Phosphorylation of filamin A regulates chemokine receptor CCR2 recycling

A key aspect of the cellular responses mediated by G-protein-coupled receptors (GPCRs) is the spatiotemporal regulation of their signaling, which occurs at the level of ligand binding but also through control of the number of receptors on the cell surface. Therefore, explaining the control of receptor sorting into different degradative or recycling endocytic pathways is fundamental to understand GPCR function. Once internalized, receptors follow different fates inside the cell: some rapidly return to the plasma membrane from the sorting endosome, others travel to the perinuclear endosomes for slower recycling or are transported to lysosomes for degradation. Receptor sorting within endosome trafficking is facilitated by dynamic actin filaments, which are generated through WASH and the Arp 2/3 complex at the surface of endosomes (Derivery et al., 2009; Gomez and Billadeau, 2009) and specialized protein interactions. Some GPCRs, such as β2-adrenergic receptor (β2AR), have specific sequences that trap the receptor in endosomal actin-enriched subdomains for transport along the short recycling route – so called ‘sequence-dependent’ recycling (Cao et al., 1999; Puthenveedu et al., 2010). The sequence-dependent pathways require PDZ-containing proteins and the retromer complex (Temkin et al., 2011). In those routes, the actin-enriched subdomains are thought to function as platforms that capture cargo proteins into retromer tubules, retrieving them from the default pathway.

The diversity of actin functions along the endocytic pathway is the result of proteins that precisely control the architecture of the actin networks and of the interacting partners among the endocytic cargo and the endocytic machinery. Some of the most prominent actin-cross-linking proteins in higher eukaryotes are the filamins (filamin A, B and C; FLNa, FLNb and FLNc, respectively) (Nakamura et al., 2011). FLNa is a dimeric protein with 24 immunoglobulin (Ig) repeats that cross-link rod-like actin filaments at right angles, creating three-dimensional networks (Hartwig et al., 1980). FLNa exists as an autoinhibitory form that involves Ig repeat 20 (Lad et al., 2007). In addition, FLNa binds to a myriad of cellular proteins of great functional diversity, acting as a scaffold and facilitating protein–protein interactions (Popowicz et al., 2006). Thereby, FLNa integrates the cellular architecture and signaling cascades that are essential for fetal development and cell locomotion. Not surprisingly, mutations in FLNa are linked to a large number of developmental diseases (Savoy and Ghosh, 2013). Among its partners are several GPCRs (Awata et al., 2001; Kim et al., 2005; Onoprishvili et al., 2003; Yu et al., 2008), including several chemokine receptors: CCR2B (one of two alternative splice forms encoded by CCR2) (Minsaas et al., 2010; Planagumà et al., 2012), CCR5 and CXCR4 (Gómez-Moutón et al., 2015; Jiménez-Baranda et al., 2007). The exact role of FLNa binding to chemokine receptors is not well understood yet, but much work indicates its function in endocytic traffic and receptor internalization (Cho et al., 2007; Kim et al., 2005; Onoprishvili et al., 2003; Seck et al., 2003). Moreover, FLNa binds to β-arrestin-2, a well-defined GPCR adaptor for clathrin-mediated internalization (Kim et al., 2005; Scott et al., 2006). Recently, FLNa has been suggested to cooperate with binding of β-arrestin-2 to the CXCR4 receptor carrying mutations in arginine residue 334 that are present in WHIM syndrome, which could have implications in WHIM-mutant CXCR4 signaling (Gómez-Moutón et al., 2015).

We have previously shown that FLNa, through predominantly Ig repeats 19–23, binds to the C-terminal tail of the monocyte chemoattractant receptor 2 (CCR2, isoform B) (Minsaas et al., 2010). CCR2 and its ligand, the monocyte chemoattractant protein 1 (CCL2), are involved in a wide variety of pathophysiological inflammatory processes (Deshmane et al., 2009; Johnson et al., 2005; Sallusto and Baggiolini, 2008) and neurodegenerative disorders (Bose and Cho, 2013). Similar to most ligand-bound GPCRs, CCL2-stimulated CCR2 is internalized in clathrin-coated vesicles (García Lopez et al., 2009), which is important to regulate its cellular signaling response (García Lopez et al., 2009; Jiménez-Sainz et al., 2003). The interaction of FLNa with the C-terminal region of CCR2B is essential for receptor internalization in endocytic vesicles and for CCL2-induced monocyte migration (Minsaas et al., 2010; Planagumà et al., 2012). Here, we show that FLNa contributes to CCR2B recycling by controlling the motility of endosomes and the entrance of the receptor into the endosomal actin microdomains that recruit cargo destined for the short-loop recycling pathway. We further show that CCL2-induced phosphorylation of FLNa at S2152 is required to facilitate receptor recycling, suggesting that GPCRs can control the membrane-traffic machinery and spatiotemporally tune its signaling. We have also demonstrated that this mechanism is shared by other GPCRs.

