An increase in integrin-linked kinase non-canonically confers NF-κB-mediated growth advantages to gastric cancer cells by activating ERK1/2

Background Increased activity or expression of integrin-linked kinase (ILK), which regulates cell adhesion, migration, and proliferation, leads to oncogenesis. We identified the molecular basis for the regulation of ILK and its alternative role in conferring ERK1/2/NF-κB-mediated growth advantages to gastric cancer cells. Results Inhibiting ILK with short hairpin RNA or T315, a putative ILK inhibitor, abolished NF-κB-mediated the growth in the human gastric cancer cells AGS, SNU-1, MKN45, and GES-1. ILK stimulated Ras activity to activate the c-Raf/MEK1/2/ERK1/2/ribosomal S6 kinase/inhibitor of κBα/NF-κB signaling by facilitating the formation of the IQ motif-containing GTPase-activating protein 1 (IQGAP1)–Ras complex. Forced enzymatic ILK expression promoted cell growth by facilitating ERK1/2/NF-κB signaling. PI3K activation or decreased PTEN expression prolonged ERK1/2 activation by protecting ILK from proteasome-mediated degradation. C-terminus of heat shock cognate 70 interacting protein, an HSP90-associated E3 ubiquitin ligase, mediated ILK ubiquitination to control PI3K- and HSP90-regulated ILK stabilization and signaling. In addition to cell growth, the identified pathway promoted cell migration and reduced the sensitivity of gastric cancer cells to the anticancer agents 5-fluorouracil and cisplatin. Additionally, exogenous administration of EGF as well as overexpression of EGFR triggered ILK- and IQGAP1-regulated ERK1/2/NF-κB activation, cell growth, and migration. Conclusion An increase in ILK non-canonically promotes ERK1/2/NF-κB activation and leads to the growth of gastric cancer cells. Electronic supplementary material The online version of this article (doi:10.1186/s12964-014-0069-3) contains supplementary material, which is available to authorized users.

The stimulation of cells by growth factors and cytokines as well as cellular interaction with ECM increase ILK activity [24]. In addition to the molecular regulation of PI3K/PTEN by ILK, Aoyagi et al. identified ILK as a new heat shock protein (HSP) 90 client protein and found that pharmacologically inhibiting HSP90 resulted in ILK degradation in a proteasome-dependent manner [25]. Furthermore, the HSP90-associated E3 ubiquitin ligase C-terminus of heat shock cognate 70 interacting protein (CHIP) causes ILK degradation [26]. Hashiramoto et al. demonstrated that HSP90 stabilized ILK and sustained AKT and ERK1/2 activation [16]. Thus, we speculate a relationship between ILK stability and the activation of its downstream kinases. Ras/MAPK pathway signaling is essential for tumorigenesis [27]. Increased ILK expression is related to high-grade gastric cancer [28], prostate cancer [29], and non-small cell lung cancer [30], although cells in these cancers commonly harbor Ras mutations [31][32][33]. Targeting ILK with siRNA decreases gastric cancer cell invasion, proliferation, and growth through an unknown mechanism [34]. Regarding the possibility that ILK acts upstream of NF-κB by regulating IKKα [13], which has been implicated in gastric tumorigenesis [35], ILK is speculated to activate cell growth through an NF-κBregulated pathway. Using gastric cancer cells (AGS, MKN45, and SNU-1), we studied the molecular regulation of ILK and identified a non-canonical pathway of ILK-regulated ERK1/2 activation for NF-κB-mediated gastric cancer cell growth, migration, and survival promotion.

