The MAP3K ZAK, a Novel Modulator of ERK-Dependent Migration, Is Upregulated in Colorectal Cancer

ORIGINAL ARTICLE The MAP3K ZAK, a novel modulator of ERK-dependent migration, is upregulated in colorectal cancer

C Rey1, B Faustin2, I Mahouche1, R Ruggieri3, C Brulard1, F Ichas4, I Soubeyran5, L Lartigue1,6 and F De Giorgi4,6

Often described as a mediator of cell cycle arrest or as a pro-apoptotic factor in stressful conditions, the MAP3K ZAK (Sterile alpha motif and leucine zipper-containing kinase) has also been proven to positively regulate epidermal growth factor receptor (EGFR) and WNT signaling pathways, cancer cell proliferation and cellular neoplastic transformation. Here, we show that both isoforms of ZAK, ZAK-α and ZAK-β are key factors in cancer cell migration. While ZAK depletion reduced cell motility of HeLa and HCT116 cells, its overexpression triggered the activation of all three mitogen-activated protein kinases (MAPKs), extracellular signal-regulated kinase (ERK), c-JUN N-terminal kinase (JNK) and p38, as well as an increase in cell motion. On the contrary, the kinase-dead mutants, ZAK-α K45M and ZAK-β K45M, were not able to provoke such events, and instead exerted a dominant-negative effect on MAPK activation and cell migration. Pharmacological inhibition of ZAK by nilotinib, preventing ZAK-autophosphorylation and thereby auto-activation, led to the same results. Activated by epidermal growth factor (EGF), we further showed that ZAK constitutes an essential element of the EGF/ERK-dependent cell migration pathway. Using public transcriptomic databases and tissue microarrays, we ﬁnally established that, as strong factors of the EGFR signaling pathway, ZAK-α and/or ZAK-β transcripts and protein(s) are frequently upregulated in colorectal adenoma and carcinoma patients. Notably, gene set enrichment analysis disclosed a signiﬁcant correlation between ZAK+ colorectal premalignant lesions and gene sets belonging to the MAPK/ERK and motility-related signaling pathways of the reactome database, strongly suggesting that ZAK induces such pro-tumoral reaction cascades in human cancers.

Oncogene (2016) 35, 3190–3200; doi:10.1038/onc.2015.379; published online 2 November 2015

INTRODUCTION constitutes the cellular basis for invasion and metastatic The mitogen-activated protein kinase (MAPK) signaling cascade dissemination of cancer cells, it remains of crucial importance to consists of a three-step activation of protein kinases that permit better understand the ERK-migrating signaling cascade. integration and transmission of numerous extracellular signals to ZAK (sterile alpha motif and leucine zipper-containing kinase) control cell response to environmental stress, growth factors and belongs to a subfamily of MAP3Ks referred to as mixed-lineage inflammatory cytokines. These signals thereby regulate vital cell kinases (MLKs). Two differential splice products of ZAK have been 8 9 functions such as cell growth, proliferation, migration, differentia- described, namely ZAK-α (also called MLTK-α or MRK-α ) and 8 9 10 tion and death.1,2 In such pathways, signals basically transit from ZAK-β (also known as MLTK-β, MRK-β and MLK-7). Identical upstream MAP3Ks to MAP2Ks that, in turn, activate the four in the N-terminus (amino-acid, aa 1–331), ZAK-α and ZAK-β classical downstream MAPKs: ERK (extracellular signal-regulated sequences completely diverge in the C-terminal region: the kinase 1/2), p38, c-JUN N-terminal kinase (JNK) or ERK5.3 Each of longest isoform ZAK-α (800 aa) has a sterile-α motif domain in these modules has shown some specificity: ERK is generally this region, which is not present in ZAK-β (455 aa).10,11 ZAK is responsive to mitogenic and differentiation stimuli, and JNK activated by phosphorylation, but until now, only one real and p38 are responsive to external and genotoxic stresses.2 upstream and/or scaffold partner has been identified for ZAK-α, Nonetheless, a high degree of redundancy and interconnections the protein kinase N α.12 Protein kinase N α enzymatic activity is exist inside and between these signaling pathways, revealing that stimulated by fatty acids (such as arachidonic acid) and Rho even more complicated regulatory mechanisms regulate cell fate GTPases.13 Interestingly, a recent report has also shown that ZAK-β and behavior. Alternative spliced variants, external factors and the activation downstream of lysophosphatidic acid and Rho C leads countless substrates of the various MAPKs add to this complexity.3 to an ERK- and p38-dependent cell-invasive process.14 This has been widely reported for ERK, a kinase for which 4200 ZAK intracellular function has been widely investigated in substrates have been indexed.4 Thus, in addition to its role in cell cardiomyocytes, where the kinase seems to have a prominent role proliferation, ERK has been reported to be involved in cell motility in cardiac cell hypertrophy.10,15,16 In cancer, however, its role has through focal adhesion kinase (FAK), calpain or myosin light chain just begun to be investigated. Although ZAK has been shown – kinase (MLCK)5,6 and more recently via the phosphatase STYX to have antitumoral properties,11,16 18 recent evidence supports (serine/threonine/tyrosine-interacting protein) and the transcrip- pro-oncogenic functions. Cho et al.19 reported that increasing tion factor Slug.7 Considering that directional cell migration ZAK-α expression level in JB6 Cl41 skin epidermal cells promotes

