i Elongation factor eEF2 kinase and autophagy jointly promote survival of cancer cells

: Cells within solid tumours can become deprived of nutrients; in order to survive, they need to invoke mechanisms to conserve these resources. Using cancer cells in culture in the absence of key nutrients, we have explored the roles of two potential survival mechanisms, autophagy and elongation factor 2 kinase (eEF2K), which, when activated, inhibits the resource-intensive elongation stage of protein synthesis. Both processes are regulated through the nutrient-sensitive AMP-activated protein kinase and mechanistic target of rapamycin complex 1 signalling pathways. We find that disabling both autophagy and eEF2K strongly compromises the survival of nutrient-deprived lung and breast cancer cells, whereas, for example, knocking out eEF2K alone has little effect. Contrary to some earlier reports, we find no evidence that eEF2K regulates autophagy. Unexpectedly, eEF2K does not facilitate survival of prostate cancer PC3 cells. Thus, eEF2K and autophagy enable survival of certain cell-types in a mutually complementary manner. To explore this further, we generated, by selection, cells which were able to survive nutrient starvation even when autophagy and eEF2K were disabled. Proteome profiling using mass spectrometry revealed that these ‘resistant’ cells showed lower levels of diverse proteins which are required for energy-consuming processes such as protein and fatty acid synthesis, although different clones of ‘resistant cells’ appear to adapt in dissimilar ways. Our data provide further information of the ways that human cells cope with nutrient limitation and to understanding of the utility of eEF2K as a potential


Introduction
Cells within solid tumours may be subject to insufficiency of nutrients such as glucose (a major metabolic fuel) and amino acids (the building blocks for proteins), especially if there is inadequate vascularisation. In order to form a tumour, such cells require mechanisms to enable them to survive such conditions. One is autophagy, the intralysosomal breakdown of existing cellular components (e.g., non-essential or damaged proteins) to provide amino acids and substrates for energy production [1]. Disabling autophagy impairs tumour growth although it can aid tumour initiation [2,3]. Autophagy is negatively regulated by mammalian target of rapamycin complex 1 (mTORC1), a protein kinase that is activated under conditions of nutrient sufficiency [4] and stimulated by the AMP-activated protein kinase (AMPK), which is turned on under conditions that increase the AMP/ATP ratio [5]. AMPK acts to reduce energy consumption, by inhibiting processes that consume ATP (or related compounds) and by promoting ones that generate ATP.
A second pro-survival mechanism involves eukaryotic elongation factor 2 kinase (eEF2K), an atypical protein kinase which phosphorylates and inhibits elongation factor 2 (eEF2), the protein which enables ribosomes to translocate along mRNAs during the elongation stage of protein synthesis [6]. Protein synthesis is a major consumer of energy (as ATP and GTP) and, within protein synthesis, almost all the energy (and amino acids; >99%) are used during elongation. Thus, by inhibiting translation elongation, eEF2K can reduce cellular dependency on nutrients. Numerous studies have shown that eEF2K helps cancer cells withstand nutrient deficiency in vitro [7,8] and aids tumour growth in vivo (reviewed [7,9]). Several studies have suggested that inhibition of eEF2K, a nonessential enzyme, may be a valid approach to cancer therapy, either by itself or in combination with other agents (see, e.g., [8,10,11]; reviewed; [7,9,12]). On the other hand, there is some evidence that eEF2K may actually impair tumour initiation (e.g., [13]).
Both autophagy and eEF2K are inhibited by signalling through the mTORC1 pathway [14] and stimulated by AMPK [15], which also inhibits mTORC1 signalling by phosphorylating tuberin (TSC2, a major regulator of mTORC1 function) [16]. Thus, two nutrient-regulated pathways that can aid tumour cell survival and tumour growth operate to inhibit a process that consumes energy and nutrients (protein synthesis) and to promote one that can generate such commodities (autophagy). mTORC1 is able to inhibit autophagy by an inhibitory phosphorylation on autophagy-related protein-13 (ATG13), which reduces the activity of the serine/threonine protein kinase ULK1 (ULK1) and in turn decreases the rate of autophagosome formation [17,18]. On the other hand, AMPK positively regulates ULK1 by directly phosphorylating it at multiple sites, which increases the recruitment of ATG proteins to the membrane domains in which autophagosome formation occurs [19]. mTORC1 has also recently been shown to reciprocally inhibit AMPK through direct phosphorylation of its catalytic -subunits [20].
A previous study has reported that eEF2K inhibits autophagy in human colon cancer, [21]. However, several other studies apparently contradict these findings and report that eEF2K promotes autophagy in human glioblastoma and breast cancer cells, as well as in mouse embryonic fibroblasts [22][23][24][25][26][27][28]. However, there is currently no information on the molecular mechanism by which eEF2K could activate or inhibit autophagy; eEF2 remains the only validated substrate for eEF2K. A study from our group investigated the possible role of eEF2K in autophagy using human lung cancer, colon cancer cells and mouse embryonic fibroblasts, and found no link between eEF2K and autophagy [29]. eEF2K and autophagy may each help cells to survive nutrient limitation by operating independently of each other, downstream of mTORC1 and AMPK.
Further information is urgently required about the interplay between autophagy and eEF2K in cancer cell survival and concerning how cancer cells might be able to cope with (acquire resistance Downloaded from http://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20210126/906579/bcj-2021-0126.pdf by guest on 30 March 2021 to) blockade of autophagy and/or eEF2K activity. Importantly, our data show that autophagy and eEF2K aid cell survival through independent mechanisms, rather than because eEF2K is required for activation of autophagy. Furthermore, we have also explored how cells can acquire tolerance, under nutrient-deprived conditions, to inhibition of autophagy and loss of eEF2K.

