Urokinase Exerts Antimetastatic Effects by Dissociating Clusters of Circulating Tumor Cells

In primary tumors, the capacity of the majority of cells means that they cannot metastasize into other organs. Only a limited number of primary tumor cells become the circulating tumor cells (CTC) that subsequently invade a patient's blood vessels, leading to the ultimate dissemination of such cells into the secondary organs. An increase of the number of CTCs generally means a greater likelihood of a poor survival outcome. For this reason, CTC detection has emerged as a useful practice to predict the emergence of cancer metastases and describe the progression of the disease (1–3). Interestingly, the presence of CTC clusters as well as single CTCs in the bloodstream has been frequently reported, while it has been confirmed that CTCs form clusters in circulatory areas (4), and particularly around primary tumors (5).

The number of CTC clusters is associated with clinically important factors such as the release of cancer cells from primary tumors and the response to chemotherapy. In addition, in previous studies in which an animal model with Lewis carcinoma and melanoma cells was used, tumor-cell aggregates disproportionately generated more lung colonies than equivalent numbers of single cells after tail-vein injections (6, 7), and the yield of spontaneous metastases was influenced by the release of clumps from fibrosarcomas (7). It is therefore of great interest that the release of cancer clumps is a parameter that is quantitatively related to the overall prognosis of spontaneous metastases.

Molnar and colleagues isolated CTC clusters by using magnetic labeling and suggested that a significant correlation exists between the number of clusters and the survival outcome of patients with colon cancer (4). Studies that used metastatic cancer cell lines revealed that many CTCs exist as clumps in blood vessels, and although the number of clumps varied widely between individuals, it seemed to be highly correlated to the extent of lung colonization (8). Furthermore, the research demonstrated that the number of CTC clusters in the circulatory system can directly predict the degree of disease progression for breast cancer.

Recently, a variety of methods and devices have been developed to detect CTCs (9, 10); for example, microfluidic assay is an effective way to count and sort those CTCs in the peripheral blood (11), while the immunomagnetic cell-separation technique can also be applied to isolate cells of a specific lineage from a variety of other immune-cell types in the blood (12). Aceto and colleagues utilized new technology platforms such as the Herringbone HBCTC-Chip and ^negCTC-iChip to isolate and observe individual CTCs and CTC clusters (13, 14). They revealed that CTC clusters originate from oligoclonal aggregates, which are released from primary tumor cells and are more metastatic than equivalent numbers of single CTCs. Although CTC clusters are rare, their significance means that the cells they are composed of must be managed carefully during cancer treatment. However, even though these studies revealed the importance of CTCs and CTC clusters, the corresponding experiments provided only cross-sectional information on their behavior and properties without detailing the corresponding spatiotemporal dynamics. For example, important questions about the associative dynamics of CTCs that underpin cluster formation or the dissociation of CTCs from such clusters remained largely unresolved. On the basis of the real-time behavior of CTCs and CTC clusters in blood vessels, we believe that the methods for ex vivo–prepared samples may have an internal limitation regarding the dissection of CTCs and CTC clusters.

As CTC clusters, which coexist with single CTCs in the bloodstream of cancer patients, represent a more critical prognostic or predictive factor of the metastatic process than single circulating cells (5), a better understanding of CTC cluster formation is urgently required to treat cancer and protect against metastasis. Live-imaging–based experiments are considered the most robust method for addressing this problem (15, 16). The use of fluorescent proteins and other developments in optical-imaging technology have made it possible to directly observe cancer cells as they spread from their original site (17). To track the motility of fluorescently labeled target cells, confocal systems that operate on a millisecond time scale are particularly useful (18, 19). In this study, we used intravital, confocal fluorescence microscopy to describe the hidden behaviors of CTC clusters in comparison with those of single CTCs. By analyzing the characteristics of CTC clusters, we confirmed a hypothesis about the crucial involvement of CTC clusters in metastasis; furthermore, we investigated the inhibitory effect of urokinase on fibrin in the CTC clusters, thereby uncovering potential pharmacologic routes for the suppression of CTC clustering.