1Laboratoire de Validation et Identification de Nouvelles Cibles en Oncologie (VINCO), INSERM U916, Institut Bergonié, Université Victor Segalen Bordeaux 2, Bordeaux, France; 2CNRS – CIRID, Bordeaux Cedex, France; 3Laboratory of Cell Signaling, The Feinstein Institute for Medical Research, Manhasset, NY, USA; 4Fluofarma, 2 rue R. Escarpit, Pessac, France and 5Département de Biopathologie, Institut Bergonié, Bordeaux, France. Correspondence: Dr L lartigue, Laboratoire de Validation et Identification de Nouvelles Cibles en Oncologie (VINCO), INSERM U916, Institut Bergonié, Université Victor Segalen Bordeaux 2, 229 cours de l’Argonne, Bordeaux 33 076, France. E-mail: [email protected] or [email protected] 6These authors contributed equally to this work. Received 4 April 2015; revised 30 August 2015; accepted 4 September 2015; published online 2 November 2015 The MAP3K ZAK controls ERK-dependent cell migration C Rey et al 3191 their neoplastic transformation post-epidermal growth factor p38 and phospho-JNK in comparison with control cells in both (EGF) or 12-O-tetradecanoylphorbol-13-acetate (TPA or PMA) conditions (Figures 2a and b), without altering total MAPKs treatment, as well as tumor formation in nude mice. More expression (Figure 2c). Interestingly, ZAK was also able to activate recently, it has also been identified as a gene preferentially ERK in the absence of serum (Figures 2b and c). In contrast to native overexpressed in gastric, breast, bladder and colorectal cancers proteins, ZAK-α and ZAK-β kinase-dead mutants’ overexpression (CRCs)20 and one of the splice gene candidates through which did not trigger any MAPK activation (Figure 2d and Supplementary PRPF6 would favor colorectal tumor growth.21 In that respect, ZAK Figure 5a). They instead behaved like dominant-negative mutants, figured among the eight genes found to be able to drive both decreasing the endogenous level of phospho-p38, phospho-JNK proliferation and Wnt signaling pathway activation in CRC cells.22 and phospho-ERK proteins. This effect not only highlighted the At last, a computational approach investigating a directed protein ability of endogenous ZAK to regulate these MAPKs, but certainly interaction network recently revealed the kinase as a strong also explained why the cells carrying these mutants migrate more positive modulator of the epidermal growth factor receptor (EGFR) slowly than their WT counterparts (Figure 1h and Supplementary signaling cascade,23 strengthening the possibility that ZAK Figure 3c). It is also to be noted that WT ZAK-α and ZAK-β constitutes a key factor of this pro-oncogenic pathway. overexpression resulted in the appearance of two bands on western In this study, we evaluated the role of ZAK-α and ZAK-β in blot membranes incubated with anti-ZAK antibody (Figure 2d, cancer progression, and its interconnection with the EGFR pathway. arrows columns 2 and 3). As shown in Figure 2d and Our results show that both ZAK isoforms regulate cancer cell Supplementary Figure 4, the upper band either disappeared after migration in a kinase-dependent manner. Most importantly, a mutation of ZAK-α and ZAK-β kinase domain (Figure 2d, columns we showed that ZAK regulates endogenous ERK activity through 4 and 5) or a phosphatase treatment (Supplementary Figure 4), which it exerts its pro-migratory effects. Public transcriptomic which presumably indicates that overexpressing ZAK results in its database, as well as tissue microarrays (TMA) analyses, further autophosphorylation and, in turn, its activation. These results were revealed ZAK-α and/or ZAK-β transcripts and proteins as being confirmed by confocal microscopy using HCT116, HT-29 and HeLa frequently upregulated in colorectal adenoma and carcinoma (CRC). cells (Figures 2e and f and Supplementary Figures 5b and c), Interestingly, when compared with the rest of the cohort by gene implying that ZAK is a central regulator of MAPK signaling in set enrichment analysis, ZAK-overexpressing patients presented a cancer cells. co-activation of the MAPK/ERK and motility-related signaling pathways, supporting a role of ZAK in activating these pro- Pharmacological inhibition of ZAK prevents its tumoral cascades in vivo. autophosphorylation and ability to induce MAPK activation and cell migration in HCT116 cells RESULTS It has previously been shown that nilotinib (Tasigna), a second- generation compound targeting BCR–ABL in chronic myeloid ZAK controls cancer cell motility 24–26 leukemia, binds ZAK with great affinity (K 8–10 nM) in vitro. α β d In order to clarify ZAK- and - intracellular activity in cancer cells, Nilotinib may thus be a pharmacological tool able to inhibit ZAK. fi we designed a series of isoform-speci c small interfering RNA To address this question, overexpressing ZAK-α or ZAK-β HCT116 (siRNA; Figure 1a and Supplementary Figure 1) and investigated cells and controls were placed in low serum conditions and the impact of their individual or combined loss on various cellular incubated with an increasing amount of nilotinib (0.3–3 μM) for α β functions. As reported in Figures 1b and c, ZAK- or ZAK- 24 h. As shown by laser scanning cytometry analysis, ZAK’s ability depletion did not have a strong effect on cell proliferation (that is, to induce phospho-c-JUN activation progressively decreased proliferation index comprised between 3.3 and 3.7 at 72 h for all upon drug treatment, and was fully abrogated at 3 μM (Figure 3a). samples) and cell death. Given this restricted effect, we then These data were strengthened by western blot studies, revealing considered ZAK to be involved in cell migration and challenged that the addition of 1 or 3 μM nilotinib totally abolished ZAK’s the cells in a wound-healing scratch test using the IncuCyte ability to increase phospho-ERK, phospho-p38 and phospho-JNK real-time imaging system (Figures 1d and e). Interestingly, level in CRC cells (Figure 3b). Interestingly, as observed with the knocking-down ZAK-α or ZAK-β reduced cell migration from 20 kinase-dead mutant K45M, no phosphorylated upper band of ZAK to 40%; whereas the combined loss of the two proteins dropped was visible on the membranes post-nilotinib treatment (white arrows). cell mobility by 60% (Figures 1d and e and Supplementary Thus, nilotinib may directly bind ZAK and prevent Figure 2). To confirm this result, HCT116 cells were transiently the process of autophosphorylation/activation induced by its over- transfected with pReceiver-ZAK-α and pReceiver-ZAK-β plasmids expression. Given these results, we then checked whether nilotinib to see whether an overexpression of ZAK would instead favor cell would annihilate ZAK intracellular function. HCT116 cells were migration (Figures 1f and g). Faint, but significant, pro-migratory transfected with pReceiver-ZAK-α and/or -ZAK-β and treated with effects were detected for ZAK-β. However, increasing ZAK-α an increasing amount of drug before cell migration monitoring protein level greatly enhanced HCT116 cell motility. Contrary to (Figure 3c, Supplementary Figure 6). At 1 μM, nilotinib markedly their WT counterparts, none of the ZAK-α and ZAK-β kinase-dead decreased ZAK-α and ZAK-α/β-overexpressing cells migration. Inter- mutants, that is, ZAK-α K45M and ZAK-β K45M, were able to estingly, although ZAK-β only caused a modest rise in cell migration, promote cell migration (Figure 1h). They even had a slight these cells exhibited a notable sensitivity to nilotinib treatment and tendency to lessen it. Similar effects were observed in HeLa cells, displayed a slower motion at drug doses as low as 0.3 μM. in which the EGFR pathway is not altered and EGF was thus used as a control (Supplementary Figure 3). EGF pathway regulates ZAK expression and activation First described as a gene candidate whose expression was ZAK controls the endogenous activity of several MAPKs, including induced by the ERK/MAPK pathway,8,23 we then evaluated ERK in KRAS-mutated HCT116 cells whether ZAK would respond to EGF. HeLa cells were thus treated To uncover which MAPK signaling modules were preferentially with recombinant human EGF, and ZAK activation and expression activated upon ZAK overexpression, we prepared protein extracts levels were monitored by western blotting using specific from control, ZAK-α and ZAK-β transfected HCT116 cells, and antibodies (phospho- and total ZAK). As observed in Figure 4a, determined the phosphorylation status of the three MAPK p38, JNK EGF engendered a peak of activation of ZAK-α 5 min after the and ERK in normal or low serum conditions. ZAK-α or ZAK-β treatment, followed by a transient, but more extended period of isoforms strikingly increased the intracellular quantity of phospho- ZAK-β activation. After quantification, it appeared that EGF

© 2016 Macmillan Publishers Limited Oncogene (2016) 3190 – 3200 The MAP3K ZAK controls ERK-dependent cell migration C Rey et al 3192 induced a 2- to 2.5-fold increase of phospho-ZAK-α and phospho- a decrease of ZAK endogenous protein level. HCT116 cells, known ZAK-β, respectively (Figure 4a, right panel). In regards to total ZAK, to carry a KRAS mutation were thus treated with a MEK inhibitor EGF provoked a signiﬁcant, albeit small, rise of the two proteins for 4, 24 or 48 h. As shown in Figure 4b, this treatment resulted in 24 h after the treatment (Figures 4a, 1.25x for ZAK-β and 1.7x for a 40% drop of ZAK-α or ZAK-β protein expression after 24 and 48 h ZAK-α proteins, data not shown). Next, we hypothesized that of contact with the MEK inhibitor. Altogether, the EGF signaling disrupting the EGFR signaling cascade in cells harboring an cascade controls both the activation and the expression of ZAK in oncogenic activation of this pathway will, on the contrary, lead to cancer cells.