CCR2B directly interacts with the Rod 1 domain of FLNa, within Ig repeat 19, through the C-terminal intracellular domain of the receptor. This interaction facilitates internalization of the ligand-activated receptor into clathrin-coated pits (Minsaas et al., 2010). We noticed that endosomal structures containing CCR2B appeared to be more peripherally distributed upon ligand stimulation in the human melanoma cell line M2 (lacking expression of FLNa) compared with the isogenic cell line (Minsaas et al., 2010). Initial experiments published previously had also indicated a defect in receptor recycling upon FLNa depletion (Minsaas et al., 2010). This suggests that FLNa plays roles in CCR2B endosomal trafficking, other than those previously reported. To analyze the role of FLNa in CCR2B trafficking through the endocytic pathway, the internalization of Flag–CCR2B was followed with live confocal microscopy in HeLa cells that had been depleted of FLNa by utilizing short hairpin (sh)RNAs (Fig. 1A; Fig. S1A). The results were consistent with previous data and showed CCR2B in peripheral endosomal structures in the absence of FLNa, 30 min after addition of CCL2 (Fig. S1B). Close observation of receptor-loaded endosomes by live confocal imaging revealed an apparent enlargement of the CCR2B-loaded organelles in cells that had been treated with two different shRNAs against FLNa (shFLNa28 and shFLNa29) (Fig. 1A; Fig. S1B). Morphometric analysis showed a significant 1.4-fold increase in the diameter of the CCR2B-loaded endosomes (P<0.001, one way ANOVA) in the absence of FLNa (Fig. 1B).

To study the nature of the enlarged endosomal particles, HeLa cells that expressed CCR2B were stimulated with CCL2 at different times, fixed and co-labeled with antibodies against Flag and EEA-1 (Fig. 1C; see Fig. S1C for lower magnification). Some internalized CCR2B particles colocalized with EEA-1 in the presence or absence of FLNa 15 min after CCL2 addition, but the localization of internalized CCR2B with EEA1 became more apparent in shFLNa (shFLNa28)- than in control (shScr)-treated cells at later time points (30 min) (Fig. 1C). This suggests that, although the arrival and exit rates of CCR2B from EEA1-positive endosomes reached the steady-state 15 min after addition of CCL2 in control cells, CCR2B continued to accumulate in the absence of FLNa over at least 15 more minutes. Quantification of fluorescence indicated that knockdown of FLNa expression increased the colocalization of CCR2B with EEA-1 endosomes 1.6-fold (P<0.05, two way ANOVA) at 30 min, compared to controls (Fig. 1D). This implied that, in the absence of FLNa, the EEA-1-labeled endosomes kept their capacity to receive incoming vesicles from the plasma membrane for at least 30 min after addition of CCL2 and that the exit of CCR2 from early endosomes was delayed.

As the receptor accumulates in enlarged early endosomes in the absence of FLNa, we studied the effect of FLNa depletion on the dynamic behavior of CCR2B-loaded endosomes using high-speed [20 frames per second (fps)] imaging. Analysis of immunolabeled CCR2B-positive endosome trajectories that lasted more than 1 min showed a reduction in endosomal velocity [25±5% (mean±s.e.m.) slower than that of controls] and distance traveled [30±5.5% (mean±s.e.m.) shorter than that in controls] in shFLNa cells (Fig. 2A). Significant differences were observed within 15 and 30 min after addition of CCL2, which were independent of the size of the endocytic structures. These experiments allowed us to observe long-range endosome trajectories directed towards the perinuclear region in control cells, whereas in shFLNa cells, CCR2B endosomes moved in random directions with their trajectories remaining close to the origin (Fig. 2B).

To analyze the balance between CCR2B entry and exit rates in and out of sorting endosomes, fluorescent resonance after photo-bleaching (FRAP) experiments were conducted in living cells expressing Rab4A–YFP, loaded with Alexa-Fluor-647-conjugated antibodies against CCR2B. Endosomes with similar levels of CCR2B–Alexa-Fluor-647 fluorescence and size were selected, and one whole endosome was bleached in its entirety. After FRAP, only 11±1.3% (mean±s.e.m.) of the receptor signal on Rab4A-containing endosomes was recovered after 2.5 min in control cells (Fig. 2C, upper panels; Movie 1). The recovery of the CCR2B signal was faster in the shFLNa cells [18±1.8% (mean±s.e.m.) after 2.5 min (Fig. 2D), compared to controls and to bleached Rab4A–YFP, which showed no differences (Fig. S2; Movie 2). This was despite FLNa depletion hindering the internalization of CCR2B and the motility of the primary endocytic vesicles. Taken together, the data strongly suggest that the exit of CCR2B from sorting endosomes was impaired in the absence of FLNa.

