Metastatic behaviour of primary human tumours in a zebrafish xenotransplantation model

Zebrafish (Danio rerio) (Tuebingen line, alb strain (Albinos) and Tg(fli1:eGFP) were handled in compliance with local animal care regulations and standard protocols of the Netherlands and Germany. Fish were kept at 28°C in aquaria with day/night light cycles (10 h dark versus 14 h light periods). The developing embryos were kept in an incubator at constant temperatures. The cloche (clo) mutant line has been previously described [17]. Heterozygous fish (clo^-/+) are kept and bred under normal conditions. 25% of offspring will consist of homozygous clo^-/- mutants which lack functional vasculature and circulation 75% will be siblings with no phenotype. Lack of circulation, an enlarged pericardium and curvature of the tail (at a later time point) are hallmarks of the cloche phenotype.

EpRas cells were cultured at 37°C in DMEM high glucose containing L-glutamine, 4% FCS and 1:100 Pen/Strep (GIBCO, Invitrogen). PaTu8988T and PaTu8898S cells were cultured in DMEM high glucose, with 10% FCS and 1:100 Pen/Strep. The EpRas cells were treated with recombinant human TGF-β1(RD systems) at a final concentration of 2 ng/ml. To induce epithelial to mesenchymal transition (EMT), cells were seeded at 70% confluency in 6-well plates and media containing TGF-β1 (2 ng/ml final concentration) was added and replaced every other day for 10 days. After this period, cells were ready for injection.

Cells were stained with either CM-Dil (red fluorescence) or DiO (green fluorescence) (Vybrant, Invitrogen). Cells were seeded in 6-well plates, grown to confluency trypsinized (without EDTA for EpRas cells or with EDTA for all other cells used). Subsequently, cells were washed with 67% DPBS (GIBCO, Invitrogen), transferred to 1.5 ml Eppendorf tubes and centrifuged 5 min, at 1500 rpm. Cells were re-suspended in DPBS containing either CM-Dil (4 ng/ul final concentration) or DiO (200 μM final concentration). Cells stained with CM-Dil were incubated 4 min at 37°C and then 15 min at 4°C. Cells stained with DiO were incubated 20 min at 37°C. After this period cells were centrifuged 5 min at 1500 rpm, the supernatant discarded and cells re-suspended in 100% FCS, centrifuged again and washed 2 times with 67% DPBS. Cells were suspended in 67% DPBS for injection into the embryos. 2 dpf zebrafish embryos were dechorionated and anesthesized with tricaine (Sigma). Using a manual injector (Eppendorf; Injectman NI2), the cell suspension was loaded into an injection needle (15 μm internal- and 18 μm external-diameter). Cells were now injected in 2 dpf albino or Tg(fli1:eGFP) zebrafish embryos. After injection, embryos were incubated for 1 h at 31°C and checked for cell presence at 2 hpt. Fish with fluorescent cells outside the implantation area at 2 hpt were excluded from further analysis. All other fish were incubated at 35°C for the following days.

Human material from surgical resection specimens was obtained at the Universitätsklinikum Greifswald according to local ethical guidelines and after obtaining informed patient consent. Tumour tissue and control tissue were cut into very small pieces using a scalp blade. A piece of tissue was then transferred to a 2 ml Eppendorf tube, washed with 67% DPBS and stained with 1:500 CM-Dil. The tissue was incubated for 6 min at 37°C and 20 min at 4°C. Washing procedures were the same as mentioned above for the cells. Before transplantation small pieces of stained tissue were further disaggregated using Dumont forceps (No.5) into a relative size of 1/5 to 1/2 the size of the yolk. Tissue pieces with the correct size were transferred to agarose plates in which the embryos were laying, ready for transplantation. For tumour and control transplantations, a glass transplantation needle was used to transfer the tissue into the yolk. With the glass transplantation needle a piece of tissue was picked up, put on top of the yolk and then pushed inside. The yolk usually sealed itself and in the majority of embryos, the tumour remained in the yolk. After transplantation, embryos were incubated for 1 h at 31°C, then embryos were checked for presence of tissue and incubated at 35°C for the following days.

