Retinoic Acid and GM-CSF Coordinately Induce Retinal Dehydrogenase 2 (RALDH2) Expression through Cooperation between the RAR/RXR Complex and Sp1 in Dendritic Cells

Retinoic acid (RA)-producing dendritic cells (DCs) play critical roles in gut immunity. Retinal dehydrogenase 2 (RALDH2) encoded by Aldh1a2 is a key enzyme for generating RA in DCs. Granulocyte–macrophage colony-stimulating factor (GM-CSF) potently induces RALDH2 expression in DCs in an RA-dependent manner, and RA alone weakly induces the expression. However, how GM-CSF and RA induce RALDH2 expression remains unclear. Here, we show that GM-CSF-induced activation of the transcription factor Sp1 and RA-dependent signaling via the RA receptor (RAR)/retinoid X receptor (RXR) complex contribute to Aldh1a2 expression. The RAR antagonist LE540 and the Sp1 inhibitor mithramycin A inhibited GM-CSF-induced Aldh1a2 expression in fms-related tyrosine kinase 3 ligand-generated bone marrow-derived DCs (BM-DCs). ERK and p38 MAPK inhibitors suppressed GM-CSF-induced nuclear translocation of Sp1 and Aldh1a2 expression. Sp1 and the RARα/RXRα complex bound to GC-rich Sp1-binding sites and an RA response element (RARE) half-site, respectively, near the TATA box in the mouse Aldh1a2 promoter. The DNA sequences around these sites were highly conserved among different species. In the presence of RA, ectopic expression of RARα/RXRα and Sp1 synergistically enhanced Aldh1a2 promoter-reporter activity. GM-CSF did not significantly induce Aldh1a2 expression in plasmacytoid DCs, peritoneal macrophages, or T cells, and the Aldh1a2 promoter in these cells was mostly unmethylated. These results suggest that GM-CSF/RA-induced RALDH2 expression in DCs requires cooperative binding of Sp1 and the RAR/RXR complex to the Aldh1a2 promoter, and can be regulated by a DNA methylation-independent mechanism.

Multiple factors, including RA, granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin (IL)-4, IL-13, Tolllike receptor (TLR) ligands, peroxisome proliferator-activated receptor-c agonists, and b-catenin-activating factors can participate in inducing Aldh1a2 expression in DCs [11,[13][14][15][16][17][18][19].GM-CSF is one of the most potent inducers of Aldh1a2 expression in DCs, and it appears to play an important role in the steady-state expression of RALDH2 in MLN-DCs [11], although its contribution can be exerted by other factors depending on the rearing conditions or the animal strains used [20].IL-4 is also a potent inducer of Aldh1a2 expression in DCs, and GM-CSF and IL-4 synergistically enhance RALDH2 expression, although IL-4 is not essential for the steady-state expression of RALDH2 in MLN-DCs [11].TLR stimulation alone induces low RALDH2 expression levels in immature DCs; however, it markedly enhances GM-CSFinduced Aldh1a2 expression and maturation [11].However, the involvement of TLR stimulation in Aldh1a2 expression in gut DCs in vivo remains controversial, as different groups have reported conflicting results [12,19,20].
There may be redundant pathways for inducing or enhancing Aldh1a2 expression, and alternative pathways may be utilized under certain circumstances, particularly in gene-knockout mice.However, RA and b-catenin do appear to be essential for Aldh1a2 expression in DCs, because a deficiency in vitamin A or b-catenin almost completely inhibits Aldh1a2 expression and RALDH2 activity in DCs [11,16].
In the present study, we assessed the molecular mechanisms involved in GM-CSF-induced and RAR-dependent Aldh1a2 expression in DCs.RA alone induces weak RALDH2 expression in fms-related tyrosine kinase 3 ligand (Flt3L)-generated bone marrow (BM)-derived immature DCs (BM-DCs); however, it is required for GM-CSF-induced RALDH2 expression in these cells [11].We found that the RAR/retinoid X receptor (RXR) complex bound to an RA response element (RARE) half-site located near the TATA box in the mouse Aldh1a2 promoter.This promoter was located within a CpG island, and contained multiple Sp1 binding sites, including one that was near the RARE half-site.Thus, we propose that Aldh1a2 expression in normal DCs requires GM-CSF/RA-dependent activation of the Aldh1a2 promoter through the cooperative binding of Sp1 and RARa/RXRa to this promoter, and is regulated by a DNA methylation-independent mechanism.

