Colorectal Tumors Require NUAK1 for Protection from Oxidative Stress

The relentless drive to proliferate exposes tumor cells to considerable metabolic stress. Proliferating tumor cells increase nutrient consumption in order to balance the competing demands of macromolecular synthesis, toward which a large proportion of nutrient metabolites are diverted, with the energetic cost of sustaining viability, measured in ATP (1). Increased metabolic activity elevates production of reactive oxygen species (ROS), altering signal transduction and, at very high levels, inflicting damage upon lipids, proteins, and nucleic acids (2). In the context of a growing solid tumor with ineffective vascularity, tumor cells are commonly deprived of their preferred nutrients and exposed to hypoxia, which also increases ROS production, adding cell-extrinsic sources of further metabolic stress. In order to survive such stress, tumor cells must adapt flexibly and continuously by modulating their rates of macromolecular synthesis, cell growth, and proliferation, in order to maintain ATP homeostasis and counteract ROS. Failure to do so leads to ATP collapse, toxic levels of ROS, and loss of viability. As such, targeted suppression of adaptive measures used by tumor cells to counteract metabolic stress may yield therapeutic benefit in cancer treatment.

NUAK1 (aka ARK5) is one of 12 kinases related by sequence homology to the catalytic α subunits of AMPK (3). Collectively, these kinases play various roles in regulating cell adhesion and polarity, cellular and organismal metabolism, and in the cellular response to various forms of stress, including oxidative, osmotic, and energetic stress (4, 5). NUAK1 is a common target of several miRNAs that are frequently suppressed in cancer, suggesting a potential role for NUAK1 in tumorigenesis (6–8). Accordingly, we previously reported that NUAK1 is required to sustain viability of cancer cells when MYC is overexpressed (9).

In contrast with the widely studied AMPK, the molecular targets and downstream pathways governed by NUAK1 are poorly defined. To date, the best-characterized substrate of NUAK1 is the PP1β subunit MYPT1 (encoded by PPP1R12A). During cell detachment, phosphorylation of MYPT1 by NUAK1 inhibits PP1β phosphatase activity toward myosin light chain. Inhibition of NUAK1 thus increases PP1β activity, delaying cell detachment and suppressing cell migration (10). Other work points to a role for NUAK1 in metabolic regulation. Muscle-specific deletion of Nuak1 protects mice from high-fat diet–induced diabetes, attributable to increased glucose uptake and increased conversion of glucose to glycogen by NUAK1-deficient skeletal muscle (11). An earlier study showed that NUAK1 protects cancer cells from nutrient deprivation–induced apoptosis (12). In the context of MYC overexpression, we showed that NUAK1 is required to maintain ATP homeostasis, in part by facilitating AMPK-dependent inhibition of TORC1-driven macromolecular synthesis (9, 13). Failure to engage this checkpoint results in cell death under conditions of metabolic stress (14–16).

Our previous work thus suggested that NUAK1 may present a good target for therapy, specifically in the context of MYC-driven cancers. Human colorectal cancer is uniformly characterized by deregulated expression of MYC, and mouse models have shown that expression of endogenous Myc is required for intestinal polyp formation upon loss of Apc, the most common tumor-initiating event in human colorectal cancer (17, 18). We asked therefore if NUAK1 is required to support tumor cell viability in colorectal cancer. Here, we show that NUAK1 is overexpressed in human colorectal cancer and that high NUAK1 expression correlates with reduced overall survival. Using genetically engineered mouse models of colorectal cancer driven by sporadic loss of Apc, we show that NUAK1 is required for both initiation and maintenance of autochthonous colorectal tumors. NUAK1 facilitates nuclear translocation of the antioxidant master regulator NRF2 by counteracting negative regulation of NRF2 by GSK3β. Depletion or inhibition of NUAK1 thus renders human colorectal cancer cells and murine colorectal tumors vulnerable to oxidative stress–induced cell death. Our data reveal NUAK1 as a candidate therapeutic target in human colorectal cancer.

We used RNA-Scope in situ hybridization to examine NUAK1 expression in a 660-sample tissue microarray of human colorectal cancer (19). NUAK1 is weakly expressed in normal human colonic epithelium, but increased expression is significantly enriched in aggressive (Dukes’ stage B and C) colorectal cancer (Fig. 1A; Supplementary Fig. S1A and S1B). In silico examination of The Cancer Genome Atlas (TCGA) colorectal adenocarcinoma cohort similarly showed significantly elevated NUAK1 expression in advanced (T stage 3 and 4) versus early (T stage 1 and 2) disease, and in patients with lymph node metastasis versus none (Fig. 1B). Meta-analysis of 17 independent cohorts comprising 947 human colorectal cancer samples, via SurvExpress (20), also revealed significantly higher NUAK1 expression in the high-risk versus low-risk group, and elevated NUAK1 expression was associated with significantly reduced overall survival and a hazard ratio of 1.49 (Fig. 1C and D). A similar reduction of overall survival was borne out by individual analysis of two large cohorts in which the outcome for the majority of patients was known (Supplementary Fig. S1C–S1F; refs. 21, 22). Elevated NUAK1 expression thus correlates with worse outcome in colorectal cancer.

