Dynamic regulation of ROCK in tumor cells controls CXCR4-driven adhesion events

Metastasis is the result of a complex set of events that culminate in tumor cell colonization of distal tissues, resulting in deadly consequences. During the metastatic process, tumor cells acquire the ability to survive and proliferate in a new tissue environment (Steeg, 2006). To access these new tissues, metastatic cells display migratory capacities not present in their normal cell counterparts (Friedl and Wolf, 2003; Wang et al., 2005). One of these migratory behaviors is the ability to be trafficked into secondary organs. Although the mechanisms that govern organ-selectivity of metastasis are multi-faceted, one important mechanism is the recruitment of tumor cells by chemokines (Ben-Baruch, 2008; Dittmar et al., 2008; Gassmann and Haier, 2008). Chemokines are potent regulators of cellular trafficking within the body, particularly in the case of directing hematopoietic cell migration. The chemokine receptor CXCR4 is a potent motomorphogen and is essential not only for hematopoiesis, but also for regeneration of multiple solid tissues (Kucia et al., 2005). The expression of CXCR4 is normally restricted to hematopoietic cells and committed stem cells; however, aberrant expression of this chemokine receptor has been shown to be correlated with tumor metastasis (Balkwill, 2004; Zlotnik, 2006). Significantly, blocking CXCR4 signaling inhibits breast tumor cell metastasis in an intravenous model system (Muller et al., 2001). It is thought that metastatic tumor cells can hijack the CXCR4 trafficking signals needed for normal tissue reorganizing activities, thereby gaining entry to distal organs as potential sites for colonization.

Rho GTPase signals are responsible for many cell migration activities (Heasman and Ridley, 2008; Titus et al., 2005; Wennerberg and Der, 2004). The Ser/Thr kinases, ROCKI and ROCKII (also known as ROCK-1 and ROCK-2), are RhoA effectors that have been associated with metastasis in animal models (Hakuma et al., 2005; Itoh et al., 1999; Lin et al., 2007; Nakajima et al., 2003a; Nakajima et al., 2003b; Narumiya et al., 2009; Ogawa et al., 2007; Takamura et al., 2001; Wong et al., 2009; Xue et al., 2007; Ying et al., 2006). Several lines of investigation implicate ROCKs in the invasive behavior of metastatic tumor cells (Cardone et al., 2005; Itoh et al., 1999; Joshi et al., 2008; Kamai et al., 2003; Kitzing et al., 2007; Li et al., 2006; Wilkinson et al., 2005). Recent studies have demonstrated that RhoA and ROCK promote invasion of matrix in response to CXCL12 (Amine et al., 2009; Azab et al., 2009; Molina-Ortiz et al., 2009). ROCKs also play a role in the regulation of invasive pseudopodia and/or invadopodia in metastatic tumor cells. ROCK activity is required for organization of the matrix surrounding advancing tumor cells to promote invasion in three dimensions (Provenzano et al., 2008) and ROCKs have been shown to be required for an ‘amoeboid’ mode of motility that is used by many metastatic cells to invade through three-dimensional matrix (Rosel et al., 2008; Sahai and Marshall, 2003; Torka et al., 2006; Wyckoff et al., 2006). Thus, one way that ROCKs are thought to contribute to metastasis is by stimulating invasiveness; either by the promoting invasive protrusions, or remodeling the matrix.

In addition to a role in tumor invasiveness, ROCKs are also known to regulate cell migration in response to extracellular cues, including those mediated by chemokines (Moyer et al., 2007; Samaniego et al., 2007; Tan et al., 2006; Vicente-Manzanares et al., 2002). Regulation of integrin adhesion is an important mechanism by which chemokines regulate cellular recruitment. CXCL12, the ligand for CXCR4, is known as a ‘capture’ chemokine because of its ability to induce arrest of circulating hematopoietic cells by the vascular endothelium (Alon and Ley, 2008; Laudanna and Alon, 2006; Ley, 2003). Although tumor cell recruitment is probably more complex than leukocyte recruitment and can involve interactions with platelets and other hematopoietic cells (Borsig, 2008; de Visser et al., 2006; Witz, 2008), tumor cell recruitment shares several features of leukocyte recruitment (Balkwill, 2004; Ben-Baruch, 2008; Dittmar et al., 2008; Miles et al., 2008). Notably, tumor cell trafficking can be directed by chemokines, and direct adhesive interactions between tumor cells and the endothelial ligands including selectins, intercellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion molecule (1VCAM-1) have been shown to contribute to tumor cell arrest (Gassmann and Haier, 2008; Glinsky et al., 2003; Jones et al., 2007; Kobayashi et al., 2007; Zetter, 1993).

Subsequent to arrest, chemokines are further thought to regulate chemotaxis by spatially coordinating integrin adhesions, such that adhesion occurs at the front and de-adhesion occurs at the back end of the migrating cell (Imhof and Aurrand-Lions, 2004; Kinashi, 2007; Worthylake and Burridge, 2001). Integrins cluster together with many associated proteins in the cytoplasm to form adhesion complexes that contain both structural and signaling molecules that regulate adhesion behavior. The highly dynamic nature of these adhesion complexes controls the formation of nascent adhesions and their maturation to more stable integrin adhesions during initial attachment and subsequent cell migration. Thus, CXCR4 has the potential to regulate recruitment through two distinct adhesion mechanisms: initial arrest of circulating tumor cells, and subsequent reorganization of the adhesion complexes to support chemotaxis.