Oncogene (2016) 3190 – 3200 © 2016 Macmillan Publishers Limited The MAP3K ZAK controls ERK-dependent cell migration C Rey et al 3193 ZAK modulates EGF-dependent migration mucosa, a subgroup of patients presenting an upregulation of ZAK- Taking into consideration that EGF elicits ZAK activation, we next α (23.6%) and ZAK-β (39.2%) emerged from the cohort (Figure 6c). wondered whether the kinase would be a central factor of the Among these patients, approximately 50–75% overexpressed both EGF-dependent migration. To address this question, several clones isoforms. All these results were confirmed in a second TMA of HeLa cells depleted for ZAK proteins were submitted to wound- constructed with the normal mucosa and tumoral tissues of 30 healing scratch tests in various growing conditions. The two most patients with metastatic CRC (Supplementary Figure 8). efficient short hairpin RNA (shRNA) clones were chosen for With a potential role in the EGFR pathway,27 we first these migration assays, namely HeLa anti-ZAK shRNA #2 and #3 hypothesized ZAK overexpression in tumors to be the result of (Figure 5a). As presented above, the absence of ZAK proteins the oncogenic activation of the EGFR axis that usually occurs in impaired HeLa cell migration in complete medium (Figure 5b and KRAS- or BRAF-mutated patients. To test this possibility, we Supplementary Figure 7A). In low serum conditions, however, determined KRAS and BRAF mutational status for all patients both WT- and ZAK-depleted cells showed little migration because included in our study (Supplementary Figure 9) and crossed the of the lack of growth factors (Figure 5c). When the cells were then data with ZAK protein level. As shown in Figure 6d, ZAK-α and challenged with EGF, only parental HeLa cells were able to ZAK-β were overexpressed regardless of the presence or absence maintain a substantial migration rate, whereas the two cell lines of KRAS or BRAF mutations, showing that ZAK overexpression lacking ZAK showed low mobility (Figure 5d and Supplementary could constitute an independent event, and, as such, a potential Figure 7B). novel mechanism able to drive the aberrant activation of EGF Interestingly, ERK was found to be less phosphorylated upon pathway and tumor development. EGF treatment in cells lacking ZAK (Figure 5e), which prompted us To address this last question, we used the GEO to investigate whether ZAK-induced cell migration would rely on database repository and analyzed available microarray data sets. ERK activity. HeLa cells were thus transiently transfected with Using Sabates-Bellver et al.28 transcriptomic study, we found that pcDNA3.1, pReceiver-ZAK-α or ZAK-α K45M and subjected to a ZAK-α and ZAK-β transcripts were markedly increased in colorectal migration test in the presence or absence of a MEK inhibitor. o ZAK-α was chosen as it had a higher effect on migration when adenoma compared with normal mucosa (Figure 6e; ***P 0.001), overexpressed. As illustrated in Figure 5f, EGF-induced migration indicating that both isoforms of ZAK are already upregulated in was perfectly abrogated by the addition of the MEK inhibitor, benign or premalignant tumors. For each isoform, we then – validating our experimental settings (Figure 5f, left panel). subcategorized patients into ZAK+ and ZAK subgroups, depend- Inactivating MEK also prevented ZAK-α from favoring cell motility ing on whether they were expressing more or less mRNA (Figure 5f, middle panel), whereas it had no effect on pcDNA3.1 or transcripts than the mean of the adenoma population ZAK K45M transfected cells (Figure 5f, right panel). Taken together, (Figure 6e; adenoma mean for ZAK-α = 9.396 and ZAK-β = 9.194) it is likely that the MAP3K ZAK is part of the EGF/MEK/ERK- and run a gene set enrichment analysis to shed light on gene sets dependent migration. that were differentially expressed between ZAK+ and ZAK– adenoma subgroups. Using the reactome database, we observed α ZAK expression is increased in colorectal tumors and associated nine gene sets that were associated with the ZAK- +population with the activation of ERK- and motility pathways and 22 with the ZAK-β+population (Po0.05). More notably, ERK- α As the EGFR pathway is frequently altered in CRC, we next and motility-related gene sets were among these modules: ZAK- fi questioned whether ZAK could be deregulated in this pathological was indeed signi cantly found associated with the gene set ‘ condition. We thus analyzed ZAK expression in a TMA of 80 reactome GRB2 SOS provides linkage to MAPK signaling for ’ β colorectal carcinomas and observed a significant increase of integrins (*P = 0.049), and ZAK- with the two genes sets ZAK-α and ZAK-β protein in the tumoral tissue compared with the ‘reactome ERK MAPK targets’ (*P = 0.024) and ‘reactome L1CAM normal mucosa (Figures 6a and b). When individual variation of interactions’ (**P = 0.002)), corroborating our data showing ZAK as ZAK expression was evaluated by dividing the median intensity a new regulating factor of the EGF migration pathway and as found in tumors by the one measured in the corresponding normal potential oncogenic factor in human tumors.

Figure 1. ZAK expression modulates HCT116 cell migration. (a) HCT116 cells were transfected with siRNA directed against ZAK-α, ZAK-β or both proteins using Oligofectamin reagent. Seventy-two hours later, cells were lyzed for western blotting to reveal ZAK-α and ZAK-β expression level using Bethyl and Abcam antibodies, respectively. (b, c) HCT116 cells were transfected with isoform-specific anti-ZAK siRNA for 72 h and harvested to evaluate cell proliferation (b) and apoptosis (c). Cell proliferation was followed by DiI staining. The index proliferation was calculated from the ratio of the DiI median fluorescence at T0 and T72 hours. Apoptosis was measured by the loss of rhodamine123 staining. Statistical analyses were done using a Kruskal–Wallis test followed by a post-multiple comparison Dunn’s test (*Po0.05 vs NT). Error bars, s.d. from three independent experiments, each done in duplicate. (d, e) HCT116 cells were transfected with siRNA directed against ZAK-α, ZAK-β or both proteins using Oligofectamin reagent. Forty-eight hours later, cells were seeded into 96-well plates and monitored for cell migration 12 h later for 12 h. Migration kinetics are shown in d, whereas specific results obtained 5 and 10 h after the scratch are shown in e. Statistical analyses were done using an analysis of variance test followed by a post-multiple comparison Holm–Sidak test (**Po0.01; ***Po0.001, ****Po0.0001 vs NT). Error bars, s.d. from one representative experiment done in eight replicates and repeated two other times in eight and four replicates, respectively. (f, g) ZAK-α and/or ZAK-β-overexpressing HCT116 cells were challenged for cell migration. HCT116 cells were transfected with an empty plasmid or one encoding ZAK-α or ZAK-β using Exgen 500 reagent. Twenty-four hours post-transfection, HCT116 cells were placed in RPMI supplemented with 0.1% of fetal bovine serum (FBS) for another day and harvested to be seeded into 96-well plates for cell migration monitoring using the IncuCyte system as described in the Materials and methods section. Migration kinetics are shown in f, whereas specific results obtained 14 and 18 h after the scratch are shown in g. Statistical analyses were done using a one-way analysis of variance (ANOVA) test followed by a Holm–Sidak’s multiple comparison in g (*Po0.05, ***Po0.001; ****Po0.0001 vs pcDNA3). Error bars, s.d. from two independent experiments, each done in quadruplicate. (h) Determination of the migration capacities of HCT116 cells transfected with ZAK-α K45M and ZAK-β K45M compared with control cells. Cells were treated exactly as depicted in f. Migration kinetics are shown on the left and middle panels, whereas specific results obtained 16 h after the scratch are shown on the right panel. Statistical analyses were done using a one-way ANOVA test followed by a Holm–Sidak’s multiple comparison (*Po0.05 vs pcDNA3). Error bars, s.d. from two independent experiments, each done in quadruplicate.