Results
ILK activity and expression are essential for NF-κBmediated cell growth Increased activity or expression of ILK enhances tumorigenesis by promoting cell growth [6]. RNAi-based ILK silencing attenuates gastric cancer cell growth [34], whereas ILK overexpression is related to gastric tumorigenesis [28]. In human gastric tumors and AGS-derived nodules in BALB/c mice, Ki-67-positive proliferating cells coexpressed ILK as demonstrated by the fluorescence-based immunostaining ( Figure 1A) and AEC-based immunostaining (Additional file 1: Supplemental materials and methods; Additional file 2: Figure S1) experiments. To investigate the possible mechanisms underlying ILKmediated gastric cancer cell growth, several gastric epithelial cell lines were characterized according to their different cell growth rates, which were higher for the AGS and SNU-1 cells and lower for the MKN45 and GES-1 cells, and used in this study (Additional file 3: Figure S2A). Compared with the MKN45 cells, the AGS and SNU-1 cells also had elevated ILK expression (Additional file 3: Figure  S2B and S2C). A lentiviral-based shRNA was used to silence ILK genetically in the AGS, SNU-1, MKN45, and GES-1 gastric epithelial cells ( Figure 1B, upper panel) as well as in A549 and H1975 human lung adenocarcinoma cells, HK-2 human renal proximal tubular epithelial cells, and THP-1 human monocytic cells (Additional file 3: Figure S2D). In these cells, ILK silencing significantly (P <0.05) decreased cell growth ( Figure 1B; Additional file 3: Figure S2E). Furthermore, treating cells with the ILK inhibitor T315 [36] significantly (P <0.05) and dosedependently retarded cell growth ( Figure 1C) without cytotoxicity (data not shown). Additionally, decreased colony formation was observed in ILK-silenced AGS cells (Additional file 3: Figure S2F). Thus, gene silencing (Additional file 3: Figure S2G) and pharmacological methods (Additional file 3: Figure S2H) to suppress ILK activity or overexpression led to cell cycle arrest at the G 1 phase. These results show a growth-promoting role of ILK.
To characterize the features of ILK-regulated cell growth, NF-κB signaling was examined because ILK can act upstream of NF-κB by regulating IKKα [13]. By immunostaining, the coexpression of ILK and phosphorylated NF-κB (Ser536) was observed in human and mouse gastric tissues ( Figure 1D), and their coexpression significantly (P <0.01) and positively correlated with the number of proliferating cells, which is indicated by 55 Figure 1 ILK expression is necessary for cell growth and NF-κB activation. (A) Representative fluorescence-based immunohistochemical staining shows the coexpression of ILK (green) and Ki-67 (red) in human gastric tumors and AGS-derived nodules in BALB/c mice. DAPI (blue) was used for nuclear counterstaining. ILK + Ki-67 + cells in the immunofluorescently stained tissue sections are presented as dot-plots of a FACS-like analysis by using TissueQuest software. The cells were calculated as the percentage and the cell number of the total cells (DAPI + cells) per field. (B) Western blot of ILK expression in cells transfected with shRNAs targeting ILK (shILK) or a control luciferase (shLuc). β-actin was used as an internal control. WST-8-based assay shows the inhibition of cell growth in ILK-silenced cells. (C) The dose-dependent effect of the ILK inhibitor T315 on the inhibition of AGS cell growth. (D) Representative fluorescence-based immunohistochemical staining shows the coexpression of ILK (green) and phosphorylated NF-κB Ser536 (red) in human gastric tumors and AGS-derived nodules in mice. DAPI (blue) was used for nuclear counterstaining. The dot-plots show the coexpression of ILK and NF-κB Ser536. (E) Logistic regression analysis showed a correlation between ILK, phosphorylated NF-κB Ser536, and Ki-67 in 93 gastric cancer specimens tested. R-squared (R 2 ) and P values are shown. (F) EMSA demonstrating NF-κB activation. (G) Luciferase reporter assay shows the activation ratio of NF-κB to control Renilla luciferase in cells treated with the NF-κB inhibitor CAPE (25 μg/mL). (H) Growth of 25 μg/mL CAPE-treated AGS cells. (I) NF-κB activation in shLuc-or shILK-transfected cells. (J) NF-κB activation after 6 h of T315 treatment in AGS cells. For cell growth, colony formation, and luciferase activity, data are mean ± SD from three independent experiments. *P <0.05, **P <0.01, and ***P <0.001 compared with Day 0 or relative control. #P <0.05, ##P <0.01, and ###P <0.001 compared with shLuc.
triple-positive cases of the total 93 gastric cancer specimens ( Figure 1E; Additional file 4: Figure S3). Immunostaining for NF-κB nuclear translocation (Additional file 3: Figure  S2I), EMSA ( Figure 1F), and promoter assays ( Figure 1G) confirmed the constitutive activation of NF-κB in the AGS cells but not in the MKN45 cells. Treating cells with the NF-κB inhibitor CAPE significantly (P <0.001) reduced NF-κB activation ( Figure 1G) and cell growth ( Figure 1H). Either ILK silencing ( Figure 1I; Additional file 3: Figure S2J) or T315 treatment ( Figure 1J) significantly (P <0.05) stopped NF-κB activity. These results demonstrated that ILK is indispensable for cell growth in the cell lines tested because it facilitates NF-κB activation in gastric cancers.
