2.1 Datasets

The RNA sequencing (RNA-seq) data of 77 ACC samples (TCGA) and 128 normal adrenal glands (GTEx) were both downloaded from the UCSC Xena database (https://xena.ucsc.edu/). Four ACC mRNA expression series (project ID: GSE10927, GSE75415, GSE12368, and GSE143383) were acquired from the GEO (http://www.ncbi.nlm.nih.gov/geo/) database.

2.2 Weighted gene co-expression network analysis

Weighted gene co-expression network analysis (WGCNA) was conducted on the RNA-seq data of 77 ACC samples and 128 normal adrenal gland samples using R package “WGCNA” (version 1.70-3) according to the previously described standard method.²⁰ One normal adrenal gland sample outlier was removed according to hierarchical clustering, and an experiential power value of 6 was used during co-expression network construction.

2.3 Functional enrichment analysis

Gene ontology (GO) functional enrichment analysis was conducted using Metascape online tools (https://metascape.org, version v3.5.20240101). Gene set enrichment analysis (GSEA) was conducted using R package “clusterProfiler” (version 4.2.2) with parameters minGSSize = 15, maxGSSize = 500, nPermSimple = 10 000. The functional annotation file for GSEA was acquired from MsigDB (https://www.gsea-msigdb.org/gsea/). PPI network analysis was conducted by STRING online tools (https://cn.string-db.org/).

2.4 Human formalin-fixed paraffin-embedded samples

Fifteen formalin-fixed paraffin-embedded (FFPE) samples of ACC, 13 paired FFPE samples of adrenocortical adenoma, and corresponding tumour-adjacent normal adrenal cortical tissue were acquired from the Department of Pathology, Shandong Provincial Hospital Affiliated to Shandong First Medical University in Jinan from 2013 to 2023. None of the patients underwent adjuvant therapy prior to surgery. The detailed clinicopathological features of the 15 ACC patients are provided in Table S1.

2.5 Plasmid, siRNA, and lentivirus construction

The pcDNA3.1-FGL1 plasmid was constructed by Genecefe Biotechnology Co., Ltd. Plasmid DNA was purified using the Endofree Plasmid Kit (QIAGEN). Two small interfering RNAs (siRNAs) targeting CENPM (siCENPM-1, siCENPM-2) and negative control (siNC) were synthesized by GenePharma Co., Ltd. Recombinant LV-luciferase-shCENPM-Puro and LV-luciferase-shScramble-Puro lentiviruses were constructed by GenePharma Co., Ltd. The siRNA and shRNA sequences are shown in Table S2.

2.6 Plasmid, siRNA transfections and lentivirus infection

The human ACC cell lines, NCI-H295R (RRID: CVCL_0458) and SW-13(RRID: CVCL_0542) were kindly provided by Cell Bank, Chinese Academy of Sciences. H295R cells were cultured in DMEM/F12 medium supplemented with 10% fetal bovine serum (FBS), and .5% solution of insulin-transferrin-selenium. SW-13 cells were cultured in Leibovitz's L-15 medium containing 10% FBS. The recombinant plasmid was transfected into SW-13 cells using PolyFast Transfection Reagent (MCE). siRNAs were transfected into H295R or SW-13 cells with GP-transfect-Mate kit (GenePharma). Recombinant lentiviruses were used to infect the SW-13 cells. Briefly, SW-13 cells were grown in six-well plates at a density of 2.0 × 10⁵ cells per well. When the cells were 50% confluent, they were infected with recombinant lentiviruses in the presence of 5 µg/mL polybrene (GenePharma, Co., Ltd.) at an MOI of 20. Stable clones were selected using 1.0 µg/ml puromycin for 10–14 days.

2.7 Immunohistochemistry and immunofluorescence staining

Human and animal FFPE tissue sections were stained with primary antibodies against IgG, CENPM, or ki67 at 4°C overnight. The Corresponding HRP-labeled secondary antibodies were used for 1 h at room temperature (RT). The average positive CENPM signal was evaluated for five randomly selected regions in each section.

Tissue sections and cell slides were stained with primary antibodies against IgG, CENPM, COL2A1, or FGL1 overnight at 4°C, and subsequently incubated with FITC- or Cy3-conjugated secondary antibodies. Cell nuclei were shown by DAPI staining. Images were captured using a fluorescence microscope (Leica; Olympus). The density of positive cells was evaluated in five randomly selected fields. The antibodies used in this study are listed in Table S3.

