Fresh primary ACC tissues were obtained from patients upon resection or biopsy at Ruijin Hospital between September 2020 to July 2023. The ACC diagnosis was confirmed by experienced pathologists, and steroidogenic factor 1 (SF1) immunostaining was performed to confirm its adrenal cortex origin. Clinicopathological information including age, sex, ENSAT stage, Ki67 index and hormonal secretion status, systemic therapies received prior to surgery or biopsy was recorded and analyzed. Hormonal secretion status was evaluated using biochemical testing of serum steroid hormone levels (e.g., cortisol, aldosterone and androgens) and 1mg dexamethasone suppression test. Informed consent was obtained from all patients, and this study was approved by local ethics committee of Ruijin Hospital (Approval number: KY320). Upon surgical or biopsy removal, pieces of tumors were fixed in formalin and paraffin-embedded for pathological diagnosis. For primary cell cultures, tumor tissues were placed in Tissue Storage Solution (Miltenyi Biotec, Cat No.130-100-008). Additional tissues were immediately snap-frozen in liquid nitrogen for later use. The overview of the tissue process pipeline was summarized in Figure 1A .

Immediately after surgery or biopsy, tumor tissues were collected in the Tissue Storage Solution (Miltenyi Biotec, Cat No.130-100-008) and transported to the laboratory on ice and isolated within 24 hours. Tumor tissues were rinsed with Hanks’ Balanced Salt Solution (HBSS; Gibco, Cat No.14175095), minced and digested with 2.0 mg/mL collagenase II (Gibco, Cat No. 17101015), 0.02 mg/mL DNase (Roche, Cat No. 11284932001) at 37°C on a shaker for up to 2 hours. Then this suspension was filtered through a 70-µm cell strainer (Falcon, Cat No.352350). After depletion of red blood cells using Red Blood Cell Lysis Buffer (Invitrogen, Cat No.00-4333-57), trypan blue staining (Gibco, Cat No.15250061) was performed for cell counting and viability assessment. After cell preparation, one portion was plated directly into 96-well plates (Corning, Cat No.3799) for mitotane sensitivity testing, whilst a small number of cells was plated in chamber slides (Millipore, Cat No. PEZGS0816) or CellCarrier Ultra plates (PerkinElmer, Cat No.6055300) for immunofluorescence staining of adrenal cortex marker, SF1 (Proteintech, Cat No.18658-1-AP). When cell amount was abundant, cells would be plated in 384-well plate for HTS and cryopreserved in liquid nitrogen for later use. Cells then were cultured in DMEM/F-12 (Gibco, Cat No. 11320033) medium supplemented with 10% fetal bovine serum (FBS; Gibco, Cat No. 10099141), 1% Penicillin-Streptomycin (10,000 U/mL; Gibco, Cat No. 15140-122) and 1% L-glutamine (200 mM; Gibco, Cat No. 25030-081).

Mitotane (MedChemExpress, Cat No. HY-13690) was dissolved in dimethyl sulfoxide (DMSO, Sigma, Cat No. D2650) to a concentration of 100mM as stock solution, aliquoted and stored at -80°C. For in vitro experiments, the final concentration of DMSO was ≤0.1%. Primary cells were plated in a 96-well plate at a density of 1.0x10^4 cells/well in triplicates and treated with mitotane (1.0μM-100μM) for 72 hours and cell viability was assessed by the Cell Counting Kit-8 (CCK-8) assay kit (Dojindo, Cat No. CK04). Mitotane-sensitive ACC cell line H295R (ATCC^® CRL­2128™) was used as the positive control. Dose-response curves, inhibition rate, half maximal inhibitory concentration (IC50) values and area under dose response curve (AUC) were calculated in Prism8.3 software (GraphPad). PDCs were arbitrarily classified as non-responders when the inhibitory effect on cell viability was less than 33% at the concentration of mitotane corresponding to the therapeutic circulating plasma concentration (14 mg/L, 50 µM) according to the previous study (10).

