Combined multidimensional single-cell protein and RNA profiling dissects the cellular and functional
来源：24直播网
Animals were maintained under specific pathogen–free conditions at the University of Oxford Biomedical Science facilities under local and United Kingdom Home Office regulations and permissions. They were housed in cages grouped to a maximum of five animals per cage on wood litter. Environmental enrichment such as tunnel, gnawing and nesting material was provided. The animal house was maintained under artificial lighting (12 h) between 7:00 am and 7:00 pm, in a controlled ambient temperature of 22 °C ± 2 °C, and relative humidity between 30% and 70%. Experiments performed were approved by the United Kingdom Home Office regulations and co-housed age- and gender-matched wild-type C57BL/6 mice were used in all experiments as a reference for genetically modified animals.
CsnbCre mice15 were crossed to the Rosa26YFP mouse line36 to induce lineage tracing in the mature mTEC compartment.
Mice heterozygous for a Foxn1 allele with a single nucleotide loss at position 1470 (designated FOXN1Δ505/WT) were generated at the Genome Engineering Facility of the MRC Weatherall Institute of Molecular Medicine, University of Oxford as previously described35.
Rag2−/− mice were bred and maintained in the mouse facility of the Department of Biomedicine at the University of Basel in accordance with permissions and regulations of the Cantonal Veterinary Office of Basel-Stadt.
For timed pregnancies 7- to 14-week-old mice were mated over-night and separated early next morning. For pregnant females the mating was considered E0.5 that morning.
Isolated thymi were cleaned from adipose tissue, separated into the two lobes, and subsequently subjected to three rounds of enzymatic digestion with Liberase (2.5 mg/ml, Roche, Cat no: 5401127001) and DNaseI (10 mg/ml, Roche, Cat no: 10104159001) diluted in PBS (Gibco, Cat no: 70011044) at 37 °C. After filtration through a 100-μm cell strainer and resuspension in FACS buffer (PBS supplemented with 2% FBS), cell number was determined using a CASY cell counter (Innovatis). For most analyses CD45+ hematopoietic cells were depleted by incubation with anti-CD45 beads (Miltenyi) as per manufacturer’s recommendations and subsequently subjected to the AutoMACS separator (Miltenyi) “depleteS” program.
Cells were counted and stained in FACS buffer containing antibodies of interest for 30 min at 4 °C in the dark. For the identification of dead cells an additional staining with propidium iodide (PI, Sigma, Cat no: P4864) or Zombie red (Biolegend, Cat no: 423110) was used. For intra-cellular staining, cells were fixed and permeabilised after cell-surface staining using the Cytofix/Cytoperm (BD Biosciences, Cat no: 554714) or the Transcription Factor Staining Buffer Set (Invitrogen, Cat no: 00-5523-00) according to the manufacturer’s protocol. Cells were analysed and sorted on a BD FACSAria III instrument (BD Biosciences). Cells were sorted into FACS buffer. Cell purities of at least 95% were confirmed by post-sort analysis. The following antibodies were used: CD4-APCCy7 (1:400, GK1.5, Biolegend), CD4-FITC (1:1000, GK1.5, Biolegend), CD5-PerCPCy5.5 (1:400, 53-7.3, Biolegend), CD8α-AF700 (1:800, 53-6.7, Biolegend), CD40-FITC (1:200, 3/23, Biolegend), CD45-AF700 (1:400, 30-F11, Biolegend), CD63-PE (1:400, NVG-2, Biolegend), CD66a-APC (1:400, Mab-CC1, Biolegend), CD66a-FITC (1:400, Mab-CC1, Biolegend), CD69-PECy5 (1:800, H1.2F3, Biolegend), CD69-PECy7 (1:200, H1.2F3, Biolegend), CD73-BV421 (1:400, TY/11.8, Biolegend), CD80-PECy5 (1:2000, 16-10A1, Biolegend), CD80-BV605 (1:400, 16-10A1, Biolegend), CD83-Bio (1:200, Michel-19, Biolegend), CD83-PE (1:400, Michel-19, Biolegend), CD104-FITC (1:400, 346-11A, Biolegend), CD117-BV421 (1:200, 2B8, BD Biosciences), CD146-APC (1:800, ME-9F1, Biolegend), EpCMA1-PerCPCy5.5 (1:800, G8.8, Biolegend), Dclk1 (1:1000, ab31704, Abcam), HVEM-APC (1:400, HMHV-1B18, Biolegend), Ly51-PECy7 (1:400, 6C3, Biolegend), MHCII-APC/Fire750 (1:2000, M5, Biolegend), UEA1-Cy5 (1:400, Vector Laboratories, in-house labelling), Sca1-BV510 (1:800, D7, Biolegend), TCRβ-FITC (1:400, H57-597, Biolegend), TCRβ-PE (1:1000, H57-597, Biolegend), Tspan8-APC (1:400, FAB6524A, R&D Systems). Biotinylated antibodies were detected using Streptavidin-BV785 (1:500, Biolegend) and unlabelled Dclk1 using anti-rabbit IgG AF647 (1:4000, Invitrogen). Data was analysed using FlowJo (version 10).
