Molecular Profiling of Pancreatic Neuroendocrine Tumors in Sporadic and Von Hippel-Lindau Patients

Von Hippel-Lindau (VHL) disease is an autosomal dominant disorder caused by a germline mutation in the VHL tumor suppressor gene, located in chromosome 3p25. Affected individuals are predisposed to develop a variety of neoplasms characterized by their marked phenotypic variability and the age-dependent penetrance (1, 2). Pancreatic neuroendocrine tumors (PanNET) develop in 10% to 17% of patients with VHL, and present specifics and differences compared with sporadic PanNET, including multiplicity, association with microadenomas and a high frequency of local tumor aggressiveness at histologic and macroscopic levels, showing peripancreatic organ invasion in 20% of the patients (1–4).

The VHL gene encodes a protein (pVHL) involved in the regulation of a transcription factor called hypoxia-inducible factor (HIF). Mutations within the VHL gene disrupt the interaction between HIF-α and pVHL, leading to constitutive HIF activation and expression of HIF targets, involved in many cellular processes, including angiogenesis and cell metabolism (1, 2). VHL also has independent effects of its role in degradation of HIF-α subunits.

In VHL disease, PanNET probably develop according to a specific pathway of carcinogenesis. To obtain further insight into this process, we used real-time quantitative reverse transcriptase PCR (RT-PCR) to quantify the mRNA expression of a large number of selected genes in VHL tumors in comparison with sporadic tumors. We particularly focused on the expression of genes related to angiogenesis, to the HIF/VHL pathway, to the processes of epithelial–mesenchymal transition (EMT) or metastasis, and involved in cellular mechanisms such as cell cycle and signaling (2, 5–8). Genes of interest were further studied at the protein level by immunohistochemistry and compared with microadenomas, which represent the early neuroendocrine tumor precursors.

Thirty-one patients with VHL (19 women and 12 men) that underwent pancreatic surgery for one or several PanNET between April 1982 and February 2009 were retrospectively studied (Table 1). In comparison with the previous publication of our group which described the pathologic data of the pancreas of 18 patients with VHL treated in Beaujon Hospital (Clichy, France), we added 13 cases in this study, 1 from Beaujon hospital (Clichy) and 12 from different cities of France (2). The median (range) age of patients at the time of surgery was 40.7 years (25–76). Surgery consisted of pancreaticoduodenectomy in 19 patients, distal pancreatectomy in 9 patients, 1 total pancreatectomy, and 2 limited resections of the pancreas.

Controls consisted of 16 non-VHL patients operated on in Beaujon Hospital for PanNET, either sporadic (n = 14) or in the context of a multiple endocrine neoplasia syndrome type 1 (MEN1; n = 2). The median (range) age of patients at the time of surgery was 53 years (33–72). Surgery consisted of 3 pancreaticoduodenectomy, 6 distal pancreatectomy, 3 median pancreatectomies, and 4 enucleations (Table 1). The tumors in VHL and non-VHL patients were matched for mRNA analysis, according to tumor grade [11 of 18 (61%) tumors of grade G2 in patients with VHL vs. 11 of 16 (68%) tumors of grade G2 in non-VHL patients], Ki-67 index (mean Ki-67, 4.77% in VHL tumors vs. 4.81% in non-VHL tumors), tumors size (mean size, 3.09 cm in VHL tumors vs. 3.11 cm in non-VHL tumors), presence of metastases [7 of 18 (38%) metastatic tumors in patients with VHL vs. 6 of 16 (37%) metastatic tumors in non-VHL patients).

For all patients the diagnosis was established on routine formalin-fixed, paraffin-embedded material [hematoxylin and eosin (H&E) sections]. The following histopathologic data were recorded (Table 1): number and size of the tumors, percentage of cell nuclei stained for Ki-67 (Dako, clone MIB1), grading according to tumor-node-metastasis (TNM) and World health Organization (WHO) 2010 classifications (9, 10). The number of microadenomas (100–5,000 μm) and of macroscopic tumors (>5 mm) according to WHO 2010 criteria were also noted (10). For small microadenomas, the expression of both CA9 and chromogranin A was required to distinguish them from large nonneoplastic islets, as described previously (2). Survival data were available for all patients. The median follow-up was 62 months (range, 2–170 months).