© 2016 Macmillan Publishers Limited Oncogene (2016) 3190 – 3200 The MAP3K ZAK controls ERK-dependent cell migration C Rey et al 3194 DISCUSSION Figures 2 and 3). Up to now, only three original articles have Although ﬁrst described as anti-proliferative and pro- reported a possible role of ZAK in cell migration or invasion, with apoptotic,11,16–18,29 ZAK is more likely to be a pro-oncogenic converse results. On one hand, ZAK-α was shown to interfere with factor.14,19,20,22,23 To add to this debate, we investigated its the Rho-GTPase-dependent migration via an interaction with the intracellular function, applying a systematic analysis of the Rho GDP dissociation inhibitor, Rho-GDIβ.30,31 This binding, repercussions of ZAK-α or/and ZAK-β knock-down or overexpres- resulting in the phosphorylation and further inactivation sion, on cell proliferation, death and migration. Of the three, of Rho-GDIβ, reduced cardiomyocytes motility although migration appeared as the hallmark the most seriously affected previous data have also suggested Rho-GDI displacement to be by ZAK expression changes (Figure 1 and Supplementary a prerequisite for the activation of the Rho-small GTPases

Figure 2. ZAK expression activates several MAPK pathways. (a) Activation of p38, ERK and JNK MAPK by ZAK-α and ZAK-β. HTC116 cells were transiently transfected with vehicle, an empty plasmid, pReceiver-ZAK-α and pReceiver-ZAK-β for 24 h and lyzed for proteins extraction. Activation of phospho-MAPK was evaluated by immunoblotting using anti-phospho-p38, anti-phospho-ERK or anti-phospho-JNK antibodies. ZAK expression was checked using an anti-ZAK antibody. (b) HCT116 cells were treated exactly as explained in a, except that they were incubated in low serum conditions for 24 h, 1 day after the transfection. Cells extracts were then processed for western blot to reveal phospho-MAPKs and ZAK expression level. (c) HCT116 cells were treated exactly as explained in a, except that they were incubated in low serum conditions for 24 h, 1 day after the transfection. Cells extracts were then processed for western blot to reveal phospho- and total MAPKs and ZAK expression level. (d) HTC116 cells were transiently transfected with an empty plasmid, pReceiver-ZAK-α, pReceiver-ZAK-β, pReceiver-ZAK-α K45M, pReceiver-ZAK-β K45M for 24 h and lyzed for proteins extraction. Activation of phospho-MAPK was evaluated by immunoblotting using anti-phospho-p38, anti-phospho-ERK or anti-phospho-JNK antibodies. ZAK expression was checked using an anti-ZAK. (e) Activation of phospho-ERK by ZAK-α and ZAK-β. HCT116 were transfected with pReceiver-ZAK-α, pReceiver-ZAK-α K45M, pReceiver-ZAK-β or pReceiver-ZAK-β K45M for 24 h, incubated in 0.1% fetal bovine serum (FBS) for another day and fixed to reveal phospho-ERK activation (red) by immunofluorescence. Images were acquired by confocal microscopy. ZAK-α and ZAK-β transfected cells (green) were detected using specific ZAK antibodies. Nuclei were stained with Hoechst 33258 dye (blue). The percentage of ZAK-overexpressing cells that exhibited an activation of phospho-ERK is displayed at the bottom left of the ‘overlay’ image (n = 60, 27, 57, 45 cells for ZAK-α, ZAK-αK45M, ZAK-β and ZAK-βK45M, respectively). (f) Activation of phospho-c-JUN by ZAK-α and ZAK-β. HCT116 were transfected with pReceiver-ZAK-α, pReceiver-ZAK-α K45M, pReceiver-ZAK-β or pReceiver-ZAK-β K45M for 24 h and fixed to reveal phospho-c-JUN activation (red) by immunofluorescence. ZAK proteins detection and further analyses were done as explained in e (n = 27, 20, 27, 20 cells for ZAK-α, ZAK-αK45M, ZAK-β and ZAK-βK45M, respectively).

Figure 3. Pharmacological inhibition of ZAK prevents its autophosphorylation and ability to induce MAPK activation and cell migration in HCT116 cells. (a) ZAK-α-induced phospho-c-JUN activation is inhibited by nilotinib in a dose-dependent manner. HCT116 were transfected with vehicle, pcDNA3.1 or pReceiver-ZAK-α and pReceiver-ZAK-β for 24 h and treated with increasing amount of nilotinib (0.3, 1 and 3 μM) for 24 h. Phospho-c-JUN activation (red) and ZAK transfected cells (green) were detected by immunofluorescence using specific antibodies as specified in Materials and Methods section. Nuclei were stained with Hoechst 33258 dye (blue). Images were acquired by laser scanning cytometry using the Icys Instrument. Representative images of ZAK and phospho-c-JUN staining in the absence or presence of 3 μM of the drug are shown on the left panel. Quantification of the phospho-c-JUN staining is displayed in the right panel (N ∼ 1500 cells for the control and pcDNA3.1 samples; N ∼ 150 cells for ZAK transfected cells, in each condition, except ZAK-α Nilo 3μM: N ∼ 60). (b) Activation of p38, ERK and JNK MAPK by ZAK-α and ZAK-β is inhibited by nilotinib. HTC116 cells were transiently transfected with vehicle, pcDNA3.1, pReceiver-ZAK- α or pReceiver-ZAK-β for 24 h and incubated in RPMI supplemented with 0.1% of fetal bovine serum (FBS). Twenty-four hours later, 1 μM (left panel) and 3 μM (right panel) of nilotinib were added to the cells for 24 h. Protein extracts were then prepared to evaluate p38, ERK and JNK phosphorylation as well as ZAK levels. (c) Nilotinib prevents ZAK-induced cell migration. Control, ZAK-α or ZAK-β-overexpressing HCT116 cells were placed in 0.1% FBS for 24 h and incubated in presence of increasing concentrations of nilotinib (0 ( = UN), 0.3, 1 and 3 μM) just prior cell migration monitoring using the IncuCyte system. Migration curves were generated from the relative wound density values and plotted for each sample in the absence and presence of the different concentrations of nilotinib used.

