YAP Enhances Tumor Cell Dissemination by Promoting Intravascular Motility and Reentry into Systemic Circulation

Metastasis comprises a complex cascade of events that culminates in the emergence of new tumors in distant locations (1). Prior research has shown that rate-limiting steps of metastasis can occur while tumor cells travel through the circulatory system, arrest at future sites of metastasis, extravasate, and grow into a new tumor, indicating the importance of understanding of these steps (2, 3).

However, these steps can be challenging to study because they are highly dynamic. While in circulation, metastatic cells can travel at velocities of hundreds of microns per second, encounter a variety of physical and chemical stresses, and engage in transient interactions with a diverse cast of blood cells (4–7). Given the complicated and dynamic nature of these events, intravital imaging is required to fully elucidate them (8).

However, intravital imaging in mice is technically challenging and remains a routine technique in only a limited number of laboratories (9). One system that offers an attractive combination of an in vivo microenvironment with straightforward imaging techniques is the zebrafish embryo (10). Zebrafish embryos have been used to study the latter steps of the metastatic cascade including travel through circulation, arrest, extravasation, and early outgrowth at the metastatic site (11–16). Recently, high temporal and spatial resolution studies in zebrafish have elucidated how blood flow dynamics influence the locations of extravasation and how tissue-specific extravasation influences metastatic tropism (17, 18).

We used the zebrafish system to test rapidly how genes known to promote metastasis in mice could influence the behavior of tumor cells in circulation. We observed that a Hippo-insensitive form of the oncogene YAP dramatically changed the behavior of tumor cells in circulation. YAP is a transcriptional coactivator that is downstream of the Hippo pathway (19). Increased activity of YAP (or its paralog TAZ) has been seen in almost every human cancer (20). In addition to promoting tumor growth and progression, YAP has been shown to promote metastasis in several tumor types (20–23). However, relatively little is known about how YAP influences the behavior of tumor cells in the circulation (22, 24, 25).

Here, we found that, while control cells remained trapped in the first capillary bed encountered, cells expressing active YAP were able to move through these vessels, which allowed them to continue to travel through systemic circulation and disseminate more widely. This increased ability to move through small vessels appears to be due to enhanced intravascular motility. YAP cells also remained in circulation longer following intravenous injection into mice, suggesting that YAP can also enhance dissemination in a mammalian system. These results suggest a novel mechanism influencing the distribution of tumor cells throughout an animal.

Zebrafish were housed as previously described (26). The flk:dsRed2 zebrafish line was originally developed in the laboratory of Dr. Kenneth Poss (Duke University, Durham, NC) and was a kind gift from Dr. Mehmet Yanik (MIT). The fli1:EGFP zebrafish line was obtained from the Zebrafish International Resource Center (Eugene, Oregon). The flk:dsRed2 and fli1:EGFP lines were crossed into the transparent casper background (a kind gift from Dr. Leonard Zon, Boston Children's Hospital, Boston, MA). Following injection with tumor cells, embryos were maintained at 34°C for the course of experiments. All zebrafish experiments and husbandry were approved by the MIT Committee on Animal Care.

Embryo injections were performed as previously described (26). Embryos were imaged on an A1R-inverted confocal microscope (Nikon) using the resonant scanner. For time-point imaging, Z stacks were acquired with a 7.4-μm step size using a 10× objective. For time-lapse imaging, Z stacks were acquired with 7.4-μm step size using a 10× objective with an additional 1.5× zoom lens for a total magnification of 15×. Whole embryos were imaged using a 4× objective to acquire Z stacks with a 15-μm step size. For time-lapse imaging, Z stacks were acquired every 2 to 3 minutes for 12 hours following injection.

Embryos were mounted for imaging at single time points using a 3D-printed pin tool as previously described (26). For time courses, single embryos were housed in wells in 48-well plates between imaging. For time-lapse imaging, embryos were mounted in 0.8% agarose with 0.02% Tricaine (Sigma) in 24-well glass-bottom plates (Mattek). Embryos were maintained at 34°C for the duration of time-lapse imaging through the use of a heated enclosure.

