cAMP Signaling Regulates DNA Demethylation by Augmenting the Intracellular Labile Ferrous Iron Pool

It is widely accepted that cAMP regulates gene transcription principally by activating the protein kinase A (PKA)-targeted transcription factors. Here, we show that cAMP enhances the generation of 5-hydroxymethylcytosine (5hmC) in multiple cell types. 5hmC is converted from 5-methylcytosine (5mC) by Tet methylcytosine dioxygenases, for which Fe(II) is an essential cofactor. The promotion of 5hmC was mediated by a prompt increase of the intracellular labile Fe(II) pool (LIP). cAMP enhanced the acidification of endosomes for Fe(II) release to the LIP likely through RapGEF2. The effect of cAMP on Fe(II) and 5hmC was confirmed by adenylate cyclase activators, phosphodiesterase inhibitors, and most notably by stimulation of G protein-coupled receptors (GPCR). The transcriptomic changes caused by cAMP occurred in concert with 5hmC elevation in differentially transcribed genes. Collectively, these data show a previously unrecognized regulation of gene transcription by GPCR-cAMP signaling through augmentation of the intracellular labile Fe(II) pool and DNA demethylation.

and activating transcription factor 1 (ATF1), can be phosphorylated by PKA and subsequently 3 bind to cAMP response elements (CRE) in gene promoters, generally to activate gene 49 transcription (Sands and Palmer, 2008).

50
Schwann cells form the myelin sheath of axons within the peripheral nervous system. cAMP 51 is a known instructive signal for Schwann cell differentiation into a myelinating phenotype 52 (Jessen et al., 1991). cAMP directly induces cell cycle arrest along with the expression of a wide 53 variety of myelination-associated genes such as Egr2/Krox20, a transcription factor that is 54 considered a master regulator of the myelin program, and myelin protein zero, the main 55 peripheral nerve myelin protein (Bacallao and Monje, 2015). However, the mechanism by which 56 cAMP regulates transcription and promotes the differentiation of Schwann cells is not fully 57 understood.

58
DNA methylation is one major epigenetic mark that regulates gene transcription. Active DNA 59 demethylation is catalyzed by ten-eleven translocation (Tet) methylcytosine dioxygenases, 60 which oxidize 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC) (Tahiliani et al., 2009; 61 Kriaucionis and Heintz 2009), and further to 5-formylcytosine (5fC) and 5-carboxylcytosine 62 (5caC) Ito et al., 2011). 5fC and 5caC are ultimately replaced by unmodified 63 cytosine (5C) to complete cytosine demethylation (Pastor et al., 2013). It is noteworthy that 64 5hmC is also a unique epigenetic mark with regulatory capacities in addition to being a DNA 65 demethylation intermediate (Shen et al., 2014). Tet belongs to the iron and 2-oxoglutarate (2OG, 66 alternatively known as a-ketoglutarate)-dependent dioxygenase family which utilizes labile Fe(II) 67 as a cofactor and 2OG as a co-substrate. We and others have shown that ascorbate, which has 68 the capacity to reduce the redox-inactive Fe(III)/Fe(IV) to Fe(II), is another cofactor for Tet 69 (Minor et al., 2013;Yin et al., 2013;Blaschke et al., 2013;Dickson et al., 2013;Chen et al., 70 2013). Thus, ascorbate has an impact on DNA demethylation by promoting the availability of 71 redox-active Fe(II) to Tet. 5mC and 5hmC are major epigenetic marks that govern both cell 72 identity and cellular phenotype transformation. Thus, it is plausible that such marks are involved 73 in the transition of Schwann cells from the immature to myelinating phenotype. 4 Similar to cAMP, ascorbate was identified as another essential factor for Schwann cells to 75 initiate and promote myelin formation (Bunge et al., 1986). Intracellular ascorbate deficiency can 76 cause hypomyelination in modeled rodents (Gess et al., 2011). It is plausible that, by regulating 77 DNA demethylation, ascorbate alters the cellular phenotype of Schwann cells towards a 78 myelinating state. Based on this shared function in enhancing Schwann cell myelination, we 79 reasoned that cAMP, like ascorbate, might also play a role in Tet-mediated DNA demethylation.

80
Here, we investigated the potential role of cAMP in DNA demethylation. We identified that cAMP

97
In the absence of cAMP treatment, 5hmC was barely detectable by dot blot in primary cultured 98 rat Schwann cells. However, after treatment with membrane-permeable cAMP (100 µM), 5hmC 99 5 signal emerged at comparable levels to cells treated with ascorbate (50 µM, Figure 1A and 1B).

100
Encouraged by this initial observation, we investigated the role of cAMP on 5hmC by 101 immunofluorescence (IF). The addition of cAMP (0 -250 µM) dose-dependently increased 102 5hmC generation in Schwann cells ( Figure 1C and 1D). The IF signal for 5hmC was sustained 103 for 1 -7 days upon continuous cAMP (100 µM) treatment ( Figure 1E and 1F), suggesting that 104 continuous cAMP treatment largely maintained 5hmC. An enhanced generation of 5hmC was 105 also verified in other cell types examined, such as HEK-293 cells, mouse embryonic fibroblasts 106 (MEF) and neuroblastoma SH-SY5Y cells, indicating that the effect of cAMP on 5hmC is not 107 limited to Schwann cells and is likely to be a general effect ( Figure 1G).

