Higher-Order Chromatin Regulation and Differential Gene Expression in the Human Tumor Necrosis Factor/Lymphotoxin Locus in Hepatocellular Carcinoma Cells

Abstract

The three-dimensional context of endogenous chromosomal regions may contribute to the regulation of gene clusters by influencing interactions between transcriptional regulatory elements. In this study, we investigated the effects of tumor necrosis factor (TNF) signaling on spatiotemporal enhancer-promoter interactions in the human tumor necrosis factor (TNF)/lymphotoxin (LT) gene locus, mediated by CCCTC-binding factor (CTCF)-dependent chromatin insulators. The cytokine genes LTα, TNF, and LTβ are differentially regulated by NF-κB signaling in inflammatory and oncogenic responses. We identified at least four CTCF-enriched sites with enhancer-blocking activities and a TNF-responsive TE2 enhancer in the TNF/LT locus. One of the CTCF-enriched sites is located between the early-inducible LTα/TNF promoters and the late-inducible LTβ promoter. Depletion of CTCF reduced TNF expression and accelerated LTβ induction. After TNF stimulation, via intrachromosomal dynamics, these insulators mediated interactions between the enhancer and the LTα/TNF promoters, followed by interaction with the LTβ promoter. These results suggest that insulators mediate the spatiotemporal control of enhancer-promoter associations in the TNF/LT gene cluster.

INTRODUCTION

Chromosomal regions harboring different tissue-specific or cellular-state-specific gene clusters may be influenced by long-range regulatory elements and higher-order chromatin organization (45, 53, 60). Recent studies suggest that transcriptional regulatory elements, such as enhancers, promoters, and chromatin insulators, contribute to gene activation and inactivation via genome accessibility and chromosomal interactions (8, 18). Among these, chromatin insulators are boundary elements that partition the genome into chromosomal subregions, probably through their ability to block interactions between enhancers and promoters when positioned between them (enhancer-blocking effect) (7, 17, 41). However, the precise mechanisms responsible for the enhancer-blocking effect and the relationship with long-range chromatin interactions remain unclear (47, 49). The CCCTC-binding factor CTCF is a highly conserved 11-zinc-finger protein that plays crucial roles at insulator sites (44). CTCF is also reported to function in transcriptional activation (62, 73) and repression (16, 36). In the IGF2/H19 locus, CTCF binds to the differentially methylated region (DMR) of the H19 gene to form a predicted chromatin loop structure (6, 22, 42). Genome-wide analyses identified the distribution of the putative CTCF-binding sites and their consensus sequences (4, 27, 28, 69). We and other groups recently determined that CTCF is enriched with cohesin in at least 14,000 sites on the human genome (46, 54, 65). CTCF and cohesin cooperatively form compact chromatin loops, leading to the colocalization of gene promoters and their common enhancer in the human apolipoprotein gene locus (40). CTCF has been reported to interact with nuclear substructures (71, 72), chromatin remodeling factors (26, 33), RNA polymerase II (10), and CTCF itself (34, 72), as well as undergoing several posttranslational modifications of the protein (12, 29, 37, 70).

Inflammation involves the activation of a highly coordinated gene expression program (43). The tumor necrosis factor (TNF) superfamily members, TNF (initially termed TNF-α), lymphotoxin α (LTα, also termed TNF-β), and lymphotoxin β (LTβ), are major proinflammatory cytokines that mediate inflammatory responses in autocrine/paracrine manners (63). TNF and LTα form homotrimers and act as soluble ligands for the TNF receptor. In contrast, LTβ forms a heterotrimer with LTα and functions as a membrane-bound ligand for the LTβ receptor. In addition to their physiological roles, the aberrant or unbalanced expression of these cytokines is linked to pathological conditions, such as tissue damage/remodeling (38), metabolic diseases (14, 20), and cancer development (19, 23). Hepatic TNF expression is closely related to steatohepatitis (64), and LTβ expression is significantly involved in liver regeneration (3) and hepatocellular carcinomas (HCCs) (23, 67). The TNF/LT genes are clustered within the major histocompatibility complex (MHC) class III region on human chromosome 6p21.3, which is the most gene-dense region of the human genome (68). Interestingly, it is reported that NF-κB does not directly interact with the proximal human TNF promoter (9, 15, 59) and that NF-κB activation induced by TNF treatment influences expression of the TNF/LT genes, resulting in the amplified inflammatory response (25). Several DNase-hypersensitive sites, generally suggestive of the presence of regulatory elements, have been found in the TNF/LT locus (5, 50, 56, 58). However, a transcriptional mechanism and higher-order chromatin regulation in the human TNF/LT locus are unknown.

Investigation of the TNF/LT locus identified at least four CTCF/cohesin-enriched insulators and a TNF-responsive TE2 enhancer in human hepatic cells. These CTCF-bound sequences possessed enhancer-blocking activities, and one of the insulators was located between the early-inducible LTα/TNF promoters and the late-inducible LTβ promoter. Chromosome conformation capture (3C) analyses determined that after TNF stimulation, these CTCF-bound insulators initially associated with the TE2 enhancer and the LTα, TNF, and LTβ promoters, followed by a persistent interaction with the TC3 insulator, the TE2 enhancer, and the LTβ promoter. These late-phase interactions were consistent with the formation of a place in which the late-inducible LTβ gene was transcriptionally active. TNF stimulation thus induces dynamic changes in higher-order chromatin organization of the overall locus, together with differential expression of the TNF/LT genes. Based on our findings that insulators mediate the spatiotemporal control of enhancer-promoter interactions, we propose a dynamic chromatin conformation model and enhancer-blocking mechanism mediated by insulators in the TNF/LT locus.

