Following genotoxic stress, cells activate a complex signalling network to arrest the cell cycle and initiate DNA repair or apoptosis. The tumour suppressor p53 lies at the heart of this DNA damage response. However, it remains incompletely understood, which signalling molecules dictate the choice between these different cellular outcomes. Here, we identify the transcriptional regulator apoptosis‐antagonizing transcription factor (AATF)/Che‐1 as a critical regulator of the cellular outcome of the p53 response. Upon genotoxic stress, AATF is phosphorylated by the checkpoint kinase MK2. Phosphorylation results in the release of AATF from cytoplasmic MRLC3 and subsequent nuclear translocation where AATF binds to the PUMA, BAX and BAK promoter regions to repress p53‐driven expression of these pro‐apoptotic genes. In xenograft experiments, mice exhibit a dramatically enhanced response of AATF‐depleted tumours following genotoxic chemotherapy with adriamycin. The exogenous expression of a phospho‐mimicking AATF point mutant results in marked adriamycin resistance in vivo. Nuclear AATF enrichment appears to be selected for in p53‐proficient endometrial cancers. Furthermore, focal copy number gains at the AATF locus in neuroblastoma, which is known to be almost exclusively p53‐proficient, correlate with an adverse prognosis and reduced overall survival. These data identify the p38/MK2/AATF signalling module as a critical repressor of p53‐driven apoptosis and commend this pathway as a target for DNA damage‐sensitizing therapeutic regimens.
There is a Have you seen? (October 2012) associated with this Article.
In response to DNA damage, cells activate a complex signalling network to prevent further cell‐cycle progression. Activation of this signalling cascade, which is collectively referred to as the DNA damage response (DDR), provides time for DNA repair, recruits repair machinery to the sites of genotoxic damage, or, if the lesions are beyond repair capacity, leads to the activation of additional pathways mediating apoptosis (Jackson and Bartek, 2009). The DDR has traditionally been divided into two major kinase branches operating through the upstream kinases ATM and ATR together with their respective effector kinases Chk2 and Chk1 (Jackson and Bartek, 2009). Moreover, p38 and its downstream substrate mitogen‐activated protein kinase‐activated protein kinase‐2 (MK2) has recently been identified as a third checkpoint effector kinase complex operating downstream of ATM and ATR and parallel to Chk1 (Manke et al, 2005; Raman et al, 2007; Reinhardt et al, 2007, 2010). p38 and MK2 are components of a general stress kinase pathway that acts in response to a variety of stimuli, including inflammatory signals, reactive oxygen species, heat shock, hyperosmolar stress in addition to DNA damage (Kyriakis and Avruch, 2001).
One of the major downstream targets of the DDR network is the transcription factor p53. DNA damage‐induced phosphorylation of p53, through ATM, Chk2, MK2 and others, at amino‐terminal sites close to the MDM2‐binding region is thought to reduce ubiquitin‐dependent degradation allowing accumulation in the nucleus, where p53 contributes to two distinct cellular responses (Toledo and Wahl, 2006): p53 promotes cell death in response to genotoxic stress through the induction of pro‐apoptotic target genes such as PUMA, BAX and BAK. In contrast, p53‐mediated transactivation of cell‐cycle‐arresting target genes, such as CDKN1A, GADD45A or RPRM, serves a protective function by allowing time to repair genotoxic lesions. The molecular cues that dictate the cellular decision between a protective p53‐mediated cell‐cycle arrest or p53‐driven apoptosis have remained largely unclear (Reinhardt and Schumacher, 2012). Numerous regulatory mechanisms, involving selective DNA binding of p53 to its target gene promoters, selective transactivation of p53‐bound target genes and differential stability of p53‐dependent transcripts, have been reported to modulate these p53‐dependent cell fate decisions, which are essential to repress the uncontrolled proliferation of incipient cancer cells carrying severely damaged genomic material (Oda et al, 2000b; Seoane et al, 2002; Wei et al, 2006; Das et al, 2007; Espinosa, 2008).
Here, we identify the apoptosis‐antagonizing transcription factor (AATF) as a critical regulator of p53‐driven apoptosis. AATF had previously been shown to be a substrate of the canonical DDR kinases ATM/ATR and Chk2 (Bruno et al, 2006) and to exert anti‐apoptotic activity through promoting expression of XIAP in response to genotoxic stress (Bruno et al, 2008). We now demonstrate that AATF negatively regulates p53‐driven apoptosis by preventing efficient DNA damage‐induced transactivation of the pro‐apoptotic p53 target genes PUMA, BAX and BAK. Moreover, we show that AATF resides in cytoplasmic protein complexes with the cytoskeletal protein myosin‐regulatory light chain‐3 (MRLC3) in resting cells. Following DNA damage, this interaction is lost through MK2‐dependent AATF phosphorylation resulting in nuclear translocation of AATF and the specific engagement of the PUMA, BAX and BAK promoters to repress p53‐dependent transcription of these proapoptotic genes. Interestingly, AATF neither binds to the promoters, nor regulates the expression of the cell‐cycle‐regulating p53 target genes CDKN1A, GADD45α or RPRM. Furthermore, our data show that AATF amplification in tumours results in worse prognosis of neuroblastoma patients. We propose that AATF functions to repress the pro‐apoptotic arm of the p53 response to promote a primarily growth‐arresting cellular response to genotoxic insults. These data suggest AATF to be an attractive drug target since AATF depletion resulted in substantially increased sensitivity to frontline chemotherapeutic drugs.