2.8 Real-time quantitative PCR

Total cellular RNA was isolated using the Trizol reagent. CDNA was obtained by reverse transcription using a cDNA synthesis kit (Toyobo). Real-time quantitative PCR was carried out using SYBR Green Realtime PCR Master Mix (Toyobo) on a QuantStudio 5 Real-Time PCR System (Thermofisher) following the manufacturer's protocol. The expression level of CENPM in the cell lines was measured using 2^−ΔΔCT. GAPDH was used as a control. Primers of CENPM and GAPDH are listed in Table S2.

2.9 Western blot

Total proteins were extracted using RIPA lysis buffer with inhibitor cocktails and quantified using the BCA protein assay kit. Proteins were separated by SDS-PAGE gel electrophoresis followed by electroblotting onto PVDF membranes. The membranes were incubated with primary antibodies against CENPM or FGL1 at 4°C overnight, and subsequently with corresponding HRP-conjugated secondary antibodies for 1 h at RT. Protein signals were detected using the Gelview 1500 pro (BLT). The antibodies used in this study are listed in Table S3.

2.10 Cell proliferation, migration, and invasion assay

Cell proliferation was detected using a colony formation assay. After siRNA transfection for 72 h, 1000 cells were seeded in each well of a six-well plate and cultured for approximately 20 days. The colonies were stained with a crystal violet solution. Cell migration was assessed using a wound-healing assay. After 72 h of siRNA transfection, cells were incubated in PBS for 10 min, scratched the wound using a 100 µL pipette tip, and cultured in a medium with 2% FBS for 72 h. Cell invasion was assessed using a transwell invasion assay. Transfected cells were placed in the upper layer of the Matrigel chamber in a serum-free medium, and a culture medium with 25% FBS was placed in the lower chamber.

2.11 DIA quantitative proteomics and bioinformatics analysis

SW-13 cells transfected with siNC or siCENPM-1 for 72 h were collected for DIA quantitative proteomic analysis. LC-MS/MS high-resolution mass spectrometry detection was performed in oebiotech Co., Ltd. using TimsTOF Pro (Bruker) and UltiMate 3000 (Thermo Fisher Scientific) systems. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (https://proteomecentral.proteomexchange.org) via the iProX partner repositor^{21, 22} with the dataset identifier PXD058018.

2.12 Co-immunoprecipitation assay

Total proteins from SW-13 cells were extracted using IP lysis buffer and then incubated with FGL1 antibody or IgG overnight at 4°C under gentle agitation. The protein–antibody complex solution was incubated with protein A/G agarose beads for 4 h at RT, and the unbound proteins were washed away. The captured proteins were separated from the beads using elution buffer and were used for western blot analysis of CENPM protein expression. The antibodies used in this study are listed in Table S3.

2.13 Metastatic ACC xenograft mouse model

Six male NPG (NOD.Cg-Prkdc^scid Il2rg^tm1Vst/Vst) mice were bought from Vitalstar Biotechnology Co., Ltd. and divided equally into two groups: LV-shCENPM and negative control. Each mouse was intravenously injected via the caudal vein with 2 × 10⁶ SW-13 cells stably transfected with LV-luciferase-shCENPM-Puro or LV-luciferase-shScramble-Puro. Twenty-eight days after injection, the mice were intraperitoneally injected with D-luciferin sodium salt (Meilunbio) and imaged by the Small Animal In Vivo Imaging System (PerkinElmer). Ethical approval for this study was granted by the Animal Research Ethics Committee of Shandong Second Medical University.

2.14 Statistical analysis

Data were analyzed using GraphPad Prism software (version 9.0) and were presented as mean ± standard deviation (SD). For those with more than two groups, analysis of variance (ANOVA) was performed to ensure p < .05. Differences between the two groups were analyzed using Student's t-test, p < .05. *p < .05, **p < .01, ***p < .001, ****p < .0001.

