Regulation of CDKN2B expression by interaction of Arnt with Miz-1 - a basis for functional integration between the HIF and Myc gene regulatory pathways

We initially observed the interaction between Miz-1 and Arnt in a yeast two-hybrid screen where a deletion mutant of Arnt lacking the transactivation domain (ArntΔQ; amino acid (aa) 1–618) was found to interact with the C-terminal part of Miz-1 (aa 470–794; data not shown). Diverse Arnt constructs (schematically presented in Figure 1A) were tested for their ability to interact with Miz-1 in mammalian cells. Cos-1 cells were transiently transfected with Arnt/Flag and Miz-1/GFP and, as demonstrated in Figure 1B (upper panel), Arnt precipitated together with Miz-1 suggesting that these two factors interact in vivo. For the interaction to occur, the basic HLH (bHLH) domain of Arnt was required, but not the PAS domain (Figure 1B, upper panel). Complex formation between Miz-1 and Myc requires amino acids V393, V394, K397 and S405 in helix II of the HLH-domain of Myc [14]. Alignment of Myc and Arnt demonstrated that with regard to charge and hydrophobicity, a similar interactions surface is present in helix II of Arnt by the amino acids L114, T115, R118 and S126 (Figure 1A). Mutation of amino acid T115 and S126 (2xmut; T115D/S126A) decreased the ability of Arnt to interact with Miz-1, and mutation of all four amino acids (4xmut; L114A/T115D/R118G/S126A) reduced complex formation below the detection limit of the assay (Figure 1B, upper panel), pointing to a similar mode of interaction between Arnt and Miz-1 as between Myc and Miz-1. Control experiments demonstrated comparable expression levels of wt Arnt and the different mutants (Figure 1B, middle panel), and moreover that Miz-1 was precipitated at similar levels in all samples, although slightly less in lanes 1 and 6 (Figure 1B, lower panel). Experiments designed to map the domain(s) of Miz-1 involved in complex formation failed, presumably because the different deletion constructs exhibited different cellular localization (i.e. cytoplasmic, nuclear or perinuclear; data not shown). Work by Peukert and colleagues indicate that Miz-1 lacks a functional nuclear localization signal, however they found that overexpression of Myc induces a redistribution of Miz-1 from the cytoplasm to the nucleus [11]. To investigate whether Arnt would alter the localization of Miz-1, Arnt/CFP and Miz-1/Flag were expressed in HEK293 cells. Consistent with previous reports, Arnt/CFP localized exclusively to the nucleus whereas Miz-1/Flag was distributed diffusely throughout the cell when expressed singlehandedly (Figure 1C) [11, 23, 24]. Notably, in cells where Arnt/CFP and Miz-1/Flag were co-expressed, Miz-1/Flag was repositioned from the cytoplasm to the nucleus (Figure 1C). Although co-localization of the two proteins in HEK293 cells was not observed under these experimental conditions, this result suggests that overexpression of Arnt induces a cellular redistribution of Miz-1 and indicates a functional significance for the Miz-1/Arnt complex.

