The proinflammatory cytokine TNF-α induces DNA demethylation–dependent and –independent activation of interleukin-32 expression

Interleukin-32 (IL-32) is a cytokine involved in proinflammatory immune responses to bacterial and viral infections. However, the role of epigenetic events in the regulation of IL-32 gene expression is understudied. Here, we show that IL-32 is repressed by DNA methylation in human embryonic kidney 293 (HEK293) cells. Using ChIP-Seq, locus-specific methylation analysis, CRISPR/Cas9-mediated genome editing, and RT-qPCR and immunoblotting assays, we found that short-term treatment (a few hours) with the proinflammatory cytokine tumor necrosis factor-α (TNF-α) activates IL-32 in a DNA demethylation–independent manner. In contrast, prolonged TNF-α treatment (several days) induced DNA demethylation at the promoter and a CpG island in the IL-32 gene in a TET family enzyme– and NF-κB–dependent manner. Notably, the hypomethylation status of transcriptional regulatory elements in IL-32 was maintained for a long time period (several weeks), causing elevated IL-32 expression even in the absence of TNF-α. Considering that IL-32 can, in turn, induce TNF-α expression, we speculate that such feed-forward events may contribute to the transition from an acute inflammatory response to chronic inflammation.


ABSTRACT
Interleukin-32 (IL-32) is a cytokine involved in proinflammatory immune responses to bacterial and viral infections. However, the role of epigenetic events in the regulation of IL-32 gene expression is understudied. Here, we show that IL-32 is repressed by DNA methylation in human embryonic kidney 293 (HEK293) cells. Using ChIP-Seq, locus-specific methylation analysis, CRISPR/Cas9-mediated genome editing, and RT-qPCR and immunoblotting assays, we found that short-term treatment (a few hours) with the proinflammatory cytokine tumor necrosis factor-α (TNF-α) activates IL-32 in a DNA demethylation-independent manner. In contrast, prolonged TNF-α treatment (several days) induced DNA demethylation at the promoter and a CpG island in the IL-32 gene in a TET family enzyme-and NF-κB-dependent manner.
Notably, the hypomethylation status of transcriptional regulatory elements in IL-32 was maintained for a long time period (several weeks), causing elevated IL-32 expression even in the absence of TNF-α. Considering that IL-32 can, in turn, induce TNF-α expression, we speculate that such feed-forward events may contribute to the transition from an acute inflammatory response to chronic inflammation.
Consistent with a role of IL-32 in inflammatory response, IL-32 expression is induced by TNF-α in various human cell types, including synovial fibroblasts, intestinal epithelial cell lines and pancreatic cancer cell lines (7)(8)(9). Reciprocally, IL-32 can also induce the expression of TNF-α and other cytokines in human THP-1 monocytic cells (1). Interestingly, although mice do not contain the IL-32 gene, ectopic treatment with human IL-32 can induce TNF-α expression in mouse Raw macrophage cells (1). Moreover, the injection of human IL-32 protein into the knee joints of wild-type mice, but not into the knee joints of Tnf gene knockout mice, provokes severe inflammation, suggesting that IL-32 exerts direct effects on joint inflammation in a TNF-α-dependent manner (2). Functionally, IL-32 promotes the differentiation of monocytes towards macrophage-like cells that display phagocytic activity, further supporting a role of IL-32 in immune response (10).
IL-32 plays important roles in inflammatory autoimmune diseases (11,12). IL-32 is highly expressed in rheumatoid arthritis synovial tissue biopsies (2), inflamed mucosa of inflammatory bowel disease (9) and chronic pancreatitis duct cells (8). These reports suggest that IL-32 is likely a cytokine involved in chronic inflammation, and it may serve as a potential therapeutic target.
As a proinflammatory cytokine, the expression of IL-32 is induced during bacterial and viral infections, and its expression improves host immunity in controlling these infections (13). For example, in patients with active Mycobacterium tuberculosis infections, IL-32 expression is induced, and it protects human macrophages and peripheral blood mononuclear cells against M. tuberculosis (14)(15)(16). Likewise, the expression of IL-32 is induced during HIV infection and influenza virus infection, as it contributes to the antiviral response (4,17,18).
DNA methylation is an important gene silencing mechanism that functions by recruiting corepressor proteins to impede the binding of DNA methylation-sensitive transcription factors (19,20). DNA demethylation can be achieved by enzyme-mediated active demethylation or by passive DNA methylation caused by interfering maintenance DNA methylation (21). TET family methylcytosine dioxygenases catalyze active DNA demethylation through the sequential oxidation of 5mC to 5hmC, 5fC and 5caC (22)(23)(24), followed by TDG-mediated base excision repair (24).
Gene expression is often regulated by sequence-specific transcription factors and epigenetic regulators. Given that IL-32 expression is regulated during inflammation, understanding whether epigenetic events occur during the induction of IL-32 expression is interesting. Here, we report that IL-32 is silenced by DNA methylation and that TNF-α induces DNA demethylation-dependent and -independent mechanisms to control IL-32 activation. We also discuss the potential significance of these mechanisms.