MATERIALS AND METHODS

Cell culture.

Hep3B, HCT116, and HeLa cells were cultured in a 1:1 mixture of Dulbecco's modified Eagle's minimum essential medium and Ham's F-12 nutrient medium (DMEM/F12; Sigma) supplemented with 10% (vol/vol) fetal bovine serum (FBS). NeHepLxHT cells were cultured in DMEM/F12 supplemented with 10% (vol/vol) FBS, 10⁻⁷ M dexamethasone, 10⁻⁷ M insulin, and 50 μg/ml G418. For TNF stimulation, Hep3B and NeHepLxHT cells were treated with recombinant human TNF-α (210-TA; R&D Systems) at concentrations of 5 ng/ml and 0.5 ng/ml, respectively. For inhibition of NF-κB signaling, BAY11-7082 (10 μM) was added to the medium for 1 h before treatment of the cells with TNF for 0.5 or 1 h.

ChIP and quantitative PCR (qPCR) analysis.

Hep3B and NeHepLxHT cells were cross-linked with 1% formaldehyde at 37°C for 10 min. Crude cell lysates were sonicated to generate DNA fragments of 200 to 500 bp. Chromatin immunoprecipitation (ChIP) was performed with anti-CTCF (07-729; Millipore), anti-RAD21 (ab992; Abcam), anti-acetylated histone H3 (06-599; Millipore), anti-acetylated histone H4 (06-866; Millipore), anti-p65 (sc-372; Santa Cruz), anti-p300 (sc-585; Santa Cruz), or anti-RNA polymerase II (phosphor-S5) antibodies (ab5131; Abcam) or with control rabbit IgG (sc-2027; Santa Cruz) (26). Cells were cross-linked for an additional 10 min when anti-p65 and anti-p300 antibodies were used.

DNA enrichment in ChIP samples was determined using qPCR analysis with an ABI Prism 7300 system (Applied Biosystems) and SYBR green fluorescence. The threshold was set to cross a point where PCR amplification was linear, and the cycle number required to reach the threshold was recorded and analyzed using the Microsoft Excel software program. PCR was performed using precipitated DNA and the input DNA. Primer sequences are listed in Table S1 in the supplemental material. Other antibodies used were anti-lamin A/C (sc-7292; Santa Cruz).

Electrophoretic mobility shift assay (EMSA).

The CTCF protein was synthesized using a coupled in vitro transcription/translation reaction with the TNT T7 Quick system (Promega), according to the manufacturer's protocol. For supershift assays, the reaction mixture was combined with 1 μl anti-CTCF antibodies (612148; BD Biosciences) (40). The sequences of the probes were as follows: H19 DMR, 5′-TGG CAC GGA ATT GGT TGT AGT TGT GGA ATC GGA AGT GGC CGC GCG GCG GCA GTG CAG GCT CAC ACA TCA CAG CCC GAG CCC GCC CCA ACT-3; TC1, 5′-TCT CCA GCA CTT CTT GCT CAG GCA GTA CCC AAA GGG GCC GCC TGG GAG CAG CAG AGA CCA GGC CCA AAG CTG CGG GCT TAC AAC AGG TTA GCC ATC CCA-3′; TC2, 5′-AGA CCC TGG TGT CCT CTC TGG CCT TAT TTA CTC CTG GTC CTC TGC CAG CCC TGC CAC CAG ATG GCC TTC TAA CTC CTT GGT TGA AAG GCC CAT CTC ATT C-3′; TC3, 5′-CCC GGT ACA GAG AGC TGC GCA GCG TGA CCG AGC GG CCC TGG GGG TCC CCG CCG CCA GGG GGC GCC CGG CCC CGG TAG CCG ACG AGA CAG TAG AGG-3′; TC4, 5′-CTT CAC CCA GGT CTC TCC AGA GAG CCT CAG GCC GCT GCC TTT ACT TAG TTC TGT GTT CAA TGC CAG AAT GCT GCC TCC TAC AGG AAG TCC ACC TGT ATT GCC CAC ACC TCC T-3′; negative control, 5′-TGG CAA AAA GAA AGG ACA GGG CTG CAA GGA GAG TAC AGA CAT GTG CTG GTG AGT GCA CTG TCT GCA TAG TTA CAC CAG AGC ATC TTA TCA ATC AGA AAC TTA TC-3′.

Luciferase reporter assay.

The reporter vector pIHLE consisted of the luciferase gene driven by the mouse H19 promoter (−818 to +6 from the transcription start site), simian virus 40 (SV40) enhancer, and a 1.8-kb AatII-HindIII fragment containing the H19 DMR insulator. The plasmid pIHLIE was constructed by inserting the 1.8-kb H19 DMR fragment between the luciferase gene and the enhancer. pIHLTE plasmids were constructed by inserting fragments of about 200 bp, including TC1, TC2, TC3, and TC4, between the luciferase gene and the enhancer (pIHLTE-1F/1R, -2F/2R, -3F/3R, and -4F/4R, respectively). For pIHLET, TC fragments were inserted downstream of the enhancer in pIHLE (pIHLET-1F/1R, -2F/2R, -3F/3R, and -4F/4R). To prepare pIHLTE with mutations (pIHLTE-1 M, -2 M, -3 M, and -4 M), base substitutions were introduced in CTCF consensus sequences at the TC1, TC2, TC3, and TC4 sites using a PCR-based mutagenesis method.

