CK1α protects WAVE from degradation to regulate cell shape and motility in the immune response Open Access

Cell shape changes require dynamic remodeling of the actin cytoskeleton. The WASP family verprolin homologous protein (WAVE) is a central Arp2/3 regulator driving lamellipodial protrusions and cell migration in most eukaryotic cells. Together with Abi, NCKAP1/Nap1, CYFIP/Sra-1 and BRK1/HSPC300, WAVE forms a conserved hetero-pentameric complex, the WAVE regulatory complex (WRC). Within the WRC, activity of WAVE towards the Arp2/3 complex is inhibited by intracomplex sequestration of its Arp2/3 activating domain, i.e. the verprolin homology, cofilin homology, acidic region (VCA) domain. One of the central WRC activators is the small RhoGTPase Rac1, which directly binds to the WRC subunit Sra-1 and activates the WRC by allosterically releasing the bound Arp2/3-activating VCA domain of WAVE. Several studies have shown that phosphorylation plays an important role in regulating WRC-Arp2/3-mediated actin filament branching and lamellipodia formation (Mendoza, 2013). WAVE proteins are phosphorylated at numerous sites, and several kinases have been identified as potential regulators (Mendoza, 2013). A previous in vitro study has identified multiple functional phosphorylation events within the acidic VCA domain of mammalian WAVE2 (officially known as WASF2) by casein kinase 2 (CK2), which are required for its activity (Pocha and Cory, 2009). However, the results are based on the overexpression of phosphorylation-deficient mutants in cultured NIH-3T3 cells in the presence of the endogenous wild type protein (Pocha and Cory, 2009). A more recent in vivo study confirmed that the C-terminal acidic region within the VCA domain of the Dictyostelium WAVE is basally phosphorylated at four phosphorylation sites by CK2 and suggested that a regulated dephosphorylation of a fraction of the cellular WAVE pool is a key step in its activation during pseudopod dynamics (Ura et al., 2012).

In this work, we analyze the role of phosphorylation of WAVE (also known as SCAR) in Drosophila in vivo. By using RNA interference (RNAi), we identified casein kinase 1α (CK1α) but not CK2 as an important regulator of lamellipodia formation and immune cell migration. Ck1α-mutant macrophages exhibit a stellate morphology and an altered migratory behavior that phenocopies wave-deficient cells. We also found that CK1α can interact physically with WAVE. Drosophila WAVE contains two conserved CK1α consensus sequences that are located in the N-terminal WHD and the C-terminal acidic domain, overlapping with two conserved CK2 phosphorylation sites within mammalian WAVE2. Phosphorylation-deficient mutations in the N-terminal but not C-terminal domain of WAVE can fully rescue the lethality of the wave mutant and the lamellipodium defects of macrophages deficient for wave. Loss-of and gain-of-function analysis, and pharmacological inhibition of CK1α further suggest that basal phosphorylation of VCA domain by CK1α is crucial for WAVE stability rather than its activity in vivo.

As previously shown, suppression of Arp2/3-mediated actin polymerization in arp2- or wave-depleted macrophages results in complete loss of lamellipodial protrusions (Rogers et al., 2003; Zobel and Bogdan, 2013). To screen systematically for candidate protein kinases that are required for lamellipodia formation, we used Drosophila macrophages as a model system, therefore combining many advantages of cultured cells with a genetic in vivo model system (Rüder et al., 2018; Sander et al., 2013). We tested 308 conditional transgenic RNAi fly lines targeting 162 kinases encoded in the fly genome (see Table S1). Transgene RNAis were specifically coexpressed with GFP in the macrophage lineage using the hmlΔ-Gal4 driver (Sinenko and Mathey-Prevot, 2004). Macrophages were isolated from third-instar larvae, and tested for their phenotypic effects on lamellipodia formation and cell spreading ex vivo.

