Molecular mechanisms underlying the development and progression of pancreatic neuroendocrine tumors (PanNETs) are still insufficiently understood. Efficacy of currently approved PanNET therapies is limited. While novel treatment options are being developed, patient stratification permitting more personalized treatment selection in PanNET is yet not feasible since no predictive markers are established. The lack of representative in vitro and in vivo models as well as the rarity and heterogeneity of PanNET are prevailing reasons for this. In this study, we describe an in vitro 3-dimensional (3-D) human primary PanNET culture system as a novel preclinical model for more personalized therapy selection. We present a screening platform allowing multicenter sample collection and drug screening in 3-D cultures of human primary PanNET cells. We demonstrate that primary cells isolated from PanNET patients and cultured in vitro form islet-like tumoroids. Islet-like tumoroids retain a neuroendocrine phenotype and are viable for at least 2 weeks in culture with a high success rate (86%). Viability can be monitored continuously allowing for a per-well normalization. In a proof-of-concept study, islet-like tumoroids were screened with three clinically approved therapies for PanNET: sunitinib, everolimus and temozolomide. Islet-like tumoroids display varying in vitro response profiles to distinct therapeutic regimes. Treatment response of islet-like tumoroids differs also between patient samples. We believe that the presented human PanNET screening platform is suitable for personalized drug testing in a larger patient cohort, and a broader application will help in identifying novel markers predicting treatment response and in refining PanNET therapy.

In contrast to many other malignancies, there are no molecular characteristics and biomarkers supporting treatment decisions in pancreatic neuroendocrine tumors (PanNETs). While molecular mechanisms underlying PanNET development and disease progression are continuously further deciphered [1, 2] and numbers of clinically approved therapies are steadily rising, the treatment options for PanNET are still primarily chosen based on clinician judgment.

The lack of appropriate models and the rarity of PanNET disease are two major factors that hinder further advances in PanNET translational research. Testing more effective therapies as well as performing predictive studies are lagging behind. Currently, only a limited number of human PanNET cell lines are available – with BON1, QGP1 and CM being the most commonly studied [3-6]. It is pertinent to note that while these cell lines were used to experimentally dissect molecular mechanisms of NETs, they do not represent well-differentiated slow-proliferating PanNETs. All of these cell lines are highly proliferative and were found to differ fundamentally in their mutational genetic background compared to PanNETs. In fact, studies have shown a strong resemblance to poorly differentiated pancreatic neuroendocrine carcinomas rather than PanNETs [7-11]. Moreover, other authors have even questioned the tissue of origin of the aforementioned cell lines, raising a significant debate as to the translational relevance of work performed using these cell lines [11, 12]. Recently, Benten et al. [13] described NT-3 as a novel cell line that better reflects well-differentiated slow-proliferating PanNETs, which present the bulk of PanNETs. Nevertheless, the full molecular profile of NT-3 cells remains to be determined so that its similarity to the primary well-differentiated slow-proliferating subtype can be established. Moreover, neither available cell lines nor genetically engineered mouse models recapitulate the spectrum of different molecular subtypes found in human primary PanNETs [2, 14]. A stronger focus on developing more personalized in vitro models for studying these tumors is therefore demanded urgently. Cultivation and expansion of patient-derived neuroendocrine cells has been challenging owing to their intrinsically poor capacity for in vitro proliferation. However, due to advances in cell culture techniques, cell models of well-differentiated slow-proliferating PanNET derived from primary tissue have recently been used to study drug response and dissecting its underlying molecular mechanisms. For example, studies on isolated human primary PanNET cells cultured in vitro indicated that such a model might be utilized to determine patient response to treatment [15-18]. However, the major limitations of all these studies are the short cultivation window of the cells, the nonphysiological 2-dimensional (2-D) format with limited cell-cell interactions, as well as their small scale in terms of investigated treatments and patient numbers.

In recent years, there have been tremendous advances in the development of 3-dimensional (3-D) tissue culture techniques, including scaffold-free setups in ultra-low attachment plates or scaffold-based encapsulation cultures to allow cell growth in 3 dimensions [19-23]. Culture of cells in 3 dimensions mimics a more physiological architecture of a tumor tissue, including cell-cell contact and allowing the development of spatial differences in the culture system with respect to proliferation, cell death and hypoxia within spheroids [21, 24-26]. Additionally, cells kept in a 3-D format can be cultured and treated longer than in 2-D monolayers [21]. Cells cultured in 3 dimensions frequently display increased therapy resistance compared to cells cultured in 2 dimensions [27-29], where 3-D culture most likely better reflects the in vivo situation [30, 31]. For this reason, lately high-throughput screenings of pharmacological compounds were preferentially performed in 3-D-cultured cells [32-34].

With the presented study, we aimed for developing a platform to collect PanNET samples from multiple surgical centers, to isolate primary cells and to cultivate these cells in 3 dimensions retaining NET characteristics and finally to measure short- and long-term in vitro treatment response.

Lead Contact and Material Availability

Further information and request for resources and reagents should be directed to and will be fulfilled by the Lead Contact Dr. I. Marinoni (Ilaria.marinoni@pathology.unibe.ch).

Primary Cell Culture

Isolated primary PanNET cells were maintained in AdvDMEM + GF medium (DMEM-F12, 5% FBS, Hepes 10 mM, 1% L-glutamine (200 mM), 1% penicillin (100 U/mL), 1% streptomycin (0.1 mg/mL), 1% amphotericin B (0.25 mg/mL) (Merck, Switzerland), 20 ng/mL EGF, 10 ng/mL bFGF (Thermo Fisher Scientific, USA), 100 ng/mL PlGF, 769 ng/mL IGF-1 (Selleckchem, USA)) and in 24-well Corning® Costar® ultra-low attachment (ULA) plates (Corning, USA) (500 µL/well, 3–5 × 105 cells/well) in a humidified cell incubator (21% O2, 5% CO2, 37°C). For drug screen cells were resuspended in fresh AdvDMEM + GF medium supplemented with 123 µg/mL growth-factor-reduced Matrigel® (Corning, USA) and plated in 96-well ULA plates (50 µL/well, 3–4 × 103 cells/well).

To set up the PanNET screening platform including drug screening, we made use of primary material from a total of 16 PanNET patients depicted in online supplementary Table 1 (for all online suppl. material, see www.karger.com/doi/10.1159/000507669). Additionally, key resources used in this study are provided in Table 1.

Table 1.

Key Resource Table

Key Resource Table
Key Resource Table

Patient Studies

All patients agreed on the use of residual material and had signed an institutional informed consent. Patient characteristics are shown in online supplementary Table 1. The study was approved by the cantonal authorities (Kantonale Ethikkomission Bern, Ref.-Nr. KEK-BE 105/2015) and the Italian ethic commission (Comitato Etico, CE 252/2019).

Patient Samples and Cryopreservation

Fresh human PanNET tissue was obtained from patients diagnosed with PanNET undergoing surgery at the Inselspital Bern, Switzerland, or at the Pancreatic Surgery Unit, Pancreas Translational and Clinical Research Center, San Raffaele Scientific Institute, Milan, Italy. Tumor tissues of 16 PanNET patients were used for 19 isolations, 12 isolations for method establishment and 7 for a proof-of-concept drug screening. Patient characteristics are summarized in online supplementary Table 1.

Upon surgical resection a pathologist processed one mirror block of the tumor tissue to 8-mm3 blocks under sterile conditions avoiding necrotic regions. These blocks were suspended in recovery cell culture freezing medium (Thermo Fisher Scientific, USA), cryopreserved using an isopropyl alcohol freezing container (Nalgene, USA) and stored in liquid nitrogen. The other mirror block was embedded in a microcassette, and fixation was performed with a PAXgene Tissue System according to the manufacturer’s instructions. In short, tissue was incubated in a PAXgene Tissue FIX Container (Qiagen, Germany) at room temperature overnight. Fixated tissue was transferred into a PAXgene Tissue FIX Container (Qiagen, Germany) at 4°C until paraffin embedding (1–2 days) or kept at –20°C if embedding was not performed instantly.

