Cell lines and plasmids
Human HNSCC cell line FaDu, human monocytic leukemia cell line THP-1, human embryonic kidney cell line 293T, BALB/c mouse breast carcinoma cell line 4T1, and C57BL/6J mouse lung carcinoma cell line LLC1 were originally sourced from the American Type Culture Collection. The HNSCC cell line OECM-1 and the C57BL/6J murine oral cancer cell line MTC-Q1 were kindly provided by Dr. Kuo-Wei Chang (Department of Dentistry, National Yang Ming Chiao Tung University of Taiwan). The pCDH-Snail, pCDH-Zeb1 and pCDH-MIR21 plasmids were generated via insertion of full-length cDNA (SNAI1: NM_005985; ZEB1: NM_030751.4, MIR21: NR_029493.1) into the pCDH-CMV-MCS-EF1-puro vector. The pcDNA3-Flag-NLRP3 plasmid was provided by Dr. Szu-Ting Chen (Institute of Clinical Medicine, National Yang Ming Chiao Tung University of Taiwan). The 3’-UTR of BRCC3 (NM_001018055) was cloned into pMIR-REPORTER to generate pMIR-BRCC3-wt, and site-directed mutagenesis was performed to generate the miR-21 binding-site-mutated pMIR-BRCC3-mut.
RNA-seq analysis of HNSCC patient samples
We used the following two datasets of RNA-seq data derived from HNSCC patients for analysis: TCGA HNSCC (n=521, website: https: //www.cancer.gov/tcga) and Taipei Veterans General Hospital (TVGH) HNSCC (21 normal oral epithelia, 35 primary tumors, and 9 metastatic tumors from 21 patients; see Table S2 for the patient characteristics). Data on TCGA HNSCC sequences were downloaded by using Xena54 and cBioPortal55 and normalized to Z score or log2(RPM+1) or log2 (norm_count+1). We applied the EMT score to categorize the patients. Briefly, EMT score is defined as the average of Z-score mesenchymal markers (VIM, FN1, CDH2, ITBG6, FOXC2, MMP2, MMP3, MMP9, SOX10, SNAI1, SNAI2, TWIST1, GSC) / average epithelial marker (CDH1, DSP, TJP1)26. We selected and categorized the top one-third EMT score patients as the EMThigh subgroup (n=172 for TCGA and n=22 for TVGH) and the lowest one-third EMT score patients as the EMTlow subgroup (n=172 for TCGA and n=22 for TVGH). The expression of the genes of interest was compared in the EMThigh versus EMTlow subgroups of TCGA and TVGH HNSCC cases.
Preparation of conditional media and macrophages
To produce conditioned media using cancer cells, HNSCC cell lines FaDu/OECM1 with overexpression or knockdown of Snail were cultured in the RPMI 1640 growth medium supplemented with penicillin/streptomycin and 0.5% FBS for 48 h. The supernatant was centrifuged at 2,000 × g for 5 min, and was subsequently collected and stored at -80°C. To prepare human peripheral blood monocyte-derived macrophages, CD14+ monocytes were isolated from the whole blood samples obtained from healthy donors and were cultured in the RPMI 1640 medium containing 10% FBS and 20 ng/ml GM-CSF for 5 days. To investigate the effects of conditioned media obtained from HNSCC cell lines that either overexpressed or exhibited knockdown of Snail on inflammasome activation, macrophages were primed with 1 μg/ml LPS for 4 h; thereafter, they were subjected to washing steps and were stimulated with 20 ng/ml IFN-γ in conditioned media for 48 h. To generate THP-1-derived macrophages, THP-1 cells were subjected to differentiation overnight with 20 ng/ml phorbol 12-myristate 13-acetate; they were subjected to washing steps and were primed with 1 μg/ml LPS for 12 h, following which they were stimulated with 5 μM nigericin in conditioned media for 1 h.
