Hypoxia in the microenvironment of many solid tumours is an important determinant of malignant progression. The ISR (integrated stress response) protects cells from the ER (endoplasmic reticulum) stress caused by severe hypoxia. Likewise, autophagy is a mechanism by which cancer cells can evade hypoxic cell death. In the present paper we report that the autophagy-initiating kinase ULK1 (UNC51-like kinase 1) is a direct transcriptional target of ATF4 (activating transcription factor 4), which drives the expression of ULK1 mRNA and protein in severe hypoxia and ER stress. We demonstrate that ULK1 is required for autophagy in severe hypoxia and that ablation of ULK1 causes caspase-3/7-independent cell death. Furthermore, we report that ULK1 expression is associated with a poor prognosis in breast cancer. Collectively, the findings of the present study identify transcriptional up-regulation of ULK1 as a novel arm of the ISR, and suggest ULK1 as a potentially effective target for cancer therapy.
The microenvironment of a solid tumour is both hostile and heterogeneous, characterized by extremes in pH, metabolite concentrations and, in particular, partial pressure of oxygen. Rapid tumour growth leads to oxygen gradients in vivo [1,2]. Oxygen levels can be extremely low, and up to 35% of tumour cells have levels below 0.1% O2 . Patients with hypoxics tumour have a significantly worse disease-free and overall survival compared with patients with well-oxygenated tumours [4,5].
Several intracellular signalling pathways known to be involved in cancer progression and adaptation are activated by hypoxia, including the well-described HIFs (hypoxia-inducible factors) [6,7]. In addition, in prolonged or severe hypoxia (<0.01% O2), dysfunctional protein folding in the lumen of the ER (endoplasmic reticulum) leads to the activation of the UPR (unfolded protein response) [8,9]. The UPR sensors PERK [PKR (double-stranded-RNA-dependent protein kinase)-like ER kinase], IRE1 (inositol-requiring enzyme 1) and ATF (activating transcription factor) 6 detect the accumulation of unfolded proteins, and elicit an adaptive response to overcome this stress .
The cytoplasmic kinase domain of activated PERK phosphorylates eIF2α (eukaryotic initiation factor 2α) and thus represses global mRNA translation [11,12], reducing the burden on the ER protein folding machinery. In addition, eIF2α phosphorylation enables the selective translation of a subset of transcripts such as the key transcriptional effector ATF4 [13,14]. ATF4 regulates the expression of genes involved in translation, amino acid import and metabolism, redox, protein secretion, glucose metabolism, DNA damage, lipogenesis, angiogenesis, and apoptosis in what has been termed the ISR (integrated stress response) [9,15–18]. Studies have highlighted the importance of the ISR in cancer cell survival and tumour maintenance. Data indicate that ATF4 and the ISR promote the survival of hypoxia, nutrient deprivation and oxidative stress in vitro [9,19–22] and tumour progression in mouse xenograft models [19,23,24].
Cancer cells rely on autophagy to survive intratumoral hypoxia, metabolic stress and starvation, and to resist radiotherapy [19,21,22,25]. Intriguingly, pancreatic cancer cells and H-Ras-expressing tumour cells have elevated levels of basal autophagy, and appear to rely on autophagy to maintain high levels of oxidative phosphorylation and a pool of viable mitochondria, in a phenomenon that has been termed ‘autophagy addiction’ [26–28].
We and others have demonstrated that some components of the autophagic machinery are subject to ISR regulation. Severe hypoxia and ER stress induce the transcription of the essential autophagy genes MAP1LC3B (microtubule-associated protein 1 light chain 3β) and ATG5 (autophagy-related 5) through the activity of ATF4 and CHOP [C/EBP (CCAAT/enhancer-binding protein)-homologous protein] respectively [19,21]. This up-regulation was shown to be crucial for maintaining high levels of autophagic flux in persistent hypoxia, and thus promotes cell survival. Similarly, we have demonstrated that ATF4 and autophagy are crucial for the resistance of cancer cells to bortezomib . In the present study, we demonstrate that ATF4 drives the transcriptional induction of the autophagy-initiating kinase ULK1 (UNC51-like kinase 1) in response to severe hypoxia and ER stress. Furthermore, we show that ULK1 and ATF4 are crucial for autophagy, mitophagy and cancer cell survival in both normoxia and severe hypoxia, and that loss of ULK1 promotes necrotic cell death. Finally, we show that elevated ULK1 mRNA expression is associated with poor prognosis in a specific group of breast cancer patients.
