Abstract
Clinical options for systemic therapy of neuroendocrine tumors (NET) are limited. Development of new drugs requires suitable representative in vitro and in vivo model systems. So far, the unavailability of a human model with a well-differentiated phenotype and typical growth characteristics has impaired preclinical research in NET. Herein, we establish and characterize a lymph node–derived cell line (NT-3) from a male patient with well-differentiated pancreatic NET. Neuroendocrine differentiation and tumor biology was compared with existing NET cell lines BON and QGP-1. In vivo growth was assessed in a xenograft mouse model. The neuroendocrine identity of NT-3 was verified by expression of multiple NET-specific markers, which were highly expressed in NT-3 compared with BON and QGP-1. In addition, NT-3 expressed and secreted insulin. Until now, this well-differentiated phenotype is stable since 58 passages. The proliferative labeling index, measured by Ki-67, of 14.6% ± 1.0% in NT-3 is akin to the original tumor (15%–20%), and was lower than in BON (80.6% ± 3.3%) and QGP-1 (82.6% ± 1.0%). NT-3 highly expressed somatostatin receptors (SSTRs: 1, 2, 3, and 5). Upon subcutaneous transplantation of NT-3 cells, recipient mice developed tumors with an efficient tumor take rate (94%) and growth rate (139% ± 13%) by 4 weeks. Importantly, morphology and neuroendocrine marker expression of xenograft tumors resembled the original human tumor.
Implications: High expression of somatostatin receptors and a well-differentiated phenotype as well as a slow growth rate qualify the new cell line as a relevant model to study neuroendocrine tumor biology and to develop new tumor treatments. Mol Cancer Res; 16(3); 496–507. ©2018 AACR.
This article is featured in Highlights of This Issue, p. 359
Introduction
Well-differentiated neuroendocrine tumors (NET) are characterized by the cardinal features of slow tumor growth and potential to secrete functionally active hormones. Because of their mostly indolent behavior, more than 50% of patients are diagnosed with metastatic disease not amenable to curative surgery (1). Despite recent advances in the development of medical treatments, long-term disease control is only achieved in a minority of patients receiving chemotherapy or targeted therapies in patients with pancreatic NET (2–4). Encouraging data on the use of peptide receptor radionuclide therapy show long-term disease stabilization but are limited to NET with high somatostatin receptor (SSTR) expression (5). In contrast, for patients with low receptor status identification of alternative NET cell–specific receptor targets is crucial. However, development and evaluation of novel therapeutic targets and optimization of existing NET therapies has been hampered by limited availability of preclinical models. Currently, there are only a few NET cell lines available (6, 7). Of these, the pancreatic cell lines BON and QGP-1 are the most widely used. Despite expression of some NET markers, BON and QGP-1 do not display a well-differentiated neuroendocrine phenotype. For example, cells do not serve as a model to study radionuclide imaging and therapy, as SSTR expression is very low and octreotide radionuclide uptake is only minor (8). Furthermore, doubling times of less than 48 hours for BON cells, are too fast to represent a relevant model for slow growing NETs (9). In xenotransplantation models, these cell lines show very rapid tumor growth with early development of tumor necrosis due to poor vascularization (10, 11).
Although the establishment of pancreatic neuroendocrine tumor cell lines with a well-differentiated phenotype and low proliferation rate would be mandatory to validate new therapeutic targets in preclinical settings, extensive research over the past 25 years has not resulted in generation of such cell lines. Accountable for this failure were undefined optimal growth conditions for NET cells as well as overgrowth by contaminating tumor fibroblasts (12). We here report successful establishment of a new pancreatic NET cell line from a lymph node metastasis of a patient with functional insulinoma. These cells, named NT-3, have now been cultured for more than 3 years and they still display a stable well-differentiated phenotype, including after xenotransplantation. Cell characterization and proliferation analyses hold promise that these cells provide a hitherto unavailable clinically relevant in vitro and in vivo model of well-differentiated neuroendocrine tumor disease.
Materials and Methods
Cell lines
Authenticated BON and QGP-1 cells were cultured in DMEM/Ham F12 and RPMI, respectively, supplemented with 10% FCS and penicillin/streptomycin. Cells regularly tested negative for mycoplasma contamination. The cells were used within 15 passages after authentication (DMSZ).
The new cell line NT-3 was generated from a surgically resected lymph node of a 33-year-old male patient with well-differentiated NET of the pancreas. The patient had stage IV disease and a debulking operation had been performed for a functionally active insulinoma. Histopathology confirmed insulin expression and the Ki-67 was determined to be 15%–20%. The local ethical review board approved the use of human tissue and informed consent from the patient was obtained before surgery. The study was conducted in accordance with the Declaration of Helsinki. Tissue culture was started after mincing of the tissue and digestion with collagenase IV (Serva). The ensuing primary mixed cell culture containing tumor and stromal cells was cultivated in RPMI medium supplemented with 10% FCS, penicillin/streptomycin, HEPES, EGF (20 ng/mL), and FGF2 (10 ng/mL; PeproTech). To eliminate fibroblasts, the primary cultures were passaged by sequential trypsinization and cultivation under low-adherent conditions. While fibroblasts attached in all culture conditions, tumor cells preferentially attached to collagen-coated culture plates. Passage numbering of tumor cell culture was started after achieving a pure tumor cell culture and successful cryopreservation. Cells from passage 15–30 have been used for the experiments.