Briefly, cells were fixed with 4% paraformaldehyde for 15 min at room temperature, and then washed twice with PBS buffer (Sangon Biotech, Cat No. B548117), followed by permeabilization with 0.1% Triton X-100 (Sigma-Aldrich, Cat No. 9036-19-5) for 15min. Next, cells were washed twice with PBS and blocked using antibody diluent (DAKO, Cat No. s3022) for 1h at room temperature. Later, cells were incubated with primary antibody against SF1 (1:100, Proteintech, Cat No.18658-1-AP) at 4°C overnight, followed by YSFluor 594-conjugated secondary antibodies (1:500, Yeasen Biotechnology, Cat No. 34212ES60). Nuclei were stained with 4,6-diamidino-2 phenylindole (DAPI) and wells were mounted using DAPI Fluoromount-G (SouthernBiotech, Cat No. 0100-20).

Cells were plated in 384-well plates (PerkinElmer, Cat No.6007680) at a density of 2000 cells per well in 50μl total volume. HTS was performed in an automated Cell::explorer HTS pro Platform (PerkinElmer). 24 hours after seeding, cells were treated with test compounds using a robot plate::handler equipped with a pintool dispensing device (PerkinElmer) for 6 days. HTS was conducted in single with four concentrations for each compound. DMSO was used as the vehicle control. Cell viability was determined using CellTiter-Glo reagent (Promega, Cat No. G7572) and luminescence was measured on an EnVision multimode plate reader (PerkinElmer). Dose-response data were analyzed. IC50 and AUC were calculated in Prism8.3 software (GraphPad).

Genomic DNA and total RNA were extracted from snap-frozen tumor tissues or patient-derived primary cell pellets using the AllPrep DNA/RNA Micro Kit (Qiagen, Cat No. 80284) according to the manufacturer’s instructions. DNA extraction from peripheral blood leukocytes was carried out using the QIAamp DNA Mini Kit (Qiagen, Cat No. 51304). DNA and RNA concentrations were evaluated on Qubit Fluorometer (Thermo Fisher Scientific).

WES was performed on the tumor DNA and matched blood DNA. Briefly, genomic DNA of tumor and paired peripheral blood samples from 9 patients was randomly sheared through ultra-sonication to generate paired-end libraries with an average insert size of ~300 bp. Exome regions were captured by the xGen Exome Hyb Panel v2 kit (Integrated DNA Technologies, Cat No. 10005153), and sequencing was performed on Illumina Novaseq 6000 platform (Illumina, San Diego, CA, USA) with 150 bp paired end strategy.

The paired-end reads from WES were mapped to human reference genome (hg19) by BWA aligner (v0.7.17) (23). Mapping results were then sorted and marked for duplications via Picard (v2.23.0) (24). Single nucleotide variants (SNVs) and small insertions and deletions (INDELs) were obtained by taking the union of three callers GATK4 Mutect2 (25), VarDict (26), and MuTect (27). All mutations were annotated by snpEff (v4.2), and ANNOVAR (v2019Dec03). All functional mutations, including missense, nonsense, splicing, nonstop SNVs, and INDELs, were obtained. Homemade pipelines were used to filter SNVs and INDELs: 1) mutations were called by more than one software; 2) variant allele frequencies (VAFs) were ≥ 10% and ≥ 4 individual mutant reads.

Sequencing coverage and copy number in the aligned sequencing reads from WES were analyzed using CNVkit (v0.9.7) (28). The sequencing coverage of WES in germline samples was assessed and used to create pooled reference data that included the technical variability at each exon region. The read depths of tumor samples were individually compared with the reference after normalization (corrected for GC content, target footprint size and spacing, and repetitive sequences). The copy number segments were inferred by the circular binary segmentation algorithm (29).

We utilized the KAPA RNA Hyper Prep Kit (Kapa Biosystems, Cat No. KK8544) for library preparation. Sequencing was performed in the Illumina Nova S4 platform. The Illumina bcl2fastq Conversion Software was used to convert base call (BCL) files into FASTQ files. The sequences were aligned to hg38 reference genome using HISAT2 (30) and gene expression levels were quantified using RSEM (31). Count correction was performed using the removeBatchEffect function from the limma R package (32) and then the batch-corrected expression matrix that used to heatmap analysis were constructed based on log-normalized transcripts per million (TPM) of each gene. P value < 0.05 and |logFoldChange| ≥ 1.5 was set as the threshold for significantly differential expression. Gene Set Enrichment Analysis (GSEA) (33) was performed based on Gene Ontology (GO) database (34, 35) and KEGG.