Ethics statement
All animal experiments were performed according to the protocols approved by the Animal Care and Use Committee of Tokushima Bunri University (Approved Number: KP13-041-001).

Mice
B10.D2 mice and C57BL/6 mice were from Japan SLC and CLEA Japan, respectively.

Real-time PCR
Total RNA was isolated from cells using RNeasy Mini Kit, and cDNA was generated using QuantiTect Reverse Transcription Kit (both from Qiagen), according to the manufacturer's instructions.The level of Aldh1a2 gene expression was determined by real-time PCR in triplicates with Power SYBR Green PCR Master Mix (Applied Biosystems) and gene-specific primers (Table S1) using an Applied Biosystems 7500 or 7900 Real-time PCR system.Quantitative normalization of cDNA in each sample was obtained by the DCT method (Aldh1a2 CT -Rplp0 CT).

Construction of plasmids
For the analysis of the Aldh1a2 promoter, the 59-flanking region of the mouse Aldh1a2 gene was cloned by PCR using mouse genomic DNA as a template and specific reverse and forward primers (Table S1).A 2.8-kb PCR product (22,600 to +182) was inserted into the pGEM-T Easy vector (Promega), according to the manufacturer's instructions.The insert sequence of the resulting vector was identical to the sequence found in NCBI, GenBank TM accession number NT_039474.8.The insert was then ligated in the promoterless firefly luciferase reporter vectors, pGL3-basic and pGL4-basic (Promega), at the KpnI and XhoI sites.Various truncated forms of the mouse Aldh1a2 promoter were also created by PCR using reverse and forward primers (Table S1).The products were inserted into the pGL3-basic vector at the KpnI and HindIII sites.Site-directed mutagenesis of putative RARE half-sites in the Aldh1a2 promoter was introduced by the QuickChange Sitedirected Mutagenesis Kit (Agilent Technologies), using reverse and forward primers (Table S1), according to the manufacturer's instructions.Various promoter vectors lacking a GC-rich region were also created by the QuickChange Site-directed Mutagenesis Kit, using reverse and forward primers (Table S1).In some experiments, PCR fragments amplified from pGL3-RALDH2 using reverse and forward primers (Table S1) were inserted into the SpeI and HindIII sites in the CpG dinucleotide-free pCpGL luciferase vector, a kind gift from Dr. M. Rehli (University Hospital of Regensburg, Germany) [21], and GT115 E. coli cells (InvivoGen) were transformed with the plasmids.The mouse Sp1 was cloned by PCR using mouse spleen (SPL) cDNA as a template and specific reverse and forward primers (Table S1).Then the PCR product was inserted into the pGEM-T Easy vector, according to the manufacturer's instructions.The PCR product was then inserted into pCMV-Myc expression vector using EcoRI and SalI sites to obtain the Sp1 expression vector pCMV-Myc-Sp1.The expression vector containing only the DNA-binding domain of Sp1 (pCMV-Myc-Sp1db) was also constructed by PCR using specific reverse and forward primers (Table S1).All the constructs were verified by sequencing using Big Dye Terminator (Applied Biosystems) on an ABI 3130 sequencer (Applied Biosystems).pSG5-RARa was a kind gift from Dr. P. Chambon (IGBMC, Universite ´Louis Pasteur) [22].pSG5-RXRa was previously described [23].

Subcellular fractionation
Cytosolic and nuclear protein extracts from BM-DCs were prepared by means of NE-PER Nuclear and Cytoplasmic Extraction Reagent (Thermo Scientific) according to the manufacturer's instructions.The purity of the cytosolic fraction and nuclear fraction was verified using anti-a-tubulin (Sigma-Aldrich) and anti-lamin B1 (MBL) Abs, respectively.