Frozen material for RT-PCR analysis was available for 18 of 31 VHL tumors and 16 of 16 non-VHL tumors, stored at −80°C (Table 1). In all cases one frozen section stained with H&E was obtained before mRNA extraction and histologic examination confirmed that selected samples contained only tumor tissue (both tumor cells and stroma) of good quality and no normal pancreatic tissue. No microdissection was carried out before mRNA extraction. We compared the expression of 52 target genes (Table 2) in VHL and non-VHL tumors. Total RNA was reverse transcribed before real-time PCR amplification (11). All of the PCR reactions were carried out using the Light Cycler 480 Real-Time PCR System and the LightCycler 480 SYBR Green I Master mix (Roche Diagnostics). Quantitative values were obtained from the crossing point (Cp), number at which the increase in the signal associated with exponential growth of PCR products begins to be detected with Light Cycler 480 System Software (Roche Diagnostics), according to the manufacturer's manuals. The precise amount of total RNA added to each reaction mix (based on optical density) and its quality (i.e., lack of extensive degradation) are both difficult to assess. We therefore also quantified transcripts of 2 endogenous RNA control genes involved in 2 cellular metabolic pathways, namely RPLP0 (also known as 36B4; NM_001002), which encodes human acidic ribosomal phosphoprotein P0, and TBP (Genbank accession NM_003194), which encodes the TATA box-binding protein. Each sample was normalized on the basis of its RPLP0 (or TBP) content. We selected RPLP0 because the prevalence of its transcripts is high and because this gene is widely used as an endogenous control for northern blot analysis (more known as 36B4; ref. 11). We also selected TBP as an endogenous control because the prevalence of its transcripts is moderate as compared with RPLP0, and because there are no known TBP retropseudogenes (retropseudogenes lead to coamplification of contaminating genomic DNA and thus interfere with RT-PCR, despite the use of primers in separate exons). Results, expressed as the N-fold differences in target gene expression relative to the RPLP0 gene, and termed N_target, was determined as N_target = 2^ΔCpsample, where the ΔCp value of the sample was determined by substracting the average Cp value of the target gene from the average Cp value of the RPLP0 gene. The N_target values of the samples were subsequently normalized to the smallest amount of target gene mRNA, detectable and quantifiable by real-time quantitative PCR assays based on SYBR Green fluorescence methodology (target gene C_t value = 35; N_target value = 1).

Primers for endogenous RNA controls and the 52 target genes were chosen with the assistance of the Oligo 5.0 computer program (National Biosciences). To avoid amplification of contaminating genomic DNA, one of the 2 primers was placed at the junction between 2 exons, if possible. In general, amplicons were between 70 and 120 nucleotides long. For each primer pair, no-template control (NTC) and no-reverse transcriptase control (RT negative) assays were conducted and produced negligible signals (usually >40 in Cp value), suggesting that primer dimer formation and genomic DNA contamination effects were negligible. RNA extraction, cDNA synthesis, and PCR reactions have been described previously (12).

Tissue microarray (TMA) blocks were constructed from representative blocks from the 31 patients with VHL and the 16 non-VHL patients. For each case at least one macrotumor, including the greatest one, was taken for the TMA (1–4 macrotumors per patient). The construction of these TMA was conducted using a tissue arrayer (Beecher Instruments) and each macrotumor specimen was represented by four 1-mm cores in the TMA block. In case of multiple tumors, the largest one was taken into account for the statistical analysis. If neuroendocrine microadenomatosis was associated, lesions >1 mm were also taken in the TMA (1–4 microadenomas per patient). In total, 3 blocks of TMA were constructed (including 50 macrotumors and 22 microadenomas). In addition to the TMA, we analyzed the normal islets in the pancreas located at a distance from the tumor in 10 of the sporadic patients.

The blocks were cut in 3-μm sections and were immunolabeled with antibodies directed against pVHL (FL181; Santa Cruz Biotechnology), CA9 (polyclonal; Novus Biological), CD34 (QBEND10; Immunotech), cyclin D1 (SP4; Thermo Scientific), vimentin (V9; Dako), E-cadherin (polyclonal; Santa Cruz Biotechnology), occludin (polyclonal; Zymed), and VE-cadherin (F-8; Santa Cruz Biotechnology).

