Bimodal genomic approach predicting Semaphorin 7A (SEMA7A) as prognostic biomarker in adrenocortical carcinoma
Anjali Dhall1, Daiki Taniyama1, Fathi Elloumi1, Augustin Luna1,2, Sudhir Varma1,3, Suresh Kumar1, Lauren Escobedo1, Mirit I. Aladjem1, Christophe E. Redon1, Nitin Roper1, William C. Reinhold1, Jaydira Del Rivero1* and Yves Pommier1*
1Developmental Therapeutics Branch & Laboratory of Molecular Pharmacology, Center for Cancer Research, National Cancer Institute, NIH Bethesda, MD 20892.
2Computational Biology Branch, National Library of Medicine, NIH, Bethesda, MD 20892. 3HiThru Analytics, LLC.
* To whom correspondence should be addressed: pommier@nih.gov, jaydira.delrivero@nih.gov
Abstract
Adrenocortical carcinoma (ACC) is a rare and aggressive endocrine malignancy with high mortality and poor prognosis. To elucidate the genetic underpinnings of ACCs, we have analyzed the transcriptome data of 112 ACC tumor samples from patients enrolled in the TCGA and NCI. Among 72 bimodally expressed genes stratifying patients into prognostic groups, we focused on SEMA7A, as it encodes a glycosylphosphatidylinositol-anchored membrane glycoprotein (Semaphorin 7a) regulating integrin-mediated signaling, cell migration and immune responses. We find that high SEMA7A gene expression is associated with poor prognosis (hazard ratio = 4.27; p-value < 0.001). In hormone-producing ACCs, SEMA7A expression is elevated and positively correlated with genes driving steroidogenesis, aldosterone and cortisol synthesis, including CYP17A1, CYP11A1, INHA, DLK1, NR5A1 and MC2R. Correlation analyses show that SEMA7A is co-expressed with the integrin-ß1, FAK (focal adhesion kinase) and MAPK/ERK (mitogen- activated protein kinase/extracellular signal regulated kinases) signaling pathways. Immunohistochemistry (IHC) staining demonstrates the feasibility of evaluating SEMA7A in ACC tissues and shows significant correlation between gene expression (RNA-Seq) and protein expression (IHC). These findings suggest SEMA7A as a candidate for further research in ACC biology, a candidate for cancer therapy, as well as a potential prognosis biomarker for ACC patients.
Translational relevance
Adrenocortical cancer (ACC) remains a challenging disease primarily due to the scarcity of reliable biomarkers for predicting patient outcomes and informing innovative therapeutic strategies, as well as its rarity, which restricts the scope of clinical trials. In our study, we performed RNAseq and IHC analyses of ACC samples sourced from The Cancer Genome Atlas (TCGA), tissue microarray slide, and National Cancer Institute (NCI) cancer patient samples. Our findings indicate that a substantial proportion of ACC tumors exhibit expression of SEMA7A, a glycoprotein involved in Semaphorin cell surface signaling. Notably, elevated levels of SEMA7A were identified as a poor prognostic biomarker and were associated with activation of the integrin- ERK-MAPK kinase signaling pathways. These results suggest that ACC tumors with high SEMA7A expression should be considered at elevated risk, and SEMA7A may serve as a potential
bioRxiv preprint doi: https://doi.org/10.1101/2025.04.03.647086; this version posted April 7, 2025. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
target for immunotherapeutic strategies, including antibody-drug conjugates, T-cell engagers, and/or small molecule inhibitors targeting the MAPK pathway.
Introduction
Adrenocortical cancer (ACC) is a rare and aggressive malignancy arising from the adrenal cortex with high morbidity and mortality rates. The estimated annual incidence of ACC is ~ 0.5 to 2 cases per million individuals worldwide. Notably, the incidence of ACC demonstrates a substantial age- related variation, with a peak occurring around the age of 50. In 40%-60% of patients, ACCs typically exhibit an aggressive biological behavior with symptoms of hormone hyperproduction (1). The prognosis for patients with ACC is generally poor, with a 5-year survival rate <15% among patients with distant metastases (2). Prognosis is closely linked to the stage at which the disease is diagnosed, with earlier-stage presentations associated with more favorable outcomes (3,4). The heterogeneous nature of ACC complicates its diagnosis and treatment, given the variability in clinical manifestations and prognostic outcomes (5). The precise molecular mechanisms underpinning ACC are not fully elucidated despite significant advancements in genomic and transcriptomic profiling. Scientific breakthroughs are warranted to identify diagnostic, prognostic, and therapeutic biomarkers, creating new opportunities for effective management strategies (6-9).
At present the clinical prognostic factors for ACC include tumor stage, cortisol secretion, and patient age. Tumor staging serves as the principal determinant of prognosis, as well as the status of surgical resection margins. Histological features such as mitotic activity exceeding 20%, a Ki-67 index greater than 12%, and a Weiss score greater than 6 have also been correlated with poor prognosis (10). Given the disease’s heterogeneity and rarity, investigations have focused on assessing various immunohistochemistry markers to identify more reliable prognostic factors. Efforts have also been made to develop genomic techniques for evaluating gene expression and alterations as potential molecular prognostic markers. Significant attention has been directed towards the Wnt/CTNNB1 and TP53 signaling pathways, which are frequently mutated in ACC, particularly with respect to CTNNB1 (beta-catenin) and P53 staining (11-13). The presence of somatic CTNNB1 mutations has been associated with recurrence in ACC patients (14).
Systemic therapies are typically employed in the adjuvant setting or for patients exhibiting metastatic or unresectable disease. Mitotane, an adrenolytic agent, has remained the most utilized medication for over fifty years. It is administered as adjuvant therapy following surgery or for inoperable or metastatic cases (15,16). In cases of advanced disease that are not suitable for surgical intervention, the utilization of cytotoxic agents in conjunction with mitotane is indicated.
