A NF-κB-Dependent Dual Promoter-Enhancer Initiates the Lipopolysaccharide-Mediated Transcriptional Activation of the Chicken Lysozyme in Macrophages

The transcriptional activation of the chicken lysozyme gene (cLys) by lipopolysaccharide (LPS) in macrophages is dependent on transcription of a LPS-Inducible Non-Coding RNA (LINoCR) triggering eviction of the CCCTC-binding factor (CTCF) from a negative regulatory element upstream of the lysozyme transcription start site. LINoCR is transcribed from a promoter originally characterized as a hormone response enhancer in the oviduct. Herein, we report the characterization of this cis-regulatory element (CRE). In activated macrophages, a 60 bp region bound by NF-κB, AP1 and C/EBPβ controls this CRE, which is strictly dependent on NF-κB binding for its activity in luciferase assays. Moreover, the serine/threonine kinase IKKα, known to be recruited by NF-κB to NF-κB-dependent genes is found at the CRE and within the transcribing regions of both cLys and LINoCR. Such repartition suggests a simultaneous promoter and enhancer activity of this CRE, initiating cLys transcriptional activation and driving CTCF eviction. This recruitment was transient despite persistence of both cLys transcription and NF-κB binding to the CRE. Finally, comparing cLys with other LPS-inducible genes indicates that IKKα detection within transcribing regions can be correlated with the presence of the elongating form of RNA polymerase II or concentrated in the 3′ end of the gene.


Introduction
Genes transcription is controlled by CREs, which are, when activated, nucleosome free regions occupied by transcription factors and identified in vivo as DNAse I hypersensitives sites (DHS) [1]. A classical view separates these elements into different categories depending on their position from the transcription start site (TSS) of genes, their sequence and their chromatin signature. However, global analyses of the transcriptome suggest that the function of these CREs is not commonly restricted to a single category, genomic regions with dual promoter and enhancer activities appearing to be widespread within the genome. For example, a recent study looking at transcription sites located outside protein-coding regions in macrophages activated by endotoxins, found 70% of extragenic RNA polymerase II (RNAPII) peaks associated with genomic regions with a chromatin signature of enhancers [2], these elements generating very low abundance non-coding transcripts, suggested to be ''junk'' RNA. However, the idea that ''enhancer-associated'' extragenic transcription would represent only noise has already been challenged by several studies. Extragenic transcripts within locus control regions (LCR), these distal regions composed of several CREs able to enhance the expression of linked genes to physiological levels in a tissue-specific and copy number-dependent manner, have been identified some time ago and are believed to play a role in the chromatin remodelling observed over these regions [3][4][5]. More recently, the link between non-coding RNA transcription from dual promoter/enhancer elements and chromatin remodelling has been established for two chicken genes mim-1 and cLys [6,7]. However these studies did not determine if these CREs are behaving simultaneously or successively as promoter and enhancer. CLys is a marker of macrophage differentiation, which rapidly responds to pro-inflammatory agents like LPS and its expression is controlled by three enhancer elements situated 26.1 kb, 23.9 kb and 22.7 kb upstream of the transcription start site, a complex promoter and a silencer element at 22.4 kb (figure 1A) [8]. We have reported that cLys expression activation was preceded by the transient transcription of LINoCR from a promoter 21.9 kb upstream of cLys TSS, this transcription being necessary for nucleosome reorganisation and eviction of the enhancer blocker protein CTCF from the silencer element [7]. Interestingly, this promoter was originally identified as a hormone response enhancer element functional in the oviduct and bound by estrogen, glucocorticoid and progesterone receptors [9,10]. However, we did not fully establish that this 21.9 kb element was driving any enhancer activity in macrophages.
