Research ArticleCancer Imaging

Intravital Microscopy Through an Abdominal Imaging Window Reveals a Pre-Micrometastasis Stage During Liver Metastasis

See allHide authors and affiliations

Science Translational Medicine  31 Oct 2012:
Vol. 4, Issue 158, pp. 158ra145
DOI: 10.1126/scitranslmed.3004394


Cell dynamics in subcutaneous and breast tumors can be studied through conventional imaging windows with intravital microscopy. By contrast, visualization of the formation of metastasis has been hampered by the lack of long-term imaging windows for metastasis-prone organs, such as the liver. We developed an abdominal imaging window (AIW) to visualize distinct biological processes in the spleen, kidney, small intestine, pancreas, and liver. The AIW can be used to visualize processes for up to 1 month, as we demonstrate with islet cell transplantation. Furthermore, we have used the AIW to image the single steps of metastasis formation in the liver over the course of 14 days. We observed that single extravasated tumor cells proliferated to form “pre-micrometastases,” in which cells lacked contact with neighboring tumor cells and were active and motile within the confined region of the growing clone. The clones then condensed into micrometastases where cell migration was strongly diminished but proliferation continued. Moreover, the metastatic load was reduced by suppressing tumor cell migration in the pre-micrometastases. We suggest that tumor cell migration within pre-micrometastases is a contributing step that can be targeted therapeutically during liver metastasis formation.


Cancer-related mortality is predominantly caused by metastatic tumor growth in secondary organs such as the liver. Metastasis is a multistep process, which requires cells to escape from the primary tumor, survive in the circulation, enter a distant organ, and grow out in this new environment (1). Furthermore, angiogenesis is required for the outgrowth of small metastases (micrometastases) into macroscopically detectable metastases (macrometastases with a diameter of >500 μm) (2). For the development of effective therapeutic agents targeting metastasis and angiogenesis, techniques are required to identify processes underlying metastatic growth and therapy resistance. Intravital microscopy (IVM) allows the visualization and analysis of tumor cell dynamics in live animals in real time and may therefore lead to the discovery of novel steps during metastasis formation, which may be used in the design of therapeutics (3).

The dynamic behavior of tumor cells that escape from the primary tumor has been extensively studied with high-resolution IVM in mice (48). A small number of studies have also imaged the colonization of organs that are prone to metastasis, such as the lungs (9), bone marrow (10), lymph nodes (11), spleen (12), and liver (13, 14). Nonetheless, these organs are anatomically inaccessible by microscopes and should be surgically exposed, which precludes long-term imaging. Therefore, the long-term dynamic aspects of colonization are largely unknown. Cranial imaging windows can be used for multiple-day imaging and have allowed investigators to image the formation of brain metastases (15). However, brain tissue, isolated from the circulation by the blood-brain barrier, has a unique environment, and therefore, the steps to colonization may not be representative for other organs, such as the lungs and the liver. Other commonly used imaging windows cannot be used to image metastasis-prone organs because of their inability to fix organs by “clamping” [dorsal skin fold chamber (16)] or by their poor abdominal fixation and window-induced organ damage [for example, mammary window (1719)]. New lung imaging windows have been designed that allow imaging for up to 3 hours (20, 21), but these cannot be used to image metastatic outgrowth over multiple days.

Short-term (<24-hour) video microscopy studies of liver colonization identified many early events in liver metastasis, such as survival and growth (13, 14). The formation of a clone should be followed longitudinally to link cellular behavior to successful colonization. For this, new long-term imaging windows are required that enable studying long-term dynamic events. Here, we have developed an imaging window for abdominal organs, which has allowed us to study cancer cell migration during the different phases of hepatic colonization in detail. We identified migration as a facilitating “pre-micrometastasis” step during the early colonization of liver metastases that can potentially be targeted therapeutically.


Characterizing response to the abdominal imaging window

The presently available imaging windows (fig. S1A) were ultimately not suitable for abdominal organs, such as the liver. Therefore, we designed the abdominal imaging window (AIW), which consists of a reusable titanium ring with a 1-mm groove on the side and a coverslip on top, tightly secured in the skin and abdominal wall by a purse-string suture (Fig. 1A, fig. S1B, and movie S1). The purse-string suture is located within the groove of the ring, and there is no direct opening to the abdomen with subsequent danger of infections.

