Abstract
Streptococcus pneumoniae (Sp), a natural colonizer of the human respiratory tract, is a diverse species with over 90 serotypes. Initial pneumococcal colonization of the human nasopharynx induces two distinct host outcomes; asymptomatic carriage or symptomatic invasive pneumococcal disease depending on the serotype and the host response. Epithelial cells are among the first to encounter both carriage and invasive serotype isolates of pneumococcus. However, the cellular processes responsible for the divergent host responses are largely unknown, as is the contribution of epithelial cells to this process. Here, we show a serotype 6B carriage isolate induces a unique inflammatory signature distinct from invasive serotype 4 (Tigr4). This inflammatory signature is characterized by activation of p65 (RelA) and requires a histone demethylase, KDM6B. At the molecular level, we show that interaction of serotype 6B with epithelial cells leads to chromatin remolding within the IL-11 promoter in a KDM6B dependent manner. We show KDM6B specifically demethylates histone H3 lysine 27 di-methyl, and this facilitates p65 access to three NF-κB sites, which are inaccessible when stimulated by IL-1β or Tigr4. Finally, we demonstrate through chemical inhibition of KDM6B, with GSK-J4 inhibitor, and through exogenous addition of IL-11 that the host response to carriage or invasive phenotypes can be interchanged. Therefore, we demonstrate that epithelial response to either carriage or invasive serotypes of S. pneumoniae is divergent and is mediated through chromatin remodeling by KDM6B.
Introduction
Streptococcus pneumoniae (Sp), a clonal species with more than 90 serotypes, naturally colonizes the upper respiratory tract of humans 1-6. Pneumococcal serotypes are found as either carriage isolates, which are asymptomatic and eventually cleared by the host, or as invasive isolates, which lead to symptomatic disease 5,7,8. Globally, invasive S. pneumoniae is a priority pathogen due to its ability to cause lethal pneumococcal infections resulting from pneumonia, sepsis, or meningitis 4,9. Importantly, colonization of the nasopharynx is a perquisite for both pneumococcal carriage and invasive disease 2,3,5,10,11.
At these initial colonization events, pneumococcus interacts with the host nasopharyngeal epithelial barrier and the innate immune system. Epithelial cells are among the first responders to pneumococcus and play a pivotal role in dictating pulmonary innate immune responses upon infection12. Recent insights using the experimental human pneumococcal carriage (EHPC) model have highlighted the essential role of NF-κB driven inflammatory responses for susceptibility, pathogenesis and transmission of pneumococcus 8,10,13-15. However, it is still largely unknown how these cellular processes are shaped at the molecular level and result in symptomatic or asymptomatic S. pneumoniae infections.
NF-κB is a master transcriptional regulator of both pro- and anti-inflammatory host responses 16-26. Briefly, NF-κB is comprised of multiple subunits that form hetero- or homodimers, of which the best characterized subunit is p65 (RelA)27,28. Activation of p65 occurs through posttranslational modifications (PTMs), such as phosphorylation of serine 536, in response to cellular sensing of inflammatory stimuli (i.e. LPS or interleukin 1 beta (IL-1β)) 28. Ultimately, activated p65 binds to a kappa-binding consensus sequence site within the nucleus to initiate transcription of NF-κB dependent genes29. However, cellular signaling alone is not enough, as a full NF-κB response also requires chromatin remodeling at the targeted inflammatory gene loci 16-26.
Chromatin, is a highly ordered structure of DNA wrapped around histone proteins. Chromatin dynamically shifts between open (euchromatin), and closed (heterochromatin) states, and these states influence gene accessibility and transcription 30-34. Switching between these two states is the result of chromatin remodeling enzymes/complexes reading, writing and erasing PTMs on histone tails. The enzymes regulating histone PTMs have been identified, and have been shown to play important roles in transcriptional responses during cellular signaling events, such as NF-κB responses 16,35. One of these enzymes is KDM6B (JMJD3), a histone demethylase, associated with NF-κB. KDM6B belongs to the Jumonji C-domain family (JMJD) of histone demethylases, of which KDM6B is the only member expressed universally outside of embryonic development 24,36. Primarily through peptide studies, KDM6B is thought to target the repressive histone marks, lysine 27 tri-methyl (H3K27me3) and di-methyl (H3K27me2) 24,37-39. To date, mounting evidence, mainly in macrophages, suggests KDM6B is essential for modulating inflammatory gene expression upon wound healing, LPS stimulation and immunological tolerance to anthrax toxin 20-23,40. However, to our knowledge, no role for KDM6B during bacterial infection has been studied to date.
Herein, we demonstrate a pneumococcal carriage isolate of serotype 6B specifically activates a unique inflammatory signature through p65. This inflammatory signature includes upregulation of KDM6B and IL-11 in human epithelial cells, and these are essential for epithelial cell integrity during challenge with 6B. We demonstrate upon challenge with serotype 6B the promoter of IL-11 is remodeled via KDM6B demethylation of H3K27me2, which allows p65 binding at three NF-κB sites upstream of the IL-11 transcription start site. We demonstrate the importance of this process in regulating epithelial cell integrity as inhibition of KDM6B leads to increased 6B induced epithelial damage; whilst exogenous addition of IL-11 partially rescues serotype 4 (Tigr4) induced cell damage. Thus, we show with chemical inhibition of KDM6B and IL-11, carriage and invasive epithelial-pneumococcal phenotypes can be interchanged.