Murine mammalian carcinoma cell lines 4T1, B16, and B16F10 were purchased from Korea Cell Bank. The cells were cultured in a RPMI-1649 medium (Hyclone) supplemented with 10% FBS, 1% penicillin, and streptomycin (10,000 U/mL). To produce cells that stably expressed GFP or RFP, lentiviral GFP and RFP vectors were transduced into 4T1 cells, and 100 μg/mL of G418 was added every other day for 4 weeks to selectively kill the cells without GFP or RFP expression. The colonies were isolated and cultured for another 2 weeks to preserve those cells with a homogeneous genetic background. Paclitaxel and urokinase were purchased from Sigma-Aldrich.

All of the animal experiments were performed in compliance with institutional guidelines and were approved by the subcommittee on research-animal care at Wonkwang University, South Korea. Both flanks of each mouse were injected with 5 × 10⁵ cells of pure GFP4T1 and RFP4T1 cell lines or mixtures of GFP4T1 and RFP4T1 cell lines. The mice were anesthetized with an intraperitoneal injection of ketamine (90 mg per kg of body weight) and xylazine (9 mg per kg of body weight) before imaging. Scissors were used to make small incisions around the tumors. The mice were placed on the heated plate of a motorized XYZ translational stage. For vasculature imaging, tetramethylrhodamine (TAMRA) dextran conjugates (5 μg/μL, 2,000,000 MW; Invitrogen) were injected intravenously; the images were typically acquired within 5 to 40 minutes after the injection. GFP-expressing cells were visualized at an excitation wavelength of 491 nm and detected through a bandpass filter of 502 nm to 537 nm (Semrock). TAMRA- and RFP-expressing cells were imaged at an excitation wavelength of 532 nm and detected using a bandpass filter of 562 nm to 596 nm (Semrock).

Urokinase was treated three times a week at 100 U in 100 μL for 4 to 5 weeks.

After the mice were anesthetized, a cardiac puncture was performed. Between 500 μL and 700 μL of the whole blood was collected. After centrifugation, the pellet was incubated with the RBC lysis buffer. The solution containing CTCs was filtered through a strainer of 40 μm mesh size (Corning Life). The cells that remained on the strainer were resuspended in 500 μL of the RPMI culture medium, whereas the rest of the fraction that flowed through the strainer was gently centrifuged again and resuspended in 1 mL of the RPMI. The GFP- and RFP-expressing cells were counted using fluorescence and bright-field microscopy.

Blood samples from patients with breast cancer were used to confirm the existence of a CTC cluster and the heterogeneity of cells in the cluster. To detect CTC with a high specificity and sensitivity, we administrated hTERT-promoter-coding adenoviral vector into blood samples (20–23). GFP was specifically expressed in the CTCs (Fig. 1A and Supplementary Fig. S1). Also, the coexistence of CD45-positive cells—presumably acting as hematopoietic cells—in the CTC clusters was confirmed (Fig. 2B).

To visualize CTCs in blood vessels and evaluate the meaning of the number of CTCs in peritumoral regions, GFP-expressing B16 and B16F10 cells were injected into both flanks of each of the test mice; the latter line is a more metastatic subline derived from B16. The injected cells gradually led to the emergence of visible tumors. At the 5-week mark following the inoculation of the tumor cells, we used intravital, confocal fluorescence microscopy to look at the peritumoral vessels after first making small incisions in the tumor area (Fig. 1C and D). As it is well known that the number of CTCs correlates to the severity of the disease, we compared the number of CTCs from GFP-expressing B16 and B16F10 cells in similar vessels (50–100 μm in diameter) to verify whether the number of CTCs in the peritumoral regions can also be as clinically meaningful as their number in circulation; there were comparatively more B16F10 CTCs than B16 CTCs in such regions (Supplementary Fig. S2A, left, and B). Furthermore, we sought to determine whether treatment with an anticancer drug can affect the number of CTCs. The application of paclitaxel, which had previously been effective for melanomas (24, 25), twice a week for 2 weeks led to a significant decrease in the number of CTCs (Fig. S2A, right, and B). These observations confirmed that, similar to the number of CTCs in circulation, the number of peritumoral CTCs seemed to accurately reflect the metastatic potential of primary tumors.

Interestingly, we occasionally observed clustered cells in peripheral vessels and the numerous GFP-negative cells seemed to be mingled within the CTC cluster (Fig. 1B). To assess the number of CTCs in clusters as a fraction of the total CTC count, movie records were used to count the cells on a weekly basis (Fig. 2A). Although the number of CTCs fluctuated every week, the average CTC count gradually increased as the tumors grew in size during the 5 weeks after the injection of the B16 and B16F10 cells (n = 5). The number of CTCs found in clusters also increased; however, they were consistently 5- to 10-fold less abundant than the single CTCs (Fig. 2B).

