ELSEVIER

Contents lists available at ScienceDirect

Pathology - Research and Practice

journal homepage: www.elsevier.com/locate/prp

PATHOLOGY RESEARCH PRACTICE

Review

Check for updates

Molecular landscape and biomarker discovery in adrenocortical carcinoma: An integrative review of bioinformatics and translational insights

Javad Omidi

Department of Chemical Engineering, Columbia University, New York, NY 10027, USA

ARTICLE INFOABSTRACT
Keywords: Adrenocortical carcinoma Biomarkers Transcriptomics Drug repositioning Systems biologyAdrenocortical carcinoma (ACC) is a rare and aggressive endocrine malignancy with limited therapeutic options and poor prognosis. Recent advances in high-throughput sequencing and integrative bioinformatics have unraveled the complex molecular landscape of ACC, highlighting critical genomic, epigenomic, transcriptomic, and immune-related alterations. This review synthesizes current evidence to provide a comprehensive overview of the key molecular mechanisms driving ACC pathogenesis. The role of recurrent mutations (e.g., TP53, CTNNB1), dysregulated cell cycle genes (e.g., CDK1, CCNB1, AURKA), non-coding RNAs, and epigenetic mod- ifications in shaping tumor behavior is discussed. Multi-omics integration and systems biology approaches have enabled the identification of robust prognostic gene signatures and protein biomarkers, offering novel tools for risk stratification. Furthermore, the tumor immune microenvironment is examined, with hypoxia, immune suppression, and checkpoint pathways highlighted as emerging targets. Finally, computational drug reposi- tioning strategies that nominate repurposed agents such as IGF1R inhibitors and BCLAF1 modulators for ther- apeutic intervention are explored. Together, these insights pave the way for precision oncology in ACC, while emphasizing the need for rigorous multi-layered validation and standardized clinical integration to enable real- world translational impact.

1. Introduction

Adrenocortical carcinoma (ACC) is a rare but highly aggressive malignancy originating from the adrenal cortex, with an estimated incidence of 0.7-2.0 cases per million annually worldwide [1]. Despite being uncommon, ACC carries a disproportionately high mortality rate, with 5-year survival rates ranging from 10 % to 47 % depending on stage at diagnosis [2].

Due to its rarity, delayed diagnosis, and limited treatment options, ACC poses significant challenges for clinicians and researchers alike. Surgical resection remains the only potentially curative treatment, yet recurrence rates remain high even in patients undergoing complete resection [3]. Systemic therapies such as mitotane or platinum-based chemotherapy offer limited benefit, and there is a pressing need for improved diagnostic, prognostic, and therapeutic tools [4].

In the past two decades, high-throughput sequencing and integrative bioinformatics have improved our understanding of ACC’s molecular landscape. Several studies have explored genetic and epigenetic alter- ations, transcriptomic signatures, and immune-related features, identi- fying key genes and pathways such as TP53, IGF2, Wnt/ß-catenin, and

the cell cycle machinery [5-7]. Notably, integrative analyses have revealed distinct gene expression profiles distinguishing ACC from benign adrenal tumors, enabling the identification of diagnostic and prognostic biomarkers [8,9]. Moreover, multi-omics studies and pro- tein-protein interaction (PPI) networks have facilitated the recognition of hub genes with prognostic value, including CDK1, CCNB1, TOP2A, BUB1B, and CEP55 [2].

The tumor microenvironment (TME) in ACC has also gained atten- tion, particularly in relation to immune infiltration, hypoxia, and im- mune checkpoint molecules such as PD-L1 and CTLA-4. These insights offer potential avenues for immunotherapy and personalized medicine [4]. Additionally, recent efforts in computational drug repositioning have identified candidate therapeutics such as Cosyntropin and IGF1R inhibitors as potential treatments for ACC, circumventing the limitations of de novo drug discovery [10].

This review synthesizes recent bioinformatics-driven studies on ACC, highlighting key molecular mechanisms, biomarkers, immune landscape features, and therapeutic opportunities. It also identifies major knowl- edge gaps and proposes future research directions. By systematically integrating and contrasting findings across a broad range of peer-

reviewed studies [1-39], this review establishes a unified and up-to-date framework for understanding the molecular and translational landscape of ACC, something not previously available in the literature. A graphical overview of the review framework and thematic structure is presented in Fig. 1, which visually maps the key molecular and clinical domains addressed in the subsequent sections.

2. Molecular pathogenesis of ACC

2.1. Genetic and epigenetic alterations

ACC displays a heterogeneous spectrum of genetic and epigenetic alterations contributing to its aggressive clinical course. One of the most recurrent genetic events involves TP53, frequently mutated in pediatric cases and linked to Li-Fraumeni syndrome in adults [5]. The oncogenic role of AURKA in destabilizing TP53 has been further highlighted as a targetable vulnerability in ACC. Recent integrative analyses also revealed a survival-associated proliferative gene module in ACC, in which AURKA, POLD1, RECQL4, CDC7, FANCI, and DTL form a TP53-centered protein interaction network. This mitotic control hub cooperates with TP53 dysfunction to drive tumor aggressiveness (Fig. 2a).

Alterations in the Wnt/ß-catenin pathway are common, particularly CTNNB1 mutations and ZNRF3 deletions. These changes promote tumor proliferation and invasiveness [5]. Additionally, overexpression of IGF2 and its embedded miR-483, along with suppression of H19/miR-675, are repeatedly observed, reinforcing the role of epigenetically modulated growth signaling. Consistent with this, Giordano et al. reported pro- nounced IGF2 overexpression in 10 of 11 ACC tumors, underscoring IGF2 as a dominant growth driver in ACC pathogenesis [9].

Emerging genomic analyses have revealed novel mutated genes such as KIF23, POLD1, and TPX2, several of which exhibit interactions with TP53 [11]. The functional importance of kinesin family genes in ACC progression and immune evasion mechanisms is now being recognized.

At the epigenetic level, DNA methylation is a major contributor to transcriptional deregulation in ACC. Guan et al. [12] categorized tumors into three methylation-based clusters with distinct survival outcomes (Fig. 2b). These subtypes demonstrate different enrichment in immune and metabolic pathways and diverge in DNA repair and mismatch repair

Fig. 1. Overview of the review structure, outlining key molecular and trans- lational aspects of ACC from pathogenesis to future directions.

Multi-Omics Integration and Systems Biology in ACC

1

Molecular Pathogenesis of ACC

6

o Genetic and Epigenetic Alterations

o Transcriptomic Signatures

2

Therapeutic Targets &

ADRENCORTICAL CARCINOMA (ACC)

Prognostic & Predictive Biomarkers

Drug Repositioning

o Gene-based Signatures o Protein Markers and Immunohistochemistry

5

Pediatric vs. Adult ACC: Comparative Genomics

Tumor Microenvironment & Immune Landscape

3

4

profiles.

Long non-coding RNAs (lncRNAs) play central roles in the epigenetic regulation of ACC. For instance, LINC00271 is significantly down- regulated in ACC and associated with chromosome segregation and Wnt signaling dysregulation [13]. Moreover, PAX8-AS1, KCNQ1OT1, and NEAT1 participate in competitive endogenous RNA networks, particu- larly through the BIRC5-miR-335-5p-PAX8-AS1 axis strongly corre- lated with poor prognosis [14]. MicroRNAs (miRNAs) also represent critical regulatory elements in ACC. Tömböl et al. [15] identified dif- ferential expression of miR-184, miR-503, and miR-511 distinguishing ACC from benign tumors. These miRs influence pathways such as G2/M checkpoint damage, indicating potential biomarker roles.

Transcriptomic profiling by Di Dalmazi et al. [16] identified novel fusion transcripts including EXOSC10-MTOR and AKAP13-PDE8A, which may impact cAMP and mTOR signaling. Such gene fusions pro- vide further insight into ACC heterogeneity and therapeutic resistance (Fig. 2c). Epigenetic modifiers such as DNMT3B, EZH2, and UHRF1 were found to be overexpressed in ACC compared to normal adrenal glands. These regulators influence chromatin remodeling and DNA methylation, underlining their oncogenic potential [7]. Telomere maintenance mechanisms (TMMs) have also emerged as relevant in ACC, where ALT (alternative lengthening of telomeres) and telomerase activation coexist, correlating with poor prognosis in specific subgroups [17]. Finally, dysregulation of lipid metabolism genes, particularly within sphingolipid and steroid synthesis pathways, has been linked to progression and worse prognosis in ACC, suggesting a potential meta- bolic vulnerability [18].

The molecular pathogenesis of ACC is strongly influenced by both genetic and epigenetic alterations, which drive tumor initiation, pro- gression, and metastatic potential. Several recent studies have employed integrative bioinformatics approaches to uncover these aberrations, of- fering valuable insights into their prognostic and therapeutic implications.

Recent mechanistic studies further demonstrate that non-coding RNA-mediated regulation plays a central role in shaping tumor signaling, immune evasion, and therapy resistance in ACC. For instance, evidence from glioma and other cancer models has shown that circRNAs modulate treatment response through Ras/Raf/ERK and PI3K/AKT cascades, influencing proliferation and resistance phenotypes [45,46]. Long non-coding RNAs similarly regulate oncogenic pathways such as Wnt, STAT3, and EZH2, driving cell cycle progression and metastatic behavior [47]. Moreover, natural antisense transcripts exert transcrip- tional and epigenetic control across multiple cancer settings, providing additional layers of regulatory complexity [48]. Key oncogenic signaling modules, particularly the PI3K-AKT-mTOR axis, have emerged as convergence points linking non-coding RNA activity with metabolic reprogramming and proliferative signaling [49-51]. Collectively, these mechanistic insights highlight ncRNA-centered regulatory networks as critical determinants of tumor behavior and suggest that similar RNA-mediated signaling architectures may drive pathogenesis, hetero- geneity, and therapeutic vulnerabilities in ACC [52].

One of the most significant epigenetic modifications implicated in ACC is N6-methyladenosine (m6A) RNA methylation, which represents the most prevalent internal modification of mRNA and plays a critical role in regulating mRNA stability, splicing, and translation. Xu et al. [19] comprehensively profiled m6A regulators in ACC and identified a prognostic signature composed of RBM15 and HNRNPC, two key genes involved in the m6A pathway. Notably, HNRNPC promotes prolifera- tion, migration, and invasion in ACC cells, supporting its oncogenic function. Moreover, patients with high m6A risk scores exhibited sup- pressed immune-related pathways and reduced expression of immune checkpoint molecules such as PD-L1 and CTLA4, indicating a link be- tween epigenetic regulation and immune evasion in ACC progression.

Another key regulatory mechanism involved in ACC is alternative splicing (AS). Xu et al. [20] conducted a genome-wide analysis of AS events using TCGA SpliceSeq data and revealed 3919 events

Fig. 2. Key genetic and epigenetic alterations in ACC. (a) Recurrent somatic and germline mutations in major driver genes including TP53, CTNNB1, and ZNRF3, as well as emerging candidates such as KIF23 and AURKA, which have been associated with tumor progression and patient prognosis [5]. (b) Epigenetic dysregulation in ACC, encompassing DNA methylation-based subtypes (ACC-C1 to ACC-C3) with distinct clinical outcomes, and chromatin/telomere alterations including over- expression of DNMT3B, EZH2, and activation of alternative lengthening of telomeres (ALT) [12]. (c) Discovery of fusion transcripts such as EXOSC10-MTOR and AKAP13-PDE8A, contributing to aberrant signaling pathways implicated in ACC pathogenesis [16]. (d) Transcriptional regulation of cell cycle by Bclaf1, which enhances the expression of CDK1 and Cyclin B1, facilitating G2/M cell cycle transition and promoting ACC cell proliferation [3].

Enrichment Score

Subtype

Enrichment Score

1

Subtype Laterality

1

TACC3

NEK2

Laterality

0.5

Gender

0

Gender

0.5

0

a

(b)

Age

-

Age

-0.5

Stage

-0.5

Stage

-1

N-Glycan Biosynthesis

-1

APICAL JUNCTION

Cher Types of OGlycan Biosynthesis

RRM2

EPITHELIAL MESENCHYMAL_TRANSITION

MYOGENESIS

Glycosphingolipid Biosynthesis

AURKA

APICAL SURFACE

Glycogen Biosynthesis

Hexcsammine Biosynthesis

HELLS

INTERFERON_ALPHA_RESPONSE

Edekaphase Metabolism

HAYANATUA NO RESPONSE

Shingolipid Metabolism

19 STATS SIGNALING

Fattone

ALLOGRAFT REJECTION

Retinoid Metabolism

16 JAK STATS SIGNALING

Mucin Type O Glycan Biosynthesis

Glycogen Degradation

INFLAMMATORY_RESPONSE

Vitamin k

EXO

FANCI

Glycosaminoglycan Biosynthesis

THEA SAMMA

INFA SIGNALING_VIA_NFKB

Bon Metabolism

BILE ACID METABOLISM

Lipoic Acid Metabolism

DPYD

OXIDATIVE PHOSPHORYLATION

Polyamine Biosynthese

DTL

PEROKISOME ADIPOGENESIS

-Glutamine and D-Glutamate Metabolism Thiamine Metabolism

FATTY ACID METABOLISM

Transsuturation

TP53

KRAS SIGNALING ON

Taurine and Hypotaurine Metabolism

ESTROGEN RESPONSE EARLY

Cartoon boende Metabolism

RESPONSE- TATE

XENOBIOTIC METABOLISM

Other Glycan Degradation

FANCD2

REACTIVE_OKYGEN_SPECIES_PATHWAY

Porphyrin and Chlorophyll Metabolism

GLYCOLYSIS

Peritose and Glucuronste Interconversions

UV-RESPOROT HOMEOSTASIS

Ascorbate and Aldrate Metabolism

WTORCT SIGNALING

SPERMATOGENESIS

Brug Metabolism by other enzymes

G2M CHECKPOINT

Nicotinamide Adenine Dinucleotide Biosynthesis

Nicotinate and Nicotinamide Metabolism

Metabolism of Xenobiotics by Cytochrome P450

CDC7

RFC5

UNFOLDED PROTEIN_RESPONSE

ONA REPAR

Metabolism

MYC TARGETS V2

Steroid Hormone Biosynthesis

WNT BETA_CATENIN_SIGNALING HEDGEHOG SIGNALING

Phenylalanine Metabolism

NOTCH SIGNALING

Cardiolipin Metabolism

RECQL4

TGF BETA SIGNALING

Codative Phosphorylation

Nitrogen Metabo st Metabolism

POLD1

POK AKT MTOR SIGNALING

ANDROGEN RESPONSE

Pantothenate and CoA Biosynthesis

(c)

Dopamine Biosynthesis

PROTEIN SECRETION

HEME_METABOLISM

Epinephrine Bicisynthesis

Norepinephone Biosynthesis

Samples in TCGA-ACC cohort

AKAP13-PDE8A

Enrichment Score

her Led Megom

Subtype Laterality Gender

1

alpha- Linoleie Ack Metabolism

Cyclooxygenase Arachidonic Acid Metabolism

0.5

Štostanoid Biosynthesis

Neomycin, Kanimysin and Gentamicin Biosynthesis

40

0

TGATACATTGGAGACCATTGCTCCTGAAAACCTCAACAAAGACTAAGCC

50

60

70

90

Stage

WP_ONA_MISMATCH_REPAIR

-0.5

Starch and Suclose Metabolism

Alachiconic Add Metabolism

REACTOME_MISMATCH_REPAIR

-1

Retinoic Acid Metabolism

Prostaglandin Biosynthesis

KEGG_MIŞMATCH_REPAIR

Glutathione Metabolism

KEGG_P53_SIGNALING_PATHWAY

Fructose and Mannose Metabolism

gluconeogenesis

BIOCARTA_RB_PATHWAY

BIOCARTA_P53_PATHWAY

Allinine, Aspartate and Glutamate Metabolism

ST_WNT_BETA_CATENIN_PATHWAY KEGG_WNT_SIGNALING_PATHWWWY

Urea Cycle

Arginine Bigsynthesis

Fatty Acid Elongation

WNT_SIGNALING

sosynthesis of Unsaturated Faty Acids

GO_CHROMATIN_REMODELING

Valine, Leucine and Isoleucine Degradation

REACTOME_CHROMATIN_MODIFYING_ENZYMES

Citric Acid Cycle

GO_NOTCH_SIGNALING_PATHWAY

Pycocylate and Dicarboxylate Metabolism

Nicoanamide Adenine Me-poism

chromosome 15

chromosome 15

L

REACTOME_SIGNALING_BY_NOTCH

PID_NOTCH_PATHWAY

Ketone Biosynthesis and Metabolism

q25.3

925.3

Samples in TOGA-ACC cohort

Pyruvate

Sulfur Metabolism

Selenocompound Metabolism

Vitamin B6 Metabolism

breakpoint

breakpoint 15.85526367

Subtype

Stage

Glycine, Serine and Threonine Metabolism

15.86064806

pane, Leucine and Isoleucine Biosynthesis

Pentose Phosphate

Coverage

ACC1

Ubiquinone and other Terpenoid-Quinone Biosynthesis

ACC2

Stage I

atty Acid Biosynthesis

Pherylalanine Tyrosine and Tryptophan Biosynthesis

0-

ACC3

Stage II

?

2

Stage III

Stage IV

Arginine and Proline Metabolism Esdine Metabolism

AKCAP13

PRESA

Age

unknown

Tryptophan Metabolism

80

Homocysteine Biosynthesis Merthion ine Cype

ENST00000561343-2

ENST00000310296.4

olate One Carbon Metabolism

60

Pyrimidine Biosynthesis

Purine Metabolism

AGTTCTGATACATTGGAGACCATTGCTOCTGAAAACCTCAACAAAGACTAAGCCCAGAAAC

5. kbp

40

Laterality

Gender

Pyrimidine Metabolism

introns not to scale

erpenod Backbone Biosynthesis

20

Left

FEMALE

herold Bios ynthesis

0

Right

MALE

Aldosterone Biosynthesis

Cortisol Biosynthesis stradiol Blosynthesis

Samples in TCGA-ACC cohort

Testosterone Biosynthesis

(d)

si NC si Bclaf1

EV

FL

si NC

si Bclaf1

EV

FL

1.5

2.0

*

*

*

*

*

145kDa

Bclaf1

Relative protein levels

*

Relative protein levels

1.0

1.5

34kDa

CDK1

1.0

58kDa

Cyclin B1

0.5

0.5

ß-actin

0.0

Bclaf1

CDK1

Cyclin B1

0.0

Bclaf1

CDK1

Cyclin B1

NCI-H295R

si NC si Bclaf1

sh NC sh Bclaf1

si NC

si Bclaf1

sh NC

sh Bclaf1

145kDa

Bclaf1

Relative protein levels

1.5

**

*

*

Relative protein levels

1.5

**

*

*

34kDa

CDK1

1.0

1.0

T

T

58kDa

Cyclin B1

₡0.5

0.5

ß-actin

₾0.0

CDK1

0.0

Bclaf1

Cyclin B1

Bclaf1

CDK1

Cyclin B1

SW-13

significantly associated with overall survival. Among these, exon skip- ping (ES) was the most prevalent, while mutually exclusive exons (ME) were rare. Prognostic AS signatures, especially those involving alter- native acceptor (AA) and retained intron (RI) events, showed high predictive accuracy (AUC > 0.9). A correlation network between splicing factors (SFs) and survival associated AS events identified EIF3A and HNRNPR as hub SFs, suggesting a tight interplay between tran- scriptional regulation and tumor aggressiveness.

At the genomic level, Zhou et al. [3] highlighted the oncogenic role of Bcl-2-associated transcription factor 1 (Bclaf1) in ACC. Using RNA-seq and experimental validation, the authors demonstrated that Bclaf1 upregulates CDK1 and Cyclin B1 (CCNB1), which are essential drivers of G2/M cell cycle progression. Silencing Bclaf1 significantly impaired ACC cell proliferation and altered cell cycle dynamics, positioning it as a potential therapeutic target (Fig. 2d).

Additional genetic aberrations were identified by Zhang et al. [21], who performed integrative bioinformatics analysis across multiple GEO datasets. The study identified a panel of hub genes, including MYC, MMP9, CDC20, and KIF2C associated with metastasis and poor survival. These genes are involved in key pathways such as cell cycle regulation, immune evasion, and extracellular matrix remodeling, highlighting the multifactorial nature of ACC progression at the molecular level. Com- plementing these findings, Ye et al. [22] constructed a competing endogenous RNA (ceRNA) network involving interactions among mRNAs, miRNAs, and lncRNAs. Notably, the HOX- A11-AS/miR-24-3p/CDK1 axis emerged as a key regulatory module, linking non-coding RNA-mediated control with cell cycle dysregulation.

This ceRNA sub-network suggests novel post-transcriptional mecha- nisms contributing to ACC tumorigenesis and prognosis.

2.2. Transcriptomic signatures

Multiple transcriptomic studies consistently show that mitotic dys- regulation is a key molecular feature of ACC. A landmark microarray study by Giordano et al. demonstrated that global transcriptional profiling robustly distinguishes ACC from adrenocortical adenomas and normal cortex, with marked overexpression of IGF2 and significantly elevated proliferation-associated markers such as TOP2A and Ki-67 identified as defining molecular features of malignancy [9]. While the methodologies employed, ranging from traditional Differentially expressed genes (DEG) screening to co-expression analysis, hypoxia modeling, and miRNA integration differ in scope and scale, they remarkably converge on a core proliferative signature that underpins tumor progression.

A foundational study by Xing et al. [2] reported 228 DEGs in ACC, enriched in cell cycle processes and chromosomal dynamics. Among the 14 hub genes identified, AURKA, EZH2, and RRM2 stood out for their association with poor prognosis. However, when the lens is shifted to pediatric ACC, Kulshrestha et al. [6] arrived at a nearly parallel conclusion. They pinpointed CDK1, CDC20, and BUB1B as critical hubs across both PPI and GGI networks.

Taking a different angle, Chen et al. [4] integrated hypoxia into transcriptomic stratification and identified a three-gene signature (CCNA2, COL5A1, EFNA3) that correlated with both immune

Fig. 3. Transcriptomic signatures in ACC derived from multi-study analyses. (a) Hypoxia-related gene expression heatmap from [4], showing stratification of ACC patients into high- and low-risk groups. The signature was constructed using LASSO-Cox regression and correlates with survival outcomes and immune microen- vironment features. (b) KEGG pathway enrichment analysis from Guo et al. [24], showing significant enrichment of cell cycle and DNA replication pathways among differentially expressed genes. Key regulators identified in these pathways include CDK1, TOP2A, and CCNB1. (c) UMAP visualization of mRNA expression profiles showing two transcriptionally distinct ACC clusters identified by unsupervised clustering [29]. These subgroups correspond to survival-associated molecular phe- notypes refined by machine learning. (d) Volcano plot showing differentially expressed genes (DEGs) between ACC and normal adrenal tissues in GSE10927. Upregulated (red) and downregulated (black) genes are indicated [30].

