Cdk5-mediated phosphorylation regulates phosphatidylinositol 4-phosphate 5-kinase type I γ 90 activity and cell invasion

Abstract

Phosphatidylinositol 4-phosphate 5-kinase type I γ (PIPKIγ90) regulates cell migration, invasion, and metastasis. However, it is unknown how cellular signals regulate those processes. Here, we show that cyclin-dependent kinase 5 (Cdk5), a protein kinase that regulates cell migration and invasion, phosphorylates PIPKIγ90 at S453, and that Cdk5-mediated PIPKIγ90 phosphorylation is essential for cell invasion. Moreover, Cdk5-mediated phosphorylation down-regulates the activity of PIPKIγ90 and the secretion of fibronectin, an extracellular matrix protein that regulates cell migration and invasion. Furthermore, inhibition of PIPKIγ activity with the chemical inhibitor UNC3230 suppresses fibronectin secretion in a dose-dependent manner, whereas depletion of Cdk5 enhances fibronectin secretion. With total internal reflection fluorescence microscopy, we found that secreted fibronectin appears as round dots, which colocalize with Tks5 and CD9 but not with Zyxin. These data suggest that Cdk5-mediated PIPKIγ90 phosphorylation regulates cell invasion by controlling PIPKIγ90 activity and fibronectin secretion.—Li, L., Kołodziej, T., Jafari, N., Chen, J., Zhu, H., Rajfur, Z., Huang, C. Cdk5-mediated phosphorylation regulates phosphatidylinositol 4-phosphate 5-kinase type I γ 90 activity and cell invasion.

Fibronectin is a large, extracellular matrix protein that binds to integrins as well as other extracellular matrix proteins, such as collagen and fibrin. Fibronectin has a key role in cell adhesion, migration, invasion, and metastasis (1–4). It has been shown that cortactin-mediated lysosomal and exosomal secretion contributes to fibronectin secretion (5, 6). However, the signaling pathways that regulate fibronectin secretion remain to be identified.

Phosphatidylinositol 4 phosphate 5-kinase type I γ (PIPKIγ90) catalyzes ATP-dependent phosphorylation of phosphatidylinositol 4-phosphate to generate phosphatidylinositol 4,5-bisphosphate (PIP₂), which functions as a vesicle docking site, thus regulating vesicle exocytosis (7, 8). It is also a precursor of several lipid second messengers, such as phosphatidylinositol 3,4,5-triphosphate, inositol 1,4,5-triphosphate, and diacylglycerol. PIP₂ also regulates focal adhesion (FA) assembly and actin polymerization by modulating the function of a number of FA and cytoskeletal proteins, such as talin, vinculin, neural Wiskott-Aldrich syndrome protein (N-WASP), gelsolin, and profilin (9–13). In addition, PIP₂ binds to a number of endocytic proteins, including the α and µ2 subunits of AP2, AP180/clathrin assembly lymphoid myeloid leukemia protein, Dab2, epsin, and dynamin (14–19), thus regulating clathrin-mediated endocytosis, where it interacts with exocyst and soluble N-ethylmaleimide-sensitive factor attachment receptor complexes to mediate vesicle exocytosis (8, 20, 21).

PIPKIγ90 is essential for cell migration, invasion, and metastasis. Depletion of PIPKIγ90 inhibits growth factor–stimulated cell migration in the MDA-MB-231 breast cancer cell line and HeLa cervical cancer cells (22, 23). PIPKIγ90 knockdown also blocks the invasion of breast cancer and colon cancer cells (24, 25). Furthermore, PIPKIγ90-depleted 4T1 breast cancer cells show significant reduction in tumor progression and metastasis (26). PIPKIγ90 also regulates neutrophil migration by controlling cell polarity as well as rear retraction (27–29) and is required for FA assembly and disassembly, key steps in cell migration (24).

PIPKIγ90 is regulated by different posttranscriptional modifications. We demonstrated that PIPKIγ90 ubiquitination at K97 by HECTD1, an E3 ubiquitin ligase that regulates cell migration, results in PIPKIγ90 degradation, thus controlling dynamic PIP₂ production to mediate FA assembly and disassembly, cell migration, invasion, and metastasis (25). It is also reported that Smurf1 ubiquitinates PIPKIγ90 at K255, causing PIPKIγ90 degradation, consequently, suppressing lung tumor growth. Ribosomal protein S6 kinase β1 (also called p70S6K1 or S6K1), one of the mechanistic target of rapamycin pathway effectors, phosphorylates PIPKIγ90 at T553 and S555, controlling PIPKIγ90 degradation and regulating the development of FAs and invadopodia. Cyclin-dependent kinase 5 (Cdk5) phosphorylates PIPKIγ90 at S650, inhibiting the PIPKIγ90-talin interaction (30), whereas Src phosphorylates PIPKIγ90 at Y644 enhancing its binding to talin and reducing talin-β integrin interaction (31). In addition, phosphorylation of PIPKIγ90 by EGFR at Y639 influences tumor cell migration and metastasis (26).

It has been reported that PIPKIγ90 regulates matrix metalloproteinase secretion and invadopodium formation (26, 32), key steps in cancer cell invasion and metastasis. These findings are consistent with the role of PIP₂ in exocytosis. Because invadopodia mediate exosomal secretion of fibronectin (6), it is likely that PIPKIγ90 also regulates fibronectin secretion.

Recent studies have shown that Cdk5, a protein kinase that regulates cell migration, invasion, and cancer metastasis, is essential for invadopodium formation (33). Cdk5 is also a key regulator of exocytosis (34). In the present study, we demonstrated that Cdk5 phosphorylated PIPKIγ90 at S453, a previously unidentified phosphorylation site, and that Cdk5-mediated PIPKIγ90 phosphorylation is essential for cell invasion. Moreover, Cdk5-mediated phosphorylation inhibited PIPKIγ90 activity and fibronectin secretion. These results suggest that Cdk5-mediated PIPKIγ90 phosphorylation spatiotemporally controls PIP₂ production and fibronectin secretion to regulate cancer cell invasion.