To study which route CCR2B follows back to the plasma membrane, we analyzed the extent of its colocalization with Rab4A and compared it with that of other well-known receptors [transferrin (Tfn) receptor, β2AR, PAR1 (also known as F2R) and ETB (also known as EDNRB)] (Fig. 3A,B). CCL2-stimulated CCR2B, as well as the other receptors, accumulated in microdomains along the Rab4A-decorated endosome membrane. Utilizing line fluorescence correlation (Fig. 3A,B), we found that Tfn was present in most Rab4A-labeled microdomains; consistent with previous reports showing the Tfn receptor can follow both the short- and the long-loop recycling routes (Mayle et al., 2012). Other GPCRs (PAR1 and Etb), which are sorted to lysosomes (Shapiro and Coughlin, 1998; Terada et al., 2014), hardly correlated with Rab4A microdomains (Fig. 3B; Fig. S3). In contrast, internalized CCR2B coincided with about 20% of the Rab4A-enriched microdomains, similar to β2AR, which mainly follows the short recycling loop from the sorting endosome to the plasma membrane. In fact, CCR2B (Flag–CCR2B) and β2AR (HA–β2AR) often colocalized or were found in close apposition on endosomal microdomains and moved together along the endosomal membrane (Fig. 3C,D), indicating that they follow the same trafficking pathway. As previously described (Temkin et al., 2011), β2AR microdomains were positive for coronin2–GFP, which marks the actin-enriched structures that trap cargo destined for the short-loop recycling pathway from sorting endosomes to the plasma membrane (Fig. 3E, lower panel). Consistent with the view that CCR2B followed the same route as coronin2, CCR2B-labeled microdomains also localized with coronin2–GFP (Fig. 3E, see also line profile).

To investigate if FLNa might be directly involved in the endosomal trafficking of CCR2, super-resolution microscopy (STED) was utilized to determine its presence on endosomes. The fluorescence signal of a functional FLNa–GFP construct (Planagumà et al., 2012) was most apparent at the cortex of HeLa cells (Fig. 4A). However, taking a sagittal plane of the cell, FLNa–GFP could be detected at discrete dynamic subdomains on the surface of vacuolar structures (Fig. 4A, lower panel). Interestingly, FLNa–DsRed colocalized with coronin2–GFP dots on the vacuolar membranes, indicating that FLNa might directly participate in the function and regulation of the actin-enriched microdomains in trapping cargo for the retromer pathway to the plasma membrane (Fig. 4B). To confirm so, we investigated whether some of these FLNa–GFP microdomains coincided with internalized fluorescence-labeled CCR2B. CCR2B was present in Rab4A–Cherry-labeled sorting endosomes in accumulations, which transiently coincided with FLNa–GFP microdomains (Fig. 4C,D; see Movies 3 and 4). The CCR2B- and FLNa-labeled accumulations were very mobile (Fig. 4C). As anticipated, FLNa–GFP localized with actin–RFP and CCR2B on receptor-loaded early endosomes (Fig. 4D). It was also possible to observe FLNa together with CCR2B on domains in Rab5A–Cherry-labeled endosomes (Fig. 4D; Movie 4). However, little coincidence was observed in Rab11–Cherry-labeled endosomes (Fig. 4D; Movie 5). These results indicate that FLNa and CCR2B colocalized on the endosomal actin and coronin2-enriched early endosomal microdomains, similar to internalized HA– β2AR (Fig. 4E, see also Fig. 5).

We next asked if FLNa is important to maintain the actin-enriched endosomal subdomains. Depletion of FLNa did not alter either the number or the intensity of the coronin2–GFP patches on the endosomes (Fig. 5A,B), suggesting that FLNa is not essential to generate or maintain the structure of coronin2–actin patches on endosomes. Strikingly though, we observed that depletion of FLNa clearly decreased the association of fluorescence-labeled CCR2B with coronin2 on Rab4A-positive endosomal membranes (Fig. 5A,C). Likewise, depletion of FLNa also diminished β2AR-receptor-loading in coronin2-positive domains (Fig. 5C). These results suggest that FLNa might be required for efficient loading of some GPCRs onto the actin-enriched sorting endosome microdomains.

If FLNa contributes to the concentration of certain GPCRs into domains for subsequent loading of transport intermediates from the sorting endosomes to the plasma membrane, its absence should alter receptor endocytic transport and recycling. We analyzed the recycling kinetics of CCR2B and β2AR in FLNa-depleted cells by using fluorescence microscopy to measure the return of the internalized receptor back to plasma membrane. Consistent with our hypothesis, knocking down FLNa produced a significant delay in the kinetics of the retrograde trafficking of the receptors to the plasma membrane (Fig. 6A,B). Meanwhile, transport of the Tfn receptor through the short or long recycling routes was unaffected in shFLNa cells (Fig. 6C,D), as shown previously (Mayor et al., 1993). Internal Tfn was normalized for the time 0 after the first stripping. The data strongly indicated that general membrane traffic from the endosomes to the cell surface was unaffected in FLNa-depleted cells, but rather that FLNa depletion specifically altered the transport of certain cargo.

