Differential Regulation of HIC1 Target Genes by CtBP and NuRD, via an Acetylation/SUMOylation Switch, in Quiescent versus Proliferating Cells

Abstract

The tumor suppressor gene HIC1 encodes a transcriptional repressor involved in regulatory loops modulating P53-dependent and E2F1-dependent cell survival, growth control, and stress responses. Despite its importance, few HIC1 corepressors and target genes have been characterized thus far. Using a yeast two-hybrid approach, we identify MTA1, a subunit of the NuRD complex, as a new HIC1 corepressor. This interaction is regulated by two competitive posttranslational modifications of HIC1 at lysine 314, promotion by SUMOylation, and inhibition by acetylation. Consistent with the role of HIC1 in growth control, we demonstrate that HIC1/MTA1 complexes bind on two new target genes, Cyclin D1 and p57KIP2 in quiescent but not in growing WI38 cells. In addition, HIC1/MTA1 and HIC1/CtBP complexes differentially bind on two mutually exclusive HIC1 binding sites (HiRE) on the SIRT1 promoter. SIRT1 transcriptional activation induced by short-term serum starvation coincides with loss of occupancy of the distal sites by HIC1/MTA1 and HIC1/CtBP. Upon longer starvation, both complexes are found but on a newly identified proximal HiRE that is evolutionarily conserved and specifically enriched with repressive histone marks. Our results decipher a mechanistic link between two competitive posttranslational modifications of HIC1 and corepressor recruitment to specific genes, leading to growth control.

HIC1 (Hypermethylated in Cancer 1), a tumor suppressor gene frequently deleted or epigenetically silenced in human cancers, encodes a transcriptional repressor (7, 20, 65). A regulatory feedback loop between HIC1 and P53 has been deciphered. HIC1 is a direct target gene of P53 (5, 24, 65). HIC1 directly represses the transcription of SIRT1, a NAD⁺-dependent class III histone deacetylase (HDAC) that deacetylates and inactivates P53, thereby modulating P53-dependent DNA damage responses (8). We have shown that SIRT1 also deacetylates HIC1. In contrast to P53, this deacetylation activates HIC1 by strengthening its transcriptional repression potential (59).

Aside from P53, SIRT1 and HIC1 have also been implicated in a feedback regulatory loop with E2F1. E2F1 is a crucial activator of SIRT1 transcription in response to DNA damage, but SIRT1 binds and deacetylates E2F1 that inhibits E2F1-mediated gene activation (30, 66). In addition, E2F1 directly activates HIC1 (27) and HIC1 directly represses the E2F1 promoter in quiescent but not in G₁ human fibroblasts, which contributes to the growth suppression induced by serum deprivation (71). Thus, HIC1 is placed at the intersection of complex regulatory loops modulating p53-dependent and E2F1-dependent cell survival, growth control, and stress responses (15).

HIC1 encodes a sequence-specific transcriptional repressor with five Krüppel-like C₂H₂ zinc fingers mediating DNA binding to a HIC1 responsive element (HiRE), (C/G)NG(C/G)GGGCA(C/A)CC (48). To date, SIRT1, ATOH1 (a proneuronal transcription factor) (4), E2F1, CXCR7 (a receptor for the chemokine CXCL12) (62), and ephrin-A1 (a cell surface ligand for Eph receptors) (72) are the only characterized direct target genes of HIC1.

Our earlier work demonstrated that HIC1 contains two autonomous repression domains, an N-terminal BTB-POZ (16) and a central region that recruits CtBP corepressor complexes through a conserved GLDLSKK motif (17, 58). Notably, hypoxia promotes association of HIC1 with CtBP (58), whereas glycolysis inhibition by 2-deoxyglucose decreases it (70). Thus, through its interaction with the redox sensor CtBP, HIC1 can specifically link SIRT1 expression to free nuclear NADH levels. SIRT1 positively regulates HIC1 repressive activity by orchestrating with HDAC4 a switch between acetylation and SUMOylation, two mutually exclusive posttranslational modifications on the same lysine K314 in the conserved ψK³¹⁴xEP motif (59). Despite its functional importance, the mechanistic details including corepressor recruitment and gene-specific regulation through this acetylation/SUMOylation switch are not known.

In the present study, by yeast two-hybrid (Y2H) screening, we identified MTA1 as a novel corepressor for HIC1. The MTA family of proteins consists of evolutionarily conserved integral subunits of the NuRD (nucleosome remodeling and histone deacetylase) complex (29). We show that endogenous HIC1 interacts with MTA1 and MBD3, two specific subunits of the NuRD complex, and recruits it to the SIRT1 promoter in WI38 fibroblasts. We further demonstrate that the acetylation/SUMOylation switch on K314 regulates the interaction with MTA1 and RBBP4 and thus the recruitment of the NuRD complex. Indeed, SUMOylation-deficient or constitutively acetylated mutants exhibiting significantly reduced binding to MTA1 displayed a decreased repression in transient reporter assays. However, this cross talk between two posttranslational modifications (PTMs) on HIC1 K314 has no effect on the recruitment of CtBP by the central region. Consistent with the role of HIC1 in growth control, we demonstrate that CtBP and NuRD complexes are differentially recruited on known HIC1 target genes in quiescent versus growing human fibroblasts WI38. In particular, HIC1/MTA1 complexes cooccupy the promoters of two new direct target genes, Cyclin D1 and p57KIP2, only in quiescent cells. Furthermore, HIC1/CtBP and HIC1/MTA1 complexes are also differentially bound on two mutually exclusive HIC1 binding sites on the SIRT1 promoter in quiescent versus proliferating WI38 cells. Notably, the evolutionarily conserved HiRE identified here is specifically enriched with repressive histone marks in quiescent cells. Thus, our results provide a unique example of promoter-specific recruitment of repression complexes regulated through competitive posttranslational modifications of HIC1.

MATERIALS AND METHODS

Y2H screen.

Yeast two-hybrid (Y2H) screening was performed by Hybrigenics, Paris, France. For bait cloning, the BTB-central region of HIC1 (positions 1 to 422) encompassing the two autonomous repression domains was PCR amplified and cloned in frame with a C-terminal LexA DNA-binding domain in a Y2H vector. The bait construct was checked by sequencing the insert and was transformed in the L40ΔGAL4 yeast strain.

A human breast tissue random-primed cDNA library, transformed into the Y187 yeast strain and containing 10 million independent fragments, was used for mating. Neither toxicity nor the autoactivation of the bait was observed in small-scale prescreens. Then, the full-scale screen was performed in conditions ensuring a minimum of 50 million interactions tested to cover five times the primary complexity of the yeast-transformed cDNA library. A total of 79 million interactions were actually tested with HIC1. After selection on medium lacking leucine, tryptophane, and histidine, positive clones were picked, and the corresponding prey fragments were amplified by PCR and sequenced at their 5′ and 3′ junctions.

Cell lines, plasmids, and antibodies.