Primary Cell Isolation and Culture

For primary cell isolation, tissue was thawed for 45–60 s in a 37°C water bath and cut into 1-mm3 pieces and washed with medium (advanced DMEM-F12, Hepes 10 mM, 1% L-glutamine, 1% penicillin-streptomycin-amphotericin B) and Dulbecco’s phosphate-buffered saline (DPBS; Thermo Fisher Scientific, USA). After aspiration of the medium, the tissue was incubated in 5 mL digestion medium (10 mg/mL collagenase IV (Worthington, USA), 0.25% Trypsin-EDTA (Sigma-Aldrich, Switzerland), 10 mg/mL DNase (Roche, Switzerland) in advanced DMEM-F12, Hepes 10 mM, 1% L-glutamine, 1% penicillin-streptomycin-amphotericin B) in a gentleMACSTM dissociator (Miltenyi Biotec, Switzerland) for 1 h at 37°C (programme TDK_1). After digestion, trypsin was deactivated with AdvDMEM (advanced DMEM-F12, 5% FBS, Hepes 10 mM, 1% L-glutamine, 1% penicillin-streptomycin-amphotericin B), and cells were filtered through a 70-µm smart strainer (Miltenyi Biotec, Switzerland) to remove debris of collagen. Red blood cells were lysed for 3 min with ACK lysis buffer (Thermo Fisher Scientific, USA) at room temperature. After 180 g centrifugation and aspiration of supernatant, the pellet was resuspended in AdvDMEM + GF medium (DMEM-F12, 5% FBS, Hepes 10 mM, 1% L-glutamine, 1% penicillin-streptomycin-amphotericin B, 20 ng/mL EGF, 10 ng/mL bFGF (Thermo Fisher Scientific, USA), 100 ng/mL PlGF, 769 ng/mL IGF-1 (Selleckchem, USA)). The cell suspension was plated in 24-well plates (cell + coated and tumor cells tested, Sarstedt, Germany) followed by a short spin for 30 s, at 200 g and incubation for 1 h (21% O2, 5% CO2, 37°C) to partially segregate fibroblasts by attachment. The supernatant was collected. For single cell dissociation the cell suspension was transferred into a 5-mL falcon tube and shortly spun down depending on cell/aggregate size. If large aggregates were present, cells were spun at 100–200 g; if smaller aggregates were present, cells were spun at 200–300 g. The cell pellet was washed with DPBS and incubated in Accutase (Thermo Fisher Scientific, USA) supplemented with DNase (10 mg/mL) (Thermo Fisher Scientific, USA). Cells were carefully dissociated by repeated (10–15×) -passage through a 1-mL syringe and 26 G 0.45 × 13 mm MicrolanceTM (BD, Switzerland) until aggregates were not visible anymore. The cell number was estimated using a hemocytometer, and cells were seeded in Adv-DMEM + GF medium in 24-well ULA plates (5 × 105 cells/mL/well). After 2 days of recovery phase, cellular aggregates were collected in 15-mL falcon tubes and differentially centrifuged at 120 g for 5 min to separate cells and aggregates from debris/apoptotic cells. Supernatant was aspirated to remove semi- and nonviable cells. Cells were counted and resuspended in fresh AdvDMEM + GF medium supplemented with 123 µg/mL growth-factor-reduced Matrigel and plated in 96-well ULA plates (50 µL/well, 3–4 × 103 cells/well). The setup consisted of 6–9 DMSO-positive control wells, 6 no-cell-negative control wells and technical triplicates for each drug concentration.

Viability Measurement

RealTime-GloTM MT Cell Viability (RTG) assay (Promega, Switzerland) was used to repeatedly monitor cell viability in 3-D human primary PanNET culture. The RTG assay was performed according to the manufacturer’s instructions, and luminescence was measured in an Infinite® 200 PRO plate reader (Tecan, Switzerland). In brief, after 3 days of sphere formation, conditioned medium of each well was supplemented with additional 50 µL of fresh AdvDMEM + GF medium containing Matrigel and RTG assay reagents (2×) to a final volume of 100 μL. Growth factors and FBS were replenished from a concentrate (0.77 µL GFs (130×) + 5 µL FBS) every 3–4 days in each well using a multichannel pipette. A 6-h RTG baseline before adding drug compounds was recorded for every well at day 0 of the drug screen. For calculating the in vitro growth curve, relative luminescence unit (RLU) values were normalized to corresponding baselines. For calculating the in vitro drug response, RLU values were normalized first to corresponding baselines followed by normalization to the DMSO control wells of a particular day as described in more details in the paragraph “curve fitting and drug sensitivity data.”

Micro-Cell-Block from Islet-Like Tumoroids

Islet-like tumoroids corresponding to 3–5 × 104 cells were collected in a 1.5-mL Eppendorf tube (either directly on the day of isolation [D0] or from 6–9 wells of a 96-well ULA plate at the end of a drug screen [D15]). Tumoroids were washed with DPBS and pelleted at 350–500 g. The supernatant was aspirated, and the cells were resuspended in human plasma derived from whole blood (Interregional Blood Transfusion SRC, Epalinges, Switzerland) and Thrombin (Diagnotec, Switzerland) (ratio 5:1) followed by 3-min incubation at room temperature. The clot was fixed with 4% PFA for 30–60 min protected from light. After a DPBS wash, the supernatant was aspirated and cells were incubated in a hematoxylin and DPBS solution (ratio 1:8) on a rocker shaker for 10–15 min at room temperature. The counterstained clot was transferred to a plastic microcassette for paraffin embedding. For immunohistochemistry the embedded material was cut into 2.5-µm-thick serial sections followed by deparaffinization, rehydration and antigen retrieval using an automated immunostainer (Bond RX, Leica Biosystems, Germany). Antigen retrieval was performed for Ki-67 (Dako M7240) with Tris for 30 min at 95°C, insulin (Sigma I-2018) and synaptophysin (Novocastra 27G12) with Tris for 30 min at 100°C. Antibodies were diluted as follows: Ki-67 1:200, insulin 1:4,000, synaptophysin 1:100. Slides were counterstained with hematoxylin. Scans were acquired with an automated slide scanner Panoramic 250 (3DHistech, Hungary) at 20× magnification. Images were acquired using QuPath software [35].

Drug Preparation

Compounds (sunitinib (S1042), everolimus (S1120) and temozolomide (S1237)) were obtained from commercial vendors and stored as stock aliquots at –80°C. A 5-point, 625-fold concentration range was used for all compounds in order to have enough data points and a sufficient large drug concentration window to calculate reliable absolute IC50 values [36]. The starting dosage for each compound was selected based on IC50 screens in cancer cell lines publicly available online (see Cancerrxgene.org, PharmacoDB, Cancer Drug Resistance DB), from literature search and/or from in vitro data from pilot human primary cell cultures and/or from PanNET cell lines (QGP1, NT3, BON1).

Curve Fitting and Drug Sensitivity Data

Drug-response curve data consisted of 6–9 DMSO-positive controls, 6 no-cell-negative controls, and 5 drug-response points for a 625-fold concentration range. For IC50 calculation RLU values that were derived from an RTG assay from short-term treatment on day 3 and long-term treatment on day 7 of each well were weighted and normalized as the following: RLU values from each 6-h RTG baseline measurement (RLUx d0) were scaled with the overall minimal value of day 0 for each plate (RLUmin d0) and transferred into a baseline weight (RLUx weight) for each well to minimize well-to-well variability:

Each RLU value from day 3 was then accordingly weighted to its baseline weight:

The percentage response from weighted RLU was calculated by normalizing each value to no-cell-negative control (0%) and DMSO-positive control (100%) intervals. These data points were fitted in a 4-parametric linear-regression model [34, 37] with two constraints, top = 100% and bottom = 0%, to estimate corresponding IC50. IC50 value differences of >4-fold were clustered in strong-responder and weak-responder groups. In case of an IC50 value >2.5-fold higher than the highest tested target-concentration samples were considered as non-responder (NR).

QuPath Image Analysis

Using QuPath software [35] digital-scanned hematoxylin-eosin and synaptophysin tissue sections were first preprocessed in the built-in visual stain editor using default settings for estimation of stain vectors to improve staining quality. Each tissue section was then superimposed with a 1,000-µm grid box. In each tissue section one representative tile out of the grid box was selected by a cytopathologist (M.T.) as a training set. Using a watershed segmentation method, positive and negative cells were automatically detected within each representative tile. Two pathologists (M.T., A.P.) then manually reconfirmed positive cell detection based on histomorphological features including cellular and nuclear shape, tumor cell nest formation, tumor columns, nuclear “salt and pepper” structure, nuclear and cytoplasmic staining intensity. A minimum of ≥1,000 cells were selected for each training set and a total of 67 parameters (including perimeter, circularity, staining optical density, etc.) were included for training of the random-trees machine learning classifier. The auto-update tool within QuPath allowed real-time reconfirmation of training efficiency/accuracy. These cell detection parameters were applied on the whole tissue slides by creating a script which performed automated cell classification/annotation. Detection results were extracted from QuPath and imported and analyzed within R.