Cell lines, treatment and lysis
Authenticity of the cell lines was confirmed by the Garvan Institute (Sydney, Australia). MDA-MB-231 (human breast carcinoma) or A549 (human lung carcinoma) cells were both cultured in growth medium -DMEM with 25 mM glucose, 1 mM sodium pyruvate and 4 mM L-glutamine (#1995065 from LifeTechnologies). A549 cells expressing short-hairpin RNA (shRNA) against eEF2K were generously provided by Janssen Pharmaceutica. PC3 (prostate carcinoma) cells were cultured in RPMI medium -11 mM glucose supplemented with 1 mM sodium pyruvate and 2 mM GlutaMax (#21870076 from LifeTechnologies). Both types of media were supplemented with 10% (v/v) foetal bovine serum (FBS) and 100 U/mL penicillin and 0.1 mg/mL streptomycin. The cells were maintained at 37°C with 5% (v/v) CO 2 . To induce eEF2K knock-down in A549 cells, 1 µM IPTG was added 5 days prior to the experiments.
During any experiment, the control conditions were treated with the corresponding vehicle in the same type of medium. For nutrient deprivation experiments, two types of media were used -DMEM without glucose and sodium pyruvate (#11966025 from LifeTechnologies) and DMEM without glucose, sodium pyruvate and glutamine (#A1443001 from LifeTechnologies). Both types of medium were supplemented with 5% (v/v) dialysed FBS and 100 U/mL penicillin and 0.1 mg/mL streptomycin.
After treatment, the medium was centrifuged at 4 °C at 1000 x g for 3 min, the cells were washed twice in 1x PBS (phosphate-buffered saline, ThermoFisher) with the PBS being centrifuged into the same tube as the media to gather dead cells. The supernatant was removed and the cells on the plate and in the tube were lysed in ice-cold lysis buffer containing 350 mM NaCl, 50 mM Tris-HCl pH 7.5, 1 mM EDTA, 1 mM EGTA, 10 mM NaF, 2.5 mM Na 2 H 2 P 2 O 7 , 50 mM β-glycerophosphate, 1% Triton X-100, 2 mM Na 3 VO 4 , 14.3 mM β-mercaptoethanol and protease inhibitor cocktail (1X from Sigma Aldrich, 05056489001). The cells were centrifuged at 4 °C at 16000 x g for 10 min and the supernatant was collected. Protein concentrations were estimated via the Bradford Assay.

SDS-PAGE and Western Blotting
SDS-PAGE and western blotting was performed as previously described [30]. Supplementary Table S1 lists all the antibodies used in this study.

CRISPR-mediated knockout of eEF2K in MDA-MB-231 and PC3 cells
The creation of eEF2K-KO MDA-MB-231 and PC3 cells was described previously [31,32]. Briefly, the gRNA was introduced into cells by nucleofection 48 h prior to isolation of transfected clones via surface-expressed CD4 using Dynabeads. Single cells were grown up to confluence and then screened by restriction digest using BpmI and by Western blotting to verify loss of the phosphorylation of eEF2 at Thr56 (of the protein sequence).

CellTox TM Green Assay
MDA-MB-231, A549 and PC3 cells were seeded at densities of 3x10 4 , 2x10 4 or 3.5x10 4 cells per well respectively in 96-well black walled opaque plates with a clear bottom and left overnight in the incubator. Once the cells had reached 90% confluence, the medium was changed to 100 µl of medium lacking either glucose/pyruvate or glucose, pyruvate and glutamine. The cells were treated with the relevant inhibitors at this time and the Express, No-Step Addition at Dosing method was followed from Promega (Kit #G8742) for 48 h. Images of each well were taken using the 4X lens of the Zeiss Axiovert 200 microscope with the GFP filter. The plates were measured at 475 nm excitation and 500-550 nm emission using the GloMax Discover Microplate Reader. The Bradford assay [33] was used to measure protein concentrations for gel loading and to normalise data.

Propidium Iodide and Calcein-AM Assay
The cells were seeded and treated in the same way as for the CellTox Green Assays. The propidium iodide and calcein-AM were added 1 h prior to measurement at 10 µM and 1 µM respectively. The propidium iodide signal was measured at 520 nm (excitation) and 580-640 nm (emission) using the GloMax Discover Microplate Reader. The calcein-AM signal was measured at 475 nm excitation and 500-550 nm emission. Calcein-AM images were taken using the 4X lens of the Zeiss Axiovert 200 microscope with the GFP filter. The Bradford Assay was used to measure protein concentrations to normalise the data.

Magic Red Assay
MDA-MB-231 cells were seeded at a density of 2.5x10 4 cells per well in a 96-well black walled opaque plates with a clear bottom and left overnight in the incubator. Once the cells had reached 90% confluence, the medium was changed to 100 µl of DMEM without glucose and pyruvate and treated with 125 nM BafA (Bafilomycin A1). To measure cathepsin L activity, the protocol from BioRad (#ICT941) was followed, with fluorescence being measured using the GloMax Discover Microplate Reader.

