Abstract
Pheochromocytomas and paragangliomas (PPGLs) are rare neuroendocrine, usually benign, tumors. Currently, the only reliable criterion of malignancy is the presence of metastases.
The aim of this study was to identify genes associated with malignancy in PPGLs.
Transcriptomic profiling was performed on 40 benign and 11 malignant PPGLs. Genes showing a significantly different expression between benign and malignant PPGLs with a ratio ≥4 were confirmed and tested in an independent series by quantitative real-time polymerase chain reaction (qRT-PCR). Immunohistochemistry was performed for the validated genes on 109 benign and 32 malignant PPGLs. Functional assays were performed with hPheo1 cells.
This study was conducted at the Department of Pathology of the Erasmus MC University Medical Center Rotterdam Human Molecular Genetics laboratory of the de Duve Institute, University of Louvain.
PPGL samples from 179 patients, diagnosed between 1972 and 2015, were included.
Associations between gene expression and malignancy were tested using supervised clustering approaches.
Ten differentially expressed genes were selected based on messenger RNA (mRNA) expression array data. Contactin 4 (CNTN4) was overexpressed in malignant vs benign tumors [4.62-fold; false discovery rate (FDR), 0.001]. Overexpression at the mRNA level was confirmed using qRT-PCR (2.90-fold, P = 0.02; validation set: 4.26-fold, P = 0.005). Consistent findings were obtained in The Cancer Genome Atlas cohort (2.7-fold; FDR, 0.02). CNTN4 protein was more frequently expressed in malignant than in benign PPGLs by immunohistochemistry (58% vs 17%; P = 0.002). Survival after 7 days of culture under starvation conditions was significantly enhanced in hPheo1 cells transfected with CNTN4 complementary DNA.
CNTN4 expression is consistently associated with malignant behavior in PPGLs.
Pheochromocytomas and paragangliomas (PPGLs) are rare chromaffin-cell tumors arising from the adrenal medulla and from the extra-adrenal sympathetic and parasympathetic paraganglia, respectively. Pheochromocytomas (PCCs) and extra-adrenal sympathetic paragangliomas (PGLs) secrete catecholamines, leading to episodes of severe hypertension, whereas parasympathetic PGLs in the head and neck region are usually not secretory but can cause problems due to local expansive growth. Although PPGLs are usually benign, a subset of at least 10% of patients may develop metastases. Metastatic behavior is more often observed when the tumor is caused by mutations in SDHB, MAX, and FH (1, 2).
Around 40% of PPGLs are caused by a mutation in one of the known susceptibility genes (1). The first PPGL-related genes were discovered two decades ago and were associated with hereditary syndromes, including multiple endocrine neoplasia syndrome (RET), von Hippel-Lindau disease (VHL), and neurofibromatosis type 1 (NF1). A decade later, mutations were identified in the SDH genes (SDHA, SDHB, SDHC, SDHD, and SDHAF2) in patients with PCC-PGL syndrome (3–7). In the last 5 years, new susceptibility genes, including EPAS1, FH, HRAS, MAX, PHD1, PHD2, and TMEM127, have been associated with a small proportion of PPGLs (8–15). Almost all PPGL-related genes are involved in the hypoxia or the AKT/mTOR pathways (16). Accordingly, several messenger RNA (mRNA) expression studies have shown that PPGLs can be divided into two main clusters according to their genotype (17–19). Cluster 1 includes VHL- and SDH-related tumors and is characterized by overexpression of genes coding for hypoxia-inducible proteins (pseudohypoxia pathway). Cluster 2 includes RET-, NF1-, TMEM127-, MAX-, and HRAS-related tumors and is associated with hyperactivation of the MAPkinase/AKT/mTOR pathway (14, 17–21). Cluster 1 tends to include more malignant PPGLs than cluster 2 (22).
PPGLs are considered malignant only if a metastasis occurs in nonchromaffin tissues (e.g., in liver, bones, lung, or lymph nodes). In the absence of curative treatment of metastatic PPGLs, 5-year survival is limited to 50% in patients with metastasis, compared with >95% in patients with benign PPGLs (23–25). Identification of markers predictive of malignant evolution should allow a more aggressive surgical treatment and closer follow-up of at-risk patients as well as early detection of recurrence or local tissue invasion. Such markers may provide new insights into the mechanisms leading to malignant transformation and thus may facilitate identification of potential targets for therapy.
Different histological grading systems, such as the Pheochromocytoma of the Adrenal Gland Scaled Score (PASS) and the recently described Grading System for Adrenal Pheochromocytoma and Paraganglioma (GAPP), have been developed to identify PPGLs with a malignant potential. PASS scoring is based on histological characteristics only, whereas GAPP grading combines histological criteria, Ki67 immunohistochemical staining, and the secretory profile of the tumor (26, 27). However, due to the poor interobserver reproducibility of PASS scoring, it was recommended not to use it for clinical prognostication (28). Furthermore, although the GAPP system includes other clinical features, it remains sensitive to interobserver variability, especially among less specialized pathologists, and has not been validated in an independent PPGL series so far. Other studies have focused on molecular differences by investigating gene expression profiles of benign and malignant tumors, but none of the differentially expressed genes was validated at the protein level (29–31).