In these cells, activation of the 21.9 kb element and subsequent transcription of LINoCR was correlated with accumulation of the protein kinase IKKa and histone H3 serine 10 phosphorylation (H3S10p) within LINoCR transcribed region [7]. IKKa is part of the IKK complex controlling the release of the NF-kB transcription factor into the nucleus to stimulate transcription of its target genes in response to pro-inflammatory stimuli. In the nucleus, NF-kB interacts with several chromatin modifiers including the histone H3K4 methyltransferase Set7/9 [11], CBP/p300 [12], TIP60 [13,14] and also IKKa for which specific chromatin modifying activity has been described [15][16][17]. At the promoter of NF-kB-dependent genes, IKKa phosphorylation of H3S10 is important to initiate transcription elongation [15,16,18]. Furthermore, IKKa binds to the phosphorylated RNA polymerase II C-Terminal Domain (RNAPII CTD) to target HP1cserine 93 (HP1cS93) [17]. HP1c is part of the HP1 family of proteins together with HP1a and HP1b in vertebrates. HP1 proteins are commonly associated with heterochromatin formation, as they are recruited to methylated lysine 9 of histone H3. They can in turn recruit methyltransferase to propagate silencing marks along chromatin [19,20]. In contrast to HP1a and HP1b, HP1c is also located in euchromatin [21] where it can be found phosphorylated at serine 93 [22], a post-translational modification abrogating the transcriptionally repressive function of HP1c [23]. HP1c also interacts with the phosphorylated RNAPII CTD and recruits the FACT (FAcilitates Chromatin Transcription) complex to RNAPII [24,25]. At NF-kB-dependent genes, HP1c controls IKKa association with chromatin and IKKa-dependent phosphorylation of the histone H3.3S31 [17].
To clarify the function of the 21.9 kb CRE, we undertook a detailed characterization of this element in macrophages. We identified a unique 60 bp region within a larger DNase I and Micrococcal Nuclease (MNase) hypersensitive domain occupied by AP1, C/EBPb and NF-kB (p65) transcription factors after LPS treatment, the latter providing a rationale for IKKa detection within the coding region of LINoCR [7]. These transcription factors act cooperatively to fully activate this CRE. Additional luciferase reporter assays and in vivo chromatin immunoprecipitation analyses reveal that this 60 bp transcription factor cluster possesses concomitant promoter and enhancer activities. In addition, in contrast with what we have described for other NF-kB-dependent genes [17], IKKa is transiently recruited to both LINoCR and cLys transcribed regions. This loss of IKKa concomitant with LINoCR transcriptional inactivation is the only change identified at the 21.9 kb CRE, for which the chromatin structure and the transcription factors occupancy are still identical during or after LINoCR transcription. This result suggests an important role of IKKa to mediate the 21.9 kb CRE promoter activity and that the 21.9 kb CRE is important to initiate transcription of cLys but not to maintain this expression after CTCF eviction from the 22.4 kb silencer. This observation is linked with HP1c poor incorporation into transcribing chromatin at cLys locus. These data may provide a paradigm for the modus operandi of CREs with dual promoter/enhancer activity and reinforce the idea that HP1c controls IKKa associated transcription.

Cell Culture
The chicken cell lines monocytes HD11 [26] and erythroblasts HD37 [27] and the mouse cell line RAW264.7 were grown in Dulbecco's modified Eagle's medium as previously described [17,28]. Mouse primary macrophages were obtained from bone marrow by culturing in Iscove's Modified Dulbecco's Medium containing 10% Foetal Calf Serum and Penicillin-Streptomycin and 10% L cell conditioned medium containing M-CSF [29] for 7 days. Where indicated, cells were treated with 1 mg/ml LPS (Sigma).