The CDKN2B promoter is an established target for the Miz-1/Myc complex. Whereas Miz-1 activates this promoter, Myc acts as a repressor through interaction with Miz-1 and displacement of p300/CBP [9, 10]. To explore the possibility that Arnt might affect this Myc/Miz-1-dependent transcriptional regulation, Arnt was overexpressed in human osteosarcoma U2OS cells together with a luciferase reporter construct containing 35 nucleotides upstream of the transcriptional start site of CDKN2B (−35CDKN2B/luc, [25]). This construct does not contain a consensus Arnt binding site, or a canonical HRE [26]. U2OS cells are frequently used in functional studies that aim to understand the molecular mechanisms underlying hypoxic transcriptional responses as this cell line respond well to low O₂ levels. As expected, based on the literature [14], Miz-1 induced luciferase activity from the CDKN2B promoter (Figure 2A). Likewise, Arnt caused an induction of luciferase activity from this promoter construct (Figure 2A). The stimulatory effects of Arnt and Miz-1 on this promoter construct did vary between experiments; however, we always observed significantly enhanced reporter gene activity after Arnt and Miz-1 overexpression. Mutation of the amino acids in Arnt required for Miz-1 interaction (2xmut, 4xmut) led to decreased reporter gene activity, but apparently, the low level of complex formation between Miz-1 and 2xmut Arnt and, although not detectable above the background in the co-IP assay, 4xmut Arnt was sufficient to drive reporter gene expression to a certain level in this system (Figure 2B). As stated above, Myc inhibits Miz-1 induced transcription from the CDKN2B promoter. The results presented in Figure 2C show that Myc also repressed Arnt-dependent reporter gene expression from this promoter, and moreover, that a mutated version of Myc that is unable to interact with Miz-1 (MycV394D) [14] failed to inhibit Arnt-induced transcription (Figure 2C). Taken together, these experiments support the concept that Arnt induces transcription from the CDKN2B promoter via interaction with Miz-1, and that Myc counteracts this activity through competition for the interaction surface of Miz-1.

To establish whether Arnt interacts with the CDKN2B promoter, chromatin-immunoprecipitation (ChIP) analyses combined with qPCR were performed. These experiments demonstrated that Arnt is enriched on the proximal CDKN2B promoter (Figure 3A) compared to a distal region of the promoter (Figure 3B) in U2OS cells. Since Arnt is an obligate partner of Hif1α and Hif2α, and thus plays an important role in hypoxia regulated gene expression [27, 28], we cultured U2OS cells in 1% O₂ for 4 or 24 hours to examine whether hypoxic conditions would affect binding of Arnt. Interestingly, recruitment of Arnt to the proximal CDKN2B promoter was reduced when the cells were cultured under hypoxia compared to normoxic conditions (Figure 3A). Unfortunately, with commercially available antibodies, we were not able to determine the occupation of Miz-1 on the CDKN2B promoter during hypoxia (or normoxia). Binding of Miz-1 to the CDKN2B promoter in normoxic cells has been shown previously [10, 15]. As a control for hypoxia induced transcription, primers spanning the HRE in the promoter of phosphoglycerate kinase 1 (PGK-1) were employed to determine Arnt interaction. As reported [29], Arnt was found to bind to the HRE of PGK-1 at normoxia, and at increased levels under hypoxic conditions (Figure 3C).

To further explore the potential role of O₂ pressure in the expression of CDKN2B, we examined whether hypoxia affected CDKN2B endogenous protein and mRNA levels. The protein level of CDKN2B decreased after 4 and 24 hours at 1% O₂ compared to the expression at normoxia (Figure 4A). Already after 2 hours, the CDKN2B mRNA level was significantly reduced, and declined further when the hypoxic exposure was prolonged to 16 or 24 hours (Figure 4B). Thus, these results are in agreement with the ChIP results in Figure 3, and indicate that as O₂ levels fall, Arnt is released from the promoter leading to decreased expression of CDKN2B. The mRNA expression of the hypoxia regulated gene PGK-1 was, in accordance with the literature [29], increased during hypoxic treatment (Figure 4C).

Expression of siRNA against Arnt (siArnt) caused a significant decrease in the CDKN2B mRNA expression, further supporting a role for Arnt in transcriptional regulation of this gene (Figure 5A, compare columns 1 and 3). As expected from previous studies [9, 10], inhibition of Miz-1 also repressed the expression of CDKN2B (Figure 5A, compare columns 1 and 5). Intriguingly, knockdown of Hif1α and Hif2α also caused reduced CDKN2B mRNA levels at normoxia (Figure 5A, compare column 1 with column 7 and 9). The effect of siHif1α was significant, despite the fact that the level of Hif1α protein at normoxia was below the detectable threshold (Figure 5C, right upper panel). In contrast, Hif2α was readily detectable at normoxia in this cell line (Figure 5C, right lower panel), as has also been reported previously [30]. In agreement with the results in Figure 4, hypoxic treatment for 6 hours decreased CDKN2B mRNA levels with around 50-60% in presence of control siRNA (Figure 5A, compare columns 1 and 2), and administration of siRNA against Arnt, Miz-1 and Hif2α under hypoxic conditions further repressed CDKN2B expression (Figure 5A, compare lane 2 with lanes 4, 6 and 10). In contrast, siHif1α had no significant effect under hypoxic conditions compared to control siRNA (Figure 5A, compare lanes 2 and 8).