Wild type macrophages acquire a round, pancake-like shape with a broad lamellipodial actin filament network (Fig. 1A). Expression of most double-stranded RNAs (dsRNAs) induced no defects in the cell morphology and lamellipodia formation (see Table S1). We identified the casein kinase 1α (Ck1α) gene as a candidate that most strongly affected lamellipodia formation and phenocopied wave-depleted cells, characterized by a prominently reduced circularity index (Fig. 1B–D,I). We only found a few more dsRNAs that affected cell shape (see Table S1).

Next, we analyzed Ck1α mutants, bearing distinct missense mutations (Ck1α^8B12, Ck1α^A and Ck1α^B). Of those, Ck1α^8B12 (Ck1α^G43D) is the only functionally characterized Ck1α allele carrying a mutation that transforms the conserved glycine residue at position 43 into an aspartic acid (G43D). It has been first described as a strong hypomorph or amorphic allele (Legent et al., 2012). Ck1α^A (Ck1α^L141M) and Ck1α^B (Ck1α^G148S) have been previously isolated in a large EMS screen, but neither allele has so far been functionally characterized (Haelterman et al., 2014). Ck1α^A carries a mutation that leads to replacement of a conserved lysine residue with methionine at position 141 (L141M), constituting the only mutation that yields removal of an H-bond – which might affect the active site of the CK1 (see 3D structure in Movie 1, dashed lines in magenta). By contrast, Ck1α^B replaces glycine with a serine at position 148 (G148S) without any obvious structural changes.

To analyze these embryonic lethal Ck1α mutations in macrophages we performed mosaic analysis with a repressible cell marker (MARCM; Wu and Luo, 2006) to generate Ck1α-mutant macrophages in a wild type animal background. Compared with control cells (Fig. 1E), isolated GFP-positive Ck1α^L141M mutant cells displayed the strongest defects in lamellipodia formation, showing a prominent stellate cell morphology compared with Ck1α^G43D and Ck1α^G148S (Fig. 1E–H, quantification of reduced circularity is shown in Fig. 1J). Thus, we confirmed CK1α as an important novel regulator of cell shape.

A possible role on cell shape has also been suggested for casein kinase 2 (CK2; Kramerov et al., 2011; Pocha and Cory, 2009), a structurally completely different enzyme that was also included in our initial RNAi screen (see also Table S1). Whereas CK1α is a monomeric serine kinase, CK2 is composed of two catalytic CK2α and two regulatory CK2β subunits that form a hetero-tetrameric (α2β2) holoenzyme (Bandyopadhyay et al., 2017). RNAi-mediated depletion of either CK2α or CK2β subunit did not significantly affect lamellipodia formation but resulted in altered blood cell homeostasis, which could be also confirmed in ck2α (also known as CkIIα)-mutant larvae (Table S1; Fig. 1K–M). Loss of ck2α function induced the formation of Atilla-positive lamellocytes at the expense of macrophages (Fig. 1L,M; quantification in Fig. 1K). Lamellocytes are giant cells that are rarely observed in healthy flies, but transdifferentiation from macrophages is dramatically induced in response to infection by parasitic wasps (Anderl et al., 2016). Transheterozygous ck2α^TIK/P1 and ck2α^TIKR/P1 mutant larvae, which lack CK2α kinase activity (Bulat et al., 2014), showed an enlarged lamellocyte compartment (≤30%) and macrophages (70%) whose average size differs compared with control macrophages (Fig. 1N). as recently suggested, these cells might be at an intermediate state of transdifferentiation (Anderl et al., 2016). Our data suggest that CK2 regulates blood cell differentiation, rather than blood cell shape.

To further examine a possible role of CK1α in cell migration, we first analyzed the migratory behavior of Ck1α^L141M mutant macrophages expressing GFP in pre-pupae. Wild type control macrophages form broad lamellipods to migrate along the epidermis (Fig. 2A). Cell trajectories show wild type macrophages migrate long distances within 20 min of acquisition (Fig. 2A; Movie 2; quantification of migration speed is shown in Fig. 2E). By contrast, Ck1α^G43D and Ck1α^L141M mutant macrophages migrate considerably slower (Fig. 2B,C,E). No significant differences were observed between (FRT, where FRT indicates the target site for the FLP recombinase) control and Ck1α^G148S mutant macrophages, suggesting that Ck1α^G148S is a hypomorph (Fig. 2D,E).

