Neuroendocrine tumors belong to the group of cancers that frequently metastasize and that are difficult to treat with anticancer drugs. They are derived from the neuroectoderm and some neuroendocrine tumor types of secret hormones (e.g., adrenalin, noradrenalin). Phaeochromocytomas and paragangliomas are rare tumors mainly deriving from neuroectodermic adrenal tissue. Those deriving from sympathetic paraganglia are termed paragangliomas. While most phaeochromocytomas and paragangliomas represent benign adenomas, some are malignant carcinomas, indicating that these tumors may have specific genetic alterations determining some for more benign and some others for more malignant phenotypes. However, even benign tumors can exert detrimental effects, e.g., paragangliomas closely located to cranial nerves and vasculature may lead to their compression or invasion (1). Phaeochromocytomas and paragangliomas are usually treated by surgery and radiotherapy rather than by drugs.

Gene signatures are in many cases used to select a group of patients for whom a particular treatment will be effective. The use of gene signatures to stratify tumors into prognostic and predictive subtypes is rapidly growing, and studies have shown increased numbers of gene signatures in the more common cancers, yet for rare and uncommon cancers gene signatures have not been studied in detail. Genetic profiling of phaeochromocytomas and paragangliomas revealed a considerably high genetic instability and quite a number of risk genes (2-7). In the present investigation, our study focused on identifying genes and gene signatures with prognostic relevance for patient survival as prognostic prediction models and tools could contribute significantly to the clinical treatment of patients. Current research is also focused on the identification of novel therapeutic targets and identifying subjects for optimal benefits of specific treatments.

The aim of the present study was to identify gene novel signatures that are useful as prognostic tools for the survival time of cancer patients. Therefore, we investigated the gene expressions deposited in the Cancer Genome Atlas (TCGA) database (8). We identified two gene signatures which have not been described before. In addition, the genes of these gene signatures may also serve as new targets for the development of novel targeted drugs in the future (9).

Compilation of risk genes. For the appropriate literature selection, we followed the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) guideline (https://www.prisma-statement.org) (Figure 1). We screened the PubMed Literature database with the keywords “risk gene AND phaeochromocytoma AND review” as well as “risk gene AND paraganglioma AND review”. At total of 174 publications were identified for phaeochromocytoma, and 145 articles for paraganglioma. These papers were visually inspected and 21 publications containing compilations of risk genes were selected for our own compilation of 186 genes (Table I). These genes are described in the literature to contribute to the development and progression of phaeochromocytoma. The expression of these genes was reported by the Cancer Genome Atlas (TCGA) (9).

Kaplan-Meier survival statistics. The Kaplan-Meier statistics is a standard technique to calculate the survival probability of cancer patients according to their clinical, biochemical, or molecular parameters. In the present study, we used the KM Plotter algorithm (https://komplot.com/analysis) as described (10,11). To avoid type I errors of multiple comparisons, we used false discovery rate corrections (12) with a cut-off of 5%. The database of the KM Plotter consists of 7,489 biopsies from different tumor types from TCGA, including phaeochromocytoma and paraganglioma.

Our workflow is shown in Figure 2. We started by analyzing all 186 genes in patients with phaeochro-mocytoma/paraganglioma or 17 other tumor types, respectively, using Kaplan-Meier statistics. Two correlating genes in phaeochromocytoma/paraganglioma and 12 commonly correlating genes in four other tumor types were subsequently subjected to combined Kaplan-Meier statistics. The prognostic relevance of these 2-gene and 12-gene signatures were determined for either all tumors or those subsets of tumors with high mutational rates. This was done to prove whether the gene signatures were also of prognostic value for survival in highly mutated tumors that are usually more aggressive than those with low mutational loads.

Analyses of phaeochromocytoma- and paraganglioma-related risk genes with the survival times of patients. We performed Kaplan-Meier statistical calculations of 186 genes for 178 biopsies from phaeochromocytoma and paraganglioma (Table I). Surprisingly, the mRNA expression of only two genes (AQP4 and FAM83H) significantly correlated with a worse prognosis of patients, i.e., shorter overall survival times (p<0.05 and FDR<5%) (Figure 3A and C). Then, we used the mean mRNA expression profiles of both genes together for a combined Kaplan-Meier analysis. As expected, a significant correlation between high gene expression and short overall survival was found (p=6.1×10^-5 and FDR=2%; Figure 3E), indicating the prognostic value of this 2-gene signature.

It is generally accepted that tumor mutations not only lead to carcinogenesis but also to tumor progression, and failure of therapy due to tumor heterogeneity and outgrowth of subpopulations with selection advantages in the tumor evolution (13). Therefore, we further compared subgroups of phaeochromocytoma and paraganglioma, i.e., tumors with high and low mutation burden. Tumors with high mutation burden and high expression of AQP4 alone, FAM83H alone, or the mean expression of both genes had significant shorter overall survival times than those with low expression of these genes (Figure 3B, D, and F). These correlations were not detectable in tumors with low mutation burden.

