Molecular mechanisms controlling CFTR gene expression in the airway

Abstract The low levels of CFTR gene expression and paucity of CFTR protein in human airway epithelial cells are not easily reconciled with the pivotal role of the lung in cystic fibrosis pathology. Previous data suggested that the regulatory mechanisms controlling CFTR gene expression might be different in airway epithelium in comparison to intestinal epithelium where CFTR mRNA and protein is much more abundant. Here we examine chromatin structure and modification across the CFTR locus in primary human tracheal (HTE) and bronchial (NHBE) epithelial cells and airway cell lines including 16HBE14o- and Calu3. We identify regions of open chromatin that appear selective for primary airway epithelial cells and show that several of these are enriched for a histone modification (H3K4me1) that is characteristic of enhancers. Consistent with these observations, three of these sites encompass elements that have cooperative enhancer function in reporter gene assays in 16HBE14o- cells. Finally, we use chromosome conformation capture (3C) to examine the three-dimensional structure of nearly 800 kb of chromosome 7 encompassing CFTR and observe long-range interactions between the CFTR promoter and regions far outside the locus in cell types that express high levels of CFTR.


Introduction
The recent development of new technologies to identify regulatory elements in non-coding regions of the human genome and elucidate their function [1] has enabled rapid advances in our understanding of many important genes. One such gene is the cystic fibrosis transmembrane conductance regulator (CFTR), which when mutated causes the common inherited disorder cystic fibrosis (CF). CFTR is a large gene encompassing 189 kb of genomic DNA [2] and showing complex cell-type specific and temporal regulation (reviewed in Ref. [3]). Of particular note is the very wide range of levels of CFTR expression in different cell types, with up to 1 ϫ 10 4 more mRNA in pancreatic ducts and colon carcinoma cell lines than in primary cultures of tracheal and bronchial epithelium [4,5]. Extensive characterization of the CFTR promoter region (reviewed in Ref. [3]) did not reveal elements that were responsible for cell-type-specific expression of the gene. Hence, we used several methods to find regions of open chromatin (DNase I hypersensitive sites, DHS) across the locus and in flanking sequences to find the critical cis-acting elements (reviewed in Ref. [3]). We identified enhancer-blocking insulator elements that flank the CFTR gene at Ϫ20.9 kb with respect to the translational start site and at ϩ15.6 kb 3' to the translational stop site which likely establish the functional limits of the locus in at least some cell types. The Ϫ20.9 kb region binds CCCTC-binding factor (CTCF) [6], which is associated with many insulator elements, although the ϩ15.6 kb element does not. However, additional DHS 3' to the locus coincide with CTCF and cohesin (Rad21) binding sites [5,7]. DNase I hypersensitive sites within the first and eleventh introns encompass intestinal/genital duct restricted enhancer elements, which cooperate in reporter gene constructs to augment CFTR promoter activity about 40-fold. Our current model for the active CFTR locus presents a looped structure, which brings the insulators and distal cis-acting enhancer elements into close association with the promoter. These direct associations were measured by quantitative chromosome conformation capture (3C) [8]. However, when presenting this model we noted its relevance to CFTR expression in cell types that express abundant CFTR, such as primary epididymis epithelial cells and the intestinal cell lines Caco2 and HT29, but also suggested that the low levels of CFTR expression in primary airway epithelial cells might be mediated by different mechanisms [5]. In an effort to elucidate these, we now concentrate on the regulatory elements that may coordinate CFTR expression in airway epithelial cells. We identify cell-typespecific DHS that are seen in primary airway epithelial cells but not in airway cell lines that express much higher levels of CFTR mRNA such as 16HBE14o-. We also describe a novel DHS close to the CFTR promoter that is unique to this cell line. Next, we investigate epigenetic modifications at the CFTR locus and flanking regions in primary airway cells: enrichment of specific histone variants that are associated with enhancers (H3K4me1) and repressed chromatin (H3K27me3) [9] are evaluated. Several of the DHS are seen to encompass regions of H3K4me1 enrichment while H3K27me3 is evident immediately 5' to the gene promoter in primary airway cells, but not in other cell types that express CFTR. Elements within the DHS that are enriched for H3K4me1 are next shown to function as cell-type-specific enhancer elements that cooperate in reporter gene assays. Finally, we examine long-range interactions across nearly 800 kb flanking the CFTR locus to determine whether airway-specific DHS that are located distal to the enhancer blocking insulator elements at Ϫ20.9 kb and ϩ15.6 kb are able to physically interact with the CFTR promoter, despite these barriers. These studies advance our understanding of the cell-type-specific regulation of CFTR and highlight mechanisms that may coordinate the low expression levels that are characteristic of primary airway cells.