HEK293T, DAOY, and NIH 3T3 cells were maintained in Dulbecco modified Eagle medium (Gibco) supplemented with 10% fetal calf serum and nonessential amino acids. WI38 were purchased from the American Type Culture Collection (14 passages) and cultured in minimal essential medium (Gibco) supplemented with 10% fetal calf serum, nonessential amino acids, and sodium pyruvate as SKNMC cells. The pTL1 HIC1, pcDNA3-FLAG-HIC1, pcDNA3-FLAG-HIC1 L225A, K314R, and E316A expression vectors have been previously described (17, 58). The plasmid encoding a Myc-tagged MTA1 was kindly provided by Rakesh Kumar (M. D. Anderson Cancer Center, Houston, TX). The pcDNA3 FLAG-HIC1 K314Q vector was derived from the wild-type FLAG-HIC1 vector by PCR mutagenesis and verified by nucleotide sequencing.

Except for the anti-HIC1 2563 or anti-HIC1 325 (17) and anti-Mi2α (53) polyclonal antibodies, commercial antibodies of the following specificities were used: FLAG from Sigma (M2 monoclonal antibody F3165); MTA1 from Santa Cruz (goat polyclonal antibodies sc-9445 for WB and mouse monoclonal sc-17773X for chomatin immunoprecipitation [ChIP]); MBD3 from IBL, Gunma, Japan (mouse monoclonal antibody 10281); Myc tag from Tebu (mouse monoclonal antibody 51826); CtBP2 from BD Biosciences (mouse monoclonal antibody 612044); SUMO1 from Santa Cruz (rabbit polyclonal sc9060); HDAC4 from Abcam (ChIP grade rabbit polyclonal ab1437); cyclin D1 from Cell Signaling (mouse monoclonal antibody 2926); p57KIP2 from BD Pharmingen (A120-1 monoclonal antibody clone) or from Santa Cruz (polyclonal sc-rabbit antibody 8298), SIRT1 from Delta Biolabs (rabbit polyclonal DB083 for WB) and from Upstate (mouse monoclonal antibody 05-707 for ChIP); HSP60 from Santa Cruz (sc13115); CtBP1 from Santa Cruz (rabbit polyclonal antibody sc9060); PCNA from Dako (mouse monoclonal M0879); H4K16Ac (rabbit polyclonal antibody 07-329), H3K9me2 (rabbit polyclonal antibody 07-441), H3K27me2 (rabbit polyclonal antibody 07-452), and H3K27me3 (rabbit polyclonal antibody 07-449) from Upstate; H3K9me3 (rabbit polyclonal antibody ab8898) from Abcam; anti-acetyl histone H3 (rabbit polyclonal antibody 06-599) from Upstate; and anti-histone H3 (rabbit polyclonal ChIP-grade antibody ab1791) from Abcam.

Acetyl-K314 antibody generation.

The rabbit antibody against HIC1 acetylated on K314 was raised using a human HIC1 13-amino-acid peptide containing an acetylated lysine at position 314 (H2N-CLL YRW MK[Ac]H EPG L-CONH2). After high-pressure liquid chromatography purification, this peptide was coupled to carrier through its additional N-terminal cysteine and used to immunize two New Zealand white rabbits (Eurogentec, Seraing, Belgium). After multiple boosts and bleeds, the serum giving the best results by enzyme-linked immunosorbent assay against the acetylated peptide and by Western blotting of Cos-7 cells transfected with HIC1 and CBP/P300 was affinity purified by a two-step purification procedure. The crude serum was passed through a column containing the acetylated peptide coupled to the carrier. The sequence-specific and acetyl-sequence-specific antibodies retained on this first column were eluted and passed on a second column containing the nonacetylated peptide coupled to the carrier. The desired K314 acetyl-specific antibodies were recovered in the flowthrough.

Transfection, coimmunoprecipitation, repression assays, and small interfering RNAs (siRNAs).

Cells were transfected in Opti-MEM (Gibco) by the PEI (Euromedex) method as previously described in either 100-mm-diameter dishes (in vivo interaction in HEK293T cells) with 2.5 μg of DNA or in 12-well plates (repression assays) with 525 ng of DNA for the RK13 cells (58). Cells were transfected for 6 h and then incubated in fresh complete medium. For coimmunoprecipitation assays, at 48 h after transfection, cells were rinsed two times in cold phosphate-buffered saline (PBS) and lysed in cold IPH buffer (50 mM Tris [pH 8], 150 mM NaCl, 5 mM EDTA, 0.5% NP-40, protease inhibitor cocktail [Roche]). Cell lysates were cleared by centrifugation (14,000 rpm, 4°C for 30 min). The supernatants were incubated overnight with 2 μg of antibody. Then, protein A/G-Sepharose beads (Amersham Biosciences) were added for 30 min. The beads were washed three times with IPH buffer. Proteins were eluted by boiling in Laemmli loading buffer and separated by SDS-PAGE before Western blotting.

For repression assays, at 48 h after transfection, cells were rinsed in PBS and lysed with the Luc assay buffer (25 mM glycyl glycine [pH 7.8], 15 mM MgSO₄, 4 mM EGTA, 1% Triton X-100). Luciferase and β-galactosidase activities were measured by using, respectively, beetle luciferin (Promega) and the Galacto-Light kit (Tropix) with a Berthold chemiluminometer. After normalization to the β-galactosidase activity, the data were expressed as the Luc activity relative to the activity of pG5-Luc with empty control vector, which was given an arbitrary value of 1. The results represent the mean values and standard deviations from two independent transfections in triplicate.

For stable transfection, NIH 3T3 cells were transfected with the empty pcDNA3-Neo or the HIC1 expression vectors using the polyethyleneimine (PEI) method as described above. At 48 h after transfection, cells were grown for 10 days in selection medium containing 1 mg of G418 (Invitrogen)/ml and directly lysed in Laemmli loading buffer before Western blot analyses.

Specific siRNAs targeting human MTA1 (siGENOME SMART pool M-004127-02) and control siRNA (siGENOME nontargeting siRNA#1 D-001210-01) were obtained from Dharmacon. The transfection of siRNA was performed with Interferin reagent (Ozyme) for 16 h according to the manufacturer's protocol. After 62 h in starvation medium, followed by 10 h in medium with serum, the cells were collected for further analyses (Western blotting and ChIP-quantitative PCR [qPCR]).

Western blot.

Western blots were performed as previously described (17). The secondary antibodies were horseradish peroxidase-linked antibodies raised against rabbit, goat, or mouse immunoglobulins (Amersham).

Quantitative reverse transcription-PCR.

Total RNA was reverse transcribed by using random primers and MultiScribe reverse transcriptase (Applied Biosystems). Real-time PCR analysis was performed by Power SYBR green (Applied Biosystems) in a MX3005P fluorescence temperature cycler (Stratagene). The results were normalized with respect to 18S RNAs used as an internal control (71).

ChIP.

ChIP was performed according to published protocols with slight modifications. Briefly, formaldehyde was added directly to the cultured cells to a final concentration of 1% for 10 min at 37°C. The cross-linking was stopped by adding glycine to a final concentration of 0.125 M. After 5 min at 37°C, cells were lysed directly in the plates by resuspension in cell lysis buffer for 5 min. The samples were then pelleted, resuspended in nuclei lysis buffer, and sonicated to chromatins with an average size of 250 bp by using a BioRuptor (Diagenode, Liege, Belgium). After preclearing with a 50% slurry of protein A/G beads preincubated with salmon sperm DNA and bovine serum albumin for 4 h at 4°C, the chromatins were incubated with relevant antibodies, normal rabbit IgG, or with no antibodies overnight. The antibody-bound chromatin was then pooled down for 30 min with protein A/G beads, washed extensively, and eluted twice using 250 μl of elution buffer. After the addition of 20 μl of 5 M NaCl, the cross-linking was reversed by overnight incubation at 65°C. The immunoprecipitated DNAs, as well as whole-cell extract DNAs (input), were purified by treatment with RNase A and then proteinase K, followed by purification on Nucleobond Extract II (Macherey-Nagel).