Results
Serotype 6B actively induces a unique inflammatory profile
The differential cellular processes driving the host response during carriage or invasive pneumococcal disease is not entirely known. To address this, we completed an exploratory microarray of human A549 epithelial cells 2hrs post-challenge with either the laboratory invasive strain Tigr4 (serotype 4) or the pneumococcal carriage strain of serotype 6B (clinical isolate; ST90 CC156 lineage F; Fig 1A). In comparison with uninfected cells, 6B differentially influenced 388 transcripts (200 upregulated and 188 downregulated); whilst Tigr4 modulated the expression of 1,205, (143 upregulated and 1,062 downregulated) (Sup. Table1). Strikingly, a large proportion of the total genes differentially regulated by 6B were inflammatory genes containing NF-κB binding sites (12% by 6B vs. 3% by Tigr4; Fig 1B). To confirm this result, we selected a panel of 41 inflammatory genes, including genes from the microarray (IL-11, KDM6B (JMJD3), PTGS2, CXCL8 (IL8), FOS and JUNB), to test by RT-PCR. Indeed, upon epithelial cell colonization in vitro, 6B induced an inflammatory profile, with significantly increased expression of CSF2 (GM-CSF), CXCL1, CXCL2, CXCL3, IL11, KDM6B and TLR9 (pV≤0.05) in comparison to Tigr4 (Fig. 1C; Sup. Table2). Thus, 6B challenged epithelial cells have a transcriptional profile that is more inflammatory than that of Tigr4.
We further cross-compared the same panel against IL-1β (Sup. Table2), a known pro-inflammatory stimulus activating p65 (RelA). Using the relative expression data obtained from RT-PCR, we performed a principal component (PCA) on the expression values for 6B, Tigr4 and IL-1β (Fig. 1D). Comparative analysis of the biplot of the first two components showed three groups, which accounted for 64.3% of the total variance. This clearly demonstrated 6B was actively inducing a unique inflammatory signature distinct from both Tigr4 and IL-1β. To determine if RT-PCR results reflected protein expression, we preformed immunofluorescence staining for KDM6B, one of the genes differentially expressed, at 2hrs post-challenge (Fig. 1E; Rep. images Sup. Fig. 1A). A549 cells were challenged with either 6B, Tigr4, or paraformaldehyde killed 6B (PFA 6B). Two hours post-challenge the nuclear ratio of KDM6B to DAPI signal intensity was quantified. For 6B there was a significant increase in nuclear KDM6B (pV≤0.001) compared to Tigr4 and uninfected cells, which mirrored our expression analysis. Furthermore, we did not see a significant increase in nuclear KDM6B following challenge with paraformaldehyde-killed 6B, which suggests that this is an active process due to pneumococcal-epithelial interaction, as paraformaldehyde fixation not only inactivates pneumococcus, but is known to maintain bacterial morphology including pili, and extracellular polymeric substances, such as capsule 41.
To determine whether expression of KDM6B in response to 6B was specific, we tested two additional JMJD methyltransferases (KDM7A and KDM8), and a non-related methyltransferase (EHMT2). KDM6B was the only one to be significantly upregulated by 6B (Sup. Fig. 1B). Together these results show 6B induces a differential transcriptional response in epithelial cells characterized by upregulation of KDM6B
6B inflammatory profile requires p65 activation and catalytically active KDM6B
A hallmark of inflammatory gene induction is the activation of p65 via phosphorylation at serine 536 (S536)27. Thus, we tested if 6B activated p65. Whole cell lysates from 2hrs post-challenge of HeLa GFP-p65 stable cell line were immunoblotted for p65 phosphorylation at S536 (Fig. 2A). In comparison to uninfected cells, both IL-1β (positive control) and challenge with 6B induced p65 phosphorylation of S536 (pV≤0.001) whereas Tigr4 did not (Fig. 2B). To determine whether p65 activation by 6B had a role in its specific inflammatory signature, we used a chemical inhibitor of p65 activation, BAY 11-7082 42-44. A549 cells were pretreated with 10µM BAY 11-7082 3hrs prior to 2hr challenge with 6B, Tigr4 or IL-1β and gene expression in comparison to uninfected cells was determined by RT-PCR. BAY 11-7082 treatment did not affect viability, or gene expression alone in comparison to untreated cells (Sup. Fig. 2A & B; gray bars). We determined expression levels for two genes significantly upregulated by 6B in comparison to Tigr4, IL-11 and KDM6B, as well as a control gene, PTGS2, which is known to be p65 dependent. During 6B challenge no significant effect between untreated (no inhibitor) and DMSO (vehicle control) was observed upon the expression of PTGS2, IL-11 and KDM6B, and under the same conditions IL-11 and KDM6B expression remained roughly two-fold higher on average in comparison to Tigr4 and IL-1β (Fig. 2C; white & light gray bars). In contrast, BAY 11-7082 inhibition of p65 during 6B, Tigr4 or IL-1β challenge resulted in reduction in PTGS2 across all samples (Fig. 2C; gray bars). Furthermore, the expression of KDM6B and IL-11 were significantly repressed only during 6B challenge in the presence of inhibitor (pV≤0.001 and 0.05 respectfully) in comparison to DMSO treated cells (Fig. 2C; gray bars). These data show p65 activation is required for IL-11 and KDM6B expression upon 6B challenge of epithelial cells.
