Immune and stromal cells from NSCLC and murine lung cancer

We provide Seurat objects with immune and stromal cells from NSCLC and murine lung cancer to enable reproduction of our analysis.<b>Isolation of stromal and immune cells from NSCLC</b>Fresh surgical specimens of NSCLC and unaffected autologous lung tissue were mechanically disaggregated and tissue was digested using a MACS human tumor dissociation kit (Miltenyi Biotec) according to the manufacturer’s protocol. Hematopoietic, erythroid and tumor cells were depleted by incubating the cell suspension with MACS anti-CD45, anti-CD235a (Glycophorin A) and anti-CD326 (EpCAM) MicroBeads (Miltenyi Biotec) and passing through a MACS LS column (Miltenyi Biotec) to enrich for non-hematopoietic cells. The flow-through was centrifuged and single cell suspensions were stained for further flow cytometric analysis or cell sorting. For the isolation of immune cells, tissues were gently smashed through a mesh using a syringe plunger and washed with RMPI1640 containing 2% FCS and 20 mM HEPES (Sigma-Aldrich) before staining for flow cytometric analysis or cell sorting.<b>Isolation of stromal and immune cells from murine lung tumors</b>Lung tissue including tumors or excised tumors (day 23) were mechanically disaggregated and tissue was digested using 1 mg/ml Dispase (Roche), 1 mg/ml Collagenase Type II (Gibco) and 25 μg/ml DNaseI (AppliChem) in combination with gentleMACS-based mechanical disruption (Miltenyi Biotech). After 30 minutes incubation at 37°C, cell suspensions were washed with PBS containing 0.5% FCS and 10 mmol/L EDTA. Enrichment of stromal cells was achieved by depleting hematopoietic and erythroid cells using MACS anti-CD45 and anti-Ter119 microbeads (Miltenyi Biotec). The flow-through was centrifuged and single cell suspensions were stained for further flow cytometric analysis or cell sorting. For the isolation of immune cells, tissues were gently smashed through a mesh using a syringe plunger and washed with PBS containing 0.5% FCS and 10 mmol/L EDTA before staining for flow cytometric analysis or cell sorting.<b>Droplet-based single cell RNA-seq analysis</b>Sorted cells were used for cDNA library generation performed following the established commercial protocol for Chromium Next GEM Single Cell 3’ GEM, Library &amp; Gel Bead Kit v2/3/3.1 (10X Genomics). Libraries were sequenced using a Novaseq 6000 sequencer at the Functional Genomic Center Zurich. Samples were collected from 7 patients and samples (CM,SM and LU) from the same patient were processed one batch. For murine cells, a total of 16 samples were processed (EYFP⁺ fibroblasts: 12 samples; GP33/34-Tetramer⁺CD8⁺ T cells: 4 samples). Gene expression estimation from sequencing files was done using CellRanger (v5.0.1) count with Ensembl GRCh38.102/GRCh38.103 releases as reference to build the index for human samples and Ensembl GRCm38.102/ GRCm38.103 releases for murine samples. Next, quality control was performed in R (v.4.1.0) using the R/Bioconductor package scater (v.1.16.0) and included removal of damaged and contaminating cells based on exceeding or low UMI counts (&gt;2.5 median absolute deviation from the median across all cells), exceedingly high or low total number of detected genes (&gt;2.5 median absolute deviation from the median across all cells) and high mitochondrial gene content (&gt; 2.5 median absolute deviations above the median across all cells). In addition, contaminating cells expressing any of the markers <i>FCER1A</i>, <i>ITGAX</i>, <i>ITGAM</i> (macrophages) and <i>CDH5</i>, <i>PECAM1</i> (endothelial cells) were removed from NSCLC stromal cell and immune cell samples. Following quality control, the human datasets included 6,676 immune cells (CM: 2,855 cells; SM: 3,821 cells) and 67,531 fibroblasts and endothelial cells (LU: 15,682 cells; CM: 26,924 cells; SM: 24,925 cells) or 5,541 CCL19+ cells (LU: 518 cells; CM: 2,996 cells; SM: 2,027 cells). For the analysis of murine fibroblasts, cells expressing transgenic Rosa26.EYFP and <i>Col1a1</i> and <i>Col1a2</i> expression were retained and contaminating cells expressing <i>Pecam1</i>, <i>Cdh5</i>, <i>Fcer1g</i> or <i>Krt8</i> were eliminated. Following quality control and contaminant removal, the LLC-gp33 dataset included 6,052 Ccl19-EYFP⁺ cells from naïve lungs (1,730 cells), lung tissue on day 15 (3,123 cells) and established tumors on day 23 (1,199 cells) distributed to Cd34⁺ AdvFB: 2,140 