Using a phylogenetic approach, we identified highly conserved sequences within intron 3 of the human TNF-α gene. These sequences form cell type-specific DNase I hypersensitivity sites and display cell type-specific DNA-protein contacts in in vivo genomic footprints. Consistent with these results, intron 3 confers specific activity upon a TNF-α reporter gene in Jurkat T cells, but not THP-1 monocytic cells. Thus, using a combinatorial approach of phylogenetic analysis, DNase I hypersensitivity analysis, in vivo footprinting, and transfection analysis, we demonstrate that intronic regulatory elements are involved in the cell type-specific regulation of TNF-α gene expression.
The primary control point of TNF-α gene transcription and protein production is the regulation of transcriptional initiation of the gene and involves the recruitment of cell type- and inducer-specific higher order structures, or enhanceosomes, to the promoter (1, 2, 3, 4). Intriguingly, the sequences contained in enhanceosome formation form a phylogenetic footprint, or a region of conserved sequences throughout the primate lineage (5) (A. Baena and A. E. Goldfeld, unpublished observations).
To identify novel potential regulatory elements in human TNF-α gene noncoding sequences, we performed a cross-species analysis of the TNF-α genomic region in the primate lineage and identified highly conserved intronic as well as other noncoding sequences. Using a five-nucleotide sliding window that was moved along the entire human TNF-α genomic sequence, the total number of pairwise differences between the human sequence and representatives of seven non-human primate species at each nucleotide position was quantified, and we identified striking regions of complete conservation, or phylogenetic footprints, particularly in intron 3. Notably, DNase I hypersensitivity (DH)3 analysis and dimethylsulfate (DMS) in vivo footprinting of three of these regions revealed that they are both accessible to DNase I and make DNA-protein contacts in a cell type-specific manner. Consistent with these results, intron 3, which contains a constitutive DH site in Jurkat T cells, but not in THP-1 monocytic cells, displays multiple T cell-specific in vivo footprints and enhances the activity of a TNF-α promoter reporter gene in Jurkat cells, but not in THP-1 cells. Thus, this highly conserved intronic sequence functions in a cell type-specific manner, similar to regulation of the proximal TNF-α promoter (1, 2, 3, 4).
Materials and Methods
Cell culture and ELISA
The human monocytic THP-1 and T lymphocytic Jurkat cell lines were cultured in RPMI supplemented with 10% FCS. Human HeLa cells were cultured in DMEM supplemented with 10% FCS. Optimum stimulation regimens for Jurkat cells were determined in an ELISA. Jurkat cells were seeded in triplicate for each time point/treatment at a density of 105 cells/well in 24-well plates. The cells were stimulated with PMA (50 ng/ml) or PMA plus ionomycin (50 ng/ml; 1 μM) for times ranging from 4–24 h, following which the supernatants from triplicate wells were pooled, cleared by centrifugation, aliquoted, and stored at −80°C. TNF-α protein levels were measured in duplicate using a Quantikine human TNF-α ELISA kit (R&D Systems, Minneapolis, MN). Similarly, optimum stimulation regimens for the THP-1 cells with tissue culture-tested Escherichia coli 0111:B4 LPS and Mycobacterium tuberculosis H37Rv whole sonicate (gift from C. C. Dascher, Harvard Medical School, Boston, MA) were determined in an ELISA. THP-1 cells were left untreated or were stimulated for 48 h with calcitriol (100 nM) to increase the expression of CD14 (6). They were then seeded in triplicate for each time point/treatment at a density of 25,000 cells/well in 24-well plates. The cells were stimulated with H37Rv whole sonicate (10 μg/ml) or LPS (1 μg/ml) for the times indicated, following which the supernatants from triplicate wells were pooled and processed as described above. The H37Rv whole sonicate contained 0.075 pg/μg endotoxin as determined by a kinetic, turbidimetric Limulus amebocyte lysate (LAL) assay (Endosafe; Charles River Laboratories, Lexington, MA).
