Cortactin Phosphorylation by Casein Kinase 2 Regulates Actin-Related Protein 2/3 Complex Activity, Invadopodia Function, and Tumor Cell Invasion

Cell invasion from the primary tumor is responsible for initiating the metastatic cascade and increasing cancer lethality (1, 2). Invasion is initiated in part through the action of invadopodia, actin-based membrane protrusions produced by tumor cells that mediate dissemination by degrading restrictive extracellular matrix (ECM) proteins through enzymatic matrix metalloproteinase (MMP) activity (3). Invadopodia contain a central filamentous (F-) actin core surrounded by an integrin-based adhesion ring complex that anchors the structure to allow focal matrix degradation and tumor cell protrusion through the basement membrane (4). Cortactin and actin-related protein (Arp) 2/3 complex are essential protein components involved in invadopodia precursor core formation required for subsequent MMP recruitment and membrane protrusion (5, 6). Cortactin overexpression is common in several cancer types including head and neck squamous cell carcinoma (HNSCC), resulting in enhanced motility, invasion, and invadopodia activity (7, 8). Cortactin binding to Arp2/3 complex activates Arp2/3 actin nucleation activity, enhancing cellular actin polymerization to form branched F-actin networks (6, 8–10). Cortactin also directly binds F-actin and bundles newly-formed filaments, providing an overall stabilizing effect on the Arp2/3-F-actin network required for invadopodia formation (5, 6, 11). Previous work has shown that a DDW motif within the cortactin N-terminal acidic (NTA) domain is central in mediating Arp2/3 activation and branched actin network formation (7, 12–14). This region is similar to the Arp2/3 binding motif found in the acidic region of the Verprolin, Central, Acidic (VCA) domain of the Wiskott–Aldrich Syndrome protein (WASp)-family of Arp2/3 nucleation promotion factors (NPF). Although the cortactin DDW motif is well established as the region responsible for Arp2/3 binding, posttranslational or other modifications of amino acids in the NTA region that regulate binding have not been reported. Tyrosine and serine phosphorylation of cortactin residues in the carboxyl-terminal region are essential for invadopodia formation, cellular invasion, and tumor metastasis through multiple mechanisms ultimately involving activation of WASp NPFs (6, 7, 15–18). In addition, comprehensive phosphorylation site mapping by mass spectroscopy has identified NTA phosphorylation sites in close proximity to the DDW motif (19). This raises the possibility that phosphorylation of one or more of these residues may serve to govern Arp2/3 binding and invadopodia function in invasive cancer.

Casein kinase (CK) 2 is a ubiquitously expressed, constitutively active serine/threonine kinase consisting of two catalytic subunits (α or α') and two β regulatory subunits (20). Increased CK2 expression correlates with cell-cycle progression, apoptosis resistance, and tumor cell motility in various cancers (20). Overexpressed CK2 enhances HNSCC tumor cell motility (20). CK2 phosphorylates the cortactin homologue HS1 at an unidentified site(s) in the NTA region (21), as well as residues near the DDW region in the NPFs neuronal (N)-WASp and WAS protein family member 2 (WASF2; WAVE2, WASp family Verprolin-homologous protein 2; refs. 22–25). Here we show that CK2 phosphorylation of threonine (T) 24 in the cortactin NTA impairs binding to and activation of Arp2/3 complex. Cortactin T24 and CK2 are required for efficient invadopodia formation and ECM degradation activity in HNSCC cell lines. Treatment of established and primary HNSCC cells with the selective CK2 inhibitor Silmitasertib impairs invadopodia function and regional HNSCC invasion. These results identify a new mechanism of invadopodia regulation that can be targeted to impair HNSCC invasion.

HNSCC cell lines OSC19 and UMSCC1 were acquired and maintained as described (26). MDA1586 cells were obtained in March 2014 from Barbara Frederick (University of Colorado, Denver, CO). All HNSCC lines were authenticated by STR profiling at the University of Arizona Genetics Core in June 2017. PCR-based mycoplasma testing (13100-01; Southern Biotech) was conducted on OSC19 and UMSCC1 lines in March 2015 and were free of contamination. The MDA1586 line was not tested for mycoplasma. HEK293T/17 cells were obtained in April 2013 from Robert Wysolmerski (West Virginia University, Morgantown, WV). NIH3T3 cells were obtained in June 2017 from Ivan Martinez (West Virginia University, Morgantown, WV) and maintained for ≤10 passages. These lines were not authenticated or tested for mycoplasma. Cells were propagated in DMEM supplemented with 10% FBS and 1% penicillin–streptomycin for <6 months. OSC19 and UMSCC1 cells stably infected with pLKO.1-puro cortactin shRNA or CK2α shRNA were generated by clonal puromycin selection following standard methods (CSNK2A1: TRCN0000380839, TRCN0000027627; CTTN: TRCN0000040275).

