The head and neck squamous cell carcinoma (HNSCC) tissue samples were obtained after informed consent from otherwise treatment-naïve patients undergoing surgical resection of their primary tumour, ensuring that the immune infiltrate was not influenced by prior therapeutic interventions such as radiotherapy. Inflamed OM tissue biopsies were obtained from individuals undergoing routine dental surgeries for a variety of inflammatory conditions such as periimplantitis, periodontitis or osseous surgery. Matched peripheral blood samples were collected from each tissue donor if possible. All study participants signed a written informed consent before inclusion in the study, and the protocols were approved by the institutional review board (IRB) at the Fred Hutchinson Cancer Research Center (IRB#6007-972 and IRB#8335). A detailed list of the samples and relevant procedure information, together with the panels and/or sequencing experiment performed is provided in Supplementary Table 1. Furthermore, cryopreserved peripheral blood mononuclear cells (PBMCs) from healthy controls (Seattle Area Control Cohort (SAC)) were obtained via the HIV Vaccine Trial network (HVTN) and used for titrations, panel development and as a longitudinal technical control for all flow cytometry acquisitions (data not shown). The human squamous cell carcinoma line SCC-15 was obtained and validated from ATCC (tested negative for mycoplasma).
Isolation of leukocytes from solid human tissues and peripheral blood
After surgical procedures, fresh tissue samples were placed immediately into a 50-ml conical tube with complete media (RP10: RPMI1640 supplemented with penicillin, streptomcyin and 10% fetal bovine serum (FBS)) and kept at 4 °C. Samples were processed within 1–4 h after collection based on optimized protocols adapted from ref. 53. In brief, tissue pieces were minced using a scalpel into small pieces and incubated with Collagenase II (Sigma-Aldrich, 0.7 mg ml−1) and DNAse (5 U ml−1) in RPMI1640 with 7.5% FBS for 30–45 min depending on sample size. Subsequently, any remaining tissue pieces were mechanically disrupted by repeated resuspension with a 30 ml syringe with a large bore tip (16 × 1.5 blunt). The cell suspension was filtered using a 70-μm cell strainer, washed in RPMI1640 and immediately used for downstream procedures.
Peripheral blood samples (1–10 ml) were collected in ACD tubes and then processed using SepMate tubes (StemCell Technologies, 85450) and Lymphoprep (Stem Cell Technologies, 07851) according to manufacturer protocols. In brief, whole blood samples were centrifuged for 10 min at 400g, and the plasma supernatant was collected separately and immediately frozen at −80 °C. Remaining cells were resuspended in 30 ml PBS and pipetted on top of 13.5 ml Lymphoprep in a SepMate tube. After centrifugation for 16 min at 1,200g, the mononuclear cell fraction in the supernatant was poured into a fresh 50-ml tube, washed with PBS and immediately used for downstream procedures. For blood samples from dental surgery patients, red blood cells were lysed using ACK-lysis buffer (Thermo Fisher, A10492-01), and the remaining white blood cells were directly used for downstream staining.
If required, cells isolated from tissue samples or from peripheral blood were frozen using either a 90% FBS/10% DMSO mixture or Cell Culture Freezing Medium (Gibco, 12648010), and stored in liquid nitrogen until used for downstream procedures.
Flow cytometry and cell sorting
For flow cytometric analysis good practices were followed as outlined in the guidelines for use of flow cytometry54 and consensus suggestions for data analysis55. Directly following isolation or thawing, cells were incubated with Fc-blocking reagent (BioLegend Trustain FcX, 422302) and fixable UV Blue Live/Dead reagent (ThermoFisher, L34961) in PBS (Gibco, 14190250) for 15 min at room temperature. After this, cells were incubated for 20 min at room temperature with 50 μl total volume of antibody master mix freshly prepared in Brilliant staining buffer (BD Biosciences, 563794), followed by two washes in fluorescence-activated cell sorting (FACS) buffer (PBS with 2% FBS). All antibodies were titrated and used at optimal dilution, and staining procedures were performed in 96-well round-bottom plates (for cell sorting in 5-ml polystyrene tubes). A detailed list of the main panels used, including fluorochromes, antibody catalogue numbers and final dilutions is provided in Supplementary Table 2 (panels designed according to best practices as described56) and Supplementary Table 3. For sorting, cells were immediately used after staining, and for analysis, the stained cells were fixed with 4% PFA (Cytofix/Cytoperm, BD Biosciences, 554722) for 20 min at room temperature, washed, resuspended in FACS buffer and stored at 4 °C in the dark until acquisition. If necessary, intracellular (CD68, granzyme B (GZMB) or CTLA4) or intranuclear staining (FOXP3, KI67, TCF1, TOX, T-bet or EOMES) was performed following the appropriate manufacturer protocols (eBioscience FOXP3/Transcription Factor Staining Buffer Set, Thermo Fisher 00-5532-00).