To determine the role of CK1α in directed wound response of macrophages, laser ablation experiments were performed in a single cell within the pupal wing (Fig. 3A–D, area encircled by dashed yellow line). Upon wounding, control macrophages (FRT control) switch from random to directed migration towards the wounding site. Cells were automatically tracked within the first 30 min post wounding and trajectories were constructed (Fig. 3A–D; quantification in Fig. 3E–H; Movie 3). Ck1α-deficient macrophages still respond to the cell damage, however, due to reduced lamellipodia formation, Ck1α^L141M mutant cells were impaired in their ability to migrate towards the ablated cell. To better characterize defects in the migratory behavior we first counted the number of cells that reached the wound within the first 30 min after wounding (Fig. 3E). Number of cells at the wound were normalized to the total number of cells within 10–80 µm of the ablation site. We also measured the mean track speed of mutant cells compared to that of FRT control cells (Fig. 3F). Both, cell numbers at the wound and cell speed were significantly reduced in Ck1α^L141M as well as in myospheroid (mys¹) and wave^Δ37 mutant cells (Fig. 3E,F; Movie 3). To further describe the impaired migratory behavior of Ck1α^L141M mutant cells we also measured the bias angle, i.e. the angle between the motion vector (a step of the cell) and the direction vector pointing towards the wound (Liepe et al., 2016; Weavers et al., 2016). A bias angle of ∼0° indicates the highest directionality to the wound, whereas cells with a value of ∼180° move in the opposite direction. For migrating wild type macrophages, the bias angle had values of <80° (shown in Fig. 3G,H as frequency distribution of the bias angle for each trajectory). By contrast, Ck1α^L141M mutant cells move with a bias angle between 80° and 180° (Fig. 3G,H). Ck1α-mutant macrophages showed reduced distance from the origin, with a reduced directionality (indicated by the cell bias angle; Fig. 3F). A similar, but more severely impaired migratory behavior was shown for wave^Δ37 and β-integrin (mys¹) mutant macrophages (Fig. 3F,G; Moreira et al., 2013). These results indicate that CK1α is required for proper lamellipodia formation and immune cell migration in vivo.

Given the similar cellular phenotypes, we next tested for a possible physical interaction between CK1α and WAVE. To this end, we transiently co-transfected Drosophila S2R+ cells with EGFP-tagged CK1α and Myc-tagged WAVE, followed by coimmunoprecipitation experiments. Pull-down assays using lysates from cells expressing tagged CK1α and WAVE revealed a physical interaction between the two proteins (Fig. 4A). To further examine whether CK1 can phosphorylate WAVE, we performed an in vitro kinase assay using recombinant human CK1δ kinase (1–317aa) – which has 65% identity to Drosophila CK1α – in the presence of purified glutathione S-transferase (GST)-tagged Drosophila WAVE protein (GST-WAVE). In this assay, ATPγS served as the phosphate donor from which mono-thiophosphate instead of phosphate is transferred to the substrate (Fig. 4B). Alkylated thiophosphate creates an epitope for thiophosphate ester-specific antibodies, which allows detection of phosphorylation (Allen et al., 2007). Our thiophosphorylation assay showed that WAVE can be phosphorylated by CK1 and its positive control Abelson kinase (Abl) (Leng et al., 2005) (Fig. 4B).