For cell culture experiments under hypoxic conditions, cells were incubated in a nitrogen-supplemented incubator with 0.5% oxygen for up to 24 hours.
Quantitative real-time PCR
Total RNA was isolated from cell cultures and tumor samples using the Nucleo Spin RNA, DNA, and protein purification Kit (Macherey-Nagel) according to the manufacturer's protocol. Total RNA was reverse-transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and a T3 Thermocycler (Biometra). Prevalidated primers for real-time PCR were purchased from Applied Biosystems (full list in Supplementary Methods). Samples in duplicate were subjected to PCR in a Step One Plus Real-Time PCR System (Applied Biosystems). Values were expressed as (Ct) values normalized to housekeeping gene GAPDH using the 2−ΔΔCt method.
Immunocytochemistry
NT-3, QGP-1, and BON cells were grown on collagen-coated chamber slides for 48 hours supplemented with or without growth factors (EGF and FGF2) and fixed with 4% paraformaldehyde. Mouse anti-Ki67 (Clone MIB-1, Dako) and mouse anti-Chromogranin-A (Clone DAK-A3, Dako,) were used as primary antibodies. Goat anti-mouse Alexa Fluor 555 (Molecular Probes by Life Technologies) was used as secondary antibody. Coverslips were mounted using Vectashield Hard Set with DAPI (Vector Laboratories). Cells were analyzed by fluorescence microscopy using a BioRevo BZ-9000 (Keyence).
Western blot analysis
Western blot analysis was performed as described previously (13). In brief, 20 μg of protein was separated by SDS-PAGE and transferred onto nitrocellulose membranes (Schleicher & Schuell). After blocking, membranes were probed with primary and horseradish peroxidase (HRP)-conjugated secondary antibodies (full antibody list in Supplementary Methods) and ensuing bands were visualized using an ECL-kit (Pierce).
ELISA
Measurement of VEGF in cell culture supernatants was performed using the R&D Systems VEGF ELISA Development Kit following manufacturer's instructions. VEGF concentrations in cell culture supernatants were normalized to total protein content of the corresponding cell culture to account for different cell numbers of the cell lines.
MTT assay
The MTT assay was carried out under standardized conditions in 96-well plates. After incubation for 24 hours, cells were treated separately with the following therapeutics: octreotide (Sandostatin, Novartis), streptozotocin (Sigma), 5-fluorouracil (5-FU; Sigma), everolimus (LC Laboratories), and sunitinib (LC Laboratories). All therapeutics were replenished once after 48 hours. 120 hours after the first addition of therapeutics, 20 μL of MTT (5 mg/mL, Fluka Analytical, Sigma) was added to each well and plates were incubated for 2 hours. Using a Tecan infinite F50 spectrophotometer (Tecan), absorbance was measured at 540 nm after solubilization of the formazan dye with DMSO. Results are displayed as mean absorbance relative to controls (untreated or DMSO control as indicated).
Cell counting
Fast growing BON and QGP-1 cells were seeded in 24-well-plates with 1 × 104 cells per well, and slower growing NT-3 with 2 × 104 cells per well, diluted in 500-μL medium supplemented with or without growth factors (EGF and FGF2). BON and QGP-1 cells were counted on days 1, 3, 7, and 10, whereas NT-3 cells were counted on days 1, 3, 7, 14, 21, 28, 35, and 42.
Soft agar assay
Soft agar assay was performed as described previously (14). In brief, 1 × 105 NT-3 cells were plated in 60-mm Petri dishes between layers of agar mixed with complete growth media. Plates were cultured up to 12 weeks with regular growth media addition. Ensuing colonies were counted by light microscopy.
Insulin measurement
Human insulin content was determined using the ADVIA Centaur Insulin Assay (REF 02230141) and ADVIA Centaur XP analyzer (Siemens Healthcare). This assay is standardized to the first human insulin international reference preparation by the World Health Organization (NIBSC code 66/304). FCS did not contain measurable insulin using this human insulin assay. In vitro samples were obtained from supernatants of NT-3 cell cultures after stimulation with 500 μmol/L 3-Isobutyl-1-methylxanthin (IBMX, Sigma) for 60 minutes. No additional growth factors were added and untreated cells were used as control. Ex vivo data were generated from serum samples of xenograft tumor mice. The manufacturer has not reported cross-reactivity to mouse insulin and we did not detect insulin in serum from untransplanted mice.