Nucleosome Mapping by Indirect End Labeling
DNase I treatment of cells and naked DNA was performed as described previously [28]. MNase digestions of HD11 and indirect end labeling were performed using isolated nuclei as described previously [30]. With 10 mg of each, different DNA preparations digested with 20 U SphI (New England Biolabs) for 3 hr at 37uC and stopped with 5 ml loading dye 20% Ficoll (Sigma), 1% SDS (Sigma), and 0.05% bromophenol blue (Sigma). The probe abutting the SphI site (23165 to 22865 bp) was prepared by PCR using a plasmid containing the full sequence of the lysozyme locus as a template with the following primers: fwd, TACTTAG-GAGGGTGTGTGTG and rev, GCACCTTGAAGATTTGTT. The probe was gel purified using a QIAquick Gel Extraction Kit (QIAGEN). Bands were quantified from the images generated on the pharosFX molecular imager using Quantity One software (BioRad).

Cloning, Mutagenesis and Transient Transfection
DNA fragments carrying the lysozyme promoter (2376 to +17 bp) and the 1.9 kb element (22132 to 21877 bp) were cloned into the luciferase vector pXPG [32]. Mutants were generated by PCR amplification in the following 50 ml reaction mixtures: 1X Pfu Turbo buffer (Stratagene), pXPG-1.9AS [7] as a template, 125 ng of both forward and reverse primers, 0.25 mM dNTPs and 2.5 U Pfu Turbo (Stratagene). PCR amplification conditions were as follows: (1) denaturation at 95uC for 30 sec, (2) 16 cycles of denaturation at 95uC for 50 sec, annealing at 55uC for 50 sec and extension at 72uC for 7 min and (3) a final extension at 72uC for 7 min. Next 1 ml of 20 U/ml DpnI (NEB) was added directly to the PCR mix, incubated for 80 min at 37uC and then heat inactivated by incubation at 80uC for 20 min. Then 50 ml of stable 3 electro-competent bacteria (Invitrogen) were transformed, incubated and plated according to manufacturer recommendations. Transfection and luciferase assays were performed as previously described [7].

Preparation of Nuclear Extracts and Electrophoretic
Mobility Shift Assay 2610 7 HD11 cells, unstimulated or stimulated with 1 mg/ml of LPS for 1 hr and nuclear extracts prepared as previously described [33]. EMSAs were performed using end-labelled, double-stranded synthetic oligonucleotides. 2 mg of nuclear extracts was diluted in EMSA buffer [33]. The buffer-diluted samples then formed complexes with either 50 ng of unlabelled competitors or 1 mg specific antibodies, anti-c-FOS (Sc-253, Santa Cruz) and anti-C/ EBPa (Sc-61, Santa Cruz) during an incubation for 15 min at 25uC before the addition of 0.5 ng 32 P c labelled probe. After incubation with the probe for 15 min at 25uC, the samples were separated on a 5% acrylamide gel (37.5:1), 0.5 X TBE, 1/1000 TEMED and 0.1% APS. The gel was fixed, dried, exposed with a K-Screens (KODAK) for 16 hrs and analysed on pharosFX molecular imager (Biorad).

Identification of a 60 bp Region within the 21.9 kb CRE Occupied by Transcription Factors after LPS Treatment in Macrophages
We have previously established that the eviction of the insulatorassociated protein CTCF from its binding site and subsequent nucleosome movement over this site was dependent on transcription of LINoCR at the cLys locus in activated macrophages [7]. LINoCR is firing from a CRE located 21.9 kb upstream of cLys TSS, this element being previously described as a hormone response enhancer in the oviduct [9,10] (figure 1a). These observations were making this CRE, one of the first characterized dual promoter/enhancer for which the associated transcription of the non-coding RNA was known to be functional. However, our previous work did not clearly establish that this element was also acting as an enhancer in macrophages. To determine, if this CRE could represent a paradigm of dual promoter/enhancer elements or if, in macrophages, it was only a promoter, we decided to undertake a more detailed characterization of this element. First, we looked at the nucleosome content in this region before and after LPS treatment in the chicken macrophage cell line HD11 compared to the 26.1 kb enhancer by chromatin immunoprecipitation (ChIP) using an antibody against total histone H3. We detected a progressive reduction in nucleosome content starting 20 min after LPS treatment and reaching a plateau after 45 min (figure 1b). Then, we performed low resolution DNase I hypersensitive site (DHS) and nucleosome (MNase) mapping analyses in response to LPS in the HD11 cell line (figure 2). Both MNase and DNase I mapping revealed a large region between 21.9 kb and 22.1 kb, becoming nucleosome free and DNase I accessible as early as 30 min post LPS stimulation (figures 2a and 2b). Additional quantification of these two southern blots indicated that within the 200 bp delimiting the 21.9 kb element a region between 60 to 100 bp was more protected against DNase I and MNase (figures 2c and 2d). Taken together, these results confirm that the 21.9 kb element is inactivated in absence of proinflammatory stimuli and highlight a small region within this element protected from MNase and DNase I digestions.