There is no reliable marker allowing malignant PPGLs to be distinguished from benign, and little is known about the mechanisms leading to malignancy in these tumors. Therefore, the aim of the current study was to identify markers associated with malignant behavior. mRNA expression arrays were performed in a large series of benign and malignant PPGLs, followed by validation by quantitative real-time polymerase chain reaction (qRT-PCR) and immunohistochemistry on an independent series of PPGLs. The effects of Contactin 4 (CNTN4) expression on cell survival were investigated in hPheo1 cells.
Materials and Methods
Samples
This study included a series of 179 tumors (142 PCCs, 37 PGLs) diagnosed between 1972 and 2015, of which 115 came from the Erasmus MC (Rotterdam, Netherlands), 36 from different Dutch and Belgian centers, and 20 from the Spanish National Cancer Research Center (Supplemental Table 1). The Erasmus MC frozen tissues were collected and stored by the Erasmus MC tissue bank, including the healthy normal adrenal tissues used for the qRT-PCR experiments. Only cases with a proven metastasis in nonchromaffin tissues were considered as malignant (n = 40). The median follow-up was 64 months (range, 1 to 396 months) for the 139 benign cases (Supplemental Table 1). Several series of tumors were used to confirm and validate the results. Transcriptomic profiling was performed on 40 benign and 11 malignant PPGLs, of which four benign and three malignant tumors were used for confirmation by qRT-PCR (Fig. 1; Supplemental Table 1). Subsequently, eight additional tumors (four benign, four malignant) that were not used for the transcriptomic profiling were analyzed by qRT-PCR to validate the results (Fig. 2). Finally, immunohistochemistry was performed on 141 tumors (109 benign, 32 malignant) for two selected proteins: Interleukin 13 Receptor Alpha 2 and CNTN4 (Fig. 1; Supplemental Table 1). All tumor samples were used in accordance with the code for adequate secondary use of tissue, code of conduct “Proper Secondary Use of Human Tissue” established by the Dutch Federation of Medical Scientific Societies (http://www.federa.org). The study was approved by the Erasmus Medical Center-University Medical Center Rotterdam Institutional Review Board.
Flow diagram of the tumors used per analysis and a summary of the corresponding results. In total, 51 tumors were included in the mRNA expression array analyses (40 benign, 11 malignant). Of these, seven PPGLs were used for the confirmation by qRT-PCR (four benign, three malignant). Validation by qRT-PCR was performed in an additional set of eight independent tumors (four benign, four malignant). For immunohistochemistry, we used 101 PPGLs (88 benign, 13 malignant), of which 82 had not been used in the previous analyses, 15 were also used in the mRNA expression array analysis, and three had been included in the validation 2 series for qRT-PCR.
Graphs of the qRT-PCR results of the (A) confirmation and the (B) validation of CNTN4. x-Axis: normal adrenal, benign PPGLs, and malignant PPGLs. y-Axis: mean of the fold change for each group relative to the expression mean of benign tumors. All results were normalized to HPRT1. The P values of the confirmation and validation were 0.02 and 0.005, respectively.
Transcriptomic profiling
Total RNA was isolated using RNA-Bee according to the manufacturer’s instructions (Amsbio, Abingdon, United Kingdom) for 40 benign and 11 malignant fresh frozen PPGLs (Supplemental Table 1). Histological evaluation ensured the presence of at least 70% neoplastic cells in the tumor specimens. Complementary DNA (cDNA) was generated with oligo-dT primer and Superscript II Reverse transcription (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer’s protocol. Gene expression analysis was performed by One-Cycle Target Labeling using the 3′IVT Express kit according to the manufacturer’s instructions (Affymetrix, Santa Clara, CA). After hybridization in the Tecan HS4800 PRO (Tecan Group Ltd., Mannedorf, Switzerland), the arrays were scanned with the Tecan Scanner (Tecan Group Ltd.). Details on methods are available on request. Raw mRNA data (CEL files) were further analyzed using “R” according to Affymetrix’s instructions (http://www.r-project.org) and an Affymetrix script from the Bioconductor website (http://bioconductor.org/). Background correction was performed according to the most commonly used “RMA” background correction method. Data were normalized according to the quantiles method. The data are available on the public database GEO (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE67066). To identify malignancy-specific expressed genes, supervised clustering was performed based on metastatic behavior of the PPGLs (benign vs malignant) using the freely available online program “Pomelo II” (http://pomelo2.bioinfo.cnio.es/), which performed a Limma test, correcting for multiple comparisons (first analysis). False discovery rates (FDRs) of <0.05 were considered as significantly different. Because pseudohypoxia-related (cluster 1) PPGLs are more frequently malignant compared with cluster 2 PPGLs, comparisons between benign and malignant PPGLs were also performed within the clusters. We performed a cluster analysis using the cluster 3 software (http://bonsai.hgc.jp/~mdehoon/software/cluster/) based on 86 hypoxia-inducible factor target genes. Subsequently, mRNA expression differences between benign and malignant PPGLs were investigated within each cluster using the Limma t test through Pomelo II.
