RhoA and ROCK Promote Migration by Limiting Membrane Protrusions*

Previously, we and others have shown that RhoA and ROCK signaling are required for negatively regulating integrin-mediated adhesion and for tail retraction of migrating leukocytes. This study continues our investigation into the molecular mechanisms underlying RhoA/ROCK-regulated integrin adhesion. We show that inhibition of ROCK up-regulates integrin-mediated adhesion, which is accompanied by both increased phosphotyrosine signaling through Pyk-2 and paxillin and inappropriate membrane protrusions. We provide evidence that inhibition of ROCK induces integrin adhesion by promoting remodeling of the actin cytoskeleton. Furthermore, we find that ROCK regulates membrane activity through a pathway involving cofilin. Inhibition of RhoA signaling allows the formation of multiple competing lamellipodia that disrupt productive migration of monocytes. Together, our results show that RhoA/ ROCK signaling promotes migration by restricting integrin activity and membrane protrusions to the leading edge.

Circulating leukocytes respond to infection or injury by dramatically altering cell shape and adhesive properties to facilitate migration from the bloodstream to the affected tissue (1,2). Leukocytes follow biochemical cues to guide the timing and location of activation (3,4). These cues include soluble inflammatory cytokines and chemokines as well as stationary adhesion molecules. Both types of signals initiate immune system responses from resting leukocytes and provide directional information for appropriate migration.
The process of cell migration can be described as the result of coordination between membrane protrusive activity at a leading edge, movement of the cell body, and retraction of the rear of the cell (5,6). New membrane extensions at the leading edge are stabilized by adhesive interactions between integrins and stationary adhesion molecules located on another cell or deposited as an extracellular matrix. Conversely, the retraction of membrane at the rear of the cell requires the disruption of the adhesive interaction either by disengagement between the ligand and receptor or dissociation of the links between the integrin and the internal cytoskeleton (7).
The actin cytoskeleton plays an important role in regulating both cell shape and adhesive properties. Actin polymerization at the leading edge can physically drive membrane protrusion (8). Interactions between the actin cytoskeleton and integrins regulate integrin activity (9,10). The links between the actin cytoskeleton and integrins can either promote or restrain integrin adhesiveness depending on the cell type and environmental context. Early investigations of integrin-actin linkages in stationary fibroblasts demonstrated that actomyosin-dependent integrin clustering was required for strong integrin adhesions (11). More recently, a study of highly motile leukocytes and lymphocytes have revealed that when these cells are circulating, interactions between integrins and cortical actin can restrain integrin activity until the appropriate signals are received (12)(13)(14)(15). Upon activation, the restraining integrin-actin linkage is broken, resulting in increased integrin mobility to allow clustering and the formation of new integrin-actin interactions that promote adhesion and signaling (16,17).
The Rho family of small GTPases regulates many facets of cytoskeletal dynamics that underlie changes in cell shape and adhesion during migration (18 -24). These signaling proteins act as molecular switches that are responsive to a range of extracellular signals that influence cell migration including those transduced by tyrosine kinase receptors, G-protein coupled receptors, and integrin adhesion molecule receptors (25)(26)(27). The targets of Rho GTPase signaling pathways modulate actomyosin contractility through regulation of myosin phosphorylation and actin dynamics by either promoting polymerization or by stimulating depolymerization and severing of existing F-actin filaments (28 -31).
Previously, we have investigated the role of RhoA in regulating cell shape and migration properties. In fibroblasts, we have shown that down-regulation of RhoA by p190RhoGAP is necessary for cell spreading and directed migration into a wound (32). In leukocytes, we have shown that RhoA signaling to a serine/threonine kinase, ROCK, 1 regulates the retraction of the tail of migrating monocytes and negatively influences integrin adhesion at the cell rear (33,34). This role for RhoA and ROCK in negative regulation of integrin adhesion was initially unexpected based on previous findings that RhoA promotes the formation of large integrin-based focal adhesions in fibroblasts. However, these large focal adhesions are a characteristic of stationary cells and are absent from leukocytes. Instead, leukocytes form small focal complexes of tethering and signaling molecules surrounding ligand-engaged integrins. These focal complexes coordinate integrin signaling during migration and are present both in leukocytes and the leading edge of migrating fibroblasts. In both cases, membrane protrusions and their accompanying focal complexes have been shown to be dependent on Rac or Cdc42 and negatively regulated by RhoA and ROCK activity (33)(34)(35)(36)(37)(38).
