Spatial metabolomics reveals glycogen as an actionable target for pulmonary fibrosis

Abstract

Matrix assisted laser desorption/ionization imaging has greatly improved our understanding of spatial biology, however a robust bioinformatic pipeline for data analysis is lacking. Here, we demonstrate the application of high-dimensionality reduction/spatial clustering and histopathological annotation of matrix assisted laser desorption/ionization imaging datasets to assess tissue metabolic heterogeneity in human lung diseases. Using metabolic features identified from this pipeline, we hypothesize that metabolic channeling between glycogen and N-linked glycans is a critical metabolic process favoring pulmonary fibrosis progression. To test our hypothesis, we induced pulmonary fibrosis in two different mouse models with lysosomal glycogen utilization deficiency. Both mouse models displayed blunted N-linked glycan levels and nearly 90% reduction in endpoint fibrosis when compared to WT animals. Collectively, we provide conclusive evidence that lysosomal utilization of glycogen is required for pulmonary fibrosis progression. In summary, our study provides a roadmap to leverage spatial metabolomics to understand foundational biology in pulmonary diseases.

Fig. 1. MALDI imaging of complex carbohydrates is a multidimensional dataset.

a Metabolic pathways of complex carbohydrate metabolism inside a cell. Anabolic metabolism of glycogen and N-linked glycans through the endoplasmic reticulum (ER) and Golgi are critical for a multitude of cellular functions. Created with BioRender.com. b Schematic of the workflow for multiplexed imaging of glycogen and N-glycans using formalin-fixed paraffin-embedded (FFPE) specimens. Tissues are sectioned onto a microscope slide and are co-treated with peptide: N-glycosidase F (PNGase F) to cleave and release N-glycans and isoamylase to cleave α-1,6-glycosidic bonds releasing linear oligosaccharide chains. Following application of α-cyano-4-hydroxycinnamic acid (CHCA) ionization matrix, samples are analyzed by MALDI and quadrupole time-of-flight mass spectrometry. Created with BioRender.com. c Left: Schematic of molecular ions recorded in each pixel after laser desorption ionization. Right: Spatial heatmap of a specific biological feature (m/z) extrapolated from total ion current (TIC) after MALDI-MSI of tissue sections. Created with BioRender.com. d TIC representing the sum of all pixels after MALDI-MSI of a human liver section. e Spatial heatmap images of selected complex carbohydrate ions (m/z) corresponding to either glycogen or N-linked glycans selected from (d). m/z values and molecular structures of selected oligosaccharide chain and N-linked glycans are on top of the heatmap. An adjacent liver section stained with hematoxylin and eosin (H&E) is presented below the heatmap for 1743 m/z. Scale bar: 2 mm. f TIC extracted from three unique pixels based on the spatial heatmap presented in (e). g Simplified flow chart for the high-dimensionality reduction and spatial clustering (HDR-SC) analysis workflow. The user input module includes MALDI-MSI and data curation, followed by the computer-based module that utilizes unsupervised clustering and spatial annotation of unique clusters. Created with BioRender.com. h Representative input data for HDR-SC presented as heatmap to show heterogeneous complex carbohydrate abundance. 5% of total pixels (columns) and mixture of matrix and carbohydrate m/z (rows) were shown as heatmap for ease of visualization. i UMAP and spatial plots of the human liver specimen by HDR-SC analysis. j UMAP and spatial plots showing matrix only cluster from the human liver specimen by HDR-SC analysis. k UMAP and spatial plots after matrix removal.

Fig. 2. High-dimensionality reduction and spatial clustering (HDR-SC) analysis of tissue sections from idiopathic pulmonary fibrosis (IPF) and COVID-19 patients.

a Hematoxylin and eosin (H&E) staining of an immediate adjacent IPF section used for MALDI-MSI for histopathology assessment. Scale bar: 5 mm. b All identified clusters visualized by UMAP (left) and spatial plots (right) in an immediate adjacent IPF section. c Annotation of spatial clusters to histopathology by a panel of board-certified pathologists in an immediate adjacent IPF section. d Zoomed in images of H&E staining of IPF section showing diffuse alveolar damage (DAD), end-stage fibrosis, and mucin aggregates in an immediate adjacent IPF section. Scale bar: 500 μm. e UMAP (left) and spatial (right) plots highlighting DAD, end-stage fibrosis, and mucin aggregate clusters. f H&E staining of an immediate adjacent COVID-19 section used for MALDI-MSI for histopathology assessment. Scale bar: 5 mm. g All identified clusters visualized by UMAP (left) and spatial plots (right) in an immediate adjacent COVID-19 section. h Annotation of spatial clusters to histopathology by a panel of board-certified pathologists in an immediate adjacent COVID-19 section. i Zoomed in images of H&E staining of COVID-19 section showing DAD, end-stage fibrosis, and mucin aggregates in an immediate adjacent COVID-19 section. Scale bar: 500 μm. j UMAP (left) and spatial (right) plots highlighting DAD, end-stage fibrosis, and mucin aggregate clusters. k Differentially expressed features among HDR clusters, columns represent pixels within each cluster and row is the glycogen or N-glycan feature overrepresented. Glycogen (yellow) and core fucosylated N-linked glycan (pink) features were overrepresented in fibrotic clusters. l Schematic of glycogen accumulation in pulmonary fibrosis. Created with BioRender.com. m H&E staining of additional IPF (n = 3, left) and COVID-19 (n = 2) tissue sections used for MALDI-MSI for histopathology assessment. Scale bar is below in (n). n Spatial distribution and heatmap of glycogen chain length +7 (1175 m/z) in additional patient tissues shown in (m). Scale bar: 5 mm. o Immunofluorescent/co-localization analysis of glycogen and alpha smooth muscle actin (α-SMA) from an adjacent 20 μm section of IPF specimen shown in (a). Tissue is stained with glycogen (red), α-SMA (green), and DAPI (blue).

