The tumor promoter‐activated protein kinase Cs are a system for regulating filopodia

Different protein kinase C (PKC) isoforms have distinct roles in regulating cell functions. The conventional (α, β, γ) and novel (δ, ɛ, η, θ) classes are targets of phorbol ester tumor promoters, which are surrogates of endogenous second messenger, diacylglycerol. The promoter‐stimulated disappearance of filopodia was investigated by use of blocking peptides (BPs) that inhibit PKC maturation and/or docking. Filopodia were partially rescued by a peptide representing PKC ɛ hydrophobic sequence, but also by a myristoylated PKC α/β pseudosubstrate sequence, and an inhibitor of T‐cell protein tyrosine phosphatase (TC‐PTP). The ability to turn over filopodia was widely distributed among PKC isoforms. PKC α and η hydrophobic sequences enhanced filopodia in cells in the absence of tumor promoter treatment. With transcriptional knockdown of PKC α, the content of PKC ɛ predominated over other isoforms. PKC ɛ could decrease filopodia significantly in promoter‐treated cells, and this was attributed to ruffling. The presence of PKC α counteracted the PKC ɛ‐mediated enhancement of ruffling. The results showed that there were two mechanisms of filopodia downregulation. One operated in the steady‐state and relied on PKC α and η. The other was stimulated by tumor promoters and relied on PKC ɛ. Cycles of protrusion and retraction are characteristic of filopodia and are essential for the cell to orient itself during chemotaxis and haptotaxis. By suppressing filopodia, PKC ɛ can create a long‐term “memory” of an environmental signal that may act in nature as a mnemonic device to mark the direction of a repulsive signal.

specifically by an amino acid sequence, 14-21, in the extreme Nterminus of the protein (Yedovitzky et al., 1997) and phosphoserine 729 in the extreme C-terminus (Xu, He, Dobson, England, & Rumsby, 2007). Restricting the location of PKCs may put constraints on their access to agonists, especially diacylglycerol (Almena and M erida, 2011).
There are additional biophysical and biochemical mechanisms that keep the enzymes from being inappropriately activated. At the Nterminal portion of the protein, there is a pseudosubstrate sequence, which binds tightly to the active site until displaced by an agonist. Second, when activated in membranes, the PKCs become susceptible to hydrolysis by two types of calpains. The C-terminal half is further degraded by a calpain isozyme that is activated at high calcium levels (Cressman, Mohan, Nixon, & Shea, 1995), [reviewed in (Franco and Huttenlocher, 2005;Steinberg, 2008)].
PKCs undergo a maturation process in which the molecules become competent or "primed" through phosphorylation at multiple sites. Phosphate addition at a site in the activation loop is essential for maturation to the competent form and is preceded or followed by phosphorylation of two sites near the C-terminal end of the molecule (Keranen, Dutil, & Newton, 1995;Lachmann, Bär, Rommelaere, & N€ uesch, 2008), [reviewed in (Griner and Kazanietz, 2007)]. Phosphate addition to these sites, called the "turn" and "hydrophobic" motifs, is required for the maturation and stabilization of the enzyme (Bornancin and Parker, 1996;Ikenoue, Inoki, Yang, Zhou, & Guan, 2008). The kinase, 3-phosphoinositide-dependent protein kinase 1 (PDK1) is thought to be responsible for activation loop phosphorylation (Fr€ odin, Jensen, Merienne, & Gammeltoft, 2000;Gao, Toker, & Newton, 2001;Newton, 2003). However, for PDK1 to dock on PKC, a negative charge on the hydrophobic motif is required (Balendran et al., 2000;Fr€ odin et al., 2000) as well as anchorage in a membrane (Cenni et al., 2002).
Certain endogenous isoforms are found to be poor in phosphorylation at this site, including PKC h (Lachmann et al., 2008). Although activation loop phosphorylation is sometimes considered a marker of PKC maturation, this site can be dephosphorylated with retention of enzymatic competency as long as the turn and hydrophobic sites of phosphorylation are occupied. Moreover, phosphorylation or other modifications elsewhere on the molecule may also be important for folding of PKC into a 'closed' and stable conformation (Freeley, Kelleher, & Long, 2011).
Phosphorylation at the hydrophobic motif also creates a site for docking the downstream effector phospholipase D1 (PLD1), which is activated by binding PKC (Hu and Exton, 2003). Consistent with PKC's importance in signal transduction, additional regulatory constraints affect its recycling and degradation. Binding of a substrate or even an inhibitor to the active site prevents dephosphorylation of one or both C-terminal sites and thereby stabilizes the enzyme (Gould et al., 2011), and the same is true for protein kinase D (PKD), also known as PKC m (Kunkel and Newton, 2015). For the cPKC isoforms, phosphate additions at the "turn" and "hydrophobic" motifs are controlled by the mammalian target of rapamycin complex (mTORC), which regulates many aspects of protein synthesis (Facchinetti et al., 2008). Alternatives have been suggested for the kinase mediating phosphorylation of the hydrophobic motif, and many different kinases may add phosphate at this site (Dong and Liu, 2005). Furthermore, regulation of PKC phosphorylation and dephosphorylation shows specificity at the isoform and signaling pathway level, as well as cell-type specificity (Freeley et al., 2011). Finally, some PKCs in their mature form are homodimers.
