Two independent but synchronized Gβγ subunit–controlled pathways are essential for trailing-edge retraction during macrophage migration

Chemokine-induced directional cell migration is a universal cellular mechanism and plays crucial roles in numerous biological processes, including embryonic development, immune system function, and tissue remodeling and regeneration. During the migration of a stationary cell, the cell polarizes, forms lamellipodia at the leading edge (LE), and triggers the concurrent retraction of the trailing edge (TE). During cell migration governed by inhibitory G protein (Gi)–coupled receptors (GPCRs), G protein βγ (Gβγ) subunits control the LE signaling. Interestingly, TE retraction has been linked to the activation of the small GTPase Ras homolog family member A (RhoA) by the Gα12/13 pathway. However, it is not clear how the activation of Gi-coupled GPCRs at the LE orchestrates the TE retraction in RAW264.7 macrophages. Here, using an optogenetic approach involving an opsin to activate the Gi pathway in defined subcellular regions of RAW cells, we show that in addition to their LE activities, free Gβγ subunits also govern TE retraction by operating two independent, yet synchronized, pathways. The first pathway involves RhoA activation, which prevents dephosphorylation of the myosin light chain, allowing actomyosin contractility to proceed. The second pathway activates phospholipase Cβ and induces myosin light chain phosphorylation to enhance actomyosin contractility through increasing cytosolic calcium. We further show that both of these pathways are essential, and inhibition of either one is sufficient to abolish the Gi-coupled GPCR-governed TE retraction and subsequent migration of RAW cells.

ing the migration direction of randomly motile cells (17). Activation of both receptor-tyrosine kinases and GPCRs controls common signaling pathways mediated through serine/threonine-specific protein kinases such as Akt and Raf, Rho GTPases including Ras-related C3 botulinum toxin substrate 1 (Rac1), Ras homolog family member A (RhoA) homolog of cell division control protein (Cdc42), and molecules calcium and diacylglycerol (DAG) as well. Through activation of such pathways, cell surface receptors can govern directional migration by modulating cellular activity at both the leading edge (LE) and trailing edge (TE), as well as by controlling basal motility (18). Nonreceptor signaling regulators, including guanine nucleotide exchange factors (GEFs) and guanine nucleotide dissociation inhibitors (GDIs) are involved in cell migration through modulating the activity of heterotrimeric G proteins. Activator of G protein signaling 3 (AGS3) is a GDI and liberates a free G␤␥ subunit by binding to G␣GDP with nanomolar affinities, whereas AGS1 is a GEF for G␣ i and promotes heterotrimer dissociation (19,20). Similarly, G␣-interacting vesicle-associated protein (GIV) and Dishevelled-binding protein (Daple) also control cell motility through the activation of G i signaling (21,22).
Asymmetrically activated GPCRs on the plasma membrane govern migration along the axis of chemokine gradient (23,24). The region of the cell on which GPCRs are more active becomes the LE. The traction forces generated at the LE pull the cell body toward the gradient, whereas propagating signals to the TE induces its retraction, facilitating the effective relocation of the entire cell (25,26). This process is termed tread-milling and involves intertwined networks of spatiotemporally coherent as well as segregated yet tightly controlled molecular and cellular events (27). Signaling activities initiated at the LE induce (i) formation of invadopodia and lamellipodia with new focal adhesions, (ii) retraction of the TE accompanied with actomyosin contractility and focal adhesion disassembly, and (iii) active relocation of internal organelles orchestrating directional movement of the entire cell. Whereas the majority of LE activities can be assigned to G␤␥ subunit-induced PI3K␥ activation and subsequent phosphatidylinositol 1,4,5-trisphosphate (PIP 3 ) production (28), it is not clear how the activation of G i -coupled GPCRs at the LE induces the retraction of the TE. The retraction of the cytoskeleton has been attributed to G␣ 12/13 subunit-mediated RhoA activation through RhoGEFs (29 -31). RhoA is initially present in the inactive GDP-bound form. Upon activation, it is converted to the active GTP-bound form. Subsequent activation of Rho family protein kinases results in phosphorylation of myosin light chain phosphatase (MLCP), thereby inhibiting its activity and leading to an increase in MLC phosphorylation. This consequently promotes actomyosin-based contractility (32). During G␣ i pathway-directed neutrophil migration, G␤␥ subunits in concert with AC9, through DAG and inositol 1,4,5-triphosphate (IP 3 ), activate mTORC2 and PKC␤II (33). This suppresses actin remodeling at the LE and phosphorylated myosin II (p-MyoII) activity at the TE. Here, AC9 and protein kinase A (PKA)-mediated inhibition of myosin light chain kinase (MLCK) and RhoA impair the TE retraction (34,35).
In this study, in addition to its LE functions, we investigated whether G i pathway-induced generation of free G␤␥ subunits also controls TE retraction. We hypothesized that during G i pathway-controlled cell migration, free G␤␥ subunits are primarily responsible for the establishment of front-back polarity, which encapsulates both LE and TE functions and thereby controls the entire process of cell migration. We tested this hypothesis using a mouse macrophage cell line, RAW264.7 (RAW), and blue opsin. Upon optical activation (OA) with 445-nm light (blue light), blue opsin induces precise spatial and temporal control of signaling and allows generation of G␣GTP and free G␤␥ subunits exclusively at the LE (36,37). Such GPCR activation mimics the state of a migratory cell experiencing a chemokine gradient. This allows examination of whether the TE retraction signal is generated at the LE. Spectral selectivity of blue opsin to blue light allows for imaging molecular and morphological responses of cells using other wavelengths, including 488 nm (green), 515 nm (yellow), and 595 nm (orange) (37). Our results show that free G␤␥ subunits are primarily responsible for controlling the entire migration process in macrophages induced by the activation of the Gi pathway. In addition to its well-known functions in the LE (38), the data demonstrate that G␤␥ subunits promote TE retraction through PLC␤-induced calcium mobilization. We show that this essential pathway, together with free G␤␥-RhoGEF-induced phosphorylation of MLCP, promotes actomyosin contractility, resulting in complete TE retraction.

