The patients were registered as part of ongoing registries and biobanks (European Network for the Study of Adrenal Tumor [ENS@T, www.ensat.org] and Excellence Network for Neuroendocrine Tumors [NeoExNet]). The study was approved by the Ethics Committees of the University of Munich and Würzburg. Written informed consent was obtained from all enrolled patients, and the experiments were performed according to relevant guidelines and protocols.

For NGS a total of 18 cryo-preserved adrenal samples were used. The discovery cohort consisted of PBMAH (n=10) and adjacent normal adrenal cortex from patients with pheochromocytoma (controls, n=8).

Additional QPCR validation was performed in RNA extracted from cryo-preserved tissue from the following samples: cortisol producing adenoma (CPA, n=9), and adrenal samples from patients undergoing bilateral adrenalectomy for persistent Cushing’s disease (BADX-CD, n=8). Normal adrenals from patients who underwent kidney surgery (normal adrenals, n=10), and adrenal samples from patients with aldosterone producing adenoma (APA, n=10) were used as controls for validation. Furthermore, the results were validated with data from 3’RNA sequencing of RNA extracted from FFPE tissues on an independent PBMAH cohort (PBMAH-FFPE, n=11).

The clinical characteristics of the groups are given in Table 1 . In total, four PBMAH patients carried ARMC5 mutations (2/10 PBMAH samples used in the discovery and 2/11 used in the validation cohort), and one patient from the validation cohort carried a KDM1A mutation.

The adrenal tissues were stored at -80°C. Total RNA isolation was carried out from all adrenal cortex samples using the RNeasy Tissue Kit (Qiagen, Germany). The isolated RNA was kept frozen at −80°C until further use. RNA yield and purity were measured using NanoDrop (Thermofisher Scientific, Germany). Quality control of the isolated RNA, RNA library preparation, and sequencing was performed by QIAGEN (Zymo Research, Irvine, US). In case of RNA from FFPE material, the tumor area was marked and 10 μm sections from five serial slides were collected under a stero microscope (Motic). Total RNA was extracted from the collected tumor using the AllPrep DNA/RNA FFPE kit (Qiagen) according to the manufacturer’s instructions. The RNA quantity and quality were determined using NanoDrop 2000 spectrophotometer (Thermo Fisher).

RNA-seq from frozen adrenal samples was performed at Qiagen, Hilden, Germany and sequencing from FFPE samples was done at the SysMed Core Facility, Würzburg. Briefly, RNA integrity and the absence of contaminating DNA were confirmed by Bioanalyzer RNA Nano (Agilent Technologies) and by Qubit DNA High sensitivity kits, respectively. QIAseq Stranded RNA Library Kits was used for library preparation. Sequencing was performed on Illumina NextSeq (single end read, 75 bp). Adapter and quality trimming were performed by the “Trim Reads” tool from CLC Genomics Workbench. Further, reads were trimmed based on quality scores. The QC reports were generated by the “QC for Sequencing Reads” tool from CLC Genomics Workbench. Read mapping and gene quantification were performed by the “RNA-seq Analysis” tool from CLC Genomics Workbench (15). In case of FFPE derived RNA, sequencing libraries were prepared using the QuantSeq 3’ mRNA-Seq protocol (Lexogen, Vienna, Austria), from 100 to 500 ng RNA. Single read sequencing (1 ×75bp) was performed on a NextSeq 550 platform (Illumina, San Diego, CA, USA). For RNA extracted from FFPE samples, sequencing libraries were constructed using the QuantSeq 3’ mRNA-Seq protocol from Lexogen (Vienna, Austria). The libraries were prepared using 100 to 500 ng of RNA. Subsequently, single-read sequencing with a read length of 75 bases (1 × 75bp) was carried out on a NextSeq 550 platform from Illumina (San Diego, CA, USA) (16).

