Factor inhibiting HIF-1 (FIH-1) modulates protein interactions of apoptosis-stimulating p53 binding protein 2 (ASPP2)

Factor-inhibiting HIF-1 (FIH-1) belongs to a family of 2-oxoglutarate- and iron(II)-dependent dioxygenases. Collectively, these enzymes have a wide range of potential substrates including proline, asparagine, and lysine residues in proteins, and also DNA because some family members act as DNA demethylases. The protein dioxygenase FIH-1 is an oxygen-dependent asparaginyl hydroxylase, which was initially reported to regulate the activity of the transcription factor complex hypoxia-inducible factor (HIF) (Lando et al., 2002a). Asparaginyl hydroxylation of HIF-α by FIH-1 inhibits the interaction of the C-terminal transactivation domain (C-TAD) of HIF-α with the transcriptional co-activator CBP/p300 (Schofield and Ratcliffe, 2004). Therefore, FIH-1-dependent hydroxylation provides an oxygen-dependent mechanism in the control of HIF activity that is independent of HIF-α stability regulation by HIF prolyl hydroxylases (PHDs) (Kaelin and Ratcliffe, 2008; Semenza, 2012). Remarkably, oxygen-dependent inhibition of the C-TAD was reported some time ago (Pugh et al., 1997) but the molecular mechanism, i.e. the function of FIH-1, was only identified later (Hewitson et al., 2002; Lando et al., 2002b; Mahon et al., 2001). Although FIH-1 and the PHDs belong to the same superfamily of dioxygenases and in general require the same set of cofactors, i.e. oxygen, 2-oxoglutarate and ferrous iron, there are striking differences between these enzymes. Firstly, FIH-1 has a higher affinity for molecular oxygen relative to the PHDs, which suggests that in mild hypoxia FIH-1 remains active while the PHDs are inactive (Koivunen et al., 2004). This is potentially important because not all HIF target genes are equally dependent on the function of the C-TAD. Secondly, the PHDs are apparently more substrate specific because very few alternative PHD substrates have been reported, so far (Cummins et al., 2006; Köditz et al., 2007). In contrast to the PHDs, FIH-1 was shown to hydroxylate substrate proteins in a highly conserved protein interaction motif termed the ankyrin repeat domain (ARD). Certainly not all ARDs are targeted by FIH-1 but asparaginyl hydroxylation has been detected in several ARD proteins, although on some occasions oligomeric peptides have been used as in vitro substrates (Cockman et al., 2006; Cockman et al., 2009; Zheng et al., 2008). Among the ARD-containing substrates of FIH-1 are cellular signalling factors such as IκBα, Notch1 and p105 (Cockman et al., 2006; Coleman et al., 2007), which are crucial components of cellular signalling pathways. However, the function of ARD hydroxylation by FIH-1 remains elusive. The large cellular pool of ARD-containing proteins could indicate that ARD hydroxylation is more widespread than anticipated previously and that a number of additional FIH-1 substrates await identification. Of note, genetic deletion of FIH-1 in mice does not lead to inadequate activation of HIF but rather induces dysregulation of energy metabolism (Zhang et al., 2010).

All members of the family of apoptosis-stimulating p53-binding proteins (ASPPs) also possess an ARD with a potential hydroxylation site for FIH-1. The ASPP protein family consists of three members, ASPP1, ASPP2 and inhibitory ASPP (iASPP) (Trigiante and Lu, 2006). ASPPs were initially identified as modulators of p53 transcriptional activity on pro-apoptotic DNA promoters. ASPP1 and ASPP2 were shown to stimulate p53-dependent apoptosis through induction of pro-apoptotic target genes, whereas iASPP has the opposite effect and suppresses p53-mediated apoptosis (Bergamaschi et al., 2003; Samuels-Lev et al., 2001). Recently, it could be shown that besides regulation of p53 activity ASPPs are also involved in the regulation of other cellular signalling pathways. iASPP was found to act as a key player in the regulation of epithelial stratification through modulation of the transcriptional activity of p63, another p53 family member known as a master transcription factor controlling this process (Chikh et al., 2011; Notari et al., 2011). ASPP1 and ASPP2, which have redundant functions in the control of p53 transcriptional activity on pro-apoptotic promoters, were recently shown to be involved in additional cellular signalling pathways, demonstrating unique functions of the proteins. For example, ASPP1 participates in the regulation of the hippo signalling pathway, which can control cell proliferation and apoptosis independent of the tumour suppressor p53 (Aylon et al., 2010; Vigneron et al., 2010). ASPP2 was shown to be involved in cell polarity signalling and the formation of tight junctions via interaction with the tight junction scaffolding protein partitioning-defective homologue 3 (Par-3). This interaction of ASPP2 and Par-3 has been shown to be particularly important during CNS development and controls the polarity and proliferation of neuronal progenitor cells (Cong et al., 2010; Sottocornola et al., 2010).