Single-stained controls were prepared with every experiment using antibody capture beads (BD Biosciences anti-mouse (552843) or anti-mouse Plus, and anti-rat (552844)) diluted in FACS buffer, or cells for Live/Dead reagent, and treated exactly the same as the samples (including fixation procedures). For each staining of experimental samples, a PBMC sample from the same healthy donor (SAC) was stained with the same panel as a longitudinal reference control (data not shown).
All samples were acquired using a FACSymphony A5 (BD Biosciences), equipped with 30 detectors and 355 nm (65 mW), 405 nm (200 mW), 488 nm (200 mW), 532 nm (200 mW) and 628 nm (200 mW) lasers and FACSDiva acquisition software (BD Biosciences). Full details on the optical configuration of the instruments used are as described19. Detector voltages were optimized using a modified voltage titration approach57 and standardized from day to day using MFI target values and 6-peak Ultra Rainbow Beads56 (Spherotec, URCP-38-2K). After acquisition, data was exported in FCS 3.1 format and analysed using FlowJo (version 10.6.x, and 10.7.x, BD Biosciences). Samples were analysed using a combination of manual gating and computational analyses approaches55, with doublets being excluded by FSC-A vs FSC-H gating. For fresh samples acquired on different experimental days with the T cell or APC panel, files were exported as compensated data and analysed combined together in a new workspace (see deposited data on www.flowrepository.org). Gates were kept the same across all samples except where changes in the density distribution of populations clearly indicated the need for sample-specific adjustment. For the APC panel, PD-L1 (V450 channel) as well as CD85k (V510 channel) were excluded from analysis because of interference or high variability from highly auto-fluorescent myeloid cells in some samples. For the T cell panel, granzyme B and TIM3 staining showed donor-specific shifts in intensity, requiring sample-specific adjustments of gates.
All cell sorting was performed either on a FACSAria III (BD Biosciences), equipped with 20 detectors and 405 nm, 488 nm, 532 nm and 628 nm lasers or on a FACSymphony S6 cell sorter (BD Biosciences), equipped with 50 detectors and 355 nm, 405 nm, 488 nm, 532 nm and 628 nm lasers. For all sorts involving myeloid cells, an 85-μm nozzle operated at 45 psi sheath pressure was used, for sorts exclusively targeting T cells, a 70-μm nozzle at 70 psi sheath pressure was used. Unless stated otherwise, cells were sorted into chilled Eppendorf tubes containing 500–1,000 μl complete RPMI, washed once in PBS and immediately used for subsequent processing.
Whole-transcriptome single-cell library preparation and sequencing
cDNA libraries were generated using the 10x Genomics Chromium Single Cell 3′ Reagent Kits v2 protocol or the v3 protocol, or using the 10x Genomics Chromium Single Cell 5′ Reagent Kit v1 protocol (see Supplementary Table 1). In brief, after sorting single cells were isolated into oil emulsion droplets with barcoded gel beads and reverse transcriptase mix using the Chromium controller (10x Genomics). cDNA was generated within these droplets, then the droplets were dissociated. cDNA was purified using DynaBeads MyOne Silane magnetic beads (ThermoFisher, 370002D). cDNA amplification was performed by PCR (10 cycles) using reagents within the Chromium Single Cell 3′ Reagent Kit v2 or v3 (10x Genomics) or the VDJ and GEX reagent kit v1 (see list of samples in Supplementary Table 1). Amplified cDNA was purified using SPRIselect magnetic beads (Beckman Coulter) according to the respective protocol. cDNA was enzymatically fragmented and size selected prior to library construction. Libraries were constructed by performing end repair, A-tailing, adaptor ligation, and PCR (12 cycles). Quality of the libraries was assessed by using Agilent 2200 TapeStation with High Sensitivity D5000 ScreenTape (Agilent). Quantity of libraries was assessed by performing digital droplet PCR (ddPCR) with Library Quantification Kit for Illumina TruSeq (BioRad, 1863040) or determined by Qubit with the dsDNA HS Assay (Q32851). Pooled Libraries were diluted to 2 nM or 3 nM and paired-end sequencing was performed on a HiSeq 2500 (Illumina) or a NovaSeq 6000 (Illumina) utilizing S1 or S2 flow cells, targeting between 25,000–50,000 reads per cell.