Graph Pad Prism (Version 8.2.1) and R statistical environment were used for data analysis and visualization in the R version 3.6.2 (2019-12-12) platform: x86_64-w64-mingw32/x64 (64-bit). Attached base packages: Grid; stats; graphics; grDevices; utils; datasets; methods; base. Other attached packages: 81) scales_1.1.0; (2) MASS_7.3–51.4; (3) reshape2_1.4.3; (4) ConsensusClusterPlus_1.50.0; (5) circlize_0.4.8; (6) ComplexHeatmap_2.2.0; (7) RColorBrewer_1.1-2; (8) Rmisc_1.5; (9) plyr_1.8.5; (10) lattice_0.20-38; (11) plotrix_3.7-7; (12) cowplot_1.0.0; (13) forcats_0.4.0; (14) stringr_1.4.0; (15) dplyr_0.8.3; (16) purrr_0.3.3; (17) tidyr_1.0.0; (18) tibble_2.1.3; (19) ggplot2_3.2.1; (20) tidyverse_1.3.0; (21) broom_0.5.3; (22) readr_1.3.1.

Hierarchical Clustering Analysis of Drug Response Profiles

Using the ConsensusClusterPlus pipeline [38], the number and membership of clusters was determined for drug response profiles based on patient-specific IC50 values of all three drug treatments. Distances were calculated using Pearson’s correlation sorted by an agglomerative hierarchical clustering algorithm. The WardD2 algorithm was used for subsampling, and the final consensus matrix was determined by group average.

Gene Expression Analysis

Mean expression values of growth factor receptors were analyzed in publicly available data of 26 PanNET patients. RNAseq data were downloaded from the ICGC Data Portal (PAEN-AU project). QC, mapping/alignment and raw count quantification is described in Scarpa et al. [2]. From RSEM data output for our downstream analysis we chose the FPKM (fragments per kilobase of exon per million fragments mapped) normalization method to account for sequencing depth and gene length for all raw read counts. A list of all available growth factor receptor was acquired from the UniProt Knowledgebase [39]. Expression values of all targets were transformed into a 0-to-1 scale for each patient to allow interpatient comparability. Mean values for each target receptor were then calculated in all of the 26 PanNET patients:

Expression value = mean (scaled0.1 [FPKM normalized raw counts]).(3)

χ2 Test and Monte-Carlo Simulation

A χ2 test of independence was conducted among all variables of interest. In order to meet requirements for χ2 test statistic and to account for relatively small expected cell frequencies our data set was resampled using a Monte Carlo simulation (replication = 1 × 105) allowing to calculate p value estimates.

Data and Code Availability

The original RNAseq data set from human primary PanNET is publicly available at the ICGC Data Portal (PAEN-AU project). The complete expression data of growth factor receptors are available in the supplementary data sheet. The code supporting the current study has not been deposited in a public repository because the analysis code was generated from generic R packages, but the code is available from the corresponding author on request.

Cryopreservation Allows a Multicenter Approach

PanNET tumors are rare, therefore a collaborative network is crucial. Here we propose a novel platform for an effective multicenter approach which permits biobanking of cryopreserved PanNET tissues from multiple surgical centers by a single central institution which performs primary cell isolation and drug testing (Fig. 1a). After tumor resection, half of the specimen was formalin-free PAXgene-fixated and paraffin embedded. These so-called mirror blocks served as controls for sample quality and were used to preassess patient-specific PanNET characteristics and tumor cell content in hematoxylin-eosin stainings and synaptophysin immunohistochemistry – a NET biomarker routinely used in clinics for diagnosis of PanNET. The other half was immediately cryopreserved in recovery freezing medium and later shipped and processed for primary cell isolation and in vitro drug screening.

Fig. 1.

Human primary PanNET cells form islet-like tumoroids and retain a neuroendocrine phenotype in vitro. a Schematic representation of human PanNET screening platform. b Venn diagram displaying usage of human PanNET patient material (outer circle) and individual patients (inner circle). The success rate for drug screening in PanNET patient material was 86% (6/7). Detailed log of cell isolation is provided in the online supplementary data file 1. c Representative hematoxylin-eosin (HE) and immunohistochemistry staining of islet-like tumoroids from B992 and original primary tumor tissue B992 (top) and islet-like tumoroids from B563m and original metastatic tumor tissue from B563m (bottom). Cultured cells were formalin-fixed and embedded after 14 days in PanNET culture medium. Formalin-fixed primary PanNET tissue or cultured cells were stained for hematoxylin-eosin (HE), synaptophysin (SYN), insulin (INS) and Ki-67. Immunohistochemistry slides were counterstained with hematoxylin. Scale bar, 50 µm.

Fig. 1.

Human primary PanNET cells form islet-like tumoroids and retain a neuroendocrine phenotype in vitro. a Schematic representation of human PanNET screening platform. b Venn diagram displaying usage of human PanNET patient material (outer circle) and individual patients (inner circle). The success rate for drug screening in PanNET patient material was 86% (6/7). Detailed log of cell isolation is provided in the online supplementary data file 1. c Representative hematoxylin-eosin (HE) and immunohistochemistry staining of islet-like tumoroids from B992 and original primary tumor tissue B992 (top) and islet-like tumoroids from B563m and original metastatic tumor tissue from B563m (bottom). Cultured cells were formalin-fixed and embedded after 14 days in PanNET culture medium. Formalin-fixed primary PanNET tissue or cultured cells were stained for hematoxylin-eosin (HE), synaptophysin (SYN), insulin (INS) and Ki-67. Immunohistochemistry slides were counterstained with hematoxylin. Scale bar, 50 µm.

Close modal

Development of a 3-D Human Primary PanNET Cell Culture Model

Within this study, we performed 19 isolations from samples of 16 PanNET patients. Patient characteristics are summarized in online supplementary Table 1. In the first part of the study (12 patients), we set up the sampling, isolation and culture conditions. Two additional isolations were used for live-cell imaging to observe tumoroid formation. In the proof-of-concept part of the study (7 patients, including 3 patients who were also included in the method and development cohort), we tested the suitability of the setup for in vitro drug screening (Fig. 1b). During method development, we successfully isolated small aggregates and single cells from cryopreserved patient material in 73% (8/11) (online supplementary data file 1). Age, sex and other clinical parameters such as WHO grade, TNM staging and Ki-67 indices did not reveal a significant association with the isolation success and/or cellular yield (online suppl. Fig. 1A).

In 3 patient samples, isolation did not yield aggregates or single cells, which we attribute to the collection of largely acellular fibrotic or necrotic tissue as revealed from hematoxylin-eosin staining of corresponding mirror blocks (online suppl. Fig. 1B). Therefore, careful macroscopic selection of viable cellular tumor regions is crucial. Upon implementation of SOPs for sampling at the surgical centers, we observed strong quality improvements. In 2 patient samples (P005, P051) yielding successful cell isolation, we detected overgrowth of fibroblasts after 15 days (online suppl. Fig. 1C). This led us to implement a low FBS concentration in the culture medium and to include a partial fibroblast reduction step during cell isolation. Thus, the cell suspension was plated on coated plastic for 2 h followed by gently rinsing for the collection of low-adhesive nonstromal cells. Two tumor cell extracts (P032x, P033) were negative for synaptophysin immunohistochemistry (IHC) staining on conventional cytospin preparations and hence excluded for further culture. However, a post hoc analysis by a cytopathologist (M.T.) revealed that these cells were tumoral cells (online suppl. Fig. 1D). Consequently, to assess tumor cell content accurately, we implemented a formalin-fixating paraffin-embedding technique termed Micro-Cell-Block. Micro-Cell-Blocks retain cellular- and tumoroid morphology and require only a low cell number. Micro-Cell-Blocks on the day of isolation (D0) served as an internal quality control to assess successful fibroblast removal and to guide decision for continuation of the drug screening pipeline. Micro-Cell-Blocks at the end of the experiment (D15) allowed to quantify tumor cell content from synaptophysin immunohistochemistry and to reconfirm target cell identity on hematoxylin-eosin staining.

