Enhanced Spatial Transcriptomics Analysis of Mouse Lung Tissues Reveals Cell-Specific Gene Expression Changes Associated with Pulmonary Hypertension

3.1. Fixed Frozen Mouse Lung Thin Section Mounting onto Xenium Slides The fluorescent in situ (FISH) followed by imaging-based spatial transcriptomics studies require thin tissue sections mounted on the defined imaging area on a slide. The Xenium pipeline, offered by 10X Genomics, is a leading commercial platform for these studies. To obtain high-quality imaging data on the Xenium platform, sample preparation is critical. Currently, the Xenium platform is validated for formalin-fixed paraffin-embedded (FFPE) thin tissue sections [7], although fresh frozen tissue sections are also compatible with the recommended protocols [8]. However, many laboratories working with various tissue sections typically follow a protocol involving PFA fixation, cryoprotection, cryosectioning of thicker sections (e.g., 40 μm), followed by immunohistochemistry (e.g., with multichannel antibody labeling) or in situ hybridization for mRNA species (e.g., RNAscope). It is unclear whether this protocol is compatible with the recommended workflow from 10X Genomics. We tested our fixed frozen thin mouse lung sections for their compatibility on the Xenium based spatial transcriptomics (Figure 1A-protocol outline, Table 1). Briefly (Figure 1B), the Mouse lung was first fixed by 4% PFA through transcardial perfusion. It was then dissected out and further post-fixed in 4% PFA overnight at 4 °C. The lung was then cryoprotected, embedded in OCT compound, and sectioned on a cryostat into 10 μm thickness. To facilitate mounting on the Xenium slide imaging area, we adopted a free-floating approach under a dissection microscope, in which a small drop of nuclease free PBS was placed on a corner of the Xenium slide imaging area. A thin lung section was transferred to the PBS meniscus using a fire-polished glass probe. By alternating the glass probe and small trimmed paint brush, the thin section was able to be flattened and guided to the desired location inside the PBS. Once a tissue section was flattened, the paint brush was used to withdraw PBS and dry the section onto a Xenium slide. Another drop of PBS was then introduced in an adjacent area to enable mounting another section. Repeating this approach, most of the imaging area of the Xenium slide was efficiently covered. Once the Xenium slide was fully mounted with sections, the tissue section layouts were recorded into metadata information for each section (e.g., mouse, age, sex, genotypes, etc.). The slide was stored in a slide mailer at −80°C for up to three weeks before being analyzed on a Xenium Analyzer, which was typically operated by a core facility or a commercial service provider. In contrast to this fixed frozen protocol, we also mounted another set of mouse lung sections using the 10X Genomics- validated ‘fresh frozen’ method. Freshly dissected lungs were embedded in OCT and sectioned on a cryostat at a thickness of 10 µm. Efforts were made to minimize tissue rolling, and the thin sections were carefully picked up from the sectioning platform of a cryostat and directly placed onto the aligned Xenium slide imaging area. Care was also taken to prevent tissue overlapping. Despite multiple attempts, obtaining a completely flat, large tissue section proved to be challenging. Additionally, a much larger area was left empty to avoid section overlapping (Figure 1C). 3.2. Initial Analyses and Clustering of Lung Cells through Identified RNA Transcripts Our previous studies employing single-cell RNA sequencing (scRNA-seq) and 10X Genomic Visium spatial transcriptomics analysis demonstrated that there was an increase in arterial endothelial cells (ECs) and a loss of distal capillary ECs in the lungs of Egln1^Tie2Cre mice, a severe pulmonary hypertension PH mice model with obstructive vascular remodeling [5,9]. Due to the limitation of losing spatial context and cell dissociation bias of scRNA-seq and the relatively lower resolution of Visium, we conducted a high-resolution Xenium in Situ Analysis using a predesigned 