DNA affinity precipitation (DNAP) assay
DNAP assay was performed as previously described [23].The biotin-labeled DNA probes (Sigma-Aldrich) were annealed to complementary oligonucleotides.COS-7 cells were transfected with 1 mg of expression vectors using LipofectAmine 2000 (Invitrogen), according to the manufacturer's instructions.Transfected COS-7 cells or nuclear fractions of BM-DCs were lysed or diluted with DNAP binding buffer [25 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1 mM EDTA, 0.25% NP-40, 1 mM DTT and Complete Protease Inhibitor Cocktail (Nacalai Tesque)], respectively.Cell debris was removed by centrifugation (20,0006g) for 10 min.Lysates were first incubated with Streptavidin-Sepharose beads (GE Healthcare) for 30 min to eliminate nonspecific binding and then incubated with 1.5 mg of poly(dI-dC) and 2 mg of biotinylated DNA probe for 1 h at 4uC.Streptavidin-Sepharose beads were then added and incubated with these mixtures for an additional 30 min at 4uC.After washing the beads three times in DNAP binding buffer, precipitated proteins were eluted in SDS-PAGE sample buffer.Samples were analyzed SDS-PAGE followed by Western blot analysis using anti-Sp1 (Santa Cruz Biotechnology), anti-Myc (Nacalai Tesque), anti-RARa (Santa Cruz Biotechnology), and anti-RXRa (Santa Cruz Biotechnology) Abs.
Transfection and luciferase assay COS-7 cells were maintained in complete DMEM medium.Cells were seeded into 12-or 6-well plates (5610 5 or 1610 6 cells/ well, respectively) and transfected with 1.25 mg of various reporter vectors, 0.5-1.5 mg of expression vectors, and 0.025 mg of pRL-TK (Promega) unless otherwise indicated using LipofectAmine 2000, according to the manufacturer's instructions.After 24 h, the cells were transferred into new 48-well plates, and then stimulated with 5 ng/ml PMA and/or 100 nM RA.Sixteen hours after stimulation, the cells were washed in PBS, and lysed in 1 6passive lysis buffer (Promega).The firefly and Renilla luciferase activities were measured by the dual luciferase assay system (Promega) in a luminometer (Turner TD-20/20), according to the manufacturer's instructions.All experiments were carried out in triplicates, and the firefly luciferase activity was normalized by the Renilla luciferase activity.

Chromatin Immunoprecipitation (ChIP) Assay
ChIP assay was performed as previously described [23].Briefly, aliquots of cultured BM-DCs (2610 6 ) were fixed with 1% formaldehyde at 37uC for 10 min.Cross-linking reactions were quenched with 150 mM glycine.Cells were washed, suspended in SDS lysis buffer, and sonicated to shear the chromatin into 200-500 bp fragments using a sonicator (Bioruptor UCW-201).After centrifugation to remove debris, aliquots were incubated with 5 mg anti-Sp1 or anti-RARa Ab or control IgG1 overnight, followed by incubation with protein G beads (Cell Signaling Technology) for 1 h at 4uC with rotation.The immunoprecipitates were sequentially washed with ChIP dilution buffer, high-salt buffer, LiCl buffer, and TE buffer.The DNA-protein complex was eluted in SDS lysis buffer, and was de-crosslinked.After proteins were digested with proteinase K, DNA was isolated and subjected to PCR analysis.An aliquot of chromatin that was not incubated with an antibody was used as the input DNA control.The binding of Sp1 or RARa to the Aldh1a2 promoter site was estimated by real-time PCR with Power SYBR Green PCR Master Mix and gene-specific primers (Table S1).The binding levels were expressed as % of input DNA and was calculated from DCT (DCT = input CT -ChIP CT) according to the following equation: % total = 2 DCT .

Analysis of DNA methylation with a bisulfite method
Genomic DNA was prepared using the DNA fast pure kit (TAKARA).DNA was denatured, modified with sodium bisulfite, purified, desulfonated using a MethylEasy Xceed kit (Human Genetic Signatures), according to the manufacturer's instructions, and applied to nested PCR for bisulfite sequencing.The primers for the first PCR and the nested PCR were shown in Table S1.PCR products were inserted into the pGEM-T Easy vector, and 10-20 clones from each sample were sequenced using Big Dye Terminator on an ABI 3130 sequencer.The methylation state for each CpG site in the amplicon sequences was analyzed by using a web-based freely available quantification tool for methylation analysis (QUMA; http://quma.cdb.riken.jp/top/index_j.html)[24].The sequences of the 59-flanking regions of the mouse Aldh1a2 gene, the human ALDH1A2 gene (GenBank TM accession number NT_010194.17), the cattle ALDH1A2 gene (NW_003104280.1), the rat Aldh1a2 gene (NW_003809872.1), the chicken ALDH1A2 gene (NW_003763854.1),and the zebrafish aldh1a2 gene (NW_003039680.2) were obtained from NCBI Genome Resources.