ILK regulates Ras activity by facilitating the complex of IQGAP1-Ras to control MAPK-activated NF-κB Because AGS cells harbor PIK3CA and KRAS mutations [37], we examined possible regulatory effects of ILK on the modulation of NF-κB activity by these 2 kinases [38]. Using a Human Phospho-MAPK Array Kit, we identified 10 kinases that were more highly expressed in the AGS cells than in the MKN45 cells. These kinases mostly acted downstream of the PI3K and MAPK signaling pathways (Additional file 5: Figure S4A). By western blotting, we confirmed an increased phosphorylation of AKT, ERK1/2, and IκBα accompanied by IκBα degradation in the AGS cells ( Figure 2A). The pharmacological inhibition of c-Raf, MEK1/2, and PI3K significantly (P <0.05) reduced cell growth ( Figure 2B), IκBα phosphorylation (Ser32) and degradation ( Figure 2C), and NF-κB activity ( Figure 2D), indicating that both PI3Kand Ras-activating signaling pathways facilitated NF-κB activation. The effects of ILK have been widely studied because of its interactions with cell growth-and NF-κBassociated AKT [4,9]. Surprisingly, ILK silencing did not affect AKT and GSK-3β phosphorylation in the AGS and SNU-1 cells but markedly reduced c-Raf and ERK1/2 activation in all cells tested ( Figure 2E; Additional file 5: Figure S4B). Without AKT deactivation, we evaluated an alternative pathway for activating NF-κB through a mechanism involving MAPK/p90RSK/IκBα signaling [38]. The knockdown of ILK reduced the multiple phosphorylation of RSK (Thr573, Thr359/Ser363, and Ser380) and IκBα phosphorylation (Ser32) and increased IκBα accumulation ( Figure 2E). Inhibiting MEK1/2 caused similar effects (Additional file 5: Figure S4C) and a cell cycle arrest at the G 1 phase (Additional file 5: Figure S4D). A Ras pulldown assay revealed that inhibiting ILK caused Ras deactivation without affecting the stability of the Ras protein ( Figure 2F). These findings demonstrated a potential non-canonical pathway for ILK to modulate NF-κB by regulating Ras/c-Raf/MEK1/2/ERK1/2/IκBα signaling. ILK can modify ERK1/2 activation under cell growth and differentiation [22]; however, the molecular regulation related to Ras signaling has not been documented [2,3]. The coexpression of ILK and phosphorylated ERK1/2 (Tyr202/Thr204) was demonstrated in human gastric tumors and AGS-derived nodules in BALB/c mice ( Figure 2G). ILK interacts with IQGAP1 [7], a Ras GTPase-activating-like protein that is dissimilar from GAP, which transforms Ras to its inactive state. Because IQGAP1 controls Ras/MAPK signaling, it is oncogenic [39,40]. The expression of IQGAP family proteins was unchanged in the AGS and MKN45 cells, even after ILK silencing (Additional file 5: Figure S4E). However, silencing oncogenic IQGAP1, but not IQGAP3, (Additional file 5: Figure S4F-S4H) effectively inhibited c-Raf/ MEK1/2/ERK1/2/RSK signaling ( Figure 2H), NF-κB activity, and cell growth ( Figure 2I). By performing a coimmunoprecipitation assay, we demonstrated a potential complex harboring ILK, IQGAP1, and Ras ( Figure 2J). Immunostaining confirmed that ILK was coexpressed at levels similar to those of IQGAP1 and Ras ( Figure 2K; Additional file 6: Figure S5). Notably, silencing ILK disrupted the IQGAP1-Ras complex ( Figure 2L and 2M). These results demonstrated that ILK facilitated the formation of the IQGAP1-Ras complex to sustain Ras activity.
Enzymatic ILK modulates the formation of the IQGAP1-Ras complex and ERK1/2-mediated cell growth Our findings showed that pharmacologically inhibiting ILK decreased cell growth, indicating the essential role of the enzymatic activity of ILK. Further inhibiting ILK by the genetic approach disrupted IQGAP1-mediated Ras signaling. Inhibiting ILK activity pharmacologically with T315 ( Figure 3A) or genetically by transfecting the enzymatic mutant ILK A262V [41] ( Figure 3B) also disrupted the formation of the IQGAP1-Ras complex in the AGS cells. Three truncated regions of ILK ( Figure 3C) were overexpressed in the MKN45 cells to identify the domain essential for sustaining the IQGAP1-Ras complex. Coimmunoprecipitation assays demonstrated that overexpressed ILK constructs harboring PH and catalytic kinase domain regions immunoprecipitated with IQGAP1 and Ras ( Figure 3D). Therefore, only kinase domain-containing ILK activated ERK1/2 ( Figure 3E). Compared with the fulllength ILK (ILK 1-452 ), transfecting the kinase-dead mutant ILK A262V did not activate ERK1/2 ( Figure 3F). Additional experiments demonstrated that only kinase domain-containing ILK increased MEK1/2-regulated NF-κB activation ( Figure 3G) followed by MEK1/2-and NF-κB-regulated cell growth ( Figure 3H) in the MKN45 cells. These results showed that enzymatic ILK mediated the formation of the IQGAP1-Ras complex to trigger ERK1/2-and NF-κB-mediated cell growth.