MATERIALS AND METHODS

Reagents

IgG-agarose was used as previously described by Huang et al. (35). Anti-PIPKIγ90 pAb (MAO-R1) was from Abcam (Cambridge, United Kingdom). Anti-CD9 rabbit pAb (C9993), anti-Flag M2 Agarose beads, anti-tubulin antibody, pLKO1 lentivirus short hairpin RNAs (shRNAs) that target PIPKIγ90, and Cdk5 were from MilliporeSigma (Burlington, MA, USA); PIPKIγ90 shRNA clone was TRCN0000037668 (A1); Cdk5 shRNA clones were TRCN0000021465, TRCN0000021466, and TRCN0000021467. Anti-Tks5 (SH3 #4) rabbit pAb was from MilliporeSigma. Anti-fibronectin mouse mAb (MA1116) was from Boster Biological Technology (Pleasanton, CA, USA). pCMV-p35 was a gift from Dr. Li-Huei Tsai (Harvard Medical School, Boston, MA, USA). pEF4-myc-His-Cdk5 and pEF4-myc-His-Cdk5-144N have been previously described (36). Alexa Fluor 488 goat anti-mouse IgG [heavy and light chain (H+L); A11001], Alexa Fluor 555 F(ab′) 2 fragment of goat anti-rabbit IgG (H+L; A21430), Alexa Fluor 680 rabbit anti-goat IgG (H+L; A27020), and Alexa Fluor 700 goat anti-rabbit IgG (H+L; A21038) were from Thermo Fisher Scientific (Waltham, MA, USA). DyLight 800 conjugated goat anti-mouse IgG (H+L) was from Thermo Fisher Scientific. Fibronectin was from Akron Biotech (Boca Raton, FL, USA). Epidermal growth factor (EGF) and stem cell factor (SCF) were from ProSpec (Rehovot, Israel). Hepatocyte growth factor (HGF) and TNF-α were from Sino Biological (Beijing, China). Growth factor–reduced Matrigel was from BD Biosciences (San Jose, CA, USA). Pfu Ultra was from Agilent Technologies (Santa Clara, CA, USA). Safectine RU50 transfection kit was purchased from Syd Labs (Natick, MA, USA). Anti-phospho-PIPKIγ (pS453) rabbit pAb (57A1) was custom-made by Syd Labs by immunizing rabbits with phospho-peptide (C) LKSS(p)PSKKGR conjugated to keyhole limpet hemocyanin. DNA primers were synthesized by MilliporeSigma.

Plasmid construction

pZZ-PIPKIγ90 was previously described (25, 35). Wild-type pFLAG-PIPKIγ90 (pFLAG-PIPKIγ90^WT) and the codon-modified plasmids pFLAG-PIPKIγ90^WT and pBabe-FLAG-PIPKIγ90^WT were previously described (32). The codon-modified plasmids pFLAG-PIPKIγ90^S453A and -PIPKIγ90^S453E were generated by pfu Ultra-based PCR with the codon-modified pFLAG-PIPKIγ90 as a template and 5′‑TCCTCCCTGAAGTCCGCGCCCTCCAAGAAG‑3′, 5′‑CTTCTTGGAGGGCGCGGACTTCAGGGAGGA‑3′, 5′‑TCCTCCCTGAAGTCCGAGCCCTCCAAGAAG‑3′, and 5′‑CTTCTTGGAGGGCTGGACTTCAGGGAGGA‑3′as primers, respectively. The codon-modified pBabe-FLAG-PIPKIγ90^S453A and pBabe-FLAG-PIPKIγ90^S453E were made by sequentially digesting the codon-modified pFLAG-PIPKIγ90^S453A and -PIPKIγ90^S453E with Age1, blunting with Klenow, and digesting with Sal1. The smaller fragments were subcloned into pBabe-neo vector, which had been treated with BamHI, Klenow, and Sal1. pGEX-4T-3-PIPKIγ90_501–668, was previously described (32). pGEX-4T-3-PIPKIγ90_501–668^S453A and -PIPKIγ90_501–668^S453E were constructed by PCR amplifying the fragments encoding residues 501–668 using primers 5′‑AATTTGGATCCGAGGACGAAGGCCGGCC‑3′ and 5′‑ATATATGAATTCTTATGTGTCGCTCTCGCCGTCGGA‑3′ and templates pFlag-PIPKIγ90^S453A and -PIPKIγ90^S453E, respectively. The PCR products were digested with BamHI and EcoRI and were inserted into the pGEX-4T-3 vector that had been digested with the same enzymes. All plasmids were sequenced by MWG Operon (Eurofins Genomics, Louisville, Kentucky, USA).

Cell culture and transfection

Chinese hamster ovary (CHO) K1 cells, MDA-MB-231 human breast cancer cells, and 293T human embryonic kidney cells were from the American Type Culture Collection (Manassas, VA, USA) and were maintained in DMEM medium (MilliporeSigma) containing 10% fetal bovine serum (FBS), penicillin (100 U/ml), and streptomycin (100 μg/ml). PC3-MM2 human prostate cancer cells were gifts from Dr. Gary Gallick (M. D. Anderson Cancer Center, Houston, TX, USA) and were maintained in Roswell Park Memorial Institute (RPMI)-1640 medium (MilliporeSigma) containing 10% FBS, penicillin (100 U/ml), and streptomycin (100 μg/ml). CHO-K1 and 293T cells were transfected with Safectine RU50 (Syd Labs), according to the manufacturer’s protocol.