To map precisely where transcription factors bind in this element, we undertook some Dimethyl Sulphate (DMS) in vivo footprinting analysis. A first set of primers provided information on transcription factors occupancy in the 22150 bp to 22000 bp region (figure 3). In untreated HD11 cells, the DMS modification pattern on both strands was indistinguishable from the nonexpressing HD37 and the G-reaction performed on naked HD11 genomic DNA. This confirms that myeloid specific transcription factors did not occupy the upstream region of the 21.9 kb element prior to LPS activation. Immediately post LPS stimulation, there was hyper-reactivity of Guanine bases at three positions, 22017/ 18, 22028 and 22038/39 bp on the anti-sense strand (figure 3a). All three footprints were present within 30 min and remained throughout the 240 min time course. Analysis of the sense strand confirmed these observations. LPS activation produced differential hypersensitivity at three sites 22040/41, 22024/23 and 22015 bp (figure 3b). However, on the sense strand, two Guanine bases, at 22040/41 and 22015 bp, were protected and one, at 22024/23 bp, was hyper-reactive. Sequence analysis associated a consensus-binding site for NF-kB between 22015 and 22017/ 18 bp and for C/EBP between 22017/18 and 22028 bp. Further upstream, although the footprint was weak, there were indications of transcription factor binding on the sense strand at 22061 and 22065 bp. The sequence surrounding this footprint revealed a potential AP1 binding site. This is consistent with the characteristic weak footprints of AP1 factors and with previous ChIP experiments detecting Fos binding to the 21.9 kb region immediately post LPS stimulation [7]. A second set of primers allowed to analyse the transcription factor occupancy post LPS stimulation of the 22000 to 21850 bp region. However, apart from the previously identified potential C/EBP and NF-kB binding sites we did not detect any reproducible DMS footprints in this region (figure S1). Having analysed the entire nucleosome free region of the 21.9 kb element it was clear that the key transcription factors were clustered within 60 bp in the upstream section of the DHS. In addition, the low-resolution analyses confirmed that the DHS and MNase sensitive domain represent the region immediately upstream of LINoCR and therefore the promoter of this ncRNA, our previous experiments having identified LINoCR TSS 22.12 kb upstream of cLys TSS [7]. Further inspection of the upstream sequence of the proposed transcription factor binding site cluster revealed a non-classical TATA box, TACATAAA, located 21 bp from the proposed AP1 binding site [34]. In summary, the DMS footprints implicate AP1, C/EBP and NF-kB binding to sites in the upstream region of the 21.9 kb dual promoter enhancer element. The positioning of the transcription factor binding sites relative to the ncRNA transcription start site and the proposed TATA box were consistent with the structure of a promoter.
NF-kB Occupies in vivo the 21.9 kb CRE and Recruits IKKa to both LINoCR and cLys Transcribing Regions Simultaneously The in vivo DMS footprinting revealed the specific sites within the 21.9 kb element at which transcription factors were binding in response to LPS. The subsequent analysis of the sequence encompassing the footprints implied that NF-kB, C/EBP and AP1 were binding. This was confirmed by electrophoretic mobility shift assays (EMSA) (figures S2 and S3a-d). In these experiments, NF-kB appeared to bind only weakly to the 21.9 kb element. However, additional EMSA experiments with a 40 bp oligonucleotides encompassing both the proposed C/EBP and NF-kB sites showed a cooperative binding of both proteins to this CRE (figure S3e).