Dimeric conformations are thought to have a prolonged residence time in membranes (Swanson et al., 2014).
The cell transmits communications from the extracellular environment in part by creating diacylglycerol. The phorbol ester tumor promoters, such as phorbol 12-myristate 13-acetate (PMA), act as surrogates of diacylglycerol. Both agonists activate PKCs and translocate them to membranes. A sustained elevation of diacylglycerol levels is sometimes produced from a phosphatidic acid precursor, and the site of production can be localized to one type of subcellular membrane to affect cellular functioning (Almena and M erida, 2011;Liu and Heckman, 1998). Thus, each PKC isoform may be activated at a different site, and there the enzyme may phosphorylate and modify substrates that, in turn, promote or prohibit its access to other compartments. This is how the effects of PKC signaling are orchestrated within the superstructure of the cell. Specifically, this type of dynamic interplay underlies mechanisms by which vesicle trafficking (Kiley, Jaken, Whelan, & Parker, 1995), chemotactic activity (Daviet, Herbert, & Maffrand, 1990), and signal transduction are regulated by PKC. For example, using low PMA levels similar to those we use here, Rehder's group showed that filopodia responded rapidly to phorbol ester, becoming shorter and fewer (Bonsall and Rehder, 1999). The opposite was found by other workers who found that a PKC inhibitor, GF 109203X, caused rapid retraction of growth cone filopodia (Fagerstrom, Påhlman, Gestblom, & Nanberg, 1996). Such discrepancies may be due to differences in the number of PKC isoforms, their varied patterns of distribution in different cell types, and/or the large number of substrates they phosphorylate. From one cell type to another, a single isoform can take on a different role in modulating cytoskeletal and adhesive structures.
The same variations have been impediments to understanding PKC's role in tumor promotion-a process which allows the emergence of a tumor from a cell population that was exposed to a small, initiating dose of carcinogen. Despite the importance of tumor promotion as a mechanism in cancer development, it has resisted scientists' attempts to explain it based on elementary regulatory features of the PKCs. By deconstructing the effects of PMA on cell phenotypes, we hope to identify the targets of activated PKC that are relevant to tumor promotion. We showed previously that filopodia declined immediately after PMA exposure (Heckman, Varghese, Cayer, & Boudreau, 2012). Stress fibers, also called actin cables, underwent a decrease briefly in PMAtreated cells but then achieved a higher steady-state level at 5 h (Li, Urban, Cayer, Plummer, & Heckman, 2006). Because stress fibers and filopodia both depend on actin bundles to form an inflexible rod-like structure at their core, the current research was carried out to investigate the relationship of the PMA-activated isoforms to the filopodia.
Our methods enable the phenotype to be classified on the basis of irreducible features that arise organically from the arrangement of the cell's parts. This type of feature classification removes the difficulty of subjective decision-making and provides information that may be valuable in understanding tumor promotion. Moreover, we use a cell line called 1000W that was developed as a model of human bronchogenic carcinoma. Its features were defined by unbiased classification methods, so that they could be related to the time course of tumor promotion (see Materials and Methods).
Filopodia act as antennae for incoming signals and are required for cells to compare the strength of adhesion from opposite directions on a substrate (Amarachintha et al., 2015). In an analysis of cancer development, we found that the loss of filopodia accounted for a greater proportion of the quantifiable morphological changes than any other feature (Heckman and Jamasbi, 1999). Filopodia typically undergo cyclical protrusion and retraction, however, and each protrusion-retraction cycle is regulated by the rates of actin assembly and actin rearward flow.
In an analysis of these rates for the nerve cell growth cone, where the filopodia are large and numerous, previous workers found that the rate of rearward flow and actin depolymerization was constant. Thus, filament disassembly at the base sufficed to ensure that retraction occurred if the rate of actin assembly at the tip of the filopodium failed to keep up with disassembly (Mallavarapu and Mitchison, 1999). The current research reveals that filopodia prevalence is regulated by PKC in two ways. The first is a promoter-mediated mechanism that decreases filopodia, and the second is a steady-state mechanism that regulates prevalence but is not responsive to activation of PKCs by tumor promoter. Stress fibers, also called actin cables, decreased briefly and then achieved a higher steady-state level at 5 h ( Figure 1c). Neurites increased gradually throughout the entire time course (Figure 1d). In these experiments, filopodia and neurites were measured by an objective method that obviated the problem of observer bias (see Materials and Methods). Figure 2 illustrates the appearance of each type of feature. Although the features' precise numerical values were not reproduced from one experiment to another, the pattern of change of each feature after exposure to tumor promoter was reproducible. Thus, there was reason to think that the prevalence of the features could be measured with some reproducibility after exposing cultured cells to different reagents.