Prior to intravenous injection into the jugular vein cannula of an un-anesthetized NOD/SCID/IL2Rγ-null mouse (NSG; Jackson Laboratory), confluent A375 empty vector (EV) and YAP-AA cells were harvested by trypsinization for 5 minutes. The trypsin was quenched with serum-containing medium and the cells were washed 3× with PBS and suspended at 25,000 cells per 50 μL in PBS. During the 20- to 30-minute cell-preparation process, mice were connected to an optofluidic cell sorter as previously described (27) for a baseline scan. ZsGreen cells (25,000) of each type (A375 YAP-AA or EV) were injected slowly into the jugular veins of two separate mice. Each mouse's blood was then scanned using the optofluidic cell sorter for the presence of fluorescent events for 3 hours after cell injection. All mouse experiments and husbandry were approved by the MIT Committee on Animal Care.

The A375 and HT-29 cell lines were obtained from ATCC and cultured in DMEM high-glucose medium supplemented with 10% fetal bovine serum (Sigma), L-glutamine (2 mmol/L, Thermo Fisher), and primocin (0.1 mg/mL, Invivogen). HUVECs were grown in EGM medium (Lonza) supplemented with the EGM-2 BulletKit (Lonza). All cell lines were sorted for fluorescent protein expression on a FACSAria cell sorter (BD) to ensure that the entire population was fluorescent. Cells were tested for Mycoplasma contamination using the Lookout Mycoplasma qPCR Detection Kit (Sigma) prior to freezing down stocks. Vials were subsequently thawed and cultured for no more than 1 month for experiments. Cell stocks were made before passage 6. Microfluidics experiments using the constriction device were performed as previously described (28). The microfluidics constriction is 6 μm, which is smaller than average diameter of the brain (9.9 μm) or tail vasculature (16.7 μm). Cells flowing through the device were traveling about twice as fast as cells flowing through the zebrafish vasculature (17, 28).

Statistical analyses were performed using GraphPad Prism (GraphPad Software).

Additional materials and methods (tumor cell burden calculations, adhesion and migration assays, viral transduction and Western blotting) are described in the Supplementary Information.

To study tumor cells in circulation and at the metastatic site, we injected cells directly into the circulation of 2-day-old zebrafish embryos via the Duct of Cuvier (DoC), a large vessel that drains directly into the heart using an established injection protocol (Fig. 1A; Supplementary Movie S1; ref. 11). As has been reported previously, most tumor cells are in the tail following injection, whereas few cells are in the brain of the embryo (17, 18). This low basal level of metastasis combined with previous studies of brain metastasis in embryos (16, 17) suggested to us that the brain might be a good location to look for enhanced metastasis. Metastasis was assayed 4 days post-injection (DPI) to allow cells time to extravasate and proliferate within the brain before imaging (Fig. 1A).

Cells were injected into flk1:dsRed transgenic embryos, which express dsRed in endothelial cells, allowing the vasculature to serve as a landmark. We generated A375 human melanoma and HT-29 human colon cancer cells that expressed histone 2B fused to EGFP (H2B-EGFP) and the far-red fluorescent protein iRFP670 to label the nucleus and cytoplasm, respectively.

Four genes known to promote melanoma metastasis in mice, BMI1, CDCP1, MCAM, YAP, or an empty vector control (21, 29–31), were overexpressed in these cells (Supplementary Fig. S1A) using a lentiviral expression system. BMI1, CDCP1, and MCAM were all wild-type proteins, whereas YAP was a mutated form insensitive to inhibition by the Hippo pathway (YAP S127A, S381A, labeled as YAP-AA in this study; ref. 21).

A375 melanoma cells overexpressing these genes or an empty vector control were assayed for brain metastasis in the zebrafish embryo system. Only YAP-AA led to an increase in brain tumor cell burden, calculated as the percent area of the brain occupied by tumor cell nuclei (see Supplementary Materials and Methods for further details; Fig. 1B and C). Similarly, YAP-AA promoted an increase in overall brain tumor size and number (Supplementary Fig. S1B and S1C). YAP-AA also promoted brain metastasis by the HT-29 colon carcinoma cell line (Fig. 1D and E; Supplementary Fig. S1E). In A375 cells, YAP-AA enhanced brain metastasis out to 8 DPI (Supplementary Fig. S1D). However, by this time, there was significant mortality, so further experiments stopped at 4 DPI. Overexpression of wild-type YAP in the A375 cell line (Supplementary Fig. S2A) did not enhance metastasis (Supplementary Fig. S2B).