Fig. 1

The AIW as a tool to study liver metastasis over multiple-imaging sessions. IVM images of tumor cells, the vasculature, and type I collagen [second harmonic generation (SHG)] were taken through the AIW surgically implanted above the liver. (A) Cartoons and pictures of the AIW. (B) High-resolution image of a C26–GFP-actin cell within the liver. Actin-rich protrusions are marked by the arrow. (C) Overview of the liver vasculature and stalled C26 tumor cells. The circled and boxed areas highlight two stalled cells. Type I collagen (purple) is shown by SHG imaging. (D) Images of the vasculature and C26 tumor cells at days 1, 2, and 5 after injection. The dotted lines outline the sinusoids. (E) Long-term (14-day) tracing of liver colonization. (F) Images of type I collagen taken at the indicated time points. A merged image highlights the colocalization (yellow). (Right) Pearson’s correlation coefficient (r) and scatter plot of the type I collagen images. The selected noncolocalizing pixels from the scatter plot, as indicated by the blue and red boxes, were replotted in the IVM image of day 1. (G) Long-term (5-day) tracing of the outgrowth of an individual cell into a micrometastasis. Scale bars, 20 μm.

Existing imaging windows remain in place for periods ranging from days to weeks (1517). To record the time until dislodgement of the AIW occurred, we implanted the window onto several abdominal mouse organs. The window did not dislodge from the abdominal wall or skin for an average of 5 weeks ± 3.9 days (SEM, n = 12), with a maximum of 63 days (fig. S2A). To detect potential discomfort caused by the AIW, we assessed the postoperative behavior of mice according to a clinical appearance scoring system (22). We measured the reactivity, appearance, and behavior of mice carrying an AIW and found a majority of normal scores (fig. S2, B to D). One day after surgery, the weight of all the animals was slightly reduced (about 5% of total body weight) but returned to normal from day 5 onward (fig. S2E). For the remaining weeks, a steady increase in weight was observed as expected for aging mice. Mice were not impaired in their mobility with the AIW, as demonstrated by their ability to stretch their bodies to get food (movie S1). Furthermore, we did not observe any visual signs of inflammation or necrosis of the skin surrounding the AIW. The white blood cell count 1 day before and 1, 4, 6, and 8 days after surgery was normal, indicating the absence of leukocytosis or leukopenia (fig. S2F). The red blood cell count was elevated 1 day after surgery, likely owing to loss of fluids during abdominal surgery, but returned to basal levels by day 4 (fig. S2G).

Liver regions that were behind the AIW and liver regions of mice without an AIW (control) were compared. We did not find any difference in the number of lymphocytes between the two groups (fig. S2H). For both the control and the AIW tissue, we observed neutrophils in the sinusoids (fenestrated capillaries) but not in between hepatocytes (fig. S2I). These data suggest the absence of a local inflammatory response upon AIW insertion. To test whether the AIW can lead to alterations in the structure of the abdominal tissue, we analyzed hematoxylin and eosin (H&E)–stained sections to identify possible necrotic zones and to look for abnormalities within the liver tissue architecture. We could not detect necrotic zones or any architectural differences between sections of nonperturbed and window-fixed livers (fig. S2J). Furthermore, there was no difference between the number of apoptotic cells in the tissue that was behind an AIW and in the tissue of control animals (fig. S2K).

We also encountered a few challenges when using the AIW. An unavoidable problem is the breakage of the coverslip, which happens in 3% of all cases. Moreover, 1 week after implantation, encapsulation of the abdominal organ by connective tissue led to a drift of the organ in 20% of the cases where the abdominal organ was not touching the coverslip of the AIW. This drift caused the imaging site to be located beyond the maximum imaging depth of our microscope (~800 μm), obscuring imaging resolution and sharpness of our images. Tissue motion caused by respiration can also result in imaging distortions. To reduce this, we used an inverted microscope and optimized the surgical procedure for each abdominal organ (Supplementary Methods). Last, an imaging box (fig. S3A) that fits the stage of our multiphoton microscope (fig. S3B) was designed to stably fix the AIW above the objective.

Repetitive tumor cell imaging with the AIW

We aimed to visualize the formation of liver metastases from individual colorectal tumor cells. For in vivo imaging of liver metastasis formation, we used a standardized liver metastasis assay (23) in which mouse colorectal tumor cells (C26 cells) expressing fluorescent proteins were injected into the splenic parenchyma of syngeneic BALB/c mice. To test whether the AIW could potentially be used to study the formation of liver metastases, we first visualized individual C26 cells in the liver parenchyma through the AIW with subcellular resolution (Fig. 1B), as exemplified by the cell protrusions enriched in green fluorescent protein (GFP)–labeled actin. Ten minutes after intrasplenic injection of C26 cells transfected with fluorescent Dendra2 (day 1), individual colorectal cells were trapped in the sinusoids of the liver (Fig. 1, C and D). These sinusoids are highly permeable and can be easily traversed by tumor cells (24). In line with this, at days 2 and 5, cells could be seen outside the sinusoids (Fig. 1D).