MATERIALS AND METHODS
Reagents and cell culture
TG (thapsigargin) (catalogue number 586005), TN (tunicamycin) (catalogue number 654380), MG115 (catalogue number 474780), bafilomycin A1 (catalogue number 196000) and pepstatin A (catalogue number 516481) were obtained from Calbiochem, and E64D (catalogue number E8640) and chloroquine (catalogue number C6628) were obtained from Sigma–Aldrich. Bortezomib was a gift from Millennium Pharmaceuticals. A431, MCF7, HT29, MDA-MB-231, HCT116 and U87 cell lines were provided by Cancer Research UK and were cultured in DMEM (Dulbecco's modified Eagle's medium; Invitrogen, catalogue number 11965) with 10% (v/v) FBS (fetal bovine serum), supplemented with 0.5 μg/ml penicillin/streptomycin (Invitrogen, catlogue number 15070) and 4 mM L-glutamine (Invitrogen, catalogue number 25030). Cell lines were incubated at 37°C in a non-humidified incubator with 95% air supplemented with 5% CO2.
For incubations in severe hypoxia (<0.01% O2), the InVivO2 400 humidified gas-sorted incubator with a glove box (Ruskin) was used. A gas mix of 5% H2, 5% CO2 and 90% N2 was fed into the system. A palladium catalyst maintained near-zero O2 concentrations by reducing trace O2. For incubations in moderate hypoxia (0.1% O2), a Heto-Holten CellHouse 170 incubator (RS Biotech) was used.
Transfections of siRNA (small interfering RNA) and DNA were performed with HiPerfect (Qiagen, catalogue number 301704) and FuGENE® 6 (Roche, catalogue number 11814443001) transfection reagents respectively, as described previously . The specific siRNA sequences used can be found in Supplementary Table S1 (at http://www.biochemj.org/bj/449/bj4490389add.htm).
Western blot analysis
RT-qPCR (real-time quantitative PCR)
In RT-qPCR experiments, the methods used by us for extraction, quantification and evaluation of the quality of RNA were as described previously . Data from RT-qPCR experiments were normalized to tubulin α6. The sequences of primers used are listed in Supplementary Table S2 (at http://www.biochemj.org/bj/449/bj4490389add.htm).
Immunocytochemistry for LC3 (microtubule-associated protein light chain 3)
Staining for LC3-positive foci was performed as described previously .
ULK1 activity assays
Endogenous ULK1 protein was immunoprecipitated from A431 and MCF7 cells and used in subsequent analysis of MBP (myelin basic protein) phosphorylation. A detailed method is given in the Supplementary Online Data.
ChIP (chromatin immunoprecipitation)
The ChIP assays were performed using the EZ ChIP Chromatin Immunoprecipitation kit (Millipore, catalogue number 17-371), according to the manufacturer's instructions, using 5 μg of rabbit anti-ATF4 antibody. The sequences of primers used for RT-qPCR of promoter fragments are shown in Supplementary Table S3 (at http://www.biochemj.org/bj/449/bj4490389add.htm).
Imagestream® analysis of autophagy
MCF7 cells were stained with a live/dead marker, Lysotracker Red, and anti-LC3 antibodies and fixed and analysed for autophagic flux using Imagestream® flow-cytometric methods as described previously .
MCF7 and A431 cells were stained with Alexa Fluor® 647-conjugated anti-AV (annexin-V) antibody (Invitrogen, catalogue number A23204) and cell death was quantified by flow cytometry according to the manufacturer's protocol and as decribed previously  using a Cyan ADP Flow Cytometer (Dako). Mitochondrial fluorescence was measured by first staining cells with Mitotracker Red FM (Invitrogen, catalogue number M22425). Mean fluorescence was used to assess changes in mitochondrial mass, as described previously .