Intriguingly, we found that Miz-1 mRNA levels were suppressed in response to hypoxia (Figure 5B compare lanes 5 and 6), as was the protein level (Figure 5C, left lower panel). A comparable response was observed when endogenous Miz-1 mRNA expression was determined in response to hypoxia (Figure 5D), suggesting that this gene is regulated by O₂ tension and substantiating the impact of Miz-1 as an important activator of CDKN2B in the normoxic state. This effect might be cell line specific, however, as a similar downregulation of Miz-1 mRNA in response to hypoxia is not observed in HCT116 or WT-8 cells [22]. The gene encoding Miz-1 contains an HRE element in the 5’ untranslated region as well as in intron 1, and it remains to be determined whether HIFs interact with these sequences. The mRNA levels of Arnt and Hif1α were also slightly reduced in response to hypoxia (Figure 5B, compare lanes 1 and 2 for Arnt and lanes 9 and 10 for Hif1α). However the corresponding protein levels increased, as they did for Hif2α, and as previously reported [31]. These results suggest that the Arnt/Miz-1 complex, possibly in association with Hifα, is involved in CDKN2B gene regulation in U2OS cells. Further supporting the scenario that such a complex has impact on transcription from the CDKN2B promoter are the results shown in Figure 5E, namely that Arnt-induced CDKN2B- promoter dependent luciferase expression is inhibited as a consequence of repressing Miz-1, Hif1α or Hif2α. All factors were relatively efficiently repressed by the corresponding siRNA, both at the mRNA level and at the protein level (Figure 5B and C).

Since Myc is an established inhibitor of CDKN2B expression, we hypothesized that Myc might replace Arnt on the proximal promoter during hypoxia. In contrast to our theory, however, ChIP experiments demonstrated decreased enrichment of Myc in response to low O₂ pressure (Figure 6A, left panel). Actually, it was previously reported that Myc is released from the CDKN2B promoter in hypoxic cells, as well as from the CDKN2A promoter, which was used as a control in our experiment (Figure 6B and [22]). CDKN2B is under tight transcriptional control by multiple signaling cascades, transcription factors and epigenetic mechanisms [32, 33], and most likely, other transcription pathways than those investigated in the present study are involved in the repression of this gene at hypoxic conditions.

The mutations in helix II of Arnt, T115D/S126A denoted “2xmut” and L114A/T115D/R118G/S126A denoted “4xmut”, were constructed using the Quick Change XL Site-directed Mutagenesis Kit (Stratagene). ArntΔbHLH (deleted of amino acids 67–146) has been described previously [49, 50] as has the plasmid Arnt/pECFP-C1 [51] and the luciferase reporter construct -35CDKN2B /Luc [25] (kindly provided by Dr. XF Wang (Pharmacology and Cancer Biology, Duke University School of Medicine, Durham, NC, USA). Myc/pcDNA3 was kindly provided by Dr. F. Haenel (Hans-Knöll-Institut für Naturstoff-Forschung, Department of Cell and Molecular Biology, 07745 Jena, Germany), and Miz-1/pcDNA3 by Dr. L.G. Larsson (Department of Microbiology, Tumor and Cell Biology, Karolinska Institutet, Stockholm, Sweden).

The human osteosarcoma U2OS (ATCC: HTB-96™) and HEK293 (ATCC: CRL-1573™) cell lines were cultured in high glucose DMEM (D5786, Sigma) supplemented with 10% bovine calf serum, penicillin (100 units/ml) and streptomycin (100 mg/ml) (Sigma). The HEK293-EBNA cell medium was supplemented with 25 mM HEPES. The cells were maintained in 5% CO₂ humidified atmosphere at 37°C. For hypoxic condition, the cells were cultured at 1% O₂ for the time points indicated in the figures.