We have previously shown that the 21.9 kb CRE cloned immediately upstream of cLys promoter in sense orientation was increasing this promoter LPS-dependent inducibility suggesting that this CRE was an enhancer in macrophages [7]. To complete this observation, we cloned the 21.9 kb CRE downstream of the luciferase polyA signal in a cLys promoter driven pXPG reporter vector, to rule out the possibility that the 21.9 Kb element cloned immediately upstream of cLys promoter generated an extended promoter. In agreement with previous observations, LPS stimulated weakly cLys promoter activity ( figure 4). Furthermore, in the resting HD11 cells, cLys promoter activity with the 21.9 kb element cloned in 39 was equivalent to cLys promoter alone. However, upon LPS incubation, cLys promoter activity was increased 4.8 fold when the 21.9 kb element was present as opposed to 2 fold LPS induction with just cLys promoter (figure 4a). Thus, the 21.9 kb element enhanced the cLys promoter's LPS inducibility by approximately 2.5 fold. These experiments established that the 21.9 kb element was a LPS-inducible enhancer in macrophages. The strict LPS dependence of both the 21.9 kb promoter and enhancer capabilities implies that the inflammatory response regulates the transcription factor(s) required for its activity.
Having determined the location, identity and binding ability of transcription factors present at the 21.9 kb element in the activated HD11 cells, their individual contribution to promoter activity was assessed in transient transfection. An extensive set of constructs containing the individual or combination of binding inactivation mutations, revealed by EMSAs, were cloned into the pXPG luciferase reporter plasmid and transfected into HD11 cells (figure 4b). The mutation preventing NF-kB binding to the 21.9 kb element completely abolishes both basal and LPS inducible activity of this promoter. In addition, mutations of C/ EBP, X or AP1 show similar impact on this promoter's basal and LPS-inducible expression, expression being further reduced by double mutants C/EBP and AP1 or X and AP1. Taken together, these results show that NF-kB is essential for the promoter activity of the 21.9 kb element but does not act alone as each individual mutation has a significant impact on the promoter activity in the transient transfection assays.
We have shown previously that C/EBPb and Fos were binding to the 21.9 kb element after LPS treatment in HD11 [7]. In this work, C/EBPb and Fos were found enriched at the 21.9 kb element from 20 min post LPS treatment in agreement with experiments above describing a quick activation of this element in response to LPS. If this element is activated early after stimulation, we did not detect any change in cLys basal mRNA level before 45 min post LPS. Because detectable changes in total mRNA level are delayed compared to transcriptional activation, we could not determine if the 21.9 kb element was first acting as a promoter and then as an enhancer after LINoCR expression was stopped or if this CRE could act simultaneously as promoter and enhancer. Using a transgenic mouse line harbouring the 21 kb cLys domain inserted into the HPRT locus [35], we first performed additional chromatin immunoprecipitation experiments looking at the NF-kB protein family member p65 in primary macrophages. As expected we detected enrichment for p65 in both cLys promoter and the 21.9 kb element after 30 min and 120 min of LPS treatment (figures 5a and S4c). Interestingly if p65 binding was stable after short-term and long-term LPS treatment, total RNAPII or elongating RNAPII (RNAPII S2p) occupancies were higher in cLys coding region (0.2 kb to 3.6 kb) after 30 min than after 120 min of LPS treatment (figures 5b, S4a and S4b). As expected, RNAPII was only detectable within LINoCR transcrib-ing region (21.9 kb to 23.2 kb) after short-term LPS treatment (figures 5b, S4a and S4b). In addition, this RNAPII enrichment was correlated with IKKa recruitment to the transcribed regions of both cLys and LINoCR (figures 5c and S4d). More detailed comparisons between p65, IKKa and RNAPII recruitment after 30 min of LPS treatment highlight that IKKa correlates with RNAPII but not with NF-kB recruitment except at the 21.9 kb element where both IKKa and p65 but not RNAPII S2p are detected (figures 6a and 6b). This observation suggests that IKKa is recruited to cLys locus by NF-kB bound to the 21.9 kb element and not to cLys promoter. In addition, this result indicates that LINoCR and cLys are transcribed simultaneously and that IKKadependent transcription is restricted to early time points.

