Reversible promoter methylation determines fluctuating expression of acute phase proteins

Acute phase reactants (APRs) are secretory proteins exhibiting large expression changes in response to proinflammatory cytokines. Here we show that the expression pattern of a major human APR, that is C-reactive protein (CRP), is casually determined by DNMT3A and TET2-tuned promoter methylation status. CRP features a CpG-poor promoter with its CpG motifs located in binding sites of STAT3, C/EBP-β and NF-κB. These motifs are highly methylated at the resting state, but undergo STAT3- and NF-κB-dependent demethylation upon cytokine stimulation, leading to markedly enhanced recruitment of C/EBP-β that boosts CRP expression. Withdrawal of cytokines, by contrast, results in a rapid recovery of promoter methylation and termination of CRP induction. Further analysis suggests that reversible methylation also regulates the expression of highly inducible genes carrying CpG-poor promoters with APRs as representatives. Therefore, these CpG-poor promoters may evolve CpG-containing TF binding sites to harness dynamic methylation for prompt and reversible responses.


Introduction
Acute phase reactants (APRs) are liver-produced plasma proteins constituting an integral part of innate defense (Gabay and Kushner, 1999;Medzhitov, 2007). They are defined by a substantial change (>25%) of their plasma concentrations in response to inflammation. IL-6 (Kopf et al., 1994) and IL-1b (Zheng et al., 1995) are chief inducers of APR expression through activation of STAT3, NF-kB and C/EBP in hepatocytes (Bode et al., 2012;Quinton et al., 2012;Poli, 1998). C-reactive protein (CRP) is the first APR to be discovered, whose plasma concentrations at baseline are less than 2-3 mg/ml, but can rapidly increase up to 1000-fold upon infection or tissue injury; the heightened levels of CRP, however, return to the baseline with the resolution of inflammation (Pepys and Hirschfield, 2003;Du Clos, 2013;Pathak and Agrawal, 2019). The mechanisms of CRP induction have been thoroughly examined by reporter assays and truncation analysis. A region of~220 bp in the proximal promoter of CRP that contains (nonconical) binding sites for STAT3, NF-k B and C/EBP-b is identified to be sufficient to mediate CRP induction by IL-6 and IL-1b (Singh et al., 2007;Young et al., 2008; Figure 1A).
Intriguingly, a promoter SNP (rs3091244) associated with plasma levels of CRP is located at 286 bp upstream the transcription start site (Szalai et al., 2005;Zacho et al., 2008;Allin et al., 2010). This SNP does not exist in binding sites of transcription factors (TFs) critical to CRP expression. Rather, the major À286C allele constitutes a CpG motif, at which DNA methylation frequently occurs; whereas the minor alleles of À286A/T disrupt the CpG motif and are associated with enhanced CRP expression. Beside the À286CpG, there are only four additional CpGs within the proximal promoter of CRP. Importantly, two of those CpGs are located at the binding sites of STAT3 and NF-kB/C/EBP-b ( Figure 1A). Given that promoter methylation affects TF recruitment (Hu et al., 2013;Yin et al., 2017) and contributes to gene silencing (Jones, 2012;Wu and Zhang, 2014;Dor and Cedar, 2018;Luo et al., 2018;Blattler and Farnham, 2013), it is notable that levels of promoter methylation and expression of CRP appear to be negatively associated albeit with Figure 1. Methylation level of CRP promoter is inversely associated with expression. (A) Schematic illustration of CRP promoter in which SNP rs3091244, CpG motifs and TF binding sites are indicated. (B) Methylation levels of CRP promoter (À550~1 bp) in pooled normal human tissues adjacent to tumors (five liver, eight colon, 10 esophagus, 10 rectum and 10 gaster) were determined by bisulfite cloning sequencing. (C) Methylation levels of CRP promoter in normal human tissues were retrieved from available GEO datasets generated by whole-genome bisulfite sequencing: Liver 01-GSM916049, Liver 02-GSM1716965, Adipose-GSM1120331, Adrenal-GSM1120325, Aorta-GSM1120329, Esophagus-GSM983649, Gaster-GSM1120333, Lung-GSM983647, Ovary-GSM1120323, Muscle-GSM1010986, Atrium-GSM1120335, Colon-GSM983645, Spleen-GSM983652, Thymus-GSM1120322 (Gene Expression Omnibus database). The bisulfite sequencing tracks of CRP promoter (left; the height of the black bars represents percentage of DNA methylation) and pooled analysis (right) are shown. (D) Methylation levels of CRP promoter in rabbit tissues