashion. The punches were 2 mm in diameter, and a maximum of 60 punches were placed on a single block. Sample sections cut on a microtome were then mounted on capillary-gap slides (Dako, Hamburg, Germany).

    Supplemental material

    Immunohistochemistry

    IHC was performed using 3-μm-thick histological TMA sections. TMA slides were stained individually with HRP-conjugated avidin biotin complex (ABC) using the Vectastain Elite ABC-HRP Kit (PK-6100, Vector Laboratories, California, USA) and NovaRed chromogen (Vector Laboratories) and counterstained with hematoxylin.25 Staining of tumor STING (#13647, Cell Signaling Tech.) and IFNβ (ab85803, Abcam) was evaluated based on immunopositivity of cells in the cytoplasm or surface of tumor cells for histo-score (H-score) according to the intensity by semiquantitative scale (0 for absent; 1 for weak; 2 for moderate; and 3 for strong staining).26 The range of the H-score was from 0 to 300. For cytosolic dsRNA evaluation, anti-dsRNA antibodies (J2 clone) were used to evaluate tumor cells according to Histo-score (0 for absent; 1 for weak; 2 for moderate; 3 for strong and 4 for super strong staining). For CD8+ tumor-infiltrating immune cells (1:500 for 2 hours at room temperature, ab4055, Abcam, Cambridge, UK), the count of intraepithelial immunopositive immune cells was evaluated under 40X microscopy. The counts of TILs are presented as the number of immune cells/mm2. The sample data were not included if the residual tissue was not enough to evaluate. Pearson’s correlation between tumor IFNβ and cytosolic dsRNA levels was analyzed by GraphPad Prism V.8.0 (GraphPad Software, Massachusetts, USA) according to their Histo-score (H-score).27

    Genomic DNA extraction and SNP genotyping

    Genotyping of SNPs was performed using the iPLEX HS panel on the MassARRAY System (Agena Bioscience, California, USA), which employs matrix-assisted laser desorption/ionization time-of-flight mass spectrometry for amplicon detection (SpectroACQUIRE, Agena Bioscience). Primers designed for PCR amplification of specific polymorphisms in TLR3 and extension reactions were prepared using MassARRAY Assay Design V.3.1 software (Agena Bioscience).

    Genomic DNA from non-tumor tissues of patients with rectal cancer was extracted from two 3 μm thick Formalin-fixed paraffin-embedded (FFPE) slides using a QuickExtract FFPE DNA Extraction Kit (QEF81050, Epicenter, Wisconsin, USA). For SNP genotyping, 10 ng of total genomic DNA was used for PCR amplification. The PCR samples contained Taq DNA polymerase (Agena Bioscience), genomic DNA (5–10 ng), PCR primers (online supplemental table S1), and dNTPs. Following standard protocols for PCR (95°C 5 s→58°C 15 s→68°C 30 s, 45 cycles), the remaining dNTPs were removed by the addition of alkaline phosphatase (Agena Bioscience), and the plates were subsequently incubated at 37°C for 40 min. Following PCR, SAP addition, and the iPLEX HS extension reaction, the samples were desalted by resin treatment for 15 min, spotted onto SpectroCHIP Arrays (Agena Bioscience), analyzed by mass spectrometry, and ultimately interpreted using SpectroTYPER V.4.0 software (Agena Bioscience).

    CRC-specific AAV2 vector generation and recombinant AAV2 virus purification

    The human IFNβ1 (aa 22-187) sequence was generated by PCR and subcloned and inserted into a pAAV2-carcinoembryonic antigen (CEA) vector (AAV-Vec), which contained the CEA promoter. The viruses were all produced using the triple transfection method, AAV-IFNβ1 and the helper plasmids pRC2-miR342 and pHelper in 293 T cells. 72 hours after transfection, the cells were collected by centrifugation, and recombinant AAV2 vectors were produced and purified using an AAVpro purification kit (#6232, Takara, Japan). AAV2 titration was quantitated by real-time PCR of vector genomes based on amplification of inverted terminal repeats (ITR) according to the manufacture’ manual (#6233 AAVpro titration kit, Takara, Japan).28 AAV-IFNβ1 was administered via intramuscular injections in the quadriceps by delivering a volume of 50 µL per side at 5×109 vg on the indicated days. Mice were sacrificed when any side of the tumor reached 20 mm.

    Flow cytometry analysis of immune cell profiles

    Tumors were dissected from the mice, weighed, and then placed in petri dishes containing blank RPMI media at room temperature to prevent dehydration. Tumors (0.3 g) were minced into small pieces with a beaver blade and homogenized by a Minilys personal homogenizer (Bertin Technologies, France), filtered through a 70 µm strainer, centrifuged, and then resuspended in blank RPMI media. Thereafter, the cell suspensions were layered over Ficoll-Paque media and centrifuged at 1025×g for 20 min. The layer of mononuclear cells was transferred into a conical tube, 20 mL of complete RPMI media was added and then gently mixed, and the sample was centrifuged at 650 g for 10 min twice. Finally, the supernatant was removed, and the TILs were resuspended in complete RPMI media.

    Then, TILs were resuspended in 500 µL of staining buffer (2% BSA, FcR blocker, and 0.1% NaN3 in PBS). The cells were stained with different surface marker panels: (1) T cells: CD45-PerCP/Cy5, CD3-APC/Fire750, CD4-FITC, CD8-PE/Cy7, CD44-PE, CD62L-APC; (2) Dendritic cells: CD45-APC/Fire750, CD3/CD19-FITC, CD11c-APC, MHC-II/PE/Cy7 CD11b-AF488 and CD8-PE; (3) IFNγ+ CD8+ T/NK cells: CD45-APC/Fire750, CD3-FITC, CD8a-PE/Cy7, CD49b-APC and IFNγ-PE and (4) FoxP3 regulatory T cells: CD3-PerCP/Cy5, CD45-PE/Cy7, CD4-APC, CD25-PE, FoxP3-AF488 and CD127-AF647. For intracellular FoxP3 staining, TILs were fixed and then permeabilized with True-Nuclear fixation and permeabilization buffer (BioLegend) for 45 min after cell surface staining. Cells were then stained with mixed antibodies for the FoxP3 regulatory T-cell panel in True-Nuclear permeabilization buffer for 30 min and washed twice with staining buffer. For IFNγ+ CD8+ T/NK cell staining, TILs were first stimulated with PMA (2 ng/mL), ionmycin (1 µg/mL) and BFA (1:1000) for 3 hours at 37°C. After simulation, TILs were washed and resuspended in staining buffer, and mixed antibodies were added to the IFNγ+ CD8+ T/NK cell panel. Then, the cells were washed, fixed with fixation buffer and permeabilized with permeabilization buffer for 20 and 5 min. The cells were then stained with mixed antibodies in permeabilization buffer for 20 min, washed twice with staining buffer, and analyzed with a Guava easyCyte Flow Cytometer (Luminex, California, USA). Isotype controls were used, including PerCP-conjugated rat IgG2b κ isotype control, APC-conjugated rat IgG2b, and PE-conjugated rat IgG2b κ isotype control. These antibodies were purchased from BioLegend.

    Library preparation, RNA sequencing and bioinformatics analysis

    Total RNA from resected tumor tissues were isolated by TRIzol reagent and purified RNA was used for the preparation of the sequencing library by TruSeq Stranded mRNA Library Prep Kit (Illumina, California, USA) following the manufacturer’s recommendations. Briefly, mRNA was purified from total RNA (1 µg) by oligo(dT)-coupled magnetic beads and fragmented into small pieces under elevated temperature. The first-strand cDNA was synthesized using reverse transcriptase and random primers. After the generation of double-strand cDNA and adenylation on 3’ ends of DNA fragments, the adaptors were ligated. The products were enriched with PCR and purified with AMPure XP system (Beckman Coulter, New Jersey, USA). The libraries were qualified by Qsep400 System (Bioptic, Taiwan) and quantified by Qubit 2.0 Fluorometer (Thermo Scientific). The qualified libraries were then sequenced on an Illumina NovaSeq 6000 platform with 150 bp paired-end reads generated by Genomics BioSci & Tech. (New Taipei City, Taiwan).

    The bases with low quality and sequences from adapters in raw data were removed using program fastp (V.0.20.0). The filtered reads were aligned to the reference genomes using HISAT2 (V.2.1.0). Differentially expressed genes were identified by DESeq2 (V.1.28.0) depends on whether having biological replicates or not. The functional enrichment analysis of Gene Ontology terms among gene clusters was implemented in an R package called clusterProfiler (V.4.0.0). For GSEA analysis, we used the GSEA tool V.3.2, with the MSigDB V.6.2 Hallmarks gene sets collection and the “classic” method for calculating enrichment scores.

    Statistical analysis

    SAS statistical software V.9.4 (SAS Institute) was used to perform the statistical analysis. All tests reported a two-sided p value with a significance level set at 0.05. Data were analyzed using one-way analysis of variance, Spearman’s correlation, and Kaplan-Meier estimations. Student’s t-test, Pearson’s χ2 and Fisher’s exact test were used for group comparisons. Cox regression analysis was used to estimate the HRs and 95% CIs for univariate and multivariate models.25 Influential factors that affected the rectal cancer patient survival rate were adjusted in the Cox models, including TRG (good response (TRG 3–4) vs poor response (TRG 1–2)), clinical response (good response (CR and partial response) vs poor response (stable disease and disease progression), and pN stage (positive vs negative). Statistical analysis was performed using SPSS V.20.0 software and GraphPad Prism V.8.0 (GraphPad Software).

    Results

    RT increases cytosolic dsRNA accumulation to activate TLR3 to promote IFN-I production

    The induction of IFN-I production for adaptive tumor-specific immunity is considered the central dogma of the immunologic effects of RT.2 7 29 To investigate whether neoCRT activates the IFN-I signaling cascade, we analyzed the mRNA data from paired pre-neoCRT biopsies and post-neoCRT surgical tissues retrieved from the GSE15781 dataset (figure 1A).30 Our results revealed a significant increase in IFN-I signatures after neoCRT treatment. Additionally, we identified gene sets associated with the immunological effects of neoCRT in patients with CRC who underwent neoCRT by GSEA (data retrieved from GSE15781). Among the top 20 upregulated pathways, 6 were immune-related gene sets, including HALLMARK_INFLAMMATORY_RESPONSE and HALLMARK_INTERFERON_ALPHA_RESPONSE and were significantly enriched (figure 1A). These findings indicate that type I IFN signatures are involved in the neoCRT-induced antitumor immune response.

