Arrest of circulating tumor cells occurs in blood vessels with permissive flow profiles
pN-range adhesion forces favor rapid and stable intravascular adhesion
Flow-mediated endothelial remodeling drives extravasation of tumor cells
Metastatic seeding is driven by cell-intrinsic and environmental cues, yet the contribution of biomechanics is poorly known. We aim to elucidate the impact of blood flow on the arrest and the extravasation of circulating tumor cells (CTCs) in vivo. Using the zebrafish embryo, we show that arrest of CTCs occurs in vessels with favorable flow profiles where flow forces control the adhesion efficacy of CTCs to the endothelium. We biophysically identified the threshold values of flow and adhesion forces allowing successful arrest of CTCs. In addition, flow forces fine-tune tumor cell extravasation by impairing the remodeling properties of the endothelium. Importantly, we also observe endothelial remodeling at arrest sites of CTCs in mouse brain capillaries. Finally, we observed that human supratentorial brain metastases preferably develop in areas with low perfusion. These results demonstrate that hemodynamic profiles at metastatic sites regulate key steps of extravasation preceding metastatic outgrowth.
Metastatic progression is a complex process resulting in the formation of lethal secondary tumors at distance from their origin (Nguyen et al., 2009). Metastatic cancer cells disseminate very efficiently throughout the body upon intravasation in the blood circulation. Recent work on breast cancer suggests that about 80% of metastases originate from early disseminated cancer cells (Harper et al., 2016, Hosseini et al., 2016). Once in the blood stream, circulating tumor cells (CTCs) may find a location favoring arrest and stable adhesion before extravasating and avoiding the hostile shear forces (Valastyan and Weinberg, 2011, Regmi et al., 2017). After extravasation, metastatic cells either remain dormant (Sosa et al., 2014) or grow successfully into life-threatening secondary tumors (Kienast et al., 2010). Although multiple mechanisms have been postulated for successful extravasation and outgrowth of metastatic cells (Chen et al., 2016b, Headley et al., 2016, Kienast et al., 2010, Strilic et al., 2016), there are only few insights on the role played by mechanical cues encountered in the blood, the main route for hematogenous metastatic dissemination.
Biomechanical forces are known to have a major impact on metastasis progression. For example, tumor cells (TCs) sense and respond to stiffening of the surrounding stroma by increasing their invasive potential (Levental et al., 2009, Mouw et al., 2014, Paszek et al., 2005). High extravascular stress caused by tumor growth (Chauhan et al., 2013, Stylianopoulos et al., 2012) and interstitial fluid pressure (Provenzano et al., 2012) lead to vascular compression that impairs perfusion and eventually promotes tumor progression, immunosuppression, and treatment resistance. Locally, invading TCs need to overcome physical tissue constraints by cellular and nuclear deformability (Harada et al., 2014, Wolf et al., 2013), possibly inducing nuclear envelope rupture and DNA damage (Denais et al., 2016), leading eventually to inheritable genomic instability (Irianto et al., 2016). Overall, while the impact of biomechanics on tumor growth and invasion are mechanistically relatively well understood, the in vivo mechanisms driving survival, arrest, and successful extravasation of CTCs, preceding metastatic growth, remain to be elucidated.
Indeed, very little is known about how CTCs arrest and adhere to the endothelium of small capillaries and leave the blood stream by crossing the vascular wall. While the “seed and soil” concept states that metastasis will occur at sites where the local microenvironment is favorable (Paget, 1989), the “mechanical” concept argues that arrest and metastasis of CTCs occur at sites of optimal flow patterns (Weiss, 1992). CTCs in the blood circulation are subjected to vascular routing (Chambers et al., 2002), collisions and associations with blood cells (Labelle et al., 2014), hemodynamic shear forces (Chang et al., 2008), and physical constraints imposed by the vessel architecture (Headley et al., 2016, Kienast et al., 2010). Only CTCs capable of overcoming or exploiting the deleterious effects of shear forces will eventually arrest, adhere to, and exit the vasculature to form a secondary tumor (Wirtz et al., 2011). Nevertheless, a direct contribution of mechanical cues to the arrest and successful extravasation of CTCs has so far only been poorly studied (Wirtz et al., 2011). Therefore, new in vivo models, where modeling, visualization, and biophysical quantification of the extravasation parameters are easily performed, are of utmost importance for assessing whether biomechanics regulate metastatic extravasation.