were determined by bisulfite cloning sequencing. Liver is the major organ expressing CRP in both humans and rabbits. Accordingly, the methylation levels of CRP promoter are lower in normal liver tissues than in other tissues. (E) Levels of CRP expression (left) and promoter methylation (right) in tumor versus normal tissues from human livers (n = 5) were determined by q-PCR and bisulfite cloning sequencing, respectively. Liver tumors exhibit higher levels of CRP expression but lower levels of promoter methylation than adjacent normal liver tissues. *p<0.05 (paired t-test). (F) Bisulfite cloning sequencing of À286C versus À286A alleles of CRP promoter in Hep3B cells at resting (Vehicle treated) or induced states (IL-6 and IL-1b treated). À286A allele was less methylated than À286C allele at both states. (G) IL-6 (10 ng/ml) and IL-1b (1 ng/ml) treatment induced CRP expression (left) and promoter demethylation (right) in Hep3B cells, while withdraw of these cytokines led to a quick drop of CRP expression and promoter re-methylation. The result of one representative experiment is shown.
undefined causality (Wang et al., 2014). In the present study, we demonstrate that the expression pattern of CRP is causally determined by reversible promoter methylation, and that this regulation may also apply to highly inducible genes with CpG-poor promoters.

Promoter methylation is inversely associated with CRP expression
To determine whether promoter methylation affects CRP expression, we first compared methylation levels of CRP promoter in different human tissues. CRP is expressed predominantly, if not solely, by the liver (Pepys and Hirschfield, 2003;Du Clos, 2013). Accordingly, methylation levels of CRP promoter in normal liver tissues were much lower than that in other tissues ( Figure 1B). Analysis of published bisulfite sequencing datasets also confirmed that CRP promoter was most demethylated in the liver ( Figure 1C). Similar results were further obtained in rabbits ( Figure 1D), wherein CRP exhibits a comparable expression pattern as in humans. Moreover, malignant liver tissues expressed more CRP than adjacent normal tissues, and they were also less methylated at CRP promoter ( Figure 1E). These data together reveal an inverse association between levels of promoter methylation and CRP expression across different tissues or cell types.
Hepatic Hep3B cell line is a conventional model to investigate APR expression (Singh et al., 2007;Young et al., 2008). CRP promoter in Hep3B cells harbors distinct alleles at the À286 position, with À286C on one allele and À286A on the other. Intriguingly, in addition to lacking the À286CpG, all other promoter CpGs on the À286A allele were much less methylated than that on the À286C allele at the resting state ( Figure 1F). Such an allelic imbalance of promoter methylation was further reinforced at the induced state. Notably, the induction of CRP by IL-6 and IL-1b was accompanied by prominent promoter demethylation ( Figure 1G). Following washout of the cytokines, however, both the expression and the promoter methylation of CRP were rapidly recovered. By contrast, the methylation level of a 5' UTR CpG remained constant during the entire time course. Therefore, levels of promoter methylation and CRP expression are also specifically and dynamically associated in the same cell type.

Promoter methylation causally determines CRP expression
To clarify whether the observed association is causal, we directly modulated methylation levels of CRP promoter and examined its consequence on expression. Treating Hep3B cells with 5-aza or RG108 to inhibit DNA methylation significantly enhanced CRP expression at the resting state, but showed little effect at the induced state ( Figure 2A; Figure 2-figure supplement 1) wherein CRP promoter also underwent active demethylation ( Figure 1G). Nevertheless, 5-aza could moderately rescue the induced expression of CRP when STAT3 or NF-kB was inhibited ( Figure 2B and C), hinting for their involvement in active demethylation of CRP promoter. Moreover, in vitro methylation before transfection markedly suppressed the reporter activity of CRP promoter in Hep3B cells ( Figure 2D). This suppression, however, was partially reversed by mutating individual CpG motifs, and was completely absent with a CpG-null mutant of CRP promoter. These results suggest that promoter methylation inhibits, whereas its demethylation enhances CRP expression, thus supporting a causal association.