Our data strongly indicate that FLNa acts as a trafficking adaptor rather than as a structural protein during endosomal trafficking to the plasma membrane. To further investigate this matter, we next asked whether the FLNa regulation of CCR2B recycling involved the FLNa actin-binding domain (ABD), which forms the orthogonal actin networks (Janmey et al., 1990). Strikingly and consistent with our hypothesis, the mouse mutant construct FLNa-ABD–DsRed, with the actin-binding domain deleted (Muriel et al., 2011), resulted in a similar level of receptor recycling to that with wild-type mouse FLNa (Fig. 6E; Fig. S3A). Confocal micrographs of mouse FLNa-ABD–DsRed denoted a prominent cytoplasmic distribution of the mutant, which hindered its visibility on vacuolar structures, which nevertheless could still be observed in some cells (Fig. S3A). These results suggest that the predominant actin cross-linking activity of FLNa might not be an important function of the protein in CCR2B recycling; however, we cannot completely rule out the possibility that the second actin-binding domain of FLNa that has been reported previously (Nakamura et al., 2007) is sufficient for receptor recycling. These results agree with our previous observations that FLNa does not affect actin–coronin2 patches but might facilitate loading of certain receptors into these microdomains.

FLNa is phosphorylated at multiple sites by several protein kinases (Nakamura et al., 2011). Among these, phosphorylation at residue S2152 (Muriel et al., 2011; Tigges et al., 2003; Travis et al., 2004) has a clear impact on FLNa function. To study whether FLNa phosphorylation is required to sustain CCR2B recycling, we analyzed the capacity of FLNa-S2152 mutants to revert receptor recycling defects in FLNa-knockdown cells by using flow cytometry to analyze shFLNa cells (Fig. 6E). Western blot analysis demonstrated that all constructs and mutants (regardless of whether they were tagged with GFP or DsRed) were expressed at similar levels (Fig. 6E,F, upper panels). We observed that the phosphorylation mimetic FLNa mutant, S2152E, supported receptor recycling similar to wild-type FLNa. However, the non-phosphorylatable FLNa-S2152A mutant only partially recovered CCR2B recycling. Consistent with this, the phospho-mimetic mutant (FLNa-S2152E–DsRed or FLNa-S2152E–GFP) was present on endosomal membranes (Fig. 6F; Fig. S3A,B), whereas the GFP- or DsRed-tagged FLNa-S1252A mutant could hardly be detected on these organelles (Fig. 6F; Fig S3A,B). This was despite the fact that both mutants were present at the cellular cortex (Fig S3B,C). This suggests that phosphorylation of FLNa at residue S2152 regulates its localization on endosomal structures.

We also investigated whether the activation of CCR2B could itself induce FLNa phosphorylation as a mechanism to control its own trafficking. For this, we utilized a specific antibody against FLNa phosphorylated at S2152. The specificity of the antibody was validated by immunoblot analysis of cells expressing the FLNa-S2152A–GFP mutant in shFLNa-transfected HEK293 cells (Fig. 7A). Only the wild-type form of FLNa but not the FLNa-S2152A mutant was detected by the antibody. Notice that only the long form of FLNa–GFP that contained the phosphorylation site was recognized by the antibody (Fig.7A, see arrows). Treatment with CCL2 induced a small but significant increase in FLNa phosphorylation, 5 and 15 min after CCL2 stimulation (Fig. 7B), which paralleled the activation of ERK by CCL2 (Fig. S4A–C). CCL2 did not induce any increase in phosphorylation of the FLNa-S2152A mutant that had been expressed in shFLNa HEK293 cells (Fig. 7A, lower panel). Different kinases have been implicated in the phosphorylation of FLNa, including several isoforms of PKC and PKA, which are also downstream effectors of GPCR signaling. Treatment with different protein kinase inhibitors targeting PKCs [GF10903, Gö6976 and bisindolylmaleimide II] (Fig. 7C–E) consistently produced a significant reduction of FLNa phosphorylation at S2152 upon CCL2 treatment. The reduction of phosphorylation of FLNa at S2152 in CCL2-stimulated cells upon treatment with GF10903 was dose dependent (Fig. 7D). In contrast, no inhibition of FLNa phosphorylation at S2152 was observed in the presence of the PKA inhibitor H-89, even at relatively high doses. β2AR stimulation also led to FLNa phosphorylation on S2152 after 5 and 15 min of isoproterenol treatment (Fig. 7F,G). Interestingly, however, phosphorylation of FLNa at S2152 induced by β2AR was not reduced in the presence of GF10903 but on the contrary, it was reduced in the presence of H-89. Again, isoproterenol stimulation did not result in any phosphorylation of the FLNa-S2152A mutant (Fig. 7F). Therefore, we can conclude that both receptors, CCR2B and β2AR, can induce phosphorylation of FLNa at S2152 upon ligand stimulation, most probably through different downstream pathways.