A phospho‐proteomic screen links MRLC3 to DNA damage kinase signalling
To identify new protein complexes that are regulated through DNA damage kinase signalling, we employed a phosphopeptide library versus protein expression library screening approach. We were interested in protein complexes that were disrupted in a phosphorylation‐dependent manner, such as the p53:MDM2 complex, which dissociates upon p53 phosphorylation by numerous DDR kinases (Lavin and Gueven, 2006; Toledo and Wahl, 2006). We used a library of partially degenerate peptides resembling the substrate motif phosphorylated by the DDR kinases Chk1/2 and MK2, corresponding to the sequence [L/I/F]‐X‐[R/K]‐[Q/X]‐X‐S/T‐X‐X‐X, where X represents any amino acid except cysteine. Both phospho‐ and non‐phospho versions of this library, hereafter denoted ϕ‐X‐R‐X‐X‐pT and ϕ‐X‐R‐X‐X‐T, were generated. Streptavidin bead‐immobilized peptide libraries were used as bait in an interaction screen against a protein library produced by in vitro transcription/translation (Elia et al, 2003; Manke et al, 2003). We screened a total of ∼200 000 cDNAs arrayed in 2000 pools containing 100 individual, in vitro‐translated and 35S‐labelled cDNAs. Most in vitro‐translated proteins did not bind to either library (Figure 1A). However, we identified 14 distinct pools, in which a 17‐kDa protein displayed preferential binding to the ϕ‐X‐R‐X‐X‐T library (Figure 1A). Progressive subdivision of these pools identified the cytoskeletal protein MRLC3 as the clone that preferentially associated with the ϕ‐X‐R‐X‐X‐T library (Figure 1A and B). To further validate this interaction, we expressed FLAG‐tagged MRLC3 in HEK293T cells and performed in vitro pull down experiments using the streptavidin‐immobilized ϕ‐X‐R‐X‐X‐T and ϕ‐X‐R‐X‐X‐pT libraries as bait. As shown in Supplementary Figure 1A, MRLC3 displayed robust binding to the ϕ‐X‐R‐X‐X‐T, but essentially no binding to the ϕ‐X‐R‐X‐X‐pT library, suggesting that Thr‐phosphorylation within the checkpoint kinase motif disrupts the interaction with MRLC3.
We next investigated the interactome of MRLC3 using yeast two‐hybrid screening. These experiments identified AATF as a likely MRLC3‐interacting protein. To confirm this interaction in mammalian cells, we performed co‐immunoprecipitation experiments in HEK293T cells co‐expressing V5.AATF and FLAG.MRLC3 or FLAG.GFP, as a control. While AATF could readily be detected in the FLAG.MRLC3 precipitates, it was undetectable in the FLAG.GFP precipitations, thus validating the interaction between AATF and MRLC3 (Supplementary Figure 1B). Since MRLC3 was identified as a protein with strong selective binding to peptides corresponding to the non‐phosphorylated forms of checkpoint kinase substrate motifs, but not to these same peptides following phosphorylation, we asked whether the AATF:MRLC3 interaction could be disrupted by phosphatase inhibition. In agreement with the results of the phospho‐proteomic screen, treatment of V5.AATF and FLAG.MRLC3‐expressing cells with the Ser/Thr phosphatase inhibitor okadaic acid, abrogated the AATF:MRLC3 interaction (Figure 1C).
We then went on to investigate whether the phosphorylation‐sensitive interaction between AATF and MRLC3 is regulated by checkpoint kinases in response to genotoxic stress and performed co‐immunoprecipitation experiments before and after DNA damage. As we had observed before, V5.AATF co‐precipitated with FLAG.MRLC3 in mock‐treated cells. In contrast, this interaction was abolished when cells were pre‐treated with UV‐C, indicating that genotoxic stress negatively regulates MRLC3:AATF complex formation (Figure 1D). Identical co‐precipitation behaviour was observed when the FLAG and V5 tags were swapped (Figure 1E). Disruption of the MLRC3:AATF complex was also observed following treatment of cells with doxorubicin, indicating that the complex is sensitive to multiple types of genotoxic stress (Supplementary Figure 1C).
To ask whether endogenous AATF and MRLC3 form similar DNA damage‐sensitive complexes, we immunoprecipitated AATF from HCT116 cells and used immunoblotting to detect co‐precipitating MRLC3. These experiments confirmed the existence of a physiological interaction between AATF and MRLC3 in resting cells (Figure 1F, lane 3). As expected, application of UV‐C or addition of doxorubicin prior to cell lysis abolished this endogenous interaction (Figure 1F and G), recapitulating the effects seen with overexpressed proteins. These data demonstrate that AATF and MRLC3 form a phosphorylation‐sensitive protein complex, which is disrupted in response to genotoxic stress, likely mediated through the activity of a basophilic checkpoint kinase.