Expression analysis by qRT-PCR
Total RNA from frozen tumor samples was extracted using TRIzol reagent (Invitrogen Life Technologies) and purified using the RNeasy Mini Kit (Qiagen, Benelux B.V.) for a confirmation set and an independent validation set of seven and eight tumors, respectively (Supplemental Table 1). First-strand cDNA was synthetized from mRNA with oligo-dT primer using the Moloney Murine Leukemia Virus Reverse transcription (Thermo Fisher Scientific) at 42°C for 60 minutes. All quantitative polymerase chain reactions were performed in triplicate on a 7500 Fast Real-time PCR System (Applied Biosystems, Waltham, MA) with an initial denaturation step of 20 seconds at 95°C, then over 40 cycles with denaturation for 3 seconds at 95°C and combined annealing/elongation for 30 seconds at 60°C using the Fast SYBR Green Master Mix (Applied Biosystems). Relative fold changes of mRNA expression levels were assessed using the 2(−ΔΔCt) method, normalizing with the hypoxanthine phosphoribosyltransferase 1 (HPRT1) signal, which was the most stably expressed housekeeping gene throughout the entire mRNA expression array series. Primer sequences are described in Supplemental Table 2. Finally, a t test was performed to compare the mean level of mRNA expression between benign and malignant samples.
Insilico analyses using the RNA sequencing data from The Cancer Genome Atlas
RNA expression data from 179 PPGLs was downloaded from The Cancer Genome Atlas (TCGA) data portal (32). Normalized counts of expression for each sample were joined into a unique expression matrix (rows = genes; columns = samples) that was submitted to edgeR (version 3.12.1), which extracted differentially expressed genes between the malignant and benign samples (33). Expression values corresponding to the CNTN4 gene were formatted to be compatible with GraphPad Prism (www.graphpad.com) to generate a scatter plot representing the distribution of the expression values of the benign and malignant samples. Finally, an unpaired two-tailed t test and a Mann-Whitney two-tailed test were performed with GraphPad Prism to investigate if CNTN4 was significantly upregulated in malignant PPGL.
Immunohistochemistry
To determine the expression of the selected genes at the protein level, immunohistochemistry was performed on four tissue micro arrays (TMAs) along with control tissue samples from normal adrenal, kidney, liver, and placenta. TMAs were generated using an automated TMA constructor (ATA-27; Beecher Instruments, Sun Prairie, WI) available at the Department of Pathology, Erasmus MC. For the four TMAs, 101 formalin-fixed, paraffin-embedded tumor blocks were selected from the tissue archive of the Erasmus MC department of pathology. The latter included 88 benign (53 sporadic, 13 RET, nine VHL, four NF1, three SDHD, three MAX, one SDHA, one SDHB, and one HRAS mutated) and 13 malignant PPGL (11 sporadic, one VHL, and one SDHB). For each tumor, representative areas were selected and marked on a hematoxylin and eosin–stained slide. Subsequently, two tissue cores (1 mm in diameter) were removed from the block and placed into the “recipient” paraffin block at a predefined position. CNTN4 immunohistochemistry was performed on 5-µM TMA sections using a dilution of 1/500 of the CNTN4 polyclonal antibody HPA021068 (Sigma-Aldrich, Saint-Louis, MO) as described by van Nederveen et al. (34). Omission of the primary antibody was used as negative control reaction. Cytoplasmic staining of the tumor cells was scored as positive. Afterward, 40 whole sections were stained for the CNTN4 protein. These included 21 benign (including two SDHD and one MAX) and 19 malignant (including four SDHB, two SDHD, one ATRX, one RET, and one SDHA). Three observers evaluated the immunohistochemistry results: E.K. and L.E. evaluated the TMAs and whole sections series, D.F.R. scored the TMAs, and R.d.K. scored the whole sections. For each tumor in the TMAs, the two cores per tumor and three observers resulted in six scores per tumor. The selected final score was the predominant one among the six scores per tumor (at least four out of six scores). If there was no predominant score, the tumor was not included in the analysis (n = 6 tumors). Finally, a Fisher’s exact test was performed using SPSS (IBM SPPS Statistics, version 20 for Microsoft Windows) to assess for significance in differences in protein expression between benign and malignant tumors. A P value <0.05 was considered statistically significant. CNTN4 staining was scored as negative or positive. The same system was applied to the whole sections, but only one score per observer was assigned for each tumor. The interobserver agreement was assessed using the kappa statistics.