Our work demonstrates a very specific role of FLNa in promoting the recycling of CCR2B from sorting endosomes to the plasma membrane. We showed that FLNa and internalized CCR2B colocalized in dynamic actin- and coronin2-enriched endosomal subdomains, which are known to concentrate cargo for the Rab4A- and retromer-dependent pathway to the plasma membrane (Puthenveedu et al., 2010; Temkin et al., 2011), which is also used by other GPCRs like β2AR and AVPR2 (V2R). Interestingly, even though filamins have a prominent role as actin cross-linkers, we also found that FLNa does not seem to play a major structural role in organizing the endosomal actin-enriched microdomains, but rather that it functions as a cargo adaptor. We showed that FLNa depletion does not apparently alter the size or number of the endosomal coronin2 microdomains, nor did it affect overall retrograde traffic, but rather it specifically prevented CCR2B and β2AR loading onto the actin-enriched microdomains and specifically delayed their transport back to the plasma membrane. Also consistent with this view, the actin-binding site of FLNa that is responsible for its actin cross-linking activity was not required to sustain efficient recycling of CCR2B and β2AR. Based on our current observations and previous data demonstrating that FLNa interacts with several chemokine receptors and integrins (Liu et al., 2015), which follow the same path to the plasma membrane, we propose that FLNa works as a receptor scaffold that links cargo directly or indirectly to actin microdomains in the sorting endosome. We found that this FLNa function is regulated by its phosphorylation at S2152, which, in turn, is promoted by GPCR-signaling to second messenger kinases.

In addition to a direct role of FLNa in CCR2B and β2AR recycling, our data demonstrated that FLNa depletion greatly reduced the long-term motility of the sorting endosomes and increased their size. Depletion of WASH (Gomez et al., 2012), or the Arp2/3 complex (Derivery et al., 2012), also causes early endosome enlargement. In these conditions, endosomes present enlarged segregated subdomains with collapsed endosomal and lysosomal networks. They do not exhibit elongated tubulations and are devoid of filamentous actin. Although FLNa depletion increased the endosome size, it did not seem to alter the actin-rich endosome domains, as the number and size of the coronin2 patches remained the same and no collapse of endosomal networks was observed, which suggest that role of FLNa differs from that seen in WASH and Arp2/3 studies. Whether the endosomal phenotypes observed by FLNa depletion are a cause or a consequence of the recycling defect, or are unrelated, is a complex question to address in the future. Thus, a defect in the recycling of certain cargo might prevent maturation of endosomes and their proper attachment to microtubules for long-range movement. Reciprocally, the enlargement of endosomes might be caused not only by the accumulation of certain cargo unable to exit the organelle but also by maintenance of the endosome capacity to accept vesicles from the plasma membrane, provided that they do not move away from the cortex. Alternatively, or additionally, depletion of FLNa might directly alter tubulin-based motility, or the architecture of microtubules (Lynch et al., 2011), and as a consequence, indirectly affect endosomal motility, size and distribution (Bayer et al., 1998) without having a dramatic effect on the actin cytoskeleton structure (Baldassarre et al., 2009). In any case, a general alteration of the endosome motility or distribution is unlikely to be the primary cause of the CCR2B recycling defects because the kinetics of Tfn transport from the endosomes to plasma membrane appeared to be unaltered in FLNa-depleted cells (Fig. 6C,D; Minsaas et al., 2010), as did its uptake (Muriel et al., 2011). We favor the hypothesis that FLNa acts as a scaffold that cooperates to load the cargo into actin-enriched domains. Concomitantly, FLNa would be needed to maintain the distribution and motility of endosomes by interacting with actin and/or microtubule networks. This hypothesis agrees with our previous results showing that upon activation, CCR2B-loaded vesicles were present in a linear distribution with directional movement, co-aligned with actin fibers and FLNa (Minsaas et al., 2010; Planagumà et al., 2012). It is interesting that FLNa regulates the linear distribution of caveolae along stress fibers (Muriel et al., 2011) and their translocation from the plasma membrane to perinuclear Rab11 structures.

