Concerted regulation of actin polymerization during constitutive secretion by cortactin and PKD2

Secretion is driven by highly dynamic processes that govern cargo sorting, carrier biogenesis and fission of various transport carriers from the Golgi compartment (Egea et al., 2006; Pfeffer, 2010). The actin cytoskeleton and its rapid changes in local network density are critical factors determining Golgi structure and biogenesis of Golgi-derived transport carriers (Egea et al., 2006). To this end, actin filaments and regulatory proteins are localized to and interact with Golgi membranes (Kienzle et al., 2014; Stamnes, 2002; Valderrama et al., 1998, 2001, 2000). One important function of the actin cytoskeleton at the Golgi compartment is the maintenance of cisternae continuity (Egea et al., 2006; Valderrama et al., 1998). Moreover, actin recruitment and polymerization facilitated by the Rho-GTPase CDC42 (Dubois et al., 2005; Luna et al., 2002; Matas et al., 2004), nucleation-promoting factors (NPFs), such as N-WASP (also known as WASL) (Alekhina et al., 2017; Bhattacharya et al., 2016; Matas et al., 2004), and nucleators, like the Arp2/3 complex (Dubois and Chavrier, 2005; Matas et al., 2004) as well as actin-regulatory proteins, for example, cortactin (Cao et al., 2005; Kessels and Qualmann, 2005) determine transport carrier budding and fission. Carrier separation is further facilitated by pinching-off vesicles through oligomerization of dynamin-2, which was also shown to interact with cortactin and recruit the nucleation promoting factor N-WASP (Cao et al., 2005; Shin et al., 2008). Additionally, cortactin facilitates post-Golgi transport processes (Cao et al., 2005). However, the molecular coordination between vesicle biogenesis and actin assembly processes during post-trans-Golgi-network (TGN) transport remains to be investigated in more detail. Arp2/3-mediated actin nucleation has been linked to AP-1 transport carrier fission from the TGN (Anitei et al., 2010), where N-WASP and Arp2/3-dependent actin polymerization towards membranes promotes tubule formation (Anitei et al., 2010). In addition, BAR-domain-containing proteins, dynamin-2 and cortactin have been reported to bind to tubular carriers in order to help maintain tubule elongation and fission (Anitei et al., 2010). Nucleation by the Arp2/3 complex requires activation through class I NPFs, like N-WASP, which utilize their respective verprolin homology (WASP homology 2, WH2) domains (Helgeson and Nolen, 2013). In resting cells, N-WASP is kept in an autoinhibited inactive state. The only physiological inhibitor of WASP-mediated actin assembly is the WAS/WASL-interacting protein family member 1 (WIP, also known as WIPF1) (Ho et al., 2004; Moreau et al., 2000; Ramesh and Geha, 2009; Takano et al., 2008). WIP can stabilize and sequester N-WASP in its inactive, auto-inhibited configuration (Ho et al., 2004; Ramesh and Geha, 2009). A number of proteins, including CDC42 or Src homology 3 (SH3) domain-containing proteins, such as Nck, Grb2, cortactin, and Toca-1, activate WASP family NPFs (Helgeson and Nolen, 2013; Ho et al., 2004; Kowalski et al., 2005; Mizutani et al., 2002; Scaplehorn et al., 2002; Tomasevic et al., 2007). N-WASP has been shown to localize to the Golgi complex, but its role at the TGN is controversial (Anitei et al., 2010; Matas et al., 2004). Cortactin is present at the TGN (Anitei et al., 2010; Kessels and Qualmann, 2005), but it is only a weak class 2 NPF, since it cannot recruit actin monomers to the Arp2/3-complex. In the presence of N-WASP, branching nucleation by cortactin is substantially increased, demonstrating a striking synergy between the two NPFs (Helgeson and Nolen, 2013). Thus, an obligatory displacement model has been proposed, whereby cortactin specifically targets nascent branch junctions to displace N-WASP, which stalls completion of nucleation by the Arp2/3 complex until it is removed by cortactin, providing an additional means of control over network branching (Helgeson and Nolen, 2013).

In order to aid in membrane-carrier budding and fission at the Golgi complex, local actin polymerization needs to be tightly controlled and precisely timed. This may be achieved by several molecular mechanisms, which include the regulation of phosphatidylinositol 4,5-bisphosphate (PIP2) levels (Higgs and Pollard, 2001; Rohatgi et al., 2000; Tomasevic et al., 2007) and regulation of CDC42 activity by GTPase-activating proteins (GAPs) or guanine nucleotide exchange factors (GEFs) (Dubois et al., 2005; Estrada et al., 2001; Ménétrey et al., 2007). Our group and others have previously described another important regulator of vesicle fission at the TGN: protein kinase D (PKD) (Baron and Malhotra, 2002; Cruz-Garcia et al., 2013; Eiseler et al., 2016; Ghanekar and Lowe, 2005; Hausser et al., 2005; Liljedahl et al., 2001; Malhotra and Campelo, 2011; Pusapati et al., 2010). PKD kinases phosphorylate substrates, such as phosphatidylinositol-4 kinase IIIβ (PI4K3β) (Hausser et al., 2005) and CERT (Fugmann et al., 2007) to alter the lipid environment in order to facilitate vesicle scission from the TGN. Moreover, we have reported a phosphorylation independent interaction of PKD2 (also known as PRKD2) with the vital GTPases ARF1 and ARL1, governing vesicle biogenesis processes at the TGN (Eiseler et al., 2016; Pusapati et al., 2010). Moreover, dimers between PKD2 and PKD3 (also known as PRKD3) isoforms have been reported to coordinate secretion from the TGN (Bossard et al., 2007). Interestingly, PKDs also control actin nucleation and polymerization via their substrate cortactin (Eiseler et al., 2010). Cortactin is phosphorylated by PKD at S298, impairing actin nucleation by the Arp2/3 complex in vitro (Eiseler et al., 2010). Here, we investigated whether PKD2 is also involved in tuning and timing actin polymerization processes during vesicle scission at the TGN via its substrate cortactin.

