Bioinformatic analyses and experimental validation of the role of m6A RNA methylation regulators in progression and prognosis of adrenocortical carcinoma

Data source

The training cohort was obtained from The Cancer Genome Atlas (TCGA) database (https://www.cancer.gov/tcga), including the gene expression data from 79 samples and the clinical data from 90 samples. The type of gene expression data was set as ‘transcriptome profiling’ and ‘gene expression quantification’. The type of clinical data was ‘BCR-XML’. To avoid the interference caused by other special tumor subtypes, the histologic types of ACC were selected as adenomas and adenocarcinomas. The specific database filtering settings were shown in Supplementary Table 1.

To find the appropriate validation cohort, we used ‘adrenal cancer’ as the searching term of GEO database and set ‘Expression profiling by array’ and ‘Homo sapiens’ as the retrieval conditions. Then, a total of 43 datasets were obtained. Among that, 12 datasets relate to ACC cell lines, 3 datasets relate to children ACC, 2 datasets relate to preclinical models, 2 datasets relate to lung metastatic tumors and 4 datasets relate to adrenal hyperplasia or Cushing's syndrome, which were both excluded. Among remaining 20 ACC datasets, only four datasets (GSE19775, GSE19750, GSE19776 and GSE33371) contain clinical information. However, GSE19775 and GSE19776 do not record the survival status, and the survival status of all samples in GES19750 is dead. Therefore, due to absent clinical information or unknown survival outcomes in other GEO datasets, we finally selected GSE33371 as the validation cohort, which contained 33 ACC samples [22]. Among that, 10 samples were excluded for their survival information was unknown and remaining 23 ACC samples were used for the validation of m6A prognostic model.

Clinical samples

5 pairs of ACC specimens and normal adjacent tissues were collected at department of urology, the second affiliated hospital of Xi'an Jiaotong University (Xi’an, China). All patients submitted informed consent for tissue use. The study protocol was approved by the Ethics Committees of the second affiliated hospital of Xi'an Jiaotong University.

Selection of m6A-related genes

There are three categories of m6A-related genes: ‘Writer’ genes including KIAA1429, METTL3, METTL14, RBM15, WTAP and ZC3H13; ‘Eraser’ genes including ALKBH5 and FTO; and ‘Reader’ genes including HNRNPC, YTHDC1, YTHDC2, YTHDF1 and YTHDF2. These 13 genes were chosen as the research objects in this study, which was consistent with the strategy adopted in previous studies [22, 23]. Besides, the correlation analysis among m6A-related genes was performed by the ‘corrplot’ package of R software, which is based on the calculation of Spearman's correction coefficient.

Identification of prognostic m6A-related DEGs

We performed the extraction and arrangement of TCGA data via Perl (Practical Extraction and Report Language) version 5.28. Combined with TCGA clinical information, 77 ACC samples were divided into localized group (n = 46) and metastatic group (n = 31). The other 2 ACC samples were excluded for their corresponding clinical information was unknown. According to 7^th AJCC TNM-stage system for ACC, patients with clinical stage I-II were classified as localized tumors (no extra-adrenal invasion), and those with clinical stage III-IV were defined as metastatic tumors (Lymphatic invasion or distant metastasis). The differentially expressed genes (DEGs) between localized and metastatic groups were determined by the ‘Limma’ package of R software.

Cox univariate regression analyses were employed to screen out the m6A-related genes with the ability to affect ACC prognosis, namely prognostic genes. Additionally, the GEPIA web server (http://gepia.cancer-pku.cn/) [23] was applied to perform overall survival (OS) analysis based on gene expression data from the TCGA and the GTEx projects, which reconfirmed the m6A prognostic genes. Finally, the intersection genes (the prognostic m6A-related DEGs) between DEGs and prognostic genes were identified via the Venn graph. These crossed genes were attempted to construct a m6A-related risk signature.

Establishment of m6A-related risk signature

We used ‘glmnet’ package to conduct the lasso regression analysis to establish a m6A-related risk signature. According to the prognostic model, the risk score of each sample was calculated, and all ACC samples were divided into high- and low-risk groups based on the median of risk score.

Survival analyses in training cohort

To evaluate the effect of m6A-related risk signature on ACC prognosis, we compared the prognostic difference between high- and low-risk groups. Subsequently, ROC curves, PCA (Principal component analysis) and t-SNE (t-distributed stochastic neighbor embedding) were used to assess the prediction accuracy of this prognostic model. Decision curve analysis (DCA) was conducted to evaluate the clinical net benefit of m6A risk signature. The relationships between the risk signature and clinicopathological features were displayed by a heatmap.