MRLC3 sequesters AATF in the cytoplasm
While MRLC3 is believed to reside predominantly in the cytoplasm, the subcellular localization of AATF is less well understood (Watanabe et al, 2007). Furthermore, it remains unclear whether AATF or MRLC3 dynamically shuttle between distinct subcellular compartments upon disruption of the AATF:MRLC3 complex. We directly investigated the spatial dynamics of MRLC3 and AATF in mouse embryonic fibroblasts (MEFs), using biochemical separation of nuclear and cytoplasmic fractions through hypotonic lysis. As shown in Figure 2A, MRLC3 was found exclusively in the cytoplasm and its subcellular distribution was not affected by UV‐C‐induced genotoxic stress. In marked contrast, AATF showed a DNA damage‐dependent dynamic re‐localization between cytoplasm and nucleus. While only minuscule amounts of endogenous AATF were detectable in the nuclei of resting cells, we observed a marked enhancement of nuclear AATF after UV‐C (Figure 2B). We confirmed these data, using immunofluorescence. As shown in Figure 2C, cells displayed strong cytoplasmic MRLC3 staining before and after UV‐C. In contrast, the AATF antibody revealed prominent granular staining in both the cytoplasm and the nucleus of resting cells (Figure 2D). Control experiments performed in AATF‐depleted cells revealed that the nuclear staining in resting cells was unspecific (Supplementary Figure 2). Following UV‐C irradiation, we observed a statistically significant loss of cytoplasmic AATF staining, consistent with depletion of cytoplasmic AATF pools and subsequent nuclear accumulation (Figure 2D and E). Confirming phosphorylation dependence, treatment of cells with okadaic acid produced a similar nuclear translocation of AATF. As shown in Figure 2F, pre‐treatment of MEFs with okadaic acid resulted in a robust nuclear accumulation of AATF, suggesting that AATF phosphorylation, likely by a DNA damage‐activated basophilic kinase, leads to disruption of cytoplasmic AATF:MRLC3 complexes with subsequent nuclear AATF accumulation.
Genotoxic stress releases AATF from MRLC3 in an MK2‐dependent manner
While the DDR is typically initiated within the nucleus, and this is the primary site of action for Chk1 (Supplementary Figure 2), our data strongly suggested that a DNA damage‐responsive basophilic kinase(s) was (directly) responsible for the disruption of cytoplasmic AATF:MRLC3 complexes in the cytoplasm. We recently identified MK2 as a checkpoint kinase sharing substrate specificity with Chk1 and Chk2, but whose primary site of action is in the cytoplasm (Reinhardt et al, 2010). Intriguingly, when we repeated our fractionation experiments with cells that had been exposed to osmotic stress, a strong MK2 stimulus (Engel et al, 1998) that has not been reported to induce Chk1/2 activation, we also observed nuclear AATF accumulation (Figure 2G). We next used in vitro kinase assays to test whether MK2 directly phosphorylated AATF. As shown in Supplementary Figure 3, MK2 readily phosphorylated AATF, while no significant 32P incorporation could be detected in control reactions lacking MK2. In silico amino‐acid sequence analyses using the Scansite software (Obenauer et al, 2003) revealed that AATF contains a sequence stretch centred around Thr‐366 that is an exact match to the kinase motif selected by MK2. Mass spectrometry confirmed phosphorylation of AATF at Thr‐366. As shown in Figure 3A, pThr‐366‐containing peptides were readily detected in Lys‐C digests from in vitro kinase assays containing AATF and MK2, suggesting that MK2 directly phosphorylates AATF on Thr‐366. To test whether the Thr‐366‐containing peptide sequence might mediate the phosphorylation‐sensitive interaction between AATF and MRLC3, we synthesized Thr‐366 and pThr‐366‐containing biotinylated peptides, spanning amino acid 357–374. Peptides were immobilized on streptavidin beads, incubated with lysates from FLAG.MRLC3 transfected cells and peptide:protein complexes were isolated by affinity purification. While we observed robust binding of MRLC3 to the non‐phosphorylated peptide, we were unable to precipitate MRLC3 with the pThr‐366 peptide, indicating that the interaction between AATF and MRLC3 is sensitive to AATF‐Thr‐366 phosphorylation, likely mediated through MK2 (Figure 3B).
We next asked whether MK2‐mediated AATF phosphorylation is sufficient to disrupt AATF:MRLC3 complexes. To address this question, we used AATF antibodies to precipitate endogenous AATF:MRLC3 complexes from HCT116 cells. Precipitates were incubated with MK2 or mock treated and then resolved on SDS–PAGE followed by immunoblotting to detect MRLC3 that remained bound to AATF throughout the course of the experiment. As depicted in Figure 3C, incubation of endogenous AATF:MRLC3 complexes with active MK2 resulted in a substantial MRLC3 release from bead‐immobilized AATF with essentially no MRLC3 being detectable in the precipitated material following the kinase reaction.
Lastly, we examined the known X‐ray crystal structure of the interaction interface between MRLC and myosin heavy chain (MHC) (Yang et al, 2007). Using sequence alignment and secondary structure predictions, we found that the peptide surrounding Thr‐366 in AATF is likely to assume a helical conformation, like that observed in the MRLC‐binding segment of MHC (Figure 3D; Supplementary Figure 4). Mapping of the MRLC‐binding segment of AATF onto the known MHC interaction grove in MRLC revealed that AATF Thr‐366 is predicted to be located in close proximity to an acidic patch within MRLC (Figure 3D). Phosphorylation of Thr‐366 in AATF is thus expected to disrupt the complex between MRLC3 and AATF.