Purification and characterization of tumor-derived exosomes
Exosome purification was performed as per methods described previously7. Briefly, exosomes were purified by differential ultracentrifugation. First, the cells were removed from the conditioned medium via centrifugation at 300 × g for 10 min. To remove large dead cells and substantial cell debris, the supernatants were successively centrifuged at increasing speeds (2,000 × g for 10 min, 10,000 × g for 30 min, the Beckman SW28 rotor). The supernatant was ultracentrifuged at 100,000 × g for 70 min to pellet the exosomes. Exosomes were subjected to washing steps in PBS and were centrifuged at the same high speed (the Beckman TLA-100.3 rotor). The morphology and size distribution of exosomes were analyzed by performing transmission electron microscopy (TEM; JEOL JEM-2000EXII, JEOL USA, INC., Peabody, MA) and nanoparticle tracking analysis (NTA, Malvern Instruments / NanoSight LM10-HS), respectively.
RNA sequencing for exosomal small RNA
For exosomal small RNA sequencing, exosomes were collected, and total RNA extraction was performed. Small RNA fractions were sequenced using the Illumina Genome Analyzer II (Illumina). ConDeTri was used to trim or remove reads according to the quality score. Qualified reads after filtering low-quality data were analyzed using miRDeep2 to clip the 3' adapter sequence, and reads shorter than 18 nucleotides were discarded before performing the alignment of reads to the human and mouse genomes from UCSC. The expression profiles of miRNAs in FaDu exosomes versus SG exosomes are listed in Supplementary Table 3.
Generation of the MIR21/mir21 knockout cell line
To generate MIR21-knockout THP1 (THP1MIR21-/-) and mir21-knockout MTCQ1 (MTCQ1mir21-/-) cells, the genomic region flanking MIR21-5p / mir21a-5p was deleted using the CRISPR/Cas9 system. Briefly, THP-1 or MTCQ1 cells were subjected to transfection with pSpCas9(BB)-2A-Puro (PX459) and pEGFP-N3 pSurrogate reporter, and the in-of-flam dTomato signal was sorted by using FACSAria (BD). The exact deleted sequence of the MIR21 / mir21 genome was cloned using the TA Cloning™ Kit (Thermo) and confirmed by conducting direct sequencing (Supplementary Fig. 4c and Fig. 4a). The depletion of miR-21/mir-21 in THP-1/MTCQ1 cells was confirmed by performing stem-loop reverse transcription-qPCR.
Proximity ligation assay
The proximity ligation assay was performed to investigate the proximity of epitopes recognized by the anti-NLRP3 and anti-ASC antibodies, representing the assembly of NLRP3 inflammasomes in macrophages. Briefly, after incubation with primary antibodies, DuoLink® In Situ PLA probes (Merck KGaA, Burlington, MA) were used and incubation was performed for 1 h at 37°C. Subsequent ligation and detection were performed using the DuoLink® In Situ Detection Reagents Red Kit (Merck KGaA, Burlington, MA, USA). Blockade, antibody hybridization, proximity ligation, and detection were performed according to the manufacturer’s recommendations. The fluorescence images were captured using the Olympus Fluoview FV10i Laser confocal microscope (Olympus Corporation, Tokyo, Japan) equipped with a 60x oil objective (Olympus UPLSAPO 60XO, NA 1.35) and analyzed using Olympus FV10-ASW Version 3.0. The PLA signal was quantified by measuring the number of red dots per cell. The total number of quantified cells from randomly selected fields is detailed in the corresponding Fig. legends.
Immunoblotting and immunoprecipitation
These procedures were performed as per previously described protocols5. The results were measured using the GE LAS-4000 biomolecular imager (GE Healthcare Inc., Marlborough, MA). The information on antibodies used in the experiments is listed in Table S14.
Luciferase reporter assay
For conducting the BRCC3 3’-UTR reporter assay, 50 ng of the wild-type or mutated reporter constructs (+985 to +1784 of the transcription start site TSS), 100 ng of a pCMV-β-gal internal control plasmid, and 3 μg of pCDNA3-miR21 or control vector were co-transfected into HEK-293T cells. Luciferase activity was measured after 48 h of incubation.