Data were presented as mean ± SEM, or as medians and interquartile ranges (IQR), whilst categorical variables were presented as percentages and absolute numbers. Statistical analysis was performed using IBM SPSS Statistics Version 26.0 and GraphPad Prism 8. All P values were two-sided, with P < 0.05 considered statistically significant.

PDCs were successfully obtained from 17 ACCs upon surgery or biopsy, including six primary tumors, two local recurrent and nine metastatic tumors (lung, liver, etc.). SF1 immunofluorescence staining was performed to confirm the adrenal cortex origin of tumor cells ( Figure 1B ). Patient characteristics are listed in Supplementary Table 1 . Of note, six patients received mitotane prior to surgical or biopsy intervention for a period of 2 months to 18 months, but underwent disease progression or lacked satisfactory response. The principal aim was to evaluate response of PDCs to mitotane and identify biomarkers to predict sensitivity ( Figure 1A ). Since the yield of the dissociation procedure varied because of differences in the size of the tumor tissue available, only when cell amount permitted, HTS would be performed to seek for other potential agents.

First of all, we performed in vitro sensitivity testing in PDCs for a 3-day mitotane exposure, allowing to exclude the impact of patient tolerance or pharmacokinetics. Cell viability inhibition at 50 µM mitotane is used to group ACCs into responders (>33% inhibition) and non-responders (≤33% inhibition). The median cell viability inhibition rate at 50 µM mitotane was 30.4% (IQR: -7.1%-47.9%). Eight patients (47%) were classified as responders with inhibition rate reached 48.4% (IQR: 39.3%-59.3%) whereas nine (53%) non-responders were scarcely inhibited by 50µM mitotane, with median inhibition rate of -1.2% (IQR: -26.4%-22.1%) ( Table 1 ; Figure 2A ). Dose-response curves showed the different potency of mitotane in the two groups as non-responders had higher IC50 values. Median IC50 for responders and non-responders were 53.4 µM (47.8-54.4µM) and 74.7 µM (70.9-98.8µM), respectively (P<0.0001). AUC were greatly increased compared to responders, and estimated AUCs were158.0 (142.1-164.3) and 213.5 (194.5-273.1) in responders and non-responders, respectively (P<0.0001) ( Table 1 ; Figures 2B, C ). Clinical response data were obtained from eight patients: all three non-responders showed clinical progressive disease; three responders showed clinical stable disease, while two responders progressed ( Table 2 ). The consistence rate between in vitro test and clinical response is 75% (6/8).

Comparison of clinicopathological characteristics between responders and non-responders demonstrated no significant differences regarding as age (57.8 ± 12.4 vs 45.9 ± 24.3, P=0.233), gender (P=0.149), ENSAT staging (P=0.704) and Ki67 index (P=0.766). Noteworthily, functional tumors with steroid hormone secretion showed a tendency of better response in vitro than non-functional ones (66.7% vs 25.0%, P=0.086) ( Table 1 ). A negative correlation was found between tumor functionality and AUC (Spearman correlation coefficient= -0.481, P=0.051) with marginal significance, in line with above findings, indicating a tendency that tumors with active hormonal function might respond better to mitotane treatment.

To identify molecular factors contributing to mitotane response, we then conducted genomic (WES) and transcriptomic sequencing (RNAseq) on tissue samples or primary cell pellets when there were no additional tissues available. In order to reveal intrinsic genetic features underlying mitotane sensitivity rather than acquired molecular features induced by mitotane treatment, a total of nine samples free from mitotane exposure were sequenced, from five responders and four non-responders classified by in vitro sensitivity testing.

Alterations in the established driver genes including SNVs and CNVs were observed in P53/RB cell-cycle pathway (8/9, 88.9%) and Wnt/β-Catenin signaling pathway (9/9, 100%) ( Figure 3A ). 4/5 responders and 3/4 non-responders harbored genetic alterations in both pathways concurrently. More specifically, somatic mutations in TP53 were found in 3 patients (2/5 responders and 1/4 non-responders) and loss of TP53 was found in 1 responder (1/5). RB1 mutation was identified in 1/5 responders and 2/4 non-responders. CNV gain or amplification in CDK4, CCNE1 and MDM2 were identified in 7 patients. It was well acknowledged that CTNNB1 mutations and ZNRF3 alterations were mutually exclusive (36). Surprisingly, we found their exclusive presence in responders and non-responders. Responder group merely harboring CTNNB1 somatic mutations (3/5) while non-responder group presenting only ZNRF3 alterations (3/4). Moreover, APC alterations were observed in 4 responders and 4 non-responders. However, whether CTNNB1 and ZNRF3 alterations render differential intrinsic sensitivity to mitotane requires further investigations.