To visualize the colonization of the liver parenchyma by fluorescent C26 cells that are trapped in the sinusoids, we retraced areas of interest over multiple-imaging sessions (14 days) using an AIW that contained a gridded coverslip (Fig. 1E and fig. S1B). For retracing with subcellular accuracy, we used type I collagen structures as reference points (visualized by SHG) (25) (Fig. 1, F and G, and fig. S4). To illustrate the ability to retrace areas using type I collagen fibers, we quantified the colocalization of the fibers of an area at days 1 and 2 by plotting the intensities of pixels of both images in a scatter plot and calculated the Pearson’s correlation coefficient (r) (Fig. 1F). The pixels in the scatter plots of well-aligned images will appear along the diagonal, whereas pixels of unaligned images will be located off the diagonal (no colocalization). When we retraced imaging regions the next day, the two images aligned well, and r ranged from 0.2 to 1, depending on the signal-to-noise ratio in the images (Fig. 1F). As a nonretraced negative control, we analyzed the same images but flipped the second image (fig. S4A). As expected, the r of the nonretraced negative control dropped toward zero (we will refer to the correlation coefficient for this negative control as rflip). This method allowed us to trace the liver colonization over the course of 5 days (Fig. 1G), as illustrated by the high r and low rflip (fig. S4B). After 5 days, type I collagen in the initial imaging area got lost due to tumor growth, preventing the retracing at subcellular accuracy.

Metastasis growing from a single founder cell

The increase in the number of cells in the growing micrometastasis can be explained either by multiple trapped cells that cluster (synergistic growth) or by the clonal growth of an individual cell (fig. S5). Earlier work suggests that most metastases are clonal (26); thus, it was expected that our C26 tumor model also grew clonally. However, we cannot exclude synergistic growth of our model. To determine whether multiple tumor cells need to cluster to initiate proliferation and growth of a metastasis or whether metastases grow from a single founder cell, we co-injected the same number of two distinctly colored C26 tumor cell populations into the spleen. The different colors enabled us to trace progeny, even when cells were located within the same imaging field (Fig. 2A). We found that 94% of the micrometastases in the liver were single-colored at day 5 (Fig. 2B). Likewise, 97% of the macrometastases (>500 μm) consisted of a single color at day 14 (Fig. 2, B and C). The color distribution in micro- and macrometastases was significantly different from the synergistic growth hypothesis (P < 0.0001, G test). These data illustrate that most metastases are grown from a single founder cell. Differently colored tumor cell populations did not differ in their ability to form macrometastases (Fig. 2D).

Fig. 2

Clonal metastatic growth of C26 liver metastases. (A) A mix of differently colored populations of C26 tumor cells was injected intrasplenically (more details in fig. S5) and imaged using the AIW at days 2 and 3. A corresponding scatter plot is shown. Scale bar, 20 μm. (B) Theoretical expected percentages of cell distribution (assuming two differently colored population of C26 cells) in metastases for clonal growth and for synergistic growth, and the experimental percentage from this study (n = 18 micrometastases in four mice, n = 39 macrometastases in four mice, P values obtained using a G test). (C) An ex vivo image of 14-day-old liver metastases, each containing only one color. The dotted lines outline metastases. (D) A graph is provided to show the percentage of macrometastases of each color in the liver (n = 39 metastases in four mice). Scale bar, 50 μm.

A pre-micrometastatic stage with high tumor cell motility and proliferation

Next, we analyzed the outgrowth of single founder cells into metastatic colonies to explore potential unidentified contributing steps in the formation of metastases. A mix of differently colored C26 cells was monitored for 5 days after injection in the spleen. Within the first 24 hours, trapped tumor cells had left the vasculature and proliferated. The center of the progeny was on average located 71.5 ± 9.3 μm (SEM) (about three cell diameters) away from the arrival site. This deviation is larger than the theoretically calculated shift of two cell diameters in the unlikely case that cells would divide asymmetrically toward one direction for this time frame, and therefore, the observed shift suggests that the cells have migrated away from the arrival site. When examining the clones at different time points, a phenotypic difference was observed (Fig. 1G). The cells of 3-day-old clones were surrounded mainly by liver parenchyma. Over the next 2 days, the cell density in the clones increased so that most cells were mainly surrounded by other tumor cells. To define these different stages, we refer to a clone as a pre-micrometastasis when most cells have at least one-half of their cell surface surrounded by liver parenchyma [Figs. 1G (day 3) and 3A]. When less than 50% of the cells within a clone are mainly surrounded by liver parenchyma and when the clone is smaller than 500 μm in diameter, we refer to this clone as a “micrometastasis” [Figs. 1G (day 5) and 3A].