LysoTracker Red Assay
MDA-MB-231 non-resistant and resistant cells were seeded at a density of 3x10 4 cells per well in 96well black walled opaque plates with a clear bottom and left overnight in the incubator. Once the cells reached 90% confluence, the medium was changed to 100 µl DMEM without glucose and pyruvate with 125 nM BafA for 1 h. At this point, the LysoTracker Red reagent was added at 156 nM (final concentration) alongside 200 µM Hoechst33342 dye. The LysoTracker signal was measured at 520 nm (excitation) and 580-640 nm (emission) using the GloMax Discover Microplate Reader. The Hoechst signal was measured at 365 nm (excitation) and 415-445 nm (emission). The LysoTracker Red signal was normalised by the Hoechst signal which stains all cells in the well.

Creation of Bafilomycin-A resistant MDA-MB-231 cells
MDA-MB-231 cells were cultured until they reached 80% confluency, upon which the medium was changed to DMEM without glucose and pyruvate supplemented with 5% (v/v) dialysed FBS, 100 U/mL penicillin and 0.1 mg/mL streptomycin. The cells were treated with 125 nM BafA for 48 h and left in an incubator at 37°C with 5% (v/v) CO 2 . After treatment, the medium was aspirated off and the cells were washed three times with 1 X PBS to remove dead cells. The medium was changed back to high glucose DMEM (#1995065 from Life Technologies) with 10% (v/v) FBS, 100 U/mL penicillin and 0.1 mg/mL streptomycin. The cells were washed with 1X PBS to remove dead cells daily until reaching 80% confluency upon which the treatment with BafA in medium without glucose and pyruvate was repeated. These treatment cycles were repeated until a significant visual increase

Label-free proteomics and phosphoproteomic analysis
Frozen MDA-MB-231 cell pellets (4 replicates of WT, eEF2K-KO #1, eEF2K-KO #2) with or without BafA, were dissolved in chilled SDC buffer (4% sodium deoxycholate in 0.1 M triethylammonium bicarbonate (TEAB) and lysed using the Precellys® tube homogeniser. The cell supernatants were heat-treated for 5 min at 95°C and further lysed using a sonicator probe at 4°C (three 30 s cycles of 1 s on/off at 80% output). The protein concentration was determined using the Nanodrop (Thermo). Equal protein amounts from each sample were used for protein digestions and phosphopeptide enrichment that was performed as according to the Easy-Phos protocol [34] with a slight modification. Phosphopeptide enrichment was performed on WT, eEF2K-KO #1 and eEF2K-KO #2 cell pellets. Following protein digestion, an equal amount of all the samples were set aside for proteomic profiling experiments. All samples (profiling, phosphopeptide fraction and non-phosphopeptide fractions) were then de-salted using POROS R3 resin (Thermo Scientific), dried, re-suspended in 0.1% (v/v) formic acid and analysed by mass-spectrometry.

LC-MS/MS and data acquisition
The peptide samples were separated by nano-liquid chromatography (nLC) using the NanoElute nLC system (Bruker Daltonics, Germany) coupled online to the trapped-ion mobility spectrometry Timeof-Flight (timsTOF) Pro mass spectrometer (Bruker Daltonics, Germany) using the Data Dependent Acquisition-Parallel Accumulation Serial Fragmentation (DDA PASEF)mode. Reverse-phase chromatography was done using a 25 cm, 75 µm ID Odyssey C18 nano column (IonOpticks, Australia) with an integrated emitter. The peptides were eluted using a 120 min gradient from 0% to 35% buffer B (acetonitrile in 0.1% formic acid) at a rate of 400 nL/min. Buffer A consisted of 0.1% (v/v) formic acid.

Data processing and analysis
The raw data files from each sample generated on the timsTOF Pro mass spectrometer was processed using the software package MaxQuant version 1.6.14 (Max Planck Institute of Biochemistry, Martinsried, Germany; [35]). The data was searched against the UniProt reference proteome for Homo sapiens with the following parameters: variable modifications -deamidation (N/Q), oxidation (M); fixed modification -methylthio (C); enzyme -trypsin; missed cleavages -2; all other parameters were default settings. Only proteins with a false discovery rate (FDR) of ≤ 1% were reported. For phosphopeptide analysis all settings were set to default except of the peptide spectrum match (PSM) was set to FDR 0.05. Samples set aside for profiling experiment was used for protein abundance analysis. Phosphopeptide analysis was performed using data obtained from the phosphopeptide fraction following enrichment. Data analysis was done using the software Perseus [36]. Briefly, raw data were cleaned to remove reverse hits and contaminants, transformed to log 2 and multi-sample ANOVA performed to identify statistically significant proteins. The post-hoc Tukey's test was used to identify significant pairs (FDR < 0.01) and the proteins were further classified into relevant GO Biological process groups for data interpretation.