Moreover, based on multivariate Logistic regression analysis, we constructed a nomogram combining ACC clinicopathological features and m6A risk signature to predict the overall survival rate (OSR) of individual at 1,2,3 year. And the prediction accuracy of the nomogram was estimated via the calibration curve.

Besides, to confirm whether risk score could be an independent prognostic factor in ACC, we performed cox univariate and multivariate analyses successively. In order to assess the application scope of m6A risk signature in ACC prognosis analysis, we compared the prognostic difference between different risk groups under the same clinical subgroups.

Survival analyses in validation cohort

The m6A prognostic model was tested in the validation cohort (GSE33371). According to the cutoff value calculated with the same prognostic model, 23 ACC samples in validation cohort were divided into high- and low-risk groups. Likewise, the prognostic difference between two groups was compared and ROC curves, independent prognostic analyses were also performed. The distribution of survival outcomes of each ACC sample was displayed using the risk plots. Besides, the heatmap of clinical features and gene expression were drawn by the ‘pheatmap’ package in R software.

Immune analyses

The immune abundances of 22 leukocyte subtypes in each ACC sample were calculated by the CIBERSORT algorithm [24]. The active levels of 13 immune-related pathways were assessed based on single-sample gene set enrichment analysis (ssGSEA), which was performed by the ‘gsva’ package of R software [25]. And then, the differences of immune infiltration levels and immune pathway activities between high- and low-risk groups were determined by the ‘Limma’ package, which revealed the effect of m6A risk signature on ACC immune microenvironment.

Further, we investigated the potential links between the m6A risk genes and immune checkpoint inhibitor (ICIs) therapy. The expression correlations between two cell-surface mediators (PD-L1 and CTLA4) and m6A risk genes were assessed using TISIDB online tool (http://cis.hku.hk/TISIDB/) [26]. Meanwhile, the impact of PD-L1 (CD274) and CTLA4 expressions on ACC prognosis, and their expressive difference between high- and low-risk groups were also ascertained. Besides, the distribution of HNRNPC and RBM15 expression across different immune subtypes were presented via TISIDB tool.

Mutation analyses

The cBioPortal database (http://cbioportal.org) provides multidimensional cancer genomics data [27]. Using ‘OncoPrint tab’ and ‘cancer types summary tab’, we summarized the genetic alteration, mutation type and frequency of m6A risk genes.

Cell culture

Two adrenocortical cancer cell lines, H295R and SW13, were purchase from Institute of Biochemistry and Cell Biology at the Chinese Academy of Sciences (Shanghai, China). H295R cells were cultured in DMEM/F12 (Dulbecco's Modified Eagle Medium) with multiple ingredients including 6.25ug/ml insulin, 6.25ug/ml transferrin, 6.25 ng/ml Selenium, 1.25mg/ml BSA, 5.35 ug/ml Linoleicacid, 2.5% Nu-Serum I and 1% Penicillin-Streptomycin (P/S) (Yipu Biotechnology co, Wuhan, China). SW13 cells were cultured in DMEM added 10% FBS (Fetal bovine serum) and 1% P/S (Yipu Biotechnology co, Wuhan, China). All cells were incubated at 37°C in a humidified atmosphere with 5% CO₂ and 95% humidity.

RNA extraction and RT-qPCR

The expressions in mRNA level of HNRNPC and RBM15 in tissues and cells were assessed by RT-qPCR. Clinical tissues were immersed in RNA wait solution (Hengya Biotechnology co, Shanghai, China) at 4°C for 24 h then saved at -80°C. Total RNA were extracted from tissues and cells through TRIzol reagent (G-Clone, Beijing, China). cDNA was synthesized by using RT (Reverse Transcription) reagent Kit (Takara, Japan). qPCR was amplificated via SYBR-Green reagent (Takara, Japan). Expression levels were normalized to GAPDH and the relative mRNA levels were calculated based on 2^−ΔΔCt method. Primer list was presented in Supplementary Table 2.