OA of blue opsin for subcellular G protein activation and directional migration of macrophages
To examine whether opsins and G proteins can be activated exclusively at the LE, the G␥9 assay was used (39). This assay employs the translocation of mCherry G␥9 (mCh-␥9) from the PM to internal membranes (IMs), generated by activated GPCRs. Global OA of HeLa cells expressing blue opsin and mCh-G␥9 resulted in the redistribution of mCh-␥9 between the PM (black curve) and IMs (red curve) with a translocation half-time (t1 ⁄ 2 ) of ϳ7.5 s (Fig. 1, A and B). Termination of OA resulted in complete reversal of the translocation process, whereas continuous OA showed sustained G protein activation, as indicated by the maintenance of accumulated mCh-␥9 in IMs (supplemental Fig. S1A). OA of blue opsin in a subcellular region of interest with an area of 2 ϫ 10 m 2 was activated and resulted in a localized disappearance of mCh-G␥9 in the PM area exposed and confined accumulation of mCh-G␥9 at the proximal IMs (Fig. 1C) (yellow arrow). No mCh-G␥9 localization changes were observed in the distal regions (blue arrow). A plot shows the reduction of mCh intensity on the activated PM regions (solid black) and a corresponding increase in adjacent IMs (solid red) (Fig. 1D). Neither distal IMs nor PMs (dashed lines) showed any fluorescence intensity change. This finding suggests that blue opsin can be used to selectively activate G proteins in a spatially confined region of a cell.
Next, an approach was devised to globally image molecular sensors while inducing localized optical activation of G proteins in single cells (supplemental Fig. S2A). This allowed generation G␤␥ subunits induce trailing-edge retraction of a G protein activity gradient across the cell, segregating the LE and TE. Due to the Gaussian type distribution of intensities of the confined light beam (supplemental Fig. S2A, middle image, blue curve), the 445-nm optical stimulus was placed 1-2 m away from the periphery of the cell, limiting G protein activity to only the intended region. Using this strategy, the RAW cells expressing blue opsin-mCh were optically activated (blue box) in a confined region (Fig. 1E). The cells showed lamellipodia formation at the activated LE and cell body retraction at the TE. This resulted in an average migration of 35 m in 20 min. This result suggests that blue opsin can be used to localize G protein activity to the LE, and such activation is sufficient to direct complete migration of RAW cells.

Role of G␣ identity in the heterotrimer upon G␤␥ subunit generation
If G␤␥ subunit is the primary regulator of G i pathway-induced cell migration, could the free G␤␥ subunits generated as a result of G s -and G q -coupled GPCR activation also result in cell migration? Activation of G q and G 12/13 , as well as G s path- Figure 1. Adopting optical activation of GPCRs to control the chemokine pathway. A, images of a HeLa cell expressing blue opsin and mCh-␥9. The white box shows the region of OA using 445-nm laser light (1 W/m 2 ) pulsed at 0.5 Hz. Before OA, cells were reconstituted with 50 M 11-cis-retinal. Upon OA of the entire cell, mCh-␥9 on the PM translocates, reducing mCh-␥9 on the PM (yellow arrow) and leading to a gain in IMs (white arrow). Upon termination of OA, mCh-␥9 returns to the PM. B, the plot shows the normalized mCh-␥9 fluorescence intensity changes on the PM and IMs. C, spatially confined OA of blue opsin in the same HeLa cell as in A. Upon OA of the white box area, mCh-␥9 on the activated PM region translocates to IMs, resulting in a localized loss of mCh-␥9. Note the gain of mCh-␥9 in immediate IM regions (yellow arrow). No mCh-␥9 changes were observed in distal regions (blue arrow) of the cell. D, the corresponding plot for C shows the mean fluorescent intensity of mCh-␥9 on the PMs and in IMs (activated and distal). The gray area shows the duration of OA. E, OA (blue box) of G i -coupled blue opsins with 445-nm light induced RAW cell migration. Cell expresses blue opsin mCh. F, a HeLa cell expressing mCh-␥9, before and after activation of G i -coupled endogenous CXCR4 receptor with 50 ng/ml SDF-1␣. The plot shows mCh intensity on the PM and in IMs. G, images and plots showing translocation of GFP-␥9 in HeLa cells also expressing G s -coupled ␤ 1 -AR before and after activation with isoproterenol (10 M). H, HeLa cells expressing G q -coupled GRPR and the sensor GFP-␥9 show no translocation of GFP-␥9 after GRPR activation with 1 M bombesin. I, localized OA (blue box) of G s -coupled CrBlue in RAW cells failed to induce migration. Blue and green lines indicate the LE and TE of the cell. To the right are the orthogonal slice views of a cross-section of these regions. The arrow (orange) shows mCh intensity changes of this cross-section over time. Scale bar, 10 m. ways, has been shown to control migration in specific cell types, including neutrophils and cancer cells (40 -42). Interestingly, the associated migration mechanisms are usually described using the activity of the accompanying G␣ subunit. To understand this, generation of G␤␥ subunits by G s -and G q -coupled GPCR activation was compared with that induced by G i -coupled endogenous CXCR4 activation by SDF-1␣ in HeLa cells. HeLa cells expressing green fluorescent protein (GFP)-G␥9 and G s -coupled ␤ 1 -adrenergic receptor (␤ 1 -AR) were exposed to 10 M isoproterenol (Fig. 1G) and showed G␥9 translocation comparable with that induced by CXCR4 activation (Fig. 1F). In contrast, activation of G q -coupled gastrin-releasing peptide receptor (GRPR) through exposure to 1 M bombesin resulted in only a marginally detectable GFP-G␥9 translocation (Fig.  1H). However, GRPR activation resulted in a robust phosphatidylinositol 4,5-bisphosphate (PIP 2 ) hydrolysis (supplemental Fig. S1B). Activation of GRPR and another G q -coupled receptor, M3 muscarinic, not only failed to induce G␥9 translocation, but also failed to induce G␥1 and G␥5 translocations (supplemental Fig. S1C). However, they exhibited robust PIP 2 hydrolysis (supplemental Fig. S1D). Therefore, regardless of the G␥ subunit in the heterotrimer, activation of G q -coupled GPCRs results in the generation of fewer free G␤␥ subunits.