Autophagy and eEF2K together protect cells against starvation for glucose and pyruvate
We initially made use of a previously reported line of A549 pulmonary adenocarcinoma cells which contain an inducible shRNA against eEF2K, allowing eEF2K expression to be 'knocked down' by treating cells with IPTG [37]. Starvation for glucose and pyruvate increased p-eEF2 levels in control A549 cells, an effect which was strongly attenuated in cells that had been pre-treated with IPTG to knock down eEF2K expression (Fig. 1A). The accumulation of LC3 demonstrates the efficacy of BafA, an inhibitor of the lysosomal v-ATPase and therefore also of autophagy [38].
Under control conditions (in medium containing glucose and pyruvate), IPTG treatment did not affect cell survival although BafA did induce an approximately 9-fold increase in cell death, as judged using the Cell Tox Green assay ( Fig. 1B; images in Supplementary Fig. S1). Absence of glucose/pyruvate led to a comparatively small 3-fold increase in cell death in A549 cells. This may be because cells in full medium continue to grow, eventually depleting the nutrients in the medium and because of their high confluence and, without autophagy, die in greater numbers. It is important to note that this is in comparison to the much smaller degree of cell death that occurs in cells growing in full medium without any treatment. Knockdown of eEF2K (and thus reduction of eEF2 phosphorylation; Fig. 1A) did not affect cell survival in the presence or absence of glucose/pyruvate. Increased levels of p62 and LC3 confirmed that BafA is indeed interfering with lysosomal function under these conditions. In full medium, and especially in medium lacking glucose/pyruvate, BafA also increased p-eEF2 levels ( Fig. 1A), perhaps reflecting its ability to impair mTORC1 signalling [39].
BafA slightly decreased cell viability in the presence or absence of glucose/pyruvate, and knockdown of eEF2K enhanced this almost two-fold in the absence of glucose and pyruvate (Fig. 1B). The lack of effect of eEF2K knockdown under glucose-replete conditions likely reflects the fact that eEF2 phosphorylation and by implication eEF2K activity are low under this condition. In contrast, the observation that BafA has a greater effect on the viability of eEF2K-knockdown cells indicates autophagy and eEF2K work together to promote cell survival under starvation conditions.
We also analysed cell viability by staining with propidium iodide (Fig. 1C), which confirmed the data from the CellTox Green assays. Lastly, as an adjunct to the CellTox Green and PI staining analysis, we employed a further method to assess cell survival, i.e., staining with the calcein-AM reagent which can only be broken down to its fluorescent product (by intracellular esterases) after entering live cells [40]. The resulting data mirrored the findings of CellTox Green and PI staining assays ( Fig. 1D; also see images in Supplementary Fig. S2). The CellTox Green and Calcein-AM staining images can be used to visualise the relative degree of cell death between treatments.
We then tested another widely used type of cancer cells, MDA-MB-231 breast cancer cells, starving them of glucose/pyruvate, which caused a marked increase in the phosphorylation of eEF2 ( Fig. 2A). This is consistent with the fact that eEF2K is activated under such conditions, at least in part via AMPK [15,41]. JAN-384 is a recently developed inhibitor of eEF2K [29]; as expected, it reduced the phosphorylation of eEF2. Conversely, BafA actually increased p-eEF2 levels. The accumulation of LC3 and p62 again demonstrates the efficacy of BafA in impairing autophagy.
In MDA-MB-231 cells kept in full medium, treatment with JAN384 had almost no effect on cell survival as assessed using the CellTox Green method (Fig. 2B) and BafA had only a minor effect, either when used alone or in combination with JAN-384 (Fig. 2B). The lack of effect of JAN-384 may reflect the fact that eEF2K has such a low level of activity under these control conditions (cf. low level of P-eEF2 in Fig. 2A).
Starving MDA-MB-231 cells of glucose and pyruvate modestly impaired their survival (Fig. 2B). JAN384 did not decrease cell survival, and nor did BafA on its own. However, when used in combination, there was a marked increase in cell death (Fig. 2B), showing that eEF2K is not required for their survival unless autophagy is also inhibited. Importantly, this again demonstrates that eEF2K and autophagy work together to aid the survival of MDA-MB-231 cells when they are starved of glucose and pyruvate.
We noted that JAN-384 treatment led to a rise in levels of LC3 and p62 ( Fig. 2A), suggesting it interferes with lysosomal function (but, based on the cell survival data, not to an extent that leads to cell death). This could indicate that eEF2K does play a role in the activation of autophagy, as indicated by some, but not all, earlier reports [23,26,27,29]. Alternatively, JAN-384 might exert an off-target effect that interferes with lysosomal function; this point is discussed further below.