IL-32 is silenced by DNA methylation in HEK293 cells
In our previous work, we performed RNA-seq experiments using HEK293 cells treated with the DNA-demethylating agent 5-aza-2'-deoxycytidine (5-aza-dC) and identified genes silenced by DNA methylation (25,26). IL-32 was one of the genes strongly activated upon 5-aza-dC treatment (Figs. 1A and 1B), suggesting that IL-32 is a gene silenced by DNA methylation in HEK293 cells. Indeed, bisulfite sequencing data revealed that both the promoter and CpG island (CGI) predicted by Sequence Manipulation Suite (27) of the IL-32 gene (Fig. 1C) are highly methylated (Fig. 1D).

Short-term TNF-α treatment induces IL-32 expression in a DNA demethylation-independent manner
Given that the IL-32 gene is induced by TNF-α treatment (7)(8)(9) and repressed by DNA methylation in HEK293 cells, we wondered whether TNF-α treatment is sufficient to overcome DNA methylation-mediated silencing. Thus, we treated HEK293 cells with 50 ng/ml TNF-α and analyzed IL-32 expression at various time points. IL-32 expression began to be induced as early as 1 h post-TNF-α treatment and was potently activated after 3 h of TNF-α treatment ( Fig. 2A).
Of course, we next asked whether IL-32 activation was accompanied by DNA demethylation. Interestingly, despite the apparent transcriptional activation, no substantial DNA demethylation at the promoter or CGI of the IL-32 gene was observed after 1 h of TNF-α treatment (Figs. 2A, 2B). These results indicate that TNF-α treatment could activate IL-32 gene expression in a DNA demethylation-independent manner.
We then examined THP-1 cells, a human monocyte-like cell line (28), and HAP1 cells, a human leukemia cell line (29). And we also observed DNA demethylation-independent activation of IL-32 expression upon short-term TNF-α treatment in these cells (Supplemental Fig. S1).

Long-term TNF-α treatment induces significant DNA demethylation of the IL-32 transcriptional regulatory region
However, we noticed a slight decrease in DNA methylation at the IL-32 promoter after 3 h of TNF-α treatment (Fig. 2B). This finding prompted us to perform longer TNF-α treatments with measurements of IL-32 expression and DNA methylation at various time points. As we anticipated, a long-term TNF-α treatment (12 d) resulted in clear DNA demethylation of the IL-32 transcriptional regulatory regions; furthermore, the accumulation of DNA demethylation was accompanied by IL-32 induction (Figs. 3A-C).

Hypomethylation triggered by long-term TNF-α treatment leads to elevated IL-32 expression after the removal of TNF-α
DNA methylation is a relatively stable epigenetic mark; therefore, we asked whether the methylation status of the IL-32 transcriptional regulatory regions could be stably maintained after TNF-α treatment. We treated HEK293 cells with TNF-α for 12 h or 12 d and then cultured the cells in a TNF-α-free medium for an additional 10-d period. Bisulfite sequencing data revealed that the promoter and CGI of the IL-32 gene remained largely hypomethylated in the cells that underwent 12 d of TNF-α treatment and 10 d of withdrawal (Fig. 4A), indicating that TNF-α induced DNA demethylation could be maintained for a considerable period of time.
Moreover, we noticed that prior exposure to long-term TNF-α treatment led to elevated basal IL-32 expression, even after 10 d of TNF-α withdrawal (Figs. 4B and 4C). These results indicated that long-term TNF-α treatment not only caused a stable epigenetic change but also led to a sustained change in the basal expression of the IL-32 gene.
To determine whether the above effect could be maintained for an even longer period, we treated HEK293 cells with TNF-α for 12 d and then cultured them in a TNF-α-free medium for 10, 18 or 30 d. RT-qPCR results showed that the upregulation of IL-32 was maintained after 10 d, 18 d, and 30 d, although the upregulated level became more moderate after 30 d (Fig. 4D). Consistently, the DNA methylation level of the IL-32 promoter and CpG island began to increase after 30 d of TNF-α withdrawal (Fig. 4E).
Taken together, these results suggest that long-term TNF-α treatment can induce heritable hypomethylation at the promoter and CpG island of the IL-32 gene, causing long-term transcriptional alteration.