Cell culture
Primary skin fibroblasts (Coriell GM08333) were grown in minimal essential media (MEM; Invitrogen, Carlsbad, CA, USA) supplemented with 15% foetal bovine serum (FBS). Primary tracheal epithelial cells were extracted from post-mortem human adult trachea as previously described with minor modifications [10]. NHBE cells, a mixture of primary human bronchial and tracheal epithelial cells (CC-2541; Lonza, Walkersville, MD, USA) were cultured in BEGM (Lonza) per the manufacturer's instructions.
Transformed human bronchial epithelial cell lines 16HBE14o- [11] and Beas2B [12] and the lung carcinoma cell line Calu3 [13] were grown in DMEM with 10% serum. All cells were grown on plastic at liquid interface.

Primer sequences
All primer sequences and locations used for DNase-chip, RT-PCR, plasmid cloning and mutagenesis and 3C are listed in the Supporting information.

DNase-chip
DNase-chip was performed as previously described, with modifications [14]. Briefly, 2-5 ϫ 10 7 cells were lysed using 0.1% NP-40 buffer. Purified nuclei were exposed to increasing amounts of DNase I (0-30 U; NEB, Ipswich, MA, USA), reactions were stopped with 0.1M EDTA, and digested chromatin was embedded into InCert agarose plugs (Lonza). Chromatin digestion was determined by pulsed field gel electrophoresis and adequately digested samples were blunt-ended with T4 DNA polymerase (NEB). Chromatin was then extracted from agarose, and blunt ends were ligated to biotinylated linkers overnight. As a control, 25 g of genomic DNA from the same cell type was ligated to linkers and processed in parallel. Chromatin was sonicated to generate 200-500 bp fragments, and biotinylated ends were captured with steptavidin Dynabeads (Invitrogen). Sheared ends were blunted with T4 DNA polymerase and ligated to non-biotinylated linkers, and samples were amplified with ligation-mediated PCR using primer oJW102C. PCR material was labelled with Cy5-dUTP and genomic DNA control labelled with Cy3-dUTP, and each hybridized to ENCODE tiling arrays (NimbleGen, human genome build 17, May 2004). Hybridization data from three (Caco2 and fibroblasts) or two (16HBE14o-) experiment was analysed with ACME statistical software [15] using a window size of 500 bp and a threshold of 0.95.

Chromatin immunoprecipitation
Primary cells were cross-linked directly on the cell culture plates, longterm cell lines were trypsinized, resuspended in DMEM and cross-linked with 0.37% formaldehyde for 10 min. Cross-linking was stopped by the addition of glycine to 0.125M. Cells were washed with cold PBS and for 1 ϫ 10 7 cells lysed in 1 ml of 1% SDS, 10 mM EDTA, 50 mM Tris/HCl pH 8.1, 1ϫ protease inhibitor cocktail (Roche, Indianapolis, IN, USA). Chromatin was sonicated to an average size of 200-500 bp.