Clonogenic assays were performed to assess survival in severe hypoxia following transfection with siRNA molecules specific for ATF4 or ULK1, as described previously . For the counting of colonies, the ColCount Colony Counter (Oxford Optronix) was used.
Caspase 3/7 protease activity was analysed using the CaspaseGlow® 3/7 assay (Promega, catalogue number G8091) according to the manufacturer's protocol. Similarly, ATP levels were measured using the Cell-Titer Glo® Luminescent Cell Viability assay (Promega, catalogue number G7570) according to the manufacturer's protocol.
Immunohistochemistry was performed on spheroids to assess clinically relevant markers of hypoxia, cell death and proliferation as described previously .
Three-dimensional growth assay
Rapid generation of single-tumour spheroids was achieved using the method developed by Ivascu and Kubbies . An Axiovert 135 Microscope (Zeiss) was used for image capture and ImageJ version 1.43 (NIH) was used to calculate spheroid volume.
HMGB1 (high-mobility group box 1) ELISA
An HMGB1 sandwich ELISA from IBL International was used to quantitatively determine the levels of HMGB1 in the cell culture medium of cells undergoing necrosis, according to the manufacturer's protocol.
For survival analysis, a published series of breast cancer patients was used. mRNA was extracted from tumour samples, and expression profiling was performed as described previously . Full details are given in the Supplementary Online Data. Written informed consent was obtained and all clinical investigations were conducted according to the ethical standards and principles expressed in the Declaration of Helsinki. Ethical approval was obtained from the local research ethics committee.
ULK1 mRNA and protein levels increase in response to severe hypoxia, ER stress, moderate hypoxia and anti-angiogenic therapy
In previous work, we identified hypoxia- and ATF4-dependent genes by microarray analysis . The top 25 of these genes were validated by RT-qPCR analysis and published . Further analysis of our data revealed that ULK1 mRNA was increased 1.76-fold by severe hypoxia (P=0.001) and then reduced to 0.60-fold after knockdown of ATF4 (P=0.004).
To confirm the microarray results identifying ULK1 as a hypoxia-inducible gene, a panel of cell lines were exposed to severe hypoxia for increasing periods of time and ULK1 mRNA levels were assessed by RT-qPCR. In A431, HT29 and MCF7 cells, ULK1 mRNA was seen to increase after 12 h in severe hypoxia, and to reach a maximum after 48 h (Figures 1A–1C). A431 and HT29 cells exhibited close to 30-fold increases in ULK1 mRNA (P<0.05 and P<0.01 respectively), whereas MCF7 cells exhibited a more modest increase of 5-fold (P<0.01).
ULK1 is induced by ER stress and hypoxia in vivo and in vitro
Consistent with the mRNA data, ULK1 protein levels in A431 and HT29 cells increased markedly after 12 and 6 h in severe hypoxia respectively (Figures 1D and 1E). In contrast, in MCF7 cells ULK1 protein levels were found to be high in normoxia, and then decreased after 48 h in severe hypoxia (Figure 1F). In U87 and HCT116 cells, ULK1 protein peaked at 6 h and then rapidly decreased with prolonged exposure to severe hypoxia (Supplementary Figures S1A and S1B at http://www.biochemj.org/bj/449/bj4490389add.htm).
Maximal induction of ULK1 mRNA and protein was preceded by the accumulation of ATF4 protein (Figures 1A–1F). As reported previously [19,21], LC3-II protein levels increased markedly after 12–24 h in severe hypoxia, indicating that autophagy was increased in this condition. Likewise, the expression of the ISR target gene CHOP showed an mRNA and protein expression profile that mimicked that of ULK1 (Supplementary Figures S1C–S1E).
To confirm that LC3-II protein levels were indicative of increased autophagic flux rather than decreased lysosomal degradation, p62 protein levels were examined. Indeed, severe hypoxia resulted in increased degradation of p62 protein in concert with LC3-II protein induction, whereas LC3-I remained constant throughout (Figures 1G and 1H).
