Multiomics analysis unveils the cellular ecosystem with clinical relevance in aldosterone-producing adenomas with KCNJ5 mutations

Abstract

Aldosterone-producing adenomas (APA), a major endocrine tumor and leading subtype of primary aldosteronism, cause secondary hypertension with high cardiometabolic risks. Despite potentially producing multiple steroid hormones, detailed cellular mechanisms in APA remain insufficiently studied. Our multiomics analysis focusing on APA with KCNJ5 mutations, which represent the most common genetic form, revealed marked cellular heterogeneity. Tumor cell reprogramming initiated from stress-responsive cells to aldosterone-producing or cortisol-producing cells, with the latter progressing to proliferative stromal-like cells. These cell subtypes showed spatial segregation, and APA exhibited genomic intratumor heterogeneity. Among the nonparenchymal cells, lipid-associated macrophages, which were abundant in APA, might promote the progression of cortisol-producing and stromal-like cells, suggesting their role in the tumor microenvironment. Intratumor cortisol synthesis was correlated with increased blood cortisol levels, which were associated with the development of vertebral fractures, a hallmark of osteoporosis. This study unveils the complex cellular ecosystem with clinical relevance in APA with KCNJ5 mutations, providing insights into tumor biology that could inform future clinical approaches.

Keywords: aldosterone-producing adenomas; cellular heterogeneity; lipid-associated macrophages; single-cell RNA-sequencing; spatial transcriptomics.

Fig. 1.

Cellular and spatial distribution profiles of APA. (A) UMAP plot of 53,007 cells from eight APA with KCNJ5 mutations categorized into nine main cell types in the scRNA-seq data. (B) UMAP plot illustrating the kernel density estimation of representative marker genes across the main cell types. (C) Dot plot showing representative marker genes across the main cell types, with dot size indicating the fraction of cells expressing specific genes, and color intensity denoting relative gene expression levels. (D) Hematoxylin and eosin–stained image of spatial transcriptomics sections from four APA, corresponding to scRNA-seq data. [Scale bar, 500 μm (Top Left).] Spatial distribution of areas for representative sample (APA2) based on: clustering and marker gene annotation (Top Right), histological annotation (Bottom Left), and the predominant cell type determined by single-cell deconvolution analysis (Bottom Right).

Fig. 2.

Cellular heterogeneity within APA. (A) UMAP plot of 27,418 subsets and reclustered tumor cells (APA cells) from eight samples, categorized into eight clusters (Left). UMAP plot colored by MC (Center). Heatmap displaying the Spearman correlation coefficient of aggregate signature scores, based on DEG, across APA cells for each cluster, with hierarchical clustering (Right). (B) Heatmap showing representative DEG for each MC, with color intensity denoting average gene expression levels. (C) Heatmap showing scaled gene set scores representing hallmarks of transcriptional intratumor heterogeneity across tumors for each MC. *, FDR < 0.05. (D) Scatter plot showing TF activity in MC2 (X axis) and MC3 (Y axis), with TF activated in MC1 represented as individual points. Colors indicate TF activated in each MC. (E) Scatter plots correlate MC2/MC3 signature score, from bulk RNA-seq data of APA (n = 26), with maximum tumor diameter, using Spearman’s test, significance at P < 0.05. (F) Multicolor immunofluorescence staining for CYP11B2 (green), CYP17A1 (magenta), and CYP11B1 (yellow), and DAPI (blue) in APA (n = 3). Representative images (APA3) are presented. (Scale bars, 200 μm.) (G) Heatmap showing representative DEG for MC1. (H) Bar plot of representative pathways enriched in MC1 from the single cell rank-based pathway enrichment analysis. (I) UMAP plot colored according to the APA cell subtypes.

Fig. 3.

Lineage dynamics and regulatory networks of APA. (A) UMAP plots are colored by APA cell subtypes, with the beginning of the trajectory marked by a black point (Left) and colored by pseudotime (Right). (B) UMAP plots of each lineage and line plots showing the expression patterns of representative genes (Left) and enriched pathways (Right) along pseudotime, with the density of expression-based APA cell subtypes above. (C) TF activated in MC1 (Glucocorticoid) cells (Left). Network diagram showing the representative target genes for FOSL2 (Right). (D) RNAscope showing coexpression patterns of CYP17A1 (green) and FOSL2 (red) (n = 2), with a representative image (APA1). [Scale bars, 2.5 mm (Left) and 25 µm (Right).] (E) Multicolor immunofluorescence staining of CYP11B2 (cyan), FOSL2 (yellow), GJA1 (magenta), and DAPI (blue) at the Left, and CYP17A1 (green), FOSL2 (yellow), GJA1 (magenta), and DAPI (blue) at the Right in APA (n = 4 for each). Representative images (APA3) are presented. (Scale bars, 20 µm.) Box plot comparing the proportion of triple-positive cells (CYP11B2/FOSL2/GJA1 or CYP17A1/FOSL2/GJA1). Each group represents 20 regions of interest (ROI) from four APA samples (*P < 0.05). (F) RT-qPCR of mRNA expression levels of FOSL2, CYP11B2, CYP17A, and GJA1 in GFP control (CTL) and FOSL2-overexpressing H295R cells (FOSL2-OE). Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001.