In our current study, we have further investigated the RhoA/ ROCK signaling pathway in regulating integrin activity. We find that ROCK negatively regulates two integrin-mediated functions: phosphytyrosine signaling and membrane protrusion. We present evidence suggesting that ROCK regulates integrin adhesion and membrane activity by suppressing cytoskeletal remodeling. Finally, we show that a biological consequence of RhoA/ROCK signaling is to promote migration by limiting membrane protrusions to the leading edge.

EXPERIMENTAL PROCEDURES
Cell Culture-THP-1 monocytes were routinely cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and antibiotics. Transfections were carried out with 15 l of LipofectAMINE 2000 (Invitrogen) and 1 g of DNA in 100 l of Dulbecco's modified Eagle's medium in 24-well plates containing 3 ϫ 10 5 cells for 24 h. Primary monocytes were isolated and electroporated as described previously (24).
Plasmids, Proteins, Pharmacological Reagents, and Antibodies-Myc-wild type cofilin, and Myc-S3A cofilin were generous gifts from Drs. Audrey Minden and Ora Bernard. The cofilin sequences were amplified by PCR and subcloned into a PET 28 vector for the generation of His-tagged fusion proteins. Fusion proteins were isolated using a His-Bind kit (Novagen) according to manufacturer's instructions. N17 RAP1a and E63 RAP1a were a kind gift from Dr. Lawrence Quillam and were subsequently subcloned into an enhanced green fluorescent protein N1 vector by Dr. William Arthur. GST and GST-C3 fusion proteins were prepared as described previously (34). Cytochalasin D, latrunculin A, jasplakinolide, and Y-27632 were purchased from Calbiochem. Concentrations of cytochalasin D and latrunculin A are indicated in the text. Jasplakinolide was used either at 1 or 10 M, each concentration giving equivalent results. Y-27632 was used at 10 M. Recombinant human ICAM-1, VCAM, and MCP-1 were purchased from R&D Systems. Human plasma fibronectin was purchased from Invitrogen. Pyk-2, paxillin, and PY-20 antibodies from Transduction Laboratories were used according to manufacturer's instructions. Cofilin antibody was purchased from Cytoskeleton Inc., and the phosphospecific cofilin antibody was a generous gift from Dr. Jim Bamburg.
Cell Adhesion, Morphology, and Migration Assays-Adhesion to 10 g/ml ICAM-1 or VCAM was performed as described previously (34). Cell morphology was assessed by F-actin staining with Alexa 594phalloidin (Molecular Probes) or by phase-contrast microscopy using a Zeiss Axiophot microscope with either a ϫ100 or ϫ63 objective. Images were obtained with a cooled CCD camera and processed in Metamorph Imaging software (Universal Imaging) and Adobe PhotoShop. Scion Image software (NIH) was used to quantitate cell area. For cells plated on fibronectin, ICAMϩVCAM or poly-L-lysine, a minimum of 10 cells from three separate experiments was analyzed. Cells transfected with Myc-S3A cofilin were detected by immunostaining, and those determined to be expressing low levels of S3A cofilin were selected for the measurement of cell area. Data were collected from three separate transfection experiments. Time-lapse videos were recorded with Scion Image software. Images were taken every 20 s for 15 min to record cell behavior at a magnification of ϫ40. Still images were further processed in Adobe PhotoShop. Transwell assays were performed by coating the 8.0-m pore membrane with a solution of 10 g/ml ICAM-1 and 3 g/ml VCAM (using concentrations previously shown to promote monocyte migration) (39), placing 10 5 monocytes in the upper chamber and 10 ng/ml MCP-1 in the lower chamber to induce migration. Primary monocytes were allowed to migrate 1 h, whereas THP-1 cell migration was assessed after 3 h. Cells adherent to the top of the Transwell filter were removed with a cotton swab, and the number of cells attached to the underside of the filter was counted by microscopy.

