Primary aldosteronism: spatial multi-omics mapping of genotype-dependent heterogeneity and tumor expansion of aldosterone-producing adenomas

Abstract

Background:

Primary aldosteronism is frequently caused by an adrenocortical aldosterone-producing adenoma (APA) carrying a somatic mutation that drives aldosterone overproduction. APAs with a mutation in KCNJ5 (APA-KCNJ5^MUT) are characterized by heterogeneous CYP11B2 (aldosterone synthase) expression, a particular cellular composition and larger tumor diameter than those with wild type KCNJ5 (APA-KCNJ5^WT). We exploited these differences to decipher the roles of transcriptome and metabolome reprogramming in tumor pathogenesis.

Methods:

Consecutive adrenal cryosections (7 APAs and 7 paired adjacent adrenal cortex) were analyzed by spatial transcriptomics (10x Genomics platform) and metabolomics (in situ matrix-assisted laser desorption/ionization mass spectrometry imaging) co-integrated with CYP11B2 immunohistochemistry.

Results:

We identified intra-tumoral transcriptional heterogeneity that delineated functionally distinct biological pathways. Common transcriptomic signatures were established across all APA specimens which encompassed 2 distinct transcriptional profiles in CYP11B2-immunopositive regions (CYP11B2-type 1 or 2). The CYP11B2-type 1 signature was characterized by zona glomerulosa gene markers and was detected in both APA-KCNJ5^MUT and APA-KCNJ5^WT. The CYP11B2-type 2 signature displayed markers of the zona fasciculata or reticularis and predominated in APA-KCNJ5^MUT. Metabolites that promote oxidative stress and cell death accumulated in APA-KCNJ5^WT. In contrast, antioxidant metabolites were abundant in APA-KCNJ5^MUT. Finally, APA-like cell subpopulations- negative for CYP11B2 gene expression- were identified in adrenocortical tissue adjacent to APAs suggesting the existence of tumor precursor states.

Conclusions:

Our findings provide insight into intra- and inter-tumoral transcriptional heterogeneity and support a role for prooxidant versus antioxidant systems in APA pathogenesis highlighting genotype-dependent capacities for tumor expansion.

Graphical Abstract

INTRODUCTION

Aldosterone-producing adenomas (APAs) are a major subtype of primary aldosteronism (PA) in which a single somatic mutation drives the constitutive upregulation of CYP11B2 gene transcription (encoding aldosterone synthase) and aldosterone biosynthesis in adrenal zona glomerulosa (zG) cells.¹ These mutations include variants in the GIRK4 K⁺ channel² (encoded by KCNJ5), the calcium channels CaV1.3^3,4 and CaV3.2⁵ (CACNA1D and CACNA1H, respectively) and the CIC-2 chloride channel⁶ (CLCN2), as well as in subunits of the ion pumps Na⁺/K⁺-ATPase^4,7 (ATP1A1), and the plasma membrane Ca²⁺ transporting ATPase⁷ (ATP2B3). In most reports, APA mutations in KCNJ5 predominate over mutations in other genes.^8,9

Cells with the potential to produce aldosterone in adrenal tissues can be localized using specific antibodies to CYP11B2 in immunohistochemistry of adrenal sections,^10–12 which has provided new insight into APA pathogenesis from a histological perspective.^11,13,14 KCNJ5-mutated APAs (APA-KCNJ5^MUT) display distinct phenotypes from other APA (APA-KCNJ5^WT). These differences include the intra-tumoral heterogeneous CYP11B2 expression that is frequently observed in the tumor areas of APA-KCNJ5^MUT by immunostaining.^15,16 Moreover, APA-KCNJ5^MUT display a predominance of lipid-rich clear cells (zona fasciculata [zF]-like cells with CYP11B1 [11β-hydroxylase] and CYP17A1 [17β-hydroxylase/17,20-lyase] expression) rather than the prevailing small compact eosinophilic cell phenotype (zG-like cells with CYP11B2 expression) in APA-KCNJ5^WT.¹⁷ Further, APA-KCNJ5^MUT are generally larger than APA-KCNJ5^WT,^9,18 suggesting divergent mechanisms of tumorigenesis.¹⁹ However, transcriptional networks that determine the APA-KCNJ5^MUT phenotype remain to be elucidated.