Restriction mapping of the human TNF-α locus
The genetic map of the chromosomal region surrounding the human TNF-α locus was compiled from GenBank accession numbers AF129756, D0012, X01394, L11015, and AF000424. The location of critical restriction sites was verified in a composite Southern blot in which genomic DNA that was digested with BamHI, NsiI, or ScaI was hybridized with a 32P-labeled probe against the TNF-α 5′-flanking region. Southern blot DNA transfer, DNA hybridization, and stringent washes were performed as previously described (7).
DH analysis of the human TNF-α locus, based on the method described by Cockerill (7), was performed in two cell lines that express the TNF-α gene in an inducible fashion (THP-1 and Jurkat) and the HeLa cell line, which does not transcribe TNF-α. Briefly, nuclei from untreated and stimulated cells were isolated in Nonidet P-40 lysis buffer and digested with increasing amounts of DNase I. DNA was phenol/chloroform-extracted, ethanol-precipitated, and digested with BamHI, NsiI, or ScaI. Samples were resolved on 0.7% agarose gels, transferred to nitrocellulose membranes, and hybridized with probes corresponding to the 5′-flanking region of the TNF-α gene and/or the 3′ end of the lymphotoxin-α (LT-α) gene.
Plasmids and transfection
TNF-α firefly luciferase reporter plasmids were constructed in pGL3 Basic (Promega, Madison, WI) by routine methods. All reporter plasmids included combinations of a TNF-α promoter element (−1243 to +181 relative to the transcription start site) and intron 3 of the human TNF-α gene. Regulatory elements were PCR-amplified from THP-1 genomic DNA, cloned into pCR4 Blunt-TOPO (Invitrogen, Carlsbad, CA), and verified by sequencing. Plasmids were isolated using EndoFree Maxiprep kits (Qiagen, Valencia, CA), and the level of endotoxin contamination was assessed using a kinetic, turbidimetric LAL assay (Charles River Endosafe). Endotoxin levels were <1 pg/ml in all experiments. THP-1, Jurkat, and HeLa cells were transfected with FuGENE 6 (Roche, Indianapolis, IN) using conditions that were optimized for each cell line according to the recommendations provided by the manufacturer. All transfections included a control Renilla luciferase plasmid (pRL-TK) and were normalized to Renilla luciferase activity. The cells were transfected in duplicate or triplicate in each of three independent experiments.
Phylogenetic analysis of the TNF-α genomic sequence in the primate lineage
The TNF-α genomic sequence extending from the transcription start site to 575 bp into the 3′-untranslated region (3′UTR) was PCR-amplified in four overlapping fragments from genomic DNA isolated from THP-1 cells and from a representative of each of seven non-human primate species (5). PCR products were cloned into pCR4 Blunt-TOPO (Invitrogen), and three independent clones were submitted for automated sequencing to assemble a consensus sequence. Alignments were performed using CLUSTALW software (www.clustalw.genome.ad.jp). Sliding window analysis of aligned primate sequences was performed as described previously (5).
DMS in vivo footprinting
Sets of three strand-specific oligonucleotides were used to analyze methylation patterns at the TNF-α locus (8). Briefly, after control and stimulated THP-1 and Jurkat cells were treated with DMS, genomic DNA was isolated and cleaved with piperidine. The resulting fragments were used as templates for a first-strand synthesis using gene-specific primer 1. Following ligation of a short linker, the target was amplified using a second gene-specific primer and a primer annealing to the ligated linker (ligation-mediated PCR). Following two additional PCR cycles with a third, end-labeled gene-specific primer, the resulting products were separated on Explorer Gel 5% gels and displayed by autoradiography.
Approximately 2.6 kb of the genomic sequence of the TNF-α gene from the transcription start site (TSS) to 575 bp into the 3′UTR was analyzed and compared in representative great apes (Pan troglodytes, Gorilla gorilla, Pongo pygmaeus, and Hylobates lar), Old World monkeys (Papio hamadryas and Macaca mulatta) and New World monkeys (Aotus trivirgatus). Unexpectedly, as shown in Table I, we found an extremely high degree of conservation within the noncoding sequences, and to characterize the spatial pattern of these regions, a sliding window analysis of the TNF-α genomic sequence was performed (data not shown). This analysis of intron 3, where the degree of sequence similarity approaches or even exceeds that of adjacent exons, revealed two regions of 100% homology (Fig. 1). Of note, these regions (marked by asterisks in Fig. 1) spanning 19 and 24 bp, respectively, contain sequences spanning 11 and 12 bp that are 100% conserved in six additional mammalian species (pig (Sus scrofa), GenBank X54859; cow (Bos taurus), AF011926; cat (Felis catus), X54000; rabbit (Oryctolagus cuniculus), M12846; rat (Rattus norvegicus), AF329987; mouse (Mus musculus), NT002588; data not shown).