YAP regulates many properties that could influence metastasis, including proliferation, survival, and extravasation (20, 22, 32). To determine how YAP-AA was enhancing metastasis, we first investigated when YAP-AA was promoting metastasis. Embryos were imaged at 10 hours post-injection, 1 DPI, 2 DPI, and 4 DPI (Fig. 2A). At the earliest time point, 10 hours post-injection (HPI), there was already a significant difference between YAP-AA and EV control brain tumor cell burden (Fig. 2B). It also appeared that the magnitude of the difference in brain tumor cell burden between control and YAP-AA cells did not change over time. To confirm that the effect of YAP-AA on metastasis occurred in the first 10 hours, the data for each fish were normalized to the 10-hour time point. When the data were analyzed in this way, there was no time-dependent difference between YAP-AA and control cells, indicating that YAP-AA's enhancement of metastasis was occurring within 10 hours of injection (Fig. 2C). YAP-AA also enhanced metastasis within the first 10 hours following injection in HT-29 cells (Fig. 2D).

Following injection, there is often a bolus of cells at the injection site that could shed cells into circulation over time. Therefore, one possibility was that YAP-AA was promoting more cells to enter circulation. However, the size of the injection site bolus was not significantly different between EV and YAP-AA cells at 10 HPI, suggesting that YAP-AA was not promoting entry into circulation (Supplementary Fig. S3A and S3B).

Additionally, we compared the total disseminated tumor cell burden in EV- and YAP-AA–injected embryos. Following entry into circulation, most tumor cells lodge in the brain or tail of the embryo (Fig. 2E). Therefore, the tumor cell burdens in the brain and tail were determined at 2 HPI and 10 HPI. At 2 HPI, there was no difference between EV and YAP-AA in tumor cell burden in the brain. Consistent with previous experiments, by 10 HPI, there was a significant difference between EV and YAP-AA tumor cell burden in the brain (Fig. 2E). In the tail, there was a slight increase in the tumor cell burden between EV and YAP-AA injected fish by 10 HPI. However, this difference was not significant (Fig. 2E). When the sum of the tumor cell burden in the brain and tail (representing the overwhelming majority of disseminated tumor cells in the fish) was calculated, there was a slight increase in the total disseminated tumor cell burden in the YAP-AA embryos (Fig. 2F). However, this difference was not statistically significant, leading us to believe that enhanced entry into circulation could not explain the marked enhancement of brain metastasis observed. Rather, YAP-AA seemed to be primarily enhancing brain metastasis.

Given that YAP-AA appeared to be specifically enhancing brain metastasis, we hypothesized that YAP-AA might be affecting arrest in the brain, extravasation in the brain, or survival in circulation. To study these processes, we took advantage of the ability to perform time-lapse imaging in living embryos. The time that cells were seen to spend in the same spot within the brain vasculature was used as a proxy for arrest. When we compared the time that individual EV and YAP-AA cells spend arrested in the same location, there was no significant difference, suggesting that YAP-AA was not enhancing arrest (Supplementary Fig. S4A). YAP-AA also did not increase the fraction of cells that had extravasated by 10 HPI (Supplementary Fig. S4B). We also probed extravasation by looking at the activity of invadopodia, ECM degrading protrusions that are required for extravasation, by performing gelatin degradation assays (33). YAP-AA cells degraded less fluorescent gelatin than did EV control cells (Supplementary Fig. S4C). These results together suggested that YAP-AA was not enhancing extravasation by A375 melanoma cells in this system. These results are in contrast with other tumor types (MDA-MB-231 and 4T1) where differences in brain metastasis in zebrafish embryos have been attributed to enhanced extravasation (16, 17) and to other work where YAP has been shown to promote extravasation (25).

