Myeloid Translocation Gene-16 Co-Repressor Promotes Degradation of Hypoxia-Inducible Factor 1

The myeloid translocation gene 16 (MTG16) co-repressor down regulates expression of multiple glycolytic genes, which are targets of the hypoxia-inducible factor 1 (HIF1) heterodimer transcription factor that is composed of oxygen-regulated labile HIF1α and stable HIF1β subunits. For this reason, we investigated whether MTG16 might regulate HIF1 negatively contributing to inhibition of glycolysis and stimulation of mitochondrial respiration. A doxycycline Tet-On system was used to control levels of MTG16 in B-lymphoblastic Raji cells. Results from co-association studies revealed MTG16 to interact with HIF1α. The co-association required intact N-terminal MTG16 residues including Nervy Homology Region 1 (NHR1). Furthermore, electrophoretic mobility shift assays demonstrated an association of MTG16 with hypoxia response elements (HREs) in PFKFB3, PFKFB4 and PDK1 promoters in-vitro. Results from chromatin immunoprecipitation assays revealed co-occupancy of these and other glycolytic gene promoters by HIF1α, HIF1β and MTG16 in agreement with possible involvement of these proteins in regulation of glycolytic target genes. In addition, MTG16 interacted with prolyl hydroxylase D2 and promoted ubiquitination and proteasomal degradation of HIF1α. Our findings broaden the area of MTG co-repressor functions and reveal MTG16 to be part of a protein complex that controls the levels of HIF1α.

Mature cells normally cover energy needs for homeostasis by ATP generated through mitochondrial respiration. In contrast, neoplastic cells characteristically show low mitochondrial respiration paralleled with increased glucose uptake, lactate export and extracellular acidification, even under good oxygen supply, termed the Warburg effect [18]. This glycolytic switch driven by activation of oncogenes and inactivation of suppressor genes [19,20] supports anabolic needs for metabolic intermediates required for neoplastic cell growth [20]. Highly proliferating normal cells also show stimulated glycolysis, which is driven by proliferative signals from growth factors [21]. We discovered that the MTG16 co-repressor inhibited the glycolytic switch and stimulated mitochondrial respiration [1]. In support of this, expression of hypoxia -inducible transcription factor 1 (HIF1) glycolytic target genes such as 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (PFKFB3) [22,23] and PFKFB4 [24] as well as the mitochondrial respiration inhibitor pyruvate dehydrogenase kinase isoenzyme 1 (PDK1) [25] were inhibited. The inhibition may involve negative regulation of HIF1 by MTG16.
HIF1 is a heterodimer composed of oxygen-regulated unstable HIF1α and ubiquitously expressed stable HIF1β subunits [26]. At hypoxic conditions, the oxygen-dependent proteasomal degradation of HIF1α is inhibited [27] permitting dimerization with HIF1α, binding of dimer to hypoxia response elements (HREs) and transcriptional activation of target genes [28,29]. HIF1 controls cellular adaptation to oxygen lack [30,31] by activating genes that coordinate stimulated glycolysis with repressed mitochondrial respiration [32]. However, HIF1α may stay active in neoplastic cells even at non-hypoxic oxygen concentrations being activated by molecules such as pyruvate, lactate, mammalian target of rapamycin (mTOR), reactive oxygen species (ROS), and oncogene gain of function or suppressor gene loss of function as reviewed in [33].
Given our findings of MTG16-mediated inhibition of glycolysis and stimulation of mitochondrial respiration, we asked whether the MTG16 co-repressor might regulate HIF1 negatively and contribute to the inhibited expression of glycolytic genes such as PFKFB3, PFKFB4 and PDK1 that are down regulated when MTG16 is elevated. For support, we first determined whether MTG16 is a HIF1α-interacting protein. Then, we investigated whether MTG16 is part of a HIF1-containing protein complex at target promoters. Furthermore, we investigated whether ectopically expressed MTG16 affected HIF1α stability. Controlled biosynthesis of ectopically expressed MTG16 was obtained by use of a doxycycline-regulated Tet-On time and dose-dependent gene expression system in B-lymphoblastic Raji cells [1].