Chromosome conformation capture (3C)
3C was performed as described previously [8,7], with minor modifications. Briefly, 1 ϫ 10 7 cells were fixed with 2% formaldehyde for 10 min. at room temperature. Cells were lysed in 5 ml cold lysis buffer [10 mM Tris (pH 8), 10 mM NaCl, 0.2% NP-40, 1ϫ protease inhibitor cocktail (Roche)] and the nuclei collected by centrifugation. Following extraction with 0.3% SDS, chromatin was digested overnight with 2000 U Hind III. Ligations were performed in a total reaction volume of 6.5 ml, using 100 U T4 DNA ligase (Roche) and incubation at 14ЊC for 4 hrs followed by 30 min. at room temperature. Cross-links were reversed by proteinase K treatment at 65ЊC overnight. Samples were purified by phenol-chloroform extraction followed by ethanol precipitation, and then resuspended in 150 l H2O. The concentration of each sample was determined by SYBR green qPCR, using the B13F/B13R primer set (amplicon found within a Hind III fragment; see Supporting Information) and comparison to a genomic DNA reference of known concentration. Samples were subsequently diluted to a concentration of 100 ng/l. A Taqman probe and reverse primer were designed that were specific to a Hind III fragment at the CFTR promoter (bait). Multiple forward primers were then designed that were each specific to different Hind III fragments across the CFTR locus (see Table S1 for primer and probe sequences and locations). Using a dilution series of digested/religated BAC DNA template, each forward primer was demonstrated to function with the 'fixed' Taqman probe and reverse primer to amplify with 100% efficiency. To quantify ligation events within 3C samples, 200 ng of 3C template was used per 20 l Taqman qPCR reaction. The ligation efficiency (or 'interaction frequency') between each fragment and the CFTR promoter was corrected for the interaction between two Hind III fragments within the ubiquitously expressed Excision repair cross-complementing rodent repair deficiency, complementation group 3 (ERCC3) locus, which has been reported to adopt the same spatial conformation in different tissues [18].

Detection of novel DHS at the CFTR locus in airway epithelial cells
We previously evaluated DHS across the CFTR locus by DNase chip in intestinal cell lines, primary male genital duct epithelial cells and primary tracheal and bronchial epithelial cells and identified both cell-type-specific and ubiquitous DHS. Here we focus on primary human tracheal and bronchial epithelial cells and compare these to airway epithelial cell lines that are frequently used in CF research. Figure 1A shows DNase chip data for skin fibroblasts that do not express CFTR, Beas2B cells that express almost undetectable levels, primary human tracheal epithelial (HTE) cells and a mixture of human bronchial and tracheal epithelial cells (NHBE), which express low levels of CFTR, 16HBE14o-and Calu3 cells that express very high levels of CFTR, which are comparable to those seen in intestinal cell lines. Relative levels of the CFTR transcript in each line are shown in Figure 1B. (These do not necessarily correlate with the levels of functional CFTR protein in each line.) The first notable feature of the DNase chip data in Figure 1A is the presence of several DHS at the 3' end of the CFTR locus and between CFTR and CTTNBP2 in the primary HTE and NHBE cells that are absent from the airway cell lines, despite the much higher levels of CFTR transcript in the latter (Fig. 1B). Also of note are several DHS that are apparently unique to HTE cells. All cell types that express CFTR show a DHS at the promoter as expected and NHBE, 16HBE14o-and Calu3 cells also show a DHS in intron 11 that we identified in intestinal cells [5]. Beas2B, NHBE and 16HBE14oϪ cells exhibit a DHS in intron 10 (10c, Refs. [19,20]) that was seen in many cells types irrespective of CFTR expression. Of particular interest are DHS seen at Ϫ44 kb and Ϫ35 kb from the translational start site in HTE, 16HBE14o-and Calu3 cells and at ϩ36.6 kb 3' to the last exon in HTE, NHBE and Calu3 cells. The DHS at ϩ48.9 kb from the 3' end of the CFTR gene, which is located in the last intron of the neighbouring CTTNBP2 locus and is seen in all the cell types analysed here, corresponds to a ubiquitous CTCF binding site that binds the cohesin complex subunit Rad21 [5]. Novel DHS were seen in Calu3 cells, in intron 1 (at 185 ϩ 20 kb) and intron 16 (at 3120 ϩ 1 kb), which have not been detected in any other cell type that we have examined. However, we previously identified a region of strong cross-species homology encompassing the 3120 ϩ 1 kb DHS and evaluated DNA-protein interactions in vitro, in the absence of a functional element [21]. Also of interest were DHS that were only detected in the primary HTE and NHBE airway epithelial cells in introns 18 (chr7:116,858,000-116,858,500), 19 (chr7:116,873,900-116, 874,800) and 23 (chr7:116,899,700-116,901,100). In an initial effort to reveal the functions of the elements located within these DHS, we first sought to determine epigenetic signatures at these sites. Because we previously characterized cell-type (intestinal) specific intronic enhancers in introns 1 and 11 of CFTR, we evaluated each of these novel airway DHS for a histone modification that is associated with enhancer elements and another associated with repressed chromatin.