Two commonly employed treatment regimens include the combination of etoposide, doxorubicin, and cisplatin with mitotane (EDP-M), and the combination of streptozotocin with mitotane (S-M). These therapeutic approaches were evaluated in an international phase-III clinical trial. The findings indicated that the EDP-M regimen was associated with higher objective response rates (ORRs) and improved progression-free survival (PFS) when compared to the S-M regimen; however, no statistically significant difference in overall survival was observed (17). Targeted therapies, including inhibitors of the insulin-like growth factor (IGF) pathway and immune checkpoint inhibitors, are currently under investigation in several clinical trials (18). Preliminary findings indicate that these treatments may provide new options for patients with advanced ACC, though further research is necessary to fully establish their efficacy and safety profiles.
To address the current challenges, research endeavors are directed towards our understanding of the genetic underpinnings of ACC (19-21). This step is crucial in the development of diagnostic and prognosis algorithms based on genomic data. Based on such approaches, several predictive characteristics have been identified, prognostic markers such as (BUB1B, PINK1, MKI67) and molecular classification derived from the clustering of genomic profiles, which proposed the presence of two primary transcriptional clusters, designated as C1A/C1B, exhibiting a strong correlation with survival outcomes. However, the C1A/C1B classifier faces challenges in its implementation as a prognostic biomarker owing to its inherent complexity (5,21-24).
Alternatively, molecular profiling approaches based on bimodal gene expression have been used to identify biomarkers possessing distinct expression distribution allowing for the classification of samples into two clearly defined expression states (25). For instance, estrogen receptor (ESR1) is a well-known bimodal expressed gene in breast cancer patients (26), and Schlafen 11 (SLFN11), a bimodal predictive genomic biomarker, predicts response to DNA- targeted chemotherapy (27). In this study, we implemented a statistical approach to identify bimodally expressed genes using transcriptome profiles of 112 ACC tumor samples from both the TCGA database and from patients enrolled in NCI clinical trials. We report that a cell surface marker, Semaphorin 7a (SEMA7A), is significantly associated with poor prognosis, differentially up-regulated in ACC in comparison with normal adrenal tissues and reflects activation of the FAK and MAPK/ERK signaling pathways in a subset of ACC patients.
Material and Methods
Dataset Collection
In this study, we utilized transcriptomic and clinical data from TCGA-ACC and NCI-ACC database. For the NCI patient samples, a single institution study was conducted in accordance with recognized ethical guidelines as per The Belmont Reports and the Department of Health and Human Services Common Rule and was approved by the NIH Institutional Review Board. Written informed consent was obtained from all participants, and consent was approved by the NIH Institutional Review Board.
RNA-seq sequencing data processed using NCI CCBR RNA-seq pipeline (https://github.com/skchronicles/RNA-seek.git) and STAR (2.7.11b) to align reads to the hg38 reference genome. RSEM was used to normalize gene expression values expressed as log2(FPKM+1) and “RemoveBatchEffect” function from the “Limma” package used to remove the impact of the library preparation protocols. In total, 112 ACC patients from TCGA-ACC (n=79) and NCI-ACC (n=33) were included in this study (see Table 1, Table S1). Additionally, normalized RNA-seq expression data from multiple TCGA cancer types (including BLCA, BRCA, CESC, CHOL, ESCA, HNSC, KIRC, KIRP, LAML, LIHC, LUAD, LUSC, PAAD, PRAD, THCA, UCS, and UVM) were obtained from cBioPortal and normal tissue expression datasets from Genotype-Tissue Expression (GTEx) database.
Bimodal Approach
To identify RNA-seq expression bimodality we employed a combination of the Hartigan Dip test (using the R Diptest package, R version 4.2.3) and Normal/Gaussian Mixture Models (with the R normix package using an Expectation-Maximization (EM) approach). For each individual gene, the expression distribution was sorted and tested for bimodality using the Dip test. If the distribution was found bimodal using Dip test with p-value <0.05, we applied the norMixEM function from the R nor1Mix package to fit a mixture of two normal distributions. Once the fitting had converged, the parameters of this model, such as the means (u1, u2), weights (W1, w2 with w1+ w2=1), and standard deviation (51, 02) for the two distributions, along with the log-likelihood, were computed per gene. We calculated an overall bimodal score, as a rough analog of a t-statistic, as the absolute difference in means divided by the larger of the two standard deviations (See Equation 1). An additional criterion was that the difference between the weights
for the two distributions was less than 0.25 to ensure the selection of genes with similar number of samples in the high and low expression distributions.
Bimodal Score =
abs(u1-12) max (ơ1,02) (1)
Statistical Analyses
Univariate survival analyses were performed using the Cox proportional hazards (Cox-PH) regression model. Gene transcripts from ACC samples were stratified into two categories high and low expression based on median expression cut-off. Significant differences in survival distributions between the high-risk (high expression) and low-risk (low expression) groups were assessed using the log-rank test, with the results presented in terms of Hazard Ratios (HR), p- values, confidence interval and concordance index. HR >1 indicates a negative impact on patient survival, while HR <1 improved survival. An HR of 1 indicates no effect on survival. The stratification of patients into high-risk and low-risk groups was visually represented using Kaplan- Meier (KM) survival curves, providing a distinct depiction of the survival probabilities over time for each group. We used the “survival” and “survminer” packages in R (version 4.2.3). Correlation analyses were performed using the Pearson correlation test. Mann-Whitney U test analyses were performed using GraphPad Prism 9.0 software (GraphPad Software Inc.).
Tissue Microarray and Immunohistochemistry (IHC) Evaluations
Tissue microarray slide AG991 was used to stain ACC and normal adrenal gland tissue samples. Immunohistochemistry (IHC) staining was performed on LeicaBiosystems’ BondRX autostainer with the following conditions: Epitope Retrieval 2 (EDTA) 20’, SEMA7A (Santa Cruz sc- 374432, 1:100 incubated 30’), and the Bond Polymer Refine Detection Kit (LeicaBiosystems #DS9800). Isotype control reagent (mouse IgG2a, BD Biosciences #553454) was used in place of primary antibody for the negative control. Slides were removed from the Bond autostainer, dehydrated through ethanol, cleared with xylene, and coverslipped. Slides were scanned using a Leica Aperio AT2 scanner (Leica Biosystems, Buffalo Grove, Illinois) at 20X magnification. The stained IHC slides were quantified using the HALO image analysis software 3.6 (Indica Labs, Albuquerque, NM, USA). Automated quantification of percent positive cells for each core was performed using cytonuclear algorithm version 2.0.5.