Overall, the recent findings on ASPP proteins demonstrate their importance for different cellular signalling pathways that influence tissue homeostasis and developmental processes. Because hypoxic conditions are often encountered by organisms during development as well as in pathological processes such as tumour formation and progression, it is important to determine whether ASPP proteins are affected by FIH-1 and thus by insufficient oxygen supply. A link between ASPP-dependent signalling pathways and the cellular oxygen-sensing machinery provides a molecular mechanism for hypoxia-dependent but HIF-independent regulation of physiological processes. Accordingly, we investigated initially whether FIH-1-dependent ASPP hydroxylation impinges on the interaction with p53 and thus affects hypoxia-induced apoptosis initiation. The relationship between hypoxia and apoptosis has been analysed in depth elsewhere (Acker and Acker, 2004; Zhou et al., 2006). However, we did not observe any effects of FIH-1 on the interaction with p53 or apoptosis. Instead, we demonstrate that FIH-1 affects the interaction of ASPP2 with Par-3 and subcellular localization of ASPP2.

Previous reports in which the substrate specificity of FIH-1 has been investigated suggest that hydroxylation of ARD-containing proteins occurs frequently although not every ARD can serve as a substrate for the enzyme (Cockman et al., 2009; Coleman and Ratcliffe, 2009). The sequence alignment of ARD-containing proteins with the FIH-1 substrates identified so far revealed potential hydroxylation sites in all three members of the ASPP protein family. The primary sequence of these domains is mostly in line with the ‘LXXXGADVNA’ consensus hydroxylation motif of FIH-1 (Fig. 1A). Very recently, it was shown that the substrate requirements of FIH-1 are not very stringent, even the hydroxylated asparagine residue can be exchanged for an aspartate or histidine residue (Fig. 1B) if appropriately positioned in the β-loop between two ankyrin repeats (Yang et al., 2011a; Yang et al., 2011b). Considering alignments of FIH-1 substrates identified so far (Fig. 1A,B), none of the amino acids in the consensus motif LXXXGADVNA is 100% conserved.

To test whether FIH-1-dependent hydroxylation of ASPP family members occurs, ARD-containing substrate protein domains were selected and generated in bacteria, or synthetic peptides were custom synthesized and used in hydroxylation assays (Fig. 1C,D). To test one pro-apoptotic ASPP family member, ASPP2 was used in the hydroxylation assays. iASPP also contains an ARD, but the amino acid sequence in the vicinity of the potential hydroxylation site of iASPP is significantly different to that of ASPP2 (</emph>Fig. 1D), whereas ASPP1 and ASPP2 show high sequence homology.

To identify novel substrates we incubated purified recombinant FIH-1 with the potential substrates. Using an N-terminal His-tagged ASPP2 protein fragment (His–ASPP2) we detected significant hydroxylation by measuring FIH-1-dependent decarboxylation of the co-substrate 2-oxoglutarate. His–HIF-1α C-TAD served as a positive control. 2-oxoglutarate-dependent dioxygenases can produce CO₂ in a substrate-independent manner. Therefore it was important to demonstrate that enzyme turnover was significantly higher compared with HIF-1αN803A, which served as a negative control (Linke et al., 2004). As previously reported for other ARD-containing proteins (Cockman et al., 2009) hydroxylation of His–ASPP2 by FIH-1 was less efficient than hydroxylation of HIF-1α (Fig. 2A). We identified the hydroxylated amino acid by mutation analysis. Mutation of N986 to alanine indeed reduced, but did not completely prevent, enzyme turnover (Fig. 2B). In general, mutation of alanine and threonine had a very similar effect. Using 20mer peptides of ASPP2 containing (wt) or lacking (N986A) the potential hydroxylation site we confirmed that FIH-1 activity was significantly reduced upon mutation of N986 to alanine (Fig. 2C). Using these peptides we further characterised the FIH-1-dependent hydroxylation of ASPP2. First, we were able to demonstrate enhanced FIH-1 activity with increasing concentrations of wild-type, but not mutant, peptide (Fig. 2C). The K_m value for 75 µM wild-type ASPP2 peptide was comparable with those for other ARD-containing FIH-1 substrates identified so far (Wilkins et al., 2009) (Fig. 2D). We could specifically inhibit enzyme turnover for wild-type but not for mutant ASPP2 peptides using the 2-oxoglutarate competitor N-oxalylglycine, which blocked hydroxylation in a concentration-dependent manner (Fig. 2E). A comparison of the ASPP2 substrate peptides with iASPP containing (wt) or lacking (N687A) the potential hydroxylation site revealed that iASPP is not hydroxylated by FIH-1 in vitro. Even increasing concentrations of iASPP peptide up to 300 µM did not lead to significant FIH-1 activity (Fig. 2F).