Effect of G s pathway activation-induced free G␤␥ on RAW cell migration
Because G s -coupled GPCR activation results in a significant translocation of G␤␥ subunits, we examined whether G s pathway activation also results in RAW cell migration mediated through free G␤␥ subunits. An optogenetic chimeric G s -coupled opsin, CrBlue, was employed to activate the G s pathway in confined regions of RAW cells (36). CrBlue was engineered by replacing cytosolic loops of blue opsin with that of the G s -coupled opsin from box jellyfish (Carybdea rastoni) (36). During the initial OA, RAW cells showed minor movement toward the optical input. However, continuous OA of CrBlue did not result in a continuous directional movement along the axis of GPCR activation (Fig. 1I). This may be due to less efficient activation of G proteins by CrBlue or opposing activities of G␣ s GTP and free G␤␥ subunits upon RAW cell migration. Increased cAMP production induced by G␣ s activates PKA, and exchange protein directly activated by cAMP (EPAC) controls cell motility (44). Whereas PKA suppresses cell migration, EPAC facilitates it (45,46). When RAW cells expressing blue opsin-mCh were treated with forskolin and isobutylmethylxanthine to increase cAMP levels, RAW cells failed to migrate upon blue opsin activation (Fig. 2, A and B). However, cells showed unperturbed LE activities, including lamellipodia formation. This finding is in agreement with reports that increasing cAMP and PKA signaling inhibits migration by attenuating RhoA and MLCK activities (8,47,48).
Because free G␤␥ subunits generated by G i pathway activation have been shown to regulate neutrophil migration through protein kinase C␤ (PKC␤II) (38), we also tested whether PKC influences RAW macrophage migration as well. PKC inhibition with a broad-spectrum PKC inhibitor, GO 6983 (which has a 7 nM IC 50 for PKC␤), does not prevent blue opsin-induced RAW cell migration ( Fig. 2 and supplemental Movies S1 and S2). Therefore, we anticipate that, during asymmetric activation of the G s pathway in RAW cells, G␣ s GTP-induced cAMP reduces the ability of free G␤␥ subunits to induce migration.

Roles of G␤␥ subunits in G i pathway-induced RAW cell motility
To isolate the contributions of G␤␥ subunits from that of G␣ i , RAW cell migration was examined in the presence of a small molecule G␤␥ subunit inhibitor, gallein (49,50). Upon localized OA of blue opsin, control RAW cells migrated with a velocity of ϳ1.8 m/min. Pertussis toxin (PTx) inhibited this migration, indicating that it is governed by the G i pathway (51) (Fig. 2D). Gallein binds to the effector-binding hot spot in G␤ (52). RAW cells incubated with 10 M gallein failed to migrate upon localized blue opsin activation, and no LE or TE activities were observed ( Fig. 2E and supplemental Movie S3). Plots show the distance traveled by the peripheries of the LE and TE as well as the velocities (Fig. 2G). Fluorescein, a gallein-like compound that does not bind G␤␥, did not show any effect on migration (supplemental Fig. S2, B and C).
To test whether gallein inhibited heterotrimer dissociation or G␣GTP signaling, we examined its effect on G␤␥ translocation and cAMP responses upon activation of ␤ 1 -AR. Translocation of GFP-␥9 was unperturbed by 10 M gallein (supplemental Fig. S3A). Additionally, in the presence of gallein, ␤ 1 -AR activation in HeLa cells showed an unperturbed cAMP production compared with the untreated cells, as measured using PKA-␦RII␤-PKA-C␣-YFP sensor (supplemental Fig. S3B). Thus, gallein inhibited migration by selectively blocking G␤␥ subunit-effector interactions.
We also tested optical inhibition of free G␤␥ subunits to examine the effect of localized G␣ i GTP activity on RAW cell motility. The optogenetic inhibitor is composed of cryptochrome 2 (CRY2) linked to the carboxyl terminus of G protein receptor kinase 2 (GRK2Ct) (53). We expressed PM-targeted cryptochrome-interacting basic-helix-loop-helix (CIB1) and cytosolic CRY2-GRK2Ct together with blue opsin in RAW cells. Using localized OA, CRY2-mCh-GRK2Ct was targeted to predefined LE (Fig. 2F). OA was continued after the addition of retinal to concurrently activate blue opsin at the LE. These cells showed neither lamellipodia formation nor migration, whereas control cells exhibited normal migration. Collectively, these data suggest that whereas free G␤␥ subunit is the primary controller, G␣GTP may not have a direct role in G␣ i -coupled GPCR-mediated RAW cell migration.