Targeted transcriptomics single-cell library preparation and sequencing
cDNA libraries were generated as described in detail58. In brief, after sorting, single cells were stained with Sample-Tag antibodies (if required, see Extended Data Fig. 8a) and or AbSeq antibodies (if required), washed three times, pooled and counted and subsequently loaded onto a nano-well cartridge (BD Rhapsody), lysed inside the wells followed by mRNA capture on cell capture beads according to manufacturer instructions58. Cell Capture Beads were retrieved and washed prior to performing reverse transcription and treatment with Exonuclease I. cDNA underwent targeted amplification using the Human Immune Response Panel primers and a custom supplemental panel (listed in Supplementary Table 3) via PCR (10–11 cycles). PCR products were purified, and mRNA PCR products were separated from Sample-Tag (and AbSeq, where applicable) PCR products with double-sided size selection using SPRIselect magnetic beads (Beckman Coulter). mRNA and Sample Tag products were further amplified using PCR (ten cycles). PCR products were then purified using SPRIselect magnetic beads. Quality of PCR products was determined by using an Agilent 2200 TapeStation with High Sensitivity D5000 ScreenTape (Agilent) in the Fred Hutch Genomics Shared Resource laboratory. The quantity of PCR products was determined by Qubit with Qubit dsDNA HS Assay (Q32851). Targeted mRNA product was diluted to 2.5 ng μl−1, and the Sample Tag and AbSeq PCR products were diluted to 1 ng μl−1 to prepare final libraries. Final libraries were indexed using PCR (6 cycles). Index PCR products were purified using SPRIselect magnetic beads. Quality of all final libraries was assessed by using Agilent 2200 TapeStation with High Sensitivity D5000 ScreenTape and quantified using a Qubit Fluorometer using the Qubit dsDNA HS Kit (ThermoFisher). Final libraries were diluted to 3 nM and multiplexed for paired-end (100 bp) sequencing on a NovaSeq 6000 (Illumina) using S1 and S2 flow cells. For the gene expression library, we targeted 5,000–20,000 reads per cell, for the AbSeq library 10,000–15,000 reads per cell, and for the Sample-Tag libraries 500–2,000 reads per cell.
Ex vivo stimulation assays
Cells were isolated from tissues or blood as described above. For some of the stimulation assays cryo-preserved cell suspensions were used after assessing good cellular viability. For the 2 h short-term stimulation assays with targeted transcriptomics (Fig. 4), CD3+ T cells (live CD45+CD19−CD3+ events) were isolated using FACS using a BD FACSAria II. Five-thousand cells were placed into each well of a V-bottom 96-well plate with 200 μl complete media. Cells were then left untreated (control), or stimulated with IL-12, IL-15 and IL-18 (each at 1 nM), or with PMA (50 ng ml−1) and ionomycin (500 ng ml−1) for 2 h at 37 °C. Cells were then washed with 1× PBS and prepared for targeted transcriptomics and staining with oligonucleotide-conjugated antibodies as described58. For the 1- to 3-day stimulation assays (Fig. 4, Extended Data Fig. 7), CD4+CD25+CD127−IL1R1+ and IL1R1− Treg cells were isolated from blood and HNSCC tissues using a FACSymphony S6 sorter (BD Biosciences), and cultured either in RP10 alone or with anti-CD3/CD28 Dynabeads (Gibco, 11161D, used at a 1:1 bead-to-cell ratio) or with anti-CD3/CD28/CD2 beads (Miltenyi, 130-092-909, Treg Suppression Inspector, also used at a 1:1 bead-to-cell ratio), either with or without recombinant IL-1β (Peprotech, 200-01B) at 50 ng ml−1. For some experiments, culture cells were subsequently stained and 250–500 viable cells were sorted on an BD S6 sorter followed by bulk RNA-sequencing (RNA-seq) analysis using the SMART-Seq v4 kit (Takara) as described further below.