379-gene mouse panel of probes, which cover major cell types in the mouse tissue, to demonstrate the feasibility of using fixed frozen mouse sections with free-floating mounting onto Xenium slides (Figure 1A). The details of preparation of multiple fixed frozen and floating lung sections on Xenium slide were illustrated in Figure 1B and described in the Methods section. We mounted 5 lung sections onto one Xenium slide imaging area (Figure 2A), of which three sections were from three Egln1^Tie2Cre (KO) mice, and two sections were from littermate controls. Xenium slides were probed and stained with cell segmentation kits, followed by processing on Xenium Analysis. We conducted further analysis using Seurat [10], Scanpy [11], and Squidpy [12] packages on the Xenium Analysis data output. Each tissue section was provided as a separate dataset, with cell segmentation already implemented on the Xenium analyzer. For each dataset, unsupervised clustering resulted in 23 clusters that were well separated into the spatial and UMAP space (Figure 2B). There was a high level of top marker gene expressions for each cluster, as shown by the heatmap (Figure 2C). After integration of the five samples, expression levels of the top marker genes for the 20 unsupervised clusters were plotted as an additional heatmap (Figure 2D, further sorted by sample ID and Clusters). Next, we explored the spatial features of the tissue sections, including cell segmentation and the spatial embedding of selective transcripts. Cell segmentation utilized immunostaining and deep learning-based cell segmentation algorithms. Using a combination of nucleus (DAPI), cellular content, and cell boundary markers provided with the Xenium Multi-Tissue Stain Mix, the staining delineated cell boundaries, and imaging information was collected by the Xenium Analyzer instrument and processed to be segmented. Cell and tissue landmarks were further delineated through H & E staining. Figure 3A shows an overall staining of all five tissue sections acquired by the Xenium analyzer. In one control lung section (Figure 3B), unsupervised clustering revealed the spatial distribution of 23 clusters of cells. A selected area with vessel and alveolar space was further shown in Figure 3C. Figure 3D shows an enlarged view of a boxed region of Figure 3C, in which contours of segmented cells of different clusters (C1 through C23) were overlaid onto the stained tissue image. The locations and relative abundance of specific transcripts can be jointly visualized with nuclei staining and H&E staining at higher magnification. Figure 3E demonstrated an enlarged view of tissue section with detected cell types (“AT2”, “AT1”, “Fibroblasts”, “Macrophages”, “Endothelial cells”, “Airway epithelial cells”), with their spatial location of the top 5 marker gene transcripts overlaid on the H & E-stained tissue image. For example, cluster 1 expressed type 2 epithelial cell (AT2) markers Lamp3, Muc1, and Sftpd, and cluster 4 expressed fibroblast markers Itga8, Mfap4, Nox4, etc. The combined staining protocol and high-resolution imaging enabled the visualization of detailed anatomical structures from selected regions of interest (ROIs). We further selected two ROIs representing severe pulmonary vascular remodeling from a KO dataset and distal vessels and alveolars from a WT dataset. Cell segmentation led to the identification of cell boundaries, and with that, the mRNA transcripts were able to be assigned to individual cells. Figure 3F (H & E-brightfield) and Figure 3G (fluorescent staining used for cell segmentation) showed an arterial vessel with severe vascular remodeling from KO mice. Figure 3J,K showed a distal vessel and alveolars from WT mice. Based on top marker gene transcripts, we identified multiple cell populations including large vessel ECs (Vwf), capillary ECs (Kdr), smooth muscle cells (Myh11), fibroblasts (C3), macrophages (Mpeg1), pericytes (Rgs5), airway epithelial cells (Scgb3a2), type 1 epithelial cell (AT1, Ager), and AT2 cells (Lamp3) (Figure 3H,I,L). 