Finally, the survival of cells in circulation was studied. During programmed cell death, the nuclei of dying cells are seen to fragment (34). Using the H2B-EGFP label in our cells, we were able to monitor nuclear fragmentation over time. No significant differences in the fraction of cells seen to undergo nuclear fragmentation were observed between EV and YAP-AA cells (Supplementary Fig. S4D). Another possibility for YAP-AA's enhancement of metastasis is through enhancing proliferation. However, given that mammalian cells take around 24 hours to divide, it seems unlikely that the difference in tumor cell burden that we see at 10 hours could be due to cell division. Collectively, these results indicate that YAP-AA is not promoting brain metastasis by significantly enhancing arrest, extravasation, survival in circulation, or proliferation in this system.

During the collection of the data described above, we observed that there appeared to be more YAP-AA cells arriving in the brain during the hours after injection (Fig. 3A and Supplementary Movies S2 and S3). When the cells in the brain were quantified over time, a steady influx of YAP-AA cells was observed over the first 10 hours, whereas few EV cells were seen to arrive (Fig. 3B). By 2.5 HPI, there was a significant increase in the number of YAP-AA cells in the brain, and this increase between YAP-AA and control cells continued to increase over time (Fig. 3B).

These observations showed that YAP-AA was promoting brain metastasis by increasing the numbers of cells that arrive in the brain in the first 10 hours. Given that YAP-AA only slightly enhanced the total number of cells in circulation, we wondered where these extra cells might be coming from. We took advantage of the small size of zebrafish embryos to image entire embryos to track all the cells in circulation over time. We observed that tumor cells were primarily clustered in the tail soon after injection (Supplementary Movies S4 and S5) as has been previously reported by others (17, 18). The EV control cells primarily remained in the tail during the course of these movies. However, over time, YAP-AA cells were seen moving in the tail and rapidly disappearing (presumably after becoming dislodged and swept away by circulation as has been previously observed in this location; refs. 17, 18). Shortly thereafter, YAP-AA cells were observed to arrive at the brain. Given these observations and the observations that (1) YAP-AA only slightly increased the total number of cells in circulation and (2) more YAP-AA cells arrive in the brain than control cells, we hypothesized that YAP-AA was affecting where cells end up in the animal by causing cells to leave the tail and reenter systemic circulation.

To determine whether YAP-AA was causing tumor cells to leave the tail, the tails of embryos were imaged every 2 minutes for 12 hours following injection. While EV control cells mostly remained stationary following arrest, YAP-AA cells moved around within the tail vasculature and eventually disappeared (Fig. 3C and D; Supplementary Movies S6 and S7).

We next sought to determine whether the cells seen disappearing from the tail were traveling to the brain using the photoconvertible protein Dendra2, which converts from red to green fluorescence upon intense illumination with a 405-nm laser (photoconverted signal is false-colored yellow in Fig. 4). A375 YAP-AA or EV control cells were engineered to express Dendra2 and iRFP670. iRFP670 is constitutively fluorescent and served as a control to label all tumor cells (false-colored green in Fig. 4). These cells were then injected into Fli1:EGFP embryos, which have fluorescent vasculature (false-colored magenta in Fig. 4).

Cells lodged in the tail were photoconverted at 2 HPI (Fig. 4A). If YAP-AA stimulates tumor cells to leave the tail and travel to the brain, there should be more photoconverted YAP-AA cells than photoconverted EV cells arriving in the brain by 10 HPI. As a control, we first confirmed that only the cells in the tail were photoconverted at the 2 HPI time point immediately following conversion (Fig. 4B). At 10 HPI, there were more photoconverted YAP-AA cells in the brains than photoconverted EV control cells (Fig. 4C and D). The fraction of photoconverted cells in the brain at 10 HPI was also significantly increased in embryos injected with YAP-AA cells (Fig. 4D). These results fit our hypothesis that YAP-AA allows tumor cells to reenter systemic circulation following initial arrest in the caudal capillary plexus.