In addition to a wide subcellular distribution near the plasma membrane, in the cytoplasm and at the Golgi complex, the Arp2/3 actin nucleation complex and cortactin have been reported to partially localize at the TGN, (Anitei et al., 2010). In accordance with these results, we also find perinuclear enrichment and partial colocalization of endogenous Arp2 and cortactin with the trans-Golgi markers TGN-46 and golgin-97 (also known as TGOLN2 and GOLGA1, respectively) (Eiseler et al., 2016), respectively (Fig. 1A,B). We also show a partial colocalization of cortactin–GFP with endogenous dynamin at golgin-97-positive compartments, which is in line with a putative implication of cortactin during vesicle fission (Anitei et al., 2010) (Fig. 1C). A role during carrier biogenesis at the TGN has also been described for the nucleation-promoting factor N-WASP (Anitei et al., 2010). N-WASP is an important activator of the Arp2/3 complex (Higgs and Pollard, 2001; Ti et al., 2011). However, a localization of N-WASP at the TGN is controversially debated, since ectopically expressed N-WASP–GFP was also shown to localize at the cis-Golgi compartment, and not the TGN (Matas et al., 2004). In order to investigate the TGN localization of endogenous N-WASP with higher resolution and contrast, we utilized Huygens classic maximum likelihood estimation (CMLE) deconvolution. Indeed, we were able to demonstrate signal enrichment at golgin-97-positive structures, enabling a putative role for N-WASP in modulating Arp2/3 complex-mediated actin dynamics at the TGN (Fig. 1D). Staining specificity was validated in N-WASP-depleted cells. (Fig. S1A). We also quantified colocalization of cortactin and N-WASP with golgin-97 in confocal image sections, respectively. Statistical analysis corroborated partial colocalization with TGN membranes, as indicated by a mean Pearson's colocalization coefficient (PCC) of 0.22 for cortactin and of 0.495 for N-WASP (Fig. 1E). The more-prominent colocalization of N-WASP with golgin-97 may be explained by direct membrane anchoring of N-WASP through interaction with PIP2 lipids (Papayannopoulos et al., 2005), whereas cortactin is not known to directly associate with Golgi membranes (Schnoor et al., 2018). To verify a broader function of the respective proteins at the TGN, we have further investigated colocalization of Arp2 with TGN46 as well as of cortactin and N-WASP with golgin-97 in differentiated polarized endothelial monolayers (Fig. S1B–D). These data indicated a partial colocalization with TGN membranes, suggesting a role for cortactin, N-WASP and the Arp complex during actin remodeling at the TGN.

During endocytosis, carrier biogenesis and scission from the plasma membrane is supported by the generation of branched actin networks that aid in carrier biogenesis and exert forces on vesicles to facilitate fission (Qualmann et al., 2000; Smythe and Ayscough, 2006). Similar actin networks are thought to drive carrier scission at the TGN, as demonstrated for AP-1 tubular carriers by Anitei et al. (2010). PKD regulates actin polymerization and formation of branched actin networks by the Arp2/3 complex via phosphorylation of the actin-regulatory protein cortactin at S298 (Eiseler et al., 2010). A connection of PKD2 to vesicle biogenesis and fission cycles at the TGN, established previously by our group and others (Eiseler et al., 2016; Hausser et al., 2005; Pusapati et al., 2010), prompted us to investigate whether phosphorylation of cortactin by PKD2 might be involved in controlling actin polymerization and constitutive secretion from the TGN. To this end, we first performed proximity ligation assays (PLA) to visualize the molecular proximity of endogenous PKD2 and cortactin at TGN structures, using confocal microscopy (Fig. 2A,B). During these experiments, furin–GFP was used to identify the TGN as indicated by quantitative colocalization with TGN46 and golgin-97 (PCC of 0.77), whereas the TGN markers showed a PCC of 0.79 (Fig. S2A). We further validated molecular proximity/interaction of PKD2–GFP and endogenous cortactin at golgin-97-positive structures using acceptor-photobleach Förster resonance energy transfer (AB-FRET). GFP and PKD2-P275G–GFP, which is no longer effectively recruited to the TGN (Pusapati et al., 2010), were used as respective negative controls for AB-FRET experiments (Fig. 2C,D). Additionally, we employed co-immunoprecipitation (co-IP) experiments to generally recapitulate an interaction of PKD2–GFP with endogenous cortactin (Fig. 2E,F). At the TGN, formation of PKD2–PKD3 dimers has been reported to coordinate secretion (Bossard et al., 2007). We have previously shown that PKD1 and PKD2 can phosphorylate cortactin (Eiseler et al., 2010). To test, whether endogenous PKD3 also phosphorylates cortactin, we performed additional experiments using PRKD3-depleted HeLa CRISPR cells. As shown in IP experiments in (Fig. S2B), depletion of PKD3 impaired phosphorylation of endogenous cortactin at S298. Thus, these data suggest that PKD2–PKD3 dimers could be also implicated in regulating cortactin functions by means of phosphorylation at S298.

The Rho-GTPase CDC42 also localizes to the TGN (Musch et al., 2001). CDC42 activates Arp2/3-complex-mediated actin polymerization through N-WASP (Bhattacharya et al., 2016; Prehoda et al., 2000; Rohatgi et al., 2000) and our data indicate that N-WASP is partially localized at the TGN. Phosphorylation of cortactin at S298 by PKD2 impairs synergistic actin polymerization in vitro (Eiseler et al., 2010). This prompted us to investigate, whether CDC42 could inhibit PKD activity and thereby the inhibitory phosphorylation of cortactin. Indeed, inhibition of CDC42 activity in HEK293T cells, using the specific CDC42 inhibitor ML141 (Hong et al., 2013), resulted in significantly enhanced PKD activity, as demonstrated by increased its increase autophosphorylation at S916 (pS916) (Fig. 3A,B). Activation of PKD2 at the TGN in HeLa cells, which predominantly express PKD2 (Yeaman et al., 2004), was quantified upon treatment with ML141 by analyzing pS916 autophosphorylation using quantitative immunofluorescence. These experiments indicated significantly enhanced PKD activity upon inhibition of CDC42 at the TGN (2.29-fold), but also in the peripheral cytoplasm of cells (2.08-fold) (Fig. 3C,D). To further address an isoform-selective PKD2 activation at the TGN, we quantified PKD2 activation by measuring the FRET between PKD2–GFP and the anti-pPKD-S742/744 activation loop antibody, labeled with Alexa-Fluor-568 secondary antibodies, with and without co-expression of constitutively active CDC42-Q61L or inactive CDC42-T17N (Fig. S2C,D). In line with the ML141 experiments, inactive CDC42-T17N significantly enhanced PKD2 activity at the TGN, as demonstrated by increased percentage FRET values with respect to vector controls, whereas constitutively active CDC42-Q61L markedly inhibited PKD activity (Fig. S2D). These data demonstrate that the important actin-regulatory GTPase CDC42 can act upstream of PKD2 at the TGN.

To determine whether CDC42, PKD2, cortactin and N-WASP are implicated in regulating actin polymerization at the TGN, we measured LifeAct–Ruby (Riedl et al., 2008) fluorescence intensities over time, as a marker for F-actin formed under steady-state conditions. A baseline for detection of fluorescence intensities was set by bleaching the LifeAct signal at the TGN, indicated by furin–GFP. To validate LifeAct–Ruby measurements, we first evaluated F-actin intensity upon co-expression of the constitutively active CDC42-Q61L as well as upon knockout of PKD2 by CRISPR/Cas9 in HeLa cells. The PKD2-knockout (KO) was verified on the protein level (Fig. S2E) and impaired secretion was demonstrated by evaluating secretion of the artificial cargo HRP (Fig. S2F). In line with the vital functions of PKD2 in modulating the lipid environment during vesicle fission at the TGN, HRP secretion from PKD2 KO cells was significantly impaired with respect to parental controls. Moreover, both active CDC42 and PKD2 knockout significantly increased mean LifeAct–Ruby fluorescence at the end of the 130 s recovery period (t=155 s), indicating enhanced steady-state F-actin generation (Fig. 3E,F; Fig. S3A,B, Movies 1–3).