The reporter vector pPL consisted of the SV40 promoter and the luciferase gene and is identical to the pGL3-Promoter vector (Promega). pTPL, pAPL, and pBPL contained the TNF promoter (−1044 to +54 from the transcription start site), LTα promoter (−924 to +43 from the transcription start site), and LTβ promoter (−971 to +12 from the transcription start site), respectively, instead of the SV40 promoter of pPL. TE1 and TE2 sequences were PCR amplified and inserted upstream of pPL, pTPL, pAPL, and pBPL (pTE1-PL, pTE2-PL, pTE1-TPL, pTE2-TPL, pTE2-APL, and pTE2-BPL). The primer sequences used to prepare the TE1 and the TE2 sequences were as follows: TE1-S, CCT GTG GCT GGA TGA AAT CT; TE1-AS, CCT GGG CAA CAA AGT GAG AC; TE2-S, CCA GGG GAG TTG TGT CTG TAA; TE2-AS, GCA GTT CGG TTC CTT GTT CT.

Reporter vectors (0.05 pmol) were transfected into Hep3B cells (1.0 × 10⁵ cells) in a 12-well plate, using FuGene6 reagent (Roche Applied Science), and analyzed using a luciferase reporter assay system (Promega) after 24 h. For dual luciferase activities (26), values are shown as means and standard deviations of the results from at least three independent experiments.

qRT-PCR.

Total RNA was isolated from cultured cells with TRIzol (Invitrogen). The cDNA synthesis used 2 μg of total RNAs that was reverse transcribed using a High Capacity cDNA reverse transcription kit (Applied Biosystems), according to the manufacturer's instructions. Quantitative PCR was performed using an ABI Prism 7300 system (Applied Biosystems) and SYBR green fluorescence. Each experiment was performed at least three times. The relative fold enrichment was quantified by normalization to β-actin or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene expression. Primer sequences are listed in Table S1 in the supplemental material.

siRNA-mediated knockdown.

Small interfering RNAs (siRNAs) for GL3, CTCF, and Rad21 were used as previously reported (40). RELA silencer select validated siRNA (s11914; Ambion) was used for p65 knockdown. siRNAs were transfected using the Lipofectamine RNAiMAX reagent (Invitrogen) for 48 h.

3C assay.

For the chromosome conformation capture (3C) assays (21, 52), formaldehyde-cross-linked chromatin from Hep3B and NeHepLxHT cells was digested with DpnII overnight, followed by ligation with T4 DNA ligase at 16°C for 4 h. To prepare control templates for standard curves, a bacterial artificial chromosome spanning the TNF/LT locus RPCI11.C-47E16 was digested with Sau3AI, which is insensitive to Dam methylase, followed by random religation. After reversing the cross-links, genomic DNA was purified by phenol extraction and ethanol precipitation. The ligated products were assessed using qPCR with an ABI Prism 7300 system (Applied Biosystems) and Thunderbird SYBR qPCR Mix (Toyobo). The efficiency of DpnII digestion was evaluated after the entire 3C treatment using qPCR to amplify uncut fragments spanning the DpnII site. More than 80% of the individual restriction sites were digested in these experiments. The 3C-qPCR data were normalized to a loading control, using internal primers located in the TNF/LT gene locus. We gained similar results after normalization with internal primers located in GAPDH (data not shown). The relative frequencies of interactions between the reference and its physically close site in the control state were finally normalized to 1. Examples of the calculation for relative interacting frequencies are described in Results. Statistical analysis was performed using Student's t test for the results of more than three independent experiments. Primer sequences are listed in Table S1 in the supplemental material.

Immunofluorescence analysis.

Cultured human cells were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) for 10 min at room temperature. Fixed cells were rinsed three times in PBS for 5 min and permeabilized with PBS containing 0.2% Triton X-100 and 0.5% normal goat serum (NGS) for 5 min on ice. Cells were rinsed three times in PBS containing 0.5% NGS for 5 min and then incubated with rabbit anti-p65 (sc-372; Santa Cruz) for 60 min followed by secondary donkey Cy3-conjugated or Alexa Fluor 488-conjugated antibodies for 60 min. Labeled cells were washed three times in PBS for 10 min each. Samples were analyzed using a fluorescence microscope system (Orca-ER1394; Olympus).

Patients and histological assessment.

A total of 38 patients (male, 29; female, 9) with HCC, who had undergone tumor resection at the National Cancer Center Hospital, Tokyo, Japan, between May 2003 and December 2005, were enrolled in the present study. The median patient age and follow-up period were 63 years and 1,719 days, respectively. Among the 38 HCC patients, 12 were immunologically positive for hepatitis C virus (HCV) infection, and 16 for persistent hepatitis B virus (HBV) infection (hepatitis B virus surface antigen positive), and 10 were negative for both HCV and HBV infection. Histological examination of noncancerous liver tissue samples revealed findings compatible with chronic hepatitis in 22 and cirrhosis in 9 and no remarkable histological findings in 7. The 38 HCCs were histologically classified into 3 well-differentiated, 27 moderately differentiated, and 8 poorly differentiated tumors. All patients were followed for more than 100 days. Clinical and pathological profiles were obtained from the medical records of the patients. This study was approved by the Ethics Committee of the National Cancer Center, Tokyo, Japan, and written informed consent was obtained from all patients.

IHC.