3.1 Identification of gene module correlated with ACC metastasis

To characterize the potential regulatory genes involved in the progression and metastasis of ACC, we performed WGCNA on ACC samples and normal adrenal gland samples (Figure 1A). All expressed genes were initially filtered using the median absolute deviation method, resulting in a total of 12 035 genes for co-expression network construction. Subsequent hierarchical clustering analysis categorized these genes into 12 distinct gene modules, each assigned a unique colour for identification, while unclustered genes were labelled as grey (Figure 1B). Eigengene adjacency analysis showed the “Salmon” gene module, comprising 363 genes, exhibited the highest eigengene adjacency to both “Cancer” and “Metastasis” phenotypes, and the lowest eigengene adjacency to “Normal” phenotypes (Figure 1C). Consistently, Pearson correlation analysis revealed that the “Salmon” module showed a positive correlation with “Cancer” (cor = .70), and “Metastasis” (cor = .41), as well as a negative correlation with “Survival” (cor = –.49), and “Normal” (cor = –.70) (Figure 1D). Additionally, the module membership (MM) of genes in the “Salmon” module was significantly correlated with their gene significance (GS) for the “Cancer” (cor = .78, p < .01) and “Metastasis” (cor = .80, p < .01) phenotypes (Figure 1E,F). Collectively, our WGCNA results revealed that the “Salmon” gene module was closely associated with the tumorigenesis and metastasis of ACC.

3.2 CENPM was the hub gene related to ACC metastasis

To further identify the key genes driving ACC metastasis, we conducted a functional enrichment analysis of genes in the “Salmon” module. GO enrichment analysis revealed that the genes in the “Salmon” module were predominantly involved in mitotic cell cycle-related processes, particularly in terms connected with chromosome segregation and spindle organization (Figure 2A). Additionally, GSEA showed that gene sets involving “Mitotic sister chromatid segregation” (NES = 1.76, p < .01), “Sister chromatid segregation” (NES = 1.72, p < .01), “Nuclear chromosome segregation” (NES = 1.53, p < .01), and “Mitotic spindle organization” (NES = 1.77, p < .01) pathways were significantly upregulated in ACC (Figure 2B).

To identify the genes within the “Salmon” module that involved the regulation of chromosome segregation in ACC, we overlapped the “Salmon” module genes with ACC upregulated genes and chromosome segregation-related genes. This analysis identified 78 genes within the “Salmon” module that were potentially related to the regulation of chromosome segregation in the ACC (Figure 2C). To further pinpoint hub genes, protein–protein interaction (PPI) network was constructed based on the 78 genes we identified. Notably, the largest PPI network was comprised of eight CENP family members (CENPI, CENPM, CENPH, CENPU, CENPK, CENPQ, CENPL, and CENPW; Figure 2D). We noticed that CENPM had the second-highest node degree (Figure 2E) and the strongest negative correlation with ACC patient overall survival time (Figure 2F), suggesting that CENPM was the hub gene related to ACC metastasis.

3.3 CENPM was upregulated in ACC, and associated with metastasis and poor prognosis of ACC patients

The RNA-seq data in TCGA and GTEx databases showed that the mRNA levels of CENPM were significantly elevated in ACC samples compared with normal adrenal gland samples (Figure 3A). Furthermore, ACC patients in stage IV displayed higher levels of CENPM than those in stage I, stage II, and stage III (Figure 3B), indicating the mRNA expression of CENPM was correlated with ACC metastasis. The microarray data in GSE10927 (Figure 3C), GSE75415 (Figure 3D), GSE12368 (Figure 3E), and GSE143383 (Figure 3F) also indicated that CENPM mRNA levels were highly upregulated in ACC patients.

The immunohistochemistry results revealed that the protein level of CENPM was enhanced in ACC patients than in normal adrenal gland tissues and adrenocortical adenoma patients. In addition, ACC patients in stage IV displayed higher levels of CENPM than those in stage II and stage III (Figure 4), suggesting that the protein expression of CENPM was also correlated with ACC metastasis.

To further elucidate the relationship between CENPM and ACC progression, survival analysis was conducted on ACC patients grouped by higher and lower CENPM expression levels. As shown by the Kaplan–Meier curves for OS and DFS, ACC patients with high CENPM levels displayed poorer overall survival (Figure 3G) and disease-free survival (Figure 3H) than those with low levels of CENPM, providing evidence that the expression of CENPM in ACC negatively correlated with the survival of patients.

3.4 Knockdown of CENPM inhibited the proliferation, migration, and invasion of ACC cells

To verify the impact of CENPM on ACC metastasis, we silenced the expression of CENPM by transfecting siRNA into two ACC cell lines H295R and SW-13. The mRNA and protein levels of CENPM were significantly downregulated after siRNA transfection (Figure 5A,B). Silencing of CENPM led to inhibition of proliferation in both H295R and SW-13 cells as shown by the colony formation assay (Figure 5C). Wound healing and transwell assays revealed that migration and invasion decreased in CENPM knockdown ACC cells (Figure 5D,E).

