Calcium signalling links MYC to NUAK1

NUAK1 is a member of the AMPK-related family of kinases. Recent evidence suggests that NUAK1 is an important regulator of cell adhesion and migration, cellular and organismal metabolism, and regulation of TAU stability. As such, NUAK1 may play key roles in multiple diseases ranging from neurodegeneration to diabetes and metastatic cancer. Previous work revealed a crucial role for NUAK1 in supporting viability of tumour cells specifically when MYC is overexpressed. This role is surprising, given that NUAK1 is activated by the tumour suppressor LKB1. Here we show that, in tumour cells lacking LKB1, NUAK1 activity is maintained by an alternative pathway involving calcium-dependent activation of PKCα. Calcium/PKCα-dependent activation of NUAK1 supports engagement of the AMPK-TORC1 metabolic checkpoint, thereby protecting tumour cells from MYC-driven cell death, and indeed, MYC selects for this pathway in part via transcriptional regulation of PKCα and ITPR. Our data point to a novel role for calcium in supporting tumour cell viability and clarify the synthetic lethal interaction between NUAK1 and MYC.

INTRODUCTION NUAK1, also known as ARK5, is one of 12 kinases related by sequence homology to the catalytic α-subunits of the metabolic regulator AMPK. 1 Perturbation of NUAK1 has revealed a diverse array of phenotypes, pointing to roles for NUAK1 in regulating cell adhesion, 2 directional migration, 3,4 neuronal axon branching, 5 glycogen synthesis, 6 replicative senescence 7 and TAU stabilization. 8 Overexpression of NUAK1 is associated with poor prognosis in multiple cancers, including colorectal, 9 (Port, et al., personal communication) ovarian, [10][11][12] and lung, 13 among others. 14 Accordingly, NUAK1 is a common target of multiple miRNAs that are frequently lost during progression to metastatic disease. [15][16][17][18][19][20] Despite the evidence that NUAK1 may contribute to multiple diseases, the signal transduction context of NUAK1 remains poorly defined.
We previously identified a role for NUAK1 in supporting viability of cancer cells when MYC is overexpressed. 21 Briefly, we showed that MYC-overexpressing cells are unable to maintain energetic homoeostasis in the absence of NUAK1, in part due to a failure to efficiently activate AMPK and slow TORC1-dependent protein translation when faced with metabolic stress. NUAK1 and AMPK thus protect cancer cells from metabolic stress, which is a hallmark of most solid tumours. 22,23 This tumour-promoting activity of NUAK1 and AMPK is somewhat paradoxical, given that both are activated by LKB1, an established tumour suppressor: LKB1 phosphorylates AMPKα subunits on Thr 172 , and NUAK1 on Thr 211 , within the conserved T-loop of the kinase domain. 1,24 Notably, AMPKα T172 is phosphorylated by CamKK2 in response to calcium signalling, 25,26 suggesting that the T-loops of these kinases may be accessible to other upstream regulators in addition to LKB1.
Here we demonstrate that NUAK1, like AMPK, is active in cancer cells in the absence of LKB1. Similar to AMPK, basal NUAK1 activity is maintained by tonic Ca 2+ signalling and activity increases in response to Ca 2+ mobilization. Unlike AMPK, NUAK1 does not appear to be regulated by CamKK2, but rather by Ca 2+ -dependent activation of PKCα. Significantly, suppression of either NUAK1 or PKCα leads to MYC-dependent cell death and MYC selects for increased Ca 2+ signalling in part via transcriptional regulation of Ca 2+ -dependent protein kinases. Our work thus reveals a novel role for Ca 2+ signalling in supporting viability of MYCoverexpressing cells via activation of PKCα and NUAK1.

RESULTS
NUAK1 is specifically required for Ca 2+ -dependent AMPK activity Depletion of NUAK1 impairs activation of AMPK in response to sustained MYC deregulation. 21 We asked whether this requirement for NUAK1 is a general feature of AMPK regulation or rather a context-dependent event. AMPK is activated by phosphorylation of the α-subunit on Thr 172 by LKB1, and activity is further enhanced upon a drop in the ATP:AMP/ADP ratio. 27 Alternatively, CamKK2 can phosphorylate AMPKα Thr 172 in response to calcium signalling. 25 Additionally, AMPK can be activated upon direct binding of pharmacological agonists, such as salicylate or A769662. 28 We therefore considered three modes of AMPK activation: indirect activation in response to energetic stress, direct activation by agonist binding and calcium-dependent activation.
