Endothelial cell-initiated extravasation of cancer cells visualized in zebrafish

Cell lines

DsRed2 (referred to as RFP) expressing HeLa cells (obtained from Anticancer) were cultured in RPMI-1640 supplemented with 10% FBS, 2 mM l-glutamine (Invitrogen), 1% Penicillin-Streptomycin (Invitrogen). The cells were incubated at 37 °C in 5% CO₂ in a humidified incubator.

Zebrafish hematogenous metastasis model

Maintenance of the transgenic zebrafish and the experimental design for this study were approved by the Animal Welfare Committee of the Hamamatsu University School of Medicine animal welfare. Zebrafish were maintained according to standard methods (Westerfield, 1993). The transgenic strain of zebrafish expressing enhanced green fluorescent protein (EGFP) under the flk1 (VEGFR2) promoter (flk1: EGFP) was obtained from the Zebrafish International Resource Center (Jin et al., 2005). Human cancer cells were microinjected into the zebrafish larvae, following the protocol reported previously with some modifications (Stoletov et al., 2010). Fish larvae were dechorionated and anesthetized with 0.006% tricaine 48 h post-fertilization (hpf; Sigma). Anesthetized larvae were then transferred onto an agarose gel for microinjection. RFP expressing cancer cells were detached from culture dishes using enzyme-free and PBS-based Cell Dissociation Buffer (13151-014, Gibco) and then were washed twice with PBS. Cancer cells were injected into the common cardinal vein using a tapered borosilicate glass capillary (1.0 mm in diameter, World Precision Instruments, Inc.) with a tip diameter of 20–40 µm (i.d.) connected to a 50 ml glass syringe. The position of the capillary tip was controlled by a manipulator (M-152, Narishige Scientific). The injected fish larvae were kept at 32 °C in the presence of 25 µg/ml dexamethasone (Sigma) for 5–7 h for immunosuppression only before observation because the presence of both tricaine and dexamethasone suppressed the heart beat of larvae. The larvae that formed severe emboli in the arterioles were selected under a fluorescence stereomicroscope (SZX16, Olympus), and used for further experiments.

Live imaging of embolus-forming cancer cells and endothelial cells

A small drop of water containing an anesthetized larva was placed in a glass-bottom dish (36 mm, Matsunami, Gifu). One ml of low temperature melting agarose (E-3126-25, BM Equipment, Tokyo) containing 0.006% tricaine was added to the dish to hold the larva in a gel attached to the glass at the bottom. The dish was filled with 2 ml water containing 0.006% tricaine. The long time dual color time-lapse recording was carried out by using a homemade imaging system driven by ImageJ software (NIH, USA) or a microscope system in an incubator (BioStation, Nikon, Tokyo). The filter wheels (FW102; Tholabs, NJ) were controlled by the ImageJ software (Research Services Branch, National Institute of Mental Health, Bethesda, Maryland, USA; http://imagej.nih.gov/) through an IJSerial plugin (http://www.eslide.net/ijstage.php) for the dual color fluorescence imaging, and time-lapse images were captured using a QuickTime Capture plugin and Time-Lapse Video macros with some modifications (http://rsbweb.nih.gov/ij/plugins/qt-capture.html). The analog monochrome camera (WAT-120N+; Watec, Tokyo) was connected to the digital converter (ADVC-300; Canopus) and a PC via IEEE1394 interface. The background noise was decreased using an image processor (ARGUS-20; Hamamatsu Photonics, Hamamatsu). The temperature on the stage was maintained at 32 °C using a transparent heating plate (Kitazato, Fuji). In a standard recording, fluorescence images were captured every 5 min for a total time of up to 11 h. The spatial evaluation of cancer cells and endothelial cells was carried out using a confocal microscope (FV1000; Olympus). 3D stack images were taken by confocal microscopy (FV1000; Olympus) and processed using image processing package Fiji (http://fiji.sc/Fiji) (Schindelin et al., 2012).

