Long-range enhancer associated with chromatin looping allows AP-1 regulation of the peptidylarginine deiminase 3 gene in differentiated keratinocyte

Abstract

Transcription control at a distance is a critical mechanism, particularly for contiguous genes. The peptidylarginine deiminases (PADs) catalyse the conversion of protein-bound arginine into citrulline (deimination), a critical reaction in the pathophysiology of multiple sclerosis, Alzheimer's disease and rheumatoid arthritis, and in the metabolism of the major epidermal barrier protein filaggrin, a strong predisposing factor for atopic dermatitis. PADs are encoded by 5 clustered PADI genes (1p35-6). Unclear are the mechanisms controlling the expression of the gene PADI3 encoding the PAD3 isoform, a strong candidate for the deimination of filaggrin in the terminally differentiating epidermal keratinocyte. We describe the first PAD Intergenic Enhancer (PIE), an evolutionary conserved non coding segment located 86-kb from the PADI3 promoter. PIE is a strong enhancer of the PADI3 promoter in Ca2+-differentiated epidermal keratinocytes, and requires bound AP-1 factors, namely c-Jun and c-Fos. As compared to proliferative keratinocytes, calcium stimulation specifically associates with increased local DNase I hypersensitivity around PIE, and increased physical proximity of PIE and PADI3 as assessed by Chromosome Conformation Capture. The specific AP-1 inhibitor nordihydroguaiaretic acid suppresses the calcium-induced increase of PADI3 mRNA levels in keratinocytes. Our findings pave the way to the exploration of deimination control during tumorigenesis and wound healing, two conditions for which AP-1 factors are critical, and disclose that long-range transcription control has a role in the regulation of the gene PADI3. Since invalidation of distant regulators causes a variety of human diseases, PIE results to be a plausible candidate in association studies on deimination-related disorders or atopic disease.

Figure 1. Identification of a strong transcriptional enhancer of the gene PADI3 in differentiated keratinocytes.

A) The CNS2 comprises an enhancer of PADI3 gene promoter activity in differentiating NHEKs. NHEKs were transfected with luciferase reporter constructs containing the indicated promoter segment with or without the CNS2 which overlaps PIE, in the presence of 0.2 or 1.5 mM calcium. Values are expressed in Relative Luciferase Units (RLU), as means+/−SEM for at least 3 independent experiments, each performed in triplicate wells.*, p<0.05. B) The CNS2 does not show any detectable enhancer activity either in A431 skin squamous cell carcinoma keratinocytes (top left), in HeLa cervix adenocarcinoma cells (top right), in immortalized HaCaT (bottom left) keratinocytes, or in human primary dermal fibroblasts (bottom right), or 3T3 transformed murine fibroblasts (not shown). C) Schematic representation of the PADI gene locus showing CNS2/PIE (asterisk), IG1 (open box) and the genes PADI1-3 (open boxes with arrows indicating the orientation of transcription). The coordinates in AJ549502 are shown in kb or bp. The telomeric (tel) and centromeric (cen) ends of the locus are shown. The 19 CNSs within IG1 are depicted by vertical bars of lenght depending upon their level of interspecies conservation (for details, see [3]). The entire mouse and human sequences of CNS2/PIE are shown. Underlined are the conserved bases between mouse and human. The centromeric end of the locus (approximatively 100 kb) including the genes PADI4 and 6 is not shown due to size limitation.

Figure 2. Hypersensitivity of the CNS2 to DNase I.

The DNase I-digestion profiles of the CNS2 including PIE, the KRT5 gene promoter region, and the two segments 114k (a LINE repeat, nt 114-549-114655) and 115k (nt 115764–115778) located at the centromeric end of IG1 are shown. The amount of template DNA was standardized by correcting for amplification of the HBE1 promoter region sequence. Inset: representative dissociation curves.

Figure 3. AP-1 factors binding and proximal CAAT motifs required for enhancer activity.