For the NCI-ACC patient specimens, formalin-fixed paraffin-embedded tissue sections were used for IHC. SEMA7A IHC was performed using a previously validated method (28). Optimized staining protocol included antigen retrieval performed by microwave heating in Antigen Retrieval Buffer (pH 6.0) (#ab93678, Abcam) for 20 min. The sections were incubated with mouse monoclonal anti-SEMA7A antibody (Santa Cruz sc-374432, 1:100 dilution in TBST (1×TBS + 0.5% Tween 20)) overnight at 4℃. The sections were then incubated for 30 minutes in N-histofine Simple Stain MAX PO(MULTI) (Nichirei Biosciences Inc., # NIC-414151F) followed by 5 minutes incubation with chromogen DAB-3S (original dilution by manufacturer, Nichirei Biosciences Inc .; 415192F) at room temperature.
Results
Overall Workflow and Implementation of the Bimodal Approach
The overall workflow for the selection of bimodal genes is provided in Figure 1A. At first, we applied the bimodal algorithm to the gene expression data collected from both TCGA and NCI. Ranking the genes based on their bimodal scores led to the selection of 200 genes having the highest bimodal distribution. Analysis on those top 200 genes retrieved 72 genes significantly associated with the survival of ACC patients. Among them 54 are significantly associated with poor prognosis, and 18 with better overall survival (Figure 1A). The list of the 72 bimodal-survival genes is provided in (Table S2) along with their annotation as belonging to the C1A/C1B molecular subtypes. We also assigned the NCI-ACC patients as C1A/C1B molecular subtypes using the ConsensusClusterPlus approach previously used for the TCGA-ACC samples (19,29) (Figure 1B, upper section labeled as molecular subtypes in red for C1A and blue for C1B).
Next, we computed a double clustering heatmap displaying both patient samples (columns in Figure 1B) and expression of the 72 bimodal-survival genes (rows). Patient distribution formed two main clusters (labeled “a” and “b” in Figure 1B). Cluster “a” is enriched for patient with advanced tumors (stage III & IV) and C1A subtype while it is the reverse for patients in cluster “b” (high fraction of localized tumors with C1B staging). Analysis of the gene distribution (rows in Figure 1B) shows an upper cluster encompassing genes that tend to be highly expressed in the cluster “b” patients and can be considered good prognosis genes. Consistently, among them, 4 are known C1B genes: CHST4 (encoding CarboHydrate sulfotransferase 4), CSDC2 (encoding Cold
Shock Domain Containing C2), NME5 (encoding Non-Metastasis cell 5 protein) and PINK1 (encoding PTEN-Induced Kinase 1, whose high expression is a well-established good prognosis biomarker in ACC (21,29). The bimodal-survival genes associated with poor prognosis (cluster “a” in Figure 1B), are distributed in 4 clusters. Two of them annotated as “al” and “a2” in Figure 1B. Cluster “al” include genes associated with DNA replication, chromatin and cell cycle: HJURP, CDK1, CDCA5, GINS2, DDIAS. Cluster “a2” includes cell surface markers, metabolism and Wnt-signaling-related genes including PDE9A, BRSK1, SYTL2, RTN4R, EFNA3, PLPP2, CRMP1, LAMC3, LEF1 and SEMA7A.
Further analyses were performed to divide the bimodal genes based on their relationship with MKI67 and identify novel predictive biomarkers that would complement MKI67. Thirty-four genes showed a low correlation with MKI67 (Figure 1C). Among them SEMA7A, which encodes the cell surface marker Semaphorin 7A showed the highest Hazard Ratio (HR = 4.27 and significant p-value < 0.001) (Figure 1C). These results reveal SEMA7A expression as a novel prognosis biomarker with a broad bimodal expression in ACC.
High expression of SEMA7A in ACC
In the TCGA-pan cancer datasets, ACC is with pancreatic adenocarcinoma (PAAD) and bladder cancer (BLCA) among the cancers expressing the highest levels of SEMA7A (Figure 2A). This high expression is specific to ACC cancer patients as the expression of SEMA7A is low in normal adrenal gland (Figures 2B and S1). The normal tissues expressing highest levels of SEMA7A are lymphoid, nervous and germinal (Figure S1). We next examined whether SEMA7A expression is associated with clinical stage, hormone production and gender. Indeed, we observed that highest expression of SEMA7A was observed in high-grade (III/IV) and hormone-producing ACC (Figure 2C). Female patients tended to have higher SEMA7A expression than males, but the difference was not significant (Figure 2C). The prognosis value of SEMA7A is shown in Figure 2D with significantly reduced survival of patients with high SEMA7A (hazard ration (HR) = 4.27 and p- value <0.001). As expected, stages III/IV and hormone-producing tumors are also at significantly higher risk (Figure 2E). Not significant difference was observed between males and female patients (Figure 2E). Correlation analyses showed a significant association of SEMA7A expression with the expression of adrenocorticotropic hormone receptor, steroidogenic enzymes, cholesterol
transporters and their transcriptional regulator genes (CYP11A1, CYP17A1, MC2R, NR5A1/SF1, DLK1 and INHA) (30-35) (Figure 2F).
Activation of the SEMA7A-integrin-ß1 downstream signaling pathways in ACC
SEMA7A is known to belong to a SEMA7A-integrin-ß1 axis that activates Mitogen-Activated Protein Kinase (MAPK) cascades; especially the ERK/MAPK signaling pathway, which plays crucial role in regulating oncogenic functions, including metastasis and tumor progression (36- 39). As shown in Figure 3A, expression of the downstream pathway genes ITGB1, AKT, PTK/FAK, ERK/MAPK1, LIMK1, CFL1 shows significant positive correlation with SEMA7A expression and with each other. These observations are consistent with the functional activation of the downstream pathways of SEMA7A in the ACC tumors expressing high SEMA7A transcripts.