ROCK Limits Membrane Protrusions-Previous work has
shown that ROCK negatively influences integrin-mediated adhesion in primary monocytes, eosinophils, and neutrophils (33,34,38). To better understand the molecular mechanism underlying this observation, we used the THP-1 monocytic cell line to study the function of ROCK on integrin adhesion, membrane activity, and leukocyte migration. THP-1 cells grow in suspension, but placing them in wells coated with the integrin sub-strates, ICAM-1 or VCAM, induces adhesion that is up-regulated Ͼ2-fold in the presence of the ROCK inhibitor Y-27632 ( Fig. 1). This indicates that integrin adhesion of the THP-1 cell line is regulated in the same manner as primary cells.
Because Y-27632-induced adhesion was correlated with a dramatic increase in cell spreading in primary monocytes, we examined the morphology of THP-1 cells treated with Y-27632. For these experiments, we used F-actin staining to observe the shape of THP-1 cells adherent to different integrin substrates. Fibronectin represents an extracellular matrix material present in wounds that engages several ␤ 1 integrins, whereas ICAM-1 and VCAM are cell adhesion molecules expressed by activated endothelial cells to recruit leukocytes through ␣ L/M ␤ 2 and ␣ 4 ␤ 1 , respectively. THP-1 cells adherent to fibronectin or ICAM ϩ VCAM displayed a rounded appearance, although the cells plated on ICAM ϩ VCAM had more fine filopodial extensions and more distinct punctate structures characteristic of podosomes ( Fig. 2A). To investigate the role of ROCK on monocyte morphology, we allowed the cells to engage the substrates in the presence of Y-27632. The area occupied by THP-1 cells plated under various conditions was measured, and the average results are depicted graphically in Fig. 2B. The inhibition of ROCK augmented cell spreading in response to integrin engagement, particularly when plated on fibronectin. In addition to a simple increase in cell spreading, the F-actin staining in Fig. 2A shows that inhibition of ROCK promoted excessive membrane activity as indicated by F-actin content and numerous lamellipodia around the perimeter of the cell. The upregulated protrusions were dependent on the integrin substrates, because cells adherent to a nonspecific charged surface (poly-L-lysine) were round and flat with no active lamellipodia or membrane ruffling as indicated by the smooth appearance and reduced levels of F-actin content.
These findings prompted us to determine whether integrin signaling was similarly misregulated by the inhibition of ROCK with Y-27632. Because tyrosine kinases are potent mediators of integrin signaling, we examined the pattern of tyrosine kinase signaling in THP-1 cells adherent to integrin substrates. concentration of phosphotryosine signal. The phosphotyrosine content of cells bound to poly-L-lysine was significantly reduced, indicating that the integrin substrates initiated the phosphotyrosine signaling. Thus, the increased membrane protrusions were accompanied by phosphotyrosine signaling triggered by integrin substrates.
To further explore the effect of inhibiting ROCK on tyrosine phosphorylation signaling, we analyzed cell lysates for the pattern of tyrosine phosphorylation by Western blot. Fig. 3B shows that when THP-1 cells adhere to fibronectin, there is an increase in tyrosine-phosphorylated proteins in the region of ϳ100 and ϳ64 -70 kDa. This adhesion-induced phosphotyrosine signaling is further enhanced when the cells are allowed to adhere in the presence of Y-2732. We next sought to determine the identity of specific proteins that may display increased tyrosine phosphorylation when ROCK is inhibited with Y-27632. Based on the molecular mass of the observed increase in phosphotyrosine-containing proteins, we evaluated the phosphorylation status of the tyrosine kinase Pyk-2 (CADTK, Cak␤, and RAFT-TK) and paxillin, which have molecular masses of 110 and 68 kDa, respectively. Again, THP-1 cells were allowed to adhere to fibronectin for 20 min followed by immunoprecipitation of Pyk-2 or paxillin and subjected to analysis of phos-photyrosine content by Western blot. In suspension, Y-27632 decreased Pyk-2 tyrosine phosphorylation, suggesting that integrin-independent signaling to Pyk-2 requires ROCK signaling. We found that the highest level of tyrosine phosphorylation for both Pyk-2 and paxillin occurred in cells adherent to fibronectin in the presence of Y-27632 (Fig. 3C).