ROCKs are important regulators of integrin adhesions, promoting the development of large integrin adhesion complexes (Riento and Ridley, 2003). Several chemokines, including CXCL12, have been shown to regulate the activity of RhoA and ROCKs; however, their specific effect on ROCK activity is context dependent. In some systems, CXCL12 induces ROCK activity whereas in others the chemokine inhibits ROCK activity (Amine et al., 2009; Arakawa et al., 2003; Moyer et al., 2007; Tan et al., 2006; Vicente-Manzanares et al., 2002). These discrepancies can probably be explained by the use of different cell types, experimental conditions, and the time course examined. CXCL12 is known to induce rapid changes in cell adhesion, cell shape and cell migratory capacities, which correspond to underlying changes in signaling pathways (Kucia et al., 2004). During a dynamic process such as cell migration, it is very likely that the effect of CXCL12 on ROCK activity varies over time and might be further influenced by the particular cellular context. Therefore, CXCL12 regulation of ROCK activity is a potential means of controlling tumor cell adhesive events. However, studies to date have not specifically investigated the role of ROCK activity in CXCL12-induced adhesive events during the early phases of cellular attachment, which is thought to be an important first step in tumor cell recruitment.

To investigate the role of ROCK in CXCR4-driven tumor cell adhesion during recruitment, we overexpressed CXCR4 in MCF7 breast tumor cells and analyzed their behavior in response to adhesion stimulated by CXCL12. To more closely mimic the conditions of a circulating tumor cell, we specifically designed our experiments to examine the transition from suspension to adhesion. Our results show that CXCL12 promotes adhesion in two ways. First, it increases initial cellular attachment to both extracellular matrix and endothelial adhesion molecules. Second, it promotes the maturation of integrin adhesion complexes following attachment. Analysis of the role of ROCK activity in these processes reveals an unexpected biphasic role for ROCK activity during adhesion. Initially, inhibition of ROCK activity is required to allow cellular attachment. Following attachment, re-activation of ROCK is required for maturation of adhesion complexes and subsequent migration. Interestingly, although CXCL12-induced adhesion maturation and migration require ROCK activity, CXCL12 reduces the level of ROCK re-activation. This suggests a model in which CXCL12 promotes adhesion maturation and migration by providing an optimal level of ROCK activity. Finally, we have examined tumor cell attachment under physiological flow conditions, and find that constitutively active ROCK blocks tumor cell attachment to an endothelial monolayer, highlighting the importance of the transient decrease in ROCK activity during tumor cell adhesion.

Chemokines, such as CXCL12, are known to increase leukocyte arrest on the endothelium by stimulating integrin adhesion. In order to identify the role of the CXCR4-CXCL12 axis in tumor cell adhesion, we used cell attachment assays to evaluate whether CXCR4 activation stimulated MCF7 breast cancer cell adhesion. We expressed CXCR4 in MCF7 cells (MCF7-CXCR4) to a level equivalent to metastatic MDA-MB 361 cells (supplementary material Fig. S1). MCF7-CXCR4 cells labeled with CellTracker™ Green were put into suspension, incubated for 5 minutes with or without 10 nM CXCL12, and then plated onto 96-well plates coated with a panel of adhesion molecules for 20 minutes. We limited the time of attachment to 20 minutes to restrict our analysis to early adhesion events as a model for tumor cell attachment. We found that CXCL12 increased the attachment of MCF7-CXCR4 cells to collagen 1 (Fig. 1A). For comparison, the attachment of MCF7-vector-control cells is also shown. We found that overexpression of CXCR4 in MCF7 cells promotes adhesion to collagen 1 even in the absence of CXCL12 stimulation. The binding to collagen 1 was very efficient and was subsequently used to normalize the attachment data and serve as a reference point for comparison between different experiments. We next analyzed binding to laminin and several endothelial cell ligands expected to be important for tumor cell adhesion and found that CXCL12 stimulated adhesion to E-selectin, ICAM-1 and VCAM-1 (between 38% and 60% of plated cells attached depending on the ligand; Fig. 1B). This is an important result because it demonstrates that stimulation of CXCR4 can enhance adhesion to endothelial ligands which might be important for tumor cell migration during metastasis.

Because ROCK is known as a potent promoter of adhesion complexes, we next wanted to determine whether basal and CXCL12-induced early adhesion events were dependent on ROCK activity. We pretreated MCF7-CXCR4 cells with Y-27632, an inhibitor of ROCKI and II. (Note that we are not distinguishing between the two ROCK isoforms because the reagents used do not discriminate between them.) We found that inhibition of ROCK with Y-27632 increased MCF7-CXCR4 cell attachment to collagen 1 (30%) and ICAM-1 (45%). In addition, we found that the ROCK inhibitor increased cell attachment even over the levels observed with CXCL12 stimulation (Fig. 1C). Inhibition of ROCK similarly increased attachment to laminin (32%), VCAM (35%) and E-selectin (50%); supplementary material Fig. S2). Because ROCK activity is associated with the formation of robust integrin adhesions, the finding that ROCK activity was not only dispensable, but actually promoted tumor cell attachment was unexpected.

Existing evidence for integrin adhesions being promoted by ROCK is largely based on studies that monitor the size of individual integrin adhesion complexes as an experimental readout; whereas our initial experiments were specifically designed to analyze attachment of whole cells. Thus, we next examined the effect of CXCL12 stimulation on individual adhesion complexes within single cells. Because we are interested in the initial phase of tumor cell attachment, we used Total internal reflectance fluorescence (TIRF) microscopy to examine the formation of adhesion complexes, marked by GFP-paxillin during the first 20 minutes after cell plating.