Expression heatmap of SEMA7A-integrin-ß1 downstream pathway genes stratifies the ACC patient samples into two subgroups, one having high expression and the other having low expression of those genes (Figure 3B). To explore the association between the downstream genes of the SEMA7A-integrin-ß1 pathway and overall survival in ACC patients, we calculated the average expression of genes involved in the SEMA7A pathway and stratified the patients into two groups based on the median expression value (cut-off). The survival analysis shows the poor prognosis of patients having high average expression of SEMA7A pathway genes (Figure 3C).
Together these results demonstrate the functionality of the SEMA7A pathway in a significant fraction of ACC tumors and the significantly poor prognosis of patients with high SEMA7A expression and SEMA7A pathway activation.
Exploration of SEMA7A protein expression in ACC
To compare SEMA7A transcripts and protein expression, we first used tissue microarray slides to determine whether SEMA7A protein expression could be measured in normal adrenal gland (n=6), adrenocortical adenoma (n=27) and ACC (n=14) tumor tissues. Figure 4A shows representative immunohistochemistry (IHC) staining images demonstrating the fraction of cells expressing SEMA7A in ACC tissues compared to normal adrenal gland tissue and adrenocortical adenoma (benign tumor). Quantification (Figure 4B) is consistent with the bimodal distribution of SEMA7A expression observed at the transcript level. Indeed, some samples show very high percentage of positive cells expressing SEMA7A while other samples display low fractions. This suggests that
SEMA7A protein levels determined by IHC can potentially be used as a biomarker to identify specific subgroups of ACC patients and score tumor aggressiveness instead of or in conjunction with RNA-Seq analyses.
To validate the IHC approach in ACC tumor samples, we compared the protein and RNA expression levels of SEMA7A in the NCI-ACC samples. We selected three samples having low and three having high SEMA7A RNA-Seq expression levels. As shown in Figure 5, the samples having low RNA-Seq levels do not show high expression at the protein level. By contrast, patients with high RNA-seq expression of SEMA7A (≥ 5) show protein expression. These results indicate a positive correlation between RNA-Seq expression and protein levels in the selected ACC samples and suggest the value of measuring SEMA7A both by RNA-Seq and IHC to classify ACC patients.
Discussion
In this study, we have implemented a bimodal approach using Dip test and Normal/Gaussian Mixture Models that led to the identification of SEMA7A as one of the genes with a broad range of bimodal distribution in ACC. Previously, Justino et al. introduced an integrated method to identify bimodal genes across multiple TCGA datasets (40). However, their analysis did not include ACC samples (26,40,41). Our aim was to explore a novel approach to classify ACC patients and capture heterogeneity across ACC tumors.
ACC presents considerable challenges due to its rarity, the complexity of its biology and its poor prognosis. Furthermore, due to the restricted number of cell lines and preclinical models, ACC patients have limited treatment options. Surgical resection continues to be the foremost approach for the treatment of localized cases, while metastasectomy has been proposed for the management of recurrent or metastatic ACC. This strategy carries significant therapeutic and palliative implications and is only associated with prolonged survival in carefully selected patients. Due to the prevalence of surgical interventions, high quality samples for biological and genomic analyses are routinely available at the Center for Cancer Research (CCR) of the National Cancer Institute at the National Institutes of Health (see Figs. 1 & 5).
To identify novel biomarkers for ACC, we have developed a method to identify bimodally expressed genes in ACC patient tumors to stratify patients based on their overall survival and discover predictive biomarkers. At first, we used RNA-seq expression profiles of ACC patient
tumors samples and computed the bimodal score and p-value for each gene. Our analysis revealed 72 bimodal genes having significant association with the overall survival of ACC patients. SEMA7A came up as the gene with the most significant association with prognosis independently of MKI67 (see Fig. 5A). In our analysis, we have observed that SEMA7A is significantly up regulated in ACC as compared to normal adrenal gland tissues (see Fig. 2). SEMA7A belongs to the Semaphorin gene family. It encodes SEMA7A (CDw108), a signaling glycoprotein (as its name indicates) anchored to the cell surface via glycosylphosphatidylinositol (GPI) linkage (42). Upon cleavage of the GPI membrane anchor, SEMA7A exists both as shed or membrane-bound. SEMA7A influences cellular dynamics in both cell-autonomous and non-autonomous manners. It enhances the mobility of melanocytes, neuronal cells and immune cells through the activation of -1 integrin signaling pathways (37,38,43). In its normal physiological role, SEMA7A binds ß1- integrin to activate downstream signaling cascades, including the pro-invasive MAPK/ERK and pro-survival PI3K/AKT pathways (44).
Recent findings have identified SEMA7A as a tumor promoter, accelerating tumor growth, increasing invasiveness, promoting adhesion to tumor-associated extracellular matrices. SEMA7A induces epithelial to mesenchymal transition, promoting metastasis in melanoma, glioblastoma, oral, kidney and breast cancers (36,45-49). Moreover, high expression of SEMA7A has been associated with significantly decreased patient survival in ER+ breast cancer patients (50). Prior to this work, no association between SEMA7A and ACC have been observed. Relevant to our study, in neuronal cells of the hypothalamus, SEMA7A expression is regulated by steroid hormones (51); an association we found also in ACC (see Fig. 2). SEMA7A is also upregulated in high grade (Stage III/IV) and hormone producing ACC patients (see Fig. 2). Our correlation analyses show that the SEMA7A-integrin-ß1 axes activates the ERK/MAPK signaling pathway in the subset of ACC patients with poor prognosis (see Fig. 3).
