MDPI

Article Molecular Cluster Mining of Adrenocortical Carcinoma via Multi-Omics Data Analysis Aids Precise Clinical Therapy

Yu Guan 1,2,3,4, Shaoyu Yue 1,2,3,4, Yiding Chen 1,2,3,4, Yuetian Pan 40D, Lingxuan An 4, Hexi Du 1,2,3,* and Chaozhao Liang 1,2,3,*

1 Department of Urology, The First Affifiliated Hospital of Anhui Medical University, 218th Jixi Road, Hefei 230022, China

2 Institute of Urology, Anhui Medical University, 81th Meishan Road, Hefei 230022, China

3 Anhui Province Key Laboratory of Genitourinary Diseases, Anhui Medical University (AHMU), 81th Meishan Road, Hefei 230022, China

4 Department of General, Visceral, and Transplant Surgery, Ludwig-Maximilians-University Munich, D-81377 Munich, Germany

* Correspondence: duhexi1989@163.com (H.D.); liang_chaozhao@ahmu.edu.cn (C.L.); Tel .: +86-18856040979 (H.D.); +86-13505604595 (C.L.)

+ These authors contributed equally to this work.

check for updates

Citation: Guan, Y .; Yue, S .; Chen, Y .; Pan, Y .; An, L .; Du, H .; Liang, C. Molecular Cluster Mining of Adrenocortical Carcinoma via Multi-Omics Data Analysis Aids Precise Clinical Therapy. Cells 2022, 11, 3784. https://doi.org/10.3390/ cells11233784

Academic Editors: Amancio Carnero, Sandra Muñoz-Galván and José M. García-Heredia

Received: 24 October 2022

Accepted: 24 November 2022 Published: 26 November 2022

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affil- iations.

CC

İ

BY

Copyright: @ 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Abstract: Adrenocortical carcinoma (ACC) is a malignancy of the endocrine system. We collected clin- ical and pathological features, genomic mutations, DNA methylation profiles, and mRNA, lncRNA, microRNA, and somatic mutations in ACC patients from the TCGA, GSE19750, GSE33371, and GSE49278 cohorts. Based on the MOVICS algorithm, the patients were divided into ACC1-3 subtypes by comprehensive multi-omics data analysis. We found that immune-related pathways were more activated, and drug metabolism pathways were enriched in ACC1 subtype patients. Furthermore, ACC1 patients were sensitive to PD-1 immunotherapy and had the lowest sensitivity to chemother- apeutic drugs. Patients with the ACC2 subtype had the worst survival prognosis and the highest tumor-mutation rate. Meanwhile, cell-cycle-related pathways, amino-acid-synthesis pathways, and immunosuppressive cells were enriched in ACC2 patients. Steroid and cholesterol biosynthetic pathways were enriched in patients with the ACC3 subtype. DNA-repair-related pathways were enriched in subtypes ACC2 and ACC3. The sensitivity of the ACC2 subtype to cisplatin, doxorubicin, gemcitabine, and etoposide was better than that of the other two subtypes. For 5-fluorouracil, there was no significant difference in sensitivity to paclitaxel between the three groups. A comprehensive analysis of multi-omics data will provide new clues for the prognosis and treatment of patients with ACC.

Keywords: adrenocortical carcinoma; multi-omics analysis; prognosis and treatment; cell signaling pathway; sensitivity to drugs

1. Introduction

Adrenocortical carcinoma (ACC) is an aggressive endocrine malignancy that originates in the adrenal cortex, accounting for approximately 5% of adrenal tumors, with an annual incidence of 0.7-2.0 cases per million people [1,2]. The onset age of ACC has bimodal characteristics, with a high incidence in the two age stages of 40-50 years old and 1-5 years old [2,3]. The disease stage is one of the most important prognostic factors. Currently, the staging system proposed by the European Adrenal Neoplasms Research Network (ENSAT) is a commonly used international standard [4]. Stages I and II are confined to the organs and can be cured by complete resection. Stages III and IV are considered aggressive and metastatic advanced tumors, and the five-year survival rate for stage IV patients is only 6-13% [2,5]. Some researchers believe that ACC may be related to the overproduction of steroid precursors, which have the characteristics of early metastasis

and recurrence [6]. Complete surgical removal of ACC is the only chance for a long- term cure [7]. Mitotane has cytotoxic effects on steroidogenic cells in the adrenal cortex; therefore, it is recommended as a clinical adjuvant therapy, but its therapeutic effect is still unsatisfactory [2,8].

Some researchers have found that insulin-like growth factor 2 (IGF2), ß-catenin (CTNNB1), and TP53 may be potential drivers of sporadic adrenocortical tumors. The IGF system has growth-promoting and differentiation functions in the adrenal glands. IGF2 overexpression is found in most ACCs and is often associated with poor outcomes [1,9]. It regulates cell proliferation and apoptosis mainly by binding to the insulin-like growth factor 1 receptor (IGF1R), especially in pediatric patients [10,11]. As a key component of the Wnt signaling pathway, ß-catenin plays an important role in the development of the adrenal cortex [12], and is a poor prognostic factor in ACC [13]. Somatic mutations are common in ACC, and some researchers have found that somatic mutations in CTNNB1 are independent predictors of poor disease-free survival and overall survival in ACC [14]. Pan-genomic studies have found that TP53 mutations, mainly exon mutations, are common in sporadic ACC cases [15].

Multiple high-throughput detection techniques have been used in multi-omics asso- ciation studies to elaborate on a single scientific subject. Numerous variables can affect cancer development. These include a wide range of information on various facts. Infor- mation may be combined at several levels and integrated, and staging prediction accuracy can be improved by multi-omics association studies [16]. Over the past 50 years, the genetic approach to cancer has taken over the profession. However, this genome-only perspective is limited and has the propensity to present cancer as a strongly heritable illness. According to new research, cancer is a multi-omics illness and is not as heritable or exclusively hereditary as previously believed. Cancer development and manifestation are influenced by the exposome, metabolome, and genome. Cancer-specific metabolism has been genetically altered to feed and support proliferating cancer cells [17]. The etiology of tumor growth may only be partially understood using a single type of molecular dataset. Several studies have used multi-omics data to categorize cancer patients and predict prog- noses [18,19]. Therefore, a crucial stage in the machine-learning-model-based prediction of survival and recurrence is learning new characteristics from multi-omics data that help predict prognosis.

Multi-omics analysis of ACC can reveal several undiscovered oncogenic alterations and guide the exploration of new therapeutic approaches. In this study, we collected clinical and pathological characteristics, DNA methylation profiles, genomic mutations, and mRNA, lncRNA, and microRNA (miRNAs) information of patients, and somatic mutation data from four ACC datasets. Multi-omics analysis using the MOVICS algorithm provided new clues for the prognosis and treatment of patients with ACC.

2. Materials and Methods

2.1. Data Collection

Multi-omics data of ACC patients, including DNA methylation, gene mutations, mRNA, miRNA, and lncRNA, were downloaded from the TCGA-ACC dataset for molec- ular subtyping. The “TCGAbiolinks” R package (version 2.25.3) which is provided in https://github.com/BioinformaticsFMRP/TCGAbiolinks.git, was used to obtain clinical features and transcriptomic expression data. Gene symbol annotation was performed as described in our previous study [20]. The miRNA expression and DNA methylation 450 matrices were downloaded from the UCSC Xena (https://xenabrowser.net/datapages, accessed on 15 April 2022). Somatic mutation data were downloaded from the cbiopportal (https://www.cbioportal.org/, accessed on 15 April 2022). After combining all available data from the different patients, 78 eligible patients were included in the TCGA-ACC cohort. In addition, we extracted data from 89 patients with ACC from the GSE19750, GSE33371, and GSE49278 cohorts [21-24]. The batch effect is an abiotic difference between at least two datasets. To eliminate the bias caused by the batch effect, we used the combat algorithm

package “SVA” (version 3.46.0). The GEO combined cohort was used as the subsequent test cohort, and the TCGA-ACC cohort was used as the training cohort.

2.2. Molecular Subtypes were Identified Using Multi-Omics Analysis

Molecular subtypes were determined using multi-omics data according to recently published guidelines for the R package “MOVICS” (version 1.0) [25]. The input for “MOVICS” is multi-omics data, and the output is the recommended molecular subtypes, presenting molecular features, prognosis, treatment sensitivity, and others; the VIGNETTE of this package is provided in https://xlucpu.github.io/MOVICS/MOVICS-VIGNETTE. html (accessed on 15 April 2022), with details of how to use it. First, univariate Cox re- gression analysis was used to evaluate factors related to overall survival (OS), including relevant biological information downloaded as above (all p < 0.05). Mutated genes were those with a mutation frequency greater than 10%. Based on the evaluation of the multi- omics data, the cluster prediction index (CPI) [26] and the gap statistic [27] were utilized to determine the correct number of subtypes. The number that resulted in the maximum value of both the gap statistic and the CPI was chosen as the optimum number of clusters for the input data. Ten clustering algorithms (iClusterBayes, moCluster, CIMLR, IntNMF, ConsensusClustering, COCA, NEMO, PINSPlus, SNF, and LRA) were then used to separate patients into various subtypes, and a combined classification using a consensus set was used to identify each subtype with a high degree of robustness.

Specifically, if tmax algorithms are specified where 2 ≤ tmax ≤ 10, the package cal- culates a matrix M ___ per algorithm, where n is the number of samples and MO = 1 when samples i and j are clustered in the same subtype; otherwise, MU = 0. After obtaining all results from specified algorithms, MOVICS calculates a consensus matrix CM = Etmax M(+), and CMij€ [0, 10]. The sample similarity among the subtypes was calculated using silhouette scores.

2.3. Characteristics of Genetic Variations among Subtypes

Tumor-mutation burden (TMB) is the number of mutations per million bases, and fraction genome alteration (FGA) refers to the percentage of gene fragments with increased or lost copy numbers in the total genome. Total neoantigen and cytolytic activity (CYT) scores from previous studies were predicted by analyzing tumor-specific mutations, splic- ing, gene fusions, endogenous reverse transcription factors, and other criteria [28]. The downloaded copy number data from FireBrowse (http://firebrowse.org/, accessed on 15 April 2022) were visualized using the MafTools R package (version 2.14.0).