Actin Remodeling Is Required for Adhesion and Membrane Protrusions-Having found that inhibition of ROCK with Y-27632 promoted increased integrin adhesion that is accompanied by increases in membrane protrusions and phosphotyrosine signaling, we then investigated the molecular mechanism responsible for increased integrin adhesion. Regulation of integrin avidity is an important mechanism controlling of integrin adhesion in leukocytes (40). It is thought that the round circulating cells maintain integrins inactive by tethering them to the band of cortical actin until appropriate signals are received. Upon activation, integrins are freed from their cytoskeletal constraints and allowed to cluster, forming qualitatively different connections with the actin cytoskeleton for signaling and migration. Evidence both for and against the model of cytoskeletal constraint has been reported, so we performed experiments to test this model with our cell system using two different drugs that destabilize the cytoskeleton, cytochalasin D, and latrunculin A. Fig. 4A shows dose-dependent adhesion of THP-1 cells to either ICAM-1-or VCAM-coated surfaces when the cells are plated in the presence of cytochalasin D or latrunculin A. Our results support the cytoskeletal constraint model because both drugs stimulate adhesion to integrin substrates at low doses. Thus, limited cytoskeletal destabilization promotes adhesion, whereas completely abolishing F-actin does not allow the formation of the subsequent cytoskeletal linkages thought to be required for productive integrin clustering and signaling. This is further supported by the observation that cells plated in the presence of 0.1 M cytochalasin D displayed a well spread morphology, whereas those plated in 1 M cytochalasin D were very small and rounded (Fig. 4B).
We then asked whether the role of Y-27632 in promoting integrin functions could be attributed to a destabilization of the cortical actin network. To address this question, we utilized another pharmacological reagent, jasplakinolide, which stabilizes F-actin structures, thereby preventing actin remodeling. ROCK Signaling Pathways That Regulate Integrin Function-Our next objective was to determine signaling pathways downstream from ROCK that influenced integrin function in a manner dependent upon cytoskeletal remodeling. The small GTPase RAP1 plays an important role in up-regulating integrin adhesion, especially in leukocytes (41)(42)(43)(44). Cross-talk between members of the small GTPase superfamily is a common regulatory mechanism, so we asked whether ROCK might exert its effect through RAP1 by examining the effect of RAP1 mutants on THP-1 cell membrane protrusions. If Y-27632 is activating RAP-1 to stimulate integrin function, transfection with constitutively active RAP1 should mimic the Y-27632 phenotype, whereas a dominant negative RAP1 should block the Y-27632 effect on morphology. Fig. 5 shows that these RAP1 mutants have little effect on the cell morphology with or without Y-27632. In addition, we found that Y-27632 did not increase the levels of GTP-RAP1 when cells were plated on fibronectin (data not shown). These results indicate that RAP1 is not a component of ROCK signaling that modulates integrin function.
Another potentially interesting candidate downstream from ROCK is the LIM kinase to cofilin signaling pathway. ROCK phosphorylates and activates LIM kinase, which phosphorylates and inactivates cofilin. Cofilin both depolymerizes and severs F-actin filaments, thereby facilitating actin remodeling (45). In this model, ROCK activity would lead to phosphorylated inactive cofilin that would tend to stabilize the existing F-actin networks, whereas Y-27632 would lead to de-phosphorylated active cofilin that would promote actin remodeling. However, LIM kinase and cofilin are regulated by additional pathways besides ROCK. To determine the contribution of ROCK signaling in cofilin regulation in our cell system, we first examined the effect of Y-27632 on cofilin phosphorylation status. We used an antibody that specifically recognizes cofilin that is phosphorylated on serine 3 to show that in THP-1 cells plated on fibronectin, there is a significant level of phosphorylated inactive cofilin. When cells are allowed to adhere to fibronectin in the presence of Y-27632, the levels of cofilin phos- phorylation are nearly abolished. This indicates that ROCK plays an important role in regulating the phosphorylation and thus activation, the status of cofilin in response to THP-1 cell adhesion.