To confirm the number of CTCs that was counted with live confocal imaging in the peritumoral regions, we compared this number with the number of CTCs obtained from the whole blood. Red fluorescent protein (RFP)-expressing 4T1 cells were used for this purpose and the previously described method was again used. To distinguish the number of cluster CTCs from the total CTC count, we filtered 500 μL of blood with a strainer after the lysis of the red blood cells (RBCs; Fig. 2C). Most of the single cells, including the blood cells and individual CTCs, flowed through the filter and remained in the throttle bottle (Fig. 2C). The CTC clusters were predominantly contained and, to a lesser extent, some CTCs and immune cells were also caught on the strainer sieves (Fig. 2D, top). As the CTCs expressed RFP and their size was somewhat larger than those of other peripheral blood monocyte cells (PBMCs), they were easily distinguished from other cells. As a result, the number of CTCs increased in correspondence with the tumor size and the CTC clusters were much less numerous compared with the single CTCs. We found that the ratio between the clusters and single CTCs was in the range of 1/5 to 1/10; that is, similar to that observed in the in vivo confocal microscopy experiments (Fig. 2E). Considering the data illustrated in Fig. 2, we concluded that the number of CTC clusters detected in the peritumoral region was small but was nonetheless meaningfully correlated to the number of CTC clusters in circulation.

When the CTC clusters were monitored in real time in a 250-μm width area, they flowed more slowly than the single CTCs, and some of them seemed almost static on the vessel surface around the tumors. We sought to make an accurate comparison of the flow-velocity values using both the single and clustered CTCs and found that the average flow speed of the single CTCs was approximately 380 μm/s, which is a little slower than the flow of the GFP-negative PBMCs (∼420 μm/second); the flow rate of the CTC clusters was approximately 50 μm/second (Fig. 3A and B). The similar pattern was observed in an orthotopic animal model with mammary fat pad (Fig. 3C and Supplementary Movie S1). The flow rate of the CTC cluster seems to be 10 times slower than single CTCs (Fig. 3D). The clusters slowly rolled within the vessel lumen while the flowing single CTCs became attached to the CTC clusters that were previously formed (Fig. 3E). Therefore, in a case wherein CTC clusters exhibit a comparatively slow speed, their adhesiveness and attachment to vessel walls may be the cause; occasionally, we also observed the GFP-positive tumor cells detach from the clusters. As we have shown in continuously recorded frames (Fig. 3E and Supplementary Movie S2), the CTC clusters exhibited dynamic compositional changes, as individual CTCs could repeatedly associate with and dissociate from the clusters (Fig. 3E).

To assess differences in the abilities of the single CTCs and CTC clusters to metastasize into secondary organs, their metastatic capacities to invade lung tissue were analyzed in an animal model. First, we selected GFP- and RFP-4T1 cells with similar tumorigenic and metastatic activities (Supplementary Fig. S3). In one experimental group, the GFP- and RFP-4T1 cells were separately injected into the opposite flanks of each of the test mice (Fig. 4A and B). In another group, mixtures of GFP-4T1 and RFP-4T1 cells were injected into both flanks, with each side receiving approximately the same number of cells (Fig. 4C and D). Five weeks after the inoculation, the mice's lungs were dissected and analyzed (Fig. 4E). We found comparable numbers of metastasized nodules in the experimental groups, implying the similar metastatic potentials of GFP- and RFP-4T1 cells (Fig. 4E). The nodules from each group were examined using a confocal microscope to observe the composition of the metastasized cells. We dissected 100 nodules from 7 mice in each group. Eleven to 16 nodules from each mouse were randomly selected based on the nodule size. When we observed each nodule under a confocal microscope, the nodules could be divided into the following three groups according to their fluorescent colors: nodules expressing exclusively GFP or RFP, and nodules simultaneously expressing both fluorescent proteins. The number of nodules in each category was counted and the totals were compared between the experimental groups (Supplementary Table S1). In the mice that received separate injections of GFP- and RFP-tagged cells, 46% of the nodules expressed only GFP and another 50% expressed only RFP; the remaining 4% of nodules contained a mixture of GFP and RFP colors (Fig. 4G and H, left bar). In the mice that were injected with mixtures of GFP- and RFP-tagged cells, the fractions of single-color nodules were slightly decreased compared with the mice that received separate injections of pure-cell populations. In the mixture-injected mice, the GFP- and RFP-containing nodules were measured at 43% and 45%, respectively. Consequently, the fraction of nodules containing both GFP- and RFP-positive cells increased to 12% (Fig. 4H). A difference of 8% in the fraction of nodules with mixed GFP- and RFP-positive cells was therefore identified between the first and second experimental groups. It is likely that every nodule of a uniform color was derived from a single site that the CTCs initially infiltrated; although, according to our observations, 8% of the nodules potentially originated from the mixed CTC clusters. Furthermore, even in those nodules that expressed only GFP or RFP, the same 8% probability regarding the coexistence of same-colored cells from different clusters could be expected, as previously mentioned. The nodules that emerged from the CTC clusters may therefore account for as much as 24% (= 3 × 8%) of all of the nodules that were observed. Our finding, whereby the GFP- and RFP-mixed nodules occurred more frequently in those animals that were injected with mixtures of GFP- and RFP-positive cells, indicates that clusters mainly form around primary tumors and not during circulation.