|.||Pan .||Gorilla .||Pongo .||Hylobates .||Papio .||Macaca .||Aotus .|
|Exon I, coding sequence||100||99||99||98||94||95||91|
|Exon IV, coding sequence||99||99||98||97||97||97||93|
|.||Pan .||Gorilla .||Pongo .||Hylobates .||Papio .||Macaca .||Aotus .|
|Exon I, coding sequence||100||99||99||98||94||95||91|
|Exon IV, coding sequence||99||99||98||97||97||97||93|
The TNF-α genomic sequence extending from the TSS to 575 bp into the 3′UTR was PCR-amplified in four overlapping fragments from genomic DNA isolated from THP-1 cells and from representative great apes (Pan troglodytes, Gorilla gorilla, Pongo pygmaeus, and Hylobates lar), Old World monkeys (Papio hamadryas and Macaca mulatta), and New World monkeys (Aotus trivirgatus). Alignments were performed using CLUSTALW software (www.clustalw.genome.ad.jp), and the degree of evolutionary conservation was assessed using default parameters. Note the extensive conservation in noncoding sequences, particularly intron 3.
Sequence extends 575 bp from the translation stop codon.
Given our previous results demonstrating that highly conserved phylogenetic footprints coincide with important regulatory regions in the human TNF-α gene promoter (5) (A. Baena and A. E. Goldfeld, unpublished observations), we next tested the hypothesis that the conserved intron 3 motifs play a role in the regulation of the gene. We first prepared and digested genomic DNA with BamHI, NsiI, and ScaI and hybridized the DNA to a probe corresponding to the 5′-flanking region of the human TNF-α gene. Restriction fragments of the expected sizes were obtained in Southern blot analysis of monocytic THP-1 cells and Jurkat T cells, while a ScaI site appeared to be missing in HeLa DNA (Fig. 2,A). A genetic map of the chromosomal region surrounding the TNF-α locus with the relevant restriction sites is shown in Fig. 2,B. We note that in addition to the entire TNF-α locus, the BamHI fragment also contains most of the LT-α gene, the entire LT-β gene, and most of the leukocyte-specific transcript 1 (LST1) gene sequence (Fig. 2 B).
We then performed DH analysis of the human TNF-α locus in these three human cell lines to test whether these regions were involved in chromatin accessibility in a cell type-specific manner. Unstimulated, PMA-stimulated, or PMA- and ionomycin (P+I)-stimulated Jurkat cells; unstimulated and LPS- or Mycobacterium tuberculosis (MTb)-stimulated THP-1 cells; and HeLa cells were tested using increasing concentrations of DNase I (Fig. 3,A). Intriguingly, consistent with the cell type- and inducer-specific regulation of TNF-α in T cells and monocytes (1, 2, 3, 4), THP-1 and Jurkat cells display distinct DH patterns within the TNF-α locus, which are summarized in the maps at the bottom of Fig. 3,A. As shown in Fig. 3,A, cell type-specific DH sites were detected in intron 3, in the proximal promoter region, and in the 3′UTR in Jurkat T cells (lanes 7–13, marked by arrows 4, 5, and 6), whereas a cell type-specific DH site was observed in the region between the TSS and the ATG of the gene in monocytic THP-1 cells (lanes 1–6, marked by arrow 1). Notably, we did not detect DH sites in the TNF-α locus in HeLa cells, consistent with the lack of induction of TNF-α in HeLa cells by any stimulus tested to date (A. E. Goldfeld, unpublished observations). Interestingly, we also identified cell type-specific DH sites in the adjacent LT-β gene. We note that this is a biologically relevant model system for the chromatin studies, since we could detect TNF-α protein levels after stimulation of the Jurkat cells with PMA or P+I and THP-1 cells with LPS or with a whole sonicate of a virulent strain of MTb (H37Rv; Fig. 3 B).