For suppression assays, IL1R1+ and IL1R1−CD4+CD25+CD127− regulatory T cells and CD4+CD25− and CD8+ Tresp cells were sorted from cryopreserved HNSCC samples. For some experiments, matched peripheral blood was included. Tresp cells were labelled with Cell Trace Violet (CTV) according to the manufacturer instructions (Thermo Fisher, C34571). In brief, 106 sorted Tresp cells were washed with PBS after the sort, and then incubated in pre-warmed PBS containing a final concentration of 5 µM freshly diluted CTV for 15 min. The reaction was quenched with prewarmed RP10. Both Tresp and Treg cells were counted twice on a BioRad TC20 cell counter. 10,000 (20,000 for some experiments) CTV-labelled Tresp cells were cultured alone, or with 10,000 Treg cells (or titred amounts of Treg cells) in a 96-well round-bottom plate at 37 °C for 4 days together with anti-CD3/CD28/CD2 beads (Miltenyi, 130-092-909, Treg Suppression Inspector). An unstimulated control well was included with every experiment. Where indicated, recombinant IL-1β (Peprotech, 200-01B) was added to achieve a final concentration of 50 ng ml−1. On the read-out day, supernatants were collected and frozen at −80 °C, and the cells were stained with a 14-colour readout panel including Live/Dead reagent (Supplementary Table 2), fixed and acquired on a BD FACSymphony A5, as described above. Cell proliferation was assessed by using the proliferation platform in FlowJo 10.7 (BD Biosciences), with percentage of divided cells (modelled, not gated) as the main readout. Supernatants were processed for Luminex analysis by the Immunomonitoring Core of the Fred Hutchinson Cancer Research Center.
Luminex analysis of tumour lysates
Luminex analysis was performed on lysates of tissues. To obtain lysates from tumour tissues, a 2 × 2 mm piece was incubated for one minute in PBS/0.1% tween. After incubation, the tissue piece was minced in the buffer and then centrifuged at 10,000 rpm for 5 min. The supernatant was collected and immediately flash-frozen on dry ice. Processing for Luminex was performed by the Immunomonitoring Core of the Fred Hutchinson Cancer Research Center.
Isolation and stimulation of mouse cells
Mouse protocols were approved by and in compliance with the ethical regulations of Fred Hutchinson Cancer Research Center’s IACUC. All animals were maintained in specific pathogen-free facilities and euthanized in accordance with institutional protocols. We received thymus, spleen, and lymph node (LN) from male Foxp3eGFP-cre-ERT2 mice (age ≥8 weeks) (from J. Lund), and mechanically dissociated thymus, spleen or lymph node through a 70-µm strainer. To enrich T cells from spleen–lymph node single-cell suspensions, we used a T cell-negative isolation based magnetic enrichment (Stemcell Technologies). For TCR stimulations, we prepared plate-bound anti-CD3 and anti-CD28 by incubating 96-well V-bottom tissue culture plates with 100 µl of 1 µg ml−1 anti-CD3 (clone: 145-2C11) and 2 µg ml−1 anti-CD28 (clone 37.51) in 1× PBS for 3 h at 37 °C. We decanted and washed residual anti-CD3/anti-CD28 solution and plated 1 × 106 isolated T cells per well in 96-well V-bottom tissue culture plates. We cultured cells in modified RP10 media (RPMI1640 supplemented with 10% FBS, 2mM l-glutamine, 100 U ml−1 penicillin-streptomycin, 1 mM sodium pyruvate, 0.05 mM β-mercaptoethanol and 1 mM HEPES). We collected cells for flow analysis at 0-, 1- and 2-day time points for flow cytometric analysis as described above. The following panel was used: anti-TCRγδ–PerCPe710 (clone eBioGL3), anti-CD4–BV786 (clone GK1.5), anti-CD8a–V500 (clone 53-6.7), anti-CD44–AF700 (clone IM7), anti-CD69–PECy7 (clone H1.2F3), anti-PD-1–BV605 (clone 28F.1A12), anti-ICOS–AF647 (clone C398.4A), anti-IL1R1–PE (clone 35F5), anti-IL1R2–BV421 (clone 4E2), anti-CD3–BUV805 (clone 17A2) and anti-FOXP3–FITC (clone FJK-16s, intranuclear post fixation).