To analyse the role of MK2 in the destabilization of AATF:MRLC3 complexes, we monitored the spatial dynamics of AATF in wild‐type or MK2/3−/−MEFs (Supplementary Figure 5) (Ronkina et al, 2007). While a prominent accumulation of nuclear AATF in MK2wt/wt MEFs was detectable (Figure 3E), we failed to monitor any nuclear enrichment of AATF in MK2/3−/− MEFs after UV‐C‐induced DNA damage (Figure 3F). These data indicate that MK2‐mediated AATF phosphorylation on Thr‐366 disrupts the interaction between AATF and MRLC3 leading to nuclear translocation of AATF induced by DNA damage.
Chromosomal gains at the AATF locus are associated with reduced survival in human tumours
Alterations in genes comprising the DDR network are frequently observed in human tumours and have been shown to promote the acquisition of further cancer‐promoting mutations or the evasion of therapy‐induced apoptosis (Hanahan and Weinberg, 2000). To test whether the AATF locus might be affected in human tumours, we made use of a cohort of 164 clinically well‐characterized neuroblastoma cases. We note that the cancer genes most frequently altered in adult neoplastic disease, such as TP53, CDKN2A or RAS, are rarely aberrant in neuroblastoma. Specifically, inactivating TP53 mutations are extremely rare in primary neuroblastomas (Maris et al, 2007). All tumour samples were obtained at diagnosis prior to any cytotoxic treatment with a poly‐chemotherapy regimen and each tumour sample was reviewed by a pathologist to ensure a minimal tumour content of 60%. We used two independent probes to detect AATF mRNA expression. As shown in Figure 4A and B, chromosomal gains on the AATF locus, determined by aCGH, were tightly correlated with increased AATF mRNA expression. We next stratified tumours according to AATF expression to ask whether AATF abundance had an impact on patient survival. As shown in Figure 4C–F, increased AATF expression was associated with reduced event‐free and overall survival in these neuroblastoma patients. These data strongly suggest that AATF is affected in human tumours and that increased AATF expression levels are associated with poor prognosis and reduced survival in neuroblastoma patients.
AATF overexpression promotes the stability of DNA damage‐induced cell‐cycle checkpoints
Fanciulli and colleagues recently reported that overexpression of AATF in HCT116 cells resulted in an enhanced stability of a doxorubicin‐induced G2/M checkpoint. This effect was p53‐dependent, since AATF overexpression in p53−/− cells did not result in any detectable difference in checkpoint stability (Bruno et al, 2006). To test whether AATF repression might lead to a destabilization of p53‐dependent DNA damage checkpoints, we subjected p53‐proficient or ‐deficient HCT116 cells expressing AATF‐specific shRNA to nocodazole trap experiments (Supplementary Figure 5B and C). In brief, addition of nocodazole 2 h following application of genotoxic agents will arrest all cells that prematurely escape DNA damage‐induced checkpoints in mitosis. The percentage of mitotically trapped cells that had escaped DNA damage‐induced checkpoints was monitored by phospho‐histone H3 staining 30 h after infliction of DNA damage. As shown in Supplementary Figure 5D, we observed ∼4% of mitotic cells in asynchronously growing p53‐proficient and ‐deficient cells expressing either control or AATF‐specific shRNA. When control shRNA‐expressing p53‐proficient HCT116 cells were treated with UV‐C in the nocodazole trap setting, <1% of these cells had escaped the G2/M checkpoint into mitosis at the end of the time course, indicating a stable UV‐C‐induced checkpoint in these cells. An almost identical checkpoint stability was observed in p53‐deficient HCT116 cells in response to UV‐C. To our surprise, RNAi‐mediated AATF depletion did not affect the stability of UV‐C‐induced checkpoints in p53‐proficient or ‐deficient HCT116 cells. We further failed to detect any significant impact of AATF depletion on the stability of doxorubicin‐ or camptothecin‐induced checkpoints in p53‐proficient or ‐deficient settings. However, when we used retroviral overexpression of AATF cDNA, we were able to detect a small, but reproducible and statistically significant increase in the stability of UV‐, doxorubicin‐ and camptothecin‐induced DNA damage checkpoints in the p53‐proficient cells (Supplementary Figure 5D). These data indicate that AATF repression does not substantially affect p53‐dependent cell‐cycle checkpoints in HCT116 cells after genotoxic stress, while AATF overexpression has a checkpoint‐stabilizing effect in p53‐proficient cells.