PI3K activation and decreased PTEN expression facilitate ERK1/2/NF-κB activation by stabilizing ILK
Because ILK is dependent on PI3K-mediated PIP3 generation for its activation [4], growth factor-and integrinmediated PI3K activation or a PI3K mutation in the AGS cells [37] could contribute to ILK expression and activation. Compared with the growth of the MKN45 cells, the growth of the AGS cells was unaffected by elevated IL-6 levels [42], and the expression of β1 and β3 integrins did not increase (Additional file 7: Figure S6). We further confirmed that PIP3 generation was higher in the PI3K-mutated AGS cells ( Figure 4A). To explore the relationships among the PI3K, ILK, and Ras signaling pathways, the AGS cells were treated with the MEK1/2, PI3K, ILK, or Ras inhibitor for different durations. At 6 h after treatment, T315 slightly inhibited MEK1/2 and ERK1/2 phosphorylation (Additional file 8: Figure S7). At 24 h after treatment, PI3K inhibition disrupted ILK expression, which was followed by MEK1/ 2/ERK1/2 deactivation 24 h after treatment ( Figure 4B). However, Ras inhibition did not deactivate AKT or ILK in the AGS cells, although Ras can mediate PI3K activation [27]. Similar to the ILK-silenced AGS cells, both MKN45 and LY294002-stimulated AGS cells showed reduced Ras activity ( Figure 4C). The translation inhibitor cycloheximide facilitated LY294002-induced ILK downregulation ( Figure 4D) and the proteasome inhibitor MG132 reversed the aforementioned effect and ERK1/2 inactivation ( Figure 4E). The tumor suppressor phosphatase PTEN negatively regulates PI3K/PDK1/AKT signaling [10,11] and ILK activity [8,9], and the AGS cells exhibited a low expression of PTEN ( Figure 2A) [43]. The forced expression of PTEN in the AGS cells not only attenuated the constitutive phosphorylation of PDK1, AKT, and PAK1 but also inhibited ILK expression, ERK1/2 phosphorylation ( Figure 4F), and NF-κB activation ( Figure 4G). These results indicated that PI3K activation and decreased PTEN expression stabilize ILK to regulate ERK1/2/NF-κB activation.