We therefore examined the functional requirement for NUAK1 in human colorectal cancer cell lines using two previously described highly selective NUAK1 inhibitors, HTH-01-015 and WZ4003. HTH-01-015 is reported to show little to no activity toward AMPK or other related kinases, whereas WZ4003 selectively inhibits both NUAK1 and the closely related NUAK2 (23). Overexpression of NUAK1 and NUAK2 in human colorectal cancer tends to be mutually exclusive and, accordingly, we detected a reciprocal pattern of NUAK protein expression in colorectal cancer cell lines (Supplementary Fig. S1G and S1H). Treatment with 5 μmol/L HTH-01-015 suppressed proliferation of multiple cell lines, and this effect was reproduced by RNAi-mediated depletion of NUAK1 (Supplementary Fig. S1I–S1L). Strikingly, treatment with 10 μmol/L HTH-01-015 was profoundly toxic in the same cell lines and correlated with a stronger reduction in phosphorylation of the NUAK1/NUAK2 substrate MYPT1, even in cells that express very low levels of NUAK1 (Fig. 1E and F). This cytotoxic effect was also observed using the dual NUAK inhibitor WZ4003, suggesting it reflects on-target activity of the inhibitors. Notably, WZ4003 gave greater suppression of phospho-MYPT1^S445, consistent with dual inhibition of NUAK1 and NUAK2, and showed somewhat greater potency, driving significant cell death at 5 μmol/L in SW480 cells (Supplementary Fig. S1M and S1N). Inhibition of NUAK1 is thus sufficient to drive apoptosis in colorectal cancer cells, and death does not require complete suppression of MYPT1^S445 phosphorylation. In contrast with the colorectal cancer lines, wild-type mouse embryonic fibroblasts (MEF) and U2OS cells were comparatively resistant to both inhibitors, especially to the NUAK1-selectve HTH-01-015 (Fig. 1G; Supplementary Fig. S1O and S1P), consistent with previous data showing that U2OS cells are refractory to NUAK1 depletion (9). Notably, both inhibitors completely suppressed MYPT1 phosphorylation in U2OS cells, indicating that NUAK1 accounts for the vast majority of MYPT1^S445 phosphorylation in this cell type. As we showed previously in MEFs (5), overexpression of MYC strongly sensitized U2OS cells to HTH-01-015–induced apoptosis (Fig. 1H). Conversely, depletion of endogenous MYC rescued colorectal cancer cells from HTH-01-015–induced apoptosis, and rescue was proportional to the degree of MYC depletion, consistent with an ectopic requirement for NUAK1 in cells with deregulated MYC (Fig. 1I).

In order to investigate the in vivo requirement for NUAK1 in colorectal cancer, we bred mice bearing a floxed Nuak1 allele (Nuak1^FL/FL; ref. 11) onto a tamoxifen (Tam)-inducible mouse model of sporadic intestinal cancer: Villin-CreER^T2; Apc^FL/+;lsl-Kras^G12D (VAK). In this model, transient Tam-dependent activation of CreER^T2 in the intestines of adult mice drives widespread deletion of one copy of Apc simultaneous with expression of oncogenic KRAS^G12D; however, tumor formation requires stochastic loss of the second copy of Apc (Supplementary Fig. S2A). In the absence of mutant KRAS, Apc-null polyps are largely restricted to the small intestine (SI), whereas in the presence of mutant KRAS, adenomas form in both large and small intestines (19).

Using a single injection of Tam to transiently activate CreER^T2, we deleted Nuak1 in the intestines of adult Villin-CreER^T2; Apc^FL/+;lsl-Kras^G12D;Nuak1^FL/FL (VAKN) mice and aged mice until symptomatic in order to compare the intestinal tumor burden with that of symptomatic VAK mice. Deletion of Nuak1 profoundly suppressed both the number and size of individual tumors in the colon of VAKN mice, compared with VAK controls (Fig. 2A–D). In contrast, we observed no significant difference in either the number or size of tumors that arose in the SI, and VAKN mice required sacrifice concurrently with VAK mice (Supplementary Fig. S2B–S2D). However, qPCR analysis of Nuak1 expression in individual polyps harvested from the SI of VAKN mice revealed enrichment of Nuak1 mRNA when compared with disease-free adjacent tissue (Supplementary Fig. S1E), indicating a failure of transient CreER activation to efficiently delete Nuak1 in the tumor-initiating population of the SI. The absence of a Nuak1^FL/FL phenotype in SI tumors thus appears to reflect technical failure but is of little clinical relevance given the rarity of SI tumors in human populations. Colonic tumors in VAKN mice presented with comparable levels of nuclear β-catenin and sporadic phospho-ERK1/2 staining, as compared with VAK tumors (Supplementary Fig. S2F); however, all tumors arising in VAKN mice retained detectable expression of Nuak1 mRNA (see Supplementary Fig. S2G for examples), suggesting a selective pressure to retain Nuak1 in colonic tumor epithelium. Importantly, deletion of Nuak1 in otherwise wild-type mice had no apparent effect on small or large intestine architecture or function (Supplementary Fig. S3A–S3C), suggesting that the requirement for Nuak1 in the adult gut is restricted to transformed tissue.

