MAP4K3/GLK Promotes Lung Cancer Metastasis by Phosphorylating and Activating IQGAP1

More than 90% of human cancer–related death is associated with tumor metastasis (1, 2). Cancer cell migration contributes to tumor metastasis (3). Understanding the fundamental mechanisms of cancer cell migration should help the development of novel therapeutic approaches for treating cancer metastasis.

IQ motif–containing GTPase-activating protein 1 (IQGAP1) is a scaffold protein that promotes multiple aspects of cell migration (4). For example, IQGAP1 weakens cell–cell adhesion, induces cytoskeletal rearrangement, and degrades extracellular matrix (5–7). Upon phosphorylation by PKC-ϵ at Ser-1443, IQGAP1 undergoes a conformational change and becomes activated (8). The Rho family GTPases Cdc42 and Rac1 directly interact with IQGAP1 and localize to the leading edge of migrating cells, leading to actin meshwork formation and cell migration (9). The regulation of cell migration by the IQGAP1/Cdc42/Rac1 system suggests that this protein complex is involved in tumor progression (10). Further studying the molecular mechanisms involved in the control of IQGAP1 activity shall provide important insights into the regulation of cancer cell migration and metastasis.

The serine/threonine protein kinase GLK (also named MAP4K3) is a member of the mitogen-activated protein kinase kinase kinase kinase (MAP4K) family (11). As upstream regulators of the MAP kinase cascades, GLK and other MAP4Ks activate c-Jun N-terminal kinase (JNK) in response to environmental stress and proinflammatory cytokines in cultured cell lines (12–17). One MAP4K family kinase, HGK (MAP4K4), is a critical regulator of cell migration, cancer invasion, and cell adhesion (18–22). HPK1 (MAP4K1) regulates cell apoptosis, cell growth, and cytokine production through binding to multiple adaptor proteins including members of the Grb2 family, Nck family, Crk family, and SLP-76 family (23). GLK regulates mTOR signaling, cell growth, apoptosis, and autophagy (24–27). GLK also upregulates NF-κB signaling by activating PKC-θ in T cells, leading to T-cell activation (28, 29). GLK signaling in T cells specifically enhances IL17A transcription by inducing the AhR–RORγt complex (30, 31). The increase of GLK protein levels in T cells is correlated with the disease severity of several human autoimmune diseases (28, 31–33). Moreover, GLK protein levels are increased in tissues of human lung cancer and hepatoma (34, 35); GLK overexpression in the cancer tissue is correlated with cancer recurrence and poor recurrence-free survival rates (34, 35). GLK overexpression in cancer cell lines may be due to the downregulation of miRNA let-7c and miR-199a-5p, which target the 3′-untranslated region of GLK (36, 37). However, the role of GLK in cancer recurrence remains unclear. Here we report a novel GLK-targeted protein, IQGAP1, which mediates GLK-induced cell migration and lung cancer metastasis.

The plasmids expressing 3×Flag-tagged or HA-tagged human GLK cDNA (NCBI accession number: NM_003618) were generated by subcloning the cDNA insert from the Flag-tagged clone into the vector pCMV6-AN-3DDK or pCMV6-AC-HA (OriGene). The plasmid expressing Myc-tagged human IQGAP1 was purchased from Addgene (#30118). The plasmid expressing GLK kinase-dead (K45E) mutant was generated using the PCR-based site-directed mutagenesis from the 3×Flag-tagged GLK clone, by mutating Lys-45 to glutamic acid at the ATP-binding domain of this kinase, as described previously (12, 38). The plasmids expressing GLK (P436/437A), GLK (P478/479A), IQGAP1 (ΔWW), IQGAP1 (S480A), IQGAP1 (S480D), and IQGAP1 (S480E) were generated by mutating the indicated residue on the 3×Flag-tagged GLK and Myc-tagged IQGAP1 plasmids. The plasmids expressing CFP-fused GLK and monomeric GFP (mGFP)-fused GLK were generated by subcloning the GLK cDNA insert from the HA-tagged GLK plasmid into the vector pCMV6-AC-mCFP and pCMV6-AN-mGFP (OriGene), respectively. The plasmids expressing YFP-fused IQGAP1 and Tomato-fused IQGAP1 were generated by subcloning the IQGAP1 cDNA insert from the Myc-tagged IQGAP1 plasmid into the vector pCMV6-AC-YFP (OriGene) and ptd-Tomato-C1 (Clontech), respectively. The GLK or IQGAP1 shRNA expression plasmids were obtained from the National RNAi Core Facility (Academia Sinica, Taiwan). For pervanadate treatment, pervanadate was freshly prepared by mixing H₂O₂ and Na₃VO₄ as described previously (39), and cells were then incubated with a final concentration of 25 μmol/L pervanadate for 1 hour at 37°C. G-LISA Activation Assay Biochem kits for Cdc42 or Rac1 were purchased from Cytoskeleton, Inc. Anti-Flag agarose beads (M2) and anti-Myc agarose beads (9E10) were purchased from Sigma. The primary antibodies used in this study were anti-IQGAP1 (BD Biosciences), anti-HA, anti-β-actin (Sigma), and anti-β-Tubulin (Sigma). Both homemade anti-GLK antibody (α-GLK-N; ref. 30) and homemade anti-GLK mAb (mAb clone C3; ref. 30) were used in this study.