Segregating the roles of the G␤␥-PI3K-PIP 3 axis on directional motility
Here we examined the significance of PI3K on cell migration relative to other G␤␥ subunit effectors. PIP 3 has been identified as one of the primary regulators of cytoskeleton remodeling during LE formation (37,54). Accumulation of PIP 3 at the LE sets a feed-forward signal amplification circuit between PI3K and Rac1, and subsequently activates WAVE complex and Arp2/3 to promote actin polymerization, generating traction forces required for cell migration (55,56). By rapidly reversing the side of OA, we previously showed that the formation of PIP 3 at the new LE and its active reduction at the former LE are G␤␥ subunits induce trailing-edge retraction synchronized (37). This may allow cells to set the migration direction along the axis of the external chemical gradient (36).
To examine whether PIP 3 generation at the LE alone can govern both the LE and TE activities, thus supporting complete cell migration, PI3K at the LE was optically activated. The inter-Src homology 2 domain of a p85b-regulatory subunit (iSH2) of PI3K was optically targeted to the user-defined LE on the PM as described previously (Fig. 3A, top) (57). This results in a significant and localized PIP 3 production at the LE (Fig. 3A, bottom). A plot shows an increase in PIP 3 at the LE. This PIP 3 response was similar to that observed in blue opsin activation-induced cell migration. However, optogenetic PI3K activation only resulted in a minor migration response.
Next, to examine whether G␤␥ subunit-induced PI3K activation/PIP 3 production controls TE retraction, blue opsin in RAW cells was activated to induce migration while pharmacologically modulating specific signaling entities in the migration pathway as follows. Inhibition of G␤␥ subunits with gallein made RAW cells completely non-responsive to migration induction ( Fig. 3, B (I)). Inhibition of PI3K with 50 nM wortmannin also prevented cell migration ( Fig. 3B (II)). Wortmannintreated cells completely ceased LE activities, but they exhibited cell-wide cytoskeleton retraction that was not observed in gallein-treated cells (Fig. 3B (II), yellow arrows). This indicates that G␤␥ subunits mediate TE retraction through a PI3K-independent pathway.
Inhibition of Rho kinase (ROCK) with 12.5 M GSK 269962 inhibited RAW cell migration, whereas LE activities, including lamellipodia formation, remained almost unperturbed (Fig. 3B (III)). Combined inhibition of both PI3K and ROCK also completely eliminated cell migration, with no LE or TE activities observed (Fig. 3B (IV)). A plot shows the percent cell size change, which captures the contractility (positive) in wortmannin-treated cells and the expansion (negative) in ROCK-inhib- whereas the LE activities remained unperturbed, demonstrating the effect of increased cAMP on cell migration. C, treatment with 10 M GO6983, a potent PKC inhibitor, did not affect RAW cell migration, suggesting that the DAG/PKC pathway may not have a significant effect on migration. D, incubation of cell with PTx (0.05 g/ml) for 12 h completely suppressed cell migration. PTx induces ADP-ribosylation and prevents G protein heterotrimer dissociation. E, RAW cells treated with the G␤␥ inhibitor, gallein (10 M, 30 min, 37°C) also failed to migrate and showed neither LE nor TE activities. F, images of RAW cells transfected with blue opsin GFP, CRY2-mCh-GRK2Ct, and CIBN-CaaX. Local OA resulted in recruitment of CRY2-mCh-GRK2Ct to the PM, near the region of OA. Here, blue opsin activation resulted in release of free G␤␥ subunits. However, the recruited CRY2-mCh-GRK2Ct is expected to sequester free G␤␥, preventing activation of its effectors. Even after 20 min of OA, initiation of cell migration was not observed. G, histograms show the extent of movement and migration velocities of the peripheries of the LE and the TE of control and pharmacologically perturbed RAW cells in A-E. *, p Ͻ 0.001. (n ϭ 3-10 cells). Scale bar, 10 m. Error bars, S.E.

G␤␥ subunits induce trailing-edge retraction
ited cells (Fig. 3B). During migration induction, RAW cells treated with ROCK inhibitor showed a complete lack of TE contractility, whereas LE activities, including lamellipodia formation, remained unchanged. Therefore, we examined whether RAW cells express free G␤␥ subunit-controlled signaling components that govern the RhoA pathway. Data from RNA-seq analysis of RAW cells showed expression of multiple types of Rho-GEFs that are potentially involved in activation of RhoA and ROCK (supplemental Fig. S4A). Among them, RAW cells possess Rho114-GEF, which has been identified as a G␤␥ subunit-activated RhoGEF (10). Contractility observed at the TE, facilitating its retraction, has been attributed to a reduction in MLCP-induced dephosphorylation of p-MLC (58). ROCK retards this process by deactivating MLCP through its phos-phorylation. Therefore, it is likely that G␤␥ subunits control the RhoA pathway to maintain a high concentration of p-MLC and facilitate TE retraction during inhibitory GPCR-induced cell migration.
Another RhoGTPase, Cdc42, has been shown to be involved in a number of migration processes and is directly and indirectly controlled by G␤␥ subunits (59). Therefore, we also examined whether Cdc42 activity could be detected during blue opsin activation-induced RAW migration. Whereas RAW cells express a significant amount of mRNA for Cdc42 and a Cdc42-GEF, FLJ00018 (ARHGEF42) (supplemental Fig. S4A), RAW cells did not show Cdc42 sensor accumulation at the LE during blue opsin-induced migration (supplemental Fig. S4C). A, a RAW cell-expressing mCh-CRY-iSH2, CIBN-CaaX and AKT-PH-Venus was optically activated to recruit mCh-CRY-iSH2 to a confined region (top). The cell shows localized PIP 3 production (bottom images and the plot), similar to that seen with blue opsin activation; however, the optogenetic activation of PIP 3 production resulted in only a minor migration response. B, RAW cells transfected with blue opsin mCh were subjected to localized OA in the presence of various inhibitors: 10 M gallein (for G␤␥), 50 nM wortmannin (PI3K), and 12.5 M GSK 269962 (ROCK). Cells were incubated with the inhibitor(s) for 30 min at 37°C before imaging and OA. I, inhibition of G␤␥ resulted in a complete suppression of migration with complete cessation of LE and TE activities. II, inhibition of PI3K prevented migration, and a reduction in cell size was also observed. III, ROCK inhibition prevented complete cell migration. Whereas no TE retraction was observed, the LE activities remained unperturbed, resulting in a cell surface area increase. IV, dual inhibition of PI3K and ROCK resulted in neither cell migration nor cell size change. This suggests that ROCK is essential for generation of inward contractile forces. The histogram shows the corresponding percent cell surface changes. C, images of RAW cells transfected with blue opsin mCh and incubated with 10 M A-7 hydrochloride, a CaM antagonist, and ML-7 hydrochloride, an MLCK antagonist, respectively, for 30 min at 37°C. I, A-7 hydrochloride inhibited cell migration induced by OA of blue opsin; however, the cell showed intact LE activities. II, MLCK inhibition also prevented RAW cell migration. The histogram shows that compared with the control, both CaM-and MLCK-inhibited cells exhibit a complete lack of TE activities, whereas comparable lamellipodia growth at the LE was observed. **, p Ͻ 0.0001 (n ϭ 10). Scale bar, 10 m. Error bars, S.E.