Fig. 3. Pulmonary fibrosis patient tissue exhibits aberrant complex carbohydrate features.

a Representative spatial distribution and heatmap of glycogen chain length +7 (1175 m/z) in normal (N) and fibrotic (F) lung patient tissue cores from TMA. Scale bar: 2 mm. b Total glycogen abundance in normal and fibrotic lung patient tissue measured by MALDI (Normal: n = 7, Fibrosis: n = 26). Values are presented as mean +/− standard error. p-value was calculated using two-tailed t-test. c Glycogen chain length abundance in normal and fibrotic lung patient tissue (Normal: n = 7, Fibrosis: n = 26). Values are presented as mean +/− standard error. d Multivariate analysis of glycogen and N-linked glycan features in normal and fibrosis lung patient samples by partial least squares-discriminant analysis (PLS-DA) displaying 95% confidence intervals. e Unsupervised clustering heatmap analysis of the top 25 glycogen and N-linked glycan features in normal and fibrotic lung patient. f Multivariate receiver operating characteristic (ROC) curve of all glycogen and N-linked glycan features between normal and fibrosis patients. g (Left) Representative spatial distribution and heatmap of 1809 m/z in N and F lung patient tissue. (Right) Relative abundance of 1809 m/z in N and F lung patient tissue. Molecular structure of the selected N-linked glycan is to the right of the graph. h (Left) Representative spatial distribution and heatmap of 1485 m/z in N and F lung patient tissue. (Right) Relative abundance of 1485 m/z in N and F lung patient tissue. Molecular structure of the selected N-linked glycan is to the right. i (Left) Representative spatial distribution and heatmap of 2012 m/z in N and F lung patient tissue. (Right) Relative abundance of 2012 m/z in N and F lung patient tissue. Molecular structure of the selected N-linked glycan is to the right. j (Left) Representative spatial distribution and heatmap of 2122 m/z in N and F lung patient tissue. Scale bar: 2 mm. (Right) Relative abundance of 2122 m/z in N and F lung patient tissue. Molecular structure of the selected N-linked glycan is to the right. g–j Values are presented as mean +/− standard error (Normal: n = 7, Fibrosis: n = 26). p-value was calculated using two-tailed t-test.

Fig. 4. Bleomycin-induced lung fibrosis shares same complex carbohydrate features as human fibrosis.

a Schematic of bleomycin-induced lung fibrosis model. Created with BioRender.com. b Zoomed in images of hematoxylin and eosin (H&E) staining of lung tissue from control (saline) and bleomycin-treated mice. Scale bar: 200 μm. c (Left) H&E staining of whole lung tissue from control (saline) (top) and bleomycin-treated (bottom) mice. (Right) Spatial distribution and heatmap of N-linked glycans: 1809, 2012, and 1485 m/z in lung tissue from control (saline) (top) and bleomycin-treated (bottom) mice. Scale bar: 2 mm. d Total abundance of N-linked glycans: 1809, 2012, and 1485 m/z in lung tissue from control (saline) and bleomycin-treated mice. Molecular structure of the selected N-linked glycans are above their respective graphs (n = 3 animals/group). Values are presented as mean +/− standard error. p-value was calculated using two-tailed t-test. e (Left) H&E staining of whole lung tissue from control (saline) (top) and bleomycin-treated (bottom) mice. (Right) Spatial distribution and heatmap of glycogen chain length +7 (1175 m/z) in lung tissue from control (saline) (top) and bleomycin-treated (bottom) mice. Scale bar: 2 mm. f Glycogen chain length abundance in lung tissue from control (saline) and bleomycin-treated mice. (n = 3 animals/group). Values are presented as mean +/− standard error. p-value was calculated using two-way ANOVA following by multiple comparisons testing. g Total glycogen abundance in lung tissue from control (saline) and bleomycin-treated mice. Values are presented as mean +/− standard error. p-value was calculated using two-tailed t-test. h Immunofluorescent/co-localization analysis of glycogen and alpha smooth muscle actin (α-SMA) from an adjacent 20; μm section of PF mouse lung shown in c. Tissue is stained with glycogen (red), α-SMA (green), and DAPI (blue) following by whole slide scanning and visualized using the HALO software. Zoomed in view shown below, and the white box in the glycogen panel represents the field of view magnified. Scale bar: 200 μm and 50 μm, respectively. i Schematics of shared glycogen and N-linked glycogen phenotype between mouse and human PF. Created with BioRender.com.