(Rho, Rac or Cdc42) and thereby for cell migration.32–34 On the Indeed, although JNK inhibitors were toxic and JNK involvement other hand, ZAK-β was shown to convey lysophosphatidic acid- could not be correctly evaluated, preliminary experiments using mediated cell invasion,14 corroborating our results and under- p38 inhibitors indicate a possible role of this well-established cell lining ZAK overexpression as a factor of aggressiveness in cancer migration factor in ZAK’s action (data not shown). cells. Interestingly, both in Korkina et al.14 and in our studies, ERK Activated by EGF and able to trigger ERK phosphorylation, ZAK appears as a central kinase able to mediate ZAK pro-invasive appeared as a novel intermediate factor of the EGF/MEK/ERK functions. Although ZAK-induced cell migration was clearly signaling network. KRAS and/or BRAF mutational status were not inhibited by a MEK inhibitor, we cannot deny, however, that a found, however, to be correlated with ZAK protein expression part of ZAK pro-migratory effects could also imply p38 or JNK. levels in patients. This raises the question of how the EGFR is

Figure 4. EGF pathway regulates ZAK expression and activation. (a) HeLa cells were seeded into six-well plates and incubated, 24 h later, in RPMI supplemented with 0.1% fetal bovine serum (FBS) for 24 h. Recombinant human EGF was added when necessary at 20 ng/ml. Cells were harvested at 5 min, 20 min, 6 h or 24 h post-EGF treatment and proteins extracted for immunoblotting against the total and phosphorylated form of ZAK-α and ZAK-β. GAPDH was used as a loading control. (b) After 24 h in culture, HCT116 cells were treated or not with 50 nM of MEK inhibitor (PD 0325901) for 4, 24 or 48 h, harvested and lyzed for protein extraction. ZAK-α and ZAK-β endogenous levels were then determined by western blot using speciﬁc antibodies. GAPDH was used as a loading control. The quantiﬁcation of ZAK-α and ZAK–β expression level, shown on the right panel, is the result of three independent experiments.

connected to ZAK? Previous data have shown that the EGF disease progression. The presence of elevated ZAK mRNA level in receptor may directly interact with c-Src and FAK to promote cell pre-cancerous lesions suggests, however, that the protein is migration upon EGF binding.35–37 Among its various downstream already deregulated at the initial stage of the disease. This was substrates, FAK was in addition shown to activate RhoGTPase then confirmed by the TMA analyses in which as many stage-I/II, through the regulation of various RhoGAP (GTPase-activating stage-III and stage-IV patients (TMA # 1) were found positive for proteins) and RhoGEF (guanine exchange factor) proteins.38 It is ZAK-α and/or -β (data not shown). This raises the question of thus tempting to speculate that ZAK would constitute a novel whether pro-migratory signals are turned up at initial stages of CR downstream kinase able to mediate pro-migratory signals carcinogenesis and are then maintained throughout disease through the successive activation of EGFR4Src/FAK4RHO4 progression. The response to this question is not trivial and is ZAK4MEK4ERK. Surprisingly, we found that ZAK-α-overexpressing still being debated. Two models are currently being compared HCT116 cells were able to dephosphorylate FAK Tyr397 (data not with explain metastasis: the ‘linear progression model’ and ‘the shown). Even if unexpected, FAK has previously been reported parallel progression model’, which mainly differ in terms of the to be dephosphorylated post-EGF-induced cell migration and timing of tumor cell dissemination. The first model suggests that invasion in EGFR-overexpressing cells.39 At this stage, these initial metastasis occurs only after an extensive expansion of the primary data indicate that ZAK-controlled cell motion is somehow related tumor and the accumulation of multiple mutations over the to FAK; however, this may either be a direct effect or the result innumerable cycles of cell division. In contrast, the second model of a negative feedback loop. Until now, ZAK has only been proposes an early separation and then development of the connected to G-protein-coupled receptors (GPCR) signaling path- primary tumor and metastases.40 It is thus likely that pro-migratory way upon lysophosphatidic acid treatment.14 Hence, our results signals already emanate from early-transformed cells. In this show a connection with a second signaling cascade coupled with schema, ZAK could be an essential factor able to favor a local EGF stimulation. Of note, both GPCR and EGFR pathways relate the dissemination of cancer cells or a wider diffusion at distant sites. MAP3K to cell migration/invasion processes. Nevertheless, this abiding high level of ZAK expression could be a Data from the literature suggested ZAK to be involved in CRC crucial element for a chemotherapeutic approach, particularly as development. Consequently, we evaluated its expression level we have identified a specific inhibitor of its kinase activity, that is, from public microarray database and TMA, and found that both the second-generation BCR–ABL inhibitor nilotinib. ZAK isoforms were greatly overexpressed in colorectal adenoma In conclusion, we demonstrated here that the MAP3K ZAK (transcriptomic analysis) and carcinoma (TMA). Based on our intervenes, in a kinase-dependent manner, as a major factor of the in vitro findings, it was then tempting to equate ZAK to a pro- EGF-controlled cell motility cascade of HeLa and HCT116 cancer metastatic factor, whose expression may be associated with cells. Largely overexpressed in a substantial pool of CRC patients,

Figure 5. ZAK modulates EGF-dependent migration. (a) Determination of ZAK-α and ZAK-β expression level after infection of HeLa cells with three different lentiviruses hosting shRNA directed against ZAK protein compared with control cells. ZAK protein level was determined by western blotting using ZAK-specific antibody. The two most efficient clones with regards to ZAK-α and ZAK-β depletion (that is, shRNA#2 and shRNA #3) were used in the subsequent experiments. (b–d) Migration kinetics of native, control and ZAK-depleted HeLa cells in complete (b), 0.1% (c) and 0.1%+EGF (100 ng/ml) (d) medium. Data were generated using the IncuCyte system. Experiments were done twice in triplicate. (e) Determination of ERK phosphorylation status post-EGF treatment in cells depleted for ZAK. Native, pxs68, pxs68 shRNA scrambled and pxs68 shRNA anti-ZAK#2 Hela cells were seeded in complete medium and then put in low serum condition (RPMI+0.1% fetal bovine serum (FBS)) for 24 h. EGF (100 ng/ml) was added for 2 h before protein extraction. ERK phosphorylation and ZAK expression level were evaluated by immunoblotting using specific antibodies. (f) Impact of the EGF pathway inhibition on ZAK-induced cell migration. HeLa cells were transfected with mock, pReceiver-ZAK-α (middle panel) or pReceiver-ZAK α K45M (right panel) and were incubated in RPMI 0.1% FBS for 24 h and plated in a 96-well plate. Mek inhibitor (50 nM) was added, where necessary, 1 h before the scratch and left in the medium all along the experiment. EGF (100 ng/ml) was used as a control for EGF-dependent migration (left panel). It was added at the time of the scratch. The surface occupied by the cells in the wound was measured every 2 h for 32 h using the Incucyte system. This value was registered and plotted as the relative wound density to construct the migration curves. Error bars, s.d. from an experiment done in quadruplicate, repeated twice.