IKKa Recruitment to NF-kB-dependent Genes can Follow Different Kinetics
IKKa enrichment profile was not temporally and spatially comparable with our previous observations made for TNF, Ccl3 or Il1b other NF-kB-dependent genes [17]. At these genes, IKKa accumulates in the 39 end, with 10 times more IKKa detected in 39 compared to the promoter region, binds to chromatin in an HP1c-dependent manner and is still detectable after 2 h of LPS stimulation. We hypothesised that IKKa was not interacting with chromatin because of the poor incorporation of HP1c to transcribing chromatin within cLys locus. Indeed, we detected only 3 to 4 times HP1c enrichment in the coding region of cLys and no enrichment at all in LINoCR-transcribing region compared to a CTCF-binding site located upstream of the murine IL6 TSS where HP1c was undetectable in macrophages (figures 5d and S4e). TNF, Ccl3 or Il1b respond to LPS with similar kinetics. To complete our analysis, we chose two additional NF-kBdependent genes with different temporal patterns of expression, BTG2 and IL10, for which expression peaks before or after 2 h of LPS treatment respectively (figure S5). ChIP experiments performed 30 min, 2 h and 4 h post LPS stimulation showed p65 binding to the promoter of both genes (figure 7a). Moreover, enrichment for HP1c correlated with RNAPII S2p at these genes in agreement with what we observed for TNF and cLys (figures 7c, 7d and S6a), indicating that the amount of HP1c detected within the coding regions of these genes shadows the rate of transcription. The analysis of IKKa recruitment to these loci unveiled a more heterogeneous association of this kinase with the different NF-kBdependent genes studied. At the BTG2 promoter, the dynamics of p65 and IKKa occupancy mimicked our observation for the cLys 21.9 Kb CRE. In parallel with BTG2 expression, IKKa, HP1c and RNAPII S2p were only detected within the gene 30 min after LPS stimulation (figures 7b-c). At the 39 end of the gene, IKKa enrichment was 4 fold higher than at the promoter whereas HP1c was also 4 fold more associated with BTG2 than with cLys but 4 fold less than with TNF (figures 7b and 7c). In contrast, the promoter of IL10, for which LPS-mediated transcription is delayed compared to BTG2, was not bound by p65 before 2 h post LPS stimulation ( figure 7a). In addition, if p65, HP1c and RNAPII S2p enrichments within this locus were comparable at 4 h with the one observed at BTG2 locus after 30 min of LPS treatment, IKKa was 2.5 to 3 fold less recruited to IL10 locus compared to the latter (figures 7a-d). These data would suggest that IKKa is mainly playing a role during the earliest stage of LPS-mediated transcription. However, we have shown previously that the presence of IKKa at transcribing NF-kB-dependent genes could be maintained after 2 h of transcription [17]. At these genes, the accumulation of IKKa downstream of the transcription end site (TES) is independent of RNAPII S2P (figure S6b). Taken together, these data highlight the fact that IKKa recruitment dynamics can obey different rules during transcription and suggest that HP1c controls this dynamics.