For sequential ChIP experiments, we started with 8-fold more cells than in the single ChIP. After the first round of immunoprecipitation, the beads were pooled by centrifugation in Tris-EDTA buffer and incubated in 100 μl of elution buffer for 10 min at 65°C. After centrifugation, the supernatant was diluted in 900 μl of immunoprecipitation buffer, followed by incubation with the second antibody as in a single ChIP experiment.

The purified DNAs were used for PCR analyses using relevant primers for SIRT1, E2F1, ATOH1, Cyclin D1, P57KIP2, and GAPDH (62).

To quantify the ratio of acetylated histone H3 relative to total histone H3 on the SIRT1 distal promoter in WI38 cells transfected with control or siMTA1, real-time ChIP PCR assays were performed according to the Q² ChIP protocol extensively described by Dahl and Collas (13). Briefly, the cross-link and sonication steps were performed as described above, whereas magnetic beads were used for the chromatin immunoprecipitation (ChIP). qPCR analyses were performed with the Power SYBR green PCR master mix and antisense (TGCCTGGAGCACAGCGTTTCTATC) and sense (TTGGTTGCCTAAAGTCACGCAGGT) SIRT1 oligonucleotides.

RESULTS

HIC1 interacts with MTA1 and the NuRD complex in vivo.

To explore in more detail the HIC1 mediated repression mechanisms acting on its target genes in quiescent cells, we first searched for novel HIC1 corepressors by Y2H screening using the two autonomous repression domains of HIC1 (residues 1 to 422) as bait. One clone and three clones contained the complete coding sequences of human CtBP1 and CtBP2, respectively, a finding in close agreement with our previous work (17, 59). We also isolated two clones corresponding to residues 397 to 473 of MTA1 (metastasis-associated protein 1) an integral subunit of the NuRD (nucleosome remodeling and histone deacetylase) complex (3, 18, 29, 37, 38) (Fig. 1 A and 2A).

FIG. 1.

HIC1 interacts with MTA1 and recruits it to the SIRT1 promoter in vivo. (A) Scheme of the human HIC1 and MTA1 proteins. The BTB/POZ domain, the central region (CR), containing the CtBP-interaction domain (CID) and the acetylation/SUMOylation switch motif (ψK³¹⁴XEP), and the five C₂H₂ zinc fingers are shown. The bait included the two autonomous repression domains of HIC1 (amino acids 1 to 422). The domains in MTA1 include BAH, ELM, SANT, and the GATA-like zinc finger. The two first cysteine of the zinc finger were not present in the isolated prey (underlined lowercase letters). (B) In vitro, HIC1 interacts with MTA1. HEK293T transfected as indicated were lysed, immunoprecipitated, and immunoblotted with the indicated antibodies to detect MTA1 and HIC1. Totals of 2% of the whole-cell extracts were similarly analyzed (Input). (C) Co-occupancy of HIC1 and MTA1 on the SIRT1 promoter. Human WI38 fibroblasts were analyzed by single ChIP with the indicated antibodies. PCR amplifications were performed with primers flanking the functional HiREs identified in SIRT1. PCR with the 5′ promoter of GAPDH was used as an internal nonbinding control.

To confirm this interaction, we performed coimmunoprecipitation analyses in transiently transfected HEK293T cells. Myc-MTA1 was coprecipitated by FLAG-HIC1 demonstrating that full-length HIC1 and MTA1 proteins associate in vivo (Fig. 1B). The endogenous interaction between these proteins is difficult to detect (data not shown). However, ChIP assays in WI38 demonstrated that endogenous HIC1 and MTA1 proteins are bound to the two adjacent HiREs in the SIRT1 promoter, a direct HIC1 target gene in human fibroblasts (8) (Fig. 1C) and sequential ChIP confirmed their colocalization on SIRT1 (see Fig. 5C).

FIG. 5.

Promoter association analyses through sequential ChIP assays. (A) Preparation of quiescent G₀ and growing mid-G₁ WI38 cell populations. Subconfluent WI38 cells were maintained in medium without serum (Quiescent Cells) for 72 h. The cells were cultured for 10 h in medium with 10% serum to obtain growing cells synchronized in mid-G₁ phase. Whole-cell extracts were analyzed by Western blotting for PCNA, cyclin D1, and P57KIP2. In the latter case, the antibody detected the full-length P57KIP2 protein (arrow), as well as an ∼40-kDa product (asterisk). HSP60 was used as a loading control. (B) HIC1, SUMOylated proteins, MTA1, SIRT1, and HDAC4 are present on SIRT1 and ATOH1. Sequential ChIP were carried out on mid-G₁ WI38 cells with the indicated antibodies. PCR amplifications were performed as described for Fig. 1D. (C) Differential regulation of known (SIRT1, ATOH1, CXCR7, and E2F1) and newly identified (Cyclin D1 and P57KIP2) HIC1 target genes by CtBP and NuRD complexes in quiescent versus growing WI38 cells. Sequential ChIP analyses were performed to analyze the co-occupancy of HIC1 and MTA1 and of HIC1 and CtBP1 on the indicated target genes. As a control, sequential MTA1/HIC1 ChIP was also performed in quiescent cells.

NuRD complexes share a core HDAC/RbAp complex with the Sin3 complex but contain specific subunits including MTA1/2 and MBD3 (Fig. 2 A) (18, 28, 34, 37). In contrast, the NODE complex (Nanog and Oct4-associated deacetylase) contains HDAC1/2 and MTA1/2 but not MBD3 (35). Coimmunoprecipitation analyses in HEK293T cells showed that HIC1 interacts with MBD3-FLAG, as well as with MBD2-FLAG (Fig. 2B) (12, 34).

FIG. 2.

HIC1 interacts with the NuRD complex. (A) Scheme of human NuRD complexes. The core components (HDAC1/2, RBBP7/RbAp46, and RBBP4/RbAp48) shared with the Sin3 complex are shown in black, whereas NuRD-specific subunits (MTA1/2, p66, ATPase Mi2, and MBD3) are shown in light gray. (B) In vitro, HIC1 interacts with MBD3 and MBD2. HEK 293T were transfected as indicated, immunoprecipitated and analyzed by Western blotting. Totals of 2% of each whole-cell extract were similarly analyzed (Inputs). (C) HIC1 interacts with the NuRD complex. Total extracts from HEK 293T transfected with the FLAG-HIC1 or a control vector (FLAG) were immunoprecipitated by anti-FLAG antibodies coupled to agarose beads. Western blot analyses showed the coimmunoprecipitation of endogenous NuRD subunits with HIC1. (D) In vivo, endogenous HIC1 interacts with MBD3. DAOY cells, a medulloblastoma cell line expressing HIC1, were lysed, immunoprecipitated, and analyzed by immunoblotting. A shorter exposure of the input lane is shown on the right. Arrows indicate the MBD3 bands. (E) MBD3 binds the SIRT1 promoter. Human WI38 fibroblasts were analyzed by ChIP as described for Fig. 1D.