Homozygous deletion of Apc (A^Hom) in the gut rapidly gives rise to a Myc-dependent “crypt progenitor” phenotype, characterized by an extension of the transit-amplifying population into the normally quiescent villi of the SI (17). This phenotype was unimpaired by deletion of Nuak1 in VA^HomN intestines (Supplementary Fig. S3D). VA^HomK transformed gut epithelium gives rise to spheroids when cultured in 3-D ex vivo, reflecting the tumor-initiating capacity of the transformed tissue (24). Primary VA^HomKN colonic epithelium showed reduced spheroid-generating capacity, compared with VA^HomK epithelium (Fig. 2E and F). Nuak1 expression was clearly detectable in the few VA^HomKN spheroids that grew, suggesting that they likely arose from cells that escaped Cre-mediated Nuak1 deletion (Fig. 2G). Interestingly, a similar reduction in spheroid-generating capacity was also observed in primary epithelium isolated from the SI (Fig. 2E and F). Accordingly, pharmacologic inhibition of Nuak1 with either HTH-01-015 or WZ4003 profoundly suppressed formation of spheroids by VA^HomK gut epithelium from both small and large intestines (Fig. 2H and I), whereas wild-type organoids were refractory to treatment over the same time frame (Supplementary Fig. S3E).

Reasoning that key physiologic roles of NUAK1 would be conserved across cell types, we exploited the fact that U2OS cells are refractory to NUAK1 suppression (in the absence of MYC overexpression) and express very little NUAK2 and performed an unbiased transcriptomic analysis after RNAi-mediated depletion of NUAK1. Metacore GeneGO pathway analysis revealed regulation of cholesterol synthesis, cell adhesion, and glutathione metabolism among the top-most pathways modulated upon NUAK1 depletion (Fig. 3A). The role of NUAK1 in regulating cell adhesion via phosphorylation of MYPT1 was described previously (10), whereas the modulation of glutathione metabolism suggested a novel role for NUAK1 in the antioxidant defense pathway. Strikingly, our transcriptomic analysis revealed a coordinated reduction in expression of a host of genes that are regulated by the antioxidant transcription factor NRF2 (encoded by NFE2L2; ref. 25), including the catalytic and regulatory subunits of the glutamate-cysteine ligase; ROS scavengers thioredoxin, peroxiredoxin, and MGST; and glutathione reductase (Fig. 3B; Supplementary Fig. S4A). Acute inhibition of NUAK1 in U2OS cells with HTH-01-015 recapitulated these results (Fig. 3C), as did CRE-mediated deletion of Nuak1 in primary MEFs (Supplementary Fig. S4B and S4C), confirming the conservation of this effect across cells types and species. RNA sequencing (RNA-seq) analysis of SW480 colorectal cancer cells upon depletion of NUAK1 revealed a strong overlap with genes modulated upon depletion of NRF2 (NFE2L2), including several antioxidant pathway genes and the miR17-92 cluster, which was recently shown to negatively regulate LKB1 upstream of NUAK1 (26), suggestive of feedback regulation (Fig. 3D and E). Similar results were obtained in HCT116 cells (Fig. 3D). Pathway analysis showed broadly similar transcriptional effects of depletion of either NUAK1 or NRF2, whereas analysis of downregulated genes revealed significant enrichment for pathways “Oxidative stress” and “NRF2 regulation of oxidative stress” in both instances (Supplementary Tables S1 and S2).

The reduced expression of NRF2 target genes suggested that NUAK1-deficient cells would be hypersensitive to oxidative stress. Accordingly, we detected elevated levels of cytosolic H₂O₂ in U2OS cells, multiple colorectal cancer cell lines, and VAK colonic spheroids, after acute treatment with HTH-01-015 (Fig. 3F–I), whereas depletion of NUAK1 sensitized U2OS cells and multiple colorectal cancer cell lines to H₂O₂-induced cell death, consistent with the inhibitor and RNAi each reducing antioxidant buffering capacity, albeit to different degrees (Fig. 3J). Similar sensitization to H₂O₂-induced cell death was also observed upon CRE-mediated deletion of NUAK1 in MEFs, providing genetic confirmation of the specificity of this effect (Supplementary Fig. S4D and S4E). Colorectal cancer lines with lower levels of NUAK1 were inherently more sensitive to ROS-induced cell death, even in the absence of NUAK1 depletion, compared with cells expressing higher levels of NUAK1, whereas the relatively modest sensitization of SW480 cells compared with U2OS cells may reflect differences in the efficiency of NUAK1 depletion or indeed the relative expression of NUAK2. Notably, depletion of NUAK1 in some colorectal cancer lines resulted in increased NUAK2 expression (Supplementary Fig. S1L). Importantly, provision of exogenous antioxidants significantly rescued human colorectal cancer cells (Fig. 3K; Supplementary Fig. S4F) and VAK spheroids (Fig. 3L; Supplementary Fig. S4G) from NUAK1 inhibitor–induced apoptosis, indicating that ROS contributes substantially to cell death in both settings. The remaining levels of cell death measured in the colorectal cancer cell lines likely reflect exhaustion of the exogenous antioxidant, evidenced by intermediate levels of ROS in cells treated for 8 hours with NUAK1 inhibitor in the presence of Trolox (Supplementary Fig. S4H and S4I).