UMSCC1 cells stably infected with pLU-Luc2 expressing luciferase were generated following standard methods. Murine cortactin rescue OSC19 and UMSCC1 cells containing cortactin shRNA stably infected with pLenti CMV Hygro cortactin constructs were generated by subsequent clonal hygromycin selection. Complete cortactin knockdown in OSC19 and UMSCC1 cells was achieved by transfection of cortactin-targeting siRNA (ON-TARGETplus SMARTpool L-010508-00-0020; Dharmacon) using a Nucleofector I (Amaxa Biosystems).

Western blotting was conducted as described (27) and visualized with autoradiography film (E3012; Denville Scientific) or captured by an Amersham Imager 600 (GE Healthcare Bio-Sciences). Antibodies used were: anti-cortactin clone 4F11 (1 μg/mL; ref. 26), anti-pS473 AKT (#4060, 1:1,000; Cell Signaling Technology), anti-panAKT (#2920, 1:1,000; Cell Signaling Technology), anti-β-actin (#8457, 1:1,000; Cell Signaling Technology), anti-CK2α (#2656, 1:500; Cell Signaling Technology), anti-DYKDDDK (FLAG) clone 2EL-1B11 (MAB3118, 1:500; Millipore), and anti-Arp3 (#07-272, 1:500; EMD Millipore). Immunoprecipitation was conducted from cells lysed in 50 mmol/L Tris Buffer pH 8.0 with 10 mmol/L EDTA and 1% NP-40 (28). Clarified lysates (1 mg) were incubated with 50 μL of FLAG M2 affinity resin (A2220; Sigma-Aldrich) for 2 hours at 4°C. Immune complexes were collected by centrifugation, washed twice with Tris buffer, separated by SDS-PAGE, and Western blotted with antibodies as described above.

Cells were plated on Oregon Green 488-conjugated gelatin (G13186; Invitrogen) coated coverslips (29). In cases of inhibitor treatment, cells were allowed to attach for 1 hour, then incubated for 12 or 24 hours with Silmitasertib (S2248; Selleckchem) as indicated. Cells were rinsed in PBS, fixed with 10% buffered formalin (SF100-4; Fisher Scientific), and labeled as described (29). Antibodies used were 4F11 (1:500) or anti-FLAG (1:500). Primary antibodies were visualized using Alexa Fluor 647 conjugated goat anti-mouse secondary antibody (A21235, 1:500; Invitrogen). F-actin was visualized with rhodamine-conjugated phalloidin (R415, 1:1,000; Invitrogen). Coverslips were mounted using ProLong Gold antifade with DAPI (P36935; Invitrogen). Images for quantifying gelatin degradation and knockdown/rescue expression were acquired with a Zeiss Axio Imager Z2 epifluorescent microscope equipped with an AxioCam MRm CCD camera and AxioVision software using LD Plan-Neofluar 40X/0.6 Corr and Plan-Apochromat 63X/1.4 oil objectives (Carl Zeiss Microscopy). Acquisition parameters were held constant within comparison groups. Confocal images were acquired using a Zeiss Axio Imager Z1 LSM510 confocal microscope with EC Plan-Neofluar 40X/1.30 and Plan-Apochromat 63X/1.4 oil objectives and Zen2009 software (Carl Zeiss Microcopy). All representative images were level adjusted to enhance contrast and brightness as needed and resized using Photoshop CC 2018 (Adobe Systems). Gelatin images were corrected for uneven illumination via bandpass filtering using ImageJ software (NIH). Degradation and invadopodia formation was quantified as described previously (29), with n ≥ 70 lentiviral infected or ≥100 inhibitor-treated cells evaluated for each condition. FLAG-stained control images were thresholded against nonspecific staining using ImageJ software. Cells above threshold values were considered positive for rescue construct expression and used for quantitation. For therapeutic treatments and RNAi stable cell lines, degradation and cell areas were determined by Image J (NIH) on an individual cell basis. Data represent the mean values normalized to control degradation area per cell area from at least three independent experiments. Invadopodia precursors were determined by colocalization of actin and cortactin at sites lacking gelatin degradation. Active invadopodia were determined by colocalization of actin, cortactin, and gelatin degradation. Data represent the mean from at least three independent experiments. Phase contrast images were acquired using a Zeiss Axiovert 200M microscope equipped with an AxioCamMR CCD camera using a Plan-Neofluar 10×/0.30 objective and AxioVision software (Carl Zeiss Microscopy).

In vitro kinase assays were performed as described (30). Briefly, 0.25, 0.5, or 1 μg of purified GST-wild-type (WT) or T24A cortactin NTA fusion proteins were incubated with 8 ng CK2α (#14-445; Millipore) and 10 μCi ³²Pγ-ATP (#NEG002A500UC; PerkinElmer) at 30°C for 10 minutes. Reactions were terminated with hot SDS sample loading buffer. Proteins were visualized by autoradiography. Purified N-WASp GST-VCA (0.5 μg) and GST (1 μg) were used as respective positive and negative controls.