Fibrin is considered to be an inducer of CTC clustering (26, 27); therefore, we sought to determine whether urokinase, a thrombolytic agent that dissolves fibrin, suppresses the clustering of CTCs in vitro and affects metastasis in vivo. We observed a reduction in the number of CTC clusters on the sieves after an in vitro incubation with 0.2 mg/mL urokinase for 10 minutes (Fig. 5A). In vivo treatment with 100 U of urokinase decreased the number of CTC clusters in the blood of treated mice compared with the numbers in the control, untreated animals, whereas the number of single CTCs was not affected (Fig. 5B and C); in vivo confocal-microscopy–based observations of the areas around the tumors showed the reproduction of this pattern (Fig. 5E and F). Collectively, these results suggest that urokinase directly breaks down CTC clusters into single CTCs.

To confirm the inhibitory effect of urokinase on CTC clustering in vivo, cell mixtures containing 50% GFP-positive and 50% RFP-positive 4T1 cells were injected into both flanks of each of the mice, as described in the experiments that are depicted in Fig. 4C, followed by the administration of urokinase once or twice a week for 3 weeks after the inoculation. The treatment with urokinase did not affect the growth of primary tumors (Fig. 5F); however, the thrombolytic agent seemed to reduce metastasis (Fig. 5H). We used confocal microscopy to analyze the color composition of the metastasized nodules. Each nodule was again assigned to one of the three groups based on the fluorescent colors that it expressed. Interestingly, the population of nodules that comprised both GFP- and RFP-tagged cells decreased from 13% to 5% after the urokinase treatment (Fig. 5I). The change of this proportion was statistically significant (P = 0.035) in accordance with the Fisher exact test; also, the ratio between the nodules that exclusively expressed RFP or GFP was preserved (Fig. 5I). Finally, to determine whether urokinase can improve the survival of animals implanted with cancer cells, we administered the mice with urokinase once or twice a week for 3 weeks after the cell inoculation. The total survival periods of the mice that were treated weekly were prolonged on average by 5 to 6 days in low-dose and by 9 to 10 days in high-dose group, thus the total survival periods of those urokinase-treated mice were improved more significantly (P < 0.01) in comparison with the untreated animal group. As the tumor cells were implanted Week 7 and the death was initiated from Week 12, it means the survival extension is more than 20% considering the diagnosed time point and life span.

Metastasis has fatal consequences for patients with cancer, while CTCs are considered the “seeds” of metastasis. To spread to distant organs, cancer cells need to enter the blood vasculature and become CTCs; this process is considered a critical step in the metastatic process. Furthermore, the flow of CTCs in the bloodstream is the last step before they disseminate into a secondary organ (1–3). Once metastasized, cancer becomes more difficult to treat, so the detection of CTCs can predict clinical output, and may be used in the monitoring of the efficacy of pharmacologic interventions against metastases. In addition, a number of CTCs can form a clump, also referred to as a “cluster” in this manuscript. According to a general understanding, CTC clusters are directly responsible for metastases, as they exhibit a higher metastatic property than individual CTCs. Many tumor cells possess strong pro-coagulant properties that promote the local activation of the coagulation system (28–30). Tumor-mediated stimulation of coagulation has been implicated in the formation of tumor stroma and the development of hematogenous metastasis (30–34).