NRF2 was recently described to contain an AMPK-substrate consensus phosphomotif (27) that could potentially be targeted for phosphorylation by NUAK1. We used immunoprecipitation (IP) of FLAG-tagged NRF2 followed by immunoblotting with a pan–phospho-AMPK-substrate antibody to assess the influence of NUAK1 inhibition on NRF2 phosphorylation levels but detected no difference (Supplementary Fig. S5A). Similarly, purified NUAK1 showed no activity toward a corresponding NRF2 peptide in vitro (Supplementary Fig. S5B). Thus, NRF2 does not appear to be a direct target of NUAK1 kinase activity.

We noticed that acute inhibition of NUAK1 resulted in decreased total NRF2 levels (Fig. 4A). NRF2 is regulated by KEAP1, which sequesters NRF2 in the cytoplasm while targeting it continuously for ubiquitin-dependent degradation (28, 29). We asked if KEAP1 is required for regulation of NRF2 by NUAK1. As expected, RNAi-mediated depletion of KEAP1 increased basal levels of NRF2, yet concomitant inhibition of NUAK1 continued to reduce total NRF2 protein levels (Fig. 4B). Accordingly, cyclohexamide time-course analysis showed that NUAK1 depletion reduces total NRF2 levels but does not affect the rate of NRF2 degradation, per se (Supplementary Fig. S5C). KEAP1 contains a number of cysteine residues that are subject to oxidation and, in the presence of ROS, oxidized KEAP1 releases NRF2, allowing it to translocate to the nucleus and activate transcription (30). We therefore examined NUAK1-depleted or HTH-01-015–treated U2OS cells for nuclear accumulation of NRF2 after acute treatment with H₂O₂ and found that loss of NUAK1 activity strongly suppressed ROS-induced nuclear accumulation of NRF2 (Fig. 4C). Accordingly, ROS-induced transcription of NRF2 targets was also suppressed upon depletion of NUAK1 (Fig. 4D). Analysis of multiple colorectal cancer cell lines likewise revealed that depletion of NUAK1 suppresses ROS-driven NRF2 nuclear accumulation, indicating that this role of NUAK1 is conserved in colorectal cancer (Fig. 4E; Supplementary Fig. S5D). Additionally, this role of NUAK1 is at least partially shared with NUAK2, as depletion of NUAK2 in SW620 cells similarly suppressed peroxide-induced nuclear accumulation of NRF2 (Supplementary Fig. S5E).

We used unbiased, SILAC-based phosphoproteomics to identify candidate mediators of NRF2 regulation upon acute inhibition of NUAK1 in U2OS cells (see schematic, Supplementary Fig. S5F). Ser⁴⁴⁵ of MYPT1 was the only site resident within a recognizable AMPK-related kinase consensus motif that was consistently reduced upon NUAK1 inhibition. This analysis also revealed reduced inhibitory phosphorylation of GSK3β at Ser⁹ and a corresponding increase in phosphorylation of multiple GSK3β targets (Fig. 5A; Supplementary Table S3). GSK3β is known to suppress nuclear accumulation of NRF2: In the presence of oxidative stress, activation of AKT inhibits GSK3β via Ser⁹ phosphorylation, allowing nuclear accumulation of NRF2 (31). We therefore examined the influence of NUAK1 depletion on ROS-driven signal transduction via AKT and GSK3β. Treatment of U2OS cells with H₂O₂ rapidly activated AKT, leading to increased GSK3β^S9 phosphorylation. Upon depletion of NUAK1, activation of AKT by H₂O₂ was unimpaired; however, the inhibitory phosphorylation of GSK3β was strongly reduced, suggesting that NUAK1 may limit dephosphorylation of GSK3β^S9 (Fig. 5B). Similar results were observed upon H₂O₂ treatment of NUAK1-depleted SW480 cells (Fig. 5C), whereas treatment with NUAK1 inhibitor suppressed GSK3β^S9 phosphorylation in SW480 cells and VAK large intestine (LI) spheroids (Fig. 5D and E). Notably, phosphorylation of MYPT1 by NUAK1 inhibits PP1β activity (10), and PP1β was previously shown to dephosphorylate GSK3β (32, 33). Strikingly, H₂O₂ led to a clear increase in NUAK1-dependent MYPT1 phosphorylation (Fig. 5B), suggesting that ROS coordinately activates AKT and inactivates PP1β (via NUAK1) in order to suppress GSK3β activity. Significantly, inhibition of GSK3β stabilized total NRF2 levels and rescued nuclear accumulation of NRF2 in NUAK1-deficienct SW480 cells (Fig. 5F; Supplementary Fig. S5G). Interestingly, depletion of PTEN similarly rescued nuclear NRF2, suggesting that the requirement for NUAK1 in this pathway can be overcome by strongly deregulated AKT signaling (Supplementary Fig. S5H).