Previous studies demonstrated KDM6B interacts with p65 for inflammatory gene activation during keratinocyte wound healing, and chIP-seq studies found LPS stimulation of macrophages lead to KDM6B regulation of specific inflammatory genes 21,23. To determine whether KDM6B had an active role in 6B induced expression of IL-11 and KDM6B, we used GSK-J4, an inhibitor of the catalytic JMJ domain of KDM6B 45. As a control, we chose expression of PTGS2, as it is associated with KDM6B and not H3K27me3, thus inhibition of the catalytic activity of KDM6B should have no effect upon its expression 21. GSK-J4 (10µM; 24hrs prior) was used to pretreat A549 cells before challenge with 6B, Tigr4 or IL-1β. GSK-J4 alone had no significant effect on cell viability, or gene expression in comparison to untreated cells, nor did it affect the transcripts of PTGS2, IL-11 or KDM6B in Tigr4 and IL-1β challenged cells (Fig. 2C and Sup. Fig. 2A & B; black bars). Whereas, when the catalytic activity of KDM6B was inhibited more than a three-fold loss of expression for both IL-11 and KDM6B was observed during 6B challenge compared to DMSO control (Fig. 2C; black bars). GSK-J4 treatment had no effect upon PTGS2 expression during 6B challenge, demonstrating KDM6B catalytic activity was specifically required for IL-11 expression (Fig. 2C; black bars). With this, we clearly show catalytically active KDM6B is required for 6B induced expression of IL-11 and KDM6B.
6B induces chromatin remodeling of the IL-11 promoter for expression
Since, KDM6B is a histone modifying enzyme localizing to chromatin, we addressed whether 6B induced expression of IL-11 required chromatin remodeling within the IL-11 promoter. We mapped and designed ChIP-qPCR primers to predicted kappa-binding sites within the IL-11 promoter using AliBaba2 software, which curates eukaryotic transcription factor DNA binding motifs from the TRANSFAC® database46. Three kappa-binding sites upstream (−2,077bp, -774bp, and -406bp) of the IL-11 transcriptional start site (TSS), and one site downstream (+83bp) were predicted (Fig. 3A). Herein, we obtained chromatin from A549 cells 2hrs post-challenge with either 6B, Tigr4, or IL-1β, and compared the recovery of p65, and KDM6B at these kappa-binding sites within the IL-11 promoter using ChIP-qPCR with and without chemical inhibition of the catalytic activity of KDM6B.
During 6B challenge there was a significant (pV≤0.001) recovery of p65 at kappa-binding sites P6 (∼25%), P3 (∼20%) and P2 (∼10%) in contrast to uninfected conditions (Fig. 3B; 6B dark blue; uninfected white). Furthermore, there was ∼15% recovery of KDM6B across the same kappa-binding sites in cells challenged with 6B (Fig. 3C; 6B dark blue; uninfected white). In contrast, there was no recruitment of p65 or KDM6B to the kappa-binding sites in IL-1β or Tigr4 challenged cells (Sup. Fig. 3B & C). Recruitment of p65 and KDM6B to these kappa-binding sites was abolished in the presence of the GSK-J4 inhibitor (Fig. 3B &C; 6B light blue; uninfected gray). This clearly showed during 6B challenge the promoter of IL-11 was rearranged in a manner requiring the catalytic activity of KDM6B.
It has been suggested, mainly through peptide studies, that the enzymatic target of KDM6B is primarily H3K27me3 39,47. Thus, we hypothesized the chromatin rearrangement within the IL-11 promoter was a result of KDM6B demethylation of H3K27me3. We used ChIP-qPCR to determine the levels of H3K27me3 and H3, for nucleosome occupancy, across the three kappa-binding sites within the IL-11 promoter. Surprisingly, H3K27me3 was not decreased, in fact there was a slight, but significant (pV≤0.05), increase at the P6 kappa-binding site in comparison to unchallenged cells (Fig. 3D; 6B dark blue; uninfected white). There was no enrichment at any kappa-site during IL-1β or Tigr4 challenge (Sup. Fig. 3D). Furthermore, there were no differences in H3 nucleosome distribution at any of the kappa-binding sites between 6B and uninfected cells (Fig. 3E; 6B dark blue; uninfected white), this was also the case for cells challenged with IL-1β or Tigr4 (Sup. Fig. 3E). In the presence of GSK-J4 the increase of H3K27me3 at the P6 and P2 kappa-binding sites was lost in conjunction with a slight but significant increase in H3 nucleosome recovery at P6 (Fig. 3E & D; 6B light blue; uninfected gray). This data showed during 6B challenge KDM6B was not demethlyating H3K27me3, and this mark seemed to increase across the promoter.
Since our data showed an active role for KDM6B enzymatic activity independent of H3K27me3, we tested another proposed substrate of KDM6B, H3K27me238. Our ChIP results showed challenge with 6B induced loss of H3K27me2 at the P6 (pV≤0.01) and variable levels at the P3 and P2 sites within the IL-11 promoter in comparison to uninfected cells (Fig. 3F; 6B dark blue; uninfected white). Strikingly, when KDM6B enzymatic activity was blocked during 6B challenge demethylation of H3K27me2 was significantly inhibited across all kappa-binding sites (Fig. 3F; 6B light blue; uninfected gray).
Together these data show: 1) upon 6B challenge of epithelial cells the promoter of IL-11 is remodeled through the cooperative role of KDM6B and p65, and 2) KDM6B enzymatic activity is directed toward H3K27me2 and independent of H3K27me3 at these kappa-binding sites.