To account for more physiological cell culture conditions a PanNET-specific culture medium was developed combining literature and human transcriptomic data from 26 low-grade PanNETs [2]. We selected growth factors that were frequently reported in the PanNET literature [13, 15, 40] and for which – except of EGFR – all the target receptors (FGFR1, IGFR1/2, FLT1/VEGFR1, EGFR) were within the upper expression quintile (<28/151) of all currently available growth factor receptors and related proteins [40] in human PanNET patients (online suppl. Fig. 1A). Additionally, PanNET culture medium was supplemented with a low percentage of extracellular matrix. Several findings showed improved in vitro culture from Matrigel complementation due to scaffolding support [41-45]. Low concentration Matrigel supplementation stabilized PanNET culture, without leading to a transient artificially increased cellular growth as seen with higher supplementation (data not shown).

Human Primary PanNET Cells Form Islet-Like Tumoroids and Retain a Neuroendocrine Phenotype in vitro

After isolation and cell culture refinement we performed live-cell imaging in two human PanNET samples. We isolated single cells from cryopreserved primary PanNET tissue (B992) and PanNET liver metastasis (B563m). Isolated cells from both patient specimens were viable. Live-cell imaging for 12 days revealed that isolated cells formed structures similar to extracted murine islets [46] which we hence termed islet-like tumoroids (online suppl. Fig. 2B). Islet-like tumoroids reached a more defined round structure after 72 h through aggregation and thereafter remained stable in volume (online suppl. Fig. 2C, suppl. video). Fourteen days after isolation, histomorphology of the islet-like tumoroids was examined and compared to corresponding mirror blocks. Islet-like tumoroids from primary and metastatic PanNET patient samples retained expression of synaptophysin, confirming that most of the cells consisted of tumor cells with preserved neuroendocrine phenotype (Fig. 1c). Furthermore, islet-like tumoroids from B992 expressed insulin as the original tumor tissue. The low percentage of Ki-67-positive tumor cells (2%) in vitro matched with the proliferation index described in the original tumor tissue (Ki-67 index <2%) (Fig. 1c, top, online suppl. Table 1). Similarly, the percentage of proliferating cells was retained in the metastatic PanNET sample (B563m), with a Ki-67 index of 12% in the original tumor tissues and 15% in cultured cells, respectively (Fig. 1c, bottom, online suppl. Table 1).

Setting Up an in vitro Drug Screening Pipeline for Islet-Like Tumoroids

After the successful pilot experiment, we sought to implement a pipeline for in vitro drug screening (Fig. 2a). Following a 2-day recovery phase after initial isolation, cells were replated from a 24-well format into a 96-well format. By this time, the majority of semi- and nonviable cells from isolation had segregated from viable cells. As seen in previous live-cell imaging analysis (online suppl. Fig. 2B, C), during 72-h incubation in the 96-well plate, cells formed islet-like tumoroids with only minor changes thereafter indicating a suitable time point for starting the drug treatment. Growth factors were replenished on days D2, D5, D8 and D12 after initial isolation. Due to low cell numbers available from PanNET specimens we selected RTG – a metabolic nonlytic assay – as a surrogate of cell viability. Pretreatment 6-h baseline measurements were recorded to normalize on an individual well basis and to correct for variability in cell number. Viability of islet-like tumoroids in each well was repeatedly (8×) monitored over a time course of 10 days before storing the cell material for further downstream analysis.

Fig. 2.

3-D human primary PanNET in vitro model for a personalized drug-screening platform. a Detailed schematic representation of in vitro 3-D drug-screening pipeline in human primary PanNET cells. GF, growth factor replenishment; thin ticks + digit, RTG viability assessment; MCB, Micro-Cell-Block. b In vitro growth curve of all screened primary PanNET samples using the metabolic surrogate assay RealTime-Glo (RTG) in 3-D human primary PanNET culture. Cells were cultured in AdvDMEM + GF and a low percentage of matrigel in 0.16% DMSO for 10 days. Normalization was calculated based on per-well 6-h RTG baseline measurement. Data represent means ± SEM (n = 7, 3 technical replicates). RLU, relative luminescence unit.

Fig. 2.

3-D human primary PanNET in vitro model for a personalized drug-screening platform. a Detailed schematic representation of in vitro 3-D drug-screening pipeline in human primary PanNET cells. GF, growth factor replenishment; thin ticks + digit, RTG viability assessment; MCB, Micro-Cell-Block. b In vitro growth curve of all screened primary PanNET samples using the metabolic surrogate assay RealTime-Glo (RTG) in 3-D human primary PanNET culture. Cells were cultured in AdvDMEM + GF and a low percentage of matrigel in 0.16% DMSO for 10 days. Normalization was calculated based on per-well 6-h RTG baseline measurement. Data represent means ± SEM (n = 7, 3 technical replicates). RLU, relative luminescence unit.

Close modal

PanNET Screening Pipeline in Control Conditions

As a proof-of-concept experiment, we tested the PanNET drug-screening pipeline with 7 patient samples. Tumor cells were successfully isolated in all 7 PanNET specimens. Quantification of tumor cells in mirror blocks of original tumor tissue showed variable tumor content within tissues and among patients (70 ± 18%, n = 7) (online suppl. Fig. 3A, B). Yet, assessment of hematoxylin-eosin stainings and synaptophysin immunohistochemistry on Micro-Cell-Blocks on the day of isolation (D0) by a cytopathologist (M.T.) reconfirmed successful selection of tumor cells after isolation and fibroblast depletion in all samples used for drug screenings (93 ± 15%, n = 7) (online suppl. Fig. 3A, B). Moreover, successful culture of tumor cells was also reconfirmed in Micro-Cell-Blocks at the end of each experiment (D15) (95 ± 11%) (online suppl. Fig. 3A, B). In all patient samples, islet-like tumoroids were formed and remained viable in 85% (6/7) for 15 days in culture.

Following the metabolic activity of untreated islet-like tumoroids over the time course of 10 days, we observed an association between in vitro proliferation and Ki-67 index in original tumor tissue in the majority of samples: the metastatic patient sample (B563m) with a Ki-67 index of 15% in the original tumor tissue displayed the highest signal increase (3.8-fold), while in 4 patient samples with lower Ki-67 indices (P049, P050, P051, B931) the signal only increased between 1.1- and 1.7-fold (Fig. 2b). In patient sample P044 this association was weak exhibiting a 1.6-fold increase despite a Ki-67 index of 18% in the original tumor tissue. P040 was the only sample with a decreasing signal in the untreated condition; hence, long-term time points (>72 h) from this particular patient sample were not included in further analysis.

In vitro Drug Response in Islet-Like Tumoroids Shows Distinctive Sensitivity Profiles

To assess whether 3-D human primary PanNET culture could be exploited to predict patient drug response in vitro, we evaluated the effect of three clinically approved PanNET treatments on cell viability [47]. 3-D human primary PanNET cultures from 7 different patients were screened for response to sunitinib, everolimus and temozolomide. A 5-point, 625-fold drug concentration range ensured a sufficient exploratory drug screening window for accurate IC50 estimation based on mathematical modeling [36]. As starting points IC50s from publicly available databases were interrogated for each drug, followed by pilot assessments of their antiproliferative effect in PanNET cell lines and murine primary cells (data not shown) as well as further literature research. Cells were treated for 10 days, and viability was repeatedly monitored at 8 time points during drug screening. Drug response profiles differed clearly among the three standard of care treatments. Dose-dependent effects of sunitinib and everolimus were observed in all tested patient samples (Fig. 3a, online suppl. Fig. 4A). Interestingly, comparing interpatient drug responses we detected varying treatment sensitivities among patients (online suppl. Fig. 4A). IC50 values determined after short-term (72 h) treatment displayed two clearly distinct groups within sunitinib treatment and within everolimus treatment harboring >4-fold differences in respective IC50 (Fig. 3b). Also consensus clustering matrix and hierarchical cluster analysis (k = 4) displayed robust response groups for short-term (72 h) treatments (Fig. 3c): a strong-responder group with samples sensitive to both treatments (P049), a group responding either primarily to everolimus (P049, P040, B563m) or to sunitinib (P050, B931, P051) – which was considered as mixed-responder group – and a weak-responder group including one sample insensitive to all treatments (P044). Importantly, in an integrative hierarchical cluster analysis, short-term treatment IC50s – for the majority of patients – clustered closely together with long-term treatment IC50s emphasizing robustness of the readouts (online suppl. Fig. 4B). Interestingly, in 1 case (B931) differences between short-term and long-term treatment were detected (online suppl. Fig. 4B).