We asked if NUAK1 is an integral part of the oxidative stress response pathway. Depletion of NRF2 with two independent siRNAs consistently reduced NUAK1 protein levels (Fig. 6A). Examination of the NUAK1 promoter revealed a near-consensus antioxidant response element (ARE) located approximately 1.2 kb upstream of the NUAK1 transcription start site, and NRF2 chromatin IPs showed specific binding of NRF2 to the putative NUAK1 ARE, albeit at much lower efficiency than to the canonical NRF2 target HMOX1 (Fig. 6B and C). Treatment of U2OS cells with H₂O₂ modestly increased NUAK1 mRNA but had much greater influence on NUAK1 protein, suggesting that posttranslational regulation may have greater functional impact (Fig. 6D and E). Time-course analysis revealed that H₂O₂ treatment rapidly increased activating phosphorylation of NUAK1 at Thr211, and consequent MYPT1^S445 phosphorylation, downstream. These changes occur within the same time frame as increased AKT phosphorylation, known to result from direct inactivation of PTEN by ROS (34), suggesting that ROS may directly modify NUAK1 (Fig. 6F). To investigate this hypothesis, we first used dimedone labeling (35) of cells expressing FLAG-tagged NUAK1 to measure cysteine oxidation after H₂O₂ treatment: Treatment with increasing doses of H₂O₂ resulted in increased dimedone labeling of FLAG-immunoprecipitated NUAK1 (Fig. 6G). Consistently, MS analysis of iodo-acetamide labeling of FLAG-NUAK1 IPs from cells treated for 5 minutes with H₂O₂ similarly revealed increased oxidation of multiple NUAK1 cysteines, as compared with untreated controls (Fig. 6H). Collectively, our data suggest a model wherein ROS-dependent activation of NUAK1 coordinates inhibition of PP1β with activation of AKT in order to counteract suppression of nuclear NRF2 by GSK3β (Fig. 6I).

The above data collectively suggest that NUAK1 may be an excellent target for therapeutic intervention in colorectal cancer. However, the relatively poor potency of the NUAK1 inhibitors used above precludes their use in vivo. We therefore used a doxycycline (dox)-inducible RNAi approach to assess the impact of acute Nuak1 suppression on preexisting tumors. We used Villin-CreER^T2 to limit expression of rtTA³ to the mouse intestine. Upon activation with dox, rtTA³ was then used to drive expression of either of two shRNAs, targeting Nuak1 mRNA from nucleotide 612 or 1533, respectively, stringently selected to specifically deplete NUAK1 as previously described (see Supplementary Methods). Supplementary Fig. S6A shows depletion of NUAK1 in MEFs upon dox-dependent expression of Nuak1 shRNA.

Tumors were initiated in heterozygous floxed Apc (VA) mice by Tam-dependent activation of CreER^T2, and tumor development in the colon was accelerated by treatment with dextran sulfate sodium salt (DSS). DSS-treated VA mice develop colonic polyps within 70 days of CreER^T2 activation with >90% penetrance (36), and this time after induction was chosen to commence dox-dependent induction of either shRNA. Mice were maintained on dox for 1 week and then harvested for analysis (for a schematic, see Supplementary Fig. S6B). DSS-treated VA mice lacking either Nuak1 shRNA or rtTA alleles were similarly administered dox, to control for effects of the antibiotic. Depletion of NUAK1 for just 1 week strongly reduced the number of tumors per mouse and moreover suppressed the size of the remaining tumors found upon examination (Fig. 7A). Similar results were obtained with both Nuak1 shRNA alleles, strongly suggesting that the observed effects reflect the “on-target” depletion of NUAK1. Of the tumors that persisted in Nuak1 shRNA-expressing mice, all expressed readily detectable levels of Nuak1 mRNA, as measured by ISH (Supplementary Fig. S6C), indicating that some tumors escape shRNA-mediated NUAK1 depletion. PEARL imaging of intestines of mice injected overnight with LI-COR ROSstar reagent revealed elevated ROS levels in colonic tumors in situ after just 2 days of NUAK1 depletion (Supplementary Fig. S6D), whereas IHC analysis showed increased oxidative damage (8-oxo-dG), increased apoptosis (TUNEL), and reduced proliferation [bromodeoxyuridine (BrdUrd)] in NUAK1-depleted tumors within the same time frame (Fig. 7B and C). Consistent with our in vitro data, transcriptomic analysis of NUAK1-depleted tumors revealed significantly reduced expression of a host of NRF2 target genes within 2 days of NUAK1 depletion (Supplementary Fig. S6E). Importantly, exogenous provision of the antioxidant N-acetyl-cysteine (NAC) in drinking water reversed the tumor-suppressive effect of NUAK1 depletion (Fig. 7D; Supplementary Fig. S6F), but had no effect on NUAK1-replete tumors (Supplementary Fig. S6G). We conclude from these results that impairment of cellular antioxidant defenses is the underlying mechanism of the tumoricidal effect of NUAK1 suppression in the gut.