Discussion
Enhancers with a ''cryptic'' promoter activity are widespread along the genome, but the functions of the produced ncRNAs, if functions at all, are still unknown [2]. Concomitantly, two dual promoter/enhancer CREs have been described, for which remodelling of the surrounding chromatin domains depends on their promoter activity and associated transcription of ncRNAs [6,7]. However in these examples, the exact dynamic of promoter and/or enhancer activity during transcription of the associated protein-coding gene is still unclear. For example, cLys 21.9 kb CRE promoter and enhancer activities have been documented but in different cell types, enhancer in the oviduct and promoter in macrophages [7,9,10]. In this study, we determined that the 21.9 kb CRE is a LPS-responsive element in macrophages controlled by a unique 60 bp transcription factors cluster occupied by AP1, C/EBP, NF-kB and a still unidentified factor, additional ChIP experiments confirming that Fos, C/EBPb [7] and NF-kB (p65) were binding in vivo to this CRE. The key inflammatory factor NF-kB is the main regulator of this CRE, which does not show any activity in absence of NF-kB binding in luciferase assays. If NF-kB is necessary, AP1, C/EBP and an undetermined factor act synergistically with NF-kB to provide full activity of this element. Such cooperative function between AP1, C/EBPb and NF-kB has been described for multiple LPS-inducible genes including IL-6, CXCL8 or IL-8 [36][37][38]. In vitro, NF-kB binding is stronger in presence of C/EBP bound next to its binding site, like observed for IL-6 and IL-8 promoters, where C/EBPb and NF-kB directly associate with each other [39]. In transient transfection, C/EBP mutant alters only slightly the 21.9 kb promoter activity suggesting that C/EBP-independent binding of NF-kB is stronger in vivo compared to EMSA or that the alteration of the 21.9 kb promoter activity in C/EBP mutant is mainly caused by the reduction of NF-kB binding. The fact that cLys contains three functional C/EBP binding sites bound by C/EBPb [7,40] and shows only a slight increase in activity after LPS treatment reinforce the second hypothesis.
This cluster of LPS-inducible transcription factors is distinct from the steroid receptor binding sites involved in regulating this CRE in the oviduct [9] and from the mapped progesterone transcription factor binding sites, which are at a greater distance from LINoCR TSS [41], but still within the 200 bp nucleosome free region. These differences would argue that this CRE accomplishes different functions in the oviduct and in macrophages. However, additional transient transfections with the 21.9 kb CRE cloned at the 39 end of a luciferase gene driven by cLys promoter confirms that the same 60 bp transcription factor cluster also drives a LPS-inducible enhancer activity in macrophages. Because our previous data suggest that cLys transcriptional activation was a strict consequence of CTCF eviction induced by LINoCR, the 21.9 kb CRE was expected to act successively as a promoter and then as an enhancer consequently to LINoCR expression being turned off. Surprisingly, this is not the model this study is revealing. ChIP experiments show p65 binding to cLys promoter and the 21.9 kb CRE after both 30 min and 120 min post LPS treatments. In contrast, IKKa is only recruited to the 21.9 kb CRE and not to cLys promoter and this recruitment is seen only after the shortest period of LPS treatment. The absence of IKKa at cLys promoter could be explained by the fact that the NF-kB heterodimer p65:c-Rel and not p65:p50 occupies cLys promoter [42]. If p65:p50 dimer recruits IKKa to the promoter of NF-kB-dependent genes [15,16], p65:c-Rel has been shown to recruit IkBb to a selected group of genes in response to LPS including TNF and IL1b [43,44]. Furthermore, IKKa is detected in both transcribing regions probably bound to the phosphorylated RNAPII CTD as reported previously [17]. This observation suggests a direct connection between cLys promoter and the 21.9 kb CRE and simultaneous promoter and enhancer activities of this element. This contact could be mediated by C/EBPb, which has been shown to form long rang interaction and DNA looping [45]. If the short distance between these two elements does not allow chromosome conformation capture (3C) analysis, the hypothesis of a direct interaction is reinforced by recent views regarding the chromatin organisation in euchromatin. These two nucleosome free regions should be indeed physically in close proximity within the 30 nm chromatin fiber structure, which has been shown to be conserved at transcribing regions [46].