Immunohistochemistry (IHC) staining of ACC tissues shows bimodal distribution of SEMA7A in ACC patients with lower levels in normal adrenal tissues and benign adrenal tumors (see Figs. 4 & 5). The positive correlation between RNA-seq and immunohistochemistry (IHC) of NCI-ACC patient samples suggests that SEMA7A could be a promising prognostic biomarker for clinical applications. SEMA7A as a potential biomarker for further investigations and for application in clinical diagnostics, particularly for forecasting outcomes in ACC patients. Finally,
bioRxiv preprint doi: https://doi.org/10.1101/2025.04.03.647086; this version posted April 7, 2025. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
SEMA7A expression on the ACC tumors position it as a potential target for the development of immunotherapies, including antibody drug conjugates (ADC) and T-cell engagers.
Acknowledgements
Our work is supported by Center for Cancer Research, the Intramural Program of the National Cancer Institute, NIH (Z01-BC 006150). We wish to thank Jonathan Hernandez, CCR Surgical Oncology Program for the surgical samples.
Author Contribution
Y.P. supervised the study. Y.P., A.D and J.D.R. devised the concept. Y.P., A.D., F.I., A.L., S.V., W.C.R. and N.R. performed data analysis. D.T., S.K., C. E. and L.E., performed immunohistochemistry staining and analysis. A.D., J.D.R., Y.P., M.I.A and S.V. wrote and critically reviewed the manuscript. All authors reviewed the results and approved the final version of the manuscript.
Declaration of Interests
The authors declare no competing interests.
bioRxiv preprint doi: https://doi.org/10.1101/2025.04.03.647086; this version posted April 7, 2025. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
References
1. Else T, Kim AC, Sabolch A, Raymond VM, Kandathil A, Caoili EM, et al. Adrenocortical carcinoma. Endocr Rev 2014;35:282-326
2. Libe R, Huillard O. Adrenocortical carcinoma: Diagnosis, prognostic classification and treatment of localized and advanced disease. Cancer Treat Res Commun 2023;37:100759
3. Sidhu S, Sywak M, Robinson B, Delbridge L. Adrenocortical cancer: recent clinical and molecular advances. Curr Opin Oncol 2004;16:13-8
4. Fassnacht M, Allolio B. Clinical management of adrenocortical carcinoma. Best Pract Res Clin Endocrinol Metab 2009;23:273-89
5. Lam AK. Adrenocortical Carcinoma: Updates of Clinical and Pathological Features after Renewed World Health Organisation Classification and Pathology Staging. Biomedicines 2021;9
6. Cheng Y, Kou W, Zhu D, Yu X, Zhu Y. Future Directions in Diagnosis, Prognosis and Disease Monitoring of Adrenocortical Carcinoma: Novel Non-Invasive Biomarkers. Front Endocrinol (Lausanne) 2021;12:811293
7. Mizdrak M, Ticinovic Kurir T, Bozic J. The Role of Biomarkers in Adrenocortical Carcinoma: A Review of Current Evidence and Future Perspectives. Biomedicines 2021;9
8. Marquardt A, Landwehr LS, Ronchi CL, di Dalmazi G, Riester A, Kollmannsberger P, et al. Identifying New Potential Biomarkers in Adrenocortical Tumors Based on mRNA Expression Data Using Machine Learning. Cancers (Basel) 2021;13
9. Mohan DR, Lerario AM, Else T, Mukherjee B, Almeida MQ, Vinco M, et al. Targeted Assessment of G0S2 Methylation Identifies a Rapidly Recurrent, Routinely Fatal Molecular Subtype of Adrenocortical Carcinoma. Clin Cancer Res 2019;25:3276-88
10. Dinnes J, Bancos I, Ferrante di Ruffano L, Chortis V, Davenport C, Bayliss S, et al. MANAGEMENT OF ENDOCRINE DISEASE: Imaging for the diagnosis of malignancy in incidentally discovered adrenal masses: a systematic review and meta-analysis. Eur J Endocrinol 2016;175:R51-64
11. Tsai WH, Dai SH, Lee CC, Chien MN, Zeng YH. A Clinicopathological Analysis of Asian Patients with Adrenocortical Carcinoma: A Single-Center Experience. Curr Oncol 2023;30:4117-25
12. Ragazzon B, Libe R, Gaujoux S, Assie G, Fratticci A, Launay P, et al. Transcriptome analysis reveals that p53 and beta-catenin alterations occur in a group of aggressive adrenocortical cancers. Cancer Res 2010;70:8276-81
13. Waldmann J, Patsalis N, Fendrich V, Langer P, Saeger W, Chaloupka B, et al. Clinical impact of TP53 alterations in adrenocortical carcinomas. Langenbecks Arch Surg 2012;397:209-16
14. Gaujoux S, Grabar S, Fassnacht M, Ragazzon B, Launay P, Libe R, et al. beta-catenin activation is associated with specific clinical and pathologic characteristics and a poor outcome in adrenocortical carcinoma. Clin Cancer Res 2011;17:328-36
15. Terzolo M, Angeli A, Fassnacht M, Daffara F, Tauchmanova L, Conton PA, et al. Adjuvant mitotane treatment for adrenocortical carcinoma. N Engl J Med 2007;356:2372-80
16. Terzolo M, Fassnacht M, Perotti P, Libe R, Kastelan D, Lacroix A, et al. Adjuvant mitotane versus surveillance in low-grade, localised adrenocortical carcinoma (ADIUVO): an international, multicentre, open-label, randomised, phase 3 trial and observational study. Lancet Diabetes Endocrinol 2023;11:720-30