Tissue samples were cut in very small pieces using a scalp blade. Cut tissue pieces were then transferred to 6 ml glass containers with 3 ml isolation media (180 ml DMEM high glucose, 20 ml 100 mM HEPES, 46 ml 5% BSA) and Collagenase (Invitrogen) (50 μl of a 6 mg/ml stock solution for each 12 ml of isolation media). Tissue was incubated in a water bath for 15 min at 37°C. The supernatant was decanted and tissue pieces were cut further into smaller pieces using a scalp blade. Tissue pieces were again incubated 15 min at 37°C in 3 ml isolation media with collagenase. Afterwards tissue pieces were transferred to 15 ml falcon tubes and cells were dissolved by pipetting up and down through serial cut blue pipette tips (5 different diameters). The cell suspension was now filtered through 2 sheets of gaze, into 2 ml Eppendorf tubes and centrifuged 5 min at 1500 rpm. The supernatant was discarded and cells re-suspended in isolation media. The described procedure was then repeated once. For injections, cells were stained with either CM-Dil or DiO and injected into 2 dpf zebrafish embryos, as described above.

Pancreatic cancer cell lines PaTu-S and PaTu-T were lysed in iced Triton-X-100 lysis buffer (0.1%) containing protease inhibitors (1 ml/mg tissue, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 0,01 M sodiumpyrophosphate, 0,1 M sodiumfluoride, 1 mM dihydrogenperoxide, 1 mM L-phenyl-methyl-sulfonyl-fluoride [PMSF] and 0,02% soybean-trypsininhibitor). Protein concentration was determined by a modified Bradford-assay (Bio Rad Laboratories, München, Germany) and equal amounts of protein were used in subsequent experiments. Cell lysates were separated by SDS-PAGE on a 7.5% polyacrylamide gel in a discontinuous buffer system and gels were blotted on nitrocellulose membranes (Hybond C, GE Healthcare Europe GmbH). After overnight blocking in NET-gelatine (10 mM Tris/HCl pH 8.0, 0.15 mM NaCl, 0.05% TWEEN 20, 0.2% gelatine) immunoblot analysis was performed followed by enhanced chemoluminescence detection (GE Healthcare Europe GmbH) using horseradish peroxidase coupled sheep anti-mouse IgG or goat anti-rabbit IgG GE Healthcare Europe GmbH). Monoclonal E-cadherin antibody (Clone 36), directed against the carboxy-terminus, was purchased from Transduction Laboratories (San Diego, CA, USA) as well as antibodies against α-, β-, and γ-Catenin. A polyclonal G3PDH antibody was purchased from Biozol (Eching, Germany).

PaTu-S and PaTu-T cells grown on glass coverslips for 24–48 h were washed 3 times with PBS, fixed for 15 min in 4% paraformaldehyde and permeabilised in 0.1% Triton- X-100 for 5 min. Blocking of unspecific binding was achieved by a 1 h incubation in 10% Aurion BSA-c (Aurion, Waageningen, The Netherlands). Following a primary antibody incubation over night (dilutions 1:100) and subsequent PBS washing steps detection was performed using dichlortriacinyl aminofluoresceine (DTAF) or Cy3-coupled sheep IgG (dilutions 1:200). Nuclei were stained by a 30 sec. incubation with DAPI (1:10000 in PBS). After a final washing step in PBS cell were mounted in Vectashield (Vector Labs, Burlingame, CA, USA). Microscopic Images were taken using an AxioCam digital microscope camera on a Zeiss Axiophot microscope.

The scratch-assay was performed as previously described by Liang et al. [18]. Cells were grown to confluency in 6-well dishes and mitomycin C was added at 10 μM for 2 h. Then the cell monolayer was scraped in a straight line with a 200 μl pipette tip. Pictures of the scratch were taken under an invert Olympus microscope at 0 h, 12 h and 24 h.

Transversal sections at 4 μM thickness were prepared as described before [19]. Coupes were directly imaged with fluorescence microscopy or differential interference (DIC) microscopy. After fluorescent pictures were taken, Hematoxylin/Eosin (HE) staining was performed as described earlier [20].

Zebrafish embryos at 2 dpt were fixed overnight in 4%PFA in PBS at 4°C. After fixing, embryos were washed with BSAc 0.1%-TritonX100 1% in PBS (blocking buffer; 3 × 10 minutes). Subsequently, the embryos were incubated for 2 hrs in blocking buffer at RT. Incubation with the primary antibody (Mouse anti-Proliferating Cell Nuclear Antigen, PCNA, from Zymed Laboratories, 1:100) was done overnight at 4°C. After washing (3 × 10 minutes) with blocking buffer, embryos were incubated with the secondary antibody (Fluorescein (DTAF)-conjugated AffiniPure Goat anti-Mouse IgG, Jackson Immuno Research Laboratories, Inc., 1:100) for 1 hr. and washed afterwards with blocking buffer (3 × 10 minutes). Embryos were mounted in 3% methylcellulose to orient them properly for imaging. Imaging was done with confocal scanning laser microscopy (Biorad 1024ES; Software: Biorad Laser sharp 2000).