Representative examples of immunohistochemistry staining of CNTN4. (A) Normal skin as positive control. (B) Negative staining. (C) Positive cytoplasmic staining.
Cell culture
Immortalized human PCC cells (hPheo1), developed by Hans K. Ghayee et al. (35), were cultured on plates coated with collagen (Sigma-Aldrich) in Dulbecco’s modified Eagle medium/F12 with 10% fetal calf serum, 5% horse serum, and 1% penicillin/streptomycin (Thermo Fisher Scientific) (35).
Transfection of CNTN4 cDNA
The plasmid p-CMV/ORF containing the CNTN4 cDNA (Bio-Connect, Huissen, Netherlands) was used to transfect the hPheo1 cell line using the jetPEI (Westburg, Leusden, Netherlands) and 15 μg of plasmid per 10-cm 70% confluent cell culture plate. The selection was performed for 10 days using 100 μg/mL of hygromycin (Sigma-Aldrich).
Functional assays
Nontransfected (NT) and transfected cells were plated (n = 300,000) in 35-mm, 6-well plates. To perform the starvation assay, cells were cultured with Dulbecco’s modified Eagle medium/F12 containing 1% of penicillin/streptomycin for 7 days. Media was changed every 3 days. After 7 days, cells were trypsinized with Trypsin-EDTA (Thermo Fisher Scientific) and counted with a Becker cell using Trypan blue solution (Sigma-Aldrich). For the cell adhesion assay, cells were cultured either on collagen-coated plates or on noncoated plates. After 24 hours, cells were counted using the Trypan blue solution (Sigma-Aldrich) and a Becker cell. The experiments were performed in triplicate, and one-way analysis of variance was performed using SPSS to investigate the functional consequences of CNTN4 expression.
Results
Transcriptomic profiling
The supervised clustering analysis (benign vs malignant PPGL) was performed and revealed 105 genes with an FDR < 0.05 (Supplemental Table 3). Of these, seven genes showed a relative overexpression ratio (fold change) of at least 4, including four genes that were overexpressed in malignant PPGLs (Table 1). It was not possible to investigate differentially expressed genes related to malignancy within different genotypes because the number of tumors per genotype was too small. Therefore, we generated clusters using an alternative approach. PPGLs were grouped based on the expression of the previously published 86 differentially expressed genes between “cluster 1” and “cluster 2” tumors, and the current PPGL series was investigated for differentially expressed genes between malignant and benign PPGLs within each cluster (36). This supervised cluster analysis resulted in two groups, the first containing 21 tumors (15 benign, six malignant) and the second containing 30 tumors (25 benign, five malignant) (Supplemental Table 4). Both groups were investigated for differences in mRNA expression between benign and malignant tumors. Within the first group, there appeared to be no significant difference between benign and malignant tumors. In the second group, 267 genes were differentially expressed (FDR < 0.05). A more stringent threshold of FDR < 0.01 and a relative fold change >4 resulted in 18 genes, of which three (CNTN4, IRX3, and SULF2) were selected because these genes were also significantly differentially expressed in the entire dataset (FDR < 0.05) (Table 1). Although the 10 selected genes have been investigated in the analyses described below, we will only describe the CNTN4 results because CNTN4 was the only gene that could be confirmed by all validation experiments and has accordingly been functionally investigated.