Fig. 5. Loss of laforin blunts complex carbohydrate perturbations during bleomycin-induced lung injury.

a bleomycin-induced lung fibrosis model in wild-type (WT) and Epm2a^−/− (LKO) mice. Created with BioRender.com. b Weight change (g) in control (saline) and bleomycin-treated WT and LKO mice during the course of the study. c Kaplan–Meier analysis of overall survival in WT and LKO bleomycin-treated mice. d Spatial distribution and heatmap of glycogen chain length +7 (1175 m/z) in lung tissue from control and bleomycin-treated WT and LKO mice. e Glycogen chain length abundance in lung tissue from control and bleomycin-treated WT and LKO mice (n = 2 animals for WT-saline, n = 3–4 animals/groups for others). Values are presented as mean +/− standard error. f Total glycogen abundance in lung tissue from control and bleomycin-treated WT and LKO mice (n = 2 animals for WT-saline, n = 3–4 animals/groups for others). Values are presented as mean +/− standard error. p-value was calculated using one-way ANOVA followed by multiple comparisons test. g Representative images of hematoxylin and eosin (H&E) staining of lung tissue from control and bleomycin-treated WT and LKO mice. h Ashcroft scoring for lung tissue from control and bleomycin-treated WT and LKO mice (n = 2 animals for WT-saline, n = 3–5 animals for other groups). Values are presented as mean +/− standard error. p-value was calculated using one-way ANOVA followed by multiple comparisons test. i Total collagen from control and bleomycin-treated WT and LKO mice (n = 2 animals for WT-saline, n = 3–5 animals for other groups). Values are presented as mean +/− standard error. p-value was calculated using one-way ANOVA followed by multiple comparisons test. j Schematic of MALDI-imaging of collagen peptides. Created with BioRender.com. k Spatial distribution and heatmap of collagen peptide 1247 m/z and 843 m/z in lung tissue from control and bleomycin-treated WT and LKO mice and total abundance. (Putative amino acid sequence, collagen subtype, and # of hydroxylate proline sites (HYP) on are displayed in white text above representative images (n = 2 animals for WT-saline, n = 3–5 animals for other groups). Values are presented as mean +/− standard error. p-value was calculated using one-way ANOVA followed by multiple comparisons test.

Fig. 6. Glycogen utilization by lysosomal-GAA during PF in vivo.

a Immunofluorescent/co-localization analysis of glycogen and acid alpha-glucosidase (GAA) from an adjacent section of human IPF lung shown in Fig. 2a. Tissue is stained with glycogen (red), GAA (green), and DAPI (blue). b Immunofluorescent/co-localization analysis of glycogen and LAMP2 from an adjacent section of mouse PF lung. Tissue is stained with glycogen (red), LAMP2 (green), and DAPI (blue). c Schematic of lysosomal salvage pathway of glycogen by GAA to provide substrates for other complex carbohydrates. Created with BioRender.com. d Schematic of bleomycin-induced lung fibrosis model in wild-type (WT) and Gaa^−/− mice (GKO). Created with BioRender.com. e Representative images of hematoxylin and eosin (H&E) staining of lung tissue from control (saline) and bleomycin-treated WT and GKO mice. f Ashcroft scoring for lung tissue from control and bleomycin-treated WT and LKO mice (n = 6–10 animals/group). Values are presented as mean +/− standard error. p-value was calculated using one-way ANOVA followed by multiple comparisons test. g Glycogen chain length abundance in lung tissue from control and bleomycin-treated WT and LKO mice (n = 6–10 animals/group). Values are presented as mean +/− standard error. h Total glycogen abundance in lung tissue from control and bleomycin-treated WT and LKO mice (n = 6–10 animals/group). Values are presented as mean +/− standard error. p-value was calculated using one-way ANOVA followed by multiple comparisons test. i Total abundance of 1444 m/z and 2012 m/z in lung tissue from control and bleomycin-treated WT and GKO mice. Molecular structure of the selected N-linked glycan is to the right of the graph (n = 6–10 animals/group). Values are presented as mean +/− standard error. p-value was calculated using one-way ANOVA followed by multiple comparisons test. j Total abundance of 1444 m/z and 2012 m/z in lung tissue from control and bleomycin-treated WT and LKO mice. Molecular structure of the selected N-linked glycan is to the right of the graph (n = 2 animals for WT-saline, n = 3–5 animals for other groups). Values are presented as mean +/− standard error. p-value was calculated using one-way ANOVA followed by multiple comparisons test. k Schematics of lysosomal salvaging of glycogen as a critical component of PF progression. Created with BioRender.com.