Purified WT or T24A cortactin proteins (2.5 μg) were bound to 4F11-conjugated protein G magnetic beads (#10003D; Life Technologies). Immune complexes were incubated in the presence or absence of activated CK2α (75 ng; #V4482; Promega) and ATP (500 nmol, #BP413-25; Fisher Scientific) at 30°C for 15 minutes. Reactions were washed twice with 10 mmol/L Tris pH 7.4, 150 mmol/L NaCl, 0.5 mmol/L EDTA. Complexes were washed once with 10 mmol/L Tris pH 7.4, 10 mmol/L EDTA, and incubated with 50 ng Arp2/3 complex (#RP01-A; Cytoskeleton) at 4°C for 30 minutes. Following incubation, binding complexes were washed once with 10 mmol/L Tris Buffer pH 7.4 with 25 mmol/L NaCl, 10 mmol/L EDTA, 1% NP-40, then boiled and Western blotted with antibodies against cortactin and Arp3.

Actin polymerization experiments were conducted as described previously (31). Reactions contained 2 μmol/L actin (10% pyrene-labeled), 75 nmol/L Arp2/3 complex, 100 nmol/L cortactin, or 50 nmol/L GST-VCA (#VCG03; Cytoskeleton), and/or varying amounts of CK2α (#14-445; Millipore) as indicated. For reactions with CK2α, GST-VCA, or cortactin mutants were preincubated with CK2α and 500 nmol ATP for 15 minutes at room temperature prior to addition to the actin polymerization reaction.

Patient-derived xenograft (PDX) tumors and cell lines were established as described (32). WVUSCC-AR2 and WVUSCC-AR5 were derived from surgical specimens of alveolar ridge HNSCC in compliance with West Virginia University Institutional Review Board approved protocol #1310105737A033. PDXs were developed in compliance with West Virginia University Institutional Animal Care and Use Committee approved protocol #15-0302.6 by placing approximately 1 mm tumor fragments into subcutaneous pockets in the flanks of anesthetized 8- to 10-week-old NOD/SCID-γ (NSG) mice. Tumor fragments were overlayed with Matrigel (354234; Corning) and incisions were closed using wound clips. Mice were weighed and monitored for tumor growth on a weekly basis. PDX tumors were passed into new NSG mice and/or used to generate cell lines once tumors reached ∼1 cm in greatest dimension.

For cell line derivation, PDX tumors were minced and digested in DMEM supplemented with 20% FBS and 1 mg/mL collagenase IV (17104019; Gibco). Digested tissues were plated onto NIH3T3 fibroblasts senesced with 4 μg/mL mitomycin C (BP2531; Fisher Scientific) and cultured in DMEM:F12 1:1 supplemented with 10% FBS, 400 ng/mL hydrocortisone (H0888; Sigma), 50 μg/mL gentamycin (15750060; Gibco), 5 μmol/L ROCK inhibitor (S1049; Selleckchem), 0.5 ng/mL recombinant human EGF (PHG0311; Gibco), and 10 ng/mL cholera toxin (C8062; Sigma). Both WVUSCC-AR2 and -AR5 were derived in August 2017 and maintained for ≤10 passages. Derived lines were verified using cytokeratin 14 staining (ab15462; Abcam). Neither STR profiling nor mycoplasma detection was performed on these cell lines. Prior to utilization in gelatin degradation or spheroid invasion assays, PDX-derived cell lines were plated directly onto cell culture dishes for one to two passages to remove the fibroblast population. Gelatin degradation and spheroid invasion assays were performed in DMEM supplemented with 10% FBS.

3D spheroid invasion assays were performed as previously described (26). A total of 1 × 10⁴ (OSC19) or 2.5 × 10⁴ (UMSCC1 and WVUSCC-AR5) cells were plated into individual wells coated with 1.5% noble agar for 24 hours (UMSCC1) or 48 hours (OSC19 and WVUSCC-AR5) to form spheroids. For each line, spheroids were collected, resuspended in 500 μL of 2 mg/mL rat tail collagen I (354236; Corning), and plated into individual wells of a 24-well plate precoated with 400 μL solidified 2 mg/mL collagen I. Plates were incubated for 1 hour at 37°C, then overlayed with 1 mL DMEM supplemented with 10% FBS and 1% penicillin–streptomycin containing DMSO or 10 μmol/L Silmitasertib. Spheroid invasion was visualized at the indicated time points by phase contrast microscopy using a Zeiss Axiovert 200M microscope equipped with an AxioCamMR CCD camera using a Plan-Neofluar 5X/0.15 objective and Axiovision software (Carl Zeiss Microscopy). Maximal radial distances for invaded cells were calculated using Axiovision software, with invasive distance determined as the difference between the initial and final maximum radius for each invaded spheroid.

Tongue tumor establishment was adapted from previous work (33). A total of 2.5 × 10⁴ luciferase expressing UMSCC1 cells were injected into the tongues of 8- to 10-week-old NSG mice (purchased from the West Virginia University Transgenic Animal Core Facility). Mice were maintained using transgenic dough diet (S3472; Bioserve) and weighed every 2 to 3 days. Tumor growth was monitored by bioluminescent imaging using 150 mg/kg d-luciferin (122796; Caliper Life Sciences) injected intraperitoneally, followed by in vivo whole-body bioluminescence imaging using an IVIS Lumina-II system and Living Image 4.0 software (PerkinElmer). Tumors were allowed to establish for 1 week, then mice were divided equally into two groups based on approximate tumor size. Mice were given 50 mg/kg Silmitasertib in DMSO or DMSO alone by oval gavage twice daily for 3 weeks. Mice were subsequently euthanized, tongues excised, processed, and stained for histologic analysis.