Immunohistochemistry (IHC) for TNF and LTβ was performed using a polymer-based method with the Envision+Dual Link system-horseradish peroxidase [HRP] (DK-2600 Glostrup; Dako). Sources and dilutions of primary antibodies were as follows: anti-TNF-α (ab9579), 1:100, Abcam; anti-LTβ (ab64835), 1:50, Santa Cruz Biotechnology. Formalin-fixed, paraffin-embedded serial tissue sections (4 μm) were placed on silane-coated slides for IHC. Sections cut through the maximum tumor diameter were selected for IHC evaluation. The sections were deparaffinized and rehydrated in xylene and grade-diluted ethanol (50 to 100%) and submerged for 20 min in 0.3% hydrogen peroxide with absolute methanol to block endogenous peroxidase activity. Antigen retrieval for TNF and LTβ was carried out by heating in target retrieval solution (Tris-EDTA buffer, pH 9; Dako Cytomation) at 121°C for 10 min by a pressure cooker. After protein blocking, the sections were incubated with each primary antibody at room temperature for 1 h, followed by incubation with Envision+Dual Link reagent at room temperature for 30 min, and visualized using 3,3-diaminobenzidine tetrahydrochloride as a chromogen. Finally, the sections were counterstained with hematoxylin. Sections were gently rinsed in PBS between incubation steps. The primary antibody was omitted from the reaction sequence as a negative control.

All sections were evaluated by two pathologists, Y. Kanai and H. Ojima, with no knowledge of any clinical or pathological information. Immunoreactivities of TNF and LTβ were defined as follows: negative, no cytoplasmic staining was observed or the intensity of cytoplasmic staining was lower than that for noncancerous hepatocytes within the same section in more than 50% of cancer cells; positive, the intensity of cytoplasmic staining was equivalent to or higher than that of noncancerous hepatocytes in more than 50% of cancer cells.

Statistical analysis.

Differences between groups were analyzed using Student's t test. A P value of <0.05 was considered statistically significant.

RESULTS

Distribution of CTCF-enriched sites in the human TNF/LT gene cluster.

CTCF-enriched sites in the human TNF/LT gene region were investigated by checking several genome-wide CTCF-binding profiles available on websites and in our published data (40, 65). At least four CTCF-enriched sites (TC1, TC2, TC3, and TC4) were identified in this locus and were conserved among the cells tested (Fig. 1A; see also Fig. S1A in the supplemental material). There were no probe sets for the TC2 site in genome tiling arrays because of the presence of frequent repeat sequences (shown by asterisks in Fig. S1A in the supplemental material). Interestingly, TC3 was located between the TNF and LTβ gene promoters, forming the possible boundary between these adjacent chromosomal subregions.

Fig 1.

CTCF-enriched sites in the human TNF/LT gene cluster locus. (A) CTCF-enriched sites in the TNF/LT locus on human chromosome 6p21.3. In addition to the NFKBIL1, LTα, TNF, LTβ, and LST1 genes, a newly identified TE2 enhancer is indicated by a red oval. Based on genome-wide CTCF-binding profiles available from websites and our published data (see Fig. S1A in the supplemental material), four enriched sites were designated TC1, TC2, TC3, and TC4. NC is used as a negative control, and TE1 is a site with no enhancer activity. (B) Direct binding of CTCF to TC sequences. Predicted CTCF-binding sequences within TC1, TC2, TC3, and TC4 sites are indicated, together with the 20-bp consensus motif (red). For EMSAs, radiolabeled duplex probes of approximately 100 bp for each TC site were incubated with anti-CTCF antibodies and synthesized CTCF. Solid and open arrowheads indicate CTCF DNA and the supershifted complexes, respectively. The H19 DMR insulator and an intergenic unrelated sequence (NC) were used as controls. (C and D) Existence of CTCF and the cofactor cohesin RAD21 at TC sites. Chromatin immunoprecipitation analyses were carried out with anti-CTCF and anti-RAD21 antibodies and control IgG, followed by quantitative PCR with specific primers for each TC site in Hep3B cells (C) or NeHepLxHT cells (D).

Based on previous reports (28, 69), each TC site contained a 20-bp consensus CTCF-binding motif (Fig. 1B). To determine if CTCF bound directly to these TC sequences, we performed electrophoretic mobility shift assays (EMSAs) using radiolabeled duplex probes of approximately 100 bp for each TC site and the in vitro transcribed/translated CTCF protein. Similar to the DMR insulator of the H19 gene used as a control (40), the TC probes formed complexes with CTCF and were further supershifted by anti-CTCF antibodies. In contrast, negative-control (NC) probes, which had sequences located downstream of the NFKBIL1 gene, did not bind to CTCF. In addition, competition assays using mutated TC probes carrying base substitutions within the consensus motif showed that mutated probes did not bind to the CTCF protein (see Fig. S1B and C in the supplemental material), indicating that CTCF specifically bound to the TC sequences.

In order to clarify the localization of CTCF and the cofactor cohesin RAD21 in hepatic cells, we performed chromatin immunoprecipitation (ChIP) analyses using anti-CTCF and anti-RAD21 antibodies, followed by quantitative PCR (qPCR) (Fig. 1C and D). We used standard cell lines: Hep3B, which originates from human HCC, and NeHepLxHT, which is a telomerase-immortalized human neonatal hepatocyte line (51). Both CTCF and RAD21 bound to the TC sites but not to the NC site. RAD21 was relatively enriched with CTCF at TC1 in the TNF/LT locus. The CTCF enrichment at the TC sites in Hep3B cells may be remarkable due to the high expression of this gene (see Fig. S1D in the supplemental material) compared with that in NeHepLxHT cells.

Differential regulation of TNF/LT genes under TNF stimulation.