Spatial transcriptomics is a powerful approach to dissect the transcriptional diversity of complex biological systems within the morphological context.^20,21 In addition, spatial metabolomics using matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI) of paraffin-embedded tissue sections has been used to investigate adrenal gland physiology²² and pathophysiology^23,24 and has shown distinct metabolome patterns of APA-KCNJ5^MUT compared with APA-CACNA1D^MUT.²³ However, the use of paraffin-embedded tissue can result in the decreased detection of functionally relevant lipid species, which are lost during the tissue embedding process.^25,26

Herein, we combined spatial transcriptomics and MSI-based spatial metabolomic profiling of adrenal tissue cryosections to define the role of transcriptomic and metabolomic reprogramming in APA pathophysiology. We hypothesized that the elucidation of biological processes which differentiate APA-KCNJ5^MUT and APA-KCNJ5^WT can identify causal molecular and cellular mechanisms of APA histological heterogeneity and tumor growth.

MATERIALS AND METHODS

The expanded methods are available in the Data Supplement. The authors declare that all supporting data are available within the article and Data Supplement or are available upon reasonable request. The study was performed in accordance with local ethics committee guidelines and was approved by the institutional review board at the Ludwig Maximilian University of Munich (ref. 379–10). All patients provided written informed consent. Details of patients, diagnosis of PA,^11,27–30 adrenal tissue samples and histology,¹⁰ sample processing,^24,31–33 genotyping,^34,35 methods for spatial transcriptomics and spatial metabolomics,^36–39 and bioinformatics are provided in the Online Supplement.

RESULTS

Intra-tumoral transcriptional heterogeneity of APAs

To delineate and compare APA intra-tumoral transcriptional heterogeneity stratified by genotype, we conducted spatial transcriptomics (Visium, 10X genomics) of adrenal cryosections from APA-KCNJ5^MUT (n = 4) and APA-KCNJ5^WT (n = 3) (Figure 1 and Table S1). First, we deconvolved APA-KCNJ5^MUT samples using a non-negative matrix factorization (NNMF) approach to quantify factors that were preferentially co-expressed by subsets of spots representing transcriptionally distinct regions within each sample specimen (Figure S1), as exemplified by NNMF of the APA section from Patient 1 (Figure 2A). Of the 3 tumor-related transcriptome signatures (factors 1, 3 and 4), factors 1 and 3 exhibited high CYP11B2 expression and defined structures with aldosterone hypersecretion. Gene ontology (GO) analysis revealed a functional difference between factors 1 and 3, with enrichment of cell death-related pathways in factor 1 (Figure 2B) and of pathways associated with metabolic processes in factor 3 (Figure 2C). The tumor region annotated by factor 4 was negative for CYP11B2 expression and associated with tissue remodeling, defined by the expression of genes of the extracellular matrix (IGFBP7, MGP), angiogenesis and cell population proliferation (Figure 2D). Factors 2 and 5 were transcriptional signatures associated with non-tumor regions of the adjacent adrenal tissue (Figure S2).

Figure 1. Adrenal tissue sections for spatial multi-omics profiling.

Illustration of adrenal samples used for spatial datasets. For patients 1–4 and 6–7, an adrenal section with an APA (section 1) and a separate section of adjacent adrenal tissue (section 2 showing adrenal zones, zG, zona glomerulosa; zF, zona fasciculata; zR, zona reticularis) was used. For patient 5, section 1 comprised APA tissue only, section 2 encompassed adjacent adrenal and an APA region. Tumor regions are outlined with a black dashed line. An overview of analytical approaches used is also shown (Upper panel). Successive cryosections were processed for spatial transcriptomic (10x Genomics Visium platform), CYP11B2 IHC to identify ROIs (Middle panel, ROIs indicated by dashed lines, scale bar, 1 mm), and MALDI-MSI. An overview of the data analysis strategy is also shown (Bottom panel). IHC, immunohistochemistry; MALDI-MSI, matrix-assisted laser desorption/ionization mass spectrometry imaging; ROIs, regions of interest. Illustration used SMART servier medical art (https://smart.servier.com/).

Figure 2. Intra-tumoral transcriptional heterogeneity of an APA with a KCNJ5 mutation.