Similarly, when A431 cells were exposed to TG or TN, potent stimulators of the ISR and autophagy , ATF4 protein accumulated after 2 h and was followed by ULK1 protein expression after 24 or 6 h respectively (Figures 1I and 1J). When A431 cells were exposed to moderate hypoxia (0.1% O2), a similar pattern of induction was observed (Figure 1K), indicating that this phenotype is not limited to severe hypoxia.
To assess whether ULK1 may be induced by intratumoral hypoxia, mRNA was isolated from U87 tumour xenografts treated with PBS or bevacizumab [humanized anti-VEGF (vascular endothelial growth factor) antibody], a treatment well known to reduce intratumoral oxygen levels [37,38]. ULK1 mRNA levels were increased nearly 2-fold by bevacizumab treatment (P<0.001), in a fashion similar to CHOP (Supplementary Figure S2A at http://www.biochemj.org/bj/449/bj4490389add.htm). Similarly, when A431 cells were grown as spheroids in vitro, ULK1 protein was found to accumulate after 3 days, and to reach a maximum after 6 days growth, correlating with expression of ATF4 and LC3-II (Supplementary Figure S2B). These spheroids were confirmed to be hypoxic by staining for pimonidazole, CAIX (carbonic anhydrase IX) and CHOP (Supplementary Figure S2C).
As mentioned above, MCF7 cells exposed to severe hypoxia strongly induced ULK1 mRNA, although ULK1 protein levels were seen to decrease. Interestingly, proteasome inhibition with bortezomib or MG115 partially blocked the reduction in ULK1 caused by prolonged exposure to severe hypoxia (Supplementary Figure S3A at http://www.biochemj.org/bj/449/bj4490389add.htm), whereas inhibitors of the lysosome or caspases failed to rescue ULK1 expression (Supplementary Figures S3B and S3C).
Severe hypoxia and ER stress induce ULK1 in an ATF4-dependent manner
Our microarray analysis identified ULK1 as a transcriptional target of ATF4 . To confirm this result, A431 and MCF7 cells were treated with siRNA to deplete ATF4. In both cell lines, this treatment reduced ATF4 protein by at least 50% after exposure to severe hypoxia (Figures 2A and 2C). ATF4 knockdown reduced CHOP protein (Figures 2A and 2C) and mRNA induction (Supplementary Figures S4A and S4B at http://www.biochemj.org/bj/449/bj4490389add.htm), indicating that the transcriptional activity of ATF4 was attenuated. Similarly, ULK1 protein expression in both A431 and MCF7 cells was abrogated in both severe hypoxia and normoxia (Figures 2A and 2C), as was ULK1 mRNA induction (Figures 2B and 2D) (P<0.05 and P<0.01 for A431 and MCF7 respectively). Likewise, the accumulation of ULK1 protein caused by TG or TN treatement was completely blocked when A431 cells were first depleted of ATF4 by RNAi (RNA interference) (Figure 2E). Interestingly, however, ectopic expression of ATF4 in normoxia was not sufficient to induce the expression of ULK1 in A431 cells to the levels seen in severe hypoxia (Supplementary Figure S5 at http://www.biochemj.org/bj/449/bj4490389add.htm).
ATF4 binds directly to the ULK1 promoter to drive ULK1 expression in severe hypoxia or due to ER stress
ATF4 binds directly to the promoter region of ULK1
Using the Genomatix® online software suite, we were able to identify several putative CREB (cAMP-response-element-binding protein)/ATF-binding sites (containing the TGACG core sequence) in the ULK1 promoter. Thus, to determine whether ATF4 binds directly to the promoter region of ULK1, ChIP was performed using antibodies specific for ATF4.
Using RT-qPCR primers spanning the region 2000 bp upstream of the transcription start site, the first exon and 500 bp downstream of translation start (Figure 2F), we observed that ATF4 ChIP samples showed the greatest enrichment in fragments corresponding to the CREB site spanning −306 bp to −285 bp relative to the ATG of exon one (Figure 2G). Furthermore, the signal obtained from this primer set increased with exposure to severe hypoxia, reflecting increased ATF4 binding with severe hypoxia (Figure 2H). A previously published set of primers designed against the CHOP promoter was used as a positive control .