Here, we demonstrate that the AMPK-related kinase NUAK1 plays a key role in protecting colorectal tumors from oxidative stress. Using a combination of genetic and pharmacologic approaches, we show that NUAK1 is required for both formation and maintenance of colorectal tumors after loss of APC; that suppression of NUAK1 reduces viability of transformed intestinal spheroids and of human colorectal cell lines; and that protecting cells from toxic levels of ROS, via facilitation of NRF2-dependent antioxidant gene expression, is a key tumor-promoting activity of NUAK1. We show that NUAK1 kinase activity is rapidly increased by ROS following cysteine oxidation and, moreover, that NUAK1 is transcriptionally regulated by NRF2, placing NUAK1 squarely within the oxidative stress response pathway. Noting that NUAK1 expression is normally highest in highly oxidative tissues (11), it thus appears that protecting cells from oxidative stress is a major physiologic role of NUAK1 that has been co-opted by tumor cells to support their survival in the typically harsh tumor microenvironment. AMPK also participates in antioxidant defense, albeit indirectly, by conserving NADPH levels via inhibition of lipid biosynthesis (37), and a recent paper has shown a genetic requirement for this activity in MYC-overexpressing melanoma (38). Although AMPK may under certain circumstances directly phosphorylate NRF2 (27), in our system, the observed level of NRF2 phosphorylation is extremely low and is not modulated by NUAK1 inhibition. As such, AMPK does not presently appear to contribute to regulation of NRF2 by NUAK1.

Instead, we show that NUAK1 facilitates nuclear import of NRF2 by counteracting negative regulation of this process by GSK3β, and that direct inhibition of GSK3β restores NRF2 nuclear import in NUAK1-deficient cells. ROS-mediated inactivation of PTEN activates AKT, resulting in direct inhibitory phosphorylation of GSK3β on Ser9 (31, 39). This phosphorylation is opposed by PP1β, which reactivates GSK3β (33). We show that activation of AKT by ROS is unaffected by NUAK1 suppression; however, NUAK1 facilitates AKT-dependent regulation of GSK3β by inhibiting PP1β via phosphorylation of the PP1β regulatory subunit MYPT1. NUAK1 is thus required to coordinate inhibition of PP1β with AKT activation in response to ROS, thereby allowing GSK3β to be switched off long enough to permit NRF2 nuclear accumulation, providing fascinating new insight into temporal coordination of Redox signal transduction. This role of NUAK1 is likely to be shared with NUAK2, which similarly suppresses PP1β via MYPT1, and, indeed, we show that depletion of NUAK2 similarly reduces nuclear NRF2 in cells that highly express NUAK2. However, further work is needed to distinguish between specific effects of NUAK1 and NUAK2 on PP1β and beyond.

This mechanism of regulation suggests that the effects of NUAK1 suppression may be quite pleiotropic and, indeed, our phosphoproteomic analysis indicated modulation of multiple GSK3β targets in addition to NRF2. Moreover, transcriptional regulation by NRF2 reaches far beyond antioxidant gene expression, as previously noted (25). However, the central role of NRF2-dependent antioxidant gene expression in supporting tumor cell viability is attested to (i) by the hypersensitivity of NUAK1-depleted colorectal cancer tumor lines to oxidative stress–induced apoptosis and (ii) by the dramatic rescue of NUAK1-depleted colonic tumors and inhibitor-treated spheroids upon provision of exogenous antioxidants. The more modest (but nonetheless significant) antioxidant rescue observed in HTH-01-015–treated colorectal cancer cells likely reflects the limits of trying to buffer against oxidative stress in standard cell culture (40). Although attempts to recapitulate the cytotoxic effects of NUAK1 inhibition in colorectal cancer cells using RNAi were unsuccessful, we believe that the effects of HTH-01-015 are specific for NUAK1 for several reasons: (i) this compound has been tested against more than 120 kinases and is extremely selective for NUAK1, although at higher concentrations it does show some activity toward NUAK2 and possibly MARK3 (23); (ii) cytotoxicity was observed only at concentrations that yielded a clear reduction in MYPT1 phosphorylation, thus indicating greater suppression of either NUAK1 or a NUAK1-like activity; (iii) cytotoxicity was reproducible with the unrelated compound WZ4003; (iv) consistent with our previous demonstration of a synthetic lethal relationship between MYC and NUAK1 (9), sensitivity to HTH-01-015 was MYC dependent and colorectal cancer cell death was rescued by MYC depletion. It is thus unclear why cytotoxicity was not observed using RNAi in the cell-culture setting, except in instances of simultaneous peroxide challenge. It may be that very low levels of residual NUAK1 suffice to suppress cell death, consistent with our data in SW620 cells, which do express very low levels of NUAK1. Additionally, the asynchronous nature of RNAi may allow cultured cell populations time to quench H₂O₂ before the threshold for loss of viability is breached and, accordingly, depletion of NUAK1 resulted in upregulation of NUAK2 in multiple colorectal cancer lines, likely dampening the impact of NUAK1 depletion. Furthermore, HTH-01-015 has been shown to partially inhibit NUAK2 at the 10 μmol/L dose that exhibited cytotoxicity in colorectal cancer lines (23) and, although we cannot entirely exclude the possibility of an off-target effect of the inhibitor, the fact that colorectal cancer cytotoxicity at this dose was significantly rescued by both antioxidant provision and by depletion of c-MYC strongly supports our interpretation that the on-target effect of the inhibitor is responsible for induction of tumor cell death. The differential sensitivity of some cells (e.g., SW480) to the NUAK1 inhibitor versus peroxide challenge after NUAK1 depletion by RNAi may thus reflect expression of NUAK2 and/or the continued biochemical activity of residual levels of NUAK1 after RNAi-mediated depletion.