TET enzymes mediate IL-32 demethylation during long-term TNF-α treatment
DNA demethylation can be achieved by passive demethylation or TET enzyme-mediated active oxidation and demethylation or both (21,30). To answer whether passive demethylation was involved in TNF-α induced demethylation, we attempted to arrest the cells at S phase and simultaneously treated the cells with TNF-α, unfortunately these cells suffered from severe cell death, and we were unable to draw a clear-cut conclusion about whether there was any involvement of passive demethylation.
To determine whether DNA demethylation at the promoter and CpG island of the IL-32 gene was mediated by TET enzymes, we generated TET1 knockout (KO), TET2 KO, TET3 KO and TET1/2/3 triple knockout (TKO) cells using the CRISPR-Cas9 system. In these cells, frameshift mutations were introduced at the carboxyl terminus of the TET family proteins to abrogate their catalytic activities (Supplemental Figs. S3, S4A, S4B).
We then performed bisulfite sequencing, and the results revealed that the DNA demethylation induced by TNF-α treatment at the IL-32 gene promoter and CGI was largely abrogated in the TET TKO cells, with the single knockouts each displaying varied partial defects (Fig. 5A). These results suggested that the TET enzymes function together to promote TNF-α-induced IL-32 gene demethylation. We also confirmed that there was no upregulation of the DNMT genes in the TET TKO cells by RNA-seq experiments (Supplemental Fig. S4C).
We next asked whether IL-32 gene demethylation mediated by the TET enzymes was responsible for the elevated IL-32 expression levels in cells recovered from long-term TNF-α treatment. Although IL-32 expression was induced by 12 h or 12 d of TNF-α treatment in all of the above cells (Supplemental Fig.  S5), elevated IL-32 basal expression was not observed in TET TKO cells withdrawn from long-term TNF-α treatment (Fig. 5B). These results are consistent with the methylation states of the promoter and CGI of the IL-32 gene in these cells and support that long-term TNF-α treatment induces DNA demethylation at the transcriptional regulatory regions of the IL-32 gene, elevating its basal expression level.

NF-κB-dependent transcriptional activation contributes to IL-32 gene demethylation and long-term elevation of its basal expression
TNF-α activates the NF-κB signaling pathway and induces nuclear translocation of the canonical p50/p65 heterodimer (31)(32)(33)(34)(35)(36). Interestingly, a p65 binding site (κB site) is located in the promoter of the IL-32 gene (Fig. 6A), and its presence was confirmed by our p65 ChIP-seq results (Fig. 6B). Therefore, we knocked out the RELA gene that encodes p65 in HEK293 cells using the CRISPR-Cas9 system (Supplemental Figs. S6A, S6B) and verified the cells using sequencing (Supplemental Fig. S6C) and Western blotting (Fig.  6C). RT-qPCR data revealed that TNF-α-mediated IL-32 activation was significantly impaired in RELA KO cells (Fig. 6D), indicating p65 is the predominant transcription factor mediating IL-32 induction in response to TNF-α. Moreover, in RELA KO cells treated with TNF-α for 12 d, the levels of DNA demethylation at the transcriptional regulatory regions of IL-32 were reduced, especially at the CGI of IL-32 gene (Fig. 6E). The impaired DNA demethylation at the CGI of IL-32 gene was accompanied with a less elevated basal expression of IL-32 in long-term TNF-α treated RELA KO cells (Fig. 6F).
These data collectively support that TNF-α-induced NF-κB signaling pathway activation leads to DNA-demethylation-independent short-term activation and DNA-demethylation-dependent elevation of IL-32 basal transcription in the absence of initial TNF-α treatment.