In order to investigate the requirement of each mode of AMPK activation for NUAK1, we made use of a recently described highly selective NUAK1 kinase inhibitor, HTH-01-015. This molecule shows little-to-no activity towards AMPK, NUAK2 or other AMPKrelated kinases (ARKs) in vitro. 29 Treatment of U2OS cells with HTH-01-015 for 1 h, 6 h or overnight reduced Ser 445 phosphorylation of the NUAK1 substrate MYPT1 in a dose-dependent manner ( Figure 1a). In contrast, acute treatment with HTH-01-015 had no effect on basal Ser 79 phosphorylation of the canonical AMPK substrate, ACC. Acute activation of AMPK in U2OS cells using the direct agonist A769662, the electron transport chain inhibitor phenformin or the widely-used Ca 2+ ionophore A23187, all increased phospho-ACC levels. Co-treatment with HTH-01-015 attenuated this increase in response to AMPK activation by Ca 2+ and by phenformin, but not by the direct AMPK agonist A769662, indicating that the requirement for NUAK1 during AMPK activation is context dependent (Figure 1b).
Suppression of Ca 2+ -dependent AMPK activity by HTH-01-015 suggested that NUAK1 may be important during LKB1independent regulation of AMPK. We therefore repeated the above analysis in HeLa cells, which lack functional LKB1 ( Figure 1c). As measured by ACC phosphorylation, activation of AMPK by either direct agonist or phenformin was much weaker in HeLa cells than in U2OS cells, and neither was affected by NUAK1 inhibition. In contrast, Ca 2+ ionophore clearly increased AMPK activity and this increase was attenuated by NUAK1 inhibition (Figure 1d), suggesting a specific role for NUAK1 in this mode of AMPK activation. NUAK1 is activated by calcium signalling LKB1 is a master regulator of the AMPK-related kinases, including NUAK1. 1 Our results in HeLa cells implied that NUAK1 is active in these cells despite the absence of LKB1. The myosin phosphatase targeting subunit of PP1β, MYPT1 (PPP1R12A), is to date the bestcharacterized substrate of NUAK1 kinase activity. 2 Acute treatment of HeLa cells with HTH-01-015 reduced MYPT1 phospho-Ser 445 levels, suggesting that NUAK1 is indeed catalytically active in these cells (Figure 2a). Depletion of NUAK1 using two independent siRNAs also reduced phospho-MYPT1 S445 , confirming the specificity of the inhibitor effect (Figure 2b). The partial reduction in phospho-MYPT1 S445 observed upon NUAK1 suppression suggested that other kinases may contribute to MYPT1 S445 phosphorylation. Indeed, NUAK2, the ARK most closely related to NUAK1, was previously reported to phosphorylate this site. 2 Accordingly, depletion of NUAK2 also reduced phospho-MYPT1 S445 levels, while combined suppression of both NUAK1 and 2 almost completely abolished MYPT1 S445 phosphorylation ( Figure 2c). Pharmacological inhibition of both NUAK1 and NUAK2, using the dual-specificity inhibitor WZ4003 29 similarly abolished MYPT1 phosphorylation, corroborating the results of the siRNA (Supplementary Figure S1A). Thus, both NUAK1 and NUAK2 are active in HeLa cells despite their LKB1-null status.
These data indicate that NUAK1 is activated in HeLa cells by an alternative upstream kinase. We first asked if CamKK2, a known upstream activator of AMPK, might similarly activate NUAK1. Treatment of HeLa cells with the CamKK2 inhibitor STO-609 strongly suppressed phosphorylation of both AMPKα T172 and ACC S79 but had no influence on phospho-MYPT1 S445 levels (Figure 2d), suggesting that CamKK2 is not upstream of NUAK1. Strikingly, treatment with calcium ionophore A23187 increased phosphorylation of both ACC S79 and MYPT1 S445 and co-treatment with STO-609 reduced ACC S79 phosphorylation but again had no effect on phospho-MYPT1 S445 , suggesting that NUAK1 is activated by a calcium-dependent kinase other than CamKK2. Accordingly, treatment of HeLa cells with two different Ca 2+ ionophores,  Ionomycin or A23187, increased MYPT1 S445 phosphorylation and this increase was attenuated by NUAK1 inhibition (Figure 2e). Conversely, treatment with the calcium chelator BAPTA strongly reduced basal levels of phospho-MYPT1 S445 (Figure 2f). Collectively, these data suggest that Ca 2+ signalling activates NUAK1 in the absence of LKB1. Interestingly, treatment with the Ca 2+ ionophore A23187 could partially rescue MYPT1 S445 phosphorylation in the presence of NUAK1 inhibitor but not in the presence of the dual NUAK1/NUAK2 inhibitor, WZ4003 (Supplementary Figure S1A). A23187 also partially rescued MYPT1 S445 phosphorylation upon suppression of either NUAK1 or NUAK2 but not both (Supplementary Figure S1B), together suggesting that NUAK2 is also activated by calcium signalling in the absence of LKB1.