CpG methylation and demethylation are mediated by DNA methyltransferases (DNMTs) and teneleven translocations (TETs), respectively (Jones, 2012;Wu and Zhang, 2014;Dor and Cedar, 2018;Luo et al., 2018). A causal association with promoter methylation would therefore predict that the expression of CRP should also be regulated by DNMTs and/or TETs. Indeed, RNAi screening revealed that knockdown (KD) of DNMT3A enhanced ( Figure 3A), whereas KD of TET2 reduced CRP expression in Hep3B cells ( Figure 3B). Knockout (KO) of DNMT3A ( Figure 3C) or TET2 ( Figure 3D) with Cas9 yielded consistent but more pronounced effects. Importantly, DNMT3A/TET2 KD or KO showed expected effects on CRP promoter methylation and on their own expression (  Figure 3E). These results together identify DNMT3A and TET2 as the negative and positive regulators of CRP expression, respectively, thus reinforcing the notion that CRP expression is causally determined by promoter methylation.
Of note, inhibitor treatment or DNMT3A/TET2 manipulations would all affect the methylation status of entire genome. To exclude any indirect effect caused by global manipulation, we specifically modulated the methylation levels of CRP promoter by dCas9-mediated targeting of the catalytic domains of DNMT3A or TET2. Enforced methylation of CRP promoter by DNMT3A-dCas9 reduced the expression of CRP in Hep3B cells, but showed little effect on that of serum amyloid A (SAA, another major human APR) and serum amyloid P component (SAP, a paralog of CRP) ( Figure 3F). By contrast, enforced demethylation of CRP promoter by TET2-dCas9 only selectively enhanced the expression of CRP ( Figure 3G). We thus conclude that DNMT3A and TET2-tuned methylation status of CRP promoter constitutes a key part of the regulatory mechanism that causally determines the expression. The effects of DNA methylation inhibitor RG-108 (25 mM, 24 hr) or 5-aza (5 mM, 12 hr) on CRP expression in Hep3B cells at the resting or induced state (n = 3). These inhibitors enhanced the resting but not the induced expression of CRP. At the induced state, the defective CRP expression caused by STAT3 (s31-201, 30 mM, 24 hr) (B) or NF-kB inhibition (BAY11-7082, 2 mM, 24 hr) (C) was partially reversed by 5-aza (5 mM, 24 hr) (n = 3). (D) In vitro vector methylation markedly inhibited reporter activities of wildtype CRP promoter (WT) following transfection into Hep3B cells (n = 3). Mutating individual CpG motif partially reversed this inhibition. As the control, in vitro vector methylation did not affect reporter activities of a CpG-null version of CRP promoter. Data are presented as mean ± SEM. *p<0.05, **p<0.01, ***p<0.001 (two-tailed t-test). The online version of this article includes the following figure supplement(s) for figure 2:

Promoter methylation of CRP dictates strength of TF recruitment
We next asked whether the promoter methylation-mediated regulation could be conferred by influencing TF recruitment. Indeed, IL-6 and IL-1b-induced demethylation of CRP promoter ( Figure 1G) was paralleled by markedly enhanced recruitment of STAT3, NF-kB p50 and C/EBP-b ( Figure 4A; Singh et al., 2007;Young et al., 2008). Moreover, in vitro methylation substantially reduced the recruitment of those TFs to vectors containing CRP promoter after transfection into Hep3B cells ( Figure 4B). The À53CpG and À108CpG are at the binding sites of p50/C/EBP-b, and STAT3, respectively ( Figure 1A). Accordingly, site-specific methylation of À53CpG selectively prevented the recruitment of p50 and C/EBP-b to CRP promoter, while site-specific methylation of À108CpG only (C) CRP expression in Hep3B cells with co-transfected Cas9 and sgRNAs targeting exon 14 (sgDNMT3A-1) or 2 (sgDNMT3A-2) of DNMT3A (n = 3). (D) CRP expression in Hep3B cells with