(a)

stage

3

CCNA2

EFNA3

COL5A1

15

0.093

0.65

0.28

2

0.0001

0.17

0.0015

EFNA3*

0.014

0.049

0.03

1

0.0035

0.021

0.045

0

0.065

0.0065

0.25

Gene expression

10

0.83

0.18

0.93

-1

-2

CCNA2 ***

-3

5

Stage I

Stage II

COL5A1 **

Stage III

Stage IV

Stage I

Stage II

Stage III

Stage IV

Stage I

Stage II

Stage III

Stage IV

Stage I

Stage II

Stage III

Stage IV

(b)

Cell cycle

(c)

(d)

30

Cellular senescence

3

ACC-UMAP1

ACC-UMAP2

25

Progesterone-mediated oocyte maturation

2

20

Oocyte meiosis

Count

1

GSE10927

-LogFDR

2

0

3

p53 signalling pathway

UMAPZ

15

4

-1

5

Antifolate resistance

6

-2

10

p.adjust

-3

DNA replication

5

0.04

-4

Folate biosynthesis

0.03

0.02

-8

-6

-4

-2

0

UMAP1

2

4

6

8

0

0.01

One carbon pool by folate

4

2

0

N

T

0.10

0.15

0.20

Log,FC O2

GeneRatio

suppression and proliferation (Fig. 3a). Rather than focusing on indi- vidual genes, Yuan et al. [23] applied Weighted Gene Co-expression Network Analysis (WGCNA) to identify co-expressed gene modules linked to clinical traits. Guo et al. [24] approached the question through Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrich- ment. Their analysis revealed pathways such as G2/M checkpoint regulation and DNA replication to be dominantly enriched, with central roles played by CDK1, CCNB1, and TOP2A (Fig. 3b).

The prognostic relevance of these transcriptomic alterations is underscored by Yin et al. [25], who demonstrated that overexpression of AURKA, TOP2A, and CDK1 correlated with reduced overall survival and higher tumor grade. Notably, not all transcriptomic changes reflect activation. Xu et al. [26] identified downregulated hub genes including EFEMP2, MFAP4, and CSRP1, associated with extracellular matrix or- ganization and differentiation. While their expression patterns were not strong predictors of overall survival, their silencing suggests a shift to- ward de-differentiation and tissue disintegration, offering a comple- mentary view to the proliferation-focused findings.

A novel subtype of ACC was proposed by Knott and Leidenheimer [27], who found that tumors overexpressing ABAT and GABRB2, involved in GABA metabolism, were associated with favorable prog- nosis. This metabolic signature, though outside the mitotic axis, suggests a biologically distinct subclass that may warrant different therapeutic strategies. Moreover, Tömböl et al. [15] provided a regulatory perspective by integrating mRNA and miRNA profiles. They identified miR-503 and miR-511 as miRNAs that post-transcriptionally target key mitotic regulators, including CDK1 and TOP2A. These findings point to a multi-layered regulatory framework, where both transcriptional and post-transcriptional mechanisms converge on a common proliferative program.

The studies indicate that ACC transcriptomic profiles are dominated by mitotic and cell cycle dysregulation. The recurring identification of CDK1, CCNB1, AURKA, and CCNA2 across methods, datasets, and tumor subtypes confirms their role as central drivers of tumor biology. Beyond classical DEG and pathway-based studies, several recent investigations have broadened the understanding of ACC transcriptomics through pan- cancer comparisons, multi-omics integration, and machine learning — driven biomarker discovery, highlighting previously underappreciated layers of transcriptomic complexity.

Crona et al. [28] approached ACC from a pan-cancer perspective, analyzing over 3300 tumors across 35 cancer types using TCGA and TARGET datasets. Their unsupervised clustering revealed that ACC forms a distinct and homogeneous transcriptomic cluster, aligning more closely with neuroendocrine tumors such as gliomas and neuroblas- tomas than with epithelial malignancies. These findings challenge con- ventional clinicopathological classifications and suggest a shared developmental origin may underlie transcriptomic similarities among neural crest-related tumors. Notably, one sarcomatoid ACC outlier clustered with soft tissue sarcomas, highlighting potential transcrip- tional reprogramming in histologically variant subtypes.

In contrast, Marquardt et al. [29] employed unsupervised clustering and random forest-based machine learning to reclassify ACC into survival-associated subgroups using only mRNA data (Fig. 3c). Their method reidentified the known C1A (poor prognosis) and C1B (better prognosis) clusters but further refined the subgrouping by identifying novel prognostic biomarkers, such as SOAT1 and EIF2A1, absent from traditional DEG lists. Interestingly, one cluster (UMAP1) aligned more closely with benign adrenal tumors, suggesting that transcriptomic distance from normal tissue may reflect malignancy grade and outcome more accurately than histology alone.

Building on multi-omics, Li et al. [30] integrated gene expression data from both GEO and TCGA to identify 490 DEGs (28 upregulated, 462 downregulated), with CCNB1 and NDC80 emerging as robust hub genes through multiple ranking algorithms (Fig. 3d). These genes, confirmed across KEGG, Reactome, and STRING analyses are central to mitotic spindle assembly and chromosomal segregation. While CCNB1

has been widely reported, NDC80 represents a less frequently empha- sized but potentially powerful biomarker, especially given its consistent upregulation and prognostic association in ACC. Their approach exem- plifies how cross-validation across platforms strengthens biomarker confidence.

Transcriptomic profiling studies in ACC consistently demonstrate coordinated dysregulation of proliferative and metabolic signaling programs. Omidi [31] reported that CKS2 (cell-cycle progression) and ACAT2 (lipid metabolic flux) are co-upregulated in tumors relative to normal adrenal tissue, reflecting a reprogrammed transcriptional state that supports accelerated growth and metabolic sustenance. This coor- dinated expression shift provides an early indication of a synergistic oncogenic transcriptional signature that becomes more pronounced at the biomarker level.

Complementing this, Di Dalmazi et al. [16] focused on genotype — transcriptome correlations, leveraging RNA-seq and mutational data from 59 adrenocortical tumors. Their analyses uncovered two tran- scriptomic clusters: one enriched in CTNNB1 and PRKACA mutations, associated with cortisol-producing adenomas; and another with wild-type status or GNAS mutations. Interestingly, several cases without known drivers still showed transcriptomic profiles close to ACC, sug- gesting undiscovered regulatory alterations, possibly involving lncRNAs or alternative splicing. Indeed, their study was among the first to sys- tematically investigate lncRNA and gene fusion landscapes in adrenal tumors, expanding the definition of transcriptomic dysregulation beyond coding genes.

These studies reveal that transcriptomic classification of ACC is no longer a one-dimensional exercise. Dimensionality reduction, cross- platform gene validation, integration of lncRNA and fusion data, and machine learning-based stratification have each provided new angles of insight. When taken with previous DEG and pathway-based findings, they illustrate that transcriptomic heterogeneity in ACC is both biolog- ically significant and clinically relevant and that unified models must accommodate molecular diversity ranging from mitotic activation to metabolic or neuroendocrine deviation.

Table 1 summarizes the key genes, their functional roles, supporting evidence, and prognostic value. The genes summarized in Table 1 converge on key biological themes that underlie ACC aggressiveness. A large subset, including CDK1, CCNB1, CDC20, BUB1B, TOP2A, and CKS2 reflects pervasive dysregulation of mitotic control, supporting the view that aberrant cell cycle progression is a defining hallmark of ACC. Additional regulators such as EZH2 and HNRNPC underscore the contribution of epigenetic remodeling and RNA-processing pathways. Meanwhile, alterations in ECM remodeling (COL5A1, EFEMP2), im- mune modulation (EFNA3), and metabolic reprogramming (ACAT2, SOAT1) indicate that tumor invasion and metabolic adaptation also shape disease progression. Notably, several of these genes demonstrate strong prognostic value, reinforcing their potential utility in risk strati- fication and the development of targeted therapeutic strategies in ACC.

2.3. Methodological overview of multi-omics analyses

Multi-omics analyses in ACC research often rely on system-level network approaches to contextualize gene expression patterns within broader regulatory architectures. Weighted Gene Co-expression Network Analysis (WGCNA) identifies sets of genes that exhibit coor- dinated expression across samples, thereby capturing shared regulatory control rather than isolated transcriptional changes. By transforming pairwise expression correlations into a network topology, WGCNA al- lows the detection of gene modules that can be associated with clinical traits, molecular subtypes, or disease outcomes. Hub genes within these modules represent central regulatory nodes that may serve as potential biomarkers or therapeutic targets by virtue of their influence on module- level behavior.

In parallel, ceRNA network modeling elucidates interactions among coding and non-coding transcripts based on shared miRNA binding

Table 1 Summary of key differentially expressed genes and their roles in ACC.
GeneFunctionEvidence SourcePrognostic Value
CDK1Cell cycle regulation, mitosisXing et al. (2019) [2]High
Kulshrestha et al. (2016) [6]
Guo et al. (2020) [24]
Yin et al. (2023)
[25]
CCNB1G2/M checkpoint controlKulshrestha et al. (2016) [6]High
Guo et al. (2020) [24]
Yin et al. (2023)
[25]
AURKAMitotic spindle formation, proliferationXing et al. (2019) [2]High
Yin et al. (2023) [25]
EZH2Epigenetic regulation (PRC2 complex)Xing et al. (2019) [2]Moderate
RRM2DNA synthesis and repairXing et al. (2019) [2]High
CCNA2Cell cycle, hypoxia signatureChen et al. (2021)High
component[4]
Guo et al. (2020) [24]
Yin et al. (2023) [25]
COL5A1ECM remodeling, hypoxia- associated geneChen et al. (2021) [4]Moderate
EFNA3Cell signaling, immune infiltrationChen et al. (2021) [24]Moderate
CDC20Chromosome segregation, mitosisKulshrestha et al. (2016) [6]High
BUB1BSpindle checkpoint regulationKulshrestha et al. (2016) [6]High
TOP2ADNA topology during mitosisGuo et al. (2020) [24]High
Yin et al. (2023) [25]
ABATGABA metabolism,Knott andFavorable
mitochondrial functionLeidenheimer (2020) [27](High)
GABRB2GABA receptor subunit, neuroendocrine roleKnott and Leidenheimer (2020) [27]Moderate
EFEMP2ECM structure, cell adhesionXu et al. (2020)Low (Disease-
[26]free survival)
MFAP4Microfibril-associated proteinXu et al. (2020) [26]Low
NDC80Kinetochore function, chromosomal segregationLi et al. (2022) [30]High
SOAT1Sterol O-acyltransferase activity, lipid metabolismMarquardt et al. (2021) [29]Moderate
EIF2A1Translation initiation factor, stress responseMarquardt et al. (2021) [29]Moderate
AKAP13-Fusion gene, cAMP signalingDi Dalmazi et al. (2020) [16]Unknown
PDE8Adisruption
CTNNB1Wnt/ß-catenin pathway activationDi Dalmazi et al. (2020) [16]Moderate
PRKACAProtein kinase A catalytic subunit, cAMP pathwayDi Dalmazi et al. (2020) [16]Moderate
GNASG-protein alpha subunit, cAMP pathwayDi Dalmazi et al. (2020) [16]Moderate
IGF2Growth factor signaling; drives autocrine proliferative loopsGiordano et al. (2003) [9]High
MKI67 (Ki-67)Proliferation index; reflects mitotic activity and tumor gradeGiordano et al. (2003) [9]High
CKS2Cell-cycle progression; mitotic checkpoint controlOmidi (2025) [31]High
ACAT2Lipid and fatty-acid metabolic reprogramming; supports tumor energy fluxOmidi (2025) [31]High

relationships. Within this framework, transcripts that compete for the same miRNAs can modulate each other’s abundance, forming interde- pendent regulatory axes across mRNAs, lncRNAs, and other RNA spe- cies. Incorporating expression correlations and miRNA-target interaction data enables the reconstruction of ceRNA networks that reflect post-transcriptional regulatory dynamics. These inferred net- works help identify regulatory hubs that contribute to ACC-specific transcriptional states and may underlie tumor progression or thera- peutic response.

To further contextualize biologically relevant gene sets, PPI analysis is used to map functional connectivity among candidate genes, high- lighting densely interconnected clusters and putative driver nodes. When these network features are integrated with survival modeling approaches such as Kaplan-Meier or Cox proportional-hazards (Cox) regression, specific hub genes can be evaluated for prognostic value in clinical cohorts. Subsequent pathway enrichment using Gene Ontology (GO) and KEGG frameworks enables functional interpretation of these gene clusters, linking molecular signatures to defined cellular pathways and oncogenic processes. This combined analytical strategy facilitates the translation of multi-omic signals into mechanistically grounded biomarkers and stratification tools relevant to precision oncology in ACC.

3. Prognostic and predictive biomarkers

3.1. Gene-based signatures

Integrative transcriptomic analyses have provided fertile ground for developing gene-based prognostic signatures in ACC, revealing both converging biological principles and methodological diversity. A comprehensive summary of these gene-based prognostic models is provided in Table 2.

Hypoxia-associated gene signatures have shown particular promise. Chen et al. [4] developed a three-gene model, CCNA2, COL5A1, and EFNA3 demonstrating that high-risk ACC patients exhibit poorer overall survival and altered immune landscapes. Lipid metabolism-related signatures were described by Subramanian et al. [18], where several genes involved in cholesterol and phospholipid metabolism (e.g., SGPL1, FDFT1, SQLE, PIK3C2B, DGAT1, PLD1) showed stage-dependent expression patterns and significant survival associations in ACC (Fig. 4a). Similarly, Guo et al. [24] identified nine hub genes, including CCNB1 and CDK1, strongly associated with poor prognosis via KEGG and PPI analyses.

A comprehensive multi-omics approach by Li et al. [30] confirmed CCNB1 and NDC80 as key prognostic genes across datasets, validated by multiple algorithms. Meanwhile, Li et al. [32] explored the TME and used immune/stromal scores to uncover 18 hub genes linked to prog- nosis, providing a broader transcriptomic context. GABAergic signaling genes were also found to have prognostic significance. Knott and Lei- denheimer [27] showed that high expression of ABAT, GABRB2, and GABRD is associated with favorable outcomes in some ACC patients, suggesting metabolic heterogeneity (Fig. 4b). Li et al. [11] focused on the kinesin family genes (KIFs) and identified KIF4A, KIF11, and PLK1 as genes whose overexpression was correlated with tumor stage and im- mune infiltration.

Unsupervised ML-based transcriptomic clustering, as demonstrated by Marquardt et al. [29], identified two prognostic ACC subtypes (UMAP1/C1B and UMAP2/C1A) with distinct expression profiles. SOAT1 and EIF2A1 emerged as novel markers in this analysis. The role of lncRNAs in prognosis was investigated by Buishand et al. [13], who found that downregulation of LINC00271 was significantly associated with poor survival and dysregulation of cell cycle pathways. Finally, integrated mRNA-miRNA-IncRNA regulatory networks were explored by Subramanian et al. [14], highlighting a prognostic axis involving BIRC5, miR-335-5p, and PAX8-AS1 as a candidate biomarker triplet.