Histone modifications across the CFTR locus in airway epithelial cells predict enhancers located within several DHS sequences
Chromatin immunoprecipitation was carried out on skin fibroblasts (a negative control for CFTR expression), primary HTE and NHBE cells, 16HBE14oϪ and Caco2 cells (a colon carcinoma cell line with very high CFTR expression levels, similar to those seen in 16HBE14oϪ), with antibodies specific for modifications seen at enhancers (H3K4me1; Fig. 2A) and in repressed chromatin (H3K27me3; Fig. 2B). More recently, the H3K27me3 mark was also shown to be enriched at poised developmental enhancers [22]. Primer sets for the SYBR green qPCR were designed to amplify regions within the DHS identified in the airway cell types together with other tissue-specific or ubiquitous DHS reported previously [5] (primer sequences are shown in Table S1). H3K4me1 was primarily evident at the CFTR promoter in the primary tracheal epithelial (HTE) cells with greatest enrichment at Ϫ2 kb and lesser enrichment at Ϫ3.4 kb and Ϫ0.5 kb with respect to the translational start site. The ChIP profile of NHBE cells at the promoter was very similar to that of Caco2 colon carcinoma cells, with highest enrichment at Ϫ2 kb and no enrichment at Ϫ0.5 kb. However, 16HBE14oϪ bronchial epithelial cells showed little H3K4me1 in the promoter region. Additional minor peaks of H3K4me1 were seen at the Ϫ35 kb DHS in both 16HBE14o-and Caco2 and intron 23, at 4374 ϩ 1.3 kb, primarily in Caco2 (with only very minor enrichment in 16HBE14oϪ). The enhancer-blocking insulator element at ϩ15.6 kb [6] shows H3K4me1 enrichment in multiple cell types, most notably HTE, NHBE and Caco2. Also of interest were minor peaks of H3K4me1 at DHS in introns 18, 19 and at ϩ36.6 kb, which were most evident in primary tracheal epithelial cells. H3K27me3 ChIP showed generally very low levels of this modification at the airway-specific DHS with minor enrichment at several elements, including in intron 18 and ϩ15.6 kb in 16HBE14oϪ cells. Of particular note in primary HTE and NHBE cells, H3K27me3 was abundant at the Ϫ3.4 kb and Ϫ2 kb promoter regions but not the Ϫ0.5 kb region. Fibroblasts, which lack CFTR expression showed H3K27me3 enrichment across all three sites evaluated within the promoter region and at the intron 11 region that is associated with enhancer function in intestinal and genital duct cells that express high levels of CFTR.

The ؊35 kb, ؊3.4 kb and intron 23 (4374 ؉ 1.3 kb) DHS contain cell-type-specific enhancers
To test our prediction that the DHS regions that were enriched for H3K4me1 encompassed enhancer elements, three DHS at Ϫ35 kb, Ϫ3.4 kb and in intron 23 (4374 ϩ 1.3 kb), were chosen for further study. Each region was amplified and cloned (in both orientations) into the enhancer site of the pGL3B 245 vector [16] in which luciferase reporter expression is driven by a 787 bp CFTR basal promoter fragment. These constructs were co-transfected with a renilla control plasmid into 16HBE14o- (Fig. 3A) and Caco2 cells (Fig. 3B) and relative luciferase expression measured. Additional constructs used in these assays were described previously [5] and contained DHS elements from intron 1 (185 ϩ10 kb), intron 10 (1716 ϩ 13.2/13.7 kb) and intron 11 (1811 ϩ 0.8 kb) of CFTR in the enhancer site of the vector. The introns 1 and 11 constructs contain intestinal-specific enhancers that are active in Caco2 cells but not in 16HBE14o-cells [16,23,24,5], while the intron 10a,b, elements do not encompass an enhancer [17,5]. In 16HBE14o-bronchial epithelial cells, the fragment encompassing the Ϫ35 kb DHS acted as a strong enhancer that significantly increased CFTR promoter activity 9-and 15-fold in forward and reverse orientations, respectively. The elements spanning the DHS at Ϫ3.4 kb and within intron 23 (4374 ϩ 1.3 kb) had more modest enhancer activity, with three-fold effect on the CFTR promoter in reverse orientation for Ϫ3.4 kb and both orientations for intron 23 and six-fold effect for Ϫ3.4 kb in the forward orientation. Two copies of the Ϫ3.4 kb element approximately doubled the enhancer activity of this element (data not shown). The Ϫ35 kb and Ϫ3.4 kb DHS regions had no enhancer activity in Caco2 cells, although the intron 23 element had a slight effect, increasing the promoter activity by 1.8-fold (Fig. 3B), which was not statistically significant. These data are consistent with the absence of the Ϫ35 kb and Ϫ3.4 kb DHS from the Caco2 cells, with a very minor DHS in intron 23 [5]. The data also suggest that DHS at Ϫ35 kb and Ϫ3.4 kb contain airway-selective enhancers while the intron 23 (4374 ϩ 1.3 kb) DHS encompasses an enhancer that may function in airway and intestinal epithelial cells, but is more effective in the airway.