Preparation of viruses and cell infection

The 293T cells were transfected with pBabe retroviral, or pLKO1 lentiviral system using Safectine RU50 transfection reagent, according to the manufacturer’s protocol. The virus particles were applied to overnight cultures of breast or prostate cancer cells for infection. Cells that stably expressed pLKO1 lentiviral shRNAs were obtained by selecting the infected cells with 1 μg/ml puromycin, and cells that were infected with pBabe retroviruses were stabilized by growing infected cells in the presence of 0.7 mg/ml neomycin for 10 d.

PIPKIγ90 phosphorylation

Flag-PIPKIγ90 (or a mutant) was cotransfected with an empty vector or a plasmid expressing active kinase into CHO-K1 cells. The cells were lysed with RIPA buffer [50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% IPEGAL (Rhodia, La Défense, France), 0.5% deoxycholate, 5 mM EDTA] containing protease inhibitor cocktail and phosphatase inhibitor cocktail. Cell lysates were analyzed by SDS-PAGE and transferred to nitrocellulose membrane. Ser453 phosphorylation was detected with an anti-PIPKIγ (pS453) rabbit pAb. To detect Ser453 phosphorylation in breast cancer cells, Flag-PIPKIγ90 and mutants were immunoprecipitated with anti-Flag agarose beads. The immune complexes were analyzed by Western blotting using the anti-PIPKIγ (pS453) antibody.

Invasion assays

Matrigel (100 µL; 1:15 dilution in serum-free DMEM medium) was added to each Transwell polycarbonate filter (6 mm diameter, 8 µm pore size; Corning Life Sciences, Corning, NY, USA) and incubated with the filters at 37°C for 5 h. Cancer cells were trypsinized and washed 3 times with DMEM containing 1% FBS. The cells were resuspended in DMEM containing 1% FBS at a density of 1 × 10⁶ cells/ml. The cell suspensions (100 µl) were seeded into the upper chambers, and 600 µl of DMEM medium containing 50 ng/ml of HGF were added to the lower chambers. The cells were allowed to invade for 15 h (or as indicated) in a CO₂ incubator, fixed, stained and quantitated as described previously (24).

In vitro PIPKIγ90 activity assays

PIPKIγ90 activity was measured as previously described (24). Briefly, pFLAG-PIPKIγ90^WT, pFLAG-PIPKIγ90^S453A, and pFLAG-PIPKIγ90^S453E were transiently expressed in CHO-K1 cells and immunoprecipitated with anti-FLAG agarose beads. The beads were washed and incubated with 100 μl of a kinase buffer containing 100 μM phosphatidylinositol 4-phosphate for 30 min at 37°C. PIP₂ formed in these assays was extracted as previously described (37), and separated by silicon thin-layer chromatography. PIP₂ was visualized by autoradiography and quantitated with a liquid scintillation counter (Beckman Coulter, Sydney, NSW, Australia). To examine the effect of Cdk5 on PIPKIγ90 activity, pZZ-PIPKIγ90 was cotransfected with an empty vector, pEF4-myc-His-Cdk5/pCMV-p35, pEF4-myc-His-Cdk5-144N/pCMV-p35, into CHO-K1 cells and immunoprecipitated with IgG agarose beads (35).

Immunofluorescence staining

The cells were trypsinized and plated on glass-bottom dishes that had been precoated with fibronectin (5 μg/ml). The cells were cultured for 24 h. The cells were fixed with a fixation buffer containing 4% paraformaldehyde and 0.05% glutaraldehyde in PBS for 20 min, permeabilized for 15 min with 0.4 mg/ml digitonin in PBS, and then blocked with 5% bovine serum albumin in PBS for 1 h. The cells were then incubated with a rabbit anti-PIPKIγ pAb and a mouse anti-PIP₂ mAb, washed with PBS, and then incubated with an Alexa Fluor 488–labeled (Thermo Fisher Scientific) goat anti-mouse and an Alexa Fluor 555–labeled (Thermo Fisher Scientific) goat anti-rabbit secondary antibody. After washing with PBS, the images of PIPKIγ and PIP₂ were acquired with an Eclipse Ti total internal reflection fluorescence (TIRF) microscope (Nikon, Tokyo, Japan) equipped with a ×60, 1.45 numerical aperture objective, CoolSnap HQ2 charge-coupled device camera (Roper Scientific, Vianen, The Netherlands). PIP₂ intensity was analyzed with NIS-Elements (Nikon).

Fibronectin secretion assays

Cells were cultured in normal culture medium (10% FBS) to 80% confluence. The cells were washed 3 times with PBS and then cultured in serum-free Opti-MEM (Thermo Fisher Scientific), including 20 ng/ml HGF for 24 h. The media were collected, concentrated with Spin-X ultrafiltration concentrators (10,000 Da cutoff; Corning), and then analyzed for the secretion of fibronectin by Western blotting.

To visualize fibronectin secretion, cells were plated on laminin (10 µg/ml)-coated glass-bottomed dishes and cultured for 36 h. The cells were fixed with 4% paraformaldehyde in PBS for 20 min, permeabilized for 15 min with 0.25% Triton X-100 in PBS, and then blocked with 5% bovine serum albumin in PBS for 1 h. The cells were stained and examined as described in previous section. Fibronectin intensities were quantitated using NIS-Elements.

Gel data quantification

Gel data were quantified by analyzing inverted images using ImageJ (National Institutes of Health, Bethesda, MD, USA) as previously described (32, 36). Data from different experiments were normalized to controls. If values from different experiments had a high variation, data sets were further normalized by dividing the numbers in a data set with a factor (e.g., 2), so that the biggest values from different experiments were similar.