A comprehensive genomic analysis by Sun and Zhang [5] revealed a

Table 2 Gene-based prognostic signatures in ACC.
Model/StudyGenes IncludedMethodPrognostic Value
Buishand et al. (2020) [13]LINC00271, HOTTIP, CHL1, HOXA11-AS1, CRNDE, FAM211A-AS1, TBXAS1 SGPL1, FDFT1, SQLE, PIK3C2B, PIK3CD, SYNJ2, DGAT1, PLA2G16, PLD1, GPD1, CHPT1Human LncRNA/mRNA Microarray, TaqMan qRT- PCR, TCGA survival analysisOnly LINC00271: High
Subramanian et al. (2021) [18]Integrated bioinformatics analysis using GEO datasets (GSE19750, GSE10927, GSE12368), DAVID, STRING, Cytoscape (MCODE), TCGA validation, survival analysis with R2 and Cox regressionHigh expression of SGPL1, FDFT1 and SQLE, and low expression of PIK3C2B, PIK3CD, SYNJ2, DGAT1, PLA2G16, PLD1, GPD1, CHPT1 associated with poor survival
Subramanian et al. (2023) [14] Sun-Zhang et al. (2024) [5] Ye et al. (2020) [22]CDK1, CCNB1, PRC1, BIRC5, GINS1, TPX2 (upregulated), RSPO3, NR2F1, TLR4, USP53, HOXA5, SLC16A9 (downregulated) A2ML1, ABCC2, ABCC9, ALDH5A1, ANGPT2, ARTN, AURKA, BMP1, C7, CCNB2, CCDC150, CDC27, CDC7, CDK1, CDKN3, CDCA7, CHAF1B, CRNDE, DIAPH3, DPYD, DTL, EML1, ETV5, EXO1, FANCD2, FANCI, GAS2L3, GINS1, HELLS, HOOK1, HTRA3, IQGAP3, KIF23, KIF2C, KNTC1, LTBP1, NEK2, PHF19, PHACTR2, PNLIPRP3, POLD1, POLQ, RECQL4, RFC5, RRM2, SDC4, SH3D19, SLC7A14, SLC9A9, TACC3, TMEM150C, TPX2, USP53 mRNAs: CDK1, CCNB1, MCM4, MAD2L1, AURKA, NCAPG, RRM2, TYMS, TPX2; miRNAs: miR-212-3p, miR-24-3p, let-7a-5p, miR-196a- 5p; lncRNA: HOXA11-ASIntegrated analysis of 8 GEO datasets (GSE143385, GSE22816, GSE49279, GSE143383, GSE19750, GSE12368, GSE90713, GSE14922) using GEO2R, miRNet, DAVID, R2 Kaplan-Meier survival analysis Integrated analysis of TCGA, GEO (GSE10927), and YKD cohorts using RNA-seq, WES, differential expression analysis, Cox regression, and protein-protein interaction analysis Differential expression from GEO (GSE12368, GSE19750), GO/KEGG via DAVID, PPI via STRING, Hub gene detection via Cytoscape, miRNA prediction via miRTarBase, lncRNA prediction via miRNet, survival analysis via cBioPortal, OncomiR, and GEPIAAll 12 hub genes significantly associated with poor overall survival in TCGA (P < 0.01)
All 45 genes in the signature were significantly associated with poor survival, even after adjusting for stage and age (P < 0.001); patients with mutations in these genes had median survival of 18.1 months vs. 80.2 months All 9 key mRNAs (CDK1, CCNB1, MCM4, MAD2L1, AURKA, NCAPG, RRM2, TYMS, TPX2), 4 miRNAs (miR-212-3p, miR-24-3p, let-7a-5p, miR-196a-5p), and the lncRNA HOXA11-AS were significantly associated with poor prognosis (P < 0.05)
Li et al. (2023) [11]KIF4A, KIF11, KIF20A, KIF22 (prognostic); PRC1, PLK1, KIF23, KIFC1, KIF5A (interaction network)Gene expression comparison using TCGA and GTEx, survival analysis via GEPIA and R packages (survival, survminer), functional enrichment via clusterProfiler, protein-protein interaction via STRING and GeneMANIA, immune infiltration analysis via TIMER2.0High expression of KIF4A, KIF11, KIF20A, and KIF22 was significantly associated with poor overall survival (OS), disease-specific survival (DSS), and progression-free interval (PFI) in ACC patients (P < 0.001)
Zhang et al. (2023) [21]CDC20, DLGAP5, KIF2C, MMP9, MYC, CCL2; Hub genes: GAPDH, MYC, VEGFA, CDC20, CCL2, MMP9, ITGAM, DLGAP5, KIF2C, FCGR3A; DEmiRs: miR-21-3p (targets DLGAP5, CCL2), miR-21-5p (targets VEGFA, GAPDH, MYC, MMP9), miR-451a (targets MYC, MMP9), let-7e-5p, let-7a-5p (both target MYC, MMP9)Integrated analysis of GEO datasets (GSE90713, GSE143383, GSE19750, GSE22816), differential expression via limma, functional enrichment via Enrichr (GO, KEGG), PPI network via STRING and Cytoscape (CytoHubba), miRNA-hub gene network via multiMiR, survival analysis via GEPIA and TCGAHigh expression of CCL2 was associated with better survival; low expression of CDC20, DLGAP5, KIF2C, MMP9, and MYC also associated with better survival (P < 0.05); expression of VEGFA, GAPDH, ITGAM, and FCGR3A was not significantly associated with prognosis
Xu et al. (2020) [12]Top AS events included UQCR11 (upregulated AT: 46528, downregulated AT: 46527), NASP (ES: 2754), EIF3A, HNRNPR, and additional genes listed in Table 2 (e.g., WASH4P, ZNF692, RHOC, CIRBP, IKBKB, NOP2, PRKAG1, USP4, METTL15, STOML1, GGCX, PSEN2, etc.)Analysis of TCGA SpliceSeq data for 92 ACC patients; identification of survival-related AS events via univariate and multivariate Cox regression and Lasso analysis; development of eight prognostic models (based on seven AS types and all AS); GO and KEGG enrichment analyses for splicing factor (SF) genes; construction of AS-SF interaction network using Cytoscape3919 alternative splicing (AS) events were significantly associated with overall survival (OS) in ACC; prognostic models based on four AS types (AA, AD, AP, RI) showed strong predictive value (AUC > 0.9); multivariate analysis confirmed AS- based risk scores as independent prognostic factors
Xu et al. (2021) [19]RBM15, HNRNPC (prognostic m6A regulators); other analyzed m6A genes: METTL3, METTL14, WTAP, KIAA1429, ZC3H13, FTO, ALKBH5, YTHDC1, YTHDC2, YTHDF1, YTHDF2Lasso regression on TCGA data to construct m6A- based risk signature (RBM15 and HNRNPC); validation in GEO dataset (GSE33371); survival analysis via Kaplan-Meier, Cox regression, ROC curves, PCA, t-SNE; immune correlation via CIBERSORT, ssGSEA, TISIDBHigh expression of RBM15 and HNRNPC was significantly associated with poor survival in both TCGA and GEO validation cohorts; m6A-related risk score (based on RBM15 and HNRNPC) was identified as an independent prognostic factor with AUC > 0.7; high-risk group had worse overall survival across multiple clinical subgroups
Li et al. (2020) [32]CD4, HLA-DRA, HCK, CD53, RAC2, HLA-DRB5, RAB37, CD93, FOLR2, GRAP2, HLA-DOA, HLA- DPA1, HPGDS, LAIR1, PTPRB, TACR1, TBXAS1, WASTranscriptome data and clinical characteristics from TCGA; immune and stromal scores calculated using ESTIMATE algorithm; DEGs identified via limma; intersect genes from immune and stromal DEGs; hub genes identified by cytoHubba (12 topological algorithms); survival analysis via Kaplan-Meier and Cox regression; immune infiltration analysis via TIMER databaseAll 18 hub genes were significantly associated with poor overall survival (OS); patients in the high-risk group (based on multivariate Cox regression models using these genes) had significantly worse survival outcomes (P < 0.0001); AUC for the prognostic model was 0.887
Marquardt et al. (2021) [29]Top genes identified by RF and associated with poor prognosis cluster ACC-UMAP2 include SLC2A1 (GLUT1), SOAT1, EIF2S1, MYC, SMAD2, BUB3, ASB4, MED27, FSCN1, GNAI3, CBX3; total top-100 genes are listed in the studyUMAP-based unsupervised clustering using mRNA expression (FPKM) data from TCGA-ACC (n = 79); random forest classifier trained on cluster labels to identify key discriminative genes; validation on ENSAT dataset; survival analysis via Kaplan-Meier and Cox regression; mutation correlation via cBioPortalUMAP-based clustering of TCGA-ACC transcriptomic data reproduced known C1A/C1B clusters associated with prognosis; ACC-UMAP1 correlated with better survival (median OS: 12.46 years) vs. ACC-UMAP2 (median OS: 7.38 years, HR = 6.27, p = 0.000029); random forest classifier identified 100 genes with the highest discriminatory power, most overexpressed in the poor prognosis cluster ACC-UMAP2
Li et al. (2022) [30]CCNB1, CDC20, CENPU, FOXM1, KIF4A, KIF11, KIF20A, MAD2L1, NCAPG, NDC80, NUF2, PBK,Differential expression analysis using TCGA and GEO datasets (GSE10927, GSE19750); functional enrichment (GO, KEGG, PANTHER, BIOCYC,16 of the 17 hub genes were significantly associated with poor prognosis in ACC (log-rank P < 0.001); only C3AR1 was not significantly (continued on next page)
Table 2 (continued)
Model/StudyGenes IncludedMethodPrognostic Value
RACGAP1, RRM2, TOP2A, TPX2 (prognostic); C3AR1 (excluded from prognostic model)REACTOME) via WebGestalt and KOBAS; PPI network and hub gene selection via STRING and Cytoscape (cytoHubba: Degree, MCC, MNC); survival analysis via GEPIA and cBioPortalcorrelated with survival; CCNB1 and NDC80 were identified as core prognostic genes
Tömböl et al. (2009) [15]Differentially expressed miRs: miR-184, miR- 503, miR-511, miR-214, miR-210, miR-375; Target genes inferred from expression correlation include CCNB2, CDC2, CDC25C, TOP2A, CDKN1C, IGF2 along with further genes reported in the studyIntegrated miRNA and mRNA expression profiling (TLDA microarray and Agilent whole-genome microarrays), tissue-specific target prediction using combined data from TargetScan, PicTar, miRBase, and expression correlation; pathway analysis via GSEA and Ingenuity Pathway Analysis (IPA)Differential expression of miR-184, miR-503 (upregulated), and miR-511, miR-214, miR-375 (downregulated) distinguished ACC from benign adrenocortical tumors with high diagnostic power; ROC analysis for dCTmiR-511-dCTmiR- 503 achieved 100 % sensitivity and 97 % specificity; prognostic relevance inferred from expression patterns, though survival data not directly analyzed
Xu et al. (2020) [26]CNN1, MYLK, CSRP1, MYH11, EFEMP2, FBLN1, MFAP4, FBLN5Differential expression analysis from GEO datasets (GSE19750, GSE12368, GSE14922), PPI network via STRING, hub gene selection via Cytoscape (MCODE), enrichment via DAVID, survival analysis via cBioPortal, gene expression verification via Oncomine, clinical correlation via UCSC Cancer BrowserExpression changes of all eight hub genes did not affect overall survival; however, downregulation of EFEMP2 was significantly associated with decreased disease-free survival; CSRP1 and MFAP4 expression levels were also associated with adverse clinicopathological features (capsular invasion, histological grade, vascular invasion)
Yan et al. (2021) [36]ASPM, AURKA, CCNB2, CDC20, CENPA, EXO1, FBXO5, HJURP, KIF2C, MKI67, NUF2, PARPBP, TACC3, TROAPMulti-step bioinformatics analysis including WGCNA on GSE76019, regression modeling (Lasso, Ridge, Elastic Net) on TCGA-ACC, survival analysis (OS, DFS), ROC/AUC validation, CNV and mutation analysis via cBioPortal, functional enrichment (GO, KEGG), PPI network via STRING and Cytoscape, classifier validation via LDA, KNN, and SVMAll 14 genes in the best model were significantly associated with poor overall survival (OS); model 2 (a = 0.6) showed the best prognostic performance with AUC = 0.844 and C-index = 0.851; six genes with highest connectivity were considered hub genes (ASPM, AURKA, CCNB2, CDC20, KIF2C, NUF2)
Yin et al. (2023) [25]CDK1, CCNA2, CCNB1, TOP2A, MAD2L1, BIRC5, BUB1, AURKAIntegrated analysis of GEO datasets (GSE12368, GSE90713, GSE143383), DEGs via GEO2R, enrichment via DAVID (GO, KEGG), PPI via STRING and Cytoscape (MCODE and cytoHubba), survival analysis via cBioPortal, expression and staging via UCSC Xena and GEPIAHigh expressions of CDK1, CCNA2, CCNB1, TOP2A, MAD2L1, BIRC5, BUB1, and AURKA was significantly associated with poor overall survival and disease-free survival (P < 0.05); alterations in these genes were also linked to advanced tumor stage in ACC
Yuan et al. (2018) [23]ANLN, ASPM, CDCA5, CENPF, FOXM1, KIAA0101, MELK, NDC80, PRC1, RACGAP1, SPAG5, TPX2WGCNA using GSE10927 dataset; DEGs via limma; hub gene selection through co-expression network and PPI (STRING, Cytoscape); validation in GSE19750 and TCGA-ACC using ANOVA, Pearson correlation, ROC curves, survival analysis via GEPIAAll 12 hub genes were significantly associated with advanced tumor stage and poor overall and disease-free survival in TCGA and GEO datasets; ROC curve AUCs ≥ 0.85 confirmed diagnostic potential
Yi et al. (2022) [37]ASPM, BIRC5, CCNB2, CDK1 (MPBs); additional 5 hub genes: DLGAP5, FOXM1, RACGAP1, TOP2A, TPX2Integrated multi-omics analysis using six GEO datasets (GSE75415, GSE12368, GSE76021, GSE19750, GSE10927, GSE76019) and TCGA-ACC; methods included DEG analysis, WGCNA, PPI network (STRING, Cytoscape), survival analysis (Kaplan-Meier, Cox), ROC, DCA, LDA, KNN, SVM classifiers, immune infiltration analysis (TIMER), and nomogram constructionAll nine hub genes were significantly associated with poor overall and disease-free survival (OS and DFS); four MPBs (ASPM, BIRC5, CCNB2, CDK1) showed the highest predictive accuracy (average classification accuracy ≥ 0.80); ASPM was confirmed as an independent prognostic factor in multivariate Cox analysis (HR = 3.262 for OS, HR = 7.335 for DFS)
Tian et al. (2020) [35]KIF18A, CDCA8, SKA1, CEP55, BUB1, CDK1, SGOL1, SGOL2Integrated bioinformatics and real-world analysis using TCGA and GEO datasets (GSE90713, GSE12368), CIBERSORT algorithm, immunohistochemistry (IHC) validation in 39 ACC patients, DEG analysis via limma, GSEA and enrichment analysis via clusterProfiler, PPI network (STRING, Cytoscape), survival analysis using Cox regressionHigher abundance of tumor-infiltrating mast cells (TIMC) was significantly associated with improved overall survival (OS) and progression- free survival (PFS) in both TCGA and FUSCC cohorts; low TIMC infiltration and overexpression of hub genes (KIF18A, CDCA8, SKA1, CEP55, BUB1, CDK1, SGOL1, SGOL2) were associated with poor prognosis
Knott and Leidenheimer (2020) [27]ABAT, ALDH5A1, GAD1 (GABA shunt); GABRB2, GABRD, GABRA3, GABRA5, GABRB3, GABRG1, GABRE (GABAA subunits); GABBR1, GABBR2 (GABAB subunits); SLC6A1, SLC6A11, SLC6A12, SLC6A13, SLC32A1 (GABA transporters)Targeted bioinformatics analysis of TCGA-ACC dataset focusing on 30 GABA system genes; transcriptomic and methylation data from cBioPortal; validation via GEO datasets and RT-PCR in NCI- H295R cell line; survival analysis via Kaplan-Meier, gene enrichment (GO/KEGG), and correlation with clinical featuresUpregulation of ABAT and ALDH5A1 was significantly associated with favorable clinical outcomes, including longer overall and progression-free survival, disease-free status, and absence of metastases; upregulation of GABRB2 was positively prognostic, while upregulation of GABRD was negatively prognostic
Sung and Cheong (2021) [17]TMM-related transcriptional regulators in ACC with ALT included CBX3, NRF1, EP300, and NFYB; enriched biological processes included mitochondrial organization and peptide secretion; no specific hub genes were listed, focus was on TMM pathway signaturesPan-cancer analysis using TCGA RNA-seq data from 31 cancer types (n = 10,704); classification of TMM types via ssGSVA; survival analysis via Kaplan-Meier and Cox regression; functional analysis via GO and transcription factor network (iRegulon); copy number variation, mutation burden, and stemness analysis includedValue: In ACC, patients with NDTMM had better survival compared to those with ALT, which was associated with poor prognosis; ALT group showed enriched mitochondrial organization, high copy number variation (CNV), and unfavorable transcription factor networks (CBX3, NRF1, EP300, NFYB); TMM classification (ALT, TEL, TEL+ALT, NDTMM) was predictive of survival across multiple cancer types
Guo et al. (2020) [24]CCNB1, CDK1, TOP2A, CCNA2, CDKN3, MAD2L1, RACGAP1, BUB1, CCNB2Integrated analysis of GEO datasets (GSE10927, GSE12368, GSE90713); DEGs via limma; functional enrichment via DAVID (GO, KEGG); PPI network via STRING and Cytoscape; validation of hub genes via GEPIA and survival data from TCGA-ACC; univariateAll nine hub genes were significantly upregulated in ACC and associated with poor overall survival in TCGA; among them, BUB1 was identified as an independent prognostic factor (P < 0.01)
Table 2 (continued)
Model/StudyGenes IncludedMethodPrognostic Value
Tian et al. (2021) [33]Prognostic proteins: FASN, FIBRONECTIN (FN), TFRC, TSC1; DEGs related to model: CENPM, NDC80, DLGAP5, SPC25, CENPF, ZWILCH, AURKB, CENPA, CDC20, CCNA2, KIF4A, BUB1B, CCNB2, UBE2C, AURKA, PLK1, CASC5, RANGAP1, BIRC5, CEP55, NEK2, SGOL2, KIF18A, CCNB1, SKA1, RRM2, ASPM, SGOL1, KIF2C, CDCA8, CENPI, KIF11, BUB1, CDCA5, CDK1, SPC24, SPAG5, NUF2and multivariate Cox regression for prognostic evaluation Multiomics analysis using RPPA data from TCPA, gene expression data from TCGA and GEO (GSE10927), and clinical data; candidate proteins selected via Kaplan-Meier and Cox regression; model built using Lasso and multivariate Cox regression; validation via IHC in FUSCC cohort and comparison with Ki-67, TP53, CTNNB1, IGF2, NR5A1; functional analysis via GSEA and DEG analysisHigh expression of FASN, FIBRONECTIN, TFRC, and TSC1 was significantly associated with poor overall survival in adult ACC patients (P < 0.05); the IPRPs model (based on these four proteins) showed independent prognostic significance with AUC = 0.933, outperforming Ki-67 and other markers across multiple cohorts (TCGA, GEO, FUSCC)
Xiao et al. (2018) [34] Chen et al. (2021) [4] Xing et al. (2019) [2]TOP2A, NDC80, CEP55, CDKN3, CDK1Integrated analysis of six GEO datasets (GSE12368, GSE14922, GSE19750, GSE33371, GSE75415, GSE90713); DEGs identified via robust rank aggregation; PPI network constructed using STRING and Cytoscape; hub genes selected based on connectivity degree > 20; expression validated using TCGA, GTEx, Oncomine, RT-PCR, Western blot, and immunofluorescence RNA-seq data from TCGA and GEO (GSE19750), hypoxia gene list from MSigDB; univariate and multivariate Cox regression, LASSO analysis to select genes; survival analysis (Kaplan-Meier, Cox), ROC curve, nomogram construction, immune infiltration via CIBERSORT, GSEA for pathway analysis DEG analysis using GEO datasets (GSE12368 and GSE19750); functional enrichment via DAVID (GO, KEGG); PPI network via STRING and Cytoscape (MCODE); survival analysis via cBioPortal; validation via Oncomine (Giordano Adrenal datasets) and UCSC Cancer BrowserHigh expression of all five hub genes was significantly associated with poor overall survival (OS) and disease-free survival (DFS) in TCGA-ACC cohort (P < 0.001); expression of these genes was also positively correlated with advanced pathological stage and T stage
CCNA2, EFNA3, COL5A1 AURKA, TYMS, GINS1, RACGAP1, RRM2, EZH2, ZWINT, CDK1, CCNB1, NCAPG, TPX2, MAD2L1, PRC1, SMC4High expression of CCNA2, EFNA3, and COL5A1 was significantly associated with poor overall survival in both TCGA and GEO cohorts (P < 0.001); the hypoxia risk score based on these genes was an independent prognostic factor with high predictive accuracy (AUC = 0.949 at 1 year) Upregulation of AURKA, TYMS, GINS1, RACGAP1, RRM2, EZH2, ZWINT, CDK1, CCNB1, NCAPG, and TPX2 was significantly associated with poor overall and disease-free survival in ACC; MAD2L1 and PRC1 were associated with worse disease-free survival; SMC4 showed no significant survival correlation
Zhou et al. (2021) [3]Core genes regulated by BCLAF1: CDK1, CCNB1; a 53-gene risk model was also identified (including CDCA8, KIF11, CENPA, AURKB, BUB1, etc.)RNA-seq and clinical data from TCGA-ACC; differential expression via limma; WGCNA to identify key modules; PPI network via STRING and Cytoscape; hub genes via degree centrality; GSEA, GO, KEGG for pathway analysis; experimental validation in NCI- H295R and SW-13 cell lines (siRNA, shRNA, qPCR, WB, flow cytometry, IHC)High expression of BCLAF1 was significantly associated with poor overall and disease-free survival (P < 0.01); BCLAF1 regulates cell proliferation and cell cycle in ACC by directly modulating CDK1 and CCNB1 at mRNA and protein levels; high-risk gene signature of 53 BCLAF1-related genes showed AUC = 1 for survival prediction
Omidi (2025) [31]CKS2, ACAT2Integrative ceRNA network modeling using TCGA- ACC and GTEx transcriptomes; differential expression; co-expression structure; survival modeling (Kaplan-Meier + CoxPH); ROC diagnostic evaluationHigh expression of CKS2 and ACAT2 associated with reduced overall survival; combined signature improves risk stratification and provides strong diagnostic performance (AUC ~ 0.90)

45-gene signature associated with survival, identified through integra- tion of mRNA profiles, mutational data, and clinical outcomes. This panel included cancer-associated genes such as POLD1, AURKA, and KIF23, which interact with TP53, a frequent germline alteration in ACC. Patients with mutations in these genes showed significantly reduced survival (median 18.1 vs. 80.2 months), highlighting their prognostic potential. Tian et al. [33] constructed the IPRPs model, a proteogenomic signature based on protein-level expression data from TCGA and real-world samples. It included FASN, TFRC, and TSC1, whose high expression was strongly associated with poor survival. Notably, the model outperformed Ki-67 in prognostic discrimination, validating its clinical utility in multiple cohorts.

Alternative splicing (AS) was investigated by Xu et al. [12], who identified 3919 survival-associated AS events and constructed four AS-based prognostic models with outstanding performance (AUC > 0.9). The top events included exon skipping and alternate donor sites, sug- gesting a novel layer of regulation in ACC pathogenesis and prognosis. Xiao et al. [34] reported a five-gene panel, TOP2A, NDC80, CEP55, CDKN3, and CDK1, as a predictive model for progression and prognosis. These genes were significantly upregulated in ACC tissues, validated via RT-PCR, western blot, and immunofluorescence, and correlated with poor survival. Xing et al. [2] identified 14 hub genes from microarray analysis and PPI network construction, including AURKA, RRM2, CDK1, and EZH2, which were functionally enriched in cell division and

p53-related pathways. These genes were significantly associated with poor prognosis and aggressive phenotypes.

Using weighted co-expression network analysis, Xu et al. [26] discovered eight downregulated hub genes such as EFEMP2, CNN1, and MFAP4, some of which (e.g., EFEMP2) correlated with shorter disease-free survival. Their loss may contribute to tumor progression. Tian et al. [35] analyzed immune infiltration and showed that genes like KIF18A, CDK1, CDCA8, and SGOL1 were upregulated in low-mast cell ACC tumors and were associated with worse outcomes. This suggests that gene-based signatures may reflect tumor immune microenviron- ments. Xu et al. [19] focused on m6A RNA methylation regulators and developed a prognostic risk score based on HNRNPC and RBM15. This model predicted survival independently of clinical stage and correlated with immune suppression.

In a pioneering multi-omic approach, Sung and Cheong [17] con- ducted a pan-cancer telomere maintenance mechanism (TMM) analysis and showed that TMM classification had prognostic implications in ACC. Although not a conventional gene expression signature, this highlights the potential of telomere biology in ACC stratification. Furthermore, Tömböl et al. [15] conducted one of the earliest studies integrating miRNA and mRNA profiles. They identified distinct miRNA-mRNA regulatory networks in ACC versus adenomas and proposed miR-503 and miR-511 as possible diagnostic and prognostic markers. Their combined ACT expression distinguished ACC from benign tumors with

Fig. 4. Gene expression-based prognostic signatures in ACC. A diverse range of transcriptomic models have been proposed to stratify ACC patients based on gene expression profiles. This figure summarizes key biomarkers and their prognostic relevance using representative visualizations from recent studies. (a) Kaplan-Meier survival curves for lipid metabolism-related genes showing significant associations between gene expression levels and overall survival in ACC [18]. (b) Kaplan-Meier survival analysis showing that upregulation of GABA pathway-related genes (e.g., ABAT, GABRB2, GABRD) is associated with improved overall and progression-free survival in ACC [27]. (c) Kaplan-Meier survival curves and ROC analyses comparing two prognostic models derived from a 14-gene signature in ACC. Model 2 shows superior predictive accuracy across independent cohorts [36]. (d) Nomogram for overall survival prediction in ACC, constructed using ASPM expression and clinical stage, based on the prognostic MPB signature (ASPM, BIRC5, CCNB2, CDK1) [37]. (e) Positive correlation between BCLAF1 expression and CDK1/CCNB1 across three independent ACC cohorts, supporting its role in promoting cell cycle progression [3]. (f) Reinforced co-expression of CKS2 and ACAT2 in ACC tumors, indicating a synergistic oncogenic axis [31].

SGPL1

SQLE

FDFT1

PIK3C2B

PIK3CD

1.UU

1.UU

-

1.UU

1

high (n=12)

1 1

0.90

1

high (n=56)

1.UU

high (n=23) low (n=56)

high (n=17)

0.90

0.90

high (n=38)

0.90

low (n=62)

low (n=67)

0.90

low (n=23)

low [n=41)

overall survival probability

0.80

overall survival probability

O RO.

overall survival probability

o no.

overall survival probability

0.80

overall survival probability

0.80

0.70

0.70

0.70.

0.70

0 70-

O RO

no

0 80

O.GO-

0.60

0.50

0.50

0.50

0.50

0.50-

0.40

0.40

0.40

0.40.

0.40-

0 30

0.30

0.30.

0.30

0.30

0.20

0.20

0.20

0 20

0.20-

0.10.

0.10

so= = 8.30-03

0.10

bent p.u.u1.

hent : 0.500

raw # 2.50-07 sont $ 1.00:00

0.10-

199 p2.76-45 torf p 1.70-42

0 10

MW A 0.451

0.00-

0.00

0

2

.

¿

A

2

#

N

=

£

0.00-

B

2

=

8

0.00-

0.00

hent p 1 000

£

£

5

2

&

A

&

8

6

2

=

FJ

*

Follow up in months

Follow up in months

8

Follow up in months

Follow up in months

#

Follow up in months

a

PLA2G16

DGAT1

PLD1

GPD1

CHPT1

high (n=28) low (n=51)

1.00

1.00

1.UU

I.UU

0.90

high (n=66)

:

1.00

0.90

high (n=39)

0.90

high (n=71)

0.90

high (n=45)

0.00

low (n=40)

overall survival probability

0 m.

low (n=13)

overall survival probability

0 m

o no.

cw (n=8)

0.80

low (n=34)

overall survival probability

0.80

0.70

0.70.

overall survival probability

0.70

overall survival probability

0.70

0.70-

O An.

0 80

DRO

O RO

0.00

0.bu

0.50

3.50

0.50

0.50

0.40

0.40

0.40

0.40

0.40

0.30

0.30

0.30

0 30

0.30

0.20

0.20

0.20-

0.20.

0.20

0.10.

me # 8.50-04

0.10

tzw p 2.02-03 tontp 8.178

0.10.

hont : 0.00

P = 0,011 bont p &L.086

0.10.

0.10

bont # 1.00g

naw p 7.48-04 bord = 0.447

0.00

4

a:

0.00

..

2

=

5

A

2

20

=

¥

&

0.00

E

A

2’

8

S

8

0.00-

o no.

8

E

2

%

$

g

£

S

2

8

8

g

E

Follow up in months

Follow up in months

Follow up in months

Follow up in months

Follow up in months

Overall survival on risk score calculated by model 1 (GSE19750)

Overall survival on risk score calculated by model 2 (GSE19750)

100%-

p = 1.7x10-3

100%-

p = 1.2 × 10-4

Overall survival

Progression-free survival

80%

80%-

100

:

1.00

L

:

60%

upregulated

upregulated

BTS

=

B

60%

Survival probability

Serathity

=

Survival probability

40%

Šemitivity

=

unaltered

40%-

3

=

20%

(b)

20%-

unaltered

=

p=0.000

B

-

AUC at 1 year. B.B

0%

0%

Bár

8

=

0

20

40

60

80

100

120

140

0

20

40

60

80

100

120

a

140

50

OS (months)

Địa

6

50

Number at risk

02

0.4

08

0.8

Number at risk

OS (months)

0.0

03

B.4

00

1.0

Time (months)

Time (months)

O

LH IN

E

8

3

150

L

1-Specificity

HI 1

A

3

do

9

to

1-Specificity

OS (months)

OS (months)

10

20

30

40

70

90

100

Events free survival on risk score calculated by model 1 (GSE76021)

Events free survival on risk score calculated by model 2 (GSE76021)

(c)

50

60

80

Points

0

ASPM

1

10.5

10

9.5

9

8.5

8

7.5

7

6.5

6

5.5

5

100

Im

:

=

Pathologic_stage

IV

Il

075

=

:

Total Points

Survival probability

=

Survival probably

8

0

10

20

30

40

50

60

70

80

90

100

110

120

130

0.50

Serutkity

Bereitsity

Linear Predictor

3

8

5.5

6

6.5

7

7.5

8

8.5

9

9.5

10

10.5

0.25

8.25

1-year Survival Probability

P=0012

=

P= 0.012

:

0.3101-455586857.76.8.850.9 0.95

00

AUG Mf 1 year: D.P

3-years Survival Probability

3

60

BỘG

tla

9

6

6

d

não

B

AUG af 5 years: D.TT

0.1

0.2

0.3

0.

0.7

0.8

0.9

Number all risk

EFS (months)

aa

02

0.4

0.8

Number of risk

EFS (months)

0.0

0.2

D.4

.

1.0

5-years Survival Probability

2

HE

2

E

EFS (months)

LA

2

==

LE

$

MR

&

EFS (months)

0.1

0.2

0.3

.50. 6 0.

0.8

0.9

e

TCGA

GSE10927

GSE90713

CKS2 vs. ACAT2

log2(CDK1 expression)

4.0-

r=0.6101

log2(CDK1 expression)

11-

8-

r=0.5698

r=0.5684

Normal (r=0.03)

p=0.0002

p=0.0001

log2(CDK1 TPM)

p<0.0001

10-

log2(ACAT2 TPM + 0.01)

n=33

8

6-

3.5

n=58

Tumor (r=0.44)

9

4.