Severe hypoxia increases cellular ULK1 kinase activity through both transcriptional and post-translational mechanisms
To date, most work has focused on post-translational regulation of the specific activity of ULK1, demonstrating an important role for AMPK (AMP-activated protein kinase)- and mTOR (mammalian target of rapamycin)-mediated phosphorylation of ULK1 in fasting and fed states respectively [39–43], rather than the transcriptional regulation of total ULK1 levels. Thus, having found that ULK1 mRNA up-regulation by ATF4 results in increased protein expression in A431 cells but not in MCF7 cells, we next sought to assess the total and specific activity of ULK1 in each case.
The total immunoprecipitated ULK1 activity total P-MBP (P- indicates phosphorylated protein)] from A431 cells increased by 103% after 24 h in severe hypoxia (Figure 3A), although, interestingly, the specific ULK1 activity (normalized to protein expression) was 35% lower (Figure 3B). Thus, in spite of decreased enzymatic activity, the ATF4-mediated increase in ULK1 protein led to increased ULK1 activity and autophagy in severe hypoxia. In MCF7 cells, on the other hand, although ULK1 protein did not accumulate in severe hypoxia, its specific activity increased by 115% after 12 h in severe hypoxia (P<0.05) (Figure 3B), yielding an overall increase in total ULK1 activity of nearly 81% at this time point (P<0.05) (Figure 3A). However, this increase in ULK1 activity was transient and, by 24 h in severe hypoxia, both specific and absolute ULK1 activity had decreased to normoxic levels (Figures 3A and 3B).
ULK1 activity increases with severe hypoxia and is dependent on ATF4
These changes in the activity of ULK1 were associated with increased AMPK signalling, as reflected by the increased level of P-AMPK and its products P-ACC (acetyl-CoA carboxylase) and P-Raptor (regulatory associated protein of mTOR), and decreased mTOR signalling, as reflected by decreased P-p70S6K (S6 kinase) (Figures 3C and 3D). Strikingly, severe hypoxia also caused a downward shift of the ULK1 band, which may be indicative of hyper-dephosphorylation, as observed previously in starvation .
ULK1 is required for hypoxia-induced autophagy
To confirm the role of ATF4 in the transcriptional regulation of ULK1 activity, A431 cells expressing an inducible lentiviral shRNA (short hairpin RNA) sequence against ATF4 were treated with doxycycline to silence ATF4 and then exposed to severe hypoxia. ATF4 was depleted at least 80% by this treatment, which caused a 50% reduction in ULK1 protein (Supplementary Figure S6C at http://www.biochemj.org/bj/449/bj4490389add.htm) and total immunoprecipitated ULK1 activity (P<0.05) (Figure 3E) after 36 h in severe hypoxia, despite no change in the specific activity of ULK1 (Supplementary Figure S6D).
ULK1 is required for hypoxia-induced autophagy and mitophagy
To investigate the requirement for ULK1 in hypoxia-induced autophagy, we depleted ULK1 using RNAi by 90% (Supplementary Figure S7A at http://www.biochemj.org/bj/449/bj4490389add.htm), and exposed A431 and MCF7 cells to severe hypoxia. Indeed, knockdown of ULK1 severely reduced the accumulation of LC3-II protein (the lipidated form of LC3, a marker for autophagy) in A431 cells (Figure 4A) and to a lesser extent in MCF7 cells (Figure 4B) in severe hypoxia. Likewise, ULK1 knockdown resulted in the accumulation of p62 protein after 24 h in severe hypoxia, indicating that ablation of ULK1 resulted in reduced autophagic flux (Figures 4A and 4B). Knockdown of ATF4, was less consistent, and even resulted in increased levels of autophagy in some cases.
Staining revealed that ULK1 knockdown abrogated the formation of LC3-positive autophagosomes in both cell lines (Figures 4C and 4D). Likewise, cells depleted of ULK1 failed to accrue autophagosomes in the presence of chloroquine (an inhibitor of autophagosome fusion with lysosmes) (Supplementary Figure S7B), again indicating that ULK1 is required for the initiation of autophagy in severe hypoxia.