To quantify invasion parameters, whole tongue histologic images were cropped to encompass the tumor invasive front and analyzed using ImageJ. Images were processed with the colour deconvolution 1.5 plugin using H&E or H&E2 presets. Resultant colour_1 images were 25% contrast enhanced before conversion into binary images. ROIs were selected for particles above 15,000 pixel units and manually verified by overlay onto the original H&E image to remove artifacts. Invasive protrusions were defined as projections at the leading edge of the tumor surrounded by stroma on three sides and identified on the binary image using the polygon selection tool. Invasive distance was determined as the difference between the farthest edge of the protrusion and the protrusion base.

Differences in mean values between groups were evaluated using Student or Welch t test. Significance was determined at P < 0.05 utilizing GraphPad Prism 7 software. Error bars represent ± SEM.

Phosphorylation of serine (S) 11, T13, and T24 in the murine cortactin NTA domain has been reported (19). The proximity of these residues to the canonical Arp2/3 binding motif consisting of amino acids 20–22 (DDW) have the potential to regulate Arp2/3 binding (Fig. 1A). To determine if these residues influence cortactin binding to Arp2/3, FLAG-tagged murine cortactin constructs were generated that contained serine to alanine (S11A) and threonine to alanine (T24A) phosphorylation-null mutations. T13 was not evaluated because it is not conserved in human cortactin. Co-immunoprecipitation studies indicate that S11A cortactin bound endogenous Arp2/3 at levels similar to WT cortactin, whereas T24A cortactin failed to effectively bind Arp2/3 despite retaining the DDW binding motif (Fig. 1B). Threonine to aspartic acid (T24D) phosphomimetic cortactin bound Arp2/3 at reduced levels compared with WT (Fig. 1B). These data demonstrate that both the DDW motif and T24 are required for optimal Arp2/3 complex binding. Furthermore, reduced Arp2/3 binding resultant from the addition of negative charge at amino acid 24 (T24D) suggests that phosphorylation may play a negative-regulatory role.

To assess the impact of T24 on Arp2/3 actin nucleation, recombinant human WT, ΔDDW, and T24A cortactin proteins were expressed in bacteria and purified (Fig. 1C). When evaluated in pyrene-labeled actin assembly assays, WT cortactin displayed slower polymerization kinetics compared with the N-WASp VCA domain, whereas the ΔDDW mutant failed to activate Arp2/3 as previously reported (Fig. 1D; refs. 16, 34, 35). T24A cortactin demonstrated intermediate activity, with reduced nucleation levels compared with WT cortactin and increased nucleation compared with ΔDDW (Fig. 1D). Taken together these data identify T24 in the cortactin NTA as a critical residue required for optimal cortactin-mediated Arp2/3 binding and activation regardless of phosphorylation status.

Cortactin is essential for initiating invadopodia formation, maturation, and ECM degradation in part due to NTA-mediated Arp2/3 binding (36, 37). To determine the role of cortactin T24 in invadopodia function, a panel of cortactin knockdown-rescue cell lines stably expressing FLAG-cortactin mutant constructs were produced in invasive UMSCC1 (Fig. 2) and OSC19 (Supplementary Fig. S1) HNSCC cell lines. Both lines spontaneously produce invadopodia and degrade ECM (26, 38). Although individual cortactin siRNA (siCTTN) and shRNA (shCTTN) treatment resulted in decreased cortactin expression and matrix degradation in each case (Fig. 2D; Supplementary Figs. S1B and S2), sequential exposure to cortactin siRNA in stable shRNA cells resulted in efficient and reliable cortactin knockdown (KD; Fig. 2; Supplementary Figs. S1 and S2). Cortactin KD cells were used for subsequent experimentation to minimize the possibility of residual endogenous cortactin masking the effects of re-expressed FLAG-cortactin mutants. FLAG-WT cortactin expression in KD cells partially restored the amount of active invadopodia formation in UMSCC1 cells (Fig. 2A–C) and fully restored ECM degradation in UMSCC1 (Fig. 2A and B) and OSC19 (Supplementary Fig. S1A–S1C) cell lines. FLAG-ΔDDW enhanced invadopodia precursor formation but failed to rescue active invadopodia and ECM degradation (Fig. 2A–C; Supplementary Fig. S1A–S1C). Similarly, both FLAG-T24A and FLAG-T24D cortactin restored invadopodia precursor formation while failing to induce invadopodia maturation above KD levels, with active invadopodia and ECM degradation levels for both mutants similar to that of FLAG-ΔDDW cortactin (Fig. 2A–C; Supplementary Fig. S1A, S1C, and S1D). These results suggest that Arp2/3 binding and activation facilitated by cortactin T24 is required for effective cortactin-mediated invadopodia formation and ECM degradation in HNSCC cells.