We collected primary lung tumor specimens from 7 non–small cell lung cancer (NSCLC) patients who underwent first pulmonary resection in the Division of Thoracic Surgery at Taichung Veterans General Hospital, Taiwan. Every patient provided written informed consent approved by the hospital's Institutional Review Board (approval number: CF13082). All experiments were performed in accordance with the guidelines and protocols approved by the Institutional Review Board, Taichung Veterans General Hospital, Taiwan. Tumor types and stages of individual specimens were determined according to the American Joint Committee on Cancer Staging Manual. All specimens, including tumor tissues and paired normal adjacent tissues of patients with NSCLC, were taken at the time of surgical resection. Portions of samples were freshly fixed with formaldehyde and then embedded with paraffin. Follow-up data were collected from chart reviews and confirmed by thoracic surgeons.

Human lung cancer and normal adjacent tissue array slides (#CC5, #CCA4, and #CCN5) were purchased from SUPER BIO CHIPS. The company provided certified documents that all human lung tissue samples were collected with patients' informed consents. The pulmonary tissue array contained 68 normal adjacent tissues and 109 tumor tissues (including small-cell carcinoma, NSCLC, mucoepidermoid carcinoma, and carcinosarcoma).

We analyzed the GLK–IQGAP1 complex in 177 pulmonary samples from tissue arrays and in 7 pulmonary resection samples from NSCLC patients. We analyzed IQGAP1 phosphorylation in 6 pulmonary resection samples from patients with NSCLC.

The murine lung cancer (luciferase-expressing Lewis lung carcinoma, LLC/luc; PerkinElmer # BW119267), human lung cancer (HCC827, H661, H1299; ATCC), and human normal lung (NL20; ATCC) cell lines were cultured in RPMI1640 medium supplemented with 10% FBS, 100 U/mL penicillin, and 100 mg/mL streptomycin (Invitrogen). The human embryonic kidney cell line HEK293T was maintained in DMEM containing 10% FBS, 100 U/mL penicillin, and 100 mg/mL streptomycin. Cell lines at passages greater than 10 were not used for the experiments in this study. All cells were grown at 37°C in a humidified atmosphere of 5% CO₂ in air. Plasmids were transfected into cells using polyethylenimine reagents. All cell lines used were tested and confirmed to be negative for Mycoplasma.

Pol II-GLK Tg mice and Pol II-GLK^E351K Tg mice in C57BL/6 background were generated using pronuclear microinjection by NHRI Transgenic Mouse Core. A full-length human GLK coding region (wild-type or E351K mutant) was placed downstream of the RNA polymerase II (Pol II) promoter (Fig. 1A; ref. 40). IQGAP1 knockout mice in C57BL/6 background were generated using embryo microinjection of TALEN mRNA by NHRI Transgenic Mouse Core. The nucleotide (nt) 161 guanine of the IQGAP1 exon 1 was deleted in the mutated allele. All animal experiments were performed in the AAALAC-accredited animal housing facilities at National Health Research Institutes (NHRI, Zhunan, Taiwan). All mice were used according to the protocols (approval number: 107002-A) and guidelines approved by the Institutional Animal Care and Use Committee of NHRI (Zhunan, Taiwan).

The mice were sacrificed by CO₂ asphyxiation. The lung from the chest was excised and cut into small fragments, followed by placing the lung fragments into the dispase-containing culture dish at room temperature. After 10 minutes, lung fragments were homogenized by gentleMACS Dissociator (Miltenyi Biotec) to obtain single-cell suspension. Mouse lung epithelial cells were negatively selected by MACS Separation. The cells were incubated with biotin-conjugated antibodies (anti-CD45, anti-CD11b, anti-CD11c, anti-CD16/32, anti-CD19, anti-F4/80, and anti-TER-119) at 4°C for 15 minutes. Ten microliters of Streptavidin MicroBeads were added into cell suspension at 4°C for 15 minutes. Cell suspension was applied onto the column and then the effluent fraction containing epithelial cells was collected.

Tissue sections were deparaffinized, and then treated for antigen retrieval by incubating the slides in boiling buffer (pH 6.0) at 85°C for 10 minutes. Nonspecific binding was sequentially blocked with 3% H₂O₂ for 10 minutes and Immunoblock-Ultra V block for 5 minutes. Tissue sections were incubated with anti-proliferating cell nuclear antigen (PCNA; 1:200; GeneTex) or anti-EGFR^del antibodies (1:200; Cell Signaling Technology) at 4°C overnight, and then incubated with HRP-conjugated secondary antibodies. The protein signals were detected using the HRP substrate 3,3′diaminobenzidine (DAB; Ultravision Quanto Detection System; Thermo Fisher Scientific, TL-060-QHL). For negative controls, primary antibodies were replaced with 2% normal serum. Tissue sections were also counterstained with Mayer hematoxylin.

For monitoring subcellular localization of GLK-mGFP and IQGAP1-Tomato in migrating cells, 2 × 10⁴ cells were seeded into 8-chamber slides 24 hours after transfection. After a further 24 hours of incubation, cells were traced using Nikon Structured Illumination Microscope (N-SIM) performed on an Eclipse Ti inverted microscope equipped with a Plan Apo ×60 water immersed objective and time-lapse live-cell imaging systems (Nikon). Motile transfected (mGFP- and Tomato-positive) cells were followed in time-lapse recording for 10 hours at an interval of 10 minutes. The images were acquired and analyzed with the NIS Elements software (Nikon).

The mass spectrometry was performed as described previously (41, 42). Briefly, protein samples were separated by SDS-PAGE and stained with silver. Specific protein bands were excised, destained, and digested with trypsin. The resulting peptide mixtures were analyzed by loading on the NanoAcquity system (Waters) connected to an LTQ-Orbitrap XL Hybrid Mass Spectrometer (Thermo Fisher Scientific) equipped with a nanospray interface (Proxeon).