We next used an S3A mutant of cofilin that cannot be phosphorylated and inactivated to examine the importance of cofilin activity in cell shape and membrane protrusions. In THP-1 cells transfected with S3A cofilin, we found that low level expression of activated cofilin led to a ϳ40% increase in cell spreading (Fig. 6B). In cells expressing high levels of activated cofilin, F-actin structures are completely abolished, leading to cells with a small rounded morphology (data not shown). This finding is consistent with our previous data using cytochalasin D and latrunculin A that showed strong dose-dependent differences on integrin function. We also examined the effect of activated cofilin on the morphology of primary monocytes. In these experiments, cells were loaded with either wild type or S3A cofilin proteins by electroporation as described previously (34). Previously, we have demonstrated that inhibiting ROCK in primary monocytes increased membrane protrusive activity, and because the cells are highly migratory, ROCK inhibition results in cells with unretracted tails (34). Fig. 6C shows that overexpression of wild type cofilin leads to a small increase in the percentage of cells with unretracted tails (22% in controls, 34% with wild type cofilin), whereas loading the cells with active cofilin results in a greater percentage of cells with unretracted tails (65%). The effect of loading primary monocytes with active cofilin is very similar to the phenotype observed with Y-27632 treatment (80% unretracted tails), albeit less potent. Together, these results indicate that cofilin is inactivated downstream from ROCK, which contributes to the promotion of membrane protrusions and inhibition of tail retraction.
ROCK Limits Membrane Protrusions and Promotes Directed Migration-This study has examined the molecular mechanisms underlying of membrane protrusions induced by inhibition of ROCK. Our next goal was to determine the biological consequence of RhoA/ROCK-regulated membrane protrusions. To study monocyte migration, we used our primary monocyte system in which cells were loaded with C3 protein, a potent inhibitor of RhoA, and subsequently ROCK activity. The loaded monocytes were plated onto coverslips and their migration was monitored by time-lapse video microscopy. Control cells loaded with GST displayed a typical migration pattern in which the membrane at the leading edge protrudes, the cell body moves forward, and the rear of the cell retracts. This results in a cell with a relatively round morphology, reflecting the balance between protrusions at the front of the cells and retraction of membrane at the rear of the cell. To change direction, the monocyte sends out a lamellipodial extension in an alternate direction and the former leading edge stops protruding and retracts (Fig. 7A). In contrast, monocytes loaded with C3 display competing membrane lamellae, pulling the cell in different directions at the same time (Fig. 7B). Membrane activity is not limited to a single leading edge, resulting in an elongated cell with unregulated protrusions along the length of the cell. Thus, RhoA activity is required to suppress membrane protrusions at the rear of a migrating monocyte.
Monocytes in which RhoA or ROCK is inhibited have robust lamellipodial protrusions, a behavior typically thought to promote migration. Yet over time, the C3-loaded monocytes do not appear to be able to migrate any significant distance, despite vigorous short term movements. To directly test whether monocytes require ROCK signaling to undergo productive migration, we used a Transwell migration assay. Primary monocytes (Fig.  8) or THP-1 cells (data not shown) were placed in the upper chamber of a Transwell and allowed to migrate toward the chemokine MCP-1 in the lower chamber. The dividing membrane was coated with the ICAM ϩ VCAM combination to simulate integrin substrates normally used by monocytes for migration. The number of cells that migrated to the lower chamber was counted by microscopy, and the results are presented in Fig. 8. As expected, MCP-1 induced Ͼ10-fold induction of monocyte migration for control cells. When the migration assay was performed in the presence of the ROCK inhibitor Y-27632, migration was reduced to ϳ3-fold. Together, these results demonstrate that RhoA/ROCK signaling is required to suppress protrusions at the rear of migrating monocytes to allow productive migration. DISCUSSION In this study, we have extended results from previous work examining the function of RhoA and ROCK signaling in leukocyte migration. We report that RhoA and ROCK signaling are required to limit membrane protrusions to a single leading edge. When ROCK signaling was inhibited with Y-27632, we found increased integrin adhesion accompanied by increased membrane activity and up-regulated phosphotyrosine signal- ing. We described an important role for cytoskeletal remodeling in regulating integrin adhesion in experiments utilizing either actin depolymerizing or stabilizing agents. An investigation into the signaling pathways that influence cytoskeletal remodeling to regulate integrin activity revealed that both ROCK and cofilin contribute to regulating membrane protrusions. Finally, we showed that in the absence of RhoA/ROCK signaling, monocytes send out competing lamellipodia that interfere with productive migration.