We next studied the dynamics of photoconverted cells in the tail vasculature. If YAP-AA is causing tumor cells to leave the tail, one would expect the photoconverted YAP-AA tumor cell burden in the tail to decrease over time while the EV photoconverted tumor cell burden in the tail would remain about the same. Indeed this is what was observed (Fig. 4E and F); at 10 HPI, there were many YAP-AA cells in the tail that were not photoconverted while almost all the EV control cells in the tail at this time point were still photoconverted (Fig. 4G and H). When the ratio of the converted to unconverted tumor cell burden in the tail was calculated, it was similar at the 2 HPI time point for both EV and YAP-AA cells (Fig. 4H). This ratio is not equal to 1 because photoconverted Dendra2 is dimmer than iRFP670. By 10 hours, this ratio had dropped for the YAP-AA cells while remaining essentially unchanged for EV control cells, again indicating that unconverted YAP-AA cells are arriving in the tail after photoconversion (Fig. 4H). This observation fits the model that YAP-AA cells are constantly trafficking through the animal, becoming transiently entrapped in a vascular bed and eventually returning to systemic circulation. This leads to the question of whether YAP-AA cells are also leaving the brain vasculature.

Although some cells were observed leaving the brain, there were also areas where cells remained sequestered for the duration of our imaging (Supplementary Movies S2 and S3). As the vessels in the brain are narrower on average than the ones in the tail (17), we suspect that this difference may account for tumor cells remaining sequestered in the brain but escaping from the tail. One other explanation for YAP-AA cells escaping the tail would be that they are less able to extravasate in the tail than EV control cells. However, over the 10-hour period of interest, most cells are intravascular in both the EV and YAP-AA conditions (Supplementary Movies S6 and S7).

Once it became apparent that YAP-AA was enhancing tumor cell transit through the tail vasculature and that these cells could then travel to the brain, we investigated how YAP-AA could be facilitating transit through the tail vasculature. We hypothesized that YAP-AA might be doing this through decreased cell–cell adhesion, enhanced deformability, or enhancing active migration. However, YAP-AA cells were more adhesive to human endothelial cells in vitro (Fig. 5A). YAP-AA cells also formed larger aggregates in a homotypic adhesion assay, suggesting that they are more self-adhesive than control cells (Fig. 5B). These two results combined suggested that YAP was not decreasing cell–cell adhesion.

We next tested whether YAP might be making cells more deformable, and therefore better able to squeeze through narrow vessels, by measuring the time tumor cells took to pass through a 6-μm constriction (passage time) in a previously described microfluidic device (28). Because tumor cells pass through this constriction in about a second, the passage times reflect a cell's intrinsic deformability as there is not enough time to engage any active processes. The passage time of YAP-AA cells was 50% longer than that of EV control cells, indicating that YAP-AA cells were worse at moving through a constriction than control cells (Fig. 5C). We believe that this increase in passage time may be due to YAP-AA cells being slightly larger than EV control cells (Fig. 5D). To control for the fact that the two cell types were run through the device separately, the transit time through a region without a constriction was analyzed; EV and YAP-AA cells of the same size had the same transit times, indicating identical running conditions (Supplementary Fig. S5A).

Given that YAP was not decreasing adhesion or enhancing deformability, we next investigated the possibility of active migration within the vasculature as it has been reported previously that tumor cells can actively migrate within the vasculature in the tail of 2-day-old zebrafish embryos (11). We first confirmed that YAP-AA enhanced migration in vitro in both the A375 and HT-29 cell lines in transwell migration assays, see Supplementary Methods (Fig. 5E). We then tested whether YAP-AA was promoting migration in vivo by performing high-speed imaging of the tumor cells in the tail. We observed that almost all the EV control cells remained rounded and in the same spot during the course of these movies (Fig. 5F and G; Supplementary Movie S8). However, the YAP-AA cells dynamically extended protrusions and moved within the vessels (Fig. 5F and G; Supplementary Movie S9). This movement does not appear to be only a passive process because it still occurs when flow is blocked by an upstream tumor cell. When the movies of tumor cells in the brain were observed (Supplementary Movies S2 and S3), YAP-AA cells also appear to migrate in the smaller vessels of the head, whereas EV cells remain in the same spot over time, suggesting that this migration is dependent on YAP and not on the architecture of the vasculature.

Collectively, our results from zebrafish suggest that YAP-AA can enhance tumor cell dissemination by promoting active migration through narrow capillary beds that entrapped control cells. This migration then allowed YAP-AA cells to reenter systemic circulation and seed additional, downstream organs.