FLNa is regulated by phosphorylation at S2152 by many protein kinases, and this phosphorylation has a direct effect on its function in regulating a variety of cytoskeleton-related events (Hastie et al., 1997; Muriel et al., 2011; Travis et al., 2004; Zhang et al., 2012). Here, we show that phosphorylation of the same residue is necessary for efficient recycling of CCR2B and β2AR. Together with the fact that the S2152A mutant of FLNa was not present in endosomes, our data suggest that phosphorylation of FLNa at S2152 facilitates the accumulation of this protein on endosomal microdomains (see model in Fig. 7I). These results agree with previous data where Myc-tagged FLNa-S2152A localizes with actin fibers at the cell surface, whereas phosphorylation of FLNa directs it towards the cytoplasm of CHP100 cells (Zhang et al., 2012). We also showed here that CCR2B and β2AR signaling can trigger phosphorylation of FLNa at S2152. Interestingly, our data also suggests that phosphorylation of FLNa at S2152 by these GPCRs follows different pathways and uses different kinases because CCR2-induced phosphorylation of FLNa at S2152 was reduced by PKC inhibitors but not by PKA inhibitors. Reciprocally, β2AR-induced phosphorylation relied more on a PKA-dependent pathway. The data are consistent with the different signaling cascades ignited by the different GPCRs, CCR2 relying mainly on G_i and G_q which lead to PKC activation (Janjanam et al., 2015; Jiménez-Sainz et al., 2003), and β2AR on G_s, leading to PKA activation. By contrast, it is well established that FLNa can be phosphorylated at S2152 by cAMP-dependent PKA (Ithychanda et al., 2015; Jay et al., 2004). Moreover, other ligand-activated GPCRs, like angiotensin II type-1 receptor (AT1R), MAS and the α_1d-adrenergic receptor, have been shown to enhance PKA-mediated filamin phosphorylation of Ig domains 16–24 in vitro (Tirupula et al., 2015). Whether GPCR-induced FLNa phosphorylation at S2152 is directly mediated by PKC cannot be confirmed at this point, but it is possible given that different isoforms of PKC have previously been demonstrated to phosphorylate filamin A in vitro (Tigges et al., 2003). Nevertheless, FLNa S2152 is probably phosphorylated by other kinases since inhibitors of PKC or PKA did not completely obliterate CCR2- or β2AR-induced phosphorylation of FLNa at S2152, and this residue has been shown to be phosphorylated by other kinases (Gómez-Moutón et al., 2015). Interestingly, the residue S2152 of FLNa is located in Ig repeat 21, which is usually hidden in the autoinhibited conformation of FLNa (Lad et al., 2007). Filamin A needs to undergo a conformational adjustment for the phosphorylation on S2152 to occur (Ithychanda et al., 2015). Together with our present data, it reinforces the concept that ligand-dependent phosphorylation of FLNa by GPCRs becomes an important pathway for FLNa activation by promoting the phosphorylation of S2152. Overall, we favor the hypothesis that a signaling response leading to activation of different second messenger protein kinases to phosphorylate FLNa, which in turn will induce its detachment from the actin structures at the cell cortex and facilitate its localization in highly dynamic actin-enriched endosomal microdomains (Fig. 7I). This mechanism probably operates for other chemokine receptors and other GPCRs.

Plasmids encoding FLNa–EGFP, Flag–CCR2B (Planagumà et al., 2012), FLNa-S2152A–DsRed and FLNa-S2152E–DsRed (Muriel et al., 2011) have been described previously. FLNa–DsRed was a gift from David A. Calderwood (Yale School of Medicine, New Haven, CT); FLNa-S2152A–GFP and FLNa-S2152E–GFP were derived from pcDNA3.1-FLNa-EGFP. Human Rab4A–YFP and human Rab5A–GFP were provided by Peter van der Sluijs (University Medical Center Utrecht, The Netherlands). pCherry-C-hRab4A and pCherry-C-hRab5A were derived from human (h)Rab4A–YFP and hRab5A–GFP (previously sequenced). Actin–RFP was from Stein O. Døskeland (Department of Biomedicine, University of Bergen, Norway). Coronin2–GFP was kindly provided by Manojkumar A. Puthenveedu (Carnegie Mellon University, Pittsburgh, PA). Primary antibodies used were: M1 anti-Flag (#F3040, Sigma; 1/1000 dilution), anti-filamin 1 (E-3; mouse sc-17749, Santa Cruz Biotechnologies; 1/1000 dilution), anti-filamin phosphorylated at S2152 (#4761, Cell Signaling; 1/1000 dilution), anti-phospho ERK1/2 and anti-ERK1/2 (p44/42 MAPK 137F5 #4695; Cell Signaling; dilution 1/1000), and anti-β-tubulin (#T6199, Sigma; 1/1000 dilution) antibodies. Alexa-Fluor-647- (#A21245, lot#651740; #A21236, lot# 716822) and Alexa-Fluor-488-conjugated (anti-mouse #A21202, lot #536050) secondary antibodies were from Invitrogen (all used at 1:1000 dilution). CF488A- and CF568-labeled M1 anti-Flag antibodies (Sigma-Aldrich) were obtained with Mix-n-Stain™ CF™ Antibody Labeling Kit (MX488AS100-1KT, Sigma-Aldrich) following the manufacturer's instructions. A mouse anti-EEA-1 antibody (#610457, BD Transduction Laboratories) was detected by an anti-rabbit secondary antibody after treating first the antibody with Affinity Pure Fab Fragment Rabbit Anti-Mouse IgG (H+L) according to the manufacturer's instructions (#315-007-003, Jackson ImmunoResearch). For western blot infrared analysis (Odyssey, LI-COR), IRDye680- and IRDye800-conjugated anti-rabbit and anti-mouse secondary antibodies were used (800CW anti-rabbit #926-32311; 800CW anti-mouse #926-32210, lot #C1102603 from Odyssey; anti-rabbit IRDye 680RD #926-68073, lot# C30724-04 from Odyssey, all used at 1:20000 dilution). CCL2 and human MCP-1 were from PeproTech (Rocky Hill, NJ); [Arg⁸]-vasopressin acetate salt and endothelin 1 were purchased from Sigma-Aldrich and isoproterenol was obtained from Calbiochem.