G␤␥ subunits induce trailing-edge retraction Calmodulin (CaM) and MLCK increase p-MLC-based contractility
Here we examined the function of the pathway components that link free G␤␥ subunits and MLCK in RAW cells. MLCK phosphorylates MLC to generate p-MLC, inducing actomyosin contractility (60). Free G␤␥ subunits have been shown to interact with and activate PLC isoforms ␤1, ␤2, ␤3, and ␦3, inducing PIP 2 hydrolysis. This results in mobilization of stored calcium to the cytosol (61,62). Ca 2ϩ -CaM binds to the M13 catalytic domain and activates MLCK (63). The activated MLCK then induces phosphorylation at Ser-19 on the regulatory light chain of MLC. Alternatively, calcium-stimulated calmodulin-dependent protein kinase kinase 2 (CaMKK2) activates adenosine monophosphate-activated protein kinase (AMPK), resulting in MLCK phosphorylation (64). These phosphorylations increase the Ca 2ϩ -CaM concentration required to activate MLCK, setting up a negative feedback loop that suppresses MLCK activity and ultimately leads to down-regulation of MLC phosphorylation (65,66). To examine the outcome of this complex relationship between cytosolic calcium and MLC phosphorylation in RAW cells, we performed the following experiments.
We first tested whether G i pathway-induced migration of RAW cells is sensitive to the CaM antagonist A7-hydrochloride (67). The inhibition of CaM suppressed blue opsin activationinduced RAW cell migration (Fig. 3, C-I). However, the cells did produce lamellipodia at the LE, similar to the untreated control cells. Next, the effect of inhibition of MLCK on cell migration was examined using its inhibitor ML-7 hydrochloride. Although cells showed lamellipodia formation at the LE, neither TE retraction nor cell migration was observed (Fig. 3C (II)). A histogram shows the distance traveled by the peripheries of the LE and TE of RAW cells treated with A-7 and ML-7, respectively, compared with control cells (Fig. 3C). Interestingly, inhibition of AMPK did not show a detectable effect on blue opsin-induced RAW cell migration (data not shown). Overall, inhibition of both CaM and MLCK did not have an obvious effect on G i pathway-induced LE signaling, whereas the cells showed no TE retraction or directional migration (Fig.  3C (I and II)).

Roles of extracellular and cytosolic calcium on TE contractility and cell migration
How cytosolic calcium influences the TE retraction of RAW cells was examined. RAW cells incubated in calcium-free Hanks' balanced salt solution (HBSS) over 30 min were unable to migrate in response to localized blue opsin activation (Fig. 4,  A and B). To examine whether this lack of migration is due to reduced cytosolic calcium prior to the migration induction, RAW cells were incubated for 30 min with BAPTA-AM, a cellpermeable calcium chelator. Compared with control RAW cells, BAPTA-AM-treated cells showed no migration, whereas their LE activities remained almost intact ( Fig. 4C and supplemental Movie S4). To confirm that the suppression of migration is not due to an unintended effect of BAPTA-AM, the same experiment was performed using thapsigargin, a sarco/endoplasmic reticulum calcium-ATPase pump inhibitor), in place of BAPTA-AM. Before the induction of cell migration, cells were incubated with 0.5 M thapsigargin to deplete stored ER calcium (68). Analogous to the BAPTA-AM experiment, cells showed no migration, whereas LE activities, including lamellipodia formation, remained unperturbed (Fig. 4D). Histograms show that depletion of cytosolic or sarcoplasmic calcium has the greatest impact on TE retraction, although LE lamellipodia formation and the overall cell migration velocities were also reduced (Fig. 4, E-G).
We then examined the link between cytosolic calcium and G␤␥-governed PI3K activity. Both control and BAPTA-AMtreated cells expressing Akt-PH-mCh (PIP 3 sensor) showed a similar and significant PIP 3 production upon localized blue opsin activation (Fig. 4H). An orthogonal view shows the mCh dynamics across a section of the cell, exhibiting PIP 3 sensor translocation from the cytosol to the PM (Fig. 4H). Nevertheless, only control RAW cells migrated (Fig. 4I and supplemental Movie S5).

G i -coupled receptor activation induces mobilization of cytosolic calcium
The ability of the G i -pathway to mobilize calcium to the cytosol through G␤␥ subunits was examined. We have previously shown that activation of G␣ i -coupled endogenous ␣ 2 -AR with norepinephrine (NE) induces an increase in cytosolic calcium in HeLa cells (69). Real-time PCR data show that RAW cells possess considerable levels of mRNA for PLC␤2, an enzyme activated by G␤␥ subunits (Fig. 5A). To confirm that G i -coupled GPCR activation induces PLC␤ activation and PIP 2 hydrolysis through free G␤␥ subunits, NE-induced PIP 2 hydrolysis in HeLa and RAW cells expressing mCh-PH was examined. OA of blue opsin resulted in significant PIP 2 hydrolysis in both cell types (Fig. 5, B and C). PIP 2 distribution remained unchanged in PTx-and gallein-treated RAW cells (Fig. 5, D and E). To examine whether the signaling circuits required to mobilize G␤␥ subunit-mediated calcium are present in RAW cells, endogenous complement component 5a receptor 1 (c5aR1), a G i -coupled GPCR, was activated, and the cytosolic calcium was probed using Fluo-4 AM. The addition of 25 M c5a resulted in a robust increase in cytosolic calcium, which peaked within 20 s (Fig. 6A). RAW cells treated with PTx failed to show c5a-induced calcium mobilization, suggesting that the G i pathway controls this process (70). To examine whether the ER calcium stores in PTx-treated cells remain intact during the addition of c5a, cells were exposed to thapsigargin. Thapsigargin has been shown to transiently increase cytosolic calcium by blocking calcium entry to the ER (71). PTx-treated RAW cells failed to produce a calcium response upon c5aR activation, but the same cells exhibited a transitory cytosolic calcium increase upon thapsigargin addition (Fig. 6B). To demonstrate that the observed calcium mobilization is independent of G␤␥ subunitmediated PI3K activation, the c5a-induced calcium response was examined in the presence of a PI3K inhibitor, wortmannin. Inhibition of PI3K did not alter the c5a-induced calcium mobilization (Fig. 6C). To directly examine whether the G␣ i pathway-mediated cytosolic calcium mobilization is controlled by free G␤␥ subunits, c5a receptors were activated in cells incubated with gallein to inhibit G␤␥ subunits. As anticipated, cells did not show a calcium increase (Fig. 6D). Cells