3.5 Knockdown of CENPM inhibited collagen-containing extracellular matrix signalling

We further conducted DIA quantitative proteomics to identify downstream effectors following CENPM knockdown. The knockdown of CENPM in ACC cells resulted in 24 upregulated proteins and 62 downregulated proteins using a twofold cut-off threshold (Figure 6A).

Unexpectedly, GO enrichment analysis revealed that the downregulated proteins following CENPM knockdown were enriched in biological processes such as collagen fibril organization and cell adhesion (Figure 6B,C). Additionally, GO cellular component analysis indicated that these proteins were predominantly localized in the collagen-containing extracellular matrix and extracellular space (Figure 6B,C). Furthermore, GSEA confirmed that the expression of genes related to cell-matrix adhesion and collagen metabolism was significantly downregulated upon CENPM silencing (Figure 6D). Collectively, these findings suggest that CENPM serves as a positive regulator of collagen-related pathways, with its knockdown leading to the dysregulation of collagen organization and cell adhesion in ACC.

Of the proteins downregulated 2⁶-fold or more, COL2A1, CILP, TNC, ASPN, and FGL1 were involved in “collagen-containing extracellular matrix” signalling (Figure 6E). The protein levels of FGL1 and CILP were validated in CENPM knockdown SW-13 cells by western blot analysis (Figure 6F–H) and the protein levels of COL2A1 were validated in CENPM knockdown SW-13 cells by immunofluorescence (Figure 6I). The results suggested that the protein levels of FGL1, CILP, and COL2A1 were downregulated in CENPM knockdown cells, which was consistent with the results of DIA quantitative proteomics.

3.6 CENPM interacted with FGL1 to regulate migration and invasion of ACC cells

The protein levels of COL2A1 and FGL1 were detected in FFPE sections of ACC and normal adrenal glands by immunofluorescence staining. In normal adrenal glands, COL2A1 and FGL1 proteins were in low levels and located in the extracellular matrix, whereas in ACC patients, COL2A1 and FGL1 proteins were in high levels and located in the cytoplasm and nucleus (Figure 7A). The co-expression of FGL1 and CENPM proteins was remarkable, whereas the co-expression of COL2A1 and CENPM proteins was not obvious (Figure 7B,C). Furthermore, the co-immunoprecipitation assay showed the physical interaction between CENPM and FGL1 (Figure 7D).

To determine whether CENPM regulates the migration and invasion of ACC cells via interacting with FGL1, a rescue experiment was performed. SW-13 cells stably infected with recombinant LV-luciferase-shCENPM-Puro virus (Figure 7E) were transfected with vector (negative control) or pcDNA3.1-FGL1 plasmids (Figure 7F). Wound healing and transwell invasion assays were performed. Overexpression of FGL1 promoted the migration and invasion of CENPM knockdown ACC cells (Figure 7G,H). These results demonstrated that CENPM regulated the migration and invasion of ACC cells via FGL1.

3.7 Knockdown of CENPM suppressed the liver metastasis of ACC in vivo

A metastatic ACC xenograft mouse model was developed by injecting SW-13 cells stably infected with LV-luciferase-shCENPM-Puro or LV-luciferase-shScramble-Puro (normal control) into NPG mice via a caudal vein (Figure 8A). Normal controls showed strong bioluminescence signals in the upper right quadrant of the abdomen, whereas weak or no signals were observed in the upper right quadrant of the abdomen in mice injected with LV-shCENPM ACC cells on day 28 (Figure 8B). A large number of white tumour nodules were observed in the livers of normal controls. In contrast, there were fewer white tumour nodules in the livers of mice that received LV-shCENPM ACC cells (Figure 8C). No visible tumour nodules were observed in the lung, spleen, and kidney, suggesting that the liver was the primary metastatic target for ACC cells. HE staining of liver specimens showed much fewer ACC metastatic tumour nodules in mice injected with LV-shCENPM ACC cells than in normal controls (Figure 8D). Knockdown of CENPM also resulted in a decrease of ki67 positive cells and a decline of FGL1 levels, as shown by ki67 staining and immunofluorescence of ACC metastatic tumour nodules, respectively (Figure 8E,F).