Comparison of the Statistical Results of the mRNA Expression Array, With Confirmation and Validation by qRT-PCR
| Gene | mRNA Expression Array | qRT-PCR Confirmation | qRT-PCR Validation | |||
|---|---|---|---|---|---|---|
| FDR | FC | P Value | FC | P Value | FC | |
| CBLN1 (Cerebellin 1 Precursor) | 0.02 | 3.85 | 0.06 | 0.02 | — | — |
| CNTN4 (Contactin 4) | 0.001 | 4.62 | 0.02 | 2.90 | 0.005 | 4.26 |
| IL13Rα2 (Interleukin 13 receptor alpha 2) | 0.01 | −4.62 | 0.01 | −6.67 | <0.001 | 3.47 |
| IRX3 (Iroquois homeobox 3) | 0.001 | 5.37 | 0.55 | 0.78 | — | — |
| ISL1 (ISL Lim homeobox 1) | 0.02 | −4.12 | 0.27 | −1.11 | — | — |
| LPHN3 (Latrophilin 3) | 0.00 | 4.23 | 0.62 | 1.41 | — | — |
| MATN2 (Matrilin 2) | 0.05 | 4.03 | 0.16 | 0.47 | — | — |
| MOXD1 (Monooxygenase DBH like 1) | 0.04 | −4.56 | 0.03 | −4.79 | 0.69 | 1.2 |
| SULF2 (Sulfatase 2) | 0.003 | 6.16 | 0.09 | 0.67 | — | — |
| WASF3 (WAS protein family member 3) | 0.03 | 3.91 | 0.20 | 2.66 | — | — |
| Gene | mRNA Expression Array | qRT-PCR Confirmation | qRT-PCR Validation | |||
|---|---|---|---|---|---|---|
| FDR | FC | P Value | FC | P Value | FC | |
| CBLN1 (Cerebellin 1 Precursor) | 0.02 | 3.85 | 0.06 | 0.02 | — | — |
| CNTN4 (Contactin 4) | 0.001 | 4.62 | 0.02 | 2.90 | 0.005 | 4.26 |
| IL13Rα2 (Interleukin 13 receptor alpha 2) | 0.01 | −4.62 | 0.01 | −6.67 | <0.001 | 3.47 |
| IRX3 (Iroquois homeobox 3) | 0.001 | 5.37 | 0.55 | 0.78 | — | — |
| ISL1 (ISL Lim homeobox 1) | 0.02 | −4.12 | 0.27 | −1.11 | — | — |
| LPHN3 (Latrophilin 3) | 0.00 | 4.23 | 0.62 | 1.41 | — | — |
| MATN2 (Matrilin 2) | 0.05 | 4.03 | 0.16 | 0.47 | — | — |
| MOXD1 (Monooxygenase DBH like 1) | 0.04 | −4.56 | 0.03 | −4.79 | 0.69 | 1.2 |
| SULF2 (Sulfatase 2) | 0.003 | 6.16 | 0.09 | 0.67 | — | — |
| WASF3 (WAS protein family member 3) | 0.03 | 3.91 | 0.20 | 2.66 | — | — |
Negative value indicates overexpressed in benign PPGLs; positive value indicates overexpressed in malignant PPGLs.
Abbreviation: FC, fold change.
Comparison of the Statistical Results of the mRNA Expression Array, With Confirmation and Validation by qRT-PCR
| Gene | mRNA Expression Array | qRT-PCR Confirmation | qRT-PCR Validation | |||
|---|---|---|---|---|---|---|
| FDR | FC | P Value | FC | P Value | FC | |
| CBLN1 (Cerebellin 1 Precursor) | 0.02 | 3.85 | 0.06 | 0.02 | — | — |
| CNTN4 (Contactin 4) | 0.001 | 4.62 | 0.02 | 2.90 | 0.005 | 4.26 |
| IL13Rα2 (Interleukin 13 receptor alpha 2) | 0.01 | −4.62 | 0.01 | −6.67 | <0.001 | 3.47 |
| IRX3 (Iroquois homeobox 3) | 0.001 | 5.37 | 0.55 | 0.78 | — | — |
| ISL1 (ISL Lim homeobox 1) | 0.02 | −4.12 | 0.27 | −1.11 | — | — |
| LPHN3 (Latrophilin 3) | 0.00 | 4.23 | 0.62 | 1.41 | — | — |
| MATN2 (Matrilin 2) | 0.05 | 4.03 | 0.16 | 0.47 | — | — |
| MOXD1 (Monooxygenase DBH like 1) | 0.04 | −4.56 | 0.03 | −4.79 | 0.69 | 1.2 |
| SULF2 (Sulfatase 2) | 0.003 | 6.16 | 0.09 | 0.67 | — | — |
| WASF3 (WAS protein family member 3) | 0.03 | 3.91 | 0.20 | 2.66 | — | — |
| Gene | mRNA Expression Array | qRT-PCR Confirmation | qRT-PCR Validation | |||
|---|---|---|---|---|---|---|
| FDR | FC | P Value | FC | P Value | FC | |
| CBLN1 (Cerebellin 1 Precursor) | 0.02 | 3.85 | 0.06 | 0.02 | — | — |
| CNTN4 (Contactin 4) | 0.001 | 4.62 | 0.02 | 2.90 | 0.005 | 4.26 |
| IL13Rα2 (Interleukin 13 receptor alpha 2) | 0.01 | −4.62 | 0.01 | −6.67 | <0.001 | 3.47 |
| IRX3 (Iroquois homeobox 3) | 0.001 | 5.37 | 0.55 | 0.78 | — | — |
| ISL1 (ISL Lim homeobox 1) | 0.02 | −4.12 | 0.27 | −1.11 | — | — |
| LPHN3 (Latrophilin 3) | 0.00 | 4.23 | 0.62 | 1.41 | — | — |
| MATN2 (Matrilin 2) | 0.05 | 4.03 | 0.16 | 0.47 | — | — |
| MOXD1 (Monooxygenase DBH like 1) | 0.04 | −4.56 | 0.03 | −4.79 | 0.69 | 1.2 |
| SULF2 (Sulfatase 2) | 0.003 | 6.16 | 0.09 | 0.67 | — | — |
| WASF3 (WAS protein family member 3) | 0.03 | 3.91 | 0.20 | 2.66 | — | — |
Negative value indicates overexpressed in benign PPGLs; positive value indicates overexpressed in malignant PPGLs.