co-transfected Cas9 and sgRNAs targeting exon 3 (sgTET2-1) or 7 (sgTET2-2) of TET2 (n = 3). (E) CRP expression in Hep3B cells with overexpressed DNMT3A or TET2 (n = 3). CRP expression in Hep3B cells with co-transfected catalytic domain of DNMT3A (F) or TET2 (G) fused to dCas9 and sgRNAs targeting CRP promoter (n = 3). The results identified DNMT3A and TET2 as the negative and positive regulators of CRP expression, respectively. Selective targeting of DNMT3A or TET2 to CRP promoter by dCas9 only regulated the expression of CRP, but did not affect that of serum amyloid A (SAA; a major human APR) or serum amyloid P component (SAP, a paralog of CRP). Data are presented as mean ± SEM. *p<0.05, **p<0.01, ***p<0.001 (two-tailed t-test). The online version of this article includes the following figure supplement(s) for figure 3:    The recruitment of STAT3, p50 and C/EBP-b to CRP promoter were all markedly enhanced at the induced versus resting state (n = 3). (B) In vitro vector methylation decreased the recruitment of STAT3, p50 and C/EBP-b to a vector containing CRP promoter at the induced state (n = 3). (C) Sitespecific methylation at À53CpG inhibited the recruitment of p50 and C/EBP-b, whereas methylation at À108CpG inhibited the recruitment STAT3 to the vector containing CRP promoter at the induced state (n = 3). Data are presented as mean ± SEM. *p<0.05, **p<0.01, ***p<0.001 (two-tailed t-test).
inhibited the recruitment of STAT3 ( Figure 4C). These CpG motifs at TF binding sites may thus act as rheostats with their methylation turning down the recruitment of critical TFs, resulting in reduced expression.

Dynamic crosstalk among TFs and promoter methylation in induced expression of CRP
We further examined how TF recruitment and promoter methylation dynamically orchestrate to regulate the induced expression of CRP. IL-6 and IL-1b induced two waves of CRP expression: the first wave lasted from 0 to 6 hr yielding the minor peak, while the second lasted from 12 to 24 hr yielding the major peak ( Figure 5A). The recruitment of STAT3 occurred during the first wave and saturated at 3 hr before the minor peak ( Figure 5B). By contrast, the recruitment of p50 was more evident during the time lag between the two waves ( Figure 5C). The recruitment of C/EBP-b, whose action depends on p50 (Cha-Molstad et al., 2000;Kramer et al., 2008;Agrawal et al., 2001), however, steadily rose till 12 hr ( Figure 5D). These would suggest that the first wave of induced CRP expression is driven by early recruited STAT3, which licenses the late recruitment of p50 that synergizes with C/EBP-b to drive the second wave. As such, STAT3 is likely the pioneer TF that binds methylated CRP promoter to initiates induction and primes demethylation.
In line with the above suggestion, STAT3 was the only TF showing appreciable early recruitment to CRP promoter upon enforced DNA methylation by TET2 KO (Figure 5B-D). This indicates that STAT3 can nevertheless be recruited to CRP promoter even when heavily methylated, consistent with the observations that vector binding of STAT3 was least sensitive to methylation ( Figure 4B and C). This also indicates that STAT3 can act largely independent of p50 and C/EBP-b to drive CRP induction, albeit with a markedly reduced amplitude ( Figure 5A). Indeed, the sole activation of STAT3 in wildtype cells was able to induce CRP expression to a level comparable to that of the minor peak, whereas the sole activation of NF-kB was completely ineffective ( Figure 5E). Despite that, NF-kB inhibition (with intact STAT3) at the induced state resulted in an even stronger methylation of CRP promoter ( Figure 5F) and a reversal of allelic imbalance ( Figures 5G and  1F). Therefore, promoter demethylation requires p50 that acts downstream of STAT3.