2.4. Comparison of Signaling Pathway Activation and Immune Infiltration

Single-sample gene-set enrichment analysis (ssGSEA) [28] package “GSVA” (version 1.1.11) was used to analyze 50 HALLMARK gene sets for each patient to reveal the ac- tivation of biological pathways. The Molecular Signatures Database (MSigDB, accessed on 15 April 2022) is one of the most widely used and comprehensive databases of gene sets for performing gene set enrichment analyses. The developers used a combination of automated approaches and expert curation to develop a collection of “hallmark” gene sets as part of the MSigDB. Each hallmark in this collection consists of a “refined” gene set derived from multiple “founder” sets that convey a specific biological state or process and display coherent expression [29]. The enrichment score (ES) represents the main result of gene enrichment analysis. In the ranking list, the ES-positive gene sets were at the top, and the ES-negative gene sets were at the bottom. The normalized enrichment score (NES) was the main evaluation index of the gene set enrichment results. The false discovery rate (FDR) is the rate of errors occurring in all discoveries with a set threshold of 0.05. We evaluated the immune infiltration of immunocytes and the activated status of the immune signature. Gene sets were collected from prior studies. The NES scores of different subgroups were calculated from the gene sets associated with immune and stromal features extracted from previous studies to demonstrate differences in the immune activation status [30]. We also

used ssGSEA to study the infiltration of 28 immune cells in tumors and calculated the infiltration score of each immune cell in each patient [31]. Metabolism-associated path- ways were obtained from the study of Possemato et al [32]. All of the above results were visualized using heat maps.

2.5. Prediction of Immunotherapy and Chemotherapy Treatment

To evaluate individual responses to immunotherapy, we used 795 specific gene sets found by other researchers in the melanoma cohort with anti-CTLA-4 or anti-PD-1 im- munosuppressive therapy as a reference [33]. Subclass mapping (SubMap) was used to analyze the similarity between the risk group and the immunotherapy subgroup and iden- tify patients who responded better to both immunotherapy agents [34]. The susceptibility to chemotherapeutic drugs was determined by estimating the half-maximum inhibitory concentration (IC50) of the samples using the Cancer Drug Sensitivity Genomics (GDSC) database and ridge regression analysis.

2.6. Statistical Analysis

The Kruskal-Wallis test was used to compare continuous data between the three groups. The relationship between these two factors was evaluated using Pearson’s correla- tion coefficient. The distribution of categorical variables among the groups was compared using the chi-square test. In the external validation cohort, the top 200 specific marker genes in the TCGA-ACC cohort were selected by nearest template prediction (NTP) analysis [35]. The log-rank test and K-M analysis were used to compare the survival rates of the high- and low-risk groups. Cox models were used to calculate the HR and 95% CI. The risk score relied on multivariate COX regression analysis to identify whether it had an independent prognostic effect and was bounded by p < 0.05. All analyses were performed using the R software (version 4.1.2) (http://www.r-project.org, Bell Laboratories, Windsor, WI, USA).

3. Results

3.1. Establishment of Molecular Subtypes

As shown in Figure 1a, when the number of clusters was three, the scores of the gap -statistical and CPI analyses were the highest. We then applied 10 multi-omics ensemble clustering algorithms to the three preset clusters and combined the results. Favorable consistency was observed among the three clusters using the ten algorithms (Figure 1b). Then, we evaluated the cluster quality via silhouette analysis, and the high silhouette width represented the robustness of the three clusters (0.74 vs. 0.64 vs. 0.47) (Figure 1c). Therefore, we redefined ACCs into three subtypes: ACC1, ACC2, and ACC3. Based on the multi-omics data in the TCGA-ACC cohort, we visualized diverse molecular features among the three subtypes, and the top ten items for each omics are listed in Figure 1d. In addition, we observed significantly different clinical outcomes among the three subtypes. ACC2 patients showed more advanced stages (59.1% vs. 15.6% vs. 50.0%, p = 0.022, Supplementary Table S1) and shorter overall survival, disease-specific survival, disease- free interval, and progression-free interval than ACC1 and ACC3 patients (all p < 0.001, Figure 2). ACC2 represented the poorest phenotype, ACC1 represented the best, and ACC3 was moderate.

Figure 1. Recognition of the adrenocortical carcinoma multi-omics classification system in the TCGA-ACC cohort. (a) CPI analysis and gap-statistical analysis results. (b) Consensus matrix for three clusters based on the 10 algorithms. (c) Silhouette-analysis evaluation of cluster quality. (d) Visualization of multi-omics data for mRNAs, IncRNAs, miRNAs, DNA CpG methylation sites, and mutant genes.

a

d

1.0-

1.0

Stage

Age

Cluster Prediction Index

Gender

0.8-

-0.8

Laterality

Stage

Gap-statistics

Subtype

Stage I

0.6-

-0.6

Stage II

Stage III

- NROB1

ERCC6L

Stage IV

ARHGAP11B

mRNA

unknown

0.4-

-0.4

KIF14

- ZWILCH

Age

BAAT

80

KCNJ14

- POLRIE

60

0.2-

-0.2

PCGF6

40

20

- DGKD

0

0.0-

0.0

Gender

2

3

4

5

6

7

8

LINC00887

TP53TG1

FEMALE

Number of Multi-Omics Clusters

EIF1B-AS1

MALE

IncRNA

- PRKCQ-AS1

Laterality

b

SNHG3

1

2

Th

Left

Subtype

Right

- AC004801.6

Sample

AL031705.1

Subtype

AC098484.2

ACC1

- FAM225A

ACC2

Similarity

- AC108463.2

ACC3

1

0.8

mRNA.FPKM

0.6

2

0.4

1

0.2

miRNA

0

-1

Subtype

MIMAT0000226

MIMATO000227

-2

ACC1

MIMAT0000691

MIMAT0004680

IncRNA.FPKM

ACC2

MIMAT0004509

ACC3

MIMAT0004672

2

MIMAT0003258

MIMAT0005951

1

MIMAT0014990

/ MIMAT0004553

0

r cg08065231

-1

og09578475

cg24189904

-2

cg07139762

cg14973588

miRNA.FPKM

Methylation

2

1

cg06391107

0

C

Silhouette plot

3 clusters Cj

og27517652

cg04175957

-1

j: n | avejecj Si

cg14385298

cg08621203

-2

M value

ACC1

1: 33 | 0.64

2

1

KMT2B

COL18A1

0

PCDH15

-1

ACC2

2: 21 | 0.74

-2

Mutation

- OBSCN

Mutated

STAB1

APOB

- MUC4

0 1

ACC3

3: 24 | 0.47

- FAT4

TLN2

DST

-0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Figure 2. Differential survival outcome in three ACC subtypes, log-rank test.

Overall survival

Disease specific survival

100

100

Survival probability (%)

Survival probability (%)

75

75

50

50

25

Overall p < 0.001

25

Overall p < 0.001

ACC1

ACC2

ACC1

ACC1

ACC2

ACC2

<0.001

ACC2

ACC1

ACC3

ACC2

<0.001

ACC2

0

ACC3

0.008

<0.001

0

ACC3

0.014

<0.001

ACC3

0

12

24

36

Time (Months)

48

60

72

84

96

108

120

0

12

24

36

48

60

72

84

96

Time (Months)

108

120

Number at risk

Number at risk

33

33

30

21

17

15

11

8

6

4

2

33

33

30

21

17

15

11

8

6

4

2

21

18

8

6

1

1

0

0

0

0

0

20

17

8

6

1

1

0

0

0

0

0

24

22

20

16

12

8

5

3

2

2

2

23

21

19

15

11

8

5

3

2

2

2

0

12

24

36

Time (Months)

48

60

72

84

96

108

120

0

12

24

36

Time (Months)

48

60

72

84

96

108

120

Disease free interval

Progression free interval

100

100

Survival probability (%)

Survival probability (%)

75

75

ACC1

ACC2

ACC2

<0.001

Overall p < 0.001

50

ACC3

0.007

0.089

50

ACC1

ACC2

25

Overall p < 0.001

++ACC3

25

ACC1

ACC1

ACC2

ACC2

ACC2

<0.001

q

ACC3

q

ACC3

<0.001

0.002

0

12

24

36

48

60

72

Time (Months)

84

96

108

120

0

12

24

36

Time (Months)

48

60

72

84

96

108

120

Number at risk

Number at risk

28

27

24

17

13

13

9

7

5

4

2

33

31

28

18

14

13

10

7

5

4

2

4

4

1

0

0

0

0

0

0

0

0

21

9

1

0

0

0

0

0

0

0

0

13

11

10

7

5

5

2

1

1

1

1

24

16

12

8

6

5

2

1

1

1

1

0

12

24

36

Time (Months)

48

60

72

84

96

108

120

0

12

24

36

Time (Months)

48

60

72

84

96

108

120

3.2. Signaling Pathway Activation in ACC Subtypes

In 50 HALLMARK terms for each patient, we found that ACC1 patients had more immune activation, such as interferon alpha response [36], interferon gamma response [37] and Kras signal up [38]. ACC2 patients had more cell-cycle-related pathway activation, such as G2M checkpoint [39], E2F targets, and DNA repair (Figure 3a). After comparing 100 pathways related to metabolism of ACCs, it was observed that ACC1 patients were enriched in drug metabolism by other enzymes [40], retinol metabolism, and pentose and glucuronate interconversion pathways. ACC2 patients were enriched in the homocysteine cycle, the methionine cycle, and pyrimidine biosynthesis. Steroid biosynthesis [41], choles- terol biosynthesis [42], and terpenoid backbone biosynthesis pathways were enriched in ACC3 patients (Figure 3b). Specifically, pathways related to DNA repair [43] were enriched in ACC2 and ACC3 patients (Figure 3c).