We next sought to determine how YAP might be driving this enhanced dissemination at the molecular level. YAP is a transcriptional regulator, yet it lacks a DNA-binding domain so it requires interactions with partner transcription factors to regulate gene expression (19). Some of these partner transcription factors have been shown to have roles in metastasis, such as the TEADs, SMADs, and β-catenin (35). YAP's promotion of metastasis has been shown to depend on interactions with different transcription factors in different contexts (21, 36). YAP can also function to regulate genes independent from transcription (37). We therefore assayed whether the promotion of metastasis in this system was dependent on YAP's ability to interact with the TEAD family of transcription factors, with other transcription factors through its WW domains, or through its transactivation domain. These possibilities were tested by overexpressing YAP with the S94A mutation, which disrupts TEAD binding (YAP-AA-S94A), mutations that disrupt binding to the WW domains (W199F, W258F: YAP-AA-WW), or a deletion of the transactivation domain that abolishes transcriptional regulatory activity (YAP-A-TA; Fig. 6A; Supplementary Fig. S6A). The YAP-AA-WW construct still promoted brain metastasis while YAP-AA-S94A and the YAP-A-TA mutants did not, indicating that YAP's ability to interact with the TEAD family of transcription factors and to regulate transcription were both required to enhance metastasis in this context while interactions mediated through the WW domains were dispensable (Fig. 6A).

YAP can regulate a large number of genes so we next were interested in narrowing down which YAP target genes may be promoting intravascular migration. Among the genes regulated by YAP are extracellular signaling molecules that could play a role in regulating motility. We had also observed in previous work that soluble signaling factors can influence the behavior of nearby cells within the vasculature of zebrafish embryos (14). We therefore tested whether YAP-AA's promotion of metastasis was a cell-autonomous process. Control A375 cells or YAP-AA cells were made to express cerulean (cyan) or iRFP670 (yellow), respectively. Control and YAP-AA cells were injected into zebrafish embryos separately or mixed together in a 1:1 ratio. When coinjected, the YAP-AA cells showed enhanced brain metastasis by 4 DPI (Fig. 6B). Coinjection did not enhance EV cell metastasis, indicating that YAP's enhancement of metastasis was confined to the YAP-AA cells (Fig. 6B). These results suggest that YAP is promoting dissemination through a cell-autonomous mechanism that is not dependent on extracellular signaling molecules.

We next sought to examine whether the results from zebrafish could be replicated in a mammalian system. We hypothesized that, if YAP-AA allowed tumor cells to travel through the first capillary bed encountered in mice, then YAP-AA cells should remain in circulation longer in a mouse than control cells following intravenous injection. We used a recently described system for studying the circulation dynamics of circulating tumor cells (CTC) in living mice to test this hypothesis (27). In this system, a catheter routes blood from the carotid artery through a custom cytometer and then returns it to the mouse via another catheter in the jugular vein (Fig. 6C). Using this system, the number of fluorescently labeled CTCs in blood can be tracked over time following a bolus injection via the jugular vein catheter. We observed significantly more YAP-AA cells in circulation over time compared with control cells (Fig. 6D). Furthermore, although the number of EV CTCs quickly drops off, the number of YAP-AA cells initially drops but slowly increases between 25 and 175 minutes post-injection. This trend is consistent with our zebrafish results, where YAP-AA cells initially becoming lodged in small capillaries and slowly reenter systemic circulation over time. These results show that YAP can increase the time tumor cells remain in circulation in a mammalian system. This increase in circulation time could allow these tumor cells to disseminate more widely throughout the animal than control cells that are mostly trapped in the earliest capillary beds they encounter.

Our data suggest that YAP can induce tumor cells that have arrested in small capillaries to migrate within these vessels to points where they can reenter systemic circulation and travel to distant organs. This observation represents a potential novel mechanism through which a gene can affect the distribution of disseminated tumor cells within an animal. This ability could conceivably then increase the fraction of disseminated tumor cells that can form metastases by allowing tumor cells to leave suboptimal metastatic sites and travel to more permissive ones.

It is often implicitly assumed that metastasizing tumor cells take a direct route from the primary tumor, through the circulatory system, to the metastatic site. Arrest in the capillaries of a distant organ is seen as the end of their trip, with cells either extravasating or dying. However, our results, and the results of others (2, 38, 39), suggest that tumor cells might take a more circuitous route. In this model, arrest within the vasculature is not an endpoint but may be a transient event that an individual tumor cell could encounter multiple times, sampling different sites for suitability to establish a metastasis.