| Actin architectural features regulated by PKC
It was possible to explain the data of Figure 1c as a function of actin bundle disassembly, as shown in previous studies. Stress fibers were negatively regulated by PKC E, which was greatly reduced in 1000W cells by PMA-initiated activation and subsequent degradation (Li et al. 2006). The role of PKC E and the mechanism of degradation are depicted graphically in Supporting Information Figures S1A and S2A, respectively. PMA affected filopodia much faster than stress fibers, but it is possible that their actin bundles were similarly affected by PKC activation. For example, PKC could be recruited to substrates in the actin core of the filopodia, which would then be rendered susceptible to dissolution after PMA. To determine whether this effect was mediated by PKC, we treated cells with agents that interfered separately and selectively with enzyme maturation and activation, according to the scheme shown in Supporting Information Figure S1B. To . Statistics on ruffling activity, filopodia, stress fibers, and neurites respectively are shown elsewhere (Heckman et al., 1996(Heckman et al., , 2004(Heckman et al., , 2012Li et al. 2006). Bars represent 6 1 standard error of the mean (SEM).
inhibit maturation, we used blocking peptides (BPs) that competed for the phosphorylation site at the hydrophobic motif. Here, the effect would be apparent after a long exposure to the agents. In addition to PKC maturation, the BP may block enzyme localization to a scaffold (see Introduction). Then, short-term exposure to PMA was used to determine whether the filopodia were still disappearing rapidly after PMA treatment.
2.2 | PKC a and g normally turn over filopodia but PKC E turns them over in response to PMA To determine whether the BP rescued the filopodia from destruction by PMA, we analyzed the coverage of the cell perimeter with filopodia in replicate samples following introduction of a hydrophobic segment specific for each of the cPKCs and nPKCs. We also analyzed the effects of BPs representing phosphorylation sites of the myristoylated alanine-rich C kinase substrate (MARCKS), a protein thought to be a universal substrate of PKC. The greatest effect was exerted by the BP for PKC E, as indicated by a 40% elevation in the coverage of the perimeter (Figure 3a). The difference was significant in the one-tailed t test at a level of probability, p < 0.025. BPs representing the homologous PKC d sequence and the phosphorylation site in MARCKS 152-162 also caused an elevation over control, but these results were not statistically significant. The finding that PKC E BP is directly engaged in filopodia destruction was not surprising, because PKC E had caused dissociation of actin cables in PMA-treated cells (Li et al. 2006). To determine whether the PKCs were downregulated in PMA-treated cells, accounting for the PMA-dependent difference in filopodia, we determined the content of the isoforms. We found no difference over times of 0-2 h treatment (Figure 4a Images are made by scanning electron microscopy, except for frame C. In frame C, actin is stained by tetrarhodamine phalloidin and viewed by epifluorescence microscopy. The features are indicated by arrows motif residues targeted by the BP were serine 657 and tyrosine 658. To determine whether they were phosphorylated in 1000W cells, we recovered PKC a under non-denaturing conditions, using an antibody specific to phospho-S657/Y658. Under these conditions, PKC was cleaved and the C-terminal portion of the molecule was recovered as double or triple bands (Figure 4g). This result was similar to those obtained by other laboratories using antibodies against dephosphorylated (England and Rumsby, 2000;Nadra et al., 2005) or phosphorylated (England and Rumsby, 2000;Mirandola et al., 2006) portions of PKC. The PKC a recovery with anti-phospho-S657/Y658 was 45% greater after 1-h PMA treatment. This suggested that phosphorylation at the hydrophobic motif was enhanced after PMA treatment because of changes in recycling or maturation.
For technical reasons, it is not possible to conclude that PKC a was specifically affected by PMA in the treated cells (see Materials and Methods). Although antibody against phospho-S657/Y658 was unlikely to recognize the dephosphorylated motif (Rybin and Steinberg, 2006), the antibody may also recover PKC b and other PKC isoforms. Nevertheless, processing at the hydrophobic site was enhanced by PMA treatment for one or more isoforms.
