Transcriptional subtypes on immune microenvironment and predicting postoperative recurrence and metastasis in human pheochromocytoma and paraganglioma
Figures
Figure 1 with 2 supplements
Divergence in transcriptional subtypes defined three PPGL subtypes across different datasets.
(A) Overview of PPGL cohort (n = 87) and TCGA cohort (n = 179) detailing the number of cases by anatomical localization: HN-PG (head and neck paraganglioma), ME-PG (mediastinal paraganglioma), RP-PG (retroperitoneal paraganglioma), BC-PG (bladder paraganglioma), and PC (pheochromocytoma). Clinical data, whole-genome sequencing (WGS), and RNA-seq were used for subtype identification and correlation analysis related to recurrence/metastasis. (B) Heatmap depicting key metabolic and biological pathways associated with each subtype and displaying the pathway enrichment analysis for three subtypes of pheochromocytoma: C1 (HIF-1), C2 (inflamed), and C3 (metabolism). Each row represents a specific biological pathway, and the color intensity represents the degree of enrichment across the subtypes. (C) Heatmap showing gene expression changes across subtypes. Specific genes like ANGPT2 and CAV1 are upregulated in C1 compared to other subtypes. (D, E) Violin plots displaying the enrichment scores for Neuroendocrine and cell cycle among the three subtypes (C1, C2, and C3). (F) Violin plots displaying CDK1 among the three subtypes (C1, C2, and C3). (G–I) Violin plots displaying the enrichment scores for epithelial–mesenchymal transition (EMT), HIF, and inflamed among the three subtypes (C1, C2, and C3). Comparisons were calculated by one-way ANOVA (D–I). Data are presented as mean ± SD. ns, p > 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.
-
Figure 1—source data 1
PPGLs samples information.
- https://cdn.elifesciences.org/articles/107108/elife-107108-fig1-data1-v1.xlsx
Figure 1—figure supplement 1
Distribution of tumor site categories—pheochromocytoma (PC) vs paraganglioma (PG)—by transcriptional subtype (C1–C3) in the aggregated dataset (TCGA + Magnus).
Figure 1—figure supplement 2
Validation of PPGL transcriptional subtype scores in TCGA and Magnus cohorts.
(A) TCGA and (B) Magnus cohorts: heatmaps of weighted gene co-expression network analysis (WGCNA)-derived gene-set scores (Score_C1, Score_C2, and Score_C3) per sample; values are z-scores, with warmer colors indicating higher concordance with the corresponding subtype.
Figure 2
Mutational landscape and molecular characterization of PPGL cohorts.
(A) Overview of the genomic alterations across the three subtypes (C1, C2, and C3) in TCGA cohorts. Genes associated with pseudohypoxic, kinase, and other pathways are shown, with individual rows representing distinct gene mutations and columns corresponding to samples. Bar plot showing the proportional distribution of variant classifications. (B) Proportional analysis of the transcriptional subtypes across C1, C2, and C3. The pseudohypoxic, kinase, WNT-altered, and wild type (WT) subtypes (lacking pathogenic mutations) are depicted, illustrating the distribution of each subtype. (C) The number of patients with mutations in the pseudohypoxic-related genes (SDHB, VHL, EGLN1, EPAS1, SDHA, and SDHD) across the three subtypes. (D) The number of patients with mutations in kinase-related genes (NF1, RET, FGFR1, HRAS, MAX, and TMEM127) across subtypes. (E) Comparisons of microsatellite instability (MSI), tumor mutation burden (TMB), and MATH (mutant allele tumor heterogeneity) scores across the three subtypes. (F) Linear regression analysis of beta coefficients (PC1 and PC2) for different cohorts including PPGL cohort, TCGA, and Magnus’s cohort. Comparisons were calculated by one-way ANOVA (E). Data are presented as mean ± SD. ns, p > 0.05.
-
Figure 2—source data 1
TCGA+SNRNA_mutation.
- https://cdn.elifesciences.org/articles/107108/elife-107108-fig2-data1-v1.xlsx
Figure 3
Immune infiltration landscape across PPGL and TCGA cohorts.
(A) Heatmap displaying the immune cell composition and clinical features, including nerve and vascular infiltration, gender, and age across C1, C2, and C3 subtype. The heatmap highlights variations of immune cell infiltration abundance across different samples in immune cell types such as dendritic cells (DC), macrophages, B cells, and natural killer (NK) cells. (B) Representative immunohistochemistry images (top) and the absolute number (bottom) of infiltrated CD8+ T cells in C1, C2, and C3 subtypes. The infiltrated number of CD8+ tumor-infiltrating lymphocytes (TILs) was determined by randomly selecting 10 fields at 40×magnification and calculating the total count of CD8+ cells. The xCell analysis in the PPGL cohort (C) and TCGA cohort (D) depicting absolute abundance of immune cells across C1, C2, and C3 subtype. Th1, T helper 1 cells; Th2, T helper 2 cells; Mφ, macrophages; HSC, hematopoietic stem cell; EC, endothelial cell. Cytotoxic lymphocytes include both NK and T cells. The forest plot illustrates tumor-infiltrating cytotoxic lymphocytes (TILs) across different transcriptional subtypes in the PPGL cohort (E) and the TCGA cohort (F), using linear regression models. Model 1 was a crude model; Model 2 was adjusted for age, gender, and primary tumor location based on Model 1; Model 3 was adjusted for MKI67, SDHB, and SSTR2 expression based on Model 2. Each model is compared to subtype C1 as the reference group. Orange represents subtype C2, and green represents subtype C3. Comparisons were calculated by one-way ANOVA (B–D). Data are presented as mean ± SD. Ns, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.
Figure 4
UMAP analysis of cell types and pseudotime trajectories in different subtypes.