3.0

8

7

6

2

2.5

0

6-

2

4

6

8

2.0

5

4

-2-

log2(BCLAF1 TPM)

2.4

2.6

2.8

3.0

3.2

3.4

4

5

6

7

8

9

log2(BCLAF1 expression)

log2(BCLAF1 expression)

10-

r=0.5939

log2(CCNB1 expression)

4.0-

r=0.4109

log2(CCNB1 expression)

2

15-

r=0.5322

log2(CCNB1 TPM)

8

p<0.0001

p=0.0175

p=0.0001

3.5

n=33

n=58

6

10

0

3.0

4

2

2.5

5

f)

4

6

8

0

2.0

0

0

2

4

6

8

2.4

2.6

2.8

3.0

3.2

3.4

4

5

6

7

8

log2(CKS2 TPM + 0.01)

9

log2(BCLAF1 TPM)

log2(BCLAF1 expression)

log2(BCLAF1 expression)

high accuracy.

Yan et al. [36] employed WGCNA and machine learning methods (Ridge, Lasso, and Elastic Net regression) to construct a multi-gene prognostic model based on TCGA-ACC and GEO datasets. From 93 survival-associated genes, two prognostic models were derived, with Model 2 demonstrating superior predictive performance in both internal and external validation cohorts. This model comprised 14 genes, of which six, BIRC5, CCNB1, CDK1, MKI67, NDC80, and TOP2A were further validated as hub genes using PPI network analysis and receiver operating characteristic (ROC) curves (Fig. 4c).

Similarly, Yi et al. [37] integrated seven independent datasets and applied multi-omics and machine learning techniques to identify ASPM, BIRC5, CCNB2, and CDK1 as meaningful prognostic biomarkers (MPBs) for ACC. These genes showed high expression in tumor tissues, were associated with poor survival, and demonstrated strong predictive value in nomogram-based models (Fig. 4d).

Yin et al. [25] confirmed the centrality of cell cycle regulation in ACC

progression. Through bioinformatics analyses of three GEO datasets, they identified CDK1, CCNA2, CCNB1, TOP2A, MAD2L1, BIRC5, BUB1, and AURKA as hub genes. These genes were significantly associated with mitotic cell cycle pathways and overall survival. Yuan et al. [23] used WGCNA to define a gene module significantly associated with clinical features of ACC. From this module, 12 hub genes were validated as prognostic markers, with enrichment in cell cycle processes, further supporting the findings of subsequent studies. Zhou et al. [3] investi- gated the oncogenic role of BCLAF1, showing that it regulates CDK1 and CCNB1, promoting cell cycle progression and ACC proliferation. Knockdown of BCLAF1 significantly inhibited ACC cell growth, sug- gesting its potential as a therapeutic target and a regulator of gene-based prognostic pathways (Fig. 4e).

Omidi [31] identified CKS2 and ACAT2 as consistently upregulated and co-regulated oncogenic drivers in ACC, with strong diagnostic ac- curacy (AUC ~ 0.90) and adverse survival association. Their reinforced co-expression in tumor samples, absent in normal adrenal tissue,

Fig. 5. Prognostic protein biomarkers and immunohistochemical validation in ACC. Key protein-level markers have been validated in ACC using a combination of western blot, qPCR, immunofluorescence, and IHC approaches across multiple studies: (a) Western blot validation showing that BCLAF1 regulates CDK1 and Cyclin B1 expression in ACC cell lines [3]. (b) Upregulation of TOP2A in ACC compared with normal adrenal tissue, validated by western blot and immunofluorescence; mRNA analyses also support elevated CEP55 expression [34]. (c) ROC analysis demonstrating that a multigene panel outperforms tumor size in distinguishing ACC from benign adrenal tumors, consistent with high diagnostic accuracy of CCNB2, IL13RA2, and HTR2B reported by RT-PCR validation (Figure redrawn from original data in [8].) (d) IHC validation of the IPRP prognostic protein panel (FASN, Fibronectin, TSC1, TFRC) in ACC tumor samples [33].

si NC si Bclaf1

EV

FL

145kDa

Bclaf1

ACC (n = 4)

Normal (n = 4)

P=0.0097

p=0.0550

P=0.0138

p= 0.0013

p= 0.0098

TOP2A

200kDa

GAPDH

37kDa

34kDa

CDK1

p = 0.0116

Relative mRNA expression

1.50

3-

58kDa

Cyclin B1

Relative protein expression of TOP2A

1.25

2-

ß-actin

1.

(a)

NCI-H295R

(b)

1.00-

0

CDK1

CDKN3

CEP55

NDC80

TOP2A

si NC si Bclaf1 sh NC sh Bclaf1

Normal(n=4)

ACC(n=4)

a

b

C

145kDa

Bclaf1

:

34kDa

CDK1

Normal

.

50pm

d

e

f

58kDa

Cyclin B1

ß-actin

ACC

50um

SW-13

Anti-FASN

Anti-FIBRONECTIN

(d)

1.0

True Positive Rate (Sensitivity)

0.8

0.6

0.4

Anti-TSC1

Anti-TFRC

0.2

Multi-Omic Predictor

Tumor Dimension

0.0

(c)

0.0 0.2 0.4 0.6 0.8 1.0

False Positive Rate (1 - Specificity)

highlights a synergistic oncogenic axis, wherein CKS2 accelerates pro- liferative progression while ACAT2 sustains the metabolic demands of lipid-supported energy reprogramming (Fig. 4f).

Zhang et al. [21] focused on metastatic ACC and identified CDC20, DLGAP5, KIF2C, and MYC among others as metastasis-associated genes with significant prognostic implications. Their survival analyses underscored the prognostic value of these genes in metastatic settings. Ye et al. [22] adopted a ceRNA-based approach, identifying RRM2, CDK1, and MCM4 as key mRNAs involved in prognostically significant ceRNA subnetworks. These genes, through miR-24-3p and IncRNA HOXA11-AS regulation, were implicated in ACC prognosis and therapeutic resistance.

These studies illustrate the rich landscape of gene-based prognostic markers in ACC, covering protein-coding genes, splicing variants, epigenetic modifiers, and non-coding RNAs, each offering complemen- tary insights for future clinical application. These diverse approaches are visually summarized in Fig. 4, including mutation-based risk, gene overexpression, m6A methylation risk signature, alternative splicing- based models, PPI-based hub genes, downregulated survival- associated genes, immune-related gene profiles, telomere maintenance classification, and miRNA-based diagnostic stratification.

3.2. Protein markers and immunohistochemistry

At the protein level, several studies have validated candidate bio- markers for ACC prognosis and diagnosis through immunohistochem- istry (IHC), western blotting, and proteomics. Zhou et al. [3] demonstrated that BCLAF1 is significantly upregulated in ACC tissues and promotes proliferation via transcriptional regulation of CDK1 and CCNB1 (Fig. 5a). Western blot and qPCR confirmed that silencing BCLAF1 leads to downregulation of these two cell cycle genes, high- lighting its role in mitotic control.

Xiao et al. [34] further validated the protein overexpression of TOP2A and CEP55 in ACC via western blot and immunofluorescence, showing their presence in both the cytoplasm and nucleus of tumor cells. High expression levels correlated with poor survival and higher patho- logical stage, making them robust prognostic markers (Fig. 5b). Simi- larly, Fernandez-Ranvier et al. [8] used RT-PCR to validate 37 differentially expressed genes between benign and malignant adrenal tumors. Among them, CCNB2, IL13RA2, and HTR2B showed high diagnostic accuracy (AUC > 0.85), strongly distinguishing ACC from benign lesions (Fig. 5c). Tian et al. [33] constructed an Integrated Prognosis-Related Proteins (IPRPs) model based on protein-level data. Proteins like FASN, fibronectin, TFRC, and TSC1 showed significant prognostic power in both TCGA and clinical cohorts. Notably, IHC validation showed that this model outperformed Ki-67 in prognostic accuracy (Fig. 5d).

Another layer of proteomic confirmation was presented by Yi et al. [37], who identified CDK1, CCNB2, ASPM, and BIRC5 as multi-platform validated prognostic biomarkers. These proteins were enriched in the cell cycle and validated via expression correlation with copy number alterations, immune infiltration, and clinical outcomes. Finally, the infiltration of tumor-associated mast cells (TIMCs), evaluated via IHC in the Fudan University cohort, revealed a positive correlation between TIMC abundance and improved survival. However, lower mast cell infiltration coincided with higher expression of poor-prognosis proteins such as CDK1 and CEP55 (Tian et al. [33]). Together, these findings support the clinical relevance of protein-level biomarkers and their integration with transcriptomic profiles for ACC characterization.

4. Tumor microenvironment and immune landscape

The tumor microenvironment (TME) of ACC represents a dynamic interplay between malignant cells, immune infiltrates, stromal compo- nents, and molecular signaling networks. Disruption in immune sur- veillance and immune suppression within the TME are key factors

contributing to ACC’s aggressiveness and poor prognosis.

One of the most comprehensive analyses of tumor-infiltrating im- mune cells in ACC was performed by Tian et al., who utilized the CIBERSORT algorithm on TCGA and GEO datasets to quantify immune cell subsets and validated their findings in real-world tissue samples. Notably, the presence of tumor-infiltrating mast cells (TIMCs) was positively correlated with favorable prognosis, while low mast cell infiltration accompanied by overexpression of genes such as CDK1 and BUB1 predicted poor outcomes (Fig. 6a) [35]. Chen et al. [4] also characterized the tumor microenvironment (TME) in ACC by applying the CIBERSORT algorithm to assess immune cell composition. Their analysis demonstrated that the high-risk cohort, defined by a hypoxia-related signature, exhibited increased infiltration of regulatory T cells (Tregs) and M2 macrophages along with reduced CD8+ T-cell levels, reflecting a distinctly immunosuppressive microenvironment.

Zhou et al. [3] provided further mechanistic insight by showing that BCLAF1, a transcription factor implicated in cell cycle progression, also correlates with immune cell infiltration. High BCLAF1 expression was associated with reduced CD8 + T cell infiltration and immune-silent phenotypes, suggesting an immunosuppressive role in the ACC micro- environment [3] (Fig. 6b). In a multi-omics study, Guan et al. classified ACC into three molecular subtypes (ACC1-ACC3). ACC1 was enriched for immune-related pathways and showed sensitivity to PD-1 checkpoint blockade, while ACC2 demonstrated high infiltration of immunosup- pressive cells and worse prognosis [12] (Fig. 6c). Li et al. [32] identified a group of 18 hub genes related to immune and stromal scores in the TME. Patients with lower immune scores, often seen in metastatic ACC, had significantly worse survival outcomes. Among the hub genes, several (e.g., KIF11, KIF4A) showed marginal positive correlation with immune infiltration [32].

Further bioinformatics analyses by Yi et al. [37] confirmed that several key prognostic genes (CDK1, ASPM, BIRC5, CCNB2) were significantly enriched in cell-cycle and immune evasion pathways. Expression of these markers was also linked to immune cell infiltration and survival prediction models, as demonstrated by their correlation with immune cell fractions and nomogram-based survival curves [37] (Fig. 6d).

Another pan-immune profiling effort by Li et al. [30] incorporated genomic, transcriptomic, and proteomic data to identify distinct im- mune subtypes in ACC, marked by variations in antigen presentation, immune checkpoint expression, and immune cell density. These classi- fications offer promising frameworks for personalized immunotherapy [30]. Zhang et al. [21] observed that hub genes involved in ACC metastasis, such as VEGFA, CCL2, and MMP9, also play roles in immune regulation and leukocyte migration, suggesting a bridge between metastasis and immune modulation. Their interaction network revealed overlapping roles in inflammatory response and immune recruitment.

Sung and Cheong [17] highlighted that telomere maintenance mechanisms (TMM), particularly the alternative lengthening of telo- meres (ALT), were associated with specific immune features in ACC. Patients with ALT activity showed distinct immune activation patterns, which may have prognostic and therapeutic implications. Arastonejad et al. [38] showed overexpression of NCAPG and NCAPH in ACC, both of which were associated with immune pathway enrichment and poor survival, pointing again to a link between chromosomal structure regulation and immune contexture. Finally, Li et al. [11] noted that certain kinesin family genes (KIFs), such as KIF20A and KIF22, were overexpressed in ACC and marginally correlated with immune infiltra- tion. These findings support the hypothesis that cell cycle regulators also impact immune surveillance and escape.

These studies underscore the heterogeneity and prognostic relevance of immune features in ACC. They highlight the therapeutic potential of integrating immune profiling with molecular classification to guide personalized treatment strategies. These findings are collectively sum- marized in Table 3, which outlines the immune landscape and check- point expression in ACC.

Fig. 6. Features of the tumor immune microenvironment in ACC. (a) High tumor-infiltrating mast cell (TIMC) abundance is associated with improved overall and progression-free survival in ACC. CIBERSORT-based immune profiling and IHC validation confirmed prognostic significance across TCGA and FUSCC cohorts [35]. (b) High BCLAF1 expression correlates with an immune-silent phenotype through downregulation of CD8+ T cell infiltration and upregulation of cell-cycle regulators [3]. (c) Immune landscape of ACC molecular subtypes. ACC1 shows strong immune activation and higher predicted sensitivity to PD-1 blockade, whereas ACC2 displays immunosuppressive infiltration and poorer prognosis [12]. (d) Associations between key prognostic markers (CDK1, ASPM, BIRC5, CCNB2) and immu- ne/stromal infiltration scores, highlighting immune-modulatory relevance of cell cycle regulators in ACC [37].

KEGG

Biological process

Cell cycle

(a)

(b)

Spliceosome

mitotic cell cycle phase transition

Steroid biosynthesis

chromosome segregation

Terpenoid backbone

Gene count

cell cycle G1/S phase transition

biosynthesis

Small cell lung cancer

50

regulation of cel cycle

-phase transition

MicroRNAs in cancer

100

organelle Sosion

Gene count

DNA replication-

150

spindle organization

200

Pancreatic cancer-

DNA conformation change

200

300

TIMChigh

12-

*** , ANOVA P=0.003

12-

cell cycle checkpoint





Oocyte meiosis-

400

Pathogenic Escherichia

250

regulation of chromosome organization

couples

TIMCs purity

9

TIMCs purity

protein-DNA complex

9

scoli infection

Chronic myeloid leukemia-

Pyrimidine metabolism-

negative regulation of cell cycle process .

Gene Ratio

Gene Ratio

organization involved in mitosis

2.4

6

..

E

-

Colorectal cancer

4

negative regulation of mitotic cell cycle

2.2

IRNA surveillance pathway-

processing in Protein Prox Celle In

3

protensomal protein catabolic process

2.0

3-

3

Adherens junction-

steroid metabolic process -

1.8

TIMClow

Endocytosis

2

response to radiation

0

0

IRNA processing

Stage I

Stage II

Stage Ill

Stage IV

T1-T2

T3-T4

10

47

40

M1

Necrosis™

Necrosis

Cellular senescence Human T-cell leukemia

positive regulation of cell cycle - establishment ce

-virus I infection

RNA transport

-maintenance of cell polarity”

DNA replication

0.00

0.01

0.02

FDR

0.03

0.04

De+00

5e-08

FDR

1e-05

TCGA

GSE10927

GSE10927

GSE90713

HALLMARK_MITOTIC_SPINDLE

Enrichment plet: HALLMARK_MITOTIC_SPINDLE

Enrichment plot: HALLMARK_G2M_CHECKPOINT

Enrichment plot: HALLMARK_Q3M_CHECKPOINT

Overall Survival

Overall Survival

E

22 22 2 22

TIMCslow

TIMCslow

Percent survival (%)

100

n=22

TIMCshigh

Percent survival (%)

100

+ :

n=10

TIMCshigh

*

50

50

-

-


n=17

n=4

p=0.019, HR(high)=3.258

p=0.018.

(high)=6.569

0

si Bclaf1

NCI-H295R

0

EV

FL

150-

GO/G1

S

All patients

0

20

40

60

80

Necrosisneg patients

100

0

20

40

60

80

I-

si NC

Time (months)

..

1.

% of total cells

- G2/M

Time (months)

-

4

R-

1.

100

8 -

1-

Progression-free Survival

Progression-free Survival

1

10

1.

J

-

50-

TIMCslow

TIMCslow

E-

1

9-

A

2.

«+

**

Percent survival (%)

100

n=22

TIMCshigh

Percent survival (%)

100

n=10

0

TIMCshigh

:

-

20

U

SINC

siBclaf1

6

-

.-

-

0

0

4

SW-13

50

50

第一

si NC

!-

si Bclaf1

EV

FL

150-

GO/G1

5

7

1 G2/M

n=17

n=4

R-

R-

I-

1.

% of total cells

100-

p=0.034, HR(high)=2.714

.030,

HR(high)=5.420

-

#

A

.

0

0

1

1

1

1

E

50

0

20

40

60

80

0

20

40

60

80

1

T

Time (months)

Time (months)

8-

R

3

0

7

SINC.

siBclaf1

#

T

E

-

(c)

(d)

Overall survival (ESTIMATE score)

Disease-free survival (ESTIMATE score)

Subtype # ACCI ACC2 ] ACC3

Subtype # ACCI # ACC2 ] ACC3

Subtype ] ACC1 # ACC2 # ACC3

ESTIMATEScore

High

Low

ESTIMATESONO

High

Low

ESTIMATE NOTE

ESTIMATE NOOTE

0.81

Anova, p = 1.9x 10-6

Anova, p = 1.2 x 10$

Anova, p = 6.5 x 10៛

1.00

1.00

Estimate Immune Score

0.4

Immune Cell Subsets

Immune Signaling Molecules

Survival probability

Survival probability

0.5

0.4

0.75

0.75

-2000

2900

0.50

0,50

0.0

0.0

.

0.0

BIRCS espression

0.25

p = 0.0039

0.25

p = 0.00015

2000

2000

:

0.00

0.00

1000

-0.4

-0.5

-0.4

0

1000

2000

3000

4000

6000

0

1000

2000

3000

4000

5000

:

ESTIMATEscore

ESTIMATEscore

ESTIMATE NON

ESTIMATE NONE

.

OS (days)

DFS (days)

·

Number at risk

Number at risk

ACC1

ACC2

ACC3

ACC1

ACC2

ACC3

-0.8

ACC1

ACC2

ACC3

sa

1

19

4

9

0

High

Low-

2

24

17

6

2

0

-2000

-2000

Subtype

Subtype

Subtype

0

1000

2000

3000

4000

5000

0

3000

OS (days)

1000

2000

DFS (days)

4000

5000

CcNB2 mpression

Low

Subtype Gender sender

1 05

Subtype

p

15

0

Laterality Gender Age

0

Subtype

-05

Overall survival (immune score)

Disease-free survival (immune score)

Stage

0.5

Stage

Central memory CD4 T cel

pra moed Plasmacytold dendritic

Cytotoxic cells ENK. meta

ADC3

Immune score

High

LOW

Immune score

High

Low

500

Mast cel

Laterality

Activated dendritic cel Macrophage

Immune noone

CYT

1.00

1.00

·

Treg cells

Right

MOSC

Survival probability

0.75

Survival probability

Immuno noone

500

500

Regulatory T cell

T cells

Gender

0.75

Activated CD& T cell

Monocyte

13 T-cell signature

FEMALE MALE

-1000

-1000

TLS

0.50

0,50

Effector memory CD4 T cel

Natural killer cell

WNTTGFB signature

Age

Type 1 T helped Cod T cel

B cell cluster

80

0.25

6 gene IFN signature

p = 0.071

0.25

Acvated BY

p = 0.049

··

:

Immature B cel Neutrochal

Macrophages

20

0.00

0.00

Type 17 T helper cel

MDSC

0

1000

2000

3000

4000

5000

0

1000

2000

3000

4000

5000

chobbnight natural kiler cell CD56dim killer

B.P. meta

Stage

Immature dendritic cell

TGFB1 activated

Immune score

05 (days)

DFS (days)

unknown stage Stage I

Number at risk

Immune score

Number at risk

-900-

Central memory CD8 T cel Memory B cell

TITR

CECM

Sa

20

32

14

8

9

0

18

12

1

9

0

Activated CD4 T cell

Stage IV

-1900

Chermna della T cell

CDB T cells

0

1000

2000

3000

4000

6000

0

OS (days)

1000

2000

3000

Natural kiler T cell

Th17 cells

DFS (days)

4000

5000

CONB2 expression

samples in TCGA-ACC cohort

samples in TCGA-ACC cohort

Overall survival (stromal score)

Disease-free survival (stromal score)

type

ACC1

ACC2

ACC3

ACC1

1

Stromal score

High

Low

Stromal score

High

Low

500

500

Stromal soure

1.00

1.00

Stromal noore

5

.

ACC2

0.8

000

0.75

0.75

-9000

-1000

Immunocyte Infiltration

Survival probability

Survival probability

4

ACC3

0.6

0.50

0.50

-1900-

-1906

3

ACC1

0.4

Low

ASIPM expression

0.25

0.25

1000

ACC2

p = 0.048

p < 0.0001

1000

2

0.2

0.00

0.00

ACC3

0

1000

2000

3000

4000

5000

0

1000

2000

3000

4000

5000

Stromal soore

Stromal NuONG

1

OS (days)

DFS (days)

pvalue

CTAL4-noR

CTLA4-R

PD1-noR

PD1-R

pvalue

Stromal score

Number at risk

Stromal score

Number at risk

500

0

Nominal p value

35

17

7

2

0

45

25

17

8

3

8

1000

PD-L1

PD-1

PD-L2

CTLA4

Bonferroni corrected

0

1000

2000

3000

4000

6000

0

1000

2000

3000

4000

5000

OS (days)

DFS (days)

-4500

Low

-15004

CCNB2 expression

Low

Table 3 Summary of immune features and checkpoint expression in ACC.
StudyMain FindingsImmune FactorsCheckpoint Expression
Chen et al.High-risk group shows tTregs, M2,Not analyzed
(2021)Tregs, 1 M2 macrophages, ĮCD8 +
[4]CD8 + T cells
Thampi et al.PD-L1 and CTLA-4General TMEPD-L1, CTLA-
(2020) [1]overexpressed in ACCoverview4
Zhou et al.BCLAF1 linked toĮ CD8 +, 1 Treg- associatedIndirect
(2021)immunosuppressive cell
[3]infiltration
Li et al.ACC subtypes defined byAPCs, NK, T cellsPD-L1, LAG3,
(2022) [30]immune gene signatures and checkpoint variationTIM3
Tian et al.TIMCs positively associated with survival; Į mast cells =1 CDK1, BUB1Mast cells, CD8 +, CDK1- linked genesNot analyzed
(2020) [35]
Zhou et al.BCLAF1 linked toCD8 + T cells 4,Indirect
(2021) [3]immunosuppressive TME; Į CD8 + T cells, t cell cycle genesTreg-associated ↑
Guan et al.ACC1: Immune-active;NK cells, Tregs,PD-1, PD-L1
(2022) [12]ACC2: Immune- suppressive; ACC1 sensitive to PD-1 blockademacrophages
Li et al.Hub genes (e.g., KIF11, KIF4A) linked to immune/ stromal scores; low immune score -> poor prognosisStromal cells, APCs, general immuneNot analyzed
(2020) [32]
Yi et al.CDK1, CCNB2, ASPMCD8 +, general immune infiltrationNot analyzed
(2022) [37]enriched in cell-cycle and immune evasion; linked to infiltration & prognosis
Zhang et al.VEGFA, CCL2, MMP9M2Indirect
(2023) [21]linked to immune regulation and metastasismacrophages, chemokine- expressing
Sung & CheongALT subtype shows distinct immune activation; TMM types linked to immune pathwaysGeneral immune activity by TMMNot analyzed
(2021) [17]
ArastonejadNCAPG, NCAPHImmuneNot analyzed
et al. (2024)overexpressed; associatedpathway-
[38]with poor survival and immune pathway enrichmentenriched genes
Li et al. (2023) [11]KIF20A, KIF22 marginally correlated with immune infiltrationKinesin family correlatedNot analyzed

5. Pediatric vs. adult ACC: comparative genomics

ACC presents as two biologically distinct diseases across age groups, with notable differences in genetic predisposition, transcriptional pro- grams, immune microenvironment, and clinical outcomes. Pediatric ACC frequently arises in the context of hereditary cancer syndromes, most prominently germline TP53 mutations, as exemplified by the well- recognized R337H founder variant in Southern Brazil, which drives early-onset tumorigenesis and underlies the marked geographic clus- tering and bimodal age distribution of disease [1]. In contrast, adult ACC is more commonly sporadic and genomically heterogeneous, involving alterations in TP53, CTNNB1, ZNRF3, MEN1, and aberrant activation of Wnt/ß-catenin and IGF2 signaling, which contribute to tumor progres- sion and therapeutic resistance (Fig. 7a) [5].