RESULTS

The residues Thr123, Ser453, and Ser650 of human PIPKIγ90 are consensus sites for Cdk5 [(S/T*)PX(R/K/H)] (Fig. 1A). To identify phosphorylation sites in PIPKIγ90, Flag-PIPKIγ90 was cotransfected with Myc-Cdk5/p35 into CHO-K1 cells. Flag-PIPKIγ90 was immunoprecipitated with anti-Flag M2 agarose beads, digested with trypsin, and phosphopeptides were enriched with a phosphopeptide-enriched kit. The peptides were analyzed by mass spectrometry. A MH⁺ ion was observed at m/z 1114.515, corresponding to PIPKIγ90 peptide 447-NSSLKSSPSK-456 with 1 phosphoryl group, suggesting that a phosphorylation site is located on this peptide. Further tandem mass spectrometry analysis of the parent ion revealed the presence of a phosphoryl group at Ser453 because the additional mass of the phosphoryl group occurs at the b7 ion and the y6 ion and all subsequent b and y ions, but not at y2, y3, b2, b3, and b4 ions, which do not contain Ser453 (Fig. 1B). The observation of the loss of 80 of the fragment SSLKS additionally confirms the phosphorylation of Ser453 by Cdk5 in cells. Thr341 and Ser476 were also phosphorylated in Cdk5/p35-transfected cells (Supplemental Fig. S1). However, because these sites are not consensus sites for Cdk5, the following studies were focused on Ser453. Ser453 and the proximal residues are conservative among different species (Fig. 1C). PIPKIγ90 phosphorylation at Ser453 by Cdk5 was confirmed by site-directed mutagenesis. As shown in Fig. 1D, substitution of Ser453 with Ala or Glu inhibited Cdk5-mediated phosphorylation of PIPKIγ90. These data indicated that Cdk5 phosphorylates PIPKIγ90 at Ser 453. To learn whether Cdk5 phosphorylates PIPKIγ90 in CHO-K1 cells, Flag-PIPKIγ90 was cotransfected with an empty vector or Myc-Cdk5/Myc-p35. The cell lysates were analyzed by Western blotting, and PIPKIγ90 phosphorylation was detected with an anti-PIPKIγ (pS453) antibody. Cdk5 significantly promoted Ser453 phosphorylation (Fig. 1E). To examine the phosphorylation of PIPKIγ90^WT, PIPKIγ90^S453A, and PIPKIγ90^S453E in breast cancer cells, FLAG-PIPKIγ90^WT,-PIPKIγ90^S453A, and -PIPKIγ90^S453E were, respectively, expressed in MDA-MB-231 human breast cancer cells by retroviral infection, and the FLAG-tagged proteins were immunoprecipitated with anti-FLAG agarose beads. The phosphorylation of these proteins was detected with the anti-PIPKIγ (pS453) antibody. PIPKIγ90^WT was phosphorylated, whereas mutation at Ser453 abolished its phosphorylation (Fig. 1F). Because HGF and EGF stimulate PIPKIγ90 phosphorylation at Thr553 and Ser555 (32), we tested whether HGF and EGF also stimulated PIPKIγ90 phosphorylation at Ser453, using SCF and TNF-α as controls. MDA-MB-231 cells stably expressing FLAG-PIPKIγ90 were serum starved and were stimulated with EGF, HGF, SCF, and TNF-α. Flag-PIPKIγ90 was immunoprecipitated with anti-FLAG-agarose beads, and PIPKIγ90 phosphorylation was detected with the anti-PIPKIγ (pS453) antibody. HGF stimulated Ser453 phosphorylation, whereas other growth factors and TNF-α had little effect (Fig. 1G and Supplemental Fig. S2). HGF stimulated Ser453 phosphorylation in a time-dependent manner, peaking at 90 min after HGF exposure (Fig. 1H). HGF-stimulated Ser453 phosphorylation was blocked by roscovitine, a Cdk5 inhibitor (Fig. 1I). Furthermore, depletion of Cdk5 significantly inhibited HGF-induced Ser453 phosphorylation (Fig. 1J). These results suggest that Cdk5 phosphorylates PIPKIγ90 at Ser453.

Figure 1.

Cdk5 phosphorylates PIPKIγ90 at S453. A) The Cdk5 consensus sequences in human PIPKIγ90 are shown. B) Tandem mass spectrometry analysis of the phosphopeptide 447-NSSLKSSPSK-456 of human PIPKIγ90 is shown. FLAG-PIPKIγ90 was cotransfected with Cdk5/p35 into CHO-K1 cells, immunoprecipitated, digested with trypsin, and enriched with a TiO2 phosphopeptide and a clean-up kit. C) Alignment of the Cdk5 consensus sequences around S453 from different species is shown. D) Cdk5 phosphorylated PIPKIγ at S453 in vitro. Recombinant GST-PIPKIγ90_158–600, -PIPKIγ90_158–600^S453A, and -PIPKIγ90_158–600^S453E were phosphorylated with active Cdk5/p35. E) Transfection of Cdk5/p35 promoted PIPKIγ90 phosphorylation at S453. Flag-PIPKIγ90 was cotransfected with an empty vector or Cdk5/p35 into CHO-K1 cells. Cell lysates were analyzed by Western blotting, and PIPKIγ90 phosphorylation was detected with an anti-PIPKIγ (pS453) antibody. F) PIPKIγ-depleted MDA-MB-231 cells were infected with retroviruses to reexpress FLAG-PIPKIγ90^WT, -PIPKIγ90^S453A, and -PIPKIγ90^S453E. FLAG-tagged proteins were immunoprecipitated, and their phosphorylation was detected with an anti- PIPKIγ (pS453) antibody. G) MDA-MB-231 cells stably expressing FLAG-PIPKIγ90 were serum-starved and stimulated with HGF, EGF, TNF-α, and SCF at 50 ng/ml for 90 min. PIPKIγ90 was immunoprecipitated, and its phosphorylation was detected with anti-PIPKIγ (pS453) antibody. Data are presented as means ± sem of 3 independent experiments. **P < 0.01 vs. control. H) MDA-MB-231 cells stably expressing FLAG-PIPKIγ90 were serum-starved and stimulated with HGF (50 ng/ml) for different times. Data are presented as means ± sem of 3 independent experiments. *P < 0.05, ***P < 0.001 vs. control. I) MDA-MB-231 cells stably expressing FLAG-PIPKIγ90 were serum-starved, treated with Cdk5 inhibitor roscovitine (10 M), and stimulated with HGF (50 ng/ml) for 90 min. Data are presented as means ± sem of 3 independent experiments. *P < 0.05, **P < 0.01. J) HGF-stimulated PIPKIγ90 phosphorylation was suppressed by depletion of Cdk5. Cdk5-depleted MDA-MB-231 cells were infected with retroviruses to express FLAG-PIPKIγ90. Resulting cells were serum-starved and then stimulated with HGF (50 ng/ml) for 20 min. Data are presented as means ± sem of 3 independent experiments. *P < 0.05.