To confirm that HIC1 interacts with NuRD, HEK293T cells overexpressing FLAG-HIC1 were subjected to anti-FLAG immunoprecipitation. Two specific subunits, MTA1 and MBD3, as well as the common ATPase Mi2α, were coprecipitated only with FLAG-HIC1 (Fig. 2C). Furthermore, endogenous HIC1 and MBD3 interact together (Fig. 2D). Finally, ChIP assays in WI38 fibroblasts demonstrated that endogenous MBD3 bound the same region of the SIRT1 promoter than HIC1 and MTA1 (Fig. 2E).

Altogether, these findings demonstrate that HIC1 interacts with NuRD components and recruits it to the SIRT1 promoter.

HIC1 recruits NuRD independently of CtBP.

Previously, we demonstrated that the HIC1 central region is a trichostatin A (TSA)-sensitive repression domain relying on both CtBP-dependent and CtBP-independent mechanisms (17, 58). The CtBP repressor complex is a huge complex mediating several coordinated histone modifications (10, 56). To determine whether the recruitment of CtBP and NuRD complexes by HIC1 could be somehow linked, we performed coimmunoprecipitation experiments with the mutant HIC1 L225A. This CtBP-deficient mutant (58) displayed an ability to complex with MTA1 comparable to that of its wild-type counterpart (Fig. 3 A). Thus, the interactions of CtBP and NuRD complexes with HIC1 are independent.

FIG. 3.

SUMOylation on HIC1 K314 favors the interaction with MTA1. (A and B) HEK 293T were transfected with the indicated HIC1 mutants. The consequences on interaction with CtBP, as well as on potential PTMs, are shown. HIC1 was immunoprecipitated, and MTA1 coprecipitation was detected by Western blotting.

SUMOylation of HIC1 on lysine 314 positively regulates the interaction with the NuRD complex.

An acetylation/SUMOylation switch motif, ψK³¹⁴XE³¹⁶P, positively regulates HIC1 transcriptional repression potential by an unknown mechanism (59). In contrast to the L225A CtBP-deficient mutant, the K314R SUMOylation- and acetylation-deficient mutant is severely impaired in its interaction with MTA1, demonstrating that this interaction may be regulated by the posttranslational modification of HIC1 K314 (Fig. 3A). To augment these results, we also tested the FLAG-HIC1 E316A mutant since this mutation in the ψKXE consensus abolishes the SUMOylation of target proteins, as shown for HIC1 (59). Consistent with the results obtained with the K314R mutant, this E316A mutant only deficient for SUMOylation is impaired in its interaction with MTA1 (Fig. 3B).

Lysine 314 is embedded into a phylogenetically conserved MKHEPX₅DE motif similar to the KEPE (KXEPX_3-6E) motif, frequently superimposed on SUMOylation sites (19). Mutation of the proline (HIC1 P317A) (59) results in a significant decrease of the interaction with MTA1 (Fig. 4 A).

FIG. 4.

The acetylation/SUMOylation switch on HIC1 K314 regulates the interaction with two components of the NuRD complex (MTA1 and RBBP4) but has no effect on the recruitment of CtBP2. (A) A mutant protein, FLAG-HIC1 K314Q, mimicking a constitutively acetylated isoform of HIC1 is shown. The P317A mutant has been previously described. HEK 293T cells were transfected as indicated. HIC1 was immunoprecipitated and MTA1 coprecipitation was detected by Western blotting. (B) A similar experiment was performed with wt HIC1, the K314R and K314Q point mutants, and HA-RBBP4. (C) wt HIC1 and the constitutively acetylated K314Q mutant were similarly tested for interaction with CtBP2. (D) HEK 293T cells were transfected with wt, E316A, and K314R HIC1 with (+) or without (−) P300. Whole-cell extracts were analyzed by immunoblotting with the specific Ac-HIC1 K314, with FLAG used as a loading control, and with P300 antibodies to detect its overexpression. (E) Overexpression of Gal4-P300, which acetylates HIC1 on K314, impairs the interaction with MTA1. HIC1 was immunoprecipitated and MTA1 coprecipitation was detected by Western blotting.

The MTA1 coding sequence lacks a consensus SUMO interacting motif (SIM) (26, 57). Thus, the interaction between HIC1 and MTA1, albeit clearly favored by HIC1 SUMOylation, does not rely on a direct noncovalent interaction between SUMO and a SIM motif in MTA1.

We then turned our attention on RBBP4(RbAp48) since a glutathione S-transferase-SUMO-2 fusion protein pulled down the CoREST1 and NuRD complexes through direct binding to a nonconsensus SIM motif found, respectively, in CoREST1 and RBBP4 (45). In transient-transfection assays, HIC1 interacts with RBBP4 and, as for MTA1, this interaction is reduced but not abolished in the presence of the SUMOylation-deficient mutant K314R (Fig. 4B). This experiment argues against a direct interaction with HIC1 via its SUMOylation and the nonconsensus SIM motif of RBBP4.

Thus, the SUMOylation of HIC1 on K314 promotes the interaction with NuRD in concert with SUMO-independent interactions with MTA1 and RBBP4.

Acetylation of HIC1 on lysine 314 disrupts the interaction with MTA1 and curtails HIC1's transcriptional repression potential.

Since K314 is competitively modified by acetylation or SUMOylation (58), we sought to determine whether these two mutually exclusive posttranslational modifications could play opposite roles in the interaction between HIC1 and MTA1. We thus constructed the K314Q mutant in which the lysine is substituted by a glutamine to mimic a constitutively acetylated lysine (1, 64). Unlike wild-type HIC1 or, to a lesser extent, the SUMOylation-deficient mutants, this K314Q mutant is significantly affected in its ability to interact with MTA1 and RBBP4 (Fig. 4A and B). In contrast, the K314Q mutation has no effects on the interaction with CtBP2 (Fig. 4C), confirming that the recruitment of the CtBP and NuRD complexes are independent. Thus, the K314Q mutation mimicking constitutive acetylation specifically impairs the recruitment of NuRD.

To confirm these results, we perturbed the acetylation pathway by overexpressing the histone acetyltransferase CBP/P300, which acetylates HIC1 on various lysine residues including K314 (59), as directly shown by a specific immunoblot against HIC1 acetylated on K314 (Fig. 4D). In agreement with the K314Q mutant, increase of CBP/P300 level decreases the interaction between HIC1 and MTA1 (Fig. 4E).

To monitor the impact of this mutation on HIC1 transcriptional repression, we performed luciferase reporter assays using HIC1-Gal4 fusions and a 5×Gal4-tk-luc reporter plasmid (59). In this assay, the K314Q BTB-RC-Gal4 chimera repressed transcription almost 5-fold less efficiently than the wild-type HIC1 (data not shown).

Collectively, these data demonstrate that the acetylation/SUMOylation switch on K314 regulates the transcriptional repression activity of HIC1 since acetylation decreases the NuRD complex recruitment, whereas SUMOylation promotes it.

MTA1, HIC1, and the SIRT1/HDAC4 complex colocalized on some HIC1 target genes.