Comparative transcriptome profiling also reveals divergent regula- tory architectures between pediatric and adult tumors. Network-based analyses in pediatric ACC identified cell-cycle-associated hubs, including CDK1, CCNB1, BUB1B, and CDC20, as central effectors of proliferative signaling, emphasizing a developmental and mitotic reprogramming phenotype in childhood disease (Fig. 7a) [6]. In contrast, adult ACC tumors exhibit transcriptional patterns linked to

chromosomal instability, mitotic checkpoint deregulation, and DNA repair stress, with recent bioinformatics analysis highlighting the NCAP gene family (NCAPG, NCAPG2, NCAPH) as key mediators of chromo- somal condensation and as predictors of poor survival outcomes in adult patients (Fig. 7b) [38].

Differences also extend to the tumor immune microenvironment. Pediatric ACC is generally characterized by low immune infiltration and a less inflamed stromal context, whereas adult ACC displays stronger hypoxia-associated transcriptional signatures, immune checkpoint expression, and tumor-associated immunosuppression, reflecting more advanced immune escape and tumor-host adaptation dynamics (Fig. 7c) [4]. These contrasting immune phenotypes have direct implications for response to immunotherapies, which have shown limited benefit in unselected ACC populations. These findings indicate that pediatric and adult ACC represent distinct molecular and immunobiological entities, despite partially overlapping oncogenic drivers. Consequently, patient stratification strategies and therapeutic decision-making should not rely on a uniform framework, but instead incorporate age-specific genomic features, immune signatures, and progression-associated biomarkers. A comparative summary of these key distinctions is provided in Table 4.

6. Therapeutic targets and drug repositioning

Despite advances in molecular profiling, targeted therapy in ACC remains a formidable challenge due to the tumor’s inherent heteroge- neity and aggressive phenotype. Nevertheless, bioinformatic and multi- omics investigations have identified a growing list of potential thera- peutic targets and repositioned drugs with translational relevance.

Among the most frequently implicated targets are cell cycle regula- tors such as CDK1, TOP2A, AURKA, and CCNB1. These genes are not only consistently overexpressed in ACC but also strongly associated with poor overall survival [3,25,37]. High-throughput network analyses have shown that these proteins function as central hubs in mitotic regulatory pathways and PPI networks [5,30]. Notably, Zhou et al. [3] demon- strated that silencing BCLAF1, a transcription factor that modulates CDK1 and CCNB1, reduced proliferation in ACC cell lines and enhanced mitotane sensitivity, indicating its dual role in tumor growth and che- moresistance (Fig. 8a).

Another promising avenue is the identification of lipid metabolism- related enzymes such as SGPL1, FDFT1, and SQLE, which are signifi- cantly upregulated in ACC and correlate with worse prognosis. These targets are embedded in sphingolipid and steroid biosynthesis pathways, and their expression profiles, along with associated Kaplan-Meier sur- vival curves, are visualized in [18].

Drug repositioning strategies have also gained traction. For instance, Shahreza et al. [10] applied the Heter-LP algorithm to predict reposi- tioned drugs targeting ACC-specific molecules such as IGF1R, TOP2A, and MAP3K3, identifying cosyntropin and sorafenib as potential can- didates. Supporting this, Yi et al. [37] found that ASPM, BIRC5, CCNB2, and CDK1 were enriched across multiple repositioning analyses, showing strong associations with drugs like vorinostat, chlorpromazine, and trifluoperazine. Guan et al. [12] further classified ACC into molec- ular subtypes using multi-omics integration. Notably, the ACC2 subtype, characterized by high tumor mutation burden and poor prognosis showed greater sensitivity to chemotherapeutic agents such as doxoru- bicin, gemcitabine, and etoposide, while ACC1 patients responded bet- ter to PD-1 blockade therapy, highlighting the potential of stratified therapeutic targeting.

Recent insights into non-coding RNA networks, such as the BIRC5-miR-335-5p-PAX8-AS1 axis, have unveiled novel vulnerabilities in ACC (Subramanian et al. [14]). A significant proportion of research has converged on cell cycle regulators as critical therapeutic targets. Multiple studies have independently identified overexpression of genes such as CDK1, CCNB1, AURKA, TOP2A, and CEP55 in ACC tumors [2, 24,34]. These genes not only function as central hubs in PPI networks but also strongly correlate with poor prognosis. For instance, Xiao et al.

TCGA

Fig. 7. Multi-aspect overview of ACC features in pediatric vs. adult cohorts. (a) Venn diagram (top) comparing differentially expressed genes (DEGs) from pediatric ACC studies vs. normal or adenoma tissues [5,6], alongside a hub-gene network (bottom) highlighting overlapping upregulated genes (eg, CCNB1, TOP2A). (b) PPI network centered on NCAP family members in adult ACC, highlighting their association with cell cycle regulation and poor survival outcomes [38]. (c) Expression heatmaps of EFNA3, CCNA2, and COL5A1 in adult ACC samples across TCGA and GEO cohorts, stratified by hypoxia-related gene signature levels [4].

ACC versus NAT

(b)

type

3

2

EFNA3

1

0

-1

-2

119

CCNA2

-3

62

331

215

COL5A1

2942

1551

high

684

GEO

low

type

2

EFNA3

1

0

ACC versus ACA

ACC versus NAT

-1

CCNA2

-2

(a)

DNAJB1

BARD1

COL5A1

TUBB3

HSPA4

SKP2

CCK1

SIRT1

CCNB1

DICE

SLE15A4

SMC2

TOP2A

DNMTI

GADDISA

BRCA2

HECTD3

c

CDC20

CCNB2

ZWINT

TACC3

CONA2

NEK6

MLEIP

PRIM2

CENPF

NCAPG

CONE1

SMARCA4

CEPSS

MKI67

NCAPG2

NCAPH

BUB1

IARS

SMC4

DOX39A

BUB1B

TOP2A

NELK

TARS

MARS

NCAPD3

NCAPG

LMNB1

TCP1

TXN

EPRS

ENO1

HSQ

C14 orf1

PLEC

NDC80

Physical Interaction

Co-expression

CAT

Co-localization

HSPA5

LOS

AURKB

NCAPH2

Pathway

PRIMI

HSF2

Predicted

Shared protein domains

NCAPD2

SR

F1

OUR

[34] demonstrated in their PPI network analysis that CDK1 had the strongest association with low overall survival (HR = 11), followed by CEP55, CDKN3, and TOP2A. Similarly, Guo et al. [24] and Xing et al. [2] used STRING-based analyses to confirm that these genes are central to mitotic regulation and tumor proliferation. Additionally, Kaplan-Meier survival curves for these genes demonstrate their association with poor clinical outcomes.

Beyond cell cycle genes, emerging targets have been discovered through integrative multi-omics approaches. Tian et al. [33] constructed a proteomic-based prognostic model and identified TFRC, TSC1, and FASN as key protein markers with strong predictive power for survival

outcomes in ACC. Kaplan-Meier plots from their study clearly support the prognostic significance of these proteins, and additional expression heatmaps and pathway enrichment results further highlight their bio- logical relevance. These proteins may serve as both prognostic in- dicators and potential therapeutic targets.

Novel metabolic and neurotransmitter-related targets have also been highlighted. Knott and Leidenheimer [27] conducted a focused tran- scriptomic analysis of the GABAergic system in ACC, revealing that upregulation of ABAT, a key enzyme in the GABA shunt was signifi- cantly associated with favorable prognosis and reduced metastatic po- tential (Fig. 8b). They presented detailed Kaplan-Meier survival

Table 4 Key differences between pediatric and adult ACC based on comparative studies.
FeaturePediatric ACCAdult ACCReferences
CommonTP53 (R337H,TP53, CTNNB1,Thampi et al. (2020) [1]
Mutationsgermline)ZNRF3, MEN1 (somatic)Sun-Zhang et al. (2024) [5]
Hub GenesCDK1,NCAPG, TOP2A,Kulshrestha et al. (2016)
CCNB1,CEP55[6] Arastonejad et al. (2024) [38]
BUB1B
Immune FeaturesLow immuneHigh PD-L1, CTLA-4, hypoxia-related expressionChen et al. (2021) [4]
infiltration
PrognosisFavorable ifPoor with recurrence/ metastasisThampi et al. (2020) [1]
localized

analyses and transcript expression comparisons, indicating that ABAT may serve as a marker of better outcome and a potential entry point for repurposing neurological drugs. In a complementary systems biology approach, Shahreza et al. [10] applied the Heter-LP computational framework for drug repositioning and constructed a drug-target inter- action network specific to ACC. Their results predicted novel in- teractions between Cosyntropin and targets such as DHCR7, IGF1R, MAP3K3, and TOP2A, genes known to be overexpressed in ACC. The network graph provides a visual foundation for validating drug reposi- tioning hypotheses based on integrated biological knowledge graphs. Finally, telomere maintenance mechanisms (TMMs) may influence drug responsiveness in ACC. Sung and Cheong [17] demonstrated that pa- tients with non-defined TMM (NDTMM) had comparatively better sur- vival outcomes, suggesting a potential biomarker for stratifying patients in therapeutic trials (Fig. 8c). Their pan-cancer survival plots highlight how different TMM categories, including ALT, telomerase, and NDTMM, correlate with outcomes in ACC.

Yan et al. [36] developed a robust multi-gene prognostic model using machine learning approaches such as Lasso and Elastic Net, ultimately identifying six hub genes with strong prognostic potential (Fig. 8d). These genes, highlighted through WGCNA and survival analysis were linked to crucial oncogenic pathways such as cell cycle regulation and p53 signaling and were further validated by CNV and mutation analyses. Complementing this, Zhang et al. [21] focused on metastatic ACC and identified ten hub genes via PPI network analysis, among which CCL2, CDC20, DLGAP5, KIF2C, MMP9, and MYC demonstrated significant correlations with patient survival. These metastasis-related targets are involved in immune evasion, cell adhesion, and microtubule dynamics, reinforcing their relevance for drug development. The convergence of findings from both studies underscores a set of potential molecular targets, laying the groundwork for precision medicine strategies and the rational repurposing of drugs for advanced or metastatic ACC.

MiRNA-centered network analyses have also begun to highlight therapeutically actionable regulatory nodes in ACC. Omidi [39] identi- fied miR-507 and miR-665 as tumor-specific hub miRNAs within the ACC ceRNA network, with miR-507 regulating cell-cycle checkpoint machinery and Rho-GTPase signaling, suggesting its potential relevance to mitosis-targeted therapeutic strategies. Notably, miR-665 exhibited radiation-responsive expression patterns, indicating a potential role in modulating treatment sensitivity. These findings underscore the possi- bility that miRNA-centered network perturbation may represent a complementary therapeutic avenue alongside conventional IGF2, Wnt/ß-catenin, and steroidogenesis-directed approaches (Fig. 8e).

Given their functional involvement in cell-cycle control and lipid metabolic flux, Omidi [31] proposed CKS2 and ACAT2 as potential therapeutic intervention points, aligning with ongoing drug develop- ment efforts targeting CDK regulatory complexes and fatty-acid syn- thesis/metabolic enzymes in endocrine malignancies. Importantly, expression levels of both CKS2 and ACAT2 remained consistently elevated across patients receiving pharmaceutical therapy alone or in

combination with radiation, suggesting that their oncogenic activity is therapy-independent and may represent fundamental biological drivers rather than treatment-specific signatures (Fig. 8f).

In summary, the integration of gene expression, survival analysis, PPI networks, and drug repositioning frameworks has yielded a set of promising targets, some of which may be actionable with existing or repurposed compounds. Fig. 8 summarizes these emerging strategies across five thematic axes: gene silencing effects, metabolic vulnerabil- ities, computationally inferred drug-target relationships, regulatory non-coding RNA networks, and subtype-specific therapeutic responses. These findings are comprehensively summarized in Table 5, which outlines the key therapeutic targets and drug repositioning opportu- nities identified in ACC.

7. Multi-omics integration and systems biology in ACC

The integration of multi-omics data, such as genomics, tran- scriptomics, epigenomics, and proteomics has significantly advanced our systems-level understanding of ACC. These approaches help identify novel molecular signatures and reveal underlying mechanisms of tumorigenesis and progression that are not evident in single-omics analyses.

Sun-Zhang et al. [5] conducted a comprehensive genomic study combining TCGA and GEO data to identify 45 survival-associated genes in ACC, many of which showed protein-protein interactions previously unexplored in this context. Notably, POLD1, AURKA, and KIF23 formed central nodes in the PPI network, highlighting critical mitotic regulation pathways. Subramanian et al. [14] integrated multiple genomic datasets and constructed a miRNA-IncRNA-mRNA network, identifying BIRC5-hsa-miR-335-5p-PAX8-AS1 as a central axis strongly associated with poor survival in ACC. Their findings underline the clinical rele- vance of non-coding RNA in prognosis. Ye et al. [22] built a ceRNA network combining lncRNA, miRNA, and mRNA layers, revealing HOXA11-AS as a key lncRNA that regulates RRM2 and CDK1 through miR-24-3p/let-7a-5p. Their multi-layered network uncovered action- able ceRNA motifs with prognostic relevance.

Xu et al. [20] analyzed alternative splicing (AS) events in 92 ACC patients from TCGA, identifying 3919 AS events significantly associated with survival. Their AS-based risk score outperformed traditional tran- scriptomic models, emphasizing the potential of splicing as a systems biomarker layer. Arastonejad et al. [38] applied network-level modeling to evaluate the role of the NCAP family genes. NCAPG, NCAPG2, and NCAPH were identified as central regulators in mitotic chromosomal organization and correlated with worse survival outcomes, reinforcing their role in aggressive ACC biology (Fig. 9a). Li et al. [11] investigated the prognostic value of kinesin superfamily (KIF) genes using a bioin- formatics approach. KIF4A, KIF11, and KIF20A were overexpressed in ACC and associated with immune cell infiltration and worse prognosis, linking mitotic machinery to immune modulation.

Zhang et al. [21] conducted a meta-analysis of metastatic ACC and identified a hub gene network including MYC, CDC20, and VEGFA. Integration of DEGs, miRNAs, and functional enrichment revealed pathways involved in metastasis, such as phagosome formation and cell adhesion. Li et al. [32] used immune and stromal scoring algorithms to reveal 18 hub genes associated with the tumor microenvironment. Their risk model based on TME-derived gene expression predicted survival with high accuracy (AUC = 0.887), linking immune contexture to omics-based prognosis. Crona et al. [28] applied pan-cancer tran- scriptomic analysis to compare ACC with 35 tumor types. ACC was found to form a unique cluster but also shared molecular proximity with neural crest-derived tumors, suggesting lineage-based systems biology relevance.

Marquardt et al. [29] applied machine learning on transcriptomic data to classify ACC into two distinct clusters overlapping with known molecular subtypes (C1A and C1B). Their model identified novel prog- nostic markers like SOAT1 and EIF2A1, enhancing predictive

Fig. 8. Therapeutic targets and drug repositioning strategies in ACC. (a) BCLAF1 knockdown decreases CDK1 and Cyclin B1 protein levels in ACC cell lines, con- firming its regulatory role in G2/M cell cycle progression [3]. (b) Expression patterns of key GABAergic pathway genes in ACC, showing elevated ABAT levels relative to other pathway components, indicating metabolic heterogeneity across tumors [27]. (c) Survival differences among telomere maintenance mechanism (TMM) subtypes in ACC and other cancers. Patients with non-defined TMM (NDTMM) show more favorable prognosis compared to ALT- or telomerase-driven tumors [17]. (d) Prognostic 14-gene model constructed using machine learning methods (Model 2), demonstrating strong predictive performance [36]. (e) Treatment-responsive expression patterns of miR-507 and miR-665 in ACC under pharmaceutical therapy vs. radiation exposure [39]. (f) CKS2 and ACAT2 remain consistently upregulated across therapy subgroups in ACC, indicating therapy-independent oncogenic activity [31].

si NC

si Bclaf1

EV

FL

Relative protein levels

1.5

Relative protein levels

2.0

20

(b)

20

p = 3.0 x 10-21

1.0

1.5

1.0

15

p = 3.1 x 10-13

15

0.5

0.5

Cyclin B1

Cyclin B1

RSEM (log2)

10

RSEM (log2)

0.0

Bclaf1

CDK1

0.0

Bclaf1

CDK1

10

5

3

si NC

si Bclaf1

sh NC

sh Bclaf1

Relative protein levels

0

5

1.5

**

Relative protein levels

1.5

**

3

1.0

1.0

-5

0

T

T

GAD1

GAD2

ABAT

ALDH5A1

upregulated

unaltered

upregulated

unaltered

0.5

T

0.5

T

¥0.0

CDK1

Cyclin B1

0.0

Bclaf1

Bclaf1

CDK1

Cyclin B1

a

Overall survival on risk score calculated by model 2 (training set)

GAD1

ABAT

ACC

GBM

HNSC

Risk score

High risk Low risk

1.00

1.00

1.00

1.00

Survival probability

0.75

Survival probability

0.75

Survival probability

0.75

0.50

.0.50

2.0.50

0.75

0.25

p = 0.04

0.25

p = 0.043

0.25

Survival probability

p = 0.046

0.00

0.00

0.00

0.50

0

50

100

Time (Day)

150

0

20

40

Time (Day)

60

80

0

50

100

Time (Day)

150

200

(c)

PAAD

SARC

0.25

1.00

1.00

p < 0.0001

Survival probability

Survival probability

(d)

0.75

0.75

TEL+ALT

ALT

0.00

0.50

0.50

0

30

60

90

120

TEL

OS (months)

Risk score

Number at risk

0.25

p = 0.04

0.25

p = 0.0074

NDTMM

High risk Low

26

26

11

16

BLON

0

5

0 2

0

30

60

90

120

(e)

0.00

0.00

OS (months)

0

20

40

Time (Day)

60

80

0

50

100

Time (Day)

150

200

1.0

0.4

A Normalized Expression (Pharma - Radiation)

2.5

Pharmaceutical

0.8

2.0

Radiation

0.2

1.5

miR-507

log10(RPM)

Sensitivity

0.6

0.0

1.0

0.5

0.4

-0.2

miR-665

0.0

-0.5

0.2

-0.4

AUC at 1 year: 0.92

miR-507

miR-665

0.0

AUC at 3 years: 0.91

AUC at 5 years: 0.95

Relative Expression Level per Therapy Group

0.0

0.2

0.4

0.6

0.8

1.0

Mean Expression (TPM)

1-Specificity

600

Pharma+Radiation

Pharma only

400

All Cases

(f)

200

IIT

0

ERP44

DEPDC1

CENPI

LSM5

SUB1

BRIX1

STC2

ACAT2

CKS2

IDO1

PLK4

LYAR

CHEK1

HOMER1

GOLGA7B

MELK

CDYL2

RAB3B

ZNF562

GJC1

NHLRC1

MRPL42

C4orf46

HSBP1

Table 5 Promising therapeutic targets and repositioned drug candidates in ACC.
Target GeneRole in ACCRepurposed DrugMechanism/InteractionEvidence Source
SGPL1Sphingolipid metabolism, cell survival, steroidogenesisFingolimodInhibits sphingosine-1-phosphate lyase (SGPL1), suppressing downstream sphingolipid signalingSubramanian et al. (2022) [18]
BIRC5Inhibits apoptosis; part of dysregulatedYM155Direct inhibitor of BIRC5 (survivin),Subramanian
mRNA-miRNA-IncRNA network associated with poor survivalinduces apoptosis in cancer cellset al. (2023) [14]
hsa-miR-335-5pDownregulated miRNA; modulates expression of key hub genes (e.g., BIRC5), associated with metastasis and poor survivalNot directly listed; therapeutic strategy involves miRNA mimic or restorationmiR-335-5p suppresses BIRC5 and other oncogenic targets; therapeutic potential in restoring tumor suppressionSubramanian et al. (2023) [14]
PAX8-AS1Dysregulated lncRNA; interacts with miR-Not specified; potential futureModulates miR-335-5p levels andSubramanian
335-5p in regulatory network affecting gene expressiontargetdownstream oncogenic geneset al. (2023) [14]
AURKAOverexpressed; associated with poor survival; strong interaction with TP53; key regulator of mitosis and cell cycleAURKA inhibitors (e.g., alisertib)Inhibition of AURKA leads to cell cycle arrest and increased radiosensitivity, particularly in TP53-mutant tumorsSun-Zhang et al. (2024) [5]
POLD1 CCL2Overexpressed; independently associated with poor survival; involved in DNA replication and repairPotential immunotherapy target (clinical trial: NCT03810339)POLD1 mutation/dysregulation associated with improved immunotherapy response across solid tumorsSun-Zhang et al. (2024) [5]
Overexpression associated with better survival; functions in immune cell recruitment and modulation of tumor microenvironmentBindaritSelective CCL2 synthesis inhibitor; reduces monocyte/macrophage infiltration, potentially affecting tumor progressionZhang et al. (2023) [21]
CDC20, DLGAP5, KIF2C, MMP9, MYCOverexpression associated with poor survival; involved in cell cycle regulation, proliferation, and metastasisIndirectly targetable; potential inhibitors include CDK inhibitors, MYC-targeted therapies, MMP inhibitorsModulate tumor cell proliferation, mitotic progression, extracellular matrix remodelingZhang et al. (2023) [21]
miR-21-3p, miR-21-5p, miR-451a, let-7a-5p, let- 7e-5pDysregulated miRNAs modulating expression of hub genes (e.g., MYC, MMP9, VEGFA, CCL2); associated with metastasis and prognosismiRNA-based therapeutics (e.g., anti-miR-21, miRNA mimics)Regulation of gene expression through post-transcriptional repression; impact angiogenesis, immune evasion, metastasisZhang et al. (2023) [21]
PDCD1 (PD-1), CD274 (PD-L1)Highly expressed in ACC1 subtype with high immune infiltration; potential immune checkpoint markers; associated with better response to immunotherapyAnti-PD-1 agents (e.g., nivolumab, pembrolizumab)Blocking PD-1/PD-L1 pathway restores T- cell function and enhances anti-tumor immunityGuan et al. (2022) [12]
ACC2-subtype-related proliferation genes (not individually listed)Associated with high tumor-mutation burden, DNA repair activation, cell cycle dysregulation; subtype ACC2 shows poorest prognosisCisplatin, doxorubicin, gemcitabine, etoposideHigher sensitivity of ACC2 subtype to these chemotherapy agents due to underlying molecular vulnerabilitiesGuan et al. (2022) [12]
ACC3-subtype markers (not individually listed)Enriched in steroid and cholesterol biosynthesis pathways; moderate prognosisAnti-CTLA-4 agents (e.g., ipilimumab)ACC3 subtype shows relative sensitivity to anti-CTLA-4 therapy via immunophenotyping (SubMap analysis)Guan et al. (2022) [12]
BCLAF1Overexpressed in ACC; promotes proliferation and cell cycle progression via regulation of CDK1 and Cyclin B1; associated with poor prognosisCDK1 inhibitors (e.g., RO-3306, dinaciclib) - indirect targeting via BCLAF1 regulation of CDK1BCLAF1 directly regulates transcription and protein levels of CDK1 and CCNB1; downregulation of BCLAF1 reduces proliferation and induces G2/M cell cycle arrestZhou et al. (2021) [3]
CCNB1, NDC80Highly overexpressed in ACC; associated with poor prognosis; central regulators of mitosis and cell cycle progressionCDK inhibitors (e.g., flavopiridol, dinaciclib) - targeting downstream cell cycle pathwaysDisruption of CCNB1-CDK1 complex or NDC80-mediated mitotic spindle assembly halts cell cycle and tumor proliferationLi et al. (2022) [30]
ASPM, AURKA, CCNB2, CDC20, CENPA, EXO1, FBXO5, HJURP, KIF2C, MKI67, NUF2, PARPBP, TACC3, TROAPOverexpressed; all 14 genes were part of the best prognostic model and significantly associated with poor survival; involved in cell cycle regulation, mitotic progression, DNA replication, and chromosomal stabilityCDK inhibitors (e.g., dinaciclib, flavopiridol), AURKA inhibitors (e. g., alisertib), PARP inhibitors (for PARPBP interaction)Disruption of cell cycle and mitotic regulators leads to growth arrest and apoptosis; high copy number variations (CNVs) and overexpression of these genes enhance proliferation and worsen prognosisYan et al. (2021) [36]
CDK1Key regulator of cell cycle; promotes mitosis and ACC proliferation; associated with poor prognosisCDK1 inhibitors (e.g., RO-3306, flavopiridol)Inhibition of CDK1 leads to G2/M arrest and apoptosis; proposed therapeutic strategy in combination with MELK inhibitorsYin et al. (2023) [25]
CCNB1, CCNA2Overexpressed in ACC; regulate cell cycle progression and mitosis; associated with advanced tumor stage and poor prognosisCDK inhibitors (e.g., dinaciclib, flavopiridol)Targeting cyclin-CDK complexes can suppress proliferation and improve survival in ACCYin et al. (2023) [25]
AURKAOverexpressed; regulates mitotic progression and is associated with TP53 mutation and poor outcomeAURKA inhibitors (e.g., alisertib, MLN8054)Inhibition of AURKA induces mitotic failure, apoptosis, and reduces tumor growth; dual targeting with Wnt/ ß-catenin suggestedYin et al. (2023) [25]
TOP2AOverexpressed; promotes tumor progression; associated with poor survival and drug resistanceEtoposide, doxorubicin (TOP2A- targeting agents)Inhibition of TOP2A disrupts DNA topology and replication, inducing apoptosisYin et al. (2023) [25]
BIRC5Inhibits apoptosis; associated with poor prognosis in ACC and other cancersYM155 (survivin inhibitor)Suppresses BIRC5 expression and restores apoptotic signaling in tumor cellsYin et al. (2023) [25]