Enrichment Score

1

Subtype

Enrichment Score

a

Subtype

Laterality

0.5

b

Laterality

1

0.5

Gender

0

Gender

0

Age

Age

-0.5

Stage

-0.5

Stage

-1

N-Glycan Biosynthesis

-1

APICAL_JUNCTION

Glycosphosphatidylinositol

EPITHELIAL_MESENCHYMAL_TRANSITION

Other Types of O-Glycan Biosynthesis

ANGIOGENESIS

Glycosphingolipid Biosynthesis

MYOGENESIS

Glycogen Biosynthesis

APICAL_SURFACE

Hexosamine Biosynthesis

INTERFERON_ALPHA_RESPONSE

Lysine Degradation

INTERFERON_GAMMA_RESPONSE

Inositol Phosphate Metabolism

KRAS_SIGNALING_UP

Shingolipid Metabolism

COMPLEMENT

Caffiene Metabolism

IL2_STAT5_SIGNALING

Retinoid Metabolism

ALLOGRAFT_REJECTION

Mucin Type O-Glycan Biosynthesis

IL6_JAK_STAT3_SIGNALING

Glycogen Degradation

INFLAMMATORY_RESPONSE

Remethylation

P53_PATHWAY

Vitamin K

COAGULATION

Glycosaminoglycan Biosynthesis

TNFA_SIGNALING_VIA_NFKB

Biotin Metabolism

APOPTOSIS

Lipoic Acid Metabolism

BILE_ACID_METABOLISM

ADP-Ribosylation

OXIDATIVE_PHOSPHORYLATION

Polyamine Biosynthesis

PEROXISOME

D-Glutamine and D-Glutamate Metabolism

ADIPOGENESIS

Thiamine Metabolism

FATTY_ACID_METABOLISM

Transsulfuration

KRAS_SIGNĀLING_DN

Taurine and Hypotaurine Metabolism

PANCREAS BETA_CELLS

Cardiolipin Biosynthesis

ESTROGEN_RESPONSE_EARLY

Sirtuin Nicotinamide Metabolism

ESTROGEN_RESPONSE LATE

Amino Sugar and Nucleotide Sugar Metabolism

XENOBIOTIC_METABOLISM

Glycosaminoglycan Degradation

REACTIVE_OXYGEN_SPECIES_PATHWAY

Other Glycan Degradation

ΗΥΡΟΧΙΑ

Porphyrin and Chlorophyll Metabolism

GLYCOLYSIS

Pentose and Glucuronate Interconversions

UV_RESPONSE_UP

Ascorbate and Aldrate Metabolism

CHOLESTEROL_HOMEOSTASIS

Retinol Metabolism

MTORC1_SIGNALING

Drug Metabolism by other enzymes

SPERMATOGENESIS

Riboflavin Metabolism

G2M_CHECKPOINT

Nicotinamide Adenine Dinucleotide Biosynthesis

E2F TARGETS

Nicotinate and Nicotinamide Metabolism

UNFOLDED_PROTEIN_RESPONSE

Metabolism of Xenobiotics by Cytochrome P450

DNA_REPAIR

Drug Metabolism by Cytochrome P450

MYC_TARGETS_V1

Metabolism

MYC_TARGETS_V2

Steroid Hormone Biosynthesis

WNT_BETA_CATENIN_SIGNALING

Tyrosine Metabolism

HEDGEHOG_SIGNALING

Phenylalanine Metabolism

Cardiolipin Metabolism

NOTCH_SIGNALING

TGF BETA SIGNALING

Oxidative Phosphorylation

UV_RESPONSE_DN

Nitrogen Metabolism

MITOTIC_SPINDLE

Glycerophospholipid Metabolism

Primary Bile Acid Biosynthesis

PI3K_AKT_MTOR_SIGNALING

ANDROGEN_RESPONSE

Pantothenate and CoA Biosynthesis

PROTEIN_SECRETION

Dopamine Biosynthesis

Epinephrine Biosynthesis

HEME_METABOLISM

Norepinephrine Biosynthesis

Ether Lipid Metabolism

C

Linoleic Acid Metabolism

alpha-Linoleic Acid Metabolism

Cyclooxygenase Arachidonic Acid Metabolism

Prostanoid Biosynthesis

Neomycin, Kanamysin and Gentamicin Biosynthesis

Galactose Metabolism

Starch and Suctose Metabolism

Kynurenine Metabolism

Arachidonic Acid Metabolism

Retinoic Acid Metabolism

Prostaglandin Biosynthesis

Glutathione Metabolism

KEGG_P53_SIGNALING_PATHWAY

Fructose and Mannose Metabolism

Gluconeogenesis

BIOCARTA_RB_PATHWAY

Glycolysis

BIOCARTA_P53_PATHWAY

Alanine, Aspartate and Glutamate Metabolism

Urea Cycle

ST_WNT_BETA_CATENIN_PATHWAY

Arginine Biosynthesis

KEGG_WNT_SIGNALING_PATHWAY

Fatty Acid Elongation

WNT_SIGNALING

Biosynthesis of Unsaturated Fatty Acids

Propanoate Metabolism

GO_CHROMATIN_REMODELING

Valine, Leucine and Isoleucine Degradation

REACTOME_CHROMATIN_MODIFYING_ENZYMES

Citric Acid Cycle

Glyoxylate and Dicarboxylate Metabolism

GO_NOTCH_SIGNALING_PATHWAY

Nicotinamide Adenine Metabolism

REACTOME_SIGNALING_BY_NOTCH

Cysteine and Methionine Metabolism

Glycerolipid Metabolism

PID_NOTCH_PATHWAY

Ketone Biosynthesis and Metabolism

Samples in TCGA-ACC cohort

Pyruvate Metabolism

Sulfur Metabolism

Selenocompound Metabolism

Vitamin B6 Metabolism

Subtype

Heme Biosynthesis

Stage

Glycine, Serine and Threonine Metabolism

Valine, Leucine and Isoleucine Biosynthesis

ACC1

Pentose Phosphate

Stage

Ubiquinone and other Terpenoid-Quinone Biosynthesis

ACC2

Fatty Acid Biosynthesis

Phenylalanine, Tyrosine and Tryptophan Biosynthesis

ACC3

Stage II

Beta-Alanine Metabolism

Stage III

Fatty Acid Degradation

Butanoate Metabolism

Stage IV

Arginine and Proline Metabolism

Histidine Metabolism

Age

unknown

Tryptophan Metabolism Homocysteine Biosynthesis

80

Methionine Cycle

Purine Biosynthesis

Folate One Carbon Metabolism

60

Pyrimidine Biosynthesis

Purine Metabolism

40

Laterality

Gender

Pyrimidine Metabolism

Terpenoid Backbone Biosynthesis

20

Left

FEMALE

Cholesterol Biosynthesis Steroid Biosynthesis

Right

MALE

Folate biosynthesis

0

Aldosterone Biosynthesis

Cortisol Biosynthesis

Estradiol Biosynthesis

Testosterone Biosynthesis

Samples in TCGA-ACC cohort

Samples in TCGA-ACC cohortEnrichment Score
Subtype1
Laterality0.5
Gender
Age0
Stage-0.5
WP_DNA_MISMATCH_REPAIR
REACTOME_MISMATCH_REPAIR-1
KEGG_MISMATCH_REPAIR

Figure 3. Differential activity of tumor-associated pathways across three ACCs subtypes in TCGA- ACC cohort. (a) Heatmap of 50 differentially activated HALLMARK pathways. (b) Heatmap of 100 pathways related to metabolism of ACCs. (c) Heatmap of DNA repair pathways.

According to the TMB data, we selected six genes (DST, FAT4, KMT2B, APOB, OBSCN, and ZCCHC6) with the highest mutation rates for demonstration. ACC2 patients had the highest rates of tumor mutation (Figure 4a). We then compared the base mutations of all patients; ACC2 also had the most mutations. C > T and T > C mutations were the most common (Figure 4b). However, ACCS had no obvious difference in genome copy numbers (Figure 4c), and the patients were divided into altered and unaltered groups according to the mutation data to compare their survival outcomes. The results showed that the mutation group had worse OS (Figure 4d). In addition, protein-protein interaction

(PPI) enrichment analysis was performed using Metascape online tools based on the BioGrid, InWeb_IM, and OmniPath databases. The molecular complex detection (MCODE) algorithm was used to identify densely connected network components, and each MCODE component was independently enriched in different pathways and biological processes. Finally, seven pathways with the best scores were retained to describe the function of DEGs in ACCs, including cell-cycle function, mitotic nuclear division, and the PID PLK1 pathway (Figure 4e). Finally, we also found that these mutated genes were related to transcription factors, such as E2F1, TP53, E2F4, and YBX1 (Figure 4f).

Figure 4. Differential tumor-mutation burden and copy number in three ACC subtypes. (a) Mutations in the six genes with the highest mutation rates. (b) Base mutations in the three ACCs cohorts. (c) Differences in genome copy numbers. (d) Differences in overall survival between mutant and non-mutant groups. (e) Functional enrichment analysis of the six genes with the highest mutation rate. (f) Relationship between mutation genes and transcription factors.

a

0

2

4

6

L

8%

DST

L

Stage

Age

Laterality

8%

FAT4

Stage I

80

Left

8%

KMT2B

APOB

Stage II

60

Right

6%

6%

OBSCN

Stage III

Stage IV

40

6%

ZCCHC6

20

Subtype

Stage

unknown

0

ACC1

Age

ACC2

Gender

ACC3

Gender

FEMALE

Laterality

MALE

Subtype

b

C

n = 32

n = 22

n = 24

Copy number-altered genome

Copy number-lost genome

Copy number-gained genome

2

kruskal.test p = 4.0 x 10-6

log10 (TMB + 1)

ACC3

H

H

1

ACC2

4

H

0

ACC1

ACC2

ACC3

C>T

ACC1

H

H

4

T>C

C>A

C>G

T>A

T>G

0.6

0.4

0.2

0.0

-0.4

-0.2

0.0

0.2

FGA (Fraction of Genome Altered)

FGL or FGG (Fraction of Genome Lost or Gained)

d

e

100%

Logrank Test p-Value: 3.001 x 10

-3

mitotic nuclear division

Probability of Overall Survival

Cell Cycle

90%

Altered group

PID PLK1 PATHWAY

80%-

Unaltered group

cellular response to DNA damage stimulus

70%

meiotic cell cycle

60%

Mitotic G1 phase and G1/S transition

50%

APC/C-mediated degradation of cell cycle proteins

40%

30%

20%

10%

0%

0

10

20

30

40

50

60

70

80

90

100

110

120

130

140

150

f

Overall Survival (Months)

Regulated by: E2F1

Regulated by: TP53

Regulated by: E2F4

Regulated by: YBX1

Regulated by: E2F3

Regulated by: TFDP1

Regulated by: KAT2B

Regulated by: MYC

Regulated by: RB1

Regulated by: SP1

Regulated by: MYCN

Regulated by: ATM

0

2

4

6

8

10

12

14

-log10(P)