In control cells, GFP-paxillin was clearly incorporated into adhesion complexes, visible by TIRF microscopy. By 20 minutes, cells were well spread and the adhesion complexes were distinct and visible, primarily at the cell periphery with a few located on the interior of the cell (Fig. 2A). We also observed that in addition to effects on adhesion complexes, CXCL12 stimulation led to increased cell protrusions and a less symmetric, more elongated morphology (Fig. 2A). CXCL12 stimulation enhanced adhesion complexes as measured by a 79% increase in the number of adhesion complexes, and a 144% increase in the area occupied by adhesions (Fig. 2B,C).

To determine the effect of ROCK on adhesion complexes, we included the ROCK inhibitor, Y-27632, during cell attachment. Inhibition of ROCK led to accumulation of numerous small adhesion complexes that were restricted to a thin band along the cell periphery (Fig. 2A). Although there was a striking effect on the morphology of the adhesions that form with Y-27632, the total area occupied by adhesions was increased by Y-27632 (Fig. 2B), which might explain why cell attachment does not require ROCK activity. The morphology of the adhesions induced by CXCL12 was blocked by Y-27632, indicating that ROCK activity was necessary to form larger adhesions (Fig. 2A). At this early time following adhesion (20 minutes), the cells treated with Y-27632 appeared larger with relatively regular edges; the asymmetry induced by CXCL12 stimulation is lost when ROCK is inhibited. Thus, these findings show that inhibition of ROCK during cell attachment leads to small clusters of GFP-paxillin-containing complexes near the periphery, analogous to the ROCK-independent adhesions formed during membrane extension (Rottner et al., 1999). These observations led us to propose that whereas ROCK is dispensable for CXCL12-driven attachment and nascent adhesion formation, re-activation of ROCK is required for adhesion complex maturation.

To test the idea that the small adhesion complexes that accumulate upon ROCK inhibition are immature adhesion complexes, we used two markers of early and mature adhesion complexes: tyrosine-phosphorylated paxillin [Y(P)-Pax] and zyxin, respectively (Zaidel-Bar et al., 2007b). We found that adhesion complexes, marked by vinculin, are differentially marked by Y(P)-Pax and zyxin following treatment with Y-27632. In untreated, control cells, adhesion complexes formed within the first 20 minutes of attachment are positive for vinculin, Y(P)-Pax and zyxin (Fig. 3A). However, when cell attachment occurs in the presence of Y-27632, the adhesions are smaller (as visualized by vinculin staining) and positive for Y(P)-Pax, but negative for zyxin (Fig. 3B). We obtained a similar result when cells were stimulated by 10 nM CXCL12. The small adhesions that formed in CXCL12 cells treated with the ROCK inhibitor were positive for Y(P)-Pax, but negative for zyxin, indicating that the enhancement of adhesions promoted by CXCL12 requires maturation that is dependent upon ROCK activity. This result highlights the distinct requirements for ROCK activity during CXCL12-promoted adhesion: ROCK activity is dispensable for both initial cellular attachment and formation of nascent adhesions, but ROCK activity is required for adhesion maturation which is probably important for cell migration.

To test whether ROCK activity was required for tumor cell migration in response to CXCL12, we monitored migration by phase contrast microscopy over a 4 hour time course (See Materials and Methods for details). MCF7-CXCR4 cells were plated onto collagen-1-coated MatTek plates in the presence or absence of CXCL12 and images were collected every 2.5 minutes for 4 hours. Fig. 4A and B show the tracks of a representative group of cells (full movies are included in supplementary material Movies 1-4). Measurement of migration by net displacement distance shows that the migration activity of the CXCL12-treated cell population was increased relative to that of the control cell population (P=0.001). The CXCL12-induced increase in migration activity was significantly attenuated in the presence of Y-27632 (P=0.014), demonstrating that ROCK activity is required for efficient cell migration. Interestingly, we found that Y-27632 did not block membrane activity because there were numerous, active membrane extensions (Fig. 4A and supplementary material Movies 3 and 4), but the morphology was abnormal. The most striking effect of CXCL12 stimulation was the increase in the percentage of cells that moved long distances. Fig. 4C shows the distribution of net cell displacement for the populations of cells under each treatment condition. Whereas 19% of the CXCL12-treated cell population migrated >80 μm, none of the control cell population migrated this far. The data also demonstrate that ROCK activity is required for CXCL12-enhanced migration, because none of the cells treated with both CXCL12 and Y-27632 migrated >80 μm.

Because initial CXCL12-stimulated tumor cell attachment is promoted by ROCK inhibition, but subsequent events, such as maturation of adhesion complexes and enhanced cell migration, are blocked by ROCK inhibition, we examined the kinetics of ROCK activity during the transition from suspension to adhesion. To measure ROCK activity, we used phosphorylation of two downstream targets of ROCK signaling, myosin phosphatase target subunit (MYPT) and cofilin, as readouts for ROCK activity during early adhesion events. As shown in Fig. 5A, cell attachment induced a brief but pronounced decrease in ROCK activity within 5 minutes of cell attachment. Following this brief decrease in activity, ROCK signaling was re-activated to basal levels in response to attachment.