    Figure 1
    Figure 1

    Radiotherapy induces IFN-I signaling in cGAS/STING-deficient cells. (A) The RNA expression of type I IFN-responsive genes from pre-neoCRT biopsy tissues (n=13) and six post-neoCRT surgical tissues (n=9) were retrieved from the GSE15781 dataset. (B) Schematic of the dsRNA and dsDNA sensing pathway for type I IFN production. (C) Paired-matched mRNA levels of genes related to dsRNA and dsDNA sensing pathways were shown (GSE15781, Wilcoxon matched-pairs signed rank test, n=12). *p<0.05. (D) Representative images of IFNβ1 and STING protein in pre-neoCRT biopsies and post-neoCRT surgical tissues. (E) Tumor IFNβ was measured by IHC and scored as Histo-score in post-neoCRT surgical tissues (n=133, paired t-test). Kaplan-Meier curves for overall survival in individuals with high or low tumor IFNβ in a cohort received neoCRT (n=133, log-rank test). The median H-score was defined as the cut-off. (F) Tumor STING was measured by IHC and scored as Histo-score in pre-neoCRT biopsies and post-neoCRT surgical tissues (n=108, paired t-test). (G) The association between STING and IFNβ within post-neoCRT surgical tissues (n=108). (H) The expression of cGAS and STING in colorectal cancer cell lines. (I) The colorectal cancer cell lines SW480, HCT116 (cGAS-deficient cells) and HT29 (cGAS/STING-proficient cells) were irradiated with 5 and 10 Gy for 48 hours. The level of CXCL10 was analyzed by qRT-PCR (n=3). ***p<0.001. One-way ANOVA test. (J) HT29 cells were infected with lentivirus carrying shcGAS and shSTING and selected by puromycin for stable cell lines. HT29shNC, HT29shcGAS, and HT29shSTING cells were irradiated with 5 Gy and harvested after 48 hours. The levels of IFNβ1 and CXCL10 were analyzed by qRT-PCR (n=3). ***p<0.001. One-way ANOVA test. (K) HT29 isogenic STING knockout cells (HT29-STINGKO) constructed with STING gRNA and HT29-WT cells constructed with non-targeting (NT) gRNA. HT29-WT and HT29-STINGKO cells were irradiated with 5 Gy and harvested after 24 hours. The levels of IFNβ1 were analyzed by qRT-PCR (n=3). **p<0.01, ***p<0.001. One-way ANOVA test. ANOVA, analysis of variance; dsDNA, double-stranded DNA; IHC, immunohistochemistry; qRT-PCR, quantitated by real-time PCR.

    IFN-I production is primarily dependent on the activation of TBK1/IRF3 signaling by dsRNA and dsDNA sensors (figure 1B).1 We observed significant upregulation of the dsRNA sensors TLR3 and MDA5, the dsDNA sensor IFI16, and the downstream genes IFNβ1 and CXCL10 in post-neoCRT surgical tissues (figure 1C). However, cGAS/STING expression is downregulated in various tumor tissues, including CRC.7 26 31 We found that even in the absence of STING expression, tumor cells from patients with CRC exhibited high levels of IFNβ after neoCRT treatment (figure 1D). High expression of IFNβ induced by neoCRT was associated with favorable survival outcomes (figure 1E,F). Although tumor STING expression was also upregulated by neoCRT (figure 1E, p=0.017), it remained relatively low. Moreover, we found that approximately 57.7% of STING-low patients exhibited high tumor IFNβ expression after neoCRT treatment (41/41+30=57.7%, figure 1G), suggesting that other molecular mechanisms, such as dsRNA sensors, may participate in IFN-I production induced by neoCRT.32

    We then irradiated HT29 (wild-type cGAS/STING), HCT116 (cGAS-deficient) and SW480 (cGAS-deficient) CRC cell lines to evaluate the production of IFN-I.7 We observed a significant increase in CXCL10 levels in all three cell lines (figure 1H,I). The upregulation was greater in the HT29 cell line than in the other two cell lines. Knockdown of cGAS and STING in HT29 cells resulted in a significant reduction in the mRNA levels of IFNβ1 and CXCL10 (figure 1J). However, partial changes in the mRNA levels of IFNβ1 and CXCL10 were detected, indicating that radiation may trigger IFN-I and CXCL10 production via cGAS/STING-independent mechanisms. Similarly, we found that the levels of IFNβ1 and CXCL10 were increased by radiation in HT29-STINGKO isogenic cells, demonstrating that alternative mechanisms involving IFN-I/CXCL10 were triggered by radiation.

    An increase in the level of cytosolic dsRNA species induced by RT can enhance the response to immunotherapy in MSS-CRC and PDAC patients.14 15 By gel electrophoresis and cytosolic dsRNA quantification, we found that irradiation led to cytosolic dsRNA accumulation with a 4–5 fold change (figure 2A). Furthermore, immunofluorescence staining with specific dsRNA antibodies (J2 clone) also revealed intense dsRNA signals in CRC cells after irradiation (figure 2B and online supplemental figure S1A).19 To investigate whether dsRNA is released into the supernatant for dsRNA sensor activation, we incubated conditioned media (CM) from irradiated CRC cells. As shown in figure 2C,D, RT-CM triggered IFNβ1 and CXCL10 upregulation. The addition of dsRNA-specific RNase III significantly reduced CM-induced IFNβ1 and CXCL10 upregulation (figure 2C,D). Notably, the single-stranded RNA-specific RNase A partially alleviated RT-CM-induced IFNβ1 and CXCL10 upregulation (figure 2C,D), suggesting that RT promotes dsRNA accumulation for IFN-I production. To further demonstrate RT-induced dsRNA accumulation, we evaluated the degree of dsRNA by specific dsRNA antibodies (J2 clone) in post-neoCRT surgical tissues from patients with rectal cancer (figure 2E). Consistently, we detected high levels of dsRNA signals, especially in good responders after neoCRT treatment, compared with poor responders (figure 2E). Furthermore, the level of dsRNA intensity was positively correlated with the level of tumor IFNβ1 after neoCRT treatment (Pearson’s correlation, p=0.0012, r=0.27; figure 2F).

    Figure 2
    Figure 2

    Radiotherapy (RT) enhanced cytosolic dsRNA accumulation for TLR3-mediated type I IFN production. (A) HCT116 cells were irradiated for 24 hours, and then RNA was extracted for electrophoresis (7.5% polyacrylamide gel electrophoresis). HCT116 cells were irradiated for 24 and 48 hours, and the content of cytosolic dsRNA was measured by dsRNA ELISA kit (n=3). *p<0.05, **p<0.01. One-way ANOVA test. (B) SW480 cells were irradiated with 5 Gy for 24 hours, and the level of dsRNA was observed by immunofluorescence staining. *p<0.05 Unpaired t-test. (C) SW480 cells were treated with conditioned medium for 24 hours and then examined by qRT-PCR (n=3). The conditioned medium was collected from irradiated cells and then incubated with RNase A and RNase III for 2 hours. The level of IFNβ1 was analyzed by qRT-PCR (n=3). ***p<0.001. One-way ANOVA test. (D) SW480 cells were treated with conditioned medium for 24 hours and then examined by qRT-PCR (n=3). The conditioned medium was collected from irradiated cells and then incubated with RNase A and RNase III for 2 hours. The level of CXCL10 was analyzed by qRT-PCR (n=3). **p<0.01. One-way ANOVA test. (E) Representative images of cytosolic dsRNA in pre-neoCRT biopsies and post-neoCRT surgical tissues. (F) The correlation between cytosolic dsRNA and tumor IFNβ expression was measured (non-linear regression model, p=0.0012, r=0.2794, n=131). (G) SW480 cells were infected with lentivirus carrying shRNA against RIG1, MDA5 and MAVS. The knockdown efficacy was examined by western blotting. (H) SW480shNC, SW480shRIG1, SW480shTLR3, SW480shMDA5 and SW480shMAVS cells were irradiated with 5 Gy. After 24 hours, the cells were harvested, and the mRNA levels of IFNβ1 and CXCL10 were examined by qRT-PCR (n=3). *p<0.05, **p<0.01. One-way ANOVA test. (I) HCT116 cells were infected with lentivirus carrying shRNA against TLR3. The knockdown efficacy was measured by immunoblotting. HCT116shNC, HCT116shTLR3#1 and HCT116shTLR3#2 cells were irradiated with 5 and 10 Gy. After 24 hours, the cells were harvested, and the mRNA levels of IFNβ1 and CXCL10 were examined by qRT-PCR (n=3). **p<0.01, ***p<0.001. One-way ANOVA test. (J) HCT116shNC and HCT116shTLR3 cells were with or without irradiation with 5 Gy. After 24 hours, CFSE-labeled SupT1 T cells were seeded in the upper wells, and the migrated T cells were examined by flow cytometry (n=3). **p<0.01, ***p<0.001. One-way ANOVA test. (K) HCT116 isogenic TLR3 knockout cells (HCT116-TLR3KO) constructed with TLR3 gRNA and HCT115-WT cells constructed with non-targeting (NT) gRNA. HCT116-WT and HCT116-TLR3KO cells were irradiated with 5 Gy and harvested after 24 hours. The levels of IFNβ1 were analyzed by qRT-PCR (n=3). *p<0.05. One-way ANOVA test. ANOVA, analysis of variance; CFSE, Carboxyfluorescein succinimidyl ester; qRT-PCR, quantitated by real-time PCR.