Fig. 3

Migration within pre-micrometastases but not within micrometastases in the liver. IVM images of type I collagen (SHG) and tumor cells were taken through the AIW. (A) Percentage of cells in the clone that has at least one-half of their cell perimeter touching liver tissue. If most of the cells reach these criteria, we refer to this clone as a pre-micrometastasis (n ≥ 11 for each condition in three or more mice per condition). (B) Representative images of cell movement over 10 hours. The yellow lines highlight the tracks of the tumor cells. Scale bars, 20 μm. (C) Scatter plot and r of the cells shown in (B). The selected noncolocalizing pixels from the scatter plot (indicating cell movement, purple and red boxes) were replotted in the t = 0 IVM image shown in (B). (D) Migration path of individually traced tumor cells (n ≥ 58 cells in three mice per condition). A few example migration tracks are shown in color. (E) Quantification of the track distance and displacement of individual tumor cells after 10 hours. Data are from the cells tracked in (D). P values were obtained using a Mann-Whitney U test. Data in (A) and (E) are means ± SEM.

To test whether cells in these phenotypically different stages display distinct behavior, we acquired 10-hour time-lapse movies of pre-micrometastases and micrometastases (Fig. 3B). The cell signals detected at 0 and 10 hours were then compared in a scatter plot (Fig. 3C). By subsequently highlighting the noncolocalizing (off-diagonal) pixels in the original microscopy image, we could see which cells had changed location between the 0- and 10-hour time points. Pre-micrometastases consisted of active cells that formed protrusions and were migratory (Fig. 3, B to D, and movie S2). Although the cells in pre-micrometastases were migratory, disintegration did not occur because the cells changed direction often. In contrast to pre-micrometastases, cells within micrometastases were largely immotile and did not form protrusions (Fig. 3, B to D, and movie S2). In line with this, individual photomarked cells within micrometastases did not move (fig. S6). Both the mean total track distance (54.7 μm) and the mean displacement (22.2 μm) of tumor cells within micrometastases were significantly less compared to pre-micrometastases (105.4 and 63.6 μm, respectively) (Fig. 3E).

Pharmacological inhibition of migration and growth of cells in pre-micrometastases

The observation that tumor cells display active migration during the pre-micrometastatic stage raises the intriguing possibility that this could contribute to the efficiency of liver metastasis formation. We have recently established that the cofilin pathway is critical for migration of C26 cells and that cofilin is activated after phospholipase C (PLC) activation (27). Although the migration of leukocytes (28), neutrophils (29), and macrophages (30) is independent of PLC, migration and invasion of many tumor cell lines, including C26 cells, can be inhibited by the PLC inhibitor U73122 (fig. S7, A and B) (27, 31). To test whether the migration of C26 cells can be inhibited during the pre-micrometastatic stage, we treated mice with U73122 or dimethyl sulfoxide (DMSO) (vehicle) after cells had left the vasculature (fig. S8A). Subsequently, we imaged in vivo migration of the tumor cells in pre-micrometastases through the liver AIW. U73122 significantly decreased the mean track distance 2-fold (Fig. 4A) and the mean displacement 1.6-fold (Fig. 4B) of tumor cells in pre-micrometastases.

Fig. 4

Migration and growth of pre-micrometastases treated with U73122. C26 tumor cells were injected intrasplenically, and their intrahepatic course was followed over time using a liver AIW. Animals were treated after cells had left the vasculature according to the scheme depicted in fig. S8. (A and B) Track distance (A) and displacement (B) of single pre-micrometastatic cells within 1 hour in treated animals (n > 282 cells in five mice per condition). (C) Growth rate of pre-micrometastases in treated animals between day 2 and day 3 (n ≥ 16 clusters in three or more mice per condition). (D) Number of micrometastases in H&E-stained liver sections from treated animals (n = 5 mice per condition). (E) At day 14, livers were isolated, and the HRA was determined (n = 15 sections in five mice per condition). (F) C26 cells transfected with Lim kinase (LIMK) or control plasmid were injected intrasplenically. At day 14, livers were isolated, and the HRA was determined (n = 5 mice per condition). Data are means ± SEM. P values were determined using a Mann-Whitney U test.

To test whether blocking cell migration in the pre-micrometastatic stage by U73122 treatment has an effect on liver metastasis formation, we analyzed the growth of pre-micrometastases. U73122 treatment significantly suppressed the growth rate of C26 pre-micrometastases in vivo by 1.7-fold (Fig. 4C), which was in contrast to the in vitro growth response (fig. S7, C and D). This suggests that U73122 may affect growth in the pre-micrometastasis indirectly by inhibiting migration. Furthermore, the diminished growth rate in vivo led to a 1.4-fold reduction in the number of pre-micrometastases that grew into micrometastases (Fig. 4D). When blocking cell migration in the pre-micrometastatic and subsequent stages (fig. S8B), the metastatic area in relation to the liver tissue area [hepatic replacement area (HRA)] was also 1.5-fold diminished (Fig. 4E). An alternative genetic approach to inhibit migration of C26 cells is the overexpression of the Lim kinase that phosphorylates and inhibits cofilin (27, 32). Similar to U73122 treatment, expression of Lim kinase reduced the HRA by 3.4-fold (Fig. 4F).