Here, we aim to address the direct impact of the blood flow on the arrest and extravasation of CTCs in vivo. We developed an original experimental approach to measure and modulate blood flow with intravascular injection of CTCs within zebrafish embryos. We observed that blood flow controls the sequential steps of arrest, adhesion, and extravasation of CTCs in vivo. In parallel, using microfluidics and optical tweezers (OT), we identified the critical adhesion force (80 pN) that CTCs require to initiate adhesion to the endothelium, which rapidly stabilizes under shear flow. This value matches the threshold dragging force measured in vivo at extravasation sites. Finally, we used our recently developed intravital correlative light and electronic microscopy (CLEM) (Follain et al., 2017, Karreman et al., 2016a, Karreman et al., 2016b) to identify endothelial remodeling as one of the major extravasation mechanisms in vivo, and that endothelial remodeling scales with flow velocities. Overall our studies demonstrate that blood flow forces at metastatic sites regulate key steps of extravasation preceding metastatic outgrowth.
Arrest and Adhesion of CTCs is Favored by Permissive Flow Velocities
To test the impact of blood flow on the arrest, adhesion, and extravasation of CTCs, we experimentally modeled metastatic seeding in endothelium-labeled zebrafish embryo (Tg(Fli1a:EGFP)) at 2 days post-fertilization by injecting fluorescently labeled TCs in the duct of Cuvier (Figures 1A and 1B ). While metastatic extravasation can be successfully tracked in this model (Stoletov et al., 2010), the zebrafish embryo further allows to combine biophysical characterization and manipulation of blood flow parameters with long-lasting and high-resolution imaging of TCs in vivo. We first quantified and mapped the position of arrested and stably adherent TCs (D2A1) in the zebrafish vasculature and noticed that CTCs preferentially arrested (and extravasated) in the caudal plexus (CP) (Figure 1C). Although arrest and extravasation can be also observed in inter-somitic vessels (ISV) (Figure 1C) and in the brain of zebrafish embryos (Figures 1C, S1A, and S1B; Video S1), the majority of the TCs arrest in the CP. We exploited the highly stereotyped vasculature of this region by compiling >10 embryos and quantitatively identifying a major hotspot of arrest in this region (Figure 1D), which sits between the caudal artery and the venous plexus. This was the case for multiple human (Jimt1, 1675, and A431), mouse (D2A1 and 4T1), and zebrafish (ZMEL) cell lines (Figure S1C). Using fast imaging of the blood flow (100 fps) within the entire zebrafish embryo, combined with PIV (particle image velocimetry) analysis, we observed decreasing flow values in the vascular region that is permissive for CTC arrest (Figures 1E and 1F; Video S2). Accurate dissection of blood flow profiles using PIV analysis showed that flow velocity progressively decreases from the anterior dorsal aorta (DA) (position 1, maximal velocity, vmax = 2,500 μm/s; Figures 1E and 1F), to a minimal flow in its most posterior region (positions 5 to 6, vmax = 500 μm/s), which we named the arterio-venous junction (AVJ). We have shown in the past that blood flow dissipates along the vascular network of the zebrafish embryo (Anton et al., 2013). In addition, the mass conservation implies that ramification of the vessels in the AVJ further contributes to the blood flow decrease. We thus set out to model this phenomenon in silico using a mathematical simulation of the blood flow in the CP (see Supplemental Information). Simulation experiments reproduced the flow drop observed in the most posterior region of the CP (Figures S1D–S1H; Video S3). Also, to determine whether flow velocity can affect the CTC arrest, we developed an in vitro approach to mimic in vivo flow profiles in microfluidic channels previously coated with endothelial cells (ECs) or human umbilical vein endothelial cells (HUVECs) (Figure 1G). We observed that adhesion of CTCs to the endothelial layer was favored by reduced flow profiles (peak velocities of 100–400 μm/s) (Figure 1G), similar to those measured in vivo in the AVJ (Figures 1D and 1E). Using higher-flow profiles that mimic flow values obtained in the anterior DA prevented efficient adhesion of CTCs to the endothelial layer (Figure 1G), as observed in vivo where no adhesion could be observed in the DA between positions 1 and 4 (Figures 1D and 1E). Adhesion efficacy (and forces) of CTCs was not affected by temperature (28°C versus 37°C; Figures S2B and S2C). Taken together, these data suggest that reduced flow profiles are permissive for stable adhesion of CTCs to the endothelium, and that the threshold velocity value for efficient adhesion of CTCs ranges from 400 to 600 μm/s.