The acetylation/SUMOylation switch on HIC1 K314 relies on a complex between SIRT1 and HDAC4 (59), raising the question of whether these proteins are present together with MTA1 on HIC1 target genes. Thus, we performed sequential ChIP experiments in WI38 cells, with a first round of ChIP with HIC1 antibodies, followed by elution of the complexes and a second round of immunoprecipitation using various antibodies. Because SUMOylation is a highly dynamic and reversible posttranslational modification (22, 23, 25, 63) and in order to optimize specific proteins detection, we used WI38 cells synchronized in mid-G₁ phase by serum starvation and release for 10 h (54) (Fig. 5 A), instead of a mixed population of asynchronously growing cells (Fig. 1C). As shown in Fig. 5B, HIC1, MTA1, SIRT1, and HDAC4 specifically bound together with SUMOylated proteins on SIRT1 promoter (8), as well as on the ATOH1 enhancer (4), but not on GAPDH (8). The absence on E2F1 promoter was expected since HIC1 directly repressed it but only in quiescent cells through the recruitment of SWI/SNF complexes (61, 71). In conclusion, our results suggest that a SUMOylation/MTA1-dependent pathway is used by HIC1 to repress some target genes.

HIC1/MTA1 and HIC1/CtBP complexes are differentially recruited on four HIC1 direct target genes, SIRT1, ATOH1, E2F1 and CXCR7.

HIC1 is involved in cell growth control, notably in the growth arrest induced by serum deprivation (71). To investigate the potential role of the HIC1/MTA1 and HIC1/CtBP complexes in this pathway, we prepared G₀ and mid-G₁ cell populations of WI38 fibroblasts (54) (Fig. 5A). As controls, in the mid-G₁ cell population we observed higher levels of the proliferative marker PCNA and of cyclin D1, a G₁-specific cyclin (Fig. 5A). Chromatins prepared from these two cell populations were analyzed by sequential ChIP experiments with HIC1 antibodies followed by MTA1, CtBP1, or rabbit IgG as a negative control. In quiescent cells, no amplification was detected for HIC1/CtBP on SIRT1, ATOH1, E2F1, and CXCR7, whereas HIC1/MTA1 complexes bound only the ATOH1 enhancer (Fig. 5C). As a control, similar results were obtained when we carried out the sequential ChIP with the MTA1 antibodies first (Fig. 5C). In growing mid-G₁ WI38 cells, the situation was similar for ATOH1 but strikingly different for the SIRT1 and CXCR7 promoters where HIC1/MTA1 and HIC1/CtBP complexes are both detected on the previously described HiRE (8, 62, 71). These complexes did not bind on E2F1 in G₁ growing cells or in quiescent cells, suggesting that the repression of E2F1 relied on SWI/SNF complexes (61, 71).

Thus, HIC1/MTA1 and HIC1/CtBP complexes are differentially recruited to subsets of HIC1 target genes in normal fibroblasts depending of the cell growth status.

Cyclin D1 and P57KIP2 are new direct target genes of HIC1/MTA1 complexes in quiescent cells.

Through gene profiling experiments, we recently reported that Cyclin D1, a key regulator of G₁ phase, is downregulated in asynchronously growing U2OS cells infected by an adenovirus expressing HIC1 (62). A HiRE around position −1623 is bound by HIC1 proteins in in vitro gel retardation assays (data not shown). In silico analyses also identified several HiRE in P57KIP2 promoter (14, 60; data not shown). We focused on P57KIP2 since it is a direct target gene of another factor isolated in our Y2H screen (Van Rechem et al., unpublished data). After analyzing the chromatins of quiescent versus mid-G₁ growing WI38 cells, we found that the HiREs in Cyclin D1 and P57KIP2 are occupied by HIC1/MTA1 (or MTA1/HIC1) but not by HIC1/CtBP only in quiescent cells (Fig. 5C). In mid-G₁ cells, neither of these two HIC1 repressive complexes is observed on P57KIP2 and Cyclin D1 (Fig. 5C). Cyclin D1 mRNAs and protein levels are low in quiescent WI38 cells and reached a maximum ca. 10 h after serum stimulation (54). We have confirmed these results by RT-qPCR analyses showing a 2.5-fold increase of Cyclin D1 mRNAs (see Fig. 8C) and by Western blotting (Fig. 5A).

FIG. 8.

Epigenetic marks deposited on the two mutually exclusive HiRE of SIRT1 in quiescent versus mid-G₁ cells. (A) Changes in enrichment of silencing epigenetic marks on the mutually exclusive HiRE of the SIRT1 promoter in quiescent versus mid-G₁ WI38 cells. ChIP were performed with antibodies directed against various histone posttranslational modifications. PCR amplifications were performed with primers amplifying the two types of HiREs in SIRT1 or the −1623 HiRE in Cyclin D1. (B) Whole-cell extracts from quiescent or mid-G₁ WI38 cells were analyzed by Western blotting for the expression of SIRT1 and HIC1. In each case, HSP60 was used as a loading control. (C) Expression levels of SIRT1 and Cyclin D1 in quiescent versus mid-G₁ WI38 fibroblasts were assessed by qPCR, and the values were normalized to 18S RNA. The P values are indicated (*, P < 0.05). (D) Knockdown of endogenous MTA1 in mid-G₁ growing WI38 cells increases the level of acetylated H3 at the SIRT1 distal promoter and upregulates SIRT1 expression. WI38 cells were transfected with nontargeting or MTA1 siRNA and synchronized in mid-G₁ phase (see Fig. 5A). MTA1 and SIRT1 levels were analyzed by immunoblotting with actin as a loading control. ChIP analyses were performed with total acetylated H3 and H3 antibodies, followed by quantitative PCR with primers amplifying the two distal HiREs.

Thus, the lack of HIC1/MTA1 and HIC1/CtBP repressive complexes on Cyclin D1 promoter in mid G₁ nicely correlates with its increase of expression culminating during this phase of the cell cycle (54).

The situation is more complex for P57KIP2, since it is a member of the CIP/KIP family of CDK inhibitors (CKIs) that are positive and negative regulators of G₁ phase progression (2, 9, 31, 47, 55). P57KIP2 expression is high during G₀ and G₁ phases and decrease during progression from G₁ to S phase (55). Nevertheless, we observed by qRT-PCR analyses a strong 14-fold decrease in P57KIP2 mRNAs levels (data not shown) between our quiescent and mid-G₁ WI38 cells despite similar protein levels (Fig. 5A). These results suggest that direct transcriptional regulation of P57KIP2 by HIC1 and other transcription factors is somehow counterbalanced by posttranscriptional mechanisms to yield similar amounts of P57KIP2 proteins in the G₀ and mid-G₁ phases (Fig. 5A). Indeed, in addition to its major role as an inhibitor of cyclin E-cdk2 complexes, P57KIP2 is also required to promote the assembly of cyclin D1-cdk4 complexes in early G₁ (2, 31, 55).

Our sequential ChIP data clearly demonstrate that Cyclin D1 and P57KIP2 are two new direct target genes of HIC1/NuRD complexes, but specifically in quiescent cells. To confirm these results, we overexpressed wild-type (wt) HIC1 and L225A (CtBP-deficient) and K314Q (NuRD-deficient) mutants in NIH 3T3 cells and then compared their ability to repress these two newly identified HIC1 target genes. Cyclin D1 and P57KIP2 protein levels decreased upon HIC1 overexpression (Fig. 6 A). We used two different anti-P57KIP2 antibodies that both detected in NIH 3T3 cells extract two bands whose intensity is significantly decreased when wt HIC1 is expressed (Fig. 6A and data not shown). The upper band with an apparent molecular mass of 57 kDa corresponds to the full-length mouse P57KIP2 protein (36). The lower ∼40-kDa band does not correspond to the related P27KIP1 protein or to any described P57KIP2 isoform and might represent a cleavage product of P57KIP2. Notably, the same band is detected in WI38 (Fig. 5A).