HeLa cells were obtained from American Type Culture Collection and always kept under low passage number from the initial clone. HEK293 (293-EBNA) were bought from Life Technologies and cultured in the presence of G418. All cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) fetal bovine serum in a humidified atmosphere under 5% CO₂. Cells were regularly tested for mycoplasma contamination. All cells were grown and transfected utilizing FuGene^® 6 (Roche) in all experiments. Cells were treated with 20 nM of CCL2 or 10 μM isoproterenol in culture medium for 5 and 15 min prior to analysis. The following concentrations of ligands were also used: 1 µM vasopressin and 10 nM endothelin. For protein kinase inhibition, cells were treated with varying concentrations of GF109203X (inhibitor of PKCα, βI, βII and γ subtypes) (ENZO Life Sciences), H-89 (general inhibitor of protein kinase A), 2 µM Gö6976 (PKCα and PKCβ inhibitor) (Sigma-Aldrich) or 2 µM bisindolylmaleimide II (general PKC inhibitor) (Calbiochem) for 60 min. Utilizing the MISSION shRNA human system (Sigma-Aldrich), FLNa shRNA (shRNA) against human FLNa (TRCN0000062528 and TRCN0000062529) was used in HeLa and HEK293 cells. Virus production was performed as described previously (Masià-Balagué et al., 2015). Cells were solubilized in N-dodecyl-D-maltoside buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM MgCl₂, 1% N-D-maltoside, 10 mM NaF, 1 mM Na₃VO₄ and protease inhibitors), as described previously (Masià-Balagué et al., 2015). The same amount of protein was loaded for each condition (measured by BioRad protein assay). Western blots were visualized and quantified by performing infrared detection (Odyssey System) in the linear range.

For live imaging, HeLa or HEK293 cells that had been transiently transfected with FLNa–GFP, FLNa–GFP-S2152A, -S2152E, Rab4A–Cherry, Rab5–Cherry, Rab11–Cherry, actin–RFP and/or coronin2–GFP and Flag-tagged CCR2B receptors were plated on glass chambers and imaged after addition of anti-Flag antibody for 15 min at 4°C followed by an incubation with anti-mouse Alexa-Fluor-647-conjugated antibody (surface labeled) for 15 min at 4°C before addition of ligand at 37°C. HEK293 cells were also transfected with plasmids encoding Flag-tagged receptors: β2AR, Etb, V2R or PAR1. HA–β2AR was transiently transfected into HEK293 cells with Flag–CCR2B. Confocal imaging was performed with a Leica TCS SP5 (Leica Microsystems, Mannheim, Germany) by using a 63×1.3 NA oil immersion lens in a sealed chamber with 5% CO₂ at 37°C. Movies were taken between 5 and 30 min after agonist addition. Each frame corresponds to 600 ms. For high-resolution fluorescence imaging, a STED microscope (Leica Microsystems) was utilized on live HeLa cells that expressed FLNa–GFP. For fixed cell imaging, HEK293 cells that had been transiently transfected with Flag–CCR2B and plated on glass chambers were treated with CCL2 (20 nM) for 15 or 30 min at 37°C before fixation, and processed for immunofluorescent staining (Planagumà et al., 2012). After fixation and permeabilization, cells were incubated with anti-Flag and anti-EEA-1 antibodies, followed by incubation with the corresponding Alexa-Fluor-conjugated secondary antibodies (as stated in each figure legend).