Humanized mouse experiments
MISTRG mice (M-CSFh/hIL-3/GM-CSFh/hSIRPαh/mTPOh/hRAG2−/−IL2Rγ−/−) were previously reported59. All animal experiments were approved by Fred Hutchinson Cancer Research Center’s Institutional Animal Care and Use Committee (protocol 50941). De-identified human fetal liver tissues, obtained with informed consent from the donors, were procured by Advanced Bioscience Resources and their use was determined as non-human subject research by Fred Hutch’s Institutional Review Board (6007-827). Fetal livers were cut in small fragments, treated for 45 min at 37 °C with collagenase D (Roche, 100 ng ml−1), and a single-cell suspension was prepared. Hematopoietic cells were enriched by density gradient centrifugation in Lymphocyte Separation Medium (MP Biomedicals) followed by positive immunomagnetic selection with anti-human CD34 microbeads (Miltenyi Biotec). Purity (>90% CD34+ cells) was confirmed by flow cytometry and cells were frozen at −80 °C in FBS containing 10% DMSO. Newborn MISTRG mice (day 1–3) were sublethally irradiated (80 cGy gamma rays in a Caesium-137 irradiator) and ∼20,000 CD34+ cells in 20 μl PBS were injected into the liver with a 22-gauge needle (Hamilton Company), as described59. Engraftment levels were measured as the percentage of human CD45+ cells among total (mouse and human combined) CD45+ cells in the blood.
The human squamous cell carcinoma line SCC-15 was obtained and verified from ATCC. Cells were grown to ∼80% confluency in DMEM/F12 supplemented with 12.5 mM l-glutamine, 15 mM HEPES, 0.5 mM sodium pyruvate and 400 ng ml−1 hydrocortisone. Approximately 0.5 million cells per mouse were resuspended in 75 µl PBS, mixed with 25 µl growth-factor-reduced Matrigel (Corning, 354230) and then injected subcutaneously under anaesthesia in the flank of humanized mice. The size of the tumours was measured weekly for 7 weeks with a calliper. SCC15 tumour tissues were processed for leukocyte isolation as described above for human tissues.
Bulk RNA-seq experiments and analysis
Bulk RNA-seq was performed on 250 sort-purified IL1R1+ and IL1-R1− Treg cells derived from either cryopreserved blood or HNSCC tissues samples after culture in conditions of no stimulation, stimulation with anti-CD3/CD28/CD2 beads, and stimulation with anti-CD3/CD28/CD2 beads and IL-1β (50 ng ml−1) for days 1, 2, and 3. In total, 88 samples were sequenced, and each condition was represented by at least 3 or more biological replicates.
Cells were sorted directly into lysis buffer from the SMART-Seq v4 Ultra Low Input RNA Kit for sequencing (Takara), immediately snap frozen on dry ice, and transferred to −80 °C storage until processed into cDNA. All samples were thawed, cells were lysed, and cDNA was synthesized and amplified per the manufacture’s instruction. After amplification, sequencing libraries were constructed using the NexteraXT DNA sample preparation kit with unique dual indexes (Illumina) to generate Illumina-compatible barcoded libraries. Libraries were pooled and quantified using a Qubit Fluorometer (Life Technologies). Sequencing of pooled libraries was carried out on a NextSeq 2000 sequencer (Illumina) with paired-end 59-base reads, using a NextSeq P2 sequencing kit (Illumina) with a target depth of 5 million reads per sample.
Base calls were processed to FASTQs on BaseSpace (Illumina), and a base call quality-trimming step was applied to remove low-confidence base calls from the ends of reads. Reads were processed using workflows managed on the Galaxy platform. Reads were trimmed by 1 base at the 3′ end then trimmed from both ends until base calls had a minimum quality score of at least 30. Any remaining adapter sequence was removed as well. To align the trimmed reads, STAR aligner (v2.4.2a) was used with the GRCh38 reference genome and gene annotations from ensembl release 91. Gene counts were generated using HTSeq-count (v0.4.1). Quality metrics were compiled from PICARD (v1.134), FASTQC (v0.11.3), Samtools (v1.2), and HTSeq-count (v0.4.1).