To determine whether the DH sites detected could be correlated with specific DNA-protein interactions in these T cells and monocytes, we next performed DMS genomic footprinting (8). Since DMS is a small chemical, the DNA region protected from modification by a bound protein is short, and thus DMS footprinting can potentially discern DNA-protein interactions with base pair precision. Examples of some of the genomic footprints we obtained are shown in Fig. 4. In the proximal promoter region, cell type-restricted constitutive genomic footprints were detected on both the coding and noncoding strands (Fig. 4, lanes 1–8, arrows 1–6, and summarized in Fig. 5,A). On the coding strand, Jurkat cells showed constitutive hypermethylation of adenosine residues adjacent to the −180 NFAT/Ets site (arrows 1 and 2), the −149 NFAT site (arrows 3 and 4), and the −117 NFAT/Ets site (arrows 5 and 6). Furthermore, as summarized in Fig. 5,A, both Jurkat and THP-1 cells had constitutive footprints between the κ3 and −84 Ets sites and immediately adjacent to the TSS. Interestingly, the genomic footprints identified in Jurkat and THP-1 cells all map in the vicinity of transcription factor binding sites that have been shown to be of critical importance in the regulation of TNF-α gene expression in T cells and monocytic cells (1, 3, 4). Furthermore, extensive constitutive footprints were identified exclusively in Jurkat cells on the noncoding strand between positions −267 and −190 relative to the TSS (data not shown). We also observed constitutive footprints at +79 on the coding strand (data not shown) and position +145 on the noncoding strand in both Jurkat and THP-1 cells (Fig. 4, lanes 9–16, arrow 7).
Remarkably, Jurkat cells displayed extensive constitutive footprints in the form of hypermethylated adenosine residues on the noncoding strand of the proximal portion of intron 3 (examples are shown in Fig. 4), four of which were also detected in THP-1 cells (Fig. 4, lanes 17–24, arrows 8–11, and summarized in Fig. 5,B). Moreover, the T cell-specific genomic footprints colocalize with the constitutive DH site that was identified in Jurkat T cells, but not in THP-1 monocytic cells. Furthermore, the majority of the genomic footprints we identified in intron 3 map to sequences that are highly conserved in the primate lineage (Fig. 5 B).
Previous studies have demonstrated the important functional roles of the proximal promoter, the region from the TSS to the ATG and the 3′UTR in TNF-α gene regulation (3, 4, 9). To test whether intron 3 might also contain a regulatory element, we examined the effect of the addition of intron 3 on the regulation of a TNF-α luciferase reporter gene. The activity of the TNF-α luciferase reporter gene containing intron 3 in PMA-stimulated Jurkat cells was increased by >50% (Fig. 6,A; p = 0.02), whereas the addition of intron 3 had no appreciable effect on TNF-α reporter activity in THP-1 cells stimulated with MTb or LPS (Fig. 6 B). Taken together, we conclude that intron 3 plays a cell type-specific regulatory role in TNF-α gene regulation.
Since the TNF-α protein product can exert both beneficial and detrimental effects in a variety of infectious and autoimmune diseases, it is not surprising that its biosynthesis is under the control of multiple complex regulatory mechanisms that must achieve a high degree of transcriptional specificity in response to an extracellular stimulus. In this study we have presented four distinct lines of evidence that demonstrate the existence of a cell type-specific regulatory element functional in Jurkat T cells in intron 3 of the human TNF-α gene: 1) intron 3 confers cell type-specific augmentation of TNF-α reporter gene transcription; 2) intron 3 displays cell type-specific accessibility of chromatin to DNase I digestion; 3) intron 3 makes cell type-specific DNA-protein contacts, as determined by in vivo DMS footprinting; and 4) there is a high degree of conservation of intron 3 in the primate lineage. Taken together, we conclude that intronic control elements function in a cell type-specific fashion in TNF-α gene transcription.