To determine the effect of BP introduction, we used sodium dodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE) to evaluate proteins from samples treated by PKC a BP or an irrelevant protein. Under denaturing conditions, an antibody directed against the extreme C-terminal sequence, which is rarely phosphorylated in living cells, recovered a full-length doublet when phosphatase inhibitors were included in the lysis solution (Figure 4h,i). A protein degradation product appeared at 37 kDa in PKC a recovered from the BP-treated sample, consistent with the known increase in degradation rate of dephosphorylated PKC (Gysin and Imber, 1996). As similar amounts of PKC a were recovered with antibodies against the dephosphorylated region around S670 and with phospho-S657/Y658 antibody, a fraction of the PKC a molecules in 1000-W cells may be dephosphorylated at S657. The origin of doublets, such as appeared in the PKC a samples, was investigated further by an immunoprecipitation technique. Phosphorylated molecules bind less SDS and are often less mobile than their PKCs were recovered by immunoprecipitation (see Materials and Methods). (a-f) PKCs recovered under denaturing conditions. Lanes 1, 2, and 3 recovered at times of 0 h, one-half hour and 2 h respectively. (g) PKC a recovered under non-denaturing conditions with phospho-PKC a (Ser657/Tyr658) antibody. Lanes 1 and 3, recovery from 30 to 120 mL of lysate after 1-h treatment with solvent vehicle. Lanes 2 and 4, recovery from 30 ml or 120 mL of lysate after 1-h treatment with PMA. (h) PKC a recovery under denaturing conditions from cells treated with PKC a BP or irrelevant protein. Lane 1 5 sample treated with PKC a BP. Lane 2 5 sample treated with irrelevant protein. A degradation product containing the extreme C-terminal is indicated by the arrow. (i) Experimentally confirmed phosphorylation sites in PKC a, adapted from http://www.phosphosite.org [Color figure can be viewed at wileyonlinelibrary.com] residues at the activation loop (Kunkel and Newton, 2015;Rybin and Steinberg, 2006). Nevertheless, the data suggest that some of the PKC a is phosphorylated at the hydrophobic motif in the steady-state, and some molecules undergo further modification during short-term PMA treatment, which may take the form of phosphate addition.
BPs for PKC b and g were also studied in a few experiments. The PKC b hydrophobic motif appeared to have no effect when interference with filopodia was analyzed. With the PKC g BP, the cells developed large, tapering protrusions that could be confused with filopodia. This effect made the samples difficult to evaluate, but it was observed when either BP or transcriptional knockdown agents were employed. Samples treated with BPs for phosphorylation sites in MARCKS and PKC d showed no significant difference compared to controls (Figure 3a,b).
It should be noted that the percentage of the perimeter covered by filopodia was a combination of two other measures. One was the fraction of cells capable of making filopodia and the second, the density of these protrusions on those cells that made filopodia. To determine whether both were affected, we expressed the results for the underlying variables separately. We also included samples from additional experiments in which the PKC E BP was tested along with other reagents, so there were more experiments to compare. Coverage values again showed significance when compared to PMA-treated control (cf. Figure 3a It was surprising that BPs enhanced filopodia in the absence of PMA, and perhaps even more surprising, that these BP sequences represented both the cPKC and nPKC classes. In comparing the effects of BPs directed against PKC E and h, however, it should be noted that activation was detected only for the isoforms E and g under the current experimental conditions (Supporting Information Figure S2). It is possible that PKC h failed to destroy filopodia because its activity was not affected by PMA under the conditions of the experiments. Previous workers also found PKC h to be resist-ant to conformational changes after PMA exposure (Kang, French, Sando, & Hahn, 2000).

| A PKC inhibitor reproduces the pattern of filopodia rescue
To clarify these findings, we introduced a PKC inhibitor (PKCI, Myr19-27) representing the myristoylated nonapeptide pseudosubstrate sequence of PKC a/b. The sequence of this peptide was identical to PKC a/b and a near match for the PKC g sequence and may be a competitive inhibitor of all PKCs. Treatment with a high concentration, which had inhibited PLD-dependent physiological effects in previous studies, showed variable effects and was often inhibitory (Figure 5A). At a low concentration, representing the IC50 for MARCKS phosphorylation, the inhibitor had no effect on filopodia. However, it rescued filopodia from PMA-initiated destruction. This suggested that the PKC targeted by the inhibitor was activated by PMA, but the insertion of PKCI into the active site prevented the enzyme from binding its usual substrates.
Although these results suggest that the cPKCs may act to destroy filopodia following PMA, they do not prove it, because the a/b pseudosubstrate sequence may also bind nPKCs. It is known that the PKC E pseudosubstrate competed with the corresponding PKC a sequence for binding partners in vitro (Liao, Hyatt, Chapline, & Jaken, 1994). Substrate specificity of the PKC isoforms is rather lax (Nishikawa, Toker, Johannes, Songyang, & Cantley, 1997). This lack of specificity would permit the PKC a pseudosubstrate to dock in the nPKC active site (Wu-Zhang and Newton, 2013). Rescue was unaffected by inclusion of C2-ceramide (Figure 5a). With a still lower (4 mM) concentration, the rescue effect was unchanged (data not shown). We attempted to inhibit the effect of nPKCs selectively by applying a myristoylated EGFR BP matching the specific sequence for PKC h substrates (Nishikawa et al., 1997). Concentrations representing the IC50 were toxic, however, and the results could not readily be interpreted.