The importance of T24 in Arp2/3 activation and invadopodia function, along with prior identification of T24 as a cortactin phosphorylation site, led us to identify the kinase(s) responsible for phosphorylating T24. Computational analysis of the sequences flanking T24 was performed by seven independent predictive algorithms, six of which suggested that CK2α had the highest probability of phosphorylating cortactin T24 (Supplementary Table S1). To test this, kinase assays were conducted using GST-tagged cortactin WT and T24A NTA fusion proteins with purified active CK2α. The N-WASp VCA domain was used as a positive control, because previous studies have shown this region to be a CK2α substrate (23, 25). Increasing amounts of GST-WT-NTA were efficiently phosphorylated by CK2α, whereas no phosphorylation was evident in GST-T24A-NTA (Fig. 3A). These data indicate that cortactin T24 can serve as a CK2α substrate, and that T24 is the only residue targeted by CK2α in the NTA region.

To determine if CK2α phosphorylation of cortactin T24 effects binding to Arp2/3 complex, recombinant human WT or T24A cortactin proteins were pre-incubated with or without CK2α, then mixed with purified Arp2/3. Phosphorylation of WT cortactin by CK2α reduced binding of Arp2/3 complex to background levels (beads alone) whereas no impact on T24A was observed (Fig. 3B). To ascertain the impact of CK2α phosphorylation on cortactin-mediated Arp2/3 activation, actin assembly assays were conducted with CK2α-phosphorylated cortactin and N-WASp VCA domain. As previously determined, CK2α phosphorylation of N-WASp VCA results in a modest reduction of Arp2/3 NPF activity (black vs. grey, Fig. 3C; ref. 25). Similarly, WT cortactin incubated with increasing amounts of CK2α prior to inclusion in polymerization assays resulted in a dose-dependent suppression of actin assembly, suggesting that CK2α phosphorylation impairs the ability of cortactin to activate Arp2/3 complex (Fig. 3C). Although Arp2/3 can be activated by direct phosphorylation from multiple kinases (39–41), CK2α had no direct effect on Arp2/3 activation (orange vs. pale green, Fig. 3C), suggesting that the inhibitory effect on Arp2/3 activity is due to phosphorylated cortactin in these assays. To determine if the CK2α-targeted T24 residue is responsible for the observed inhibitory effect on Arp2/3 activity, polymerization assays were conducted with WT and T24A cortactin proteins following incubation with CK2α. Although CK2α inhibited the ability of WT cortactin to activate Arp2/3 complex (blue vs. red, Fig. 3D), preincubation of CK2α with T24A cortactin exhibited no additional inhibitory effect on Arp2/3 nucleation activity (green vs. purple, Fig. 3D). Collectively these data indicate that cortactin phosphorylation at T24 by CK2α reduces the ability of cortactin to bind and activate Arp2/3 complex-mediated branched actin network formation.

To determine if CK2α impacts invadopodia function, CK2α expression was stably knocked down in OSC19 and UMSCC1 cells using anti-CK2α shRNAs targeting two different regions in the CK2α transcript. Both shRNAs reduced CK2α expression to nondetectable levels in each cell line (Fig. 4A). Neither line expressed the alternative CK2α' isoform in control or CK2α knockdown cells (data not shown). CK2α knockdown in OSC19 cells reduced the level of active invadopodia by 71% to 88% and ECM degradation by 63% to 73%, whereas knockdown in UMSCC1 cells reduced active invadopodia formation by 68% to 92% and ECM degradation by 36% to 60% (Fig. 4B–D). Although these data indicate that CK2α is a key mediator of invadopodia maturation and function, neither invadopodia formation nor matrix degradation was entirely abolished. This would indicate that alternative signaling pathways impinging on cortactin and other invadopodia proteins remain active in the absence of CK2α expression.

Silmitasertib (CX-4945) is an orally bioavailable small molecule ATP-competitive inhibitor that targets CK2α kinase activity and is currently undergoing clinical trials in multiple cancer types (NCT01199718, NCT02128282, NCT00891280; ref. 20). To determine the impact of Silmitasertib on HNSCC tumor cell-mediated invadopodia formation and ECM degradation, established HNSCC cell lines were treated with increasing Silmitasertib concentrations and evaluated for effects on invadopodia activity and ECM degradation. Dose-dependent decreases in ECM degradation were observed at concentrations above 0.5 μmol/L in all evaluated HNSCC lines (Fig. 5A and B). The greatest impairment of gelatin degradation was seen at 10 μmol/L (Fig. 5A and B), comparable to effective CK2-specific growth-inhibitory doses in several cancer cell lines (20, 42). At this concentration, active invadopodia formation was significantly diminished in OSC19 and UMSCC1 cells, whereas MDA1586 cells displayed nonsignificant decreases (Fig. 5C). To determine if the invadopodia inhibitory effect of Silmitasertib was directly due to altering CK2α-mediated cortactin T24 phosphorylation, antibodies against phosphorylated cortactin T24 peptides (pT24) were designed and purified. Attempts by two different commercial vendors failed to generate a pT24-specific antibody (not shown). Therefore, we evaluated the phosphorylation status of S473 (pS473) in AKT following Silmitasertib treatment as a surrogate marker for drug efficacy, as this site is known to be phosphorylated by CK2α (20). MDA1586 cells treated with 1 or 10 μmol/L Silmitasertib had decreased pS473 AKT after 24 hours, the same timeframe used in ECM degradation assays (Supplementary Fig. S3). OSC19 and UMSCC1 cells had decreased pS473 AKT after treatment with 10 μmol/L Silmitasertib for 12 hours, the time used for matrix degradation assays in these lines (Supplementary Fig. S3).