Non-negative matrix factorization expression-based analysis identified the top genes defining factors (transcriptionally distinct regions) in adrenal sections with an APA. The figure shows the analysis of patient 1 carrying a KCNJ5 mutation; the same approach was used for the analysis of patients 2 to 7. A, Spatial CYP11B2 expression and spatial distribution of tumor-related factors, colour-codes indicate gene expression levels. Factor 1, 3 and 4, are tumor-related factors: factor 1 and factor 3 comprised gene sets preferentially co-expressed in subsets of CYP11B2 expressing tumour regions. CYP11B2, PCP4, MT3, GAPDH and HSD3B2 were the top genes defining factor 1; FDX1, HSPE1, FTH1, HSPD1, and STAR, as well as CYP11B2 defined factor 3; conversely factor 4 defined tumor regions without CYP11B2 gene expression with enrichment of angiogenesis and cell proliferation. Scale bar, 1 mm. B-D, Geno Ontology analysis showing enriched biological processes associated with factors related to “aldosterone hypersecretion” (factor 1 and factor 3) and “tissue remodeling” (factor 4).

CYP11B2 heterogeneity characterized by a lower proportion of cells with CYP11B2-positive immunostaining is often seen in APA-KCNJ5^MUT.^15,16 To identify molecular profiles which distinguish diverse CYP11B2 expression levels in APA-KCNJ5^MUT, tumor spatial transcriptomic spots were annotated into 2 clusters after subdivision into high and low CYP11B2 expression (Figure S3A). Gene and pathway enrichment analysis predicted by PROGENy showed that high CYP11B2 expressing transcriptome populations displayed upregulated steroidogenesis-related genes- like PCP4, FDX1, MC2R- and steroidogenesis-related pathways- like VEGF signaling, p53 signaling, and androgen signaling (Figure S3B–D). In the CYP11B2-expressing regions of Patients 4 and 6, a single APA-KCNJ5^WT tumor-related transcriptome signature was identified. In contrast, 3 distinct transcriptome signatures were annotated in the CYP11B2-expressing tumor regions of Patient 2 (factors 2, 3 and 4) (Figure S4 and Figure S5). Together these data validate the previously reported tumor cell heterogeneity in APAs¹⁶ and illustrate the intra-tumoral diversity of transcriptional programs.

Spatial profiling of APA reveals 2 distinct transcriptional profiles in CYP11B2-immunopositive adrenal regions.

Hierarchical cluster analysis identified common transcriptional signatures (or factors) across all 7 APAs and highlighted 2 distinct transcriptional signatures in CYP11B2-expressing regions (referred to here as CYP11B2-type 1 and CYP11B2-type 2) (Figure 3A). The CYP11B2-type 1 and CYP11B2-type 2 signatures each comprised genes which were markers of different cell types of the adrenal cortex (from the zG, zF and zona reticularis [zR]) (Table S2). Thus, the CYP11B2-type 1 region primarily encompassed genes expressed in the zG, such as KCNJ5 and VSNL1, and showed functional enrichment of hallmark gene sets- from MSigDB (Molecular Signatures Database hallmark- associated with cell stress (defined by the TP53 pathway, reactive oxygen species pathway and apoptosis). The CYP11B2-type 2 region expressed the zF and zR genes CYP17A1, CYP11B1, and SULT2A1, and showed enrichment of pathways related to mTORC1 and hypoxia-associated signals (Figure 3B). CYP11B2-type 1 was dominant in APA-KCNJ5^WT, consistent with their predominance of compact eosinophilic cells with CYP11B2 and KCNJ5 expression (zG-like cells)^7,9. CYP11B2-type 2 was dominant in APA-KCNJ5^MUT coherent with their higher proportion of lipid-rich clear cells with CYP11B1 and CYP17A1 expression (zF-like cells)^7,9 (Figure 3C).

Figure 3. Co-integrated spatial transcriptomics and metabolomics of APAs.