Growth rate of micro- and macrometastases unaffected by U73122

If U73122 affects growth in the pre-micrometastasis by inhibiting migration, then we hypothesize that U73122 would not inhibit growth in the micro- and macrometastases because migration of cells is diminished at these stages. Growth is a balance of proliferation and cell death; thus, we determined the number of proliferating and apoptotic cells in liver sections of 14-day-old macrometastases of mice that followed a prolonged treatment protocol (fig. S8B). We did not find significant differences in the number of proliferating (Ki67-positive) cells in metastases of these two animal groups (Fig. 5A) or in the mitotic indices (Fig. 5B). There were no significant differences in cleaved caspase-3 or terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL)–positive cells in the two treatment groups (Fig. 5, C and D). To test for potential effects of U73122 on microvasculature density, we stained the sections for CD31. There was no significant difference between the two treatment groups in microvasculature (Fig. 5E). Moreover, U73122 did not inhibit growth of micrometastases into macrometastases because the size of the 14-day-old macrometastases was not reduced (Fig. 5F).

Fig. 5

Proliferation, survival, and vascularization of C26 macrometastases in the presence of U73122. Mice were injected intrasplenically with C26 cells and treated with DMSO or U73122 until macrometastases were formed. (A) Liver sections of treated animals were stained for proliferation using a Ki67 antibody. The number of proliferating cells was determined in macrometastases (n > 40 fields in three mice per condition). (B) The mitotic index was determined in liver sections of mice injected intrasplenically with C26–H2B-Dendra2 tumor cells (n > 40 fields in three mice per condition). (C) Liver sections of treated animals were stained for apoptotic cells using a cleaved caspase-3 (casp-3) antibody, and cells were quantified in the macrometastases as a percentage of the total amount of cells (n > 40 fields in three mice per condition). (D) Liver sections of treated animals were stained for TUNEL. The number of apoptotic cells was determined in macrometastases (n > 40 fields in three mice per condition). (E) Liver sections of treated animals were stained for CD31+ blood vessels. The number of blood vessels within macrometastases was quantified (n > 8 metastases in three mice per condition). (F) In treated animals, the size of macrometastases was determined in H&E-stained liver sections (n > 38 metastases in three mice per condition). (G) Mice were subcutaneously injected with C26 tumor cells and treated with U73122 or DMSO daily. The tumor growth rate was measured over time (n = 3 mice per condition). (H) The bar diagram shows the normalized HRA of 14-day-old macrometastases of animals treated from day 6 onward according to the scheme depicted in fig. S8C (n = 5 mice per condition). Data are means ± SEM. P values were determined using a Mann-Whitney U test.

These in vivo data agree with the in vitro results showing that U73122 treatment does not directly target proliferation and apoptosis of C26 cells. In line with this, the growth of subcutaneous C26 tumors was not affected by U73122 treatment (Fig. 5G). This implies that the inhibition of growth of pre-micrometastases [with the subsequent effect on the HRA (Fig. 4E)] by U73122 is indirectly achieved by inhibition of migration. Indeed, treatment only during the micrometastatic and macrometastatic stages (fig. S8C), where cells are not migratory (from day 6 onward), did not affect the HRA (Fig. 5H). Collectively, our results suggest that cell migration within pre-micrometastases is a contributing step of metastatic liver colonization.

Long-term imaging of abdominal organs at subcellular resolution

To extend the use of the AIW, we examined whether we could visualize abdominal organs other than the liver at subcellular resolution in living mice using multiphoton microscopy. We were able to visualize single cells in the small intestine, spleen, pancreas, and kidney (Fig. 6). As an example of how the AIW could contribute to stem cell research, we have visualized stem cells positive for the leucine-rich repeat–containing heterotrimeric guanine nucleotide–binding protein (G protein)–coupled receptor 5 (Lgr5) in the crypts of the small intestine, which reside between the paneth cells in the epithelial layer (Fig. 6, A and B). We visualized an individual Lgr5+ stem cell division over the course of 3.5 hours using the AIW (fig. S9 and movie S3). To confirm division, we performed lineage-tracing experiments in GFP-labeled stem cells, where, at day 1, expression of a red fluorescent label in one of the Lgr5+ cells was activated using tamoxifen (Supplementary Methods). On day 2, this new label was present in two daughter cells, but not in the surrounding cells (Fig. 6C). To track these stem cells over multiple days, we used the vasculature as visual (fiducial) landmarks.