RNA interference, RT-qPCR

The siRNA sequence targeting human VEGF-A (referred here as VEGF) was the same as that designed in a previous report (Takei et al., 2004): 5′-GGAGUACCCUGAUGAGAUCdTdT-3′ (sense), 5′-GAUCUCAUCAGGGUACUCCdTdT-3′ (antisense). MISSION siRNA Universal Negative Control (Sigma) was used as a nonsense control siRNA. Relative mRNA amounts were quantified using iQ™ SYBR^® Green Supermix (Bio-Rad Laboratories, Inc). Forward 5′-CCTGGTGGACATCTTCCAGGAGTA-3′ and reverse 5′-CTTGGTGAGGTTTGATCCGCATAA-3′ primers were used to detect VEGF, and forward 5′-AACGGATTTGGTCGTATTGGGC-3′ and reverse 5′-TTCTCAGCCTTGACGGTGCCAT-3′ primers were used to detect glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA. For comparative, quantitative analysis, transcript levels were normalized to the level of GAPDH and changes were determined. The comparative quantitation method (ΔΔCt) was used to compare the different samples and transformed to absolute values with 2ˆ(−ΔΔCt) for obtaining relative fold changes. All assays were performed in triplicates. VEGF siRNA-treated RFP-HeLa cells were injected into the blood vessels 17–19 h after transfection. The cancer cell-injected larvae were kept at 32 °C in the presence of 25 µg/ml dexamethasone for 5–7 h for immunosuppression, and these cells were then observed for 11 h within 37 h after the transfection, so that VEGF was efficiently depleted during the observation period.

Anti-VEGF treatment

Sunitinib (Toronto Research Chemicals, Toronto), an orally active VEGFR tyrosine kinase inhibitor, was dissolved in dimethyl sulfoxide (DMSO) to make a stock solution of 10 mM. The stock solution was diluted in water to attain a concentration of 5 µM. Sunitinib treatment was started immediately after cancer cell injection. The injected fish larvae were kept at 32 °C in the presence of 25 µg/ml dexamethasone (Sigma) and 5 µM sunitinib for 5–7 h for immunosuppression before observation. The cancer cell-injected larvae were then held in glass-bottom dishes with low temperature melting agarose on which water containing 0.006% tricaine and 5 µM sunitinib was added.

In vitro observation of live cells

The cells were seeded in polymer-bottom dishes (Bio Medical Science, Japan), and time-lapse images were captured every minute for 12 h using BioStation (Nikon).

Immunofluorescence

The cells in the polymer-bottom dishes were fixed (4% paraformaldehyde for 15 min), permeabilized (1% Triton X-100 for 15 min) and blocked (PBS containing 2% bovine serum albumin (Sigma, St Louis) for 30 min) for non specific immunostaining. The cells were then incubated with anti-vinculin mouse monoclonal antibody (ab18058, 1:200, Abcam, City) overnight at 4 °C, subsequently with anti-mouse-IgG conjugated with Alexa 488 (Invitrogen), and finally with 1 µg/ml Hoechst 33342 (Dojindo, Kamimashiki). The immunostained cells were observed under a confocal microscope (FV1000; Olympus, Hachioji).

Chemotactic cell migration assay

The cells were seeded in BD Falcon™ FluoroBlok™ 24-Multiwell Insert Systems (351157; BD, NJ). The bottom wells were filled with chemoattractant-rich NIH3T3 conditioned medium. The cells were stained with calcein-AM (Dojindo) after 24-h incubation, and the number of stained cells in the bottom wells was counted in fluorescence images captured using an inverted fluorescence microscope.

Electron microscopy

Fish larvae were fixed for scanning electron microscopy (SEM) by perfusion with 2% paraformaldehyde and 2.5% glutaraldehyde in PBS using the microinjection system for approximately 5 min, and then transferred into the same fixation buffer. Samples were dehydrated in a graded ethanol series, then frozen in liquid nitrogen and cracked into several pieces. These samples were then collected and freeze dried with t-butanol, and stained with OsO₄, and examined with a scanning electron microscope (S-4800; Hitachi, Tokyo).

Extravasation of embolus-forming human cancer cells

This study first examined the potential for the extravasation of human cervical cancer cells (HeLa). RFP-expressing HeLa (RFP-HeLa) cells were microinjected into the circulation, following the protocol previously reported (Stoletov et al., 2010). Unlike the previous study reporting that injected cancer cells are arrested in the thinner intersegmental vessels (Stoletov et al., 2010), RFP-HeLa cells formed severe emboli mostly in the thicker caudal artery immediately after injecting into the circulation. The clusters of cancer cells in the emboli extravasated and adhered to the tissue outside the blood vessels after 17–20 h (Fig. 1A). The spatial distribution of the adhering-cancer cells and blood vessel-forming endothelial cells was evaluated using a confocal microscope 1 day after the cancer cell injection to confirm that embolus-forming cancer cells actually extravasated and developed adhesion to the tissue outside the blood vessels. The optically sliced images clearly showed that endothelial tube structures were devoid of the mass of cancer cells, and all of the RFP-HeLa cells adhered to the tissue outside the blood vessels (Fig. 1B). Therefore, RFP-HeLa cells, which formed severe emboli in thicker caudal artery of zebrafish, have the ability to efficiently extravasate and adhere to the tissues surrounding the blood vessels.