A, Spatial-weighted correlation analysis and Ward.D2 hierarchical clustering of 7 APA (P01-P07) identified 2 distinct signatures with CYP11B2 expression (CYP11B2-type 1 and CYP11B2-type 2). B, HALLMARK gene set enrichment of CYP11B2-type 1 and CYP11B2-type 2. C, Distribution of KCNJ5 mutated (APA-KCNJ5^MUT) and wild type APA (APA-KCNJ5^WT) according to CYP11B2-type 1 and CYP11B2-type 2 transcriptomic signatures. D, MetaboAnalyst 5.0 pathway enrichment analysis of discriminative metabolites in APA-KCNJ5^MUTversus APA-KCNJ5^WT. Annotated pathways selected by unadjusted P<0.05 (hypergeometric test). Circles indicate pathway impact (horizontal axis, topology analyses; vertical axis, enrichment). E, CYP11B2 immunohistochemistry and MALDI-MSI showing spatial distribution of pro-ferroptotic metabolites in 2 APAs with or without a KCNJ5 mutation. Scale bars, 1 mm.

Spatial metabolomics was used to further characterize genotype-related heterogeneity. We used in situ MALDI-MS imaging of all adrenal specimens (7 APAs and their paired adjacent cortex) to quantify and visualize metabolites and identify the enriched metabolomic profiles in tumor regions. We identified 207 discriminative masses: 162 enriched in APA-KCNJ5^WT and 45 abundant in APA-KCNJ5^MUT. APA-KCNJ5^WT were characterized by metabolomic signatures related to linoleic acid metabolism and the biosynthesis of polyunsaturated fatty acids (PUFAs eg., adrenic acid and docosapentaenoic acid) and an abundance of phosphatidylethanolamine PE (40:4), PE (38:4), and PE (40:6) (Figure 3D and 3E; Table S3). In contrast, APA-KCNJ5^MUT exhibited abundant metabolites of glutamine and glutamate metabolism. The accumulation of ferroptosis-promoting PUFAs and PE in APA-KCNJ5^WT support a role for activated cell death mechanisms to restrict tumor cell expansion.^40,41 In contrast, elevated pools of glutamine and glutamate in APA-KCNJ5^MUT reflect the presence of active metabolic signals favorable for tumor survival and growth.^42,43

Spatial profiling of APA and adjacent adrenal tissue uncovers mechanisms underlying genotype-related differences in tumor size.

We investigated the transcriptomic and metabolomic divergence of APA from paired adjacent adrenal tissue. For this, 14 cryosections were processed for spatial multi-omics profiling comprising the 7 APAs described above, plus a section of adrenocortical tissue adjacent to each APA. The basic structure of APA and adjacent adrenal was captured by NNMF,⁴⁴ which identified 3 transcriptomic structural landscapes classified as adjacent zF and zR, adrenal medulla, and a mixture of APA and adjacent zG demonstrating that different morphological regions can have the same basic structure with shared transcriptional programs (APA-KCNJ5^MUT, Figure S6 and Figure 4A; APA-KCNJ5^WT, Figure S7 and Figure 4B). Multiple biological processes of metabolism and steroid biosynthesis were identified in APA-KCNJ5^MUT whereas abundant cell death pathways with anti-tumoral metabolic signatures were highlighted in APA-KCNJ5^WT (Figure 4C and 4D). These observations were further supported by spatial metabolomic profiling which identified the enrichment of glucose metabolism in APA-KCNJ5^MUT versus adjacent adrenal cortex demonstrating active pathways for protection from steroidogenesis-mediated oxidative stress, including the oxidative arm of the pentose phosphate pathway and pentose glucuronate interconversions (Figure 4E and Figure S8; Table S4).

Figure 4. Spatial transcriptomics and metabolomics of APAs and adjacent adrenal tissue.

A-B, Heatmap of top genes contributing to each factor (transcriptionally distinct region) of APA-KCNJ5^MUT and APA-KCNJ5^WT. Genes indicated in red represent zonation markers of the zona glomerulosa (zG) (CYP11B2, DAB2, and KCNJ5); markers of the zona fasciculata (zF) or zona reticularis (zR) (CYP11B1 and CYP17A1); and of the adrenal medulla (CHGB, CHGA and TH). C-D, Gene ontology analysis (biological process) of tumor-related factors (factor 2 in APA-KCNJ5^MUT, Figure S6; factor 3 in APA-KCNJ5^WT, Figure S7). E-F, pathway enrichment analysis of upregulated metabolites in APA-KCNJ5^MUT (left) or APA-KCNJ5^WT (right) versus adjacent adrenal (unadjusted P<0.05, hypergeometric test).