(continued on next page)

Table 5 (continued)
Target GeneRole in ACCRepurposed DrugMechanism/InteractionEvidence Source
ASPMOverexpressed; associated with poor OS and DFS; regulates mitosis and tumor progressionLucanthone, Monobenzone, 8-Aza- guanine, Thioguanosine, ResveratrolEnriched through CMap and DSigDB; drugs predicted to reverse ASPM-driven oncogenic signatureYi et al. (2022) [37]
BIRC5Inhibits apoptosis; high expression linked to poor OS and DFS; affects immune infiltrationDasatinib, Lucanthone, Monobenzone, 8-Azaguanine, Thioguanosine, ResveratrolTargeted through multiple compounds from DSigDB and CMap; involved in cell cycle controlYi et al. (2022) [37]
CCNB2Promotes cell cycle progression; overexpression linked to poor survival and advanced stageDasatinib, Lucanthone, Monobenzone, 8-Azaguanine, Thioguanosine, ResveratrolCell cycle-associated gene enriched for inhibitors disrupting mitotic checkpointsYi et al. (2022)
[37]
CDK1 ABATDrives mitotic entry; highly expressed in ACC; poor prognostic indicator Upregulated in ~40 % of ACC tumors; strongly associated with improved overall and progression-free survival; linked to reduced metastasis, lower cortisol levels, and favorable metabolic phenotypeDasatinib, Lucanthone, Monobenzone, 8-Azaguanine, Thioguanosine, Resveratrol Vigabatrin (Sabril®)Cell cycle arrest induced by CDK1 inhibition; multiple compounds suggested via enrichment analysisYi et al. (2022) [37]
Irreversible inhibitor of ABAT; suppresses brain metastasis in preclinical models; modulates GABA shunt activity, mitochondrial metabolism, and immune- related pathwaysKnott & Leidenheimer (2020) [27]
GABRB2Upregulated in ~36 % of tumors; positively correlated with ABAT expression and favorable survival outcomes; may form functional GABA A receptors in ACC cellsBrexanolone®, Klonipin®, Solfoton® (positive allosteric modulators of GABA A receptors)Enhances GABA A receptor activity; potential to modulate proliferation and immune signalingKnott & Leidenheimer (2020) [27]
GABRDUpregulated in ~15 % of tumors; negatively correlated with ABAT; associated with poor survival and increased recurrenceNone specified; potential candidate for modulation via GABA A receptor-targeting agentsEnhances inhibitory signaling; high expression may reflect suppressed ABAT activity and aggressive tumor phenotypeKnott & Leidenheimer (2020) [27]
DNMT1Upregulation associated with hypermethylation and suppression of ABAT expression; poor prognostic factorDemethylating agents (e.g., decitabine, azacytidine)Reverses methylation-driven silencing of ABAT; potentially restores favorable metabolic and immune phenotypeKnott & Leidenheimer (2020) [27]
CBX3, NRF1, EP300, NFYBMaster transcriptional regulators enriched in ACC with ALT subtype; associated with poor prognosis and elevated mitochondrial biogenesis, CNV, and interferon-gamma productionNo specific drug reported; suggested potential for TMM- targeted therapy in ALT-ACCRegulate expression of peptide secretion and mitochondrial pathways; influence prognosis through modulation of telomere maintenance mechanism (TMM)Sung & Cheong (2021) [17]
CCNB1, CDK1, TOP2A, CCNA2, CDKN3, MAD2L1, RACGAP1, BUB1, CCNB2All genes significantly upregulated in ACC; involved in mitotic spindle checkpoint, chromosome segregation, and cell cycle regulation; high expression correlated with worse overall survival; BUB1 was identified as an independent prognostic factorCDK inhibitors (e.g., dinaciclib, flavopiridol), TOP2A inhibitors (e. g., etoposide, doxorubicin), AURKA/B inhibitors (for BUB1- related mitotic checkpoint modulation)Disruption of mitotic checkpoint and cell cycle progression through inhibition of CDKs, topoisomerase IIx, and checkpoint regulators; potential for targeted suppression of highly proliferative ACC cellsGuo et al. (2020) [24]
FASN, FN (Fibronectin), TFRC, TSC1Overexpressed in high-risk ACC group; independently associated with poor prognosis; part of the IPRPs model (Integrated Prognosis-Related Proteins); higher expression linked to shorter overall survivala) FASN -> TVB-2640, C75, Orli- stat (FASN inhibitors) b) TFRC -> Anti-TFRC antibodies (e.g., ch128.1Av), ferritin-based nanoparticles c) TSC1 -> Targeted indirectly via mTOR inhibitors (e.g., rapamycin, everolimus)a) FASN promotes lipid synthesis and tumor growthTian et al. (2021) [33]
b) TFRC involved in iron uptake, supporting tumor proliferation c) TSC1 suppresses mTORC1 pathway; its overexpression dysregulates growth signals
TOP2ASignificantly overexpressed in ACC; associated with poor overall and disease-free survival; promotes proliferation and invasivenessAclarubicinInhibits topoisomerase II alpha activity, disrupting DNA topology and replication; induces apoptosis in tumor cellsXiao et al. (2018) [34]
NDC80, CEP55, CDKN3, CDK1All significantly overexpressed and associated with advanced pathology stages and poor prognosis; involved in mitotic progression, chromosome segregation, and cell cycle regulationa) CDK1 - CDK inhibitors (e.g., RO-3306, flavopiridol) b) CEP55 - Potentially targetable via PI3K/AKT pathway modulationa) CDK1: Inhibition induces G2/M arrest b) CEP55: Regulates PI3K pathway and metastasis-related signals (e.g., FOXM1, MMP-2)Xiao et al. (2018) [34]
AURKA, TYMS, GINS1,Overexpressed in ACC; significantly associated with poor overall and disease-free survival; involved in cell division, DNA replication, mitotic checkpoint, chromosomal stability, and angiogenesisa) AURKA -> Alisertib b) TYMS - 5-FU, Pemetrexed c) RRM2 - Triapine, Didox d) EZH2 -> Tazemetostat e) CDK1, CCNB1 - Flavopiridol, RO-3306 f) TOP2A-related genes (e.g., TPX2) -> Etoposidea) AURKA: Inhibition blocks mitotic spindle formation b) TYMS: Inhibition disrupts DNA synthesis c) EZH2: Epigenetic silencing via histone methylation; inhibition reactivates tumor suppressors d) CDK1/CCNB1: Cell cycle arrest at G2/ MXing et al.
RACGAP1, RRM2, EZH2, ZWINT, CDK1, CCNB1, NCAPG, TPX2(2019) [2]
DHCR7, IGF1R, MC1R, MAP3K3, TOP2APredicted disease-related targets by Heter- LP; IGF1R and TOP2A previously linked to ACC; others (DHCR7, MC1R, MAP3K3) show plausible association with adrenal signaling and tumor phenotypeCosyntropinCosyntropin stimulates adrenal cortex function; proposed as adjunct to mitotane for adrenal support; identified as novel ACC drug via Heter-LP methodLotfi Shahreza et al. (2020) [10]
hsa:1584 (mitotane target)Existing mitotane target; validates network prediction accuracySpironolactoneIdentified as additional drug interacting with mitotane's known target (hsa:1584); potential ACC therapy candidateLotfi Shahreza et al. (2020) [10]
hsa:2230Predicted novel target related to flavin adenine dinucleotide; no prior literatureFlavin adenine dinucleotideInteraction suggested by Heter-LP; role in mitochondrial metabolism and energyLotfi Shahreza et al. (2020)
links; proposed for experimental follow-uppathways[10]

(continued on next page)

Table 5 (continued)
Target GeneRole in ACCRepurposed DrugMechanism/InteractionEvidence Source
miR-507Tumor-suppressive regulatory hub; maintains mitotic checkpoint integrity; higher expression correlates with improved survivalRestoration-based strategies (miRNA mimic delivery; epigenetic re-activation; nanoparticle miRNA replacement)Suppresses oncogenic targets involved in cell cycle progression, mitotic spindle assembly, Rho-GTPase signaling, and genomic stability regulationOmidi (2025) [39]
miR-665Context-dependent regulatory miRNA; downregulated in ACC; displays radiation- responsive dynamicsTreatment-sensitization approach (miR-665 expression monitoring to stratify radiotherapy response)Involved in immune-signaling modulation, endothelial remodeling, and treatment-associated stress response signalingOmidi (2025) [39]
CKS2Promotes cell-cycle acceleration and mitotic progression; associated with poor survivalCDK inhibitors (e.g., Palbociclib, Ribociclib)CKS2 enhances CDK complex activity, facilitating G1/S and G2/M transitions; CDK inhibitors can reduce proliferative drive in CKS2-high tumorsOmidi (2025) [31]
ACAT2Drives lipid metabolic reprogramming and supports tumor energy productionFatty-acid synthesis inhibitors (e.g., TVB-2640 - FASN inhibitor; Etomoxir - CPT1 inhibitor)ACAT2 fuels lipid-derived acetyl-CoA and B-oxidation, supporting tumor growth; metabolic inhibitors may starve high- ACAT2 tumorsOmidi (2025) [31]

capabilities of omics-driven classifiers. Guan et al. [12] applied the MOVICS algorithm to multi-omics datasets (mRNA, lncRNA, miRNA, somatic mutations, methylation) from TCGA and GEO, categorizing ACC into three distinct molecular subtypes. These subtypes exhibited sig- nificant differences in prognosis, immune infiltration, and drug sensi- tivity, thereby guiding personalized treatment decisions (Fig. 9b). Li et al. [30] employed a comprehensive multi-omics machine learning framework integrating gene expression, mutation, and immune land- scape data. Their approach identified CEP55, CENPF, and BIRC5 as robust biomarkers across multiple platforms, capable of distinguishing immune-high and immune-low ACC subtypes.

Yan et al. [36] constructed and validated multi-gene prognostic models using WGCNA and machine learning on transcriptomic data. Six hub genes including MKI67 were identified, which were linked to copy number variations and survival outcomes (Fig. 9c). Yi et al. [37] expanded this concept by integrating WGCNA with immune infiltration and drug prediction, pinpointing ASPM, BIRC5, CCNB2, and CDK1 as multi-omics prognostic markers enriched in the cell cycle and responsive to small-molecule drugs. Chen et al. [4] introduced a hypoxia-related transcriptomic signature that independently predicted prognosis and correlated with immune cell infiltration patterns. High-risk patients had distinct immune suppressive profiles, showcasing the value of inte- grating tumor microenvironmental features. Di Dalmazi et al. [16] performed RNA-Seq on ACC and adenoma samples, demonstrating how transcriptomic clustering aligns with mutation profiles. Their results suggest the existence of transitional states between adenoma and car- cinoma, supporting a continuum model of tumorigenesis.

Tömböl et al. [15] performed the first miRNA-mRNA co-profiling in ACC. Using tissue-specific miRNA-target predictions, they identified regulatory networks linked to G2/M checkpoint control, revealing novel insights into miRNA-mediated pathogenesis. Guo et al. [24] combined multiple gene expression datasets to identify 9 hub genes enriched in cell cycle and senescence pathways, showing strong correlation with poor survival in ACC patients. Tian et al. [33] proposed a protein-based prognostic model (IPRPs) incorporating data from RPPA and gene expression profiles. Their model outperformed traditional biomarkers like Ki-67 and was validated across three independent cohorts.

Yuan et al. [23] conducted co-expression and PPI network analysis on GEO and TCGA data, revealing 12 hub genes involved in mitotic progression and strongly associated with ACC prognosis. These findings support the central role of mitotic dysregulation in ACC pathophysi- ology (Fig. 9d). Recent systems-level ceRNA network analyses have highlighted tumor-specific regulatory rewiring in ACC. Omidi [39] identified miR-507 and miR-665 as central hub miRNAs within the ACC transcriptomic network, demonstrating strong differential network centrality and prognostic relevance, particularly for miR-507. This study underscores the value of multi-omics-integrated network modeling for

discovering non-coding RNA biomarkers beyond standard differential expression (Fig. 9e).

These studies illustrate that multi-omics integration and systems biology approaches in ACC are unveiling complex regulatory networks and convergent biomarkers, paving the way for precision oncology in this rare malignancy. A comparative overview of the included multi- omics studies is provided in Table 6.

8. Discussion

8.1. Challenges and future directions

Although ACC is well-characterized as a rare and clinically aggres- sive malignancy with a bimodal age distribution and poor prognosis [40-42], it continues to present major therapeutic and molecular chal- lenges. Surgical resection, currently the only curative option offers limited efficacy, with high recurrence rates and 5-year survival ranging from 16 % to 47 % [42]. In parallel, significant advances in the molec- ular understanding of ACC have not yet translated into meaningful clinical benefit. While the omics revolution has enabled deep profiling of ACC at genetic, epigenetic, transcriptomic, and immunologic levels, several structural and methodological barriers continue to hinder diagnostic refinement, therapeutic innovation, and biomarker valida- tion [43,44].

One of the most pressing challenges is the rarity of ACC, which severely limits access to large, well-annotated patient cohorts. Most current studies rely on TCGA and ENSAT datasets that, while valuable, comprise fewer than 100 tumor samples. This restricts statistical power, limits the ability to capture population heterogeneity, and complicates the identification of robust prognostic subgroups [5,30]. Moreover, pediatric ACC remains drastically underrepresented, rendering cross-age comparisons dependent on indirect inference [1,6]. Moving forward, the establishment of international biobanks and cross-institutional data harmonization will be essential to overcome these limitations.

A second challenge is the persistent disconnect between computa- tional prediction and experimental validation. Numerous studies have proposed gene signatures using algorithms such as LASSO regression [2], PPI-based hub analysis [24,37], or ceRNA network construction [22], yet relatively few of these findings have been validated by RT-qPCR, immunohistochemistry, or functional assays. In some cases, proposed biomarkers (e.g., BIRC5, CCNB1) recur across studies, but are rarely tested in independent patient cohorts. This bottleneck highlights the urgent need for integrated pipelines that combine in silico discovery with in vitro and in vivo validation. This need for integrated computa- tional and experimental workflows is supported by recent studies that combined in silico co-expression analyses with RT-qPCR validation to

Fig. 9. Multi-omics integration and system-level modeling in ACC. (a) Functional enrichment and PPI network of NCAPG, NCAPG2, and NCAPH. (A) GO analysis highlights their roles in chromosomal segregation and condensation (q-value < 0.01). (B) The PPI network reveals interactions with key mitotic regulators (e.g., SMC2, AURKB) [38]. (b) Molecular subtyping of ACC using integrated DNA methylation, mRNA, miRNA, and lncRNA data, revealing three subtypes with distinct survival and pathway signatures [12]. (c) Performance evaluation of the multi-gene prognostic model in TCGA (training) and GEO (validation) cohorts, including calibration curves, decision curve analysis, ROC/AUC evaluation, and Kaplan-Meier survival stratification, confirming strong predictive accuracy and clinical applicability [36]. (d) Co-expression module analysis identifying hub genes enriched in mitotic regulation and associated with poor prognosis [23]. (e) Tumor-specific ceRNA network structure in ACC showing miR-507 and miR-665 as central regulatory hubs [39].

Gene Ontology

regulation of chromosome

Overall survival

Disease specific survival

segregation

regulation of chromosome

separation

100

100

chromosome separation

positive regulation of

8

Survival probability (%)

Survival probability (%)

chromosome segregation

75

75

positive regulation of

chromosome separation

regulation of chromosome

condensation

p.adjust

50

50

condensed chromosome

0.01

0.02

nuclear chromosome

0.00

25

Overall p < 0.001

25

Overall p < 0.001

condensed nuclear chromosome

0.04

8

ACC1

ACC2

-ACC1

ACC1

ACC2

ACC2

+ACC1

chromosomal region

ACC2

<0.001

ACC3

ACC2

00.001

ACC2

Count

0

ACC3

0.008

<0.001

0

ACC3

0.014

<0.001

ACC3

DNA packaging complex

5

0

12

24

36

48

60

72

84

96

108

120

0

12

24

36

48

60

2

84

96

108

120

chromosome, centromeric region

10

Time (Months)

Time (Months)

Number at risk

Number at risk

single-stranded DNA binding

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

Q

magnesium ion binding

24

22

20

16

12

8

5

3

2

2

2

23

21

19

15

11

8

5

3

2

2

2

DNA-directed 5’-3’ RNA

S

0

12

24

36

48

60

72

84

96

108

120

0

12

24

36

48

80

72

84

96

108

120

polymerase activity

Time (Months)

Time (Months)

5’-3’ RNA polymerase activity

0

a

RNA polymerase activity

Disease free interval

Progression free interval

·

100

100

0.2

GeneRatio

0.4

0.6

Survival probability (%)

Survival probability (%)

TOP2A

75

75

NCAPH2

gulation of chromosome segregation

ACC1

ACC2

egulation of chrom

condensation

ACC2 <0.001

Overall p < 0.001

NEK6

50

ACC3 0.007

0.089

regulation of chvo

me separation

50

ACC1

ACC2

MKI67

chromosome separation

ACC3

category

25

Overall p < 0.001

25

chromosome separation

+ACC1

ACC1

positive regulation of chromosome segregation

ACC2

ACC2

ACC2 <0.001

positive population of chromosome segregation

positive regulation of chromosome separation

ACC3

NCAPD3

q

ACC3

<0.001

0.002

regulation of chromosome condensation

- regulation of chromasame organization

0

12

24

36

48

60

72

84

96

108

120

0

12

24

36

48

60

72

84

96

120

regulation of chromosome segregation

Time

108

(Months)

Time (Months)

CENPF

regulation of chromosome separation

Number at risk

Number at risk

sister chromatid segregation

28 27 24

17

13

13

9

7

5

4

2

33

31

28

18

14

13

10

7

5

4

2

size

4

4

1

0

0

0

0

0

0

0

Q

9

13

1

O

Q

O

0

O

5.0

11

10

7

5

5

2

1

1

1

1

21

O

0

Q

24

16

12

8

6

5

2

1

1

1

1

NDC80

Valet chromatid segregali

7.5

0

12

24

36

48

80

72

120

72

10.0

(b)

84

Time (Months)

96

108

0

12

24

36

48

60

84

96

108

Time (Months)

120

12.5

NCAPH

Module membership vs. gene significance cor=0.79, p=1.7e-191

AURKB

SMC2

NCAPD2

ZH

ER CO

E

KA

IN

H

%

NCAPG2

Q

&

D

c

00

NCAPG

SMC4

0℃

ND

NTO

C

L

ME

8

5

F

DE

ÉP

POCO

N

FUC

C

F1

Gene significance for stage

CN

U

DC

T

NO

VN

VD

UF

2

2

2

TIM

N

OD

NO

OC

G

0

TO

KJ

0

0

=

Q

0

OC

M

DR

S

C

N

3

3

c

PX

F

BE

®

MAN

F

0

:

CZ

FC

B

UE

R

0

¥

OTO

SAP

S

:

e

EL

2

A

NL

MC

F

EK

3

a

IN

68

S

:

CN

LK

D

EN

EN

CI

CN

OL

CE

®

8

F

0

8

3

=

3

SE

0.0

0.2

0.4

0.6

0.8

Q#

3.7

a.

..

LE

Q.

am

18

a3

As

am

4.