cells; Hhip⁺ AdvFB: 865 cells; Npnt⁺ AlvFB: 968 cells; SMC/PC: 1,276 cells; Rgs5⁺ PRC: 302 cells; Sulf1⁺ TRC: 501 cells). The mCOV-gp33-Flt3l immunized lung tumor dataset included 9,410 Ccl19-EYFP⁺ (TLS TRC: 248 cells; Sulf1⁺ TRC: 467 cells; Rgs5⁺ PRC: 351 cells; Cd34⁺ AdvFB: 2,902 cells; Hhip⁺ AdvFB: 2,864 cells; Npnt⁺AlvFB: 2,193 cells; SMC/PC: 385 cells). Cells with high expression of stress-associated genes (<i>Trp53inp1</i>, <i>Slc43a1</i>, <i>Usp54</i>, <i>Map3</i><i>k</i><i>6</i>, <i>Ddit4</i>, <i>Fabp4</i>) were removed from the murine samples. For the analysis of GP33/34⁺CD8⁺ T cells, contaminating cells expressing <i>Fcer1g</i> (macrophages) and high levels of mitochondrial genes were removed. Following quality control, the dataset contained 4,171 GP33/34-tetramer⁺CD8⁺ T cells (DTR^- tumor: 1,774 cells; DTR⁺ tumor: 2,397 cells, effector memory T cells: 1,701 cells; Ccr7⁺ T cells: 593 cells; effector T cells: 209 cells; cycling T cells: 739 cells; exhausted T cells: 929 cells).<b>Bioinformatics analysis</b>Downstream analysis was performed using the Seurat R package (v.4.3.0.1). For each experiment, acquired fibroblast and/or immune cell datasets were first merged and pre-processed including normalization, detection of the top 2,000 highly variable genes, scaling and centering, dimensionality reduction with PCA, TSNE and UMAP, graph-based clustering and cluster determination via calculation of unbiased cluster markers, as implemented in the scater R/Bioconductor package (version 1.16.0). Clusters were characterized based on the expression of calculated cluster markers and canonical marker genes as reported in previous publications. For FRC subset analysis <i>CCL19</i>-expressing fibroblasts were re-embedded and subsets were defined according to marker genes and tissue origin (e.g. SMC/PC or AdvFB in unaffected lung tissue and PRC or TRC in tumor margin). Functional signatures were further derived by running Gene Ontology (GO) enrichment analysis on subset-specific or condition-specific genes using the clusterProfiler R/Bioconductor package (v.3.15.3). Integration was performed using the IntegrateData function from the Seurat package.For interactome analysis, immune cells and <i>CCL19</i>+ fibroblasts were first merged and characterized using normalization, scaling, UMAP dimensionality reduction, graph-based clustering and calculation of unbiased cluster markers. <i>CD8</i>⁺ T cells were reembeded from total immune cell datasets and used for determination of interactomes with <i>CCL19</i>⁺ fibroblasts. Clusters were characterized based on the expression of calculated cluster markers and canonical marker genes. CellChat (version 1.6.1) was used as a tool to infer ligand–receptor interactions based on scRNA-seq data. CellChat applies a signaling molecule interaction database (CellChatDB) to predict intercellular communication patterns based on differentially over-expressed receptors and ligands. Known ligand-receptor complexes based on the KEGG Pathway database are used by the software to compute the communication probability on a pathway level. For the construction of multiple dimension scaling (MDS) plots, positions of cells were calculated by taking the mean Gaussian kernel for coordinates of cells from the calculated MDS representation for each celltype. Differentiation trajectories of <i>CCL19</i>⁺ FRCs in NSCLC were inferred using the monocle3 R/Bioconductor package (version 1.0.0). In brief, cells were clustered and the trajectory graph was learned using the LearnGraph function with default parameters. Pseudotime was inferred using the OrderCell function with root nodes defined from the mural SMC/PC and AdvFB clusters of the unaffected lung. Differentiation trajectories of murine Ccl19-EYFP⁺ FRCs were inferred using the slingshot R package (version 2.8.0) with AdvFBs and SMC/PC clusters of naive lungs defined as starting points and the TRC and PRC tumor specific clusters set as end points. Changes in gene expression along the inferred trajectories were analysed using the FitGAM function of the tradeSeq R/Bioconductor package (version 1.14.0) with default parameters to fit generalized additive models (GAMs) for the 1000 most variable genes. Lastly, significant associations between gene expression and pseudotime were determined using the StartVsEndTest function to calculate differential gene expression between the start and end of <i>Cd34</i>⁺AdvFB to TLS TRC trajectory (T3).