Furthermore, through the use of Imagestream® analysis of autophagic flux  (co-localization of endogenous LC3 protein and lysosomal stain), we found that ablation of ULK1 reduced flux in MCF7 cells by 90% in severe hypoxia (P<0.001) and, intriguingly, by more than 80% in normoxia (P<0.001) (Figures 4E and 4F). Thus ULK1 is involved in both basal autophagy and hypoxia-induced autophagy.
ULK1 has been shown previously to play an important role in the regulation of mitophagy in starvation and red blood cell maturation [33,43]. When mitochondrial mass was assessed by flow cytometry, it was found that depletion of ULK1 or ATF4 increased mean mitochondrial fluorescence by 40% and 20% respectively in both normoxia and hypoxia (P<0.001) (Figure 4G), suggesting that ULK1 and ATF4 are required for both constitutive and hypoxia-induced mitophagy.
ULK1 and ATF4 are required for growth and survival in normoxia and severe hypoxia
Having established that ULK1 is required for autophagy and mitophagy in both normoxia and severe hypoxia, we next sought to assess the effects of their loss on cell death. Depletion of ULK1 by RNAi markedly increased AV staining in A431 and MCF7 cells in both normoxia and severe hypoxia (P<0.001) (Figures 5A and 5B). Depletion of ULK1 and ATF4 reduced clonogenicity by 65% and 80% respectively in both normoxia and severe hypoxia (P<0.001) (Figure 5C). Likewise, knockdown of ULK1 or ATF4 in severe hypoxia reduced cellular ATP levels by more than 75% (Figure 5D).
ULK1 and ATF4 are required for cancer cell survival in both normoxia and severe hypoxia
To determine whether the effector caspases were activated by ablation of ULK1 or ATF4, caspase 3/7 activity was measured. Loss of ATF4 in A431 cells nearly doubled the observed caspase 3/7 activity, indicating that apoptosis was being activated (P<0.001) (Figure 5E). On the other hand, ULK1 knockdown actually reduced caspase 3/7 activity in both normoxia and severe hypoxia (P<0.01 and P<0.001) (Figure 5E), suggesting that cell death in this case is due to a caspase-3/7-independent mechanism, such a necrosis.
One common marker for necrosis is HMGB1, an architectural chromatin-binding factor [44–46]. Indeed, knockdown of ULK1 or ATF4 in A431 cells exposed to normoxia or severe hypoxia resulted in an increase in HMGB1 in the medium (Figure 5F) and decreased intracellular HMGB1 (Figure 5G). This result implied that knockdown of either ULK1 or ATF4 results in caspase-3/7-independent cell death, although it is not possible to identify the mechanism as necrosis without further study.
Further evidence for non-apoptotic cell death was found when cell morphology was examined in spheroids. Spheroids derived from ATF4-knockdown cells showed a high degree of nuclear fragmentation and more caspase 3 cleavage than the scrambled control, whereas spheroids formed from the ULK1-knockdown cells showed almost no nuclear fragmentation and less caspase 3 cleavage than the scrambled control (Supplementary Figure S8A at http://www.biochemj.org/bj/449/bj4490389add.htm).
Depletion of ULK1 or ATF4 reduces spheroid growth
We next assessed the effect of ULK1 or ATF4 knockdown on growth in a three-dimensional spheroid model . Spheroids derived from A431 cells depleted of ULK1 or ATF4 showed significantly retarded growth (P<0.01) (Figures 6A and 6B). ULK1- and ATF4-depeleted spheroids harvested on day 3 contained 84 and 90% AV-positive cells respectively, whereas control spheroids contained 26% AV-positive cells (P<0.01) (Figure 6C). Staining for pimonidazole, CAIX and CHOP indicated that the ULK1- and ATF4-knockdown spheroids were much less hypoxic than the scrambled controls (Figure 6D). Staining for Ki67, a marker of proliferation, showed no difference between groups (Figure 6D).