KDM6B and IL-11 contribute to epithelial cell integrity
We next wanted to address the role of KDM6B during 6B colonization of epithelial cells. Interestingly, previous works demonstrate KDM6B and p65 are both required for keratinocyte wound healing 23. Using these findings coupled with our own data showing 6B induced KDM6B and p65 recruitment to the IL-11 promoter in epithelial cells, we hypothesized that KDM6B and IL-11 were involved in maintaining epithelial cell integrity during pneumococcal colonization. In order to separate epithelial membrane permeability induced mainly by the pneumolysin toxin, a pore forming cholesterol dependent cytolysin (CDC), from cell death, we coupled the LDH cytotoxicity assay with Trypan blue exclusion. Combining these assays allowed us to separate cells with only damaged plasma membranes, which are permissible to Trypan, from dead cells that also release lactate dehydrogenase48,49. Herein, we used Trypan blue exclusion and LDH cytotoxicity assays in the presence of KDM6B inhibitor GSK-J4 (10µM) or DMSO (vehicle control) 24hrs prior to challenge with either Tigr4 or 6B (Fig. 4A). We observed no difference in either epithelial integrity or cell viability between uninfected cells with and without GSK-J4 inhibitor (Fig. 4B &E). Furthermore, epithelial integrity and viability was not compromised during 6B challenge in comparison to uninfected cells (Fig. 4B &E). In contrast, challenge with Tigr4 resulted in ∼60% epithelial membrane damage (Fig. 4B), and ∼45% cell death (Fig. 4E). Strikingly, inhibition of KDM6B catalytic activity affected epithelial integrity of 6B challenged cells, there was a significant ∼20% (pV<=0.001) increase in plasma membrane permeability (Fig. 4B), and ∼15% (pV<=0.01) increase in LDH release in comparison to the respective controls (Fig. 4E). These results suggest KDM6B plays a role in cell integrity only upon 6B challenge.
Since KDM6B is necessary for both cell integrity and regulates IL-11 expression, we further tested the role of IL-11 during pneumococcal colonization of epithelial cells. We next determined if exogenous recombinant human IL-11 was sufficient to rescue epithelial integrity loss seen during Tigr4 challenge. At the time of challenge, the inoculums of Tigr4 and 6B were supplemented with recombinant human IL-11 prior to addition to A549 cells. After 2hrs Trypan blue exclusion and LDH release assays were performed (Fig. 4C-E). IL-11 at the time of Tigr4 challenge partially rescued cell integrity by ∼20% (pV<=0.001), in comparison to untreated controls (Fig. 4D). There was no significant effect of exogenous IL-11 on uninfected or 6B challenged cells (Fig. 4D). Furthermore, LDH release showed exogenous IL-11 lowered Tigr4 cytotoxicity by ∼15% (pV<=0.01; Fig. 4E). Together this data shows IL-11 contributes to maintaining epithelial cell integrity during pneumococcal colonization.
Our data suggested the carriage isolate of serotype 6B induced KDM6B and IL-11 to maintain epithelial integrity. We hypothesized other pneumococcal carriage isolates could also induce IL-11 expression, while invasive ones would not. To test this, we compared additional carriage isolates of either serotype 19A or 19F, and two invasive serotype 1 isolates harboring either a hemolytic or non-hemolytic pneumolysin allele. Isolates of serotype 19A and 19F upregulated IL-11 expression in A549 epithelial cells in comparison to uninfected cells, whereas the serotype 1 isolates did not (Fig. 4F). To determine if IL-11 upregulation was specific to pneumococcal carriage isolates, or potentially upregulated by commensal organisms, we tested five additional oral microbiome commensals. Indeed, Streptococcus gordonii, Streptococcus sanguinis, Streptococcus oralis, Eikenella corrodens and Fusobacterium nucleatum also upregulated the expression of IL-11 in immortalized gingival keratinocytes (Sup. Fig. 4A). Together, our data show pneumococcal carriage isolates and commensal organisms induce IL-11 expression upon colonization, suggesting a common response is induced by colonizing bacteria. All together, we show pneumococcal carriage isolate 6B requires active KDM6B during in vitro colonization of human epithelial cells to mitigate epithelial cell damage, and IL-11 partially rescues epithelial cell integrity during Tigr4 challenge.
KDM6B is essential for local containment of carriage of 6B in vivo
Having defined an essential role for KDM6B during serotype 6B colonization of epithelial cells in vitro, we hypothesized local inhibition of KDM6B during 6B colonization of the murine nasal epithelium would promote 6B to escape from the nasopharynx due to loss of epithelium integrity. We challenged mice with 3 - 4×106 CFU of either 6B or Tigr4 mixed with either DMSO (vehicle control) or 5mM GSK-J4. The plated inoculums showed no significant effect of DMSO or GSK-J4 on bacterial viability (Sup. Fig. 5). Bacterial burden in the nasal lavage (NL), bronchoalveolar lavage fluid (BALF), lungs, and spleens of mice 24 and 48hrs post-inoculation were quantified by conventional colony forming unit (CFU) enumeration (Fig. 5A & B). Bacterial burdens from 6B and Tigr4 challenged DMSO animals showed on average one log more bacteria across all organs in comparison to Tigr4 by 24hrs (Fig. 5A; 6B light blue; Tigr4 gray). By 48hrs, infection with Tigr4 showed a progression of bacteria towards internal organs. The loosely attached bacteria in the NL and BALF decreased, while the burden in the lung and spleen increased in comparison to 24hrs. However, 6B CFU numbers either remained constant or decreased in all samples (Fig. 5B; 6B light blue). With this data we concluded 6B was primarily contained within the murine nasal cavity, whereas by 48hrs post-inoculation Tigr4 had escaped the nasopharynx and begun to disseminate from the lungs.