    To clarify which RNA sensor participates in dsRNA-mediated IFN-I production induced by RT, we generated lentiviruses carrying shRNAs against RIG1, TLR3, MDA5 and MAVS to transduce SW480 cells (cGAS-deficient). After puromycin selection, the cells were irradiated to analyze the levels of IFNβ1 and CXCL10 mRNA (figure 2G,H). Knockdown of RIG1, TLR3 and MAVS partially attenuated RT-induced IFNβ1 and CXCL10 upregulation (figure 2H), suggesting that RIG1 and TLR3 may participate in sensing cytosolic dsRNA accumulation for RT-induced IFN-I production. Similar results were observed in HCT116 cells (cGAS-deficient, figure 2I). Knockdown of the dsRNA sensor TLR3 significantly attenuated RT-induced IFNB1 and CXCL10 expression (figure 2I). These results indicated that the dsRNA-mediated signaling pathway may be involved in RT-induced IFN-I production. Furthermore, we found that fewer SupT1 T cells migrated to TLR3-deficient cells after radiation by performing coculture experiments (figure 2J), suggesting that radiation decreased the immunogenicity of TLR3-deficient cancer cells. Additionally, we found that there were fewer migrated T cells in coculture with RIG1-deficient cells after radiation than in coculture with TLR3-deficient cells (figure 2J and online supplemental figure S1B). Consistent with these results, the levels of IFNβ1 and CXCL10 were reduced in HCT116-TLR3KO isogenic cells after radiation (cGAS-deficient, figure 2K), suggesting that TLR3 is critical for recognizing extracellular dsRNA to increase cancer immunogenicity after radiation.

    To demonstrate the importance of the dsRNA sensor TLR3 for RT-induced IFN-I production, we treated cells with the synthetic TLR3 activator poly (I:C) and the TLR3/dsRNA complex inhibitor CU-CPT4a33 together with RT. TLR3, IFNβ1, MX1 and CXCL10 mRNA levels were profoundly increased by irradiation in SW480 and HCT116 cells (figure 3A and online supplemental figure S1C). The synthetic TLR3 activator poly (I:C) also enhanced the radiation-induced increase in TLR3 and IFNβ1 mRNA levels (figure 3A). With the administration of the TLR3/dsRNA complex inhibitor CU-CPT4a, the radiation-induced increase in TLR3 and IFNβ1 mRNA levels were markedly attenuated (figure 3A). MX1 and CXCL10 were strongly upregulated by irradiation (figure 3A). Treatment with a TLR3/dsRNA complex inhibitor suppressed irradiation-induced MX1 and CXCL10 upregulation (figure 3A). Similar results were observed in the HCT116 cell line (online supplemental figure S1C). The TLR3 downstream transcription factor interferon regulatory factor 3 (IRF3) was phosphorylated under RT treatment (figure 3B). Blockade of TLR3/dsRNA significantly alleviated IRF3 phosphorylation after irradiation (figure 3B). These results suggest that the TLR3 signaling pathway participates in RT-induced IFN-I signatures.

    Figure 3
    Figure 3

    TLR3 and IFNAR1 are indispensable for irradiation-induced type I IFN and CXCL10 production. (A) SW480 cells were irradiated and treated with 2.5 µg/mL poly (I:C) and 10 µM TLR3 inhibitor together for 24 hours. The mRNA levels of TLR3, IFNβ1, CXCL10 and MX1 were analyzed by qRT-PCR (n=3). *p<0.05, **p<0.01 and ***p<0.001. One-way ANOVA test. (B) SW480 cells were irradiated and treated with 2.5 µg/mL poly (I:C) and 10 µM TLR3 inhibitor together for 24 hours. The protein levels of p-IRF3, IRF3, p-STAT1, and STAT1 were examined by immunoblotting. *p<0.05, **p<0.01, ***p<0.001. One-way ANOVA test. (C) SW480 cells were transfected with pCMV6-vector, pCMV-TLR3-WT, or pCMV-TLR3-L412F for 24 hours and then irradiated with 5 Gy. After 24 hours, we analyzed the protein level by immunoblotting. (D) The quantification of p-IRF3/IRF3 and p-STAT1/STAT1 is shown. *p<0.05, **p<0.01. One-way ANOVA test. (E) HCT116 cells were infected with lentivirus carrying shRNA against IFNAR. The knockdown efficacy was measured by immunoblotting. HCT116shNC, HCT116shIFNAR1#1 and HCT116shIFNAR1#2 cells were irradiated with 5 and 10 Gy. After 24 hours, the cells were harvested, and the mRNA level of CXCL10 was examined by qRT-PCR. ***p<0.001. One-way ANOVA test. (F) HCT116shNC and HCT116shIFNAR1#2 cells were irradiated with 5 Gy. After 24 hours, conditioned medium was harvested to analyze the level of CXCL10 by ELISA. ***p<0.001. One-way ANOVA test. ANOVA, analysis of variance; qRT-PCR, quantitated by real-time PCR.

    Previous studies have demonstrated that defects in the leucine 412 residue of TLR3 significantly impair its affinity for dsRNA34 and are associated with poor clinical outcomes in patients with breast cancer and CRC.9 12 Therefore, we generated a TLR3-L412F mutant to investigate its impact on RT-induced signaling. As shown in figure 3C, the downstream proteins IRF3 and STAT1 were significantly activated by irradiation. Forced expression of TLR3-WT clearly promoted IRF3 and STAT1 activation, but dominant-negative TLR3-L412F expression inhibited their activation under RT exposure (figure 3C,D). Furthermore, knockdown of type I IFN receptor 1 (IFNAR1) decreased CXCL10 mRNA levels and released CXCL10 (figure 3E,F). Taken together, these results indicate that irradiation led to the accumulation of released dsRNA, which autonomously promoted IFN-I signaling and CXCL10 expression via IFNAR1.

    Blockade of the TLR3/dsRNA complex significantly alleviated radiation-induced intratumoral T lymphocyte infiltration to facilitate tumor progression in vivo

    To explore the effect of dysfunctional TLR3 on the therapeutic efficacy of RT, BALB/c mice were inoculated with syngeneic CT26 colon carcinoma cells and received local RT (5 Gy for two fractions) in combination with the TLR3/dsRNA complex antagonist CU-CPT4a (figure 4A).34 We found that the administration of a TLR3/dsRNA complex antagonist significantly delayed RT-induced tumor regression. The density of cytosolic dsRNA was also inhibited by the TLR3 inhibitor (figure 4B), suggesting that the deficiency of the dsRNA sensor TLR3 strongly influenced RT-induced antitumor immunity. Furthermore, these data corroborate the notion that the therapeutic efficacy of RT relies on TLR3-mediated IFN-I signaling in neoplastic cells. The levels of Ifnβ1, Cxcl10 and Mx1 were upregulated in resected tumor tissues from RT-treated mice (figure 4C,D). Moreover, the mRNA level of Cxcr3, which is the receptor of CXCL10 on T lymphocytes, was also increased, suggesting that immune cells may infiltrate tumors via the CXCL10-CXCR3 interaction (figure 4D). However, pharmacological inhibition of TLR3 decreased the enrichment of RT-induced IFN-I signatures (figure 4C,D), suggesting that a decrease in CXCL10 may lead to the recruitment of fewer T lymphocytes within the TME. Indeed, we observed fewer intratumoral CD3+ TILs, CD8a+ TILs and cytotoxic GzmB+CD8a+ TILs in the resected tumors from the RT/TLR3i-treated mice than in those from the RT-treated mice (figure 4E,F). Additionally, we found that both Cxcr3+CD8a+ TILs and Cxcr3+CD11b+ myeloid cells infiltrated tumors after RT treatment. However, the density of Cxcr3+CD11b+ myeloid cells was less than that of Cxcr3+CD8a+ TILs within tumors. Pharmacological inhibition of TLR3 significantly decreased the RT-induced infiltration of Cxcr3+CD8a+ TILs and Cxcr3+CD11b+ myeloid cells (figure 4G). The phosphorylation of IRF3 and STAT1 was profoundly increased in the resected tumors from the RT-treated mice (figure 4H). Blockade of TLR3 significantly reduced RT-induced IRF3 and STAT1 phosphorylation (figure 4H). Taken together, these results indicated the importance of TLR3 in RT-induced CXCL10 production via IRF3/STAT1 signaling for the recruitment of CXCR3+ immune cells, which are mainly CXCR3+CD8+ T cells.

    Figure 4
    Figure 4

    TLR3 signaling is important for type I IFN and CXCL10 production to recruit intratumoral TILs in vivo. (A) Tumor growth of CT26-driven colon carcinoma established in BALB/c mice (n=8/group) that were treated with radiotherapy (5 Gy for two fractions) or radiotherapy/TLR3 inhibitor (0.4 mg/kg TLR3 inhibitor 1 hour before and after irradiation). Tumor volume was recorded every 3 days (mean±SE). *p<0.05. The resected tumor tissues were harvested on day 25 for further analysis. (B) The representative image of cytosolic dsRNA in resected tumor tissues which was observed by anti-dsRNA (J2 clone) immunofluorescence staining (n=3). The cytosolic dsRNA fluorescence intensity was examined. ***p<0.001. One-way ANOVA test. (C) Resected tumors were extracted, and the mRNA levels of Ifnβ1 and Cxcl10 were analyzed by qRT-PCR (n=4). *p<0.05. One-way ANOVA test. (D) Resected tumors were extracted, and the mRNA levels of Mx1 and Cxcr3 were analyzed by qRT-PCR (n=4). *p<0.05. One-way ANOVA test. (E) The representative images of CD3+ TILs and CD8a+ TILs which were analyzed by immunohistochemistry. The density of CD3+TILs and CD8a+ TILs was counted under high-power-field microscopy (n=6). **p<0.01, ***p<0.001. One-way ANOVA test. (F) The representative images of GzmB+CD8a+ TILs which were analyzed by immunofluorescence staining. The density of GzmB+CD8a+ TILs was counted under high-power-field microscopy (n=6). **p<0.01, ***p<0.001. One-way ANOVA test. (G) The representative images of Cxcr3+CD8a+ TILs and Cxcr3+CD11c+ DCs and dsRNA which were by immunofluorescence staining. The density of Cxcr3+CD8a+TILs and Cxcr3+CD11c+ DCs was counted under high-power-field microscopy (n=6). **p<0.01, ***p<0.001. One-way ANOVA test. (H) The resected tumors were extracted and analyzed by immunoblotting (n=3). Quantitative analysis of the immunoblotting results. ***p<0.001. One-way ANOVA test. ANOVA, analysis of variance; qRT-PCR, quantitated by real-time PCR; TIL, tumor-infiltrating lymphocyte.