3.3. ACC1 Patients May Benefit More from Anti-PD-1 Therapy, and Chemotherapy Is More Suitable for ACC2 Patients

To further evaluate the immune status of the three subtypes, we compared their esti- mated immune scores, immune cell subsets, and immune signaling molecules, which have been reported to serve as biomarkers for immunotherapy [31,44]. As shown in Figure 5a, the ACC1 subtype exhibited the highest estimated immune score (p = 11.1 x 10-8), im- mune cell subsets (p =6.9 x 10-8), and immune signaling molecules (p = 1.6 × 10-6), which was consistent with the results of the pathway analysis. We investigated the infiltration landscape of different immune cells among the three subtypes. Compared to ACC2 and ACC3 subtypes, the ACC1 subtype had higher filtration of immunocytes such as central- memory CD4 T cells, plasmacytoid dendritic cells, mast cells, macrophages, regulatory T cells, activated CD8 T cells, and T helper cells. Based on 18 immune-related signatures that have been published [45], the ACC1 subtype exhibited higher immune scores in most signatures, including cytotoxic cells, T.NK. meta, CYT, treg cells, T cells, 13 T-cell signatures, TLS, WNTTGFB signatures, B cell cluster, 6 gene IFN signatures, macrophages, and MDSC (Figure 5b). PD-L1 is a special protein in tumor cells that can bind to PD1 on effector T cells to induce T cell exhaustion, which is a pivotal factor implicated in tumor immune escape [46]. We found that the ACC1 subtype expressed more PD-L1 and PD-1 than the ACC2 and ACC3 subtypes (all p < 0.05, Figure 5c). Therefore, the ACC1 subtype repre- sented the high immune-infiltration phenotype, which was also called “hot tumor” and indicated a potential response to anti-PD-1 therapy [47]. Furthermore, SubMap analysis showed that patients with the ACC1 subtype would benefit more from anti-PD-1 therapy (Bonferroni p < 0.05, Figure 5d). In addition, we evaluated the susceptibility of the three subtypes to chemotherapy. The results showed that ACC2 patients had better sensitivity to cisplatin (p=3.9 × 10-6), doxorubicin (p=3.4 × 10-6), gemcitabine (p=1.7×10-7), and etoposide (p = 3.3 x 10-6) (Figure 6). Collectively, this novel ACC molecular classification may facilitate the selection of appropriate treatments for different patients. The treatment of ACC1 patients with anti-PD-1 therapy and ACC2 patients with cisplatin, doxorubicin, gemcitabine, or etoposide appears to confer more clinical benefits.

3.4. Extra Validation for Molecular Subtypes in GEO Cohorts

To further validate the results in the TCGA-ACC cohort, three GEO cohorts were enrolled: GSE19750, GSE33371, and GSE49278. A combat algorithm was first conducted to eliminate the batch effect of the three GEO cohorts to make the data more comparable (Figure 7a). To distinguish the three subtypes, the top 300 specific genes for each subtype were selected to represent the separation of the three subtypes (Figure 7b). Consistently, we compared the OS of different subtypes and found that ACC2 patients had the worst prognosis (p < 0.001, Figure 7c), which was consistent with the results in the TCGA-PRAD cohort. ACC2 patients had worse survival than ACC1 and ACC3 (17.6% vs. 62.5% vs. 43.5%, p < 0.01) and a worse proportion of advanced-stage disease (55.9% vs. 18.7% vs. 26.0%, p = 0.031, Supplementary Table S2). In the pathway enrichment analysis, ACC1 patients had more activation of immune and drug-metabolism pathways, while ACC2 patients had more cell-cycle-related pathway activation, and pathways related to DNA repair were enriched in ACC2 and ACC3 patients. These functional results were similar to those of the TCGA-ACC cohort (Figure 8). Among the 18 immune-related signatures, ACC1 patients also scored highest for most items, including cytotoxic cells, T.NK. meta, CYT, treg cells, T cells, 13 T-cell signatures, TLS, B cell cluster, 6 gene IFN signatures, macrophages, and MDSC. The results showed that ACC1 represented a high immune activation phenotype and was potentially susceptible to targeted immunotherapy. Furthermore, it is worth mentioning that the ACC3 subtype also exhibited relatively higher immune scores, despite being well below ACC1 (Figure 9a). SubMap analysis showed that the patients with ACC1 were sensitive to anti-PD-1 therapy. We observed that ACC3 patients were sensitive to anti-CTLA-4 therapy (Figure 9b). Moreover, drug-sensitivity tests showed that ACC2 patients had the highest drug sensitivity to cisplatin, doxorubicin, gemcitabine, etoposide,

and paclitaxel (Figure 9c) (all p < 0.01). In addition, our system remained an independent prognostic factor in the four ACC patient cohorts after adjustment for other major clinical characteristics (all p < 0.05, Table 1). This shows the reliability of our classification.

Figure 5. Differences of immune infiltration in three ACC subtypes. (a) Score of related indicators of immune infiltration. (b) Heatmap of infiltration landscape of different immune cells between the three subtypes. (c) Expression of the three subtypes to PD-L1 and PD-1. (d) SubMap analysis of anti-PD-L1 therapy. ** p<0.01, *** p < 0.001, **** p < 0.0001.

A

Subtype ACC1 ACC2 E ACC3

Subtype

ACC1

ACC2

ACC3

Subtype E ACC1 ACC2 E ACC3

0.8+

Anova, p = 1.9x 10-6

Anova, p = 1.2 x 10”

5

Anova, p = 6.5 x 10-6

Estimate Immune Score

Immune Cell Subsets

Immune Signaling Molecules

0.4

0.5

0.4

0.0

0.0

0.0

-0.4

-0.5

-0.4

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

-0.8

ACC1

ACC2

ACC3

Subtype

Subtype

Subtype

B

Subtype

1

Subtype

1

Laterality

0.5

Laterality

0.5

Gender

0

Gender

0

Subtype

Age

Stage

-0.5

Age

ACC1

-1

Stage

-0.5

-1

ACC2

Central memory CD4 T cell

Cytotoxic cells

ACC3

Plasmacytoid dendritic cell

Mast cell

T.NK. meta

Laterality

Activated dendritic cell

CYT

Left

Macrophage MDSC

Treg cells

Right

Regulatory T cell

T cells

Gender

T follicular helper cell

Activated CD8 T cell

13 T-cell signature

FEMALE MALE

Monocyte

Effector memeory CD4 T cell

TLS

Natural killer cell

WNTTGFB signature

Age

Effector memeory CD8 T cell

80

Type 1 T helper cell

B cell cluster

60

Activated B cell

6 gene IFN signature

Immature B cell Neutrophil

40

Macrophages

20

Type 17 T helper cell

MDSC

0

CD56bright natural killer cell

CD56dim natural killer cell Eosinophil

B.P. meta

Stage

TGFB1 activated

unknown

Immature dendritic cell

Central memory CD8 T cell

TITR

Stage I

Memory B cell

C ECM

Stage II

Activated CD4 T cell

Stage III

Type 2 T helper cell

CD8 T cells

Stage IV

Gamma delta T cell

Natural killer T cell

Th17 cells

samples in TCGA-ACC cohort

samples in TCGA-ACC cohort

C

1

type

ACC1

ACC2

ACC3

D

ACC1

5




ACC2

0.8

Immunocyte Infiltration

ACC3

0.6

4

3

ACC1

0.4

ACC2

2

0.2

ACC3

1

pvalue

CTAL4-noR

CTLA4-R

PD1-noR

PD1-R

pvalue

0

Nominal p value

PD-L1

PD-1

PD-L2

CTLA4

Bonferroni corrected

Figure 6. Differences in chemotherapy susceptibility between the three subtypes.

5.0-

9

o

A

Estimated IC50 of Doxorubicin

·

S

O

P

Estimated IC50 of Cisplatin

Estimated IC50 of 5-Fluorouracil

-1

3

UP

3

6

·

4.5

9

O

D

8

S

0

0

8

· D

Q

6

el

0

O

o

0

8

Ba

S

0

D.

.

0

&

·

4.0

8

8

a

4

0

O

4

®

8

-2-

·

·

o

®

·

2-

8

LO

3.5.

·

0

3.0-

Kruskal-Wallis, p = 3.9 x 106

-3-

Kruskal-Wallis, p = 3.4 x 106

Kruskal-Wallis, p = 0.068

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

Estimated IC50 of Gemcitabine

0

3-

Estimated IC50 of Etoposide

-3.0-

O

O

1

8

o ·

·

0

Estimated IC50 of Paclitaxel

9

O

5

8

8

·

D

8

3.5

5

6

2-

188

-2

9

P

o

9

Ht

%

0 4

0 00%

O

8

.

0

.

0

A

.

0

0

8

O

.

a

0

D

O

O

·

:+

C

4.0-

·

9

9

-3

o

0

8

b

1-

P

O

00

0

o

0

.

·

-4.5-

-4-

Krusk

-Wallis, p = 1.7 x 107

0-

Krusk

-Wallis, p = 3.3 x 10-6

-5.0-

Krusk

-Wallis, p = 0.056

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

Figure 7. Recognition of the adrenocortical carcinoma multi-omics classification system in the GEO cohort. (a) The combat algorithm eliminates the queue batch effect. (b) Representing the ACCs in the GEO cohort. (c) Differential overall survival outcome in reproduced ACCs of GEO cohorts, log-rank test.

a

Raw PCA for combined expression profile

Combat PCA for combined expression profile

8

8

Comp 2: 14.4% variance

3

Comp 2: 5.7% variance

40

40

2

20

GSE19750

GSE33371

GSE49278

0

-

0

-20

-20

-40

GSE19750

GSE33371

GSE49278

-150

-100

-50

0

50

-60

-40

-20

0

20

40

b

Comp 1: 63.8% variance

Comp 1: 7.3% variance

C

ACC1

ACC2- ACC3

100

Overall p < 0.001

75

template features

Survival probability (%)

50

25

I

ACC1

ACC2

ACC2

<0.001

-

0

ACC3

0.011

0.011

0

12

24

36

48

60

72

84

96

108

Time (Months)

120

Number at risk

32

31

29

26

26

23

21

18

16

12

10

1

34

18

10

5

5

3

3

3

3

3

2

p - value

ACC2

ACC3

0

23

19

17

14

8

7

7

5

4

4

2

ACC1

0

12

24

36

48

60

72

84

96

108

120

class predictions

Time (Months)