As a direct proof of ASPP2 hydroxylation in vivo, we analysed full-length overexpressed ASPP2 purified from HEK293T cells, using LC-MS/MS. Using chymotrypsin as a protease, we could reproducibly detect a peptide consisting of the potential hydroxylation site at N986 (GVNVNAADSDGWTPL, [M+2H]²⁺ = 758.3573). In addition to the signal at m/z 758.36, a signal with a mass difference of +16 ([M+2H]²⁺ = 766.3546) was detected at a slightly earlier retention time, corresponding to the hydroxylated peptide. Fig. 3A shows the signal intensity in the MS survey scan for both peptides at the respective peak in the elution profile. In wild-type HEK293T cells, both the hydroxylated and the unmodified peptide were detected, with a higher intensity for the hydroxylated peptide. In order to show that the modification is indeed introduced by FIH-1, we analysed ASPP2 from cells overexpressing FIH-1 and could show a significant increase in the relative amount of hydroxylation, with the unmodified peptide being barely detectable. In contrast, after treatment with the hydroxylase inhibitor dimethyloxalylglycine (DMOG) the unmodified variant was detected almost exclusively (Fig. 3A,B). The identity of these peptides was verified by MS/MS analysis (Fig. 3C), resulting in positive identification by the Mascot search engine with a false discovery rate of 1% and a mascot score above 40. The b-ions defining N986 as the site of hydroxylation are clearly visible. Taken together, these results show that ASPP2 is hydroxylated in vivo at N986 and that the amount of hydroxylation is dependent on FIH-1 activity.

Interaction between enzyme and substrates was tested by co-precipitation of in vitro transcribed and translated, radioactively labelled FIH-1 following pulldown of glutathione S-transferase (GST)-tagged test protein fragments incubated with the enzyme. Pulldown of GST–ASPP2 led to the highest amount of FIH-1 co-precipitated. Mutation of the hydroxylation site to alanine or threonine reduced binding of the enzyme to the substrate. The same effect could be observed for GST–HIF-1α upon mutation of the hydroxylation site N803. Overall, pulldown of GST–HIF-1α co-precipitated less FIH-1 than GST–ASPP2 did, indicating stronger interaction of ASPP2 with the enzyme (Fig. 4).

To assess the physiological effects of FIH-1-dependent hydroxylation of ASPP2 we generated stable FIH-1 knockdown cell lines by lentiviral transduction with shRNA (termed shFIH-1 cells). On the protein level we achieved a knockdown efficiency of constantly more than 80% (Fig. 5); this efficiency was stable for at least 3 weeks. Control cell lines were transduced with the empty vector pLKO.1 or unspecific shRNA (shscrambled). Using these cell lines we examined whether ASPP2 protein levels are altered upon suppression of FIH-1 through lentiviral transduction, hypoxic or anoxic conditions. In HCT116 colon carcinoma (Fig. 5) as well as in H441 lung carcinoma cells we could not detect any significant changes of ASPP2 protein levels under suppression of FIH-1, indicating that enzymatic hydroxylation of ASPP2 by FIH-1 has no regulatory effects on the total protein amount.

To further examine the physiological consequences of FIH-1-dependent hydroxylation of ASPP2 we tested whether proliferation or apoptosis of shFIH-1 cells are altered relative to control cells. Regarding proliferation we could not detect any reproducible changes upon suppression of FIH-1 by means of an MTT test (Fig. 6A) or cell counting (data not shown) in HCT116 and H441 cells. For apoptosis induction we used prolonged stimulation of the cells with the chemotherapeutic drugs etoposide and doxorubicin as well as ionizing radiation. We could not detect any FIH-1-dependent difference in the apoptotic response of HCT116 cells to chemotherapy or irradiation (Fig. 6B). To extend the phenotypic analysis we also compared adhesion and migration of the cell lines. We observed a minor increase in the number of adherent cells after FIH-1 suppression, which could also be detected upon knockdown of ASPP2 (Fig. 6C). Migration of FIH-1-depleted cells was slightly decreased, which cannot be attributed to ASPP2 modification because ASPP2 knockdown did not mimic the effect (Fig. 6D). Taken together these results suggest that hydroxylation of ASPP2 by FIH-1 did not have gross effects on apoptosis induction or migration of the cell cultures.

In HCT116 cells we tested whether protein interactions identified so far for ASPP2 are affected by lentiviral suppression of FIH-1. Following reciprocal co-immunoprecipitation of ASPP2 and Par-3 from whole cell lysates we could demonstrate that the interaction of these proteins was impaired upon FIH-1 knockdown, leading to less co-precipitation of the interaction partner (Fig. 7A,B) whereas total protein levels were not affected in shFIH-1 cells. As the localisation of ASPP2 and Par-3 to cell–cell contacts has previously been shown to be interdependent (Cong et al., 2010; Sottocornola et al., 2010), we tested whether the impaired interaction upon suppression of FIH-1 leads to changes in the localisation of ASPP2. Indeed, indirect immunofluorescence showed that the protein was partly redistributed into the cytoplasm in HCT116 shFIH-1 cells, whereas control cells showed an ASPP2 localisation pattern that was clearly focussed to the cell–cell contacts (Fig. 7C). Because the interaction of the proteins was not completely disrupted upon FIH-1 knockdown, a fraction of ASPP2 still localised to cell–cell contacts. H441 cells, which have a more epithelial phenotype than HCT116 cells, also showed an altered localisation of ASPP2 upon suppression of FIH-1. In this case the protein was only partly redistributed into the cytoplasm and the localisation pattern of the ASPP2 fraction remaining at the cell boundaries was clearly disorganised, compared to control cells (Fig. 7D). In both cell lines, we found the effects on the localisation of ASPP2 to be dependent on the hydroxylase activity of the enzyme, because the inhibition of FIH-1 activity by treatment with the hydroxylase inhibitor DMOG, also reduced the cell-contact localisation of the substrate (Fig. 7C,D). The effect of FIH-1 knockdown was not detectable in CaCo2 and DLD1 cells, which demonstrates that ASPP2 localization is affected by a cell-type-specific component. Additionally, we examined whether the redistribution of ASPP2 we observed in a p53 wild-type background affects the interaction with the tumour suppressor, but we did not detect any differences in the amount of ASPP2 co-precipitating with p53 in HCT116 cells (Fig. 7E). Therefore, we propose a model in which FIH-1-dependent hydroxylation of ASPP2 triggers the interaction with Par-3 while p53-dependent signalling is not affected by the protein modification (Fig. 8).