    Patients carrying the TLR3-L412F variant exhibited recruitment of fewer CD8+ TILs and had poor survival outcomes after neoCRT treatment

    TLR3 polymorphisms and the IFN-I signature are considered strong prognostic factors for the therapeutic efficacy of chemotherapy in breast cancer and CRC.9 12 Therefore, we analyzed whether a loss-of-function TLR3 variant (L412F/rs3775291) influences the therapeutic efficacy of neoCRT in patients with advanced CRC. By genotyping, we found that 37% of patients had the GA/AA genotype (a TLR3 variant), which was significantly associated with the response to neoCRT. 23 of 44 (52.3%) patients with the TLR3 variant achieved a better pathologic response to neoCRT (TRG3-4), whereas 55 of 74 (74.3%) patients with TLR3-WT achieved a better pathologic response to neoCRT (p=0.017, online supplemental table S2). Furthermore, the TLR3 variant was also associated with disease status (online supplemental table S2). Patients with the TLR3-L412F mutation had a high risk of local regional recurrence (p=0.047), cancer-specific death (p=0.015) and distant metastasis (p=0.055, online supplemental table S2).

    We then examined the density of intratumoral CD8+ TILs in pretreatment biopsies and post-neoCRT surgical specimens.35 Kaplan-Meier survival analysis indicated that CD8+ TILs (84% vs 64%, p=0.042) and wild-type TLR3 (76% vs 50%, p=0.011) were associated with improved disease-free survival (DFS) (online supplemental table S3, figure 5A,B). Additionally, patients with the TLR3-L412F mutation had a lower density of CD8+ TILs after neoCRT treatment (figure 5C, p=0.0216). A significant correlation between TLR3-L412F and the density of CD8+ TILs within the TME was found in the post-neoCRT surgical tissue (online supplemental table S4). Patients with wild-type TLR3 had a high density of CD8+ TILs after neoCRT (OR 4.875, 95% CI 1.537 to 15.461; p=0.005; online supplemental table S4).

    Figure 5
    Figure 5

    Patients with the TLR3 loss-of-function variant (L412F/rs3775291) were associated with less infiltration of CD8+TILs, poor therapeutic response and survival outcome. (A) Kaplan-Meier curves for 5-year DFS in individuals with high or low CD8+TILs density in the post-neoCRT surgical tissues of LARC patient who received neoCRT in a retrospective study (log-rank p=0.042, n=111). (B) Kaplan-Meier curves for 5-year DFS in individuals with wild-type or defective TLR3 variant (L412F/rs3775291) in the cohort receiving neoCRT treatment (log-rank p=0.011, n=118). (C) Representative image of CD8+ TILs within the TME in wild-type and mutant (L412F/rs3775291) TLR3 before and after neoCRT treatment. The density of CD8+ TILs within the TME in the post-neoCRT tissues is shown (unpaired t-test, p=0.0216, n=100). (D) Kaplan-Meier curves for 5-year DFS in individuals with TLR3 variant (L412F/rs3775291) and high/low CD8+ TILs density in the cohort receiving neoCRT treatment (log-rank p=0.0063, n=100). (E) The forest plot of multivariate analysis including pN stage, TRG, TLR3 variant, CD8 counts and TLR3/CD8 (n=100). Multivariate Cox analysis. (F) Scheme of advanced CRC patient enrolment (n=24). T0: diagnosis day; T1: patients received 1 week of treatment; T2: surgery day. (G) The level of type I signatures including IFNα2, IFNβ1 and CXCL10 was measured in the serum of patients with advanced CRC carrying wild-type TLR3 (n=9, paired t-test). *p<0.05 and **p<0.01. (H) The density of CD8+TILs within TME was measured in the patients with advanced CRC carrying wild-type TLR3 (n=9, paired t test). *p<0.05. (I) The level of type I signatures including IFNα2, IFNβ1 and CXCL10 was measured in the serum of patients with advanced CRC carrying mutant TLR3 (n=15, paired t-test). (J) The density of CD8+TILs within TME was measured in the patients with advanced CRC carrying mutant TLR3 (n=15, paired t test). (K) The association between the level of tumor-intrinsic CXCL10 mRNA level and the response to neoCRT was evaluated in the external dataset retrieved from NCBI GEO dataset (GSE46862). MI: minimal response (TRG1); MO: moderate response (TRG2); NT: near complete response (TRG3) and TO: total response (TRG4). (n=64, one-way ANOVA test). (L) Kaplan-Meier curves for DFS in patients with colorectal cancer with high or low CXCL10 mRNA (cut-off=median TPM), which was retrieved form the GEPIA database (Gene Expression Profiling Interactive Analysis, http://gepia.cancer-pku.cn/). Log-rank test (n=270). ANOVA, analysis of variance; CRC, colorectal cancer; DFS, analysis of variance; TILs, tumor-infiltrating lymphocytes; TME, tumor microenvironment.

    Next, we assessed the differences in survival between groups classified by these two factors. For patients with both wild-type TLR3 and a high CD8+ TIL count within the TME prior to neoCRT treatment, the estimated DFS was 92%. Kaplan-Meier survival analysis indicated that patients with both wild-type TLR3 and CD8+ TILs had a better disease-free status (95.2% vs 63.7%, p=0.007; figure 5D). Univariate analysis revealed that TLR3-L412F (rs3775291) and CD8+ TILs were significantly associated with 5-year DFS (online supplemental table S5). According to the univariate analysis of pre-neoCRT biopsy specimens, patients harboring TLR3-L412F had an increased risk for poor 5-year DFS (HR 2.195, 95% CI 1.175 to 4.098, p=0.014), and those with a low CD8+ TIL count also had an increased risk for poor 5-year DFS (HR 2.41, 95% CI 1.003 to 5.794, p=0.049) compared with patients with wild-type TLR3 and a high CD8+ TIL count (online supplemental table S5). Patients with both TLR3-L412F and a low CD8+ TIL count had a high risk for poor 5-year DFS (HR 2.287, 95% CI 1.161 to 4.506; p=0.017). Moreover, in the univariate analysis of post-neoCRT surgical tissues, patients with both TLR3-L412F and a low CD8+ TIL count had an increased risk for poor 5-year DFS (HR 2.663, 95% CI 1.298 to 5.462; p=0.008). These results indicate that the TLR3 variant and CD8+ TILs have significant prognostic value for patients with advanced CRC. After adjustment for pN stage and TRG in multivariate analysis, the prognostic value of the TLR3-L412F variant and CD8+ TILs was observed (figure 5E and online supplemental table S5). Patients with both the TLR3-L412F variant and a low CD8+ TIL count within the TME presented an increased risk of poor DFS (HR 2.675, 95% CI 1.225 to 5.840, p=0.014; online supplemental table S5) and a low 5-year DMFS rate (HR 2.627, 95% CI 1.203 to 5.735, p=0.015; online supplemental table S6) after adjustment for pN stage and TRG. These results show that the combination of the defective TLR3 variant with CD8+ TILs is an independent prognostic factor for patients with advanced CRC (online supplemental table S5).

    The defective TLR3 variant influenced serum IFN-I signature enrichment and the response to neoCRT in a prospective study

    To further evaluate the predictive value of the TLR3 variant on the response to RT via the IFN-I signature, we prospectively enrolled 24 patients to examine the serum levels of IFN-I signatures, including IFNα2, IFNβ1 and CXCL10, and the density of CD8 within the TME at different time points of the neoCRT regimen (figure 5F). We found that the levels of serum IFNα2 and IFNβ1 before the neoCRT regimen (T0) and after the neoCRT regimen (T2) were not different. However, the serum IFNα2 (T1 vs T0: p=0.0431, Tukey’s multiple comparisons test) and IFNβ1 (T1 vs T0: p=0.086, Tukey’s multiple comparisons test) levels were markedly increased after neoCRT treatment for 1 week. By stratifying patients with the TLR3 variant, we found that serum IFNα2 and IFNβ1 levels were greater in patients with wild-type TLR3 than in patients carrying TLR3-L412F (figure 5G,I). Moreover, the density of CD8+ immune cells was markedly greater in patients with wild-type TLR3 than in patients with TLR3-L412F (figure 5H,J). These results showed that patients carrying the TLR3 variant exhibit attenuated RT-induced IFN-I production for immune cell recruitment. Interestingly, mRNA data retrieved from GSE46862 also showed that patients with high CXCL10 mRNA levels had a better response to neoCRT (figure 5K: minimal response; MO: moderate response; NT: near total response; TO: total response).36 Kaplan-Meier survival analysis of data from the Gene Expression Profiling Interactive Analysis (GEPIA) database (http://gepia.cancer-pku.cn/)21 showed that patients with CRC with high CXCL10 expression have more favorable disease-free survival (DFS) than patients with low CXCL10 expression (figure 5L). These results indicated that TLR3-mediated IFN-I signaling, which promotes CXCL10 production, determines the therapeutic efficacy of neoCRT by regulating antitumor immunity.

    Administration of CRC-specific IFNβ delivery effectively rescued the antitumor immunity of RT

    For therapeutic consideration, it was of interest to determine whether TLR3 signaling-deficient tumors could be treated with a type I IFN-based strategy. BALB/c mice were inoculated with syngeneic CT26 colon carcinoma cells and received 5 Gy for two fractions in combination with a TLR3/dsRNA complex inhibitor. Recombinant IFNβ proteins were locally administered by intratumoral injection (figure 6A). Intratumoral supplementation with IFNβ strongly promoted RT-induced tumor regression, and survival was significantly prolonged, even in the TLR3 blockade subgroup (figure 6B,C).