Two processes of extravasation in human cancer cells

The process of extravasation by the RFP-HeLa cells was observed using a long-time dual color time-lapse recording system to study mechanisms of extravasation of human cancer cells. The embolus-forming RFP-HeLa cells and endothelial cells could be observed for more than 11 h, which was long enough for these cells to exhibit slow behaviors. There were two distinctive processes of extravasation observed in RFP-HeLa cells (Fig. 2A). Some of the cancer cells actively invaded the vessel wall and penetrated the wall in the process broadly accepted as the process of extravasation (Stoletov et al., 2010; Weis et al., 2004) (referred to as cancer cell invasion; Fig. 2A, left; Movie S1). In contrast, other masses of cancer cells seemed to be quiescent. These cells did not invade the endothelial cell layer. Instead, a new leaf of endothelial cells appeared and extended over the embolus-forming cells. The endothelial cells eventually covered the cancer cells on the vessel wall. Simultaneously, the original layer of endothelial cells gradually disappeared and the cluster of cancer cells spread to the tissue outside the blood vessels (referred to as endothelial covering; Fig. 2A, right; Movie S2). Intriguingly, 9 out of 17 extravasations were carried out in the absence of cancer cells’ active invasion, while 5 were clearly accompanied by invasion of cancer cells. In the remaining 3 extravasations, embolus-forming cancer cells showed invasion and surrounding endothelial cells also newly extended over the cancer cells (Fig. 2B). The spatial distribution of embolus-forming cancer cells and newly spreading endothelial cells was further analyzed using 3D confocal microscopic images (47 slices, step size: 1 µm) that were taken at 10 h postadministration. Side views of the stack image and 3D reconstructed image clearly showed that some endothelial cells were extending over the cancer cells or penetrating into the cluster of cancer cells (Fig. 2C; Movie S3). These results suggest that efficient activation of endothelial cells by cancer cells is also important for extravasation, in addition to high motility of cancer cells, which has been regarded as one of the most crucial factors associated with malignancy.

Normal proliferating fibroblast 3T3 cells (NIH, USA) showed neither the invasion type nor the endothelial covering type extravasation in this model. Six clusters of these cells were in the vessels of 6 different individual fish for more than 11 h. All of them formed severe emboli similar to those of cancer cells, but showed neither the development of strong adhesions to the vascular walls nor the process of extravasation mentioned above. These cells were eventually pushed away by the blood flow. This suggests a difference to exist in the expression of adhesion molecules between tumorigenic cells and normal cells, and adhesion to the endothelial walls is a key to induce extravasation for the circulating cancer cells after forming an embolus in the blood vessels.

Effects of VEGF depletion on cancer cell properties and extravasation

Malignant tumors actively induce new blood vessels from surrounding tissues by secreting vascular endothelial growth factor (VEGF) to receive nutrients and oxygen. In addition, cancer cells may intravasate through this newly-formed leaky vasculature, leading to the metastasis in remote places (Senger et al., 1993; Esser et al., 1998). Hence we examined the effects of depletion of VEGF expressed in RFP-HeLa cells to study the role of VEGF in tumor cell extravasation. The cells were transfected with siRNA prior to their injection into the vessel of the fish. The efficient depletion of VEGF was confirmed by real-time quantitative PCR (RT-qPCR). All the 3 family members of VEGFA, which are secreted by cancer cells, have the same target sequence so that all the VEGFA (referred to as VEGF, here) were depleted in the cancer cells. Expression of VEGF was depleted to 29 ± 6% in comparison to the control obtained by using nonsense siRNA-treated cells (Fig. 3A). The effect of VEGF depletion on the cancer cell itself was then evaluated morphologically in vitro. The VEGF-depleted cells tended to aggregate with their motility decreased uniformly. They were flattened and their adhering surface became very large (Fig. 3B; Movie S5). Quantitative evaluation indicated that cell motility was reduced to 36.0 ± 7.7% in comparison to the control obtained by using nonsense siRNA-treated cells as reported previously (Bachelder et al., 2003) (Fig. 3C).

Vinculin, a focal adhesion protein, was immunostained in nonsense siRNA-treated and VEGF siRNA-treated RFP-HeLa cells to examine the effects of VEGF depletion on the adhesive property of cancer cells. Focal adhesions were uniformly formed at the cell periphery in VEGF depleted-cells, but only locally in control siRNA-treated cells so as to form a highly polarized shape (Fig. 3D). These results suggest that VEGF is involved not only in the activation of endothelial cells, but also in the migration of cancer cells via cell polarization.