Interestingly, enforced DNA demethylation by DNMT3A KO not only tripled CRP expression during the entire course of induction, but eliminated the time lag between the two waves ( Figure 5A). The augmented amplitude can be explained by the enhanced recruitment of the three TFs, while the altered dynamics may correspond to the shifted timing of p50 recruitment ( Figure 5B-D). As such, p50 selectively recruited during the time lag could be responsible for promoter demethylation to prime the second wave. Accordingly, C/EBP-b appears to be the major effector that responds to promoter demethylation: its overexpression did not demethylate CRP promoter ( Figure 5H and I), but when combined with blockage of DNA methylation, it drove the resting expression of CRP to a level approaching to that induced by IL-6 and IL-1b ( Figure 5J). C/EBP-b KO, however, lowered the induced expression of CRP by~70% ( Figure 5K). These together demonstrate a stepwise induction of CRP where TFs and promoter methylation dynamically orchestrate ( Figure 5L).

Reversible methylation regulates expression of genes with CpG-poor promoters
Having established the regulation of CRP expression by reversible promoter methylation, we wondered whether the same regulation can be applied to other APRs. Indeed, SAA behaved similarly as CRP. Treating Hep3B cells with IL-6 and IL-1b resulted in a drastic increase in the expression of SAA ( Figure 6A) and a reduction in methylation levels of its promoter ( Figure 6B). These were, however, quickly recovered following cytokine withdraw. By contrast, neither the expression nor the promoter methylation of SAP was affected by treatment or withdraw of IL-6 and IL-1b ( Figure 6A and C). Moreover, DNMT3A KO also markedly enhanced the induced expression of SAA, but barely affected that of SAP ( Figure 6D). These results suggest that reversible promoter methylation may be a general mechanism underlies the induction of APRs.
The promoters of most mammalian genes contain a high frequency (observed number/expected number >0.6) of CpGs termed CpG islands (CGIs) that are resistant to DNA methylation (Saxonov et al., 2006). The CpG frequency of CRP promoter, however, is exceptionally low (~0.23). Interestingly, a low CpG frequency appears to be general feature of APR promoters ( Figure 7A). The recruitment of STAT3 to CRP promoter was still evident in TET2 KO cells. DNMT3A KO resulted in a stronger amplitude and altered dynamics of CRP induction. The recruitment to CRP promoter was enhanced for all the three TFs in DNMT3A KO cells, whereas the timing of recruitment was altered only for p50. (E) CRP expression in Hep3B cells treated with vehicle, 1 ng/ml IL-1b, 10 ng/ml IL-6 or their combination for 48 hr (n = 3). As IL-1b is unable to induce IL-6 production in Hep3B cells (Kramer et al., 2008), the effects of STAT3 and p50 can be largely dissociated by treating cells with one single cytokine (Kramer et al., 2008;Ganapathi et al., 1991;Ganapathi et al., 1988). IL-1b could not induce CRP expression, suggesting p50 is not required for the first wave of CRP induction. (F) Methylation levels of CRP promoter in Hep3B cells at the induced state treated without (Vehicle) or with inhibitors of STAT3 (30 mM s31-201) or NF-kB (2 mM BAY11-7082) for 24 hr. (G) Ratios of methylation levels on À286C versus À286A alleles in Hep3B cells at the induced state treated with the NF-kB inhibitor (2 mM BAY11-7082) for 24 hr. Methylation levels (H) and allelic methylation of CRP promoter (I) in Hep3B cells expressing a control or a C/EBP-b vector at the resting state. C/EBP-b overexpression showed no effect on methylation status of CRP promoter. The result of one representative experiment is shown. (J) CRP expression in Hep3B cells with or without C/EBP-b overexpression under the indicated conditions for 48 hr (n = 3). (K) CRP expression in Hep3B cells without (sgNT) or with C/EBP-b KO (sgC/EBP-b) following induction with 10 ng/ml IL-6 and 1 ng/ml IL-1b for 48 hr (n = 3). The dramatic effects of C/EBP-b KO or Figure 5 continued on next page We then extended our analysis to genes with CpG-poor promoters. In Hep3B cells treated with IL-6 and IL-1b, strongly induced genes tended to manifest lower CpG ratios in their promoters Figure 5 continued overexpression suggest that this TF is the major effector that respond to promoter methylation status and determine the amplitude of CRP expression. (L) A schematic illustration of how TF recruitment and promoter methylation dynamically orchestrate to regulate the induction of CRP. Data are presented as mean ± SEM. **p<0.01, ***p<0.001 (two-tailed t-test).  Figure 7C). Importantly, genes with CpG-poor promoters showed significantly stronger changes in both expression and promoter methylation ( Figure 7D and E; Figure 7-figure supplement 1). Therefore, dynamic methylation may also regulate the expression of highly inducible genes with CpG-poor promoters.

