Changes in H3K27ac following lipopolysaccharide stimulation of nasopharyngeal epithelial cells
- 149 Downloads
The epithelium is the first line of defense against pathogens. Notably the epithelial cells lining the respiratory track are crucial in sensing airborne microbes and mounting an effective immune response via the expression of target genes such as cytokines and chemokines. Gene expression regulation following microbial recognition is partly regulated by chromatin re-organization and has been described in immune cells but data from epithelial cells is not as detailed. Here, we report genome-wide changes of the H3K27ac mark, characteristic of activated enhancers and promoters, after stimulation of nasopharyngeal epithelial cells with the bacterial endotoxin Lipopolysaccharide (LPS).
In this study, we have identified 626 regions where the H3K27ac mark showed reproducible increase following LPS induction in epithelial cells. This indicated that sensing of LPS led to opening of the chromatin in our system. Moreover, this phenomenon seemed to happen extensively at enhancers regions and we could observe instances of Super-enhancer formation. As expected, LPS-increased H3K27ac regions were found in the vicinity of genes relevant for LPS response and these changes correlated with up-regulation of their expression. In addition, we found the induction of H3K27ac mark to overlap with the binding of one of the NF-kB members and key regulator of the innate immune response, RELA, following LPS sensing. Indeed, inhibiting the NF-kB pathway abolished the deposition of H3K27ac at the TNF locus, a target of RELA, suggesting that these two phenomena are associated.
Enhancers’ selection and activation following microbial or inflammatory stimuli has been described previously and shown to be mediated via the NF-kB pathway. Here, we demonstrate that this is also likely to occur in the case of LPS-sensing by nasopharyngeal epithelial cells as well. In addition to validating previous findings, we generated a valuable data set relevant to the host immune response to epithelial cell colonizing or infecting pathogens.
KeywordsH3K27ac Enhancers LPS Epithelial cells RELA
Histone 3 Lysine 27 acetylation
Log 2 (Fold change)
Nuclear Factor kappa B
Transcription Start Site
The epithelium constitutes a natural barrier against pathogens and is particularly important in fighting infection. Indeed, epithelial cells are the first to detect microbes and are crucial in mounting an effective innate immune response as well as interacting with more specialized immune cells to initiate the adaptive immunity processes . Especially, the epithelial cells in the respiratory tract are critical in regards to defense against pathogens entering the body through inhalation .
These cells recognize specific microbial components through pattern recognition receptors, mainly Toll-like receptors (TLR) . Upon recognition, signaling pathways are activated and lead to the expression regulation of genes such as cytokines, chemokines and antimicrobials by transcription factors . Regulation by transcription factors goes together with chromatin organization which helps coordinate gene expression. Epigenetic changes following stimulation has received a vast interest recently and was found to be determined by the cell type targeted as well as the environmental signals applied. Indeed, selection of regulatory regions, particularly enhancers, following stimulation consists of interplay between lineage-determining and signal-dependent transcription factors that together achieve specific cell response to a particular stimulus .
Among the various changes and marks added onto histone proteins following re-organization after stimulation, Histone 3 Lysine 27 acetylation (H3K27ac) seems to be very dynamic and has been shown to be important in regulation of the immune response [6, 7]. This mark indicates active enhancers and promoters .
The changes in chromatin following microbial stimuli has been mostly studied in immune cells, most often in mouse macrophages [7, 9, 10]. Here, we studied a human nasopharyngeal epithelial cell line, Detroit 562 cells, used extensively in studying airborne infectious diseases such as infection with Neisseria meningitidis , Streptococcus pneumoniae  and influenza virus  among others. We challenged these cells with Lipopolysaccharide (LPS), a potent gram negative bacterial endotoxin that targets TLR4, and described the changes in H3K27ac mark induced by the stimulation.
We have recently investigated the response of one of the master regulator of innate immunity, Nuclear Factor kappa B (NF-kB) member RELA, to the same stimulus and in the same cell line . Thus, we were able to examine LPS-induced H3K27ac regions identified with regards to binding of RELA, one of the NF-kB member activated downstream of TLR4. This factor has been previously shown to be crucial for selecting enhancers and recruiting co-factors following stimulation [15, 16].
The data set described here characterizes the response to LPS by epithelial cells. LPS is commonly used as a model for bacterial infection and nasopharyngeal epithelial cells are particularly relevant for airborne pathogen infection. Thus, this study may provide insights into the host response to bacteria responsible for infectious respiratory diseases.