Confocal pictures were taken either with the Biorad Confocal microscope 1024ES (Zeiss microscope) combined with Krypton/Argon laser, or the dual laser scanning confocal microscope Leica DM IRBE (Leica) or with the Nikon TE300 confocal microscope and a coherent Innova 70C laser (Chromaphor, Duisburg, Germany). Pictures were further taken by DIC microscopy using the Axioplan 2 microscope with an AxioCam MR5 camera (Carl Zeiss). Further, fluorescent stereomicroscope pictures were taken with the Leica DFC 420C camera attached to a Leica MZ16FA microscope. Two hours post implantation the embryos were anesthetized with tricaine and positioned laterally, with the site of the implantation to the top, on 3% methylcelulose, on a slide with depression. Each time two rows of twenty embryos were screened. Two hours post implantation every embryo that showed cells outside the area of implantation was discarded and not considered for the experiment.

Mouse mammary epithelial cells (EpH4) transformed with oncogenic Ras (EpRas) have been used to establish a mouse tumourigenesis model over a decade ago [21]. In these EpRas cells, TGF-β signalling causes epithelial to mesenchymal transition (EMT) which transforms cells to a highly invasive phenotype and enables distant metastasis formation when transplanted into nude mice [22]. Initially, we evaluated metastasis formation using this well-characterized system in the zebrafish cell xenograft model. We transplanted fluorescently labelled EpRas cells into the yolk sac of 2 day old zebrafish embryos to study metastatic behaviour in vivo. EpRas cells that had been stimulated with TGF-β for 10 days prior to injection, showed metastatic behaviour in the zebrafish, comparable to results previously reported in mice [21‐23]. Following EMT the cells invaded embryonic tissue, entered the circulation and homed in at distant tissues and organs. EpRas TGFβ treated cells were found in blood islands, brain, caudal fin, caudal vein, gill arches, heart, intestine, liver, mandible, optic cup (eye), otic cup, pericardium, somites, swim bladder. However, they had a tendency to invade and home in to muscle tissue, head structures, caudal fin and blood islands (Fig. 1 and Additional File 1). To a lesser extent, we observed invasion of these cells into the liver or other organs of the gastrointestinal tract. In contrast, unstimulated EpRas control cells remained at the place of injection in the yolk and neither invaded the developing zebrafish nor did they enter blood circulation (Fig. 1 and Additional file 2). In three independent experiments the average percentage of migrating cells observed for the EpRas TGF-β-stimulated cells was 46.6% (SD +/- 2.0; p-value < 0.001) compared to 0.5% (SD +/- 0.7; p-value < 0.001) for the parental EpRas cells (see Additional file 1). Furthermore, the TGF-β stimulated EpRas cells formed tumour cell masses in the developing zebrafish (Fig. 1), which resemble the formation of metastases in nude mice[23]. In the zebrafish, cells begin to invade the embryo already several hours after injection (on average 4 hours post injection)(see Additional File 3) and tumour cell masses are visible as early as 3 days post implantation (dpi).

For optimal visualization, we used the transgenic zebrafish line, Tg(fli1:eGFP)[11, 24], which expresses GFP under the fli1 promotor (an early endothelial marker) and therefore exhibits a green fluorescent vasculature [11, 24]. In a time lapse movie (see Additional file 4; rate: 1 frame/minute) we show an example of fluorescently labelled EpRas TGF-β cells (3 dpi) which have invaded the zebrafish body, have translocated into the vasculature and have colonized at distant sites in the zebrafish larvae (5 dpf). Some cells are visible in the blood stream whereas others have extravasated from the vasculature. Evaluation of metastasis formation in the zebrafish model is therefore significantly faster than in currently used mouse models, where it may take several weeks until metastases become detectable [23]. The sensitivity of the zebrafish tumour xenograft model further allows observation of individual cells and their daughter cells in vivo.