    Figure 6
    Figure 6

    Local IFNβ1 overcomes the impact of TLR3 dysfunction and significantly enhances the therapeutic efficacy of radiotherapy. (A) Tumor growth of CT26-driven colon carcinoma established in BALB/c mice (n=5/group) that were treated with local radiotherapy (5 Gy for two fractions), radiotherapy/TLR3 inhibitor (0.4 mg/kg TLR3 inhibitor 1 hour before and after irradiation), and radiotherapy/TLR3 inhibitor/IFNβ (10,000 U/mouse). (B) Tumor growth was recorded as the mean tumor volume±SEM (n=5). *p<0.05. (C) The survival time was recorded every 3 days, and the mice were sacrificed if the average tumor diameter reached 20 mm. (D) Diagram of the novel engineered AAV-IFNβ with a CRC-specific promoter and the administration protocol. A total of 3×105 CT26 cells were subcutaneously injected into the right leg, and 2×105 CT26 cells were subcutaneously injected into the left back 2 days later. Local radiotherapy was given on days 10 and 13, and AAV (1×108 vg in 50 µL PBS, intramuscular injection) was administered on days 9, 13, 16 and 20. AAV-Vec: empty AAV vector with CRC-specific CEA promoter; CEAp: CEA promoter; gray box: inverted terminal repeats (ITR). (E) CT26 tumor-bearing mice (n=5–6/group) received AAV and local radiotherapy (5 Gy for two fractions) on the indicated days. Tumor volume was measured every 3 days, and tumors were harvested on day 31. **p<0.01, ***p<0.001. CR: complete response. (F) The resected irradiated tumors were weighed (n=3–5, variable due to CR). **p<0.01, ***p<0.001. One-way ANOVA test. (G) The tumor-infiltrating DCs and CD8+ immune cells in resected primary tumors were evaluated by immunofluorescence staining. (H) Quantification of tumor-infiltrating CD8+ immune cells in resected primary tumors. *p<0.05, **p<0.01. One-way ANOVA test. (I) Quantification of tumor-infiltrating CD11c+ DCs in resected primary tumors. *p<0.05, **p<0.01. One-way ANOVA test. (J) The tumor volume of non-irradiated tumors (abscopal tumors) was measured every 3 days, and the tumors were harvested on Day 31. *p<0.05, **p<0.01. (K) The resected abscopal tumors were weighed (n=4–5, variable due to CR). **p<0.01, ***p<0.001. One-way ANOVA test. (L) The tumor-infiltrating DCs and CD8+ immune cells in resected abscopal tumors were evaluated by immunofluorescence staining. (M) Quantification of tumor-infiltrating CD8+ immune cells in resected abscopal tumors. *p<0.05, **p<0.01. One-way ANOVA test. (N) Quantification of tumor-infiltrating CD11c+ DCs in resected abscopal tumors. *p<0.05 and **p<0.01. One-way ANOVA test. ANOVA, analysis of variance.

    To further evaluate the importance of local IFNβ in RT-mediated eradication of distant tumors, we engineered AAV2 with IFNβ1 to constitutively maintain tumor-intrinsic IFNβ1 levels under the control of the CRC-specific CEA promoter (figure 6D). Transduction with AAV2-CEAp-IFNβ1 (hereafter referred to as AAV-IFNβ1) at a multiplicity of infection (MOI)=1000 showed that the level of IFNβ1 was positively correlated with the level of CEACA5 mRNA in CRC cell lines (online supplemental figure S1D–F). Moreover, tumor-intrinsic IFNβ1 was significantly elevated and retained for 3 weeks (online supplemental figure S1G). The serum IFNβ1 concentration slightly increased after the administration of AAV-IFNβ1 and decreased to a normal level within 2 weeks (online supplemental figure S1G), suggesting that the administration of AAV-IFNβ1 (1×108 vg, intramuscular injection) induced cancer-specific expression and was positively correlated with the CEA concentration. We then evaluated whether AAV-IFNβ1 enhances the therapeutic efficacy of RT and potentially induces systemic antitumor immunity to eradicate distant tumors in CRC cells. Therefore, 2 days after subcutaneous inoculation of CT26 cells into the right leg of BALB/c mice, CT26 cells were subcutaneously inoculated on the left back of BALB/c mice. Local RT (5 Gy) was administered on the indicated days, and AAV-IFNβ1 (1×108 vg) was given by intramuscular injection (figure 6D). As shown in figure 6E, we found that the tumor volume was substantially reduced when AAV-IFNβ1 was administered in combination with RT. Significant tumor regression was observed (figure 6E). Moreover, ~50% of the mice achieved a CR after administration of AAV-IFNβ1 plus RT. The resected primary tumor weight was also reduced on day 28 (figure 6F). Tumor-infiltrating CD11c+ dendritic cells (DCs) and CD8+ lymphocytes were significantly increased in resected primary tumors after the administration of AAV-IFNβ1/RT (figure 6G,I). Moreover, the tumor volume of distant non-irradiated tumors (abscopal tumors) was also markedly decreased by AAV-IFNβ1/RT treatment (figure 6J). The resected abscopal tumor weight was also reduced on day 28 (figure 6K). The levels of tumor-infiltrating DCs and CD8+ lymphocytes were significantly increased in resected abscopal tumors after the administration of AAV-IFNβ1/RT (figure 6L,N).

    We then analyzed the immune cell profiles of the resected primary tumors, including CD4, CD8, DC, Treg and NK cells, by flow cytometry (figure 7 and online supplemental figure S2–S5). We found that the density of tumor-infiltrating CD8+ and effector/memory CD8+ T cells (CD44+CD62L-CD8+CD3+CD45+ T cells) significantly changed after RT (figure 7A,B, online supplemental figure S2A). There was no significant change in the percentage of tumor-infiltrating CD4+ or effector/memory CD4+ T cells (CD44+CD62L−CD4+CD3+CD45+ T cells; online supplemental figure S2B,C). The densities of tumor-infiltrating CD11c+ DCs and cDC1s (CD11b−CD8+ CD11c+MHC-II+CD45+ DCs) were significantly changed after RT (figure 7C and online supplemental figure S3A, AAV-Vec vs AAV-Vec/RT). However, there was no significant increase in the number of tumor-infiltrating CD11c+ DCs or cDC1s after administration of AAV-IFNβ1 alone (figure 7C and online supplemental figure S3B). When AAV-IFNβ1 was combined with RT, the infiltration of tumor-infiltrating cDC1s was remarkable (online supplemental figure S3B). Furthermore, we found that the average intensity (mean fluorescence intensity) of IFNγ in tumor-infiltrating CD8+ T lymphocytes and CD49b+ NK cells also markedly increased after the administration of AAV-IFNβ1/RT (figure 7D,E and online supplemental figure S4). There was no significant change in Foxp3+CD25+CD4+ regulatory T cells (online supplemental figure S5, AAV-Vec/RT vs AAV-IFNβ1/RT). However, the ratio of IFNγ+CD8+ T/Foxp3+ T cells was markedly increased after the administration of AAV-IFNβ1/RT (figure 7F). Gene expression profiling of tumors from mice revealed profound upregulation of proinflammatory cytokines such as Ifng, Prf1, Gzmb, Cxcl9, Cxcl10, and Cxcl11 and T-cell markers such as Cd3e, Cd8a and Cd4 in tumor-bearing mice treated with AAV-IFNβ1/RT compared with those treated with AAV-Vec/RT (figure 7G). GSEA also revealed that immune-related signatures, including the HALLMARK_IFNα, IFNγ, IL6_JAK_STAT3, inflammatory and TGFβ1 signatures, were more enriched in the AAV-IFNβ1/RT group than in the other groups (online supplemental figure S6A). HALLMARK_IFNγ response signatures were more enriched in the AAV-IFNβ1 group compared with the AAV-Vec group (online supplemental figure S6B), and the antitumor immune response was enhanced in the AAV-IFNβ1/RT group (online supplemental figure S6C). Moreover, the therapeutic efficacy of AAV-IFNβ1 plus RT dramatically decreased with CD8 depletion (figure 7H), suggesting that cancer-intrinsic IFNβ1 reshaped cancer immunogenicity to trigger systemic antitumor immunity via CD8. We further analyzed the expression of ITR and IFNβ1 in different tissues to evaluate their biodistribution and specificity (online supplemental figure S7A–C). These results demonstrated that IFNβ1 expression was significantly increased in tumor cells. Moreover, AAV-IFNβ1 plus RT had no major effects on toxicity (online supplemental figure S7D). Taken together, these results indicated that supplementation with cancer-intrinsic IFNβ1 altered cancer immunogenicity to trigger systemic antitumor immunity to increase the benefit of RT.

    Figure 7
    Figure 7

    Cancer-intrinsic IFNβ1 expression by AAV transduction significantly reshaped cancer immunogenicity to enhance RT-induced immunity. (A) The tumor-infiltrating CD8+CD3+CD45+ T cells were analyzed by flow cytometry (one-way ANOVA test, n=3–5). *p<0.05, **p<0.01. One-way ANOVA test. (B) The tumor-infiltrating CD44+CD62L-CD8+CD3+CD45+ TEM cells were analyzed by flow cytometry (one-way ANOVA test, n=3–5). *p<0.05, **p<0.01. One-way ANOVA test. (C) The tumor-infiltrating CD11c+MHC-II+ DCs were analyzed by flow cytometry (one-way ANOVA test, n=3–5). *p<0.05, **p<0.01. One-way ANOVA test. (D) The tumor-infiltrating IFNγ+CD8+CD3+CD45+ T cells were analyzed by flow cytometry (one-way ANOVA test, n=3–5). *p<0.05. One-way ANOVA test. (E) The tumor-infiltrating IFNγ+CD49b+CD45+ NK cells were analyzed by flow cytometry (one-way ANOVA test, n=3–5). *p<0.05. One-way ANOVA test. (F) The ratio of IFNγ+CD8+ T cells/Foxp3+ Treg cells. *p<0.05. One-way ANOVA test. (G) Gene expression in whole tumors from mice on day 30 after the indicated treatments. Fold-change over isotype control values represented. The results are representative of two independent experiments (n=2 mice per group). (H) Diagram of AAV-IFNβ1, local RT and anti-mouse CD8a antibodies administration protocol. A total of 3×105 CT26 cells were subcutaneously injected into the right leg. Local RT was given on days 10 and 13, and AAV (1×108 vg in 50 µL PBS, intramuscular injection) was administered on days 9, 13, 16 and 20. Anti-mouse CD8a antibodies were given by intraperitoneal injection (100 µg/mouse). Tumor volume was measured every 2–3 days, and tumors were harvested on day 28 (n=5). **p<0.01, ***p<0.001. (I) The proposed mechanism by which engineered CRC-specific AAV-IFNβ therapy overcomes dysfunctional TLR3-mediated IFN-I attenuation to enhance the response to RT. AAV, adenovirus-associated virus; ANOVA, analysis of variance; RT, radiotherapy.