A) Two conserved AP-1 sites were identified in PIE. Alignment of the sequences of the conserved telomeric or centromeric AP-1 putative binding sites (bold) in PIE in human, mouse, rat, dog, cow, armadillo (Dasypus novemcinctus), hedgehog (Echinops telfairi) and platypus (Ornithorhynchus anatinus). Sequence retrieval and analyses were performed using the Ensembl web site (http://www.ensembl.org), the Vista suite (http://genome.lbl.gov/vista/index.shtml), and the program Dialign-TF (http://www.genomatix.de) with default parameters. Note that the centromeric site is more highly conserved that the telomeric site. B) AP-1 and CAAT recognition motifs are mandatory for enhancer activity. Luciferase (luc) plasmids were engineered with invalidated telomeric (AP-1t) and centromeric (AP-1c) AP-1 sites in PIE (lines 1–3), or invalidated CCAAT or GC boxes in the PADI3 promoter (cross) (lines 4–7). Line pC2-PADI3-luc: shown is the ratio between the luciferase activity of the PADI3 promoter segment with PIE versus that without PIE. Lines 1–3: shown are the ratios between the luciferase activities of the mutated constructs versus that of the construct containing the PADI3 promoter segment without PIE. Lines 4–7: shown are the ratios between the luciferase activities of the mutated constructs with PIE versus that of their counterparts without PIE. Values are expressed as means+/−SEM for 2 independent experiments, each performed in triplicate wells. C, D) Bound c-Fos/c-Jun are required for PIE function. Electrophoretic mobility shift assays (C) showed that both telomeric (AP-1t) and centromeric (AP-1c) AP-1 sites are expressed and able to bind nuclear factors from differentiating NHEKs, generating a shifted band, which is no longer detected in the presence of unlabeled wild type competitor. When antibodies specific to c-Jun, JunB or c-Fos were used, the shifted band was either supershifted or absent due to the formation of antibody-containing higher molecular weight complexes which do not enter the gel. The binding of c-Jun and c-Fos to PIE in living NHEKs was confirmed by chromatin immunoprecipitation (D). Immunoprecipitated chromatin from NHEKs grown in the presence of 1.5 mM calcium showed about 3-fold relative enrichment in c-Jun or c-Fos bound to PIE as compared to that from NHEKs grown in the presence of 0.2 mM calcium (grey bars). No enrichment was observed using antibodies to JunB or the irrelevant anti-hemaglutinin (HA), nor when we queried the unrelated CNS3 which is centromeric to the CNS2 (nt 109913–109973) (white bars). Data are shown as means+/−SEM for 2 independent experiments performed in triplicate wells. E) Blocking c-Jun or the ras/raf/ERK cascade impairs the activity of PIE. Differentiating NHEKs were co-transfected with reporter plasmids containing the PADI3 gene minimal promoter region fitted or not with PIE, and an expression vector coding for dominant negative mutants of c-Jun (TAM67), ras (ras-), raf (raf-), ERK1 and 2 (ERK1- and ERK2-, respectively) or the GFP as a control (GFP) (top). For these co-transfections, the luciferase activities generated by the plasmids containing PIE were compared with that generated by the plasmids containing only the PADI3 gene promoter. Parallel co-transfection experiments using the plasmid pAP1-luc showed dramatic loss of AP-1 activity in NHEK transfected with the dominant negative mutants (bottom). For these co-transfections, the luciferase activities generated by pAP1-luc co-tranfected with mutants were compared with that generated by pAP1-luc transfected alone. The plasmid pCR2.1 was used to adjust for the mass of the input DNA. Values are expressed as means+/−SEM for 2 independent experiments performed in triplicate wells.

Figure 4. Increased frequency of interaction between PIE and the PADI3 promoter indicating chromatin looping.