However, the addition of GSK-J4 changed the bacterial distribution of 6B. Indeed, GSK-J4 treated animals challenged with 6B showed increased burden across all samples in comparison to the 6B DMSO control group at 24hrs (Fig. 5A). Additionally, the recovered bacteria from the NL and BALF in the GSK-J4 6B challenged group was not significantly different to the Tigr4 DMSO group (Fig. 5A). However, after 24hrs there was no significant difference in the recovery of Tigr4 from the NL, BALF, lungs or spleens between DMSO or GSK-J4 treated animals (Fig. 5A; Tigr4 DMSO white, Tigr4 GSK-J4 gray). By 48hrs post-challenge 6B GSK-J4 animals maintained a significantly (pV<=0.05) high bacterial burden in the BALF compared to 6B DMSO treated animals (Fig. 5B; 6B DMSO light blue, 6B GSK-J4 dark blue). Importantly in animals treated with GSK-J4 6B was recovered from the spleen, an organ which bacteria were mostly undetected in DMSO control animals. GSK-J4 treated animals in the Tigr4 group also showed an increase in bacterial burden at 48hrs post-challenge in the NL, BALF, lung and spleen. Altogether, these data show KDM6B activity is specifically required for containment of 6B during colonization of the murine nasal cavity, and is potentially a negative regulator of Tigr4 dissemination.
Discussion
Colonization of the nasopharynx is an essential process that precedes asymptomatic pneumococcal carriage or symptomatic pneumococcal disease, with the bacteria first encountering the epithelial barrier 3,5. The molecular and transcriptional processes that define carriage at this stage are largely unknown. Towards this end, we completed a human microarray of A549 epithelial cells challenged with either 6B or Tigr4 pneumococcal strains. We show the pneumococcal carriage isolate 6B differentially regulated 388 genes, with a primary enrichment for NF-κB associated genes in comparison to Tigr4. A further study of NF-κB signaling demonstrated 6B activated p65, in contrast to Tigr4. Direct comparison of NF-κB regulated genes shows that 6B induced a unique inflammatory signature that included KDM6B and IL-11 expression, in contrast to Tigr4. We demonstrate molecularly that carriage pneumococcus, through the activity of KDM6B, induces remodeling of the IL-11 promoter to reveal three NF-κB sites, which are not accessible during IL-1β or Tigr4 stimulation. Together, this is the first demonstration that pneumococcal carriage remodels chromatin within epithelial cells to support a unique inflammatory signature.
Our findings, in conjunction with recent works by Weight et al., support the idea that pneumococcal carriage, in contrast to invasive pneumococcus, is actively inducing a host response to promote confinement to the nasopharynx 15. Our results strongly suggest KDM6B and its regulation of IL-11 transcription are key components modulating the host-pneumococcal response during colonization by carriage and invasive S. pneumonie strains. In both in vivo and in vitro experiments with chemical inhibition of KDM6B, we were able to interchange carriage and invasive phenotypes through a host driven mechanism. Since KDM6B differentially regulates multiple inflammatory genes we cannot rule out the possibility there are other genes with concurrent or synergistic functions with IL-11. However, our IL-11 rescue experiments with Tigr4 suggest a role for IL-11 in locally maintaining a permissive/tolerogenic epithelial-pneumococcal host response. Interestingly, IL-11 is known to influence mucus production, wound healing of gastric ulcers and resistance of endothelial cells to immune mediated injury 50-53. Promoting the confinement of a carriage pneumococcal strains within the nasopharynx, is also reflected in our microarray data, as 52 genes associated with wound healing gene ontology were upregulated by 6B in comparison to Tigr4 (Sup. Table 1). Additionally, 39 of the 52 upregulated wound healing genes were also associated with KDM6B and/or H3K27me3 determined from the ChIP-seq studies of macrophages (Sup. Table 1). Therefore, our findings suggest the initial pneumococcal-epithelial cell interaction plays an important role in driving a host response leading to divergent asymptomatic or symptomatic phenotypes during pneumococcal disease. Within this principal, we propose, in contrast to invasive serotypes, that early colonization of nasal epithelium by carriage serotypes actively induces “tolerogenic inflammation” through upregulation of wound healing cascades as a means to counter balance an early deleterious pro-inflammatory host response (i.e. neutrophil influx), thus preserving a prolonged niche within the host. Interestingly, we find that a signature gene, IL-11, was also induced by other pneumococcal carriage isolates and by several commensal organisms, suggesting this is a common response to colonizing bacteria.
The lysine demethylase KDM6B has mainly been characterized in cellular development, however a few studies suggest that this particular histone demethylase also fine-tunes inflammatory responses and wound healing downstream of p65 largely through unknown mechanisms 21-24,54-56. We are the first to report both a biological and molecular role for KDM6B and H3K27me3/2 in regulation of a specific gene locus, IL-11, during bacterial colonization. Surprisingly, although KDM6B was shown to primarily target H3K27me3 and to a lesser extent H3K27me238,39,47, our results suggest KDM6B is selectively demethlyating H3K27me2 and not H3K27me3 at the IL-11 promoter. These results are consistent with previous observations of Da Santa et al., who reported gene regulation by KDM6B independently of H3K27me3 21. With this observation, we hypothesize that KDM6B is differentially regulating inflammatory gene expression through selective demethylation of H3K27me2 through either an unknown regulatory element or posttranslational modifications to KDM6B. Future ChIP-seq and proteomic studies with biological stimuli, such as pneumococcus, will yield substantial insight into possible KDM6B complexes, and the dynamics of H3K27me3/2 in epigenetic control of inflammatory signaling cascades.
Through our study of p65 activation by pneumococcus, we find that serotype 6B activated p65 to similar levels as IL-1β, however, the ensuing transcriptomic responses are very different. Combining these observations with active remodeling of the IL-11 promoter strongly suggests that under 6B stimulation there are additional p65 interacting partners or posttranslational modifications (PTMs), in conjunction with phosphorylation of serine 536. Such data would support a novel biological role for “NF-κB barcode hypothesis”, where a signature barcode of PTMs on NF-κB mediates a specific gene expression pattern 57,58. While we have established a link between p65, KDM6B and IL-11 expression, identification of the p65 PTMs and interacting partners necessary for the inflammatory signature of 6B will advance our understanding of not only carriage pneumococcal host responses, but also p65 regulation during tolerogenic inflammatory responses.