Fig. 4.

Spatial transcriptomic and genomic heterogeneity within APA. (A) UMAP plot of 9,976 spots within the APA area from four samples (Top) and spatial cluster distribution (Bottom) classified into 12 clusters based on signature scores for APA cell subtypes from scRNA-seq data. (B) Box plot illustrates the signature score for APA cell subtypes in spatial clusters. Group differences were analyzed using the Kruskal–Wallis test, with post hoc Bonferroni correction (*P < 0.05). (C) Concentration areas of APA cell subtypes with a representative image (APA2). (D) Multisampling WES was performed on four FFPE samples from APA. Five regions per sample were randomly selected and dissected using a 3 mm biopsy punch (Left). Variant allele frequency (VAF) distribution of somatic mutations detected in five distinct tumor regions of a representative sample (APA4) is shown (Right). The KCNJ5 driver mutation (red) is ubiquitously present across all regions. Shared passenger mutations (green) are detected in two or more regions, while region-specific passenger mutations (gray) are unique to individual regions. The y axis represents VAF, and the x axis shows the distribution of mutations in each region (Right).

Fig. 5.

LAM as a key player of the TME in APA. (A) UMAP plot of 6,963 myeloid cells showing five myeloid cell subtypes (Top). Stacked bar plot of cell subtype proportions in APA and NT, adjusted P < 7.1e-06 (0.05/6963) for significance (Bottom). (B) Heatmap showing representative DEG in myeloid cell subtypes. (C) Pie charts displaying myeloid cell subtype proportions in bulk RNA-seq data from APA (n = 26) and NT (n = 6), analyzed using the Student's t test, P < 0.05. (D) Multicolor immunofluorescence staining of CD68 (red), GPNMB (yellow), and DAPI (blue) was performed on the APA and NT areas (n = 8 and n = 4, respectively). Representative images (APA3) are presented. White arrows indicate the LAM. [Scale bars, 500 µm (Left) and 20 µm (Right).] Box plot comparing the proportion of LAM (CD68/GPNMB) in the APA and NT areas (*P < 0.05). (E) Heatmap of scaled enrichment scores for representative pathways in the main myeloid cell subtypes.

Fig. 6.

Cellular crosstalk within the TME in APA. (A) Scatter plot depicting the outgoing and incoming interaction strengths of APA cells and nonparenchymal cells. (B) Heatmap showing interaction strengths per cell subtype (Left) and dot plot ranking of sender cells by interaction strength, with LAM as the receiver (Right). (C) Spatial feature plot of the LAM signature score based on the DEG per spot in a representative image (APA2) (Left). Scatter plots correlating the LAM signature score with the signature scores of APA cell subtypes per spot, using Spearman’s test, with significance indicated at r > 0.20 and P < 0.05 (Right). (D) Multicolor immunofluorescence staining for CD68 (red), GPNMB (yellow), CYP17A1(green), and DAPI (blue) was performed on APA (n = 8). Representative images (APA3) are presented. White arrows indicate the LAM. [Scale bars, 500 µm (Top) and 20 µm (Bottom).] (E) Scatter plots correlate LAM proportion in myeloid cell subtypes, from bulk RNA-seq data of APA (n = 26), with serum cortisol levels and maximum tumor diameter, using Spearman’s test, significance at P < 0.05.

Fig. 7.

Intratumor cortisol synthesis and clinical impact in APA. (A) Schematic of steroid pathways, highlighting the 28 steroid metabolites analyzed. (B) PLS-DA scores plot differentiating APA patients with and without VF (n = 11 and n = 16, respectively); the dots represent individual samples. (C) Dot plot showing the VIP score of component 1 in the PLS-DA model. *, VIP score > 1. (D) Box plot of tumor volume adjusted-tissue cortisol levels in APA (n = 9), CPA (n = 11), and controls (n = 5) (*P < 0.05). (E) Scatter plot correlating tumor volume adjusted-tissue cortisol levels with DST cortisol, using Spearman’s test (*P < 0.05). (F) Representative MS images of cortisol in the APA (APA7), CPA, and control (GN).