Methylation of vectors and oligonucleotides
The pCpGL vector containing a 0.53-kb fragment (2373 to + 182) of the Aldh1a2 promoter and the double-stranded oligonucleotides for DNAP assay were methylated with CpG methyltransferase (M.SssI) and S-adenosylmethionine (New England Biolabs), according to the manufacturer's instructions.Mock methylation reactions did not contain M.SssI.Methylated and mock methylated vectors and oligonucleotides were purified by phenol-chloroform extraction and ethanol precipitation [25].

Statistical analysis
Statistical comparisons were carried out by using the two-tailed unpaired Student's t-test.The p value , 0.05 was considered significant.

The transcription factor Sp1 participates in inducing Aldh1a2 expression
Mouse Aldh1a2 that encodes RALDH2 comprises 14 exons that are divided by 13 introns spanning more than 50 kb of genomic DNA, and it is localized on chromosome 9 [26,27] (Figure S1A).Several CpG islands were found in Aldh1a2 and its upstream region using NCBI MapViewer.We analyzed a 7,000-bp fragment containing TSS and the first exon with the CpG island searcher, and identified a 1,445-bp (2816 to +629) region as a CpG island (Figure S1B).The 59-flanking region of the first exon included typical mammalian promoter consensus elements, a TATA box (2 38), GC-rich regions, a sterol regulatory element-binding protein (SREBP)-binding region (2417), and NF-kB-binding sites (2647, 2622, and 2575) in accord with a previous report [27].In this CpG island, there were no typical RARE sites, although we found two putative RARE half-sites (275 and 249) near the TATA box (Figure 1A).Putative STAT-binding sites (22217, 22172, and 2 2153) were also found far upstream from these sites.
The transcription factor Sp1 is known to bind to GC-rich promoter regions [28].We found that mithramycin A, which inhibits Sp1 binding to GC-rich regions [29], inhibited GM-CSFinduced Aldh1a2 mRNA expression and RALDH2 protein expression in Flt3L-generated BM-DCs (Figure 1B).However, it markedly reduced BM-DC viability only in the absence of GM-CSF (data not shown).
To assess the effect of Sp1 on Aldh1a2 promoter activity, we searched for a suitable cell line to construct a reporter assay system.We first tested several DC-like and macrophage-like cell lines for their capability to express Aldh1a2 or their utility for ectopic gene expression.None of these cell lines allowed sufficient transfection or provided signals from the transfectants for our analysis.Finally, we chose irrelevant COS-7, a fibroblast-like cell line derived from African Green Monkey kidney cells, and transfected these cells with an Sp1 expression vector and a pGL3 reporter vector that included a 2.8-kb fragment (22,600 to +182) of the 59-flanking region of Aldh1a2 [pGL3-RALDH2 (2 2,600)] (Figure 1A).Aldh1a2 promoter activity was enhanced by the ectopic expression of Sp1 in a dose-dependent manner (Figure 1C and Figure S2).A 1,727-bp deletion (22,600 to 2872) in the 59-flanking region of this gene significantly enhanced Sp1induced Aldh1a2 promoter activity (Figure 1C).There may have been a negative regulatory region in this deleted fragment.Deleting the next 500 bp (2873 to 2372) slightly reduced Aldh1a2 promoter activity.Although further deletion significantly reduced this activity, a short construct that contained a 0.31-kb fragment (2154 to +156) still showed significant responsiveness to Sp1.These results suggest that the 59-flanking region up to 2373 is nearly sufficient for Sp1-dependent induction of Aldh1a2 promoter activity.