A meta-analysis conducted by Brouwer et al., highlighted single nucleotide polymorphisms (SNPs) associated with NFKBIA, NFKBIE, and TIRAP correlated with protection, whereas SNPs within NEMO (IKBKG) or IRAK4 associated predominantly with increased susceptibility to disease 59. Analysis of KDM6B and H3K27me3 ChIP-seq data from LPS stimulated macrophages, shows these protective genes, NFKBIA, NFKBIE, and TIRAP, are also associated with KDM6B and/or H3K27me3, whilst NEMO and IRAK4 are not 21. Since NFKBIA, and NFKBIE are known to inhibit NF-κB through sequestration within the cytoplasm 16,60,61, one could hypothesize KDM6B is a chromatin level negative regulator that balances inflammatory signaling in conjunction with p65 across a unique “p65-KDM6B” axis. In this context, KDM6B serves as the molecular “regulator or brake” responsible for modulating the host response based upon the severity, or degree of inflammatory signal input. This role is evidenced by our in vivo studies showing chemical inhibition of KDM6B in vivo results in hypervirulence of Tigr4 and the escape of a carriage serotype 6B isolate from the murine nasal cavity.
Overall, our data demonstrates the first biological role of KDM6B in bacterial colonization. We further reveal catalytically active KDM6B is required for host tolerance to pneumococcal carriage isolate 6B. We further show exogenous IL-11 is partially sufficient to rescue Tigr4 induced cell damage in vitro. While we have only begun to scratch the surface of the molecular pathways involved in this process, it is clear characterizing pneumococcal carriage can not only identify new means to combat pneumococcal disease, but reveal new mechanisms involved in commensal organism meditated inflammatory processes.
Materials and Methods
Bacterial strains, growth conditions and CFU enumeration
Clinical isolates of serotypes 6B (ST90; CNRP# 43494), 19A (ST276; CNRP# 45426) and 1 (non-hemolytic; ST306; CNRP# 43810) were obtained from the Centre National de Référence des Pneumocoques (Emmanuelle Varon; Paris, France). Serotype 19F (BHN100; ST162 Birgitta Henriques Normark, Karolinska Institutet 62), serotype 4 Tigr4 (Thomas Kohler, Universität Greifswald), and serotype 1 (ST304 hemolytic; M. Mustapha Si-Tahar, Université de Tours). Experimental starters were prepared from frozen master stocks struck on 5% Columbia blood agar plates (Biomerieux Ref# 43041) and grown overnight at 37°C with 5% CO2 prior to outgrowth in Todd-Hewitt (BD) broth supplemented with 50mM HEPES (Sigma) (TH+H) at 37°C with 5% CO2 in closed falcon tubes. Midlog bacteria were pelleted, and diluted to 0.6OD600 /mL in TH+H media supplemented with Luria-Bertani (BD) and 15% glycerol final concentration. Aliquots were made and frozen at -80°C for experiments. All experiments performed with frozen experimental starters of S. pneumoniae less than 14 days old. For experiments, starters were grown to midlog phase in TH+H broth at 37°C with 5% CO2 in closed falcon tubes, pelleted at 1,500xg for 10mins at room temperature (RT), washed in DPBS, and concentrated in 1mL DPBS prior to dilution at desired CFU/mL using 0.6OD600 /mL conversion factors in either cell culture media or DPBS for animal studies (conversion factors Sup. Table 3). For paraformaldehyde (PFA) killed bacteria the concentrated bacteria prior to dilution was incubated 4% PFA for 30mins at RT, washed in DPBS, and diluted to desired CFU/mL using 0.6OD600 /mL conversion factors. Bacteria CFU enumeration was determined by 96well dilution plating.
Cell culture and In vitro challenge
A549 human epithelial cells (ATCC ref# CCL-185) were maintained in F12K media (Gibco) supplemented with 1x GlutaMax (Gibco) and 10% heat inactivated fetal calf serum (FCS) at 37°C with 5% CO2. Stable HeLa GFP-p65 were generated using the sleeping beauty system, and maintained in DMEM supplemented with 1x GlutaMax (Gibco) 10% heat inactivated FCS 63. A549 or HeLa GFP-p65 cells used until passage 15. For in vitro challenge studies, A549 or HeLa GFP-p65 cells were plated in tissue culture treated plates at 2×105 cells (6well; for 72hrs), 5×104 cells (24well; for 48hrs), or 1×104 cells (96well; for 48hrs). Bacterial inoculums diluted in cell culture media was added to cells, and bacterial-epithelial cell contact synchronized by centrifugation at 200xg for 10mins at RT, then moved to 37°C with 5% CO2 for 2hrs. For inhibitor studies, cell culture media was aspirated, and replaced with filter sterilized culture media containing inhibitor volume matched DMSO (Sigma), GSK-J4 (Sigma ref# SML0701), or BAY 11-7082 (Sigma ref# B5556) at 10µM final concentration for 24hrs or 3hrs respectively prior to bacterial addition. Human IL-11 (Miltenyi Biotec ref# 130-094-623) and human IL-1β (Enzo Life Sciences ref# ALX-522-056) were used at 100 ng/mL and 10 ng/mL final concentration respectively in cell culture media.