Mutational analysis and genotypic fingerprinting
Mutation analysis of hotspot regions in 22 genes that are frequently mutated in cancer as well as genotypic fingerprinting was performed by multiplex-PCR based targeted high-throughput sequencing (HTS) using the Ion AmpliSeq Colon and Lung Cancer Research Panel v2 and the Ion AmpliSeq Sample ID Panel and subsequent sequencing with the Ion Torrent system (Thermo Fisher Scientific) according to the manufacturer's recommendations (full list of mutational sides investigated in Supplementary Methods). Targeted exome mutational analysis of MEN1, VHL, TSC1, TSC2, DAXX, ATRX was performed after enriching the coding and flanking intronic regions using the Agilent in solution technology according to manufacturer's protocol (Agilent) with some slight modifications and sequencing using the Illumina HiSeq 4000 system (detailed protocol for data analysis, see Supplementary Methods).
Xenograft animal model and MRI
To generate subcutaneous tumors, 2 × 106 NT-3 cells in 0.1-mL RPMI containing 1% FCS were mixed with 0.1-mL extracellular matrix gel (Matrigel) and injected subcutaneously into flanks of NOD/SCID mice (The Jackson Laboratory). Intraportal cell transplantation was performed as described previously (15) using 2 × 106 NT-3 cells in 0.2-mL RPMI with 2% FCS. Animals experiments were approved by the local authorities (UKE Forschungstierhaltung and Behoerde für Gesundheit und Verbraucherschutz, Hamburg, approval number 125/15). MRI measurements (protocol see Supplementary Methods) of xenografted mice were performed on 4-week intervals for volumetric analysis. At the end of experiments, mice were sacrificed under animal welfare guidelines, blood was collected and necropsy was performed. Excised tumors were divided in half and either fixed in formalin for paraffin embedding or snap-frozen and stored at −80°C for molecular analyses.
IHC
Tissue sections from formalin-fixed paraffin-embedded xenograft tumors were dewaxed, blocked with blocking solution (Dako), and incubated with the following primary antibodies: Chromogranin A, synaptophysin, and Ki-67 (all from Dako). Positive staining was visualized using a corresponding secondary HRP-coupled antibody and the DAB staining system (Dako). Slides were counterstained with hematoxylin (Sigma). Staining of primary human tumor tissue was done as described before using the following primary antibodies: Chromogranin A, Synaptophysin, VMAT1, SSTR2A, SSTR5, Ki-67, and insulin (16).
Statistical analysis
All graphs and statistical analysis have been performed using GraphPad Prism 4 utilizing standard two-tailed unpaired Student t test. P values <0.05 were considered significant (*, P < 0.05; **, P < 0.005; ***, P <0.001). Data are presented as mean ± SEM of at least three replicates.
Results
Generation and morphology of the cell line
After mechanical dissection and collagenase digestion of the original surgical specimen, cell cultures contained tumor cells in small grape-like clusters as well as various components of stromal cells. Within 8 weeks of sequential trypsinization and serial passaging, contaminating fibroblasts were eliminated and a pure tumor cell culture emerged. These cells were now called NT-3. NT-3 formed islet-like spheres upon culture in nonadherent conditions and small grape-like clusters in adherent conditions on collagen-coated plates after single cell plating (Supplementary Fig. S1A–S1D). Successful cryopreservation and verification of the neuroendocrine phenotype (e.g., chromogranin A and synaptophysin expression) was achieved 5 months after initiating the primary cultures, when enough cells were available for analysis. At this time point, passage numbering was started. Genotypic fingerprinting of 8 unlinked, nonexonic single-nucleotide polymorphisms (SNP) with consistent high minor-allele frequencies in both, DNA derived from the original tumor and from NT-3 cells, confirmed the origin of our cell line.
Currently, NT-3 have been in continuous culture since 42 months and 58 passages. The cells have retained their morphology and well-differentiated neuroendocrine phenotype throughout (Supplementary Fig. S2A). Screening for oncogenic mutations in hotspots of 22 genes by targeted high-throughput parallel sequencing ruled out mutations associated with high-grade neuroendocrine carcinoma (e.g., ras, Smad4, EGFR, and PIK3CA) and confirmed the absence of TP53 mutations (Supplementary Fig. S2B). Targeted sequencing of genetic mutations associated with pNET (i.e., VHL, MEN1, TSC1, TSC2, DAXX, and ATRX) revealed a homozygous missense mutation of MEN1 (chromosome 11, position 64572018; c.1621A>G; p.T541A), while other mutations in these genes were not found.
Neuroendocrine differentiation
One of the prominent features of NET is their high degree of differentiation and functional activity. The available pancreatic NET cell lines BON and QGP-1 have not been vigorously characterized for their neuroendocrine differentiation status. Therefore, we analyzed expression of neuroendocrine markers and neuroendocrine-related transcription factors in these cells and compared them with our novel NT-3 cell line.