Hydroxylation is increasingly recognized as a key post-translational modification with profound effects on the interaction profile of proteins. Well-documented examples of this process are collagen where prolyl and lysyl hydroxylation stabilize the peptide triple helix, and the α-subunits of the transcription factor HIF. In the case of HIF, prolyl hydroxylation induces interaction with von-Hippel-Lindau protein and proteasomal degradation while asparaginyl hydroxylation prevents interaction with the transcriptional co-activator p300/CBP. The latter process was the first reaction recognized to be dependent on FIH-1. Subsequently, it was demonstrated that a subgroup of ARD-containing proteins also undergo hydroxylation by FIH-1 although a clear cellular function or an effect on protein interaction of substrate proteins has not been demonstrated so far. In the current report we provide evidence showing that FIH-1-dependent hydroxylation of the ARD of ASPP2 leads to significant changes in the interaction profile and to an alteration of subcellular localization. Regarding the effects of ARD hydroxylation, our study indicates a new role for FIH-1 in the regulation of protein interaction, and, by extrapolation, that the interaction profiles of other ARD-containing substrates could also be altered upon modification by the enzyme. The study was based on the observation that ASPP2 shows significant sequence homology when aligned with known FIH-1 substrates, and the demonstration that ASPP2 is indeed a substrate of FIH-1. We detected a single hydroxylation site for FIH-1 in the ARD of ASPP2 N986. Importantly, in vivo hydroxylation of ASPP2 was demonstrated, by mass spectrometry, to be dependent on FIH-1 activity. In contrast, the conserved protein family member iASPP was not modified by FIH-1 at the corresponding asparagine residue, demonstrating the specificity of the hydroxylation reaction despite the frequent occurrence of ARD hydroxylation by FIH-1.

Recent studies also provided evidence for a distinct mechanism of substrate recognition that differs between ARD-containing substrates and HIF-1α. It was shown previously that in the case of mouse Notch1, increased substrate length dramatically improves the binding affinity of FIH-1, whereas this effect was much less prominent for HIF-1α substrates. Overall, the binding affinity of Notch substrates for FIH-1 was higher than that of HIF-1α (Wilkins et al., 2009). We observed that the situation in our interaction assays was similar, with stronger binding of FIH-1 to ASPP2 protein fragments than to HIF-1α. In contrast to the HIF-1α C-TAD, which was reported to be present in an unstructured state in solution and to become structured upon binding of the enzyme (Dames et al., 2002), the ARD pre-exists in a structured state in solution (Coleman et al., 2007; Kohl et al., 2003). Nevertheless, the interaction interface of an ankyrin repeat in complex with FIH-1 was shown to be similar to that of HIF-1α and the enzyme with a largely extended conformation of the substrates (Coleman et al., 2007; Elkins et al., 2003). Therefore, ARD-containing proteins have to undergo a conformational change to enable accessibility of the hydroxylation site and adopt the extended conformation necessary for the reaction. This discrepancy in substrate structure could give rise to the different hydroxylation efficiencies we observed for the HIF-1α and ASPP2 protein fragments. Hydroxylation of the HIF-1α protein fragment was one order of magnitude more efficient than that of ASPP2 which could be due to the necessity of unfolding to enable hydroxylation. Moreover, hydroxylation of ASPP2 was generally more efficient if short peptides of 20 amino acids (aa) were used for the hydroxylase activity assay instead of protein fragments of 240 aa, again arguing for a role of secondary and tertiary structure in limiting reaction efficiency.

However, as the stability of the folded ARD is also determined by its primary amino acid sequence, the residues surrounding the target asparagine may also determine hydroxylation efficiency. In this context, Wilkins et al. very recently observed that the sequence of the hydroxylation site contributes to distinct characteristics of different substrates. They reported that amino acids proximal to the target asparagine influence the kinetics of ARD-hydroxylation by FIH-1 and propose that single amino acid exchanges in this region can prevent maintenance of the conformation necessary for hydroxylation in the active site of the enzyme (Wilkins et al., 2012). This is expected to apply to our situation as well where ASPP2 and iASPP differ in the primary amino acid sequence at the −3, +1 and +2 position relative to the target asparagines, which presumably renders ASPP2 a substrate of FIH-1 while iASPP is not hydroxylated. Despite the promiscuous character of FIH-1 the substrate requirements are thus proved to be distinct also in the case of ARD-containing proteins. The discovery of two highly conserved proteins, one of which can serve as substrate while the other cannot, further contributes to the definition of a consensus FIH-1 hydroxylation site and facilitates the prediction of additional new substrates.