Discussion
By using genetic and pharmacological approaches, we confirmed the proliferation-promoting role of ILK in vitro in gastric cancer cells. The function of ILK is highly related to NF-κB activation. An in vivo model of gastric cancer in mice showed the essential role of ILK in tumor growth [34], and increased ILK [28] and NF-κB [35] activity or expression is related to gastric tumorigenesis. Although the potent mechanisms for ILK-regulated cell proliferation have been previously documented [6], we further investigated the molecular basis of ILK-mediated gastric cancer cell growth, which is related to NF-κB activation. To the best of our knowledge, clinical observations in this study indicated a positive correlation between ILK expression, NF-κB activation, and cell proliferation in gastric cancers. Protection from apoptosis [13] and the stimulation of cell proliferation are oncogenic effects of ILK that are generally achieved by facilitating NF-κB activation. Based on these rationales, the molecular regulation of the ILK/NF-κB pathway was examined. According to the present study results (summarized in Figure 8), increased ILK activity or expression, which is controlled by PI3K/HSP90-mediated protein stabilization, triggers a non-canonical pathway of IQGAP/Ras/c-Raf/MEK1/2/ ERK1/2/RSK/NF-κB signaling to stimulate cell growth, migration, and survival.
Aberrant ILK is involved in tumorigenesis, and it is speculated that ILK canonically stimulates AKT to promote NF-κB-mediated oncogenic processes, such as antiapoptosis [13] and survival [14]. In tumorigenesis, NF-κB can promote survival and proliferation, angiogenesis, adhesion/invasion/metastasis, and inflammation. However, the mechanisms underlying NF-κB activation in cancer cells are multifaceted; both AKT and MAPKs are crucial for NF-κB activation [38]. Of note, AGS cells have mutated PIK3CA and KRAS [37]. Inhibiting PI3K and Ras activity attenuated NF-κB activation. We therefore hypothesized a role for ILK to canonically or non-canonically act upstream of AKT and ERK1/2, respectively. ILK-mediated AKT phosphorylation at Ser473 was confirmed in lung adenocarcinoma A549 cells but not in gastric cancer cells. Surprisingly, ILK silencing effectively negatively affected ERK1/2 activation in all cells tested. In addition to the kinase activity of ILK, scaffold functions of ILK associated with intracellular molecules also need further investigation to verify its roles in regulating AKT and ERK1/2 signaling. ERK1/2 mediates NF-κB activation through a mechanism involving MAPK/p90RSK/IκBα signaling [38]; similarly, we demonstrated that this pathway was required for ILKmediated NF-κB activation in gastric cancer cells. ILK has a possible upstream role in ERK1/2 activation [16,22,23,48], but no rational mechanisms exist for examining this regulation [2,3].
An emerging and widely demonstrated role for IQGAP1 is its control over diverse biological functions by interacting with various cellular factors [53]. For MAPK signaling, IQGAP1 can directly regulate Ras/ERK1/2 activation during tumorigenesis [39,40]. Ras is a small GTPase that can hydrolyze GTP into GDP in a GTP-bound protein to inactivate the protein, a process accelerated by GAPs, and this process can be reversed with a guanine nucleotide exchange factor to reactivate the Raf/MAPK pathway [27]. IQGAP is a Rho-GTP-binding protein and has a region similar to that of Ras GAP. However, IQGAP has no GAP function but can stabilize the GTP-bound protein in an activated state [54,55]. A global analysis of the ILK interactome has shown a strong ILK-IQGAP1 interaction [7]; however, no studies have illustrated the axis of ILK/ IQGAP1/Ras signaling. Wickstrom et al. demonstrated an ILK-IQGAP1 association through the non-ANK repeats of ILK and the IQ motif of IQGAP1 for facilitating integrin signaling [56]. To our knowledge, the present study is the first to show that increased or decreased activity or expression of ILK did not affect IQGAP1 expression, although silencing ILK disrupted the IQGAP1-Ras interaction. These three proteins formed a novel complex around the cell membrane, particularly in cell-cell junctions. Regarding IQGAP1 sustaining Ras activity in gastric cancer cells, IQGAP1 but not IQGAP3 mediated cell growth through an ERK1/2-regulated NF-κB activation pathway. The present study reveals a novel axis of ILK/ IQGAP1 for Ras/ERK1/2 signaling; however its cellular significance requires further investigation. The abnormal expression of IQGAP1 is related to poor prognosis in gastric cancers [57]. The formation of an ILK-IQGAP1 complex may have multiple biological effects; IQGAP1 has been implicated in diverse cellular functions through a mechanism involving the interaction of cytoskeletal components, small GTPases, kinases, and receptors [53].
Besides using the genetic approach, pharmacologically inhibiting ILK with T315 or an ILK kinase-dead mutation (ILK A262V ) abolished the IQGAP1-Ras interaction and thus deactivated ERK1/2. The enzymatic activity of ILK is speculated to be crucial for the axis of ILK/ IQGAP1/Ras signaling. Further results showed the importance of the PH-like domain and the kinase domain in mediating the formation of ILK/IQGAP1 complex as well as ERK1/2 activation, NF-κB activation, and cell growth. The mechanism underlying the ILK-mediated IQGAP1-Ras interaction is unknown; ILK phosphorylates its substrates as well as acts as an adaptor for protein-protein interactions [2,3]. Together with the findings that the non-ANK repeats of ILK are required for binding with the IQ domain of IQGAP1 [53,56], it is reasonable to hypothesize that ILK initiates the phosphorylation of the IQGAP1-Ras complex and/or that ILK confers an adaptor-mediated interaction for this complex.
Although increased ILK expression contributes to tumorigenesis, the mechanisms for ILK overexpression remain unknown. ILK upregulation and activation are related to β1 integrin signaling [1], and integrins may regulate gastric tumorigenesis, particularly by modulating adhesion and metastasis [58,59]. Thus, integrin overexpression causes ILK upregulation and activation. However, in the gastric cancer cell lines AGS and MKN45 that we tested, there were no differences in the expression or activation of β1 or β3 integrin. In addition to integrin signaling, growth factors generally activate PI3K for cell growth advantages. The expression of cell growth-associated IL-6 is higher in the AGS cells than in the MKN45 cells; however, its potential role in gastric cancer cell proliferation has been excluded. In addition to PIK3CA mutations [37], growth factors involved in regulating ILK expression need further investigation.
Surprisingly, pharmacologically inhibiting PI3K decreased ILK expression along with ERK1/2 deactivation. This result indicates an upstream role for PI3K-regulated ILK stabilization linked to Ras/ERK1/2 signaling. In the AGS cells, ILK overexpression may result from a natural mutation in PIK3CA that generates PIP3 to stabilize ILK expression. The most important finding in this study is the mechanism of ILK-dependent ERK1/2 activation demonstrated in the AGS cells, although these cells harbor PI3CA and KRAS mutations [37]. ILK may be required to sustain Ras activity, although Ras is automatically activated. The PH-like domain is essential for ILK to interact with PIP3 [4]. Once PI3K-driven PIP3 generation mediates ILK recruitment, ILK signaling can be initiated after protein stabilization. It is speculated that aberrant PI3K activation can cause ILK upregulation. The level of PTEN, a tumor suppressor related to PIP3 downregulation, is markedly decreased in several cancers, including gastric cancers. Restoring PTEN expression in the AGS cells effectively abolished both AKT and ERK activation and decreased ILK expression. PI3K/PTEN/PIP3-mediated ILK stabilization is therefore important for ERK1/ 2 activation. PDK1 activates PAK1 [60], and PAK1 phosphorylates ILK [61]. Here, the pharmacological inhibition of PI3K inhibited PDK1 and PAK1 activation (Additional file 14: Figure S13A); however, PAK1 knockdown did not alter ILK expression but decreased c-Raf/ERK1/2 phosphorylation, NF-κB activation, and cell growth (Additional file 14: Figure S13B-13D). These results indicated that PI3Kcontrolled PAK1 does not contribute to ILK expression. HSP90 is a chaperone that associates with client proteins [62][63][64][65]. HSP90 regulates ILK stability [25] through an interaction at amino acids 377-406 within the kinase domain of ILK. HSP90 can maintain the tumor-like character of rheumatoid synovial cells by stabilizing ILK, ERK, and AKT [16]. However, the potential regulation of ILK/ERK/ AKT signaling by HSP90 remains undetermined. Our study further demonstrated that the HSP90-associated E3 ligase CHIP was required for ILK stabilization, whereas inhibiting PI3K and HSP90 caused CHIP-proteasomemediated ILK degradation. Radovanac et al. demonstrated that ILK stability is negatively regulated by CHIP ubiquitination in fibroblasts when HSP90 is inhibited [26]. The molecular stabilization of ILK by HSP90 and CHIP remains unclear, but HSP90 interacts with the kinase domain of ILK [25,26], and CHIP-mediated ubiquitylation occurs at the ANK repeats of ILK [26]. CHIP regulates ILK stability in conjunction with others, and this regulation is important for ILK-regulated ERK1/2 and NF-κB activation and cell growth. Regarding the multiple oncogenic effects of ILK, the results of this study and those of previous studies reveal the molecular mechanism of ILK stabilization and expression; furthermore, we showed that ILK promotes cancer cell growth, migration, and survival responses through ERK1/2-regulated NF-κB activation.
The limitation of this study is that the changes observed were in the AGS cells, which have PIK3CA and KRAS mutations [37] and decreased PTEN expression. To confirm the finding that ILK facilitates ERK1/2 activation in response to physiological and pathological stimuli, we showed that hydrogen peroxide and H. pylori infection caused ILK-regulated ERK1/2 activation. In addition, the exogenous administration of EGF confirmed the requirement of ILK and IQGAP1 in EGF-induced ERK1/2 activation, NF-κB activation, cell growth, and migration [47,48]. Crosstalk between EGFR and integrins facilitates gastric cancer cell invasion and proliferation [66]. Moreover, the overexpression of WT EGFR and EGFR mutated at domain VIII may induce PI3K/HSP90-regulated ILK stabilization followed by ERK1/2 activation [48,52]. Regarding the critical role of Ras/ERK1/2 in cell growth advantages [27], the aforementioned results elucidate the significance of ILK in cell growth advantages, at least in gastric cancers, through a non-canonical mechanism involving the facilitation of ERK1/2 activation.