Cells seeded on sterile cover slides were cotransfected with Flag-tagged GLK and Myc-tagged IQGAP1 expression plasmids, followed by fixation, permeabilization, and blocking. In situ proximity ligation assay (PLA) assays were performed using the Duolink In Situ-Red kit (Sigma) according to the manufacturer's instructions. Briefly, cells were incubated with anti-Flag and anti-Myc antibodies, followed by species-specific secondary antibodies conjugated with oligonucleotides (PLA probes). After ligation and amplification reactions, the signal from each pair of PLA probes in close proximity (<40 nm; GLK–IQGAP1 interaction) was visualized as an individual red spot and analyzed by Leica DM2500 upright fluorescence microscope. For cell line experiments, at least five different fields were randomly selected, and the number of red spots per cell was counted. Each experiment was repeated at least three times.

For experiments using human pulmonary tissues, tissue sections were deparafinized, antigen retrieved, and nonspecific-binding blocked, followed by in situ PLA assays using first antibodies for IQGAP1 (1:4,000, CUSABIO) plus either GLK (1:3,000, mAb clone C3), or phospho-IQGAP1 Ser-480 (1:2,000, Allbio Science). The mAb for phosphorylated IQGAP1 Ser-480 was generated by immunization of a mouse with phospho-peptides (human IQGAP1 epitope: ⁴⁷³NTVWKQL[pS] SSVTGLT⁴⁸⁷). The tissue sections were then incubated with species-specific secondary antibodies conjugated with oligonucleotides (PLA probes), followed by ligation and amplification reactions. The number of PLA signals per tissue (3.14 mm²) was counted.

All experiments were repeated at least three times. The associations between metastasis and GLK transgene were evaluated using the Fisher exact test. To evaluate normality of each column data, Kolmogorov–Smirnov and Shapiro–Wilk tests were performed. The statistical significance between two unpaired groups was analyzed using the two-tailed Student t test (for normally distributed data) or using the two-tailed Mann–Whitney U test (for nonnormally distributed data). Cluster analyses (hierarchical clustering and subsequent k-means clustering) were used to divide patients into subgroups. Kaplan–Meier survival analyses were performed to show the difference in the survival between subgroups (e.g., PLA signal-High vs. PLA signal-Low). The log-rank test was used to calculate the significance of the survival distributions between two groups. Data were calculated using SPSS 19 software. A P value of <0.05 was considered statistically significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001). All statistical analyses of clinical data were further independently verified by two biostatisticians at Institute of Population Sciences of National Health Research Institutes (Zhunan, Taiwan).

To study the role of GLK in cancer progression, we generated the whole-body GLK transgenic (Pol II-GLK Tg) mice using RNA polymerase II (Pol II) promoter–driven human GLK cDNA (Fig. 1A). GLK overexpression was confirmed by real-time PCR (Fig. 1B). Because GLK overexpression is correlated with cancer recurrence of human NSCLC and hepatoma (34, 35), we characterized whether Pol II-GLK Tg mice spontaneously develop lung cancer or liver cancer using IHC analysis. The data showed that 1-year-old Pol II-GLK Tg mice did not develop any lung cancer or liver cancer (Supplementary Fig. S1A). To study the effect of GLK on cancer progression, Pol II-GLK Tg mice were bred with a genetically modified lung cancer mouse line, the lung-specific pulmonary surfactant protein A (SPA) promoter-driven EGFR-deletion mutant transgenic (SPA-EGFR^del Tg) mouse line (43). Consistent with the published results (43), all 1-year-old SPA-EGFR^del Tg mice (9/9) indeed developed lung cancer (PCNA positive; Fig. 1C; ref. 44), and so did SPA-EGFR^del;Pol II-GLK Tg mice (15/15; Fig. 1C). IHC analysis using anti-EGFR-deletion mutant antibodies showed that EGFR^del-expressing cells were detected in the lung cancer of both SPA-EGFR^del Tg mice and SPA-EGFR^del;Pol II-GLK Tg mice (Fig. 1D). We next studied whether GLK transgene induces lung cancer (EGFR^del-positive) metastasis to other organs in SPA-EGFR^del;Pol II-GLK Tg mice. We performed IHC using anti-EGFR deletion–mutant antibodies on the tissues of the cervical lymph nodes, the liver, and the brain from wild-type and three different transgenic mice. For regional metastasis to cervical lymph nodes, all but one (14/15) SPA-EGFR^del;Pol II-GLK Tg mice displayed numerous metastatic EGFR^del-expressing lung cancer cells in cervical lymph nodes. In contrast, only 3 of 9 SPA-EGFR^del Tg mice showed a few metastatic EGFR^del-expressing lung cancer cells in cervical lymph nodes (Fig. 1E; Supplementary Fig. S1B). For distant metastasis, all SPA-EGFR^del;Pol II-GLK Tg mice displayed metastasis of EGFR^del-expressing lung cancer cells to the brain (14/15) or liver (15/15) (Fig. 1E). In the 9 control SPA-EGFR^del Tg mice, only one SPA-EGFR^del Tg mouse (1/9) developed both brain metastasis and liver metastasis, only one (1/9) developed liver metastasis, and remaining 7 mice did not develop any detectable distant metastasis (Fig. 1E). These results suggest that GLK induces distant metastasis of lung cancer to the brain and liver.