We observed high mRNA expression of chromogranin A (CgA) in NT-3, and, to a lesser extent, in BON but only low expression in QGP-1 (Fig. 1A). Correspondingly, NT-3, and to a lesser extent, also BON stained positive for CgA on immunofluorescence (Fig. 1B). The neuronal marker synaptophysin was expressed in all three cell lines confirming their neuroendocrine origin and again NT-3 showed a much higher expression than BON and QGP-1 (Fig. 1A). Another prominent feature of neuroendocrine cells is their uptake and processing of monoamines. We detected expression of vesicular monoamine transporter (VMAT) 1 and VMAT2 in NT-3, while QGP-1 expressed neither of them and BON only weakly expressed VMAT1 (Fig. 1A). In addition, all cell lines expressed enzymes required for serotonin biosynthesis (e.g., tryptophan hydroxylase 1 and dopa-decarboxylase) to various extent (Supplementary Fig. S2C). Reassuringly, the original NET tumor of the patient was also positive for CgA, synaptophysin, and VMAT1 (Fig. 1C).
Neuroendocrine phenotype of NT-3. A, Expression analyses of CgA, synaptophysin, and VMAT1 in BON, QGP-1, and NT-3 cells showed a high degree of neuroendocrine differentiation in NT-3 cells. B, Strong expression of CgA in NT-3 was confirmed by immunocytochemistry. C, Staining of the primary surgical tumor specimen for CgA, synaptophysin, and VMAT1 corresponded to the expression profile of NT-3 cells. D, Increased expression of neuroendocrine-related transcription factors PDX-1, neurogenin3, NeuroD, Isl-1, Pax4 and Pax6 supported the well-differentiated phenotype of NT-3 compared with BON and QGP-1. E, Expression and IBMX-stimulated release of insulin in NT-3 confirmed functionality. Correspondingly, IHC of the primary human tumor showed insulin expression. Scale bar, 50 μm.
Neuroendocrine cell differentiation in the pancreas during embryonal development is promoted by multiple transcription factors, most of which are still expressed and functionally relevant in adult islet cells (17). We observed high expression of PDX-1, Isl-1, Neurogenin3, NeuroD, Pax4, and Pax6 in NT-3 cells, whereas BON only express NeuroD and QGP-1 only expresses Isl-1 (Fig. 1D). As a marker of functionality, NT-3 expresses and releases insulin, corresponding to the functional insulinoma of the patient (Fig. 1E). Insulin expression in the original patients' tumor was confirmed by IHC showing moderate staining intensity with some strongly stained tumor cells. Interestingly, insulin expression by NT-3 was modifiable by the omission or addition of the growth factors EGF/FGF-2.
The finding of growth factor–dependent insulin expression prompted us to investigate this effect also for the abovementioned neuroendocrine markers, and we detected profoundly reduced expression upon growth factor supplementation for CgA, synaptophysin, and all neuroendocrine-related transcription factors (Fig. 2A and data not shown). Correspondingly, NT-3 cells showed a reduction in E-cadherin and an increase in vimentin expression when treated with growth factors, suggesting epithelial-to-mesenchymal shifting of their phenotype (Fig. 2B). This was, in part, reflected by the appearance of a small spindle-shaped subpopulation in NT-3 cultures treated with growth factors (Fig. 2C), which completely disappeared after subsequent withdrawal of growth factors. In contrast to this high plasticity of NT-3, QGP-1 showed only minimal changes upon growth factor treatment, whereas BON cell phenotype was completely unaffected by addition or omission of growth factors (data not shown).
Plasticity of NT-3 cells. A, Supplementation of the cell culture media with growth factors EGF and FGF2 decreased expression of neuroendocrine markers in NT-3 cells. B, This was accompanied by downregulation of E-Cadherin and upregulation of vimentin. C, These changes are reflected by appearance of a spindle-shaped subpopulation (arrows) in growth factor–supplemented cell cultures. **, P < 0.005; ***, P < 0.001. Scale bar, 100 μm.
Cell proliferation
Well-differentiated NETs are characterized by a low proliferation rate and slow tumor growth. Cultured BON and QGP-1 cells had a doubling time of less than two days (BON 1.5 ± 0.4 days, QGP-1 1.6 ± 0.3 days), which was unaffected by addition of growth factors (Fig. 3A; Supplementary Fig. S3A). In contrast, NT-3 grew slowly with a doubling time of 10.9 ± 0.7 days (Fig. 3A). Their growth was dependent on the addition of growth factors, as we did not observe growth of NT-3 cells plated at low cell density without growth factors (Fig. 3B).