108
We next tested whether the elevation of endogenous cAMP imparts a similar effect as 109 exogenously applied cAMP. Treating cells with forskolin, which directly activates 110 transmembrane adenylate cyclase (AC) to produce endogenous cAMP, promoted 5hmC 111 generation in Schwann cells as did bicarbonate, an activator of soluble AC ( Figure 1H). The 112 production of 5hmC was also observed when cells were treated with phosphodiesterase (PDE) 113 inhibitors caffeine or IBMX, both of which prevent cAMP degradation ( Figure 1H). Conversely, 114 no 5hmC signal was observed after cells were treated with AMP (100 µM) ( Figure 1H).

115
Collectively, these observations suggest that endogenous cAMP is indeed involved in 5hmC

119
To understand how cAMP enhances 5hmC generation, we first examined the transcription of

131
Released puromycin is incorporated in cellular proteomes, which can be visualized using a 132 convenient IF assay (Spangler et al., 2016). Treatment with cAMP at a dose as low as 1 µM 133 increased intracellular labile Fe(II), which reached even higher levels after treatment with 10 -

159
The molecular mechanism by which cAMP alters labile iron in principle could be related to iron

177
In an attempt to understand how cAMP signaling causes endosome acidification, we 178 examined the major known targets of cAMP, which include PKA, CNGCs, and Epac. PKA

221
It has been previously shown that cAMP shifts the transcriptome (Montminy, 1997). Indeed, 222 7,020 transcripts were differentially transcribed in Schwann cells in response to elevated cAMP 10 as shown by RNA-seq ( Figure 7A and 7B). Of the differential transcripts, 54% were upregulated 224 and 46% downregulated, which is concordant with the bi-directional transcriptional regulation of 225 5hmC . Furthermore, genome-wide 5hmC profiles were also altered by cAMP 226 treatment as revealed by hMeDIP-seq ( Figure 7C). In total, 66,963 5hmC peaks were 227 upregulated and 10,026 peaks were downregulated by cAMP treatment. By integrating RNA-228 seq and hMeDIP-seq, we found that 4,071 differential transcripts (58% of total differential 229 transcripts) correlate with altered 5hmC peaks located within promoter regions or gene bodies 230 ( Figure 8A). Overall, cAMP increased 5hmC level mainly in gene bodies of these differential 231 transcripts, suggesting that changes in 5hmC could be responsible for the differential 232 transcription or involved in its regulation.   cofactors, it appears also to have been appropriated for cellular signaling in the particular case 270 of enzymes that employ labile Fe(II) as an essential co-factor, such as the iron and 2OG-271 dependent Tet and Jumonji C domain-containing histone demethylases. The changes to labile 272 cytosolic Fe(II) measured by Trx-Puro likely also reflect the nuclear Fe(II) pool, since there is a 273 rapid equilibrium between these two Fe(II) pools, presumably via nuclear pores (Ma et al., 2015).
12 Thus, the results described above suggest that cAMP increases the accessibility of Fe(II) to Tet 275 by augmenting labile Fe(II) in the cytosol and nucleus.

276
An increase in labile Fe(II) can be induced by cAMP in as little as 2 hours, indicating that no 277 new protein synthesis is required. The molecular mechanism by which cAMP alters labile Fe(II) 278 appears to be related to iron uptake rather than storage. Cellular iron uptake is a multistep

312
One recent study showed that cAMP has an impact on DNA methylation by mainly 313 increasing the expression of Tet or DNA methyltransferases (Fang et al., 2015). This may not 314 be a general effect but more likely a cell specific effect since we observed a decreased

486
Statistical significance was determined using 3 alternative differential expression calculators: 490 differential transcripts, edgeR called 8,029 differential transcripts, and CuffDiff called 8,076 491 differential transcripts with a total of 7,020 transcripts called as differential by all three programs.

523
Reads were trimmed with trim_galore to remove low quality bases from reads (scores <20 in 524 Phred+33 format), and Illumina adapters. After quality control was checked sequence reads 525 were aligned to the rat genome (Rnor_6.0, Ensembl.org) using BWA (Li et al., 2009a). All    (Anders et al., 2015). Statistical significance was determined 546 using edgeR as was used for RNA-seq, except without using the calcNormFactors function so 547 as to normalize to total read counts rather than counts within peaks. This normalization was 548 necessary to account for a global change in 5hmC levels. Differential expression was also 549 calculated using DESeq2. To minimize false positives, we considered only peaks with a 550 minimum of 2X fold change and below an adjusted P-value (false discovery rate, FDR) below

Competing interests 616
The authors declare that no competing interests exist.            upregulated 5hmC at promoter and gene body regions of differential transcripts. (B) 33.8% of differential transcripts were associated with both 5hmC peaks and PKA-dependent transcription factors (TF). 24.2% of differential transcripts were associated with 5hmC peaks only and 22.7% of differential transcripts were associated with TF only. The rest of the differential transcripts (19.3%) was not associated with either 5hmC peaks or TF. (C) cAMP (100 µM) increased transcription and 5hmC, mainly in the gene bodies, of Egr2 and Pmp2 shown by UCSC Genome Browser views of hMeDIP-seq and RNA-seq reads (n = 1 experiment with 3 biological replicates).