G␤␥ subunits induce trailing-edge retraction
treated with the gallein-like control compound fluorescein (which does not bind G␤␥) did exhibit calcium mobilization upon c5AR activation (supplemental Fig. S2D).
The pathway by which free G␤␥ subunit-induced calcium mobilization in RAW cells was mapped as follows. Cells were incubated in calcium-free buffer for 1 h. These cells did not exhibit calcium mobilization upon c5aR activation (Fig. 6E).
G␤␥ subunit-induced PIP 2 hydrolysis was examined after blocking inositol triphosphate receptor (IP3R) using 10 M 2-aminoethoxydiphenyl borate (2APB). In the presence of 2APB, c5a failed to induce calcium mobilization in RAW cells (Fig. 6D). To examine whether calcium enters through an ion channel, c5a-induced calcium was examined in RAW cells after chelating extracellular calcium by adding BAPTA to the cell

G␤␥ subunits induce trailing-edge retraction
culture medium (Fig. 6E). Results show that the addition of BAPTA did not perturb (at 150 s) the C5a activation-induced calcium response. Collectively, these data suggest that free G␤␥ subunits increase cytosolic calcium through the following pathway: PLC␤ activation 3 PIP 2 hydrolysis 3 IP 3 generation 3 IP3R activation. Because we induced cell migration through the activation of G i -coupled blue opsin at the LE, blue opsin's ability to induce calcium mobilization in RAW cells was also examined. The cells were preincubated with the calcium indicator Fluo4-AM and continuously imaged using 488-nm light at 1 Hz to capture Fluo-4 fluorescence before and after retinal addition. Upon the addition of retinal to activate opsin, blue opsinexpressing RAW cells showed a calcium response in a similar manner to that induced by the c5a addition (Fig. 6F). Because of the broad absorption spectra of blue opsin that overlaps in its right shoulder with 488 nm (supplemental Fig. S4B), we employed this wavelength for both imaging of Fluo4 and blue opsin activation.

Discussion
We have previously shown that localized OA of blue opsin can activate heterotrimeric G proteins and induce directional cell migration (37,72). Both LE and TE activities are essential for cell migration. Our optogenetic strategy allowed for delivery of identical optical cues to individual cells, restricting G protein activation only to the LE. This enabled mapping signal propagation from the LE to the TE in real time in living cells. Our data indicate that G␣ i GTP is not likely to play a direct role in G i -mediated RAW cell migration. Nevertheless, further investigations are required to confirm this. The lack of RAW cell migra-tion upon localized activation of G s -coupled CrBlue suggests that G␣ s GTP signaling may obstruct G␤␥ subunit-induced RAW cell migration (Fig. 1F). The loss of the ability of blue opsin to induce RAW cell migration in the presence of elevated cAMP further validates these findings. These observations are also in concert with the previously reported cAMP-mediated inhibition of cell migration (47,48). Elevated cAMP can prevent contractile force generation through PKA-mediated inhibition of MLC phosphorylation. Although free G␤␥ subunits can enhance the activities of AC isoforms 2, 4, and 7 (73,74), this process requires G␣ s GTP and therefore eliminates any possibility of cAMP generation during the activation of the G i pathway. Collectively, these observations suggest that cAMP is likely to negate the ability of free G␤␥ subunits from the G s pathway to induce RAW cell migration. Further, gallein induces complete inhibition of cell migration, demonstrating that free G␤␥ subunits play the primary role in controlling the complete cell migration induced by G i pathway activation.
Free G␤␥ subunits interact with and control multiple effectors, many of which can be linked to cell migration. Analysis of RNA-seq data from RAW cells shows the expression of these signaling proteins in RAW cells, validating the existence of the proposed pathway (Fig. 5A). Among them, activation of PI3K␥, Rac1, and Cdc42 has been shown to be involved in the LE activities (75). Optogenetic activation of Cdc42 in RAW cells resulted in complete cell migration together with accumulation of the Cdc42 sensor at the LE (76). However, with the same sensor, we did not observe detectable Cdc42 activity during blue opsin activation-induced RAW cell migration. This can  3). B and C, images of a HeLa cell expressing mCh-PH and a RAW cell expressing mCh-PH and blue opsin, respectively, before and after GPCR activation. Activation of endogenous ␣ 2 -AR with 400 M NE in HeLa cells and activation of blue opsin using 445-nm light in RAW cells both resulted in PIP 2 hydrolysis. The plots show the corresponding mCh-PH changes in the cytosol. D and E, images of RAW cells expressing blue opsin and mCh-PH, treated with PTx (overnight incubation with 0.05 g/ml PTx at 37°C) and gallein (30-min incubation with 10 M at 37°C), respectively. Both PTx-and gallein-treated cells failed to exhibit PIP 2 hydrolysis upon OA of blue opsin. The plots show the corresponding mCh-PH changes in the cytosol. Scale bar, 10 m.