The abundancy of chimeric DNA containing PIE and the promoter region of the genes PADI1 or PADI3 were explored by 3C analysis using specific primers (top). The levels of PIE-PADI3 promoter chimeric DNA in extracts from NHEKs grown in the presence of 1.5 mM calcium were approximately 20-fold greater than that in NHEKs grown in the presence of 0.2 mM calcium, or dermal fibroblasts. Amplification of BAC-based negative control for random association events (see methods) yielded no product (not shown). All amplifications were performed in triplicate wells. Results were normalized with respect to the abundancy of the PADI3 promoter sequence in the template. Data are expressed in Relative Units (RU) as means+/−SEM for 2 independent experiments, each performed in triplicate wells.*, p<0.05. Representative dissociation curves (inset) show the major amplimer peaks (arrow) along with secundary peaks (arrowhead) indicative of primer-dimer formation in some experiments. Such possible primer-dimer was not detectable in ethidium-bromide stained agarose gel of the end-point PCR reactions and did not seem to interfere with calibration curves.

Figure 5. NDGA blocks calcium-induced increase of the PADI3 mRNA levels and impairs c-Jun or c-Fos recruitment at PIE.

Quantitative RT-PCR analysis of the steady-state levels of PADI1-3 (A) and involucrin (B) mRNAs was performed using unstimulated or calcium-stimulated NHEKs grown in the presence of 20 µM NDGA in 0.1% DMSO, or 0.1% DMSO, and the β2-microglobulin mRNA as the reference. Data are expressed as means+/−SEM for 2 independent experiments, each performed in triplicate wells.*, p<0.05. A) NDGA blocks calcium-induced increase of the mRNA levels of PADI3, but not that of PADI 1, 2. Treatment with 0.1% DMSO had no effect on the PADI3 mRNA levels (2^(−ΔΔCt)) in calcium-stimulated NHEKs. Treatment with 20 µM NDGA resulted in a dramatic decrease of PADI3 mRNA levels in calcium-stimulated NHEKs. B) NDGA blocks calcium-induced increase of the involucrin mRNA levels. In untreated NHEKs, calcium stimulation resulted in a two fold increase in involucrin mRNA levels. In NHEKs treated with the carrier (DMSO), calcium stimulation associated with a five fold increase in involucrin mRNA levels. In NHEKs treated with NDGA, calcium stimulation associated with only a 1.5 fold increase in involucrin mRNA levels. Together these results suggested that treatment with NDGA blocked calcium-induced increase of the PADI3 and involucrin mRNA levels. RU, relative units. C) NDGA impairs c-Jun or c-Fos recruitment at PIE in calcium-stimulated NHEKs. ChIP analysis of PIE occupancy by c-Jun/c-Fos was performed in the presence of NDGA or the carrier (DMSO). Immunoprecipitated chromatin from NHEKs grown in the presence of 1.5 mM calcium showed strongly decreased occupancy of PIE by c-Jun (20% remaining) or c-Fos (34% remaining) but not non specific factors (74% remaining) in the presence of NDGA as compared to the same cells grown in the presence of the carrier. Immunoprecipitated chromatin from NHEKs grown in the presence of 0.2 mM calcium showed no decrease in the occupancy of PIE by c-Jun (99% remaining) or c-Fos (130% remaining) or non specific factors (140% remaining) in the presence of NDGA as compared to NHEKs grown in the presence of the carrier. Data are expressed as Relative Units (RU) and shown as means+/−SEM for 2 independent experiments performed in triplicate wells. *, p<0.05.

Figure 6. Possible model for PADI3 promoter regulation.

The binding of cJun and/or cFos dimers to the AP-1 sites of PIE is enhanced in calcium-stimulated keratinocytes and increases the rate of transcription of PADI3 (broken arrow). Transcription factors with functional association (cJun, cFos, NF-Y) are shown as open circles. Sp1/3 factors are denoted as grey circles. Transcription factors binding sites are indicated (capital letters). The arrows show the transcription orientation of the genes PADI1-2 (white boxes). TBP: TATA-box binding protein; pol II: RNA polymerase II.