Proliferation. A, Growth curves of tumor cells cultured with growth factor supplementation show a fast growth rate of BON and QGP-1, while NT-3 exhibit a very slow growth over a period of 7 weeks. B, NT-3 plated at low density without growth factor addition did not proliferate. C and D, Staining for Ki-67 was used to calculate proliferative indices in BON, QGP-1, and NT-3 cells with or without growth factor supplementation. Scale bar, 50 μm.
Clinical tumor grading of NET is assessed by IHC staining for the nuclear proliferation factor Ki-67. BON and QGP-1 have a Ki-67 labeling index of 80.0% ± 3.3% and 82.6% ± 1.0% respectively, corresponding to highly proliferative G3 carcinomas. In contrast, nuclear Ki-67 in NT-3 is only 2.0% ± 0.2%, corresponding to a slowly proliferative G1–G2 tumor. Addition of growth factors slightly increased Ki-67 index in BON by 7%, compared with a profound increase in NT-3 by 325% to an index of 14.6% ± 1.1% (Fig. 3C and D), now similar to the patient's original tumor and xenograft tumors (see below). Cellular anchorage-independent growth, which represents a sign of malignancy, was confirmed in NT-3 by soft agar assay. Out of 1 × 105 cells plated in soft agar, 2,200 formed cell clusters of 10–20 cells with a latency of 12 weeks, corresponding to a colony-forming frequency of approximately 2% (Supplementary Fig. S3B).
Therapeutic targets
SSTRs are crucial therapeutic targets for somatostatin analogues and peptide radio receptor therapy in NET treatment. NT-3 shows high mRNA expression of SSTR1, SSTR2, SSTR3, and SSTR5, among those, the expression of SSTR-1 is particularly strong. In contrast, BON and QGP-1 express these receptors either only weakly or not at all (Fig. 4A). None of these cell lines expresses SSTR4. Western blotting confirmed high expression of SSTRs and showed that the most clinically relevant SSTR2 is exclusively expressed in NT-3, while SSTR3 and SSTR5 are also expressed in BON (Fig. 4B). Correspondingly, the patient's primary tumor tissue showed positive staining for SSTR2A and SSTR5 on IHC (Supplementary Fig. S4A and S4B).
Therapeutic targets in NT-3. A and B, Western blot and qPCR analysis of SSTR subtypes revealed higher expression of SSTR1 and SSTR2 in NT-3 compared with BON and QGP-1. C, Expression of VEGF A, VEGF B and angiopoetin mRNA is at least 3-fold higher in NT-3 cells compared to BON and QGP-1. The expression of VEGF A is further increased after 6 and 24 hours in low oxygen environment (D) and VEGF release into the supernatant after 24h of hypoxia was confirmed by ELISA (E). Ensuing hypoxic conditions in tumor cells were verified by increase in carboanhydrase IX (CAIX) expression after 6 and 24 hours by qPCR (F) and after 24-hour by Western blot analysis (G).
Another therapeutic target in clinical NET treatment are angiogenic factors. In standard culture conditions, the mRNA expression of VEGF A and B in NT-3 is more than 3-fold higher than in BON and QGP-1 (Fig. 4C). In addition, NT-3 cells also express the proangiogenic growth factor angiopoietin 2 (ANG2) to a higher extent than BON and QGP-1 cells (Fig. 4C). Upon cell transfer into a hypoxic environment (0.5% oxygen), we observed a further increase in VEGF (Fig. 4D), but not in ANG2 expression (not shown). ELISA confirmed secretion of VEGF into cell culture supernatants under hypoxic conditions (Fig. 4E). Establishment of a relevant hypoxic environment was confirmed by upregulation of carboanhydrase IX, a prototypic hypoxia-induced gene target (Fig. 4F–G).
These results demonstrate that the new NT-3 cell line should meet the criteria for a suitable model to study antiproliferative and antiangiogenic treatments.
Treatment response
Current guidelines for the treatment of advanced pancreatic NET recommend somatostatin analogues, chemotherapy with streptozotocin/5-FU, and the targeted therapies everolimus or sunitinib. To evaluate the potential of our new NT-3 cells for studying upcoming novel therapies, we aimed to establish the efficacy of current treatments compared with QGP-1 cells. Upon treatment with 100 nmol/L octreotide, we observed a significant (−34.8%, P < 0.001) reduction in NT-3 cell numbers after a treatment for 5 days (Fig. 5A). In contrast, the low SSTR-expressing QGP-1 cells did not show significant reduction of cell numbers (−4.6%, P = 0.22). Treatment of cell cultures with 50 nmol/L everolimus resulted in a 31.5% and 26.2% reduction of NT-3 and QGP cell numbers after 5 days (P < 0.001), respectively (Fig. 5B). Increasing the dose up to 500 nmol/L did not further increase treatment response (data not shown). Treatment with sunitinib in doses up to 500 nmol/L did not decrease viability or proliferation in either cell line (data not shown). To evaluate treatment response to standard cytotoxic chemotherapy, we analyzed the dose-dependent effect on cell viability after incubation with streptozotocin and 5-FU. After 5 days, streptozotocin treatment at 1 mmol/L resulted in a viability decrease of −43.3% ± 1.1% in NT-3 (P < 0.001) compared with a lesser effect of −18.9% ± 7.5% (P = 0.10) in QGP-1 (Fig. 5C). Increasing the dose to 10 mmol/L substantiated the difference in susceptibility toward streptozotocin. In contrast, both cell lines responded well to 5-FU; treatment with 10 μmol/L for 5 days resulted in decreased viability of −30.1% ± 8.8% (P = 0,014) in NT-3 compared with −49.7% ± 6.0% in QGP-1 (P = 0.0002; Fig. 5D).