Conclusion
Based on the finding that PI3K/PTEN/HSP90-regulated ILK upregulation induces non-canonical IQGAP1/Ras/ ERK1/2-mediated NF-κB activation and growth advantages in gastric cancer cells, targeting this pathway may be beneficial when used in combination with other anticancer agents. Our hypothesis requires further in vivo investigation with an appropriate animal model.

Gastric cancer specimens
In this study, 150 cores, including reactive and cancerous (different grades and stages) tissues of the stomach, were obtained from a commercial tissue microarray (BioChain Institute, Inc., Hayward, CA, USA) built for immunohistochemical interrogation. AGS-derived gastric nodules in BALB/c mice were accordingly arrayed [67]. For animal studies, 6-to 8-week-old male wild-type BALB/c mice were purchased from Jackson Laboratory (Bar Harbor, ME, USA) and fed standard laboratory food in the Laboratory Animal Center of National Cheng Kung University. The animals were raised and handled according to the guidelines established by the National Science Council, Taiwan. The experimental protocols adhered to the rules of the Animal Protection Act of Taiwan and were approved by the Laboratory Animal Care and Use Committee of National Cheng Kung University.
Cell lines and cell culture AGS (CRL-1739, ATCC; derived from a biopsy specimen of an untreated human gastric adenocarcinoma harboring KRAS, PIK3CA, CDH1, and CTNNB1 mutations), metastatic MKN45 (JCRB0254, The RIKEN Cell Bank, Japan; established from a poorly differentiated adenocarcinoma of the medullary type from the stomach of a 62year-old woman bearing an E-cadherin mutation), and SNU-1 human gastric adenocarcinoma cells (CRL-5971, ATCC; derived from a metastatic ascites site of a poorly differentiated primary stomach carcinoma harboring KRAS and MLH1 mutations), and GES-1 gastric epithelial immortalized cells, kindly provided by Dr. Pei-Jung Lu, National Cheng Kung University, were routinely grown in plastic cell culture dishes in Ham's F-12 nutrient mixture, DMEM, or RPMI 1640 (F-12, RPMI; Invitrogen Life Technologies, Rockville, MD, USA) with L-glutamine and 15 mM HEPES supplemented with 10% heat-inactivated fetal bovine serum (FBS; Invitrogen Life Technologies), 50 units of penicillin, and 50 μg/mL of streptomycin and maintained in a humidified atmosphere with 5% CO 2 and 95% air.