Expectedly, Gene Ontology analysis on the genes assigned to LPS-increased H3K27ac regions revealed enrichment for relevant biological processes terms such as “cellular response to Lipopolysaccharide” (−log10 P-value = 13.35) or “response to bacterium” (-log10P-value = 13.03), coming up as some of the top hits (Fig. 1c and Additional file 3 for all terms). These results confirm that the LPS-regulated H3K27ac regions identified are reliable and consistent with the stimulus applied.
Next, we annotated the H3K27ac peaks identified to investigate whether they were more likely to lie in specific genomic regions. When looking at all peaks (found in Control and/or LPS condition), most were found within Intronic (42.4%), Intergenic (36.1%) and Promoter (13.2%) regions, consistent with H3K27ac being a mark of active promoters and enhancers . While LPS-increased peaks also showed the same trend with 44.6% of Intronic, 47.4% of Intergenic and 4.0% of Promoter peaks, there was significantly more intergenic (p < 0.0001) and less Promoter (p < 0.0001) peaks compared to non LPS-increased H3K27ac regions. In addition, there was also less exonic (p = 0.004) and 5’UTR (p = 0.008) peaks (Fig. 1d). This observation is consistent with other reports showing a crucial role for enhancers in cellular response to stimulus . This was further supported when looking at the distribution of enhancers and promoters along the H3K27ac regions ordered according to their variation in signal after LPS stimulation. Promoters were rarer among the 300 increased and decreased peaks compared to the unchanged H3K27ac peaks while enhancers showed the opposite trend (Additional file 4 B and Additional file 2 B), suggesting that promoters might be less relevant in regulating the response to LPS but necessary to regulate constitutive cell function which do not vary under stimulation.
In addition, we performed super-enhancers analysis as it was shown previously that stimulation can re-organize cellular super-enhancers . Using H3K27ac ChIP-seq signal, we identified 1057 and 1119 super-enhancers in Control and LPS condition respectively (Additional file 5 A). As previously shown, these super-enhancers were much bigger than other enhancers (Additional file 5 B). Gene Ontology analysis revealed relevant biological processes such as “immune system process” (Raw P-value = 2.7e-22) or “regulation of apoptotic process” (Raw P-value = 1.7e-18) as top terms for the LPS super-enhancers that were not present in the top Gene Ontology results for Control super-enhancers (Additional file 5 C). Moreover, we found 111 super-enhancers present in LPS condition but not in Control, while 49 super-enhancers were exclusively found in Control (Fig. 1e). One of the gained super-enhancers after LPS stimulation occurred upstream of the CSF2 locus (Additional file 5 D) which encodes a cytokine activating macrophages and dendritic cells . These results suggest that LPS stimulation not only opens chromatin but also induces formation of super-enhancers around certain immune genes.
One of the most striking changes in H3K27ac followed by gene up-regulation was found at the TNF locus described later on. However, other genes showing more modest LPS-increased H3K27ac and up-regulation could be observed. This was the case of the ICAM1 gene showing an active region close to its promoter at rest for which the H3K27ac signal increased after LPS stimulation (Additional file 6 D). This effect was accompanied by a gene expression increase of about 2 fold (Additional file 6 C).
We first performed motif analysis on the H3K27ac peaks identified in Control and LPS conditions to examine any differences between the two. Interestingly, the top motifs identified were common and enriched similarly in both conditions and consisted mainly of the AP-1 family of transcription factors (Jun, Fosl2, Fra 1, ATF3, BATF, AP-1 – Additional file 8) also activated downstream of TLR stimulation . In addition, we found the motif for RELA (also known as p65) subunit of NF-kB to show great variation between the enrichment in Control and LPS H3K27ac peaks (Additional file 8 – highlighted in yellow). The peaks identified under LPS stimulation were more likely to contain a motif for RELA compared to the control peaks (Fig. 3b). Moreover, the motif for p50/p52 – other NF-kB subunits – was also enriched in LPS but not in Control peaks (Additional file 8 – highlighted in green). Similarly, motifs analysis in the LPS-increased H3K27ac peaks revealed a great enrichment of RELA as well as p50/p52 motifs (Additional file 9). Furthermore, de novo motif analysis in the same set of peaks identified a consensus matching perfectly the NF-kB and RELA motifs (Fig. 3c). These observations suggest that NF-kB plays an important role in regulating gene expression following LPS detection, especially the RELA and p50/p52 subunits. This is consistent with current knowledge about the NF-kB canonical pathway involving RELA:p50 dimers that has been shown to be activated downstream of TLR4 following LPS binding .