| Protein tyrosine phosphatase (PTP) inhibitors reproduce both patterns-Steady-state elevation and filopodia rescue
It has long been known that PKC is associated with a phosphotyrosinecontaining protein that is a binding partner of integrin in filopodia (Wu, Wang, Mason, & Goldberg, 1996). Moreover, a similar substrate was thought to reside upstream of Rho GTPase where it could regulate focal adhesion assembly (Schoenwaelder and Burridge, 1999). Although neither the kinase responsible for phosphorylation nor the substrate is known, c-Src (Rous sarcoma virus nonreceptor tyrosine kinase) may be a substrate of PKC (Thuringer et al., 2010). Thus, it is possible that PKC regulates filopodia upstream of c-Src. In preliminary work, we found that phenylarsine oxide (PAO), a compound that modifies thiol groups, hence inactivating any PTPs with the XCysXXCysX motif, enhanced filopodia (De, 2015). PAO has many effects on the cytoskeleton, and in some cases affects RhoA GTPase, which has vicinal cysteines within 2.5 | Comparison of PKC E and cPKCs with respect to PMA-stimulated turnover The above data suggested that PMA activates and translocates the PKC E isoform to a membrane where it phosphorylates a substrate that destroys filopodia. The PKCI effect, however, suggested that PKC E may not be the sole isoform responding to PMA ( Figure 5A). We compared the possibility of parallel effects of PKC a/b with those of PKC E by reducing the levels of these isoforms selectively and then analyzing filopodia prevalence. Because PKC molecules had a long half-life under steady-state conditions, it was impossible to eliminate them completely by transcriptional knockdown. Nevertheless, the effects of knocking down one isoform could be modeled to good effect, as long as the amounts of the other isoforms in the re-balanced mixture were known (Table 1, Figures S3 and S4 in Supporting Information). We measured filopodia by an unbiased method in order to better detect the differences (see Materials and Methods). In the absence of PMA, the knockdown of any one isoform had little effect on the cell phenotype. After Effect of specific inhibitors of TC-PTP and PTP-MEG2 with and without PMA. The mean of two experiments is shown 6 1 SEM. *difference significant at p < 0.05, **difference significant at p < 0.025, *** difference significant at p < 0.001. When combined data for two concentrations of PTP-MEG2 were tested against control, the difference was significant at p < 0.004 a 2-h exposure to PMA, however, the filopodia prevalence in the PKC a knockdown sample was lower than in either the control or PKC E knockdown samples (Figure 6a). The differences were statistically significant, as shown by the designation of the sample means by different letters on Table 2. With PKC b knockdown, the filopodia prevalence was significantly lower than the control but was indistinguishable from the value for PKC E knockdown (Table 2). This suggested that PKC a protected filopodia from destruction by the PKCs whose content was left unchanged by treatments that knocked down PKC a transcription.
Thus, it was unlikely that PKC a participated in promoter-stimulated turnover. To confirm that the procedure was effective, the quantities of PKC a and E were determined after knockdown of the respective isoforms (Table 3, Supporting Information Figure S2D,F).
To visualize these interactions, the unbiased measure of filopodia (factor 4) was regressed against the content of the remaining isoforms in samples selectively reduced in PKC a, b, or E. The content was  Therefore, the knockdown of PKC a content linked the mass of the remaining PKCs to the destruction of filopodia (Figure 6b). This effect may be indirect, however. For example, PKC a knockdown may merely decrease the number of molecules competing with other PKCs for PMA, or decrease the competition with other PKCs for occupancy of a scaffold.
The above result was investigated further to determine whether the resistance to filopodia destruction was related to ruffling activity, which was known to be inversely correlated with factor 4 (Heckman et al., 2012). By measuring ruffling frequency in the same cells as shown in Figure 6a,b, we found that the decrease in PKC a content caused a marked activation of ruffling. The R 2 for PKC a-depleted samples was 0.42. The positive slope of this regression curve indicated that 42% of the increased ruffling could be attributed to an increase in the content of residual PKCs, which consisted mainly of the PKC E and g isoforms (Table 1). Therefore, PKC a conserved filopodia against PMAstimulated destruction. This effect may be indirect, e.g. by counteracting the stimulation of ruffling by E and g isoforms or other PMA receptors. Because the remaining PKCs largely consisted of E and g, and both were activated by PMA, it is probable that one or both enhanced ruffling. As the R 2 levels for the other isoforms were <0.15, it was unlikely that any other PKCs, even PKC b, acted like PKC a (Figure 6c).
PKC b was at the lowest level of any isoform in 1000W cells (Table 1).
Thus, the data suggested that PKC a antagonized the removal of filopodia by one or more of the other PKC isoforms, probably by PKC E.
Although their levels rebounded as the PKC content in the pool declined after 2 h of PMA treatment, filopodia never regained the prevalence observed at time zero (Figure 6a). Depletion of either PKC a or E from the pool of PKC molecules caused a significant elevation in filopodia prevalence at 5 h (Table 2). This did not persist over longer times, however, as factor 4 values for samples between 2 and 10 h uniformly showed R 2 values <0.12 (Figure 6d). Although fluctuations in filopodia prevalence were independent of the size of PKC pool at later times after PMA exposure, ruffling frequency continued to show a positive relationship to PKC content. Ruffling frequency showed R 2 values greater than 0.2 (except for the epsilon knockdown) and positive slope.