It is known that CXCL12 and other chemokines can regulate RhoA and ROCK activity, however, the kinetics of regulation are dependent upon the specific experimental system used (Arakawa et al., 2003; Moyer et al., 2007; Tan et al., 2006; Vicente-Manzanares et al., 2002). To determine how CXCL12 affected the ROCK activity during tumor cell attachment, MCF7-CXCR4 cells were pretreated with 10 nM CXCL12 5 minutes prior to plating onto collagen-1-coated dishes, and the phosphorylation state of MYPT and cofilin were measured as indicators of ROCK activity. We found that CXCL12 decreased the overall levels of ROCK activity, although the pattern of ROCK activity in response to attachment was retained (Fig. 5A). Quantification of the blots is provided in supplementary material Fig. S4A,B. We next measured the levels of GTP-RhoA using an ELISA-based activity assay (Fig. 5B,C). As with ROCK activity, cellular attachment induced a transient decrease in RhoA activity, followed by re-activation. Pretreatment with 10 nM CXCL12 decreased the activity of RhoA at early time points of tumor cell adhesion, mirroring the effect of CXCL12 on ROCK activity. These findings are different from that observed in adherent cell cultures, where CXCL12 has been shown to enhance RhoA and ROCK activity (Amine et al., 2009; Azab et al., 2009; Molina-Ortiz et al., 2009). For comparison, we performed analogous assays in steady state adherent cells and found that CXCL12 increased ROCK activity in this context, consistent with other studies (supplementary material Fig. S5). Thus, the influence of CXCL12 on ROCK activity is dependent upon the specific cellular context, and in newly attaching cells, CXCL12 reduces the activity of ROCK.

Because there is often a reciprocal relationship between ROCK and Rac activity, we also measured the effect of CXCL12 on Rac activation, following cell attachment. We found that 5 minutes after attachment there was a brief, 25% increase in Rac activity that returned to baseline by 10 minutes (supplementary material Fig. S4C). However, 10 nM CXCL12 had no effect on Rac activity under these conditions, suggesting that although initial adhesion is correlated with a transient increase in Rac activity, the ability of CXCL12 to promote tumor cell adhesive events does not occur through modulation of Rac.

It is important to note that phosphorylation of MYPT and cofilin represents the balance of phosphorylation by ROCK and dephosphorylation by cognate phosphatases. The rapid kinetics of changes in ROCK target phosphorylation are an indication of how dynamically ROCK signaling pathways are regulated. Lower MYPT phosphorylation levels suggest that CXCL12 tips the balance more towards dephosphorylation; although preserving the biphasic pattern of ROCK activity in response to adhesion. Even though CXCL12 treatment resulted in less ROCK activity, the reduced activity was sufficient for maturation of adhesions (Fig. 3). Cell migration is dependent upon a balance between integrin adhesion and myosin contractility, such that the most efficient migration occurs when the balance between adhesion and contraction is optimal (Gupton and Waterman-Storer, 2006). Because ROCK influences both integrin adhesion and myosin contractility, the reduced ROCK activity in CXCL12-treated cells might correspond to the optimal levels of ROCK activity needed to enhance cell attachment, adhesion maturation and migration of MCF7-CXCR4 cells.

Together our data show that CXCL12 induces attachment to a wide range of adhesive ligands, including both extracellular matrix and endothelial cell adhesion molecules. The goal of the study was to determine if CXCR4-CXCL12 signaling could influence adhesive activity of tumor cells transitioning from suspension to adhesion. Thus, we measured the adhesion of the MCF7-CXCR4 cells under physiological flow conditions following stimulation with 10 nM CXCL12.

MCF7-CXCR4 cells were pumped through a laminar flow chamber in the presence or absence of CXCL12, as described in the Materials and Methods. We found that the flowing tumor cells attached to the collagen 1 under shear stresses ranging from 5.0 to 40 μN/cm². Based on published data, we selected 20 μN/cm² for subsequent analysis (Liang and Dong, 2008; Ngo et al., 2008). Our data shows that CXCL12 leads to a marked, 2.8-fold (180%) increase in the number of cells attached under flow (Fig. 6A). In fact, the increase is significantly more than what was observed in the static attachment assays (26%; Fig. 1). This is probably because shear stress provides a more stringent environment for attachment, and the requirement for CXCL12 signaling becomes more essential. Inhibition of ROCK activity with Y-27632 did not block adhesion induced by CXCL12, similar to what we observed with the static adhesion assays in Fig. 1. Microscopic analysis of the morphology of the newly adherent cells (Fig. 6B) showed that CXCL12 enhances cell spreading, whereas Y-27632 impairs spreading under flow conditions. Thus, CXCL12 and Y-27632 both promote attachment under flow, but only CXCL12 supports cell spreading.

We have used collagen 1 matrix as an initial surface to measure adhesion under flow conditions because we find it to form a robust surface for MCF7-CXCR4 cell adhesion and it provides an important condition for comparing our static attachment assay results. However, to more closely mimic the conditions for tumor cell recruitment, we performed analogous experiments in which the lower surface of our laminar flow cell was composed of an endothelial cell monolayer. A confluent monolayer of endothelial cells (HUVECs) was stimulated with TNFα for 3 hours prior to the introduction of MCF7-CXCR4 cells (see Materials and Methods for details). In these experiments, the MCF7-CXCR4 tumor cells were fluorescently labeled with CellTracker™ Green prior to the attachment assay, so they could be easily discerned from the endothelial monolayer (Fig. 7B). Fig. 7A shows that stimulation of suspended tumor cells with 10 nM CXCL12 increased the number of tumor cells attached to the activated endothelium by 2.3-fold. Thus, tumor cells expressing CXCR4 have a greatly enhanced chance of adhering to activated endothelium when stimulated with CXCL12. Inhibition of ROCK with Y-27632 also increased attachment to the endothelium by nearly twofold, and did not block CXCL12-induced attachment (Fig. 7A), showing that ROCK activity is not required for initial tumor cell attachment under flow conditions.