The fact that CK2α phosphorylates multiple targets aside from cortactin (20) raises the possibility that the inhibitory effect of Silmitasertib may be due to impairing phosphorylation of additional proteins involved in invadopodia function. To determine the extent of cortactin-specific CK2α phosphorylation in invadopodia-mediated ECM degradation, UMSCC1 cells expressing FLAG-WT, FLAG-T24A, and FLAG-T24D cortactin were treated with 1 μmol/L Silmitasertib and evaluated for additional suppressive effects on matrix degradation (Fig. 5D and E). Silmitasertib diminished ECM degradation in cells expressing FLAG-WT cortactin by 51%, similar to the reduction observed in nontransfected cells (Fig. 5D vs. B). Neither FLAG-T24A nor FLAG-T24D expressing cells treated with Silmitasertib demonstrated reductions in ECM degradation levels from baseline vehicle-treated controls (Fig. 5D and E). Although these results do not entirely negate alternative CK2-dependent signaling pathways in invadopodia regulation, it does suggest that cortactin T24 is the primary CK2α target in governing HNSCC invadopodia function.

To evaluate the anti-invadopodia effect of Silmitasertib in more translationally-relevant models, PDXs were derived from a well- and a moderately-differentiated HNSCC surgical sample. PDX tumors maintained original patient tumor architecture, displaying collective invasion and keratin pearls characteristic of differentiated HNSCC (Fig. 6A). Primary cell lines derived from these PDX tumors form invadopodia and spontaneously degrade gelatin within 24 hours (Fig. 6B). Both lines exhibit tight colony morphology under cell culture conditions consistent with HNSCC lines derived from epithelial HNSCC (Fig. 6B; refs. 32, 43). Treatment of WVUSCC-AR2 and WVUSCC-AR5 with Silmitasertib yielded similar results to those observed in established lines, with gelatin degradation impaired 38% to 62% in WVUSCC-AR2 cells and 56% to 66% in WVUSCC-AR5 cells at and above 0.5 μmol/L (Fig. 6C and D). Similarly, 10 μmol/L Silmitasertib treatment suppressed active invadopodia by 76% in WVUSCC-AR2 and 91% in WVUSCC-AR5 (Fig. 6E). Collectively these data indicate that CK2α kinase activity is essential for maximal ECM degradation ability in HNSCC.

To determine if Silmitasertib impacts HNSCC invasion, we initially utilized a 3D in vitro assay designed to model collective invasion typically observed in differentiated HNSCC. Tumor spheroids generated from OSC19, UMSCC1, and WVUSCC-AR5 PDX lines were embedded between layers of collagen I. WVUSCC-AR2 cells failed to form spheroids and could not be used in this assay. Spheroids were treated with 10 μmol/L Silmitasertib or vehicle (DMSO) for 48 hours (Fig. 7A and B). Silmitasertib significantly reduced 3D collective invasion in all assayed lines (Fig. 7B). We next evaluated the ability of Silmitasertib to control invasion in the tongues of mice harboring orthotopic tumors. Luciferase-expressing UMSCC1 cells injected into the tongues of NSG mice formed detectable tumors within 1 week (Fig. 7C). Mice were divided into two groups containing similar tumor size as determined by in vivo bioluminescence, with one group treated with vehicle and one with Silmitasertib. Mice dosed twice daily for 3 weeks displayed a nonsignificant reduction in tumor growth compared with controls, similar to previous single agent xenograft studies (Fig. 7C; ref. 20). To negate potential bias due to unequal tumor size, four equivalent tumors from each group were selected for further assessment. In-depth evaluation of tumor margins from tongues excised after 4 weeks revealed alterations in invasive characteristics of Silmitasertib-treated mice (Fig. 7D–I). Tumors in drug-treated mice had a less ragged appearance at the invasive front, exhibiting shorter and smaller collective cell protrusions into the tongue stroma compared with control mice (Fig. 7D–G). In addition, tongues from Silmitasertib-treated mice had reduced perineural invasion of nerves adjacent to the invasive front (Fig. 7H and I). No difference was seen in the size or invasive distance of detached collective groups within the invasive front (Supplementary Fig. S4). Collectively these data support a role for CK2α signaling in driving several pro-invasive behaviors associated with poor patient outcome in HNSCC (2, 44).