For the sham-treated control, 33% of the increased ruffling could be attributed to differences in PKC content ( Figure 6e). As PKC E and PKC g were known to be activated by PMA exposure (see Supporting Information Figure S2), they could continue to stimulate ruffling.
Indeed, PKC E knockdown decreased ruffling throughout the acute phase of ruffling stimulation (Figure 6f). The analysis was consistent with an interpretation that PKC E activation was responsible for most of the filopodia destruction, and it destroyed filopodia by enhancing ruffling activity.   (Jansen et al., 2001). Of all isoforms of PKC, PKC E is most frequently found as an oncogene in various types of cancer. The relationship between PKC E and malignancy is discussed in recent reviews (Garg et al., 2014;Urtreger, Kazanietz, & Joffe, 2012). In contrast, PKC h acts as a tumor suppressor gene in the skin (Kashiwagi, Ohba, Chida, & Kuroki, 2002).
One mechanism by which the BP rescues filopodia may have to do with the capability of PKC E to enhance PDK1 autophosphorylation (Garczarczyk et al., 2009). PKC E's hydrophobic motif must be phosphorylated, however, in order for PDK1 to bind (see Introduction).  Good et al., 1998). Although it is unlikely that a decrease in AGC kinases overall could be responsible for the "rescue" effect, which was specific to a kinase or kinases activated by PMA, it is possible that the rescue was caused by BP-mediated repression of PKC E maturation. To date, we have not been able to implicate another kinase in addition to PKC E in filopodia destruction. Another possible mechanism by which the PKC E BP conserves filopodia is that, if the hydrophobic BP becomes phosphorylated, it competes for the site where PKC E is localized at the Golgi region, which requires phosphorylation at the hydrophobic motif (Xu et al., 2007).
The above mechanisms appeared specific to PKC E, as filopodia were not rescued by the BPs for other nPKCs found in 1000W cells.
The contrast between the isoforms is especially noteworthy in regard to PKC h. PKC h, like PKC E, is often localized to the Golgi apparatus.
PKC h acts downstream of a b1g2 or b3g2 heterodimer from G protein-coupled receptor and forms a complex which drives the activation of PKD and the generation of complexes between PKC and PKD localized to the TGN membrane (Añel and Malhotra, 2005). It is thought that both PKC h and E can phosphorylate residues in the activation loop of PKD, driving activation in some cell types. However, a PDK1/PKC h/PKD complex can be formed that exerts negative regulation on PKC h (Brändlin, Eiseler, Salowsky, & Johannes, 2002a). The PKC h/PKD complex is unique among the nPKCs in two ways: (1) it is essential for Golgi integrity (Añel and Malhotra, 2005) and (2)  The mechanism of action of the PKC a BP remains unknown. It is unlikely that PKC a or b was hyperactivated by the promoter treatment, as they are activated synergistically by diacylglycerol and calcium (Liu and Heckman, 1998), and the latter is generally at low concentrations in the cell. The role of PKC a in opposition to PKC E also remains HECKMAN ET AL.

| 307
unknown. The role of PKC a may be passive, i.e., it may compete for a scaffold or receptor that is also occupied by PKC E. There is no evidence in these studies that the PKC a BP targets the same cellular trafficking functions as that against PKC h. The effects of the PKC a BP, however, suggested destabilization of the folded structure, which would be expected in the absence of a stabilizing phosphorylation at the hydrophobic motif (Gould and Newton, 2008;Newton, 2010). The lack of phosphate at the hydrophobic site prevents PKC maturation to the competent state and increases its rate of degradation (Gysin and Imber, 1996). Thus, the BP may work by decreasing the amount of competent protein, as well as increasing its turnover. Our determinations of PKC mass recovery with phospho-PKCa (Ser657/Tyr658) antibody with and without PMA treatment suggested that PMA caused an increase in competent molecules. This may have contributed to the effect by which the alpha isozyme counteracted the PKC E-mediated destruction of filopodia in PMA-treated cells (Figure 6a). Neither activity nor competency, however, can be predicted from phosphorylation states, as mentioned previously by other researchers (Freeley et al., 2011;Kunkel and Newton 2015).

To gain additional information as to what other PKCs responded
to PMA in the system, we used a myristoylated BP inhibitor of cPKCs.
The inhibitor rescued filopodia from PMA-initiated destruction but had little effect on filopodia in the steady state. As mentioned above, there was no way of excluding that it inhibited PKC E. C2-ceramide treatment did not affect the rescue of filopodia by the PKC a/b inhibitor.
C2-ceramide selectively inhibits the translocation of cPKCs to membranes (Jones and Murray, 1995) and stimulates a serine/threonine phosphatase that dephosphorylates residues at the activation loop (Kitatani, Idkowiak-Baldys, & Hannun, 2007), so the lack of effect in the experiments could be because the inhibitor actually targeted PKC E. Although PKC g was known to be activated following treatment with nanomolar concentrations of PMA (Li et al., 2006), PKC g BP appeared to have little effect on filopodia.