MYC drives increased PKC activity Calcium regulates multiple kinases including Ca 2+ /Calmodulindependent kinases 1-4; CamKK1 and 2; and conventional isoforms of protein kinase C (cPKC). Noting our previously described link between NUAK1 and MYC overexpression, we wondered if MYC might drive expression of a calcium-dependent kinase upstream of NUAK1. Oncogenic transformation of mouse embryo fibroblasts (MEFs) with MYC specifically increased expression of PKCα and CamKKβ, along with the inositol tri-phosphate receptor ITPR1, which regulates calcium release from the endoplasmic reticulum ( Figure 3a). Notably, MYC was previously shown to bind the promoters of all three genes in diverse cell types, including MEFs. 30 MYC overexpression strongly enhanced PKC activity, as measured by phosphorylation of the canonical PKC substrate MARCKS, 31 and modestly but reproducibly enhanced Ca 2+ -dependent activation of AMPK (Figures 3b and c). MARCKS phosphorylation was suppressed by BAPTA or by treatment with the PKCα/β inhibitor Gö6976, 32 suggesting that deregulated MYC specifically increases activity of Ca 2+ -dependent PKC isoforms ( Figure 3b). HeLa cells express high levels of MYC 33 and depletion of MYC in HeLa cells reduced p-MARCKS levels (Figure 3d), suggesting that this consequence of MYC overexpression is conserved across species.

NUAK1 is activated by PKCα
We next asked if targeted suppression of PKC impairs NUAK1 activity. Inhibition of PKCα and β isoforms with Gö6976 strongly reduced p-MYPT1 S445 in a dose-dependent manner ( Figure 4a). Notably, this effect was transient, as p-MYPT1 S445 levels rebounded within 16 h of Gö6976 treatment, and this was mirrored by a recovery in overall PKC activity ( Figure 4b). SiRNA-mediated depletion of PKCα also reduced p-MYPT1 S445 levels to a degree that was similar to NUAK1 inhibition but less than that observed after 1 h treatment with the highest concentration of Gö6976 tested, which may reflect promiscuity of the PKC inhibitor at this dose ( Figure 4c; note that a lower concentration of Gö6976, 0.5 μM, was used for all subsequent experiments). No effect on p-MYPT1 S445 was observed using siRNA targeting PKCβ1 (not shown).
In light of our data showing that both NUAK1 and NUAK2 contribute to Ca 2+ -induced MYPT1 S445 phosphorylation, we asked if the effects of PKCα depletion were mediated by either NUAK1, NUAK2, or both. The reduction of p-MYPT1 S445 achieved upon NUAK1 depletion was minimally influenced by co-depletion of PKCα (compare lane 2 with lane 4), consistent with a role for PKCα upstream of NUAK1. In contrast, suppression of MYPT1 S445 phosphorylation by NUAK2 depletion was strongly enhanced by co-depletion of PKCα (compare lanes 5 and 6), suggesting that NUAK2 resides in a distinct pathway ( Figure 4d). Interestingly, depletion of PKCα consistently reduced expression of NUAK1 ( Figure 4d). This effect was observed using two independent siRNAs targeting PKCα and neither siRNA influenced NUAK1 mRNA levels (Supplementary Figure S2), strongly suggesting that the effect does not reflect off-target activity of the siRNAs used. Proteasome inhibition largely rescued NUAK1 levels upon depletion of PKCα, suggesting that PKCα promotes NUAK1 protein stability ( Supplementary Figures S2B and C).
To examine the effects of acute calcium signalling on NUAK1 activation, we requisitioned an affinity-purified phosphopeptide antibody against T211-phosphorylated NUAK1, and overexpressed either wild type or T211A mutant, FLAG-tagged, NUAK1 in HeLa cells. In FLAG immunoprecipitates, the antibody strongly detected a band migrating at the correct size for NUAK1 only in lysates from WT but not from T211A mutant-overexpressing cells. Identical results were obtained using a commercial anti-phospho-AMPKα T172 antibody that cross-reacts with overexpressed phospho-NUAK1 T21134 (Supplementary Figure S2). For both antibodies, the intensity of this band increased within 10 min of Ca 2+ ionophore treatment and decreased upon acute treatment with PKC inhibitor (Figure 4e and Supplementary Figure S2).    (Figure 4f), no such rescue was evident in p-NUAK1 T211 levels.