Next, we determined steady-state F-actin at the TGN upon inhibition of CDC42 in parental and PKD2 KO cells using two specific CDC42 inhibitors: ML141 (10 µM) (Hong et al., 2013; Surviladze et al., 2010) and casin (5 µM) (Florian et al., 2012). We verified the inhibition of CDC42 by active CDC42-PBD-pulldown assay (Fig. S3C). Indeed, inhibition of CDC42 activity by both inhibitors significantly impaired actin polymerization at the TGN in PKD2KO cells, where F-actin turnover was reduced back to basal cell levels (Fig. 3G,H; Fig. S3D). In parental cells, Casin significantly reduced F-actin formation, whereas ML141 treatment showed a non-significant reduction of actin polymerization at the TGN. Thus, our data indicate that CDC42 can act upstream of PKD2 to control actin dynamics at the TGN.

We also investigated, whether ectopic expression of PKD2 inhibits actin polymerization in parental and PKD2 KO cells. Overexpression of PKD2 strongly decreased LifeAct–Ruby signals at the TGN in parental cells. In PKD2 KO cells, F-actin signals were significantly increased and re-expression of PKD2 was able to partially reverse this phenotype (Fig. 4A,B; Fig. S3E,F, Movies 4–7).

We then tested, whether enhanced generation of F-actin at the TGN in PKD2 KO cells was significantly reversed to parental with siLacZ (parental/siLacZ) levels by depletion of cortactin (siCTTN) (Fig. 4C,D; Fig. S3G,H, Movies 8-11). Indeed, cortactin knockdown in PKD2 KO cells significantly impaired steady-state actin polymerization to basal levels. In parental cells, depletion of cortactin also reduced formation of F-actin; however, changes were not statistically significant. Thus, these data demonstrate that cortactin is indeed implicated in enhanced actin polymerization at the TGN, controlled by PKD2.

We have previously demonstrated in vitro that cortactin that is not phosphorylated at S298A enhances synergistic actin polymerization by the Arp2/3-complex in the presence of an NPF (Eiseler et al., 2010), such as N-WASP. Phosphorylation of cortactin at S298 was further shown to generate a 14-3-3β-binding site (Eiseler et al., 2010). As the binding and biological effects of 14-3-3 proteins are strictly dependent on the presence of a phosphorylated residue (Mackintosh, 2004; Muslin and Xing, 2000; Sroka et al., 2016; Wille et al., 2018), we were unable to use serine to glutamate phosphorylation-mimicking mutations in all further experiments.

To evaluate a role for N-WASP in synergistic nucleation at the TGN, cortactin and cortactin-S298A were ectopically expressed in control and WASL (N-WASP) knockdown cells. Expression of cortactin-S298A significantly enhanced LifeAct fluorescence compared to cells with vector alone and or expressing wild-type cortactin. Strikingly, knockdown of WASL significantly reversed F-actin generation in cortactin-S298A-expressing cells compared to vector controls (Fig. 4E,F; Fig. S4A,B), pointing to a regulation by N-WASP and PKD2-phosphorylated cortactin.

The formation of actin at the Golgi complex, and in particular at the TGN, is not easily visualized or quantified (Gurel et al., 2014). As a second, independent method to evaluate relative steady-state actin polymerization (i.e. that was not dependent on the expression of LifeAct), we measured actin–GFP incorporation into phalloidin–Alexa-Fluor-568-labeled filaments at TGN-46-positive structures with AB-FRET (Sroka et al., 2016). We opted to utilize this previously published FRET approach (Sroka et al., 2016) to acquire specificity for nascently generated F-actin signals, since neither actin–GFP nor phalloidin–Alexa-568 alone displayed strong enrichment at the TGN (Fig. S4C,D). Indeed, in line with our LifeAct-FRAP experiments, we found that expression of constitutively active CDC42-Q61L significantly enhanced actin incorporation into filaments, whereas inactive CDC42-T17N impaired actin polymerization below vector levels. The latter changes were however not substantial enough to generate significant differences from the control condition. Of note, knockdown of WASL upon co-expression of CDC42-Q61L impaired actin polymerization to levels that were no longer significantly different from those in the vector controls (Fig. S4C), confirming validity of the FRET approach (Sroka et al., 2016) as a second method to determine relative actin turnover at the TGN. In additional experiments, only cortactin-S298A, but not wild-type cortactin significantly enhanced percentage FRET values and thus steady-state actin polymerization, whereas knockdown of WASL in cortactin-S298A-expressing cells reduced actin–GFP incorporation to vector control levels (Fig. S5A,B).

In summary, our analysis of actin polymerization at the TGN suggest that active PKD2 inhibits F-actin formation by phosphorylation of cortactin. We also show, that activation of CDC42 inactivates PKD2 and triggers signaling events that facilitate Arp2/3-complex-mediated synergistic actin polymerization in the presence of non-S298-phosphorylated cortactin in conjunction with the resident NPF N-WASP.

Since we were able to show cortactin-S298-phosphorylation-dependent changes in actin turnover/polymerization at the TGN, we were prompted to investigate whether this would also translate into differences in constitutive secretion. To this end, we performed secretion experiments using the artificial substrate secreted horseradish peroxidase (HRP) (Cruz-Garcia et al., 2013; Fugmann et al., 2007) upon knockdown of PRKD2 (PKD2) and CTTN (cortactin) using two different shRNAs in HEK293T cells (Fig. 5A; Fig. S6A). In line with our previous data (Eiseler et al., 2016; Pusapati et al., 2010), depletion of PRKD2 (shPRKD2) impaired constitutive secretion of HRP. Moreover, knockdown of CTTN (shCTTN) also significantly impaired constitutive secretion (Fig. 5A; Fig. S6A). Similar to what was found for steady-state F-actin generation at the TGN, only the non-phosphorylatable cortactin-S298A enhanced constitutive secretion of HRP (Fig. 5B; Fig. S6B). To corroborate these data, we further measured constitutive secretion of the endogenous cargo MMP2 (Eiseler et al., 2016) from NIH 3T3 fibroblasts (Fig. 5C,D). Again, only non-S298-phosphorylatable cortactin significantly increased constitutive MMP2 secretion and concomitantly decreased cellular MMP2 levels (Fig. 5D).