Transfection
Raji cells (8x10 6 ) were electroporated with plasmid in 0.4 ml of culture medium using the Bio-Rad Electroporation Apparatus (Bio-Rad Laboratories, Hercules, CA) with electrical settings of 960 mF and 260V. Antibiotic was added after 48 h for selection of resistant recombinant clones, which were isolated, expanded into mass cultures and screened for expression.

Generation of stable doxycycline inducible MTG16 clones
The Tet-On 3G tetracycline inducible gene expression system (Clontech, Ozyme, Saint Quentin en Yulines, France) was used for generation of stable doxycycline inducible clones of MTG16 inserted under the TRE3G promoter (P TRE3G ) in B-lymphoblastic Raji cells (Raji/ MTG16 Tet-On 3G cells) as previously described [1]. Incubation with 10-20 ng/ml of the tetracycline analog doxycycline induces Tet-On 3 G trans activator binding to tet operator repeats within P TRE3G leading to transcriptional MTG16 activation. MTG16 biosynthesis was seen after 3 to 4 h of induction at a very low concentration (20 ng/ml) of doxycycline making unspecific effects unlikely (data not shown).

Co-immunoprecipitation
Co-immunoprecipitation was performed as described earlier [40]. Raji-MTG16 cells were incubated overnight under 4% O 2 with doxycycline for expression of MTG16. Alternatively, COS-7 cells were transfected with MTG16 and HIF1α and harvested after 24 h. The cells were homogenized with lysis buffer (250 mM NaCl, 20 mM Na-phosphate /pH 7.0/, 30 mM Na-pyrophosphate /pH 7.0/, 5 mM EDTA, 0.1 mM Na 3 VO 4 , 10 mM NaF and 0.1%NP-40) supplemented with protease inhibitors on ice for 1h. Lysates were cleared by centrifugation at 13000 rpm for 1 h at 4°C and incubated overnight with Protein A Sepharose and antibody. The Protein-A-Sepharose beads were then washed three times in lysis buffer to remove unbound non-specific protein. Thirty-μl sample buffer (114 mM Tris HCL /pH 6.8/, 15% glycerol, 105 mM SDS, 238 mM β-mercaptoethanol and 0.013% Bromophenol blue) was added followed by boiling 5 min, centrifugation and Western blotting.

GST pulldown assays
GST and GST fusion proteins were expressed in Escherichia coli JM109 cells by transformation with GST, GST-HIF1α and GST-MTG16 in pGEX-4T3 vectors. Recombinant bacterial cells expressing GST, GST-HIF1α and GST-MTG16 were lysed in NTN buffer (100 mM NaCl, 20 mM Tris and 0.1%NP-40) by sonication for 10 min. Lysates were cleared by centrifugation 10 min at 13000 rpm. GST-HIF1α and GST-MTG16 were bound to glutathione-Sepharose beads during rotation 2 h at 4°C. The beads were washed five times in NTN buffer containing 1.0 mM EDTA. Then, whole cell lysate was mixed with beads and incubated during rotation overnight at 4°C. After washing five times with NTN buffer containing 1.0 mM EDTA, bound proteins were eluted with sample buffer during heating at 95°C for 5 min and examined by Western blotting.

Statistical analysis
The significance of difference between samples was determined by the unpaired Student´s t tests or the one-or two-way ANOVA followed by post hoc tests using the Graphpad Prism version 5.0a Software (GraphPad Software, Inc., CA), unless stated differently. Single, double and triple asterisks represent P<0.05, P<0.001 and P < 0.0001, respectively. Data are presented as means±SEM.