FIG. 6.

The K314Q point mutant mimicking constitutive acetylation is unable to repress Cyclin D1 and P57KIP2 in NIH 3T3 cells. (A) HIC1 but not the K314Q mutant markedly repressed the expression of P57KIP2 and cyclin D1 proteins. NIH 3T3 cells were transfected with the pcDNA3 FLAG-HIC1 (wt), FLAG-HIC1 L225A, and FLAG-HIC1 K314Q expression vectors containing a neomycin resistance gene. After 10 days in medium containing 1 mg of G418/ml, whole cells extracts were analyzed by SDS-PAGE. HIC1 overexpression and upregulation of endogenous P57KIP2 and cyclin D1 proteins were analyzed by immunoblotting. The antibody directed against P57KIP2 detected the full-length protein (arrow) and an ∼40-kDa product (asterisk), exactly as in WI38 cells (Fig. 5A). Actin was used as a loading control. (B) SUMOylation-dependent association of HIC1 and MTA1 on P57KIP2. Human SKNMC cells were transfected with an expression vector for the de-SUMOylase SSP3 and analyzed by single ChIP. PCR amplifications were performed on P57KIP2 and on GAPDH, as a control.

In contrast, cyclin D1 and P57KIP2 were markedly upregulated in NIH 3T3 expressing L225A and K314Q mutants as predicted by their biochemical properties.

Finally, we carried out ChIP assays on SK-N-MC Ewing/PNET cells overexpressing or not the de-SUMOylase SSP3 (SENP2) that inhibits HIC1 SUMOylation (59). In control SK-NM-C cells, endogenous HIC1 proteins (62), as well as MTA1, are detected on P57KIP2 promoter (Fig. 6B). Upon overexpression of SSP3, HIC1 still occupies the P57KIP2 promoter, whereas MTA1 is no more detected (Fig. 6B). Thus, the SUMOylation of HIC1 and perhaps the SUMOylation of other components might be required to stably recruit NuRD at the P57KIP2 promoter.

Taken together, our results demonstrate that Cyclin D1 and P57KIP2 are two new HIC1 direct target genes in quiescent but not in mid-G₁ growing fibroblasts.

HIC1/MTA1 and HIC1/CtBP complexes are present on mutually exclusive sites on the SIRT1 promoter in quiescent versus growing cells.

SIRT1 encodes an NAD-dependent class III deacetylase playing critical roles in cell metabolism and in the regulation of cell survival upon stresses (30, 41, 43). Accordingly, its expression is tightly regulated through protein and mRNA stabilization, through the control of mRNA via microRNAs, and through direct transcriptional effects implicating, among many others, HIC1 (8, 41, 70). In close relationship with the present study, glycolysis inhibition in primary human fibroblasts has been found to increase SIRT1 transcription through the dissociation of the HIC1/CtBP complex (70). As a consequence of all of these regulatory mechanisms, the SIRT1 protein level increases when serum-starved human lung fibroblasts or MEFs are stimulated with serum (52), a result that we confirmed in WI38 cells (Fig. 8C). Thus, the lack of HIC1/CtBP and HIC1/MTA1 complexes on the previously described HiRE (8, 70) in SIRT1 in quiescent cells, in contrast to their presence in growing cells, presented a conundrum (Fig. 5C).

However, a reappraisal of these sites by phylogenetic footprinting of the human, murine, and rat SIRT1 genes demonstrates that even though their arrangement is grossly conserved, their sequences and localizations are highly divergent (Fig. 7 A). Upon closer inspection of the SIRT1 promoters, we identified a new putative HiRE, close to the transcription initiation site and downstream of a conserved E2F1 binding site (66). Notably, this “proximal” −8 HiRE site is perfectly conserved in the three SIRT1 genes in striking contrast with the “distal” −1116 and −1039 HIC1 binding sites (8) (Fig. 7A).

FIG. 7.

HIC1 recruits the CtBP and NuRD complexes on mutually exclusive sites of the SIRT1 promoter in quiescent versus mid-G₁ growing WI38 cells. (A) Identification of a new, evolutionarily conserved, HIC1 binding site on the SIRT1 promoter. In silico analyses demonstrated that the two previously described “distal” HiRE (black boxes) are not well conserved and identified a new proximal HiRE conserved in the three genomes (gray box). The transcription start site (+1, bent arrow) and the ATG initiation codon (+54) are indicated. The “proximal” HIC1 site is adjacent to a conserved E2F1 binding site (plain oval), whereas a site not found in the murine genome is shown as a dotted oval. (B) HIC1 occupies the distal and proximal sites on the SIRT1 promoter in a mutually exclusive way. ChIP was performed on WI38 cells growing in complete medium (0 h) or after 24, 48, or 72 h of serum starvation. (C) Differential binding of HIC1/MTA1 and HIC1/CtBP complexes on the SIRT1 distal and proximal promoters. Sequential ChIP analyses were performed on quiescent and mid-G₁ WI38 cells, followed by PCR with primers amplifying the two sites in SIRT1.

To explore the functional relevance of these sites, we first performed single ChIP in WI38 cells over 72 h of serum starvation. In asynchronously growing cells, HIC1 bound the distal HiRE sites as described previously (8, 70) (Fig. 7B). After 24 h without serum, HIC1 no longer occupies the distal sites, which correlates well with the concomitant transcriptional upregulation of SIRT1 described in similar stress conditions in primary human fibroblasts (70). Strikingly, after 48 h, HIC1 reappears on the SIRT1 promoter but on the newly identified proximal site, where it remains stably bound till the end of the starvation period (72 h) (Fig. 7B). These experiments demonstrate that the proximal site is functional but also reveals an unexpected complex and dynamic binding of HIC1 on SIRT1 promoter upon serum withdrawal.

When we analyzed by sequential ChIP the presence of HIC1/MTA1 and HIC1/CtBP complexes on these two types of HiRE in the SIRT1 promoter, an exactly inverse correlation was observed in quiescent versus mid-G₁ WI38 cells, in close agreement with our single ChIP experiments (Fig. 7B and C). Indeed, in quiescent cells both HIC1/MTA1 or MTA1/HIC1 and HIC1/CtBP complexes are bound to the phylogenetically conserved proximal HiRE but not to the distal sites. Conversely, in proliferating cells, these two repressive complexes are found on the distal HiRE sites but not on the proximal site (Fig. 7C and see Fig. 9).

FIG. 9.