ZAK could be a novel interesting therapeutic target in this 10% fetal bovine serum at 37 °C in a humidified CO2 incubator. pathology, notably because (i) its elevated expression correlates Transfections were made using Exgen 500 (Euromedex, Souffelweyer- with MAPK activation and migration pathways and (ii) nilotinib, an sheim, France). For EGF treatment (20–100 ng/ml, R&D Systems, Lille, anticancer agent already used in clinic, appears as a potent France), cells were pre-incubated in RPMI-1640 0.1% fetal bovine serum for specific inhibitor of the kinase. 24 h. MEK1/2 inhibitor III (50 nM, PD 0325901, Calbiochem/VWR, Fontenay- sous-Bois, France) was added 1 h before EGF treatment. Nilotinib was purchased from Sequoia Research Products (Pangbourne, UK). MATERIALS AND METHODS Cell lines and reagents Plasmids, siRNA and shRNA HCT-116, HT-29 and HeLa cells were cultured in RPMI-1640 (Gibco, pReceiver-ZAK-alpha (EX-X0399-MO2) and ZAK-beta (EX-V0290-M02) Invitrogen/ThermoFisher Scientific, Illkirch, France) supplemented with plasmids were purchased from Genecopoia (Tebu-bio, Le Perray-en-

Yvelines, France). siRNA (Eurogentec, Angers, France) targeting ZAK-α siRNA#2—5′-CATGCAAGCCAAGCAGAAT-3′) (200 nM) were transfected (siRNA#1—5′-GAAGTGGATTGTAGGAATA-3′ and siRNA#2—5′-GGTGCCCAT using Oligofectamin reagent (Invitrogen/ThermoFisher Scientiﬁc). ShRNA TAAGTATCAA-3′) or ZAK-β (siRNA#1—5′-GGAGGATAATGACATGGAT-3′ and depleting both ZAK-α and ZAK-β (shRNA 3267: 5′-ACGAGAGAT

*** *** normal mucosa normal mucosa adenoma 10 adenoma 10 9 9 8 8

7 7 mRNA expression (log2) mRNA mRNA expression (log2) mRNA 6 normal mucosa adenoma normal mucosa adenoma

MEAN MEAN Normal Mucosa 8.371 Normal Mucosa 8.457 Adenoma 9.393 Adenoma 9.194 Figure 6. ZAK is overexpressed in colorectal adenoma and tumors compared with healthy mucosa. (a) Expression level of ZAK protein in normal tissue (NT) and tumoral tissue (TT) evaluated by immunofluorescence on a TMA composed of 80 CRC tissue sections, using an anti- ZAK-α (left panel) and anti-ZAK-β (right panel) antibody. Data are presented using a box and whiskers plot (whiskers being min to max). Statistical analyses were done by applying a paired t-test (*Po0.05; ****Po0.0001). The mean and median fluorescence intensity of ZAK-α and ZAK-β are reported in the table below each plot. (b) ZAK intracellular distribution in NT or TT. The TMAs used in (a) and stained with anti- ZAK-α or ZAK-β antibodies were used to acquire representative images of the intracellular localization of the two different isoforms of ZAK in tumors. Images of ZAK-α and ZAK-β were acquired in the green channel by confocal microscopy. Hoechst 33258 was used as a nuclear marker (blue). (c) Fold change in ZAK-α (left panel) and ZAK-β (right panel) expression level between the NT and TT for each patient calculated by dividing, for each patient, ZAK median intensity of fluorescence in the TT by ZAK median intensity of fluorescence in the NT. (d) Table reporting KRAS and BRAF status in ZAK-α and ZAK-β+ patients and the proportion of such mutated patients in these two cohorts compared with the total population. (e) mRNA expression level of ZAK-α (left panel) and ZAK-β (right panel) in colorectal adenoma and surrounding normal mucosa from 32 patients. Data were obtained from advanced analysis of the public database published by Sabates-Bellver et al.,28 as explained in the Materials and Methods section. Statistical analysis was done using a paired t-test (***Po0.001).

Oncogene (2016) 3190 – 3200 © 2016 Macmillan Publishers Limited The MAP3K ZAK controls ERK-dependent cell migration C Rey et al 3199 TAACCATTCCAA-3′ and shRNA 3268: 5’-CCATAACCATACAACACACAT-3′) Transcriptomic and gene set enrichment analyses (Open Biosystems/GE-Healthcare, Velizy-Villacoublay, France) were ZAK mRNA expression was evaluated from a public database using cloned into pLKO.1 lentiviral plasmid before virus production and cell Sabates-Bellver et al.’s microarray data sets.28 The transcription profiles infection. were downloaded from the Gene Expression Omnibus (GEO) repository (accession number GSE8671).41 The two following probes: 222757_s_at Live-cell kinetic assays for cell proliferation and migration and 225662_at were chosen (based on the maximum interquartile range and the best percentage of identity between the probe and ZAK sequence) Cells were seeded onto six-well plates before being transfected. Forty- α β eight hours later, they were plated into a 96-well plate and transferred into to monitor ZAK- and ZAK- expression, respectively. Expression data were the IncuCyte real-time imaging system (Essen Instruments, Hertfordshire, analyzed on a log2 basis. Gene set enrichment analysis was done using the Broad Institute Website as described in Subramanian et al.42 and Mootha UK). For cell growth measurement, high-quality images were acquired and 43 processed by a contrast-based confluence algorithm. For cell migration, a et al. We used reactome gene sets of the C2 collection (that is, 674 strip of cells (grown at confluence) was removed from each well before functional gene sets) in this study. image acquisition (every 2–4 h for 16–50 h). Cell migration was analyzed using the IncuCyteFLR software (Essen BioScience, Hertfordshire, UK) and Clinical data and CRC samples expressed in terms of percentage relative wound density, a metric taking Two patient cohorts were analyzed: the first included 80 patients with into consideration the spatial cell density in the wound area relative primary CRC (previously described Rey et al.44 and Soubeyran et al.45), and to the spatial cell density outside of the wound area at every time point the second included 30 patients with metastatic CRC. In all cases, (http://www.essenbioscience.com/essen-products/software/cell-player- fragments of tumor and normal mucosa obtained at surgery were quickly cell-migration-software-module/). In that respect, relative wound density is frozen in liquid nitrogen for DNA extraction. The remaining material was self-normalizing for changes such as cell proliferation and/or pharmaco- fixed in Holland Bouin's (cohort#1 = TMA#1) or formol (cohort#2 = TMA#2) logical effects. solution and paraffin embedded for pathological evaluation and TMA construction. Flow cytometry analyses of cell proliferation and cell death Cell proliferation and apoptosis were evaluated using DiI (1,1′-dioctadecyl- TMA preparation and immunostaining 3,3,3′,3′-tetramethylindocarbocyanineperchlorate) and Rhodamine 123 TMA construction was previously described.44,45 For staining, a 5 μm dyes (Molecular Probes, Invitrogen/ThermoFisher Scientific), respectively. section of the TMA placed on a glass slide was de-paraffinized in toluene Cells were incubated with a solution of 5 μM DiI diluted in 0.3 M sucrose and incubated with anti-ZAK primary antibodies (from R Ruggieri’s under mild stirring for 15 min at 37 °C, rinsed, and seeded in a 96-well plate laboratory; Bethyl Laboratories or Abcam) for 1 h at room temperature, before siRNA transfection. Three days later, cells were stained with 100 ng/ ml of Rhodamine 123 at 37 °C for 30 min, trypsinized and resuspended in after proteolytic epitope retrieval in citrate buffer (pH 6.0). Alexa Fluor 488 Krebs Ringer buffered saline plus 4% fetal bovine serum. We used the goat anti-rabbit or anti-mouse (Molecular Probes) secondary antibodies FACSCalibur Flow Cytometer (BD Biosciences, Rungis, France) for analyses. were incubated for 1 h at room temperature. 4,6-Diamidino-2-phenylindole μ fl Proliferation was evaluated by calculating the proliferation index from the (1 g/ml) was added for 15 min. Immuno uorescence signals were FL2 median intensity at T0 and T72 hours. collected with an automated laser scanning cytometer (iCys; Compucyte). Quantification was done using a phantom ( = pseudocells) segmentation. For each spot of tissue, cellular areas were selected based on nuclear Western blot analysis staining (4,6-diamidino-2-phenylindole) and used to evaluate ZAK-α and Cells were lysed for 1 h at 4 °C, in 50 mM Tris-Hcl, 150 mM NaCl, 1% NP40 ZAK-β fluorescence intensities. They were plotted as histograms from supplemented with a cocktail of protease (Roche, Sigma Aldrich, Lyon, which the median intensity value was extracted. The average median France) and phosphatase inhibitors (1.25 mM sodium orthovanadate, intensity for each tumor was calculated from the three cores of tissues 30 mM sodium fluoride). Lysates were pelleted 15 min at 13200 r.p.m. available on the TMA. Data were computed and analyzed using Excel and and supernatants kept for protein quantification (Bradford assay). Proteins GraphPad Software (La Jolla, CA, USA). were then separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis, and transferred to polyvinylidene difluoride membranes for immunoblotting with the appropriate primary antibody: mouse anti- CONFLICT OF INTEREST α ZAK and phospho-ZAK (kindly provided by R Ruggieri); rabbit anti-ZAK- The authors declare no conflict of interest. (Bethyl Laboratories/Euromedex, Souffelweyersheim, France); mouse anti- ZAK-β (Abcam 57318, Paris, France); anti-actin (Sigma, Lyon, France); anti- GAPDH (Santa Cruz, Heidelberg, Germany); anti-phospho (Thr180/Tyr182) ACKNOWLEDGEMENTS and total p38, anti-phospho (Thr183/Tyr185) and total SAPK/JNK and anti- phospho- (Thr202/Tyr204) and total p44/42 (ERK1/2) MAPKs (Cell Signaling This work was supported by the French National Institute for Medical Research Technology, Ozyme, St Quentin en Yvelines, France). Membranes were (INSERM), Aquitaine Region, French Ministry of Research, Institut Bergonié, then rinsed, incubated with a horseradish peroxidase-conjugated anti- Association for Cancer Research (ARC), the Ligue Contre le Cancer, and Cancéropôle ‘ ’ mouse or anti-rabbit antibody (Amersham/GE Healthcare Europe GmbH, Grand Sud-Ouest, Agir Cancer Gironde and The French National Research Agency Velizy-Villacoublay, France), and developed with the Immobilon western (ANR). We thank Nicolas Faur, Raphael Moustié and Assia Chaibi for their technical detection system (Millipore, Molsheim, France). help and Pippa McKelvie-Sebileau of Institut Bergonié for medical editorial assistance.