AATF negatively regulates p53‐dependent apoptosis
AATF repression has been shown to result in increased apoptosis in response to DNA damage (Di Certo et al, 2007; Bruno et al, 2008). One possible explanation for this observation could be that AATF might be involved in repressing the pro‐apoptotic arm of the p53 response to indirectly promote the induction of a p53‐dependent cell‐cycle arrest following DNA damage. In addition to transactivating cell‐cycle‐regulating genes, such as CDKN1A or GADD45α, p53 has also been shown to transactivate pro‐apoptotic genes, such as PUMA, BAX and BAK (Reinhardt and Schumacher, 2012). Thus, we asked whether AATF affected DNA damage‐induced apoptosis in p53+/+ and p53−/− HCT116 cells expressing either control or AATF‐specific shRNA. Cells were treated with camptothecin, doxorubicin or UV‐C and apoptosis was quantified by FACS 24 h later. The percentage of apoptotic cells following the different treatments did not significantly differ in p53−/− cells that expressed either control or the AATF‐specific shRNA (Figure 5A). In striking contrast, AATF depletion in p53+/+ cells resulted in a dramatically increased sensitivity to all three DNA‐damaging agents (Figure 5A). Camptothecin treatment of control cells resulted in 13.4±2.1% apoptotic cells 24 h later, which rose to 38.5±4% in AATF‐depleted p53+/+ cells (Figure 5A). At 24 h after doxorubicin application, we observed that 16.4±2.2% of control cells were apoptotic, in contrast to 30.5±5.5% of AATF‐depleted cells (Figure 5A). Similarly, UV‐C irradiation of AATF‐depleted p53+/+ cells resulted in 35.7±5.1% of apoptotic cells, compared with only 24.7±3.6% in control cells (experiments in panel A n=16, P<0.05 Student's t‐test). These experiments were then extended to an additional panel of human cancer cell lines containing wt or mutant forms of p53. H23, H2122 (both p53 mutant), SW1573 and H460 (p53 wild type) were transduced with control or AATF‐specific shRNA and exposed to UV, camptothecin or doxorubicin, as above. Similar to the HCT116 cells, these cell lines segregated in their phenotypic response based solely on their p53 status. Loss of AATF resulted in significantly enhanced rates of apoptosis in response to all three damaging agents only in cell lines with intact p53, but not in cells carrying mutant p53 (Supplementary Figure 6). Interestingly, MRLC depletion significantly increased resistance to UV‐, camptothecin‐ and doxorubicin‐induced apoptosis in p53+/+ cells, consistent with the observation that MRLC3 sequesters anti‐apoptotic AATF in the cytoplasm in an inactive state (Supplementary Figures 5C and 6). Thus, loss of AATF appears to enhance the p53‐dependent apoptotic response after DNA damage. To finally demonstrate that MK2‐mediated AATF phosphorylation on Thr‐366 was critical for the release of AATF from MRLC and apoptosis regulation, we repeated the assays in p53+/+ HCT116 cells expressing either wild‐type AATF (AATFWT), a non‐phosphorylatable AATF Threonine‐366 to Alanine mutant (AATFTA) or a phospho‐mimicking AATF Threonine‐366 to Aspartate mutant (AATFTD). As shown in Figure 5B, camptothecin, doxorubicin and UV‐C induced moderate levels of apoptosis in cells overexpressing AATFWT, comparable to those of empty vector‐transduced control cells (compare Figure 5A and B). Overexpression of AATFTA slightly increased the number of apoptotic cells when compared with AATFWT‐transduced cells, consistent with cytoplasmic trapping of anti‐apoptotic AATF, however, this failed to reach statistical significance potentially due to the effects of endogenous AATF (Figure 5B, white and light grey bars). In contrast, overexpression of AATFTD prevented efficient induction of apoptosis in response to camptothecin, doxorubicin and UV‐C. While camptothecin treatment resulted in 21.0±2.6% of apoptotic cells in AATFWT‐overexpressing cells, this number dropped to 13.8±2.4% in cells transduced with the AATFTD mutant. We observed a similar drop in the rates of apoptosis when we compared AATFWT and AATFTD‐expressing cells that were treated with doxorubicin (16.0±2.0% versus 9.4±2.9%, respectively) or UV‐C (26.7±2.1% versus 13.8±2.4%, respectively (Figure 5B), experiments in Figure 5A and B n=16, P<0.05, Student's t‐test. As might be expected, we failed to observe such a resistance‐mediating effect of AATFTD overexpression in p53−/− cells (Supplementary Figure 7). These data are consistent with a role for AATF in antagonizing p53‐driven apoptosis following DNA damage in a manner that depends on MK2‐mediated AATF phosphorylation and nuclear translocation in response to genotoxic stress.
AATF regulates the functional outcome of the p53 response
To elucidate how AATF is involved in determining the outcome of p53 activation, we next monitored the DNA damage‐mediated induction of a panel of well‐established p53 target genes. We used immunoblotting to analyse the expression of the proapoptotic genes PUMA, BAX and BAK, as well as the growth‐arresting genes CDKN1A, GADD45α and RPRM in p53‐proficient HCT116 cells expressing control or AATF‐specific shRNAs. AATF depletion resulted in a substantially enhanced expression of the proapoptotic genes PUMA, BAX and BAK in response to UV‐C, while no difference in the expression of the cell‐cycle‐regulatory genes CDKN1A and GADD45α could be observed (Figure 5C). Only a miniscule, but consistent, induction of RPRM was seen in AATF‐depleted cells upon UV‐C (Figure 5C). We next performed chromatin immunoprecipitation (ChIP) to investigate whether AATF selectively binds to the promoter regions of the p53 target genes whose gene products it appears to regulate following genotoxic stress. We designed a set of primers to amplify the genomic DNA of the promoter regions covering known p53‐binding sites (indicated by the cartoons accompanying Figure 5D–I). As shown in Figure 5D, <8% of the PUMA promoter DNA subjected to the ChIP experiment could be recovered with AATF‐specific antibodies in the absence of genotoxic stress. In contrast, UV irradiation prior to immunoprecipitation allowed the recovery of 49.2±8.9% of the input DNA, indicating enhanced AATF binding to the PUMA promoter in response to genotoxic stress. Deletion of MK2/3 almost completely prevented this interaction of AATF with the PUMA promoter region, indicating a requirement for MK2 to promote AATF binding to the PUMA promoter. Furthermore, we could document a similar MK2/3‐dependent interaction of AATF with the promoters of the proapoptotic p53 target genes BAX and BAK in response to UV‐C (Figure 5E and F). In contrast, while there was a small and reproducible, yet statistically not significant, increase in the AATF occupancy on the CDKN1A promoter, similar to what has been described by others (Bruno et al, 2006), we failed to detect an increase in AATF occupancy on the GADD45α and RPRM promoters in response to DNA damage despite extensive efforts (Figure 5G–I). Figure 5G–I n=9, P<0.05, Student's t‐test. These data strongly suggest that AATF dampens the p53‐dependent expression of a set of selective apoptosis‐inducing genes likely by exerting transcriptional repression on the promoter regions, while having little if any effect on a set of canonical p53‐regulated genes involved in mediating cell‐cycle arrest upon genotoxic stress.