Treatment response in NT-3 compared with QGP-1. Treatment efficacy of various treatments was assessed in NT-3 and QGP-1 cells after 5 days of treatment by MTT assay. Octreotide at 100 nmol/L (A) compromised NT-3 viability, whereas everolimus at 50 nmol/L was equally effective in NT-3 and QGP-1 (B). Addition of genotoxic chemotherapeutics STZ (C) and 5-FU (D) showed a stronger effect for streptozotocin in NT-3 than QGP-1 cells at all tested doses, but similar responses for 5-FU in both cell lines.
Xenograft animal model
On the basis of our previous experience with BON and QGP-1 xenografts (11), we transplanted 2 × 106 NT-3 tumor cells into the flanks of NOD/SCID mice and we achieved a tumor take rate of 94% (15/16). NT-3 tumor growth in vivo was very slow, with the first palpable nodes detected approximately 6 weeks after cell transplantation. Because of the small tumor size during the first 12 weeks, traditional caliper measurements were unreliable and we employed serial MRI of subcutaneous tumor nodules to monitor their growth (Fig. 6A). The tumors were akin to human neuroendocrine tumors on MRI imaging, with a strong enhancement in T2-weighted sequences (Fig. 6A). Median NT-3 tumor growth rate (as measured by tumor volume) was +139% ± 13% per 4 weeks (Fig. 6B). The proliferative index corresponded well to the proliferative index of the original tumor (15%–20% Ki-67 labeling rate; focally up to 25%). mRNA expression profiles of neuroendocrine markers in subcutaneous tumors were in line with expression profiles of NT-3 cells in vitro (Fig. 6C). In contrast to our previous xenograft experiments with BON and QGP-1 cells (11), we did not detect any necrotic areas within the tumor, while macroscopic appearance of NT-3 tumors showed high vascularization (Fig. 6D). We detected high mRNA expression of human CgA, synaptophysin, and the somatostatin receptors SSTR1, 2, 3, and 5. In addition, serum analyses of tumor-bearing mice after 16 weeks demonstrated detectable levels of human insulin (median 118 mU/L, range 6–412 mU/L), and insulin expression in xenograft tumor cells was verified by IHC (Fig. 6E). These results confirmed the functional well-differentiated phenotype of the tumor xenograft.
Xenograft tumor model. Subcutaneous tumors in NT-3 transplanted NOD/SCID mice were first detected by MRI (A). Growth curves were calculated from serial imaging of tumors over 16 weeks (B). Expression profiles of tumor xenografts resembled the well-differentiated phenotype of NT-3 cells in vitro (C). Normal murine liver tissue of control mice (no xenograft) served as negative controls. Xenograft tumors were highly vascularized and correspondingly expressed high levels of human VEGF A (D). Analysis of human insulin content in mouse serum 16 weeks after transplantation of tumor cells confirmed the functionally active phenotype of NT-3 in vivo. Correspondingly, IHC showed insulin expression in xenografts (E). Histology of dissected xenograft tumors confirmed neuroendocrine identity and displayed a morphology and proliferation rate akin to the original human tumor (F). Scale bar, 1 cm for MRI images and 200 μm (10×) and 50 μm (40×) for microscopic images.
Histology of dissected xenograft tumors at the end of animal experiments recapitulated the insular and nesting growth pattern of the original patient's tumor with highly abundant stroma (Fig. 6F). Positive staining for CgA and synaptophysin confirmed the neuroendocrine identity of xenografts. Also, IHC for human cytokeratins revealed CK18 and CK19 expression in both the original tumor and in NT-3 xenografts (Supplementary Fig. S5). Although not being a focus of this study, initial pilot experiments using intraportal injection of NT-3 cells led to development of small liver tumors in one out of three transplanted mice. These were also rich in stroma tissue and were positive for CgA, indicating that our new cell line will also be suitable for future experiments on orthotopic NET liver metastases (Supplementary Fig. S6).