Discussion
Though TFs critical to CRP expression have been identified, how their actions are coordinated remains unclear. This study demonstrates a previously unrecognized, epigenetic mechanism wherein methylation status of CRP promoter responds to and further modifies the effects of distinct TFs. At the induced state, the pioneered binding of STAT3 to CRP promoter in hepatocytes drives a minor wave of induction, and further licenses the subsequent recruitment of NF-kB p50. The two TFs probably work together to tip the balance of TET2 and DNMT3A at CRP promoter, leading to its demethylation. As significant cell proliferation was not noted, the methylated cytosine might be eventually removed by base excision repair mechanism . Consequently, the demethylated promoter enhances the recruitment of C/EBP-b to drive the major wave of CRP Hep3B cells were treated with 10 ng/ml IL-6 and 1 ng/ml IL-1b for 24 hr, and gene expression profiles were determined by RNA-seq. Genes with larger expression changes exhibited lower promoter CpG ratios. There are 10979, 2130, 454, and 121 genes in <2 fold, 2~5 fold, 5~15 fold, and >15 fold categories, respectively. Mouse liver tissues were collected at the resting or turpentine-induced state. Their transcriptome and methylome were then determined and compared. (C) Genes whose expression changed by over 2-fold between the two states exhibit lower promoter CpG densities. There are 12338, 2876, 444, and 51 genes in <2 fold, 2~5 fold, 5~15 fold, and >15 fold categories, respectively. With the increase in promoter CpG density, genes show markedly reduced changes in levels of their expression (D) (There are 3096, 3903, 3903, and 3907 genes in Q1, Q2, Q3, and Q4 categories, respectively) and promoter methylation (E) (There are 3745, 3738, 3741, and 3742 genes in Q1, Q2, Q3, and Q4 categories, respectively). Statistical analysis was performed using K-S test. The online version of this article includes the following figure supplement(s) for figure 7: induction. During the recovery phase, however, the loss of activated STAT3 and p50 results in a rapid remethylation of CRP promoter and termination of induction. At the resting state, however, the relatively hypomethylated promoter of CRP in the liver versus other tissues likely favors C/EBP-b recruitment, contributing to its tissue-specific, basal expression. These may form the basis for CRP, a putative pattern recognition receptor (Du Clos, 2013;Bottazzi et al., 2010), to constitute an integral part of immune surveillance in both homeostasis and inflammation.
In addition to CRP, we further show that reversible methylation also appears to be involved in regulation of highly inducible genes carrying CpG-poor promoters with APRs as representatives. In this regard, it is worth noting that DNA methylation is a relative stable epigenetic modification . Though dynamic changes in global or local DNA methylation status have been demonstrated in processes of development, aging and disease, these (gradual) changes, once occurred, are largely persistent or irreversible Dor and Cedar, 2018;Luo et al., 2018;Halder et al., 2016;Sellars et al., 2015;Domcke et al., 2015;Flavahan et al., 2016;Dmitrijeva et al., 2018;Koch et al., 2018;Horvath and Raj, 2018) with its causality in determining gene expression even being questioned (Bestor et al., 2015). There are, however, only rare cases where the expression of specific genes, such as pS2 (Métivier et al., 2008;Kangaspeska et al., 2008) and IL-10 (Hwang et al., 2018), is regulated by rapid and reversible changes in DNA methylation. Our findings here identify an important scenario in which reversible promoter methylation plays a critical role in determining the expression pattern of a class of proteins featured by CpG-poor promoters in response to inflammatory stimuli.