As illustrated in Fig. 2, we observed the adherence of CTC clusters to blood vessels. It was most likely that the clusters that appeared to contain GFP-negative cells as well as GFP-positive cells were composed of CTCs. Although platelets can be one of the components of CTC clusters, it is plausible that the GFP-negative cells that we observed were actually macrophages rather than platelets, as their diameter was 10 to 20 μm (35), whereas a platelet's diameter is normally under 2 μm (36, 37). Tumor-associated macrophages induce the formation of CTC clusters around blood vessels near the tumor, establishing gradients of the EGF within the tumor environment; this process can also lead to the attraction of tumor cells toward blood vessels (38, 39).

Fibrinogen is likely to play an important role in the emergence of CTC clumps. The majority of solid tumors in humans and experimental animals contain considerable amounts of fibrinogen-related products—mostly cross-linked fibrin—suggesting that fibrin and fibrinogen are important in tumorigenesis and metastasis (40–44). Research has found that fibronectin has also been upregulated in several types of malignant tumors, and its expression positively correlates to an invasive and metastatic phenotype. Low-molecular weight (LMW) heparin directly leads to fibrin lysis, and patients treated with LMW heparin have a significant survival benefit; however, research has still not conclusively verified if the antimetastatic effect of heparins and other antithrombotic agents is clinically beneficial (45, 46). The ability of thrombolytics such as urokinase to break cell aggregates is well documented and this property is sometimes used to treat intravascular coagulation in cancer patients (47). The pharmacologic mechanism-of-action of this drug targets cell-aggregation–promoting fibrin. To decrease the number of CTC clusters, urokinase, which can break down blood clots, was intravenously injected into those mice that had previously developed lung cancer (Fig. 5), and the injections successfully increased the survival rate of those mice with lung cancer metastases; therefore, our findings demonstrate an ample therapeutic potential of the manipulations of CTC clusters that could minimize cancer metastases. Moreover, our data suggest that the antimetastatic effect of coagulation inhibition may partially come from the disassembling of CTC clusters. In a previous study, plakoglobin seemed to be a promising therapeutic drug for inhibiting the formation of CTC clusters (5). As the antimetastatic effect of siRNA against plakoglobin caused an inhibition of cell aggregation in circulation, the direct manipulation of fibrinogen may be another promising method to control CTC clusters, in addition to the targeting of plakoglobin.

Improvements in optical-imaging techniques have now enabled us to obtain far more precise observations of metastasizing cells, and confocal imaging of tumors that express fluorescent proteins enables researchers to conduct detailed observations of cell motility. Although it is not easy to observe a CTC cluster in circulation with a microscope, our intravital imaging reliably detected CTC clusters that emerged near primary tumors. The development of the CTC clusters in the peritumoral region was confirmed by an analysis of the results obtained from those cells expressing GFP, RFP, or a mixture of both fluorescent proteins (Fig. 4). Our calculations showed results that are comparable with those obtained by Aceto and colleagues in radiolabeled cells.

Therefore, while the previous studies on CTCs relied on the isolation of such cells from the blood, our imaging-based observations helped us to directly understand the dynamics of CTCs and the clustering process in blood vessels. This analysis on the CTC cluster with the patients may give a benefit to predict disease progression. And the trial with in vivo or ex vivo screening system that looks for an active intervention into CTC cluster formation may lead a novel drug discovery for the treatment of cancer before it starts disseminating into the secondary organs.

No potential conflicts of interest were disclosed

Conception and design: J.W. Choi, S.H.A. Yun

Development of methodology: J.W. Choi, J.K. Kim, Y.J. Yang, S.H.A. Yun

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.W. Choi, P. Kim

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.W. Choi, J.K. Kim, P. Kim

Writing, review, and/or revision of the manuscript: J.W. Choi, J.K. Kim

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.K. Kim, P. Kim, K.-H. Yoon

Study supervision: J.W. Choi, K.-H. Yoon, S.H.A. Yun

This work was supported by Basic Science Research Program (2014R1A1A1002431) and Global Frontier Project grants (NRF-2013M3A6A4072626), through the National Research Foundation of Korea (NRF), and funded by the Ministry of Science, ICT & Future Planning. This work was also supported by a grant (HI12C0110) from Korean Health Technology R&D Project, Ministry of Health and Welfare, Republic of Korea.

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.