We next studied the underlying mechanism of GLK-induced lung cancer metastasis. Knockdown of miRNA let-7c or miR-199a-5p promotes cell migration of cancer cell lines and also upregulates GLK levels (36, 37), suggesting that GLK may induce lung cancer metastasis by promoting cancer cell migration. We first evaluated whether GLK indeed promotes lung cell migration by transwell migration assays. We used GLK overexpression and shRNA knockdown approaches to study the role of GLK in regulating migration of two human lung cancer cell lines. HCC827 lung cancer cells expressed lower levels, while H661 lung cancer cells expressed higher levels, of endogenous GLK proteins than those of the normal lung cell line NL20 (Supplementary Fig. S2A). The HCC827 and H661 cells were transfected with a GLK-expressing plasmid and GLK-specific shRNA constructs, respectively. The migrated cell number was increased by GLK overexpression (Supplementary Fig. S2B) but attenuated by GLK shRNA knockdown (Supplementary Fig. S2C). Overexpression or knockdown of GLK was confirmed by immunoblotting analysis (Supplementary Fig. S2D and S2E). MTT assays showed that neither overexpressing nor downregulating GLK affected cell proliferation during the incubation time period (16 hours) of cell migration assays (Supplementary Fig. S2F and S2G), excluding the possibility that the observed phenotypes were due to changes in cell proliferation. The involvement of GLK in cell migration was further evaluated using primary lung epithelial cells from previously generated GLK-deficient mice (28) and newly generated GLK transgenic (Pol II-GLK) mice. The expected GLK protein levels in GLK-deficient and transgenic mice were confirmed by immunoblotting analysis (Supplementary Fig. S3A). The primary lung epithelial cells were isolated from mice and subjected to transwell migration assays. The number of primary lung epithelial cells that migrated through transwells in a uniform field was increased for cells from GLK transgenic mice (Supplementary Fig. S3B and S3C), whereas the migrated cell number was decreased for cells from GLK-deficient mice, compared with that of wild-type mice (Supplementary Fig. S3B and S3C).

The regulation of cell migration by GLK was further verified by examining cell motility of primary lung epithelial cells by time-lapse live-cell imaging using Nikon Structured Illumination Microscopy. The migrating cells were defined by their displacement lengths that are more than 15 μm during a 3-hour period. The lung epithelial cells isolated from GLK transgenic mice showed a marked increase in the percentage of migrating cells compared with that of wild-type mice (Supplementary Fig. S3D and S3E; Supplementary Movie S1). Conversely, GLK-deficient lung epithelial cells showed a decrease in the percentage of migrating cells compared with that of wild-type mice (Supplementary Fig. S3D and S3E; Supplementary Movie S1). Moreover, the average migrating length of GLK transgenic lung epithelial cells was significantly increased, whereas that of GLK-deficient lung epithelial cells was significantly decreased (Supplementary Fig. S3F). These results indicate that GLK plays an important role in regulating cell migration.

To identify the GLK-targeted molecule involved in GLK-regulated cell migration, we characterized GLK-interacting proteins in anti-Flag immunocomplexes isolated from Flag-GLK–transfected HEK293T cells. Following immunoprecipitation of GLK, the GLK-interacting proteins were resolved by SDS-PAGE and visualized by silver staining (Fig. 2A). The seven most prominent protein bands enhanced in GLK-transfected cells were sliced and then digested by trypsin, and the resulting protein peptides were subjected to mass spectrometry analysis. Moreover, four pervanadate-induced tyrosine phosphorylation sites (Tyr-366, Tyr-379, Tyr-574, and Tyr-735) on GLK proteins were identified by mass spectrometry analysis (Fig. 2A, right). We identified several putative GLK-interacting proteins, including myosin, IQGAP1 (Fig. 2B), vimentin, drebrin, and HSP70 (ordered by database search scores from highest to lowest). Among these proteins, IQGAP1, a positive regulator of cell migration, was selected for further study, whereas myosins, HSPs, and cytoskeletal proteins are common contaminant proteins detectable by mass spectrometry. Next, we confirmed the interaction between GLK and IQGAP1 using reciprocal coimmunoprecipitation assays (Fig. 2C and D). GLK was coimmunoprecipitated with Flag-tagged IQGAP1 proteins with an anti-Flag antibody (Fig. 2C and D). This coimmunoprecipitation between GLK and IQGAP1 was abolished by GLK (Y735F) mutation (Fig. 2E), suggesting that Tyr-735 phosphorylation of GLK protein is important for the interaction between GLK with IQGAP1. In situ PLA with a combination of PLA probes corresponding to Flag (Flag-tagged GLK) and Myc (Myc-tagged IQGAP1) showed strong PLA signals in cells overexpressing both proteins than those overexpressing each alone (Fig. 2F and G). The PLA signals suggest a direct interaction (<40 nm) between GLK and IQGAP1. Moreover, the fluorescence resonance energy transfer (FRET) assay using CFP-tagged GLK and YFP-tagged IQGAP1 fusion proteins showed a direct interaction (1–10 nm) between these two molecules (Fig. 2H). To further confirm the direct interaction, coimmunoprecipitation experiments were performed using purified proteins. Flag-tagged GLK and Myc-tagged IQGAP1 proteins from HEK293T cell lysates were purified by eluting immunocomplexes with Flag and Myc peptides, respectively. The coimmunoprecipitation assays showed an interaction between purified GLK and IQGAP1 proteins (Fig. 2I). The data from three different approaches (PLA, FRET, and purified proteins) suggest that GLK interacts directly with IQGAP1.