Analysis of APA-KCNJ5^WT versus the adjacent adrenal cortex identified enrichment of omega-3 (for example, alpha-linolenic acid and docosahexaenoic acid) and omega-6 (for example, docosapentaenoic acid and gamma-linolenic acid) PUFAs (Figure 4F and Figure 5; Table S5) whereas metabolites of glutathione metabolism were preferentially localized in APA-KCNJ5^MUT (Figure S9). Therefore, the enrichment of PUFAs in APA-KCNJ5^WT and glutathione in APA-KCNJ5^MUT highlight prooxidant versus antioxidant responses indicating genotype-related diverse capacities for tumor expansion.

Figure 5. MALDI-MSI identification of upregulated omega-3 and omega-6 polyunsaturated fatty acids in KCNJ5 mutated APA compared with adjacent adrenal cortex.

A, corresponding H&E-stained tissues of KCNJ5 mutated APA (Upper) or KCNJ5 wildtype APA (Lower) and adjacent adrenal gland used for MALDI-MSI. Outlined regions in black indicate APA tumor areas with positive CYP11B2 immunostaining. B-C, representative images showing spatial distribution of omega-3 polyunsaturated fatty acids (B) and omega-6 polyunsaturated fatty acids (C). Outlined regions in white represent the borders of the tissue section.

APA-like subpopulations in the adjacent adrenal cortex

Tissue-covered transcriptomics spots in APA and adjacent cortex stratified by APA genotype were integrated and corrected for batch effects. TSNE (t-distributed stochastic neighbor embedding) identified 6 unique transcriptome subpopulations in adrenals (APA + adjacent cortex) from patients with APA-KCNJ5^MUT and 7 unique subpopulations in adrenals from patients with APA-KCNJ5^WT (Figure 6). Subpopulations were annotated based on CYP11B2 immunostaining (Figure 1), expression of marker genes (Figure S10, Figure S11 and Table S6), and gene ontology analyses. The adrenals with APA-KCNJ5^MUT comprised 2 clearly distinct tumor clusters, termed clusters APA-like 1 and APA-like 2. APA-like 1 displayed CYP11B2 gene expression whereas APA-like 2 did not (Figure S10). Both APA-like clusters were also located in the adjacent cortex to APA-KCNJ5^MUT (Figure 6A, C and E). Adrenals with APA-KCNJ5^WT exhibited a single tumor cluster- termed APA-like 3 - which expressed the CYP11B2 gene (Figure S10) and was in the tumor region and the adjacent cortex (Figure 6B, D and F). Notably, although the signature of APA-like subpopulations 1 and 3 comprised the CYP11B2 gene transcript, CYP11B2 gene expression was present only within tumor regions and not in the corresponding subpopulations in the adjacent adrenal cortex. Comparisons of tumor samples versus the adjacent cortex of patients with a different APA genotype, for example, tumor specimens from APA-KCNJ5^MUT compared with the adjacent cortex of APA-KCNJ5^WT and vice versa, consistently identified APA-like 1, 2, and 3 subpopulations (Figure S12). However, the APA-like 1 subpopulation in the tumor regions of APA-KCNJ5^MUT was undetectable in the adrenal cortical tissue adjacent to APA-KCNJ5^WT (Figure S12E).

Figure 6. Spatial transcriptomics establishes APA-like subpopulations in the adjacent adrenal cortex.

A-B, t-distributed stochastic neighbor embedding representations of the integrated spatial transcriptomes of APAs and adjacent adrenal tissues stratified by APA-KCNJ5 mutation status and tissue type. Panels A and B show integrated spatial transcriptomic spots from APAs and adjacent adrenal tissues (left image of A and B) as well as separated into APAs (boxed image, top right of A and B) and adjacent adrenal tissues (boxed image, bottom right of A and B). Colors represent transcriptome subpopulations as indicated. A, APA-KCNJ5^MUT (n = 4) and their adjacent adrenal cortex (n = 4). B, APA-KCNJ5^WT (n = 3) and their adjacent adrenal cortex (n = 3). C-D, representative spatial transcriptomics sections showing the distribution of spots in APAs and adjacent adrenals from patient 7 with a KCNJ5 mutation (male) and patient 6 without a KCNJ5 mutation (female). E-F, stacked plots showing the proportion of each transcriptome cluster in APAs and adjacent adrenals. Left panels and right panels indicate adrenal tissues from patients with a KCNJ5 mutation (represented in light blue, Left,) or without a KCNJ5 mutation (represented in light yellow, Right). All APA-like signatures (APA-like 1, 2 and 3) are common to tumor areas and adjacent tissue and may represent precursor APA subpopulations.