MAPK activation is required for GM-CSF-induced Aldh1a2 expression and Sp1 nuclear translocation
It is known that optimal Aldh1a2 expression in CD103 + MLN-DCs depends on either the MEK1/2-ERK pathway or the p38a MAPK pathway [17,30], and that GM-CSF induces the activation of MAPKs, including ERK and p38 MAPK [31].Accordingly, the GM-CSF-induced Aldh1a2 expression in BM-DCs was inhibited by the MEK1 inhibitor PD98059 and the p38 MAPK inhibitor SB203580 (Figure 2A).It is also known that ERK and p38 MAPK activation contributes to Sp1 nuclear translocation [32,33].Indeed, we found that Sp1 nuclear localization was enhanced by GM-CSF stimulation and was inhibited by treating BM-DCs with PD98059 and SB203580 (Figure 2B).These results suggest that ERK and p38 MAPK activation is essential for GM-CSF-induced Sp1 nuclear translocation and Aldh1a2 expression in BM-DCs.
To activate ERK and MAPK in our reporter assay system, we added PMA to a culture of COS-7 cells that were transfected with the reporter vector that contained the 0.53-kb fragment (2373 to +156) of the 59-flanking region of Aldh1a2 [pGL3-RALDH2 (2 373)] and an Sp1 expression vector.PMA treatment enhanced Sp1-induced Aldh1a2 promoter activity (Figure 2C).Deleting a 16bp (256 to 241) region, but not a 22-bp (299 to 278) or 17-bp (2139 to 2123) region, resulted in a dramatic reduction in Sp1induced reporter activity in the presence or absence of PMA.The results suggest that the short GC-rich region near the TATA box is essential for Sp1-and MAPK-dependent Aldh1a2 expression.To examine whether Sp1 directly bound to the GC-rich region, we performed a DNAP assay with biotinylated DNA probes, A (2 161 to 2119), B (2118 to 282), and C (281 to 240), which corresponded to the fragments of the 59-flanking region of Aldh1a2 (Figure 3A).Sp1 bound to all these probes, although it bound less efficiently to Probe A than to the other probes (Figure 3A).Because Probe C contained the essential region for Sp1-dependent Aldh1a2 expression, we used Probe C to assess the DNA-binding capacity of Sp1 in the nuclei of BM-DCs.
Stimulating Flt3L-generated BM-DCs with GM-CSF enhanced the Sp1 protein levels in the nuclei of these cells and enhanced their binding capacity to Probe C (Figure 3B).Taken together, these results suggest that MAPK-dependent Sp1 nuclear translocation and its binding to the short GC-rich region near the TATA box are involved in GM-CSF-induced Aldh1a2 expression in DCs.

RAR and RXR participate in inducing Aldh1a2 expression
Probe C contained the two putative RARE half-sites (275 and 249).To test whether RARa or RXRa could bind to these putative RARE half-sites, we performed DNAP assays using biotinylated wild-type Probe C and its mutant DNA probes, Probe C(RARE-h mt1) and Probe C(RARE-h mt2); these mutant probes contained a mutation in the upstream putative RARE half-site and a mutation in the downstream putative RARE half-site, respectively (Figure 4A).RXRa alone but not RARa alone could bind to Probe C and Probe C(RARE-h mt1).RARa and RXRa mutually enhanced their binding to Probe C and Probe C(RARE-h mt1).RARa, RXRa, or their combination could not bind to Probe C(RARE-h mt2) (Figure 4A).These results suggest that RARa and RXRa can bind to the RARE half-site located adjacent to the TATA box.
RAR-mediated signaling appears to be essential for Aldh1a2 expression in DCs.In vitamin A-deficient mice, Aldh1a2 expression and RALDH2 activity in MLN-DCs and LP-DCs are markedly reduced [11,[17][18][19].Accordingly, GM-CSF-induced Aldh1a2 expression in BM-DCs was markedly suppressed by the RAR pan-antagonist LE540 (Figure 4B), as previously reported [11].RA alone induced low Aldh1a2 expression levels in Flt3L-generated BM-DCs; however, it significantly enhanced GM-CSF-dependent Aldh1a2 expression in a dose-dependent manner (Figure 4C), in accordance with our previous observation [11].Without the addition of any additional RA to the culture medium, retinol or a minute amount of RA in fetal bovine serum in the culture medium may have weakly stimulated RAR.
To assess the involvement of RARa and RXRa in the RAdependent activation of Aldh1a2 promoter activity, we transfected COS-7 cells with RARa and RXRaexpression vectors and a pGL3 reporter vector that contained a 1-kb fragment (2873 to + 182) of the 59-flanking region of Aldh1a2 [pGL3-RALDH2 (2873)] (Figure 4D).Aldh1a2 promoter activity was induced by the ectopic expression of RARa and RXRa and was markedly enhanced by RA stimulation.The mutation (mt2) in the RARE half-site significantly reduced this RA-induced promoter activity.These results suggest that RA induces Aldh1a2 transcription through an RARa/RXRa complex bound to the RARE half-site adjacent between the critical Sp1-binding region and the TATA box.