Intravital imaging studies in mice have observed that the arrest of individual tumor cells can be dynamic, with tumor cells often arresting temporarily before being carried along by blood flow some time later (12, 26, 40). Additionally, early studies, which tracked tumor cells in circulation in mice, suggested that cells seen departing following stable arrest in one organ can travel to other organs over the course of a few hours (2, 41). At longer time scales, experiments show that tumor cell transit in animal models is more complex than just a linear stream from primary tumor to metastatic site. Instead, tumor cells can even metastasize from one primary tumor to another contralateral tumor or from a metastasis back to the primary tumor in a process called reseeding (39).

The results of our experiments in mice, that YAP-AA cells remain in circulation longer than EV control cells, are consistent with activation of YAP allowing tumor cells to travel through the first capillary bed encountered. As the tumor cells in our experiments were injected intravenously into the jugular vein, the first capillary bed they encountered would be the lungs. Therefore, the fact that there were many more YAP-AA cells in circulation over time suggested that the YAP-AA cells were able to move through the capillaries in the lung to return to systemic circulation (Fig. 6D). Furthermore, the slow increase in YAP-AA CTC counts over time is concordant with the results in zebrafish embryos where initially all YAP-AA cells are trapped in the tail vasculature and return to circulation over time (Supplementary Movie S5).

Additional support for a dynamic model of arrest in the vasculature can be adduced from our result that YAP-AA cells are more adhesive to endothelial cells than control cells (Fig. 5A). If arrest were entirely a passive process, then the more adhesive YAP-AA cells would be expected to be rapidly removed from circulation in our mouse experiments. The fact that YAP-AA cells circulate for much longer than control cells in Fig. 6D suggests that arrest is more complex than just passive adhesion to the endothelium.

One outstanding question is whether YAP-AA cells would form more tumors in organs downstream of the lungs in mice such as the liver. We suspect that this would be the case given that overexpressing activated YAP has previously been shown to greatly enhance the metastatic potential of poorly metastatic tumor cells and even primary cells in the lung (21). However, YAP can also influence the seeding and growth of metastases by regulating tumor cell survival and proliferation (22), which would make any increase in metastasis seen in a long-term experiment difficult to interpret.

The ability to move through the first capillary bed encountered may also be required to account for the presence of metastasis in some instances. For example, colon cancer frequently metastasizes to the liver, which is the first capillary bed downstream through the circulatory system (42). However, colon cancer can also metastasize to the lungs. Given the layout of the circulatory system and the liver vasculature, it seems likely that in order to reach the lungs, tumor cells would first have to travel through capillaries in the liver (43). Genes that aid in this transit through the liver vasculature could therefore lead to more tumor cells reaching the lung and increase the number of metastases seen there.

A number of observations suggest that YAP-AA is promoting intravascular migration. First, our high-speed movies show YAP-AA cells actively extending protrusions and moving through the vessels in the tail in an elongated state that resembles active migration (Fig. 5F and G). Second, the results from the microfluidic constriction device experiment suggested that the YAP-AA cells were not intrinsically better at squeezing through a channel (Fig. 5C). Third, as has been previously reported, YAP greatly enhanced migration in 2D transwell migration assays (Fig. 5E; ref. 21).

Ours is also not the first report of intravascular migration by tumor cells in zebrafish embryos. For example, MDA-MB-435 melanoma cells were observed to migrate within the vasculature of zebrafish embryos following overexpression of Twist1 (11). Time-lapse imaging found individual tumor cells with a rounded morphology crawling actively within the vasculature. This crawling was confirmed to be an active process as it could occur against the direction of blood flow (11). Additional human and mouse tumor cell lines have been reported to crawl along the lumen of the vasculature in zebrafish embryos (18) as well as chicken embryos (33).

Finally, a caveat to our experiments is that they were performed by overexpressing a constitutively active form of YAP. Ideally, we would also have shown that knocking down YAP in these cells would lead to a decrease in brain metastasis. However, the low baseline metastasis observed (Fig. 1B–E) makes it difficult to observe any decrease in metastasis in this system.