Fig. 3.

Effect of standard of care pharmacological treatments on cell viability in 3-D human primary PanNET culture. a Representative in vitro viability curves using the metabolic surrogate assay RealTime-Glo (RTG) in 3-D human primary PanNET culture (P050) treated with 0.16% DMSO (control, Ctrl) or indicated treatment sunitinib (SUN), everolimus (EVE) and temozolomide (TEM) for 10 days. Normalization was calculated based on per-well 6-h RTG baseline measurement and corresponding DMSO control of the respective day. For all tested compounds a 5-point, 625-fold concentration range was used based on vast literature research and in-house in vitro preliminary studies. Data represent means ± SEM (n = 1 per patient, 3 technical replicates). RLU, relative luminescence unit. b In vitro dose response curves of screened PanNET patient displaying IC50 for SUN, EVE and TEM after short-term (72 h) treatment. Treatment responses (means ± SEM) were fitted into a 4-parameter logistic regression model in GraphPad software to calculate absolute IC50. Data represent fitted curve means (n = 7). Dotted line, absolute IC50. c Heat map comparing absolute IC50s for SUN, EVE and TEM in 3-D human primary PanNET culture after short-term (72 h) treatment. The heat map was derived using the WardD2 clustering method with displaying Pearson’s clustering distance using ComplexHeatmap R-package [38]. The color code represents the scaled IC50 (Z score) for each drug. Each row represents an individual patient.

Fig. 3.

Effect of standard of care pharmacological treatments on cell viability in 3-D human primary PanNET culture. a Representative in vitro viability curves using the metabolic surrogate assay RealTime-Glo (RTG) in 3-D human primary PanNET culture (P050) treated with 0.16% DMSO (control, Ctrl) or indicated treatment sunitinib (SUN), everolimus (EVE) and temozolomide (TEM) for 10 days. Normalization was calculated based on per-well 6-h RTG baseline measurement and corresponding DMSO control of the respective day. For all tested compounds a 5-point, 625-fold concentration range was used based on vast literature research and in-house in vitro preliminary studies. Data represent means ± SEM (n = 1 per patient, 3 technical replicates). RLU, relative luminescence unit. b In vitro dose response curves of screened PanNET patient displaying IC50 for SUN, EVE and TEM after short-term (72 h) treatment. Treatment responses (means ± SEM) were fitted into a 4-parameter logistic regression model in GraphPad software to calculate absolute IC50. Data represent fitted curve means (n = 7). Dotted line, absolute IC50. c Heat map comparing absolute IC50s for SUN, EVE and TEM in 3-D human primary PanNET culture after short-term (72 h) treatment. The heat map was derived using the WardD2 clustering method with displaying Pearson’s clustering distance using ComplexHeatmap R-package [38]. The color code represents the scaled IC50 (Z score) for each drug. Each row represents an individual patient.

Close modal

Current murine and human cell line models do neither accurately represent human well-differentiated slow-proliferating PanNETs, nor distinct human molecular subtypes, nor interpatient variability. In this study, we present a human PanNET screening platform allowing multicenter sample collection of cryopreserved patient specimens with an 86% success rate in primary cell isolation and cell culture. Isolated cells of well-differentiated slow-proliferating PanNET can be cultured in 3 dimensions and screened in vitro assessing response profiles to standard of care treatments for PanNETs. Since the cell number was the major limiting factor, we established protocols that are optimized for minimal amounts of resection specimens.

We present cryopreservation as a solution to make multicenter studies possible, thereby overcoming the issue of the rarity of PanNET samples. While difficult to implement in different centers, this generation of “living cell repositories” is promoted as innovative biobanking setting [48] and increasingly used in translational cancer research [49]. To account for more physiological conditions, growth factor supplementation for our PanNET culture medium was based on a combination of literature research [13, 15, 40], transcriptomic analysis of growth factor receptors as well as pilot experiments testing different growth factor concentrations and combinations. Final PanNET culture medium composition was selected according to best retention of viability during a 15-day period to minimize selection. This approach is clearly different from classical organoid approaches, where the culture medium selects for stem cell-like cells, and where these cells are kept individually in a biomatrix in order to produce clonal organoids [50, 51]. The aim of our presented “tumoroid” model is to in vitro treat a similar tumor cell composition as present in the patient. Following this approach, we can obtain a remarkable success rate of 86%, but we acknowledge that classical organoid models have many other advantages such as the potential to intervene mechanistically [52-54]. With a retention of ±70% of isolated cells in experiments before drug screening we believe that selection bias is minimal and that we are capable of treating the majority of cells representing the original tumor [55]. Compared to the limited number of studies using primary PanNET tissues in 2-D culture [15-18], we observe a reproducibly extended life span of the isolated cells up to 15 days. Longer experiments would also be possible at least in a subgroup of tumors; however, typically we did observe major responses already during the first 3 days of treatment.

We show that primary cells isolated from PanNET express original tumor characteristics and retain their neuroendocrine phenotype after 15 days. Interestingly, isolated cells form islet-like tumoroids in vitro. Similarly, nonneoplastic endocrine pancreatic cells are physiologically structured as islets. Kojima et al. reviewed the history of abundant findings which revealed that single cell suspension of endocrine pancreatic tissue from several species form islet-like structures and reconstitute their original architecture in vitro [56]. Currently, we do not know whether this reflects an endocrine specific phenotype or an even broader epithelial phenotype.

With the conditions presented, isolation and culture were successful in 6 out of 7 patients (86%). Only 1 sample (P040) showed a loss of viability after 7 days and was excluded from further long-term treatment analysis. During 3-D culture, all G1 PanNET patient samples expectedly displayed only marginal growth, whereas the metastatic patient sample B563m (Ki-67 index of 15%) showed the highest growth in vitro (3.8-fold in 10 days). One G2 sample (P044, Ki-67 index of 18%) exhibited a somewhat lower growth of 1.6-fold in 10 days. It seems not surprising to infrequently observe a rather weak association of Ki-67 index and in vitro proliferation. Indeed, a clear linear correlation of Ki-67 index to tumor growth rate has not been demonstrated clinically, to our knowledge. Biologically, proliferation represents only one aspect out of many: we neither have knowledge about a different fraction of cell death within our PanNET specimen nor do we know exact durations of cell cycles for the isolated cells from individual patient specimens.

In vitro treatment with clinically approved chemotherapeutics for advanced PanNET disease revealed overall distinctive response profiles and drug sensitivities based on IC50s. Comparing short-term (72 h) versus long-term treatment (7 days and 15 days) showed identical results in the majority of samples, indicating that these different time windows are potentially of minor importance. However, in 1 tumor (B931) we observed differences between short-term and long-term treatment. While it might not be important to use long-term treatment for detecting primary response in sunitinib and everolimus treatment, prolonged treatment could be of potential importance for other chemotherapeutics. In our series, we do not see a clear response to temozolomide in all of the 7 PanNETs examined. A possible explanation for this is the mechanism of action of temozolomide, which is strongly linked to cellular proliferation. Cytotoxicity of temozolomide is mediated by O6-methylguanine adducts, which can mis-pair with thymine during DNA replication. The resulting futile cycles of DNA processing induce cytotoxic double-strand DNA breaks that trigger apoptosis [57, 58]. Due to the low proliferation rate of our samples a time window of 10 days may be still too short for a detection of measurable effects. In line with that, the metastatic patient sample (B563m) that proliferated in vitro shows at least a faint response to temozolomide in our screen – even if the IC50 estimation is still far from our tested drug concentration window (online suppl. Fig. 4A) and even if the sample has been scored as weak responder. We can exclude nonpotency of the chemotherapeutic compound itself since our implemented 625-fold concentration window (0.46–288.00 μM) covers a sufficiently large drug window tested in PanNET and glioblastoma cell lines (data not shown) to eliminate this as a potential bias.

In other tumor entities it has been shown that ex vivo drug response correlates with patients’ response in primary cell culture approaches similar to our setup (e.g. esophageal adenocarcinoma, breast cancer, and head and neck squamous cell carcinoma) [59-61] and in patient-derived xenograft models [62, 63]. Further studies are needed to evaluate whether the observed in vitro sensitivity will correlate with clinical response in PanNET patients as well. To answer this question, first a prolonged clinical follow-up is crucial. Secondly, a larger patient cohort will be needed to perform correlation analysis and to have enough statistical power. Full clinical follow-up data of all enrolled patients are therefore collected. Chemotherapy-specific in vitro treatment duration and concentration range with the highest predictive value will be defined by comparison of clinical data with the in vitro drug response. Due to the nonlytic approach, the islet-like tumoroids are collected after the experiment and are available for next-generation-sequencing end point analysis, as is the original tumor material from patients. The presented PanNET screening platform might therefore serve as a basis for developing personalized treatment of PanNET patients, performing molecular fingerprinting on the original tissue to be able to potentially detect predictive markers.