To examine the transcriptional regulation of the TNF/LT genes, we performed quantitative reverse transcription (RT)-PCR (qRT-PCR) analyses with Hep3B and NeHepLxHT cells stimulated by TNF-induced NF-κB activation (Fig. 2A; see also Fig. S2A and B in the supplemental material). Expression of LTα and TNF mRNAs was markedly increased in Hep3B cells 1 h after stimulation, but LTβ mRNA was not simultaneously induced. Moreover, TNF expression seemed to be variable after the 1-h peak, while LTα and LTβ expression did not peak until 24 h after TNF treatment. Early induction of the LTα and TNF genes also occurred in NeHepLxHT cells, with subsequent expression of the LTβ gene. The patterns of TNF/LT expression differed between these cell lines, probably due to the constitutively low activation of the NF-κB pathway in Hep3B cells (see Fig. 4A) (11, 55).

Fig 2.

Differential regulation of TNF/LT genes under TNF stimulation. (A) Effect of TNF stimulation on TNF/LT expression in Hep3B cells. qRT-PCR analyses were performed with Hep3B cells under TNF treatment. (B) Nuclear translocation of NF-κB induced by TNF stimulation. The subcellular localization of the p65 subunit of the NF-κB heterodimer was analyzed by immunofluorescent staining of TNF-stimulated Hep3B cells, together with the use of BAY11-7082, an inhibitor of NF-κB activation. (C and D) NF-κB-dependent expression of the TNF/LT genes. TNF-induced expression of the TNF/LT genes was examined by qRT-PCR analyses in Hep3B (C) or NeHepLxHT (D) cells in combination with NF-κB inhibition. *, P < 0.05; **, P < 0.01; ***, P < 0.005.

Fig 4.

Characterization of TNF-responsive enhancer in the human TNF/LT locus. (A) Enhancer activity of TE2. The luciferase reporter vectors pPL, pTPL, pAPL, and pBPL contained the SV40 promoter, TNF promoter, LTα promoter, and LTβ promoter, respectively. The candidate enhancers TE1 and TE2 were inserted in these vectors upstream of the promoter. Hep3B cells were transfected with the reporter vectors and treated with TNF for 3 h (solid bars). Luciferase activities were normalized to basal pPL, pTPL, pAPL, and pBPL. The values are given as means and standard deviations of the results from more than three independent experiments. P, SV40 promoter; TP, TNF promoter; AP, LTα promoter; BP, LTβ promoter. (B to E) The chromatin state of the TE2 enhancer in TNF-stimulated Hep3B cells. ChIP assays were performed with antibodies against p65/NF-κB (B), p300 (C), RNA polymerase II (D), or acetylated histone H3 (E) or H4 (F). The MCP1 enhancer (ME) was used as a positive control. The values are given as means and standard deviations of the results from more than three independent experiments. *, P < 0.05; **, P < 0.01; ***, P < 0.005.

Nuclear translocation of NF-κB is critical for its activation (24), and we therefore investigated its subcellular localization under TNF stimulation, using immunofluorescent staining of p65, a subunit of the NF-κB heterodimer (Fig. 2B; see also Fig. S2C in the supplemental material). Cytoplasmic p65 translocated to the nucleus at 30 min after stimulation, and this translocation was inhibited by the addition of BAY11-7082, a specific inhibitor of IκBα phosphorylation (48). The translocated p65 was found to decrease at 1 h after the stimulation (see Fig. S2D in the supplemental material). The expression status of the TNF/LT genes was analyzed in parallel using qRT-PCR analyses (Fig. 2C and D). TNF-induced expression of TNF, LTα, and LTβ was attenuated by NF-κB inhibition. Since the use of BAY11-7082 had cytotoxic effects at late time points after TNF stimulation, we carried out siRNA-mediated knockdown of p65 (see Fig. S2G and H in the supplemental material). The induction of the TNF, LTα, and LTβ genes was consistently inhibited by depletion of p65, indicating that the TNF/LT genes are regulated by NF-κB in the TNF-treated hepatic cells. Expression of the neighboring NFKBIL1 gene was unaffected by the stimulation. TNF treatment caused no significant cell damage throughout the study (see Fig. S2E and F in the supplemental material). Thus, the TNF/LT genes are differentially induced by TNF-activated NF-κB signaling.

CTCF-dependent enhancer-blocking activity in the TNF/LT gene locus.

Previous studies demonstrated that the H19 DMR insulator contains multiple CTCF-binding sites, which are essential for enhancer-blocking activity (6, 22, 26). Luciferase reporter assays were performed with Hep3B cells to test the enhancer-blocking effects of TC1, TC2, TC3, and TC4 (Fig. 3). The presence of TC1, TC2, TC3, and TC4 between the enhancer and promoter reduced the luciferase activities to approximately 60% of those for the control pIHLE vector (pIHLTE-1F, pIHLTE-2F, pIHLTE-3F, and pIHLTE-4F). TC sequences in the opposite direction showed similar results (pIHLTE-1R, pIHLTE-2R, pIHLTE-3R, and pIHLTE-4R), indicating that the TC sites possess enhancer-blocking activities that are independent of the orientation of the sequences. To exclude the possibility that the TC sites exhibit silencer-like activities, the TC sequences were placed downstream of the enhancer (pIHLET-1F, pIHLET-1R, pIHLET-2F, pIHLET-2R, pIHLET-3F, pIHLET-3R, pIHLET-4F, and pIHLET-4R). Luciferase activity was not reduced by TC sites in this position, suggesting that TC sites do not possess silencer-like functions. The use of mutant TC sites lacking CTCF-binding function, as described above (see Fig. S1B and C in the supplemental material), demonstrated no enhancer-blocking effects (pIHLTE-1M, pIHLTE-2M, pIHLTE-3M, and pIHLTE-4M), further suggesting that the insulator activities of the TC sites depend on CTCF. These results suggest that TC1, TC2, TC3, and TC4 are functional insulators.