MTG16 is a HIF1α-interacting protein
In order to detect a direct interaction between HIF1α and MTG16 and to map possible interaction domains, we used bacterially produced fusions of HIF1α with glutathione-S-transferase (GST). GST-HIF1α was recovered on glutathione-Sepharose beads and examined by SDS-PAGE. The binding of MTG16 species to HIF1α in cell lysates from COS-7 cells transfected with MTG16 was investigated. Results showed full length MTG16 to be pulled down by GST-HIF1α ( Fig 1B). A number of MTG16 truncation constructs were used to map sites of interaction with HIF1α. MTG16 residues 1-290, 1-354, 1-405 bound to GST-HIF1α ( Fig 1B). However, a construct consisting of residues 290 to 653 (lacking the NHR1 domain) was not pulled down by GST-HIF1α. As follows, MTG16 residues 1-290 containing NHR1 (residues 171-268) are sufficient for binding to HIF1α. Binding specificity was confirmed by lack of interaction of MTG16 with GST ( Fig 1C). Furthermore, in the reverse experiment, full length HIF1α was pulled down by GST-MTG16 ( Fig 1D).
To detect interaction between MTG16 and HIF1α co-immunoprecipitation assays were used. Ectopic interactions were examined in COS-7 cells transiently transfected with MTG16 and HIF1α. Results from IP-Western analyses showed that MTG16 co-precipitated HIF1α and the reciprocal IP-Western experiment showed HIF1α to co-precipitate MTG16 (Fig 2A).
Interaction between HIF1α and MTG16 was also examined in Raji/MTG16 Tet-On 3G cells, which were incubated with doxycycline to induce expression of MTG16. The experiments were performed during exposure to 4% O 2 to increase the level of HIF1α. Results from IP-Western analyses showed that MTG16 co-precipitated HIF1α (Fig 2B, top panel) and the reciprocal experiment showed that HIF1α co-precipitated MTG16 (Fig 2B, lower panel). These data demonstrate binding between MTG16 and HIF1α.
MTG16 is detectable at HIF1α response elements (HREs) in PFKFB3, PFKFB4 and PDK1 promoters in-vitro As MTG16 does not bind directly to DNA, indirect association might be mediated via a protein complex containing HIF1α. Electrophoretic mobility shift/supershift assays (EMSA) were used MTG16 Co-Repressor Inhibits HIF1 Activity to examine the capacity for association with HREs in PFKFB3, PFKFB4 and PDK1 promoters. Biotinylated oligonucleotide probes of HRE core consensus sites with flanking regions (Fig 3) were incubated with nuclear extracts prepared from Raji-MTG16 cells incubated at 4% O 2 to increase HIF1 and with doxycycline for 12 h to induce production of MTG16 in order to investigate interactions between probe and protein. As a result, nuclear extract proteins showed a shift indicating binding of protein to the probes (Fig 3). Binding was specific as no shift was observed in the presence of excess unlabeled competitor probe. Furthermore, protein bound to the labeled probes was "super-shifted" by α-MTG antibody but not with a control antibody. Our results show that MTG16 becomes associated with HREs within PFKFB3, PFKFB4 and PDK1 promoters. In addition to the specific shift is an unspecific band-shift for the PDK1 and PFKFB3 promoters, which is unaffected or only weakly affected by the competitor probe (Fig 3). Nonetheless, the band contains MTG16 as it is lost by super-shifting with α-MTG antibody. In conclusion, EMSA demonstrated a capacity of MTG16 to associate with HREs in PFKFB3, PFKFB4 and PDK1 promoters in-vitro.
To get further support for an interaction between MTG16 and HIF1α on promoters, ChIP assays were used. We asked whether MTG16 was present in HIF1-containing complexes. Thus, ChIP assays were performed to detect MTG16, HIF1α and HIF1β co-occupancy in-vivo at HREs in the promoters of the HIF1 target genes. The assays were performed with chromatin isolated from Raji/MTG16 Tet-On 3G cells incubated at 4% O 2 (to activate HIF1α) and with doxycycline 12 h (to induce expression of MTG16) after which immunoprecipitation with antibodies towards MTG16, HIF1α or HIF1β was performed. By the use of primers specific for HRE sites with flanking regions, qPCR showed amplification products generated by precipitation with either α-MTG, α-HIF1α or α-HIF1β, indicating binding in-vivo of MTG16, HIF1α and HIF1β near HRE sites of PFKFB3, PFKFB4 and PDK1 (Figs 4 and 5).