-

AP

2

Module Membership in turquoise module

==

==

==

0.05 0.10 0.15 0.20 0.25 0.30

hub genes in co-expression network

(d)

DEGs in common datasets

-

1

With risk score Wehcut risk

1

With risk somne

0

With The score Without risk 800ne

3

One-way ANOVA tumor grade/stage

0

Net Benefit

&

ROC curve

0

Net Benefit

Net Benefit

3

(AUC ≥ 0.85)

0

1

O

1

6

1

0

8

0

0

6

18

60

63

0

6

(12.8%)

(42.6%)

(44.7%)

0

12

-0.05

8

0

2

10

3

D.DO

0.05

0.10

0.15

0.0

0.1

0.2

0.4

0.5

D.D

0.1

02

0.4

0.5

2

Threshold Probability

Threshold Probability

Threshold Probability

6

3

0

0

Confidence intervals

Overall survival curve (TOGA-ACC)

0

0

1 200

hub genes in

protein-protein interaction network

0

1

N

1.00

Survival analysis

-

(OSE19750 and TOGA ACC)

Pearson’s correlation of tumor grade/stage

8

0.75

Size of each lat

Semmitvity (%)

Survival probability

8

V

150

AUC: 80.1% (70.2%-09.0%)

un

ma

0.25

P < 0.0001

·

0.00

Kao VOS DUTCH ING

AUG

0

1000

2000

3000

8

pe AUC 6442

OS (days)

4000

5000

Nomogram

Number at risk

0

C-index = 0.872

5

12

10

4

0

0

1

1000

4000

100

80

60

40

20

000

0

OS (days)

9000

Specificity (%)

Confidence intervals

0

Overall survival curve (GSE19750)

hsa-mir-665

1-2

8

1.00

Survival probability

0.75

Sernitivity (%)

, 8

.50

AUC: 86.4% (72.6%-100.0%

5

0.25

0.00

miRNA

hsa-mir-507

2

AUG

0

2000

05 (days)

4000

6000

Nomogram

mRNA

Number at risk

-

C-index = 0,882

8

2

E

2

2

Q

000

100

BO

60

40

20

0

Speciforty(%)

2000

OS (days)

6000

Specificity (%)

e)

Table 6 Table 6 (continued)
Summary of multi-omics findings and their implications in ACC.StudyOmics Layers UsedKey FindingsContribution to ACC Research
StudyOmics LayersKey FindingsContribution to ACC
UsedResearchXu et al. (2020) [20] Arastonejad et al. (2024) [38]Alternative splicing (AS) transcriptome data + clinicalIdentified 3919First comprehensive
SubramanianGene expression (mRNA)Identified a novel regulatory networkSystems-level insight intosurvival-associated AS events across seven AS types (e.g., ES, RI, AA,multi-omics study
et al.focusing on splicing
(2023)+ miRNA(BIRC5-miR-ncRNA-mRNAregulation in ACC;
[14]+ lncRNA335-5p-PAX8-AS1) strongly associated with poor overall survival; uncovereddysregulation; prioritization of novel multi-omic biomarkers anddata + splicing factors (SFs) expression profilesetc.); built eight prognostic signatures using Lasso and Cox regression with AUChighlighted the prognostic significance of AS patterns and SFs;
12 hub genes, 5 miRNAs, and 4therapeutic targets> 0.9; established AS-SF regulatoryintroduced novel candidate
lncRNAs through integrative network analysisnetwork involving hub AS events and SFs such as EIF3A andtherapeutic targets based on splicing dysregulation
Sun-ZhangmRNAIdentified 45 genesProvided a robustHNRNPR
et al. (2024) [5]expression (RNA-Seq), DNA somatic mutations (WES), clinical survival datawhose expression and mutation status significantly correlate with survival; constructed a multi- parameter genemulti-omics framework for stratifying ACC patients by survival risk; highlighted new prognosticTranscriptome (mRNA expression microarray), survival data, pathological staging, PPI network, GO enrichmentIdentified 3 NCAP family genes (NCAPG, NCAPG2, NCAPH) as significantly overexpressed in ACC; high expression linkedIntroduced NCAP family genes as novel multi-omics biomarkers and diagnostic/ prognostic
signature integrating expression, mutation, and survival associations; revealed novel protein-protein interactions (e.g., AURKA-POLD1, KIF23-TP53)markers and therapeutic candidates via systems-level interaction analysisto poor survival and advanced stages; constructed a 23-node PPI network; GO enrichment revealed roles in chromosomal segregation and condensation; ROCindicators in ACC; provided a systems- level view of chromosomal regulation in ACC pathogenesis and precision diagnostics
Ye et al. (2020)Transcriptome (mRNA)Identified 9 key hub genes (e.g., CDK1,Provided a multi- layer ceRNALi et al. (2020) [32] Crona et al. (2018) [28] Marquardt et al. (2021) [29]analysis showed strong discriminatory
[22]+ miRNA + lncRNARRM2), 4 prognostic miRNAs (miR- 212-3p, miR-24-3p,regulatory network highlighting ncRNA-mediatedTranscriptome (mRNA expression from TCGA), immune and stromal cell estimation (ESTIMATE algorithm), PPI network, survival datapower (AUC > 0.85)
Identified 1122 intersect DEGs based on immune and stromal scores; selected 18 hub genes via PPI network (e.g., CD4, HLA-DRA, HCK, PTPRB); lower immune scores correlated with worse survival andIntegrated TME profiling with transcriptomic data; constructed a prognostic gene signature with high predictive accuracy (AUC =0.887); highlighted immunological
let-7a-5p, miR-196a- 5p), and 1 lncRNA (HOXA11-AS) via integrated analysis; constructed ceRNA sub-networks like RRM2-miR-24-3p/ let-7a-5p-HOXA11- AS associated withdysregulation in ACC; proposed novel prognostic biomarkers and therapeutic targets based on multi- omics crosstalk
dysregulation in
poor prognosisrecurrence; immune-ACC and proposed
Li et al. (2023)Transcriptome (mRNAIdentified 4 key KIF genes (KIF4A, KIF11,Provided integrative systemsrelated pathways such as MHC class II, T-cellnovel immune- related biomarkers
[11]expression), clinical data, PPI network, immune cell infiltration analysisKIF20A, KIF22) significantly upregulated in ACC and associated with advanced stage andbiology perspective on kinesin superfamily in ACC; proposed KIFs as novel biomarkers and potential therapeutic targets by combining transcriptomic, immunologic, and network-level evidenceTranscriptome (RNA-seq) from TCGA and TARGET; unsupervised clustering; principal component analysis (PCA); consensus clusteringreceptor signaling, and cytokine signaling enriched ACC forms a homogenous transcriptomic cluster across 3319 pan- cancer samples; unexpectedly clusters near neural crest tumors (GBM, LGG, NBL, PNET, PPGL); identified 78 discriminative transcripts, althoughProvided transcriptomic evidence that ACC has closer molecular proximity to neural crest tumors than previously understood; opened new perspectives for classifying ACC beyond traditional
poor survival; constructed gene-gene and protein-protein networks; linked KIFs to immune infiltration and functional pathways like mitosis and antigen presentation
Zhang et al. (2023)Transcriptome (mRNA) + miRNA + PPI networks + clinicalIdentified 10 hub genes (e.g., CDC20,Provided a systems- level map of metastatic mechanisms in ACC; integratedno strong shared GO signatures between ACC and neural crest tumorsclinicopathological criteria; suggested
[21]DLGAP5, KIF2C, MYC) and 5 key miRNAs (e.g., miR-potential for extrapolating therapeutic
survival data21-3p, let-7a-5p) linked to ACC metastasis; built a DEmiR-hub genetranscriptomic, regulatory, and survival data to prioritize multi-strategies from gliomas and neuroblastomas to ACC
network; showed prognostic impact of several genes and miRNAs on patient survivalomic biomarkers for metastasis and prognosisTranscriptome (RNA-seq from TCGA and ENSAT), mutational data, supervisedIdentified two robust transcriptomic clusters (ACC-UMAP1 and ACC-UMAP2) matching known C1A/C1B subtypes;Offered an unbiased machine-learning- based stratification of ACC; validated clusters across
datasets; proposed
(continued on next page)
Table 6 (continued)Table 6 (continued)
StudyOmics Layers UsedKey FindingsContribution to ACC ResearchStudyOmics Layers UsedKey FindingsContribution to ACC Research
Guan et al. (2022) [12]machine learning (Random Forest), UMAP clustering100 differentially expressed genes identified via RF (e.g., SLC2A1, SOAT1, EIF2S1, MYC,novel biomarker candidates for prognosis and therapy based on transcriptomic signatures and mutational correlation Provided one of the mostmiR-mRNA target prediction + pathway analysis (IPA, GSEA)introduced a novel tissue-specific miR target prediction strategy integrating inverse expressionfunctional miR target discovery; revealed post- transcriptional regulatory mechanisms driving ACC pathogenesis; proposed miRs as both diagnostic biomarkers and potential
FSCN1); ACC-UMAP2 linked to poorer survival and higher mutation rates in TP53 and CTNNB1patterns and filtering; found G2/M DNA damage checkpoint as the top dysregulated pathway; validated
mRNAIdentified three robustmiR-503 and miR-511
+ lncRNAACC molecularas potentialmodulators of the
+ miRNAsubtypes (ACC1-3) using 10 multi-omics clustering algorithms; ACC2 had worst survival and highest mutation burden; ACC1 showed strongest immune activation and response to anti-PD-1 therapy; distinct metabolic and DNA repair pathways enriched in subtypescomprehensive multi-omics classifications in ACC; enabled precise therapeutic stratification and prediction of immunotherapy and chemotherapy response; established a novel and clinically actionable molecular taxonomyYan et al. (2021) [36]biomarkers distinguishing ACCcell cycle machinery
+ DNA methylation + somatic mutations + clinical data
Transcriptome (from TCGA and GEO), survival data, CNV and mutation data, PPI network, machine learning models (Ridge, Elastic Net, Lasso),from benign tumors (AUC~100 % sens/ spec) Constructed and validated 11 multi- gene prognostic models; selected a 14- gene signature (e.g., ASPM, AURKA, CCNB2, CDC20, MKI67) with strong predictive power (AUC > 0.84, C-indexProvided a robust, externally validated, multi- omics-based prognostic model; demonstrated the clinical utility of nomogram and machine-learning- based prediction;
Li et al. (2022) [30] Di Dalmazi et al. (2020) [16] Tömböl et al. (2009) [15]TranscriptomeIdentified 490 DEGsDelivered an integrated multi-GSEA= 0.851); Model 2offered a panel of novel ACC
(from GEO and(28 upregulated, 462chosen as best
TCGA), PPI network,downregulated); 17 hub genes (e.g.,omics pipeline combining transcriptomics, interactome, pathway biology, and survival/ mutation analysis; provided two robust diagnostic/ prognostic biomarkers for potential clinical translation Delivered a multi- layer integrative analysis revealing the interplay between transcriptome, mutation status, and steroidogenicYuan et al. (2018) [23]Transcriptome (microarray) + clinical data (tumor grade, survival) + WGCNAperforming; CNVs and mutations in these genes significantly influenced mRNA expression and patient survival; six hub genes identified via PPI network Identified 12 real hub genes (ANLN, ASPM, CDCA5, CENPF, FOXM1, KIAA0101, MELK, NDC80, PRC1, RACGAP1, SPAG5, TPX2) through a stepwise validation pipeline; allbiomarkers strongly associated with survival and actionable for precision medicine First study applying WGCNA to ACC; provided a comprehensive systems-level identification and validation of robust
pathway enrichment (KEGG, Reactome, PANTHER, BioCyc), survival data (GEPIA), mutationalCCNB1, NDC80, TPX2, FOXM1, KIF11, RRM2) were strongly associated with poor prognosis; CCNB1 and NDC80 selected as top core genes through multi-step prioritization (Degree, MCC, MNC); enriched in cell cycle and p53 signaling
analysis
(cBioPortal)
pathways+ PPI network + validation in
TranscriptomeIdentified distinctprognostic biomarkers; suggested hub genes as candidate drivers of proliferation and disease progression in ACC
(RNA-seq)transcriptomicTCGA, GEO,
+ lncRNAs + gene fusion detection + somatic mutationclusters among ACAs and ACCs, correlating with hormonal activity and driver mutations (e.g., PRKACA, GNAS,GEPIAassociated with tumor grade and poor prognosis; enriched in cell cycle, DNA replication, and mitotic pathways
analysis (WES &
RNA-seq)CTNNB1, TP53); ACC samples showedphenotype; provided new pathogenetic insights including gene fusions and lncRNA profilesYi et al. (2022) [37]Transcriptome (RNA-seq + microarray), WGCNA, DEG analysis, PPI network,Identified 9 hub genes (e.g., ASPM, BIRC5, CCNB2, CDK1, TOP2A, FOXM1); 4 MPBs (ASPM, BIRC5, CCNB2, CDK1) withProvided a robust integrative pipeline combining
distinct expression profiles, enriched gene fusions, and lncRNA
transcriptomics, machine learning, and immune
miRNA (miR) expressiondownregulation; genes like TOP2A and CCNB1 were selectively overexpressed in ACCs; fusion events (e.g., AKAP13-PDE8A) and altered steroidogenesis pathways observed Identified six miRs with significantdistinguishing benign from malignant tumors Delivered the first integrative miR-mRNA analysis in ACC; introduced a new method forsurvival data, mutation & CNV data, drug signature enrichment, immune infiltration analysisstrong predictive power for OS (AUCs > 0.8 for 1-, 3-, 5-year survival); MPBs correlated with CNVs, immune infiltration, tumor grade, and stage; drug signatures like vorinostat, trifluoperazine, and alpha-estradiol predicted to target MPBsanalysis; proposed clinically applicable nomograms for survival prediction; revealed multi- omics-based MPBs as promising diagnostic, prognostic, and therapeutic targets in ACC
profiling + mRNAexpression differences in ACC (e.g., miR-Guo et al. (2020) [24]Transcriptome (three GEO datasets), GO/ KEGG functionalIdentified 200 consensus DEGs across GSE10927, GSE12368, andDelivered an integrative cross- cohort
transcriptomics5031, miR-1841, miR-
+ tissue-specific5114, miR-2144);transcriptomic (continued on next page)
Table 6 (continued)
StudyOmics Layers UsedKey FindingsContribution to ACC Research
annotation, PPI network, TCGA survival dataGSE90713; nine hub genes (CCNB1, CDK1, TOP2A, CCNA2, CDKN3, MAD2L1, RACGAP1, BUB1, CCNB2) enriched in cell cycle and p53 pathways; all hub genes significantly upregulated and associated with poor overall survival; BUB1 identified as independent prognostic factoranalysis for biomarker discovery in ACC; validated multi- gene panel associated with mitotic dysregulation and adverse prognosis; highlighted the utility of combining public datasets for robust gene signature identification
Tian et al. (2021)Proteomics (RPPA from TCPA), transcriptome (TCGA, GEO),Developed a prognostic modelFirst large-scale proteogenomic analysis in ACC; introduced a robust multi-omics-based
[33](IPRPs) based on 4 key proteins: FASN, Fibronectin (FN),
real-world IHC data, clinical dataTFRC, TSC1; validated in TCGA, GEO (GSE10927), and FUSCC cohorts; model outperformed Ki-67 and ß-catenin in prognostic accuracy (AUC = 0.933); hub genes (e.g., CCNB1, CDK1, AURKA, PLK1) linked to cell cycle pathways; GSEA showed enrichment in chromosome separation and metaphase-anaphase transitionprognostic model validated across cohorts; provided protein-level biomarkers (not just mRNA), reinforcing translational potential for clinical application and precision medicine
Chen et al. (2021) [4]Transcriptome (RNA-seq), hypoxia-related gene sets, immune infiltration (CIBERSORT), survival dataDeveloped a 3-gene hypoxia-associated signature (CCNA2, EFNA3, COL5A1) predictive of overall survival; high-risk score correlated with worse prognosis, immune suppression (e.g., lower PDL1 and CTLA4), and increased resting NK cellsProvided a clinically actionable hypoxia-based prognostic model; linked hypoxia signaling to immune microenvironment remodeling; validated performance across TCGA and GEO cohorts with high AUC values (0.949-0.871)
Omidi (2025) [39]Gene expression (mRNA) + miRNA expression (multi-dataset) + network topology analysisIdentified miR-507 and miR-665 as tumor-specific regulatory hub miRNAs in ACC; revealed regulatory network rewiring between normal and tumor states; demonstrated miR- 507 association with improved survival and miR-665 as radiation- responsiveProvides systems- level multi-omics insight into miRNA- centered regulatory architecture; prioritizes miR-507 and miR-665 as clinically relevant network biomarkers with diagnostic and therapeutic relevance in ACC
Omidi (2025) [31]mRNA expression (TCGA-ACC + GTEx)Identified CKS2 and ACAT2 as co- upregulated, transcriptionally co-Provides systems- level evidence for a synergistic proliferative-
integrated with regulatory network topology and clinicalregulated oncogenic drivers; demonstrated reinforced tumor- specific regulatory coupling and strongmetabolic oncogenic axis, highlighting clinically actionable biomarker
Table 6 (continued)
StudyOmics Layers UsedKey FindingsContribution to ACC Research
survival/prognostic and diagnostic performance (AUC~ 0.90)candidates and network-defined therapeutic vulnerabilities in ACC
diagnostic
modeling

confirm IncRNA-oncogene regulatory interactions in colorectal cancer [53].

Therapeutic translation faces additional barriers. Mitotane remains the only FDA-approved drug for ACC, despite limited efficacy and sub- stantial toxicity. Recent computational repositioning studies have nominated agents such as Cosyntropin, Linsitinib, and CDK1 inhibitors [3,10,25], but no repositioned drug has yet entered advanced clinical trials for ACC. Moreover, few preclinical models exist to test drug effi- cacy in relevant tumor systems, and systems pharmacology validation remains largely unexplored. Bridging this translational gap will require preclinical screening of prioritized candidates in xenograft or organoid models, as well as biomarker-informed patient stratification in trial designs.

Another significant obstacle is the fragmentation of omics data. While transcriptomics is widely used in ACC research, integration with proteomics, epigenomics, copy number variation, and immune land- scapes is rare and often lacks standardized workflows. Promising studies have demonstrated the power of multi-omics integration in stratifying ACC subtypes and identifying convergent markers [4,30], yet such ap- proaches are not yet the norm. Multi-cohort, multi-layered frameworks will be critical for defining clinically meaningful subgroups and pre- dicting therapy response with higher accuracy.

Additionally, systems-level interpretation of gene-gene interactions remains in its infancy. Most existing analyses rely on linear pathways or static PPI networks, which fail to capture temporal dynamics or feed- back mechanisms. Emerging evidence on NCAP family genes [38], TPX2/FOXM1 circuits [37], and ceRNA-mRNA-IncRNA crosstalk [22] suggests that combinatorial and non-linear regulatory logic underlies aggressive ACC phenotypes. To model this complexity, future studies should incorporate dynamic systems biology approaches using time-series, perturbation-based, or single-cell datasets.

The immune landscape of ACC also remains incompletely charac- terized. While immune subtyping [12] and TME-associated gene panels [21,32] have advanced our understanding, most computational models lack integration of checkpoint markers, immune suppression signatures, or therapy response predictors. Given the limited success of immune checkpoint inhibitors in ACC so far, personalized immunogenomic classifiers are urgently needed to optimize patient selection and com- bination therapies.

Finally, open science and data-sharing practices must be strength- ened to accelerate collective progress. Although initiatives such as ENSAT and A5 have provided foundational resources, widespread public access to raw sequencing data, clinical metadata, and reproducible pipelines remains limited. Harmonized data formats, metadata stan- dards, and open-access repositories would greatly improve collabora- tion, benchmarking, and cross-cohort validation.

In conclusion, ACC research stands at the threshold of a new era driven by bioinformatics, systems biology, and translational innovation. To fully realize this potential, the field must bridge existing methodo- logical gaps, validate predictions through experimental pipelines, and expand collaborative infrastructure. These coordinated efforts will be essential to move from descriptive omics to actionable insights that improve diagnosis, prognosis, and therapy in ACC. A schematic sum- mary of these key challenges and proposed solutions is presented in Fig. 10, illustrating the translational bottlenecks across biomarker dis- covery, therapeutic development, and systems-level integration.

Fig. 10. Schematic summary of major challenges and future strategic directions in ACC research. Key bottlenecks include limited patient cohorts, insufficient biomarker validation, incomplete therapeutic translation, fragmented multi- omics integration, and static modeling frameworks. Proposed strategic solu- tions encompass multi-institutional biobanking, integrated in silico-wet-lab validation pipelines, ACC-specific preclinical model systems, comprehensive multi-omics integration, and dynamic systems biology frameworks. Arrows illustrate the translational progression from current limitations toward action- able and clinically relevant research solutions.

Challenges

Strategic Directions

Rare Datasets

Multi-Institutional Biobanks

Prediction Validation Gap

Integrated In Silico-Wet-Lab Pipelines

Limited Therapy Transition

ACC-Specific Preclinical Models

Fragmented Omics

Multi-Omics Integration

Static Modeling

Dynamic Systems Biology Frameworks

8.2. Multi-omics integration in clinical decision-making

Despite extensive molecular characterization efforts, the trans- lational impact of multi-omics discoveries in ACC remains limited. However, evidence from recent studies underscores the potential of in- tegrated transcriptomic, epigenetic, and immunogenomic approaches to refine patient stratification and guide individualized therapeutic stra- tegies. For example, lncRNA signatures have demonstrated robust ca- pacity to delineate tumor subtypes and correlate with clinically meaningful phenotypes, suggesting that non-coding RNA networks may function as stable biomarkers in precision oncology frameworks [54]. Similarly, miRNA-based regulatory axes such as miR-489-mediated suppression of oncogenic signaling highlight the feasibility of leveraging post-transcriptional modulators to enhance therapeutic responsiveness and overcome malignant phenotypes [55]. These find- ings reinforce the premise that non-coding RNA-centered multi-omics integration can yield clinically actionable biomarker panels for risk prediction and treatment tailoring in ACC. Similar findings have been reported for other oncogenic lncRNAs, where coordinated dysregulation across transcriptomic datasets and validation in clinical tissues has confirmed their role in tumor progression and metastatic behavior [56].

In parallel, growing evidence indicates that the tumor microenvi- ronment, particularly macrophage polarization dynamics plays a deci- sive role in modulating therapeutic resistance and shaping clinical outcomes. Non-coding RNAs have been shown to regulate macrophage phenotype transitions and immune evasion circuits, suggesting that incorporation of immune-omics layers into ACC profiling may improve prediction of immunotherapy responsiveness [57]. Furthermore, trans- lational pipelines involving siRNA-based gene targeting and nanoparticle-enabled delivery systems illustrate practical strategies to

overcome biological instability, systemic degradation, and off-target constraints that currently limit the therapeutic implementation of RNA-directed interventions [58]. However, successful clinical trans- lation will require standardized delivery platforms, validated pharma- cokinetic profiles, and controlled evaluation in preclinical ACC-specific disease models.

Finally, insights from hematologic malignancies demonstrate that lncRNAs can function as diagnostic and prognostic determinants and can predict therapy resistance, supporting their inclusion in multi-layered biomarker frameworks designed for real-world clinical decision- making [59]. Integrating these lessons into ACC research may facili- tate the transition from descriptive genomic catalogs to clinically deployable precision medicine tools, particularly if coupled with coor- dinated prospective validation efforts, inclusion of pediatric ACC co- horts, and harmonized biorepository infrastructure.

8.3. Methodological limitations and future validation needs

Current findings should be interpreted considering several method- ological constraints. The rarity of ACC limits the availability and di- versity of well-annotated patient cohorts, which may introduce sampling bias and restrict the statistical power needed to resolve clini- cally meaningful subgroups or molecular phenotypes. This challenge has been consistently emphasized in recent genomic and multi-omics char- acterizations of ACC, where small cohort sizes constrained the gener- alizability of inferred regulatory programs and subtype-specific alterations [39,60]. Moreover, although integrated transcriptomic and regulatory analyses offer valuable hypotheses regarding oncogenic drivers and miRNA-mediated network remodeling, retrospective in sil- ico inference cannot fully recapitulate dynamic regulatory states, tumor-microenvironment interactions, or treatment-induced adaptation [61,62]. To ensure translational rigor, the candidate biomarkers and regulatory modules proposed here will require multi-layered experi- mental validation, including RT-qPCR and immunohistochemical confirmation in independent ACC cohorts, followed by functional perturbation assays (e.g., siRNA or CRISPR-mediated silencing) to establish mechanistic relevance [63]. In addition, evaluation in patient-derived organoid and xenograft systems will be necessary to determine phenotypic consequences and therapeutic exploitability [64]. Ultimately, prospective, biomarker-stratified clinical studies are essen- tial to determine whether these molecular signatures can reliably inform risk classification and guide targeted or immunomodulatory treatment decisions in real-world clinical settings.