17. Fassnacht M, Terzolo M, Allolio B, Baudin E, Haak H, Berruti A, et al. Combination chemotherapy in advanced adrenocortical carcinoma. N Engl J Med 2012;366:2189-97
18. Pegna GJ, Roper N, Kaplan RN, Bergsland E, Kiseljak-Vassiliades K, Habra MA, et al. The Immunotherapy Landscape in Adrenocortical Cancer. Cancers (Basel) 2021;13
19. Zheng S, Cherniack AD, Dewal N, Moffitt RA, Danilova L, Murray BA, et al. Comprehensive Pan-Genomic Characterization of Adrenocortical Carcinoma. Cancer Cell 2016;29:723-36
20. Mohan DR, Borges KS, Finco I, LaPensee CR, Rege J, Solon AL, et al. beta-Catenin- Driven Differentiation Is a Tissue-Specific Epigenetic Vulnerability in Adrenal Cancer. Cancer Res 2023;83:2123-41
21. Assie G, Jouinot A, Fassnacht M, Libe R, Garinet S, Jacob L, et al. Value of Molecular Classification for Prognostic Assessment of Adrenocortical Carcinoma. JAMA Oncol 2019;5:1440-7
22. Morimoto R, Satoh F, Murakami O, Suzuki T, Abe T, Tanemoto M, et al. Immunohistochemistry of a proliferation marker Ki67/MIB1 in adrenocortical carcinomas: Ki67/MIB1 labeling index is a predictor for recurrence of adrenocortical carcinomas. Endocr J 2008;55:49-55
23. Almeida MQ, Bezerra-Neto JE, Mendonca BB, Latronico AC, Fragoso M. Primary malignant tumors of the adrenal glands. Clinics (Sao Paulo) 2018;73:e756s
24. Zhang F, Zhang F, Liu Z, Wu K, Zhu Y, Lu Y. Prognostic Role of Ki-67 in Adrenocortical Carcinoma After Primary Resection: A Retrospective Mono-Institutional Study. Adv Ther 2019;36:2756-68
25. Ertel A. Bimodal gene expression and biomarker discovery. Cancer Inform 2010;9:11-4
26. Moody L, Mantha S, Chen H, Pan YX. Computational methods to identify bimodal gene expression and facilitate personalized treatment in cancer patients. J Biomed Inform 2019;100S:100001
27. Murai J, Thomas A, Miettinen M, Pommier Y. Schlafen 11 (SLFN11), a restriction factor for replicative stress induced by DNA-targeting anti-cancer therapies. Pharmacol Ther 2019;201:94-102
28. Takashima T, Sakamoto N, Murai J, Taniyama D, Honma R, Ukai S, et al. Immunohistochemical analysis of SLFN11 expression uncovers potential non-responders to DNA-damaging agents overlooked by tissue RNA-seq. Virchows Arch 2021;478:569- 79
29. de Reynies A, Assie G, Rickman DS, Tissier F, Groussin L, Rene-Corail F, et al. Gene expression profiling reveals a new classification of adrenocortical tumors and identifies molecular predictors of malignancy and survival. J Clin Oncol 2009;27:1108-15
30. Chida D, Nakagawa S, Nagai S, Sagara H, Katsumata H, Imaki T, et al. Melanocortin 2 receptor is required for adrenal gland development, steroidogenesis, and neonatal gluconeogenesis. Proc Natl Acad Sci U S A 2007;104:18205-10
31. Ferraz-de-Souza B, Lin L, Achermann JC. Steroidogenic factor-1 (SF-1, NR5A1) and human disease. Mol Cell Endocrinol 2011;336:198-205
32. Rubtsov P, Karmanov M, Sverdlova P, Spirin P, Tiulpakov A. A novel homozygous mutation in CYP11A1 gene is associated with late-onset adrenal insufficiency and hypospadias in a 46,XY patient. J Clin Endocrinol Metab 2009;94:936-9
33. Sigala S, Bothou C, Penton D, Abate A, Peitzsch M, Cosentini D, et al. A Comprehensive Investigation of Steroidogenic Signaling in Classical and New Experimental Cell Models of Adrenocortical Carcinoma. Cells 2022;11
34. Ansell PJ, Zhou Y, Schjeide BM, Kerner A, Zhao J, Zhang X, et al. Regulation of growth hormone expression by Delta-like protein 1 (Dlk1). Mol Cell Endocrinol 2007;271:55-63
35. Nai-Yun Sun SK, Yoo Sun Kim, Diana Varghese, Arnulfo Mendoza, Rosa Nguyen, Reona Okada, Karlyne Reilly, Brigitte Widemann, Yves Pommier, Fathi Elloumi, Anjali Dhall, Mayank Patel, Etan Aber, Cristie Contreras-Burrola, Rosie Kaplan, Dan Martinez, Jennifer Pogoriler, Amber K. Hamilton, Sharon J. Diskin, M. Maris, Robert W. Robey, Michael M. Gottesman, Jaydira Del Rivero, Nitin Roper. Identification of DLK1, a Notch ligand, as an immunotherapeutic target and regulator of tumor cell plasticity and chemoresistance in adrenocortical carcinoma. 2024
36. Song Y, Wang L, Zhang L, Huang D. The involvement of semaphorin 7A in tumorigenic and immunoinflammatory regulation. J Cell Physiol 2021;236:6235-48
37. Pasterkamp RJ, Peschon JJ, Spriggs MK, Kolodkin AL. Semaphorin 7A promotes axon outgrowth through integrins and MAPKs. Nature 2003;424:398-405
38. Suzuki K, Okuno T, Yamamoto M, Pasterkamp RJ, Takegahara N, Takamatsu H, et al. Semaphorin 7A initiates T-cell-mediated inflammatory responses through alpha1beta1 integrin. Nature 2007;446:680-4
39. Garcia-Areas R, Libreros S, Amat S, Keating P, Carrio R, Robinson P, et al. Semaphorin7A promotes tumor growth and exerts a pro-angiogenic effect in macrophages of mammary tumor-bearing mice. Front Physiol 2014;5:17