To demonstrate the role of IQGAP1 in GLK-induced cell migration, we studied whether IQGAP1 knockout attenuates GLK-promoted cell migration of primary lung epithelial cells. We first generated IQGAP1 knockout mice using TALEN technology (Fig. 3A); IQGAP1 knockout was characterized by immunoblotting analyses (Fig. 3B). IQGAP1 knockout mice were then bred with GLK transgenic mice. The primary lung epithelial cells were isolated from the offspring or the parental GLK transgenic (Pol II-GLK) mice and subjected to transwell migration assays. The migrated cell number of primary lung epithelial cells from GLK transgenic mice was drastically decreased by IQGAP1 homozygote knockout (Fig. 3C and D), and was modestly decreased by IQGAP1 heterozygote knockout (Fig. 3C and D). The migration dynamics of primary lung epithelial cells from the offspring were further examined by time-lapse live cell imaging using Nikon Structured Illumination Microscopy. The percentage of migrating epithelial cells of GLK transgenic mice was increased compared with that of wild-type mice (Fig. 3E and F). The GLK-induced migration of GLK transgenic epithelial cell was significantly reduced by IQGAP1 homozygote knockout (Fig. 3E and F; Supplementary Movie S2), and modestly reduced by IQGAP1 heterozygote knockout (Fig. 3E and F; Supplementary Movie S2). Moreover, the migration lengths of primary lung epithelial cells were increased by GLK transgene compared with those of wild-type cells (Fig. 3G), whereas the GLK-induced migration lengths were decreased by IQGAP1 knockout (Fig. 3G). Next, we studied whether similar results can be obtained using the HCC827 lung cancer cell line (Supplementary Fig. S4). The HCC827 cells were transfected with GLK plasmid alone or together with each of two different IQGAP1 shRNA constructs. As compared with control cells, the migration ability of HCC827 cells was enhanced by GLK transfection alone but reduced by IQGAP1 knockdown with individual IQGAP1 shRNAs even in the presence of GLK overexpression (Supplementary Fig. S4A). GLK overexpression and IQGAP1 knockdown were confirmed by immunoblotting analysis (Supplementary Fig. S4B). No major difference was seen in cell growth between each group during the time period of cell migration assays (Supplementary Fig. S4C). Overall, these data suggest that GLK promotes cell migration through IQGAP1.

We then investigated how GLK promotes IQGAP1-mediated cell migration. First, we studied whether GLK colocalizes with IQGAP1 in migrating cells. The GLK-mGFP and IQGAP1-Tomato proteins were coexpressed in HCC827 lung cancer cells and monitored by time-lapse live-cell imaging using either confocal microscopy (Fig. 4A) or super-resolution N-SIM microscopy (Fig. 4B–D). Data showed that GLK was localized diffusely throughout the cell, including the plasma membrane (Fig. 4A). The IQGAP1 distribution pattern was similar to that of GLK (Fig. 4A). Notably, super-resolution N-SIM imaging showed that GLK and IQGAP1 were colocalized mainly at the cell membrane (Fig. 4B). During cell migration, the cell formed lamellipodia and spike-like filopodia at the migratory front of the cell (45). The super-resolution N-SIM microscopy showed a striking colocalization of GLK and IQGAP1 predominantly at filopodia and lamellipodia of the cell during the extension step prior to cell body movement. (Fig. 4C; Supplementary Movie S3). Moreover, time-lapse super-resolution live-cell imaging also showed colocalization of GLK and IQGAP1 predominantly at the leading edge of the migrating cells (Fig. 4D, Supplementary Movie S4 and S5). These data indicate that GLK and IQGAP1 colocalize predominantly at the leading edge and may cooperate to promote cell migration.

WW domains of proteins recognize proline-rich motifs and trigger downstream signaling pathways (46–48). GLK and other MAP4K members contain an N-terminal kinase domain, a C-terminal Citron homology (CNH) domain, and several proline-rich motifs in the middle (11, 12). MAP4K1 (also named HPK1) interacts with its interacting proteins via the proline-rich motifs (23). In addition, IQGAP1 binds to the MAP kinase ERK2 via its WW domain (49), which preferentially recognizes ligands containing proline-rich sequence (47). We thus asked whether GLK and IQGAP1 interact with each other through the proline-rich domain and the WW domain, respectively.

We tested whether IQGAP1 interacts with one or both of the two potential WW domain-recognized proline-rich regions of GLK. We generated three GLK mutants (P436/437A, P478/479A, and P436/437A;P478/479A) by substitution of the Pro-436/437 and/or Pro-478/479 residues to alanine within the two proline-rich regions (Supplementary Fig. S5A–S5D). The IQGAP1 WW domain mutant (ΔWW) was also generated (Supplementary Fig. S5A–S5D). Different pairs of these mutants and their wild-type constructs were then cotransfected into HEK293T cells. Overexpression of wild-type or mutant GLK and IQGAP1 proteins was confirmed by immunoblotting analysis (Supplementary Fig. S5D). In situ PLA assay with a combination of PLA probes corresponding to Flag (Flag-tagged GLK) and Myc (Myc-tagged IQGAP1) showed strong PLA signals in cells overexpressing both proteins. The PLA signals were reduced by overexpression of either GLK (P436/437A) or GLK (P478/479A) mutant (Supplementary Fig. S5B and S5C). Moreover, the interaction between GLK and IQGAP1 was completely abolished by double mutations (P436/437A;P478/479A) of both GLK proline–rich regions (Supplementary Fig. S5B and S5C). In addition, the PLA signals were reduced by overexpression of IQGAP1 (ΔWW) mutant (Supplementary Fig. S5B and S5C). Collectively, these data indicate that Pro-478/479 and Pro-436/437 regions of GLK mediate its binding to the WW domain of IQGAP1.