A compound termed A484954 has also been reported as a specific inhibitor of eEF2K and has previously been used as an alternative to JAN384 [42], although its potency as an eEF2K inhibitor in cells is very weak, concentrations of 30-100 M being needed to substantially decrease p-eEF2 levels. However, it was important to address the roles of eEF2K in cell survival and autophagy without interference from possible off-target effects of A484954 and JAN384, whose specificities remain incompletely characterised. To do so, we employed CRISPR-Cas9 genome editing to inactivate the copies of the EEF2K gene in MDA-MB-231 breast carcinoma cells. The gene targeting strategy and data for selected clones of targeted cells are shown in Supplementary Figs. 3-6. Screening by PCR identified several targeted clones which showed indels in the expected region of the EEF2K gene (Supplementary Figs. 3B, C, 4, 5, and 6A, B). Following treatment with H 2 O 2 , which indirectly activates eEF2K, there was a marked rise in p-eEF2 in control cells but not in knock-out (KO) cells, e.g., clones A and D (Fig. 3A, B). These data are consistent with eEF2K being the only kinase that phosphorylates eEF2 at the inhibitory site, Thr56 [43], and show that these clones do indeed lack functional eEF2K. It is important to note that the initiator methionyl residue is missing from the mature form of eEF2, and the numbering of residues is based on the protein sequence, rather than that of the open-reading frame [43].
This advance allowed us to test the effect of genetic knockout of eEF2K on the properties of these cells. Two clones of eEF2K-KO MDA-MB-231 cells (termed 'KO #1' and 'KO #2') were chosen for further study. Loss of eEF2K did not affect cell survival either in the presence or the absence of glucose/pyruvate ( Fig. 3C; see corresponding westerns in Fig. 3B) but did lead to higher levels of cell death when cells were also treated with BafA. Inhibition of eEF2K with A484954 (A484) yielded similar results to those seen for eEF2K-KO cells. Treatment with A484 alone did not significantly alter cell survival but in combination with BafA treatment, it led to a much greater increase in cell death. Interestingly, in the presence of glucose, combination treatment with A484954 and BafA further increased death in eEF2K KO, but not WT, cells. This could be attributed to the fact that KO cells never possessed any eEF2K protein, while, in cells treated with A484954, eEF2K had only been inhibited for a limited period (48 h). Inspection of the Cell Tox Green-stained cells under the microscope confirmed these data ( Supplementary Fig. S7). Increased levels of p62 and LC3 (Fig. 3B), as well as use of the reagent 'Magic Red' which measures lysosomal cathepsin (protease) activity, confirmed that BafA does indeed impair lysosomal function in eEF2K-KO and control cells ( Supplementary Fig. S8). Glucose/pyruvate starvation increased the levels of p-eEF2, and as expected, this effect was completely lost in eEF2K-KO cells (Fig. 3B). Unlike JAN384, KO of eEF2K did not lead to an increase in LC3 levels (Fig. 3B) suggesting that the effect of JAN384 is not related to its ability to inhibit eEF2K, but to an off-target effect; these data also indicate that eEF2K does not regulate autophagy, at least in MDA-MB-231 cells. A484954 did not significantly alter the levels of LC3 or p62 (Figs. 3B) indicating it is free from this off-target effect.
We again also analysed cell viability by staining with propidium iodide (Fig. 3D), which showed that, while knocking out eEF2K or treating with BafA individually did not affect viability of cells in glucose/pyruvate-free medium, BafA did increase the death of the KO (but not WT) cells. This confirms the findings of the CellTox Green approach.
Lastly, we used Calcein-AM to assess cell survival in MDA-MB-231 cells. The resulting data for WT and eEF2K-KO MDA-MB-231 cells mirrored the findings for CellTox Green ( Fig. 3E; also see images in Supplementary Fig. S9). The data for both MDA-MB-231 and A549 cell lines confirm that autophagy and eEF2K work together to aid cell survival during glucose and pyruvate starvation, while eEF2K plays little role on its own under these conditions. It was also of interest to assess whether the eEF2K inhibitors JAN-384 and A484954 had any offtarget effects that affected cell survival. To study this, we made use of eEF2K-KO MDA-MB-231 cells where any effects of these compounds cannot be due to inhibition of eEF2K. CellTox Green assays were performed on eEF2K KO cells in medium lacking glucose and pyruvate, treated with BafA and either JAN-384 or A484954. There was no significant difference in cell survival between eEF2K-KO cells treated only with BafA and those treated with BafA and either of the eEF2K inhibitors ( Supplementary Fig. S10 A, B). This indicates that neither compound has an off-target effect that influences cell survival, at least under these conditions.
A limitation of the CellTox Green method is that it does not directly report the percentage of dead vs. live cells. However, the 'lysis solution control', in which all cells have been permeabilised, can be used as a reference to compare the proportion of cell death in the various treatments. The data in Supplementary Fig. S10A, B imply that, for example, about 60-75% of cells die when eEF2K-KO cells are treated with BafA under Glc-starved conditions.
It is also of note that knocking out eEF2K did not alter the levels of either FLIP or MCL-1, which are involved in the apoptotic pathway ( Supplementary Fig. S10C), consistent with the idea that eEF2K assists survival of these cells by downregulating protein synthesis, thus reducing cellular energy consumption, rather than affecting levels of proteins involved in cell survival, although further studies on the many additional proteins that control cell death would be needed to be certain.