(A) UMAP plot showing the distribution of various cell types across three subtypes (C1, C2, and C3) and normal tissues. Each subtype is uniquely colored, with consistent coloring for the same subtype. Corresponding mutated genes are indicated. PC/AT, pheochromocytoma/adrenal tumor; NAMs, neuroblastic cells; ECs, endothelial cells; SCLCs, Schwann cell-like cells. (B) UMAP plot highlighting specific immune cells across three subtypes and normal tissues, with corresponding marker genes indicated. (C) UMAP plot highlighting specific immune cells in each subtype: C1, C2, C3, and normal tissues. (D) Bar plot represents the proportion of each immune cell type in the C1, C2, C3, and normal subtypes. (E) The dot plot represents a ligand–receptor interaction plot generated using CellChat, showcasing the communication network between various immune cell types across three different subtypes (C1, C2, and C3). The rows represent different immune cell populations, while the columns depict various ligands and receptors involved in cell–cell communication. The color gradient from red to green indicates interaction strength, with red representing stronger interactions and green representing weaker ones. The size of the dots reflects the significance level of each interaction. Distinct regions of enhanced interactions are highlighted with colored boxes, emphasizing subtype-specific signaling patterns. (F) Pseudotime analysis revealing the developmental trajectories of cell populations in different subtypes. Arrows indicate the progression paths of cell differentiation, suggesting different developmental stages and lineage decisions across C1, C2, C3, and normal subtype.
Figure 5 with 1 supplement
The specific genetic features of different PPGL subtypes.
(A) Violin plots depicting the expression levels of ANGPT2, PCSK1N, and GPX3 gene across three identified subtypes in PPGL cohort. (B) Immunohistochemistry images showing the protein expression of ANGPT2, PCSK1N, and GPX3 across subtypes C1, C2, and C3. Scale bars represent 50 µm. (C) Survival curves comparing the progression-free survival (PFS) of patients across three subtypes of PPGLs, integrating data from both the PPGL and TCGA cohorts. (D) The forest plot displaying the hazard ratios (HR) with 95% confidence intervals (CI) for PFS across PPGL subtypes using Cox regression analysis. Model 1 was a crude model; Model 2 was adjusted for age, gender, and race based on Model 1; Model 3 was further adjusted for tumor location and pathogenic mutation type based on Model 2. Each model is compared to subtype C1 as the reference group. Orange represents subtype C2, and green represents subtype C3. Comparisons were calculated by one-way ANOVA (A). Data are presented as mean ± SD. ns, p > 0.05; ****p < 0.0001.
Figure 5—figure supplement 1
Effect sizes of marker genes for PPGL transcriptional subtype classification.
The forest plots (A) show the effect sizes (β values) of marker genes as continuous variables for distinguishing subtype C1 from non-C1, (B) for distinguishing subtype C2 from non-C2, and (C) for distinguishing subtype C3 from non-C3, all based on linear regression.
Figure 6
The characteristics of ANGPT2 knockout.
(A) The ROC curve illustrates the diagnostic ability to distinguish ANGPT2 expression in PPGLs, specifically differentiating subtype C1 from non-C1 subtypes. The red dot indicates the point with the highest sensitivity (88.9%) and specificity (83.3%). AUC, the area under the curve. (B–D) Violin plots display the distribution of cell cycle scores, epithelial–mesenchymal transition (EMT) scores, and HIF scores in PPGLs between ANGPT2high and ANGPT2low Groups, using a cutoff value of 1.445 for ANGPT2 expression as shown in panel A. (E) Immunoblots showing ANGPT2 and Actin expression in ANGPT2 wild-type (ANGPT2WT) and knockout (ANGPT2‒/‒) rat PPGL cell line PC12. (F) The graph shows a series of tumors harvested from BALB/c nude mice xenografted with either ANGPT2WT or ANGPT2‒/‒ PC12 tumors over 16 days. (G) The tumor growth volume of ANGPT2WT or ANGPT2‒/‒ PC12 tumors in Balb/c nude mice. Data are presented as mean ± SD (n=7 mice per group). Statistical significance was assessed by unpaired two-tailed Student’s t-test. (H) Heatmap illustrates changes in gene expression related to cell proliferation, tumor metastasis, and hypoxia between ANGPT2WT or ANGPT2‒/‒ PC12 cells. (I–L) The dotplots illustrate EMT scores, tumor metastasis scores, cell proliferation scores, and hypoxia scores in ANGPT2 KO tumors. Comparisons were calculated by t tests (B–D, I–L). Data are presented as mean ± SD. ***p < 0.001; ****p < 0.0001.
-
Figure 6—source data 1
Original files for western blot analysis displayed in Figure 6E.
- https://cdn.elifesciences.org/articles/107108/elife-107108-fig6-data1-v1.zip
-
Figure 6—source data 2
Zip file containing original western blots for Figure 6E, indicating the relevant bands.
- https://cdn.elifesciences.org/articles/107108/elife-107108-fig6-data2-v1.zip
Author response image 1
Extended Data Figure A-B.
(A) The ROC curve illustrates the diagnostic ability to distinguish PCSK1N expression in PPGLs, specifically differentiating subtype C2 from non-C2 subtypes. The red dot indicates the point with the highest sensitivity (93.1%) and specificity (82.8%). AUC, the area under the curve. (B) The ROC curve illustrates the diagnostic ability to distinguish GPX3 expression in PPGLs, specifically differentiating subtype C3 from non-C3 subtypes. The red dot indicates the point with the highest sensitivity (83.0%) and specificity (58.8%). AUC, the area under the curve.
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Transcriptional subtypes on immune microenvironment and predicting postoperative recurrence and metastasis in human pheochromocytoma and paraganglioma
eLife 14:RP107108.
https://doi.org/10.7554/eLife.107108.3