Discussion
In vitro and in vivo evaluation of future therapies in preclinical models remains the mainstay of successful development of novel treatments ultimately improving patient care in cancer patients (18, 19). The use of cell lines has recently been challenged by the announcement of the National Cancer Institute (NCI) to abandon their NCI-60 panel of cell lines and to encourage the use of patient-derived xenografts. Nevertheless the NCI still recommends the use of well-characterized novel primary cell cultures to explore future therapeutic options, and our novel cell line and the corresponding xenograft model fulfill these criteria.
So far, testing of potentially new targets in neuroendocrine tumor disease has been impaired by the lack of suitable preclinical models. The available genetic NET models display either a rapid and short-term fatal disease course, that is, Rip1-Tag2 mice (20), or a very benign course without frequent development of disseminated tumor disease, such as β-cell–specific MEN1 knockout mice (21). Likewise, the reported xenograft models of QGP-1 and BON cells show extremely rapid tumor growth with development of large necrotic areas and severe tumor disease in mice within a few weeks (10, 11). Hence, there is great need for a suitable and clinically relevant in vivo tumor model to assess future therapies in NET disease. The here reported novel pancreatic neuroendocrine tumor cell line NT-3 closely resembles the patient's original tumor: it has a well-differentiated phenotype, is functionally active, and has a slow growth rate in vitro and in vivo, thereby recapitulating the cardinal features of neuroendocrine tumors and providing a hitherto unavailable relevant preclinical neuroendocrine tumor model.
Currently, only few NET cell lines are available. Of these, the neuroendocrine cell lines derived from small intestinal NETs (KRJ-1, GOT1, P-STS) probably display a well-differentiated phenotype. Because of their limited availability, a thorough characterization of their biology and validation as a relevant NET model in vivo is still pending. From a clinical point of view, treatment of pancreatic NET is more challenging and associated with a worse prognosis than small intestinal NETs. The two available pancreatic neuroendocrine cell lines BON and QGP-1 have so far been widely used as a bona fide NET model. Documented fast growth rates of these tumor cells have put their usefulness as NET models into question. Our detailed characterization of BON and QGP-1 further substantiates such concerns as expression of neuroendocrine markers in these cell lines was low. Given that both cell lines have been in use for more than 25 years, it appears likely that the cells have acquired a more malignant phenotype in culture. Although the currently detected doubling times of 36 and 38 hours for BON and QGP-1 cells, respectively, are higher than the originally reported doubling times of 60 hours (BON) (22) and 84 hours (QGP-1; ref. 23), these cells proliferate much faster, even in their early passage numbers, compared with NT-3 with a doubling time of more than 240 hours under growth factor–stimulated conditions. The recent discovery of mutations in RAS and TP53 genes in BON and QGP-1 added to these concerns as well-differentiated NET rarely harbor such mutations (24–26). In contrast, our new NT-3 cell line does not harbor mutations in RAS or TP53 genes thereby resembling the nonmutated status of well-differentiated NET for these oncogenes. It is open to speculation, whether potentially effective new therapeutics for NET patients have been dismissed or development has been prematurely stopped due to failure in BON and QGP-1 cells.
We found a missense mutation in the MEN1 gene in NT-3 cells. This mutation is listed as a known SNP (rs2959656) with a minor allele frequency between 6% and 16% in different populations. Although the SNP has been described in sporadic insulinoma, parathyroid adenoma, and hemangioblastoma (27), the pathogenic role of this variant is controversial due to the high frequency in the general population. Targeted sequence analysis of “classical” pNET-associated genes, that is, VHL, MEN1, TSC1, TSC2, DAXX, and ATRX did not reveal any further mutations, which is in line with findings from Jiao and colleagues reporting mutations in either of these genes in only 68% of patients with sporadic pancreatic NETs (26). The lack of such mutations in NT-3 offers exciting new opportunities for future analyses as targeted knockdown will help to establish the role of either of the above-mentioned genes with regard to tumor progression and therapeutic response of pNETs. Furthermore, detailed genetic and proteomic analyses could identify yet unknown mechanism for the development and progression of pNETs.
One of the main obstacles in obtaining primary NET cultures is the unsolved question of optimal growth conditions for these cells (12). We here report an essential role for the supplementation of growth factors EGF and FGF-2 to establish proliferating cell cultures. Earlier studies in BON and QGP-1 cells also found a growth stimulatory role for EGF, FGF-2, TGF-α, and IGF-1 (28–30). Recently, a c-MET stimulating antibody has been identified to promote primary NET cell growth in vivo (31). As expression of these growth factors apart from IGF-1 and TGF-α is generally low in NET cells, the tumor microenvironment most likely contributes these growth stimulatory factors (28, 32, 33). Indeed, our NT-3 xenograft model of NT-3 supports this hypothesis as NT-3 cells, surrounded in vivo by abundant stroma, show a proliferation rate comparable with the in vitro growth rate with growth factor stimulation. For further exploration of such interesting new aspects in NET tumor biology, we have conserved tumor-associated fibroblasts from the original patient's tumor, as recently recommended by the NCI for novel primary tumor cell cultures, to study the interaction of NET cells with their microenvironment.