CREB and the cAMP-response element (CRE) at the IL-32 promoter are not required for elevated IL-32 basal expression upon long-term TNF-α treatment
The CpG site within a CRE of the IL-32 promoter was reported to be demethylated during influenza A virus infection, which increased transcription factor CREB binding (4). We wondered whether this CpG site within CRE was also a target for TNF-α-induced demethylation, playing a role in the long-term activation of IL-32 gene. Therefore, we examined the CRE in the IL-32 promoter (Supplemental Fig. S7A) and confirmed its demethylation by TNF-α treatment (Figs. 2B, 3C, 4A and 5A). We next asked whether this CRE mediates the upregulation of IL-32 transcription through long-term TNF-α stimulation. Frameshift mutations were introduced in both alleles of the CREB1 gene to disrupt CREB binding to CRE (Supplemental Fig. S7B). However, the RT-qPCR results revealed normal IL-32 activation by TNF-α in CREB1 KO cells (Supplemental Figs. S7C, S7D).
In addition, we also mutated this CRE within the IL-32 promoter from TGACGTCA to TTTCGTCA (Supplemental Fig. S7E). Again, RT-qPCR revealed a largely normal elevation of IL-32 basal expression after long-term TNF-α treatment (Supplemental Fig.  S7F).
Collectively, these data suggested that the long-term effect of TNF-α treatment is not solely dependent on the DNA demethylation of the CpG site within the CRE of the IL-32 promoter.

DISCUSSION
Signaling events triggered by environmental cues are well known for their roles in transcriptional regulation. In most cases, the majority of transcriptional changes triggered by signals are reset, and target gene expression returns to its initial basal level upon withdrawal of the environmental cues that initiated the signaling events (56,57). However, sometimes, signaling events can also trigger lasting epigenetic changes that facilitate a long-term effect (56-60), which is an interesting field termed "signal to chromatin" (61)(62)(63).
DNA methylation is certainly one of the most stable epigenetic marks that can mediate a lasting effect. In recent years, increasing evidence has supported the role of transcription factor binding in facilitating DNA demethylation in neighboring regions (37-50) as well as the role of signaling events in stimulating DNA demethylation (64). However, cases reporting a full axis from signal to TF to DNA demethylation to a lasting transcriptional change in the absence of the initiating signal are still limited (58). Here, we report one such casean axis involving a TNF-α signal, NF-κB pathway activation and association of p65 at the IL-32 promoter, TET enzyme-mediated IL-32 gene demethylation and the long-term activation of IL-32 expression (Fig. 7).
In addition to reporting the abovementioned case, the discovery of DNA demethylation-dependent and -independent mechanisms involved in activating IL-32 expression may have additional significance worthy of further investigation. As a TNF-α target, IL-32 has been reported to reciprocally induce the expression of TNF-α in certain cell types (1). We suspect that under certain in vivo situations, a strong acute inflammation event or the accumulative effect of several acute inflammation events may lead to the demethylation of the IL-32 gene and a lasting elevation of IL-32 basal expression, which may in turn stimulate TNF-α expression in these cells or neighboring cells. Such a self-reinforcing feedforward loop may well contribute to the conversion from acute inflammation to chronic inflammation.
Understanding the potential mechanisms governing the conversion from acute inflammation to chronic inflammation is highly important due to its relevance to human health. Although our current study does not offer a clear answer for this important question, it provides an interesting direction for future exploration. One obvious difficulty in following up this study is the lack of a mouse model. The IL-32 gene does not exist in rodents (11), and follow-up studies will likely focus on human diseases. Therefore, one key question is what kind of pathological conditions may be relevant to our observations. We reason that chronic inflammatory diseases and autoimmune diseases are potential candidates on which to focus.
TNF-α antagonists including soluble receptors and antibodies have excellent efficacy in the treatment of chronic inflammatory diseases (e.g., rheumatoid arthritis and inflammatory bowel disease) (65,66). Establishing a connection between TNF-α-induced demethylation and the long-term activation of proinflammatory genes, including but not limited to IL-32, in any of the above diseases would be highly interesting.
To offer a mechanistic answer for TNF-α-induced long-term gene activation in the absence of TNF-α, the current model is missing one piece. We reason that the long-term effect of TNF-α was due to DNA demethylation that facilitated the association of transcription factor(s) sensitive to DNA methylation. However, in this case, we do not yet know the identity of such transcription factor(s). The CREB binding site in the CRE of the IL-32 promoter and its association with CREB provided an ideal candidate, particularly because this site was found to be demethylated in A549 cells infected with influenza virus (4) and the association of CREB with CRE is DNA methylation sensitive (67,68). However, in our case, this site does not appear to be the sole answer, because neither mutation of the CREB gene nor mutation of the CRE site in the IL-32 promoter caused sufficient changes (Supplemental Fig. S7). Future studies in this direction are of great interest.
We also performed HPLC-MRM MS/MS experiments at various time points following TNF-α treatment and observed a gradual subtle decline of the global 5mC level (Supplemental Fig. S8). Obviously, TNF-α treatment induced DNA demethylation is not restricted to IL-32 gene. The identification of other potential targets and their biological significance are interesting topics for future investigation.