We next analyzed the potential effect of CK1α on endogenous WAVE protein. First, we used the cell-permeable, CK1α-specific ATP-competitive inhibitor D4476 (Rena et al., 2004). Remarkably, inhibition of CK1α activity by D4476 resulted in a significant reduction of WAVE protein levels in S2 cells (Fig. 4C, quantification is shown in Fig. 4D). This was prevented by addition of the proteasome inhibitor MG132, suggesting that phosphorylation of WAVE by CK1α protects WAVE from ubiquitin-mediated degradation (Fig. 4C,D). In addition, we found that CK1α depletion through RNAi results in substantial reduction of WAVE protein levels in larval macrophages (Fig. 4E). Quantification of nine independent experiments is shown in Fig. 4F. To further analyze whether inhibition of ubiquitin-mediated degradation can also revert the lamellipodium defects evoked by depletion of CK1α, we screened for proteasome components and tested various RNAi lines. Ubiquitin-dependent degradation is a multi-step process that involves members of the cullin protein family as part of E3 ligase complexes (Morreale and Walden, 2016) as well as the 26S proteasome, consisting of a 20S catalytic core and a 19S regulatory complex (Saeki, 2017). Upon inhibition of proteasomal degradation by targeting either the 20S catalytic core (Prosβ5, Prosβ7) or the 19S regulatory complex (Rpn11, Rpn6), we found that the lamellipodium defects induced in response to RNAi of Ck1α were significantly rescued, whereas knockdown of rpn11, rpn6, prosβ5 and prosβ7 alone did not show significant differences regarding cell shape (Fig. 5A,C). We also tested the six known Drosophila members of the cullin protein family (Cul1, 2, 3, 4, 5 and 6), which function as scaffolds to assemble distinct E3 ligase complexes (Ketosugbo et al., 2017). Interestingly, RNAi-mediated knockdown of Cul1 and Cul2, but not Cul3 to Cul6 (RNAi had already been validated), rescued cell shape defects of macrophages evoked by RNAi of Ck1α (Fig. 5B,C). Knockdown of any cullin protein alone did not show significant differences in cell shape (Fig. 5B,C). Taken together, these data further provide evidence that CK1α protects WAVE from ubiquitin-mediated proteasomal degradation to ensure the proper formation of lamellipodia.

We also tested whether forced overexpression of CK1α might increase WAVE protein levels. For this, we generated a stably transfected S2 cell line expressing CK1α under the control of a Cu²⁺-inducible metallothionein promoter (pMT). However, induced expression of full-length CK1α did neither induce a phosphorylation-dependent mobility shift of endogenous WAVE nor did it yield increased levels of WAVE protein (Fig. S1A,B), suggesting that prominent basal phosphorylation by CK1α already stabilizes endogenous WAVE levels.

CK1α is a monomeric, serine/threonine-specific protein kinase that recognizes the canonical consensus sequence [S(p)/T(p)-X-X-S/T] [where S(p)/T(p) indicates a phosphorylated residue (Flotow et al., 1990) and X represents any amino acid]. In addition, non-canonical consensus sequences recognized by CK1 family members have been described previously, i.e. the SLS motif as found in β-catenin (Marin et al., 2003). WAVE proteins contain a conserved SLS motif in the N-terminal WAVE homology domain (WHD) and a canonical CK1α consensus sequence in the C-terminal acidic region comprising VCA (Fig. 6A, marked in red). To further explore the physiological relevance of phosphorylation, we performed rescue experiments in fly. We generated different mutant WAVE variants, in which the serine residues within the N-terminal WHD (i.e. S44, S45 and S47) or the C-terminal VCA domain (i.e. S591, S593, S596, S598 and S602) were replaced with unphosphorylatable alanine residues yielding SA mutants (N-terminal SA^3x and C-terminal SA^5x mutant variants) (Fig. 6A) or with phosphomimetic aspartic acid residues yielding SD mutants (i.e. N-terminal SD^3x- and C-terminal SD^5x mutant variants) (Fig. 6A). To ensure equal expression rates we integrated these transgenes into the same landing site (68E) using the ΦC31-mediated transgenesis strategy (Bischof et al., 2007).