    Discussion

    Here, we showed that cancer-intrinsic cytosolic dsRNA accumulation and release by RT triggered IFN-I production, leading to systemic antitumor immunity and favorable outcomes via TLR3 (figure 7I). Pharmacological inhibition of TLR3 significantly attenuated the therapeutic efficacy of RT by decreasing cancer immunogenicity and antitumor immunity. A dysfunctional TLR3 variant (TLR3-L412F) was associated with decreased recruitment of CD8+ TILs within the TME and poor patient survival outcomes after neoCRT treatment. CRC-specific IFNβ1 overexpression (via engineered and safe AAVs) enhanced cancer immunogenicity to boost RT-mediated antitumor immunity to eradicate distant abscopal tumors. This study showed that supplementation with AAV-based cancer-intrinsic IFNβ1 delivery can overcome poor immunogenicity and increase the response to RT.

    RT is widely used for cancer treatment owing to its ability to damage DNA and lead to cell death.37 Pharmacological inhibitors of DNA damage response proteins, including ATR, ATM and PARP,38 39 have been developed to enhance the therapeutic efficacy of RT.18 However, increasing evidence indicates that the efficacy of RT largely relies on type I IFN-dependent innate and adaptive immunity. RT-induced dsDNA fragments are recognized by cGAS to trigger STING-mediated production of IFN-I and several proinflammatory cytokines for anticancer immunity.6 38 Indeed, we detected greater IFN-I signaling in post-neoCRT surgical tissues than in pre-neoCRT biopsies, and CXCL10 in pre-neoCRT biopsies was associated with the response to neoCRT in two independent datasets (GSE15781 and GSE46862).30 36 Although our results were inconsistent with those of previous meta-analysis studies in which no predictive gene signature was identified to evaluate the neoCRT response,40 these discrepancies may be due to the use of different analytic strategies. First, we only evaluated the cancer-intrinsic expression of the IFN-I signature in the GSE15781 dataset and did not integrate the gene signatures of normal tissues before and after neoCRT treatment, which is different from the analytical strategy of Snipstad.30 Second, we examined the level of tumor CXCL10 in four subgroups from pre-neoCRT biopsies (TRG1-4) in the GSE46862 dataset, which was different from the findings of previous studies (non-responders: TRG1/2 vs responders: TRG3/4).36 40 Therefore, cancer-intrinsic IFN-I signature may be representative of cancer immunogenicity for RT-induced antitumor immunity. Defects in cGAS/STING evade immune surveillance and attenuate the antitumor immune response in several malignancies, such as CRC and lung cancer.7 26 41 In our previous studies, we found that most patients with CRC were deficient in either cGAS or STING, which might be the result of epigenetic regulation.26 42 Apart from the role of RT-induced cytosolic dsDNA in antitumor immunity,43 recent studies have demonstrated that the formation of dsRNA derived from endogenous retroviruses (ERVs)44 and mitochondrial dsRNA45 46 can be stimulated by RT to activate the innate antiviral RIG-1/MDA5/MAVS pathway, which in turn induces the IFN-I response to boost antitumor immune responses. Lee et al indicated that radiation-induced ERV dsRNA expression and the subsequent immune response play critical roles in clinical RT, and manipulation of epigenetic regulators and the dsRNA-sensing innate immunity pathway could be promising strategies for enhancing the efficacy of RT and cancer immunotherapy.47 Additionally, significant expression of RNA species and a high density of tumor-infiltrating NK cells enhance the efficacy of RT and ICB treatment in patients with poorly immunogenic cancers such as microsatellite-stable CRC and PDAC.13 14 More recent studies also demonstrated that the RNA methyltransferases FTSJ3 and METTL3 suppress dsRNA-induced IFNβ signaling to promote immune evasion and suppress the response to ICBs,48 49 suggesting that dsRNA levels can be used as potential biomarkers of the response to RT and ICBs.

    Here, we found that the genetic status of the RNA sensor TLR3 is associated with the infiltration of immune cells and the therapeutic response to neoCRT treatment in patients with advanced CRC. The presence of wild-type TLR3 was positively correlated with a high density of CD8+ TILs within the TME, suggesting that chemoradiotherapy may promote TLR3-mediated IFN-I signaling for CD8+ TIL recruitment and tumor-specific immune responses via dsRNA. A dysfunctional TLR3 variant (TLR3-L412F) may attenuate dsRNA recognition, resulting in a decreased tumor-specific immune response to avoid immunosurveillance for tumor regrowth. Consistent with our results, the TLR3-L412F SNP (L412F/rs3775291) is associated with an increased risk of cancer and poor survival outcomes in several malignancies, such as CRC,12 NSCLC,50 nasopharyngeal carcinoma,51 oral cancer,52 and hepatocellular carcinoma.53 Castro et al also indicated that patients with CRC harboring TLR3-L412F have significantly shorter survival than patients with CRC harboring wild-type TLR3.12 This TLR3-L412F variant is located adjacent to the glycosylation site (Asn413) within the ligand binding surface for receptor activation and function.54 Thus, the TLR3-L412F variant may affect ligand binding or dimerization of TLR3, resulting in the inactivation of TLR3-mediated IFN-I and recruitment of CD8+ TILs. However, the functional relationship between these dsRNA species and the associated immune cell response still needs further investigation.

    TLR3 and CXCL10 levels were also correlated with improved survival outcomes in multiple malignancies.11 55 Therefore, TLR3 agonists have been extensively tested as immunostimulatory agents in patients with cancer.56 57 Salaun et al indicated that patients with breast cancer with high TLR3 levels have a better therapeutic response to RT with a TLR3 agonist than to adjuvant chemotherapy alone in a clinical trial,55 suggesting that the coadministration of TLR3 agonists may increase the response to RT to trigger bona fide ICD to activate tumor-specific immune responses.56 Takeda et al also reported that a TLR3-specific adjuvant increased the therapeutic efficacy of ICBs.58 These results suggest that the TLR3-mediated IFN-I response can be considered a therapeutic strategy to increase the clinical benefit of RT and immunotherapy.59 60 However, we cannot exclude the possibility that RIG-1 plays a role in cytosolic dsRNA recognition. Our results showed that digestion of dsRNA by dsRNA-specific RNase III in conditioned medium and treatment with a competitive dsRNA/TLR3 complex significantly decreased the RT-induced IFN-I response. However, we also found that RIG-1 knockdown significantly attenuated the RT-induced IFN-I response in cGAS-deficient cells, suggesting that RIG-1 may directly recognize cytosolic dsRNA to facilitate IFN-I production to different extents.

    Here, we found that local supplementation with IFNβ via either intratumoral injection or cancer-specific transduction of AAV-IFNβ1 overcomes impaired TLR3 signaling to increase the response to RT and prolong the survival period in vivo. Moreover, administration of CRC-specific AAV-IFNβ remarkably augmented the therapeutic response to RT by inducing high infiltration of dendritic cells and the activation of IFNγ+CD8+ T cells and IFNγ+ NK cells, resulting in non-irradiated (abscopal) tumor regression, suggesting that immunomodulation by tumor IFNβ potentiates the response to RT through antitumor immunity. Our previous study demonstrated that engineered mesenchymal stem cells (MSCs) with AAV-sTRAIL-IFNβ1 significantly reshaped cancer immunogenicity and increased the therapeutic efficacy of RT and ICBs.59 We exploited the tumor-homing characteristic of MSCs, and we found that local supplementation with sTRAIL and IFNβ1 improved the infiltration of immune cells such as CD4+ and CD8+ T cells to facilitate the antitumor immune response. Similarly, Zhang et al reported that MSCs equipped with IFNα promote antitumor immunity in T cells.61 Taken together, the present results demonstrate that cancer-intrinsic IFNα/β-based immunotherapy may be alternative therapeutic strategies to overcome the impact of TLR3 deficiency and promote the antitumor immune response mediated by RT in CRC.

    These findings suggest an important role for the TLR3 variant in patients with advanced CRC who receive neoCRT treatment. In patients with defective TLR3, dsRNA recognition, which facilitates the antitumor immune response, may be impaired, leading to worse local control and shorter DFS. However, cancer-specific administration of IFNβ via AAV-based methods may provide an alternative therapeutic strategy to overcome the impact of TLR3 deficiency and increase the response to RT.

    Data availability statement

    Data are available on reasonable request. The original dataset is available on request from the corresponding author.

    Ethics statements

    Patient consent for publication

    Ethics approval

    This study involves human participants and was reviewed and approved by the Institutional Review Board (IRB) of China Medical University Hospital (protocol number: CMUH105-REC2-072). Participants gave informed consent to participate in the study before taking part. Animal experiments approved by the China Medical University Institutional Animal Care and Use Committee (Protocol No.: CMUIACUC-2021-108-1).

    Acknowledgments

    We thank the National RNAi Core Facility at Academia Sinica in Taiwan for providing shRNA reagents and related services. This study was partially based on data from the China Medical University Hospital cancer registry. Experiments and data analysis were performed in part through the use of the Medical Research Core Facilities Center, Office of Research and Development at China Medical University, Taichung, Taiwan, R.O.C.