Differentially expressed RNAs identified through NGS were validated by QPCR. RNA concentration was assessed using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific), and reverse transcription was performed on 50 ng of RNA using Superscript VILO reverse transcriptase (Thermo Fisher Scientific) according to the manufacturer’s instructions. The selection of a suitable housekeeping gene was carried out using the BestKeeper tool to determine the most stably expressed housekeeping gene (p-value <0.001) (15). In frozen adrenal samples of the discovery cohort, controls (n=8) and PBMAH (n=10), housekeeping genes ACTB, GAPDH, and PPIA were evaluated. For adrenocortical cell lines (NCI-H295R; n=10, CU-ACC2; n=10, and MUC1; n=10), the housekeeping genes ACTB and PPIA were assessed. PPIA was identified as the most stable reference gene for human adrenal samples, while ACTB was chosen as the reference gene for adrenocortical cell lines ( Figure S1 ). QPCR was conducted using TaqMan Fast Universal PCR Master Mix (Thermo Fisher Scientific) on a Quantstudio 7 Flex Real-Time PCR System (Thermo Fisher Scientific), following the manufacturer’s TaqMan mRNA assay protocol. TaqMan probes used are listed in Table S1 , and negative control reactions lacked cDNA templates. Each QPCR reaction used 5 ng of cDNA and was analyzed in technical triplicates. The PCR was carried out in a 20 μL mixture, with 10 μL of Master Mix, 1 μL of TaqMan probes, 5 μL of cDNA, and 4 μL of nuclease-free water. The qPCR was performed following the recommended thermocycling conditions on the Quantstudio 7 Flex Real-Time PCR System (Thermo Fisher Scientific) under the Fast mode, with an initial activation at 95°C for 20 sec and 40 cycles of 95°C for 1 sec and 60°C for 20 sec. Gene expression levels were quantified using the relative quantification method (17), normalized to the reference gene, to enable easier comparisons.

ACTH stimulation tests were performed in 16 female mice (C57BL/6). Briefly, Jfemale mice representing each time point of injection). Briefly, 13-week-old mice were intraperitoneally injected with 1mg/kg of ACTH (Sigma Aldrich, Germany) and adrenals collected after 10, 30, and 60 min of injections (4 mice per time point). In addition, control adrenals were collected from mice at baseline conditions (0 min). Details of the experiment are published in (18). Gapdh was used as housekeeping gene in the QPCR. All mice were maintained in accordance with facility guidelines on animal welfare and approved by Landesdirektion Sachsen, Germany.

The identified PPARG pathway was further characterized in vitro using three adrenocortical cell-lines (NCI-H295R, CU-ACC2 and MUC1), derived from patients with malignant adrenocortical carcinoma (ACC). The human NCI-H295R cell line, derived from a primitive ACC in a female patient, was obtained from the American Type Culture Collection (ATCC) and cultured as indicated by ATCC. MUC-1 cell line, established from a neck metastasis of an EDP-M-treated (etoposide, doxorubicin and cisplatin plus oral mitotane) male patient, was kindly given by Dr. Hantel and cultured as suggested (19). CU-ACC2 cell lines were obtained from Dr. Katja Kiseljak-Vassiliades (20). A detailed description of these cell lines can be found in Sigala et al. (21). The activation of PPARG in the cell lines was assayed for cell viability, gene expression changes and effect on steroidogenesis. Rosiglitazone, which is a known drug in diabetes therapy, was used as a PPARG activator.

Cells (50.000 cells/well) were seeded in 96-wells-plates and treated with increasing concentrations of ACTH (2.5–20 nM; Alfasigma), solubilized in water, and rosiglitazone (5–40 µM; Cayman Chemical). Rosiglitazone was solubilized and serially diluted using DMSO. The drug solutions were prepared 200 times more concentrated to dilute the DMSO at a ratio of 1:200 in the well. Both compounds were tested alone and in combination. Cell viability was evaluated by WST-1 assay according to the manufacturer protocol (Roche). All the subsequent sets of experiments were conducted treating the cells for 48h in serum-free medium. For treatments on CU-ACC2 cell line also the hydrocortisone has been removed.

Cells (1.5x10⁶ cells/well) were seeded in 6-wells-plates and treated with ACTH (2.5 nM), and 5, 10, 20 µM of rosiglitazone in a final volume of 3 ml. These doses were chosen based on the cell viability results. After 48h the media were collected and stored at -20°C until analysis performed using liquid chromatography tandem mass spectrometry (LC-MS/MS) as described by Schweitzer et al. (22). The cells were scraped from the bottom of the wells to isolate the mRNA using the Maxwell RSC Simply RNA Kit (Promega). RNA concentration was determined using a NanoDrop 2000 spectrophotometer (Thermo Fisher) and 1000 ng RNA were reverse transcribed with the High-Capacity cDNA Reverse Transcription Kit, Applied Biosystems). QPCR was performed using TaqMan gene expression probes (Thermo Fisher Scientific) for PPARG (Hs01115513) and for ACTH receptor, also known as the melanocortin receptor 2 (MC2R), (Hs00300820). Endogenously expressed ACTB (Hs99999903) was used as housekeeping gene for normalization. For each QPCR reaction, 5 ng cDNA were used, and each sample was analyzed in technical triplicates. All transcripts were amplified using TaqMan Gene Expression Master Mix (Thermo Fisher) using the CFX96 real-time thermocycler (Bio-rad) and the Bio-rad CFX Manager 2.0 software. Cycling conditions were 95°C for 3 min, followed by 39 cycles of 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s. Fold change was calculated using the delta cycle threshold (dCt) method, normalized to housekeeping gene ACTB.