Knockdown of ULK1 or ATF4 reduces spheroid growth
To determine whether a similar effect could be observed by pharmacological inhibition of autophagy, A431 spheroids were prepared and treated with a single dose of chloroquine or bafilomycin A1. Spheroids treated in this way showed a similar, although less complete, growth delay (Supplementary Figure S8B).
High levels of ULK1 correlate with a poor prognosis in breast cancer
To examine the potential importance of ULK1 in cancer prognosis and treatment, the expression of ULK1 mRNA was examined in a cohort of 152 breast cancer patients with long-term follow up. Univariate analysis revealed that patients with the highest level of ULK1 mRNA had significantly lower relapse-free survival (log-rank test, P=0.021) (Figure 7). This association remained significant when Cox multivariate analysis including oestrogen-receptor status, patient age, nodal status, tumour size and tumour grade was performed (P<0.034).
High levels of ULK1 correlate with a poor prognosis in breast cancer
Hypoxia is a common feature of solid tumours, and is an essential determinant of prognosis and response to therapy. The activation of the ISR and autophagy represents a mechanism of cancer cell resistance to hypoxia. In the present study we demonstrated that the essential autophagy gene ULK1 is a link between these two mechanisms. We showed that ULK1 is transcriptionally up-regulated in severe hypoxia and in response to ER stress by the direct activity of ATF4 at the ULK1 promoter. Previous work has shown that ATF4 and CHOP can support high levels of autophagic flux in severe hypoxia through the transcriptional up-regulation of autophagic machinery proteins MAP1LC3B and ATG5, but does not support de novo autophagy itself [19,21]. Thus the results of the present study represents the first example in which ATF4 and the ISR are shown to regulate the expression of an autophagy-initiating factor and thus to potentially play an important role in the initiation of de novo autophagy.
It is well known that ULK1 kinase activity is subject to substantial post-translation regulation . Indeed, to date most work has focused on the phosphorylation events that regulate changes in the specific activity of existing ULK1 protein. In the present study, we found that the total activity of ULK1 is also subject to substantial transcriptional regulation by ATF4. When we exposed A431 cells to severe hypoxia the increase in total ULK1 kinase activity was dependent on the transcriptional up-regulation of ULK1 by ATF4, although the specific activity of ULK1 remained unchanged, the massive accumulation of ULK1 protein resulted in greatly increased total kinase activity. In contrast, in MCF7 cells the specific kinase activity of ULK1 markedly increased without protein induction, suggesting that activity is regulated by post-translational mechanisms. In this context, ATF4 appeared to play a role in the maintenance of ULK1 protein level rather than hypoxic up-regulation. Thus, depending on cell type, either transcriptional up-regulation or post-translational phosphorylation may be the primary mode of regulating ULK1 kinase activity. In support of this notion, recent work has demonstrated that ULK1 expression and autophagy induction in response to DNA damage were increased through transactivation of ULK1 by p53 .
In the present study, we demonstrated that ULK1 is required for autophagy in severe hypoxia, and ATF4 is responsible for increasing or maintaining ULK1 kinase activity in a cell-dependent context. We found that knockdown of ULK1 reduced autophagy induction in severe hypoxia when measured by Western blotting and immunofluorescence. To our surprise, when we examined autophagic flux by ImageStream® analysis, we found that this dependence also occurred in normoxic conditions, and that ablation of ULK1 or ATF4 in normoxia, severe hypoxia or three-dimensional cell culture (spheroids) resulted in a similar reduction in autophagic flux, and a similar enhancement of cell death. Furthermore, RNAi against ATF4 reduced ULK1 levels in normoxia as well as hypoxia, suggesting that both ATF4 and ULK1 are active in the ‘unperturbed’ normoxic condition. In this vein, it has become apparent that autophagy is crucial for cancer cell growth and the survival of the oncogenic changes that lead to cancer development-persistent autophagy is necessary to maintain a viable pool of mitochondria in H-Ras-expressing cancer cells and pancreatic cancer cells under otherwise unstressed conditions, a phenomenon that has been termed ‘autophagy addiction’ [26,27,49,50]. Importantly, it appears that ULK1 is dispensable in untransformed cells under normal growth conditions and is necessary for survival only in starvation conditions. ULK1-knockout mice are viable, showing only mild defects in haemopoesis . Furthermore, ULK1−/− MEFs (mouse embryonic fibroblasts) expressing a stable shRNA construct against ULK2 are viable under normal growth conditions. It is only when these cells are starved of amino acids and serum that the ULK1/ULK2 deficiency results in cell death .