Live-cell confocal imaging was performed on transiently transfected Rab4A–Cherry and/or Flag–CCR2B (surface labeled) control (scrambled shRNA) and shFLNa HEK293 cells at 37°C. Images were recorded as 16-bit sequences with a pinhole of 246 µm, giving a z-slice of 2 µm, and acquired at 3-s intervals. A 100-mW LASOS argon laser, set to 25% intensity, was used to photobleach a Rab4A endosome, and recovery was monitored for 26 cycles (10 s) with reduced laser power. A 488-nm line was used to follow the Rab4A-positive endosome and a 633-nm line was used to bleach the entire CCR2B endosome (4 µm²). Monitoring of receptor fluorescence recovery was done for 2.5 min. The mobile fraction (MF) was estimated as MF=(F_i–F_b)/(F₀–F_b)*100, where F_i is the final intensity upon recovery over the period of the experiment, F₀ is the intensity before bleaching, and F_b is the intensity immediately after bleaching. Data were analyzed for 50 endosomes under each condition and were acquired on three different days. The background was subtracted and corrected by measuring the fluorescence decay of a non-bleached area in the same cell and normalized to the Rab4A signal (Rabut and Ellenberg, 2005).

Quantification of receptor recycling was performed as previously described (Patel et al., 2011) using confocal microscopy. Cells that expressed Flag–CCR2B or Flag–β2AR were incubated with anti-Flag antibody at 4°C for 15 min, and endocytosis was initiated by addition of CCL2 or isoproterenol for 15 min at 37°C. Cells were acid-washed to eliminate traces of the antibody bound to the cell surface receptor and allowed to recover with DMEM for various time periods, washed with ice-cold PBS-Ca²⁺, fixed and incubated with Alexa-Fluor-647-conjugated antibody without permeabilization. Fluorescence intensities were averaged over individual cells and analyzed using the FiJi distribution of ImageJ. In each experiment, a minimum of four replicates were analyzed for each condition. The percentage of antibody that was recycled to plasma membrane was calculated as 1 minus the internal fluorescence values and was normalized against values at time 0 (post agonist) (results in Fig. 6A,B).

For the Tfn flow-cytometry-based recycling assay, HeLa cells were incubated with Alexa-Fluor-568-labeled Tfn at 0.02 mg/ml for 2 or 60 min at 37°C to allow internalization. At 2 min, only early endosomes are loaded with Tfn, and the kinetics of short-loop recycling can be measured. At 60 min, mainly recycling endosomes are loaded with Tfn and, therefore, traffic from the recycling endosomes is mostly measured. After acid stripping, to eliminate surface-bound fluorescent Tfn, the internalized fraction was chased at 37°C with 2 mg/ml of unlabeled Tfn for the indicated times. At each time point, cells were acid-washed, washed with ice-cold PBS and immediately detached using trypsin. Fluorescence intensity profiles of cell populations (10,000 cells per sample) were measured using a FacsAria I SORP (Becton Dickinson, San Jose, CA). Internal Tfn was normalized for the time 0 after the first stripping. The percentage of Tfn recycled was calculated as 1 minus the internal fluorescence values.

Analysis of CCR2B recycling was performed in HEK293 shFLNa cells that had been transiently expressing mouse FLNa–DsRed to rescue the recycling delay, and also cells expressing FLNa-ABD–DsRed, FLNa-S152A–DsRed and FLNa-S2152E–DsRed (Fig. 6E). Conjugation of anti-Flag M1 antibodies with CF488 was performed following the manufacturer′s instructions (Sigma-Aldrich). Conjugated CF488A–M1 anti-Flag antibody was incubated with cells for 15 min at 4°C prior to addition of 20 nM CCL2 and incubation for 15 min at 37°C to allow internalization. Then, cells were acid-washed and allowed to recycle back the receptor to the plasma membrane for the indicated times. Cells were again acid-washed, to exclusively eliminate the antibodies attached to the receptors reaching the plasma membrane, and trypsinized before analysis using a FacsAria I SORP instrument (Becton Dickinson). Green fluorescence intensity profiles were measured only in cell populations that expressed DsRed (5000 cells per sample). The fluorescence intensity measured is then a reflection of internalized receptor that remained intracellularly, which decreased with time upon chase.

Data are presented as mean values or as fold induction±s.e.m. as indicated in each figure legend. The sum of all the individual stimulatory effects were compared to that of the combined effects by using the unpaired t-test with two-tailed P-values, or one-way ANOVA (Fig. 1B) or two-way ANOVA (Fig. 1D) with GraphPad Prism. Differences were considered statistically significant when *P<0.05, **P<0.01 and ***P<0.001.

We thank E. Rebollo, T. Zimmerman (Centre for Genomic Regulation, Barcelona, Spain), A. Echarri [Centro Nacional de Investigaciones Cardiovasculares, Barcelona, Spain (CNIC)] and D. M. Pavón (CNIC) for scientific and technical assistance, and D. A. Calderwood for providing FLNa–DsRed.