Cell transfection

Specific small interfering RNAs (siRNAs) were synthesized to inhibit HNRNPC expression and pcDNA 3.1 was employed to carry HNRNPC overexpression sequence. si-HNRNPC and pc-HNRNPC were designed by GenePharma Biotechnology (Shanghai, China). The siRNA sequences were listed in Supplementary Table 2. Lipofectamine 2000 (Invitrogen, Thermo Fisher, Waltham, MA, USA) was used to transfect H295R and SW13 cells. The transfection efficiency was determined by Western blot after cultivation for 48 h.

MTT assay

MTT assay was performed to evaluate the proliferative ability of adrenal cancer cells. Six groups ACC cells were seeded in 96-well culture plates with the concentration of 5 × 103/well. After incubating for 24 h, 48 h and 72 h, cells were added with MTT Reagent (Solarbio Life Science co, Beijing, China) and cultured for 4 h, at 37°C. Then, DMSO was added in each well after discarding the medium. The absorbance was measured by a microplate reader (ThermoFisher, Waltham, MA, USA) at 490 nm.

Transwell migration and invasion assay

The migrative and invasive abilities of ACC cells was measured via Transwell chamber (Corning, NY, USA). When conducting invasion assay, 100μL DMEM-diluted Matrigel (Corning, NY, USA) was added to each well before seeding cells. But no Matrigel was added during migration assay. Transfected cells (5 × 10⁴/well) were seeded in upper chamber with serum-free medium and complete DMEM containing 20% FBS was added in the lower chamber. After 48 h incubation and discarding the medium of upper chamber, cells on the upper surface of membrane were removed by twice PBS washing and cotton swab wiping out. The cells adhering to the lower surface of membrane were fixed in paraformaldehyde for 20 minutes and stained by 0.1% crystal violet for 20 minutes. The migrative and invasive cells were observed under a microscope.

Western blot

Well-proliferative cells were lysed by RIPA (ThermoFisher, Waltham, MA, USA). Protein concentration was detected using BCA Protein Assay Reagent (Pierce, Appleton, Wisconsin, USA). Then, 50 μg protein samples were separated by 10% SDS-PAGE and transferred to nitrocellulose membranes. After blocking the membranes with 10% fat-free milk for 2 h at room temperature, the membranes were incubated with primary antibodies against HNRNPC (HPA051075, Sigma, Louis, MO, USA) at 4°C overnight. After membranes washing, the HRP secondary antibody (goat anti-rabbit IgG, ThermoFisher, Waltham, MA, USA) was added and incubated for 2 h at room temperature.

Statistical analysis

All statistical analyses were performed using R software (version 3.6.2). The Kolmogorov–Smirnov test was used to analyze the relationships between m6A-related risk signature and the clinicopathological features of ACC. The difference in overall survival rate (OSR) between different groups was compared based on the Kaplan–Meier method. All statistical results with a p-value <0.05 were considered as significant.

Ethics approval and consent to participate

The study protocol was approved by the Ethics Committees of the second affiliated hospital of Xi'an Jiaotong University. All patients submitted informed consent for tissue use.

Availability of data and materials

The datasets used and/or analyzed in the current study are available from the corresponding author upon reasonable request.

Thirteen regulatory genes play a crucial role in m6A methylation

Three types of m6A regulatory genes were included in the subsequent bioinformatic analyses. The functions of these regulatory genes and the process of m6A methylation were shown in Figure 2A. A multicomponent methyltransferase complex consisting of ‘Writer’ genes is responsible for m6A methylation. METTL3 and METTL14 constitute the core portion of m6A complex, in which only METTL3 possesses the activity of methyltransferase, while, METTL14 provides a structural support for RNA binding [28]. WTAP acts as an adaptor protein to bridge METTL3 and METTL14 [29]. Besides, other ‘Writer’ genes (KIAA1429, RBM15 and ZC3H13) are proven to serve as auxiliary subunits required for m6A methylation. ‘Eraser’ genes, FTO and ALKBH5 have the ability of demethylase, therefore, which can reverse the m6A methylation [30]. All bioproducts of ‘Reader’ genes can recognize and bind to the m6A-modified sites on RNA. These genes regulate mRNA abundance through affecting RNA splicing, transport, translation and changing structural stability of RNA [28].

Figure 2. Expression of m6A regulators in ACC. (A) The process and molecular functions of m6A RNA methylation. (B) The expressive heatmap of m6A-related genes. The gene symbols are on the right of the heatmap. L, localized tumor (n = 46). M, metastatic tumor (n = 31). High expression is shown in red and low expression is green. (C–E) Differential expression of METTL3, HNRNPC and RBM15 between localized and metastatic tumor samples. Gene expression is measured by FPKM. FPKM, Fragments Per Kilobase per Million. ^*P < 0.05, ^**P < 0.01.