Acute changes in DNA methylation have also been analyzed at a genome-wide scale in previous studies. One study examined mouse neurons activated by electroconvulsive stimulation, and found that methylation changes were preferentially occurred in CpG-poor regions (Guo et al., 2011). Though promoters were underrepresented in these regions, their methylation changes were nevertheless modestly anticorrelated with gene expression (Guo et al., 2011). Such CpG content-dependent changes in DNA methylation, however, were not observed in another study examining neurons activated by contextual learning (Halder et al., 2016). Moreover, DNA methylation changes in human dendritic cells following infection rarely occurred at promoters (Pacis et al., 2019;Pacis et al., 2015), and were claimed to be a consequence of gene expression (Pacis et al., 2019). Those findings thus argue that the regulation of reversible methylation on inducible expression of genes with CpG-poor promoters may be context-dependent, and that rigorously controlled case study should be integrated into genome-wide investigation to conclude on causality.
Interestingly, genes showing inducible expression in macrophages activated by endotoxin have been classified into two groups: nucleosome remodeling-independent genes with CpG-rich promoters, and nucleosome remodeling-dependent genes with CpG-poor promoters (Ramirez-Carrozzi et al., 2009). Despite their distinct requirements for SWI/SNF complexes, preassembled Pol II and new protein synthesis, both groups of genes exhibit a single-wave kinetics of induction (Ramirez-Carrozzi et al., 2009;Hargreaves et al., 2009;Ramirez-Carrozzi et al., 2006). This is in contrast with CRP and probably other major APRs, which show a two-wave kinetics of induction with the second wave licensed by promoter demethylation. However, nucleosome remodeling might still act downstream in the second wave, as C/EBP-b has recently been shown to promote CRP expression by recruiting BRG1 via MKL1 (Fan et al., 2019). Conversely, promoter methylation could also contribute to inducible expression of nucleosome remodeling-dependent genes in macrophages by directly regulating TF recruitment (Hu et al., 2013;Yin et al., 2017;Thomas et al., 2012).
Despite that most CpGs in mammalian promoters not associated with CGIs are usually methylated Luo et al., 2018), TETs and DNMTs can nevertheless mediate active demethylation and de novo methylation, respectively. This indicates that the methylation status of part of the genome may depend on the balance of the two types of enzymes (Jones, 2012;Wu and Zhang, 2014;Dor and Cedar, 2018;Luo et al., 2018;Blattler and Farnham, 2013). Cellular signaling able to tip the balance of TETs and DNMTs could thus represent a regulatory mechanism that finely and reversibly tunes the expression of certain genes in response to environmental cues. In case of CRP, DNMT3A and TET2 appear to be involved in the regulation. Though the mechanism of TF-induced local demethylation is not fully understood (Luo et al., 2018), TFs including NF-kB and EGR1 have been reported to evoke DNA demethylation in neurons (Jarome et al., 2015) for example by recruiting TET1 (Sun et al., 2019). Future study is warranted to elucidate how TFs regulate the balance of TETs and DNMTs and to discover scenarios where regulation by reversible DNA methylation plays a prominent role.
In some experiments, vectors were either entirely methylated with CpG methyltransferase (NEB, Ipswich, MA; catalog number: M0226L; lot number: 0311608) or site-specifically methylated at À53 or À108 CpG sites of CRP promoter through vector PCR using appropriately methylated primers before transfection.

Statistical analysis
Data were presented as mean ± SEM. Statistical analysis was performed by the two-tailed Student's t-test, one-way ANOVA with Tukey post hoc or K-S tests as appropriate. Values of p<0.05 were considered significant.