Permissive Flow Profiles Promote Stable Adhesion of CTCs to the Endothelium
Adhesion of CTCs to the endothelium is an important feature that precedes their extravasation (Reymond et al., 2013). Furthermore, mechanical constraints imposed by cell size and vessel topology likely favor the initial arrest of CTCs (Cameron et al., 2000, Kienast et al., 2010, Luzzi et al., 1998). We set out to test whether such features also contribute to arrest and adhesion in the zebrafish embryo. Our models of CTCs are composed of multiple human (Jimt1, 1675, A431), mouse (D2A1, 4T1), and zebrafish (ZMEL) cell lines, whose mean diameters range from 10.23 to 5.25 μm (Figure 2A). We validated these models by comparing their values to average diameters of human CTCs isolated from breast and lung cancer patients (ranging from 7 to 29 μm) (Figure 2A). We then accurately measured mean vessel diameters of the CP of the zebrafish embryo (Figure 2B). We noticed that diameters of the highly perfused vessels (red, Figure 2B) each displayed a minimal value that exceeds the maximum size for our CTC models (10.23 μm; Figure 2B). When injected in the embryo, our CTCs mostly arrested in the flow-permissive hotspot (AVJ), suggesting that flow drop significantly affects the arrest of CTCs (Figure 2C). Interestingly, when we correlated the position of arrested CTCs within the vasculature to the vessel size, we observed that 49% of them were located in vessels with diameters <10.23 μm (Figure 2D). These cells could have been trapped in low-size vessels or crawled, upon arrest, into these vascular regions before extravasating. Indeed, we have observed that intravascular CTCs are capable of efficiently crawling on the vessel wall, suggesting that CTCs establish firm adhesions with the endothelium (Video S4). Moreover, stiff 10-μm polystyrene beads circulate continuously when injected in the embryo, suggesting that size restriction, although it does participate, is not a major determinant of cell arrest (Figure S2A). These observations led us to test the role of adhesion molecules in driving successful, flow-dependent arrest of CTCs. Thus, we took advantage of our in vitro microfluidic approach and used the OT technology to trap and stick CTCs to the EC monolayer. Doing so, we identified an average value of 80 pN for the very early adhesion forces (<1 min after adhesion of the CTCs to the endothelial layer) required for the attachment of CTCs to ECs (Figure S2C; Video S5). Interestingly, applying Stoke’s law to measure the correspondence between flow and force intensity, we noted that a value of 80 pN represents an average flow value of 450 μm/s, which agrees with threshold flow values that we measured both in vivo and in vitro (Figure 1). Thus, CTCs very quickly establish adhesion forces that allow them to sustain flow velocities of 450 μm/s.