G␤␥ subunits induce trailing-edge retraction
either be a result of higher affinity of free G␤␥ subunits to other effectors or limited sensitivity of the assay to detect a minor Cdc42 activation, or both.
The minor RAW cell migration observed with optogenetic control of PI3K indicates that additional pathways controlled by free G␤␥ subunits, other than PI3K and Cdc42, are required for G i pathway-mediated complete RAW cell migration, especially for the processes involved in TE retraction. The ability of PI3K-inhibited cells to show a cell-wide contractility, while failing to exhibit LE activities, further supports the involvement of additional pathways (Fig. 2F). The RNA-seq data indicates the presence of multiple types of Rho-GEFs in RAW cells that can be involved in the activation of RhoA and subsequently ROCK (supplemental Fig. S4A). Among them, we found Rho114-GEF, which is activated by free G␤␥ subunits (10). The deficiency in cell migration observed after inhibition of G␤␥ subunits and lack of TE retraction after ROCK inhibition suggest that G␤␥ subunits control TE retraction in RAW cells through the activation of RhoGEFs alone or together with other mechanisms.
ROCK promotes p-MLC-mediated contractility to induce TE retraction by deactivating MLCP through its phosphorylation (77). Whereas ROCK can also phosphorylate MLC (78), MLCK is crucial for MLC phosphorylation. Interestingly, realtime PCR data suggest that RAW cells express a considerable amount of PLC␤2 and CaM, obviously linking G␤␥ subunits with MLCK though Ca 2ϩ -CaM (Fig. 5A). Further, G i -coupled GPCR activation induces calcium mobilization, which is sensitive to PTx, gallein, and 2APB. This clearly demonstrates the existence of a functional G␤␥-PLC␤-IP 3 -IP3R-calcium signaling axis. Inhibition of cell motility upon either elimination or reduction of cytosolic calcium using BAPTA-AM and thapsigargin, respectively, indicates that calcium is an essential player in this mechanism. Unperturbed LE activities together with inhibited directional migration in cells lacking cytosolic calcium indicate that the LE and the TE activities are controlled by distinct yet G␤␥ subunit-controlled molecular mechanisms. By inactivating MLCP and through direct phosphorylation of MLC, ROCK promotes actomyosin contractility. If the TE retraction is mainly controlled by the RhoA-ROCK pathway, cells should be able to show a significant migratory response in the presence of low concentrations of cytosolic calcium. However, data presented here reject this hypothesis. Regardless of the dual control of cytoskeleton (77,78), data here show that ROCK is unable to support complete cell migration in the absence of cytosolic calcium. This indicates the existence of a calcium-mediated mechanism that acts as a gatekeeper for the TE retraction. Data from experiments with CaM as well as MLCK inhibition suggest that, independent of PI3K and RhoA pathway activation, G␤␥ subunits are likely to control MLC phosphorylation as well. The complete inhibition of cell migration observed upon inhibition of either the Ca 2ϩ -CaM or RhoA-ROCK pathway suggests that neither pathway is redun- , which inhibits sarco/endoplasmic reticulum (SR) calcium transport. The plot shows individual cell calcium responses after c5a (black traces) followed by thapsigargin addition (red traces). C, images of RAW cells incubated with 50 nM wortmannin, before and after c5aR activation. The images and plot show that PI3K inhibition has no effect on the c5aR-induced calcium response. D, images of RAW cells before and after exposure to various inhibitors: G␤␥ inhibitor, 10 M gallein; cytosolic calcium chelator, 10 M BAPTA-AM; IP3R antagonist, 5 M 2APB. Cells were incubated with inhibitors for 30 min at 37°C. All of these agents inhibited the c5aR-induced calcium response. E, images of RAW cells incubated in calcium-free buffer containing 5 M BAPTA, before and after c5AR activation. Lack of extracellular calcium partially reduced the c5aR-induced calcium release (compare with A). F, images of RAW cells expressing blue opsin-mCh before and after opsin activation by 50 M 11-cis-retinal. Blue opsin induced a robust calcium response only in the presence of retinal (red traces), compared with cells lacking the chromophore (black trace). Scale bar, 10 m.

G␤␥ subunits induce trailing-edge retraction
dant, and therefore both are mandatory for TE retraction during G i pathway-governed RAW cell migration.
Overall, our results suggest that G␤␥ subunits can solely control complete RAW cell migration through direct regulation of three interconnected yet independently operated mechanisms. The first mechanism includes previously defined G␤␥ 3 PI3K␥ 3 activation of Rho family GTPases, including Rac1 and Cdc42, which control LE functions. Specific RhoGEFs, including P114-RhoGEF, which controls RhoA 3 ROCK to increase MLC phosphorylation, come under the second G␤␥ subunit-controlled mechanism. From the data presented here, we show a third G␤␥ subunit-controlled mechanism that regulates TE retraction through direct activation of PLC␤ 3 calcium 3 Ca 2ϩ -CaM 3 MLCK activation, subsequently increasing MLC phosphorylation. Additionally, the data also show that, whereas the RhoA-mediated mechanism is essential, it is not sufficient for cells to achieve complete TE retraction (Fig.  7). Therefore, the G␤␥ 3 PLC␤ 3 Ca 2ϩ -CaM 3 MLCK axis is indispensable for MLC phosphorylation-induced actomyosin contractility. Nevertheless, evidence is available for cells diversely utilizing cytosolic calcium to govern cytoskeleton remodeling via mechanisms other than contractility (79,80). For instance, gonadotropin-releasing hormone-expressing neurons employ calcium and the RhoA/ROCK pathway to pro-mote outward actin in the leading process as a supporting mechanism to migrate (81).
The local excitation global inhibition (LEGI) model that includes excitable signaling networks has been proposed to examine the differential LE and TE activities during cell migration (37,82,83). The present study demonstrates that cytosolic calcium is an integral component of the cell migration signaling networks governed by the G i pathway. The locally generated calcium at the LE is likely to diffuse through the cytosol to distal regions and promote contractility. Therefore, calcium generated as a result of G␤␥ subunit-induced PLC␤ activation can be considered as a candidate for the long-sought global inhibitors described in the LEGI model. Although calcium-induced contractility can be global, because of the active cytoskeleton remodeling at the LE, retractions appear to be visible only in distal regions. Considering the conservation of migration mechanisms across cell types, other migratory cells, including cancer cells, may employ similar signaling networks to execute their migration. Therefore, the novel mechanisms as well as the synergistic cross-talk between pathways presented here may not only help to advance understanding of cell migration, but may also identify molecular targets for therapy development.