Because we previously showed that the intestinal-specific enhancer elements within intron 1 and intron 11 function cooperatively when cloned together into the enhancer site of the pGL3 245 vector, we next generated constructs in which the Ϫ35 kb, Ϫ3.4kb and intron 23 elements were cloned into pGL3B 245 together with the intron 11 enhancer region. However, we saw no evidence of cooperation between these elements in either 16HBE14o-or Caco2 cells (data not shown). In contrast, when we combined the Ϫ35 kb, Ϫ3.4 kb and intron 23 elements together in the enhancer site of the pGL3 245 vector they showed cooperative interactions, with a 60-fold increase in enhancer activity in comparison to the promoter alone. The Ϫ35 kb and intron 23 elements combined independently in pairs with the Ϫ3.4 kb element also showed cooperative activity though the values were threefold lower than the three elements combined.

Longer range interactions across 800 kb spanning the CFTR locus
Because we observed cell-type-specific DHS in airway epithelial cells at Ϫ44 kb, Ϫ35 kb, ϩ21.5 kb and 36.6 kb DHS, which were distal to the enhancer blocking insulators characterized at Ϫ20.9 and ϩ15.6 kb with respect to the CFTR gene, we next used chromosome conformation capture (3C, Ref. [8]) to investigate whether these sites associated with the CFTR promoter by a looping mechanism. We also evaluated the expression of the flanking genes ASZ1 and CTTNBP2 in these cell types to determine whether the more distal DHS could be involved in the regulation of these loci. Semi-quantitative RT-PCR and microarray analysis demonstrated that ASZ1 was not expressed in NHBE, HTE and 16HBE14o-cells, although trace amounts were detected in Caco2 (data not shown). Thus, the 5' DHS are more likely to be associated with regulatory elements for CFTR than ASZ1. In contrast, CTTNBP2 transcripts were evident in variable amounts in NHBE, HTE and Caco2 cells but were not detected in 16HBE14o-. Thus, we cannot exclude the possibility that the DHS 3' to CFTR contain regulatory elements that influence CTTNBP2 expression instead of, or in addition to CFTR.
For the 3C experiments, we used a 'bait' Taqman probe and reverse primer located at the CFTR promoter (described previously, Ref. [7]) and forward primers close to the 3' end of Hind III fragments spanning from hg 17, chr7:116,387,846-117,180,423, a genomic distance of 790,420 kb. Real-time PCR reactions using the probe/reverse primer and each of the forward primers enabled quantification of ligation events (subsequently referred to as 'interaction frequency') between the CFTR promoter and specific distal regions within each sample. Hence, the forward primers in Hind III fragments enabled us to measure directly whether elements within these regions were physically associated with the CFTR promoter region. Figure 4 shows that, although no long-distance interactions are evident in skin fibroblasts, where the CFTR promoter is inactive, in NHBE cells there is a very slightly elevated interaction of the promoter with the Ϫ35 kb DHS and a fragment located at ϩ20 kb with respect to the end of the transcript, although these are likely not significant. In contrast, 5' to the locus in 16HBE14oϪ cells, although no interaction is event with the Ϫ35 kb DHS region, the Ϫ79.5 kb DHS [25] and a Hind III fragment encompassing Ϫ163 kb showed higher interaction frequencies with the CFTR promoter. 3' to the locus, strong interactions with the promoter were seen at ϩ20 kb and to a lesser extent at ϩ109 kb. A similar pattern of distal interactions was seen in Caco2 cells that we showed previously to demonstrate strong looping of intragenic and more proximal flanking regions of the CFTR locus [5]. Thus, despite the presence of enhancer blocking insulators at Ϫ20.9 and ϩ15.6 kb 5' and 3', respectively, to the CFTR gene, a higher order of chromatin interactions is seen in cell types that express high transcript levels, which brings more distal (Ͼ100 kb away from the locus) genomic regions into close association with the active CFTR promoter.