For transient transfection experiments, U2OS cells were plated at a density of 6.5×10⁴ cells per well onto 12-well plates and transiently transfected the following day using XtremeGene9 (Roche). Cells were transfected with reporter plasmid (0.3 μg, −35CDKN2B/Luc) and expression plasmids (0.1 ug) as indicated in the figures. The total amount of DNA was kept constant by compensating with the plasmid psp70 (New England Biolabs). 24 hrs after transfection start, the cells were washed once with PBS and lysed in luciferase lysis buffer (10 mM Tris–HCl pH 8.0, 4 mM EDTA, 150 mM NaCl, 0.65% NP40). Cell extracts were then assayed for luciferase activity on a LUCY-3 luminometer (Anthos, Austria). Co-immunoprecipitation (co-IP) assays were conducted using Cos-1 cells. The cells were grown in 100 mm dishes to 70–80% confluency and transfected with indicated plasmid combinations using X-tremeGene 9 (Roche). The manufacture’s protocol was followed using a total amount of 5 μg plasmid per dish. Transfected cells were incubated for 36 hours at 5% CO₂ in 37°C before preparation of whole cell extract. The cells were washed with ice-cold PBS before lysis buffer (25 mM Hepes, 100 mM NaCl, 5 mM EDTA, 20 mM glycerol phosphate, 0.5% Triton, 20% glycerol; 300 μl) containing protease inhibitors (100 μM (Na₃VO₄), 2 mM DTT, 1 mM PMSF, 5 μg/μl aprotinin and 5 μg/μl leupeptin)) was added. The extracts were then sonicated (2 × 5 sec/amplitude 40), and centrifuged at 13000 rpm (16000 × g) for 5 minutes at 4°C to pellet cell debris. The supernatants were stored at −80°C.

Cell lysates (800 μg total protein in a volume of 400 μl lysis buffer; see above) were pre-cleared with recombinant Protein G Agarose (10 μl, cat #15920-010, Invitrogen). Anti-GFP saturated rProtein G Agarose was prepared by incubating 50/50 rProtein G Agarose/TBS slurry (800 μl) with anti-GFP (56 μg) and incubated for 2 hours on a rotation platform at 4°C. The precleared extracts were then incubated with the anti-GFP saturated rProtein G-Agarose on a rotating platform o/n at 4°C. The beads were vigorously washed three times with ice-cold lysis buffer. 2× loading buffer (30 μl; 125 mM Tris–HCl pH 6.8, 4% (w/v) SDS, 20% (w/v) Glycerol and 0.004% (w/v) BromPhenol Blue) was added to the beads and the samples were incubated at 95°C for 5 minutes. Proteins were resolved by 8% SDS-PAGE and subjected to immunoblotting.

The co-IP membranes were incubated with Flag antibody (1/3000, diluted in 1× PBS-T (0.1% (v/v) Tween in 1 × PBS), (M2; Stratagene)) or GFP antibody (1/3000, diluted in 1 × PBS-T, (JL- 8; BD Biosciences)), and subsequently with an HRP-conjugated goat anti-mouse secondary antibody (1/10000 in 1× PBS-T) using the Supersignal Chemoluminiscent substrate (Thermo scientific) for detection. For visualization of the immunostaining patterns, the membranes were immediately exposed in a Luminescent Image analyzer (LAS 3000, Fujifilm) for 1–15 minutes. Protein levels of Arnt, Miz-1, Hif1α and Hif2α upon their downregulation by specific siRNA were also analysed by immunoblotting using antibodies against Arnt [52], Miz-1 (Santa Cruz), Hif1α (Abcam), Hif2α (Abcam) and beta-actin (Abcam). CDKN2B expression was detected using the C-20 antibody from Santa Cruz (sc-612).