Fig 3.

CTCF-dependent enhancer-blocking activity of TC sequences. pIHLTE plasmids were constructed by inserting fragments of approximately 200 bp containing wild-type or mutant-type TC (lacking the CTCF binding function) between the promoter and the enhancer in pIHLE. The H19 DMR insulator was used as a control. For pIHLET, TC fragments were inserted downstream of the enhancer in pIHLE. The luciferase activities from pIHLE were normalized to 100. The values are given as means and standard deviations of the results from more than three independent experiments. Luc, luciferase gene; P, H19 promoter; Enh, SV40 enhancer; DMR, H19 DMR insulator; TC1-TC4, CTCF-enriched sites; MT1 to MT4, the mutant TC sequences. *, P < 0.05; **, P < 0.01; ***, P < 0.005.

Characterization of a TNF-responsive hepatic enhancer in the human TNF/LT locus.

In order to understand the overall regulatory mechanisms in the TNF/LT locus, we investigated the role of transcriptional enhancers in hepatic cells. Based on several DNase-hypersensitive sites in the locus (56), modified histones, p300 binding, previously reported enhancers (HSS-9 and HSS+3) in mouse T cells (58), and κB-responsive elements conserved among humans, mice, and rats (30, 31), we chose two candidates, named TE1 and TE2, which were located about 3.5 kb upstream of the LTα gene and just downstream of the TNF gene, respectively (Fig. 1A; see also Fig. S1A in the supplemental material). Luciferase reporter assays were performed with Hep3B cells to determine if TE1 and TE2 act as enhancers (Fig. 4A). Compared to the control (pPL) and TE1 (pTE1-PL), TE2 significantly increased transcription from the SV40, TNF, LTα, and LTβ promoters (pTE2-PL, pTE2-TPL, pTE2-APL, and pTE2-BPL), probably because of the constitutively low activation of NF-κB in Hep3B cells. Under TNF stimulation, these promoter activities were further elevated. These results indicate that TE2 has a TNF-responsive enhancing effect on the TNF/LT gene promoters. In addition, the effect of TE2 on the LTα promoter seemed to be weaker than that on the TNF promoter. The TNF-inducible enhancer activities of TE2 were also detected in other cell lines (see Fig. S3A in the supplemental material).

NF-κB p65 cooperates with histone acetyltransferase p300 (74), which functions as a transcriptional coactivator that accumulates in active enhancer elements (61). To validate the role of TE2 as an active enhancer, we investigated recruitment of p65 and p300 to TE2 by TNF stimulation in Hep3B cells, using ChIP-qPCR assays (Fig. 4B and C). A previously demonstrated enhancer of the MCP-1 gene (ME) was used as a control (57). Recruitment of p65 and p300 to TE2 occurred at 0.5 h after TNF stimulation. Interestingly, RNA polymerase II (Pol II) and acetylated histone H4 were also significantly enriched at TE2 (Fig. 4D to F). In contrast, histone H3 acetylation showed no remarkable changes (Fig. 4E). It was previously reported that various stimuli, such as serum, interleukin 1β (IL-1β), gamma interferon (IFN-γ), and TNF induced the acetylation of histone H4 but not histone H3 (2, 13, 32). Similar data were obtained in NeHepLxHT cells (see Fig. S3B in the supplemental material). These results indicate that TE2 is an active enhancer, which has four putative κB-binding motifs (see Fig. S3C in the supplemental material), under TNF-stimulated conditions in hepatic cells.

CTCF and the cofactor cohesin are involved in transcriptional regulation in the TNF/LT gene cluster.

RNA interference-mediated knockdown in Hep3B cells was used to determine if CTCF and cohesin, which are enriched at the TC insulators, were involved in transcriptional regulation in the TNF/LT locus. Western blot and qRT-PCR analyses showed that CTCF and RAD21 were depleted at both the protein and RNA levels (Fig. 5A and B). ChIP-qPCR confirmed that the amounts of CTCF and RAD21 were significantly reduced at each TC site in the knockdown cells (see Fig. S4A and B in the supplemental material). The effect of the knockdown on the constitutively low activation of the TNF/LT genes in Hep3B cells was tested by qRT-PCR analyses (Fig. 5C and D). The loss of CTCF reduced TNF expression and increased LTβ expression, while RAD21 depletion increased NFKBIL1, TNF, and LTβ expression, suggesting that CTCF and cohesin have overlapping but certain distinct roles. Indeed, cohesin was reported to be able to behave as a transcriptional regulator, independent of CTCF (46, 54, 65).

Fig 5.

CTCF-mediated insulators are involved in transcriptional regulation in the TNF/LT gene cluster. (A and B) RNA interference-mediated knockdown of CTCF (A) and the cofactor cohesin RAD21 (B). Western blot and qRT-PCR analyses were carried out with Hep3B cells. As previously demonstrated (40), more than two distinct siRNAs against CTCF or RAD21 and control siRNAs were used in the experiments. (C and D) Effects of CTCF and RAD21 knockdown on the transcriptional status of the TNF/LT genes. Using qRT-PCR analyses, the transcriptional levels of these genes were analyzed relative to that of β-actin and were normalized with the control GL3. (E) Effect of CTCF knockdown on TNF/LT expression in TNF-stimulated NeHepLxHT cells. CTCF siRNAs were introduced into NeHepLxHT cells for 48 h, followed by TNF treatment for the indicated time period. Values are given as means and standard deviations of the results from more than three independent experiments. *, P < 0.05; **, P < 0.01; ***, P < 0.005.