Ubiquitous re-expression of N-terminal phosphomutant or phosphomimetic variants (SA^3x or SD^3x) fully rescued embryonic lethality of wave mutants, indicating that the SLS motif is dispensable for WAVE function (Fig. 6B). By contrast, phosphorylation of the C-terminal acidic domain of WAVE is crucial for WAVE function. With only a small number of progenies, the phosphomutant variant (SA^5x) failed to rescue the lethality of the wave mutant, whereas the C-terminal phosphomimetic variant (SD^5x) completely restored viability (Fig. 6B). A similar phenotypic analysis in macrophages further confirmed that basal phosphorylation of the C-terminal acidic domain is required for WAVE function (Fig. 6C). Defects regarding lamellipodia formation in wave-mutant macrophages can be substantially rescued by re-expressing wild type, N-terminal phosphomimetic SD^3x or phosphomutant SA^3x WAVE protein (Fig. 6C). However, C-terminal phosphomutant SA^5x WAVE failed to rescue the lamellipodium defects (Fig. 6C). Cells expressing the SA^5x variant still showed strongly reduced circularity and mean track speed compared to cells rescued either by wild type, SA^3x or SD^5x protein (quantification in Fig. 6D and E). Similarly, we still found significant defects in the migratory directionality of mutant macrophages expressing phosphomutant SA^5x WAVE compared to wild type or the phosphomimetic SD^5x variant (quantification in Fig. 6F). Finally, we tested whether defects in lamellipodia formation evoked by RNAi of Ck1α can be rescued by overexpression of the phosphomimetic SD^5x variant. Indeed, overexpression of the phosphomimetic SD^5x but not the phosphomutant SA^5x WAVE variant substantially rescued cell shape defects of macrophages deficient for Ck1α (Fig. 6G). Likewise, re-expression of wild type WAVE did not rescue defects evoked by RNAi of Ck1α, suggesting that basal phosphorylation of the acidic domain of WAVE is essential for its stability in vivo.

We finally tested whether a phosphomimetic WAVE SD mutant of the VCA domain exhibits a higher protein stability or actin polymerization activity in vivo. Here, we used the Drosophila wing imaginal disc as an in vivo model to measure the effect of the WAVE overexpression. We used the en-Gal4 driver, which only induces expression in the posterior compartment of wing imaginal discs, whereas the anterior compartment serves as a negative control (Fig. 7A–E, schematic in F). Expression of an EGFP transgene served as an additional negative control (Fig. 7A,A′) and that of wave RNAi transgene as a positive control (Fig. 7E,E′) for changes in F-actin levels. We found that phosphomimetic WAVE-SD^5x (Fig. 7C,C′) is more stable compared to WAVE-wild type (WAVE-WT (Fig. 7B,B′) and phosphomutant WAVE-SA^5x (Fig. 7D,D′,G). Protein levels of WAVE-WT and WAVE-SA^5x were not significantly different (Fig. 7G). Moreover, despite the fact that phosphomimetic WAVE-SD^5x is more stable than the wild type protein, we found no increased activity (F-actin induction) in WAVE-WT and WAVE SD^5x cells (Fig. 7H). Thus, these data suggest that phosphorylation of the VCA domain promotes WAVE stability rather than its actin nucleation activity in vivo.