We note that an Ets-binding element in intron 3 has previously been implicated in regulation of the murine TNF-α gene by LPS in a single murine monocytic cell line, RAW264.7 (10). We were unable to detect a functional role for intron 3 in human TNF-α gene regulation by LPS or MTb in monocytic THP-1 cells, but did see a functional role for intron 3 in Jurkat T cells (Fig. 6), consistent with the detection of a DH site (Fig. 3) and in vivo footprints in this region in the T cells (Fig. 4). Furthermore, we were unable to detect a DH site in intron 3 in LPS- or mock-stimulated murine monocytic J774 cells (R. Barthel and A. E. Goldfeld, unpublished). Here we also show cell type-specific DH sites in the proximal promoter, in sequences between the TSS and the ATG, and in the 3′UTR, regions that have previously been shown to play important functional roles in TNF-α gene regulation (1, 3, 4, 9, 11, 12). Consistent with our findings, we note that constitutive DH sites have been identified in the IL-2 gene that are cell lineage specific and that distinguish IL-2 producer and nonproducer T cells (13).
Approximately 1% of the genome exists as discrete regions of chromatin that provide increased access to factors that interact with DNA, which can indicate the existence of specific structural or regulatory elements, such as enhancers, locus control regions, silencers, and boundary elements (14). Moreover, these DH sites, which can be detected by sensitivity to DNase I, have frequently provided the first evidence of the existence of distant regulatory elements located even several kilobases upstream or downstream of genes, or in some cases within introns (15, 16, 17, 18, 19, 20, 21).
Interestingly, we also observed a correlation between the presence of DH sites and the ability of a specific cell type to express the gene product for both the LST1 and LT-β genes in specific cell types. The LST1 gene is constitutively expressed in leukocytes and dendritic cells and encodes a small protein that modulates the immune response and cellular morphogenesis (22, 23). Only THP-1 and Jurkat cells, which both express the LST1 gene product in a constitutive fashion, display a constitutive DH site in the same segment of the 5′-flanking region (see DH sites 3 and 9 in Fig. 3,A). LT-β, which is expressed on the surface of activated T cells, B cells, and lymphokine-activated killer cells and plays a critical role in spleen and lymph node organogenesis (24), displayed a PMA-inducible DH site in the 5′-flanking region of the gene in Jurkat T cells, but not THP-1 or HeLa cells (see DH site 8 in Fig. 3 A).
Boffelli et al. (25) have shown recently that phylogenetic shadowing of primate sequences that includes sequence from as few as four to six primates in addition to humans is sufficient for the identification of functional elements in the human genome, many of which are likely to be missed by human-mouse comparisons. Consistent with these observations, our previous analysis of TNF-α sequences in the primate lineage demonstrated that such an approach is a powerful tool for the identification of candidate regions important for gene regulation. Our analysis of 2.6 kb of genomic TNF-α sequence in representative great apes, Old World monkeys, and New World monkeys demonstrates a high degree of conservation in intronic sequences, and remarkably, intron 3, which contains a DH site and confers functional activity in vitro upon a TNF-α promoter reporter, shows among the highest degree of evolutionary conservation within TNF-α noncoding sequences (Table I). Notably, recent large scale sequence alignment studies (26) have suggested that there appears to be a significant overrepresentation of transcription factor binding sites in introns. These bioinformatics models are supported, for example, by the identification of regulatory elements within introns of the genes encoding IL-4 and GATA-3 (15, 16) and by the data presented here, which extend this concept and support a cell type- and inducer-specific role for intronic regulatory elements in the transcription of TNF-α, which encodes a key molecule in the immune response (27).
We thank Renate Hellmiss for the artwork, and Alla Tsytsykova and Andres Baena for helpful discussions and comments on the manuscript. We are also grateful to Jessica Leung for help with the sliding window analysis.
This work was supported by National Institutes of Health Grants GM56492 and HL59838, the American Heart Association (to A.E.G.), and a postdoctoral fellowship from the Heiser Program of the New York Community Trust (to R.B.).
Abbreviations used in this paper: DH, DNase I hypersensitivity; DMS, dimethylsulfate; LAL, Limulus amebocyte lysate; LST1, leukocyte-specific transcript 1; LT-α, lymphotoxin-α; MTb, Mycobacterium tuberculosis; P+I, PMA and ionomycin; TSS, transcription start site; UTR, untranslated region.