We previously reported (25) that depletion of PIPKIγ90 using shRNA inhibited the invasion of MDA-MB-231 cells and that reexpression of PIPKIγ90 restored the invasion of PIPKIγ90-depleted cells. To assess the potential role of PIPKIγ90 phosphorylation in cancer cell invasion, the Matrigel-invasive capabilities of PIPKIγ90-depleted MDA-MB-231 cells that express FLAG-PIPKIγ90^WT, FLAG-PIPKIγ90^S453A, or FLAG-PIPKIγ90^S453E, were measured. Reexpression of PIPKIγ90^WT in PIPKIγ90-depleted cells restored cell invasion to an extent comparable to the invasion of cells expressing empty pLKO.1 vector. In contrast, reexpression of -PIPKIγ90^S453A or PIPKIγ90^S453E in PIPKIγ-depleted cells did not rescue their invasion (Fig. 2A). To examine whether that pathway also regulates the invasion of other cancer cell line, we examined the effects of phosphorylation site mutants PIPKIγ90^S453A and PIPKIγ90^S453E on the invasion of PC-3MM2 prostate cancer cells. Reexpression of PIPKIγ90^WT in PIPKIγ90-depleted cells rescued the cell invasive capacities, whereas that of PIPKIγ90^S453A and PIPKIγ90^S453E did not (Fig. 2B). UNC3230, a specific inhibitor of PIPKIγ, also inhibited the invasion of PC3-MM2 cells (Supplemental Fig. S3). Depletion of Cdk5 using shRNAs inhibited the invasion of MDA-MB-231 cells (Fig. 2C). Furthermore, expression of PIPKIγ90 in Cdk5-depleted cells had an additional inhibition on cell invasion (Fig. 2D). These results suggest that Cdk5-mediated PIPKIγ90 phosphorylation is crucial for cell invasion.

Figure 2.

Cdk5-mediated PIPKIγ90 phosphorylation regulates cell invasion. A) PIPKIγ90^WT restored the invasive capacity of PIPKIγ-depleted cells, but PIPKIγ^S453A and PIPKIγ^S453E did not. PIPKIγ-depleted MDA-MB-231 cells were infected with retroviruses to express codon-modified FLAG-PIPKIγ90, -PIPKIγ90^S453A, or PIPKIγ90^S453E and then selected with neomycin. Matrigel invasion was performed with 50 ng/ml HGF in the lower chambers. Cells that express shRNA control were used as controls. Data are presented as means ± sem, n = 4. *P < 0.05, **P < 0.01 vs. shRNA control. B) PIPKIγ-depleted PC3-MM2 cells were infected with retroviruses that express codon-modified FLAG-PIPKIγ90, -PIPKIγ90^S453A, or PIPKIγ90^S453E. Resulting cells were examined for their Matrigel invasive capacities with cells that express shRNA as a control. Data are presented as means ± sem, n = 3. *P < 0.05 vs. shRNA A1. C) Depletion of Cdk5 by shRNA inhibited the invasion of MDA-MB-231 cells. Cell invasion was performed with 50 ng/ml HGF in the lower chambers. Data are presented as means ± sem of 3 independent experiments. *P < 0.05, **P < 0.01. D) Expression of PIPKIγ90 inhibited the invasion of Cdk5-depleted cells. Cdk5-depleted MDA-MB-231 cells and the shRNA control cells were infected with retroviruses to express FLAG-PIPKIγ90. Resulting cells were examined for their Matrigel invasive capacities with the parental cells as controls. Data are presented as means ± sem of 4 independent experiments. **P < 0.01.