WAVE proteins contain a conserved C-terminal VCA domain that directly binds to and activates the Arp2/3 complex, driving branched actin polymerization. A previous study suggested that CK2 phosphorylates serine of mammalian WAVE2 at positions 482, 484, 488, 489 and 497 within the acidic domain of the VCA domain that promotes Arp2/3 complex activity in vitro (Pocha and Cory, 2009). Consistently, overexpression of a GFP-WAVE2 5A construct inhibits lamellipodial protrusion in transfected NIH3T3 cells (Pocha and Cory, 2009). Whether a phosphomimetic WAVE SD mutant of the VCA domain exhibits increased actin polymerization activity in vitro has not been addressed so far. In this work, we found that overexpression of phosphomimetic WAVE-SD^5x and wild type WAVE equally induces F-actin in epithelial tissue. This suggests that increased stabilization did not necessarily result into increased actin nucleation activity. However, inhibiting the phosphorylation of the VCA domain (see WAVE-SA^5x variant) clearly leads to significantly reduced F-actin induction as compared to overexpression of wild type WAVE (WAVE-WT). Therefore, basal phosphorylation of the VCA domain not only seems to be required for protein stability but also seems to promote WAVE activity. However, the phosphomimetic SD^5x variant behaves like the wild type WAVE as it fully rescues wave-mutant lethality and any defects in lamellipodia formation. Thus, our data more closely resemble previous observations of WAVE2 in NIH3T3 cells (Pocha and Cory, 2009). Our findings also imply that dephosphorylation is a crucial regulatory step in the regulation of WAVE. Our data further suggest that basal phosphorylation by CK1α protects WAVE against increased ubiquitin-dependent degradation. However, the question remains how phosphorylation of the VCA domain by CK1α impacts on WAVE stability? A recent study has demonstrated that WAVE2 undergoes ubiquitylation in a T-cell activation-dependent manner that is followed by proteasomal degradation dependent on the VCA domain (Joseph et al., 2017). The main WAVE2 ubiquitylation site has been mapped to Lys45, a highly conserved residue within the WHD, and is required for WRC integrity and stability (Joseph et al., 2017). A model has been proposed in which activation of WAVE triggers a conformational change that releases the sequestered VCA domain, exposes the WHD (Lys45), to ubiquitylation and, thereby, promotes WAVE degradation (Joseph et al., 2017). Overall, our data provide first in vivo evidence for CK1α-dependent phosphorylation of WAVE as a so far unknown and crucial mechanism to control WAVE functions (Fig. 8).

Our data also show that CK1α – but not CK2 – is an important regulator of WAVE function in vivo. In contrast to Ck1α-mutant cells, macrophages deficient for the catalytic CK2α subunit formed normal lamellipodia but showed increased differentiation of lamellocytes at the expense of macrophages. Our data suggest that CK2 regulates differentiation, rather than shape and locomotion, of blood cells. Accordingly, previous data revealed that phosphorylation of dMi-2, the catalytic subunit of the nucleosome remodeling and deacetylase (NuRD)/Ush complex, by CK2 modulates nucleosome remodeling activity, which might contribute to the repression of blood lineage-specific genes (Bouazoune and Brehm, 2005; Lenz et al., 2021).

Fly husbandry and crossing were carried out according to the standard methods. Crosses were maintained at 25°C, UAS-Gal4-based experiments including RNAi were performed at 29°C. The following fly lines were used: CK1α RNAi BL 25786 and VDRC 16645, hmlΔ-Gal4, UAS-eGFP (Sinenko and Mathey-Prevot, 2004); en-Gal4 (Bloomington Stock Center). Transgenic pUASp-wave WT, pUASp-wave SD^3x, pUASp-wave SA^5x, pUASp-wave SD^5x and pUASp-wave SA^5x flies were generated using ΦC31-mediated transgenesis (M{3xP3-RFP.attP}ZH-86F (Bischof et al., 2007).

To generate MARCM-ready stocks the following fly lines from Bloomington Stock Center were used: hsFLP, tubP-GAL80, w*, FRT19A; Pin/CyO; (BL# 5133); FRT19A;; (BL# 1709); mys¹FRT19A/FM7c;; (BL# 23862), y¹w* Ck1α^[A] FRT19A/FM7c, Kr-GAL4, UAS-GFP (BL#57084), y¹w* Ck1α^8B12 FRT19A/FM7i, twi-GAL4, UASt-GFP; sna^[Sco]/CyO (BL# 63802); y¹w* Ck1α[A] FRT19A/FM7c, Kr-GAL4, UAS-GFP (BL# 64459); hsFLP, w*, y; tubP-GAL80, FRT40A; hmlΔ-GAL4, UAS-eGFP/TM6B; hsFLP; wave^Δ37, FRT40A/Cyo; y* w*;; ck2α^[TikR]/TM3 (BL# 24511); y* w*;; ck2α^TIK/TM3 (BL# 24512); y* w*;; ck2α^P1/TM6B, Tb1 (Kyoto 141869). MARCM (Mosaic Analysis with a Repressible Cell Marker; Wu and Luo, 2006) experiments were performed as follows.