The observations that ROCK activity is dispensable for tumor cell attachment, coupled with the decrease in ROCK activity during early adhesion time points, led us to hypothesize that the transient decrease in ROCK activity that occurs following tumor cell adhesion is required for cell attachment. To test the hypothesis that attachment under flow conditions requires the inhibition of ROCK, we expressed a fluorescently tagged constitutively active mutant of ROCK (GFP-CA ROCK) in MCF7-CXCR4 cells and measured the attachment of the active-ROCK-transfected cells as compared with GFP control transfectants under flow conditions. Fig. 6C shows that GFP-CA ROCK expression reduced attachment to collagen 1 by 51%, as compared with GFP controls, and prevented an increase in attachment in response to CXCL12. The morphology of the GFP-CA ROCK cells that did attach to the substratum was relatively round and unspread, even in the presence of CXCL12, demonstrating that too much ROCK activity prevents both attachment and spreading in response to CXCL12 (Fig. 6D).

We then examined attachment to endothelial cell monolayers and found that tumor cells expressing GFP-CA ROCK reduced basal attachment by 27%, as compared with GFP controls. Strikingly, GFP-CA ROCK expression completely blocked attachment in response to CXCL12 (Fig. 7C). These results demonstrate that CXCL12 promotes the binding of flowing tumor cells to an activated endothelial monolayer in a manner that requires inhibition of ROCK. This is an important result because it demonstrates that inhibition of ROCK is required for tumor cell attachment. In other words, not only is ROCK activity unnecessary, but ROCK inactivation must occur to allow initial cellular attachment. The finding that active ROCK potently blocks attachment to endothelial cells in response to CXCL12, coupled with the findings that CXCL12 reduces RhoA and ROCK activity levels, support the model that inhibition of ROCK is a previously unrecognized signaling event during chemokine-promoted recruitment of CXCR4-expressing tumor cells.

To better understand the signaling mechanisms that promote metastasis, we designed a system to examine the role of ROCK signaling in CXCR4-driven breast tumor adhesive events. Interestingly, we found that ROCK plays a biphasic role in the regulation of adhesion during tumor cell attachment. Initially, ROCK inhibition is required for cell attachment; but subsequently, re-activation of ROCK is required for maturation of adhesion complexes and migration (Model, Fig. 8). These two distinct requirements for ROCK activity during adhesive events are reflected in the kinetics of ROCK activity. Specifically, analysis of ROCK activity following adhesion shows two distinct phases: a rapid initial decrease in ROCK activity followed by subsequent re-activation (Fig. 5). Interestingly, CXCL12 reduces the activity levels of ROCK at both phases. This suggests that CXCL12 signaling acts in concert with attachment signals to fine-tune ROCK activity for optimal tumor cell adhesive events, as too much or too little activity blocks tumor cell behavior. To more accurately model the environment of tumor cell recruitment, we performed adhesion assays under conditions of shear stress and examined adhesion to extracellular matrix and endothelial cell monolayers (Figs 6, 7). Under conditions of shear stress, the ability of the chemokine CXCL12 to stimulate tumor cell adhesion and spreading was even more pronounced than in the static adhesion assays (Figs 1, 6). Further analysis of cell attachment under flow conditions showed that expression of active ROCK blocked attachment, demonstrating that not only is ROCK activity dispensable for cellular attachment, but inhibition of activity is required for attachment (Figs 6, 7).

Although many factors are likely to contribute to the recruitment of tumor cells to specific organs during metastasis, one mechanism is that adhesion promoted by CXCL12 plays a fundamental role in the process (Balkwill, 2004; Ben-Baruch, 2008; Dittmar et al., 2008; Miles et al., 2008; Witz, 2008). CXCL12 is a known ‘capture’ chemokine for hematopoietic cell recruitment, and as such functions as a gatekeeper for cellular trafficking into specific tissues. The aberrant trafficking characteristics of malignant tumor cells are often thought to be via appropriation of mechanisms normally used during hematopoietic cell recruitment by the vascular endothelium (Kucia et al., 2005; Miles et al., 2008); although tumor cell recruitment involves other factors, including interactions with hematopoietic cells in the bloodstream (Borsig, 2008; de Visser et al., 2006). Direct adhesive interactions between tumor cells and the endothelial lining of the vasculature have been shown to contribute to tumor cell arrest (Gassmann and Haier, 2008; Gassmann et al., 2009; Glinskii et al., 2005; Glinsky, 2006), and there are multiple examples of tumor cells using endothelial adhesion molecules, including selectins, ICAM-1 and VCAM-1, during metastasis (Kobayashi et al., 2007; Zetter, 1993). Following initial cell attachment, the dynamics of the adhesion complexes are important for subsequent cell migration, and the proteins making up the adhesion complexes can be regulated by chemokine signals to promote migration (Kinashi, 2007). To investigate specific mechanisms by which CXCL12 signaling can promote metastasis, we focused on the regulation of tumor cell adhesive events by CXCL12.