Some m6A-related genes are differential expressive in metastatic ACC samples.

The expression differences of m6A genes between localized and metastatic ACC samples were shown in Figure 2B–2E. A violin plot of differential expression of m6A genes was presented in Supplementary Figure 1. Among 13 m6A regulators, only METTL3, HNRNPC and RBM15 were upregulated in metastatic ACC samples, which suggested that these m6A DEGs may serve as oncogenes in ACC progression. Besides, the correlations among m6A genes were presented in Supplementary Figure 2. All regulatory genes displayed positively co-expression and METTL14-YTHDC1 possessed the largest correlation coefficient (0.67).

Screening the m6A genes associated with ACC prognosis

Through cox univariate analysis of TCGA data, 5 genes were proven to be related to ACC prognosis, namely METTL3 (p = 0.030, HR = 1.089), METTL14 (p = 0.043, HR = 0.767), WTAP (p = 0.041, HR = 1.089), HNRNPC (p = 0.001, HR = 1.043) and RBM15 (p < 0.001, HR = 2.057) (Figure 3A). Moreover, we reconfirmed the m6A prognostic genes via GEPIA database (Figure 3B–3N). Similar with TCGA analytical results, METTL3 (p = 0.05, HR = 2.1), HNRNPC (p = 0.0033, HR = 3.3) and RBM15 (p < 0.011, HR = 2.7) can still affect ACC survival outcomes (Figure 3D, 3F, 3H). However, the ectopic expression of METTL14 and WTAP did not bring a prognostic difference (Figure 3G–3I).

Figure 3. Screening of m6A prognostic genes. (A) Cox univariate prognostic analyses of m6A-related genes based on TCGA database. (B–N) Prognostic difference of m6A-related genes between high- and low-expression groups based on GEPIA database. Group cutoff is set as median of gene expression.

Construction and assessment of m6A-related risk signature.

According to the above analyses, we finally identified three m6A genes with both discernable expression and affecting prognosis via Venn diagram (Figure 4A). Subsequently, these m6A prognostic DEGs, METTL3, HNRNPC and RBM15, were selected in the Lasso regression analysis. When the partial likelihood deviance reached maximum, the value of log (λ) ranged from –3 to –2. At this time, there were two variables whose coefficients did not decay to zero, therefore, these two genes were applied to construct the prognostic model (Supplementary Figure 3). The m6A-related risk score3 = 0.458*RBM15+0.017*HNRNPC.

Figure 4. Construction of m6A-related risk signature. (A) Identification of m6A-related prognostic DEGs. (B) The survival difference between the high- (n = 39) and low-risk groups (n = 38). (C) Time-dependent ROC curves of m6A risk signature. (D) PCA plot of m6A risk signature. (E) t-SNE plot of m6A risk signature. (F) ROC curves of m6A risk signature and ACC clinical parameters. (G) Risk plots of m6A risk signature. (H) The summary of HNRNPC and RBM15 mutations. (I) The summary of mutation types. (J) The mutation types of HNRNPC. (K) The mutation types of RBM15. (L) The heatmap of m6A risk signature. The relationships between m6A-related risk levels and the clinicopathological characteristics of ACC were determined by Kolmogorov–Smirnov test. Overexpression is presented in red and low expression is green. DEGs, differentially expressed genes; ROC, receiver operating characteristic curve; PCA, principal component analysis; s-SNE, t-distributed stochastic neighbor embedding; ^*P < 0.05, ^**P < 0.01.

The risk score of each ACC sample was calculated through m6A risk signature (Figure 4G). According to the median of risk score, 77 ACC samples were divided into high- and low-risk groups. High risk group resulted in worse survival outcome bringing a 38.3% of 5-year OSR, comparing with 64.6% of that in low risk group (Figure 4B). Moreover, high risk levels were tightly associated with undesirable clinicopathological status (Figure 4L), which indicated m6A risk signature was involved in malignant progression of ACC.

Besides, the ROC curves confirmed a great prediction accuracy of m6A risk signature in ACC prognosis, and the accuracy increased slightly with the extension of follow-up time (Figure 4C). In contrast to T, N and clinical stages, m6A risk signature also presented a better prediction accuracy (Figure 4F). PCA and t-SNE analytical results both indicated that the novel risk signature could successfully cluster all ACC samples into two different prognosis groups (Figure 4D–4E). In a word, the m6A risk signature is reliable in theory.