To date, the physiological consequences of ARD-containing protein hydroxylation by FIH could not be revealed for any of the substrates identified. In our study we compared the phenotype of cells expressing FIH-1 to that of cells where the expression was reduced by lentiviral transduction of shRNA directed against the enzyme. Using this approach we could not detect differences in proliferation of the cancer cell lines, although we observed an induction of p53 and p21, which was previously described by Pelletier et al. to be accompanied by reduced proliferation in different wild-type p53 cell lines (Pelletier et al., 2012). Moreover, we were not able to detect differences in the interaction of ASPP2 and p53 or an effect on adhesion or migration of the cells in the presence or absence of FIH-1. This is in line with our observation and the results of Pelletier et al. concerning the mortality of the cells, which was not altered upon FIH-1 suppression either under control conditions or under the influence of apoptotic stimuli. This seems in contrast to a recent publication showing that FIH-1 suppression can increase renal cancer cell apoptosis thereby triggering survival of these cells (Khan et al., 2011). Another group demonstrates induction of tumour growth through FIH-1 overexpression in a xenograft mouse model using osteosarcoma cells whereas suppression of FIH-1 had no effect on proliferation of the cells (Kuzmanov et al., 2011). Taken together, these reports indicate that the function of FIH-1 apparently depends on the cell type examined.

Previous studies on ARD-containing FIH-1 substrates showed that neither the half-life nor thermodynamic stability of long-lived (e.g. Ranbankyrin-5) or short-lived proteins (e.g. IκBα) were altered upon hydroxylation by FIH-1 (Devries et al., 2010; Singleton et al., 2011) although studies using consensus ankyrin repeat proteins detected stabilization of the domain through hydroxylation by FIH-1 (Hardy et al., 2009; Kelly et al., 2009). With respect to ASPP2, we were unable to demonstrate differences in total protein levels of ASPP2 upon suppression of FIH-1, suggesting that hydroxylation does not have a gross regulatory function in terms of ASPP2 stability or production.

Although the stability of the ARD has not been shown to alter the stability of the protein in any case examined so far, the total conformation of the ARD-containing protein can change upon hydroxylation by FIH-1, thereby modulating the affinity to protein interaction partners. In proof of this hypothesis, we demonstrated, for the first time, a difference in protein interaction and localization for an ARD-containing protein under suppression of FIH-1. In HCT116 colon carcinoma shFIH-1 cells the interaction of ASPP2 with Par-3 was significantly reduced, leading to an altered intracellular localization of ASPP2. H441 pulmonary adenocarcinoma cells with a more epithelial phenotype also showed a clearly disorganized localization of ASPP2 at the cell boundaries. In addition, these effects were dependent on hydroxylation as demonstrated in experiments in which the hydroxylase inhibitor DMOG was used. In contrast, the interaction of ASPP2 with the tumour suppressor p53 was not affected by FIH-1-dependent hydroxylation. These results point to a specific role of FIH-1 in the formation of cell–cell contacts and maintenance of tight junctions via regulation of the ASPP2–Par-3 interaction, which was shown to be important in this context. Therefore, we attempted to detect differences in monolayer permeability or transepithelial electrical resistance (TER) under suppression of FIH-1 in cultured tumour cells. However, few cultured cell lines develop an interpretable TER, and despite considerable effort we were unable to produce cells that exhibited stable lentiviral suppression of FIH-1 and a sufficiently high TER simultaneously. Therefore, it remains unclear at present whether the reduced ASPP2–Par-3 interaction indeed affects tight junctions, or whether an effect depends on a distinct, potentially pathophysiological situation such as severe hypoxia. Interestingly, the FIH-1 knockout mice show a phenotype of metabolic disorders and no developmental defects (Zhang et al., 2010). This, however, does obviously not preclude an effect under pathophysiological conditions. In contrast, ASPP2 knockout mice display manifest defects in the CNS due to irregular polarity and proliferation of neuronal progenitor cells (Sottocornola et al., 2010) that are potentially related to defects of tight junction integrity.

The physiological role of the ASPP2–Par-3 interaction has been characterized in epithelial cells exclusively, where it generally regulates the early formation of cell–cell contacts and the maintenance of tight junctions (Cong et al., 2010). It is now interesting to further explore the role of this interaction itself and the impact of FIH-1 especially under pathophysiological conditions in which hypoxia plays a role. In this context, it is particularly interesting to assess the function of the ASPP2-Par-3 complex in endothelial cells because hypoxia is known to increase the permeability of, for example, the blood brain barrier after ischemic injury (Lochhead et al., 2010; Natah et al., 2009). In vitro, we could not demonstrate FIH-1-dependent effects on proliferation, apoptosis induction, migration or adhesion of cultured tumour cells. However, the integrity of in situ epithelia or the function of the blood–brain barrier – since Par-3 has been connected with CNS development – cannot be reproduced adequately in vitro. Owing to the limitations of in vitro models, our data may provide guidance for further studies using in vivo disease models that are required to assess the physiological consequences of the altered interaction between ASPP2 and Par-3.