    References

    1. ↵
      1. Rodriguez-Ruiz ME,
      2. Vitale I,
      3. Harrington KJ, et al
      . Immunological impact of cell death signaling driven by radiation on the tumor microenvironment. Nat Immunol 2020;21:120–34. doi:10.1038/s41590-019-0561-4
      OpenUrlPubMed
    2. ↵
      1. Yamazaki T,
      2. Kirchmair A,
      3. Sato A, et al
      . Mitochondrial DNA drives abscopal responses to radiation that are inhibited by autophagy. Nat Immunol 2020;21:1160–71. doi:10.1038/s41590-020-0751-0
      OpenUrlCrossRefPubMed
    3. ↵
      1. Demaria S,
      2. Coleman CN,
      3. Formenti SC
      . Radiotherapy: changing the game in immunotherapy. Trends Cancer 2016;2:286–94. doi:10.1016/j.trecan.2016.05.002
      OpenUrl
    4. ↵
      1. Rodríguez-Ruiz ME,
      2. Vanpouille-Box C,
      3. Melero I, et al
      . Immunological mechanisms responsible for radiation-induced abscopal effect. Trends Immunol 2018;39:644–55. doi:10.1016/j.it.2018.06.001
      OpenUrlCrossRefPubMed
    5. ↵
      1. Huang KC-Y,
      2. Chiang S-F,
      3. Lin P-C, et al
      . Tnfalpha modulates PANX1 activation to promote ATP release and enhance P2RX7-mediated antitumor immune responses after chemotherapy in colorectal cancer. Cell Death Dis 2024;15:24. doi:10.1038/s41419-023-06408-5
    6. ↵
      1. Mackenzie KJ,
      2. Carroll P,
      3. Martin C-A, et al
      . cGAS surveillance of micronuclei links genome instability to innate immunity. Nature 2017;548:461–5. doi:10.1038/nature23449
      OpenUrlCrossRefPubMed
    7. ↵
      1. Xia T,
      2. Konno H,
      3. Ahn J, et al
      . Deregulation of STING signaling in colorectal carcinoma constrains DNA damage responses and correlates with tumorigenesis. Cell Rep 2016;14:282–97. doi:10.1016/j.celrep.2015.12.029
      OpenUrlCrossRefPubMed
    8. ↵
      1. Alexopoulou L,
      2. Holt AC,
      3. Medzhitov R, et al
      . Recognition of double-stranded RNA and activation of NF-kappaB by toll-like receptor 3. Nature 2001;413:732–8. doi:10.1038/35099560
      OpenUrlCrossRefPubMedWeb of Science
    9. ↵
      1. Sistigu A,
      2. Yamazaki T,
      3. Vacchelli E, et al
      . Cancer cell-autonomous contribution of type I interferon signaling to the efficacy of chemotherapy. Nat Med 2014;20:1301–9. doi:10.1038/nm.3708
      OpenUrlCrossRefPubMed
    10. ↵
      1. Pfirschke C,
      2. Engblom C,
      3. Rickelt S, et al
      . Immunogenic chemotherapy sensitizes tumors to checkpoint blockade therapy. Immunity 2016;44:343–54. doi:10.1016/j.immuni.2015.11.024
      OpenUrlCrossRefPubMed
    11. ↵
      1. Sato Y,
      2. Motoyama S,
      3. Wakita A, et al
      . TLR3 expression status predicts prognosis in patients with advanced thoracic esophageal squamous cell carcinoma after esophagectomy. Am J Surg 2018;216:319–25. doi:10.1016/j.amjsurg.2018.01.038
      OpenUrlCrossRef
    12. ↵
      1. Castro FA,
      2. Försti A,
      3. Buch S, et al
      . TLR-3 polymorphism is an independent prognostic marker for stage II colorectal cancer. Eur J Cancer 2011;47:1203–10. doi:10.1016/j.ejca.2010.12.011
      OpenUrlCrossRefPubMed
    13. ↵
      1. Roulois D,
      2. Loo Yau H,
      3. Singhania R, et al
      . DNA-Demethylating agents target colorectal cancer cells by inducing viral mimicry by endogenous transcripts. Cell 2015;162:961–73. doi:10.1016/j.cell.2015.07.056
      OpenUrlCrossRefPubMed
    14. ↵
      1. Parikh AR,
      2. Szabolcs A,
      3. Allen JN, et al
      . Radiation therapy enhances immunotherapy response in microsatellite stable colorectal and pancreatic adenocarcinoma in a phase II trial. Nat Cancer 2021;2:1124–35. doi:10.1038/s43018-021-00269-7
      OpenUrlCrossRef
    15. ↵
      1. Westcott PMK,
      2. Sacks NJ,
      3. Schenkel JM, et al
      . Low neoantigen expression and poor T-cell priming underlie early immune escape in colorectal cancer. Nat Cancer 2021;2:1071–85. doi:10.1038/s43018-021-00247-z
      OpenUrl
    16. ↵
      1. Morel KL,
      2. Sheahan AV,
      3. Burkhart DL, et al
      . EZH2 inhibition activates a dsRNA-STING-interferon stress axis that potentiates response to PD-1 checkpoint blockade in prostate cancer. Nat Cancer 2021;2:444–56. doi:10.1038/s43018-021-00185-w
      OpenUrl
    17. ↵
      1. Huang KC-Y,
      2. Chiang S-F,
      3. Yang P-C, et al
      . ATAD3A stabilizes GRP78 to suppress ER stress for acquired chemoresistance in colorectal cancer. J Cell Physiol 2021;236:6481–95. doi:10.1002/jcp.30323
      OpenUrl
    18. ↵
      1. Chen J-Y,
      2. Lin P-Y,
      3. Hong W-Z, et al
      . Activation of STING by the novel liposomal TLC388 enhances the therapeutic response to anti-PD-1 antibodies in combination with radiotherapy. Cancer Immunol Immunother 2024;73:92. doi:10.1007/s00262-024-03692-8
    19. ↵
      1. Suresh MV,
      2. Thomas B,
      3. Machado-Aranda D, et al
      . Double-stranded RNA interacts with toll-like receptor 3 in driving the acute inflammatory response following lung contusion. Crit Care Med 2016;44:e1054–66. doi:10.1097/CCM.0000000000001879
      OpenUrl
    20. ↵
      1. Subramanian A,
      2. Kuehn H,
      3. Gould J, et al
      . GSEA-P: a desktop application for gene set enrichment analysis. Bioinformatics 2007;23:3251–3. doi:10.1093/bioinformatics/btm369
      OpenUrlCrossRefPubMedWeb of Science
    21. ↵
      1. Tang Z,
      2. Li C,
      3. Kang B, et al
      . GEPIA: a web server for cancer and normal gene expression profiling and interactive analyses. Nucleic Acids Res 2017;45:W98–102. doi:10.1093/nar/gkx247
      OpenUrlCrossRefPubMed
    22. ↵
      1. Huang C-Y,
      2. Chiang S-F,
      3. Ke T-W, et al
      . Cytosolic high-mobility group box protein 1 (HMGB1) and/or PD-1+ TILs in the tumor microenvironment may be contributing prognostic biomarkers for patients with locally advanced rectal cancer who have undergone neoadjuvant chemoradiotherapy. Cancer Immunol Immunother 2018;67:551–62. doi:10.1007/s00262-017-2109-5
      OpenUrl
    23. ↵
      1. Dworak O,
      2. Keilholz L,
      3. Hoffmann A
      . Pathological features of rectal cancer after preoperative radiochemotherapy. Int J Colorectal Dis 1997;12:19–23. doi:10.1007/s003840050072
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      1. Huang KC-Y,
      2. Chiang S-F,
      3. Ke T-W, et al
      . DNMT1 constrains IFNbeta-mediated anti-tumor immunity and PD-L1 expression to reduce the efficacy of radiotherapy and immunotherapy. Oncoimmunology 2021;10:1989790. doi:10.1080/2162402X.2021.1989790
    25. ↵
      1. Wang X,
      2. Sheu JJ-C,
      3. Lai M-T, et al
      . RSF-1 overexpression determines cancer progression and drug resistance in cervical cancer. Biomedicine (Taipei) 2018;8:4. doi:10.1051/bmdcn/2018080104
    26. ↵
      1. Chen T-W,
      2. Hung W-Z,
      3. Chiang S-F, et al
      . Dual inhibition of TGFbeta signaling and CSF1/CSF1R reprograms tumor-infiltrating macrophages and improves response to chemotherapy via suppressing PD-L1. Cancer Lett 2022;543:215795. doi:10.1016/j.canlet.2022.215795
      OpenUrlCrossRef
    27. ↵
      1. Lin Y-S,
      2. Chiang S-F,
      3. Chen C-Y, et al
      . Targeting CD73 increases therapeutic response to immunogenic chemotherapy by promoting dendritic cell maturation. Cancer Immunol Immunother 2023;72:2283–97. doi:10.1007/s00262-023-03416-4
      OpenUrlCrossRef
    28. ↵
      1. Huang KC-Y,
      2. Lai C-Y,
      3. Hung W-Z, et al
      . A novel engineered AAV-based neoantigen vaccine in combination with radiotherapy eradicates tumors. Cancer Immunol Res 2023;11:123–36. doi:10.1158/2326-6066.CIR-22-0318
      OpenUrl
    29. ↵
      1. Zhang F,
      2. Manna S,
      3. Pop LM, et al
      . Type I interferon response in radiation-induced anti-tumor immunity. Semin Radiat Oncol 2020;30:129–38. doi:10.1016/j.semradonc.2019.12.009
      OpenUrl
    30. ↵
      1. Snipstad K,
      2. Fenton CG,
      3. Kjaeve J, et al
      . New specific molecular targets for radio-chemotherapy of rectal cancer. Mol Oncol 2010;4:52–64. doi:10.1016/j.molonc.2009.11.002
      OpenUrlPubMed
    31. ↵
      1. Zheng J,
      2. Mo J,
      3. Zhu T, et al
      . Comprehensive elaboration of the cGAS-STING signaling axis in cancer development and immunotherapy. Mol Cancer 2020;19:133. doi:10.1186/s12943-020-01250-1
    32. ↵
      1. Gebremeskel S,
      2. Johnston B