We also compared the two established human pancreatic tumour cell lines, PaTu8988-S and PaTu8988-T [25] (referred to herein as PaTu-S and PaTu-T) in their invasive and metastatic potential in a single zebrafish. Both sister cell lines originate from liver metastases of the same human pancreatic adenocarcinoma [25]. E-Cadherin expression in PaTu-S cells (Fig. 2A and 2B(a)) correlates with the maintenance of functional cell-cell contacts (Fig. 2B) and a reduced tendency of cells to migrate (Fig. 2C)[26, 27]. Whereas PaTu-S cells show localization of E-cadherin/β-catenin complexes at the plasmamembrane (Fig. 2B(e)), PaTu-T cells lack E-Cadherin expression (Fig. 2A and 2B(b)) and β-catenin is mainly localized in the cytoplasm (Fig. 2B(d). This observation is paralleled by enhanced migratory capabilities of PaTu-T cells compared to PaTu-S cells, which we confirmed in an in vitro migration assay ('scratch assay' [28] (Fig. 2C and Additional file 5).

We then labelled PaTu-S cells with green fluorescence and PaTu-T cells with red fluorescence. When implanted successively in the yolk sac of the same zebrafish embryo, green PaTu-S cells remained in the yolk, whereas red PaTu-T cells displayed invasion and metastatic behaviour (Fig. 2D and Additional file 1). Similar results were observed when cells were mixed prior to injections (data not shown).

PaTuT cells were found in the brain, caudal vein, gill arches, gut, heart, intestine, liver, operculum, pericardium, somites, swim bladder. They showed a tendency to invade and home in to organs of the gastrointestinal tract. Micrometastasis was often observed in organs such as the liver, the gut and the intestine. Invasion and homing in of these cells into muscle tissue was observed to a lesser extent. This behaviour was qualitative different from what we observed for TGF-β treated EpRas cells. We further tested the metastatic behaviour of PaTu-T cells in homozygous cloche mutants (cloche^-/^-), which lack a functional vasculature and circulation ([17] Additional File 6). In contrast to control zebrafish, no metastatic behaviour was observed in the cloche^-/^- fish, indicating that the observed invasion/migration of PaTu-T cells indeed involves metastasis formation through the vascular system. Zebrafish were followed until three days post injections (Fig. 2E; see Additional file 1 and see Additional file 7).

We then pursued our primary goal to employ the zebrafish also as a simple, fast and effective test system for metastasis formation of primary human tumours. After informed patient consent small fragments of tumour explants from pancreas, colon and stomach carcinoma, as well as tumour-free areas from the same resection specimen were fluorescently labelled with CM-DIL and directly xenotransplanted into the yolk sac of zebrafish embryos. Tumour and non-tumour control cells were followed live by laser scanning confocal microscopy. Tumour cells started to invade the embryo on average 12 hours post injection and micrometastasis formation was visible as early as 24 h post injection. In parallel, we also investigated and compared the invasive and metastatic behaviour of tumour cells that had been dissociated from primary human tumours by collagenase digestion prior to transplantation.

In total, pancreatic tumours of four different patients were analysed. Three had carcinomas of the pancreas head and one had an adenocarcinoma of the ampulla vateri with infiltration of the pancreas (cancer grades and pTMN stages for all tumours are shown in table 1). On average 59.8% (SD +/- 5.2) of transplanted pancreatic tumour fragments showed invasion and migration in the developing zebrafish (table 1 and Fig. 3). Evaluation criteria for invasion and migration were that at least 5 cells had to be identified outside of the yolk and detectable within the developing zebrafish (table 1). Development of micrometastases was assessed by the presence of daughter cells at 3 dpt. The results are listed in table 1 and in Fig 4 examples of micrometastases in different tissues detected in sections of 5 dpf zebrafish are shown at high resolution (Fig. 4B, C, D, G, H, K and 4L). An example showing proliferating pancreatic tumour cells and the initial formation of a micrometastasis is given in Fig 5(E). Cell division is further indicated by PCNA immunostaining of invasive tumour cells in the zebrafish embryo (see Additional file 8). After 5 dpf embryos fall under strict local animal experiment regulations, therefore most embryos were not followed for longer periods. It is likely that the number of micrometastases would still increase over time.