To know whether Cdk5-mediated phosphorylation regulates PIPKIγ90 activity, we measured the kinase activities of PIPKIγ90^WT and the phosphorylation site mutants. FLAG-PIPKIγ90^WT, -PIPKIγ90^S453A, and -PIPKIγ90^S453E were transfected into CHO-K1 cells and immunoprecipitated with anti-FLAG antibody-conjugated agarose beads or normal IgG agarose beads. The activities of PIPKIγ90 and mutants were determined by measuring PI(4,5)P₂ production with PI(4)P and [γ-³²P]ATP as substrates. PI(4,5)P₂ was separated by thin-layer chromatography, imaged by autoradiography, and quantified by liquid scintillation counting. A very low kinase activity was detected in IgG-agarose beads that were incubated with FLAG-PIPKIγ90^WT-transfected lysates (data not shown). Substitution of Ser453 with alanine or glutamic acid significantly reduced PIPKIγ90 activity in vitro (Fig. 3A). To examine whether mutation at Ser453 also causes reduction in PIPKIγ90 activity in breast cancer cells, PIPKIγ90-depleted MDA-MB-231 cells that reexpress FLAG-PIPKIγ90^WT, -PIPKIγ90^S453A, or -PIPKIγ90^S453E were plated on fibronectin-coated glass-bottom dishes, fixed, and costained with a mouse anti-PIP₂ mAb (2C11) and a rabbit anti-PIPKIγ pAb. Reexpression of PIPKIγ90^WT caused an increase in PIP₂ levels, whereas reexpression of PIPKIγ90^S453A or PIPKIγ90^S453E did not (Fig. 3B). Surprisingly, PIPKIγ90 that localized to FAs did not costain with PIP₂. To test the role of Cdk5 in regulating PIPKIγ90 activity, ZZ-PIPKIγ90 was cotransfected with an empty vector, Myc-Cdk5/Myc-p35, or Myc-Cdk5-144N/Myc-p35 into CHO-K1 cells. ZZ-PIPKIγ90 was immunoprecipitated with IgG agarose, and the activity of PIPKIγ90 was determined as described above. Coexpression of Cdk5/p35 inhibited the activity of PIPKIγ90, whereas cotransfection of p35 and Cdk5-144N, a kinase dead mutant, did not influence the activity of PIPKIγ90 (Fig. 3C). These data suggest that Cdk5-mediated phosphorylation negatively regulates PIPKIγ90 activity.

Figure 3.

Cdk5-mediated phosphorylation inhibits the activity of PIPKIγ90. A) FLAG-PIPKIγ90^WT, -PIPKIγ90^S453A, and -PIPKIγ90^S453E were transfected into CHO-K1 cells and immunoprecipitated with anti-FLAG agarose beads. Left, top) Expression levels of PIPKIγ90 are shown. The activities of PIPKIγ and the mutants were determined with PI(4)P and [γ-³²P]ATP as substrates. PIP₂ was visualized by autoradiography and quantitated by liquid scintillation counting. Data are presented as means ± sem of 3 independent experiments. **P < 0.01, ***P < 0.001 vs. control. B) PIPKIγ-depleted MDA-MB-231 cells that reexpress FLAG-PIPKIγ90^WT, -PIPKIγ90^S453A, and -PIPKIγ90^S453E were plated on fibronectin (5 µg/ml)-coated, glass-bottom dishes and cultured for overnight. The cells were costained with an anti-PIP₂ mouse mAb and an anti-PIPKIγ rabbit pAb. PIP₂ intensity was quantitated with NIS-Elements. Data are presented as means ± sem of 3 independent experiments; >30 cells were analyzed in each group. **P < 0.01 vs. control. C) ZZ-PIPKIγ90^WT was cotransfected with the empty vector, Myc-Cdk5/Myc-p35, or Myc-Cdk5-144N/Myc-p35 into CHO-K1 cells. ZZ-PIPKIγ90 was immunoprecipitated with IgG-agarose beads. Left, top: the expression levels of PIPKIγ90 are shown. The activity of PIPKIγ90 was determined as described for A. Data are presented as means ± sem of 3 independent experiments. *P < 0.05 vs. control.

Because of the crucial role of fibronectin in cell invasion (1, 38–41), we set out to determine whether the Cdk5-PIPKIγ90 pathway regulates fibronectin secretion. To examine whether the phosphorylation site mutants of PIPKIγ90 influence fibronectin secretion, we examined the fibronectin secretion of PIPKIγ90-depleted MDA-MB-231 cells that were rescued with PIPKIγ90^WT, PIPKIγ90^S453A, and PIPKIγ90^S453E. The cells were grown in 10-cm dishes to 80% confluence, washed 3 times with PBS, and incubated with 10 ml of serum-free Opti-MEM with 20 ng/ml HGF for 24 h. The supernatants were collected, concentrated with Spin-X ultra-concentrators and analyzed with Western blotting. Depletion of PIPKIγ90 significantly reduced fibronectin secretion, and reexpression of PIPKIγ90^WT restored and further promoted fibronectin secretion (Fig. 4A). The fibronectin secretion in the control, PIPKIγ90-depleted, and the PIPKIγ90-rescued cells correlated with the PIPKIγ levels in those cells (Fig. 2A). Reexpression of PIPKIγ90^S453A and PIPKIγ90^S453E partially or fully rescued fibronectin secretion, as compared with the control cells, but were unable to promote fibronectin secretion when compared with that of the WT (Fig. 4A). To understand the role of Cdk5 in fibronectin secretion, Cdk5-depleted MDA-MB-231 cells were analyzed for fibronectin secretion with shRNA vector cells as a control. Depletion of Cdk5 enhanced fibronectin secretion (Fig. 4B), suggesting that Cdk5 is a negative regulator of fibronectin secretion. To examine the role of PIPKIγ activity in fibronectin secretion, MDA-MB-231 cells were treated with DMSO or UNC3230, a specific PIPKIγ inhibitor, and then treated with HGF in serum-free medium for 24 h. The supernatants were collected, and the fibronectin secretion was analyzed. UNC3230 significantly inhibited fibronectin secretion in a dose-dependent manner (Fig. 4C). These results suggest that Cdk5-mediated phosphorylation of PIPKIγ90 negatively regulates fibronectin secretion.

Figure 4.

