Single cell RNAseq (cTEC sort) dataset

Description

Methods

Material Method

Thymic stromal cells were isolated from week 2-old Fam83h^wt/wt and Fam83h^-/- animals (see above). cTECs were identified as CD45- Ter119- EpCAM+ UEA- Ly.51+ cells and single-cell sorted into 384-well plates already containing 0.3 µL of lysis buffer and 3 µL of silicone oil (100 cSt, Sigma-Aldrich), using the Formulatrix Mantis. As a negative control, a subset of wells received no sorted cells.

Generation of smart-seq3xpress libraries (scRNA-seq)

Smart-seq3xpress libraries were prepared following published protocol involving reverse transcription and 14 cycles of PCR pre-amplification after sample storage at −80°C. Indexed libraries were generated by transferring cDNA into plates with dried primers, followed by Tn5 tagmentation and PCR amplification for 16 cycles. Libraries were purified using SeraMag and SPRIselect beads, quality-checked on a Fragment Analyzer, quantified by Qubit, and sequenced on an Illumina NextSeq 2000 with 2 × 109 bp paired-end and index reads.

Analysis of scRNAseq Data

Alignment of Raw Data

The raw read files were processed with zUMIs (version 2.7.9e), applying published parameters and settings documented in the supplemental file (Smartseq3xpress_SS3x_manuscript_all.yaml). Reads were aligned with STAR (version 2.7.3a) against mouse genome GRCm38, using gencode.vM8 annotation. Several downsampled count matrices were created, and for downstream analyses, a range of 10,000–20,000 reads per cell was used.

Data Processing

Subsequent steps were carried out in Seurat (version 5.2.1), focusing on 5′-end tagged reads. Reads mapping to spike-ins (ERCC and SIRV) were excluded, and cells were filtered based on quality metrics (e.g., gene count >1200, RNA count >2200, spike-in <12%, mitochondria <6%). SCTransform with glmGamPoi was used for normalization. Dimensionality reduction and clustering were done using the first 30 principal components, UMAP, and a resolution of 0.8. For integration, 50-dimension CCA via the IntegrateLayers function was applied. Seurat objects were converted to h5ad format and uploaded to Scarfweb. Data is available at https://www.nygen.io.