We also analyzed the effects of CTCF knockdown on TNF/LT genes in TNF-treated NeHepLxHT cells in which the TNF/LT genes are normally silenced (Fig. 5E). The loss of CTCF reduced TNF expression and accelerated LTβ induction in the stimulated cells (Fig. 5E; see also Fig. S4C to 4E in the supplemental material). These results suggest that CTCF/cohesin-mediated insulators are involved in the transcriptional regulation of the TNF/LT gene cluster. It is notable, however, that TNF stimulation itself did not affect the degrees of CTCF and RAD21 enrichment at each TC site (see Fig. S4F and G in the supplemental material), suggesting that higher-order chromatin regulation may be involved in the expression of the TNF/LT genes upon TNF stimulation. We assessed the knockdown effects with no significant cell damage throughout the study (see Fig. S4H and I in the supplemental material).

Dynamics of higher-order chromatin conformation in the TNF/LT locus.

3C assays were performed with Hep3B and NeHepLxHT cells to investigate higher-order chromatin regulation in the TNF/LT locus, where TE2 enhancer, gene promoters and TC insulators were identified as functional elements (Fig. 6; see also Fig. S5 in the supplemental material). Use of the 4-bp-recognizing restriction enzyme DpnII allowed us to examine these elements separately. Based on qPCR analyses of the intramolecular ligation products, the relative interacting frequencies of the reference site (yellow bar) with other 7 DpnII fragments containing each element in the TNF/LT locus were measured, as further described in Fig. S6 in the supplemental material. TE2 and TC2 were mainly chosen as the reference sites because of their effectiveness in the experiments. The efficiency of DpnII digestion of individual sites was > 80%, and samples without ligation gave no PCR-amplified products. We determined if CTCF knockdown affected the chromatin conformation of the TNF/LT locus in Hep3B cells (see Fig. S5A and B in the supplemental material). Compared with the basal control state, the frequencies of interactions of the referenced TE2 or TC2 with other fragments were mostly reduced to <50% in the CTCF-depleted cells, suggesting that CTCF is involved in the basal conformation of the locus.

Fig 6.

Dynamic changes in higher-order chromatin conformation of the TNF/LT locus under TNF stimulation. DpnII digestion was used to design 3C analyses to allow the examination of individual fragments containing each TC site, TNF/LT gene promoter, and TE2 enhancer. (A) The relative interacting frequencies between the reference TE2 fragment (yellow bar) and other DpnII fragments were determined by qPCR analyses of at least three distinct samples from Hep3B cells under TNF treatment. The relative frequencies of interactions between the reference TC2 (yellow bar) and other DpnII fragments in Hep3B cells (B) or between the reference TE2 (yellow bar) and other DpnII fragments in NeHepLxHT cells (C) are shown. In the right panel, the radar chart shows the average relative frequencies of interactions between the reference (central yellow circle) and each functional element. PCR amplification using internal primers located in the TNF/LT locus was used for a loading control to normalize the amount of DNA fragments. Efficiencies of DpnII digestion and subsequent ligation were confirmed at each restriction site used. The relative frequencies of interactions between the reference and its physically close site in the control state were normalized to 1 (TE2-TNF [A and C] or TC2-TC1 [B]). Control basal state, blue; TNF-expressing state, magenta; TNF/LTβ-expressing state, green. TC sites, TNF/LT gene promoters, and TE2 enhancer are indicated by the same color bars in the locus (upper panel) and the 3C data. The values are given as means and standard deviations of the results from more than three independent experiments. *, P < 0.05; **, P < 0.01; ***, P < 0.005.

To clarify the spatiotemporal chromatin dynamics of the TNF/LT locus, we then examined the frequencies of interaction between these regulatory elements under TNF stimulation (Fig. 6). 3C assays were carried out in the cells under the basal control state, TNF-expressing state (0.5 h after stimulation), and TNF/LTβ-expressing state (24 or 3 h after stimulation). Compared with results for the basal control state, the frequencies of TE2 interaction with other sites tested in the locus were significantly augmented in TNF-expressing Hep3B cells (Fig. 6A), suggesting that intrachromosomal interaction occurred in the locus. Interestingly, TE2 maintained an interaction with the LTβ promoter and TC3 in the TNF/LTβ-expressing state while remaining separate from other elements. We also examined the frequencies of TC2 ligation with other fragments and found that TC2 enhanced the interaction with other fragments in the TNF-expressing state (Fig. 6B). However, TC2 maintained its close localization with the TNF and LTα promoters, but not with other fragments, in the TNF/LTβ-expressing state. Using the TE2 fragment as a reference, similar data were obtained in TNF-stimulated NeHepLxHT cells (Fig. 6C), except for some interactions of TE2 with the TC3, TC1, and LTα promoter. Using the TC2 fragment as a reference, we did not clearly detect the interactions with other fragments in NeHepLxHT cells. Collectively, these data suggest that the enhancer-promoter interactions are selectively controlled by intrachromosomal association and subsequent dissociation of the TNF/LT locus upon activation of TNF signaling. To further demonstrate interactions between TC insulators in chromatin reorganization, we assessed their relative frequencies of interaction in these cells using TC4 as a reference (see Fig. S5C and D in the supplemental material). These TC sites consistently showed association in the TNF-expressing state and subsequent dissociation in the TNF/LTβ-expressing state (modeled in Fig. S7 in the supplemental material).

Expression of TNF and LTβ in human HCC tissues.