Previous work has shown that autophagy inhibition can cause necrosis in apoptosis-deficient cells . Alternatively, there are a number of systems in which inhibition of autophagy can lead to a reduction in apoptosis or cell death [28,48,51–55]. Moreover, recent work has implicated ULK1/ATG13-mediated autophagy in promoting DNA-damage-induced cell death, although the authors failed to show a direct role for ULK1 in this process . In the present study, we found that ablation of endogenous ULK1 predisposed cells to die by an unidentified caspase-3/7-independent mechanism, whereas loss of ATF4 resulted in apoptotic cell death. One explanation for this result could be that autophagy is required for the maintenance of a low level of ATP to initiate caspase activation. Previous work has indicated that mitophagy is required for apoptosis induction in certain contexts . Alternatively, it may also be that loss of ULK1 results in elevated ROS (reactive oxygen species) and lysosomal dysfunction, leading to necroptosis [56–58]. In either scenario, the residual pool of ULK1 remaining following ATF4 knockdown might be sufficient to drive caspase-dependent death. In any case, further study is necessary to identify the mechanism of this caspase-3/7-independent cell death.
We found a positive correlation between ULK1 mRNA levels and reduced relapse-free survival in a cohort of breast cancer patients, implying that ULK1 up-regulation may be a mode of breast cancer cell survival and tumour progression. Likewise, the ISR is important in cancer cell survival and tumour maintenance. Tumour xenografts deficient in ATF4 or expressing a dominant-negative construct of eIF2α show greatly retarded growth and enhanced levels of cell death [19,23,24]. Data indicate that ATF4 and the ISR promote cancer cell survival of hypoxia, nutrient deprivation and oxidative stress in vitro [9,19–22], in part through the maintenance of high levels of autophagic flux through the up-regulation of MAP1LC3B and ATG5 [19,21]. The present study demonstrates for the first time that ATF4 can transcriptionally up-regulate the autophagy-initiating kinase ULK1 and that ULK1 is crucial for autophagy, mitophagy and cancer cell survival in normoxia, severe hypoxia and spheroids. Thus, on the basis of the results of the present study and those of others [26,27], we speculate that a specific ULK1 inhibitor may be an effective killing agent against a subset of cancers, with limited toxicity towards normal cells.
AMP-activated protein kinase
activating transcription factor
carbonic anhydrase IX
C/EBP (CCAAT/enhancer-binding protein)-homologous protein
eukaryotic initiation factor 2α
high-mobility group box 1
integrated stress response
microtubule-associated protein light chain 3
microtubule-associated protein 1 light chain 3β
myelin basic protein
mammalian target of rapamycin
PKR (double-stranded-RNA-dependent protein kinase)-like ER kinase
regulatory associated protein of mTOR
real-time quantitative PCR
short hairpin RNA
small interfering RNA
UNC51-like kinase 1
unfolded protein response
Luke Pike conceived and conducted all experiments unless otherwise specified. Adrian Harris supervised the project and research group. Dean Singleton assisted with Western blots for LC3 in Figures 1 and 4, as well as experiments pertaining to viral shRNA knockdown of ULK1. Francesca Buffa prepared Figure 7 in its entirety. Olga Abramczyk and James Murray conducted ULK1 kinase activity experiments. Kanchan Phadwal and Anna Katharina Simon conducted Imagestream analysis of autophagy. Ji-Liang Li conducted tumour xenograft studies with bevacizumab treatment, from which RNA was extracted and analysed.
We thank Dr Karim Bensaad and Dr Alan MacIntyre for their critical reading of this work before submission.
This work was supported by Cancer Research UK, the Rhodes Trust, the NIHR (National Institute for Health Research) Oxford Biomedical Research Centre, and the Natural Sciences and Engineering Research Council of Canada.