Monoclonal antibodies against ASPP2 (611354) and HIF-1α (610959) were purchased from BD Biosciences (Heidelberg, Germany). The rabbit polyclonal antibody raised against β-actin (A2103) was from Sigma (Munich, Germany). Rabbit polyclonal antibodies against FIH-1 (NB100-428) and Par-3 (07-330) were from Novus Biologicals (Littleton, CO, USA) and Millipore (Billerica, MA, USA), respectively. The monoclonal p53 antibody (OP43) was purchased from Merck (Darmstadt, Germany). The goat polyclonal, HRP-conjugated antibodies against rabbit (PO448) and mouse (PO447) antibodies were from DAKO (Hamburg, Germany). The secondary antibodies Alexa Fluor 488 and Atto 647 came from Invitrogen and Sigma, respectively. Etoposide, doxorubicin, Ac-DEVD-AMC, puromycin, polybrene, chloramphenicol, ampicillin and kanamycin were purchased from Sigma (Munich, Germany). Dimethyloxalylglycine (DMOG) was from Enzo Life Sciences (Farmingdale, NY, USA).

E. coli Stbl3 (Invitrogen, Darmstadt, Germany) served as a host for production of the lentiviral vectors pLKO.1 (8453, Addgene), pLKO.1-shRNA-FIH-1 (TRCN0000064823, Sigma), containing the 21-nucleotide sequence GCCCTTGTTGAACACAATGAT corresponding to nucleotides 1086–1107 of the FIH-1 mRNA (GenBank acc. no. NM_017902) and pLKO.1-shRNA-ASPP2 (TRCN0000061797, Sigma) containing the 21-nucleotide sequence GCCTTTCTTATCTAATCCTTA corresponding to nucleotides 2551–2571 of the ASPP2 mRNA (GenBank acc. no. NM_005426). pLKO.1-scramble was purchased from Addgene (plasmid ID 1864). Expression plasmids for N-terminal His-tagged full-length FIH-1 [pET28a(+) FIH-1] and HIF-1α (737–826) (pET32a HIF-1α 737–826 wt and N803A) were kindly provided by Christopher J. Schofield (University of Oxford, UK) (Hewitson et al., 2002) and Daniel J. Peet (University of Adelaide, Australia) (Linke et al., 2004), respectively. For expression of ASPP2 in mammalian cells the plasmid pcDNA3.1/ASPP2-V5 (C-terminal tag) containing the human ASPP2 mRNA sequence (GenBank acc. no. AJ318888) was kindly provided by Xin Lu (Samuels-Lev et al., 2001). Overexpression of FIH-1 was achieved by transient transfection with pcDNA3-FIH-1. For bacterial expression His-tagged ASPP2 (852–1091) plasmids with an N-terminal tag were generated by PCR cloning. For mutation studies bacterial expression plasmids for mutant N-terminal His-tagged ASPP2 (852–1091) were generated by overlap extension PCR (Ho et al., 1989). The generated plasmids were used as templates for PCR cloning of bacterial expression constructs for N-terminal GST–ASPP2 (852–1091). The PCR products were transferred into pGEX-4T-1 (GE Healthcare) using BamHI and XhoI. We also generated plasmids for bacterial expression of GST–HIF-1α (737–826) by PCR cloning. In this case pET32a HIF-1α 737–826 wild type or N803A served as templates and pGEX-4T-1 was used as a recipient vector. All oligonucleotide sequences and cloning details are available on request. The integrity of each construct was verified by DNA sequencing.

HCT116 wild-type and p53^−/− cell lines were kindly provided by B. Vogelstein (Bunz et al., 1998) and maintained in McCoy’s 5A medium (Promocell) supplemented with 10% FBS and antibiotics. HEK293T cells were cultured in high-glucose DMEM (Invitrogen) supplemented with 10% FBS, 2 mM glutamine and antibiotics. Transient transfections were performed using GeneJuice (Novagen) in a ratio of 3∶1 (µl reagent/µg DNA). Lentiviral particals were produced</emph> in HEK293T cells as described previously (Brockmeier et al., 2011). For transduction 2×10⁵ HCT116 or H441 cells were incubated for 16–24 hours with 2×10⁶ transduction units and 8 µg/ml polybrene. Infected cells were selected with 6 µg/ml and 2 µg/ml puromycin, respectively.

For western blotting, whole cell extracts were prepared in RIPA lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 0.1% SDS, 1% Nonidet P40, 0.5% sodium desoxycholate, 2 mM EDTA) containing protease inhibitor cocktail (Roche, Mannheim, Germany). Cell lysates were separated using reducing 7.5–10% SDS gels. The PVDF membranes were blocked with 5% skimmed milk in TBST. The HRP-conjugated secondary antibodies were detected with an ECL kit (GE Healthcare, Munich, Germany) and an FX7 chemoluminescence documentation system (Peqlab, Erlangen, Germany). For co-immunoprecipitation, whole-cell extracts were prepared in NP40 lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 0.5% Nonidet P40, 2 mM EDTA). Cell lysates were incubated with immunoprecipitation antibodies for 2 hours at 4°C. Immune complexes were recovered by incubation with protein-G–Sepharose beads (Sigma) for 1 hour at 4°C. The beads were washed, resuspended in sample buffer and boiled for 7.5 minutes at 95°C. The supernatant was analysed by immunoblotting. Co-immunoprecipitations were repeated independently at least three times.