AATF promotes doxorubicin resistance in vivo
To investigate whether the MK2‐dependent apoptosis‐antagonizing role of AATF could also be observed in vivo, we performed xenograft experiments. 106 HCT116 cells stably expressing either control, or AATF‐specific shRNAs were injected into the flanks of NCRnu/nu mice and doxorubicin treatment was initiated upon tumour formation. Control tumours showed only a mild volume reduction from 1430 (±331) to 967 (±258) mm3, while AATF‐depleted tumours showed a remarkable doxorubicin response with tumour volumes decreasing from 1295(±422) to 408(±125) mm3. The differences in mean tumour volumes between control and AATF shRNA‐expressing tumours reached statistical significance after only 9 days of treatment (Figure 6A). Using immunohistochemistry, we could show that Puma was specifically induced in doxorubicin‐treated AATF‐depleted tumours, while little Puma staining was seen in the controls (Figure 6B). These data lend further support to an apoptosis‐antagonizing role of AATF through repression of p53‐mediated Puma transcription after DNA damage.
To further probe the importance of MK2‐mediated AATF phosphorylation on Thr‐366, we repeated the xenograft assays with HCT116 cells that were transduced to overexpress AATFWT, AATFTA or AATFTD. Doxorubicin was initiated when AATFWT‐overexpressing tumours were 1450(±267) mm3 and AATFTA‐overexpressing tumours were 1577(±141) mm3 in volume. AATFWT‐ and AATFTA‐overexpressing tumours subsequently regressed to 212(±98) and 305(±103) mm3 in volume, respectively. In contrast, tumours expressing the phospho‐mimicking AATFTD mutant displayed remarkable doxorubicin resistance. AATFTD‐expressing tumours measured 1250(±250) mm3 in volume at the initial dose of doxorubicin but showed only a mild volume reduction to 907(±170) mm3 over the treatment cycle (Figure 6C). These data provide in vivo evidence that Thr‐366 in AATF serves as a molecular switch that, when phosphorylated, promotes resistance against genotoxic stress, likely through the selective abrogation of transactivation of the proapoptotic p53 target genes PUMA, BAX and BAK.
Based on these observations, one might predict that p53‐proficient human tumours display an enrichment of nuclear AATF, as this would be critical to counteract p53‐driven apoptosis. In contrast, p53‐defective tumours might not show such an AATF accumulation, as nuclear AATF localization is likely not to be selected for in the absence of functional p53. To directly test this hypothesis in a relevant human tumour entity, we analysed a panel of 79 individual human endometrial cancer specimens. We used a well‐established immunohistochemistry‐based algorithm to stratify these tumours based on their p53 status. Staining for p53 was regarded as aberrantly enhanced when >70% of tumour cells showed strong staining. Since the vast majority of p53 mutations stabilize the mutant protein, gross overexpression in cancer cells is suitable to identify the vast majority of the tumour‐associated p53 aberrations, with only rare false‐negative and probably no false‐positive cases (given the internal control of normal protein levels in noncancerous cells on each section) (Jiang et al, 2009). Tumour specimens were further stained with antibodies detecting AATF and MK2 (Figure 7A). As shown in Figure 7B, tumours expressing wild‐type levels of p53 showed a robust enrichment of nuclear AATF, with the majority of samples displaying grade II/III nuclear AATF staining. This enrichment of nuclear AATF staining was significantly underrepresented in tumours lacking wild‐type p53 (Fisher's exact test, P< 0.05), strongly suggesting that nuclear localization of AATF is selected for in p53‐proficient tumours, potentially to repress p53‐driven apoptosis. Intriguingly, the activation and translocation of MK2 could also be observed in most human tumour samples, which also displayed nuclear AATF staining, further substantiating the hypothesis that AATF translocates into the nucleus in an MK2‐dependent manner (Figure 7A).