Fig. 6

Subcellular visualization of abdominal organs. In all images, type I collagen is shown in purple, the vasculature in red, and other cell types or structures as indicated. (A) A cartoon showing intestinal stem cells and paneth cells within a crypt. (B) IVM images of the small intestine in which the cell-cell adhesions (left) or LGR5+ stem cells (right) are visible. Scale bar, 20 μm. (C) Tracing was obtained by stochastically inducing the expression of one of the confetti colors [red fluorescent protein (RFP)] in the Lgr5-expressing stem cells using tamoxifen (TAM). The red-labeled stem cell proliferates and produces two red-labeled stem cells the next day (indicated by the white arrows). Scale bars, 10 μm. (D) Images of insulin-producing β cells that were transplanted under the kidney capsule. Left panel shows an overview of multiple islets, and the right panel shows a higher magnification of a single islet. The asterisk highlights the normal kidney vasculature. Scale bar, 50 μm. (E) Images of a 28-day tracing experiment of islets of Langerhans that were transplanted under the kidney capsule. Images of the same islet at 1, 5, 12, and 28 days are shown. Scale bar, 100 μm. (F) IVM images of T and B cells (left image) and of OT1-GFP CD8+ T cells and poly-actin cyan fluorescent protein (CFP) CD8+ T cells in the spleen (right image). Inset shows magnification of selected area. Scale bars, 20 μm. (G) IVM images of OT1 GFP CD8+ cells (green) and poly-actin CFP CD8+ cells (shown in red) were made of the same mouse on day 1 (left) and day 7 (right) after it was challenged with an ovalbumin (OVA) peptide in the presence of CpG. (Right) A bar diagram shows the ratios of the GFP/CFP cells normalized to day 1. The OVA-challenged group consisted of three mice, and the control group of one mouse. Data are means ± SEM.

In another demonstration of long-term imaging through the AIW, we visualized the engrafment of GFP-labeled islets of Langerhans in the kidney (Fig. 6D) and then traced the cells for 28 days through the AIW (Fig. 6E). The vascularization of the islets containing β cells was clearly visible at day 8 after transplantation (Fig. 6D).

We were also able to visualize insulin-producing β cells through a pancreas AIW using a mouse insulin promoter–GFP (MIP-GFP) transgenic mouse model (fig. S10A), with pancreatic acinar cells visible at subcellular resolution (fig. S10B). In our last demonstration with the AIW, we intravenously transferred fluorescent B and CD8+ T lymphocytes into the spleen and then followed the acute CD8+ T cell response to antigen (ovalbumin peptide) challenge in vivo over multiple days (Fig. 6, F and G, and fig. S11) (Supplementary Methods). Tracking specific immune responses is an active area of imaging research, and using the spleen AIW, we were able to quantify a T cell response 7 days after antigen challenge (Fig. 6G and fig. S11A). Additional information can be extracted from the IVM data, as we found individual CD8+ T cells to be predominantly migratory and clustered CD8+ T cells to be nonmigratory (fig. S11B and movie S4), which likely reflects antigen recognition (33).


Long-term imaging of metastatic outgrowth and cancer treatment holds promise for discovering novel processes that may be exploited therapeutically (3). Here, we describe the development of an AIW and have used it to image metastatic outgrowth in the liver at subcellular resolution. This high-resolution IVM technique can potentially be used to evaluate tumor recurrence after chemotherapy, which is a poorly understood phenomenon. We have used the AIW to identify a distinct intervention step during metastatic colonization to prevent formation of liver metastases. In this step, individual C26 tumor cells can proliferate into pre-micrometastases. In these pre-micrometastases, the cell density is low and cells display migratory behavior. Subsequently, the growing clones condense into micrometastases, in which cell migration is strongly diminished. Our imaging study indicates that growth of pre-micrometastases is linked to migration and that interfering with migration at this stage prevents liver metastasis formation, which could potentially retard tumor progression and improve patient survival rates.

In our study, we found that tumor cell migration supports the growth of pre-micrometastases but not of micrometastases. Processes such as growth depend on the dynamic interplay between cell intrinsic properties and their microenvironment (34). The different microenvironments experienced by tumor cells in pre-micrometastatic and micrometastatic stages may provide an explanation for the observed differential dependency on migration. Migration is possibly required for optimal positioning of the growing clone. Several migration-inducing genes have recently been linked to hepatic colonization of human colorectal cancer (35, 36). These genes potentially affect extravasation and niche finding, and additionally, they may affect the growth of pre-micrometastases.

Here, we were able to reduce migration in the pre-micrometastases by therapeutic intervention with the PLC inhibitor U73122, which led to reduced metastatic growth. Although U73122 reduced the metastatic burden by ~40%, this drug will not be the first choice as inhibitor for clinical purposes. We used U73122 because it effectively blocks migration without directly affecting proliferation and cell survival. However, the inhibition of metastatic load was not complete. A drug that affects other metastatic steps, such as proliferation or survival, in addition to down-regulating migration, will be much more effective in reducing metastatic outgrowth. Nevertheless, migration is a well-studied process, and many inhibitors of this process have already been developed, which may serve as chemotherapeutic agents.