CLys expression is maintained several hours after that LINoCR has been shut down [7], but this expression is IKKa independent. The reasons, for IKKa disappearance from cLys locus, are unclear since the transcription factors and especially NF-kB (p65) still occupy their binding sites at the 21.9 kb CRE several hours post LPS treatment. It could be explained by the inactivation of the transcription factors, bound to this regulatory element, by posttranslational modifications. For example, p65 can be activated by Msk1-dependent phosphorylation of its serine 276 and deactivated by PP2A phosphatase without affecting p65 DNA binding [47,48]. Post-translational modifications also regulate C/EBPb activity [49][50][51], C/EBPb preventing NF-kB phosphorylation and thus its activation in TNF tolerant cells [52]. Finally, the composition of the AP1 dimer alters the transcriptional activation capability of this transcription factor [53]. Together, these results suggest that the transcription factors bound cluster within the 21.9 kb CRE can be inactivated without observable changes in DNA binding. In this model, the 21.9 kb CRE would initiate transcription of cLys and would concomitantly abrogate CTCF enhancer blocker activity. The maintained expression of cLys would thereafter be controlled by the three-enhancer elements upstream of CTCF, the 21.9 kb element playing a minor role or no role at all after CTCF eviction. Such a transient role of IKKa is observed for BTG2, for which p65 binding to the promoter is maintained after loss of IKKa and in absence of transcription. However IL10 expression, which is delayed compared to the other analysed genes, is concomitant with p65 binding to its promoter arguing for a role of this protein in transcription of late LPS responsive genes. NF-kB and IKKa binding to the IL10 promoter is induced by the HIV-1 TAT protein but observed 30 min after stimulation [54]. In contrast, when compared to BTG2, IKKa recruitment to IL10 locus after LPS stimulation is poor suggesting that NF-kB (p65) can activate genes independently of IKKa or that p65 is playing a minor role in IL10 expression, as suggested by other studies [55,56]. Nevertheless, the presence of IKKa can be measurable after 15 min of LPS stimulation and maintained beyond 2 h within transcribing regions in correlation with NF-kB (p65) promoter occupancy as observed for TNF, Ccl3 or IL1b [17]. This ''extended'' detection of IKKa correlates with a dense distribution of HP1c throughout the transcribing regions of these genes [17]. As described for IKKa, HP1c directly interacts with the elongating polymerase [24,25] and closely mimics the distribution of the latter during transcription. In addition, chromatin accumulation of IKKa downstream of the TNF transcription end site in activated macrophages is HP1c-dependent [17]. This chromatin association of IKKa is not observed for cLys and BTG2 for which the distribution of this kinase with both HP1c and RNAPII S2p are correlated.
In macrophages, the cLys 21.9 kb CRE is driven by a 60 bp transcription factors cluster and is a NF-kB/IKKa-bound dual promoter and enhancer with both activities being simultaneous. Its promoter activity is associated with chromatin remodelling of the transcribed region and CTCF eviction and its enhancer activity is characterised by initiation of cLys transcription. In addition, we have determined that IKKa association with coding regions can be restricted to the earliest stage of LPS-mediated macrophages activation, in contrast with genes, like TNF, which display chromatin associated IKKa in the 39 end of the gene [17]. These two scenarios suggest a NF-kB-dependent recruitment of IKKa during the initiation step of transcription, followed for some genes  Figure S1 In vivo DMS footprinting of the distal part of the 21.9 kb promoter. HD11 cells were, in order from left to right, either unstimulated or LPS (1 mg/ml) stimulated for 30 min, 60 min or 240 min. Cells were then treated with DMS before the isolation of genomic DNA for hot piperidne cleavage and LM-PCR analysis. The HD37 erythroid cell line which do not express clys and the naked HD11 genomic DNA, G reaction, reference sequence are also shown