Analyses of phaeochromocytoma-and paraganglioma-related risk genes with the overall survival times of patients with 20 other tumor types. The 186 risk genes may not be specific exclusively for this tumor type but may also play a role in other tumor types. Therefore, we addressed the question whether these risk genes may also be related to the overall survival times of patients suffering from other tumor types. The TCGA-based data repository of the KMPlotter contained several other tumor types in addition to phaeochromocytoma and paraganglioma with a total number of 7,489 biopsies. We calculated the overall survival times of patients not only with phaeochromocytoma and paraganglioma but also with these other tumor types with a total number of 3,162 Kaplan-Meier statistics calculations. The plots shown in Figure 4 and Figure 5 depict the 181 correlations which were statistically significant (yellow marked, p<0.05, FDR <5%). All others did not fulfill these criteria (blue marked). Of the 181 statistically significant correlations, 56 were associated with patient survival of one cancer entity, 22 with two cancer types, five with four and eight with five tumor types. The mRNA expression of one gene correlated with six tumor types.

We were interested to see which tumor types were most frequently associated with significant Kaplan-Meier correlations. As shown in Figure 6, renal papillary cell carcinoma, renal clear cell carcinoma, and hepatocellular carcinoma of the liver were the most frequently correlating tumor types with 27 or more significant associations, followed by lung adenocarcinoma, uterine corpus endometrial carcinoma and pancreatic ductal carcinoma (with 18 or more significant correlations). As also mentioned above, phaeochromocytoma and paraganglioma were only associated with two genes.

Based on the results shown in Figure 4, we investigated which tumor types correlated most frequently with the set of 186 genes. Out of the 18 tumor entities studied, the four most frequently appearing cancer types were renal papillary cell carcinoma, renal clear cell carcinoma, liver hepatocellular carcinoma, and lung adenocarcinoma (Figure 7). The expression of 18-32 genes correlated with worse overall survival of patients with these four tumor types. Therefore, we selected these four tumor types for our further analyses.

Overall survival analyses with a 12-gene signature. We investigated the accumulation of significant correlations in these tumor types in more detail. Renal papillary cell carcinoma, renal clear cell carcinoma, liver hepatocellular carcinoma, and lung adenocarcinoma showed significant correlations with 12 genes in common (BUB1, BUB1B, CDK1, CENPA, CKAP2L, IQGAP3, MKI67, NDC80, PBK, RRM2, TOP2A, and TTK) (Table II, Figure 4 and Figure 5). Then, we used the mean mRNA expression of these 12 genes and performed Kaplan-Meier analyses. As shown in Figure 7 A-D, the high expression of this 12-gene signature correlated with shorter overall survival in these four tumor entities compared to its low expression (p<0.0001; FDR=1%).

We also performed Kaplan-Meier statistics for high or low mutation burden and found a significant correlation for patients with highly mutated hepatocellular carcinoma. Their high mean expression of the 12-gene signature correlated with shorter survival, while low expression was associated with longer survival (p<0.00001; FDR=1%).

Refractory-free survival analyses with a 12-gene signature. While overall survival times are measured from initial diagnosis of the tumor to the death of the patient, it is also interesting to study the refractory-free survival, i.e., the time from initial diagnosis to the reappearance of a tumor after therapy. Refractory-free survival is usually associated with better life quality of a patient compared to overall survival. Therefore, we performed Kaplan-Meier statistics of the four tumor types mentioned above (renal papillary cell carcinoma, renal clear cell carcinoma, liver hepatocellular carcinoma, and lung adenocarcinoma) regarding refractory-free survival times of the patients. As shown in Figure 8, the high mean mRNA expression of these 12 genes significantly correlated with shorter refractory-free survival times of patients with hepatocellular carcinoma (Figure 8A and B) or renal papillary cell carcinoma (Figure 8C and D). This was true if all tumors were subjected to survival analysis (Figure 8A and C) (p<0.001; FDR ≤2%). Tumors with many mutations are generally accepted as being more aggressive than those with fewer mutations. Therefore, we investigated whether this 12-gene signature is also predictive of shorter survival in tumors with high mutation rates. Indeed, high gene expression also indicated lower survival times of patients with high mutation rates (Figure 8B and D) (p<0.0001; FDR ≤1%).

The aim of the present investigation was, first, to investigate whether genes whose expression has been correlated to the onset of phaeochromocytoma and paraganglioma may also be relevant for the outcome of this disease, i.e., for the survival of patients. Since these genes are not exclusively specific for phaeochromocytoma/paraganglioma, we assumed that they might also be relevant for the survival of patients suffering from other tumor types than phaeochromocytoma/paraganglioma. Therefore, a second aim of his study was to see whether the genes with prognostic relevance for the survival of phaeochro-mocytoma or paraganglioma patients may also predict survival of patients with other tumor types. Based on literature mining, we compiled 186 risk genes for phaeochromocytoma/paraganglioma. As risk genes may contribute to carcinogenesis and tumor progression, their final role for the lifetime expectancy of cancer patients is not that well understood. To identify prognostic biomarkers for patient survival that may also serve as possible targets for future drug developments, we subjected the mRNA expression of these genes in 178 biopsies of phaeochromocytoma and paraganglioma deposited in the TCGA database to Kaplan-Meier survival analyses. Unexpectedly, the expression of only two genes (AQP5 and FAM83H) significantly correlated with short survival times of patients. Since the 186 genes are not exclusively related to the development of phaeochromocytoma/ paraganglioma alone, we also investigated their prognostic value in 17 other tumor types. Interestingly, we observed a cluster of 12 genes that commonly appeared in four tumor types (renal papillary cell carcinoma, renal clear cell carcinoma, hepatocellular carcinoma, and lung carcinoma).