40. Justino JR, Reis CFD, Fonseca AL, Souza SJ, Stransky B. An integrated approach to identify bimodal genes associated with prognosis in cancer. Genet Mol Biol 2021;44:e20210109
41. Pascual-Ahuir A, Fita-Torro J, Proft M. Capturing and Understanding the Dynamics and Heterogeneity of Gene Expression in the Living Cell. Int J Mol Sci 2020;21
42. Yamada A, Kubo K, Takeshita T, Harashima N, Kawano K, Mine T, et al. Molecular cloning of a glycosylphosphatidylinositol-anchored molecule CDw108. J Immunol 1999;162:4094-100
43. Reilkoff RA, Peng H, Murray LA, Peng X, Russell T, Montgomery R, et al. Semaphorin 7a+ regulatory T cells are associated with progressive idiopathic pulmonary fibrosis and are implicated in transforming growth factor-beta1-induced pulmonary fibrosis. Am J Respir Crit Care Med 2013;187:180-8
44. Garcia-Areas R, Libreros S, Iragavarapu-Charyulu V. Semaphorin7A: branching beyond axonal guidance and into immunity. Immunol Res 2013;57:81-5
45. Ma B, Herzog EL, Lee CG, Peng X, Lee CM, Chen X, et al. Role of chitinase 3-like-1 and semaphorin 7a in pulmonary melanoma metastasis. Cancer Res 2015;75:487-96
46. Zhang S, Kong F, Zheng L, Li X, Jia L, Yang L. SEMA7A as a Novel Prognostic Biomarker and Its Correlation with Immune Infiltrates in Breast Cancer. Int J Gen Med 2024;17:4081-99
47. Elder AM, Tamburini BAJ, Crump LS, Black SA, Wessells VM, Schedin PJ, et al. Semaphorin 7A Promotes Macrophage-Mediated Lymphatic Remodeling during Postpartum Mammary Gland Involution and in Breast Cancer. Cancer Res 2018;78:6473- 85
48. Wang Y, Li C, Qi X, Yao Y, Zhang L, Zhang G, et al. A Comprehensive Prognostic Analysis of Tumor-Related Blood Group Antigens in Pan-Cancers Suggests That SEMA7A as a Novel Biomarker in Kidney Renal Clear Cell Carcinoma. Int J Mol Sci 2022;23
49. Fu Y, Sun S, Bi J, Kong C. Construction of a risk signature for adrenocortical carcinoma using immune-related genes. Transl Androl Urol 2020;9:1920-30
50. Crump LS, Wyatt GL, Rutherford TR, Richer JK, Porter WW, Lyons TR. Hormonal Regulation of Semaphorin 7a in ER(+) Breast Cancer Drives Therapeutic Resistance. Cancer Res 2021;81:187-98
51. Parkash J, Messina A, Langlet F, Cimino I, Loyens A, Mazur D, et al. Semaphorin7A regulates neuroglial plasticity in the adult hypothalamic median eminence. Nat Commun 2015;6:6385
bioRxiv preprint doi: https://doi.org/10.1101/2025.04.03.647086; this version posted April 7, 2025. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
bioRxiv preprint doi: https://doi.org/10.1101/2025.04.03.647086; this version posted April 7, 2025. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
| Variable | Number of Patients (total = 112) |
|---|---|
| Age | |
| Median age < 49 years | 56 |
| Median age ≥ 49 years | 56 |
| Gender | |
| Female | 77 |
| Male | 35 |
| Tumor Stage | |
| Stage I | 12 |
| Stage II | 50 |
| Stage III | 18 |
| Stage IV | 29 |
| Not Available | 3 |
| Vital Status | |
| Alive | 75 |
| Dead | 37 |
| Median overall Survival (OS) | |
| < 43 months | 58 |
| ≥ 43 months | 54 |
| Hormone Producing | |
| Yes | 64 |
| No | 35 |
| Not Available | 13 |
A
C
Dataset Collection
18 Bimodal Genes (Good Prognostic Markers)
54 Bimodal Genes (Bad Prognostic Markers)
Correlation with MKI67
TCGA
NCI
Positive High (15) >0.3
Genes significantly associated with ACC patient survival
Cell Surface Metabolism Wnt signaling
112 Adrenal Cortical Carcinoma (ACC)
Positive Low (37) < 0.3
Genomic data
Clinical data
Negative (20)
Survival analysis (Median Cutoff)
RTN4R
p-value
TDGF1
RNA-Seq Expression Data log2 (FPKM+1) ~50,000 Genes
<0.0001
Top 200 bimodal genes
ARMH4
<0.002
PDE9A
0.01
CRMP1
0.02
Bimodal Approach
LEF1
0.05
0.4
ADGRE5
Statistical Test
0.3
> DipTest
> Normal/Gaussian mixture models (norMixEM) with k -means clustering algorithm
Density
LAMC3
0.2
EFNA3
PLPP2
0.1
SEMA7A
Rank Genes based on Bimodal Score (mu, w, sigma) and p-value < 0.05
0.0
0
1
2
3
4
8
Gene Expression
10
12
14
Hazard Ratio (HR)
B
a
b
Data source Tumor stage
Data source
U
Mol. subtype
NCI
CHST4
TCGA
CSDC2
C1B
Good Prognosis
NMÉ’S
C18
C1B
Tumor stage
28TB16
ZFYVE28
PINK1
I/II
ABMCX
C1B
ASTKO3 FASTKIDS
III/IV
Y
VATIL
Molecular Subtype
A2010883.3
AL 121895 2
AC243984.4
C1A
a1
PVT1
AC028315.2
C1B
AC005383.2
DNA replication Chromatin Cell cycle
AC231981.1
High
3
EKAPZL
2
MT
C1A
Bad Prognosis
1
LPP2
2299.1
0
NBP
P2
-1
-2
1
Low
-3
a2
871
:
Gene Expression
CAZA
Cell surface Wnt/ß-catenin signaling Metabolism
BRSK
OGRES
AS1
ATNAR
CENA
AC010996.1
CACNB2
Figure 1: (A) Study architecture and bimodal gene selection. (B) Unsupervised hierarchical clustering heatmap displaying 72 bimodal genes associated with ACC patient survival, tumor stage, molecular subtype (C1A/C1B), and PINK1, BUB1B, MKI67 signature. (C) Classification of the 72 bimodal genes into 3 groups based on correlation analysis with MKI67. Subgroup a1 is enriched for DNA replication markers and subgroup a2 for cell surface markers including SEMA7A.