Enrichment Score

Enrichment Score

a

1

b

Subtype

1

Side

0.5

Subtype

Side

0.5

Gender

0

Gender

0

Age

-0.5

Age

-0.5

Stage

-1

Stage

INTERFERON ALPHA_RESPONSE

INTERFERON GAMMA_RESPONSE IL6_JAK_STAT3_SIGNALING

Aldosterone Biosynthesis

-1

Cortisol Biosynthesis

Estradiol Biosynthesis

INFLAMMATORY_RESPONSE

Testosterone Biosynthesis

ALLOGRAFT_REJECTION

Glycerolipid Metabolism

TNFA_SIGNALING_VIA_NFKB

Pyruvate Metabolism

APOPTOSIS

Propanoate Metabolism

KRAS_SIGNALING_UP

Biosynthesis

COAGULATION

Fatty Acid Degradation

COMPLEMENT

Valine, Leucine and Isoleucine Degradation

IL2_STAT5_SIGNALING

Terpenoid Backbone Biosynthesis

MITOTIC_SPINDLE

Cholesterol Biosynthesis

Biosynthesis

G2M_CHECKPOINT

Steroid Hormone Metabolism

E2F TARGETS

Steroid Hormone Biosynthesis

DNA REPAIR

Fatty Acid Elongation

MYC_TARGETS_V1

Biosynthesis of Unsaturated Fatty Acids

MYC_TARGETS_V2

Ketone Biosynthesis and Metabolism

ESTROGEN_RESPONSE_EARLY

Butanoate Metabolism

ESTROGEN RESPONSE_LATE

Glutathione Metabolism

APICAL_SURFACE

Metabolism of Xenobiotics by Cytochrome P450

MYOGENESIS

Drug Metabolism by Cytochrome P450

Ascorbate and Aldrate Metabolism

APICAL JUNCTION

Histidine Metabolism

EPITHELIAL_MESENCHYMAL_TRANSITION ANGIOGENESIS

Thiamine Metabolism

Oxidative Phosphorylation

TGF_BETA_SIGNALING

Arginine and Proline Metabolism

UV_RESPONSE_DN

Beta-Alanine Metabolism Purine Metabolism

HEDGEHOG_SIGNALING

WNT_BETA_CATENIN_SIGNALING

Pyrimidine Metabolism

Pentose and Glucuronate Interconversions

NOTCH_SIGNALING

Glyoxylate and Dicarboxylate Metabolism Glycine, Serine and Threonine Metabolism Folate biosynthesis Urea Cycle

BILE ACID_METABOLISM

OXIDATIVE_PHOSPHORYLATION

ADIPOGENESIS

FATTY ACID_METABOLISM

Primary Bile Acid Biosynthesis

P53_PATHWAY

Nicotinamide Adenine Dinucleotide Biosynthesis

XENOBIOTIC_METABOLISM

Nicotinate and Nicotinamide Metabolism

REACTIVE_OXYGEN SPECIES_PATHWAY

Selenocompound Metabolism

Lipoic Acid Metabolism

PROTEIN_SECRETION

Nicotinamide Adenine Metabolism

HEME METABOLISM

Polyamine Biosynthesis

SPERMATOGENESIS

D-Glutamine and D-Glutamate Metabolism

KRAS_SIGNALING_DN

Nitrogen Metabolism

PANCREAS_BETA CELLS

Alanine, Aspartate and Glutamate Metabolism

ANDROGEN_RESPONSE

Arginine Biosynthesis

MTORC1_SIGNALING

Folate One Carbon Metabolism

CHOLESTEROL_HOMEOSTASIS

Vitamin K

PEROXISOME

Ubiquinone and other Terpenoid-Quinone Biosynthesis

PI3K_AKT_MTOR_SIGNALING

Gluconeogenesis Glycolysis

HYPOXIA

Starch and Suctose Metabolism

GLYCOLYSIS

Galactose Metabolism

UNFOLDED_PROTEIN_RESPONSE

Neomycin, Kanamysin and Gentamicin Biosynthesis

UV_RESPONSE_UP

Glycogen Biosynthesis

Samples in GEO cohort

Glycogen Degradation

Hexosamine Biosynthesis

C

Enrichment Score

1

Biotin Metabolism

Arachidonic Acid Metabolism

Cyclooxygenase Arachidonic Acid Metabolism

Prostanoid Biosynthesis

Prostaglandin Biosynthesis

Transsulfuration

Retinoic Acid Metabolism

Phenylalanine, Tyrosine and Tryptophan Biosynthesis

WP_DNA_MISMATCH_REPAIR

Ether Lipid Metabolism

Linoleic Acid Metabolism Heme Biosynthesis

REACTOME_MISMATCH_REPAIR

KEGG_MISMATCH_REPAIR

Porphyrin and Chlorophyll Metabolism

N-Glycan Biosynthesis

GO_CHROMATIN_REMODELING

Amino Sugar and Nucleotide Sugar Metabolism

REACTOME_CHROMATIN_MODIFYING_ENZYMES

Fructose and Mannose Metabolism

Glycerophospholipid Metabolism

KEGG_P53_SIGNALING_PATHWAY

Caffiene Metabolism

BIOCARTA_RB_PATHWAY

Drug Metabolism by other enzymes

Pantothenate and CoA Biosynthesis

BIOCARTA_P53_PATHWAY

Citric Acid Cycle

GO_NOTCH_SIGNALING_PATHWAY

Pentose Phosphate Riboflavin Metabolism Remethylation

REACTOME_SIGNALING_BY_NOTCH

PID_NOTCH_PATHWAY

Sirtuin Nicotinamide Metabolism

Cardiolipin Metabolism

ST_WNT_BETA_CATENIN_PATHWAY

Pyrimidine Biosynthesis

KEGG_WNT_SIGNALING_PATHWAY

Lysine Degradation

C

Kynurenine Metabolism

WNT_SIGNALING

Tryptophan Metabolism

Samples in GEO cohort

Homocysteine Biosynthesis

Methionine Cycle

Subtype

Stage

ADP-Ribosylation

Glycosphosphatidylinositol

ACC1

Stage I

Cardiolipin Biosynthesis Purine Biosynthesis

ACC2

Other Types of O-Glycan Biosynthesis

Stage II

Dopamine Biosynthesis

ACC3

Epinephrine Biosynthesis

Stage III

Norepinephrine Biosynthesis

Glycosphingolipid Biosynthesis

Stage IV

Glycosaminoglycan Degradation

Other Glycan Degradation

Age

unknown

Shingolipid Metabolism

Valine, Leucine and Isoleucine Biosynthesis

80

Inositol Phosphate Metabolism

Glycosaminoglycan Biosynthesis

60

alpha-Linoleic Acid Metabolism

Sulfur Metabolism

Gender

Vitamin B6 Metabolism

40

Laterality

Taurine and Hypotaurine Metabolism

Left

FEMALE

Tyrosine Metabolism

20

Phenylalanine Metabolism

0

Right

MALE

Retinoid Metabolism

Retinol Metabolism

Cysteine and Methionine Metabolism

Mucin Type O-Glycan Biosynthesis

Samples in GEO cohort

Subtype0.5 0
Side
Gender-0.5
Age-1
Stage

Figure 8. Differential activity of tumor-associated pathways across three ACCs subtypes in GEO cohort. (a) Heatmap of 50 differentially activated HALLMARK pathways. (b) Heatmap of 100 pathways related to metabolism of ACCs. (c) Heatmap of DNA repair pathways.

Figure 9. Differences in immune infiltration and chemotherapy susceptibility of three subtypes in GEO cohort. (a) Heatmap of infiltration landscape of different immune cells between the three subtypes. (b) SubMap analysis of anti-PD-L1 therapy. (c) Differences in chemotherapy susceptibility between the three subtypes.

a

Subtype

1

b

Side

0.5

Nominal p value

Gender

0

ACC1

1

Age

Stage

-0.5

Subtype

-1

ACC1

WNTTGFB signature

ACC2

ACC2

0.8

TITR

ACC3

TGFB1 activated

Laterality

0.6

C ECM

Left

ACC3

CD8 T cells

Right

Cytotoxic cells

Gender

Bonferroni corrected

0.4

T cells

FEMALE

ACC1

CYT

MALE

B.P. meta

Age

0.2

B cell cluster

80

ACC2

Th17 cells

60

6 gene IFN signature

40

20

13 T-cell signature

ACC3

0

TLS

Stage

Treg cells

pvalue

CTAL4-noR

CTLA4-R

PD1-noR

PD1-R

unknown

Macrophages

Stage

T.NK. meta

Stage II

MDSC

Stage III

Stage IV

C

Samples in GEO cohort Kruskal-Wallis, p = 1.3 x 10-6

4.0-

-0.5-

Kruskal-Wallis, p = 1.4 x 10-8

4.5

Kruskal-Wallis, p = 0.78

Estimated IC50 of Cisplatin

Estimated IC50 of Doxorubicin

Estimated IC50 of 5-Fluorouracil

3.5

-1.0-

4.0

3.0

-1.5

3.5

-2.0-

2.5-

3.0

-2.5-

2.0-

2.5

-3.0

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

Kruskal-Wallis, p = 3.2 x 10-9

Kruskal-Wallis, p = 7.4 x 109

Kruskal-Wallis, p = 0.0091

0-

3

-3.35-

Estimated IC50 of Gemcitabine

Estimated IC50 of Etoposide

Estimated IC50 of Paclitaxel

-1

2-

-3.40

-2-

-3.45

1.

-3

-3.50

-4-

0-

-3.55

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

Table 1. Prognostic value of ACC subtype after adjusting for clinicopathological parameters.
HR95% CIp Value
TCGA-ACC Cohort
Age1.013(0.986-1.041)0.351
Gender, male vs. female1.455(0.614-3.451)0.394
Laterality, right vs. left1.533(0.664-3.54)0.317
Stage
Stage II vs. stage I2.883(0.301-27.628)0.358
Stage III vs. stage I6.28(0.641-61.55)0.115
Stage IV vs. stage I16.164(1.49-175.301)0.022
Stage unknow vs. stage I2.451(0.117-51.349)0.564
ACC subtype
ACC2 vs. ACC145.146(7.393-275.694)0
ACC3 vs. ACC14.661(0.877-24.779)0.071
GEO cohort
Age1.01(0.99-1.031)0.31
Gender, male vs. female1.236(0.649-2.354)0.518
Laterality
Right vs. left1.15(0.54-2.45)0.717
Unknow vs. left1.2(0.51-2.822)0.677
Stage
Stage II vs. stage I2.765(0.351-21.81)0.334
Stage III vs. stage I8.223(0.909-74.351)0.061
Stage IV vs. stage I11.723(1.504-91.399)0.019
Stage unknow vs. stage I4.067(0.453-36.548)0.21
ACC subtype
ACC2 vs. ACC14.959(2.241-10.97)0
ACC3 vs. ACC12.578(1.048-6.341)0.039

4. Discussion

The common age of onset for patients with ACC is between 50 and 70 years. Al- though most ACC is considered sporadic and the cause is unknown, a small number of cases are thought to be associated with genetic predisposition, including Lynch syn- drome, Li-Fraumeni syndrome, multiple endocrine neoplasia type 1, and familial ade- nomatous polyposis [48-51]. Ripley et al. [52] found that the first-line treatment for re- current or metastatic ACC depends on the patient’s underlying state and tumor charac- teristics. For patients who tolerate systemic chemotherapy, etoposide, doxorubicin, and cisplatin combined with mitotane (EDP-M) is superior to streptomycin-mitotane. ACC is an aggressive form of cancer, with an overall 5-year survival rate of 16-47%. The 5-year survival rates from stages I to IV were 81%, 61%, 50%, and 13%, respectively [4,53,54]. Through targeted gene analysis, mutations in TP53 or CTNNB1 have been found to be associated with molecular alterations in major ACC signaling pathways, higher tumor stage, and poorer disease-free survival (DFS). Activation of the Wnt/CTNNB1 pathway is associated with a high mitotic rate and a low survival rate. However, these markers did not show independent prognostic value in multivariate analyses, including tumor grade [55].