Because GLK directly binds to IQGAP1, we speculated that GLK may be a kinase that regulates IQGAP1-mediated cell migration. To determine whether GLK phosphorylates IQGAP1, in vitro kinase assay was conducted using purified proteins of GLK, GLK kinase-dead (K45E) mutant, and IQGAP1. IQGAP1 phosphorylation was induced by GLK but not GLK kinase-dead (K45E) mutant (Fig. 5A). Following SDS-PAGE fractionation and mass spectrometry analysis, Ser-480 was identified as the GLK-phosphorylated residue on IQGAP1 (Fig. 5B). Next, we tested whether the GLK-induced IQGAP1 Ser-480 phosphorylation regulates the activation of Cdc42 or Rac1, as well as the interaction of IQGAP1 with Cdc42 or Rac1. Immunoprecipitation data showed that active (GTP-binding) Cdc42 proteins were increased in GLK plus IQGAP1-overexpressing cells; conversely, active Cdc42 protein levels were attenuated by overexpression of GLK plus IQGAP1 (S480A) mutant (Fig. 5C, bottom). In contrast, active (GTP-binding) Rac1 protein levels were not increased by GLK plus IQGAP1 overexpression (Fig. 5D, bottom). These results were further supported by ELISA results of Cdc42 and Rac1 activation (Fig. 5E). In addition, coimmunoprecipitation data showed that the interaction of IQGAP1 with either Cdc42 or Rac1 was not affected by the IQGAP1 (S480A) mutation (Fig. 5C and D, top). To evaluate the role of IQGAP1 Ser-480 phosphorylation in IQGAP1-mediated cell migration, HCC827 cells were cotransfected with IQGAP1 (S480A) phosphorylation-defective mutant and GLK plasmids. The transwell migration assays showed that the migrated cell number of GLK-overexpressing cell was decreased by overexpression of IQGAP1 (S480A) mutant [Fig. 5F (top right), G, and H]. Conversely, overexpression of two IQGAP1 Ser-480 phosphomimetic (S480D and S480E) mutants induced a higher cell migration ability than that of overexpression of IQGAP1 or IQGAP1 (S480A) mutant in HCC827 lung cancer cells (Supplementary Fig. S6A–S6C). These results suggest that IQGAP1 Ser-480 phosphorylation is responsible for IQGAP1 activation and IQGAP1/Cdc42-mediated cell migration.

Our results suggest that GLK interacts with and phosphorylates IQGAP1. We next studied the interaction between GLK proline regions and IQGAP1 WW domain indeed controls the GLK-IQGAP1–induced cell migration. HCC827 lung cancer cells cotransfected with GLK and IQGAP1 displayed enhancement of migration than that of vector-transfected control cells, whereas cotransfection of GLK plus IQGAP1 (ΔWW) mutant abrogated the enhanced cell migration (Fig. 5F and G). Overexpression of GLK (P436/437A), (P478/479A), or (P436/437A;P478/479A) mutant attenuated GLK-induced cell migration (Fig. 5F and G). Overexpression of GLK and IQGAP1 was confirmed by immunoblotting analysis (Fig. 5H).

We next studied whether GLK promotes metastasis of lung cancer through IQGAP1-mediated cancer cell migration. To shorten the time (12 months) required for the development of lung cancer metastasis in SPA-EGFR^del;Pol II-GLK Tg mice, we generated a GLK-mutant transgenic mouse line (Pol II-GLK^E351K Tg), which expressed a constitutively activated GLK (E351K) mutant (Fig. 6A–C). Notably, the GLK (E351K) mutation was reported in the Supplementary information of a previous publication (50); however, the functional consequence of GLK (E351K) mutation has not been demonstrated until this study (Fig. 6A and B). Overexpression of GLK (E351K) mutant was confirmed by real-time PCR (Fig. 6D). Next, we bred Pol II-GLK^E351K Tg mice with SPA-EGFR^del Tg mice to generate SPA-EGFR^del;Pol II-GLK^E351K Tg mice, which displayed enhanced GLK protein levels in the lung compared with those of SPA-EGFR^del Tg mice (Fig. 6E). SPA-EGFR^del;Pol II-GLK^E351K Tg mice (8/8) indeed developed lung cancer (Fig. 6F) and regional/distant metastasis at a younger age (7-month-old) than that of SPA-EGFR^del;Pol II-GLK Tg mice. All 7-month-old SPA-EGFR^del;Pol II-GLK^E351K Tg mice displayed distant metastasis of EGFR^del-expressing lung cancer cells to the brain and/or liver [both brain and liver (6/8), brain only (1/8), and liver only (1/8); Fig. 6G]. In contrast, all SPA-EGFR^del;Pol II-GLK^E351K;IQGAP1^−/− mice did not develop any distant metastasis (0/6) at 7-month-old age (Fig. 6G). These data suggest that GLK induces distant metastasis through IQGAP1 in SPA-EGFR^del lung cancer model. To verify this notion, we studied the interaction between GLK and IQGAP1, as well as IQGAP1 Ser-480 phosphorylation in tissues of human NSCLC.