eEF2K protects some, but not all, types of cells against starvation for glutamine
We then tested the effect of omitting another key nutrient, the amino acid glutamine, which plays an important role in many cells by providing a source of both carbon (for oxidative metabolism) and nitrogen (for biosynthesis; [44]). Starvation for glutamine increased p-eEF2 levels in A549 cells (Fig.  1A) and impaired cell survival under control conditions. Survival fell further upon knocking down eEF2K (Fig. 1B). Treatment with BafA did not significantly alter cell viability of either control or eEF2K knock-down cells (Fig. 1B). This suggests that under starvation for glutamine, glucose and pyruvate, eEF2K plays an important role in cell survival, whereas autophagy does not. It is important to note that MDA-MB-231 cells survived for only a short period of time when starved of glutamine, therefore no significant differences between the various treatments were observed. Hence, glutamine starvation data for MDA-MB-231 cells is unavailable.
To assess whether there are differences in the roles that eEF2K and autophagy play in promoting survival of different types of cancer cells, we extended our studies to a third cell type, PC3 (prostate cancer-derived) cells. To this end, we again used CRISPR-Cas9 genome editing to disrupt the EEF2K gene, using the same targeting and screening strategy for these human cells as for MDA-MB-231 cells. Data for the validation of eEF2K-KO clones are shown in Supplementary Figs. S3D, E, 6C, D, 11, and 12.
As shown in Fig. 4A, whereas starvation for glucose/pyruvate or glutamine increased levels of p-eEF2 in WT or control cells, phosphorylation of eEF2 was absent from the two knockout clones. Treatment with BafA increased the levels of p62 or LC3, confirming its effectiveness in inhibiting autophagy.
As assessed by CellTox Green staining, neither knockout of eEF2K nor treatment with BafA affected the viability of PC3 cells in medium that contained glucose and pyruvate ( Fig. 4B; see also images in Supplementary Fig. S13; analysis of LC3 shows that BafA is effective in interfering with autophagy). Surprisingly, in medium lacking glucose/pyruvate, eEF2K-KO cells showed less death than WT cells; a similar pattern was seen for cells deprived of glutamine, with WT cells showing much greater cell death than either clone of eEF2K-KO cells (Fig. 4B). Cell viability was also assessed by staining with propidium iodide (Fig. 4C) and calcein-AM (Fig. 4D, Supplementary Fig. S14), yielding data which were in good agreement with the CellTox Green data. This shows that eEF2K can play different roles in cell survival depending on the cancer cell type.
Cells can acquire the ability to escape their dependence on eEF2K and autophagy for their survival An important limitation of cancer treatment is the acquisition of resistance to therapeutic agents [45,46]. It was therefore of substantial interest to assess whether cancer cells can become resistant to disabling of eEF2K and/or autophagy, and, if so, by what mechanism(s) this arises. To create such cells, we subjected WT or eEF2K-KO MDA-MB-231 cells to repeated starvation for glucose and pyruvate, concurrent to treatment with BafA for 48 h. The cells were allowed to recover and were then treated under the same conditions until there was a visible difference in cell survival between these cells and cells that did not undergo this process. Chloroquine (CQ), a compound that enters lysosomes and binds protons, thereby raising the pH of its lumen and impairing lysosomal processes such as autophagy was also used to confirm the resistance of these cells to autophagy inhibition [38]. It was important to use two mechanistically distinct ways of interfering with lysosomal function to be sure that resistance was truly due to overcoming the need to autophagy for cell survival rather than, e.g., the v-ATPase becoming resistant to BafA. At different stages during this selection process, cells were analysed by CellTox Green (Fig. 5B); in this way, we were able to derive cells that were able to survive glucose/pyruvate starvation better than control cells. The approach is schematized in Fig. 5A.
Surviving cells were then cultured further, to select those which had acquired the ability to survive nutrient-deprived conditions without eEF2K and/or active autophagy (Fig. 5A). As shown in Fig. 5B-D, we derived WT or eEF2K knockout cells whose survival was not impaired by treatment with CQ or BafA. To confirm that resistance was not simply due to lysosomal function becoming resistant to BafA or CQ, we confirmed that each of these agents could still raise the intralysosomal pH of resistant cells, which was indeed the case of both drugs, in both resistant clones, since BafA or CQ each markedly decreased the Lysotracker (a dye that labels the acidic lumen of lysosomes) signal in all cases (Fig. 5C). This was further confirmed by showing that LC3 II accumulated in response to either agent (Fig. 5D). Interestingly, the LysoTracker Red assay also revealed that the resistant WT and both KO cell lines had increased autophagic flux under control conditions. This trend continued even when the cells were treated with CQ or BafA, although by far the largest increase in the flux was seen with KO #2, suggesting the two KO lines have differing characteristics, a desirable feature for studying resistance mechanisms.
Interestingly, after a week in full growth medium, the resistant cells start to gradually lose their ability to counteract the effects of being in medium without glucose/pyruvate and concurrent inhibition of autophagy. By the second week of growth after the initial testing for resistance, such resistance was completely lost, as indicated by the CellTox Green data (Fig. 6A, B). The analysis shows that in week 2 (1 week after recovery from treatment) the resistance of the cells to glucose/pyruvate starvation and BafA or CQ had decreased. By week 3, the resistance of the cells had returned to the levels seen for cells that had not undergone the repeated treatments, showing the same levels of cell death after starvation of cells for glucose/pyruvate either alone or in combination with either BafA or CQ.