To study liver colonization, we used a standardized liver metastasis assay in which the colorectal carcinoma cell line C26 was intrasplenically injected. The C26 cell line generates carcinomas with an undifferentiated phenotype. Hence, it may be a good model for metastasis development by undifferentiated colorectal carcinoma. Whether the processes described in this report also play a role during metastasis formation by well-differentiated colorectal carcinomas and/or by other tumor types has yet to be understood. With the recently developed protocols for isolating and expanding tumor-initiating cells directly from tumor resection specimens (37, 38), this now seems feasible. Such studies should reveal the contribution of tumor cell migration during liver colonization across a large panel of human colorectal tumors. The demonstration of such a contribution would form an incentive to start evaluating the added value of migration-targeting drugs in the treatment of metastatic (colorectal) cancer.

Because inhibiting migration will only selectively target metastatic growth at the pre-micrometastatic stage, one can argue against the effectiveness of inhibiting migration as a treatment strategy when patients present themselves with a metastasized tumor. However, similar to tumor cells in the primary tumor, tumor cells within metastases invade and enter blood vessels, leading to circulating tumor cells that seed secondary metastases (39, 40). In the clinical setting, it has also been observed that, even after successful resection of the primary tumor, tumor cells are detected in the blood that may seed new metastases (41). This suggests that new metastases are constantly initiated. Interference of migration during the pre-micrometastasis phase could potentially inhibit the outgrowth of these new clones, thereby blocking the expansion of metastasis, ultimately leading to prolonged survival of patients. The success of this potential therapeutic approach could be evaluated preclinically using the AIW.

Materials and Methods

Cell culture and generation of stable cell lines

C26 colorectal tumor cells were cultured in Dulbecco’s modified Eagle’s medium + GlutaMAX (GIBCO, Invitrogen Life Technologies) supplemented with 5% (v/v) fetal bovine serum (Sigma), streptomycin (100 μg/ml), and penicillin (100 U/ml) (Invitrogen Life Technologies). Cells were kept at 37°C in a humidified atmosphere containing 5% CO2. C26–H2B-Dendra2, C26-Dendra2, C26-mCherry, and C26–LifeAct-GFP were generated with standard lentiviral transfection (Supplementary Methods). Afterward, cells were sorted by flow cytometry and grown as a polyclonal population, or single-cell clones were selected and combined to form a polyclonal population.

Animal models

All experiments were carried out in accordance with the guidelines of the Animal Welfare Committee of the Royal Netherlands Academy of Arts and Sciences, the Netherlands. Female BALB/c (10 to 12 weeks), C57BL/6 mice (10 to 12 weeks), and MIP-GFP mice were purchased from The Jackson Laboratory. 129P2/OlaHsd and FVB/n (The Jackson Laboratory) mice were crossed to obtain female 129P2/OlaHsd;FVB/n mice and were used to measure AIW postsurgical behavior and blood count. Female E-cadherin–mCFP, Lgr5-EGFP-Ires-CreERT2, and R26R-confetti mice were a gift from H. Clevers (Hubrecht Institute). Mice were housed under standard laboratory conditions and received food and water ad libitum. U73122 (Santa Cruz Biotechnology; dissolved in 100% DMSO) and DMSO were both dissolved in phosphate-buffered saline (PBS) [final concentration, 5% (v/v) DMSO], and animals were treated by intraperitoneal injection every other day with 200 μl of U73122 (120 μg/ml) or 5% DMSO.

Liver metastasis assay

C26 cells were harvested by brief trypsinization. Colorectal liver metastases were induced as described previously (23). In brief, single-cell suspensions were prepared to a final concentration of 7.5 × 104 cells/100 μl of PBS. Through the incision made for implanting the AIW, cells were injected into the parenchyma of the spleen. To circumvent outgrowth of tumor cells in the spleen and to prevent the tumor cells from leaving the spleen at later time points, we removed the spleen 10 min after injection of the tumor cells. We did not observe arrival of new cells in the liver after splenectomy. For the mice used for intravital imaging, a liver AIW was implanted immediately after the splenectomy.

Intravital microscopy

Mice were sedated with isoflurane inhalation anesthesia [1.5 to 2% (v/v) isoflurane in O2] and placed within a custom-designed imaging box. The isoflurane was introduced through a face mask and ventilated by an outlet on the other side of the box. The imaging window was placed through a hole in the bottom of the box. The imaging box and microscope were kept at 32°C by a climate chamber. Mouse vitals were monitored during imaging with the MouseOx system (STARR Life Sciences Corp.). Imaging was performed on an inverted Leica TCS SP5 Acousto-Optical Beam Splitter two-photon microscope with a chameleon Ti:Sapphire–pumped optical parametric oscillator (Coherent Inc.). For more details on microscopy, see Supplementary Methods.