The animal experiments were approved by the Institutional Animal Care and Utilization Committee of National Yang Ming Chiao Tung University (IACUC certificate No. 1090514) and Taipei Veterans General Hospital (IACUC certificate No. 2013-169). The in vivo caspase activity assay (Fig. 4b-c), tumor measurement (Fig. 4d-f), and single-cell RNA-sequencing analysis (Fig. 6, Supplementary Fig. 7, Supplementary Fig. 8) were performed using the MTCQ1-C57BL/6J syngeneic murine oral squamous cell carcinoma model41. For the in vivo caspase-1 activity assay, 1 × 106 MTCQ1-WT or MTCQ1mir21-/- cells were inoculated into the subcutaneous region of C57BL/6 mice for 14 days. The mice received intraperitoneal injection of cisplatin (5 mg/kg/day) or phosphate-buffered saline (PBS) since the 14th day for 4 consecutive days. Mice were sacrificed on the 19th day. The tumor specimens were collected, and the F4/80+ tumor-associated macrophages (TAMs) were isolated to determine caspase 1 activity (F4/80+ FLICA+ cells, ImmunoChemistry). To measure the size and weight of the tumors, 1 × 106 MTCQ1-WT or MTCQ1mir21-/- cells were inoculated into the subcutaneous region of C57BL/6J Nlrp3-/- mice. The developed tumors were measured regularly and allowed to grow until the average volume reached 50 mm3 (volume = width2 × length/2). Mice received PBS or 5 mg/kg cisplatin daily every 3 days for a total of six doses. The mice were sacrificed on the 18th day, and the weights of the tumors were recorded. For single-cell RNA-sequencing (scRNA-seq) analysis, 1 × 106 MTCQ1-WT or MTCQ1mir21-/- cells were injected into the subcutaneous region of C57BL/6 mice for 14 days (n=3 for each group). Intraperitoneal injection of cisplatin 5 mg/kg was administered on the 14th day, and the tumors were harvested for sc-RNA-seq analysis.
In vivo PLA was performed, and seral Il1-β levels were detected in both syngeneic 4T1-BALB/c and LLC-C57BL6/J models (Fig. 5a-e, Supplementary Fig. 6). Tumor size and weight were measured, and the infiltrated T cells were analyzed using the LLC-C57BL6/J model (Fig. 5f-i). For in vivo PLA in BALB/c mice (Fig. 5b,d), 2.5 × 105 4T1 cells expressing Snail or a control vector were intravenously injected into the tail veins of the mice. The tumor-bearing mice received a single injection of 5 mg/kg cisplatin on the 14th day after tumor-cell injection. The mice were euthanized on the 17th day after the tumor cell injection. F4/80+ macrophages were isolated from the lungs, and PLA was performed using these cells. We used the breast tumor orthotopic model to examine serum Il-1β levels. A total of 5 × 105 4T1 cells expressing Snail or a control vector were injected into the mammary fat pad of BALB/c mice. The tumor-bearing mice received a single intraperitoneal injection of cisplatin (5 mg/kg body weight) on the 10th day after tumor-cell injection. The mice were euthanized on the 13th day after the tumor cell inoculation. The sera were harvested, and IL-1β levels were detected using a mouse IL-1β ELISA kit (Supplementary Fig. 6b). For the LLC-C57BL/6 model, 5 × 105 LLC1 cells transfected with shRNA against murine Snail or a control sequence were inoculated into the wild-type C57BL/6J subcutaneous area Nlrp3-/- mice. The tumor-bearing mice received a single intraperitoneal injection of 5 mg/kg cisplatin on the 10th day after tumor cell injection. The mice were euthanized on the 13th day after the tumor cell inoculation. The sera were harvested, and Il-1β levels were detected. To measure the tumor size in the LLC-C57BL/6 model (Fig. 5f-g), 5 × 105 LLC cells receiving shRNA against Snail or a control sequence were injected into the subcutaneous region of the C57BL/6J wild-type or Nlrp3-/- mice. The developed tumors were measured regularly and allowed to grow until the average volume reached 50 mm3. Mice received PBS or 5 mg/kg/day for 4 consecutive days for a total of four doses. The mice were sacrificed on the 16th day, and the weights of the tumors were recorded. For quantification of CD8+IFN-γ+ T-cell infiltration and Ifgr expression in tumors (Fig. 5h), wild-type C57BL/6 mice were inoculated subcutaneously with 2.5 × 105 LLC-Ctrl or LLC-shSnail cells. The developed tumors were measured regularly and allowed to grow until the average volume reached 50 mm3. Mice received PBS or 5 mg/kg cisplatin daily for four consecutive days in the LLC1 model. The mice were euthanized on the 16th day post-cisplatin injection. The tumor-infiltrating CD8+ IFN-γ+ T cells were analyzed using flow cytometry. The expression of Ifgr in tumors was examined using RT-qPCR (Fig. 5i).