3.2 | Tyrosine kinase substrate associated with PKC As mentioned above, there is previous evidence for a tyrosine kinase substrate closely associated with focal contacts. TC-PTP binds to and is activated by the integrin a1 subunit, possibly at the focal contact (Mattila et al., 2005). Because the proto-oncogene, c-Src, and PKC are transient components of the focal contacts, the kinase may be c-Src. If so, several candidate substrates exist. One, p130Cas (Crk-associated substrate), is both a scaffold for c-Src and a substrate whose many phosphorylation sites are subject to processive phosphorylation (Pellicena and Miller, 2001). A second, FAK (focal adhesion kinase), is required for the turnover of focal adhesion sites (Westhoff, Serrels, Fincham, Frame, & Carragher, 2004). Other probable substrates of c-Src, e.g., PKL/GIT2 (paxillin kinase linker/G-protein coupled receptor kinase-interacting protein) or paxillin, may be candidates for the unknown phosphorylated component of filopodial focal contacts. Unfortunately, no specific inhibitor of c-Src was available. Experiments with highly selective PTP inhibitors confirmed the existence of two patterns. Thus, PTP inhibitors could be grouped into classes. A number of them enhanced filopodia in the steady-state with-out rescuing them from PKC-mediated destruction. The second class, exemplified by TC-PTP, replicated the effect of PKC E BP.

| Role of promoter-specific turnover in cell motility
The data suggested that, in the epithelial cells employed here, PKC E is activated by PMA and stimulates ruffling which, in turn, removes filopodia. Moreover, the data showed that, at the levels normally present in cytoplasm, PKC a suppressed ruffling and thereby inhibited the destruction of filopodia by PMA. In turning over filopodia, PKC E probably circumvents the role of both PKC a and h. Currently, we have little understanding of how this promoter-specific turnover contributes to carcinogenesis and tumor promotion. Nevertheless, as cells lose sensors during their conversion into cancer cells, they may have a weaker response to contact and less directional persistence, or they may treat all surfaces as unfavorable for attachment. Our finding that PKC a can suppress ruffling and thereby affect promoter-specific turnover of filopodia reinforces the need for investigating PKCs as a system. As R 2 values around 0.3 represent minor effects, the suppression of ruffling observed is small. Nevertheless, it is clear that there is some opposition in the cells between PKC a and PKC E. PKC a can act as a tumor suppressor or an oncogene, depending on the context, so the same duality has been found for its in vivo activity (Griner and Kazanietz, 2007).
In the context of signaling, what benefit would accrue to a cell if it ensured the long-term suppression of filopodia? It may be a mnemonic device. In previous work, we showed that cells require filopodia in order to orient themselves on a haptotactic gradient (Amarachintha et al., 2015). Cells may activate PKC E selectively on the less adhesive margin of the cell and thereby reduce the prevalence of filopodia.
When a cell meets a surface that is unfavorable for attachment, it may retract filopodia through the promoter-specific turnover mechanism while setting a marker to "remember" the direction of unfavorable contact. In nature, the marker is probably the sustained formation of diacylglcerol from phosphatidylcholine via PLD and generation of phosphatidic acid, which are required to maintain PKC in the membrane (Lopez-Andreo, Gomez-Fernandez, & Corbalan-Garcia, 2003).
This mechanism may be invoked during development of the nervous system, accounting for the importance of PKC in neural guidance (Bonsall and Rehder, 1999;Cheng, Mao, & Rehder, 2000).

| Treatment with BPs, small inhibitory RNAs (siRNAs), and ODNs
We designed peptide sequences corresponding to the phosphorylation sites within the hydrophobic sequence of protein kinase Cs (Table 4).
We also investigated MARCKS as a candidate for regulating filopodia.
To this end, BPs were also designed to block the three sites, 152, 153, and 163, phosphorylated in the effector domain (Heemskerk, Chen, & Huang, 1993). Peptides were delivered into the cells with BioPORTER reagent (Genlantis, San Diego, CA), a lipid formulation that facilitates the transfer of peptides and proteins into the cytoplasm. It was previously found to be harmless to 1000W cells (Heckman et al., 2009 (Li et al., 2006). PTP inhibitors were synthesized as previously described (Zhang et al., 2009). The inhibitory constants (K i ) for the compounds shown, TC-PTP and PTP-MEG2, were 4 and 34 nM, respectively.

| Fixation, microscopy, and analysis
Dishes containing Ge-coated substrates were fixed for 10 min with warm (378C) formaldehyde made up from paraformaldehyde in cytoskeletal buffer (pH 7.4). Samples were rinsed with phosphate-buffered saline and stored refrigerated until analyzed.