Immunofluorescence and Trypan blue bright field microscopy
To quantify nuclear KDM6B, A549 cells were seeded on acid washed and UV treated coverslips in 24well plates as described above, 2hrs post-challenge media was aspirated, cells washed in DPBS, and fixed with 2.5% PFA for 10mins at RT. Coverslips were blocked and permeabilized overnight in 5% BSA 0.5% Tween20. Coverslips were incubated for 1hr at RT with KDM6B (1:500; abcam ref# ab38113) diluted in 5% BSA 0.5% Tween20, washed in 0.5% Tween20, and incubated for 1hr with Alexa Fluor 488 secondary. After secondary, coverslips were washed in 0.5% Tween20 and mounted using Prolong Gold with DAPI (Invitrogen). Confocal microscopy images were acquired on a Ziess axio observer sinning disk confocal. Nuclear KDM6B intensity per cell was quantified within an ROI generated from the DAPI signal in Fiji 64. For Trypan exclusion microscopy, A549 cells were seeded in 96well plates as described above. 2hrs post-challenge culture media was aspirated, cells washed in DPBS, and Trypan blue (Thermo) added for 10mins at RT. Trypan blue was removed, and cells fixed with 2.5% PFA for 10mins at RT. PFA was removed and fixed cells washed in DPBS prior to imaging on a EVOS FL (Thermo). Trypan positive cells were scored manually as % of total cells in an imaged field.
A549 epithelial microarray
A549 cells were infected as described above, and total RNA harvested using RNeasy kit (Qiagen). RNA quality was confirmed using a Bioanalyzer (Agilent). Affymetrix GeneChip human transcriptome array 2.0 was processed as per manufacturer’s instructions. Data was analyzed using TAC 4.0 (Applied Biosystems).
LDH assay
LDH assays were performed on cell culture supernatants as per manufacturer’s instructions (Pierce LDH cytotoxicity kit (Thermo ref# 88953). LDH absorbance was read using Cytation 5 (BioTek) at manufacturer’s recommended excitation and emissions.
ChIP and ChIP-qPCR
Detailed ChIP buffer components are in supplemental methods. In brief, 8×106 A549 cells were cross-linked in tissue culture plates with 1% formaldehyde for 10mins at RT, then quenched with 130mM glycine for 5mins at RT. Cells were washed in DPBS, gently scraped, and transferred to an eppendorf. Harvested cells were pelleted at 200xg, supernatant aspirated and frozen - 20°C. To obtain chromatin, cell pellets were thawed on ice and lysed for 30mins on ice in nuclear isolation buffer supplemented with 0.2% Triton X-100. Nuclei pelleted, supernatant aspirated and suspended in chromatin shearing buffer for sonication with a Bioruptor (Diagenode) to 200-900bp size. Sheared chromatin was cleared by centrifugation, then sampled for size using 2% agarose gel electrophoresis and quantification using Pico488 (Lumiprobe ref# 42010). ChIP grade antibodies to p65 (L8F6) (CST ref #6956), KDM6B (abcam ref# ab38113), H3K27me3 (abcam ref# ab6002), H3 (abcam ref# ab195277), or H3K27me2 (diagenode ref# C15410046-10) were used at manufacturer’s recommended concentrations and bound to DiaMag beads (diagenode ref # C03010021-150) overnight with gentle rotation. Quantified chromatin was diluted to 10µg per immunoprecipitation condition was added to antibody bound DiaMag beads overnight with gentle rotation. Beads were washed with buffers 1-6 (supplemental methods), decrosslinked by boiling for 10mins with 10% Chelex, treated with RNase and proteinase K, then purified using phenol-chloroform extraction followed by isopropanol precipitation. Recovered DNA was suspend in molecular grade water, and 1µL used for Sybr Green reactions as per manufacturer’s instructions on a BioRad CFX384 (BioRad). % recovery was calculated as 2 raised to the adjusted input Ct minus IP Ct multiplied by 100. For histone marks, H3K27me3/2, the % recovery was normalized to the % recovery of H3. qPCR primers listed in Sup. Table 3.
RNA isolation and RT-PCR
Total RNA isolated using TRIzol (Life technologies ref#15596-026) extraction method as per manufacturer’s recommendations. Recovered RNA was suspended in molecular grade water, nano dropped and 5µg converted to cDNA using Super Script IV as per manufacturer’s instructions. cDNA was diluted to 20ng/µL in molecular grade water and 1µL used for Sybr Green reactions as per manufacturer’s instructions on a BioRad CFX384 (BioRad). RT-PCR primers listed in Sup. Table 3. Relative expression was calculated by ΔΔCt method to GapDH 65.
Immunoblots and quantification
Cell culture media was removed, washed in DPBS and whole cell lysates harvested with Laemmli buffer 66. Lysates were boiled 10min, and frozen at -20°C. Whole cell lysates were ran on 8% polyacrylamide SDS PAGE gels, transferred to PVDF membrane (BioRad), blocked in 5% BSA, then probed for p65 (CST ref #6956), p65 phosphorylation at serine 536 (CST ref# 3033), or actin AC-15 monoclonal (sigma ref# A5441) as per manufacturer’s recommendations. Appropriate secondary-HRP conjugated antibodies were used with clarity ECL (BioRad) developing reagents. Membranes were developed on ChemiDoc Touch (BioRad)
In vivo animal studies
All protocols for animal experiments were reviewed and approved by the CETEA (Comité d’Ethique pour l’Expérimentation Animale - Ethics Committee for Animal Experimentation) of the Institut Pasteur under approval number Dap170005 and were performed in accordance with national laws and institutional guidelines for animal care and use. Wildtype C57BL/6 female 8-9 week old mice were purchased from Janvier Labs (France). Animals were anesthetized with a ketamine and xylazine cocktail prior to intranasal challenge with 20µL of 3 - 4×106 CFU of Tigr4 or 6B. Bacterial inoculums were made as described above with minor modification. In brief, after 0.6OD600 bacterial were concentrated and diluted in either filter sterilized DMSO/DPBS or 5mM GSK-J4/DPBS. 24 and 48hrs post-inoculation animals were euthanized by CO2 affixation. The nasal lavage was obtained by blocking the oropharynx, to avoid leakage into the oral cavity and lower airway, and nares flushed with 500µL DPBS. Bronchoalveolar lavage fluid (BALF), lungs, and spleens were collected and placed in 1mL DPBS supplemented with 2x protease inhibitor cocktail (Sigma ref # P1860). CFU were enumerated as described above on 5µg/mL Gentamicin Columbia Blood agar selection plates.