3.3. High Resolution Spatial Transcriptomics Revealed Major Cell Types in the Mouse Lung The data delivery deck from the Xenium analyzer can be visualized using Xenium Explorer, which also allows spatial coordinates of selected ROIs and cell metadata information to be exported for further customized analysis. The delivered data can readily be analyzed with the Scanpy and Squidpy pipelines, which allow cell label transfer when mapping to the LungMap Mouse Reference v1.1 dataset [13,14]. We then performed unsupervised clustering for the transcripts from 2 WT and 2 KO datasets and identified 28 clusters and predicted 40 lung cell types, including AT1, AT2, CAP1, CAP2, etc. (Figure 4A). Annotated cell types onto their spatial coordinates were plotted in Figure 4A. For the four samples (control_1, control_2, KO_1 and KO_2), the UMAP embedding of all the cells after unsupervised clustering was shown in Figure 4B. In addition, annotated cell types from the same UMAP space were shown in Figure 4C. We further plotted the cell type transition from the 28 unsupervised clusters to the selected 20 most abundant annotated cell type clusters using the Sanky plot (Figure 4D). Marker genes in these annotated cell types were shown in heatmaps (Figure 4E). The cell proportion analysis (including the total number of 300,848 cells assigned to the predicted 40 annotated cell types) showed that AT2, AT1, AF1, ECs (AEC, VEC, CAP1, CAP2, EPC), pericytes and LECs are relatively abundant compared to other cell types (Figure 4F). The cell proportions observed in spatial transcriptomics were relatively more similar to the observed in vivo cell proportions obtained from scRNA-seq analysis [15]. 3.4. High Resolution Spatial Transcriptomics Identified Cell Type Specific Changes in the PH Lungs Next, we performed the Seurat V5 pipeline to further compare the cell type-specific changes, including the cell proportion change and differential gene expression (DEG) analysis between control and KO mice. Similar cell type annotation and spatial features were observed between Scanpy and Seurat analysis using UMAP and Spatial Plots (Figure 5A,B). We then evaluated the cell proportion change between control and KO mice. We found that many cell types were increased or decreased in the PH mice. For example, EC subpopulations such as EPC, AEC and VECs, macrophage subpopulations such as alveolar macrophages (AM) and interstitial macrophages (IM), fibroblasts subpopulations AF1 and SCMF, pericytes, and lymphatic ECs were increased in the KO mice, whereas CAP1, CAP2, T cells were reduced in the KO mice compared to WT mice (Figure 5C,D). We then performed the cell type specific DEGs analysis between control and KO mice. We found that mesenchymal markers, including Des, Mfap4, Pi16 were upregulated in the EC subpopulations (AEC, CAP1, CAP2, and VEC) from KO mice compared to control (Figure 5E), suggesting that ECs underwent endothelial mesenchymal transition in PH. Moreover, we also observed the increase of myofibroblast markers Myh11 in the KO fibroblast populations (AF1, AF2, SCMF) compared to WT (Figure 5F), indicating the activation of myofibroblasts in the KO lungs. 3.5. Increased Arterialization in Distal Microvasculature in PH by High Resolution Spatial Transcriptomics Our recent studies demonstrated that there was an increase of arterialization in the distal lung in the PH using scRNA-seq and Visium studies [9,16]. We further extracted the ECs from the Xenium data and confirmed that there is a loss of capillary ECs (including CAP1 and CAP2) and an increase in arterial EC (AEC) and endothelial progenitor cells (EPCs) [13] in PH (Figure 6A). Spatial mapping also demonstrated that the AECs and EPCs were increased in the distal microvasculature in the KO lungs compared to WT lungs (Figure 6B). We also plotted the individual markers for AEC, CAP1 and CAP2, and we found that AEC marker Sox17 was significantly increased (^∗∗∗ p < 0.001, Wilcoxon test), whereas CAP1 marker Plvap and CAP2 marker Prx were both reduced in the KO lungs (^∗∗∗∗ p < 0.0001, Wilcoxon test, Figure 6E). Immunostaining and RNASCOPE analysis further validated that CAP2 cells labeled by CD31 and Ednrb were significantly reduced in the KO lungs (Figure 6F).