Model. (A) HIC1 repressive complexes and epigenetic marks found on the distal and proximal SIRT1 promoter in quiescent versus mid-G₁ WI38 fibroblasts. The brown circle represents a yet-undefined component of the NuRD complex that could directly interact with SUMO covalently linked to HIC1 K314. RBBP4 interacts directly with the MTA1 GATA zinc finger. (B) Dynamic binding of the HIC1/CtBP and HIC1/NuRD complexes on the two mutually exclusive HiRE on SIRT1. (a) In asynchronously growing WI38 fibroblasts, HIC1 recruits CtBP and NuRD complexes on the previously described distal HiRE (8). (b) After 24 h of serum starvation, when SIRT1 mRNAs and proteins levels increase (42, 71), HIC1 is no more detected on the SIRT1 promoter. (c) After 48 h, HIC1 is detected again on SIRT1 but on the proximal site identified herein. The presence of the NuRD and CtBP complexes was not investigated. (d) However, as shown by sequential ChIP, both complexes are present on the conserved proximal site after 72 h in starvation medium. (e) After serum replenishment for 10 h, the HIC1/CtBP and HIC1/NuRD complexes are relocalized on the distal sites.

In conclusion, our results demonstrate that HIC1/MTA1 and HIC1/CtBP repressive complexes bind to mutually exclusive sites in SIRT1 promoter depending on whether the human WI38 fibroblasts are quiescent or proliferating.

Silencing histone modifications, hypoacetylated H4K16 and H3K27me3, are enriched at the proximal HiRE of the SIRT1 promoter in quiescent WI38 cells.

Given that SIRT1 protein levels are lower in quiescent cells (52) (Fig. 8 B), we sought to determine whether silencing histone modifications could be specifically enriched on this new “proximal” HiRE of the SIRT1 promoter as a consequence of stable repression mechanisms. To that end, we performed single ChIP assays for various histone modifications considered as activating (H4K16ac) or as repressive marks (H3K9me2, H3K9me3, H3K27me2, and H3K27me3) on quiescent or mid-G₁ growing WI38 cells.

In addition to SIRT1, we also analyzed Cyclin D1 as an HIC1 target gene that is clearly transcriptionally upregulated after serum replenishment and whose expression culminates in mid G₁ (54). As expected, these experiments highlighted major changes on the regions encompassing the two types of HiREs in the SIRT1 promoter, although we failed to detect any H3K9me2 and H3K27me2 epigenetic marks (data not shown). In quiescent WI38 cells, the distal HiRE is devoid of repressive histone marks but enriched with activating marks such as acetylated H4K16. In contrast, two typical repressive marks, hypoacetylated H4K16 and H3K27me3, are found on the “proximal” HiREs in quiescent cells (Fig. 8A). Although NuRD can facilitate Polycomb binding (42), the presence of H3K27me3 on this region is probably not related to the recruitment of NuRD by HIC1. Indeed, Cyclin D1 promoter, which is bound by HIC1/MTA1 complexes in quiescent cells (Fig. 5C), is not enriched in H3K27me3 but rather in another repressive mark, H3K9me3 (Fig. 8A).

When cells are replenished with serum, the repressive marks are lost on Cyclin D1, as shown by the lack of H3K9me3 and on the SIRT1 proximal promoter as shown by the lack of H3K27me3 and the reappearance of acetylated H4K16 (Fig. 8A).

In order to relate this observation to protein levels, we analyzed cell extracts by Western blotting. HIC1 protein levels remain constant, whereas the level of SIRT1 protein expression is significantly lower in quiescent cells which corresponds to the association of the proximal HiRE with silencing epigenetic marks (52) (Fig. 8B). In addition, RT-qPCR analyses of SIRT1 RNA levels showed a 1.4-fold increase in mid-G₁ cells compared to quiescent cells (Fig. 8C). Thus, the mutually exclusive binding of HIC1 repressive complexes on the two types of HiRE in the SIRT1 promoter described here (Fig. 7 and 9) highlights an additional dynamic and direct transcriptional mechanism to the complex cellular regulation of SIRT1 (30, 41). Similarly, the 2.5-fold upregulation of Cyclin D1 mRNAs (Fig. 8C) and increase in protein levels in mid-G₁ growing WI38 cells (Fig. 5A) (54) is in good agreement with the presence of HIC1/MTA1(NuRD) complexes on the Cyclin D1 promoter in quiescent cells and its absence in mid-G₁ growing cells (Fig. 5C).

To establish a direct functional link between HIC1/MTA1 complexes and histone modifications, we then attempted to knock down endogenous HIC1 in WI38 by RNA interference approaches. Unfortunately, we failed to sufficiently reduce endogenous HIC1 protein levels using an siRNA approach. As an alternative, in mid-G₁ growing cells, when HIC1/MTA1 complexes bind the distal SIRT1 promoter, MTA1 knockdown results in a 60% increase of acetylated histone H3, as well as in SIRT1 protein levels (Fig. 8D).

In conclusion, our results identify HIC1 as a new transcriptional regulator involved in the correct timing of Cyclin D1 expression within G₁ phase. In addition, they reveal a new layer in the complex cellular regulation of SIRT1 through the identification of a conserved HIC1 binding site specifically enriched for histone silencing marks only in quiescent cells and the dynamic relocalization of HIC1 repression complexes to the distal sites during cell cycle reentry.

DISCUSSION

In this report, we established connections between two competitive posttranslational modifications on HIC1 K314, acetylation and SUMOylation and the differential recruitment of NuRD repressive complexes. These complexes bound in quiescent cells two novel HIC1 target genes Cyclin D1 and p57KIP2 and a newly identified HIC1 binding site in SIRT1.

SUMOylation is a central mechanism regulating activity of transcription factors (22, 23, 25, 68). HIC1 is one among few examples, including the transcription factors Sp3 and MEF2 (22), as well as the transcriptional coactivators PGC-1α (51) and PARP1 (40), where a lysine is competitively modified by acetylation and SUMOylation. In most instances, SUMOylation acts as a major switch converting the transcription activator into a repressor or attenuating the coactivator function, mainly by direct binding between SUMO and multiple corepressor complexes (44, 45). In contrast, mutation of the SUMOylation site of HIC1 has been shown to blunt its transcriptional repression potential by an unknown mechanism (59). Here, we identified MTA1 as a novel HIC1 corepressor, regulated through these competitive PTMs. MTA1, MTA2, and the related MTA3 proteins are mutually exclusive, specific subunits of the NuRD complex. Their homologous N-terminal parts contain several functional domains interacting with core complex components (49). Their more divergent C-terminal parts are involved in specific interactions with small motifs found in transcriptional repressors. For example, BCL6, a BTB/POZ and zinc finger transcriptional repressor whose structural organization is highly similar to HIC1, interacts directly with the C-terminal part of MTA3 through a KKYK³⁷⁹ motif localized in its central region (1, 21). Likewise, the transcriptional cofactors FOG1/2, BCL11B, and transcription factors of the Sall1 family interact with MTA1 through an N-terminal MSRRKQxKP motif (32). These motifs are highly divergent, but a common theme is their regulation by posttranslational modifications. Indeed, phosphorylation by protein kinase C (PKC) of serine 2 in the Sall1 repression motif disrupts its interaction with MTA1 (33). Similarly, p300-mediated acetylation of K379 in the KKYK³⁷⁹ motif abolishes the interaction with the cell-type-specific MTA3 subunit, which is indispensable for BCL6 to repress genes involved in plasma cell differentiation, notably PRDM1 (1, 21, 46). HIC1 is unique since the interaction with two components of the NuRD complex, MTA1, initially isolated as a “prey” in the Y2H screen, and RBBP4, is impaired but not abolished by acetylation of K314 and favored by its deacetylation/SUMOylation induced by a SIRT1/HDAC4 dual-deacetylase complex (59). Although clearly favored by SUMOylation, the interaction between HIC1 and MTA1 or RBBP4 is not due exclusively to a direct interaction between SUMO bound on HIC1 K314 and a consensus SUMO-interacting motif in MTA1 or even the unusual SIM motif recently described in RBBP4 (45), as demonstrated by the interaction still observed with the K314R mutant (Fig. 3A and 4B). Thus, it would be important to identify which NuRD core unit or associated cofactor is involved in direct interaction with HIC1. Nevertheless, an important implication of the cross talk between these two competitive posttranslational modifications of HIC1 and their resulting positive or negative impact on corepressors recruitment could be the establishment of rapid cycles of repression-derepression of HIC1 target genes in response to signaling pathways. SUMOylation is a very labile posttranslational modification concerning only a small fraction of the substrate but with very important biological consequences (25). In agreement with this model (25), the rapid but transitory SUMOylation of a small portion of HIC1 could favor its SUMO-dependent incorporation into the NuRD complex. Then, SUMO-independent interactions with MTA1 and RBBP4 would stabilize the HIC1/NuRD complex on relevant target genes. A rapid dissociation of the complex could be achieved through the acetylation of HIC1, which disrupts the interaction with MTA1 and RBBP4. Future challenges will be to identify the signaling pathways regulating on the one hand the SUMOylation of HIC1, notably via the HDAC4/SIRT1 complex and on the other its acetylation, since they have opposite functional consequences on HIC1 transcriptional repression potential.