Immunofluorescence and image analysis REFERENCES Cells were seeded onto 96-well plates the day prior transfection, incubated 1 Schaeffer HJ, Weber MJ. Mitogen-activated protein kinases: specific messages in low serum conditions (when needed) and fixed with 3.7% formaldehyde from ubiquitous messengers. Mol Cell Biol 1999; 19: 2435–2444. for 10 min at room temperature. After three washes in phosphate-buffered 2 Wagner EF, Nebreda AR. Signal integration by JNK and p38 MAPK pathways in saline, cells were permeabilized in 0.5% Triton X-100 for 5 min. Nonspecific cancer development. Nat Rev Cancer 2009; 9: 537–549. binding was blocked with 0.2% gelatin/phosphate-buffered saline for 3 Keshet Y, Seger R. The MAP kinase signaling cascades: a system of hundreds of 30 min, and staining was performed using a specific primary antibody components regulates a diverse array of physiological functions. Methods Mol Biol (previously described, and anti-phospho-c-JUN Ser63, Cell Signaling 2010; 661:3–38. Technology) overnight at 4 °C and a fluorescent secondary antibody 4 Yoon S, Seger R. The extracellular signal-regulated kinase: multiple substrates conjugate (Alexa Fluor 488, 532 or 594 anti-mouse or rabbit IgG, Molecular regulate diverse cellular functions. Growth Factors 2006; 24:21–44. Probes, Invitrogen/ThermoFisher Scientific) for 1 h at room temperature. 5 Huang C, Jacobson K, Schaller MD. MAP kinases and cell migration. J Cell Sci 2004; Nuclei were stained with Hoechst 33258 dye (Molecular Probes). Image 117: 4619–4628. acquisition was done using the Zeiss LSM 510 Meta microscope (Le Pecq, 6 Rajalingam K, Wunder C, Brinkmann V, Churin Y, Hekman M, Sievers C et al. France) or the iCys laser scanning cytometer (CompuCyte, Boston, Prohibitin is required for Ras-induced Raf-MEK-ERK activation and epithelial cell MA, USA). migration. Nat Cell Biol 2005; 7:837–843.