ChIP assays were also carried out with chromatin isolated from Raji/MTG16 Tet-On 3G cells incubated at 20% O 2 ( Fig 5) and with doxycycline for 8 or 24 h to induce expression of MTG16. The results indicated binding in-vivo of MTG16, HIF1α and HIF1β near HRE sites of PFKFB3, PFKFB4 and PDK1 after 8 h but not after 24 h of incubation with doxycycline ( Fig  5). The lack of MTG16 binding of HIF1α and HIF1β at 24 h is consistent with the MTG16mediated total downregulation of HIF1α that was observed at this time point in cells exposed to 20% O 2 (shown below) and serves as an additional control for specificity of MTG16 binding to HIF1α. In as much as the HIF1α expression was much lower at 20% O 2 compared to 4% O 2 , the results of Fig 5 may be less biologically relevant. However, down regulation of PDK1, PFKFB3 and PFKFB4 was observed previously during the conditions of 20% O 2 [34] consistent with a correlation between binding of the MTG16-HIF complex and transcriptional repression. HIF1 activates many glycolytic target genes in addition to PFKFB3, PFKFB4 and PDK1 such as lactate dehydrogenase A (LDHA), hexokinase (HK) and phosphofructokinase-1 (PFK1) [41]. The promoters of these genes were also strongly downregulated by MTG16 (S1 Fig). Therefore, we determined whether MTG16 also interacted in-vivo with bound HIF1α on the LDH, HK and PFK1 promoters. ChIP assays showed binding of MTG16, HIF1α and HIF1β near HRE sites also at these promoters (Fig 6). The endogenous co-occupancy of glycolytic gene promoters by both HIF1α and MTG16 is consistent with the interaction that we detected between these two proteins.

MTG16 promotes HIF-1α degradation
The HIF1α mRNA levels in Raji/MTG16 Tet-On 3G cells incubated with doxycycline to produce MTG16 showed no significant decrease (data not shown) indicating the co-repressor not to regulate HIF1α gene expression. However, the MTG16 interactions shown by IP-Western and ChIP analyses may diminish HIF1α protein stability and thereby suppress HIF1 activity. The α-MTG antibody was used in Western blotting because of its sensitivity for detection of MTGs [38]. α-MTG did not detect any MTG isoforms in Raji cells when MTG16 was not ectopically expressed (Fig 7). Therefore, this antibody detects only MTG16 in these cells when induced by incubation with doxycycline. HIF1α stability was investigated both at 20% and 4% O 2 . For detection of HIF1α by Western blotting in experiments performed at 20% O 2 , film exposures were prolonged to compensate for a lower level of HIF1α compared to conditions at 4% O 2 as indicated in legends to Figs 7 and 8. Results from time course experiments demonstrated that MTG16 destabilized HIF1α during exposure to 20% O 2 ( Fig 7A); HIF1α became undetectable between 8 and 16 h of incubation with doxycycline. Results from similar experiments performed during exposure to 4% O 2 (to increase the level of HIF1α) showed that HIF1α decreased after 18 to 24 h of incubation with doxycycline, as compared to incubation without doxycycline (Fig 7B). However, the decrease during hypoxia was less pronounced as compared to 20% O 2 , indicating dependency on the level of O 2 and HIF1α.