Next, we evaluated the role of N-WASP during constitutive secretion by measuring HRP release. Ectopic expression of cortactin-S298A significantly increased constitutive secretion compared to wild-type cortactin, whereas secretion was significantly impaired upon depletion of endogenous WASL. Indeed, knockdown of WASL impaired constitutive secretion by the cortactin-S298A mutant to levels that were no longer significantly different from wild-type cortactin (Fig. 5E; Fig. S6C). This suggests that cortactin and N-WASP are likely to cooperate not only during regulation of synergistic actin polymerization at the TGN (Fig. 4D,F) but also in the regulation of PKD2-controlled constitutive secretion (Fig. 5E). Moreover, these data are supported by evaluating secretion of endogenous MMP2 from NIH 3T3 fibroblasts. Here, knockdown of the mechano-GTPase dynamin-2 (DNM2), N-WASP (WASL) and cortactin (CTTN) significantly impaired secretion by 63.1%, 55.5% and 46.8%, respectively (Fig. S6D,E).

Finally, if vesicle release from TGN membranes were regulated by PKD2 and cortactin, depletion of cortactin would be expected to also result in cargo retention. To evaluate such an effect upon knockdown of CTTN, we quantified the fluorescence intensity of endogenous MMP2 at golgin97-positive structure in WI38 fibroblasts, which express high levels of MMP2 (Fig. 5F,G; Fig. S6F). In line with the HRP and MMP2 constitutive secretion data, we demonstrate increased MMP2 fluorescence at the TGN upon depletion of cortactin, suggesting that cortactin is required for proper cargo release (Fig. 5F,G).

Our data demonstrated a functional role for synergistic actin polymerization initiated by cortactin and N-WASP at the TGN. In vitro pyrene-actin polymerization in the presence of the Arp2/3-complex, a GST-tagged VCA domain (minimal fragment of WASP) and purified cortactin or cortactin-S298A, indicated cortactin and its phosphorylation by PKD regulate synergistic actin polymerization with class I NPFs (Eiseler et al., 2010). However, cortactin is also implicated in the activation of full-length N-WASP, by relieving N-WASP autoinhibition (Kowalski et al., 2005). Thus, we next examined how phosphorylation of cortactin at S298 might regulate Arp2/3 complex activity at different levels. To this end we employed a CFP–N-WASP–YFP FRET biosensor (Lorenz et al., 2004b; Oser et al., 2009). This sensor indicates the extent of activated N-WASP by a concomitant loss of FRET during unfolding (Lorenz et al., 2004b). The biosensor was co-expressed with wild-type cortactin, cortactin-S298A or a vector control in HEK293T (Fig. 6A; Fig. S7A) and HeLa cells (Fig. S7B,C). Active N-WASP was determined by FRET in total cell lysates. We detected significantly increased levels of N-WASP in an active, open conformation upon co-expression of cortactin-S298A, in respect to wild-type cortactin. This suggests that PKD phosphorylation might be implicated in controlling N-WASP activation. We also evaluated N-WASP activity regulated by cortactin and cortactin-S298A in the perinuclear region of intact MCF-7 cells. The N-WASP-FRET biosensor was co-expressed with the respective constructs in MCF-7 cells, which are highly responsive to heregulin, and cells were stimulated to induce actin polymerization and secretion (Döppler et al., 2013; Eiseler et al., 2010). Again, N-WASP activity was significantly enhanced and more sustained upon co-expression of non-phosphorylatable cortactin-S298A (Fig. 6B). These data indicate that cortactin and its phosphorylation by PKD are not only implicated in modulating synergistic actin polymerization (Eiseler et al., 2010), but also in controlling the class I NPF activity of N-WASP.

Given our above findings, we were interested to investigate how the activation of N-WASP is controlled at a molecular level. In vivo, N-WASP is sequestered in its inactive, autoinhibited state by the inhibitory binding protein WIP. In vitro, TOCA-1 (transducer of CDC42-dependent actin-assembly protein-1 homolog, also known as FNBP1L) is a potent, CDC42-dependent activator of the N-WASP–WIP complex, requiring a functional SH3 domain (Ho et al., 2004; Moreau et al., 2000). In addition, SH3 domains from a number of other proteins, such as Grb2 and Nck1/Nck2 are known to promote N-WASP activation (Campellone et al., 2012; Ho et al., 2004; Moreau et al., 2000). Similar functions have been described for the SH3 domain of cortactin (Kowalski et al., 2005). Thus, we examined further whether phosphorylation of cortactin at S298 by PKD2 might alter the molecular complex between N-WASP and WIP at the TGN.

Initially, we performed colocalization studies with mCherry–WIP and endogenous N-WASP in HeLa cells. Although, we did not detect an enrichment of WIP at golgin-97-positive structures, a partial colocalization was demonstrated utilizing Huygens confocal deconvolution microscopy (CLME) (Fig. 6C). To investigate localization and molecular proximity of endogenous proteins, we additionally performed PLAs for N-WASP and Arp2 (Fig. 6D,E), as well as N-WASP and WIP (Fig. 6F,G). PLA staining indicated a wide distribution of Arp2–N-WASP and N-WASP–WIP signals throughout the cell, which also included furin–GFP-positive structures (Fig. 6D–G). Thus, WIP is indeed present at the TGN. We therefore performed co-IP experiments with ectopically expressed proteins to assess changes in N-WASP–WIP interaction upon co-expression of vector, cortactin and cortactin-S298A. Ectopically expressed N-WASP was detected in endogenous WIP immunoprecipitations (IPs) when wild-type cortactin was co-expressed, but this interaction was lost upon co-expression of cortactin-S298A (Fig. 6H). Since it was previously reported that cortactin can also bind to WIP (Kinley et al., 2003), and enhanced its ability to act as an NPF, we were prompted to conduct co-IPs of cortactin or cortactin-S298A–FLAG with mCherry–WIP (Fig. 7A,B). Concomitant with a loss of N-WASP–WIP interaction, the binding of WIP to non-phosphorylated cortactin-S298A was significantly increased (Fig. 7A,B). We further verified that binding of WIP to cortactin was dependent on its SH3 domain (Kinley et al., 2003) (Fig. S7D). Molecular proximity of cortactin and WIP at the TGN was corroborated by PLA staining at furin–GFP-positive structures (Fig. 7C,D). To determine regulation of N-WASP activity by WIP and PKD2-cortactin at an endogenous level, we again performed co-IP experiments in parental and PKD2 KO HeLa cells. In line with experiments described above, PKD2 knockout resulted in a significantly impaired interaction of endogenous N-WASP and WIP, suggesting that this pathway is indeed utilized in cells (Fig. 7E,F). Thus, our data point to a mechanism whereby active PKD2 phosphorylates cortactin at S298 to block the activation of the N-WASP–WIP complex at the TGN (Figs 6D–H and 7A–F) until the lipid environment is ready for carrier fission. In order to acquire further quantitative evidence for such a regulation at the TGN, we performed AB-FRET interaction studies with N-WASP–GFP and endogenous WIP at TGN46-postive structures in HeLa cells. Cortactin–FLAG, cortactin-S298A–FLAG or empty vector were co-expressed to alter the N-WASP–WIP complex composition (Fig. 7G; Fig. S7E). In HeLa cells, N-WASP–GFP is localized in the cytoplasm, the perinuclear region and the nucleus (Mizutani et al., 2004; Suetsugu and Takenawa, 2003; Wu et al., 2004), which is comparable to what is seen for the endogenous protein, but TGN enrichment is less prominent, likely due to ectopic expression. We were therefore prompted to exclude strongly expressing cells from AB-FRET experiments in Fig. 7G and Fig. S7E. In line with the co-IPs and PLA experiments, AB-FRET analysis gave a significantly decreased percentage FRET value upon expression of cortactin-S298A for the interaction of N-WASP with endogenous WIP at the TGN with respect to that found upon expression of wild-type cortactin (Fig. S7D,E). Moreover, we were able to detect an enhanced interaction of cortactin-S298A–GFP with endogenous WIP labeled with Alexa-Fluor-568-conjugated secondary antibodies at TGN46-postive structures (Fig. 7H; Fig. S7F). Thus, our AB-FRET experiments further verify and quantify the PKD-phosphorylation-dependent interaction of cortactin and WIP at the TGN and indicate a coordinated N-WASP activation through WIP sequestration.