As a transcription factor recruiting several proteins, HIC1 could control different biological functions through distinct biochemical mechanisms. As a paradigm, in germinal center B cells, BCL6 interacts (i) with the NCoR/SMRT or BCoR corepressors through its BTB/POZ domain to target genes involved in proliferation and survival while they undergo immunoglobulin somatic hypermutation; (ii) with the MTA3/NuRD through its central region to target PRDM1, thus mediating a differentiation blockade; and (iii) finally with CtBP through the BTB/POZ and central region to mediate its own negative autoregulation (11, 39). When we addressed the role of HIC1 in growth control through sequential ChIP experiments, we demonstrated that HIC1/MTA1 and HIC1/CtBP complexes are differentially recruited in quiescent and normal growing cells on SIRT1 and ATOH1 but are not found on E2F1. In contrast, HIC1/MTA1 complexes are specifically bound in quiescent cells on HIC1 binding sites (HiRE) identified by in silico analyses in the promoter of a key regulator of the G₁ phase, Cyclin D1, but not in mid-G₁ growing cells, which is fully in agreement with the exquisitely regulated early G₁-phase expression of Cyclin D1. The situation is less clear for P57KIP2, a member of the CIP/KIP family of CDK inhibitors (2, 55). P57KIP2 is the only CIP/KIP gene whose targeted disruption results in developmental abnormalities, leading to death soon after birth (67). Thus, in addition to its role in cell cycle regulation, P57KIP2 is essential for normal development, exactly as HIC1 since homozygous Hic1^−/− mouse embryos display severe developmental defects culminating in perinatal death (6). The repression of P57KIP2 that we observed between quiescent and growing fibroblasts could thus be only partially related to cell cycle regulation but also to a yet-unidentified developmental process. The same holds true for the repression by HIC1 of the proneuronal gene ATOH1 observed here in WI38 fibroblasts compared to its functionally relevant repression during normal cerebellar development, whose loss due to HIC1 promoter hypermethylation participates in medulloblastoma development (4, 50). Alternatively, the HIC1-mediated repression of the P57KIP2 CDK inhibitor and of Cyclin D1, albeit apparently contradictory, could be nevertheless relevant to the correct progression of G₁ phase. Indeed, at low levels P57KIP2 binds to the CDK-cyclin heterodimers and promotes their assembly, whereas at high levels it abrogates CDK activity (2, 9, 31, 47, 55). Thus, it would be interesting to investigate the binding of HIC1 to the promoters of P57KIP2 and possibly of other CIP/KIP family members at earlier times after serum addition. These experiments would indicate whether, in addition to posttranslational modifications, direct transcriptional regulation by HIC1 could also contribute to the regulation of CIP/KIP levels after mitogenic stimulation. Nevertheless, our results identify a new link between HIC1 and cell growth through its posttranslational modifications, which regulate the specific recruitment of NuRD complexes to Cyclin D1 and probably to other growth control regulators, including P57KIP2.

SIRT1, the first HIC1 target gene characterized (8), is a key regulator of cancer, senescence, metabolism, and longevity. SIRT1 is involved in an intricate network of transcriptional and posttranslational regulations, notably with P53, E2F1, and HIC1 (15). The two adjacent HiREs previously described in the SIRT1 promoter have been implicated in the P53-dependent response to DNA damage after etoposide treatment (8), as well as in the response to glycolysis inhibition through the redox sensor CtBP (70). Our sequential ChIP analyses in synchronized growing mid-G₁ WI38 fibroblasts confirmed these results and further demonstrated the presence of HIC1/MTA1 and HIC1/CtBP complexes (Fig. 5 and 8D). It is worth noting that our coimmunoprecipitation experiments with the various point mutants in HIC1 central region affecting the interaction with NuRD or CtBP suggested that both complexes can bind simultaneously the HiRE (Fig. 3 and 4). However, these distal sites are not well conserved in rodent genomes. In contrast, our results revealed a hitherto-unsuspected new proximal HiRE that is evolutionarily conserved and bound by HIC1/MTA1 and HIC1/CtBP complexes in quiescent cells with the concomitant appearance of stable epigenetic repressive marks, hypoacetylated H4K16 and H3K27me3. When quiescent cells reenter the cell cycle after serum addition, the relocalization of HIC1 repression complexes on the distal sites is correlated with the loss of these repressive epigenetic marks, a transcriptional upregulation of SIRT1 mRNAs and an increase in SIRT1 protein levels (Fig. 7, 8, and 9). Thus, these two types of HIC1 binding sites might perform different functions. The distal HiRE could participate in a dynamic species-specific fine-tuning of SIRT1 expression in growing cells to counterbalance the increase in SIRT1 expression induced by metabolic stress or by E2F1 after genotoxic stress. Similarly, SIRT1 deacetylase activity is increased by SUMOylation in many mammals but not in mice due to a K-to-R mutation in the SUMOylation consensus, demonstrating that the complex SIRT1 regulation exhibits salient differences even between closely related species (69). In contrast, the phylogenetically conserved proximal HiRE identified here could be generally used to stably lock SIRT1 transcription when cells cease to divide due to cell cycle withdrawal, as observed with most of the cells in adult organisms that are kept in G₀ or in terminally differentiated cells that cannot divide anymore. Indeed, HIC1 is highly expressed in normal tissues and primary cells but rapidly inactivated in the first steps of tumorigenesis (65).

The present study has unveiled a link between two competitive posttranslational modifications of HIC1 and transcriptional repression through gene-specific corepressor recruitment participating in the complex regulation of genes crucial for cell growth and survival. Future studies will be required to understand how various external stimuli or stresses can interfere with this cross talk between posttranslational modifications of HIC1 and thereby interfere with its tumor suppressor function.