We next examined the effects of VEGF depletion on extravasation in vivo. Like the normal RFP-HeLa cells, the VEGF-depleted cells immediately formed emboli after being injected into blood vessels, and then the endothelial cells migrated over the embolus-forming cancer cells, although the process was markedly delayed (Fig. 3E). Strikingly, the process of cancer cell invasion through the endothelial cells was completely suppressed during the 11-h observation period (Figs. 3E and 3F; Movie S6). Despite the severe inhibition of cancer cell invasion, six extravasation events were still observed in VEGF-depleted cells (Fig. 3F), thus suggesting that the clinically adopted anti-angiogenic strategy of VEGF targeting is insufficient for the prevention of metastasis. However, since RNA interference does not completely suppress target molecule expression compared to that of gene knockout, we cannot rule out the possibility that a residual amount of VEGF could be enough to activate surrounding endothelial cells.

Effects of a multi-targeted anti-angiogenic kinase inhibitor on cancer cell properties and extravasation

VEGF depletion in RFP-HeLa cells only delayed the endothelial covering-type extravasation, while completely inhibiting the cancer cell invasion-type extravasation. Therefore, we anticipated that a potent inhibitor of tumor angiogenesis via VEGF signaling could completely inhibit both types of extravasation. Sunitinib (Sutent^®; Pfizer Inc.), which effectively suppresses angiogenesis by inhibiting signaling from the receptors for VEGF and for platelet derived growth factor (PDGF) (Bergers et al., 2003; Pietras & Hanahan, 2005), was orally administered to the zebrafish larvae. The thinner intersegmental vessels were severely deteriorated by the overnight treatment with sunitinib (Fig. 4A). However, the thicker blood vessels were not affected morphologically, suggesting that the effect of sunitinib is limited to the newly-formed vulnerable vessels. Endothelial cells in the mature vasculature are independent of VEGF signaling for survival (Gerber et al., 1999; Lee et al., 2007). Sunitinib inhibits the VEGF receptor, so the treatment of cells with sunitinib should affect the cancer cells like VEGF depletion. As expected, the cells examined in vitro adhered to an obviously larger area of the substrate, became less motile, and tended to aggregate in the presence of sunitinib, as the VEGF-depleted cells did (Fig. S1A). In addition, the sunitinib treatment reduced the cell motility by 35.0 ± 5.5% (Fig. S1B). Focal adhesions were uniformly formed at the cell periphery in sunitinib-treated cells, as in VEGF-depleted cells (Fig. S1C).

The effect of sunitinib on extravasation was examined in vivo. Sunitinib treatment completely suppressed the process of cancer cell invasion-type extravasation as expected. In an unexpected development, the incidence of endothelial covering-type extravasation of RFP-HeLa cells was not affected by treatment with sunitinib, although the process was markedly delayed, similarly to the extravasation of VEGF-depleted cells (Fig. 4B; Movie S7). The area of cancer cells that were covered by the endothelial cells in the presence of sunitinib was markedly larger than that in normal or VEGF-depleted RFP-HeLa cells in most events of extravasation, and the blood vessel walls concurrently, but slowly, moved toward the embolus-forming RFP-HeLa cells (Fig. 4B; Movie S7). Therefore, the sum of the area of extravasated cancer cells was calculated from 7 movies that were recorded in RFP-HeLa cells with and without sunitinib treatments to quantitatively compare the whole volume of cancer cells extravasated (Fig. S1D). Although the incidence of endothelial covering-type extravasation in the presence of sunitinib (9 incidences in 7 larvae) was the same as that observed without treatment of sunitinib (Figs. 2B and 4C), sunitinib treatment increased the total volume of extravasated cancer cells to 153% of the volume of extravasated cells without sunitinib treatment during 11-h observation period (Fig. 4D). These results suggest that the anti-angiogenic inhibitor sunitinib can completely suppress the cancer cell invasion-type extravasation without having strong effects on the incidence of endothelial covering-type extravasation. Paradoxically, the total volume of the extravasated cancer cells increased in the presence of sunitinib, thus suggesting that sunitinib accelerates the process independently of the VEGF signaling. Moreover, the endothelial cells apparently activated by sunitinib were studied morphologically using a scanning electron microscope. Many protrusions were observed on the luminal face of endothelial walls of arteries in zebrafish, and interestingly, many small holes were found in some regions on the endothelial wall (Fig. 4F). These holes are reminiscent of fenestrated endothelial walls in tumor vasculature or normal microvasculature of mice (Inai et al., 2004; Kamba et al., 2006; Mazzone et al., 2009). Sunitinib-treated zebrafish larvae showed no hole on the endothelial walls of arteries with fewer protrusions (Fig. 4H), and also showed prominently thicker endothelial walls (Fig. 4G).