We then used the RELA ChIP-seq data we generated previously  to investigate overlap between RELA binding and LPS-increased H3K27ac regions. Interestingly, 82.3% (515 out of 626) of the LPS-increased peaks overlapped with RELA binding sites in the same conditions (Fig. 3d). Moreover, RELA binding signal at the LPS-increased peaks was significantly higher than in other H3K27ac peaks (Fig. 3e). This was supported when plotting RELA signal along the H3K27ac peaks ordered according to their difference in signal between LPS and Control condition where the two data aligned (Additional file 4 C and quantification in Additional file 2 C). Furthermore, RELA signal at super-enhancers was significantly higher than at other enhancers (Fig. 3f) and it was also higher at LPS-only super-enhancers compared to Control-only super-enhancers (Additional file 7 C). Taken together these results support the role of RELA in coordinating the opening of the chromatin following stimulation [15, 16] as well as the formation of super-enhancers .
In this study, we identified LPS-induced H3K27ac sites in epithelial cells and reported 626 LPS-increased ones. These represent a small portion of the regulatory regions identified with this histone mark. They were highly enriched in enhancer regions, consistent with the important role of enhancers in cellular responses , and made up 1.7% of all activated enhancers. Several other studies have looked at enhancer activation following microbial or inflammatory stimuli, they reported various numbers of stimulus-induced activated enhancers but all consisted of a small fraction of all activated enhancers identified [7, 16, 21]. For instance, in the study by Ostuni et al., LPS induction was investigated in mouse macrophages and reported 6308 activated enhancers out of ~ 65,000 regulatory regions identified. They also reported 9004 enhancers that were repressed following LPS stimulation, contrasting with our study where we were not able to identify any LPS-decreased H3K27ac sites. Loss of activated enhancers following inflammatory stimuli has been documented and is usually associated with down-regulation of basal cell function [15, 20]. Differences in the number of stimulus-induced activated and repressed enhancers across studies and with our study may be due to the divergence of organisms and cell types used. Indeed, cell type specificity is known to influence enhancer selection  as well as gene expression response . Alternatively, variation in the stimulus as well as the duration of induction could lead to differences in the chromatin remodeling observed between studies, especially considering that response to infection is a time-dependent process .
Furthermore, we compared the LPS-induced H3K27ac and RELA binding following the same stimulation. We found them to be highly correlated and most LPS-increased H3K27ac peaks to overlap that of RELA. Moreover, inhibiting the NF-kB pathway abolished the deposition of the histone mark following LPS stimulation, suggesting an important role for this transcription factor in regulating the above mentioned epigenetic change. These results are in line with that of another study looking at IL-1 signaling, a pleiotropic cytokine inducing inflammation through the NF-kB pathway. This study showed that RELA regulates IL-1 inducible H3K27ac, particularly via upstream members of the signaling pathway, TAK1 and IKK2 . In addition, RELA is known to interact with chromatin modifying complexes such as p300  resulting in acetylation of histones surrounding the sites where RELA binds DNA . This could also explain the deposition of the H3K27ac surrounding RELA binding sites as seen in our results at the TNF locus and its impairment following NF-kB inhibition.
We also found RELA binding to be highly bound to Super-enhancers identified following LPS stimulation. This is consistent with previous reports showing that NF-kB is involved in chromatin rearrangement after TNFα induction , potentially by redistributing co-factors at gained super-enhancers .
Additionally, we found AP1 members motifs to be enriched in both Control and LPS H3K27ac peaks. The AP1 complex is activated downstream of TLR and is able to interact with NF-kB  but it also constitutively binds open chromatin, especially at intergenic and intronic regions  thus it is not surprising to find these motifs enriched in both conditions here.
Finally, we showed that H3K27ac deposition correlated with increased gene expression, consistent with the activating role of this mark . Although the correlation coefficient between H3K27ac changes and gene expression regulation is rather limited, it is similar to what others investigating the same mark have reported . Traditionally, regulation of inflammatory genes has been seen as a first wave of early genes that do not require chromatin remodeling expressed quickly after induction and a second wave of late genes showing slower kinetics due to changes in the chromatin that are necessary to make them accessible . However, in our study, TNF – which is considered an early gene in response to LPS stimulation  – did show H3K27ac changes close to its promoter which was not observed at resting condition suggesting that some degree of chromatin rearrangement was necessary to induce gene expression in this case. This is further confirmed by the suppression of its up-regulation following LPS when H3K27ac deposition is abolished following NF-kB inhibition. The rapidity of TNF expression following LPS stimulation could indicate that the promoter was already in a poised state where other histone marks were present .