To obtain cells for scRNA-seq analysis, 1 × 106 MTCQ1-WT/MTCQ1mir21-/- cells were inoculated into the subcutaneous region of C57BL/6 mice. Cisplatin (5 mg/kg) was administered intraperitoneally on the 14th day, and tumors were harvested on the 17th day. Tumors were dissociated using a tumor dissociation kit (Miltenyi Biotec), and dead cells were removed using a dead cell removal kit (MACS). CD45+ tumor-infiltrating immune cells were isolated using microbeads (Miltenyi Biotec). Cell viability detected via trypan blue staining was over 85% for subsequent sequencing. For library construction and sequencing, we used the droplet-based scRNA-seq (10x Genomics Chromium Single Cell 3’ Reagent Kit v3.1 no. 1000121) for single-cell library preparation. After the conduction of reverse transcription, cDNA sequencing was performed using the Illumina Novaseq 6000 (Illumina, Inc.). The QC and filtering steps were performed using a loupe browser. Briefly, data on cells with low total counts and high mitochondrial gene expression were filtered. scRNA-seq reads were processed using a 10x Genomics Cell Ranger pipeline and analyzed using the Partek Flow software (Partek). Clustering of cells in our dataset was performed using a t-distributed stochastic neighbor embedding (t-SNE) algorithm in Partek Flow. To identify the t-SNE sub-clusters present in different types of immune cells, the average gene expression of each cluster was identified using CIPR43. GO analysis of the differentially expressed genes between clusters was performed using DAVID56 and GSEA57. Partek Flow was used for gene-specific analysis (GSA), macrophage re-clustering, macrophage trajectory analysis, and volcano plot generation.
Tissue processing and data generation for Visium spatial gene expression
The sample used for visium analysis was obtained from the primary tumor of a 65 year-old male suffered from stage 4 oral squamous cell carcinoma. Tissue sections prepare were follow by Visium Spatial Tissue Optimization User Guide Rev D (10x Genomics, CG000239). In brief, Frozen samples were cryosectioned on Visium Tissue Optimization Slides. Histology images were taken using Olympus IX83 (10X PH Objective). Library construction was according to the Visium Spatial Gene Expression User Guide. Libraries were loaded at 250 pM and sequenced on a NovaSeq 6000 System (Illumina) using a NovaSeq S4 Reagent Kit (300 cycles, 20012866, Illumina), at a sequencing depth followed the formula provided by the manufacturer (10X Genomics): Calculate total sequencing depth≧ coverage area x total spots on the capture area x 50000 read pairs/ spots. (Coverage area: 20.8%, 36.1%, 26.8%, 31.5% / Sequencing depth: 96M, 170M, 126M, 146M read pairs) Data processing of Visium data, raw FASTQ files and images were output with Space Ranger software (Version 1.2.1), hg38 reference genome was used for gene alignment. Analyzing of gene sets (inflammasome related genes and EMT genes) and visualization image were produced by Loupe Browser 5.1.0. Analyzing of gene sets of the inflammasome-related genes58 and EMT genes26, and visualization image were produced by Loupe Browser 5.1.0.