Tracking of C26 tumor cells

C26 cells (7.5 × 104) were injected into the spleens of BALB/c mice. Each day, a large overview image of stitched high-resolution images (tile scan) was taken through the liver AIW (minimum of 3 mm × 3 mm). Maximum projections of the three-dimensional volumes (150 to 300 μm deep) were generated from the IVM images. Next, the SHG signal was manually overlaid for the various positions containing cells with ImageJ software [National Institutes of Health (NIH)].

To determine track distance and displacement of cells in the liver parenchyma, we imaged intrahepatic tumor cells and their metastases with the AIW at day 2 (pre-micrometastases) and day 5 (micrometastases). Where indicated, mice were treated with DMSO or U73122 according to the treatment schedule. Three-dimensional volumes (z stacks) were collected every hour for 10 hours (Fig. 3, C and D) or every 10 min for 3 hours (Fig. 4, A and B), and cells were tracked manually with an ImageJ plug-in (NIH). The XYZ position was determined over time, and the displacement and track distance were calculated by Microsoft Excel.

To monitor the in vivo growth rate of pre-micrometastases, we tracked the intrasplenically injected C26 cells and their pre-micrometastases over multiple days. Cells per pre-micrometastasis were counted 1 and 2 days after injection, and the growth rate was calculated with the following formula: LOG(cell count day 1/cell count day 2)/LOG(2).

Statistical analysis

A Student’s t test was used to determine whether there was a significant difference between two means. P < 0.05 was considered significant. If there was no normal distribution present (tested with a Shapiro-Wilk test) or if n was less than 20, then a nonparametric Mann-Whitney U test was performed. For the clonal analysis, a G test (based on the log-likelihood ratio) was performed to determine whether our findings were different from the theoretical findings based on the synergistic growth hypothesis. A one-way or two-way analysis of variance (ANOVA) was used to compare three or more samples.

Supplementary Materials

Materials and Methods

Fig. S1. Comparison of imaging windows.

Fig. S2. The AIW does not have deleterious effects in mice.

Fig. S3. An imaging box fixes the AIW and exposes it to the objective.

Fig. S4. Scatter plots of retraced areas during IVM.

Fig. S5. Clonal or synergistic metastatic growth within the liver.

Fig. S6. Photo marking and tracking of an individual cell within a micrometastasis.

Fig. S7. Inhibition of in vitro migration with U73122.

Fig. S8. Timeline of mouse experiments treated with U73122.

Fig. S9. Intravital imaging of intestinal stem cells through the AIW.

Fig. S10. Intravital imaging of pancreatic tissue through the AIW.

Fig. S11. Intravital imaging of OVA-reacting T cells in the spleen through the AIW.

Movie S1. Normal mobility of representative mice carrying an AIW.

Movie S2. Migration of cells in a pre-micrometastasis and in a micrometastasis.

Movie S3. Intestinal Lgr5+-GFP stem cells in a crypt.

Movie S4. Movement of OVA-challenged T cells.

References and Notes

  1. Acknowledgments: We thank J. de Rooij and S. I. J. Ellenbroek for critically reading this manuscript; F. Kruiswijk for taking photographs of the surgically implanted AIWs; H. Beghtel and J. Korving for their help with the histochemistry; S. van der Elst for assistance with the fluorescence-activated cell sorting experiments; S. Joosten, J. Heuvelmans, and P. Krieken for technical assistance; the Hubrecht Imaging Center for imaging support; and H. Clevers for providing the Lgr5-GFP and confetti mice. Funding: This work was supported by VIDI fellowship 91710330 (J.v.R. and E.B.) from the Dutch Organization of Scientific Research (NWO), grants from the Dutch Cancer Society [L.R. (2009-4621) and E.J.A.S. (2009-4367)], an equipment grant (175.010.2007.00) from the NWO, and grants from the Dutch Diabetes Research Foundation (E.J.P.d.K.) and the Diabetes Research Foundation Netherlands (C.J.M.L., L.v.G., and E.J.P.d.K.). Author contributions: L.R. performed most of the experiments, analyzed the data, and wrote the manuscript. E.J.A.S. performed the experiments, analyzed the data, and contributed to the writing of the manuscript. C.G., N.V., and T.N.S. performed and interpreted the imaging immune response experiments in the spleen. C.J.M.L., L.v.G., and E.J.P.d.K. performed and interpreted the imaging experiments of the kidney. E.B. provided mice. A.Z. supported the experiments. D.S. performed stainings. D.A.R. performed the C26 proliferation experiments. R.S. performed stainings and supported the experiments. A.d.G. supported the imaging experiments. I.H.B.R. helped to improve the surgical procedures. O.K. helped to design and supervise the overall project and helped write the manuscript. J.v.R. conceived and initiated the study, designed and supervised the overall project, and wrote the manuscript. All authors had the opportunity to discuss the results and comment on the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All codes of custom-written software described are available upon request from J.v.R.

Stay Connected to Science Translational Medicine

Navigate This Article