Actin polymerization has been suggested to support biogenesis and scission of transport carriers from various locations, for example, during endocytosis (Anitei et al., 2010; Kessels and Qualmann, 2005; Qualmann et al., 2000). At the TGN, the molecular composition and upstream regulation of branching actin polymerization still needs to be elucidated in more detail. In our present work, we show that the actin regulatory protein cortactin and its phosphorylation by PKD2 at S298 (Eiseler et al., 2010) are vital regulators of Arp2/3-complex-initiated actin polymerization at the TGN (Figs 2 and 4; Figs S4C,D and S5). In addition, N-WASP and its sequestering protein WIP (Ho et al., 2004) were shown to be localized at TGN structures by using a combination of colocalization studies, PLA experiments and FRET (Figs 1D,E, 6C–G and 7G; Fig. S7E). We show that phosphorylation of cortactin at S298 impairs the interaction of WIP and cortactin (Fig. 7A,B,H; Fig. S7F). WIP is known to stabilize the inactive autoinhibited conformation of N-WASP, preventing unsolicited nucleation by the Arp2/3-complex (Ho et al., 2004). Actin polymerization and N-WASP activity at the TGN are controlled downstream of the small GTPase CDC42 (Musch et al., 2001) (Fig. 3E–H; Fig. S3A-D). Active CDC42-GTP can directly activate N-WASP (Ho et al., 2004; Tomasevic et al., 2007) and disable a second level of inhibition by inactivating PKD2, thereby shifting the balance to non-S298-phosphorylated cortactin. Consequently, N-WASP sequestration in the autoinhibited state by WIP is disrupted (Figs 6H, 7E–H; Fig. S7E,F) and WIP binding to cortactin is enhanced (Fig. 7A,B,H; Fig. S7F), thus facilitating synergistic actin nucleation. We propose that the directional force generated by this precisely timed polymerization supports separation of transport carriers from the donor membrane, as we were able to demonstrate enhanced constitutive secretion upon expression of the non-S298-phosphorylated cortactin (Fig. 5B–D). Secretion was N-WASP dependent (Fig. 5E; Fig. S6D,E), and we show TGN retention of endogenous MMP2 cargo upon cortactin depletion (Fig. 5F,G).

Thus, phosphorylation of cortactin by PKD2 might regulate the timing of carrier fission steps (Fig. 8). Here, our data are in line with previously published work demonstrating a role for active PKD in modifying the TGN lipid environment to a fission-competent state (Fugmann et al., 2007; Hausser et al., 2005). As long as PKD is active, there is inhibition of actin polymerization at the TGN. Once PKD activity is terminated, synergistic actin polymerization is facilitated by non-S298-phosphorylated cortactin.

In addition to PKD activity, dimer formation between PKD2 and PKD3 isoforms was reported to coordinate secretion from the TGN (Bossard et al., 2007). We have previously shown that PKD1 and PKD2 can phosphorylate cortactin at S298 (Eiseler et al., 2010). Moreover, our data suggest that PKD3 can also phosphorylate endogenous cortactin (Fig. S2B). Thus, it is possible that PKD2 and PKD3 homo- or hetero-dimers at the TGN could also be implicated in the coordinated regulation of actin polymerization. However, this needs to be investigated in much more detail by determining actin polymerization upon abrogation of the PKD dimer formation in a follow-up study.

In our current manuscript, we were prompted to investigate the regulation of actin polymerization downstream of PKD2–cortactin at the TGN, since we have previously demonstrated an interaction of PKD2 with ARF1 and ARL1, vital governing GTPases during the vesicle biogenesis cycle (Eiseler et al., 2016). Once the lipid environment has been modified accordingly, active PKD needs to be inactivated to allow for the coordinated initiation of synergistic actin nucleation by the Arp2/3 complex. This is accomplished via active CDC42 (Fig. 3A–D; Fig. S2C,D). Whether cross-signaling by RhoGTPases and a link to microtubules, as suggested by Eisler et al. (2018), is associated with this process needs to be determined in future studies. In such a context, a putative link between CDC42 or RhoA activity and PKD2 activation as well as regulation through the ARF1 master switch for vesicle biogenesis (Dubois et al., 2005) will have to be further investigated in detail.

So far, it is also unknown how inactivation of PKD by CDC42 at the TGN is achieved at the molecular level. An intermediate mediator might be the lipid metabolite diacylglycerol (DAG). DAG is required for the recruitment of PKDs to TGN membranes, and co-activation of PKDs by novel protein kinase C (nPKC) upstream kinases (Wang, 2006). Formation of DAG at the TGN can be accomplished by different biosynthesis pathways, such as sphingomyelin synthase (SMS) (Subathra et al., 2011) or through phospholipase D1 (PLD1) (Ktistakis et al., 1995; Sarri et al., 2011). Phospholipase C β3 was also described to be implicated in DAG generation at the TGN (Diaz Anel, 2007). How the activity of the respective DAG biosynthesis pathways is linked to CDC42 activity is currently unclear, for example, the PLD1 isoform was reported to be regulated by several activators, including Rho family GTPases and PKC (Brown et al., 1995; Walker et al., 2000). Since CDC42 activation has also been shown to be DAG dependent (Vaughan et al., 2014), extensive research will be required to rigorously delineate biosynthesis pathways during PKD activation at the TGN downstream of CDC42.