The assay was performed essentially as described [51]. Briefly, HEK293 cells were grown on cover slips and transfected with the expression plasmids Flag/Miz-1 and pECFP/Arnt using the Lipofectamine Reagent (Invitrogen). 24 hours after transfection, cells were fixed in 4% paraformaldehyde/PBS for 15 min at RT. Blocking was performed with 10% FBS/PBS for 15 min at RT. Cells were then subjected to incubation with anti-Flag [1:100 (M2; Stratagene) in 1% BSA/0.1% Triton-X100/PBS] and Texas-red conjugated goat anti-mouse antibody (Molecular Probes). Cells were washed 4× for 10 min with PBS and subjected to confocal laser scanning microscopy.

U2OS cells were cultured at 21% O₂ or reduced O₂ (1% O₂) for 4 or 24 hours, and ChIP analyses combined with quantitative PCR were performed as previously described [53] using antibodies against Arnt [52], Myc (Santa Cruz) or IgG (Jackson ImmunoResearch). PCR primers flanking the proximal promoter of the CDKN2B gene (−168/-19) were F: 5′- cgcatgcgtcctagcatctttg -3′ and R: 5′- gaattccgttttcagctgggcc -3′ giving a product of 149 bp. Distal CDKN2B promoter primers (−3874/-3744) were F: 5′- ggtgggccctaatccaatctgac -3′ and R: 5′- gcacatggccatcctactgctg -3′ giving a product of 130 bp (see [25] for nucleotide numbering of the CDKN2B promoter). PCR primers flanking the proximal promoter of the CDKN1A gene (+6/+106) were F: 5′- tgtgtgagcagctgccgaagtc -3′ and R: 5′- tgccgccgctctctcacct -3′ giving a product of 100 bp. Distal CDKN1A promoter primers (−2337 /-2158) were F: 5′- catccctatgctgcctgcttcc -3′ and R: 5′- cctgtctcctaccatccccttcct -3′ giving a product of 179 bp. Primers spanning the HRE site of the PGK-1 promoter (−75/+101) were F: 5′- gacagcgccagggagcaatg -3′ and R: 5′- gct ccggaggcttgcagaatg -3′ giving a product of 176 bp. The comparative Ct method was used for relative quantification of target DNA amplification using SYBR Green.

U2OS cells were transfected with control siRNA (10 nM; 5′- UUCUCCGAACGUGUCACGU -3′) or siRNA against Arnt (5′-GGAACAAGAUGACAGCCUATT -3′), Miz-1 (5′- GCCUCAUCAGCCUGCUGAATT -3′), Hif1α (5′- GAAGAACUAUGAACAUAAATT -3′) and Hif2α (5′- CGGAUAGACUUAUUGCCAATT -3′) (Qiagen) using the HiPerFect transfection reagent (Qiagen) according to the manufacturer’s instructions. The cells were cultured for 48 hours and then placed at 21% or 1% O₂ for 6 hours before harvested.

RNA was prepared from U2OS cells using Trizol (Life Technologies). RNA (1 μg) was reversely transcribed using Superscript II Reverse Transcriptase kit (Life Technologies) according to the manufacturer’s recommendations. TaqMan quantitative RT-PCR was performed using the ABI 7300 system with TaqMan master mix and pre-design primer/probes for human Arnt, Miz-1, Hif1α and Hif2α (Arnt: cat# Hs01121918_m1, Miz (ZBTB17) cat#: Hs01114794_g1, HIF1α cat#: Hs00153153_m1, HIF2α cat#: Hs01026142_m1, CDKN2B cat#: Hs00793225_m1, PGK-1 cat# Hs00943178-g1, HPRT cat#: 4326321E; Applied Biosystems). Hypoxanthinephophoribosyltransferase (HPRT) was used as an endogenous control in ∆∆CT analyses. All measurements were performed twice in duplicates.

One-way analysis of variance (ANOVA) followed by Bonferroni adjustment was used for multiple comparisons of data presented in Figures 2B, 3B, 4B and 5D. ANOVA with Dunnett’s T3 adjustment was used when variances were unequal (Figures 2C, 3A and C, 4C, 6A and B). Student’s t-test was used in Figures 2A, 5A, B and E.