DISCUSSION

The biological mechanisms driving the predominant APA-KCNJ5^MUT phenotype of larger tumor diameter, particular cell composition and heterogeneous CYP11B2 immunostaining, are unknown. ^{4,9,15,16,18,19} To address this knowledge gap, this study had three main goals: (i) to characterize the intra- and inter-tumor transcriptional heterogeneity of APAs; (ii) to delineate the transcriptional and metabolomic mechanisms through which KCNJ5 mutation status contributes to tumor expansion; (iii) to spatially distinguish pre-APA-like cells in the adrenal cortex adjacent to APAs.

The histological heterogeneity of CYP11B2 immunostaining in APA-KCNJ5^MUT has been previously reported.^15,16 In this context, tumor regions with varying levels of CYP11B2 expression in APA-KCNJ5^MUT share identical somatic KCNJ5 variants with similar variant allele frequencies,^15,45 indicating that transcriptional diversity between tumor regions with different CYP11B2 expression levels should occur in a spatial context. Here, we establish APA heterogeneity at the transcriptional level through the unbiased analysis of the transcriptome landscape. Our findings suggest that heterogenous CYP11B2 expression may be caused by differential regulation of specific signaling pathways because biological pathways related to the stimulation of aldosterone production, such as VEGF,⁴⁶ p53,⁴⁷ and androgen⁴⁸ pathways, are dominant in tumor regions of APA-KCNJ5^MUT with higher CYP11B2 expression.

Defining the functional significance of clear versus compact eosinophilic cells in APAs using spatial transcriptomics is of particular interest because it has been challenging to define the transcriptional programs of these different cell subpopulations, in part because APAs comprise a mixture of the 2 cell types.⁴ Our bioinformatics analyses identified 2 distinct APA transcriptional signatures associated with CYP11B2 expressing regions, referred to here as CYP11B2-type 1 and CYP11B2-type 2, across genotype groups. The distribution of these signatures matched that of the previously described APA morphology (small compact versus clear cells) with diverse relative proportions of these cell types according to KCNJ5 mutation status and tumor diameter.^15,17,49 In the spatial context of our study, the CYP11B2-type 1 transcriptome signature was enriched in compact eosinophilic cells and was characterized by increased cell stress-associated signaling activity, suggesting that microenvironmental stress conditions may restrict their capacity for cell proliferation. Conversely, the CYP11B2-type 2 transcriptome signature which predominated in APA-KCNJ5^MUT- which have a higher proportion of lipid-rich clear cells- exhibited enhanced mTORC1 and hypoxia-associated signaling activity, consistent with the promotion of metabolic adaptations for proliferation under hypoxic conditions.^50,51

We resolve enriched cell stress transcriptional signals and accumulated PUFAs and pro-ferroptotic arachidonoyl-PE species in APA-KCNJ5^WT relative to their adjacent cortex and to APA-KCNJ5^MUT. Prior studies have demonstrated that an abundance of PUFAs and pro-ferroptotic arachidonoyl-PE species can activate cell death pathways such as ferroptosis in various diseases.^40,41,52,53 In contrast, enriched metabolites of the oxidative pentose phosphate pathway and of glutathione metabolism in APA-KCNJ5^MUT highlight active antioxidant defense mechanisms to eliminate stress overload signals and maintain tissue homeostasis.^54–56 Therefore, taken together, these data can help explain the restricted growth of APA-KCNJ5^WT tumors and the sustained growth of APA-KCNJ5^MUT and help account for the genotype-related differences in APA tumor size.