A decrease in HIF1α protein could result from either decreased biosynthesis or increased degradation. HIF1α is hydroxylated on one of two conserved prolyl residues predominantly by MTG16 Co-Repressor Inhibits HIF1 Activity prolyl hydroxylase domain-containing protein 2 (PHD2) [42] as a requisite for its polyubiquitination and proteasomal degradation. Therefore, we investigated if HIF-1α was ubiquitinated in response to MTG16. Raji-MTG16 cells co-transfected with HA-HIF1α and FLAG-ubiquitinin were incubated during exposure to 4% O 2 with doxycycline to induce production of MTG16. Examination of immunoprecipitated HA-HIF1α by α-FLAG antibody showed that HA-HIF1α ubiquitination indeed was increased by MTG16 (Fig 8A and 8B).
Concomitant interactions of MTG16 with HIF1α and PHD2 may increase the rate of hydroxylation. We investigated binding of MTG16 to PHD2 by use of IP-Western analyses in Raji-MTG16 cells incubated during exposure to 4% O 2 for 12 h with doxycycline to induce MTG16 production. Both MTG16 (precipitated with α-MTG) and HIF1α (precipitated with α-HIF1α) co-precipitated PHD2 in whole cell lysate (Fig 8C). This result indicates that MTG16 interacted not only with HIF1α but also with PHD2. As these co-immunoprecipitation experiments were carried out with ectopically synthesized proteins they are unlikely to reflect MTG16 Co-Repressor Inhibits HIF1 Activity unspecific interactions. Further support for MTG16-mediated prolyl hydroxylation of HIF1α was achieved by immunoblotting with an antibody that specifically recognizes hydroxyproline at residue 564 of HIF1α (Hyp-564-HIF1α). Hydroxylation of Hyp-564 was detected in Raji-MTG16 cells within 4 h of incubation with doxycycline at both 20% O 2 ( Fig 8D) and 4% O 2 (Fig 8E). The expected down regulation of HIF1α with time after induction of MTG16 is seen in both cases. The results indicate that MTG16-mediated hydroxylation may be followed by proteasomal degradation of HIF1α. In support of this, the proteasomal inhibitor MG132 strongly inhibited MTG16-mediated HIF1α decrease both during exposure to 20% (Fig 8F) and 4% O 2 (Fig 8G).
Collectively, our results demonstrate that MTG16 reduces HIF1α protein levels through the prolyl hydroxylation-dependent pathway for ubiquitination and following proteasomal degradation.

Discussion
Published results [1] showing that MTG16 can inhibit glycolysis and stimulate mitochondrial respiration by affecting expression of some HIF1 target genes (PFKFB3/4 and PDK1), suggested a possible negative regulation of HIF1, which controls expression of glycolytic genes in response to hypoxia / tumorigenesis. In this study, we demonstrate that the MTG16 corepressor is an opposing regulator of HIF1 by serving as a HIF1α-interacting protein with an amplifying effect on HIF1α hydroxylation and proteasomal degradation.
HIF1 is usually not activated at non-hypoxic conditions except in certain neoplastic cells such as those used for the experiments in this work. The relevance of MTG16-mediated reduction of HIF1α observed under the non-hypoxic conditions may be questioned, as HIF1α was much lower than in the cells exposed to hypoxia. However, the complete degradation of HIF1α at 20% O 2 concomitant with loss of MTG16 co-occupancy of glycolytic gene promoters strengthened the role of a HIF1α interaction. Total degradation of HIF1α at 20% O 2 by MTG16 might reflect a low HIF1α level, but degradation was observed at mild hypoxia too when the HIF1α level is vigorously increased. The significance of this finding is strengthened by the reduced expression of HIF1α target genes observed in the presence of MTG16 during mild hypoxia (S1 Fig).