The documented insulin expression and secretion of NT-3 cells warrants evaluation as a potential model for human pancreatic beta cells. Until now, no such human model exists and most basic diabetes research has been performed in a rat insulinoma cell line (INS1E), with obvious limitations such as differences in protein expression and functional status between human and rodent islet cells (34, 35). The only human cell line derived from an insulinoma (called CM) has already lost insulin expression and secretion during in vitro culture (36). Recently, a genetically engineered human beta cell line has been presented, but still needs to prove easy access, handling, and a stable phenotype for basic research (37, 38).
Well-differentiated pancreatic NETs are currently treated with somatostatin analogues or STZ/5-FU, and these substances are also effective for growth reduction in our NT-3 cells in vitro. Indeed, it has been shown that STZ is particularly potent in SSTR-positive NET (39), pointing toward an increased susceptibility in tumors with a well-preserved neuroendocrine phenotype due to a hitherto undiscovered mechanism. In contrast, 5-FU, which is a general cytotoxic chemotherapeutic agent and therefore also active against dedifferentiated tumors, is also effective in QGP-1. Everolimus was similarly effective in both, NT-3 and QGP-1, as the mTOR pathway is activated in well- and less differentiated tumors (40).
Two of the most effective treatments for patients with NET (e.g., somatostatin analogues and peptide radionuclide therapy) are dependent on expression of SSTRs, in particular, SSTR2 and SSTR5, on the tumor cell surface (41). Until now, improvements in SSTR targeting have been impaired by the lack of suitable and widely available preclinical models. Our novel NT-3–derived animal model has the potential to overcome these limitations as we observed high expression of SSTRs in vitro and in xenograft tumors. Recently, clinical pilot studies were initiated to investigate combinations of radionuclide SSTR-targeted therapies and chemotherapies in NET (42). Making well-differentiated patient-derived cells with relevant SSTR expression, that is, our NT-3 model, widely accessible to translational researchers will aid such attempts in providing a platform to optimize timing and dosing of therapies. Assessing the effect of cotherapeutics on SSTR expression in NT-3 will help choosing effective drug combinations. As a substantial fraction of NET tumors, especially pancreatic NET, do not express sufficient levels of SSTR2 and SSTR5, these patients are currently not amenable to effective treatments. As NET cells do not only express SSTR, but also many other neuroendocrine-specific cell surface receptors (e.g., for incretin hormones like GIP and GLP-1), searching for and evaluating such receptors as therapeutic targets might stimulate development of novel therapies (43).
In summary, we have successfully established a well-differentiated and slow growing pancreatic NET cell line. NT-3 cells and the corresponding xenograft animal model will overcome current limitations in developing and testing novel therapies for this difficult to treat disease.
Disclosure of Potential Conflicts of Interest
D. Benten reports receiving commercial research support as research funding. J. Schrader reports receiving a commercial research grant from Novartis and has received speakers bureau honoraria from Novartis and IPSEN. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: D. Benten, J.R. Izbicki, A.W. Lohse, J. Schrader
Development of methodology: D. Benten, D. Perez, J. Schrader
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Benten, Y. Behrang, L. Unrau, V. Weissmann, G. Wolters-Eisfeld, S. Burdak-Rothkamm, F.R. Stahl, M. Anlauf, P. Grabowski, M. Möbs, J. Dieckhoff, B. Sipos, C. Eggers, D. Perez, M. Bockhorn, J. Schrader
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Benten, Y. Behrang, L. Unrau, G. Wolters-Eisfeld, S. Burdak-Rothkamm, M. Anlauf, B. Sipos, M. Bockhorn, A.W. Lohse, J. Schrader
Writing, review, and/or revision of the manuscript: D. Benten, Y. Behrang, M. Möbs, B. Sipos, M. Bockhorn, A.W. Lohse, J. Schrader
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Dieckhoff, M. Fahl, M. Bockhorn, J.R. Izbicki, A.W. Lohse
Study supervision: D. Benten, J.R. Izbicki, A.W. Lohse, J. Schrader
Acknowledgments
This work was funded by Forschungsförderungsfond Medizin of University Medical Center Hamburg-Eppendorf (to J. Schrader) and a research grant from the Theranostic Research Network Bad Berka (to J. Schrader). D. Benten and L. Unrau received grant support from Deutsche Forschungsgemeinschaft (SFB 841, project C7). G. Wolters-Eisfeld received grant support from Deutsche Forschungsgemeinschaft (WO 1967/2-1).
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.
Footnotes
Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).
- Received March 26, 2017.
- Revision received August 28, 2017.
- Accepted December 20, 2017.
- ©2018 American Association for Cancer Research.