The immune system plays an important role in the surveillance and elimination of cancer cells, and immune evasion through various mechanisms is considered one of the characteristics of cancer [56]. Priming and activation of peripheral immune cells lead

to a T-cell inflammatory phenotype, including expansion of CD8+ cytotoxic T cells, in- terferon signaling, and local production of chemokines [57]. Rooney et al. [28] demon- strated that cytolytic immune activity, measured by industrial expression of perforin 1 and granzyme B genes, was associated with higher mutation counts. Their prediction of antigenic epitopes in a range of solid tumor malignancies supports the idea that tumor types with a high mutational burden are more susceptible to immunotherapy strategies. Several researchers have demonstrated that tumor-infiltrating lymphocytes (TILs) are as- sociated with improved clinical outcomes in ovarian-cancer patients [58-60]. Importantly, blocking PD-1, LAG-3, or CTLA-4 with gene ablation or blocking antibodies alone leads to compensatory upregulation of other checkpoint pathways, enhancing their ability to locally suppress T cells, which in turn can be overcome by a combination of blocking strategies [61].

According to Jouinot et al., the Ki67 index and targeted methylation measures of MS-MLPA can be utilized in conjunction with ENSAT staging and clinically common prog- nostic indicators [62]. Nevertheless, CpG islands are infrequently methylated, especially those connected to gene promoters. Further research is required to ascertain the extent to which DNA methylation of CpG islands controls gene expression [63]. The genetic changes identified in the targetable pathway indicate a potential route for novel treatments aimed at common chemotherapy-resistant cancers [64]. Dysregulation of miRNA subsets in ACC may contribute to the development of this malignancy. Additionally, it has been demonstrated that ACC patients with high expression of miRNA-related subsets have poorer survival rates, indicating the potential prognostic utility of these subsets [65]. Ac- cording to certain researchers, the ACC-related genes TP53 (8 of 41 tumors, 19.5%) and CTNNB1 (4 of 41 tumors, 9.8%) both exhibited somatic mutations. Somatic mutations in recurrent ZNRF3 and TERT sites and genes created by ACC are mutually exclusive. Additionally, according to gene ontology, Wnt signaling is the most often altered pathway in ACC [66].

Researchers have discovered that lncRNA SNHG3 is associated with miR-577/SMURF1 in prostate cancer and miR-139-5p/TOP2A in renal cell carcinoma. It is also associated with CDK6, Bax, Bcl-2, N-cadherin, E-cadherin, and vimentin [67,68]. When Xp21 is lost, NR0B1, which causes X-linked AHC, GK, which causes glycerol kinase deficiency, and in certain circumstances, DMD are also lost (resulting in Duchenne muscular dystrophy). When the Xp21 deletion expanded proximally to encompass DMD or when a larger loss extended distally to include IL1RAPL1 and DMD, developmental abnormalities were observed in men with Xp21 deletion [69]. DNA methylation collaborates with histone changes and miRNAs to control transcription. Additionally, research has shown that DNA methylation controls miRNA expression [63]. Some scientists have discovered that there is no connection between transcription expression and its target factors in E. coli. Furthermore, the static gene regulatory networks (GRNs) currently in use are insufficient to explain transcriptional regulation. This suggests that, when examining the cell at a systemic level, one cannot expect to observe a causal link between the expression of tran- scription factors and their targets [70]. The expression of hundreds of transcripts is often cataloged by RNA-seq measurements, but most are redundant (i.e., strongly correlated) or noisy. In addition, the number of samples available is less than the number of fea- tures owing to the expenses associated with conducting experiments, which makes it simple for conventional machine learning and statistical algorithms to overfit the biological data [71].

The biological processes of a tumor are extremely complex, and different types of features are associated with each other. Therefore, it is crucial to interpret the hetero- geneity of tumors using multi-omics analysis. We used 10 algorithms to determine the ACC multi-omics system by consensus clustering, which made the system more stable and convincing. We found that immune-related pathways were more activated, and drug metabolism pathways were enriched in ACC1 subtype patients. In addition, ACC1 patients are sensitive to PD-1 immunotherapy and have the lowest sensitivity to chemotherapeu-

tic drugs. Patients with the ACC2 subtype had the worst survival prognosis and the highest tumor-mutation rate. Meanwhile, cell-cycle-related pathways, amino acid syn- thesis pathways, and immunosuppressive cells were enriched in ACC2 patients, steroid and cholesterol biosynthetic pathways were enriched in patients with the ACC3 subtype, and DNA-repair-related pathways were enriched in subtypes ACC2 and ACC3. We as- sumed that between ACC1 and ACC2, ACC3 is the transition type. According to the results, even though ACC2 and ACC3 have a comparable distribution of tumor stage, patients who belonged to ACC2 had much worse prognoses than those who belonged to ACC3. In comparison to other subtypes, ACC3 exhibits more WNT pathway activation, greater steroid and cholesterol production, greater copy number change, and a lack of the OBSCN and ZCCHC6 mutations that were found in ACC1 and ACC2. The lowest level of immune pathway activation is then met by the ACC3 subtype. The sensitivity of the ACC2 subtype to cisplatin, doxorubicin, gemcitabine, and etoposide was better than that of the other two subtypes. For 5-fluorouracil, there was no significant difference in the sensitivity to paclitaxel between the three groups. We believe that multi-omics analysis in ACC can provide patients with more accurate clinical treatment and better prognosis prediction.

Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells11233784/s1. Table S1: Summarization of clinical features of ACC subtypes in TCGA-ACC cohort; Table S2: Summarization of clinical features of ACC subtypes in GEO cohort.

Author Contributions: Conceptualization, Y.G., S.Y. and C.L .; methodology, Y.G. and Y.C .; software, Y.P. and L.A .; validation, Y.G., H.D. and C.L .; formal analysis, Y.G .; data curation, S.Y .; writing, Y.G. and S.Y .; review and editing, H.D .; supervision, C.L .; funding acquisition, C.L. All authors have read and agreed to the published version of the manuscript.

Funding: This work is supported by the National Natural Science Foundation of China (81802827 and 81630019), the Anhui Natural Science Foundation of China (2108085QH315) and the scientific research project of the Education Department of Anhui Province (YJS20210270).

Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.

Data Availability Statement: Multi-omics data of ACC patients including DNA methylation, gene mutations, mRNA, miRNA, and LncRNA were downloaded from TCGA-ACC dataset (https://portal. gdc.cancer.gov, accessed on 15 April 2022) and GEO dataset (https://www.ncbinlm.nih.gov/geo/, accessed on 15 April 2022); the miRNA expression and DNA methylation 450 matrix were download from the UCSC Xena (https://xenabrowser.net/datapages, accessed on 15 April 2022). Download the somatic mutation data from cbiopportal (https://www.cbioportal.org/, accessed on 15 April 2022). Further enquiries can be directed to the corresponding author. We declare that the data and materials in this study will be provided free of charge to scientists for noncommercial purposes.

Conflicts of Interest: The authors declare no conflict of interest.

References

1. Lehmann, T .; Wrzesinski, T. The molecular basis of adrenocortical cancer. Cancer Genet. 2012, 205, 131-137. [CrossRef] [PubMed]

2. Else, T .; Kim, A.C .; Sabolch, A .; Raymond, V.M .; Kandathil, A .; Caoili, E.M .; Jolly, S .; Miller, B.S .; Giordano, T.J .; Hammer, G.D. Adrenocortical carcinoma. Endocr. Rev. 2014, 35, 282-326. [CrossRef]

3. Michalkiewicz, E .; Sandrini, R .; Figueiredo, B .; Miranda, E.C .; Caran, E .; Oliveira-Filho, A.G .; Marques, R .; Pianovski, M.A .; Lacerda, L .; Cristofani, L.M .; et al. Clinical and outcome characteristics of children with adrenocortical tumors: A report from the International Pediatric Adrenocortical Tumor Registry. J. Clin. Oncol. 2004, 22, 838-845. [CrossRef] [PubMed]

4. Fassnacht, M .; Johanssen, S .; Quinkler, M .; Bucsky, P .; Willenberg, H.S .; Beuschlein, F .; Terzolo, M .; Mueller, H .- H .; Hahner, S .; Allolio, B .; et al. Limited prognostic value of the 2004 international union against cancer staging classification for adrenocortical carcinoma: Proposal for a revised Tnm classification. Cancer 2009, 115, 243-250. [CrossRef] [PubMed]

5. Fassnacht, M .; Kroiss, M .; Allolio, B. Update in adrenocortical carcinoma. J. Clin. Endocrinol. Metab. 2013, 98, 4551-4564. [CrossRef] [PubMed]

6. Datta, J .; Roses, R.E. Surgical management of adrenocortical carcinoma: An evidence-based approach. Surg. Oncol. Clin. N. Am. 2016, 25, 153-170. [CrossRef] [PubMed]

7. Nekic, A.B .; Knezevic, N .; Tomsic, K.Z .; Kraljevic, I .; Balasko, A .; Polovina, T.S .; Solak, M .; Dusek, T .; Kastelan, D .; Croatian ACC Study Group. The effect of surgeon expertise on the outcome of patients with adrenocortical carcinoma. J. Pers. Med. 2022, 12, 100. [CrossRef] [PubMed]

8. Berruti, A .; Grisanti, S .; Pulzer, A .; Claps, M .; Daffara, F .; Loli, P .; Mannelli, M .; Boscaro, M .; Arvat, E .; Tiberio, G .; et al. Long-term outcomes of adjuvant mitotane therapy in patients with radically resected adrenocortical carcinoma. J. Clin. Endocrinol. Metab. 2017, 102, 1358-1365. [CrossRef] [PubMed]