Each substrate was labeled with a code number, mounted on a glass slide, and sealed with nail polish, then observed by phase contrast in a Zeiss Axiophot light microscope using a Plan-Neofluar 1003/1.30 objective lens. Samples were imaged in a raster pattern, and all single cells with clearly visible edges were scored. Any cell with 25% or more of the edge rounded up was excluded from the analysis (De, 2015). Because the 1000W cells adhered tightly to the substrate, retraction fibers were rarely seen, and all the filopodia were adnate under conditions of the experiments. Occasionally, wavy filopodia which were thought to represent retracting filopodia were seen (Figure 3). The images of Figure 3 represent the full range of morphologies found in the samples. Filopodia of the 1000W cells never rose vertically from the substrate, nor did the agents used here ever cause them to loop back and adhere to the cell body through their tips. Structures were excluded from the analysis if they showed a decrease in actin bundling (Figure 3j) or were too large ( Figure 3k). The standard for selecting filopodia was based on illustrations of cells rich in filopodia as defined by automated pattern processing and analysis software (Heckman et al., 2012). For each sample, counts were made by two or three independent observers.
It should be noted that the number of filopodia on the cell perimeter was subject to considerable variation in repeat experiments. This was due in part to varying conditions of attachment, spreading and growth of the cells, which cannot be precisely replicated in repeat experiments. A predictable pattern of decrease in filopodia occurs, however, as shown in Figure 1b. Thus, raw coverage could vary from 5% to 20%, and the data were typically presented as a ratio between the counts for each sample to baseline represented by sham-treated control. The number of cells counted on each sample is shown in Supporting Information Table S1.

| Unbiased classification of cell features
The above method used to quantify filopodia relied mainly on a subjective evaluation of the cell perimeter. A more general method of classifying cell features, based on the mathematical deconstruction of the cell into its native parts, was used in the transcriptional knockdown experiments, however. This enabled features to be recognized qualitatively and their prevalence in a sample to be evaluated on a quantitative basis.
The initial classifications were derived from images acquired by Tolansky interference, which allowed information from three dimensions to be included in the data compilation (Heckman and Jamasbi, 1999). Each of the three lowermost interference contours was evaluated with respect to 34 geometrical variables, and each of these 102 variables was rendered dimensionless by dividing them by the value of a dimensioned variable, i.e., area, perimeter, or length of the major axis of the ellipse of concentration. By extracting principal components and computing the latent factors, i.e. theoretical variables that account for the covariance of variables in the primary variable dataset, we reduced the variable set to 20 factors. For samples collected in experiments, each primary variable was multiplied by a positive or negative loading constant to convert the values into the factor score. SAS software (SAS Institute, Cary, NC) was used together with software written by the laboratory (Heckman et al., 2009) for data conversion into factor values.
Valid solutions for the edge features could be obtained from images generated by scanning electron microscopy (Heckman et al., 2009) or phase microscopy (Amarachintha et al., 2015). Cells were prepared for scanning electron microscopy as previously described (Heckman, Kanagasundaram, Cayer, & Paige, 2007). By extracting the edge contour of each cell and processing the digital information as above, we could detect small differences and analyze features without relying on subjective criteria. In these experiments, the prevalence of filopodia (factor 4) was analyzed starting with samples imaged in a Hitachi S2700 scanning electron microscope. Edge contours were converted into digital format and contour extraction performed as previously described (Heckman et al., 2009). At least 30 cells were sampled for each experimental treatment.
The current method was compared with other methods of semiautomated feature analysis. The other methods are designed to automate the analysis of features that are already known, and are mainly applied to neurons where the filopodia are large and concentrated in growth cones or dendrites. One such method (Hendricusdottir and Bergmann, 2014) requires the operator to manually identify the tip and base of the filopodium in one of a series of time-lapse images. The direction of travel of the dendrite must also be determined manually. We note that another of these methods (Costantino et al., 2008) requires manual selection of a threshold for converting gray-level to binary format. Manual setting of a derivative value for the Laplacian of two-dimensional Gaussian for edge detection is also required. Finally, the efficacy of these methods must be confirmed by tracing the same structures manually or using other programs for tracking dynamic changes. The objective of these studies was to analyze dynamic changes in length, as well as vertical and lateral tilt of filopodia. These, as well as an additional technique designed to determine the presence of proteins along the length of the filopodium (Saha et al., 2016), are not suitable for data sets comprising thousands of filopodia. These methods were not congruent with the objective of the current studies, which was to understand prevalence in order to identify mechanisms for filopodia regulation. In the current research, subjective evaluations were used to identify agents that affected filopodia regulation. Completely unbiased feature identification was essential to identify morphological changes occurring in oncogenic transformation.

| Statistics
Values for filopodia obtained by phase microscopy were tested and found to occupy a normal distribution. Calculations and statistical analysis were done using Excel, except for the Student's t-test, which was done using the GraphPad online service. Statistical tests on samples processed by unbiased classification were done with SAS software (see Unbiased classification of cell features).