In this study, we have measured actin turnover at the TGN by using LifeAct-Ruby FRAP (Figs 3E–H and 4), since LifeAct signals were considerably enriched at furin-GFP-positive structures (Fig. S8A,B). The enrichment of LifeAct signals at the TGN is in line with a previous study comparing different probes for F-actin live imaging at subcellular compartments (Belin et al., 2014). The use of LifeAct for FRAP analysis has been controversial owing to fast exchange rates on filaments (Belin et al., 2014; Riedl et al., 2008). However, in their experiments those authors have not investigated actin turnover at the Golgi compartment, where filaments are very short, transient and have very high turnover rates (Gurel et al., 2014), instead they measured stress fibers and lamellipodia. In our hands, FRAP experiments with LifeAct–Ruby were able to resolve differences in relative actin polymerization at the TGN. In contrast, in actin–Ruby FRAP experiments, even enhanced actin polymerization upon overexpression of constitutively active CDC42-Q61L was not detectable (data not shown). We suggest that this may be the result of insufficient enrichment at the TGN, as well as fast diffusion by actin monomers, which can also be observed by a fast, steep increase in fluorescence recovery (Fig. S8C,D).

Our data indicate that regulation of branching actin polymerization at the TGN is mediated by cortactin and its inhibitory phosphorylation via PKD2 in conjunction with N-WASP–WIP and the Arp complex. Thus, knockout of PKD2 resulted in enhanced steady-state actin polymerization at the TGN and this phenotype was reversed upon depletion of cortactin (Fig. 4C,D; Fig. S3G,H).

Nevertheless, we cannot exclude the possibility that, at least partially, other actin remodeling pathways controlled by PKD isoforms might be involved in regulating actin turnover at the TGN. One candidate is the slingshot–cofilin pathway, which generates barbed ends and monomers by severing filaments (DesMarais et al., 2005; Lorenz et al., 2004a). Furthermore, PKD1, slingshot and cofilin have been reported to generate Golgi outposts in neurites of neuronal cells and thus might regulate actin polymerization at the TGN (Quassollo et al., 2015). However, to generate forces that can drive separation of transport carriers from donor membranes, branching actin polymerization mediated by the Arp2/3 complex is required (Anitei et al., 2010). In addition to regulating synergistic actin polymerization with N-WASP, cortactin has been reported to stabilize actin branch junctions (Helgeson and Nolen, 2013), which may further facilitate vesicle release from TGN membranes. Moreover, in vitro and in invadopodia, cortactin has been demonstrated to directly interact with cofilin and inhibit its actin severing activity (Oser et al., 2009). Thus, it is not likely that both pathways contribute to actin polymerization at the same spatial location or at the same time.

Our current study is also consistent with the detailed description of molecular complexes involved in synergistic nucleation, proposed as an obligatory displacement model for WASP family NPFs (Helgeson and Nolen, 2013). Here, active N-WASP dimers initiate actin nucleation, which is then stalled, as long as N-WASP is occupying the Arp2/3 complex. Completion of the nucleation process was shown to require displacement of N-WASP from Arp2/3 by cortactin, freeing N-WASP for additional nucleation events (Helgeson and Nolen, 2013). Since we have shown enhanced synergistic nucleation for non-phosphorylatable cortactin-S298A in the presence of a VCA minimal fragment (Eiseler et al., 2010), it is likely that this cortactin mutant somehow affects displacement of WASP NPFs from the Apr2/3 complex. In contrast to the class I NPF N-WASP (Alekhina et al., 2017; Helgeson et al., 2014), cortactin is a weak class 2 NPF, unable to recruit actin monomers to the Arp2/3 complex (Helgeson and Nolen, 2013). Since recruitment of actin monomers is thought to induce conformational changes required for efficient nucleation (Helgeson and Nolen, 2013), monomer recruitment might be facilitated by interaction of the monomer-binding protein WIP with non-S298-phosphorylated cortactin during nucleation in vivo (Kinley et al., 2003). How these processes can be fully integrated in the obligatory displacement model (Helgeson and Nolen, 2013) remains to investigated.

Since binding of WIP to cortactin is dependent on its interaction with the cortactin SH3 domain (Kinley et al., 2003), we propose that 14-3-3 protein binding to the pS298 consensus site (Eiseler et al., 2010) induces conformational changes that inhibit efficient SH3-mediated interactions. This interpretation is in line with published data where SH3-dependent binding of interaction partners is masked or inhibited by 14-3-3-proteins bound to phosphorylated consensus motifs (Freeman and Morrison, 2011).

In summary, we here present a novel role for PKD2 and its phosphorylation of cortactin in the coordination and timing of synergistic actin nucleation at the TGN with the NPF N-WASP. We were further able to measure the respective changes in steady-state actin polymerization at the TGN and we also demonstrate an increase in constitutive secretion that is dependent on the same proteins (Fig. 8).

HEK293T, HeLa and MCF-7 cells were recently acquired from ATCC. WI-38 cells were recently acquired from EACC via Sigma-Aldrich (Taufkirchen, Germany). NIH3T3 fibroblasts were kindly provided by Franz Oswald (University Clinics Ulm, Germany). The HUVEC endothelial cell line was kindly provided by Steffen Just (University Clinics Ulm, Germany). HEK293T, HeLa and MCF-7 cells were cultured in DMEM supplemented with 10% fetal calf serum (FCS) and 1% penicillin and streptomycin. WI-38 were cultured in DMEM with 15% FCS, 1% non-essential amino acids and 1% penicillin and streptomycin. HUVECs were cultured in MV2 endothelial growth medium with supplement and 1% penicillin and streptomycin (Promo Cell, Heidelberg, Germany). All cells were kept at 37°C in 5% CO₂. All cells were regular tested for contamination.

A full list of antibodies used in this study is available in Table S1. In Table S2 DNA expression vectors, lentiviral shRNAs and siRNAs are listed.

HeLa cells were acquired from the ATCC. PRKD2 isoform-selective CRISPR constructs containing gRNA and Cas9–GFP in a single expression vector (U6gRNA-Cas9-2A-GFP) were acquired from Sigma-Aldrich (Taufkirchen, Germany). The CRISPR construct gRNA sequence was designed as following: PRKD2, HS0000083056, target site, 5′-GCAGGAGGTTGGCCGACGTGGG-3′; PRKD3, HS0000321772, target site, 5′-GAGAGTATGTGGACGAATCTGG-3′. HeLa CRISPR lines were produced by transfecting cells with the respective CRISPR constructs using Lipofectamin 3000 (Thermo Fisher Scientific, St Leon-Rot, Germany). GFP-positive cells were subsequently transferred as single cells into 96-well culture plates by fluorescence-associated cell sorting (FACS). After 3 weeks of growth, cells from single clones were harvested, lysed as described in Eiseler et al. (2016) and analyzed for PKD isoform expression by western blotting. Positive hits were further verified by sequencing of PCR products from genomic DNA.