RNAseq was performed to investigate gene expression signatures. A total of 1612 genes were differentially expressed (|logFoldChange| ≥ 1.5, P<0.05) between responder and non-responder tumors ( Figure 3B ). Evidence has accumulated that mitotane dysregulated lipid metabolism and raised the potential correlation between mitotane responsiveness and capacity of handling lipids (37, 38). From our transcriptome data, to be noted, expression of genes involved in steroidogenesis (CYP11B1) and lipid metabolism (CYP27A1, ABCA1, PLIN2, PLIN4, NR1H3, etc) were significantly upregulated in mitotane-sensitive tumors ( Figure 3C ), implying elevated capacity for handling intracellular lipids. Consistently, functional enrichment analysis using GSEA showed pathways associated with lipid metabolism are significantly upregulated in responders and non-responders including lipid metabolic process, lipid catabolic process, lipid oxidation, cholesterol metabolic process and steroid metabolic process, etc, possibly underlying tumor functionality ( Figure 4A ). On the other hand, Wnt signaling pathway and cell cycle process were significantly downregulated in the responder group ( Figure 4B ).

To further investigate marker genes correlated to in vitro mitotane responsiveness, Spearman correlation analysis was performed between gene expression levels and response data of AUC. A list of genes previously reported as key regulators of lipid metabolism (uptake, biosynthesis, storage, and lipolysis, efflux, etc), steroidogenesis as well as genes priorly proposed to be potentially predictive for mitotane response were analyzed. We failed to find correlation between RRM1, SOAT1, CYP2W1 mRNA expression level and mitotane responsiveness ( Supplementary Figure 1 ). Of note, oxysterol synthetic enzyme, CYP27A1 and cholesterol efflux pump, ABCA1 were negatively correlated to AUC ( Figure 4C ), denoting the higher expression of CYP27A1 and ABCA1, the lower of the AUC value and the better of in vitro responsiveness to mitotane. As a pivotal mechanism for preventing intracellular free cholesterol accumulation, it was tempting to speculate that higher CYP27A1 and ABCA1 implied higher intracellular free cholesterol at baseline, which required enhanced conversion and efflux ability, thus more susceptible to mitotane.

These findings indicated that the dependence on the higher capacity for lipid metabolism to maintain intracellular lipid balance conferred ACC more vulnerable to mitotane.

In order to uncover potential therapy for ACC especially those mitotane non-responders, we designed and set up a compound library containing 40 compounds in four concentrations. Primarily, drugs or compounds were chosen based on the following criteria: 1) The drug was FDA-approved or in clinical trials; 2) The compound had been reported effective in ACC models or proposed as potential targeted anti-cancer treatments for ACC (39–41). Drugs and highest concentrations used in HTS as well as references are listed in Supplementary Table 2 . Aiming to establish a differential cytotoxicity assay, a 6-day treatment with compounds were performed in PDCs. Eight patient-derived ACC cells (four responders and four non-responder) were tested in a proliferative assay against our in-house library in 384-well plates. Surprisingly, both mitotane responsive and non-responsive ACC cells were extremely vulnerable to disulfiram treatment. Antihelminthic agent, niclosamide, and proteasome inhibitor, bortezomib, which were previously reported effective in ACC cell lines (42, 43), were identified efficacious in 6/8 and 5/8 PDCs, with estimated IC50 ranging from 0.22μM to 0.77μM and 10nM to 50nM, respectively. Furthermore, doxorubicin and cisplatin were effective in 3/8 and 2/8 ACCs, respectively. Additionally, PI-103, a PI3K/mTOR inhibitor, was active in 4/8 ACCs and primary culture derived from Patient 13 demonstrated sensitivity to multi-targeted tyrosine kinase inhibitor, sunitinib and anlotinib ( Figures 5A, B ).