ChIP-seq
ChIP experiments were performed with HEK293 cells using previously described procedures (69). ChIP-seq libraries were constructed with a Kapa hyper prep kit (Kapa Biosystems, Cat# KK8504) and NEBNext multiplex oligos for Illumina (index primers set 1) (NEB, Cat# E7335). Libraries were sequenced via NovaSeq using the 150 bp paired-end mode.

Bioinformatics
50-bp single-end reads were generated by BGISEQ-500 platforms for mRNA-seq experiments (BGI, Shenzhen). Sequencing qualities were evaluated with FastQC software and aligned to human genome hg38 using STAR aligner. RPKM values were quantified using Cuffdiff (v2.0.2). FPKM values were added to a pseudo-value of 0.5 to avoid being divided by zero. ChIP-seq reads were generated by Illumina NovaSeq-6000 platforms (paired-end, 150 bp). Adaptors were removed by Trim_galore software, and then aligned to hg38 genome sequences (< 2 bp mismatches allowed) with Bowtie2. Uniquely mapped reads were kept and then were extended to the average fragment size. Genome profile files were generated with IGV tools and linearly normalized to the same depth of 10 million reads.

IL-32 locus-specific methylation analysis
To perform IL-32 promoter and CpG island (Supplemental Table 1) locus-specific methylation analysis, purified genomic DNA was treated with an EpiTect Bisulfite Kit (Qiagen, Cat# 59104), and the converted DNA was amplified using locus-specific nested PCR primers (Supplemental Table 2). Purified PCR products were cloned, sequenced and then analyzed using a BiQ Analyzer (70).

Primers for RT-qPCR
The sequences of primers used for RT-qPCR include the following: IL-32, forward TGGCGGCTTATTATGAGGAGC and reverse CTCGGCACCGTAATCCATCTC; GAPDH, forward CTGGGCTACACTGAGCACC and reverse AAGTGGTCGTTGAGGGCAATG.

Data availability
All high-throughput sequencing data have been deposited under the GEO accession number GSE121361.
G.L.). This work was also supported by grants from the Chinese Ministry of Science and Technology (2017YFA0504200, 2016YFA0100400 and 2015CB856200), the China Natural Science Foundation (31521002, 31425013, 31730047 and 31571344), the Chinese Academy of Sciences (XDB08010103, XDBP10, QYZDY-SSW-SMC031) and. Z.Zhang. is supported by the Youth Innovation Promotion Association (2017133) of the Chinese Academy of Sciences.    . IL-32 basal expression is upregulated after long-term TNF-α treatment and is accompanied by sustained hypomethylation at the promoter and CGI. A, Locus-specific bisulfite sequencing data showed that the hypomethylation status of the IL-32 promoter and CGI can be maintained after 10 d of TNF-α withdrawal. B, A time-course experiment revealed that the IL-32 basal expression level is upregulated after long-term TNF-α treatment and TNF-α withdrawal. Averages from three independent experiments are shown, and error bars represent standard deviation in the RT-qPCR results. C, Western blot results showed cells treated long-term with TNF-α display a higher basal protein expression level of IL-32. D, RT-qPCR results revealed that the upregulation of IL-32 expression can be maintained for at least 30 d after TNF-α withdrawal. Averages from three independent experiments are shown, and error bars represent standard deviation. E, Bisulfite sequencing data revealed that cells subjected to 12 d of TNF-α treatment maintained relatively low methylation levels at the promoter and CGI of the IL-32 gene even after 30 d of TNF-α withdrawal. Figure 5. TET enzymes mediated DNA demethylation, leading to upregulated IL-32 basal expression upon long-term TNF-α treatment. A, Locus-specific bisulfite sequencing results showed that TET enzymes are responsible for the DNA demethylation events during long-term TNF-α treatment. B, RT-qPCR results showed that the upregulated IL-32 expression that occurred after long-term TNF-α treatment is dependent on TET enzymes. Averages from three independent experiments are shown, and error bars represent standard deviation.