Immunohistochemical/immunocytochemical staining
Tissue blocks were fixed overnight at 4°C with 4% neutral buffered paraformaldehyde solution, dehydrated, cleared with HistoClear II (National Diagnostics, Atlanta, GA, USA), and embedded in wax. For immunohistochemical staining, tissue sections were deparaffinized, rehydrated, incubated with 3% H 2 O 2 in methanol for 15 min, and subjected to heat-induced antigen retrieval by boiling for 10 min in 0.01 M citric acid. For immunocytochemical staining, cells were fixed in 3.7% formaldehyde in PBS for 10 min. After washing twice with PBS, the tissue sections and cells were mixed with primary antibodies in antibody diluents (Dako Corporation, Carpinteria, CA, USA) and incubated overnight at 4°C. The following day, samples were washed with PBS and incubated with or without HRP-or fluorescence-labeled secondary antibodies at room temperature for 1 h. For immunohistochemistry, antibodies against ILK, Ki-67, pNF-κB, NF-κB, and pERK1/2 were used. HRP-reactive sections were washed with PBS, developed with an AEC substrate, counterstained with hematoxylin, and visualized using a microscope (IX71; Olympus, Tokyo, Japan). For confocal microscopy, DAPI (5 μg/mL) was used for nuclear staining. The sections were then visualized using a confocal laser scanning microscope (Digital Eclipse C1si-ready; Nikon, Tokyo, Japan). We alternatively used the Tissue-FAXS system (TissueGnostics, Vienna, Austria) to analyze the immunostaining of the tissues. In HistoQuest software (TissueGnostics, Tarzana, CA, USA), we plotted the X and Y axes for the expression of the indicated proteins. Therefore, the upper-right region of dot-plot shows double-positive expression but not coexpression. The stained tissue microarray slides were digitized using the TissueFAXS system. HistoQuest software was used for detecting and quantifying stained regions. Two markers, the AEC master marker (ILK, pNF-κB, or Ki-67) and the hematoxylin non-master marker, were analyzed and calculated.

Cell viability and cytotoxicity assays
To measure cell growth, cell viability was determined using a colorimetric assay (Cell Counting Kit-8; Dojindo Molecular Technologies, Kumamoto, Japan) according to the manufacturer's instructions. A microplate reader (SpectraMax 340PC; Molecular Devices Corporation, Sunnyvale, CA, USA) was used to measure the absorbance at 450 nm, and the data were analyzed using Softmax Pro software (Molecular Devices Corporation). The relative growth rate was calculated by normalization to the control group. To evaluate cell damage, lactate dehydrogenase (LDH) activity was determined using a colorimetric assay (Cytotoxicity Detection Kit; Roche Diagnostics, Lewes, UK) according to the manufacturer's instructions. Aliquots of culture media were transferred to 96-well microplates. SpectraMax 340PC was used to measure the absorbance at 620 nm with a reference wavelength of 450 nm, and the data were analyzed using Softmax Pro software.

Western blotting
Harvested cells were lysed in a buffer containing 1% Triton X-100, 50 mM Tris (pH 7.5), 10 mM EDTA, 0.02% NaN 3 , and a protease inhibitor cocktail (Roche Boehringer Mannheim Diagnostics, Mannheim, Germany). After a freeze-thaw cycle, cell lysates were centrifuged at 10,000 × g at 4°C for 20 min. The lysates were boiled in a sample buffer for 5 min. Proteins were then subjected to SDS-PAGE and transferred to PVDF membranes (Millipore, Billerica, MA, USA) using a semi-dry electroblotting system. After blocking with 5% skim milk in PBS, the membranes were incubated overnight with a 1:1,000 dilution of primary antibodies at 4°C. The membranes were then washed with 0.05% PBS-Tween 20 and incubated with a 1:5,000 dilution of HRPconjugated secondary antibody at room temperature for 1 h. After washing, the membranes were soaked in ECL solution (PerkinElmer Life and Analytical Sciences, Inc., Boston, MA, USA) for 1 min and exposed to an Xray film (BioMax; Eastman Kodak, Rochester, NY, USA). The relative signal intensity was quantified using ImageJ software (version 1.41o; W. Rasband, National Institutes of Health, Bethesda, MD, USA). The changes in the ratio of proteins compared with the normalized value of untreated cells (indicated protein/β-actin or phosphorylated protein/total protein/β-actin) are also determined. One set of representative data obtained from three independent experiments is shown and the data shown as the mean ± SD values from three independent experiments (Additional file 15: Figure S14).