Immunohistochemistry was carried out using an automated immunohistochemical stainer according to the manufacturer's guidelines (Streptavidin-peroxidase with an automate Ventana Benchmark, Ventana). Immunostaining of paraffin sections was conducted after dewaxing and rehydrating slides. Antigen retrieval was conducted by pretreatment with high temperature. Substitution of the primary antibody with PBS was used as a negative control.

pVHL, CA9, cyclin D1, vimentin, E-cadherin, and occludin were evaluated by calculating a score obtained by multiplying the intensity (negative, 0; weak, 1; moderate, 2; and strong, 3) by the percentage of stained cells. The pattern of expression (cytoplasmic, membranous, and nuclear) was noted. All cores were evaluated separately and a mean score was calculated for each tumor. Tumors with a score greater than the median value were considered at high protein expression. The heterogeneity of the distribution of the scores was analyzed in scatter plots.

CD34 and VE-cadherin staining allowed evaluation of the microvascular density (MVD). The cores were evaluated separately at high magnification (× 25 objective; area of a core: 0.78 mm²). In each tumor, the core with the highest count was documented. Tumors with an MVD greater than the median value were considered at high MVD. The heterogeneity of the distribution of the MVD evaluated by CD34 and VE-cadherin was analyzed in scatter plots.

Comparison of immunohistochemical characteristics between the groups were conducted using the Kruskal–Wallis test for continuous data and the χ² test or the Fisher exact test for categorical data. Levels of gene mRNA expression between patients with or without a VHL mutation were compared using the Mann–Whitney U test and a ROC (receiver operating characteristics) analysis. The association between histoprognostic factors and levels of gene mRNA expression was searched in the patients with VHL by using the Mann–Whitney U and Spearman tests. Data were analyzed with the SAS 9.1 statistical software for Windows (SAS Institute Inc.). All statistical tests were 2-sided. The critical level of statistical significance was set at P < 0.05.

Nineteen (36%) of the 52 genes were significantly upregulated in the tumors in the patients with VHL (Table 3). The upregulated genes were directly related to HIF signaling (CA9/carbonic anhydrase IX; HIF2A/hypoxia-inducible factor 2 α subunit; GLUT1/glucose transporter 1), angiogenesis (CDH5/VE-cadherin; VEGFR1/vascular endothelial growth factor; EDNRA/endothelin receptor type A; ANGPT2/angiopoietin 2; CD34; VEGFR2/kinase insert domain receptor; VEGFA/vascular endothelial growth factor A; ANGPT1/angiopoietin 1), the processes of EMT (VIM/vimentin) and metastasis (LAMA4/lamininα4; CXCR4/chemokine c-x-c receptor 4), growth factors and receptors (PDGFB/platelet-derived growth factor β polypeptide; IRS1/insulin receptor substrate 1; ERBB1/epidermal growth factor receptor), and cell cycle (CCND1/cyclin D1; CDKN2A/cyclin-dependent kinase inhibitor 2A). We obtained the same 19 genes when we compared the VHL tumors with the group of non-VHL/non-NEM1 (n = 14) tumors.

Three (6%) of the 52 genes were significantly downregulated in the tumors of the patients with VHL (Table 4). These genes were involved in EMT process (OCLN/occludin) and signaling pathway (RPS6KB1/ribosomal protein S6 kinase-polypeptide 1; GADD45B/growth arrest and DNA damage inducible β). Two of these 3 genes were confirmed when we compared the VHL tumors with the group of non-VHL/non-NEM1 (n = 14) tumors (only GADD45B did not reach significance any more with this analysis, P = 0.13).

Among the 19 upregulated genes, we found that high LAMA4, EDNRA, and PDGFA expression correlated with high pT (P = 0.04, P = 0.02, and P = 0.05, respectively) and that high HIF1A expression correlated with greater tumor size (P = 0.01).

The scores (range, mean) in VHL tumors, non-VHL tumors, and microadenomas are given in Table 5. The homogeneity of the distribution of the results is presented in scatter plots for each molecule analyzed (Fig. 1).

• pVHL: Low cytoplasmic expression was correlated with VHL disease (P < 0.0001); the 19 upregulated and 3 downregulated genes in VHL tumors correlated with lower and higher expression of pVHL, respectively (P < 0.05 for all the 22 genes). There was no statistical difference between the expression in tumors and microadenomas in the patients with VHL. In the normal non-VHL islets, the cytoplasmic score ranged from 200 to 300.