Our previous work linked the selective requirement of tumor cells for NUAK1 to MYC overexpression, and this link is borne out here by the rescue of NUAK1 inhibitor–induced death upon depletion of MYC from colorectal cancer cell lines. In the intestine, loss of Apc leads to β-catenin–dependent overexpression of endogenous MYC. Although deregulated MYC is alone insufficient for intestinal tumor formation (41), it is nonetheless required for β-catenin–driven polyposis and, significantly, is also required for the elevation of ROS levels observed in vivo upon loss of Apc (42).

Colorectal tumors will thus have evolved in the face of continuous oxidative stress, and cells derived therefrom would likely be better buffered against oxidative stress than cells (e.g., U2OS) that lack MYC deregulation. Accordingly, we show that U2OS cells depleted of NUAK1 are exquisitely sensitive to a peroxide challenge and that this is phenocopied by MYC overexpression. Note that the absence of NUAK2 expression from U2OS cells likely increases their reliance upon NUAK1. NUAK1 thus functions in two major tumor-protective pathways, ATP homeostasis and the oxidative stress response, that are rapidly engaged to support viability upon MYC overexpression (43). As such, NUAK1 appears to be more intimately linked with the downstream metabolic consequences of MYC deregulation than with the absolute levels of MYC protein per se, and we recently linked MYC deregulation to calcium-dependent activation of NUAK1 in LKB1-deficient cells (13).

Exploiting the heightened sensitivity of tumor cells to ROS is emerging as a plausible strategy for cancer therapy (44, 45). Recently, intravenous injection of very high doses of dihydroascorbate was shown to suppress colorectal tumor formation by saturating ROS scavengers, and subsequent work suggests that this strategy may indeed show clinical benefit (46, 47). With increasing evidence linking elevated NRF2 to aggressive disease (48, 49), disabling antioxidant defenses via transient inhibition of NUAK1 may offer a new strategy for improving therapeutic outcomes in cancer.

All experiments involving mice were approved by the local ethics committee and conducted in accordance with UK Home Office license numbers 70/7950 and 70/8646. Mice were housed in a constant 12-hour light/dark cycle, fed, and watered ad libitum. Mice bearing dox-inducible shRNAs targeting Nuak1 are described in the Supplementary Materials section. All mice were maintained on mixed (FVBN × C57Bl/6 × 129/SV) background, and littermate controls were used for all experiments. To induce allele recombination, transient activation of CreER^T2 in the intestine was performed on mice ages 6 to 12 weeks via single IP injection of 50 mg/kg Tam. For survival analysis, humane endpoints were defined as exhibition of two or more symptoms: >15% weight loss, pale feet, lethargy, and bloody stool. Where indicated, 1.75% DSS, m.w. 35–50 kDa (M.P. Biochemicals) was administered in drinking water for 5 days, commencing 4 days after allele induction, followed by distilled water for 1 week, then tap water. Doxycycline (Sigma; in H₂O) was administered by oral gavage in 2 mg daily boluses, from day 64 to day 70 after induction. N-acetylcysteine (Sigma; 4% w/v solution) was administered in drinking water, starting 3 days before shRNA induction, and replaced every 3 to 4 days until sacrifice. All mice were sacrificed using a schedule 1 procedure. ROSstar 650 reagent (LI-COR) was injected IV the day before tissue harvesting and signal was detected by PEARL imaging.

Primary spheroid cultures of intestinal crypts were established as previously described (50) from the SI and colon of VA^HomK and VA^HomKN mice: Adult mice were induced as above, and tissues were harvested 4 days later. Intestines were flushed with ice-cold PBS and opened longitudinally, and villi were removed using a glass coverslip. Intestines were incubated in EDTA/PBS (2 mmol/L for SI; 25 mmol/L for LI) for 30 minutes at 4°C. Excess solution was discarded, and loose intestine fragments were collected by manual trituration in 3 PBS washes. The crypt-enriched fractions were passed through a 70 μmol/L cell strainer and pelleted at 600 rpm for 2 minutes in a table-top centrifuge. Resuspended crypts were counted by hemocytometer, then seeded in Matrigel (BD Bioscience) with advanced DMEM/F12 media (Invitrogen), supplemented with 10 mmol/L HEPES; 2 mmol/L glutamine; 0.1% FBS; penicillin/streptomycin, N-2 and B-27 supplements (1×, Invitrogen). Alternatively, for quantification of primary spheroid formation, isolated crypts were further incubated in Cell Dissociation solution (Thermo) until a single-cell suspension was achieved. Cells were then counted and seeded at normalized density as above. Growth factors Noggin (100 ng/mL) and EGF (50 ng/mL; Peprotech) were added to primary cultures but removed from subsequent passages. Spheroids were counted manually 3 or 4 days after seeding. Wild-type organoid cultures were prepared similarly but additionally supplemented with R-Spondin (500 ng/mL; R&D Systems). Established crypt cultures were split 1 to 2 times per week by manual disruption followed by incubation in Cell Dissociation solution (Thermo) until a single-cell suspension was achieved. Cells were then counted and reseeded at normalized density. NUAK1 inhibitors, HTH-01-015 (Apex Biotech) or WZ4003 (Medchem Express) in DMSO, were added to single-cell suspensions at the indicated concentrations. Trolox [(±)-6-Hydroxy-2,5,7,8-tetramethylchromane-2-carboxylic acid; Sigma] was added to single-cell suspensions at a final concentration of 500 μmol/L for 16 hours prior to HTH-01-015 and replenished daily for 3 days. ROS detection was performed by confocal fluorescent microscopy using 5 μmol/L CellRox green (Thermo; 3 hours at 37°C) after overnight treatment of preformed spheroids with HTH-01-015.