Statistical analysis
All experiments, unless otherwise noted, were repeated 2-4 times with the statistical test in figure legends. P values were calculated using GraphPad Prism software. For RT-PCR all statistics were calculated on either the ΔΔCt or ΔCt depending on the desired comparison. PCA plots using the “prcomp” function of the base stats package in R on scaled and mean centered log2 transformed data. Microscopy data was collected from analysis of 20-50 cells for nuclear staining, or 200-300 cells for brightfield per biological replicate per group. Animal studies used the minimum number of animals required to reach power based on post-hoc CFU analysis calculated using G*Power software.
Supplemental Methods
AlamarBlue cytotoxicity
A549 cell viability was determined using AlamarBlue (Thermo ref# DAL1025) as previously described 67. AlamarBlue absorbance was read using Cytation 5 (BioTek) at manufacturer’s recommended excitation and emissions.
Oral commensal in vitro infection and RT-PCR
ChIP buffer solutions as follows
Author Contributions
Conceived and designed all experiments: MGC and MAH. Preformed experiments: MGC, EP (KDM6B microscopy and HeLa GFP-p65 western blot), and OR (animal studies). Analyzed data: MGC and EP. DPM performed all oral commensal infections and RT-PCR. LB generated the stable HeLa GFP-p65 cell line under the supervision of JE. MGC wrote the original manuscript draft. MGC and MAH edited and reviewed the manuscript. MAH supervised the research. All authors approved the final manuscript.
Conflict of interest statement
The authors declare no conflict of interest.
Figure Legends and Tables
Supplemental Figure 1: KDM6B microscopy and demethylase RT-PCR. A) Representative images of nuclear KDM6B (green) merged with nucleus stained with DAPI (false colored red for visual display). Scale = 10µm. B) Total RNA 2hrs post-challenge with 6B or Tigr4 from A549 cells. Demethylase panel RT-PCR shown as ΔCt (n=3; 3 technicals per biological replicate). Bar graph ± Std. All data analyzed by One way-ANOVA with Tukey’s multiple comparison post-hoc test, **= pV≤0.01, ns=not significant.
Supplemental Figure 2: Tigr4 and IL-1β inhibitor RT-PCR with viability data. A549 cells untreated or treated with 10µM BAY 11-7082, 10µM GSK-J4, or DMSO vehicle control (n=4). A) Cell viability determined by AlamarBlue. Expressed as % of untreated cells. No significant difference observed. B) Transcript levels for IL-11, KDM6B and PTGS2 determined by RT-PCR. Displayed as ΔCt for comparison to untreated (n=4). Bar graph ± Std. All data analyzed by One way-ANOVA with Tukey’s multiple comparison post-hoc test, no significant difference observed.
Supplemental Figure 3: ChIP-PCR IL-11 locus for Tigr4 and IL-1β. Chromatin obtained from untreated and 10µM GSK-J4 treated A549 cells 2hrs post-challenge with Tigr4 or IL-1β. 10 µg chromatin input used for ChIP of p65, KDM6B, H3K27me3 and histone H3 (H3), followed by ChIP-qPCR at primer locations (P6, P3 & P2) spanning the NF-κB sites upstream of the transcriptional start site (TSS). A) Schematic of IL-11 promoter with ChIP-qPCR primer locations (P6, P3 & P2) and the NF-κB sites. B - E) % recovery of input for p65, KDM6B, H3K27me3 normalized to H3, or H3 bound at P6, P3 & P2 in untreated and GSK-J4 treated samples (n=3 untreated; n=2 GSK-J4 treated. Tukey box and whisker plot with dots representing outliers. No significant difference observed in comparison to 6B or uninfected cells from Fig. 3.
Supplemental Figure 4: IL-11 RT-PCR with oral commensals. A) Total RNA harvested from immortalized gingival keratinocytes challenged with S. gordonii, S. sanguinis, S. oralis, E. corrodens or F. nucleatum. Transcript levels for IL-11 determined by RT-PCR and represented as relative expression to uninfected (n=1).
Supplemental Figure 5: Tigr4 and 6B bacterial inoculates from animal challenges. Tigr4 and 6B inoculums for animal intranasal challenge model in DMSO (vehicle control) or 5mM GSK-J4 (n=2). Conventional CFU enumeration shows no significant difference. Analyzed by One way-ANOVA with Tukey’s multiple comparison post-hoc test, ns=not significant.
Acknowledgments
We would like to thank Emmanuelle Varon, Birgitta Henriques Normark, M. Mustapha Si-Tahar, and Thomas Kohler for their generous gifts of S. pneumonie strains. We are thankfully Gregory Dore (Institut Pasteur, Nuclear Organization and Oncogenesis Group – Oliver Bischof) for processing the microarray Affymetrix GeneChips. Biostatistics and R discussions with Jose Nabuco (Institute Pasteur) and Jeremy Camp (University of Veterinary Medicine Vienna) were greatly appreciated. Michael G. Connor is support by the Pasteur Foundation Fellowship. G5 Chromatin and infection is support by the Institut Pasteur, and the Agence National de la Recherche (ANREpiBActIn).