Abbreviation: FC, fold change.
Confirmation and validation of the relative mRNA overexpression
To confirm the relative overexpression observed by transcriptomic profiling, qRT-PCR was performed for the 10 selected genes on two normal adrenals, four benign PPGLs, and three malignant PPGLs all included in the transcriptomic profiling. To further validate the results obtained with the confirmation set, qRT-PCR was performed on eight independent PPGLs (four benign, four malignant). The overexpression of CNTN4 in malignant PPGLs was confirmed and validated (Fig. 2).
In silico analyses using the TCGA RNA sequencing data
CNTN4 expression differences between benign and malignant PPGL were analyzed using the RNA sequencing data from 179 PPGLs (including 14 malignant primary tumors) from the TCGA. CNTN4 was significantly overexpressed in malignant PPGLs, with a mean fold change of 2.7 (FDR = 0.022) (Supplemental Fig. 1).
Detection of the protein expression
Immunohistochemistry was subsequently performed for CNTN4 using TMA, which included 88 benign and 13 malignant PPGLs. A significantly higher frequency of CNTN4-positive staining was observed in the malignant group (P = 0.002): 58% of the malignant PPGLs were scored positive vs 17% of the benign tumors (Table 2). Representative examples of positive and negative staining are shown in Fig. 3. The sensitivity and specificity of the CNTN4 staining were 64% and 83%, respectively. Moreover, CNTN4 was differentially expressed according to genotype: CNTN4 was absent in RET-mutated PPGL (P = 0.03) and positive in a high proportion of SDH-mutated PPGLs (83%) compared with other genotypes (P = 0.002) (Table 2).
Statistical Analyses of CNTN4 Staining
| CNTN4 | ||
|---|---|---|
| TMA | Full Slides | |
| Benign vs malignant PPGLs | P = 0.002 (81 benign and 13 malignant) positive: 58% of malignant vs 17% of benign | P = 0.17 positive: 44% of malignant vs 24% of benign |
| MAX-mutated PPGLs vs other genotypes | (two MAX-mutated and 91 others) | (one MAX-mutated and 39 others) |
| NF1-mutated PPGLs vs other genotypes | P = 0.34 (three NF1-mutated and 90 others) | — |
| RET-mutated PPGLs vs other genotypes | P = 0.03 (14 RET-mutated and 79 others) positive: 0% of RET-mutated vs 73% of others | (one RET-mutated and 39 others) |
| SDH-mutated PPGLs vs other genotypes | P < 0.001 (six SDH-mutated and 87 others) positive: 83% of SDH-mutated vs 18% of others | P = 0.001 positive: 86% of SDH-mutated vs 22% of others |
| VHL-mutated PPGLs vs other genotypes | P = 0.31 (10 VHL-mutated and 83 others) | — |
| CNTN4 | ||
|---|---|---|
| TMA | Full Slides | |
| Benign vs malignant PPGLs | P = 0.002 (81 benign and 13 malignant) positive: 58% of malignant vs 17% of benign | P = 0.17 positive: 44% of malignant vs 24% of benign |
| MAX-mutated PPGLs vs other genotypes | (two MAX-mutated and 91 others) | (one MAX-mutated and 39 others) |
| NF1-mutated PPGLs vs other genotypes | P = 0.34 (three NF1-mutated and 90 others) | — |
| RET-mutated PPGLs vs other genotypes | P = 0.03 (14 RET-mutated and 79 others) positive: 0% of RET-mutated vs 73% of others | (one RET-mutated and 39 others) |
| SDH-mutated PPGLs vs other genotypes | P < 0.001 (six SDH-mutated and 87 others) positive: 83% of SDH-mutated vs 18% of others | P = 0.001 positive: 86% of SDH-mutated vs 22% of others |
| VHL-mutated PPGLs vs other genotypes | P = 0.31 (10 VHL-mutated and 83 others) | — |
Statistical Analyses of CNTN4 Staining
| CNTN4 | ||
|---|---|---|
| TMA | Full Slides | |
| Benign vs malignant PPGLs | P = 0.002 (81 benign and 13 malignant) positive: 58% of malignant vs 17% of benign | P = 0.17 positive: 44% of malignant vs 24% of benign |
| MAX-mutated PPGLs vs other genotypes | (two MAX-mutated and 91 others) | (one MAX-mutated and 39 others) |
| NF1-mutated PPGLs vs other genotypes | P = 0.34 (three NF1-mutated and 90 others) | — |
| RET-mutated PPGLs vs other genotypes | P = 0.03 (14 RET-mutated and 79 others) positive: 0% of RET-mutated vs 73% of others | (one RET-mutated and 39 others) |
| SDH-mutated PPGLs vs other genotypes | P < 0.001 (six SDH-mutated and 87 others) positive: 83% of SDH-mutated vs 18% of others | P = 0.001 positive: 86% of SDH-mutated vs 22% of others |
| VHL-mutated PPGLs vs other genotypes | P = 0.31 (10 VHL-mutated and 83 others) | — |