Cells were grown on glass plates coated with rat collagen-I (Biozol, Eching, Germany). Cells were fixed with 4% paraformaldehyde in PBS for 30 minutes and permeabilised with 0.5% Triton X-100 in PBS for 3 minutes. After blocking with 2% BSA in PBS for 15 minutes, the cells were incubated with primary antibodies for 2 hours. Alexa-Fluor-488- and Atto-647-conjugated secondary antibodies were used for detection. Nuclei were counterstained with DAPI. Slides were analysed using a Zeiss LSM510 confocal microscope with a 63×/1.2 NA oil immersion lens (Carl Zeiss, Heidelberg, Germany).

Expression was performed in BL21-CodonPlus (DE3)-RIPL competent E. coli cells (Stratagene) and induced as follows. His–HIF-1α: 1 mM IPTG, 37°C; 4 hours; His–FIH-1: 1 mM IPTG, 30°C; 6 hours; GST–ASPP2: 1 mM IPTG, 30°C; 6 hours; His–ASPP2: 0.5 mM IPTG, 20°C; 16 hours; GST–HIF-1α: 1 mM IPTG, 30°C; 6 hours. Cell pellets were disrupted in lysis buffer containing CellLytic™ B (Sigma), 0.2 mg/ml Lysozyme and protease inhibitor cocktail (Roche) in PBS (for GST fusion proteins) or in 20 mM Tris-HCl (pH 8), 150 mM NaCl (for His-tagged proteins). For glutathione affinity purification, lysates were incubated with glutathione–Sepharose 4B (GE Healthcare) at 4°C for 1 hour. The proteins were eluted with 20 mM Tris-HCl (pH 8.8), 150 mM NaCl, 25 mM reduced glutathione. For nickel affinity purification, salt concentration of the cleared lysate was increased to 500 mM NaCl, then 5 mM imidazole, 0.5 mM dithiothreitol (DTT) and 1 mM phenylmethylsulfonyl fluoride were added. The lysates were incubated with Ni-NTA Agarose (Qiagen) at 4°C for 1 hour. Bound protein was eluted with 20 mM Tris-HCl (pH 8), 500 mM NaCl, 250 mM imidazole. After affinity purification, protein solutions were desalted using PD-10 columns (GE Healthcare). When necessary the protein concentration was increased using Amicon centrifugal filter devices (Millipore). Purified proteins were stored at −20°C or −80°C.

FIH-1 activity was assessed using the hydroxylation-coupled decarboxylation of 2-oxo[1-¹⁴C]glutarate. ¹⁴CO₂ released from the reaction was precipitated using a Ca(OH)₂-soaked filter paper placed in the top of the tube above the reaction. Hydroxylation was performed in a final volume of 40 µl containing 7.5 µM His–FIH-1, 10–300 µM substrate, 0.3% BSA (3 mg/ml), 50 mM Tris-HCl (pH 7), 0.5 mM DTT, 4 mM ascorbate, 20 µM FeSO₄, 40 µM 2-oxo[1-¹⁴C]glutarate (Perkin Elmer). After 2 hours at 37°C, ¹⁴CO₂ released from the reaction was detected by scintillation counting using Ultima Gold XR (Perkin Elmer) in a Tri-Carb liquid scintillation counter (Perkin Elmer). Control experiments showed that ¹⁴CO₂ production was linear over time for at least 20 minutes. At least three independent samples were analysed. All results were confirmed in independent sets of experiments. Statistical significance was tested by one-way ANOVA analysis using GraphPad Prism 5 software.