Cells were isolated and CD45 depletion plus backbone staining were performed as described. The surface backbone panel included antibodies directed against CD45 (1:400, 30-F11, AF700, Biolegend), EpCAM1 (1:800, G8.8, PerCPCy5.5, Biolegend), Ly51 (1:400, 6C3, PECy7, Biolegend), MHCII (1:2000, M5, BV510, Biolegend), CD40 (1:200, 3/23, PECy5, Biolegend), CD80 (1:400, 16-10A1, BV605, Biolegend)), CD86 (1:800, GL-1, BV650, Biolegend), Sca1 (1:1000, D7, BV785, Biolegend), Podoplanin (1:200, 8.1.1, APC, Biolegend), CD31 (1:1000, 390, FITC, Biolegend), the Ulex europaeus agglutinin I (UEA1) lectin labelled with biotin (1:1000, Vector Laboratories), followed by secondary streptavidin-BV421 (1:1000, Biolegend) staining and Zombie red (1:1000, Biolegend) staining. Subsequently, the stained cells were distributed across the three 96-well plates provided with the LEGENDScreen kit (Biolegend, Cat no: 700009), each well containing a unique PE-labelled exploratory antibody as well as isotype controls and blanks. PE-labelled antibodies targeting GP2, Tspan8, CD177 and F3 were used as additional exploratory surface antibodies. Due to the low cell numbers obtained after CD45 depletion only ¼ of the recommended quantity of exploratory antibodies was used. Plates were incubated at 4 °C for 30 min in the dark. Thereafter, fixation was performed using the Cytofix buffer (BD Biosciences, Cat no: 554714) for 1 h at 4 °C in the dark. As an additional backbone marker, cells were stained intracellularly for anti-AIRE (1:400, 5H12, AF750, Invitrogen) in Cytoperm buffer (BD Biosciences, Cat no: 554714) and one well stained with anti-FOXN137 (1:1600, 2/41, PE, kind gift from Hans-Reimer Rodewald) as an additional exploratory marker, over-night at 4 °C in the dark. The next day cells were resuspended in 100 μl FACS buffer before analysis.
For the Infinity Flow computational analysis of the LEGENDScreen datasets, the acquired fcs files were gated on CD45 negative cells or specifically on EpCAM1+ TEC using the FlowJo software. The newly exported fcs files were then used as the dataset for the Infinity Flow pipeline as recently published21. The augmented data matrices generated during this process were then further analysed using the Seurat package for hierarchical clustering of the cells and differential expression analysis24, mostly following the workflow presented in https://satijalab.org/seurat/articles/pbmc3k_tutorial.html. We used default parameters, except for the data normalization method, normalization.method = “CLR”. Values below zero were set to zero to allow for log normalization. Markers were filtered by hand to exclude T-cell related and focus on stromal cell related genes (Supplementary Table 1). The top 20 PCA dimensions were used for clustering and UMAP projections and a clustering resolution of 0.275 was used.
We compared the Infinity Flow data matrices with the scRNAseq dataset of reference 13 by identifying the most closely related genes for each Infinity Flow protein. However, some of the antibodies bind to protein complexes, and here we chose the most abundant transcript related to such a complex—for example, we chose H2-Ab1 RNA transcripts for MHC class II protein detection. Furthermore, UEA1 detection was identified via Fut1 RNA expression, since FUT1 synthesises the glycan target of UEA1; for the complete assignment see Supplementary Table 2. We compared the Infinity Flow fluorescence values with the scRNAseq normalised log counts. Clusters from each dataset were then compared using the SingleR package in R38, with the Wilcox ranked sum test (using the SingleR option de.method = “wilcox”).
Frozen thymus tissue sections (7 µm) were fixed in acetone and stained using antibodies specific for CD69 (1:100, H1.2F3, Biolegend), Ly51 (1:200, 6C3, Biolegend), K8 (1:500, TROMA-1, NICHD supported Hybridoma Bank), K14 (1:500, Poly19053, Biolegend). Images were acquired using a Leica DMi8 microscope.