To examine whether the expression of TNF and LTβ is differentially regulated in vivo, we carried out immunohistochemical (IHC) analyses of HCC tissues (Fig. 7). Immunoreactivities of TNF and LTβ were assessed by comparison with the intensity of cytoplasmic staining of noncancerous hepatocytes within the same section. Representative images are shown in Fig. 7A, and the data for each tissue are summarized in Table S2 in the supplemental material. As summarized in Fig. 7B, neither TNF nor LTβ expression was detected in 16 out of 38 HCCs studied (42.1%), while both were densely stained in 31.6% of the cancer tissues. Interestingly, TNF alone was highly expressed in 10 of the 38 cancer tissues (26.3%), while LTβ alone was not detected in any cases. There may be at least two transcribed states in vivo, a TNF-expressing state and a TNF/LTβ-expressing state. We analyzed the correlation between the IHC data and clinical features and found no significant correlations between TNF and/or LTβ expression status and viral status, histological findings (differentiation grade of cancer, presence of chronic hepatitis or cirrhosis), or overall survival of the patients (data not shown). Although it is currently unknown whether the data for HCC tissues are related to higher-order chromatin states of the TNF/LT locus (shown in Fig. 6), these results suggest that differential expression of TNF and LTβ occurs in vivo.

Fig 7.

Expression of TNF and LTβ in human hepatocellular carcinoma tissues. (A) Representative immunohistochemical staining of human HCC. When the intensity of cytoplasmic staining was equivalent to or higher than that for noncancerous hepatocytes in >50% of cancer cells, the case was defined as positively stained. Three representative cases of the 38 cancer tissue samples tested are shown. Hematoxylin-and-eosin staining (upper) and immunostaining for TNF (middle) and LTβ (lower) are shown. Scale bar, 500 μm. (B) Percentages of TNF- and LTβ-stained cancer tissues. Cases with neither TNF nor LTβ expression (TNF− LTβ−), expression of both (TNF+ LTβ+), and TNF expression alone (TNF+ LTβ−) were found in 42.1%, 31.6%, and 26.3% of the cancer tissues, respectively. No cases expressed LTβ alone. The data for each tissue are summarized in Table S2 in the supplemental material.

DISCUSSION

The present study demonstrates the significance of the spatiotemporal regulation of gene activities and higher-order chromatin dynamics in the human TNF/LT locus. We identified four CTCF-dependent insulators (TC1, TC2, TC3, and TC4) and an enhancer (TE2) in hepatic cells. The well-known H19 DMR insulator contains four CTCF binding sites, while each TC site has single CTCF binding sequence with moderate enhancer blocking activities (Fig. 3). The LTα/TNF promoters and TE2 were located between TC2 and TC3, while the LTβ promoter was between TC3 and TC4, which may play a role in differential regulation of these three genes. The LTα/TNF genes were immediately induced by TNF stimulation in a fashion sensitive to inhibition of NF-κB signaling, while the LTβ gene was expressed later, as seen in other cell types (1, 39). Our previous report on the human apolipoprotein gene locus suggested that CTCF insulators play an essential role in clustered gene control (40). Furthermore, the current study shows that insulator interactions are likely to mediate intrachromosomal association and subsequent dissociation following TNF signaling. The dynamic enhancer-promoter associations and differential expression in the TNF/LT locus may be directed by the NF-κB-related regulatory molecules.

From the viewpoint of enhancer-promoter-insulator associations, we propose a spatiotemporal dynamics model in the human TNF/LT locus (see Fig. S7 in the supplemental material). In the basal state, CTCF-bound TC sites, the TE2 enhancer, and the TNF/LT promoters are located some distance apart in the chromatin structure. After TNF signaling activation, in the TNF-expressing state, the TC insulators, TE2, and TNF/LT promoters become colocalized and form a compact chromatin structure, resulting in interactions between TE2 and the TNF and LTα promoters. Because the LTβ gene is not fully induced at this stage, the LTβ promoter is likely to be sequestered by forming a possible chromatin loop between TC3 and TC4 (see Fig. S5C and D in the supplemental material). In addition, TC sites may be involved in stabilizing the interaction between TE2 and the TNF promoter because of the decrease of TNF expression in CTCF-depleted cells (Fig. 5C and E). In the TNF/LTβ-expressing state, TE2 significantly maintained its interaction with the LTβ promoter despite a reduced association with other elements. Thus, sequential chromatin conformation changes may contribute to switching of the enhancer-promoter interaction. Posttranslational modifications of CTCF and changes in the interacting molecules may be involved in the mechanism of intrachromosomal dynamics in the TNF/LT locus (47).

Our study revealed that TNF signaling can induce spatiotemporal remodeling of the clustered gene region and that CTCF insulators are likely to mediate higher-order control of transient enhancer-promoter interactions in the TNF/LT locus. Previous studies of the TNF/LT locus in hematopoietic cells suggested the presence of certain regulatory elements in intron 3 of the TNF gene and in the final exon of the LTβ gene (5, 66). The sequences, including the TC3 site, showed silencer activity in human T cells, though our study indicated that TC3 had a CTCF-dependent enhancer-blocking function, suggesting that the regulatory elements may differ among cell types. We showed that CTCF-mediated higher-order chromatin is involved in TNF/LT gene regulation. Persistent NF-κB activation in chronic inflammation may result in the chromatin conformation of the TNF/LT locus being deregulated and maintained in the TNF/LTβ-expressing state as an epigenetic memory. Indeed, constitutive NF-κB activation was recently noted to cause LTβ expression in inflamed hepatocytes and HCC cells in vivo (35), and LTβ was demonstrated to be an inducer of HCC (23). The proposed higher-order chromatin conformation of the TNF/LT locus may be involved in these in vivo situations.