To study the interaction of GLK with IQGAP1 in NSCLC tissues, we collected clinical lung tissues from 7 human patients with NSCLC who underwent pulmonary resection. We also employed a commercially available pulmonary tissue array containing 85 NSCLC tissues (including 49 squamous cell carcinoma, 17 adenocarcinoma, 11 bronchioloalveolar carcinoma, and 8 large cell carcinoma) and 68 normal adjacent tissues, as well as 3 small-cell lung carcinoma tissues. These tissues were subjected to in situ PLA with a combination of paired PLA probes corresponding to GLK and IQGAP1. The data showed multiple strong PLA signals in most (81/92) of NSCLC tissues but not in any small cell carcinoma tissues (Fig. 7A and B). Most (61/68) of normal adjacent tissues from NSCLC patients did not display any PLA signals, while 7 of 68 normal adjacent tissues showed much less PLA signals. The few PLA signals in the normal adjacent tissues of patients with NSCLC may be metastatic cancer cells migrating from original lesion. These results suggest that the GLK–IQGAP1 complex in pulmonary tissue may be a diagnostic biomarker for NSCLC. Next, we studied whether the GLK–IQGAP1 complex could act as a potential prognostic biomarker for NSCLCs. Patients with NSCLC (squamous cell carcinoma and adenocarcinoma), whose survival data were available, were divided into four PLA signal subgroups after cluster analyses. The two subgroups with highest PLA signals contained only one and two patients, respectively, thus, these two subgroups were excluded from further analysis. The remaining two subgroups (n = 63) were subjected to Kaplan–Meier survival analysis using the survival data (available from the provider). Patients with NSCLC with high PLA signals showed poor survival during follow-up periods (n = 63, PLA signal-High vs. PLA signal-Low, P = 0.069; Fig. 7C). The higher P value (P = 0.069) may be due to the exclusion of three data with highest PLA signals. Nevertheless, patients with NSCLC with more GLK–IQGAP1 complexes have a lower survival rate than that of the patients with less GLK–IQGAP1 complexes. Because 40% to 60% of NSCLC patients die of cancer recurrence after cancer resection (51), we studied whether the GLK–IQGAP1 complex is associated with NSCLC metastasis. The cancer cells with the GLK–IQGAP1 complex particularly accumulated on/near the vascular wall in the lung (Fig. 7D); GLK–IQGAP1 complex–positive cells also existed in lumen of the blood vessel (Fig. 7D). Moreover, the bone, lymph node, or soft tissue section with metastatic carcinoma displayed GLK–IQGAP1 complex–positive cells (Fig. 7D); the cancer cells were verified using PCNA staining (Fig. 7D). Merged images show that the GLK–IQGAP1 complex–positive cells in these tissues were indeed cancer cells (Fig. 7D). The data suggest that lung cancer cells with the GLK–IQGAP1 complex tend to be metastatic. Next, we examined the GLK-induced IQGAP1 Ser-480 phosphorylation in human NSCLC tissues. After several failed attempts, we finally obtained a mAb against IQGAP1 Ser-480 phosphorylation. However, the immunostaining signal using phospho-IQGAP1 mAb was not strong enough to provide a discernible signal. To enhance the specificity and staining signal of anti-phospho-IQGAP1 mAb, we performed PLA that amplifies phosphorylation signals (52) like a PCR with a combination of paired PLA probes corresponding to IQGAP1 and phospho-IQGAP1 Ser-480. The antibody specificity was demonstrated using IQGAP1 S480A mutant–expressing cells and IQGAP1-knockout lung cancer tissues (SPA-EGFR^del;Pol II-GLK^E351K;IQGAP1^−/−; Supplementary Fig. S7A and S7B). Using human NSCLC tissues, we found multiple PLA signals of IQGAP1 Ser-480 phosphorylation in 82.7% (72/87) of tumor tissues tested (Fig. 7E and F). The phospho-IQGAP1 PLA signals coexisted with PCNA staining in the same cells (Fig. 7E). Moreover, patients with NSCLC were divided into PLA signal-High and PLA signal-Low subgroups after cluster analyses. Kaplan–Meier survival analysis showed that patients with NSCLC with high phospho-IQGAP1 PLA signals had poor survival during follow-up periods (n = 63, PLA signal-High versus PLA signal-Low, P = 0.037; Fig. 7G). Collectively, our findings suggest that GLK promotes cell migration and cancer metastasis by direct binding to and phosphorylating IQGAP1 (Graphical Abstract).

Cell migration plays a critical role in cancer progression and cancer metastasis. Identification of signaling molecules that regulate cell migration should help development of novel therapeutic approaches for cancer metastasis. GLK overexpression is correlated with cancer recurrence of human lung cancer or hepatoma (34, 35). Here we report that GLK is a key kinase controlling IQGAP1-mediated cell migration and cancer metastasis. Our data showed that cell migration of primary lung epithelial cells was enhanced by GLK transgene but inhibited by GLK deficiency or IQGAP1 knockout. Moreover, distant metastasis of the lung cancer mouse model was significantly enhanced by GLK transgene, whereas GLK-induced distant metastasis of lung cancer was abolished by IQGAP1 knockout. Most importantly, the GLK-IQGAP1 complex was induced in tumor tissues of patients with NSCLC, and the number of the protein complex was correlated with poor survival of patients with NSCLC. These findings suggest that the GLK–IQGAP1 pathway is a therapeutic target for cancer metastasis or cancer recurrence.

A key finding of this study is that GLK induces IQGAP1-mediated cell migration by directly phosphorylating IQGAP1 at Ser-480 residue. It has been proposed that IQGAP1 activity may be regulated by autoinhibition through a C-terminal intramolecular interaction; PKCϵ-induced phosphorylation of IQGAP1 at Ser-1443 inhibits its intramolecular interaction to facilitate IQGAP1 function in cytoskeletal regulation (8). Whereas GLK phosphorylated IQGAP1 at Ser-480 located in the N-terminal region. A previous report indicates that IQGAP1 promotes cell motility by activating Rac1 and Cdc42 (53). The GLK-induced Ser-480 phosphorylation of IQGAP1 promoted Cdc42 activation and cell migration without affecting the interaction of IQGAP1 with Cdc42. Notably, the GLK-induced IQGAP1 Ser-480 phosphorylation did not regulate the activation of Rac1, which is responsible for directional/persistent migration instead of Cdc42-regulated random migration (54). Consistent with the selective regulation of Cdc42 by the GLK–IQGAP1 pathway, GLK-promoted cell migration of primary lung epithelial cells was nondirectional cell migration. Furthermore, IQGAP1 Ser-480 phosphorylation was indeed detectable in tumor tissues of human patients with NSCLC. These results indicate that IQGAP1 activity can be regulated by multiple phosphorylation events. Thus, IQGAP1 may interact with different sets of effector proteins in response to activation by distinct phosphorylation sites of IQGAP1.