While most phaeochromocytomas and paragangliomas are benign and only a fraction exerts malignant features such as metastasis, the other four tumor types are malignant. Hence, it can be speculated that the two genes we identified in phaeochromocytoma and paraganglioma may be more associated with benign tumor growth, while the other 12 identified genes are related to malignant cancer growth. Interestingly, the majority of proteins encoded by these 12 genes have functions in mitosis or DNA metabolism, while the two genes with prognostic significance for phaeochromocytoma had other functions (water homeostasis, cell migration) (Table II). Aberrant mitosis and DNA metabolism are well-known characteristics of cancer, and it is plausible that genes related to these two biological processes have prognostic value for malignant rather than for benign tumor growth.

The MKI-67 gene represents a surrogate marker to monitor the proliferative capacity of tumors. Its encoding protein Ki-67 is widely used as prognostic marker in many cancer types including the carcinoma types studied here (14). A high Ki-67 score is associated with high proliferation, and rapid proliferating tumor cells are generally more susceptible to chemotherapy than slowly growing ones.

On the other hand, AQP4 and FAM83H may be valuable biomarkers to predict survival of phaeochromocytoma/ paraganglioma with limited malignant potential. A role of AQP4 has been described on other tumors of the brain such as low-grade glioma (15,16). AQP4 aggregation influences plasma membrane dynamics to alter cell proliferation, invasiveness, migration, and apoptotic potential in glioma cells.

FAM83H has a role for cell migration. Hence, this gene may play a role in metastatic phaeochromocytoma. FAM83H and other members of the FAM83 family are involved in diverse cancer types and play a role for poor survival prognosis (17). The 12 genes associated with shorter survival times of the four carcinoma types are partwise well-known for their prognostic role (e.g., BUB1, CDK1, RRM2, TOP2A), partwise their prognostic value is not well studied yet (e.g., AQP4, FAM83H, CKAP2L, TTK), and our investigation provides further evidence to consider them as valuable biomarkers in the future. As the majority of investigations come from Chinese authors, it comes as no surprise that tumor types that are very common in China (i.e., hepatocellular carcinoma, lung adenocarcinoma) have been more frequently studied than others (i.e., kidney cancers) (Table II).

We were not only interested identifying single prognostic markers but also studying entire gene signatures. The clustering of 12 genes in the four carcinoma types is remarkable, and forming the mean mRNA expression value of these 12 genes even improved the statistical significance in the Kaplan-Meier statistics making this gene signature a powerful tool to predict the survival probability of patients. Although other signatures have been reported in the literature related to several pathways and mechanisms (e.g., ferroptosis, immune response, glycolysis, etc.) (18-27), the mitosis-related 12-gene signature described in the present study is novel and not described before.

We also investigated the mean mRNA expression of AQP4 and FAM83H and found an improved statistical significance to predict survival of phaeochromocytoma and paraganglioma patients compared to the expression of both genes alone. Hence, this may serve as a novel 2-gene signature for these tumor types.

Novel biomarkers are not only valuable for the survival prognosis of patients but also for development of novel strategies for individualized treatment options. Mitosis and the DNA are also important treatment targets for classical anticancer drugs such as Vinca alkaloids, taxanes, alkylating agents, platin derivatives, and antimetabolites. Hence, the proteins encoded by the genes identified here may serve as novel targets for drug development.

During the past years, it became more and more clear that the current armamentarium of targeted drugs addressing the currently known drug targets can improve therapy success of cancer patients to some extent, but satisfying long-term cures are not reachable in many cases. Therefore, new targets for new drugs are urgently required.

The biomarkers we found in our investigations were mainly related to the mitotic spindle and DNA metabolism. Mitosis and the DNA are also important treatment targets for classical anticancer drugs such as Vinca alkaloids, taxanes, alkylating agents, platin derivatives, and antimetabolites. However, the proteins identified in this study are novel candidates for targeted treatment. Some of the proteins encoded by the genes we identified are already used as drug targets (e.g., TOP2A, CDK1) (28,29), others are recognized but not yet largely exploited for drug discovery (30-35). Hence, there is considerable potential to identify novel inhibitors in the future for the proteins encoded by the genes identified in the present investigation.

The Authors declare no conflicts of interest.

Conceptualization, T.E.; methodology, E.O.; formal analysis, T.E.; investigation, E.O. and T.E.; data curation, E.O. and T.E.; writing - original draft preparation, E.O. and T.E.; writing - review and editing, T.E.; supervision, T.E.; project administration, T.E. Both Authors have read and agreed to the published version of the manuscript.

No artificial intelligence (AI) tools, including large language models or machine learning software, were used in the preparation, analysis, or presentation of this manuscript.