| CNTN4 | ||
|---|---|---|
| TMA | Full Slides | |
| Benign vs malignant PPGLs | P = 0.002 (81 benign and 13 malignant) positive: 58% of malignant vs 17% of benign | P = 0.17 positive: 44% of malignant vs 24% of benign |
| MAX-mutated PPGLs vs other genotypes | (two MAX-mutated and 91 others) | (one MAX-mutated and 39 others) |
| NF1-mutated PPGLs vs other genotypes | P = 0.34 (three NF1-mutated and 90 others) | — |
| RET-mutated PPGLs vs other genotypes | P = 0.03 (14 RET-mutated and 79 others) positive: 0% of RET-mutated vs 73% of others | (one RET-mutated and 39 others) |
| SDH-mutated PPGLs vs other genotypes | P < 0.001 (six SDH-mutated and 87 others) positive: 83% of SDH-mutated vs 18% of others | P = 0.001 positive: 86% of SDH-mutated vs 22% of others |
| VHL-mutated PPGLs vs other genotypes | P = 0.31 (10 VHL-mutated and 83 others) | — |
We performed a CNTN4 immunohistochemistry staining on a second validation set of 40 tumors using whole sections (21 benign, 19 malignant) (Fig. 3). Using the consensus scores, a higher frequency of malignant tumors (44%) was positive compared with benign (24%) PPGLs, although the difference was not statistically significant (P = 0.17). Three observers evaluated the staining, and interobserver agreement was evaluated using kappa scores. The agreement between E.K and L.E was almost perfect (κ = 0.9; P < 0.001), whereas the agreement between R.d.K and E.K and between R.d.K and L.E was fair (κ = 0.3). Considering only the tumors with full agreement between the three assessors (15 benign, nine malignant), 67% of the malignant PPGLs were scored positive vs 20% of the benign tumors. This was statistically significant (P = 0.02). Furthermore, CNTN4 staining was positive in 86% of SDH-mutated PPGLs vs 22% of non–SDH-related OOGLs (P = 0.001).
Starvation assays
To investigate the functional effect of CNTN4, hPheo1 cells were used because they do not express CNTN4 (Supplemental Fig. 2). After transfection followed by an antibiotic selection, we generated a stably transfected cell line expressing CNTN4 at continuous levels (Supplemental Fig. 2). After culturing the cells for 7 days in starvation media, a large number of the NT cells were floating, whereas the hPheo1 cells overexpressing CNTN4 were confluent and well attached. The experiment was performed in triplicate, and cells from both conditions were counted. We observed that twice the number of cells were present in the hPheo1 cells overexpressing the CNTN4 compared with the NT cells (averages from the three experiments: 1,022,777 cells vs 496,111 cells; P = 0.02) (Fig. 4A).
Functional assays using NT and CNTN4-overexpressing hPheo1 cells. (A) Comparison between the number of NT and CNTN4-overexpressing hPheo1 cells after 7 days of starvation. The ratio relative to the number of cells plated at day 0 is 1.65 for the NT cells and 3.41 for the CNTN4-overexpressing cells. (B) Comparison between the number of NT and CNTN4-overexpressing hPheo1 cells after 1 day of culture on plastic. The ratio relative to the number of cells cultured on coated plates is of 2.2 for the NT cells and 3.6 for the CNTN4-overexpressing cells. *Statistically significant (P < 0.05).
Cell adhesion assays
The NT and stably transfected CNTN4-overexpressing hPheo1 cells were used for the cell adhesion experiment, in which we investigated the influence of cell coating on cell growth. After 24 hours, we counted almost the same number of overexpressing CNTN4 cells between the coated plates and the noncoated ones (1,177,000 and 1,050,000), and we counted 1.4 times less NT on the plastic than on the coated plates (1,181,000 and 860,000) (P < 0.001) (Fig. 4B).
Discussion
The aim of the current study was to identify genes associated with malignancy in PPGLs. To identify these genes, we performed transcriptomic profiling and found overexpression of CNTN4 in malignant PPGLs. These results were confirmed in a selection of samples, validated in an independent set of PPGLs by qRT-PCR, and further validated using RNA expression data from TCGA. In addition, immunohistochemistry showed that CNTN4 staining was more frequently present in malignant than in benign PPGLs. Finally, two types of functional assays showed that CNTN4 expression in hPheo1 cells resulted in a substantial survival benefit.