9. Jouinot, A .; Armignacco, R .; Assie, G. Genomics of benign adrenocortical tumors. J. Steroid Biochem. Mol. Biol. 2019, 193, 105414. [CrossRef]

10. Giordano, T.J .; Thomas, D.G .; Kuick, R .; Lizyness, M .; Misek, D.E .; Smith, A.L .; Sanders, D .; Aljundi, R.T .; Gauger, P.G .; Thompson, N.W .; et al. Distinct transcriptional profiles of adrenocortical tumors uncovered by DNA microarray analysis. Am. J. Pathol. 2003, 162, 521-531. [CrossRef] [PubMed]

11. Xu, Y .; Qi, Y .; Zhu, Y .; Ning, G .; Huang, Y. Molecular markers and targeted therapies for adrenocortical carcinoma. Clin. Endocrinol. 2014, 80, 159-168. [CrossRef]

12. Kim, A.C .; Reuter, A.L .; Zubair, M .; Else, T .; Serecky, K .; Bingham, N.C .; Lavery, G.G .; Parker, K.L .; Hammer, G.D. Targeted disruption of beta-catenin in Sf1-expressing cells impairs development and maintenance of the adrenal cortex. Development 2008, 135, 2593-2602. [CrossRef] [PubMed]

13. Ragazzon, B .; Libé, R .; Gaujoux, S .; Assié, G .; Fratticci, A .; Launay, P .; Clauser, E .; Bertagna, X .; Tissier, F .; de Reyniès, A .; et al. Transcriptome analysis reveals that P53 and {Beta}-catenin alterations occur in a group of aggressive adrenocortical cancers. Cancer Res. 2010, 70, 8276-8281. [CrossRef] [PubMed]

14. Libé, R .; Groussin, L .; Tissier, F .; Elie, C .; René-Corail, F .; Fratticci, A .; Jullian, E .; Beck-Peccoz, P .; Bertagna, X .; Gicquel, C .; et al. Somatic Tp53 mutations are relatively rare among adrenocortical cancers with the frequent 17p13 loss of heterozygosity. Clin. Cancer Res. 2007, 13, 844-850. [CrossRef]

15. Zheng, S .; Cherniack, A.D .; Dewal, N .; Moffitt, R.A .; Danilova, L .; Murray, B.A .; Lerario, A.M .; Else, T .; Knijnenburg, T.A .; Ciriello, G .; et al. Comprehensive pan-genomic characterization of adrenocortical carcinoma. Cancer Cell 2016, 29, 723-736. [CrossRef]

16. Li, W .; Liu, B .; Wang, W .; Sun, C .; Che, J .; Yuan, X .; Zhai, C. Lung cancer stage prediction using multi-omics data. Comput. Math. Methods Med. 2022, 2022, 2279044. [CrossRef]

17. Wishart, D. Metabolomics and the multi-omics view of cancer. Metabolites 2022, 12, 154. [CrossRef]

18. Ding, M.Q .; Chen, L .; Cooper, G.F .; Young, J.D .; Lu, X. Precision oncology beyond targeted therapy: Combining omics data with machine learning matches the majority of cancer cells to effective therapeutics. Mol. Cancer Res. 2018, 16, 269-278. [CrossRef]

19. Francescatto, M .; Chierici, M .; Rezvan Dezfooli, S .; Zandona, A .; Jurman, G .; Furlanello, C. Multi-omics integration for neuroblas- toma clinical endpoint prediction. Biol. Direct. 2018, 13, 5. [CrossRef]

20. Meng, J .; Guan, Y .; Wang, B .; Chen, L .; Chen, J .; Zhang, M .; Liang, C. Risk subtyping and prognostic assessment of prostate cancer based on consensus genes. Commun. Biol. 2022, 5, 233. [CrossRef]

21. Demeure, M.J .; Coan, K.E .; Grant, C.S .; Komorowski, R.A .; Stephan, E .; Sinari, S .; Mount, D .; Bussey, K.J. Pttg1 overexpression in adrenocortical cancer is associated with poor survival and represents a potential therapeutic target. Surgery 2013, 154, 1405-1416, discussion 16. [CrossRef] [PubMed]

22. Legendre, C.R .; Demeure, M.J .; Whitsett, T .; Gooden, G.C .; Bussey, K.J .; Jung, S .; Waibhav, T .; Kim, S .; Salhia, B. Pathway implications of aberrant global methylation in adrenocortical cancer. PLoS ONE 2016, 11, e0150629. [CrossRef] [PubMed]

23. Heaton, J.H .; Wood, M.A .; Kim, A.C .; Lima, L.O .; Barlaskar, F.M .; Almeida, M.Q .; Fragoso, M.C .; Kuick, R .; Lerario, A.M .; Simon, D.P .; et al. Progression to adrenocortical tumorigenesis in mice and humans through insulin-like growth factor 2 and beta-catenin. Am. J. Pathol. 2012, 181, 1017-1033. [CrossRef] [PubMed]

24. 4. Assié, G .; Letouzé, E .; Fassnacht, M .; Jouinot, A .; Luscap, W .; Barreau, O .; Omeiri, H .; Rodriguez, S .; Perlemoine, K .; René-Corail, F .; et al. Integrated genomic characterization of adrenocortical carcinoma. Nat. Genet. 2014, 46, 607-612. [CrossRef]

25. Lu, X .; Meng, J .; Zhou, Y .; Jiang, L .; Yan, F. Movics: An R Package for multi-omics integration and visualization in cancer subtyping. Bioinformatics 2020, 36, 5539-5541. [CrossRef]

26. Chalise, P .; Fridley, B.L. Integrative clustering of multi-level omic data based on non-negative matrix factorization algorithm. PLoS ONE 2017, 12, e0176278. [CrossRef]

27. Hastie, T .; Tibshirani, R .; Walther, G. Estimating the number of data clusters via the gap statistic. J. R. Stat. Soc. B 2001, 63, 411-423.

28. Rooney, M.S .; Shukla, S.A .; Wu, C.J .; Getz, G .; Hacohen, N. Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell 2015, 160, 48-61. [CrossRef]

29. Liberzon, A .; Birger, C .; Thorvaldsdottir, H .; Ghandi, M .; Mesirov, J.P .; Tamayo, P. The molecular signatures database (Msigdb) hallmark gene set collection. Cell Syst. 2015, 1, 417-425. [CrossRef]

30. Meng, J .; Lu, X .; Zhou, Y .; Zhang, M .; Ge, Q .; Zhou, J .; Hao, Z .; Gao, S .; Yan, F .; Liang, C. Tumor immune microenvironment-based classifications of bladder cancer for enhancing the response rate of immunotherapy. Mol. Ther. Oncolytics 2021, 20, 410-421. [CrossRef]

31. Yoshihara, K .; Shahmoradgoli, M .; Martínez, E .; Vegesna, R .; Kim, H .; Torres-Garcia, W .; Trevino, V .; Shen, H .; Laird, P.W .; Levine, D.A .; et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat. Commun. 2013, 4, 2612. [CrossRef] [PubMed]

32. Possemato, R .; Marks, K.M .; Shaul, Y.D .; Pacold, M.E .; Kim, D .; Birsoy, K .; Sethumadhavan, S .; Woo, H .- K .; Jang, H.G .; Jha, A.K .; et al. Functional genomics reveal that the serine synthesis pathway is essential in breast cancer. Nature 2011, 476, 346-350. [CrossRef] [PubMed]

33. Chen, P .- L .; Roh, W .; Reuben, A .; Cooper, Z.A .; Spencer, C.N .; Prieto, P.A .; Miller, J.P .; Bassett, R.L .; Gopalakrishnan, V .; Wani, K .; et al. Analysis of immune signatures in longitudinal tumor samples yields insight into biomarkers of response and mechanisms of resistance to immune checkpoint blockade. Cancer Discov. 2016, 6, 827-837. [CrossRef] [PubMed]

34. Hoshida, Y .; Brunet, J.P .; Tamayo, P .; Golub, T.R .; Mesirov, J.P. Subclass mapping: Identifying common subtypes in independent disease data sets. PLoS ONE 2007, 2, e1195. [CrossRef] [PubMed]

35. Hoshida, Y. Nearest template prediction: A single-sample-based flexible class prediction with confidence assessment. PLoS ONE 2010, 5, e15543. [CrossRef]

36. Schneider, B .; Fukunaga, A .; Murry, D .; Yoder, C .; Fife, K .; Foster, A .; Rosenberg, L .; Kelich, S .; Li, L .; Sweeney, C. A Phase I, pharmacokinetic and pharmacodynamic dose escalation trial of weekly paclitaxel with interferon-alpha2b in patients with solid tumors. Cancer Chemother. Pharmacol. 2007, 59, 261-268. [CrossRef]

37. Yu, X .; Zhu, D .; Luo, B .; Kou, W .; Cheng, Y .; Zhu, Y. Ifngamma enhances ferroptosis by increasing Jakstat pathway activation to suppress Slca711 expression in adrenocortical carcinoma. Oncol. Rep. 2022, 47, 97. [CrossRef]

38. Hermsen, I.G.C .; Haak, H.R .; De Krijger, R.R .; Kerkhofs, T.M.A .; Feelders, R.A .; De Herder, W.W .; Wilmink, H .; Smit, J.W.A .; Gelderblom, H .; de Miranda, N .; et al. Mutational analyses of epidermal growth factor receptor and downstream pathways in adrenocortical carcinoma. Eur. J. Endocrinol. 2013, 169, 51-58. [CrossRef]

39. Tömböl, Z .; Szabó, P.M .; Molnár, V .; Wiener, Z .; Tölgyesi, G .; Horányi, J .; Riesz, P .; Reismann, P .; Patócs, A .; Likó, I .; et al. Integrative molecular bioinformatics study of human adrenocortical tumors: Microrna, tissue-specific target prediction, and pathway analysis. Endocr. Relat. Cancer 2009, 16, 895-906. [CrossRef]

40. Terzolo, M .; Zaggia, B .; Allasino, B .; De Francia, S. Practical treatment using mitotane for adrenocortical carcinoma. Curr. Opin. Endocrinol. Diabetes Obes. 2014, 21, 159-165. [CrossRef]