Taken with the data above, these data suggest that Ca 2+ signalling regulates NUAK1 in HeLa cells via activation of PKCα, while MYPT1 S445 phosphorylation is regulated both via NUAK1 and via a distinct pathway involving Ca 2+ -dependent, Gö6976-refractory, activation of NUAK2.
The PKCα-NUAK1 pathway supports viability of MYC-overexpressing cells We previously showed that MYC-overexpressing cells require NUAK1 to sustain viability. 21 (Figure 5a). Partial inhibition of NUAK1 with 5 μM HTH-01-015 was surprisingly well tolerated, suggesting that a threshold level of NUAK1 activity is sufficient to prevent cell death. Similar results were obtained using siRNAmediated NUAK1 depletion, in that death was only induced upon very strong suppression of NUAK1 expression (Figure 5b). Death induced by 10 μM HTH-01-015 was significantly attenuated by reducing MYC levels with either of two MYC-targeting shRNAs (Figure 5c), consistent with our previous demonstration of MYC 'dose-dependence' for the synthetic lethal interaction with NUAK1. 21 Consistent with a role for PKCα upstream of NUAK1, depletion of PKCα with either of two siRNAs also drove pronounced HeLa cell apoptosis (Figure 5d), while treatment of HeLa cells with Gö6976 significantly enhanced killing by a sublethal dose of NUAK1 inhibitor (Figure 5e). Note that Gö6976 treatment alone did not kill HeLa cells, likely owing to the transient nature of PKC inhibition by this compound (Figure 4b).
We asked if death induced upon loss of PKCα mechanistically mirrored that induced by loss of NUAK1. Under conditions of energetic stress, cancer cells activate a metabolic checkpoint in order to limit mTORC1-driven macromolecular synthesis, via phosphorylation of RAPTOR-Ser 792 by AMPK. 35 Failure to engage this checkpoint results in death of stressed cells 24,36 and our previous work showed that NUAK1 is required for efficient checkpoint activation. 21 Dynamic analysis of this checkpoint in HeLa cells revealed a complex response to Ca 2+ ionophore, with p-RAPTOR S792 increasing steadily over time whereas phospho-S6K T389 and phospho-4EBP1 T37/46 levels, downstream of mTORC1, rose initially before declining (Figure 5f), consistent with Ca 2+ simultaneously activating the mTORC1 pathway as well as the inhibitory AMPK-Raptor pathway. 37 Importantly, depletion of either PKCα or NUAK1 reduced both basal and Ca 2+ -activated phosphorylation of RAPTOR-Ser 792 , suggesting that failure to efficiently engage the metabolic checkpoint may contribute to death in both instances (Figure 5g). Consistent with this hypothesis, treatment of HeLa cells with the mTORC1 inhibitor Rapamycin significantly rescued cells from death induced by depletion of either NUAK1 or PKCα, and the degree of rescue was similar in both instances (Figure 5h). Although these data do suggest that other downstream pathways likely contribute to cell death, they strongly support the core observation that NUAK1 and PKCα act in a similar manner to support cell viability.

NUAK1 regulates RAPTOR via both AMPK-dependent and independent mechanisms
Confirming the requirement for NUAK1 to restrain mTORC1 activity, S35 -Methionine labelling showed increased protein translation in NUAK1-depleted HeLa and U2OS cells (Figure 6a), as shown previously. 21 We therefore examined RAPTOR regulation by NUAK1 in greater detail. Activation of AMPK by Ca 2+ ionophore (A23187), phenformin or salicylate in U2OS cells all lead to increased RAPTOR S792 phosphorylation. In contrast with the selective requirement for NUAK1 during AMPK regulation of ACC, RAPTOR S792 phosphorylation was reduced by NUAK1 inhibition under all conditions examined (Figure 6b). Depletion of NUAK1 also significantly reduced both basal and AMPKactivated RAPTOR S792 phosphorylation, confirming the specificity of this effect (Figure 6c). Inhibition of NUAK1 reduced AMPKdependent RAPTOR S792 phosphorylation in immortalized Prkaa1 FL/FL ;Prkaa2 FL/FL double floxed MEFs. Strikingly, phospho-RAPTOR S792 was still detectable in the same MEFs after CRE recombinase-mediated deletion of AMPKα1 and α2, albeit at reduced levels, and NUAK1 inhibition further reduced detection, indicating that NUAK1 can regulate RAPTOR in the absence of functional AMPK (Figure 6d). Accordingly, deletion of NUAK1 in Nuak1 FL/FL MEFs also reduced both basal and AMPK-activated RAPTOR S792 phosphorylation (Figure 6e). Together these data show that efficient restraint of mTORC1 via inhibitory phosphorylation of RAPTOR requires both NUAK1 and AMPK.