HNRNPC and RBM15 infrequently mutated in ACC

Using the cBioPortal database, only four out of 89 ACC samples (4.49%) were found to have HNRNPC and RBM15 mutations (Figure 4H). Regarding mutation type, amplification was the most predominant type for all samples (2 cases) and the only type in HNRNPC genetic alteration (Figure 4I–4K). Overall, HNRNPC and RBM15 infrequently mutated in ACC, suggesting that their aberrant expressions may result from epigenetic modification.

M6A-related risk signature contributes to ACC prognostic analysis

Based on cox univariate and multivariate analyses, only T stage (p = 0.005, HR_mean = 3.665) and risk level (p = 0.036, HR_mean = 2.457) were identified as the independent prognostic factors of ACC (Figure 5A–5B). Meanwhile, the risk signature can distinguish the prognostic difference of patients with Stage I-II (p = 0.029), T3-4 (p = 0.026) and N0 (p = 0.001), but not with Stage III-IV (p = 0.288), T1-2 (p = 0.169) and N1 (p = 0.749) (Figure 5D–5I).

Figure 5. The effect of m6A-related risk signature on ACC prognosis in TCGA cohort. (A) The results of cox univariate prognostic analysis. (B) The results of multivariate prognostic analysis. (C) The calibration curves. (D–I) The prognostic differences of ACC patients with the same clinical subgroups. (J) The DCA curve of m6A risk signature. ‘Simple’ curve (red) represents the prognostic model composed of age and clinical stage. ‘Complex’ curve (green) represents the prognostic model composed of age, clinical stage and m6A risk score. (K) The nomogram is used to predict the OSR of ACC patient at 1, 2, 3 year. DCA, decision curve analysis.

Moreover, when we inserted m6A risk signature in ACC prognostic analysis, the net benefits of clinical decision-making were increased (Figure 5J), which indicated that the novel risk signature can benefit prognostic prediction of ACC. To easily predict the overall survival rate (OSR) of individual at 1,2,3 year, we constructed a nomogram using T stage and m6A risk level, the two independent prognostic factors of ACC (Figure 5K). For example, an ACC patient with T2 stage (20 points) and 35 points of risk score (30 points) will obtain a 55 total points, whose 2-year OSR is predicted to be higher than 80%. Besides, calibration plots showed that the nomogram did well compared with an ideal model (Figure 5C). Altogether, m6A-related risk signature can provide important supplement and new idea for ACC prognostic analysis.

M6A-related risk signature is confirmed in the validation cohort

To verify the prognostic value of m6A-related risk signature, GSE33371 was selected as the validation cohort. Similar to training cohort, high risk group led a worse survival outcome (p = 0.039) (Figure 6A–6C). Meanwhile, m6A risk signature possessed a better prognostic accuracy than ACC clinicopathological features (AUC = 0.698) (Figure 6B). Moreover, m6A risk level (HR = 3.938, p = 0.045) and Weiss score (HR = 4.666, p = 0.043) were both identified as ACC-independent prognostic factors (Figure 6D–6E). Besides, high m6A risk was closely with unfavorable tumor weight and Weiss score, but not with tumor diameter and clinical stage (Figure 6F). Altogether, the prognostic value of m6A risk signature was successfully verified in the validation cohort.

Figure 6. The validation of m6A-related risk signature in GSE33371 cohort. (A) The survival difference between the high- and low-risk groups. (B) Multivariable ROC curves in the validation cohort. (C) The risk plots in the validation cohort. (D–E) identification of independent prognostic factors in the validation cohort. The results of cox univariate regression are shown in green and those of cox multivariate regression are red. (F) The heatmap of m6A risk signature in validation cohort. ^*P < 0.05.