Another important finding of this study is the direct binding of GLK to IQGAP1. Our data showed that two proline regions (Pro-436/437 and Pro-478/479) of GLK and the WW domain of IQGAP1 were required for the interaction between GLK and IQGAP1. Consistently, the transwell cell migration assays showed that overexpression of either GLK (P436/437;478/479A) mutant or IQGAP1 (ΔWW) mutant reduced the cell migration of lung cancer cells (Fig. 5F and G). Interestingly, using primary epithelial cells, we found that IQGAP1 KO abolished GLK transgene-induced cell migration in transwell migration assays (Fig. 3C and D); however, cotransfection of GLK (P436/437;478/479A) and IQGAP1 (ΔWW) did not efficiently abolish lung cancer cell migration. This phenomenon may be due to the homodimer formation (25, 55) between exogenous mutant GLK proteins, and endogenous wild-type GLK proteins or between exogenous mutant IQGAP1 proteins and endogenous wild-type IQGAP1 proteins. WW domains are divided into four groups (group I to group IV) with different binding preferences for proline-rich motifs (56): the group I binds Pro-Pro-X-Tyr motifs (where X is any amino acid); the group II binds Pro-Pro-Leu-Pro motifs; the group III binds polyproline motifs flanked by Arg or Lys; and the group IV binds phospho-Ser/Thr-Pro–containing motifs. Because of the ability of binding to each other's cognate ligands, the group II and III WW domains have been redefined as one single group II/III WW domain (57). Our data showed that GLK binding to the WW domain of IQGAP1 was mediated by two proline-rich regions Pro-436/437 (⁴³²PPPLPP⁴³⁷) and Pro-478/479 (⁴⁷⁷RPPPPR⁴⁸²) of GLK, which match to the recognition sequence of the group II and group III WW domains, respectively. This finding indicates that IQGAP1 contains the group II/III WW domain.

Our results of N-SIM time-lapse live-cell imaging showed that IQGAP1 displayed a polarized distribution in migratory lung cancer cells, and that GLK was highly colocalized with IQGAP1 at filopodia and lamellipodia of the polarized membrane during cell migration. The data support that GLK cooperates with IQGAP1 to promote cell migration. Previous reports indicate that IQGAP1 localizes at the leading lamellipodia of migrating cells and promotes cell motility by activating Rac1 and Cdc42 (53). Our results showed that GLK-induced IQGAP1 Ser-480 phosphorylation enhances Cdc42 activation and subsequent cell migration. Cancer cell migration contributes to cancer progression and cancer metastasis. Consistently, the extent of interaction between GLK and IQGAP1 was correlated with poor survival of patients with NSCLC. The lung cancer cells containing the GLK–IQGAP1 complex accumulated around tumor blood vessel, suggesting their tendency to extravasation into the tumor blood vessel. The distal metastatic carcinoma tissues also showed GLK–IQGAP1 complex–containing cells. Collectively, GLK may bind to and activate IQGAP1 at the leading edge of migrating cells, leading to Cdc42-mediated cell migration and cancer metastasis.

In conclusion, GLK plays a crucial role in promoting cell migration and cancer metastasis by directly binding to and phosphorylating IQGAP1. These findings suggest that the GLK–IQGAP1 complex is a potential therapeutic target for cancer recurrence.

No potential conflicts of interest were disclosed.

Conception and design: H.-C. Chuang, T.-H. Tan

Development of methodology: H.-C. Chuang, C.-C. Chang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.-C. Chuang, C.-C. Chang, C.-F. Teng, C.-H. Hsueh, L.-L. Chiu, P.-M. Hsu, M.-C. Lee, C.-P. Hsu, Y.-R. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-C. Chuang, C.-C. Chang, C.-F. Teng, C.-H. Hsueh, L.-L. Chiu, P.-M. Hsu, P.-C. Lyu, T.-H. Tan

Writing, review, and/or revision of the manuscript: H.-C. Chuang, C.-C. Chang, T.-H. Tan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.-C. Chang, Y.-C. Liu

Study supervision: H.-C. Chuang, T.-H. Tan

We thank members of Tan Lab for technical assistance, including Ms. Ching-Yi Tsai for IHC staining, PLA assay, and G-LISA activation assay, as well as Ms. Ting-Shuan Kung for mouse breeding and IHC staining. We thank the Transgenic Mouse Core (NHRI, Taiwan) for generation of transgenic and knockout mice. We thank the Core Facilities of National Health Research Institutes (NHRI, Taiwan) for tissue sectioning/H&E staining, confocal microscopy, live-cell imaging, and super-resolution N-SIM microscopy. We thank the Laboratory Animal Center (AAALAC accredited) of NHRI for mouse housing. We thank Institute of Biological Chemistry of Academia Sinica for mass spectrometry. We also thank Dr. Shao-Chun Hsu and the Imaging Core Facility of the Institute of Cellular and Organismic Biology, Academia Sinica for using the software Imaris (Version 9.1.2). This work was supported by grants from the National Health Research Institutes, Taiwan (IM-105-PP-01 and IM-105-SP-01 to T.-H. Tan) and Ministry of Science and Technology, Taiwan (MOST-106-2321-B-400-013 to T.-H. Tan). T.-H. Tan is a Taiwan Bio-Development Foundation (TBF) Chair in Biotechnology.

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.