Discussion
Variation in the utilization of cis-acting regulatory elements is not uncommon as a mechanism for achieving cell-type-specific regulation of gene expression. Here we identify novel cis-acting elements associated with DHS in the active CFTR gene in primary airway epithelial cells (HTE and NHBE) or airway epithelial cell lines (16HBE14o-, Calu3 and Beas2B). These sites were not evident in colon carcinoma cell lines such as Caco2 that express very high levels of CFTR, more than 10,000-fold more than is seen in primary tracheal and bronchial epithelial cells (when estimated by qRT-PCR). However, these different DHS profiles cannot be accounted for solely by mechanisms that coordinate expression of high levels of CFTR because the bronchial epithelial cell lines 16HBE14oand Calu3 both express abundant CFTR, equivalent to that seen in Caco2 cells, but lack the DHS seen in primary airway epithelial cells. Of particular interest are DHS at Ϫ44 kb, Ϫ35 kb with respect to the translational start site and in introns 18 (chr7: 116,858,000-116,858,500), 19 (chr7:116,873,900-116,874,800) and 23 (chr7:116,899,700-116,901,100) that are evident in primary HTE and NHBE cells. The Ϫ44 kb and Ϫ35 kb DHS are also evident in 16HBE14oϪ and Calu3 cells. Also of interest was the DHS at Ϫ3.4 kb, which was distinct from the major promoter DHS in 16HBE14oϪ cells. Fig. 4 Long-range interactions between the CFTR promoter and specific DHS measured with q3C. The organization of the ASZ1, CFTR and CTTNBP2 loci are displayed above the graphs. The grey bar represents the 'bait' region of the CFTR promoter, which includes a primer and Taqman probe adjacent to the 5' Hind III site; this Hind III fragment spans the identified CFTR transcriptional start sites. The x-axis in each graph represents the location of the 3' end of each assayed Hind III fragment relative to the translational start site; the y-axis represents interaction frequency relative to the interaction frequency between two Hind III fragments within the ubiquitously expressed ERCC3 gene. Data for each cell type are from a single representative 3C experiment (each experiment performed at least twice in duplicate), error bars represent S.E.M. of at least two PCR reactions for each fragment.
The enrichment of H3K4me1 within DHS at Ϫ35 kb, Ϫ3.4 kb and intron 23 (4374 ϩ 1.3 kb) predicted that these sites might encompass enhancer elements, despite the fact that the Ϫ35 kb and intron 23 sites are evident in the primary airway cells, which show very low levels of CFTR expression. All three elements showed enhancer activity when cloned into a pGL3 vector, in which luciferase expression is driven by the CFTR basal promoter, and transfected into 16HBE14o-cells. The Ϫ35 kb site showed the most robust enhancer activity, with about 15-fold increase over the basal promoter alone in 16HBE14o-cells, and the element is inactive in Caco2 colon carcinoma cells, consistent with absence of the DHS from this line. For comparison it is notable that the intron 11 strong intestinal enhancer, which we described previously [5], augments CFTR promoter activity in Caco2 cells at a level comparable to that of the Ϫ35 kb enhancer in 16HBE14ocells (Fig. 3B), but lacks activity in 16HBE14o-cells. Together, these data demonstrate the cell-type selectivity of each site. The intron 23 element has modest enhancer activity in comparison and this is evident in 16HBE14oϪ cells but is not significant in Caco2 cells, which show a very minor DHS at this site [5]. The Ϫ3.4 kb DHS is of additional interest since it has moderate enhancer activity in 16HBE14o-cells but not in Caco2 where the DHS is not evident. However, this activity is orientation-dependent in the enhancer site of pGL3B only showing significance when the element is cloned in the forward orientation. Thus, it is not behaving as a classical enhancer and may encompass another type of regulatory element. Also of interest is the observation that the Ϫ35 kb, Ϫ3.4 kb and intron 23 elements cooperate when inserted together into the enhancer site of the pGL3B 245 construct. However, these elements do not cooperate with the intestinal-specific DHS in introns 1 and 11, which we showed previously to cooperate with each other in intestinal cells. This further supports our suggestion that different regulatory mechanisms control CFTR expression in the airway and the intestinal epithelium.