U2OS (2009), HCT116, SW620, and SW480 (all in 2013) cell lines were obtained from the ATCC and cultured in DMEM supplemented with penicillin/streptomycin and 10% FBS. Cells were expanded initially upon receipt and aliquoted into frozen stocks. Upon resuscitation, cells were passaged as required and discarded after no more than 3 months of continuous culture. Cell lines were periodically validated using the Promega Geneprint 10 authentication kit, most recently in August 2017. All cell lines in culture were tested every 3 months for Mycoplasma.

Whole-transcriptome analysis was performed by Illumina RNA-seq. The following datasets are available through ArrayExpress: U2OS ± NUAK1 shRNA, accession number E-MTAB-6244; SW480 ± siNRF2 or siNUAK1, accession number E-MTAB-6264; Apc/DSS-induced colonic tumors ± shNUAK1, accession number E-MTAB-6265. A full description of methodology is provided in the Supplementary Materials.

All experiments were performed at least 3 times except where noted in the text. Raw data obtained from quantitative real-time PCR, FACS, and spheroid generation assays were copied into Excel (Microsoft) or Prism (Graphpad) spreadsheets. All mean and SEM values of biological replicates were calculated using the calculator function. Graphical representation of such data was also produced in Excel or in Prism. Box and spider plots were generated using Prism. Statistical significance for pairwise data was determined by the Student (unpaired) or paired t test, as indicated. For multiple comparisons, ANOVA was used with a post hoc Tukey test. *, P < 0.05; **, P < 0.01; ***, P < 0.001. For Kaplan–Meier plots, Mantel–Cox log-rank P values are presented; for tumor enumeration, Mann–Whitney tests were performed.

Additional methods are described in the Supplementary Materials.

No potential conflicts of interest were disclosed.

Conception and design: J. Port, N. Muthalagu, H. Esumi, O.J. Sansom, D.J. Murphy

Development of methodology: J. Port, N. Muthalagu, F. Ceteci, B. Kruspig, M. Mezna, O.J. Sansom

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Port, N. Muthalagu, M. Raja, F. Ceteci, T. Monteverde, B. Kruspig, S. Lilla, L. Neilson, K. Gyuraszova, S. Svambaryte, A. Bryson, D. Sumpton, M. Drysdale, H. Esumi, G.I. Murray, S.R. Zanivan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Port, N. Muthalagu, M. Raja, F. Ceteci, B. Kruspig, A. Hedley, G. Kalna, S. Lilla, K. Gyuraszova, D. Sumpton, S.R. Zanivan, D.J. Murphy

Writing, review, and/or revision of the manuscript: J. Port, N. Muthalagu, G. Kalna, H. Esumi, O.J. Sansom, D.J. Murphy

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Port, N. Muthalagu, M. Brucoli, J. Tait-Mulder, M. Mezna, A. McVie, C. Nixon, M. Drysdale, G.I. Murray

Study supervision: O.J. Sansom, D.J. Murphy

Other (performing the experiments): N. Muthalagu

The authors wish to thank the staff of the CRUK Beatson Institute Biological Services Unit for animal husbandry and assistance with in vivo experiments; the staff of the CRUK BI Histology core facility and William Clark of the NGS core facility; David McGarry, Rene Jackstadt, Jiska Van der Reest, Justin Bower, and Heather McKinnon for many helpful discussions, and countless colleagues at the CRUK BI and Glasgow Institute of Cancer Sciences for support; Prem Premsrirut & Mirimus Inc. for design and generation of dox-inducible Nuak1 shRNA–expressing mice; and Nathanael Gray for initial provision of NUAK1 inhibitors. Funding was provided by the University of Glasgow and the CRUK Beatson Institute. J. Port was supported by European Commission Marie Curie actions C.I.G. 618448 “SERPLUC” to D.J. Murphy; N. Muthalagu was supported through Worldwide Cancer (formerly AICR) grant 15-0279 to O.J. Sansom and D.J. Murphy; B. Kruspig was funded through EC Marie Curie actions mobility award 705190 “NuSiCC”; T. Monteverde was funded through British Lung Foundation grant APHD13-5. The laboratories of S.R. Zanivan (A12935), O.J. Sansom (A21139), and M. Drysdale (A17096) are funded by Cancer Research UK. O.J. Sansom was additionally supported by European Research Council grant 311301 “ColoCan.”