We are well aware that the present protocol still bears limitations. It depends on surgical resection specimens of PanNET metastases and high-stage tumors; however, most of these patients are diagnosed via biopsies of liver metastases. With further experience the protocol has the potential of being adapted to biopsy specimens as well – yet – some biological role must be proven first to ethically justify additional biopsies. The composition of growth factors could be further refined, and the culture system does not factor in stromal and inflammatory features potentially contributing also to tumor response.

In conclusion, we present a 3-D human primary PanNET screening platform as a new preclinical model, which reflects the characteristics of an individual tumor and has the capability to detect differential treatment response. Therefore, this model has the potential to pave the way towards more personalized medicine for PanNET patients in the future, including better patient stratification and identification of novel and experimental treatments.

We thank the Tissue Bank Bern (Bern, Switzerland) and the Translational Research Unit (Institute of Pathology, Bern, Switzerland) for their technical, material and administrative support and the Cytopathology (Institute of Pathology, Bern, Switzerland) for technical support performing formalin-fixation and paraffin embedding in tissue and cell culture material. We also thank Dr. Joel Zindel, Mr. Philipp Zens and Mr. Hassan Sadozai for their constructive feedback on the manuscript drafts.

Subjects (or their parents or guardians) have given their written informed consent, and the study protocol was approved by the cantonal authorities (Kantonale Ethikkomission Bern, Ref.-Nr. KEK-BE 105/2015) and the Italian ethic commission (Comitato Etico, CE 252/2019).

All the authors declare no competing financial interest. There is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

The study was supported by Wilhelm Sander to Ilaria Marinoni and a Swiss Cancer League Grant (KLS 3360-02-2014) to Aurel Perren. Valentina Andreasi’s PhD scholarship and Francesca Muffatti’s research fellowship were supported by Gioja Bianca Costanza legacy donation.

A.P., I.M., T.W. and S.L.A.M. designed the study. T.W. and S.L.A.M. developed the tumoroid methodology. T.W., M.S. and S.L.A.M. performed experiments and data acquisition. S.L.A.M. analyzed and visualized data. M.T. analyzed, scored and quantified immunohistochemistry. M.F., C.D., M.S.L., F.M., S.P., C.K.F. and B.G. provided patient biopsies and patient clinical information. R.M., M.S.L., A.D., V.A., M.C.Z., C.K.F., B.G. and M.S. provided administrative, technical or material support. A.P., I.M., T.W. and S.L.A.M. wrote the paper.