Also, in our experiments, YAP transcriptional activity was high, apparently independent of the state of the Hippo pathway. Given the failure of wild-type YAP to promote brain metastasis in our hands (Supplementary Fig. S2), it seems likely that YAP is normally repressed by the Hippo pathway within the vasculature in our system. However, there are a number of mechanisms that might activate YAP activity within the vasculature in other contexts.

YAP and TAZ have been shown to be activated by fluid shear stress (44, 45), and this activation can promote tumor cell motility (45). Another way YAP could be activated within the vasculature is through the physical force on the nucleus as tumor cells are deformed by being pushed into narrow capillaries (46). It has been observed that tumor cells arrested in the vasculature undergo large deformations of their nuclei as they are pushed into narrow vessels by the blood flow so it seems likely that YAP/TAZ activity could be enhanced in these cells (12, 47). Finally, tumor cells arrested at metastatic sites are known to interact with platelets and neutrophils (7, 48, 49), which have recently been shown to promote YAP/TAZ activity (50). As zebrafish lack circulating thrombocytes (the zebrafish equivalent of platelets) during the time period of our experiments, this mechanism for YAP activation would not have been available (51). Additionally, YAP activity could be constitutively high in some tumor cells as it has been shown that multiple oncogenic signaling pathways, such as Src, can lead to increased YAP activity and YAP-mediated metastasis (22, 23).

In summary, we found that YAP-AA promoted brain metastasis in zebrafish embryos. Through time-lapse imaging, we were able to assess YAP-AA's influence on arrest, extravasation, survival in circulation, and travel through the vasculature to determine how it was promoting metastasis in this system. We found that, while control cells arrested in the first capillary bed encountered and remained trapped there, YAP-AA induced tumor cells to migrate through these vessels and reenter circulation, leading to more widespread dissemination and metastasis formation. Our results in mice were consistent with these results and suggest that YAP can enhance widespread dissemination in mammals as well. This ability to transit through the first capillary bed encountered represents a novel mechanism by which a gene can influence the global dissemination pattern of tumor cells and potentially increase the number of metastases in distant organs.

D.C. Benjamin reports grants from Howard Hughes Medical Institute, Ludwig Center for Molecular Oncology, NIH, and NCI during the conduct of the study. J.H. Kang reports grants from Samsung Scholarship during the conduct of the study. B. Hamza reports grants from Ludwig Center for Molecular Oncology, Koch Institute Support Grant P30 CA14051 from the NCI, NIH grant 1R01 CA184956, and Ludwig Center for Molecular Oncology Graduate Fellowship during the conduct of the study. S.R. Manalis reports grants from Ludwig and NCI during the conduct of the study. R.O. Hynes reports grants from Howard Hughes Medical Institute and Ludwig Center for Molecular Oncology, MIT during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

D.C. Benjamin: Conceptualization, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, project administration, writing–review, and editing. J.H. Kang: Software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing–review, and editing. B. Hamza: Software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing–review, and editing. E.M. King: Validation and investigation. J.M Lamar: Resources, supervision, methodology, writing–review, and editing. S.R. Manalis: Supervision, funding acquisition, methodology, writing–review, and editing. R.O. Hynes: Conceptualization, supervision, funding acquisition, methodology, writing-original draft, writing–review, and editing.

We thank Adam Amsterdam for his guidance on experimental design and zebrafish husbandry. We thank Jess Hebert for his scientific advice, revisions to the manuscript, and assistance with mouse experiments. We also thank Eliza Vasile and Jeff Wyckoff of the Koch Institute Microscopy Core for their guidance on microscopy and assistance with imaging experiments. We thank Chenxi Tian for helpful discussions and assistance with mouse experiments. We also thank Alex Miller and Kelsey DeGouveia for their assistance with in vivo cytometry experiments. This work was supported by the Howard Hughes Medical Institute, the MIT Ludwig Center for Molecular Oncology, and by the NIH (1RO1 CA184956 and 1U54CA217377). D.C. Benjamin was supported by the MIT Biology Department NIH training grant (5-T32-GM007287) and by the Howard Hughes Medical Institute. This work was supported in part by the Koch Institute Support (Core) Grant (P30-CA14051) from the NCI.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.