© 2016 Macmillan Publishers Limited Oncogene (2016) 3190 – 3200 The MAP3K ZAK controls ERK-dependent cell migration C Rey et al 3200 7 Reiterer V, Fey D, Kolch W, Kholodenko BN, Farhan H. Pseudophosphatase STYX 26 Rix U, Remsing Rix LL, Terker AS, Fernbach NV, Hantschel O, Planyavsky M et al. modulates cell-fate decisions and cell migration by spatiotemporal regulation A comprehensive target selectivity survey of the BCR-ABL kinase inhibitor of ERK1/2. Proc Natl Acad Sci USA 2013; 110: E2934–E2943. INNO-406 by kinase profiling and chemical proteomics in chronic myeloid 8 Gotoh I, Adachi M, Nishida E. Identification and characterization of a novel MAP leukemia cells. Leukemia 2010; 24:44–50. kinase kinase kinase, MLTK. J Biol Chem 2001; 276: 4276–4286. 27 Choi HS, Choi BY, Cho YY, Zhu F, Bode AM, Dong Z. Phosphorylation of Ser28 in 9 Gross EA, Callow MG, Waldbaum L, Thomas S, Ruggieri R. MRK, a mixed lineage histone H3 mediated by mixed lineage kinase-like mitogen-activated protein kinase-related molecule that plays a role in gamma-radiation-induced cell triple kinase alpha. J Biol Chem 2005; 280: 13545–13553. cycle arrest. J Biol Chem 2002; 277: 13873–13882. 28 Sabates-Bellver J, Van der Flier LG, de Palo M, Cattaneo E, Maake C, Rehrauer H 10 Bloem LJ, Pickard TR, Acton S, Donoghue M, Beavis RC, Knierman MD et al. et al. Transcriptome profile of human colorectal adenomas. Mol Cancer Res 2007; Tissue distribution and functional expression of a cDNA encoding a novel mixed 5: 1263–1275. lineage kinase. J Mol Cell Cardiol 2001; 33: 1739–1750. 29 Yang JJ, Lee YJ, Hung HH, Tseng WP, Tu CC, Lee H et al. ZAK inhibits human lung 11 Liu TC, Huang CJ, Chu YC, Wei CC, Chou CC, Chou MY et al. Cloning cancer cell growth via ERK and JNK activation in an AP-1-dependent manner. and expression of ZAK, a mixed lineage kinase-like protein containing Cancer Sci 2010; 101: 1374–1381. a leucine-zipper and a sterile-alpha motif. Biochem Biophys Res Commun 2000; 30 Huang CY, Yang LC, Liu KY, Chang IC, Liao PH, Chou JI et al. ZAK negatively 274 – : 811 816. regulates RhoGDIbeta-induced Rac1-mediated hypertrophic growth and cell 12 Takahashi M, Gotoh Y, Isagawa T, Nishimura T, Goyama E, Kim HS et al. Regulation migration. J Biomed Sci 2009; 16: 56. of a mitogen-activated protein kinase kinase kinase, MLTK by PKN. J Biochem 31 Huang CY, Yang LC, Liu KY, Liao PH, Chou JI, Chou MY et al. RhoGDIbeta-induced 133 – 2003; : 181 187. hypertrophic growth in H9c2 cells is negatively regulated by ZAK. J Biomed Sci fi 13 Amano M, Mukai H, Ono Y, Chihara K, Matsui T, Hamajima Y et al. Identi cation of 2009; 16: 11. a putative target for Rho as the serine-threonine kinase protein kinase N. Science 32 DerMardirossian C, Schnelzer A, Bokoch GM. Phosphorylation of RhoGDI by Pak1 271 – 1996; : 648 650. mediates dissociation of Rac GTPase. Mol Cell 2004; 15:117–127. 14 Korkina O, Dong Z, Marullo A, Warshaw G, Symons M, Ruggieri R. The MLK-related 33 DerMardirossian C, Bokoch GM. GDIs: central regulatory molecules in Rho GTPase kinase (MRK) is a novel RhoC effector that mediates lysophosphatidic acid activation. Trends Cell Biol 2005; 15: 356–363. (LPA)-stimulated tumor cell invasion. J Biol Chem 2013; 288: 5364–5373. 34 Griner EM, Churchill ME, Brautigan DL, Theodorescu D. PKCalpha phosphorylation 15 Huang CY, Chueh PJ, Tseng CT, Liu KY, Tsai HY, Kuo WW et al. ZAK re-programs of RhoGDI2 at Ser31 disrupts interactions with Rac1 and decreases GDI activity. atrial natriuretic factor expression and induces hypertrophic growth in H9c2 Oncogene 2013; 32: 1010–1017. cardiomyoblast cells. Biochem Biophys Res Commun 2004; 324: 973–980. 35 Goi T, Shipitsin M, Lu Z, Foster DA, Klinz SG, Feig LA. An EGF receptor/Ral-GTPase 16 Yang JJ. Mixed lineage kinase ZAK utilizing MKK7 and not MKK4 to activate the signaling cascade regulates c-Src activity and substrate specificity. EMBO J 2000; c-Jun N-terminal kinase and playing a role in the cell arrest. Biochem Biophys Res 19:623–630. Commun 2002; 297:105–110. 36 Ray RM, Li C, Bhattacharya S, Naren AP, Johnson LR. Spermine, a molecular switch 17 Wang X, Mader MM, Toth JE, Yu X, Jin N, Campbell RM et al. Complete inhibition regulating EGFR, integrin beta3, Src, and FAK scaffolding. Cell Signal 2012; 24: of anisomycin and UV radiation but not cytokine induced JNK and p38 activation 931–942. by an aryl-substituted dihydropyrrolopyrazole quinoline and mixed lineage kinase 37 Sieg DJ, Hauck CR, Ilic D, Klingbeil CK, Schaefer E, Damsky CH et al. FAK integrates 7 small interfering RNA. J Biol Chem 2005; 280: 19298–19305. growth-factor and integrin signals to promote cell migration. Nat Cell Biol 2000; 2: 18 Yang JJ. A novel zinc finger protein, ZZaPK, interacts with ZAK and stimulates the – ZAK-expressing cells re-entering the cell cycle. Biochem Biophys Res Commun 249 256. 2003; 301:71–77. 38 Schaller MD. Cellular functions of FAK kinases: insight into molecular mechanisms 123 – 19 Cho YY, Bode AM, Mizuno H, Choi BY, Choi HS, Dong Z. A novel role for mixed- and novel functions. J Cell Sci 2010; : 1007 1013. lineage kinase-like mitogen-activated protein triple kinase alpha in neoplastic cell 39 Lu Z, Jiang G, Blume-Jensen P, Hunter T. Epidermal growth factor-induced tumor transformation and tumor development. Cancer Res 2004; 64: 3855–3864. cell invasion and metastasis initiated by dephosphorylation and downregulation 21 – 20 Liu J, McCleland M, Stawiski EW, Gnad F, Mayba O, Haverty PM et al. of focal adhesion kinase. Mol Cell Biol 2001; : 4016 4031. Integrated exome and transcriptome sequencing reveals ZAK isoform usage in 40 van Zijl F, Krupitza G, Mikulits W. Initial steps of metastasis: cell invasion and 728 – gastric cancer. Nat Commun 2014; 5: 3830. endothelial transmigration. Mutat Res 2011; :23 34. 21 Adler AS, McCleland ML, Yee S, Yaylaoglu M, Hussain S, Cosino E et al. 41 Barrett T, Troup DB, Wilhite SE, Ledoux P, Rudnev D, Evangelista C et al. NCBI GEO: fi An integrative analysis of colon cancer identifies an essential function for PRPF6 mining tens of millions of expression pro les--database and tools update. Nucleic in tumor growth. Genes Dev 2014; 28:1068–1084. Acids Res 2007; 35: D760–D765. 22 Firestein R, Bass AJ, Kim SY, Dunn IF, Silver SJ, Guney I et al. CDK8 is a colorectal 42 Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA et al. cancer oncogene that regulates beta-catenin activity. Nature 2008; 455:547–551. Gene set enrichment analysis: a knowledge-based approach for interpreting 23 Vinayagam A, Stelzl U, Foulle R, Plassmann S, Zenkner M, Timm J et al. A directed genome-wide expression profiles. Proc Natl Acad Sci USA 2005; 102: 15545–15550. protein interaction network for investigating intracellular signal transduction. Sci 43 Mootha VK, Lindgren CM, Eriksson KF, Subramanian A, Sihag S, Lehar J et al. Signal 2011; 4: rs8. PGC-1alpha-responsive genes involved in oxidative phosphorylation are 24 Manley PW, Drueckes P, Fendrich G, Furet P, Liebetanz J, Martiny-Baron G et al. coordinately downregulated in human diabetes. Nat Genet 2003; 34:267–273. Extended kinase profile and properties of the protein kinase inhibitor nilotinib. 44 Rey C, Soubeyran I, Mahouche I, Pedeboscq S, Bessede A, Ichas F et al. HIPK1 drives Biochim Biophys Acta 2010; 1804: 445–453. p53 activation to limit colorectal cancer cell growth. Cell Cycle 2013; 12: 1879–1891. 25 Rix U, Hantschel O, Durnberger G, Remsing Rix LL, Planyavsky M, Fernbach NV 45 Soubeyran I, Mahouche I, Grigoletto A, Leste-Lasserre T, Drutel G, Rey C et al. et al. Chemical proteomic profiles of the BCR-ABL inhibitors imatinib, nilotinib, Tissue microarray cytometry reveals positive impact of homeodomain interacting and dasatinib reveal novel kinase and nonkinase targets. Blood 2007; 110: protein kinase 2 in colon cancer survival irrespective of p53 function. Am J Pathol 4055–4063. 2011; 178:1986–1998.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)