Deleted in hemocytes only (DEMON) (Moreira et al., 2013) males were crossed with mutant or FRT19a control virgin flies and placed at 25°C for 48 h. The progeny was submitted to three 1-h heat shocks every 24 h at 37°C. Between heat shocks, crosses were maintained at 29°C. Each heat shock was carried out in a 37°C water bath, followed by 1 h at 18°C to extend the G2 phase and improve MARCM efficiency. Female third-instar larvae containing GFP-expressing hemocytes were then selected for further analysis.

Drosophila S2R⁺ cells were propagated in 1× Schneider's Drosophila Medium, as described previously (Stephan et al., 2008). S2R⁺ cells were transfected as described previously (Nagel et al., 2017).

Drosophila S2R⁺ cells were transfected with 11.37 µg pMT-Ck1α-6×c-myc plasmid expressing Myc-tagged CK1α (Lam et al., 2018). For selection of co-transfected cells, we used the pCoHygro (Invitrogen) selection vector (1.26 µg) expressing the hygromycin resistance gene. Three days after transfection, medium was replaced and cells were cultured in the presence of 300 μg/ml hygromycin-B (ThermoFisher). Stably transfected cells were screened for hygromycin resistance over 5 weeks. During this time, the medium was changed every 4–5 days. Expression of CK1α was induced by addition of CuSO₄ (final concentration of 500 µM). Cells were harvested at indicated time points.

The CK1α inhibitor D4476 (Sigma) was resuspended in DMSO to 5000 µM stock dilution. D4476 was then diluted in cell culture medium and added to cells at final concentrations of 100 µM in 24-well tissue culture plates. Cells were treated for 4 h. The proteasome inhibitor MG132 (Sigma) was resuspended in DMSO to 1 mM stock dilution. Cells were treated with 10 µM MG132 for 4 h. Control-treated cells were treated with equal volumes of DMSO under conditions identical to those of drug treatment.

Phosphorylation assays were conducted using recombinant glutathione S-transferase (GST)-tagged Drosophila WAVE protein (GST-WAVE) as substrate (0.2 µg/µl in 50 mM Tris buffer pH 8.0). First, GST-WAVE was incubated with ATPγS. Which serves as the phosphate donor from which mono-thiophosphate instead of phosphate is transferred to the substrate. Second, alkylation of thiophosphorylated serine, threonine or tyrosine residues is allowed by addition of para-nitrobenzylmesylate (PNBM). Alkylated thiophosphate creates an epitope for thiophosphate ester-specific antibodies, which allows to detection phosphorylation (Allen et al., 2007). Kinases used were CK1 (truncated human δ isoform) (Graves and Roach, 1995) and Abl [truncated form of murine leukemia virus v-Abl (Foulkes et al., 1985)], both purchased from NEB. Reactions were mixed and incubated for 30 min at 30°C before PNBM was added for alkylation. Alkylation with PNBM occurred for 120 min at 30°C.

Co-immunoprecipitation experiments were performed as described previously (Bogdan et al., 2004). Samples were used for SDS PAGE and western blot analysis as described below.

Protein extracts were separated by SDS-PAGE and analyzed by western blotting. The following antibodies were used: anti-WAVE (1:1000; Bogdan et al., 2004), anti-Tubulin (1:3000, DSHB AA4.3) and anti-thiophosphate ester (rabbit, ab133473, Abcam). The following secondary antibodies were used diluted 1:5000 in 10% milk TBS-T: goat anti-mouse IgG (H+L), HRP (ThermoFisher) and goat anti-guinea Pig IgG (H+L), HRP. Western blots were quantified using Image studio lite 5.2 software from Li-Cor and statistically analyzed using GraphPad Prism 8 software.