HNSCC patient samples
The Institutional Review Board approved the study proposed by Taipei Veterans General Hospital (TVGH-IRB certificate No. 2014-03-004AC; No. 2017-05-013AC; No. 2018-06-001BC). Five independent sets of samples were used for the experiments. The patient characteristics are presented in the corresponding supplementary tables. The first group comprised nine cases of multiplex immunofluorescence staining (Fig. S1A, Table S1). The second group comprised frozen samples of 65 microdissected tumors derived from 21 HNSCC patients and contralateral normal oral epithelia (Table S2). Bulk RNA sequencing was performed for tumors and normal samples (Fig. 1B). The third group included 50 samples and the contralateral normal oral epithelia (Table S11). The relative expression levels of miR-21, CXCL10, and IFNG were determined by analyzing the fold changes in the tumor tissues compared with their normal counterparts (Fig. 7B). The fourth group included 19 HNSCC patients who received chemotherapy at the Taipei Veterans General Hospital. The tumor tissues were subjected to staining with an anti-Snail antibody for the categorization of patients (Table S12). The serum levels of IL-1β before and after chemotherapy were determined before and 1 d after chemotherapy (Fig. 7C-D). The fifth group comprised six paraffin-embedded samples from HNSCC patients receiving treatment at the Taipei Veterans General Hospital (Table S13). The samples were used for PLA/immunofluorescence staining to determine the NLRP3 inflammasome activity in TAMs (Fig. 7E).
Quantitative PCR was performed using the StepOnePlus real-time PCR system (Applied Biosystems Inc., Foster City, CA, USA). The primer sequences used for real-time PCR are listed in Table S14.
The paraffin-embedded tissue sections were subjected to deparaffinization and retrieval, followed by subjection to washing steps with water and blockade with 3% hydrogen peroxide. The samples were washed first with water and subsequently with PBS, after which they were subjected to blockade and stained with antibodies, and were subsequently subjected to enzymatic avidin-biotin complex (ABC)-diaminobenzidine (DAB) staining (Leica Biosystems, Wetzlar, Germany). Nuclei were subjected to counterstaining with haematoxylin. All comparative images were obtained using an identical microscope and camera settings (Olympus BX51; Olympus Corporation, Tokyo, Japan). The information on the antibodies used in the experiments is listed in Table S14
Multiplex immunofluorescence staining of HNSCC samples
To perform staining of multiple markers, namely CD4, CD8a, CD163, CD68, CD66b, and PanCK, via multiplex immunofluorescence staining, samples were stained using the Opal 7-Color IHC Kit (Akoya Biosciences) and the corresponding antibodies. After the performance of staining, the Vectra Polaris Automated Quantitative Pathology Imaging System (Akoya Biosciences) was used to scan multispectral data using the Form Tissue Analysis Software (Akoya Biosciences). ImageJ was used to unmix the overlapping signals and to measure the positive signal of each slide. A total of nine patients and ten multispectral images from each patient were involved and analyzed. The information on the antibodies used in the experiments is listed in Table S14.
Data and code availability
The accession numbers for the data reported in this paper are GEO: GSE99474, GSE172326 (review token: mxqtioykdbebtip), GSE178537 (review token: mnytwiwedxkfvsf), and GSE 181300 (review token: wjuhsswmvnylxub). The survival and gene expression data of TCGA HNSCC cohort were based on UCSC Xena dataset54 and cBioPortal55.
Statistical analyses were performed using GraphPad Prism8 (GraphPad Software). Two-sided independent Student’s t-test (normal distribution) Pearson’s correlation test was used to analyze the correlation between the two continuous factors. All statistical data were derived from at least three independent biological replicates, and each experiment contained at least two technical replicates. P ≤ 0.05 was considered statistically significant (*: ≤0.05, **: ≤0.01, ***: ≤0.001).