We next asked whether analysis of human tumour cells might lend support to our hypothesis that AATF is a p53 co‐factor that selectively blunts the pro‐apoptotic arm of the p53 response, using a panel of lung cancer cell lines, which were stratified based on p53 status and AATF copy number (Figure 7C). When we analysed cisplatin‐induced apoptosis in these cell lines, we found that focal copy number gains at the AATF locus were associated with a significant protection from cisplatin‐induced apoptosis in p53‐proficient cells (Figure 7D), similar to what we had observed in the neuroblastoma samples described in Figure 4. Thus, AATF amplification emerges as a potential mechanism that is employed by human cancer cells to antagonize the detrimental effects of p53‐driven apoptosis.
Here, we identified AATF as a key repressor of the pro‐apoptotic arm of the p53 response. Moreover, our data provided mechanistic details as well as clinical evidence implicating AATF as a potential drug target for cancer therapy. An important role for AATF in the DDR signalling networks had been suggested before (Bruno et al, 2002, 2006). Previous studies demonstrated that phosphorylation of AATF by the canonical checkpoint kinases ATM/ATR and Chk2 was important for AATF protein stabilization and involved in the control of cell‐cycle progression (Bruno et al, 2006). However, much of this understanding of the role of AATF in DDR resulted from genetic gain and loss of function experiments (Bruno et al, 2002). We now report data corroborating the initial observations, showing an enhanced stability of DNA damage‐induced cell‐cycle checkpoints upon overexpression of exogenous AATF. The effects of AATF overexpression depend at least partially on the activation status of the DDR network with AATF promoting cell‐cycle progression in the absence of DNA damage and enforcing a cell‐cycle arrest in response to genotoxic stress. These contrasting effects of AATF signalling correlated with distinct promoter binding of AATF in the different scenarios (Bruno et al, 2002, 2006). Expanding on these initial observations with variable control of cell cycle, we now show that AATF serves as a master regulator of the p53‐dependent apoptotic programme. Consistent with a clinically important role of apoptosis regulation, AATF expression and localization was indeed regulated in human tumours and gene amplification correlated with worse clinical outcome and reduced overall survival following genotoxic poly‐chemotherapy. Intriguingly, chemotherapy‐naive neuroblastomas are known to express wild‐type copies of TP53 in the vast majority of cases. Thus, these data corroborate the in vitro experiments performed by Fanciulli and colleagues showing that AATF overexpression protected p53‐proficient HCT116 cells from chemotherapy‐induced apoptosis (Bruno et al, 2006).
Recent studies demonstrated that AATF overexpression results in the induction of antiapoptotic target genes even in the absence of genotoxic stress. Furthermore, AATF depletion prevented the doxorubicin‐induced expression of XIAP, one of the identified target genes (Bruno et al, 2008). However, the mechanisms regulating AATF‐mediated transcriptional control remained largely unclear. We now provide a molecular rationale for the observations previously reported. We demonstrated that access of endogenous AATF to the genomic DNA is at least in part regulated through cytoplasmic sequestration of substantial pools of functionally inactive AATF in resting cells through a direct interaction with the cytoskeletal protein MRLC3. MK2‐mediated AATF phosphorylation on Thr‐366 acts as a molecular switch to disrupt cytoplasmic AATF:MRLC complexes. In response to different genotoxic agents, liberated AATF translocates to the nucleus, where it engages the promoters of the pro‐apoptotic p53 target genes PUMA, BAX and BAK, repressing their expression. In addition, there was a consistent increase in the binding of AATF to the CDKN1A promoter as demonstrated previously (Bruno et al, 2006). However, in contrast to previous reports, we failed to demonstrate an appreciable increase in cell‐cycle regulating p21 protein expression upon AATF knockdown. This might be the result of different doses of genotoxic agents or different time points of analysis.
AATF was previously shown to be stabilized through ATM/ATR‐ and Chk2‐mediated phosphorylation (Bruno et al, 2006). We now report that AATF is also a substrate of the p38/MK2 stress kinase pathway. The p38/MK2 module has been shown to be activated after DNA damage (Kyriakis and Avruch, 2001) and may also involve the canonical DDR kinases ATM and ATR (Raman et al, 2007; Reinhardt et al, 2007). However, in response to UV irradiation, the p38/MK2 pathway is activated independently of ATM and ATR (Reinhardt et al, 2007). These observations suggest that there might be a certain stress specificity to MK2‐mediated AATF phosphorylation with maximum AATF effects only when ATM/ATR, Chk2 and MK2 are simultaneously activated. Hyperosmotic stress as a p38/MK2‐specific stimulus produced a robust nuclear translocation of AATF, indicating that p38/MK2 activity is sufficient for AATF re‐localization. However, further experiments are required to test whether mere nuclear accumulation of AATF is sufficient for its apoptosis‐repressing function.
The data presented in this study provide a molecular mechanism that helps explain the regulation of the p53 response through dynamic spatial control of the transcriptional co‐factor AATF. The proposed mechanism of AATF‐mediated control of p53 signalling (Figure 8) is in line with a model in which the transcriptional activity of p53 is regulated by co‐factors engaging a subset of p53‐targeted promoters to selectively regulate the transcriptional activity of these promoters, thus shaping the functional outcome of p53 activation. We could show that p53 and AATF engage the PUMA, BAX, BAK, CDKN1A, GADD45α and RPRM promoters independent of each other. Thus, a model is emerging in which DNA damage‐mediated nuclear AATF accumulation on the promoters of the pro‐apoptotic genes PUMA, BAX and BAK represses the transactivation of these genes, while not restricting the expression of the cell‐cycle‐regulating genes CDKN1A, GADD45α and RPRM. The net result is a selective repression of p53‐driven apoptosis in response to genotoxic stress. These observations are in good agreement with a previously described role for the p38/MK2 pathway in mediating a prolonged cell‐cycle checkpoint arrest following DNA damage (Bulavin et al, 2001; Raman et al, 2007; Reinhardt et al, 2010).