We first investigated how ROCK signaling contributed to CXCR4-promoted tumor cell attachment. To test the requirement for ROCK activity, we measured cell attachment in the presence of the ROCK inhibitor Y-27632. We found that ROCK activity was not required for attachment in response to CXCL12. In complementary experiments, we expressed a constitutively active mutant of ROCK, which markedly reduced cell attachment to both matrix and endothelial monolayers under flow conditions (Figs 6,7). Previous studies have shown that RhoA activity is decreased upon integrin engagement; however, these studies did not measure the activity of specific RhoA effectors nor the effect on cell attachment. Thus, our results extend the previous findings by demonstrating that adhesion triggers a transient decrease in ROCK activity, and identify ROCK inhibition as a previously unrecognized requirement for attachment of tumor cells.

Although inhibition of ROCK was necessary for initial tumor cell attachment, subsequent re-activation of ROCK was required for downstream responses to CXCL12. Paxillin-containing adhesion complexes were enhanced by CXCL12 through a mechanism that required ROCK activity (Fig. 2). Adhesion complexes can contain at least 150 different proteins, involving both structural and signaling components (Zaidel-Bar et al., 2007a). Investigations into regulation of integrin adhesion complexes have shown that several cytoskeletal and signaling systems are coordinated to control adhesion dynamics. Integrin adhesions are regulated by the structure of the matrix outside the cell, as well as actin and microtubule dynamics, myosin II contractility, and signaling modifiers within the cell (Broussard et al., 2008; Gupton and Waterman-Storer, 2006; Webb et al., 2004) (Choi et al., 2008; Even-Ram et al., 2007; Vicente-Manzanares et al., 2007). The interplay between these systems leads to the formation of nascent adhesions in the lamellipodium, followed by maturation of the nascent adhesions into larger, dynamic complexes which support forward movement of the lamella to advance cell motility (Giannone et al., 2007; Gupton and Waterman-Storer, 2006; Vicente-Manzanares et al., 2007). Our results show that ROCK activity is required for the maturation of nascent adhesions in response to CXCL12 and subsequent cell migration (Figs 3,4). The requirement for ROCK activity during adhesion maturation and cell migration is reflected in the activity assays that show ROCK is re-activated following the initial decrease in activity (Fig. 5). Together, our data show that although ROCK activity is inhibited during cell attachment, its activity must return for CXCL12 to regulate adhesion complexes and cell migration.

Our results show that ROCK has a biphasic response to adhesion: a transient decrease, followed by a return to activity levels greater than or equal to baseline. This is similar to the pattern of RhoA activity in response to integrin engagement (Arthur et al., 2000; Ren et al., 1999) where a transient decrease in RhoA activity was shown to be necessary for cell spreading, although the effects on cell attachment were not assessed. Arthur et al. identified activation of p190RhoGAP as a mediator of the decrease in RhoA activity (Arthur et al., 2000). Interestingly, CXCL12 decreases the overall levels of ROCK activity, while retaining the transient decrease pattern induced by adhesion (Fig. 5B). One possibility is that CXCR4 signaling activates tyrosine kinases, such as Src or Arg, that activate p190RhoGAP, leading to inhibition of RhoA-GTP loading and lower ROCK activity (Bradley et al., 2006; Chang et al., 1995; Fincham et al., 1999). In this way, CXCR4 could act in concert with integrin signaling to modulate ROCK activity to regulate tumor cell adhesion events that support motility.

We found that CXCL12-induced migration requires ROCK activity; however, CXCL12 reduces the level of ROCK activity generated by adhesion. This seemingly paradoxical result suggests that CXCL12 leads to an optimal level of ROCK activity to enhance migration (Model, Fig. 8). In the context of current models for leading edge behavior during migration (Choi et al., 2008; Giannone et al., 2007; Gupton and Waterman-Storer, 2006), CXCL12 might fine-tune ROCK activity to regulate cycles of lamellipodial and adhesion dynamics, favoring advance of the lamella and persistent migration. Further studies are designed to probe how CXCR4 regulates lamellipodial and adhesion dynamics, as a means to understanding mechanisms by which modulation of ROCK activity contributes to tumor cell recruitment.

Metastasis is clearly an important health concern and CXCR4 and ROCK have each been identified as contributing factors. ROCK is a multifunctional kinase: it receives input from a wide range of extracellular signals and is responsible for cellular responses including regulation of the actomyosin cytoskeleton, cell growth and survival, and gene transcription. The mechanisms by which extracellular signals direct ROCK to act at the right place and at the right time to achieve the appropriate biological outcome are presently undefined. Likewise, the effects of CXCR4 are pleiotropic. Although ROCK and CXCR4 might impinge on multiple steps of tumor cell metastasis, our studies implicate tumor cell adhesion to the endothelium as an effect of aberrant CXCL12 signaling, and identify ROCK-dependent regulation of cell adhesion events as a possible mechanism by which CXCR4 promotes tumor cell recruitment and metastasis. Continued investigation into how CXCR4 regulates ROCK signaling to promote tumor cell adhesion and recruitment has the potential for identifying target molecules for blocking CXCR4-driven metastasis.