• CA9: High membranous expression correlated with VHL disease (P ≤ 0.0001; Fig. 2A and B). There was no statistical difference between the expression in tumors and microadenomas in patients with VHL. Islets in non-VHL pancreas did not express CA9.

• CD34: High MVD correlated with VHL disease (P = 0.0008; Fig. 2C and D). The MVD was comparable when assessed by CD34 or VE-cadherin staining (P > 0.05).

• Cyclin D1: The staining was cytoplasmic or nuclear. High cytoplasmic and nuclear cyclin D1 expressions were both correlated with VHL disease (P = 0.01 and P = 0.001, respectively; Fig. 2E and F). There was no statistical difference between the expression in tumors and microadenomas in patients with VHL. Islets in non-VHL pancreas did not express Cyclin D1.

• Vimentin: High cytoplasmic expression was correlated with VHL disease (P ≤ 0.0001; Fig. 3A and B). There was no statistical difference between the expression in tumors and microadenomas in the patients with VHL. Islets in non-VHL pancreas did not expressed Vimentin.

• E-cadherin: Lower membranous expression was correlated with VHL disease (P = 0.0007; Fig. 3C and D). There was no statistical difference between the expression in tumors and microadenomas in the patients with VHL. In the normal non-VHL islets, the score ranged from 150 to 250.

• Occludin: The staining was mainly cytoplasmic with variable membranous enhancement. High expression was correlated with absence of VHL disease (P ≤ 0.0001; Fig. 3E and F). There was no statistical difference between the expression in tumors and microadenomas in the patients with VHL. In the normal non-VHL islets the cytoplasmic, focally membranous staining score ranged from 100 to 150.

• VE-cadherin: High MVD correlated with VHL disease (P = 0.04). The MVD was comparable when assessed by CD34 or VE-cadherin staining (P > 0.05).

In this article, we have identified a set of genes differentially expressed in PanNET comparing VHL and sporadic patients, mainly involved in angiogenesis, but also in several pathways related to cell cycle and signaling growth factors, extracellular matrix remodeling, and EMT. The up- or downregulation of these genes was significantly correlated with the protein (pVHL) expression status in these tumors. In VHL disease, mutation in the gene leads to considerable decrease of pVHL expression, but we also found that a small subset of sporadic PanNET expressed low levels of pVHL, which could represent somatic VHL alterations, as previously reported (13).

It is well known that in VHL disease, mutation in the VHL gene results in increased expression of HIF, which regulates many angiogenic pathways (14). This is in accordance with the fact that all the tumors developed in patients with VHL, including PanNET, are highly vascularized (2, 15). CD34 and VE-cadherin (cadherin-5) expression is related to tumor angiogenesis (16). We found strong mRNA and protein expression of these molecules in VHL tumors and microadenomas, as we analyzed both tumor cells and their stroma in our samples. VEGF is an important target of HIF and a well-known potent mediator of angiogenesis, overexpressed in VHL associated tumors (17–19). Our results showing a strong mRNA expression of VEGF-A and of its receptors VEGFR1 and VEGFR2 in VHL PanNET are in accordance with those reported in other tumor types associated to VHL disease, suggesting VEGF as a prime candidate for a therapeutic target (20–22). Interestingly, the use of sunitinib, a potent tyrosine kinase inhibitor of VEGF pathway, may have value in the treatment of VHL-related tumors including pheochromocytoma (23). In the presence of VEGF, both angiopoietin 1 and 2 are capable of augmenting angiogenesis (24–27). This parallels well with our results, showing high amounts of both angiopoietin 1 and 2 mRNA in VHL PanNET, as previously observed in renal clear cell carcinomas with alteration of the VHL protein (28). Our results support targeting this system as an effective antiangiogenic therapy in such tumors. Endothelin 1 is a hypoxia-inducible HIF-1 target gene capable of stimulating proliferation and migration, inducing VEGF production and increasing the expression of proto-oncogenes (29–31). We found higher expression of endothelin receptor type A in VHL tumors, which suggests that the regulation of the endothelin pathway is probably highly specific in PanNET in the clinical setting of VHL disease.