Sp1 and RARa/RXRa cooperatively contribute to RAdependent Aldh1a2 expression
Sp1 can directly interact with RARa and RXRa [34,35].We examined whether Sp1 and RARa/RXRa mutually enhanced their binding to Probe C. As shown in Figure 5A, when Sp1 or the combination of RARa and RXRa was expressed in COS-7 cells, each of these could bind to this probe.However, when Sp1, RARa, and RXRa were expressed together, the binding of each  2873) reporter vector or that containing the mutated RARE half-site (mt2) with or without the 0.5 mg of expression vectors, pSG5-RARa and pSG5-RXRa.One day after transfection, cells were stimulated with or without 100 nM RA, and luciferase activities were measured.The relative promoter activities were calculated by arbitrarily defining the activity of pGL3-baic alone without RA as 1.Data in (B, C, and D) are presented as mean + SD (B and D) or mean 6 SD (C) of triplicate cultures.Statistical significance between two groups was determined by the Student's t test (*p,0.05,**p,0.01,***p,0.001;NS, not significant).Data are representative of at least three independent experiments.doi:10.1371/journal.pone.0096512.g004component was significantly enhanced.These results suggest that Sp1 and RARa/RXRa mutually enhance their binding to the Aldh1a2 promoter region.Accordingly, Sp1 and RARa/RXRa cooperatively enhanced Aldh1a2 promoter-reporter activity in the presence of RA (Figure 5B).A truncated form of Sp1 that consisted only of the DNA-binding factor could not enhance the reporter activity in the presence or absence of RARa/RXRa (Figure 5B), and it suppressed Sp1-induced promoter activity in a dose-dependent manner (Figure S3).Mithramycin A and the MAPK pathway inhibitors PD98059 and SB204580 inhibited Aldh1a2 expression that was induced by the combination of RA and GM-CSF (Figure S4A).These results suggest that Sp1 and RARa/RXRa cooperatively contribute to GM-CSF/RA-induced Aldh1a2 expression, depending on MAPK activation.
Accordingly, ChIP assay results indicated that both RA and GM-CSF enhanced the binding of both Sp1 and RARa to this promoter region in BM-DCs, although no additional enhancement was found after stimulation with the combination of GM-CSF and RA in this assay (Figure 5C).The results suggest that adding either GM-CSF alone or RA alone more or less induces both Sp1 activation and RARa activation.Accordingly, mithramycin A, PD98059, and SB204580 also partially suppressed RAinduced Aldh1a2 expression (Figure S4B).
Interestingly, the sequence of the RARE half-site adjacent to the TATA box of mouse Aldh1a2 was identical to those of different species, including rat, human, cattle, chicken, and zebrafish (Figure 6).Furthermore, the sequences around the RARE halfsites in these species were highly conserved.These data suggest that the RARE half-site and the GC-rich Sp1-binding region are important for Aldh1a2 expression.