Cell culture and transfections
HeLa cells (ATCC, Manassas, VA) were cultured in minimum essential medium (CellGro) supplemented with 10% dialyzed fetal bovine serum (Atlanta Biologicals) in the presence of 1% penicillin-streptomycin in 60-mm tissue culture dishes at 37°C in a 5% CO 2 humidified incubator and subcultured every 2 or 3 days. At 75% confluence, adherent cells were detached after versene-EDTA (CellGro) and centrifuged at 1000 ϫ g for 3 min; versene-EDTA was aspirated before cells were resuspended in regular culture medium at a cell density of 1 ϫ 10 6 / ml. RAW mouse macrophage cell line (ATCC, Manassas, VA) was cultured in RPMI 1640 (10-041-CV; Corning, Manassas, VA) with 10% fetal bovine serum (Atlanta Biologicals, Flowery Branch, GA), L-glutamine, and 1% penicillin-streptomycin in a humidified incubator at 37°C and 5% CO 2 . The cells were subcultured every 2-3 days when they achieved 70 -80% confluence. For experiments, cells were seeded on 35-mm glass-bottomed dishes (In Vitro Scientific) at 8 ϫ 10 4 /ml cell density and cultured in RPMI 1640 supplemented with 10% dialyzed fetal bovine serum and antibiotics. Cells between passages 3 and 25 were used for experiments. HeLa and RAW cells were transfected using the transfection reagent PolyJet TM (Signagen) according to the manufacturer's protocol. Transfection was performed on cells plated on glass-bottomed dishes. After 5 h of incubation with the transfection reagent, the medium was replaced with 1 ml of fresh culture medium. Cells were imaged 12-24 h after transfection. RAW cells were also transfected by electroporation using the D-032 protocol on an Amaxa Nucleofector device (Lonza, Basel, Switzerland). Each electroporation was performed on 1.25 ϫ 10 6 cells in 200 l of Nucleofector solution, followed by the immediate addition of 800 l of warm culture medium. The cells were then plated on glass-bottomed

Live cell imaging and data analysis
Experiments were performed with a ϫ60, 1.4 numeric aperture oil objective or a ϫ10, 0.3 numeric aperture objective in a spinning-disk XD confocal TIRF imaging system that is composed of a Nikon Ti-R/B inverted microscope, a Yokogawa CSU-X1 spinning disk unit (5000 rpm), and an Andor FRAPPA unit for photoactivation of manually selected regions of the opsins in real time, all controlled using Andor iQ version 3.1 software (Andor Technologies, Belfast, UK), with 445-, 488-, 515-, and 594-nm solid-state lasers and an iXon ULTRA 897BV back-illuminated deep-cooled EMCCD camera. Blue opsin mCh and Akt-PH-mCh were imaged using 594-nm excitation and 630-nm emission settings, blue opsin-mTurquoise was imaged using 445-nm excitation and 478-nm emission, and Fluo-4 AM was imaged using 488-nm excitation and 515-nm emission, respectively. For OA of blue opsin, cells were reconstituted with 50 M 11-cis-retinal. OA was then performed with 445-nm blue light (white or blue boxes; 0.8 W/m 2 ) pulsed at 0.5 Hz covering the entire cell or user-defined subcellular regions. Iterative OAs were performed at 3-s intervals, adjusting the location of the optical stimuli with respect to the cellular response. Axial focus drifts were circumvented using the Perfect Focus System (Nikon) during cell migration experiments. Mean pixel fluorescence intensity changes in the entire cell or in selected areas of the cell were determined using the Andor iQ version 3.1 software. Time-lapse images were analyzed using the analytical tools accompanying the Andor iQ 3.1 software;, further processing of data was conducted using Origin pro (OriginLab Corp.).

Cytosolic calcium measurements
For intracellular calcium measurements, cells were cultured on cell culture dishes at 37°C with 5% CO 2 . Experiments were performed 12-24 h after plating. Cells were washed twice with HBSS with calcium, pH 7.2, and incubated for 50 min at room temperature with the fluorescent calcium indicator Fluo-4 AM (0.28 nM). The fluorescence intensity of Fluo-4 AM (488 nm) was continuously imaged at 1-s intervals using 488-nm excitation and 515-nm emission with confocal microscopy. Fluo-4 AM fluorescence intensities obtained from regions of interest were normalized to initial values.

Real-time quantitative PCR assay
The RAW cells were homogenized, and mRNA was isolated with the GeneJET RNA purification kit according to the manufacturer's protocol. Synthesis of cDNA was performed using the RADIANT cDNA synthesis kit. The gene expression was quantified using real-time PCR using the RADIANT Green LO-ROX quantitative PCR kit (Alkali Scientific, Pompano Beach, FL). DNA amplification was carried out using Icycler (Bio-Rad), and the detection was performed by measuring the binding of the fluorescence dye SYBR Green to double-stranded DNA. All of the primers were ordered from IDT-DNA (supplemental Table S1). The relative quantities of target gene mRNA against an internal control, ␤-actin, were calculated using the ⌬⌬CT method. The difference (⌬CT) between the mean values in triplicates of target genes and those of ␤-actin were calculated using Origin 9 (OriginLab Corp.), and the relative quantified value was expressed as 2 Ϫ⌬⌬CT .
Author contributions-P. S. conducted most of the experiments and analyzed the results. D. K. conducted experiments on optical control of PI3K using iSH2, G␥9 translocation by SDF1␣-induced CXCR4 activation. K. R. conducted experiments on G␥9 translocation with bombesin, PIP 2 hydrolysis, G␣ q -induced cell shape change, and CrBlue-induced cell migration. K. S. conducted experiments on the G␥9 translocation with N. E. A. K. and P. S. conceived the idea for the project and wrote the manuscript.