      . Concepts and mechanisms underlying chemotherapy induced Immunogenic cell death: impact on clinical studies and considerations for combined therapies. Oncotarget 2015;6:41600–19. doi:10.18632/oncotarget.6113
      OpenUrlPubMed
    33. ↵
      1. Takemura N,
      2. Kawasaki T,
      3. Kunisawa J, et al
      . Blockade of TLR3 protects mice from lethal radiation-induced gastrointestinal syndrome. Nat Commun 2014;5:3492. doi:10.1038/ncomms4492
    34. ↵
      1. Gosu V,
      2. Son S,
      3. Shin D, et al
      . Insights into the dynamic nature of the dsRNA-bound TLR3 complex. Sci Rep 2019;9:3652. doi:10.1038/s41598-019-39984-8
    35. ↵
      1. Chen T-W,
      2. Huang KC-Y,
      3. Chiang S-F, et al
      . Prognostic relevance of programmed cell death-ligand 1 expression and CD8+ TILs in rectal cancer patients before and after neoadjuvant chemoradiotherapy. J Cancer Res Clin Oncol 2019;145:1043–53. doi:10.1007/s00432-019-02874-7
      OpenUrlCrossRef
    36. ↵
      1. Gim J,
      2. Cho YB,
      3. Hong HK, et al
      . Predicting multi-class responses to preoperative chemoradiotherapy in rectal cancer patients. Radiat Oncol 2016;11:50. doi:10.1186/s13014-016-0623-9
    37. ↵
      1. Sharma RA,
      2. Plummer R,
      3. Stock JK, et al
      . Clinical development of new drug-radiotherapy combinations. Nat Rev Clin Oncol 2016;13:627–42. doi:10.1038/nrclinonc.2016.79
      OpenUrl
    38. ↵
      1. Feng X,
      2. Tubbs A,
      3. Zhang C, et al
      . ATR inhibition potentiates ionizing radiation-induced interferon response via cytosolic nucleic acid-sensing pathways. EMBO J 2020;39:e104036. doi:10.15252/embj.2019104036
    39. ↵
      1. Wong WK,
      2. Guerra Liberal FDC,
      3. McMahon SJ
      . DNA repair inhibitors potentiate fractionated radiotherapy more than single-dose radiotherapy in breast cancer cells. Cancers (Basel) 2022;14:3794. doi:10.3390/cancers14153794
    40. ↵
      1. Momma T,
      2. Okayama H,
      3. Kanke Y, et al
      . Validation of gene expression-based predictive biomarkers for response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer. Cancers (Basel) 2021;13:4642. doi:10.3390/cancers13184642
    41. ↵
      1. Kitajima S,
      2. Ivanova E,
      3. Guo S, et al
      . Suppression of STING associated with LKB1 loss in KRAS-driven lung cancer. Cancer Discov 2019;9:34–45. doi:10.1158/2159-8290.CD-18-0689
      OpenUrlAbstract/FREE Full Text
    42. ↵
      1. Konno H,
      2. Yamauchi S,
      3. Berglund A, et al
      . Suppression of STING signaling through epigenetic silencing and missense mutation impedes DNA damage mediated cytokine production. Oncogene 2018;37:2037–51. doi:10.1038/s41388-017-0120-0
      OpenUrlCrossRefPubMed
    43. ↵
      1. Yum S,
      2. Li M,
      3. Chen ZJ
      . Old dogs, new trick: classic cancer therapies activate cGAS. Cell Res 2020;30:639–48. doi:10.1038/s41422-020-0346-1
      OpenUrlPubMed
    44. ↵
      1. Pan D,
      2. Bao X,
      3. Hu M, et al
      . SETDB1 restrains endogenous retrovirus expression and antitumor immunity during radiotherapy. Cancer Res 2022;82:2748–60. doi:10.1158/0008-5472.CAN-21-3523
      OpenUrl
    45. ↵
      1. Arnaiz E,
      2. Miar A,
      3. Dias Junior AG, et al
      . Hypoxia regulates endogenous double-stranded RNA production via reduced mitochondrial DNA transcription. Front Oncol 2021;11:779739. doi:10.3389/fonc.2021.779739
    46. ↵
      1. Miar A,
      2. Arnaiz E,
      3. Bridges E, et al
      . Hypoxia induces transcriptional and translational downregulation of the type I IFN pathway in multiple cancer cell types. Cancer Res 2020;80:5245–56. doi:10.1158/0008-5472.CAN-19-2306
      OpenUrlAbstract/FREE Full Text
    47. ↵
      1. Lee AK,
      2. Pan D,
      3. Bao X, et al
      . Endogenous retrovirus activation as a key mechanism of anti-tumor immune response in radiotherapy. Radiat Res 2020;193:305–17. doi:10.1667/RADE-20-00013
      OpenUrlCrossRef
    48. ↵
      1. Zhuang Q,
      2. Dai Z,
      3. Xu X, et al
      . RNA methyltransferase FTSJ3 regulates the type I interferon pathway to promote hepatocellular carcinoma immune evasion. Cancer Res 2024;84:405–18. doi:10.1158/0008-5472.CAN-23-2049
      OpenUrl
    49. ↵
      1. Guirguis AA,
      2. Ofir-Rosenfeld Y,
      3. Knezevic K, et al
      . Inhibition of METTL3 results in a cell-intrinsic interferon response that enhances antitumor immunity. Cancer Discov 2023;13:2228–47. doi:10.1158/2159-8290.CD-23-0007
      OpenUrlCrossRef
    50. ↵
      1. Dai J,
      2. Hu Z,
      3. Dong J, et al
      . Host immune gene polymorphisms were associated with the prognosis of non-small-cell lung cancer in Chinese. Int J Cancer 2012;130:671–6. doi:10.1002/ijc.26067
      OpenUrlPubMed
    51. ↵
      1. Moumad K,
      2. Lascorz J,
      3. Bevier M, et al
      . Genetic polymorphisms in host innate immune sensor genes and the risk of nasopharyngeal carcinoma in North Africa. G3 (Bethesda) 2013;3:971–7. doi:10.1534/g3.112.005371
      OpenUrl
    52. ↵
      1. Zeljic K,
      2. Supic G,
      3. Jovic N, et al
      . Association of TLRr2, TLR3, TLR4 and CD14 genes polymorphisms with oral cancer risk and survival. Oral Dis 2014;20:416–24. doi:10.1111/odi.12144
      OpenUrl
    53. ↵
      1. Li G,
      2. Zheng Z
      . Toll-like receptor 3 genetic variants and susceptibility to hepatocellular carcinoma and HBV-related hepatocellular carcinoma. Tumour Biol 2013;34:1589–94. doi:10.1007/s13277-013-0689-z
      OpenUrl
    54. ↵
      1. Sun J,
      2. Duffy KE,
      3. Ranjith-Kumar CT, et al
      . Structural and functional analyses of the human toll-like receptor 3. Role of glycosylation. J Biol Chem 2006;281:11144–51. doi:10.1074/jbc.M510442200
      OpenUrlAbstract/FREE Full Text
    55. ↵
      1. Salaun B,
      2. Zitvogel L,
      3. Asselin-Paturel C, et al
      . TLR3 as a biomarker for the therapeutic efficacy of double-stranded RNA in breast cancer. Cancer Res 2011;71:1607–14. doi:10.1158/0008-5472.CAN-10-3490
      OpenUrlAbstract/FREE Full Text
    56. ↵
      1. Aranda F,
      2. Vacchelli E,
      3. Obrist F, et al
      . Trial watch: toll-like receptor agonists in oncological indications. Oncoimmunology 2014;3:e29179. doi:10.4161/onci.29179
    57. ↵
      1. Li Y,
      2. Jiang X,
      3. Luo T, et al
      . TLR3 agonist nanoscale coordination polymer synergizes with immune checkpoint blockade for immunotherapy of cancer. Biomaterials 2022;290:121831. doi:10.1016/j.biomaterials.2022.121831
      OpenUrl
    58. ↵
      1. Takeda Y,
      2. Kataoka K,
      3. Yamagishi J, et al
      . A TLR3-specific adjuvant relieves innate resistance to PD-L1 blockade without cytokine toxicity in tumor vaccine immunotherapy. Cell Rep 2017;19:1874–87. doi:10.1016/j.celrep.2017.05.015
      OpenUrlPubMed
    59. ↵
      1. Huang KC-Y,
      2. Chiang S-F,
      3. Chang H-Y, et al
      . Engineered sTRAIL-armed MSCs overcome STING deficiency to enhance the therapeutic efficacy of radiotherapy for immune checkpoint blockade. Cell Death Dis 2022;13:610. doi:10.1038/s41419-022-05069-0
    60. ↵
      1. Koerner J,
      2. Horvath D,
      3. Herrmann VL, et al
      . PLGA-particle vaccine carrying TLR3/RIG-I ligand Riboxxim Synergizes with immune checkpoint blockade for effective anti-cancer immunotherapy. Nat Commun 2021;12:2935. doi:10.1038/s41467-021-23244-3
    61. ↵
      1. Zhang T,
      2. Wang Y,
      3. Li Q, et al
      . Mesenchymal stromal cells equipped by IFNalpha empower T cells with potent anti-tumor immunity. Oncogene 2022;41:1866–81. doi:10.1038/s41388-022-02201-4
      OpenUrl

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • KC-YH and KSCC contributed equally.

    • Contributors KC-YH, H-YC, W-ZH, P-CY, S-FC, and P-CL conducted and performed the experiments. WT-LC, T-WK and T-WC participated in the collection and IHC evaluation of the advanced CRC patients. KC-YH performed the statistical analysis. J-AL, A‑CS, J-YC and S-LP conducted animal experiments for radiotherapy. WT-LC and KSCC supervised this study. KC-YH and KSCC analyzed the data and wrote the manuscript.

    • Funding This study was supported in part by China Medical University Hospital (DMR-112-139) and Ministry of Science and Technology (MOST111-2628-B-039-010, MOST111-2314-B-039-078 and MOST111-2314-B-039-039, Taiwan).

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.