CNTN4 belongs to the neuronal immunoglobulin (Ig) superfamily, which is involved in cell surface interactions and is thought to have a role in guidance of axonal growth. This superfamily consists of six contactins, each of which is composed of six extracellular Ig-like domains and four extracellular fibronectin type III domains attached to the membrane by a glycosylphosphatidyl inositol anchor (37). CNTN4 mRNA overexpression in SDH-related PCCs was once reported (17). This is in accordance with our results. Interestingly, overexpression of another contactin family member, Contactin 1, has been shown to be associated with malignancy in different types of cancers, including lung adenocarcinomas, gastric cancers, and gliomas (38–40). In our study, we observed an important survival benefit due to overexpression of CNTN4, which could explain the role of CNTN4 in malignancy. To understand the mechanism of this survival benefit, we investigated the AKT and ERK signaling pathways by Western blotting with phospho-AKT and phospho-ERK antibodies (data not shown) because these pathways have previously been associated with increased survival under metabolic stress conditions (41). Yet, these signaling pathways do not seem to be differentially activated between cells overexpressing CNTN4 and the wild-type cells. The exact molecular mechanisms behind increased survival of CNTN4-overexpressing hPheo1 cells remains unclear.
In agreement with Burnichon et al. (17), CNTN4 was expressed in a higher proportion of SDH-related PPGLs compared with PPGLs harboring mutations in RET and other susceptibility genes. Although it is tempting to speculate that differential expression of CNTN4 may at least partly explain the differential risk of metastatic dissemination of PPGLs according to genotype (42), these findings need to be consolidated in larger groups of patients harboring germline mutations in different susceptibility genes.
Our study has some limitations. The main limitation is that some of the apparently benign PPGLs might develop metastases with time. In this regard, it might be interesting to continue the follow-up of patients with an apparently benign PPGLs, especially those with positive CNTN4 staining. Another concern is the small number of malignant tumors. Although our work needs replication in larger cohorts, this limitation reflects the rarity of metastatic evolution in PPGLs and may thus prove difficult to overcome.
In summary, mRNA expression profiling, qRT-PCR, and immunohistochemistry consistently show that CNTN4 (over)expression is associated with malignant PPGLs and thus may be a predictive marker for malignancy. Furthermore, CNTN4 overexpression contributes to the survival of hPheo1 cells under metabolic stress conditions, supporting its role in the malignant behavior of these tumors. After confirmation in other cohorts, it would be interesting to assess the predictive value of CNTN4 staining within an existing scoring system, such as PASS and GAPP, in which additional parameters are taken into account, such as SDHB mutation status, extra-adrenal localization, histological characteristics (increased cell density, necrosis, invasion), and/or more recently identified altered vascular growth patterns (26, 27, 43).
Abbreviations:
- cDNA
complementary DNA
- CNTN4
Contactin 4
- FDR
false discovery rate
- GAPP
Grading System for Adrenal Pheochromocytoma and Paraganglioma
- mRNA
messenger RNA
- NF1
neurofibromatosis type 1
- NT
nontransfected
- PASS
Pheochromocytoma of the Adrenal Gland Scaled Score
- PCC
pheochromocytoma
- PGL
paraganglioma
- PPGL
pheochromocytoma and paraganglioma
- qRT-PCR
quantitative real-time polymerase chain reaction
- TCGA
The Cancer Genome Atlas
- TMA
tissue micro array
- VHL
von Hippel-Lindau.
Acknowledgments
We thank Professor Anne Mourin (Cliniques Universitaires Saint-Luc, UCL, Brussels, Belgium) for continuous support and contributions to patient and tumor recruitment, Dr. Nicolas de Saint-Aubain (Institut Jules Bordet, Brussels, Belgium) for providing tumors, Dr. Nisha Limaye (de Duve Institute, UCL, Brussels, Belgium) for help and advice, Francesca Severino (Cliniques Universitaires Saint-Luc, UCL, Brussels, Belgium) for help in maintenance and improvement of the UCL PPGL database, and Dr. Adriaan P. de Bruïne (Department of Pathology, Maastricht University Medical Centre, Netherlands) and Dr. Marijke R. van Dijk (Department of Pathology, University Medical Centre Utrecht, Netherlands) for contributing tumor samples.
Financial Support: These studies were partly supported by Grant F.R.S.-FNRS-Télévie and Grant FRSM 3.4510.11F (Fonds National à la Recherche Scientifique to A.P.) and by a grant from the Foundation against Cancer, Belgium (to M.V.). L.E. was supported by a F.R.S.-FNRS-Télévie fellowship. This work was also supported by the Netherlands Organization for Health Research and Development (Grant 920-03-314), the Vanderes Foundation (project 139), and the European Science Foundation (project 3824; to E.K.) by Project PI14/00240 from Fondo de Investigaciones Sanitarias, Instituto de Salud Carlos III, co-financed by FEDER 2014–2020.
Disclosure Summary: The authors have nothing to disclose.
References
Author notes
These authors contributed equally to this study.