R version 4.2.0 was used for statistical analyses of NGS data. To identify RNAs differentially expressed, generalized linear model (GLM, a flexible generalization of ordinary linear regression that allows for variables that have distribution patterns other than a normal distribution) in the software package edgeR (Empirical Analysis of DGE in R) was employed to calculate p-values (23, 24). P-values were adjusted with the Benjamini–Hochberg false discovery rate (FDR) procedure (25). GraphPad Prism Version 8 was used for statistical analysis of QPCR. To identify RNAs differentially expressed in QPCR, the dCt method (target gene’s Ct minus housekeeping RNA’s Ct) was used in Microsoft Excel 2016 (Microsoft, Redmond, WA, USA). For intergroup comparison of QPCR data, ANOVA test with Benjamini–Hochberg false discovery rate (FDR) procedure was used (26). P<0.05 and FDR<0.05 were considered significant. Pathway mapping for the significant genes was performed using the ShinyGO program (27).

Transcriptome analyses of PBMAH adrenal samples (n=10) identified a total of 1104 genes to be significantly differentially expressed in comparison to the control group (n=8) in the discovery cohort (l2fc>|2|, FDR<0.05). Almost 70% of the genes were downregulated (769 genes), indicating an overall transcriptionally repressed state in PBMAH samples ( Figure 1 , Table S2 ). Hierarchical clustering based on the top upregulated significant genes was performed and revealed a good discrimination between (27)en PBMAH samples and controls. Interestingly, PBMAH samples with ARMC5 mutations clustered together while the ARMC5^wt PBMAH samples clustered separately and were further divided in two additional clusters ( Figure S2 ).

To better understand the transcriptional profile of PBMAH, all significant genes (n=1104 genes) were used for pathway mapping to identify which genes are enriched in different pathways ( Figure 2A ). This pathway mapping analysis is grounded in the concept of fold enrichment ( Figure 2A ), wherein fold enrichment is computed as the percentage of genes in the input list relative to the corresponding background percentage. This calculation serves as a metric of the statistical significance of pathway dysregulation. Overall, the top pathway hits from the PBMAH transcriptome generated two major pathway clusters ( Figure 2B ) (1): “peroxisome proliferator-activated receptors” (PPARs) signaling, which is widely acknowledged for its anti-glycemic and anti-lipolytic effects (28), and (2) pathways related to neuronal function and diseases including “nicotine addiction” and “neuroactive ligand-receptor interaction”. To validate the pathway enrichment based on the transcriptome data, the significantly altered PPARG and its top downregulated target genes (ADIPOQ, APOA1, FABP4, PCK1 and PLIN1) were chosen. A significant (p<0.01) downregulation of PPARG and its target genes ADIPOQ, FABP4 and PLIN1 was found in the PBMAH samples in comparison to controls ( Figure 3 ). For the neuronal pathway cluster, five genes (DRD2, GRIA2, GRIA4, GRIN2A and SCTR) shared most among the pathways were chosen for further analyses but failed to show significant dysregulation ( Figure S3 ).

The observed downregulation of PPARG and its target genes by QPCR and NGS were further validated in two ways. In the first step, DESeq2 based gene expression analyses from FFPE RNA from an independent PBMAH cohort (n=11) ( Table 2 ) showed that PPARG and its target genes were consistently downregulated in the validation cohort of PBMAH. In the second step, the expression of PPARG and its target genes was verified in adrenal samples from other types of CS, normal adrenals and APAs ( Figure 4 ). Significant downregulation of PPARG was observed in all the CS samples in comparison to both controls, APAs and normal adrenals ( Figure 4A ). Interestingly, significant downregulation of PPARG and its target genes, ADIPOQ ( Figure 4B ) and FABP4 ( Figure 4C ) was also observed in APA in comparison to normal adrenals, but not as pronounced as in CS subtypes. In contrast, the significant downregulation of PLIN1 was specific to CS groups ( Figure 4D ). Taken together, the candidate PPARG pathway downregulation observed in the adrenal samples of PBMAH by NGS could be validated in all adrenals with cortisol excess.

To analyze whether Pparg expression is influenced by ACTH, an ACTH stimulation study was done in mice and Pparg expression was assayed at different timepoints. No changes in Pparg expression were found in any of the timepoints ( Figure 5 ).

Next, the effect of ACTH stimulation on MC2R mRNA expression was tested in adrenocortical cell lines. Only NCI-H295R cells were found to show a decrease in MC2R expression upon ACTH treatment ( Figure 6 ). The activation of PPARG in the cell lines was assayed for cell viability, gene expression changes, and effect on steroidogenesis. Rosiglitazone, an insulin sensitizer formerly used in diabetes therapy, was used as a PPARG activator.