DISCUSSION
Suppression of NUAK1 is synthetic lethal with MYC overexpression, suggesting that NUAK1 may present an attractive target for treatment of MYC-driven cancers. 21,38 A thorough understanding of the signal transduction context of NUAK1 will be crucial to determine if such a strategy is feasible in human subjects. Here we show that NUAK1 is active in HeLa cells despite the absence of LKB1. We show modulation of NUAK1 activity by calcium perturbation, and present evidence that PKCα participates in NUAK1 activation in response to Ca 2+ signalling. Importantly, Ca 2+ -dependent activation of the AMPK-mTORC1 metabolic checkpoint requires both PKCα and NUAK1, and depletion of either drives pronounced apoptosis, suggesting a positive role for this pathway in tumour maintenance. Our specific findings are summarized in Figure 7.
It is widely thought that the tumour suppressive function of LKB1 is mediated by one or more of the AMPK-family kinases. 27,39 Loss of LKB1 would thus be predicted to result in loss of ARK activity, downstream. Accordingly, deletion of Stk11, encoding Lkb1, in wild-type MEFs was shown to suppress activity of AMPK and all related ARKs, as measured in cell-free kinase assays using a peptide substrate optimized for AMPK. 1 However, several of the ARKs, including Nuak1, showed only weak activity towards the peptide used, suggesting it was a suboptimal substrate for these kinases. Indeed, subtle differences in peptide substrate sequences have revealed distinct preferential phosphorylation patterns of AMPK and MARK kinases. 40 Thus, in vitro kinase assays with a onesize-fits-all peptide substrate likely fail to accurately reflect Additionally, several independent groups have definitively shown that AMPK is directly phosphorylated by CamKK2, reflecting an alternative pathway to AMPK activation. 25,26,41,42 Activation of AMPK by CamKK2 is particularly important in prostate cancer and in the physiological regulation of skeletal muscle and vascular endothelial cell function. [43][44][45] Interestingly, the ARK SIK2 was recently shown to be activated by an as-yet unidentified Ca 2+ -dependent kinase in Ovarian cancer cells. 46 Our demonstration that NUAK1 and NUAK2 are similarly regulated by Ca 2+ -dependent signalling thus fits an emerging pattern of calcium regulating multiple ARKs, either alongside or in the absence of LKB1. This regulation may have particular relevance in LKB1-deficient disease settings.
Our data speak to the complexity of signal transduction through AMPK, NUAK1 and the related ARKs. Indeed, AMPK is often discussed as if it were a single entity. Rather, up to 12 different permutations of trimeric AMPK complexes can assemble from the 2α, 2β and 3γ-encoded subunits, not accounting for splice variants. 22 It is likely that the different AMPK complexes may respond differentially to distinct upstream stimuli, and indeed in terms of their activity towards specific downstream substrates. Our demonstration of a specific requirement for NUAK1 in Ca 2+ -dependent AMPK activity towards ACC, and a more general requirement for NUAK1 in AMPK activity towards RAPTOR, point towards a highly contextual requirement for NUAK1 and may indicate that NUAK1 modulates the activity of a specific subset of AMPK complexes. On top of this, the 11 related ARKs can exhibit both overlapping and private substrate specificities. This is reflected by our demonstration of an AMPK-independent role for NUAK1 in RAPTOR regulation, and by phosphorylation of MYPT1 by NUAK1, NUAK2 and potentially by additional ARKs. Consistent with this, we also find Ca 2+ -dependent phosphorylation of the canonical AMPK substrate ACC even after complete suppression of CamKKβ-dependent AMPK activity in HeLa cells. Clearly, considerably more work will be needed to disentangle these complex signalling networks. Whereas calcium has long been recognized to drive MYC expression 47 and more recently to regulate MYC function, 48,49 the reciprocal regulation of calcium signalling by MYC has not garnered much attention. MYC was shown to increase calcium signalling during B-cell differentiation by suppressing expression of the calcium exporter PMCA. 50 ChIP-SEQ analysis has revealed MYC binding to the promoters of ITPR1-3, PRKCA and CamKK2 in diverse cell types, 30 consistent with our observation that MYC promotes expression of these genes. The pronounced increase in phosphorylation of the PKC substrate MARCKS compared with the much more modest effect of MYC overexpression on PKCα levels suggests that regulation of this pathway by MYC is only partially explained by the observed transcriptional effects. Nevertheless, our data do suggest that MYC actively selects for increased cellular sensitivity to calcium, and does so in part to promote NUAK1 activity, maintain metabolic homoeostasis and thereby sustain cell viability. The relative contribution of calcium signalling to NUAK1 activation likely depends on several factors including the strength of calcium signalling, whether LKB1 is present or absent and, if present, the relative levels of PKCα and LKB1 upstream.