Proteolysis of restrictive tissue barriers is essential to all steps in the metastatic cascade and increasing evidence indicates that the proteolytic activity of invadopodia is required for invasive breaching of ECM barriers (1). Signals that govern invadopodia are dynamic and highly regulated, requiring coordinated activity of several oncogenic pathways in parallel to achieve maximal efficiency (1). Active assembly and turnover of cellular F-actin networks is required for initiating invadopodia formation and subsequent maturation, involving recruitment and activation of membrane-bound and secreted MMPs to mediate ECM proteolysis (1). Breakdown of existing F-actin networks, in conjunction with Arp2/3 activation in invadopodia, is responsible for actin network turnover necessary for productive branched F-actin formation that drives membrane protrusion. Early recruitment of cortactin is necessary for invadopodia initiation, where phosphorylation of C-terminal tyrosines 421, 470, and 486, along with serines 405 and 418, occurs downstream of growth factor and integrin signaling (45). These phosphorylation events create binding sites for scaffolding platforms that recruit N-WASp and WAVE2, ensuring that activation of Arp2/3 is maintained throughout the invadopodia cycle (6). Cortactin also recruits the F-actin severing protein cofilin to invadopodia, providing the necessary machinery for cyclic actin network breakdown and regrowth during invadopod extension (16). In addition to these important C-terminal functions, the ability of the NTA region to directly bind Arp2/3 is also essential for invadopodia formation and ECM degradation, presumably through direct Arp2/3 activation and prolonged stabilization of Arp2/3-F-actin networks (10, 16).

Regarding the ability of the cortactin NTA to bind Arp2/3, this study reveals two distinct findings. First, that T24 is required for Arp2/3 binding, because mutation of this residue ablates (A) or reduces (D) the association of cortactin with Arp2/3 in co-immunoprecipitation assays. T24 is two residues C-terminal to the well-defined DDW Arp2/3 interaction motif that is conserved in other Arp2/3 NPFs (6). This places T24 in close proximity to contribute to Arp2/3 binding. Although the crystal structure of the cortactin NTA bound to Arp2/3 complex has not been reported, previous chemical crosslinking and three-dimensional reconstruction studies indicate that the NTA region primarily binds the Arp3 subunit in the complex, with the most N-terminal residues spanning the Arp2–Arp3 interface to contact the Arp2 subunit (Visual Overview; refs. 9, 34). Thus, T24 may contribute essential hydrogen bonding with recipient polar residues near the basic side chains on Arp3 involved in electrostatic binding to the acidic NTA residues. This concept is supported by the requirement of W22 in the DDW motif for Arp2/3 binding, along with the lack of an equivalent threonine in the VCA region of WASp proteins (16, 34, 35). Although T24A cortactin appears incapable of binding Arp2/3, it can activate Arp2/3 actin nucleation in conditions where ΔDDW cortactin remains inactive (Fig. 1D). This may be due to T24 having a lower affinity for Arp3 than the DDW region and/or the high cortactin molar levels required to stimulate NPF activity in vitro. Nonetheless, the requirement for cortactin T24 in invadopodia-mediated ECM degradation supports an essential biological role for this residue in a non-phosphorylated context.

Second, our study shows that T24 phosphorylation serves to negatively regulate cortactin binding to and activation of Arp2/3 complex (Fig. 3B–D). Addition of the phosphate group to T24 likely imparts a steric and electrostatic disruption, preventing the DDW and other interacting residues in the NTA from initiating and/or maintaining binding to Arp3. Steric interference may be the predominant mechanism, since T24D cortactin can bind Arp2/3 at reduced levels (Fig. 1B), indicating that negative charge alone is insufficient to completely prevent Arp2/3 association. Phosphorylation of T24 therefore serves to block the ability of cortactin to bind Arp2/3, similar to the regulation of the Arp2/3-F-actin regulatory proteins coronin 1B and cofilin (46, 47). Although we did not determine if phosphorylation of T24 serves to release cortactin from existing Arp2/3-F-actin branchpoints, phosphorylation of Arp2/3-bound cortactin T24 may be an additional mechanism to dissociate cortactin from these regions. Such a mechanism could function in facilitating breakdown and recycling of invadopodia Arp2/3-F-actin networks.

The ability of CK2α to phosphorylate cortactin T24 supports a wider role for this kinase in actin regulation. Previous work identified T24 as a putative but unconfirmed CK2α phosphorylation site in the NTA region of myeloid-cell specific cortactin homolog HS1 (21). In addition, CK2α phosphorylates two adjacent sites in the A domain of WASp family proteins (22–25). These studies report conflicting results on Arp2/3 activity, with phosphorylation of WASp enhancing and in N-WASp and WAVE2 inhibiting VCA-stimulated Arp2/3-mediated actin polymerization. Our results with CK2α-phosphorylated N-WASp VCA domain also shows an inhibitory effect on Arp2/3 activation (Fig. 3C), suggesting that CK2α effects may be differential to specific WASp family proteins.