bioRxiv preprint doi: https://doi.org/10.1101/2025.04.03.647086; this version posted April 7, 2025. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
A
8
TCGA Pan-Cancer
B
8
0
0
0
0
0
SEMA7A Expression
8
0
8
0
SEMA7A Expression
0
6
00
8
Kom o
8
0 0
8
@
4
4
8
:
2
0 00 0
00 0
2
0
0
0
0
0
8
0
8
0
0
0
0
PAAD
BLCA
ACC
UCS
HNSC
LUSC
LUAD
CESC
UVM
ESCA
CHOL
LAML
BRCA LIHC
0
0
KIRC
THCA
PRAD
KIRP
ACC Normal TCGA GTEx
C
Clinical stage
Hormone production
Gender
D
8-
8
8
SEMA7A Expression
1
SEMA7A expression
*
**
ns
6
6
6
Survival probability
0.75
Low
High
n=56
4
4
4
0.50
n=56
2
2
2
0.25
HR = 4.27
0
0
p-value = 0.0001 ***
0
0
I/II
III/IV
No
Yes
0
50
100
150
Female
200
Male
Time (Months)
E
Clinical Stage
1
Hormone Production
Gender
1
Survival probability
I/II
III/IV
Survival probability
No
1
0.75
n=62
0.75
Yes
Female
n=36
Survival probability
0.75
Male
n=35
0.50
0.50
0.50
0.25
HR = 5.5
n=47
0.25
HR = 3.01
n=64
0.25
n=77
p-value = 0.02*
HR = 1.05
0
p-value = 1.01e-05 ***
0
p-value = 0.883
0
50
100
Time (Months)
150
200
0
50
100
150
200
0
Time (Months)
0
50
100
150
200
Time (Months)
F
14
12
12
10
CYP17A1
10
10
CYP11A1
8
8
8
INHA
6
6
6
4
R=0.25
4
p-value = 0.0078
R=0.23
4
2
2
p-value = 0.01
2
R=0.468
0
0
“p-value = 1.899e-07
0
2
4
6
0
2
4
6
0
SEMA7A
SEMA7A
0
2
4
6
SEMA7A
14
8
12
8
10
6
8
6
DLK1
NR5A1
MC2R
4
:
6
4
4
2
R=0.206
2
R=0.267
2
p-value = 0.02
p-value = 0.004
R=0.159
0
0
0
p-value = 0.09
0
2
4
6
0
2
4
6
0
2
4
6
SEMA7A
SEMA7A
SEMA7A
A
SEMA7A
0.22*
ITGB1
AKT
0.49 ***
0.45 ***
! 0.63 ***
PTK/FAK
ERK/ MAPK1
LIMK2
CFL (Cofilin)
Regulation of actin cytoskeleton Invasiveness or Invasion/metastasis
0.69 ***
0.21*
B
0.67 ***
SEMA7A
ITGB1
1.00
AKT
0.75
PTK/FAK
0.50
ERK/MAPK1
0.25
LIMK2
0.00
CFL
Individual patients
C
1
Survival probability
SEMA7A pathway (Low)
0.75
n=56
0.50
SEMA7A pathway
0.25
n=56
(High)
HR = 2.55
p = 0.014*
0
0
50
100
150
200
Months
Figure 3: (A) MAPK/ERK signaling pathway is co-regulated via SEMA7A, (B) Heatmap shows the expression of neighboring genes of MAPK/ERK pathway in ACC patients. (C) Kaplan- Meier survival plot, based on average expression of SEMA7A pathway genes, divided into two groups based on median expression cut-off.
A
Normal adrenal gland
B
33%
37%
57%
SEMA7A
100
positive
(> 50%)
SEMA7A positivity (%)
Adrenocortical adenoma
50
0
ACC
Normal (n= 6)
Adenomas (n= 27)
ACC (n= 14)
50 um
A
8
NCI-5
6
NCI-7
MKI67 (log 2 (FPKM +1))
NCI-11.
·NCI-9
4
NCI=1
NCI-18
2
0
0
2
4
6
8
B
SEMA7A (log 2 (FPKM +1))
NCI-1
NCI-11
NCI-5
NCI-18
NCI-7
NCI-9
CE
8
available under aCC-BY-NC-ND 4.0 International license.
was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made bioRxiv preprint doi: https://doi.org/10.1101/2025.04.03.647086; this version posted April 7, 2025. The copyright holder for this preprint (which
Supplementary Data
GTEx Normal Bulk Tissue Expression (SEMA7A)
Pancreas
Heart - Left Ventricle
Liver
Muscle - Skeletal
Heart - Atrial Appendage
Artery - Tibial
Kidney - Cortex
Thyroid
Figure S1: Distribution of SEMA7A expression across a wide range of normal human tissues obtained from GTEx database.
Artery - Aorta
Brain - Nucleus accumbens (basal ganglia)
Stomach
Adipose - Visceral (Omentum)
Fallopian Tube
Colon - Transverse
Minor Salivary Gland
Adipose - Subcutaneous
Colon - Sigmoid
Breast - Mammary Tissue
Adrenal Gland
Kidney - Medulla Esophagus - Gastroesophageal Junction
Prostate
Esophagus - Muscularis
Cervix - Ectocervix
Pituitary
Cervix - Endocervix
Artery - Coronary
Vagina
Uterus
Bladder
Esophagus - Mucosa
Whole Blood
Brain - Caudate (basal ganglia)
Brain - Amygdala
Brain - Hypothalamus
2.5-
Brain - Putamen (basal ganglia)
Small Intestine - Terminal Ileum
Lung
Brain - Anterior cingulate cortex (BA24)
Skin - Not Sun Exposed (Suprapubic)
2.0-
Skin - Sun Exposed (Lower leg)
log10(TPM+1)
Ovary
Brain - Hippocampus
.5-
Brain - Substantia nigra
Brain - Cortex
Brain - Frontal Cortex (BA9)
1.0-
Cells - Cultured fibroblasts
Nerve - Tibial
Brain - Cerebellar Hemisphere
0.5-
Testis
Brain - Cerebellum
Brain - Spinal cord (cervical c-1)
Spleen
Cells - EBV-transformed lymphocytes