To test the role of adhesion molecules, we targeted β1 integrins (ITGB1), known to play a central role in tumor metastasis (Seguin et al., 2015) and promote stable adhesion of CTCs to the endothelium before extravasation (Gassmann et al., 2009, Schlesinger and Bendas, 2015). We determined whether compromising stable adhesion mediated by β1 integrins would affect the ability of CTCs to arrest under permissive flows. We depleted ITGB1 in our cells using a small interfering RNA approach (Figure 2E) and assessed their stable adhesion 3 hr post-injection (hpi) using a heatmapping procedure over several embryos. Interestingly, while the adhesion hotspot in the AVJ is conserved, the overall number of adhesion event is significantly reduced (Figure 2E). Similar observations were made in our microfluidic system, where only 20% of ITGB1-depleted CTCs stably adhere to the EC monolayer using the same perfusion parameters (Figure 2E). Comparable values were obtained when adhesion (or detachment) of perfused CTCs to (from) ECs was forced using OT in microfluidic channels (Figure S2D; Video S6), indicating that permissive flow profiles, combined with proper adhesion forces, allow stable adhesion of CTCs to the endothelium.
To further support this hypothesis, we assessed whether adhesion events could be observed in vivo using intravital CLEM (Goetz et al., 2014), which allows to combine live imaging of xenografted CTCs in the zebrafish embryo with electron microscopy (EM). Interestingly, CTCs that were arrested in the zebrafish vasculature only 15 min post-injection (mpi) displayed finger-like adhesive contacts with ECs (Figure 2G). In addition, ECs displayed long protrusions when in contact with arrested CTCs (upper panels, Figure 2G). These observations suggest that integrin-dependent adhesion forces between early arrested CTCs and ECs quickly exceed stripping forces from the blood flow. We aimed to further demonstrate the contribution of early adhesion forces to flow-dependent arrest of CTCs and performed OT in vivo. Although OT can very efficiently trap circulating red blood cells (RBCs) in the vasculature of living zebrafish embryo (Video S7) and detach adhered CTCs from ECs in vitro(Figure S3B), OT was inefficient for detaching arrested CTCs in vivo (Figure 2H; Video S7). The inability to detach >35 arrested CTCs with OT, demonstrates that early adhesion forces rapidly exceed 200 pN (which is the technical limit of our OT setup). These results suggest that low flow forces (∼80 pN) enable the arrest and stable adhesion of CTCs in vivo.
Pharmacological Tuning of Hemodynamic Forces Modulates Pacemaker Activity
Since the arrest of CTCs occurs at the site of permissive flow patterns, we investigated whether tuning flow forces would affect the arrest efficacy. We first tuned the zebrafish pacemaker activity (PMA) and subsequent flow forces using pharmacological treatments. We selected lidocain (Vermot et al., 2009) and isobutylmethylxanthine (IBMX) (Luca et al., 2014) to decrease or increase the (PMA), respectively (Figures 3A and 3G ). Upon treatment, we assessed cardiac PMA and measured an average decrease and increase of 20% for lidocain and IBMX, respectively. Using fast imaging combined with PIV analysis we determined the resulting velocities in three positions of the DA for several embryos (ISV 1, 4, and 8) and observed that lowering or increasing PMA with lidocain and IBMX, respectively, significantly altered flow profiles (Figures 3B and 3H; Video S8). In brief, while lidocain treatments led to lower velocities with longer times under 400 μm/s (Figures 3B and S3A), IBMX significantly increased the maximum velocities of flow pulses and decreased the overall duration of the flow under 400 μm/s (Figures 3H and S3A). We confirmed the impact of the two drugs on flow profiles using in silico3D flow simulation (Figures S3C–S3F; Video S9). We further assessed the impact of tuning PMA on hemodynamic forces by trapping RBCs using OT in the AVJ region in vivo (Figure S3B; Video S10). We measured the forces exerted on trapped RBCs, both at the vessel wall and in its center, and extracted the corresponding flow profiles based on the Poiseuille law for each condition (Figure S3C). While lidocain significantly reduced flow forces in the center of the vessel, IBMX increased flow forces at the vessel wall and in the center (Figures 3C and 3I). Importantly, before taking on experiments aiming to study the behavior of CTCs in distinct flow profiles, we demonstrated that our pharmacological treatments had no impact on the vasculature architecture and permeability (Figures S4A–S4D), on the migratory and adhesive properties of TCs in vitro (Figures S4E–S4G), or on adhesion efficacies in our microfluidic approach (Figure S4I).
See full article..