The cortactinΔSH3 mutation was generated by PCR on pCR3.V62-Met-Flag-Cortactin template (Table S2) using forward primer 5′-ACGCGGAATTCAGGAAAGCTTCGACA-3′ and reverse primer 5′-AAGGGCCCCTAGTTCTCGTACTCATCGTA-3′. The amplicon was inserted in pcDNA3 using EcoRI and ApaI. The cortactinΔSH3-S298A mutation was introduced by PCR with forward primer 5′-CTGGCCAAGCACGAGGCCCAGCAAGAC-3′ and reverse primer 5′-GTCTTGCTGGGCCTCGTGCTTGGCCAG-3′, pcDNA3-FLAG-CortactinΔSH3 was used as template.

Samples were lysed in lysis buffer 1 (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM MgCl₂, 1% Triton X-100) plus Complete protease^® and PhosStop inhibitors^® (Roche, Basel, Switzerland). Cell debris was removed by centrifugation (13,000 g for 10 min). Protein content in the lysate was measured using Bradford assay reagent (BioRad, Hercules, CA). For IP, the same amount of lysate was used for each condition and incubated with an appropriate antibody concentration for 2 h at 4°C. Protein-A–Sepharose beads (GE Healthcare Life Sciences, Chicago, IL) were added and further incubated for 2 h at 4°C. Samples were washed four times with 900 µl lysis buffer 1 and subjected to SDS-PAGE on 10% acrylamide gels. Samples were transferred onto nitrocellulose membranes. Blots were blocked with 2% BSA in TBS Tween-20. Primary antibodies were applied in blocking solution overnight at 4°C. The membranes were incubated with the appropriate secondary antibodies conjugated to horseradish peroxidase (HRP) for 1 to 2 hs at room temperature; signal was detected using SuperSignal West Dura reagent (Thermo Fisher Scientific). Quantitative analysis of western blots was performed by measuring integrated band densities using NIH ImageJ (National Institutes of Health, Bethesda, MD).

For western blot analysis, 5×10⁵ HEK293T cells were seeded in six-well plates and processed as indicated. For confocal microscopy 8×10⁴ HeLa cells were seeded on cover slips in 12-well plates. The next day, the growth medium was replaced with growth medium containing 10 µM ML141 (Sigma-Aldrich) or the same amount of DMSO. Cells were incubated for 1 h at 37°C, washed with PBS and either fixed for microscopy or lysed for western blotting. For FRAP experiments cells were prepared as described below. At 1 h prior to imaging medium was changed to Phenol Red-free medium containing 10 µM ML141, 5 µM Casin (Biotechne, Wiesbaden-Nordenstadt, Germany) or DMSO, respectively. ML141 was previously characterized by Surviladze et al. (2010) to have an IC₅₀ for CDC42 of 0.2 µM, and for Rac1, Ras, Rab2a and Rab7 of greater than 100 µM (Surviladze et al., 2010). Casin (as Pirl1-related compound 2) was described to have an IC₅₀ of 2 µM in an PIP2-stimulated actin polymerization assay (Peterson et al., 2006). Furthermore, casin was shown to inhibit CDC42 activity at 5 µM in murine hematopoietic stem cells (Florian et al., 2012).

The assay was performed with a CDC42 Activation Assay Biochem Kit from Cytoskeleton, Inc. (Denver, CO) according to the manufacture's manual. At 2 days prior to the assay, 15×10⁵ WI-38 fibroblasts were seeded in 145 mm plates; 3.5 mg of cleared total cell lysate was used for the experiments. GTP-γS- and GDP-loaded CDC42 were used as respective control. Bound GTP-loaded CDC42 was analyzed by western blotting.

For confocal imaging, cells were processed as described previously (Eiseler et al., 2016). In brief, 8×10⁴ HeLa cells were seeded on sterilized 15×15 mm cover slips in 12-well plates. The next day cells were transfected using Lipofectamin 3000 (Thermo Fisher Scientific), treated with inhibitors and then fixed. Fixation was performed with 3.75% formaldehyde for 20 min at room temperature. Samples were permeabilized with 0.1% Triton X-100 in PBS for 30 s or 2 min. After blocking (5% FCS, 0.05% Tween-20 in PBS) for 20 min, samples were stained with primary antibody for 2 h at room temperature or overnight at 4°C. Samples were washed three times with PBS and secondary antibodies conjugated with an Alexa fluorophore were applied for 2 h at room temperature. Coverslips were mounted in Fluoromount-G (Thermo Fisher Scientific).

AB-FRET experiments were performed as described previously (Becher et al., 2017; Eiseler et al., 2016; Sroka et al., 2016). Cells were analyzed with a LSM710 confocal laser-scanning microscope (Zeiss, Jena, Germany) and TCS-SP8-HCS (Leica, Mannheim, Germany) equipped with HC PL APO 40×/1.30 Oil CS2, HC PL APO 40×/1.10 W CORR CS2 or HC PL APO 63×/1.20 W CORR CS2 immersion objectives. Images were acquired in sequential scan mode. Gain and offset were set in a way that all cellular structures were imaged within the linear range of detectors having no saturated pixels and black level background values. AB-FRET experiments were performed with transiently transfected HeLa cells, fixed, processed and stained as stated above. This method allows the determination of protein–protein interactions indicated by an apparent molecular proximity below 10 nm (Becher et al., 2017; Eiseler et al., 2012, 2016; Gu et al., 2004; Kunkel and Newton, 2009; Oser et al., 2010; Sroka et al., 2016; Valente et al., 2012; Wille et al., 2014; Wouters et al., 1998; Zeug et al., 2012). The FRET acceptor in a region of interest (ROI) was bleached by an intensive laser line and increase in donor fluorescence, indicating FRET, was measured. AB-FRET measurements were carried out by acquiring pre- and post-bleach images of donor and acceptor using an automated time series. Donor dyes in cells were in all instances GFP-tagged proteins. Acceptors were labeled with Alexa-Fluor-568-conjugated antibodies or Phalloidin–Alexa-Fluor-568 (F-actin). As an internal threshold, 8-bit images (256 gray values) were used for recording raw data to only register substantial changes in fluorescence intensities as respective gray value changes. FRET values were acquired from at least three independent experiments. Cells were randomly sampled from slides. In experiments with ectopic transgenes, cells with very strong ectopic expression were not recorded to avoid artifacts. Image analysis was performed using NIH ImageJ. Quantitative FRET analysis was executed by calculating the mean percentage FRET and s.e.m. for each cell from non-thresholded raw data for equally sized sub-ROIs randomly placed on TGN-marker positive structures within the bleached region [%FRET=((Donor_post−Donor_pre)/Donor_post)×100]. Cells that failed to show FRET in the bleach-region were set to zero during analysis. Results are presented as scatter graphs with single percentage FRET values. Numbers on top of graphs indicate mean percentage FRET values and s.e.m. for all evaluated ROIs. Numbers in brackets indicate the number of cells with FRET in the bleach-region. Raw data for all AB-FRET experiments is provided in Table S3.