Recombinant ICAM-1, VCAM-1, CXCL12 and TNFα were purchased from R&D Systems (Minneapolis, MN). Anti-phospho-myosin phosphatase (MYPT; Thr696) was purchased from Millipore (Billerica, MA). Anti-phospho-ERK (1:2000) and anti-phospho-cofilin [Ser3, -cofilin, -ERK, actin and -MYPT (all at 1:1000) were obtained from Cell Signaling (Danvers, MA)]. Y-27632, anti-zyxin (1:100) and anti-vinculin (1:400) were obtained from Sigma-Aldrich (St Louis, MO). FITC-labeled phospho-paxillin antibody (Tyr118; used at 1:50 dilution) was purchased from Biosource (Camarillo, CA). CellTracker cell-permeable dye and Lipofectamine 2000 were both purchased from Invitrogen (Carlsbad, CA). Collagen 1 and reduced-growth-factor Matrigel were obtained from BD Biosciences (Bedford, MA). An N-terminal GFP cloning vector was generated by inserting the GFP gene [from EGFP-C1 (Clontech, Mountain View, CA)] into pcDNA3 (Invitrogen) then it was converted to a Destination vector with a conversion kit (Invitrogen) following the manufacturer's instructions (kindly provided by Keith Burridge, UNC Chapel Hill, NC). The resulting nGFP-DEST vector was then used to subclone the constitutively active mutant of ROCKII (pCAG-myc-Δ3-ROCK2 used as a template) as a GFP-fusion protein.

MCF7 breast adenocarcinoma cell lines were routinely cultured in DMEM supplemented with 10% FCS, GlutaMax, insulin, non-essential amino acids, and antibiotics (BD Biosciences). The MCF7-CXCR4 and MCF7-vector control cell lines were generated by lentiviral transduction of the human CXCR4 gene, or empty vector, followed by selection with G418 to generate a stable clonal cell line. HUVEC from pooled donors and EGM2 medium were purchased from Lonza (Basel, Switzerland). Primary cell human umbilical vein endothelial cells (HUVEC) cultures were used between passages 3 and 6 and monolayer cultures obtained by seeding at a density of 2500 cells/cm² such that the cells formed a confluent monolayer upon plating (Worthylake et al., 2001).

Ninety-six-well plates were coated with the indicated basement membranes (all at 10 μg/ml) for 2 hours prior to assay. ECM-coated assay plates were then blocked with blocking buffer [serum free medium (SFM) plus 1% BSA] at 37°C for 30 minutes prior to assay. MCF7-CXCR4 cells were labeled with CellTracker (cell-permeable, fluorescent dye, 1:5000 dilution) for 30 minutes followed by starvation with blocking buffer for 2 hours. Cells were then incubated with or without 10 μM Y-27632 for 10 minutes, and/or 10 nM CXCL12 (5 minutes). Trypsinized cells were seeded in 96-well plates (60,000 cell/well) and allowed to adhere at 37°C for 20 minutes, they were then washed with PBS, and fixed in 3.7% formaldehyde for 15 minutes. Fluorescence was detected using a fluorescence plate reader (Molecular Devices, Sunnyvale, CA; ex_492nm/em_517nm). Adhesion assays were performed three or more times and each data point is the average of three wells per experiment. Background binding to blocked-only wells was subtracted, and the relative level of adhesion determined by normalizing to collagen 1 control binding, which was included in every experiment as a standard.

Flow experiments were performed with the FCS2 closed chamber from Bioptechs (Butler, PA). Microaqueduct slides (Bioptechs) were coated with collagen 1 (10 μg/ml) overnight at 37°C, 5% CO₂. 40 mm coverslips were placed on the microaqueduct slide separated by inside gaskets of 0.1. Shear stress (τ) was calculated using the equation τ=(6Qμ)/(wh²), where Q is the volumeric flow rate, μ is the viscosity of the fluid, w is the slide width, and h is the height of the flow chamber (Rops et al., 2007). 5×10⁵ cells/ml were incubated with or without CXCL12 (10 nM, 5 minutes) or Y-27632 (10 μM) in SFM, which remained in the solution during the flow experiments. In some cases, cells were transfected with nGFP-DEST vector alone, or nGFP-CA ROCK (transfection described above) and used in flow-based attachment assays after 48 hours. MCF7-CXCR4 cells were flowed through the chamber at a shear stress of 2 dynes/cm² for 20 minutes at 37°C, after which images from ten regions of the chamber were taken and the number of attached cells per field were counted manually. Images were acquired with a 10× (NA=0.45) objective on an Olympus IX81 microscope, equipped with a Hamamatsu OrcaER camera, run by Slidebook software.

To assess ROCK activity during adhesion, cells were serum-starved for 2 hours, then detached using 0.05% trypsin, pooled and centrifuged at 800 g, for 5 minutes, and then incubated alone or with 10 nM CXCL12 for 5 minutes at 37°C. Cells were then plated in incubation medium onto collagen-1-coated (10 μg/ml) culture dishes and allowed to adhere for 5-60 minutes at 37°C. The cells were then washed with PBS to remove any unattached cells and the remaining adherent cells were lysed with ice-cold modified RIPA buffer with protease and phosphatase inhibitors (Sigma). Equal amounts of protein were loaded on 10% polyacrylamide gels, transferred to PVDF membranes, blocked for 1 hour with 5% milk-TBS, and incubated with phosphospecific antibodies to MYPT or cofilin.

For analysis of adherent cells (supplementary material Fig. S5) MCF7-CXCR4 cells were plated 2 days prior to the experiment, then serum starved for 2 hours prior to the addition of 10 nM CXCL12 for the indicated times. Following CXCL12 treatment, cells were then washed with PBS and lysed with ice-cold modified RIPA buffer with protease and phosphatase inhibitors. Equal amounts of protein were loaded onto polyacrylamide gels for immunoblot analysis. For analysis of suspension cells (supplementary material Fig. S5) MCF7-CXCR4 cells were suspended as for the attachment assays, incubated for 10 minutes, then 10 nM CXCL12 was added for the indicated times. To demonstrate that the phosphorylation of MYPT and cofilin were dependent upon ROCK activity, we performed the same assay in the presence of Y-27632 (supplementary material Fig. S5). Following incubation, cells were pelleted at 500 g for 5 minutes and lysed with ice-cold modified RIPA buffer with protease and phosphatase inhibitors (Sigma). Equal amounts of protein were loaded onto polyacrylamide gels for immunoblot analysis.