1.
Di Domenico
A
,
Wiedmer
T
,
Marinoni
I
,
Perren
A
.
Genetic and epigenetic drivers of neuroendocrine tumours (NET)
.
Endocr Relat Cancer
.
2017
Sep
;
24
(
9
):
R315
34
.
[PubMed]
1351-0088
2.
Scarpa
A
,
Chang
DK
,
Nones
K
,
Corbo
V
,
Patch
AM
,
Bailey
P
, et al;
Australian Pancreatic Cancer Genome Initiative
.
Whole-genome landscape of pancreatic neuroendocrine tumours
.
Nature
.
2017
Mar
;
543
(
7643
):
65
71
.
[PubMed]
0028-0836
3.
Baroni
MG
,
Cavallo
MG
,
Mark
M
,
Monetini
L
,
Stoehrer
B
,
Pozzilli
P
.
Beta-cell gene expression and functional characterisation of the human insulinoma cell line CM
.
J Endocrinol
.
1999
Apr
;
161
(
1
):
59
68
.
[PubMed]
0022-0795
4.
Evers
BM
,
Ishizuka
J
,
Townsend
CM
 Jr
,
Thompson
JC
.
The human carcinoid cell line, BON. A model system for the study of carcinoid tumors
.
Ann N Y Acad Sci
.
1994
Sep
;
733
1 Molecular and
:
393
406
.
[PubMed]
0077-8923
5.
Iguchi
H
,
Hayashi
I
,
Kono
A
.
A somatostatin-secreting cell line established from a human pancreatic islet cell carcinoma (somatostatinoma): release experiment and immunohistochemical study
.
Cancer Res
.
1990
Jun
;
50
(
12
):
3691
3
.
[PubMed]
0008-5472
6.
Kaku
M
,
Nishiyama
T
,
Yagawa
K
,
Abe
M
.
Establishment of a carcinoembryonic antigen-producing cell line from human pancreatic carcinoma
.
Gan
.
1980
Oct
;
71
(
5
):
596
601
.
[PubMed]
0016-450X
7.
Doihara
H
,
Nozawa
K
,
Kojima
R
,
Kawabata-Shoda
E
,
Yokoyama
T
,
Ito
H
.
QGP-1 cells release 5-HT via TRPA1 activation; a model of human enterochromaffin cells
.
Mol Cell Biochem
.
2009
Nov
;
331
(
1-2
):
239
45
.
[PubMed]
0300-8177
8.
Gragnoli
C
.
The CM cell line derived from liver metastasis of malignant human insulinoma is not a valid beta cell model for in vitro studies
.
J Cell Physiol
.
2008
Aug
;
216
(
2
):
569
70
.
[PubMed]
0021-9541
9.
Jonnakuty
C
,
Gragnoli
C
.
Karyotype of the human insulinoma CM cell line—beta cell model in vitro?
J Cell Physiol
.
2007
Dec
;
213
(
3
):
661
2
.
[PubMed]
0021-9541
10.
Lopez
JR
,
Claessen
SM
,
Macville
MV
,
Albrechts
JC
,
Skogseid
B
,
Speel
EJ
.
Spectral karyotypic and comparative genomic analysis of the endocrine pancreatic tumor cell line BON-1
.
Neuroendocrinology
.
2010
;
91
(
2
):
131
41
.
[PubMed]
0028-3835
11.
Siddique
ZL
,
Drozdov
I
,
Floch
J
,
Gustafsson
BI
,
Stunes
K
,
Pfragner
R
, et al
KRJ-I and BON cell lines: defining an appropriate enterochromaffin cell neuroendocrine tumor model
.
Neuroendocrinology
.
2009
;
89
(
4
):
458
70
.
[PubMed]
0028-3835
12.
Boora
GK
,
Kanwar
R
,
Kulkarni
AA
,
Pleticha
J
,
Ames
M
,
Schroth
G
, et al
Exome-level comparison of primary well-differentiated neuroendocrine tumors and their cell lines
.
Cancer Genet
.
2015
Jul-Aug
;
208
(
7-8
):
374
81
.
[PubMed]
2210-7762
13.
Benten
D
,
Behrang
Y
,
Unrau
L
,
Weissmann
V
,
Wolters-Eisfeld
G
,
Burdak-Rothkamm
S
, et al
Establishment of the First Well-differentiated Human Pancreatic Neuroendocrine Tumor Model
.
Mol Cancer Res
.
2018
Mar
;
16
(
3
):
496
507
.
[PubMed]
1541-7786
14.
Sadanandam
A
,
Wullschleger
S
,
Lyssiotis
CA
,
Grötzinger
C
,
Barbi
S
,
Bersani
S
, et al
A Cross-Species Analysis in Pancreatic Neuroendocrine Tumors Reveals Molecular Subtypes with Distinctive Clinical, Metastatic, Developmental, and Metabolic Characteristics
.
Cancer Discov
.
2015
Dec
;
5
(
12
):
1296
313
.
[PubMed]
2159-8274
15.
Falletta
S
,
Partelli
S
,
Rubini
C
,
Nann
D
,
Doria
A
,
Marinoni
I
, et al
mTOR inhibitors response and mTOR pathway in pancreatic neuroendocrine tumors
.
Endocr Relat Cancer
.
2016
Nov
;
23
(
11
):
883
91
.
[PubMed]
1351-0088
16.
Mohamed
A
,
Romano
D
,
Saveanu
A
,
Roche
C
,
Albertelli
M
,
Barbieri
F
, et al
Anti-proliferative and anti-secretory effects of everolimus on human pancreatic neuroendocrine tumors primary cultures: is there any benefit from combination with somatostatin analogs?
Oncotarget
.
2017
Jun
;
8
(
25
):
41044
63
.
[PubMed]
1949-2553
17.
Mohamed
A
,
Blanchard
MP
,
Albertelli
M
,
Barbieri
F
,
Brue
T
,
Niccoli
P
, et al
Pasireotide and octreotide antiproliferative effects and sst2 trafficking in human pancreatic neuroendocrine tumor cultures
.
Endocr Relat Cancer
.
2014
Oct
;
21
(
5
):
691
704
.
[PubMed]
1351-0088
18.
Wiedmer
T
,
Blank
A
,
Pantasis
S
,
Normand
L
,
Bill
R
,
Krebs
P
, et al
Autophagy Inhibition Improves Sunitinib Efficacy in Pancreatic Neuroendocrine Tumors via a Lysosome-dependent Mechanism
.
Mol Cancer Ther
.
2017
Nov
;
16
(
11
):
2502
15
.
[PubMed]
1535-7163
19.
Aboulkheyr Es
H
,
Montazeri
L
,
Aref
AR
,
Vosough
M
,
Baharvand
H
.
Personalized Cancer Medicine: An Organoid Approach
.
Trends Biotechnol
.
2018
Apr
;
36
(
4
):
358
71
.
[PubMed]
0167-7799
20.
Broguiere
N
,
Isenmann
L
,
Hirt
C
,
Ringel
T
,
Placzek
S
,
Cavalli
E
, et al
Growth of Epithelial Organoids in a Defined Hydrogel
.
Adv Mater
.
2018
Oct
;
30
(
43
):
e1801621
.
[PubMed]
0935-9648
21.
Nath
S
,
Devi
GR
.
Three-dimensional culture systems in cancer research: focus on tumor spheroid model
.
Pharmacol Ther
.
2016
Jul
;
163
:
94
108
.
[PubMed]
0163-7258
22.
Tuveson
D
,
Clevers
H
.
Cancer modeling meets human organoid technology
.
Science
.
2019
Jun
;
364
(
6444
):
952
5
.
[PubMed]
0036-8075
23.
Xu
H
,
Lyu
X
,
Yi
M
,
Zhao
W
,
Song
Y
,
Wu
K
.
Organoid technology and applications in cancer research
.
J Hematol Oncol
.
2018
Sep
;
11
(
1
):
116
.
[PubMed]
1756-8722
24.
Huang
L
,
Holtzinger
A
,
Jagan
I
,
BeGora
M
,
Lohse
I
,
Ngai
N
, et al
Ductal pancreatic cancer modeling and drug screening using human pluripotent stem cell- and patient-derived tumor organoids
.
Nat Med
.
2015
Nov
;
21
(
11
):
1364
71
.
[PubMed]
1078-8956
25.
Pampaloni
F
,
Reynaud
EG
,
Stelzer
EH
.
The third dimension bridges the gap between cell culture and live tissue
.
Nat Rev Mol Cell Biol
.
2007
Oct
;
8
(
10
):
839
45
.
[PubMed]
1471-0072
26.
Bresciani
G
,
Hofland
LJ
,
Dogan
F
,
Giamas
G
,
Gagliano
T
,
Zatelli
MC
.
Evaluation of Spheroid 3D Culture Methods to Study a Pancreatic Neuroendocrine Neoplasm Cell Line
.
Front Endocrinol (Lausanne)
.
2019
Oct
;
10
:
682
.
[PubMed]
1664-2392
27.
Ekert
JE
,
Johnson
K
,
Strake
B
,
Pardinas
J
,
Jarantow
S
,
Perkinson
R
, et al
Three-dimensional lung tumor microenvironment modulates therapeutic compound responsiveness in vitro—implication for drug development
.
PLoS One
.
2014
Mar
;
9
(
3
):
e92248
.
[PubMed]
1932-6203
28.
Luca
AC
,
Mersch
S
,
Deenen
R
,
Schmidt
S
,
Messner
I
,
Schäfer
KL
, et al
Impact of the 3D microenvironment on phenotype, gene expression, and EGFR inhibition of colorectal cancer cell lines
.
PLoS One
.
2013
;
8
(
3
):
e59689
.
[PubMed]
1932-6203
29.
Riedl
A
,
Schlederer
M
,
Pudelko
K
,
Stadler
M
,
Walter
S
,
Unterleuthner
D
, et al
Comparison of cancer cells in 2D vs 3D culture reveals differences in AKT-mTOR-S6K signaling and drug responses
.
J Cell Sci
.
2017
Jan
;
130
(
1
):
203
18
.
[PubMed]
0021-9533
30.
McMillin
DW
,
Negri
JM
,
Mitsiades
CS
.
The role of tumour-stromal interactions in modifying drug response: challenges and opportunities
.
Nat Rev Drug Discov
.
2013
Mar
;
12
(
3
):
217
28
.
[PubMed]
1474-1776
31.
Weiswald
LB
,
Bellet
D
,
Dangles-Marie
V
.
Spherical cancer models in tumor biology
.
Neoplasia
.
2015
Jan
;
17
(
1
):
1
15
.
[PubMed]
1522-8002
32.
Pauli
C
,
Hopkins
BD
,
Prandi
D
,
Shaw
R
,
Fedrizzi
T
,
Sboner
A
, et al
Personalized In Vitro and In Vivo Cancer Models to Guide Precision Medicine
.
Cancer Discov
.
2017
May
;
7
(
5
):
462
77
.
[PubMed]
2159-8274
33.
Tiriac
H
,
Belleau
P
,
Engle
DD
,
Plenker
D
,
Deschênes
A
,
Somerville
TD
, et al
Organoid Profiling Identifies Common Responders to Chemotherapy in Pancreatic Cancer
.
Cancer Discov
.
2018
Sep
;
8
(
9
):
1112
29
.
[PubMed]
2159-8274
34.
van de Wetering
M
,
Francies
HE
,
Francis
JM
,
Bounova
G
,
Iorio
F
,
Pronk
A
, et al
Prospective derivation of a living organoid biobank of colorectal cancer patients
.
Cell
.
2015
May
;
161
(
4
):
933
45
.
[PubMed]
0092-8674
35.
Bankhead
P
,
Loughrey
MB
,
Fernández
JA
,
Dombrowski
Y
,
McArt
DG
,
Dunne
PD
, et al