M6A-related risk signature affects immune microenvironment of ACC

The immune abundances of 22 leukocyte subtypes in each ACC sample were calculated based on CIBERSORT algorithm (Supplementary Figure 4). Further, we determined the difference of immune cellular proportion between high- and low-risk groups (Figure 7A). High m6A risk can facilitate infiltration levels of T cells CD4 memory activated (p = 0.037), NK cells resting (p = 0.027), Macrophages M0 (p < 0.001), Dendritic cells resting (p = 0.010), Dendritic cells activated (p = 0.007) and Eosinophils (p < 0.001) in ACC; Inversely, it can reduce that of B cells naive (p = 0.021), T cells gamma delta (p = 0.026), Macrophages M1(p = 0.011) Macrophages M2 (p = 0.018) and Mast cells resting (p = 0.007). Moreover, except for MHC (major histocompatibility complex) class I and Type I IFN (interferon) Response, the activities of other 11 immune pathways were all retarded in high m6A-risk level (Figure 7B).

Figure 7. The effect of m6A-related risk signature on ACC immune microenvironment. (A) Comparison of the infiltrating levels of 22 immune cells between different risk groups. (B) Comparison of the activity scores of 13 immune-related pathways between different risk groups. (C–D) The relationships between HNRNPC expression and CD274 (PD-L1), CTLA4 expressions. (E–F) The relationships between RBM15 expression and CD274 (PD-L1), CTLA4 expressions. (G–H) The prognostic differences of ACC patients in TCGA cohort between high- and low-CD274 or CTLA4 expression. (I–J) The expressive difference of CD274 or CTLA4 between high and low m6A-risk group. (K–L) The distributions of m6A risk genes in different PAAD immune subtypes. DC, dendritic cell; APC, antigen-presenting cells; CCR, cytokine-cytokine receptor; IFN, interferon.

High m6A risk level may be unfavorable to ICIs therapy.

Immune checkpoint inhibitors (ICIs) bring an important breakthrough for genitourinary tumors, especially for advanced bladder cancer [31]. Therefore, we further investigated the potential associations between m6A risk level and ICIs therapy. It is now established that the therapeutic effect of ICIs therapy was tightly related to antigen-presenting process [32], immune checkpoint [33] and cytolytic effect [34]. However, these immune processes were all suppressed by high m6A risk level (Figure 7B), suggesting that high m6A risk level may hinder ICIs treatment.

Commonly, patients with PD-L1 overexpression had better response to anti-PD-1 therapy, thus enjoyed better survival [35]. Similar results were observed in this study, ACC patients with high PD-L1 expression (training cohort) had better prognosis than that with low expression (Figure 7G). However, this phenomenon did not appear in CTLA4 (Figure 7H). As for the expressive associations of immune-checkpoint markers with m6A risk genes, the expressions of PD-L1 (CD274) and CTLA4 negatively correlated with that of HNRNPC, while failed to be correlate with that of RBM15 (Figure 7C–7F). And their expressions in high m6A risk group were both lower than those in low risk group (Figure 7I–7J). In a word, high m6A risk not only suppressed immunotherapy-related biological processes, but also repressed the expressions of immune-checkpoint markers. Besides, it was observed that there were no significant differences of HNRNPC and RBM15 expressions between different immune subtypes (Figure 7K–7L).

HNRNPC promotes the proliferation, migration and invasion of ACC cells.

According to GEPIA database, HNRNPC was reconfirmed to be upregulated in tumor samples (Figure 8A). However, there was no expressive difference of RBM15 between tumor and normal samples (Figure 8B). Similarly, overexpression of HNRNPC was detected in clinical tumor tissues by qPCR but not for RBM15 (Figure 8C–8D). Therefore, we focused on the biofunctions of HNRNPC in adrenal cancer cells.

Figure 8. The expression and biofunction of HNRNPC in ACC. (A–B) The differential expression of HNRNPC and RBM15 between tumor and normal samples based on GEPIA database. (C–D) The differential expression of HNRNPC and RBM15 in 5 pairs of clinical specimens. (E) Detection of transfection efficiency. (F–G) HNRNPC can promote proliferation of H295R and SW13 cells. (H) HNRNPC can promote migration and invasion of H295R and SW13 cells. ^*P < 0.05.

siRNAs and pcDNA 3.1 were used to alter the protein levels of HNRNPC in H295R and SW13 cells. Western blot test revealed that si-HNRNPC can effectively suppress the expression of HNRNPC and pc-HNRNPC can promote it (Figure 8E). Further, silencing HNRNPC can retard the proliferative, migrative and invasive abilities of adrenal cancer cells (Figure 8F, 8G, 8H). Conversely, overexpression of HNRNPC facilitated the proliferation, migration and invasion of adrenal cancer cells (Figure 8F, 8G, 8H). These results reiterated that HNRNPC may serve as an oncogene in ACC.