However, p53 serves a role as a double‐edged sword: Undoubtedly, p53 has enormous potential as a drug target, since p53‐mediated transcription of pro‐apoptotic genes, has been shown to effectively induce apoptosis (Miyashita et al, 1994; Selvakumaran et al, 1994; Zhan et al, 1994; Oda et al, 2000a; Nakano and Vousden, 2001). However, p53‐dependent induction of CDKN1A, 14‐3‐3σ and other cell‐cycle‐regulating genes has a protective effect, through prevention of mitotic catastrophe due to premature checkpoint collapse in tumour cells experiencing chemotherapy‐induced genotoxic stress (Asada et al, 1999; Abbas and Dutta, 2009; Jiang et al, 2009; Koster et al, 2010). Thus, it is desirable to identify pathways regulating the functional outcome of the p53 response. Here, we identify AATF as a critical modulator selectively repressing p53‐driven apoptosis in a phospho‐dependent manner by preventing the induction of the pro‐apoptotic p53 target genes, PUMA, BAX and BAK while allowing unrestricted CDKN1A, GADD45α and RPRM expression. Thus, the p38/MK2/AATF signalling module emerges as a powerful apoptosis‐antagonizing pathway specifically intercepting p53‐driven apoptosis (Figure 8). Our data presented here provide a rationale for the development and clinical use of inhibitors of the p38/MK2/AATF pathway to harness the full apoptotic potential of p53, which is non‐mutated in ∼50% of all human tumours (Jiang et al, 2009).
Materials and methods
Cell culture and chemicals
HEK293T cells were cultured in DMEM supplemented with 10% FBS and transfected with the calcium phosphate method, p53+/+ and p53−/− HCT116 cells were cultured in McCoy's 5A plus 10% FBS and transfected with Genejuice (Novagen). WT and MK2/3 DKO MEFs, SW1573, NCI‐H460, NCI‐H23, NCI‐H2122, U2OS cells were grown in DMEM plus 10% FBS. Okadaic acid, camptothecin and doxorubicin were purchased from Sigma. UV‐C was applied using a Stratalinker. Hypertonic stress was applied through prewarmed Ringer solution containing 582 mosm/l sucrose 10 min prior to harvesting.
Antibodies and RNAi
Antibodies against FLAG (Sigma), V5 (Chemicon), cl. caspase‐3 (BD Biosciences), pHH3 (Upstate), AATF (Sigma, Abnova and homemade), fibrillarin (Abcam), β‐tubulin (Dev. St. Hybridoma Bank), Chk1, pSer‐345 Chk1, MK2, pThr‐334 MK2, Puma, Bax, p53, Gadd45α (Cell Signalling), MRLC (Santa Cruz and homemade), p21 (Santa Cruz), p53 (Dako) were used for immunoprecipitation, immunoblotting, immunofluorescence immunohistochemistry and FACS. AATF and luciferase shRNA sequences were previously described (Bruno et al, 2006; Reinhardt et al, 2007). MRLC shRNA targeting sequence was 5′‐TAA CTT CAA ATT CCA GCC AAA‐3′.
Conflict of Interest
The authors declare that they have no conflict of interest.
Supplementary Figures, Legends and Methods Combined (by MS)
Source Data Fig. 1
Source Data Fig. 2
Source Data Fig. 3
Source Data Fig. 5
This work was supported by the Deutsche Forschungsgemeinschaft (BE2212, SFB‐829 to TB, RE2246/1‐1, RE2246/2‐1, SFB‐829, SFB‐832 to HCR, SFB‐566 to MG, SFB‐832 to RKT and to BS, TH‐1386/3‐1 to RKT and MLS), the Volkswagenstifung (Lichtenberg Programme to HCR), the National Institutes of Health (GM68762, CA112967, ES015339 to MBY), the Deutsche Krebshilfe (107954 to RKT), the Ministry of Science, NRW (MIWT, 4000‐12 09 to RKT and 313‐005‐0910‐0102 to HCR), the BMBF (0313921 to BS and TB, 01GS08100 to RKT), The Max Planck Society (M.I.F.A.NEUR8061 to RKT), EU Framework Programme CURELUNG (HEALTH F2 2010 258677 to RKT), Stand Up To Cancer American Association for Cancer Research Innovative Research Grant (SU2C AACR IR60109 to RKT), by the Behrens‐Weise Foundation (to RKT) and by an anonymous foundation to RKT and the Köln Fortune Programme (KH, SC). MLS is a fellow of the International Association For the Study of Lung Cancer (IASLC). HCT116 cells were from B Vogelstein. AATF shRNA constructs were from M Fanciulli. We thank H Zentgraf for support in generating monoclonal antibodies. We thank the members of our laboratories for helpful discussions.
- Copyright © 2012 European Molecular Biology Organization