The question arises as to how the primary airway cells (HTE and NHBE) maintain CFTR expression at very low levels, when they exhibit DHS that encompass strong enhancer elements. A trivial explanation would be that although expression levels are very low in the airway cell cultures as a whole, small numbers of cells may achieve high levels of transcription by recruitment of these enhancer elements. Alternatively, a partial explanation may be provided by our ChIP data on the repressive chromatin mark H3K27me3, which is abundant at the Ϫ3.4 kb and Ϫ2 kb promoter regions but not the Ϫ0.5 kb region in these cells. In fibroblasts, where the CFTR locus is inactive, H3K27me3 enrichment is seen at all three sites evaluated within the promoter region but in contrast, there is no evidence for this histone modification in cells with a highly active CFTR locus, such as 16HBE14oϪ and Caco2. The significance of the H3K27me3 mark at promoter regions in the context of models of repressive transcription has been discussed elsewhere [22]. It is possible that there is a competition occurring at the CFTR promoter between Trithorax group (TrxG) activating proteins and Polycomb group (PcG) repressive proteins including polycomb repressive complexes 1 and 2 (PRC1 and PRC2). The latter protein generates the H3K27me3 mark, which can spread across chromatin domains. In contrast, TrxG proteins catalyse a different histone modification (H3K4me3) that activates transcription. Possibly fine control of this competition at the CFTR promoter can regulate low levels of expression (high H3K27me3 at the Ϫ3.4 and Ϫ2 kb DHS, but not at Ϫ0.5 kb), turn off promoter activity completely (high H3K27me3 spreading from the Ϫ2 kb site to the Ϫ0.5 kb site), or enable high levels of promoter activity where PRC2 is totally replaced by TrxG proteins and there is no evidence of H3K27me3.
Another question that arises from our data is how regulatory elements that lie distal to the enhancer blocking insulators that flank the CFTR locus can influence promoter activity. We previously identified and characterized a 5' insulator associated with a DHS at Ϫ20.9 kb to the gene, which binds CTCF and another ϩ15.6 kb 3' to the locus which does not [6]. The Ϫ20.9 kb site is not evident on the DNase chip data shown in Figure 1A due to the presence of adjacent repetitive elements which are excluded from the microarray. However, we showed the site was present in multiple cell types [25,26] and it is seen on DNase seq analysis of both HTE and NHBE cells (unpublished data). However, despite the presence of these enhancer blocking insulators, the long range 3C data shown in Figure 4 suggest that at least in airway cell types that express high levels of CFTR such as 16HBE14o-, chromosome looping enables regions that are up to ~100 kb 5' and 3' to the locus to be brought into close association with the gene promoter. These data are consistent with observations on HeLa and Caco2 cells [27] that demonstrated interaction of regions Ϫ80 kb 5' with the CFTR promoter. Moreover, they provide an explanation for the mechanism of action of the Ϫ35 kb enhancer element identified here in airway-specific regulation of CFTR expression, and demonstrate how it bypasses the Ϫ20.9 kb insulator element to interact with the promoter. Perhaps the classical enhancer-blocking assay [28] that we used to define the function of the CFTR insulators measures a function that is not active over these extremely long genomic distances in vivo.
In conclusion, we identified a novel set of cooperating enhancer elements that are associated with DHS within and flanking the CFTR gene in primary human tracheal and bronchial epithelial cells and are not found in many other cell types that express CFTR. We showed that though these sites may lie outside the insulator elements that flank CFTR they can still interact directly with the CFTR promoter by chromatin looping. Moreover, we showed that CFTR in primary airway epithelial cells appears to be regulated by different mechanisms from those seen in airway cell lines which have more than 10,000 times as much CFTR transcript. We suggest a mechanism involving local histone modifications at the CFTR promoter that may cause partial transcriptional repression in these primary airway cells, thus maintaining transcripts in low abundance. Future work will identify cell-type-specific trans-acting factors that mediate the function of these airwayspecific elements.