QuPath: open source software for digital pathology image analysis
.
Sci Rep
.
2017
Dec
;
7
(
1
):
16878
.
[PubMed]
2045-2322
36.
Sebaugh
JL
.
Guidelines for accurate EC50/IC50 estimation
.
Pharm Stat
.
2011
Mar-Apr
;
10
(
2
):
128
34
.
[PubMed]
1539-1604
37.
Garnett
MJ
,
Edelman
EJ
,
Heidorn
SJ
,
Greenman
CD
,
Dastur
A
,
Lau
KW
, et al
Systematic identification of genomic markers of drug sensitivity in cancer cells
.
Nature
.
2012
Mar
;
483
(
7391
):
570
5
.
[PubMed]
0028-0836
38.
Wilkerson
MD
,
Hayes
DN
.
ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking
.
Bioinformatics
.
2010
Jun
;
26
(
12
):
1572
3
.
[PubMed]
1367-4803
39.
UniProt Consortium
.
UniProt: a worldwide hub of protein knowledge
.
Nucleic Acids Res
.
2019
Jan
;
47
D1
:
D506
15
.
[PubMed]
0305-1048
40.
Hilfenhaus
G
,
Göhrig
A
,
Pape
UF
,
Neumann
T
,
Jann
H
,
Zdunek
D
, et al
Placental growth factor supports neuroendocrine tumor growth and predicts disease prognosis in patients
.
Endocr Relat Cancer
.
2013
May
;
20
(
3
):
305
19
.
[PubMed]
1351-0088
41.
Hughes
CS
,
Postovit
LM
,
Lajoie
GA
.
Matrigel: a complex protein mixture required for optimal growth of cell culture
.
Proteomics
.
2010
May
;
10
(
9
):
1886
90
.
[PubMed]
1615-9853
42.
Kleinman
HK
,
Martin
GR
.
Matrigel: basement membrane matrix with biological activity
.
Semin Cancer Biol
.
2005
Oct
;
15
(
5
):
378
86
.
[PubMed]
1044-579X
43.
Kutschka Ingo
, Chen Ian Y., Kofidis Theo, Arai Takayasu, von Degenfeld Georges, Sheikh Ahmad Y., et al. Collagen Matrices Enhance Survival of Transplanted Cardiomyoblasts and Contribute to Functional Improvement of Ischemic Rat Hearts. Circulation.
2006
Jul;114(1_supplement):I–167.
44.
Poincloux
R
,
Collin
O
,
Lizárraga
F
,
Romao
M
,
Debray
M
,
Piel
M
, et al
Contractility of the cell rear drives invasion of breast tumor cells in 3D Matrigel
.
Proc Natl Acad Sci USA
.
2011
Feb
;
108
(
5
):
1943
8
.
[PubMed]
0027-8424
45.
Zheng
W
,
Zhang
W
,
Jiang
X
.
Biomimetic Collagen Nanofibrous Materials for Bone Tissue Engineering
.
Adv Eng Mater
.
2010
Sep
;
12
(
9
):
B451
66
. 1438-1656
46.
Montesano
R
,
Mouron
P
,
Amherdt
M
,
Orci
L
.
Collagen matrix promotes reorganization of pancreatic endocrine cell monolayers into islet-like organoids
.
J Cell Biol
.
1983
Sep
;
97
(
3
):
935
9
.
[PubMed]
0021-9525
47.
Kaderli
RM
,
Spanjol
M
,
Kollár
A
,
Bütikofer
L
,
Gloy
V
,
Dumont
RA
, et al
Therapeutic Options for Neuroendocrine Tumors: A Systematic Review and Network Meta-analysis
.
JAMA Oncol
.
2019
Apr
;
5
(
4
):
480
9
.
[PubMed]
2374-2437
48.
Bolck
HA
,
Pauli
C
,
Göbel
E
,
Mühlbauer
K
,
Dettwiler
S
,
Moch
H
, et al
Cancer Sample Biobanking at the Next Level: Combining Tissue With Living Cell Repositories to Promote Precision Medicine
.
Front Cell Dev Biol
.
2019
Oct
;
7
:
246
.
[PubMed]
2296-634X
49.
Alkema
NG
,
Tomar
T
,
Duiker
EW
,
Jan Meersma
G
,
Klip
H
,
van der Zee
AG
, et al
Biobanking of patient and patient-derived xenograft ovarian tumour tissue: efficient preservation with low and high fetal calf serum based methods
.
Sci Rep
.
2015
Oct
;
5
(
1
):
14495
.
[PubMed]
2045-2322
50.
Sato
T
,
Vries
RG
,
Snippert
HJ
,
van de Wetering
M
,
Barker
N
,
Stange
DE
, et al
Single Lgr5 stem cells build crypt-villus structures in vitro without a mesenchymal niche
.
Nature
.
2009
May
;
459
(
7244
):
262
5
.
[PubMed]
0028-0836
51.
Sato
T
,
Stange
DE
,
Ferrante
M
,
Vries
RG
,
Van Es
JH
,
Van den Brink
S
, et al
Long-term expansion of epithelial organoids from human colon, adenoma, adenocarcinoma, and Barrett’s epithelium
.
Gastroenterology
.
2011
Nov
;
141
(
5
):
1762
72
.
[PubMed]
0016-5085
52.
Fujii
M
,
Shimokawa
M
,
Date
S
,
Takano
A
,
Matano
M
,
Nanki
K
, et al
A Colorectal Tumor Organoid Library Demonstrates Progressive Loss of Niche Factor Requirements during Tumorigenesis
.
Cell Stem Cell
.
2016
Jun
;
18
(
6
):
827
38
.
[PubMed]
1934-5909
53.
Matano
M
,
Date
S
,
Shimokawa
M
,
Takano
A
,
Fujii
M
,
Ohta
Y
, et al
Modeling colorectal cancer using CRISPR-Cas9-mediated engineering of human intestinal organoids
.
Nat Med
.
2015
Mar
;
21
(
3
):
256
62
.
[PubMed]
1078-8956
54.
Drost
J
,
van Boxtel
R
,
Blokzijl
F
,
Mizutani
T
,
Sasaki
N
,
Sasselli
V
, et al
Use of CRISPR-modified human stem cell organoids to study the origin of mutational signatures in cancer
.
Science
.
2017
Oct
;
358
(
6360
):
234
8
.
[PubMed]
0036-8075
55.
Weeber
F
,
Ooft
SN
,
Dijkstra
KK
,
Voest
EE
.
Tumor Organoids as a Pre-clinical Cancer Model for Drug Discovery
.
Cell Chem Biol
.
2017
Sep
;
24
(
9
):
1092
100
.
[PubMed]
2451-9456
56.
Kojima
N
.
In vitro reconstitution of pancreatic islets
.
Organogenesis
.
2014
Apr-Jun
;
10
(
2
):
225
30
.
[PubMed]
1547-6278
57.
Koumarianou
A
,
Kaltsas
G
,
Kulke
MH
,
Oberg
K
,
Strosberg
JR
,
Spada
F
, et al
Temozolomide in Advanced Neuroendocrine Neoplasms: Pharmacological and Clinical Aspects
.
Neuroendocrinology
.
2015
;
101
(
4
):
274
88
.
[PubMed]
0028-3835
58.
Fu
D
,
Calvo
JA
,
Samson
LD
.
Balancing repair and tolerance of DNA damage caused by alkylating agents
.
Nat Rev Cancer
.
2012
Jan
;
12
(
2
):
104
20
.
[PubMed]
1474-175X
59.
Lohse
I
,
Al-Ali
H
,
Volmar
CH
,
D Alvarez Trotta
A
,
Brothers
SP
,
Capobianco
AJ
, et al
Ex vivo drug sensitivity testing as a means for drug repurposing in esophageal adenocarcinoma
.
PLoS One
.
2018
Sep
;
13
(
9
):
e0203173
.
[PubMed]
1932-6203
60.
Majumder
B
,
Baraneedharan
U
,
Thiyagarajan
S
,
Radhakrishnan
P
,
Narasimhan
H
,
Dhandapani
M
, et al
Predicting clinical response to anticancer drugs using an ex vivo platform that captures tumour heterogeneity
.
Nat Commun
.
2015
Feb
;
6
(
1
):
6169
.
[PubMed]
2041-1723
61.
Naipal
KA
,
Verkaik
NS
,
Ameziane
N
,
van Deurzen
CH
,
Ter Brugge
P
,
Meijers
M
, et al
Functional ex vivo assay to select homologous recombination-deficient breast tumors for PARP inhibitor treatment
.
Clin Cancer Res
.
2014
Sep
;
20
(
18
):
4816
26
.
[PubMed]
1078-0432
62.
Garralda
E
,
Paz
K
,
López-Casas
PP
,
Jones
S
,
Katz
A
,
Kann
LM
, et al
Integrated Next Generation Sequencing and Avatar Mouse Models for Personalized Cancer Treatment.
Clin Cancer Res. 2014 Mar;clincanres.3047.
2013
.
63.
Weroha
SJ
,
Becker
MA
,
Enderica-Gonzalez
S
,
Harrington
SC
,
Oberg
AL
,
Maurer
MJ
, et al
Tumorgrafts as in vivo surrogates for women with ovarian cancer.
Clin Cancer Res. 2014 Jan;clincanres.2611.
2013
.

Simon Leonhard April-Monn and Tabea Wiedmer contributed equally to this work and shared first authorship. Aurel Perren and Ilaria Marinoni shared last authorship.

Copyright / Drug Dosage / Disclaimer
Copyright: All rights reserved. No part of this publication may be translated into other languages, reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording, microcopying, or by any information storage and retrieval system, without permission in writing from the publisher.
Drug Dosage: The authors and the publisher have exerted every effort to ensure that drug selection and dosage set forth in this text are in accord with current recommendations and practice at the time of publication. However, in view of ongoing research, changes in government regulations, and the constant flow of information relating to drug therapy and drug reactions, the reader is urged to check the package insert for each drug for any changes in indications and dosage and for added warnings and precautions. This is particularly important when the recommended agent is a new and/or infrequently employed drug.
Disclaimer: The statements, opinions and data contained in this publication are solely those of the individual authors and contributors and not of the publishers and the editor(s). The appearance of advertisements or/and product references in the publication is not a warranty, endorsement, or approval of the products or services advertised or of their effectiveness, quality or safety. The publisher and the editor(s) disclaim responsibility for any injury to persons or property resulting from any ideas, methods, instructions or products referred to in the content or advertisements.