9. Conclusion

This review integrates emerging molecular, transcriptomic, and immunogenomic insights to provide a comprehensive overview of the current state of ACC research. Studies have repeatedly identified key oncogenic drivers, such as CDK1, CCNB1, FOXM1, and BIRC5 as being central to tumor progression and poor prognosis [2,3,37]. The increasing recognition of non-coding RNAs, including HOXA11-AS and LINC00271, has further expanded the landscape of potential biomarkers and regulatory axes in ACC [13,22]. Recent ceRNA- and network-based studies have also highlighted synergistic dual oncogenic pathways involving CKS2 and ACAT2, as well as the emergence of miR-507 and miR-665 as central regulatory nodes in ACC, providing refined molec- ular stratification and prognostic insight [31,39].

Moreover, the integration of bioinformatics and systems biology has enabled deeper insights into the complex interaction networks within ACC tumors. Multi-omics clustering approaches have successfully stratified patients into molecular subtypes with distinct survival out- comes and therapeutic vulnerabilities [12,29]. These integrative models, when combined with immune profiling and protein-based risk signatures, provide new opportunities for patient-specific management strategies [32,33].

Drug repositioning has emerged as a promising avenue for expanding therapeutic options in ACC, with computational studies identifying agents such as Linsitinib, Roscovitine, and Cosyntropin that may target ACC-specific vulnerabilities [4,10]. However, translation into clinical application remains limited, underscoring the need for functional vali- dation and ACC-specific preclinical models.

By synthesizing recent discoveries from molecular, transcriptomic, and immunogenomic studies, this review highlights both the progress and the gaps in ACC research. Continued development of integrative pipelines, validation strategies, and cross-institutional collaborations will be critical in transforming molecular discoveries into clinically actionable tools. In this regard, systems-level and multi-omics frame- works hold particular promise for shaping the next generation of di- agnostics and therapeutics in ACC. However, realizing this potential will require coordinated multi-center biobanking efforts, ACC-specific organoid and xenograft model development, and prospective biomarker-stratified clinical trial designs, to bridge the remaining translational gap between computational discovery and real-world pa- tient care.

CRediT authorship contribution statement

Javad Omidi: Writing - review & editing, Writing - original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization.

Declaration of Generative AI and AI-assisted technologies in the writing process

During the preparation of this work the author used ChatGPT (OpenAI) in order to improve English language and readability. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References

[1] Shwetha Thampi, R.Kumaravel Veena, H.Maxwell Jessica, Adrenocortical carcinoma: a literature review, Curr. Probl. Cancer 6 (2020) 100519.

[2] Xing Kai, Yuquan Ji, Pengfei Liu, Chunxiao Liu, Xiaorong Li, Jiahui Wan, Screening and identification of key biomarkers in adrenocortical carcinoma based on bioinformatics analysis, Oncol. Lett. 18 (5) (2019) 4667-4676.

[3] Hui Zhou, Mengchen Zou, Xinyi Ding, Fei Zou, Role of Bclaf1 in promoting adrenocortical carcinoma proliferation: a study combining the use of bioinformatics and molecular events, Cancer Manag. Res. 13 (2021) 6785-6795.

[4] Qi Chen, Fang Mingzhi, Yang Hongjian, Lin Jiatian, He Min, A hypoxia signature for predicting prognosis and tumor immune microenvironment in adrenocortical carcinoma, J. Oncol. 2021 (2021) 5512504.

[5] Alexander Sun-Zhang, C.C. Juhlin, T. Carling, U. Scholl, M. Schott, C. Larsson, S. Bajalica-Lagercrantz, Comprehensive genomic analysis of adrenocortical carcinoma reveals genetic profiles associated with patient survival, ESMO Open 9 (7) (2024).

[6] Aashish Kulshrestha, P.Sahu Aditya, Vikas K. Verma, Rashmi K. Ambasta, Pradeep K. Dwivedi, Network analysis reveals potential markers for pediatric adrenocortical carcinoma, Oncol. Lett. 12 (6) (2016) 4667-4674.

[7] Abdullah M. Alshabi, Xiang Li, Xiang Ling, Xing Qin, Lei Bao, Identification of invasion and proliferation related genes in adrenocortical carcinoma microenvironment based on bioinformatic methods, Aging 11 (23) (2019) 11157-11177.

[8] Gustavo G. Fernandez-Ranvier, Orlo H. Clark, Wen T. Lin, Julie A. Duh, Patricia H. Kebebew, James S. Ginzinger, Electron Kebebew, Identification of biomarkers of adrenocortical carcinoma using genomewide gene expression profiling, Arch. Surg. 143 (8) (2008) 841-846.

[9] Giordano, J. Thomas, Dafydd G. Thomas, Rork Kuick, Michelle Lizyness, David E. Misek, Angela L. Smith, Donita Sanders, et al., Distinct transcriptional profiles of adrenocortical tumors uncovered by DNA microarray analysis, Am. J. Pathol. 162 (2) (2003) 521-531.

[10] Mohammad L. Shahreza, Mahboobeh Jafari, Somayyeh Zolghadr, Ali Masoudi- Nejad, A computational drug repositioning method applied to rare diseases: adrenocortical carcinoma, Sci. Rep. 10 (1) (2020) 16569.

[11] Xiao Li, Yanghao Tai, Shuying Liu, Yating Gao, Kaining Zhang, Jierong Yin, Huijuan Zhang, Xia Wang, Xiaofei Li, Dongfeng Zhang, Bioinformatics analysis: relationship between adrenocortical carcinoma and KIFs, Biotechnol. Genet. Eng. Rev. 39 (2) (2023) 575-585.

[12] Yu Guan, Shaoyu Yue, Yiding Chen, Yuetian Pan, Lingxuan An, Hexi Du, Chaozhao Liang, Molecular cluster mining of adrenocortical carcinoma via multi- omics data analysis aids precise clinical therapy, Cells 11 (2022) 3784.

[13] Floryne O. Buishand, Yi Liu-Chittenden, Yu Fan, Amit Tirosh, Sudheer K. Gara, Dhaval Patel, Daoud Meerzaman, Electron Kebebew, Adrenocortical tumors have a distinct, long, non-coding RNA expression profile and LINC00271 is downregulated in malignancy, Surgery 167 (2020) 224-232.

[14] Chitra Subramanian, Reid McCallister, Mark S. Cohen, Multi-genomic analysis of 260 adrenocortical cancer patient tumors identifies novel network BIRC5-hsa-miR- 335-5p-PAX8-AS1 strongly associated with poor survival, Surgery 173 (2023) 43-51.

[15] Zsófia Tömböl, Peter M. Szabó, Viktor Molnár, Zoltán Wiener, Gergely Tölgyesi, János Horányi, Peter Riesz, Peter Reismann, Attila Patócs, István Likó, Rolf- Christian Gaillard, András Falus, Károly Rácz, Peter Igaz, Integrative molecular bioinformatics study of human adrenocortical tumors: microRNA, tissue-specific target prediction, and pathway analysis, Endocr. Relat. Cancer 16 (2009) 895-906.

[16] Guido Di Dalmazi, Claus Scholz Barbara Altieri, Silviu Sbiera, Michaela luconi, jens waldman, darko kastelan, filippo ceccato, iacopo chiodini, giorgio arnaldi, anna riester, andrea osswald, felix beuschlein, sascha sauer, martin fassnacht, silke appenzeller, and cristina 1. ronchi, rna sequencing and somatic mutation status of adrenocortical tumors: novel pathogenetic insights, J. Clin. Endocrinol. Metab. 105 12 (2020) e4459-e4473.

[17] Ji-Yong Sung, Jae-Ho Cheong, Pan-cancer analysis of clinical relevance via telomere maintenance mechanism, Int. J. Mol. Sci. 22 (20) (2021) 11101.

[18] Chitra Subramanian, Mark S. Cohen, Identification of novel lipid metabolic biomarkers associated with poor adrenocortical carcinoma prognosis using integrated bioinformatics, Surgery 171 (2022) 119-129.

[19] Fangshi Xu, Yibing Guan, Yubo Ma, Li Xue, Peng Zhang, Xiaojie Yang, Tie Chong, Bioinformatic ANalyses and Experimental Validation of the Role of M6a Rna Methylation Regulators in Progression and Prognosis of Adrenocortical Carcinoma, Aging 13 (8) (2021) 11919-11936.

[20] Ning Xu, Zhi-Bin Ke, Xiao-Dan Lin, Fei Lin, Shao-Hao Chen, Yu-Peng Wu, Ye- Hui Chen, Yong Wei, Qing-Shui Zheng, Identification of survival-associated alternative splicing events and signatures in adrenocortical carcinoma based on tega spliceseq data, Aging 12 (6) (2020) 4996-5009.

[21] Yiyi Zhang, Ying Zhu, Nie Tang, Integrative bioinformatics analysis reveals the key molecular players in metastatic adrenocortical carcinoma, Res. Sq. (2023).

[22] Senlin Ye, Haohui Wang, Wei Li, Lu Yi, Identification of mRNA-miRNA-lncRNA competing endogenous RNA network in adrenocortical carcinoma using bioinformatics analysis, Res. Sq. (2020).

[23] Yuan, Guofeng Qian Lushun, Liang Chen, Chin-Lee Wu, Han C. Dan, Yu Xiao, Xinghuan Wang, Co-expression network analysis of biomarkers for adrenocortical carcinoma, Front. Genet. 9 (2018) 328.

[24] Jinshuai Guo, Yinzhong Gu, Xiaoyu Ma, Lu Zhang, Huimin Li, Zhongyi Yan, Yali Han, Longxiang Xie, Xiangqian Guo, Identification of hub genes and pathways in adrenocortical carcinoma by integrated bioinformatic analysis, J. Cell. Mol. Med. 24 (2020) 4428-4438.

[25] ] Mengsha Yin, Yao Wang, Xinhua Ren, Mingyue Han, Shanshan Li, Ruishuang Liang, Guixia Wang, Xiaokun Gang, Identification of key genes and pathways in adrenocortical carcinoma: evidence from bioinformatic analysis, Front. Endocrinol. 14 (2023) 1250033.

[26] Fangshi Xu, Peng Zhang, Miao Yuan, Xiaojie Yang, Tie Chong, Bioinformatic screening and identification of downregulated hub genes in adrenocortical carcinoma, Exp. Ther. Med. 20 (2020) 2730-2742.

[27] ] Erika L. Knott, Nancy J. Leidenheimer, A targeted bioinformatics assessment of adrenocortical carcinoma reveals prognostic implications of gaba system gene expression, Int. J. Mol. Sci. 21 (22) (2020) 8485.

[28] I Joakim Crona, Samuel Backman, Staffan Welin, David Taïeb, Per Hellman, Peter Stålberg, Britt Skogseid, Karel Pacak, RNA-sequencing analysis of adrenocortical carcinoma, pheochromocytoma and paraganglioma from a pan- cancer perspective, Cancers 10 (12) (2018) 518.

[29] André Marquardt, Laura-Sophie Landwehr, Cristina L. Ronchi, Guido Di Dalmazi, Anna Riester, Philip Kollmannsberger, Barbara Altieri, Martin Fassnacht, Silviu Sbiera, Identifying new potential biomarkers in adrenocortical tumors based on mRNA expression data using machine learning, Cancers 13 (18) (2021) 4671.

[30] Xiunan Li, Jiayi Li, Leizuo Zhao, Zicheng Wang, Peizhi Zhang, Yingkun Xu, Guangzhen Wu, Comprehensive multiomics analysis reveals potential diagnostic and prognostic biomarkers in adrenal cortical carcinoma, Comput. Math. Methods Med. 2022 (2022) 2465598 (Article ID).

[31] Javad Omidi, Synergistic dual oncogenic role of CKS2 and ACAT2: enhanced regulatory coherence and biomarker potential in adrenocortical carcinoma, Cancer Genet. (2025), https://doi.org/10.1016/j.cancergen.2025.10.103.

[32] Xiao Li, Yang Gao, Zicheng Xu, Zheng Zhang, Yuxiao Zheng, Feng Qi, Identification of prognostic genes in adrenocortical carcinoma microenvironment based on bioinformatic methods, Cancer Med. 9 (3) (2020) 1161-1172.

[33] Xi Tian, Wen-Hao Xu, Aihetaimujiang Anwaier, Hong-Kai Wang, Fang-Ning Wan, Da-Long Cao, Wen-Jie Luo, Guo-Hai Shi, Yuan-Yuan Qu, Hai-Liang Zhang, Ding- Wei Ye, Construction of a robust prognostic model for adult adrenocortical

carcinoma: results from bioinformatics and real-world data, J. Cell. Mol. Med. 25 (8) (2021) 3898-3911.

[34] Xi Tian, Wenhao Xu, Yuchen Wang, Aihetaimujiang Anwaier, Hongkai Wang, Fangning Wan, Yu Zhu, Dalong Cao, Guohai Shi, Yiping Zhu, Yuanyuan Qu, Hailiang Zhang, Dingwei Ye, Identification of tumor-infiltrating immune cells and prognostic validation of tumor-infiltrating mast cells in adrenocortical carcinoma: results from bioinformatics and real-world data, OncoImmunology 9 (1) (2020) 1-12.

[35] Xin Yan, Zi-Xin Guo, Dong-Hu Yu, Chen Chen, Xiao-Ping Liu, Zhi-Wei Yang, Tong- Zu Liu, Sheng Li, Identification and validation of a novel prognosis prediction model in adrenocortical carcinoma by integrative bioinformatics analysis, statistics, and machine learning, Front. Cell Dev. Biol. 9 (2021) 671359.

[36] Xiaochun Yi, Yueming Wan, Weiwei Cao, Keliang Peng, Xin Li, Wangchun Liao, Identification of four novel prognostic biomarkers and construction of two nomograms in adrenocortical carcinoma: a multi-omics data study via bioinformatics and machine learning methods, Front. Mol. Biosci. 9 (2022) 878073.

[37] He Xiao, Deqiang Xu, Ping Chen, Guang Zeng, Xinghuan Wang, Xinhua Zhang, Identification of five genes as a potential biomarker for predicting progress and prognosis in adrenocortical carcinoma, J. Cancer 9 (23) (2018) 4484-4495.

[38] Mahshid Arastonejad, Daniyal Arab, Somayeh Fatemi, Pezhman Golshanrad, Unveiling the significance of ncap family genes in adrenocortical carcinoma and adenoma pathogenesis: a molecular bioinformatics exploration, Cancer Inform. 23 (2024) 1-10.

[39] Javad Omidi, miR-507 and miR-665 as central MicroRNA regulators in the ceRNA network of adrenocortical carcinoma: a systems biology approach, Hum. Gene (2025) 201498.

[40] Martin Fassnacht, Matthias Kroiss, Bruno Allolio, Update in adrenocortical carcinoma, J. Clin. Endocrinol. Metab. 98 (12) (2013) 4551-4564.

[41] Tobias Else, Alex C. Kim, Aaron Sabolch, Victoria M. Raymond, Asha Kandathil, Elaine M. Caoili, Shruti Jolly, Barbra S. Miller, Thomas J. Giordano, Gary D. Hammer, Adrenocortical carcinoma, Endocr. Rev. 35 (2) (2014) 282-326.

[42] Rossella Libé, Adrenocortical carcinoma (ACC): diagnosis, prognosis, and treatment, Front. Cell Dev. Biol. 3 (2015) 45.

[43] Guillaume Assié, Eric Letouzé, Martin Fassnacht, Anne Jouinot, Windy Luscap, Olivia Barreau, Hanin Omeiri, et al., Integrated genomic characterization of adrenocortical carcinoma, Nat. Genet. 46 (6) (2014) 607-612.

[44] Siyuan Zheng, D.Cherniack Andrew, Dewal Ninad, A.Moffitt Richard, Danilova Ludmila, A.Murray Bradley, M.Lerario Antonio, et al., Comprehensive pan-genomic characterization of adrenocortical carcinoma, Cancer Cell 29 (5) (2016) 723-736.

[45] Mehrdad Hashemi, Sahar Rezaei Sara Etemad, Romina Rajabi Setayesh Ziaolhagh, Soheila Abdi Parham Rahmanian, et al., Progress in targeting PTEN/PI3K/Akt axis in glioblastoma therapy: Revisiting molecular interactions, Biomed. Pharmacother. 158 (2023) 114204.

[46] Alireza Mafi, Seyedeh Mahdieh Khoshnazar, Amirhossein Shahpar, Noushin Nabavi, Neda Hedayati, Mina Alimohammadi, Mehrdad Hashemi, Afshin Taheriazam, Najma Farahani, Mechanistic insights into circRNA-mediated regulation of PI3K signaling pathway in glioma progression, Pathol. Res. Pract. 260 (2024) 155442.

[47] Mahshid Deldar Abad Paskeh, Mohammad Hossein Gholami Atefeh Mehrabi, Ehsan Ranjbar Amirhossein Zabolian, Adnan Ranjbar Hossein Saleki, et al., EZH2 as a new therapeutic target in brain tumors: molecular landscape, therapeutic targeting and future prospects, Biomed. Pharmacother. 146 (2022) 112532.

[48] Mehrdad Hashemi, Sophie Mousavian Roshanzamir, Sima Orouei, Pouria Daneii, Rasoul Raesi, Haleh Zokaee, Pooria Bikarannejad, et al., Shedding light on function of long non-coding RNAs (lncRNAs) in glioblastoma, Non Coding RNA Res. 9 (2) (2024) 508-522.

[49] Alireza Mafi, Neda Hedayati, Samaneh Kahkesh, Sara Khoshayand, Mina Alimohammadi, Najma Farahani, Kiavash Hushmandi, The landscape of circRNAs in gliomas temozolomide resistance: insights into molecular pathways, Non Coding RNA Res. 9 (4) (2024) 1178-1189.

[50] Sajad Najafi, Shing Cheng Tan, Pourya Raee, Yazdan Rahmati, Yahya Asemani, E. Hui, Clarissa Lee, Kiavash Hushmandi, et al., Gene regulation by antisense transcription: a focus on neurological and cancer diseases, Biomed. Pharmacother. 145 (2022) 112265.

1] Ashkan Pouyan, Masoud Ghorbanlo, Masoud Eslami, Majid Jahanshahi, Ehsan Ziaei, Ali Salami, Khatere Mokhtari, et al., Glioblastoma multiforme: insights into pathogenesis, key signaling pathways, and therapeutic strategies, Mol. Cancer 24 (1) (2025) 58.

[52] Neda Hedayati, M. Rezaee, R. Mousavi Dehmordi, M. Alimohammadi, A. Mafi, Recent insights into the angioregulatory role of long non-coding RNAs and Circular RNAs in Gliomas: from signaling pathways to clinical aspects, Curr. Med. Chem. (2024).

[53] Fatemeh Norouzinasab, Niloufar Salimian, Khatere Mokhtari, Mohammadarian Akbari, Mazaher Maghsoudloo, Maliheh Entezari, Afshin Taheriazam, Najma Farahani, Mehrdad Hashemi, Discovery of LINC01614 associated with the SPP1 gene in colorectal cancer, Pathol. Res. Pract. 266 (2025) 155761.

[54] Sepideh Mirzaei, Mahshid Deldar Abad Paskeh, Farid Hashemi, Amirhossein Zabolian, Mehrdad Hashemi, Maliheh Entezari, Teimour Tabari, et al., Long non-coding RNAs as new players in bladder cancer: Lessons from pre- clinical and clinical studies, Life Sci. 288 (2022) 119948.

[55] Mahshid Deldar Abad Paskeh, Sepideh Mirzaei, Sima Orouei, Amirhossein Zabolian, Hossein Saleki, Negar Azami, Kiavash Hushmandi, et al., Revealing the role of miRNA-489 as a new onco-suppressor factor in different cancers based on pre-clinical and clinical evidence, Int. J. Biol. Macromol. 191 (2021) 727-737.

[56] I Niloufar Salimian, Maryam Peymani, Kamran Ghaedi, Sepideh Mirzaei, Mehrdad Hashemi, Diagnostic and therapeutic potential of LINC01929 as an oncogenic LncRNA in human cancers, Pathol. Res. Pract. 244 (2023) 154409.

[57] Maliheh Entezari, Mehrdokht Sadrkhanloo, Mohsen Rashidi, Sholeh Etehad Asnaf, Afshin Taheriazam, Mehrdad Hashemi, Milad Ashrafizadeh, et al., Non-coding RNAs and macrophage interaction in tumor progression, Crit. Rev. Oncol. /Hematol. 173 (2022) 103680.

[58] Sepideh Mirzaei, Mohammad Hossein Gholami, Hui Li Ang, Farid Hashemi, Ali Zarrabi, Amirhossein Zabolian, Kiavash Hushmandi, et al., Pre-clinical and clinical applications of small interfering RNAs (siRNA) and co-delivery systems for pancreatic cancer therapy, Cells 10 (12) (2021) 3348.

9] Anuradha Kirtonia, Milad Ashrafizadeh, Ali Zarrabi, Kiavash Hushmandi, Amirhossein Zabolian, Atefe K. Bejandi, Reshma Rani, et al., Long noncoding RNAs: a novel insight in the leukemogenesis and drug resistance in acute myeloid leukemia, J. Cell. Physiol. 237 (1) (2022) 450-465.

[60] Taylor Kalomeris, Majd Al Assaad, Jesus Delgado-de la Mora, Gunes Gundem, Max F. Levine, Baris Boyraz, Jyothi Manohar, et al., Whole genome profiling of primary and metastatic adrenocortical carcinoma unravels significant molecular events, Pathol. Res. Pract. 266 (2025) 155725.

[61] Walaa A. El-Dakroury, Heba M. Midan, Ahmed I. Abulsoud, Shereen Saeid Elshaer, Ahmed A. El-Husseiny, Doaa Fathi, Nourhan M. Abdelmaksoud, et al., miRNAs orchestration of adrenocortical carcinoma-particular emphasis on diagnosis, progression and drug resistance, Pathol. Res. Pract. 248 (2023) 154665.

[62] I Rin Yamada, Fumi Kawakami, Naoko Maeda, Shoji Nagao, Mayuko Yamamoto, Takeshi Motohara, Eiji Kondoh, Yoshiki Mikami, Yoshihiro Komohara, Immune cell densities, HLA status, and PD-L1 expression in stage I neuroendocrine neoplasm of the uterine cervix, Pathol. Res. Pract. (2025) 156146.

[63] Gail M. Seigel, Onyekwere Onwumere, Moira Sauane, Sual Lopez, Enyuan Shang, Habiba Habiba, Stephen Redenti, Hans E. Grossniklaus, Rajendra Gharbaran, ALYREF/THOC4 expression and cell growth modulation in retinoblastoma, Pathol. Res. Pract. 260 (2024) 155392.

[64] Stefano Marletta, Anna Caliò, Francesco Pierconti, Shuko Harada, George J. Netto, Pietro Antonini, Diego Segala, et al., SFPQ: TFE3-rearranged PEComa: differences and analogies with renal cell carcinoma carrying the same translocation, Pathol. Res. Pract. (2025) 155963.