Pupal macrophages were isolated as described previously (Ruder et al., 2018). Wing imaginal discs were dissected from third-instar larvae and collected in ice-cold PBS. The discs were fixed with 4% PFA for 45 min at RT on a rotary mixer. 3% (w/v) BSA in PBS, 0.3% Triton X-100 was used as blocking solution. Samples were incubated in primary antibody over night at 4°C. 0.3% PBT was used for every washing step. Primary antibodies used were anti-WAVE (1:1000; Bogdan et al., 2004) and anti-Atilla (1:10; Kurucz et al., 2007). Secondary antibody was goat anti-guinea pig 488 (1:1000, Thermo Fisher Scientific). Actin staining was carried out using Alexa Fluor 568 conjugated to phalloidin (1:100) during the secondary antibody incubation for 2 h. Discs were mounted in Fluoromount G (Thermo Fisher Scientific) and stored at 4°C.

Structure illumination microscopy images were taken with an ELYRA S.1 microscope (Cell Observer SD, 63×/1.4 oil-immersion objective). Confocal fluorescence images were taken using a Leica TCS SP8 with an HC PL APO CS2 63×/1.4 oil objective. Live imaging of macrophage cultures was performed using a Zeiss CellObserver Z.1 with a Yokogawa CSU-X1 spinning disc scanning unit and an Axiocam MRm CCD camera (6.45 µm×6.45 µm). Ablation experiments were done using a 355 nm pulsed UV laser (Rapp, Optoelectronics), as reported previously (Sander et al., 2013; Rüder et al., 2018).

Molecular visualization, editing and analysis, and image production were carried out using the UCSF Chimera package (Pettersen et al., 2021).

Cell morphology was analyzed by using FIJI shape descriptor parameter. Circularity ranges were between 0 (infinitely elongated polygon) and 100 (perfect circle), 4π×area÷perimeter² (Zdilla et al., 2016).

Tracking of migrating macrophages was performed using the spots module of Imaris 9.3 (Bitplane; https://imaris.oxinst.com/versions/9-3) software. The reference frame module was set at the ablation site. After automatic tracking, all time-lapse movies were checked and were manually corrected if neccessary. The mean track speed was measured by using the Imaris software and values were analyzed with Graph Pad Prism. The bias angle between the vector towards the ablated cell and the direction vector of the cell was calculated in R software (R Studios Version 1.4). The angle between the vector directly towards the ablation cell and the direction vector of the cell at each time point was calculated using ⁠,

i.e. the scalar product of vectors a and b divided by the multiplication product of the length of each vector ⁠.

Polar histograms were generated using the package ggplot2. For directed migration only cells within a 10–80 µm radius of the wounding site were analyzed. Results were statistically analyzed with GraphPad Prism 8.

Confocal microscopy images were processed and quantified with FIJI software. F-actin and WAVE intensities were quantified within the same plane. The intensities of three different regions (10×10 µm size) within the posterior and anterior compartment were quantified for each experiment. The integrated density of each square was measured using FIJI software. The mean value of each side was taken to calculate the quotient of posterior over anterior intensity.

Results were statistically analyzed with GraphPad Prism 8.

We thank the Bloomington Stock Center and VDRC for fly stocks. We thank Moritz Sander (Institute for Neurobiology, University of Münster, Germany) and Benedikt Nagel (Institute of Physiology and Pathophysiology, Philipps-University Marburg, Germany) for performing initial RNAi experiments. We thank Doris Wagner and Kirsten Ramlow (both Institute of Physiology and Pathophysiology, Philipps-University Marburg, Germany) for establishing and testing stable S2 cells and antibodies, Katja Rust and Mailk C. Bischoff (Institute of Physiology and Pathophysiology, Philipps-University Marburg, Germany) for critical reading of the manuscript.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258891.