We observed a strong correlation among increased integrin adhesion, increased integrin-dependent phosphotyrosine signaling, and inappropriate lamellipodial protrusions. We propose that the misregulated integrin engagement initiates phosphotyrosine signaling, which promotes membrane protrusion. Evidence for a direct link between integrin signaling and lamellipodia formation comes from our identification of Pyk-2 and paxillin as two proteins whose adhesion-induced tyrosine phosphorylation levels are influenced by ROCK activity. Both Pyk-2 and paxillin have been implicated in promoting protrusive activity in leukocytes and fibroblasts (46 -50). This is further supported by studies from the Cabanas laboratory showing that increased Pyk-2 signaling in neutrophils treated with C3 was downstream from integrin clustering (51). Our data highlight a previously unappreciated role for ROCK signaling in suppressing integrin adhesion, signaling, and membrane protrusions away from the leading edge.
Circulating leukocytes must maintain tight regulation over the state of integrin adhesiveness. Leukocytes inhibit integrin function in the resting state to prevent unwarranted inflammatory responses but quickly activate integrin adhesion in response to infection or injury. A rigid band of cortical actin maintains the round shape of circulating leukocytes, and it has been proposed that this cortical actin constrains integrin mobility, rendering them unable to cluster with other integrins (14,40). Our data support the model for cytoskeletal remodeling as a molecular mechanism that regulates integrin adhesion in leukocytes. We found that low doses of two actin destabilizing agents, cytochalasin D and latrunculin A, promote integrin adhesion, whereas the jasplakinolide, a F-actin stabilizing agent, prevents adhesion.
Despite the pharmacological evidence for the role of actin remodeling in integrin activation, the in vivo signaling pathways that would regulate actin dynamics have not been well characterized. Here, we implicate ROCK in governing actin remodeling, because inhibition of ROCK has the same effect as low doses of actin destabilizing agents on integrin adhesion. In addition, the combination of the ROCK inhibitor with an actin stabilizer had counteracting effects on adhesion and membrane protrusion. Further investigation of signaling pathways downstream from ROCK showed that the actin remodeling protein, cofilin, is involved in regulating membrane protrusions in THP-1 cells and tail retraction in primary monocytes. Cofilin is known to promote actin remodeling by virtue of its actin sev- ering and depolymerizing activity (45). Furthermore, cofilin activity has been implicated in promoting lamellipodial protrusions in response to growth factors (52,53). The effects of expressing non-phosphorylatable, activated cofilin were not as striking as those observed with inhibition of ROCK, which indicates a role for additional signaling molecules. These results indicate that regulation of integrin function by actin remodeling was regulated by ROCK signaling through a pathway involving inactivation of cofilin.
To evaluate the biological consequence of RhoA and ROCK regulation of integrin functions, we analyzed the behavior of primary monocytes in addition to THP-1 cells. Using time-lapse video microscopy, we observed that inhibition of RhoA signaling leads to the formation of multiple competing lamellae. These monocytes appear to be unable to choose a leading edge and do not appear to migrate any appreciable distance, despite robust membrane protrusive activity. We measured the capability of the monocytes to migrate in a Transwell assay, which demonstrated the ROCK dependence of MCP-1-induced migration. This finding is consistent with results reported by Ashida et al. (54) that show that Y-27632 inhibits migration of THP-1 cells. Our results confirm those findings and build upon them by implicating two integrin-dependent functions, phosphotyrosine signaling and membrane protrusions, as molecular mechanisms underlying the block in directed migration.
Combining data from the present studies with that from many different investigators and systems, we propose a model in which spatial regulation of RhoA activity is required for productive migration. In this model, RhoA must be actively inhibited at the leading edge to allow lamellipodial protrusions. This is supported by other studies in fibroblasts that show that the inhibition of RhoA/ROCK signaling with C3, Y-27632, or overexpression of p190RhoGAP promotes cell spreading, protrusion, and polarity (32,36). Furthermore, p190RhoGAP is localized to sites of membrane protrusions and ruffling (55,56). Although RhoA must be inhibited at the leading edge, our model also proposes that RhoA must be activated at the rear of the cell to promote migration. Myosin-based contractility has previously been implicated in the tail retraction of migrating neutrophils (57), and the RhoA/ROCK signaling pathway is known to regulate myosin phosphorylation. However, our previous studies in migrating monocytes did not reveal a major role for myosin contractility in RhoA-dependent tail retraction. Our current data suggest an additional mechanism that RhoA and ROCK are required to suppress lamellipodia formation away from the leading edge and that failure to limit membrane activity inhibits productive migration.