Protein bands were excised and destained by alternately incubating them three times with 10 mM NH₄HCO₃ and 5 mM NH₄HCO₃/50% acetonitrile (ACN) (v/v) for 10 minutes each. Proteins were reduced with 10 mM DTT for 45 minutes at 56°C and alkylated with 55 mM iodoacetamide for 30 minutes at room temperature in the dark. In-gel digestion was carried out with 150 ng chymotrypsin (Promega) per band in 10 mM ammonium bicarbonate overnight. Peptides were extracted twice from the gel using 15 µl 0.05 trifluoroacetic acid/50% ACN (v/v), the extracts combined and ACN removed in vacuo. Peptides were separated using an Ultimate 3000 RSLCnano HPLC system (Thermo Fisher Scientific) online-coupled to a QExactive mass spectrometer (Thermo Fisher Scientific). For peptide separation, a C18 RP nano LC column (75 µm inner diameter, 500 mm, particle size 2 µm; Acclaim PepMap, Thermo Fisher Scientific) and a binary solvent system consisting of 0.1% (v/v) formic acid (solvent A) and 0.1% (v/v) formic acid in 84% (v/v) acetonitrile (solvent B) with the following linear gradient were used: 5–40% solvent B in 41 minutes and 40–95% solvent B in 1 minute. Columns were operated at a constant temperature of 60°C. The LC was coupled to the mass spectrometer using a nano-electrospray ion source (Thermo Fisher Scientific) and distal coated Silica Tips (FS360-20-10-D, New Objective). To provide high mass accuracy, lock masses (derived from a set of distinctive air contaminants) were routinely used for internal calibration of each MS spectrum acquired. Survey scans (550–1200 m/z) were recorded with a resolution of 70,000 for m/z = 200. For fragmentation, an inclusion list with masses for possible chymotryptic hydroxylated ASPP2 peptides was used. In case none of those masses was found, data-dependent fragmentation of the three most intensive signals was performed. Fragmentation was carried out in the HCD cell and the spectra recorded in the orbitrap with a resolution of 35,000 for m/z = 200. The AGC was set to 200,000 ions and a maximum fill time of 600 mseconds. HCD fragmentation was carried out at a normalized collision energy of 27%. Dynamic exclusion was enabled with an exclusion time of 5 seconds to prevent fragmentation of previously selected precursors. Raw files were processed using ProteomeDiscoverer 1.3 (Thermo Fisher Scientific). MS/MS spectra were searched using the Mascot search engine (Matrix Science), against the Uniprot complete proteome set for Homo sapiens (26/10/2012, 68,109 entries) with the following settings: precursor tolerance: 5 p.p.m., fragment tolerance: 20 m.m.u., chymotrypsin specificity, three missed cleavages allowed, oxidation of methionine and hydroxylation of asparagine as variable modifications, carbamidomethylation of cysteine as a fixed modification. Peptide identifications were filtered at 1% false discovery rate (FDR) on peptide level by the peptide validator algorithm of Proteome Discoverer 1.3. In addition to database searching, MS/MS spectra for ASPP2 hydroxylated or non-hydroxylated peptides were inspected manually. For quantitative comparison of peptides, elution profile areas from extracted ion chromatograms (5 p.p.m. mass tolerance) of the respective masses were calculated in XCalibur (Thermo Fisher Scientific).

In vitro transcription and translation was performed with rabbit reticulocyte lysate (L1170; Promega). For radioactive labelling of generated proteins EasyTag™ L-[³⁵S]methionine (NEG709A, PerkinElmer) was added. To detect interaction with FIH-1, unlabelled FIH-1 substrates were incubated as for the decarboxylation assay with 20 µM zinc acetate and 40 µM N-oxalylglycine instead of FeSO₄ and 2-oxo[1-¹⁴C]glutarate. Following addition of 10 µl ³⁵S-labelled FIH-1 at 37°C for 2 hours, 50 µl glutathione–Sepharose 4B were added. After incubation for 1 hour at 4°C and washing, the beads were resuspended in sample buffer and boiled for 7.5 minutes at 95°C. After centrifugation, the supernatant was used for detection of FIH-1 by autoradiography or stained with Coomassie Blue to determine whether interacting substrate was present.

Release of fluorescent 7-amino-4-methylcoumarin (AMC) from the caspase-3-specific substrate acetyl Asp-Glu-Val-Asp 7-amido-4-methylcoumarin (Ac-DEVD-AMC; Sigma A1086), was quantified to detect apoptosis. Cells were lysed in NP40 buffer [50 mM Tris pH 7.3, 150 mM NaCl, 1% (v/v) Nonidet P40]. Cell lysates containing 50 µg protein were incubated with a final concentration of 50 µM Ac-DEVD-AMC in caspase substrate buffer [50 mM Hepes pH 7.3, 100 mM NaCl, 10% (w/v) sucrose, 0.1% (w/v) CHAPS, 10 mM DTT] at 37°C. AMC fluorescence was measured every 10 minutes at an excitation of 360 nm and emission of 460 nm over 4 hours in a fluorescence reader (FLx800, Biotek, Bad Friedrichshall, Germany). For analysis of the results, replicate values (n = 3 per experiment) of a single timepoint in the linear range of the reaction were plotted.

To measure cell proliferation, 2×10³ HCT116 or H441 cells were plated into 96-well plates. At the indicated times the cells were incubated with 0.5 mg/ml 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) for 4 hours at 37°C and disrupted with MTT lysis buffer. Absorbance was measured at 540 nm in a multimode reader (Epoch, Biotek).

To measure adhesion 4×10⁴ HCT116 or H441 cells were plated into collagen-coated 96-well plates in serum-free medium containing 0.5% BSA. Cells were left to adhere for 2 hours at 37°C. The plate was washed twice with PBS and adherent cells were incubated with 2.5 µM Calcein-AM for detection of living cells. Fluorescence was quantified every 5 minutes at an excitation of 485 nm and emission of 525 nm over 2 hours in a fluorescence reader (FLx800, Biotek). For analysis of the results, replicate values (n = 10 per experiment) of a single timepoint in the linear range of the reaction were plotted.

Migration of cells was monitored by scratch assays. Cells were grown to confluent monolayers in collagen-coated culture dishes before scratching. Closure of the wounds was measured as a reduction of distance between the edges over time. For statistical analysis, distances were measured at 11 different positions of each scratch.

The authors would like to thank Kirsten Goepelt, Frank Splettstoesser and Melanie Baumann for excellent technical assistance. We are also grateful to Joachim Fandrey for many helpful discussions and invaluable support.