Our data report previously unappreciated APA-like molecular regions in the adrenal cortex adjacent to APAs. We identified 2 APA-like transcriptome subpopulations common to APA-KCNJ5^MUT and their paired adjacent cortex (APA-like 1 and APA-like 2), and a single APA-like subpopulation in the tumor and adjacent cortex of APA-KCNJ5^WT (APA-like 3). Spatial transcriptomics spots of these APA-like subpopulations in APA and their adjacent cortex were clustered according to their similar gene expression profiles. CYP11B2 transcripts were absent from the APA-like 2 subpopulation in APA-KCNJ5^MUT, which is consistent with the heterogeneous CYP11B2 immunostaining observed in these tumors.¹⁵ In contrast, CYP11B2 transcripts were part of the APA-like 1 and APA-like 3 signatures; however, the corresponding subpopulations expressed the CYP11B2 gene only in tumor regions and not in the adjacent cortex. Thus, APA-like 1 and APA-like 3 are distinct from aldosterone-producing micronodules¹¹ and might represent distinct precursor APA-like states. In addition, the APA-like-1 subpopulation identified in APA-KCNJ5^MUT and the adjacent cortex was apparently absent from the adjacent cortex to APA-KCNJ5^WT thus indicating that this signature could be a unique feature of adrenals harboring a APA-KCNJ5^MUT. However, because tumors may influence their surrounding microenvironment, we cannot rule-out the possibility that the APA-like 1 signature in the adjacent cortex to APA-KCNJ5^MUT resulted from modulatory effects on the tumor microenvironment by the APA itself. Further investigation of larger cohorts is needed to confirm our findings and assess the potential influence of gender in the context of KCNJ5 mutations.

In conclusion, we demonstrate intra-tumoral transcriptional heterogeneity of APA-KCNJ5^MUT and the transcriptome signatures of the component compact eosinophilic cells and lipid-rich clear cells. We propose that APA tumor cells require metabolic adaptations to oxidative stress signals to survive and proliferate and we provide evidence to support the existence of novel molecular zones in the adrenal cortex which might represent tumor precursor states.

PERSPECTIVES

Conventional techniques that apply bulk or single cell RNA sequencing cannot effectively dissect the tissue organization of gene expression levels. In contrast, spatially resolved transcriptomics profiling (spatial RNA sequencing) combined with state-of-the-art bioinformatics platforms can quantifiably assess gene expression at the cellular level within the context of tissue morphology. We applied this approach to surgically removed adrenal glands from patients operated for an APA. We established transcriptomics signatures which define the distinct cell compositions of APAs and the molecular and cellular diversity of these tumors according to KCNJ5 mutation status. Further, the integration of spatial transcriptomics and metabolomics datasets demonstrated that the cellular response to oxidative stress might direct genotype-related APA size differences. In the adrenocortical tissue adjacent to APAs, APA-like transcriptomic signatures were uncovered without aldosterone synthase transcription thereby suggesting the existence of novel tumor precursor cells prior to the acquisition of autonomous aldosterone production. Overall, we establish how spatial omics technologies can provide a blueprint for understanding the spatial architecture and genotype-specific tumorigenesis of APAs.

NOVELTY AND RELEVANCE.

What Is New?

We integrated spatial transcriptomics and spatial metabolomics profiles with CYP11B2 immunohistochemistry of surgically removed adrenals from patients operated for an APA.

What Is Relevant?

• Despite high inter-tumoral transcriptional heterogeneity, 2 specific transcriptional signatures common to all APAs were identified that characterize compact eosinophilic cells or lipid-rich clear cells.

• Antioxidant metabolites are enriched in APAs with a KCNJ5 mutation and prooxidant metabolites are enriched in APAs without a KCNJ5 mutation.

• Novel APA-like molecular subpopulations without CYP11B2 expression are identified in the adrenal cortex that suggest the existence of tumor precursor states.

Clinical/Pathophysiological Implications

Our findings advance the understanding of the transcriptional context of inter- and intra-tumoral APA heterogeneity and provide novel insight into the genotype-dependent tumor expansion capabilities of APAs.

Nonstandard Abbreviations and Acronyms

APA

aldosterone-producing adenoma

PA

primary aldosteronism

zG

zona glomerulosa

zF

zona fasciculate

zR

zona reticularis

MALDI-MSI

matrix-assisted laser desorption/ionization mass spectrometry imaging

KCNJ5

gene encoding GIRK4 (G-protein-gated inwardly rectifying potassium channel 4)

NNMF

non-negative matrix factorization

PUFA

polyunsaturated fatty acid

PE

phosphatidylethanolamine