Cdk5-mediated phosphorylation regulates fibronectin secretion. A) Secreted fibronectin in PIPKIγ-depleted MDA-MB-231 cells that reexpress FLAG-PIPKIγ90^WT, -PIPKIγ90^S453A, and -PIPKIγ90^S453E were collected from serum-free supernatants, concentrated, and determined by Western blotting. The fibronectin secretion was examined in PIPKIγ-depleted cells and empty PLKO.1 vector-transfected cells, and was used as controls. Data are presented as means ± sem of 3 independent experiments. *P < 0.05, ***P < 0.01. B) Cdk5-depleted MDA-MB-231 cells were examined for fibronectin secretion, and cells transfected with empty pLKO.1 vector ere used as controls. Data are presented as means ± sem of 3 independent experiments. *P < 0.05 vs. control. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. C) Effect of UNC3230 on the fibronectin secretion of MDA-MB-231 cells. Data are presented as means ± sem of 3 independent experiments. *P < 0.05, **P < 0.01 vs. control.

To visualize fibronectin secretion in cells, PIPKIγ-depleted MDA-MB-231 cells that reexpress FLAG-PIPKIγ90^WT, -PIPKIγ90^S453A, and -PIPKIγ90^S453E were plated on laminin and stained for fibronectin. Fibronectin secretion was examined using TIRF microscopy. Secreted fibronectin appeared as round dots, and depletion of PIPKIγ reduced fibronectin secretion; reexpression of PIPKIγ90^WT in the depleted cells restored fibronectin secretion, whereas that of PIPKIγ90^S453A or PIPKIγ90^S453E did not (Fig. 5A, B). To know which molecules are colocalized with secreted fibronectin, MDA-MB-231 cells were plated on laminin and costained with anti-fibronectin antibody and anti-Zyxin (FA marker), anti-Tks5 (invadopodium marker), and anti-CD9 (exosome marker) antibodies. As shown in Fig. 5C, secreted fibronectin did not colocalize with Zyxin. However, it significantly colocalized with Tks5 and CD9, suggesting that Tks5 and CD9 may be involved in fibronectin secretion.

Figure 5.

Determining fibronectin secretion by TIRF microscopy. A) PIPKIγ-depleted MDA-MB-231 cells that express FLAG-PIPKIγ90^WT, -PIPKIγ90^S453A, and –PIPKIγ90^S453E were plated on laminin-coated (10 µg/ml), glass-bottom dishes and cultured for 36 h. The cells were fixed and costained with an anti-fibronectin antibody and ActinFarRed (phalloidin) 647. Fibronectin secretion was observed using TIRF microscopy. Scale bar, 10 µm. B) Fibronectin (Fn) intensities were determined in PIPKIγ-depleted MDA-MB-231 cells that express FLAG-PIPKIγ90^WT, -PIPKIγ90^S453A, and -PIPKIγ90^S453E in PIPKIγ-depleted cells, and cells transfected with empty pLKO.1 vector were used as controls. Data are presented as means ± sem of 30 cells from 3 independent experiments. ***P < 0.01 vs. control. C) Colocalization of fibronectin with Tks5 and CD9, but not Zyxin. MDA-MB-231 cells were plated on laminin, cultured for 36 h, fixed, and costained with anti-fibronectin/-Zyxin, anti-fibronectin/-Tks5, anti-fibronectin/-CD9 antibodies, respectively. The colocalization of fibronectin with those molecules was examined by TIRF microscopy. Scale bar, 10 µm.

DISCUSSION

Cdk5 regulates FA assembly and disassembly and, consequently, cell migration and invasion by phosphorylating talin and PIPKIγ90 (36, 42), and we recently demonstrated that PIPKIγ90 ubiquitination and degradation regulate FA dynamics to regulate cell migration and invasion (24, 25). In the present study, we demonstrated that Cdk5-mediated PIPKIγ90 phosphorylation regulates PIPKIγ90 activity to control fibronectin secretion and, consequently, cell invasion.

We demonstrated that Cdk5 phosphorylates PIPKIγ90 at S453 in vitro and in cells. It has been reported that Cdk5 phosphorylates PIPKIγ90 at Ser650 to inhibit the PIPKIγ90–talin interaction. Because there are several other potential Cdk5 phosphorylation sites on PIPKIγ90, we have employed mass spectrometry to identify Cdk5 phosphorylation sites on PIPKIγ90 by cotransfecting PIPKIγ90 and Cdk5/p35. Surprisingly, we did not detect any phosphorylation at Ser650, although the peptide containing Ser650 was detected in mass spectrometric analysis. Instead, we identified Ser453 as a phosphorylation site on PIPKIγ90 in Cdk5/p35-transfected cells (Fig. 1B). Mutation at Ser453 abolished the phosphorylation of PIPKIγ90 by Cdk5 in vitro (Fig. 1D). Furthermore, transfection of Cdk5/p35 promoted Ser453 phosphorylation, whereas Cdk5 shRNAs or roscovitine, a specific inhibitor of Cdk5, blocked HGF-induced Ser453 phosphorylation (Fig. 1I, J). The results suggest that Ser453 on PIPKIγ90 is a major Cdk5 phosphorylation site. Cdk5 is a p35-activated protein kinase, which is enriched in neuronal cells and regulates neuronal migration (43–46). New studies have demonstrated that it also regulates the migration and invasion in many other cell types, including cancer cells (36, 42, 47–49). Cdk5 is amplified and overexpressed in human pancreatic cancer tissues (50). It has been reported that Cdk5 is essential for TGF-β–induced epithelial–mesenchymal transition and breast cancer progression (51). We demonstrated here that depletion of Cdk5 using shRNAs inhibited the invasion of MDA-MB-231 cells. PIPKIγ90 produces PIP₂ and is essential for cancer cell invasion and metastasis (25, 26, 32). Reexpression of PIPKIγ90 in PIPKIγ90-depleted cells restored the invasion, whereas that of PIPKIγ90^S453A or PIPKIγ90^S453E did not (Fig. 2). The incapacity of PIPKIγ90^S453E to rescue the invasion of PIPKIγ90-depleted cells suggests that the glutamate residue does not mimic Cdk5-mediated phosphorylation of PIPKIγ90. Because cell invasion is a dynamic process that requires phosphorylation and dephosphorylation, the glutamate residue cannot fulfill that dynamic process. Similar results were reported in other phosphorylation pathways (36, 52). The similar effects of PIPKIγ90^S453A and PIPKIγ90^S453E on cell invasion in MDA-MB-231 cells and PC-3-MM2 cells (Fig. 2A, B) could be attributed to the equivalent activity of the 2 mutants (Fig. 3). It has been reported that protein kinase D1 phosphorylates PIPKIγ90 at Ser448 (53), a residue close to Ser453, suggesting that the region around Ser453 is a regulated hot spot. It is likely that Cdk5 and protein kinase D1 phosphorylate PIPKIγ90 to regulate the invasion of MDA-MB-231 cells under different conditions. These results suggest that Cdk5-mediated PIPKIγ90 phosphorylation regulates cell invasion.