Cortactin T24A and T24D expression in KD HNSCC lines impairs active invadopodia formation and subsequent ECM degradation (Fig. 2A–C; Supplementary Fig. S1A and S1C), supporting in vitro evidence that T24 phosphorylation prevents Arp2/3 binding and activation necessary for invadopodia activity. The importance of CK2α-mediated cortactin T24 phosphorylation is further supported by the inability of Silmitasertib to further suppress ECM degradation in cortactin T24A or T24D expressing cells (Fig. 5D and E). CK2α-specific knockdown or pharmacologic inhibition in cells would therefore be expected to reduce phosphorylation of cortactin and WASp family proteins, resulting in net increased Arp2/3 activity as observed from in vitro studies. The fact that CK2α knockdown or kinase blockade also impairs invadopodia formation and function is paradoxical, but can be explained by cyclic CK2α phosphorylation and dephosphorylation to regulate cortactin binding to and activation of Arp2/3 in invadopodia during maturation and elongation. Dephosphorylation of cortactin T24 and WASp NPFs by unknown phosphatases would enable these NPFs to promote Arp2/3 activation, actin polymerization, and invadopodia activity. Following network breakdown during actin turnover, CK2α phosphorylation would return NPFs to the inactive state to await the next cycle of actin assembly. Although similar cyclic regulation of Src and cofilin has been shown to be essential for invadopodia function and provides support for this model (38, 46), confirmation of T24 phosphorylation within invadopodia leading to altered actin dynamics along with identification of the T24 phosphatase are necessary to fully confirm this proposed mechanism.

Locoregional control of HNSCC dissemination is problematic, where perineural invasion and invasive metastatic spread to cervical lymph nodes accelerates patient decline (44, 48, 49). Cortactin and CK2α expression are elevated in HNSCC and individually correspond with poorer patient outcomes (7, 20). CK2 knockdown and Silmitasertib treatment in HNSCC cells has antiproliferative and antimetastatic properties in vitro and in mice, providing support for the utility of Silmitasertib in HNSCC (20). The anti-invasive effect of Silmitasertib shown here is likely due in part to disruption of the invadopodia actin assembly through combined inhibition of multiple Arp2/3 NPFs. We note that T24 cortactin mutants, CK2α knockdown, and Silmitasertib treatment in HNSCC cells do not completely abolish invadopodia formation, degradation activity or invasion. Activation of invadopodial Arp2/3 NPFs through alternative scaffolding or lipid-based signaling pathways that can bypass CK2 inhibition can account for this residual activity (1). However, our results indicate that CK2α inhibition by a single agent does significantly impair collective HNSCC invasion. Given the current paucity of treatment options for invasive HNSCC, combining Silmitasertib with additional drugs known to impair invadopodia function by blocking additional invadopodia and motility pathways (8) has the potential to provide novel treatment options for controlling invasive spread of late-stage HNSCC harboring elevated cortactin and/or CK2 expression.

S.A. Weed reports receiving a commercial research grant from Gilead Sciences. No potential conflicts of interest were disclosed by the other authors.

This laboratory was supported in part by a grant from Gilead Sciences during the period when the work was conducted. Gilead grant funds were not used to support any aspect of the project reported in this manuscript. Gilead Sciences had no input on the findings disclosed herein.

Conception and design: S.M. Markwell, E.T. Interval, S.A. Weed

Development of methodology: S.M. Markwell, E.T. Interval, D.A. Schafer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.M. Markwell, A.G. Ammer, E.T. Interval, J.L. Allen, B.W. Papenberg, R.A. Hames, J.E. Castaño, D.A. Schafer, S.A. Weed

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.M. Markwell, A.G. Ammer, E.T. Interval, B.W. Papenberg, S.A. Weed

Writing, review, and/or revision of the manuscript: S.M. Markwell, E.T. Interval, S.A. Weed

Study supervision: S.A. Weed

Other (figure design): B.W. Papenberg

Other (I provided reagents for the actin assembly assays and guidance to Dr. Markwell as he performed the assays in my lab.): D.A. Schafer

The authors thank Jianjiong Gao (University of Missouri) for sequence prediction analysis. The assistance of the West Virginia University Microscope Imaging Facility, Animal Models and Imaging Facility, and Biochemistry Protein Core are gratefully acknowledged.

This work was supported by pilot grants from NIH U54GM104942, P20RR016440, and P20GM103434 (to S.M. Markwell, A.G. Ammer, J.L. Allen, B.W. Papenberg, R.A. Hames, and S.A. Weed). Additional support was provided by the West Virginia University Department of Neurobiology and Anatomy (to S.M. Markwell, S.A. Weed), Department of Biochemistry (to S.M. Markwell, S.A. Weed), Department of Otolaryngology, Head and Neck Surgery (to E.T. Interval, J.E. Castaño), and the Dorothy D. Radford Endowed Fund of the West Virginia University Cancer Institute (to S.A. Weed). Instrumentation support was provided by NIH grants U54GM104942, P20RR016440, P30RR032138/GM103488, P20GM103434, and P20RR016477. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.