41. Suzuki, S .; Minamidate, T .; Shiga, A .; Ruike, Y .; Ishiwata, K .; Naito, K .; Ishida, A .; Deguchi, H .; Fujimoto, M .; Koide, H .; et al. Steroid metabolites for diagnosing and predicting clinicopathological features in cortisol-producing adrenocortical carcinoma. BMC Endocr. Disord. 2020, 20, 173. [CrossRef]

42. Liang, F .; Kapoun, A.M .; Lam, A .; Damm, D.L .; Quan, D .; O’Connell, M .; Protter, A.A. B-type natriuretic peptide inhibited angiotensin Ii-stimulated cholesterol biosynthesis, cholesterol transfer, and steroidogenesis in primary human adrenocortical cells. Endocrinology 2007, 148, 3722-3729. [CrossRef] [PubMed]

43. Lavoie, J.M .; Csizmok, V .; Williamson, L.M .; Culibrk, L .; Wang, G .; Marra, M.A .; Laskin, J .; Jones, S.J.M .; Renouf, D.J .; Koll- mannsberger, C.K. Whole-genome and transcriptome analysis of advanced adrenocortical cancer highlights multiple alterations affecting epigenome and DNA repair pathways. Cold Spring Harb. Mol. Case Stud. 2022, 8, a006148. [CrossRef] [PubMed]

44. Meng, J .; Zhou, Y .; Lu, X .; Bian, Z .; Chen, Y .; Zhou, J .; Zhang, L .; Hao, Z .; Zhang, M .; Liang, C. Immune response drives outcomes in prostate cancer: Implications for immunotherapy. Mol. Oncol. 2021, 15, 1358-1375. [CrossRef]

45. Zhang, X .; Lan, Y .; Xu, J .; Quan, F .; Zhao, E .; Deng, C .; Luo, T .; Xu, L .; Liao, G .; Yan, M .; et al. Cellmarker: A manually curated resource of cell markers in human and mouse. Nucleic Acids Res. 2019, 47, D721-D728. [CrossRef] [PubMed]

46. Iwai, Y .; Ishida, M .; Tanaka, Y .; Okazaki, T .; Honjo, T .; Minato, N. Involvement of Pd-L1 on tumor cells in the escape from host immune system and tumor immunotherapy by Pd-L1 blockade. Proc. Natl. Acad. Sci. USA 2002, 99, 12293-12297. [CrossRef] [PubMed]

47. Liu, Y.T .; Sun, Z.J. Turning cold tumors into hot tumors by improving T-cell infiltration. Theranostics 2021, 11, 5365-5386. [CrossRef]

48. Fassnacht, M .; Dekkers, O.M .; Else, T .; Baudin, E .; Berruti, A .; de Krijger, R.R .; Haak, H.R .; Mihai, R .; Assie, G .; Terzolo, M. European society of endocrinology clinical practice guidelines on the management of adrenocortical carcinoma in adults, in collaboration with the European Network for the study of adrenal tumors. Eur. J. Endocrinol. 2018, 179, G1-G46. [CrossRef] [PubMed]

49. Raymond, V.M .; Everett, J.N .; Furtado, L.V .; Gustafson, S.L .; Jungbluth, C.R .; Gruber, S.B .; Hammer, G.D .; Stoffel, E.M .; Greenson, J.K .; Giordano, T.J .; et al. Adrenocortical carcinoma is a lynch syndrome-associated cancer. J. Clin. Oncol. 2013, 31, 3012-3018. [CrossRef] [PubMed]

50. Wasserman, J.D .; Novokmet, A .; Eichler-Jonsson, C .; Ribeiro, R.C .; Rodriguez-Galindo, C .; Zambetti, G.P .; Malkin, D. Prevalence and functional consequence of Tp53 mutations in pediatric adrenocortical carcinoma: A children’s oncology group study. J. Clin. Oncol. 2015, 33, 602-609. [CrossRef]

51. Raymond, V.M .; Else, T .; Everett, J.N .; Long, J.M .; Gruber, S.B .; Hammer, G.D. Prevalence of germline Tp53 mutations in a prospective series of unselected patients with adrenocortical carcinoma. J. Clin. Endocrinol. Metab. 2013, 98, E119-E125. [CrossRef] [PubMed]

52. Ripley, R.T .; Kemp, C.D .; Davis, J.L .; Langan, R.C .; Royal, R.E .; Libutti, S.K .; Steinberg, S.M .; Wood, B .; Kammula, U.S .; Fojo, T .; et al. Liver resection and ablation for metastatic adrenocortical carcinoma. Ann. Surg. Oncol. 2011, 18, 1972-1979. [CrossRef] [PubMed]

53. Fassnacht, M .; Libe, R .; Kroiss, M .; Allolio, B. Adrenocortical carcinoma: A clinician’s update. Nat. Rev. Endocrinol. 2011, 7, 323-335. [CrossRef] [PubMed]

54. Stigliano, A .; Chiodini, I .; Giordano, R .; Faggiano, A .; Canu, L .; Della Casa, S .; Loli, P .; Luconi, M .; Mantero, F .; Terzolo, M. Management of adrenocortical carcinoma: A consensus statement of the Italian society of endocrinology (Sie). J. Endocrinol. Investig. 2016, 39, 103-121. [CrossRef]

55. Geraghty, M.T .; Kearns, W.G .; Pearson, P.L .; Valle, D. Isolation and characterization of an ornithine aminotransferase-related sequence (Oat13) mapping to 10q26. Genomics 1993, 17, 510-513. [CrossRef] [PubMed]

56. Hanahan, D .; Weinberg, R.A. Hallmarks of cancer: The next generation. Cell 2011, 144, 646-674. [CrossRef]

57. Gajewski, T.F .; Schreiber, H .; Fu, Y.X. Innate and adaptive immune cells in the tumor microenvironment. Nat. Immunol. 2013, 14, 1014-1022. [CrossRef] [PubMed]

58. Sato, E .; Olson, S.H .; Ahn, J .; Bundy, B .; Nishikawa, H .; Qian, F .; Jungbluth, A.A .; Frosina, D .; Gnjatic, S .; Ambrosone, C .; et al. Intraepithelial Cd8+ tumor-infiltrating lymphocytes and a high Cd8+/regulatory T cell ratio are associated with favorable prognosis in ovarian cancer. Proc. Natl. Acad. Sci. USA 2005, 102, 18538-18543. [CrossRef] [PubMed]

59. Zhang, L .; Conejo-Garcia, J.R .; Katsaros, D .; Gimotty, P.A .; Massobrio, M .; Regnani, G .; Makrigiannakis, A .; Gray, H .; Schlienger, K .; Liebman, M.N .; et al. Intratumoral T cells, recurrence, and survival in epithelial ovarian cancer. N. Engl. J. Med. 2003, 348, 203-213. [CrossRef] [PubMed]

60. Hwang, W.T .; Adams, S.F .; Tahirovic, E .; Hagemann, I.S .; Coukos, G. Prognostic significance of tumor-infiltrating T cells in ovarian cancer: A meta-analysis. Gynecol. Oncol. 2012, 124, 192-198. [CrossRef] [PubMed]

61. Huang, R.Y .; Francois, A .; McGray, A.R .; Miliotto, A .; Odunsi, K. Compensatory upregulation of Pd-1, Lag-3, and Ctla-4 limits the efficacy of single-agent checkpoint blockade in metastatic ovarian cancer. Oncoimmunology 2017, 6, e1249561. [CrossRef] [PubMed]

62. Jouinot, A .; Assié, G .; Libe, R .; Fassnacht, M .; Papathomas, T .; Barreau, O .; De La Villeon, B .; Faillot, S .; Hamzaoui, N .; Neou, M .; et al. DNA methylation is an independent prognostic marker of survival in adrenocortical cancer. J. Clin. Endocrinol. Metab. 2017, 102, 923-932. [CrossRef] [PubMed]

63. Moore, L.D .; Le, T .; Fan, G. DNA methylation and its basic function. Neuropsychopharmacology 2013, 38, 23-38. [CrossRef]

64. Ho, A.S .; Kannan, K .; Roy, D.M .; Morris, L.G.T .; Ganly, I .; Katabi, N .; Ramaswami, D .; Walsh, L .; Eng, S .; Huse, J.T .; et al. The mutational landscape of adenoid cystic carcinoma. Nat. Genet. 2013, 45, 791-798. [CrossRef] [PubMed]

65. Özata, D.M .; Caramuta, S .; Velázquez-Fernández, D .; Akçakaya, P .; Xie, H .; Höög, A .; Zedenius, J .; Bäckdahl, M .; Larsson, C .; Lui, W .- O. The role of microrna deregulation in the pathogenesis of adrenocortical carcinoma. Endocr. Relat. Cancer 2011, 18, 643-655. [CrossRef] [PubMed]

66. Juhlin, C.C .; Goh, G .; Healy, J .; Fonseca, A.L .; Scholl, U.I .; Stenman, A .; Kunstman, J .; Brown, T.C .; Overton, J.D .; Mane, S.M .; et al. Whole-exome sequencing characterizes the landscape of somatic mutations and copy number alterations in adrenocortical carcinoma. J. Clin. Endocrinol. Metab. 2015, 100, E493-E502. [CrossRef]

67. Zhang, C .; Qu, Y .; Xiao, H .; Xiao, W .; Liu, J .; Gao, Y .; Li, M .; Liu, J. Lncrna Snhg3 promotes clear cell renal cell carcinoma proliferation and migration by upregulating Top2a. Exp. Cell Res. 2019, 384, 111595. [CrossRef]

68. Li, T .; Xing, Y .; Yang, F .; Sun, Y .; Zhang, S .; Wang, Q .; Zhang, W. Lncrna Snhg3 sponges Mir-577 to up-regulate smurf1 expression in prostate cancer. Cancer Med. 2020, 9, 3852-3862. [CrossRef] [PubMed]

69. Achermann, J.C .; Vilain, E.J. Nr0b1-Related Adrenal Hypoplasia Congenita. In Genereviews®; Adam, M.P., Everman, D.B., Mirzaa, G.M., Pagon, R.A., Wallace, S.E., Bean, L.J.H., Gripp, K.W., Amemiya, A., Eds .; University of Washington: Seattle, WA, USA, 1993.

70. Larsen, S.J .; Rottger, R .; Schmidt, H .; Baumbach, J.E. Coli gene regulatory networks are inconsistent with gene expression data. Nucleic Acids Res. 2019, 47, 85-92. [CrossRef] [PubMed]

71. Rana, P .; Thai, P .; Dinh, T .; Ghosh, P. Relevant and non-redundant feature selection for cancer classification and subtype detection. Cancers 2021, 13, 4297. [CrossRef] [PubMed]