Proteomic analysis of 'resistant and sensitive' cells reveals numerous changes in levels of specific proteins
To try to explore the mechanisms by which cells become resistant to BafA either in the presence or absence of eEF2K, proteomic profiling using mass spectrometry was conducted using lysates from control or BafA-resistant WT or eEF2K-KO MDA-MB-231 cells. A label-free approach was used to identify and quantify relative protein abundance changes between the cell lines of interest. A total of 3804 proteins were identified within the proteome of MDA-MB-231 cell lines analysed (Supplementary Table S2). Of these, 622 proteins showed statistically significant (FDR <0.01) changes in their protein abundance in at least one cell line/condition (Supplementary Table S3). The relevant protein changes are listed in Table 1.
In all three BafA-resistant lines, we observed lower levels of a number of proteins involved in biosynthetic processes which require resources such as metabolic energy. Consistently, we saw lower levels of several amino acyl-tRNA synthetases, although the exact members of this family which changed differed somewhat between WT and KO lines (Fig. 7A). These enzymes catalyse the attachment of specific amino acids to their cognate tRNAs for protein synthesis, a step which requires the equivalent of two ATP molecules. The MS data also revealed lower levels in resistant cells of several translation factors and other proteins that are required for protein synthesis (Fig. 7B), a process that consumes a high proportion of cellular energy [47] as well as amino acids. The chaperone HSPA5 (a member of the hsp70 family, also termed BiP or GRP78, which aids protein folding in the endoplasmic reticulum) was also consistently lower in resistant cells. Some other chaperones were also lower in resistant cells, although among them only DNAJB1 was lower in resistant WT cells and both eEF2K-KO clones ( Supplementary Fig. S15).
Fatty acid synthase (FASN), which catalyses a key step in the nutrient-intensive process of fatty acid synthesis, was lower in all three resistant lines compared to the 'parental' cells (Fig. 8). Western blot analysis revealed that FASN protein levels did indeed trend lower in BafA-resistant WT and KO#1 cells (Fig. 9A, B). Cytosolic acetyl-CoA acetyltransferase (ACAT2), an enzyme that catalyses the condensation of two molecules of acetyl-CoA to form acetoacetyl-CoA (a necessary step for cholesterol synthesis) was also decreased in all three resistant cell lines. Interestingly, levels of the enzyme ACADVL (mitochondrial very long-chain specific acyl-CoA dehydrogenase) were higher in all three resistant lines; this enzyme plays a key role in fatty acid oxidation (i.e., the 'reverse' of the process in which FASN takes part).
We also observed that several proteins associated with the function of lysosomes, the organelle where autophagy takes place, (e.g., AP1B1 [AP-1 complex subunit beta-1], NPC1 [Niemann-Pick disease, type C1] and CTSD [cathepsin D]) were elevated in some or all (CTSD) of the resistant cells ( Supplementary Fig. S15).
To confirm some of the proteomics data, western blots were performed on the same cell lysates that were used for the proteomics experiment (Fig. 9). The western blot data showed similar trends to the proteomics data. Namely, it was observed that FASN tended to decrease in the resistant cells, as well as HSPA5 (Fig. 9A, B). Interestingly, both procathepsin D (the immature form of cathepsin D) and cathepsin D itself were significantly higher in all three of the resistant cell lines compared to their non-resistant counterparts as judged from the MS data ( Supplementary Fig. S15) and by immunoblot analysis (Fig. 9A, B). Levels of the autophagosome cargo protein p62 also trended higher in resistant cells, although not significantly (MS analysis: Supplementary Fig. S15 and  immunoblot data, Fig. 9A, B).
The availability of eEF2K-KO cells provided us with the opportunity to assess whether deletion of eEF2K resulted in loss of phosphorylation of any sites/proteins other than Thr56 in eEF2; such additional events could reveal novel substrates for eEF2K. We therefore also performed phosphopeptide enrichment and phosphoproteome analysis by LC-MS/MS for WT and eEF2K-KO cells. A total of 16,074 phosphosites were detected; 10,737 of which were Class I sites (siteprobability score >0.75) (Supplementary Table S4). The analysis revealed the complete absence of the phosphopeptide, AGETRFTDTR, corresponding to the major phosphorylation site in eEF2 for eEF2K (Supplementary Table S5; Thr56; [43]). This confirms that phosphorylation of eEF2 is lost in eEF2K-KO cells. In addition, 31 other sites were found to be phosphorylated only in the WT cells and not in either the eEF2K-KO-1 or eEF2K-KO2 cells (Supplementary Table S6). Some of these could be novel substrates for eEF2K, although further experiments are needed to verify this (which are beyond scope of the current study). Notably, none of these sites correspond to the sites found to be phosphorylated in a study that showed possible new substrates for eEF2K [48].