Perinatal cTEC (CD45−EpCAM1+MHCII+Ly51+CD83+CD40+) and non-perinatal cTEC (CD45−EpCAM1+MHCII+Ly51+CD83−CD40−) were sorted from the thymi of 2-week-old C57BL/6 mice and put in co-cultures with CD69− DP thymocytes sorted from the same thymi, respectively. Cells were transferred in a 1:1 TEC to DP ratio into 1.5 mL tubes containing 1 mL Iscove’s modified Dulbecco’s medium (IMDM, Gibco, Cat no: 12440053) supplemented with 10% FBS, 100 units/mL penicillin and 100 μg/mL streptomycin (Sigma, Cat no: P4333-100ML) and 1× GlutaMAX supplement (Gibco, Cat no: 35050061). Co-cultures were maintained at 37 °C in a humidified atmosphere containing 10% CO2 for 48 h and then analysed by FACS. As a control DP cells were also cultured without the addition of TEC.
6- to 7-week-old Rag2−/− animals were injected intraperitoneally with 50ug of anti-CD3ε (clone KT3, in-house produced) or HBSS. Four weeks post injection thymi were analysed for the appearance of DP thymocytes and for changes within their cTEC compartment.
Cells were isolated from six thymi of 1-week- and three thymi of 16-week-old C57BL/6 mice and depleted of CD45+ cells by AutoMACS. Subsequently cells were stained for CD45 (1:400, 30-F11, AF700, Biolegend), EpCAM1 (1:800, G8.8, PerCPCy5.5, Biolegend), Ly51 (1:400, 6C3, PECy7, Biolegend), Ter119-FITC (1:400, TER119, Biolegend) and with PI. In addition cells were stained with antibodies coupled to oligonucleotides directed against CD9 (MZ3, Totalseq-A, Biolegend), CD40 (3/23, Totalseq-A, Biolegend), CD49a (HMα1, Totalseq-A, Biolegend), CD54 (YN1/1.7.4, Totalseq-A, Biolegend), CD63 NVG-2, Totalseq-A, Biolegend), CD73 TY/11.8, Totalseq-A, Biolegend), CD83 (Michel-19, Totalseq-A, Biolegend), CD117 (2B8, Totalseq-A, Biolegend), CD146 (human with cross reactivity to mouse) (P1H12, Totalseq-A, Biolegend), CD200 (OX-90, Totalseq-A, Biolegend), CD274 MIH6, Totalseq-A, Biolegend), HVEM (HMHV-1B18, Totalseq-A, Biolegend), Ly6D (49-H4, Totalseq-A, Biolegend), Ly6C/Ly6G (Gr1) (RB6-8C5, Totalseq-A, Biolegend), MadCAM1 (MECA-367, Totalseq-A, Biolegend), Podoplanin (8.1.1, Totalseq-A, Biolegend), CD80 (16-10A1, Totalseq-A, Biolegend), CD86 (GL-1, Totalseq-A, Biolegend), MHCII (M5, Totalseq-A, Biolegend), Sca1 (D7, Totalseq-A, Biolegend), CD31 (390, Totalseq-A, Biolegend), EpCAM1 (G8.8, Totalseq-A, Biolegend), CD36 (HM36, Totalseq-A, Biolegend), CD133 (15-2C11, Totalseq-A, Biolegend), CD157 BP-3, Totalseq-A, Biolegend), CD300LG (ZAQ5, Totalseq-A, Biolegend), and the Ulex europaeus agglutinin I (UEA1) lectin labelled with biotin, followed by secondary staining with streptavidin-PE (Totalseq-A, Biolegend) coupled to an oligonucleotide. CD45−Ter119−EpCAM1+ and CD45−Ter119−EpCAM1− cells were sorted in a 70–30% ratio into a 1.5 mL tube containing FACS buffer for the 1-week-old and 16-week-old samples, respectively. For both timepoints an estimate of 28,000 total cells were loaded on two wells of a 10x Genomics Chromium Single Cell Controller. After single-cell capture cDNA and library preparation were performed according to the manufacturer’s instructions using a Single-Cell 3’ v3 Reagent Kit (10x Genomics) with the changes as described in26 to capture cDNA and produce libraries from antibody-derived oligos (ADT). Sequencing was performed on one lane of the Illumina NovaSeq 6000 system with a mix of 90% cDNA library and 10% ADT library resulting in 151nt-long paired-end reads.