MYC is a paradigm driver of apoptosis when expressed at high levels 51,52 whereas conventional PKC isoforms inhibit apoptosis in many cell types. 53 Suppression of PKCα or β induces apoptosis, [54][55][56] whereas overexpression has been shown to suppress death induced by MYC or withdrawal of IL3. 57,58 Both PKCα and β isoforms have been shown to promote Ser 473 phosphorylation of AKT, 59,60 which inhibits canonical MYCinduced apoptosis, primarily via suppression of pro-apoptotic BH3 protein expression/function. [61][62][63] Importantly, we showed previously that MYC-overexpressing cells continue to require NUAK1 even when active AKT is overexpressed, 21 pointing to a role for NUAK1 in protecting tumour cells from non-apoptotic cell death and suggesting that calcium and PKCα may govern multiple pathways that promote tumour cell survival. Targeted suppression of these pathways may thus have therapeutic benefit in multiple cancers where MYC is deregulated. 64
Immunoprecipitation and immunoblotting FLAG-NUAK1 wild type, mutant (T211A) or empty vector transiently overexpressed HeLa cells were rinsed with ice-cold phosphate-buffered saline and then lysed in Lysis Buffer containing 50 mM Tris-HCl (pH 7.5), 1% NP-40, 0.27 M sucrose and phosphatase/protease inhibitors. Cell lysates (1 mg) were incubated overnight at 4°C with anti-FLAG M2 Affinity gel (Sigma, Irvine, UK; A2220). Immunoprecipitated were washed twice with Lysis Buffer containing 0.15 M NaCl, twice with 50 mM Tris-HCl (pH 7.5) plus phosphatase inhibitors and resuspended in sodium dodecyl sulphate sample buffer. For whole-cell extracts, cells were rinsed with ice-cold phosphate-buffered saline and then lysed in situ with lysis buffer containing 150 mM NaCl, 50 mM Tris (pH 7.5), 1% NP-40, 0.5% sodium deoxycholic acid, 1% sodium dodecyl sulphate plus protease and phosphatase inhibitor cocktails. Lysates were then sonicated to reduce viscosity and diluted in sodium dodecyl sulphate sample buffer. Immunoprecipitated and whole-cell extracts were resolved by sodium dodecyl sulphatepolyacrylamide gel electrophoresis, transferred to nitrocellulose membranes for subsequent incubation with primary antibodies overnight at 4°C. Densitometry analysis of individual immunoblots was performed using ImageJ (NIH, Bethesda, MD, USA).
Cell death analysis HeLa cells were treated or transfected as for figure legends and on the day of the analysis the supernatant was collected, cells were rinsed in phosphate-buffered saline and harvested by trypsinization. Cells were then centrifuged at 300 g for 5 min at 4°C and pellet incubated in 200 μl Annexin V binding buffer (10 mM HEPES pH 7.4, 140 mM NaCl, 2.5 mM CaCl 2 ) containing APC-Annexin V (Biolegend, San Diego, CA, USA) for 10 min at room temperature. Propidium iodide was added prior to analysis by FACSCalibur (BD Biosciences, Wokingham, UK) flow cytometry.
Statistical analysis Raw data were uploaded into Prism (Graphpad, La Jolla, CA, USA) or Excel (Microsoft, Reading, UK) spreadsheets for generation of graphs. All experiments were performed on at least three occasions, except where noted, and mean and s.d. values from biological replicates are presented. Statistical significance was determined by T-test and one-way or two-way ANOVA as per figure legends. * denotes Po 0.05; **o0.01; ***o 0.001.