A quality filter was applied to retain libraries in which the fraction of aligned reads examined compared to total FASTQ reads was >70%, the median coefficient of variation of coverage was less than 0.85, and the library had at least 1 million reads. All sequenced samples passed these quality filters. Non-protein coding genes and genes expressed at less than 1 count per million in fewer than 10% of samples were filtered out. Expression counts were normalized using the TMM algorithm. For differential gene expression analysis, the linear models for microarray data (Limma) R package after Voom transformation was used; this approach either outperforms or is highly concordant with other published methods. Linear models were generated, and donor identity was included as a random effect. For differential gene expression comparisons, genes with a false discovery rate (FDR) of less than 0.1 and an absolute expression fold-change greater than 1 were considered differentially expressed.
Pre-processing for whole transcriptome analysis (WTA) and targeted transcriptomics data
Raw base call (BCL) files were demultiplexed to generate Fastq files using the Cell Ranger mkfastq pipeline within Cell Ranger (10x Genomics). Whole-transcriptome Fastq files were processed using the standard Cell Ranger pipeline (10x genomics) within Cell Ranger 2.1.1 or Cell Ranger 3.0.2. In brief, Cell Ranger count performs read alignment, filtering, barcode and unique molecular identifier (UMI) counting, and determination of putative cells. The final output of Cell Ranger (the molecule per cell count matrix) was then analysed in R using the package Seurat60,61 (3.0) as described below. For targeted transcriptomics data, Fastq files were processed via the standard Rhapsody analysis pipeline (BD Biosciences) on Seven Bridges (www.sevenbridges.com). In brief, after read filtering, reads are aligned to a reference genome and annotated, barcodes and UMIs are counted, followed by determining putative cells. The final output (molecule per cell count matrix) was also analysed in R using Seurat60,61 (version 3.0) as described below. For 5′ VDJ sequencing experiments, the output after Cell Ranger vdj was analysed using the Loupe VDJ browser v3 (10x Genomics). For the SMART-Seq v4 experiments, Fastq files were aligned to the GRCh38 reference genome as described in more detail above.
Seurat workflow for targeted transcriptomics and WTA data
The R package Seurat60,61 was used for all downstream analysis, with custom scripts based on the following general guidelines for analysis of scRNA-seq data62.
In brief, for whole-transcriptome data, only cells that had at least 200 genes (v2 kits) or 800 genes (v3 kits), and depending on sample distribution less than 7–15% mitochondrial genes were included in analysis. All acquired samples were merged into a single Seurat object, followed by a natural log normalization using a scale factor of 10,000, determination of variable genes using the vst method, and a z-score scaling. Principal component analysis was used to generate 75 principal components, followed by data integration using Harmony30. The dimensionality reduction generated by Harmony was used to calculate UMAP, and graph-based clustering with a resolution between 0.2 and 0.6. For cell annotation, we applied SingleR as a purely data-driven approach32, and used the expression of typical lineage transcripts to verify the cell label annotation. For all subsequent analysis steps, the integrated Seurat object was separated into two objects containing all T cells or all APCs, respectively, and UMAP calculation as well as clustering steps were repeated.
For targeted transcriptomics data36, separate cartridges from the same experiment were merged (if applicable), and only cells that had at least 30 genes were included in downstream analysis. After generating a Seurat object, a natural log normalization using a scale factor of 10,000 was done, followed by determination of variable genes using the vst method, and a z-score scaling. Principal component analysis was used to generate 75 principal components, followed by data integration using Harmony30. The dimensionality reduction from Harmony was used for subsequent UMAP calculation and graph-based clustering with tuned resolution. Protein phenotyping data was stored in a separate slot as described in the Seurat tutorial for CITE-seq data, and normalized using the centred log ratio (CLR) method36. For some figures, the count matrices were exported as FCS files using the package Premessa, and then imported into FlowJo 10.7.x. Appropriate arcsinh transformations were applied in a channel-specific manner, and transcript or protein expression was plotted and quantified using two-dimensional plots.