To measure Rho and Rac GTP levels, cells were suspended and then replated as described for the ROCK activity assay (see above) in the presence or absence of 10 nM CXCL12. Rho and Rac activity were measured by G-LISA assay (Cytoskeleton, Boulder, CO), with strict adherence to the manufacturer's guidelines. The G-LISA reports levels of active RhoA and Rac normalized by protein input levels. Additionally, equivalent amounts of cellular lysates were taken from RhoA activity lysates and were analyzed by immunoblotting to ensure RhoA expression did not change during the time frame of the experiment.

MCF7-CXCR4 cells were plated onto collagen-1-coated coverslips (10 μg/ml) and allowed to adhere at 37°C for 20 minutes with 10 nM CXCL12, 10 μM Y-27632. As a control, we also tested an alternative ROCK inhibitor, H1152 (supplementary material Fig. S3). Following the adhesion period, cells were washed with ice-cold PBS and then fixed with 4% paraformaldehyde in PBS. After washing three times with PBS, cells were permeabilized with 0.2% Triton X-100 in PBS, washed once with PBS, and 4% normal goat serum was used as a blocking agent. Fixed cells were then incubated with primary antibodies [anti-phospho-paxillin (1:50), anti-vinculin (1:400) or anti-zyxin (1:100)] for 2 hours, followed by a 1-hour incubation with the appropriate secondary antibody (goat anti-rabbit or anti-mouse Alexa Fluor 488 or Alexa Fluor 594; Jackson ImmunoResearch, West Grove, PA). In some experiments, cells were stained for F-actin with Alexa Fluor 594-phalloidin. Coverslips were then mounted onto glass slides using Prolong AntiFade Gold (Invitrogen). Images were obtained using an Olympus IX81 with a 60× (NA=1.4) oil immersion objective, equipped with a Hamamatusu EM camera, and processed using SlideBook software (Intelligent Imaging Innovations, Denver, CO). Images were further processed using Adobe Photoshop, using only linear adjustments of the signal.

Cells were transfected overnight with GFP-paxillin using Lipofectamine 2000 transfection reagent (Invitrogen) at a ratio of 1:3 (μg/μl) DNA:Lipofectamine according to manufacturer's protocol. The following day, cells were starved in SFM for 2 hours, detached from plates, and replated for 20 minutes on collagen-1-coated Bioptechs Delta T dishes. During attachment and imaging, cells were maintained at 37°C using a Bioptechs Delta T temperature-controlled stage adaptor and heated lid. Images were collected on an Olympus IX71 microscope with a 60× TIRFM objective (NA=1.49). Samples were excited with an argon laser (488 nm) and fluorescent images captured with a Hamamatsu OrcaER digital camera, using Slidebook Software.

MCF7-CXCR4 cells were detached, pretreated with CXCL12 (10 nM, 5 minutes), Y-27632 (10 μM, 10 minutes) or Y-27632 (10 μM, 10 minutes) followed by CXCL12 (10 nM, 5 minutes), and plated onto collagen-1-coated MatTek dishes. Phase-contrast images of newly adherent cells were taken at 2.5-minute intervals beginning 20 minutes after initial plating, and lasting for 4 hours using an Olympus IX71 microscope with a 20× objective (NA=0.87). Adhering cells were maintained at 37°C, 5% CO₂ using a LiveCell Environmental Chamber (NEUE Group, Ontario, NY). Cell position in sequential images was determined using Slidebook software and x-y coordinates of individual cells were plotted with starting points adjusted to (0,0). Net displacement was calculated from the distance from the starting (x,y) position (t=20 minutes) to final (x,y) position (t=4 hour).

MCF7, MCF7-CXCR4 and MDA-MB 361 cell cultures were dissociated, then divided into samples that were fixed or fixed and permeabilized, all steps were performed with BD Biosciences Dissociation, Fixation and Permeabilization solutions for Flow Cytometry analysis following the manufacturer's instructions. Each of these was then incubated with anti-CXCR4 (10 μg/ml MAB171, R and D Systems), or no primary antibody, followed by anti-mouse-FITC (Jackson ImmunoResearch). 1×10⁶ cells/ml were analyzed by flow cytometry using a BD CantoII flow cytometer, and the mean fluorescence intensity value was used for comparison of total CXCR4 expression (fixed and permeabilized samples) and surface CXCR4 expression (fixed only) between the different cell types.

We appreciate all of the input from the members of the Cell Adhesion and Migration Group at LSU Health Sciences Center. In addition, we would like to thank Chris Turner for the GFP-paxillin construct, Stephania Cormier and Dahui You for the flow cytometry analysis, and Don Mercante for the statistical analyses. The project was supported by grant number P20RR020160 from the National Center for Research Resources, grant number P20RR018766 from the National Center for Research Resources, grant number LEQSF(2006-09)-RD-A-16 from the Louisiana Board of Regents, and Seed Grant Funding from the Louisiana Cancer Research Consortium. The content is solely the responsibility of the authors and does not necessarily represent the official views of the granting agencies. Deposited in PMC for release after 12 months.