Single Nuclei Sequencing of Early Postnatal Lung Specimens

Clinical interventions and inflammatory signaling shape the transcriptional and cellular architecture of the early postnatal lungThese are fully processed, integrated and annotated datasets from:23 histologically normal early postnatal (0 to 2 year) distal lung specimens: (FullEarlyPostnatalAtlas.RDS).4 distal lung specimens from patients diagnosed with "Evolving" or "Organized" Bronchopulmonary Dysplasia: (FullBronchopulmonaryDysplasiaAtlas.RDS).5 distal lung specimens from patients diagnosed with Pulmonary Interstitial Glycogenosis: (FullPulmInstGlycogenosisAtlas.RDS).3 distal lung specimens from 16, 21 and 23 weeks post-conception: (16_21_23weekAdditionalSamples.RDS).Code used in analysis of this data is available at: http://github.com/jason-spence-lab/Frum-et-al.-2025a.git METHODSSubmission of samples for single nucleus RNA-sequencing (snRNA-seq)Samples were stored in LN2 until preparation for snRNA-seq. Small pieces from each sample were shaved of a single region specimen and then minced into small rice grain sized fragments using a No.1 Scapel in a 6 cm dish on dry ice. Samples were transferred to 1.5 mL tubes, dissociated by pestle and then nuclei were purified using the 10X Chromium Nuclei Isolation Kit (10X Genomics, Cat#1000493) following the manufacturer’s recommendations, counted using a Countess Automated Cell Counter (v2) and resuspended at 1000 cells/µl in PBS + 1% BSA (Mitenyl, Cat#). The University of Michigan Advanced Genomics Sequencing Core (prepared libraries using the Chromium Next GEM Single Cell 3’ GEM, Library and Gel Bead Kit v3.1 (10X Genomics, Cat#PN1000128) targeting 7500 nuclei per specimen. snRNA-sequencing libraries were sequenced to a projected average read depth of 80,000 reads per nuclei using a NovaSeq 6000 with S4 300 cycle reagents. Computational Analysis of snRNA-seq dataAmbient RNA CorrectionReads were mapped to human genome (GRCh38-2020-A) and gene expression matrices generated using CellRanger v7. Raw matrices were used as input for CellBender v0.30, which re-called nuclei and corrected for ambient RNA at a false-positive rate of 0.01. CellBender corrected gene expression matrices were imported into Seurat v5 in RStudio v1.4 running R v4.1.Preprocessing/QC FilteringOnly nuclei within the following thresholds were considered for further analysis: between 500 to 7500 features, more than 1000 unique molecules sequenced, less than 5% mitochondrial RNA reads and less than 7.5% ribosomal reads. Data Integration, Dimensional Reduction, Clustering For each specimen data was normalized using Seurat::NormalizeData() and then all specimens were integrated using a mutual nearest neighbors batch correction implemented by SeuratWrappers::RunFastMNN() in SeuratWrappers v0.3.5 using 2000 features. The first 30 dimensions of the mutual neighbors reduction was used to generate a Uniform Manifold Approximation and Projection (UMAP) by Seurat::RunUMAP(). Nearest-neighbor graph construction was performed using Seurat::FindNeighbors(). Louvain clustering was performed at a resolution of 1.0 using Seurat::FindClusters(). Additional QCClusters were inspected for mutually exclusive expression of major cell class markers (Epithelial: CDH1, EPCAM;Endothelial: PECAM1, Immune: PTPRC, Mesenchymal: PDGFRA, PDGFRB). Clusters of cells at the center of the UMAP coexpressing markers of multiple major cell classes were removed. We speculate these data points are doublets or ambient RNA. After removal, Data Integration, Dimensional Reduction and Clustering were performed again. Cell Type AnnotationClusters were first annotated as Epithelium, Mesenchyme, Immune or Endothelium based on unique expression of the major cell class markers identified above (Fig. 1B). Each of these annotations was subclustered, using the following top dimensions of the integrated mutual nearest neighbors reduction calculated by Seurat::RunFastMNN() on the complete dataset as input for Seurat::RunUMAP() and Seurat::FindNeighbors(): Epithelium: 20, Mesenchyme: 25, Immune: 16, Endothelium: 15. Louvain clustering was performed using Seurat::FindClusters() at the following resolutions: Epithelium: 0.8, Mesenchyme: 0.4, Immune: 0.3, Endothelium: 0.15. At this point clusters were annotated to minor cell classes based on known markers (i.e. Airway vs Alveolar, Lymphoid vs. Myeloid, etc) (Fig. 1C, D). Some of these minor cell classes were further subclustered to achieve a cell type level annotation (Vessels, Lymphoid, Myeloid), while all others were annotated on the cluster structure evident at the first round of subclustering. Cell type annotations (Fig. 1E) were consistent with known markers (Fig. 1H) of cell type identity.Full Preprint: https://doi.org/10.1101/2025.10.17.683116