For all differential gene expression analyses we utilized the Seurat implementation of MAST (model-based analysis of single-cell transcriptomes) with the number of UMIs included as a covariate (proxy for cellular detection rate (CDR)) in the model34. For calculating the T helper scores (Extended Data Fig. 5f, g) we used the AddModuleScore function of Seurat (see Github script on https://github.com/MairFlo/Tumor_vs_Inflamed/blob/main/OM_HNSCC_scRNAseq_Harmony). The genes used were as follows: TH1: IFNG, TBX21, IL12RB1 and IL12RB2; TH2: TNFSF11, GATA3 and IL4; TH17: RORC, CCR6, IL17A, IL17F, IL23R, IL22, AHR, IL26, CCL20; TC: CD8B, CD8A, TNF, IFNG, IL2, GZMB, PRF1, GZMA and FAS. Tex: TCF7, TOX, HAVCR2, PDCD1 and LAG3; Treg: FOXP3, CTLA4, IL2RA, IL2RB and ENTPD1.
NicheNet analysis was adapted from the vignette described at https://github.com/saeyslab/nichenetr35. In brief, the separate Seurat objects containing APCs (described above) were subsetted to contain only HNSCC-derived cells, and the Seurat object containing T cells only HNSCC and OM-derived cells. During multiple separate NicheNet runs, different T cell subsets were set as ‘receiver’ (that is, CD4 non-Treg clusters 0 and 2, CD8 T cell clusters 1, 3 and 4 and Treg cluster 5; Extended Data Fig. 5a) and all myeloid cell clusters (except the pDC and mast cell cluster; Fig 2b) as ‘sender’ populations. For the receiver cell population, a DEG test was performed to find genes enriched in HNSCC vs OM samples, with the key parameters being set as follows: genes expressed in at least 10% of the cells of the respective T cell clusters, and filtered after the DEG test for an adjusted P-value of less than 0.05 and average log fold change more than 0.25. For the sending cell population, all ligands expressed in at least 5% of the cells in the respective APC cluster were considered. NicheNet analysis was performed based on the vignette to infer receptors, filter for documented links and generate a circus plot of the top ligand-receptor interactions for the respective cellular populations. Scoring of the predicted targets was based on a Pearson correlation coefficient as described in the NicheNet vignette. Circos plots were generated as described in the vignette35 to visualize links between ligands on APCs and receptors on the T cell subsets.
For the T cell panel, FAUST was used to discover and annotate phenotypes in 22 samples (11 HNSCC and 11 OM). FAUST was applied to CD45+ live lymphocytes identified through manual gating. The MR1–tetramer, CD45 and the Live/Dead marker were excluded from the FAUST analysis to account for the manual analysis. After tuning, FAUST selected the markers CD8, CD4, CD3, CD45RA, CD27, CD19, CD103, CD69, CD28, HLADR, GZMB, PD-1, CD25, ICOS, TCRγδ, CD38 and TIM3 for discovery and annotation of phenotypes. Counts of the discovered phenotypes labelled CD3+ and CD19− were tested for association with tissue type using a binomial generalized linear mixed-effects model with a subject level random effect. Fifty phenotypes were associated with tissue type at the FDR-adjusted 0.05 level.
For the APC panel, FAUST was used to discover and annotate phenotypes in 32 samples (16 HNSCC and 16 OM). FAUST was applied to CD45+ live CD19−CD3− cells identified through manual gating. The markers CD3, CD19, CD45, PD-L2 and CD85k and the Live/Dead marker were excluded from the FAUST analysis to account for the manual analysis as well as observed autofluorescence in the detectors used for PD-L2 and CD85k. After tuning, FAUST selected the markers CD1c, CD11b, CD11c, CD14, CD16, CD32, CD38, CD40, CD68, CD80, CD86, CD123, CD141, CD163, CD206, CX3CR1, HLADR, PDL1 and SIRPA for discovery and annotation of phenotypes. Counts of the discovered phenotypes annotated as HLADR+ were tested for association with tissue type using a binomial generalized linear mixed-effects model with a subject level random effect. 21 phenotypes were associated with tissue type at the FDR-adjusted 0.05 level.
Unless stated otherwise, all data are represented as mean ± s.d. Statistical analyses between blood, OM and HNSCC samples were performed using one-way ANOVA with Tukey’s multiple comparisons test. P-values are shown in full, except if smaller than 0.0001. Statistical analysis was performed using GraphPad Prism (v9).
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.