Discussion
Here we have studied the roles of two signalling processes that are controlled by the availability of nutrients and metabolic energy in the ability of cells to withstand nutrient withdrawal. eEF2K controls the rate of elongation during protein synthesis, a process which has heavy demands for energy (ATP and GTP) and for amino acids. On the other hand, autophagy breaks down macromolecules to create substrates for the generation of metabolic energy as well as free amino acids [49]. Both are activated through the energy-sensing AMPK pathway which is turned on when energy or some nutrients levels are low. Both eEF2K and autophagy are inhibited by mTORC1 signalling, which is active when levels of amino acids and cellular energy/glucose are high.
Our data show that disabling eEF2K in A549 and MDA-MB-231 cells does not compromise cell survival during depletion of nutrients (glucose and pyruvate), suggesting additional mechanisms can fulfil the needs of the expensive process of protein synthesis. One clear candidate for this is autophagy. While inhibition of autophagy alone had a small effect on survival of these cells, simultaneous inhibition of eEF2K and autophagy had a much greater effect. This suggests that autophagy is supplying at least a substantial fraction of the material required for protein synthesis. Thus, at least in these types of cells, autophagy and eEF2K operate in a mutually complementary manner to cells to withstand nutrient deprivation. In this context, it is notable that autophagy and eEF2K are each regulated both by mTORC1 signalling and by AMPK [6,14,50], with eEF2K being controlled by glucose availability both directly and independently of AMPK [5,51]. eEF2K controls the most nutrient-consuming stage of protein synthesis, elongation, and is also the only component downstream of mTORC1 and AMPK known to regulate general protein synthesis, other targets for these pathways acting to modulate translation of specific mRNAs or subsets of mRNAs [52,53]. Interestingly, cyclin-dependent kinase CDK1 (which regulates the cell cycle) is a further example of a common regulator of eEF2K [54] and of autophagy [55].
Although several reports have suggested that eEF2K promotes autophagy [22][23][24][25][26][27][28], we were previously unable to confirm this [29]. The present data showing that eEF2K and autophagy together contribute to aiding survival of A549 and MDA-MB-231 cells again indicates they act independently to promote the survival of these cells under nutrient-deprivation. We did observe that, in MDA-MB-231 cells, JAN-384 tended to increase the levels of LC3 and p62, which indicates impairment of autophagic flux. In contrast, however, the knockout or knockdown of eEF2K did not affect p62 and LC3 levels, in line with our group's previous findings for A549 cells [37]. Our Magic Red data also show that cathepsin L activity was not affected in eEF2K-KO MDA-MB-231 cells. The LysoTracker Red assay also showed that autophagic flux was not significantly altered in eEF2K-KO cells. Combined, these data show that eEF2K does not promote autophagy in any of the cell-types tested here. It thus appears that JAN-384 itself may interfere with autophagy independently of its effects on eEF2K.
Tumour cells can acquire resistance to therapeutic agents or other adverse conditions, such as insufficient nutrients, through a variety of mechanisms (see e.g., [56]; reviewed [57,58]). We generated cells which could withstand nutrient withdrawal and inhibition of autophagy (by BafA), both on wildtype and eEF2K knock-out backgrounds. eEF2K has previously been shown to help cancer cells withstand nutrient starvation [8]. We then conducted mass spectrometry analysis to determine changes in the levels of specific proteins in the resistant as compared to the 'parental' non-resistant cells.
When comparing resistant to non-resistant cells, it was found that the abundance of a number of amino acyl-tRNA synthetases (and a ligase) were decreased, namely AARS (alanyl tRNA ligase), YARS, LARS, WARS, and TARS (i.e., the tyrosyl, leucyl, seryl, tryptophanyl and threonyl tRNA synthetases). Furthermore, a variety of translation factors (PABPC1, eIF3B, eIF4A1, eIF4A3, eIF4G1 and eIF5) also showed lower levels in resistance cells, as did RAN and RANGAP1, proteins that are linked to RNA transport, and the ribosomal proteins RPS3 and RPS8. Protein synthesis [47] and ribosome biogenesis consume a substantial proportion of metabolic energy in dividing cells; our data suggest that rates of these processes may be slowed in resistant cells to save energy, thus allowing cells to survive for longer under conditions of nutrient deprivation. This offers further evidence that slowing down protein synthesis (for example, due to activation of eEF2K) helps cells cope with shortage of nutrients and metabolic energy.
Interestingly, proteins involved in protein folding (chaperones, including HSPA5, HSPA4L, DNAJB1, ERO1L) were all decreased in resistant cell lines. This could make sense if, as seems likely from the above data, the cells had slower rates of protein synthesis and thus did not require as many chaperones to ensure correct protein folding. Also, several chaperones such as heat shock proteins consume ATP, so their downregulation may help cells to conserve this valuable resource.
Similarly, proteins involved in fatty acid synthesis (FASN, ACAT2), another energy-consuming process, were also reduced in resistant cells. Conversely, levels of ACADVL, a rate-limiting enzyme of fatty acid oxidation [59], was increased in resistant cells. ACSL4 (long chain fatty acid-CoA ligase 4), which catalyses the conversion of fatty acids to acyl-CoA for their degradation was decreased. ADH5 (alcohol dehydrogenase 5), which participates in the oxidation of hydroxy-fatty acids [60,61], was also decreased. The decreases in ADH5 and ACSL4 could be associated with the decrease in fatty acid synthesis since, if cells are making fewer fatty acids, they may also not need as many enzymes of fatty acid breakdown to degrade them. As fatty acid synthesis is not vital for cells during nutrient deprivation, these changes could potentially save (or generate) cellular energy.
Levels of certain enzymes involved in the metabolism of glucose were also higher, notably of the Misoform of pyruvate kinase (PKM), which catalyses a key step in glycolysis, and lactate dehydrogenase A (LDHA). PKM catalyses the conversion of phosphoenolpyruvate to pyruvate. LDHA converts pyruvate to lactate, a key reaction that allows the continuation of (anaerobic) glycolysis (by regenerating NAD + ) without the entry into the Krebs cycle of carbon atoms derived from glucose, and their subsequent loss as CO 2 . These changes are expected to allow glucose-deprived cells to retain more of their valuable carbon atoms than would be the case if pyruvate instead entered the Krebs cycle and its carbon atoms lost as CO 2 . This is reminiscent of the 'Warburg effect', a feature of many tumour cells [62,63]. Notably, the levels of ALDH2 (mitochondrial aldehyde dehydrogenase) were also decreased. This enzyme oxidises acetaldehyde to acetate, which can then be converted to acetyl-CoA by acetyl-CoA synthetase (ATP is required for this). Furthermore, the levels of DLAT (mitochondrial dihydrolipoyllysine-residue acetyltransferase component of pyruvate dehydrogenase complex), an enzyme which catalyses the final step of the conversion of pyruvate to acetyl-CoA were decreased. If the cells adapt by utilising anaerobic glycolysis to synthesise lactate, they would have less need to make acetyl-CoA, thereby saving ATP by reducing the rate of conversion of acetate to acetyl-CoA.
Furthermore, according to western blots (Fig. 9), p62, procathepsin D and cathepsin D levels were increased. This could potentially explain the higher level of basal autophagic flux observed in resistant cells (Fig. 5C). Cells may use this higher autophagic flux to gather amino acids from unneeded proteins as soon as they were placed into the glucose-free medium, prior to the point of time when BafA fully blocks autophagy.
Our data show that the acquisition of resistance to nutrient starvation is reversible, suggesting it is not due to stable genetic changes such as mutations. It might reflect epigenetic modifications that alter the expression of specific genes or subsets of genes. This reversibility may have implications for the design of anti-tumour therapies.
A striking feature of our data is that, while eEF2K aids the survival of A549 and MDA-MB-231 cells under nutrient-depleted conditions, the opposite is evident for PC3 cells, where loss of eEF2K aids their survival; the reasons for this surprising result are unclear and require further study. These findings do match features of the literature where, while disabling eEF2K generally impairs cells survival or tumour growth (see, e.g., [8,29,64]), some studies report the converse [10,13]. Our observation that the acquisition of resistance to BafA is temporary indicates that, as is now the case for some therapeutic agents, intermittent dosing may be more effective than continuous dosing, as tumour cells that initially become resistant to a drug may subsequently revert to being sensitive, such that later treatment with an agent will be more effective.                   Marker