The MTG family of co-repressors is involved in transcriptional repression [2]. The MTG isoforms share the evolutionary highly conserved Nervy Homology Regions (NHRs) 1 to 4 [43]. The finding that the N-terminal MTG16 residues including NHR1 were sufficient for the binding of HIF1α is consistent with this domain being a structural scaffold for multiple transcription factors [44]. On the other hand, the MTG16-mediated inhibition of glycolytic gene expression required intact NHR2-3 [1] consistent with that the importance of the latter modules for oligomerization, protein-protein interaction with other co-repressors such as N-CoR and SMRT [15] and interactions with HDACs [15,44,45] required for transcriptional repression. The finding by EMSA that MTG16 associated with HREs within glycolytic gene promoters and the findings by ChIP assays that HIF1α-MTG16 interaction also takes place at these promoters near HRE sites indicates that MTG16 may affect transcriptional activity of HRE-bound HIF1α. Furthermore, recruitment of MTG16 to HREs was suggested to be HIF1α dependent because it was lost at 24 h of incubation with doxycycline at 20% O 2 (Fig 5) when HIF1α was degraded. HIF1α is normally kept inactive during non-hypoxic conditions by post-translational hydroxylation at conserved critical prolyl residues catalysed in particular by the prolyl-hydroxylase-domain (PHD)-containing enzyme prolyl hydroxylase PHD2 [46]. Hydroxylation is followed by ubiquitination by the von-Hippel-Lindau tumour suppressor protein (VHL) and proteasomal degradation of HIF1α [47]. As PHD proteins require oxygen for catalytic activity, HIF1α is normally stabilized and active when PHD proteins become inactivated during hypoxia [48]. However, in contrast to normal, certain neoplastic cells show O 2 independent constitutive expression of HIF1α [33]. MTG16-mediated degradation was observed both at 20% O 2 and during hypoxia. That MTG16 in both cases augmented recruitment of PHD2 to HIF1α and amplified its degradation through a PHD/ VHL-dependent pathway is consistent with the combined interaction that we discovered between HIF1α and PHD2. The hydroxylation of Proline-564 of HIF1α by MTG16 observed not only at 20% O 2 but also at mild hypoxia is consistent with the HIF1α degradation observed. Earlier studies have indicated hydroxylation of HIF1α at Proline-564 during conditions of mild hypoxia [49,50].
Both O 2 /PHD/VHL-dependent and-independent mechanisms regulate HIF1α protein levels. For example, the OS-9 protein is a negative O 2 -dependent regulator of HIF1 that increases prolyl hydroxylation by interacting with HIF1α and PHDs [51]. Hsp70 promotes ubiquitination and proteasomal degradation of HIF1α by recruiting the ubiquitin ligase CHIP [52]. The tumour suppressor Sirtuin 6 acts as a HIF1α co-repressor at glycolytic target genes through deacetylation of Lys9 of Histone 3 [53]. Furthermore, the mitochondrial tumour suppressor Sirtuin 3 promotes destabilization of HIF1α [54,55]. Our results add MTG16 as a novel negative regulator of HIF1α . possibly acting by a PHD/VHL-dependent mechanism. MTG16 is known as a chromatin repressor, which can bind other co-repressors and recruit HDAC [13] with an epigenetic role in gene transcription. Our results indicate an additional function for MTG16 on HIF1-mediated transcription merely by degrading HIF1α. It remains to be investigated whether the MTG isoforms MTGR1 and MTG8 also contribute to HIF1α hydroxylation and proteasomal degradation. Furthermore, it may be of interest to investigate whether the leukemia fusion protein AML1-MTG16 [10] retains the MTG16 ability for degradation of HIF1α.
The present results suggest a mechanism through which MTG16 could contribute to prolonged downregulation of glycolytic genes. This is consonant with MTG16-mediated inhibition of the expression of the glycolytic genes and the PDK1 gene concomitant with a metabolic switch towards mitochondrial respiration [1]. In summary, we identified the co-repressor MTG16 to be a novel binding partner of HIF1α and a negative regulator of its stability. Our results depict molecular mechanisms that have a potential to modulate functional interactions between key proteins regulating the metabolism of neoplastic cells.
Supporting Information S1 Table. Primers used for ChIP-PCR amplification. Primers were designed using Primer 3 plus software.