Invasive cells were found for pancreatic tumours in blood islands, caudal fin, caudal vein, gut, heart, hindbrain, intestine, liver, mesonephric duct, mesonephric tubule, mandible, operculum, pericardium, somites, swim bladder, for the colon tumour in caudal vein, gut, heart, intestine, liver, pericardium; and for the stomach tumours in the caudal vein, gill arches, heart, intestine, liver, mandible, otic cup, pericardium;

Similar to preferences observed for PaTu-T cells, tumour cells of implanted gastrointestinal tumor tissue fragments had a tendency to invade and home in to organs of the gastrointestinal tract. Micrometastasis were often observed in the liver, the gut and the intestine. Homing in of these cells in muscle tissue and the formation of micrometastasis was rarely observed. Although we observed qualitative differences for the preferential homing in of the two cell lines tested (PaTu-T and TGF-β treated EpRAS cells), a future study, with a larger cohort of tumour specimen and tumour types is necessary to determine tumour specific preferences.

Fig 3 shows examples of fish embryos directly after transplantation (A, B) and at 1- and 3-dpt (C-G). Cell invasion and micrometastasis formation of pancreatic tumour cells is clearly detectable 24 h after transplantation (E). Similar results were also obtained for transplanted tissue fragments of a colon adenocarcinoma (43.9% invasion and migration) and two moderately differentiated adenocarcinomas of the stomach (average 53.5%, SD +/- 1.2) (p-values in all tumour experiments were < 0.001). As a control for the tumours we used colon and gastric mucosal fragments or peritumoural, non-transformed tissue of the respective tissue explants from the same patients. In the pancreas the non-transformed tissue controls mostly showed histological manifestations of chronic pancreatitis and only one was considered as having a normal pancreatic histology.

Histological sections of control pancreas and of pancreatic cancer tissue of human patients are shown in Fig 3(J, K). In all control transplantations of chronic pancreatitis specimen and of fragments of normal pancreas, colon and stomach tissue cellular invasion and migration was never observed (see Additional file 1 and Fig. 3D, F). In addition, we also tested the metastatic behaviour of a benign tumour. Tissue fragments of adenomateous colonic polyps (0,4 cm and 1 cm) were investigated, which had not yet invaded through the lamina muscularis mucosae. No metastasis was observed for either of them in the zebrafish embryo (Additional file 1). Comparable results were seen when cells were dissociated from tumour or control tissue samples prior to their injection into the zebrafish. All four primary pancreatic tumour cells showed cellular invasion and migration with an average of 48.8% (SD +/- 9.0) (see Additional file 1). In the case of dissociated primary colon and stomach tumour cells 44.4% and 35.3% of cell injections, respectively, resulted in cellular invasion and migration (see Additional file 1).

Xenotransplantation experiments of tumour fragments as well as the injection of isolated primary tumour cells allowed to discriminate between non-invasive chronic pancreatitis and infiltrating pancreatic adenocarcinoma. In real-time, we show an example of pancreatic tumour cells in the zebrafish 1 day after a pancreatic cancer fragment was transplanted (see Additional file 9). This movie shows circulating tumour cells and tumour cells which have extravasated into the perivascular tissue of a 3 day old zebrafish. A moving primary human tumour cell passing through the caudal vein and into an intersegmental vessel is also visible. In a supplemental movie we show how a pancreatic tumour cell is slowly traversing through the caudal vein of a 3 dpt Tg(fli1:eGFP) zebrafish (see Additional file 10). In Fig 4 histological sections of zebrafish embryos transplanted with a pancreatic human tumour are shown at 3 dpt. Examples of micrometastases in the liver (B, C and D) and near the mesonephric duct (G and H) are shown at higher magnifications. A cell mass is further shown in K and L.

Some xenograft models that have been established in the mouse involve the orthotopic transplantation of specific human tumours and tumour cells [29, 30]. Surgical orthotopic implantations (SOI) of tumour cells or of resected primary tumour fragments into immune deficient mice have proven useful for studying their growth and metastatic potential [31].

Here we transplanted freshly dissociated primary pancreatic tumour cells and normal pancreatic control cells into the liver of zebrafish larvae. For these experiments, we used the Tg(fli1:eGFP) transgenic zebrafish line (with the advantage of a fluorescent vasculature), which allows an exact localization and injection into the highly vascularised liver.

Primary cells of control pancreatic tissue, when transplanted into the liver, remained at the site of the injection and did not invade the developing zebrafish nor did they enter the blood circulation (Fig. 5). In contrast, primary tumour cells invaded the neighbouring tissue, entered the circulation and migrated and homed in at distant tissues and organs (Fig. 5).

Furthermore we investigated in our zebrafish xenotransplantation model the effects of protease inhibitors on the invasiveness of implanted tumour cells and tumour fragments. Two different protease inhibitors were able to inhibit the invasiveness of tumour cells and of a primary pancreatic tumour (see Additional file 1).