For Huygens deconvolution, images of the TGN from fixed cells embedded in Fluoromount G (Thermo Fisher Scientific; refractive index of 1.4), were acquired using a Leica TCS-SP8-HCS confocal laser-scanning microscope. Imaging conditions were set as follows: objective HC PL APO 63×/1.20 W CORR CS2; size: 4096×4096 or 2048×2048 pixels, pinhole: 0.6 AU, scan speed: 100–200 Hz bidirectional, z-stack encompassing several TGN sections (5–18) using a z-Galvo table. Image acquisition was performed with highly sensitive HyD GaAsP-hybrid detectors and the Huygens deconvolution software module. Subsequently, image stacks were loaded into the Huygens Essential Deconvolution software and processed by the Huygens optimized classical maximum likelihood estimation algorithm (CMLE).

HeLa cells (10⁵) were seeded in 35 mm glass-bottom µ-dishes (Ibidi, Martinsried, Germany). The next day, cells were transfected with the indicated constructs using Lipofectamin 3000 (Thermo Fisher Scientific). To quantify actin polymerization at the TGN we recorded either LifeAct–RubyN143C or actin–Ruby fluorescence intensities at furin–GFP-positive TGN structures. Before the experiment, the growth medium was replaced with Phenol Red-free RPMI or DMEM plus 10% FCS and 1% penicillin-streptomycin. FRAP experiments were performed using a Leica TSC-SP8-HCS confocal laser scanning microscope using a tabletop incubator (Oko Labs, Ottaviano, NA, Italy) at 37°C and 5% CO₂. Images (512×512 pixels; scan speed 200 Hz) were acquired every second with a pinhole of 2 AU to compensate for dynamic movement of Golgi stacks during acquisition. Five pre-bleach images were acquired before bleaching of furin–GFP-positive TGN structures using a 561 nm laser line (AOTF 100%) to set a baseline for fluorescence intensity measurements at the TGN. Subsequently, LifeAct–RubyN143C or actin–Ruby fluorescence intensities, respectively, were recorded for 130 s. Mean of ROI fluorescence intensities at the TGN were determined using detection masks generated around the furin–GFP signal.

HeLa cells (3×10⁴) were seeded on sterilized 12 mm diameter coverslips in 24-well plates. The next day cells were transfected with a plasmid encoding for Furrin-GFP using Lipofectamin 3000 (Thermo Fisher Scientific). The following day samples were fixed for 20 min with 3.7% PFA at room temperature. Cells were permeabilized with 0.1% Triton X-100 in PBS for 2 min. PLA staining was performed according to manufactures protocol with Duolink In Situ Red Starter Kit Mouse/Rabbit (Sigma-Aldrich).

The extent of N-WASP activity was determined by quantifying the presence of open, unfolded N-WASP conformations using a CFP–N-WASP–YFP FRET activity sensor that loses Förster energy transfer after relieving autoinhibition (Lorenz et al., 2004b; Oser et al., 2009). The sensor was initially characterized by Lorenz et al. (2004b). HEK293T (1.5×10⁵) cells were co-transfected with sensor and the indicated constructs (ratio: 1 to 3) in a 12-well plate format. After 24 h cells were lysed in 100 µl FRET lysis buffer (20 mM Tris-HCl pH 7.4, 100 mM NaCl, 5 mM MgCl₂ and 0.5% Triton X-100) with Complete^® protease and PhosStop^® inhibitors (Roche) for 20 min. After lysates have been cleared by centrifugation 80 µl of cleared lysates were measured in a black non-binding 96-well plate using a Tecan M-200 Pro (Tecan, Halver, Germany) plate reader (CFP, Ex: 433 nm, Em: 475 nm; FRET, Ex 433 nm, Em: 527 nm). FRET was calculated by generating mean CFP:FRET ratios for n=12 transfected samples. Expression of transgenes was verified by western blotting.

The CFP–N-WASP–YFP FRET activity sensor was co-transfected with the indicated constructs in MCF-7 cells on 35 mm µ-dishes (Ibidi, Martinsried, Germany). After 24 h, cells were serum-starved in Phenol Red-free RPMI medium for 4 h and then subjected to live imaging using a LSM710 confocal laser scanning microscope (Zeiss, Jena, Germany) at 37°C and 5% CO₂ employing a 63× high NA objective. Images were acquired in ROI scan mode centered on the perinuclear region of a cell with enriched presence of the N-WASP sensor (512×512 pixels; pixel dwell: 6.3 µs; Ex: 405 nm laser, Em: CFP: 470-501 nm; Em: FRET: 524-550 nm, 1 frame/s; pinhole 30 µm to compensate for dynamic Golgi movements). After 60 s, cells were stimulated with 100 ng/ml Heregulin (PeproTech, London, UK). Fluorescence intensities in the clipped scan ROIs were analyzed by NIH ImageJ and presented as CFP:FRET ratios.

To determine the extent of constitutive secretion, we measured release of the artificial cargo HRP (Cruz-Garcia et al., 2013; Fugmann et al., 2007) from HEK293T cells. HEK293T cells were modified either by transduction with lentiviruses coding for shRNAs (Table S2) to generate semi-stable knockdown cell lines upon selection with puromycin 6 µg/ml for 1 week or by transiently co-transfecting cells with the indicated DNA expression constructs and HRP as well as siRNAs (ratio: 3 to 1) in a 6-well format. After transfection of siRNAs, cells were incubated for 48 h. Subsequently, cells were carefully washed three times and incubated with 600 µl of serum-free, Phenol Red-free RMPI medium for 4 h. Supernatants were harvested, cleared and 5 µl were incubated with 200 µl freshly prepared ECL solution (Luminol 224 mg, Coumaric acid 37.1 mg, 0.05% H₂O₂ in 100 ml 50 mM Tris-HCk pH 8.0) in white non-binding 96-well plates for 5 min to reach equilibration. The presence of HRP in cell culture supernatants was subsequently detected by measuring luminescence using a Tecan M-200 Pro plate reader. In general, five or six independent experiments with three technical replicas were quantified.

Statistical analyses were performed using Prism software version 6 or 8 for Windows (GraphPad Software, San Diego, CA). Graphs depict mean±s.e.m. for all conditions. Normality was tested with a D'Agostino and Pearson omnibus test. Statistical significance is indicated by asterisks (ns, not significant; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001).

The authors want to acknowledge use of the Imaging Core Facility for ‘Multiphoton and Confocal Microscopy’ and FACS Core Facility at Ulm University. We thank Ralf Kemkemer, Franz Oswald, Johan van Lint, Steffen Just and Angelika Hausser for providing valuable research reagents. We thank Claudia Ruhland, Edith Glass and Susanne Fluhr for excellent technical assistance.