Lentiviral-based RNAi transfection
Protein expression was downregulated using lentiviralbased short hairpin RNA (shRNA) targeting the indicated sequences of the different genes, as summarized in Additional file 16: Table S1. Luciferase shRNA (shLuc) was used as a negative control. shRNA clones were obtained from the National RNAi Core Facility, Institute of Molecular Biology and Genomic Research Center, Academia Sinica, Taipei, Taiwan. Lentiviruses were produced by the RNAi Core Facility, National Cheng Kung University. Cells were transduced by lentiviruses with an appropriate multiplicity of infection in complete growth medium supplemented with polybrene (Sigma-Aldrich). After transduction for 24 h and puromycin (Calbiochem, San Diego, CA, USA) selection for 3 days, protein expression was monitored by western blotting. IQGAP1 expression was silenced using the commercial siRNA IQGAP1-HSS113014, containing the following siRNA target sequences: 5′-UUUAGCUGCAGGAAUCUGUA GGGCC-3′ and 5′-GGCCCUACAGAUUCCUGCAGC UAAA-3′ (Invitrogen). Transfection was performed by electroporation by using a pipette-type microporator (Microporator system; Digital Bio Technology, Suwon, Korea). After transfection, cells were incubated for 18 h in RPMI 1640 at 37°C before infection. A nonspecific scrambled siRNA (Stealth RNAi ™ siRNA Negative Control Kit, 12935-100; Invitrogen) was used as the negative control.

Luciferase reporter assay
To analyze NF-κB promoter activity by a luciferase reporter assay, transient transfection was performed using the GeneJammer transfection reagent (Stratagene, La Jolla, CA, USA). In short, cells were cotransfected with 0.2 μg of an NF-κB-promoter-driven firefly luciferase reporter and 0.01 μg of a Renilla luciferase-expressing plasmid (pRL-TK; Promega). Twenty hours after transfection, the cells were lysed and harvested for firefly and Renilla luciferase using the Dual-Glo luciferase assay system (Promega). For each lysate, firefly luciferase activity was normalized to Renilla luciferase activity to assess transfection efficiencies.

Ras pull-down assay
Ras activation assays were performed according to the affinity precipitation protocol provided by the manufacturer (Ras Activation Assay Kit 28820; Single Oak Drive, Temecula, CA, USA). Cell lysates were incubated with Raf-1 RBD for 45 min at 4°C and centrifuged to pellet agarose beads. The agarose beads were washed, and the pellets were resuspended in 2× Laemmli sample buffer and boiled for 5 min. The supernatant was collected, and cellular proteins were resolved by SDS-PAGE and analyzed by immunoblotting.

Coimmunoprecipitation
For coimmunoprecipitation, 100 μg cell lysate from cells was incubated overnight at 4°C with 5 μg protein G (Amersham Biosciences, Uppsala, Sweden) and 2 μg of antibodies. The expression of interacting proteins was determined by Western blotting.

PI3K activity assay
A PIP3 Mass ELISA (K-2500 s, Echelon Biosciences, Salt Lake City, UT, USA) was performed to detect PI3K activity in cells according to the manufacturer's instructions.

Wound healing assay
For cell migration assays, the confluent monolayers of cells were wounded by scraping a pipette tip across the monolayer. The cells were washed with PBS and incubated with appropriate media. Images at 100× magnification were taken at wounding and 12 h later by using a microscope (IX71; Olympus). Cell migration was assessed using ImageJ software and was reported as the number of cells that migrated into the scraped region.

Apoptosis assay
Apoptosis was assessed using nuclear propidium iodide (PI; Sigma-Aldrich) staining and flow cytometry (FACS-Calibur; Becton Dickinson, San Jose, CA, USA) with excitation at 488 nm and emission in the FL2 channel (565-610 nm). Samples were analyzed using CellQuest Pro 4.0.2 software (Becton Dickinson), and quantification was performed using WinMDI 2.8 software (The Scripps Institute, La Jolla, CA, USA). Apoptosis levels were reported as the percentages of sub-G 1 phase cells.

Statistical analyses
Values were expressed as mean ± standard deviation (SD). Significant differences between groups were assessed using one-way ANOVA followed by Dunnett's post hoc test as appropriate, Student's t test, or analysis of variance. Analyses were performed using GraphPad Prism 4 software (GraphPad Software Inc., La Jolla, CA, USA). After the densities of the detected proteins were quantified using HistoQuest software, logistic regression analysis of the graded expression of ILK, phosphorylated NF-κB Ser536, and Ki-67 in 93 gastric cancer specimens tested was performed. After the AEC-based immunohistochemistry assay, no AEC staining was considered negative, <25% staining was considered grade 1, between 25% and 50% staining was considered grade 2, and >50% staining was considered grade 3. Data analysis was performed using multinomial logistic regression and the pseudo Rsquared test in GraphPad Prism 4 software (GraphPad Software Inc.). The exact P-values are listed in the corresponding figure legends. Statistical significance was set at P <0.05.

Supplemental materials and methods
The additional information was also attached (Additional file 1: Supplemental materials and methods).

Additional files
Additional file 1: Supplemental materials and methods.
Additional file 2: Figure S1. Expression of ILK in gastric tumors. The representative AEC-based immunohistochemical staining of ILK and Ki-67 in 2 gastric tumor patients. Hematoxylin is used for nuclear counterstaining.