Because PIPKIγ90 is an enzyme that produces PIP₂, we examined the role of Cdk5-mediated phosphorylation in regulating PIPKIγ90 activity. Substitution of Ser453 with Glu significantly inhibited PIPKIγ90 activity (Fig. 3A), suggesting that Cdk5-mediated phosphorylation inhibits PIPKIγ90 activity. Surprisingly, substitution of Ser453 with Ala also inhibited PIPKIγ90 activity, suggesting that Ser453 is very critical for the activity of PIPKIγ90, so that substitution of Ser453 with any other residues will disrupt the activity. That result is consistent with the PIP₂ staining intensity in MDA-MB-231 cells expressing PIPKIγ90^WT, PIPKIγ90^S453A, and PIPKIγ90^S453E. The critical role of Ser453 in regulating PIPKIγ90 activity is supported by a recent report that protein kinase D–mediated phosphorylation of PIPKIγ90 at Ser488, a residue close to Ser453, inhibits the activity of PIPKIγ90 (53). Furthermore, cotransfection with Cdk5/p35 reduced PIPKIγ90 activity. These results suggest that Cdk5-mediated phosphorylation negatively regulates PIPKIγ90 activity. PIP₂ staining, however, did not significantly colocalize with PIPKIγ90 at FAs. This is probably because PIP₂ at FAs is sequestered by vinculin and talin, thus not being accessible to the antibody.

Fibronectin is a key component of the extracellular matrix and is crucial for cell migration and invasion (1, 2, 38, 40). However, the molecular mechanisms that regulate fibronectin secretion are less understood. Depletion of PIPKIγ90 suppressed fibronectin secretion, whereas reexpression of PIPKIγ90 restored and promoted fibronectin secretion (Figs. 4A and 5A). Moreover, inhibition of PIPKIγ90 activity with UNC3230 also inhibited fibronectin secretion. These data suggest that PIPKIγ90 activity is essential for fibronectin secretion, which is consistent with the role of PIP₂ in vesicle docking (7, 8, 54).

It has been reported that Cdk5 negatively regulates insulin secretion (55, 56). Our data show that depletion of Cdk5 significantly promoted fibronectin secretion (Fig. 4B), suggesting that Cdk5 is also a negative regulator of fibronectin secretion. Moreover, PIPKIγ90^S453A and PIPKIγ90^S453E were unable to rescue or promote fibronectin secretion (Figs. 4A and 5A), suggesting that Cdk5-mediated phosphorylation of PIPKIγ90 negatively regulates fibronectin secretion. The fibronectin secretions in Figs. 4A and 5A are slightly different, probably because experimental conditions were different: the cells were grown on tissue culture dishes in Fig. 4A but on laminin-coated, glass-bottom dishes in Fig. 5A.

Secreted fibronectin did not colocalize with Zyxin, a FA protein, suggesting that FAs are not involved in fibronectin secretion. However, fibronectin significantly colocalized with Tks5, an invadopodia marker, and with CD9, an exosome marker (Fig. 5C). It has been reported that exosomes, which include fibronectin as cargo, are secreted through invadopodia (6, 57), suggesting that fibronectin may be secreted by invadopodium-like structures.

PIP₂ is crucial for cell migration and invasion (58–60). The highly regulated PIP₂ production is controlled by PIPKIγ90 ubiquitination (25, 61) and Cdk5-mediated phosphorylation. PIP₂, in turn, regulates the secretion of fibronectin, which, consequently, modulates cell invasion (1, 2, 41, 62). During cell invasion, fibronectin secretions are regulated spatially and temporarily. For example, adhesion rings surrounding invadopodia are essential for their maturation (63). It is likely that PIPKIγ90 mediates fibronectin secretion at the adhesion rings, whereas Cdk5-mediated PIPKIγ90 phosphorylation suppresses fibronectin secretion in the centers of invadopodia, thus regulating invadopodium maturation. That hypothesis is supported by the critical role of Cdk5 in invadopodium maturation (33). Future studies should refine this hypothetical model.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

Cdk5

cyclin-dependent kinase 5

CHO

Chinese hamster ovary

EGF

epidermal growth factor

FA

focal adhesion

FBS

fetal bovine serum

H+L

heavy and light chain

HGF

hepatocyte growth factor

N-WASP

neural Wiskott-Aldrich syndrome protein

PIP₂

phosphatidylinositol 4,5-bisphosphate

PIPKIγ90

phosphatidylinositol 4-phosphate 5-kinase type I γ 90

SCF

stem cell factor

shRNA

short hairpin RNA

TIRF

total internal reflection fluorescence

WT

wild type

AUTHOR CONTRIBUTIONS

L. Li, T. Kołodziej, N. Jafari, and J. Chen performed the experiments and data analysis; H. Zhu and Z. Rajfur contributed the manuscript discussion and writing; and C. Huang directed the research, performed the experiments, and wrote the paper.