We identified promoters and enhancers activated following sensing of LPS, a gram negative bacterial endotoxin, via TLR4. We found that deposition of H3K27ac marks correlated with increased gene expression as well as recruitment of RELA, one of the NF-kB members particularly important for the innate immune response. In addition to validate previous studies showing epigenetic changes following external signals, this data set may be valuable in the context of respiratory infectious diseases.
Cell culture and treatment
Detroit 562 cells were purchased from ATCC and cultured in RPMI medium (Gibco) supplemented with 10% fetal bovine serum performance (Gibco), 100 U/ml penicillin and 100μg/ml streptomycin (Gibco), and 1 mM sodium pyruvate (Gibco). LPS from E.coli B4:111 (Sigma) at 1 μg/mL was used to stimulate the cells. Inhibition of NF-kB was done treating the cells with BAY 11–7082 (ChemCruz) at 100uM (stock solution at 50 mM in DMSO) for one hour prior and at 90uM (dilution 9/10 in the medium containing the ligand) during the LPS treatment (Fig. 4 b and c). RELA activation assay, RT-qPCR and ChIP-qPCR procedures are described in Additional file 11.
Cells were treated with LPS or with fresh medium (Control - no treatment) for 80 min in Detroit 562 cells. ChIP-seq was performed according to the library-on-beads protocol from Wallerman et al.  using the NEBnext DNA library prep kit (New England Biolabs) as described previously  with the following modifications. 5 μg of H3K27ac antibody (Active motif, reference 39,138) was mixed with 50 uL of Dynabeads Protein G before immunoprecipitation. Sequencing was performed on an Illumina NextSeq sequencer (1x76bp). Biological duplicates consisting of two independent experiments were sequenced and analyzed.
Sequencing reads were mapped to the human genome hg19 with BWA, quality control and filtering were then performed using samtools. Multi-aligned or unmaped reads, PCR duplicates and reads with a MAPQ< 20 were filtered out giving around 50–110 millions passed reads per sample (Additional file 1). For each sample, peak calling was performed with Dfilter  against input DNA using relaxed threshold (−lpval = 1). The irreproducibility discovery rate (IDR) method  was used to identify reproducible peaks in both conditions separately. Peaks with an IDR < 0.05 between duplicates were selected in Control and LPS and consisted of the final peaks in each condition. Any peaks overlapping the blacklisted regions found in wgEncodeDacMapabilityConsensusExcludable.bed (http://hgdownload.cse.ucsc.edu/goldenpath/hg19/encodeDCC/wgEncodeMapability/) were further removed from the final list (Additional file 1).
Homer  was used to generate bedGraph files to visualize the signal in the UCSC genome browser with the command makeTagdirectory followed by makeUCSCfile (Fig. 4a and Additional file 10 C). In the same program, the command mergePeaks.pl allowed the comparison of different sets of peaks (Figs. 1a, e and 3d). The output merged peaks of the comparison between Control and LPS H3K27ac peaks consisted of the common as well as condition-specific peaks and were referred to as “all H3K27ac peaks”. The command annotatePeaks.pl in Homer was used to annotate the peaks (Fig. 1d, Additional file 4 B and Additional file 2 B).
Differential binding analysis was done on all H3K27ac peaks using each duplicate’s bam files from LPS and Control H3K27ac ChIP-seq as well as corresponding input DNA bam files as background. GC and read count normalization was first performed, then for every one of H3K27ac peaks (row), tag count for each condition was normalized by the median of the row, and the fold change over the median for each condition was generated and used in the heatmap representation. A threshold of 1.5 (signal of more than 1.5 time the median) was applied and clustering was then performed with K-mean clustering, dividing the data into 4 groups (Fig. 1b).
Throughout the manuscript, LPS-increased H3K27ac peaks refer to reproducible peaks showing an increase in signal in the LPS compared to the Control H3K27ac ChIP-seq in the differential binding analysis (Group 1 in Fig. 1b). Non LPS-increased H3K27ac peaks refer to all other H3K27ac peaks not present in this group of peaks.