The dataset was analysed by the Bioinformatics Core Facility, Department of Biomedicine, University of Basel. cDNA reads were aligned to ‘mm10’ genome using Ensembl 102 gene models with the STARsolo tool (v2.7.9a) with default parameter values except the following parameters: soloUMIlen=12, soloBarcodeReadLength=0, clipAdapterType=CellRanger4, outFilterType=BySJout, outFilterMultimapNmax=10, outSAMmultNmax=1, soloType=CB_UMI_Simple, outFilterScoreMin=30, soloCBmatchWLtype=1MM_multi_Nbase_pseudocounts, soloUMIfiltering=MultiGeneUMI_CR, soloUMIdedup=1MM_CR, soloCellFilter=None. ADT libraries were also processed using the STARsolo tool with default parameters except soloCBmatchWLtype=1MM_multi_Nbase_pseudocounts, soloUMIfiltering= MultiGeneUMI_CR, soloUMIdedup=1MM_CR, soloCellFilter=None, clipAdapterType=False, soloType=CB_UMI_Simple, soloBarcodeReadLength=0, soloUMIlen=12, clip3pNbases = 136.
Further analysis steps were performed using R (v4.1.2). Note that cell filtering was done based only on the analysis of the gene expression, not ADT abundance. Cells were considered as high-quality cells if they had at least 2000 UMI counts, which is the threshold derived from the distribution of UMI counts across cells, forming a data set of 9953 cells.
Multiple Bioconductor (v3.14) packages including DropletUtils (v1.14.2), scDblFinder (v1.8.0), scran (v1.22.1), scater (v1.22.0), scuttle (1.4.0) and batchelor (v1.10.0) were applied for the further analysis of the dataset mostly following the steps of the workflow presented at https://bioconductor.org/books/release/OSCA/. Normalised39 log-count values for the gene expression were used to construct a shared nearest-neighbour graph40, which nodes, i.e., cells, were clustered by ‘cluster_louvain’ method from the R igraph package41. Counts reflecting the ADT abundance in cells were also log-normalised and clustered in a similar manner.
The data set was subjected to the cell-type annotation using the Bioconductor package SingleR (v1.8.1) and samples from the Immunological Genome Project (ImmGen) provided by the Bioconducter package celldex (v1.4.0) as the reference. Clusters of cells mostly assigned to ‘Epithelial cells’ (5834 cells) were filtered (Supplementary Fig. 6d). Note that one of the clusters (cluster A, Supplementary Fig. 6a) was excluded at this step, because it was mostly composed of cells with elevated percentage of reads mapping to mitochondrial and ribosomal genes and lower number of counts.
The gene expression of filtered cells was re-analysed by removing the batch effect formed by the combination of the sample of origin and the number of counts per cell (cells with >12,000 counts and cells with 12,000 counts) and re-clustered (Fig. 5a, b). Cells were also subjected to the cell-type annotation using scRNAseq transcriptional profiles of single TEC as the reference data set13 (Fig. 5d, f). The scoreMarkers function of the scran package was applied to find marker genes of clusters 1–3. The standardised log-fold change across all pairwise comparisons ‘mean.logFC.cohen’>1 was used as the significance threshold defining the set of marker genes.
A t-SNE dimensionality reduction was used for visualizing single cells on two dimensions. T-SNE coordinates were calculated using the runTSNE function from the scater package and default parameters. For the visualization of cells based on the gene expression, coordinates of principal components and 2000 most variable genes with excluded mitochondrial and ribosomal genes were used as the input. For the visualization of cells based on the ADT abundance, coordinates of principal components and all ADTs were used as the input.
Triplicates of Sca1−CD63−CD66a+CD117+ tuft-like and Sca1−CD63−CD66a−CD117− non-tuft-like mTEC cells were sorted from thymi of 6-week-old C57BL/6 mice into trizol (Invitrogen, 15596026). Subsequently samples were submitted to ultra-low input bulk RNAseq. Reads were trimmed using Trimmomatic (version 0.36) to remove adapter sequences and aligned to the mouse genome (mm10) using STAR (version 2.7.3a)42,43. HTSeq (version 0.12.4) was used to assign reads to genes with the option “intersection-nonempty”44. Differentially expressed genes were identified using edgeR (FDR < 0.05)45. Spearman correlation coefficients were calculated between Sca1−CD63−CD66a−CD117− non-tuft-like mTEC bulk RNAseq samples and scRNAseq data from reference 13 for all differentially expressed genes with log2 fold change ≥1.
GraphPad Prism (version 9) was used to perform all statistical analyses, except for the bulk RNAseq and CITEseq datasets. The statistical tests used are described in the figure legends, and exact p-values are shown within each figure. Non-significant differences are not specified.
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