EMT permits epithelial cells to acquire mesenchymal like properties via disruption of intercellular adhesion. This process is linked to tumor invasiveness, aggressiveness, and drug resistance (5). Among proteins involved in EMT we found deregulation of vimentin, which is the main intermediate filament of mesenchymal cells. It is upregulated in EMT and is associated with the acquisition of invasion (32). We confirmed strong upregulation of vimentin in PanNET in patients with VHL at both the protein and RNA levels. Vimentin has been shown to be upregulated in VHL-deficient cells as well as in renal clear cell carcinomas and hemangioblastomas in VHL disease (7, 33–35). Transmembrane proteins also play a role in EMT, such as E-cadherin, crucial for the maintenance of structural integrity of epithelia. Its downregulation is a key molecular change in EMT, a process known as “cadherin switch” (5, 36). Lower expression of E-cadherin in VHL tumors at the protein but not at the mRNA level is in favor of a posttranscriptional mechanism of regulation (5). In addition to downregulation of adherens junctions in EMT, tight junctions are involved in this process in in vivo VHL defective sporadic renal clear cell carcinoma, by both HIF-dependent and -independent mechanisms (6). VHL loss of function has a negative effect on the mRNA and protein expression of the tight junction component occludin in PanNET, and also in microadenomas, such as for E-cadherin. Junctions disruption could therefore contribute to tumorigenesis, from the earliest premalignant lesions. This is in accordance with the fact that downregulation of these cell adhesion molecules is involved in malignant transformation in other tumor types, but we found no correlation with clinicopathologic data in our series (37).

Metastasis-related genes are involved in tumor progression. By expressing CXCR4 chemokine receptor, human tumors acquire properties to invade tissue barriers, migrate to secondary organs, and form metastases (38) Loss of pVHL function causes increased expression of CXCR4 (39–41). In accordance with these data, we found strong mRNA expression of CXCR4 in VHL tumors and similar results have been reported in other types of tumors developed in patients with VHL (40, 41). It has recently been shown that CXCR4 triggers expression of metastasis-associated matrix metalloproteinase (MMP) 2/MMP 9, but no difference was found in MMP 2 or MMP 9 mRNA expression between VHL and sporadic PanNET (39). Overexpression of laminin α4 chain has been suggested to correlate with malignancy, migration, and metastatic capacity in different tumor types (42–44). The higher expression of this molecule in our VHL tumors is in line with the earlier discussion on EMT and with the higher risk of malignant course of PanNET in patients with VHL disease (4, 45).

Target genes involved in cell trafficking or acting as growth factors or receptors previously implicated in sporadic NET carcinogenesis and that may represent useful chemotherapeutic targets were also explored. Recent articles have reported high levels of PDGFRB and EGFR in aggressive PanNET (8, 46, 47). Our study reveals that both PDGFRB and the EGFR (ERBB1) are expressed at higher levels in VHL tumors, which supports further investigation of specific targeted anti-PDGFR and EGFR treatment in patients with VHL presenting with a PanNET. Insulin receptor substrate 1 (IRS1), an important mediator in insulin-like growth factor1/insulin signaling, is an adaptor molecule in signal transduction that play a role in development of PanNET (48–51). Interestingly, IRS1 was expressed at higher levels in VHL tumors. GADD45β, which regulates apoptosis, has previously been implicated in the pathogenesis of malignant PanNET (8). Our results point out that it is overexpressed in sporadic tumors and, in contrast, probably not involved in VHL tumorigenesis.

In conclusion, PanNET that develop in the clinical setting of VHL disease express a specific set of genes detected from the very early stages of microadenomas, which lead to consider them as a distinctive subtype of tumors. These specific genes are involved in several metabolic pathways, mainly related to angiogenesis, EMT, and extracellular matrix remodeling, but also to cell signaling. The number of targets regulated directly or indirectly by VHL is likely to be large in PanNET and many but not all are regulated by the HIF system which makes a dominant contribution to alterations in transcription in VHL PanNET.

No potential conflicts of interest were disclosed.

The authors thank the following pathologists, whose names do not appear among the authors, who provided cases and pathological data for this study (*pathologists members of the TENpath network): G. Belleannée* (Bordeaux), S. Casnedi* (Strasbourg), A. Chevallier (Nice), A. Gomez* (Toulouse), P. Guinebretière (Saint-Denis), M.-F. Heymann* (Nantes), C. Laboisse* (Nantes), G. Monges* (Marseille). They are grateful to Sylvie Mosnier, Sandrine Couroble and Véronique Albano for their excellent technical assistance.

Anne Couvelard received a grant from the «GTE: Groupe d'Etude des Tumeurs Endocrines».

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.