Gene Ontology was carried out with GREAT, genes were associated to the peaks using the “single nearest gene” rule with a threshold of 1000Kb and all H3K27ac (peaks identified in LPS and/or Control ChIP-seq) were used as background (Fig. 1c and Additional file 3).
Motif analysis was performed with Homer using findMotifsGenome.pl command. Known motif enrichment in the LPS and Control H3K27ac peaks was done against the whole genome (Fig. 3b and Additional file 7) while known and de novo motif enrichment in the LPS-regulated H3K27ac peaks was compared to all H3K27ac peaks (Fig. 3c and Additional file 8).
Super enhancers analysis was performed as described previously . All H3K27ac peaks identified in LPS and/or Control condition were used as input and the H3K27ac ChIP-seq signal (merged bam file from duplicates) used to rank the enhancers (Additional file 5).
Integration of ChIP-seq with gene expression data
The Gene expression data used in this study consists of RNA-seq experiments performed in control and LPS (treatment for 100 min with 1μg/mL) condition in the same Detroit 562 cells  and available under the accession number GSE91019 in the NCBI’s Gene Expression Omnibus (GEO) depository.
Distance to the closest TSS was retrieved from the annotation analysis performed with the command annotatePeaks.pl in Homer. Only the peaks not located at a gene’s promoter (-2Kb to +2Kb around TSS) are considered when comparing LPS-increased and non LPS-increased H3K27ac peaks in Fig. 3a while the whole distribution can be found in Additional file 9 A.
The Rnachipintegrator program (https://github.com/fls-bioinformatics-core/RnaChipIntegrator) was used to assign H3K27ac peaks to their closest gene within 1000 Kb of the peak edge. Log2FC or FPKM values for the genes associated were then extracted from the RNA-seq data in order to investigate their expression (Fig. 3b and Additional file 4 D and Additional file 2 D). Log2FC was also used to interrogate the correlation between H3K27ac signal and gene expression. In this analysis, log2FC H3K27ac was calculated from the normalized read count in each condition (see Additional file 11 for details). Moreover, the following thresholds were applied to remove noise: Log2FC H3K27ac < 0.38 (FC = 1.3) and Log2FC Gene expression < 0.58 (FC = 1.5).
Integration of H3K27ac ChIP-seq with RELA binding
The RELA DNA binding data used in this study is available under the accession number GSE91018 in the NCBI’s GEO repository. It consists of RELA ChIP-seq experiments performed under Control or LPS (treatment for 80 min at 1μg/mL) condition in the same Detroit 562 cells .
RELA ChIP-seq was used to quantify RELA binding at the various H3K27ac peaks identified in this study. Normalized read count in this data-set was obtained with the command annotatePeaks.pl from Homer, using the option –d referring to the duplicates’ merged RELA ChIP-seq data set’s tag directory. Read counts were used to compare RELA binding at LPS-increased and non LPS-increased H3K27ac regions (Fig. 3e) and at super-enhancers and other enhancers (Fig. 3f) as well as to quantify the signals in the integrated analysis (see Additional file 11 for details).
Comparison of proportions in Figs. 1d and 3b was tested using the online 2 × 2 Chi-square test from Vassarstats (http://vassarstats.net/odds2x2.html) and the Pearson P-value was reported. Significance between distributions (Figs. 2a, b, 3e and f) was investigated using a non –paired Wilcoxon rank-sum test (wilcox.test in R). Finally, correlation between H3K27ac signal and gene expression was tested with a Pearson test (cor.test in R), correlation coefficient as well as P-value is reported (Fig. 2c).
We thank Kumar Vibhor for his help on the ChIP-seq analysis, particularly the differential binding analysis. The Next Generation Sequencing Platform and the Research Pipeline Development team at the Genome Institute of Singapore made the sequencing possible as well as data processing and mapping of the reads.
This work was funded by the Agency for Science and Technology, Singapore.
Availability of data and materials
The datasets generated and/or analysed during the current study are available in the NCBI’s GEO repository under the accession number: GSE104635, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE104635
SD and LB designed the study, LB performed the experiments and the analyses, MH provided support and comments of the results. LB wrote the manuscript with the help of SD and MH. All authors read and approved the final manuscript.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 13.Pappas C, Aguilar PV, Basler CF, Solórzano A, Zeng H, Perrone LA, Palese P, García-Sastre A, Katz JM, Tumpey TM. Single gene reassortants identify a critical role for PB1, HA, and NA in the high virulence of the 1918 pandemic influenza virus. Proc Natl Acad Sci. 2008;105(8):3064–9.CrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.