CpG methylation in the Aldh1a2 promoter region inhibits Sp1-dependent Aldh1a2 promoter activation
Because DNA methylation of CpG islands within proximal promoters is often associated with transcriptional regulation, we examined whether CpG methylation in the promoter region affected Sp1-dependent Aldh1a2 transcription.The 59-flanking region (2373 to +156) of Aldh1a2 was inserted into a reporter vector that lacked CpG dinucleotides in its vector backbone (pCpGL) to avoid interference from methylation of the vector backbone.This reporter vector either remained unmethylated or was methylated in vitro using CpG methyltransferase (M.SssI).COS-7 cells were transfected with the methylated or unmethylated pCpGL-reporter vector together with the Sp1 expression vector (pCMV-Myc-Sp1) or an empty pCMV-Myc vector.Methylation of the Aldh1a2 promoter region dramatically inhibited Sp1dependent Aldh1a2 promoter activity (Figure 7A), but did not affect Sp1 binding to the DNA oligonucleotides Probe B and Probe C derived from the Aldh1a2 promoter region (Figure 7B).These results indicate that Aldh1a2 transcription silencing by CpG methylation of the promoter region is not due to direct interference with Sp1 binding to the promoter region.
Flt3L-generated BM-DCs can be subdivided into two major subsets on the basis of their surface phenotypes: cDCs and pDCs.cDCs have been classically defined as B220 -CD11c hi DCs, whereas pDCs have been defined as B220 + CD11c int DCs (Figure 7C).In MLNs, Aldh1a2 expression is found in cDCs; however, there is only minimal expression in pDCs [11,36].Accordingly, GM-CSF induced Aldh1a2 expression in BM-cDCs but weakly induced this expression in BM-pDCs (Figure 7D).
We next examined whether methylation of the CpG sites in the promoter may control the responsiveness for Aldh1a2 expression.A 395-bp region (2424 to 230) near TSS within the aforementioned 1,445-bp-long CpG island (Figure S1B) in BM-cDCs and BM- pDCs was assessed for the DNA methylation status by sodium bisulfite sequencing.We found that the CpG island was mostly unmethylated not only in cDCs but also in pDCs (Figure 7E).We also assessed the DNA methylation status of the 395-bp region in SPL-DCs cultured with or without GM-CSF.SPL-DCs express Aldh1a2 after GM-CSF stimulation [11].The CpG island was mostly unmethylated in SPL-DCs, with or without GM-CSF stimulation (Figure S5).Furthermore, the 395-bp region was also unmethylated in normal resident peritoneal macrophages and naı ¨ve CD4 + T cells (Figure S5), although these cells did not express Aldh1a2 after GM-CSF stimulation in vitro (data not shown).These results indicate that in these normal immune cells, Aldh1a2 expression is regulated by a DNA methylation-independent mechanism.
On the other hand, the 395-bp region in the murine macrophage-like cell line RAW264 and the murine DC-like cell line DC2.4 was hypermethylated (Figure S5).Both these cell lines did not express Aldh1a2 after GM-CSF stimulation; however, after incubation with the DNA methylation inhibitor 5-aza-29-deoxycytidine for 6 days, RAW264 cells began to express Aldh1a2 after stimulation with GM-CSF or LPS (data not shown).These results suggest that DNA methylation of the GC-rich regions in the Aldh1a2 promoter contributes to its unresponsiveness under certain circumstances.

Discussion
The results of the present study indicated that the binding of Sp1 and RARa/RXRa to a GC-rich region and an RARE halfsite, respectively, in a short 59-flanking region of the TATA box was critical for the GM-CSF/RA-dependent induction of Aldh1a2 expression.These short 59-flanking regions were well conserved among different species, which suggested the importance of this region for initiating Aldh1a2 transcription.DNAP and Aldh1a2 promoter reporter assays using COS-7 cells indicated that Sp1 and the RARa/RXRa complex cooperatively enhanced their binding to this promoter and the reporter activity.
However, a ChIP assay using BM-DCs indicated that Sp1 and RARa binding was induced after adding either GM-CSF alone or RA alone.Because we found that at least residual RAR-mediated signaling appeared to be essential for GM-CSF-induced Aldh1a2 expression in BM-DCs and that mithramycin A partially inhibited RA-induced Aldh1a2 expression, it is likely that adding GM-CSF or RA more or less induced the activation of both Sp1 and RAR/ RXR.Thus, even after adding GM-CSF or RA alone, Sp1 and RAR/RXR may also be able to cooperatively induce this promoter activation in BM-DCs.However, our ChIP assay showed that there was no additional coordinate effect when both GM-CSF and RA were added.Thus, it is possible that Sp1 and RARa bind to the Aldh1a2 promoter transiently or coordinately allow another transcription factor to bind to this promoter.Because b-catenin-mediated signaling is essential for LP-DCs to express Aldh1a1 and Aldh1a2 mRNA [16], b-catenin may contribute to deliver an additional signal for Aldh1a2 expression.Accordingly, GM-CSF has been suggested to stabilize the bcatenin protein, possibly by inactivating glycogen synthase kinase-3b during macrophage differentiation [37].Nonetheless, any additional GM-CSF-dependent signal remains to be investigated.
GM-CSF and RA are involved in RALDH2 expression not only in mouse DCs but also in human DCs [38,39].In small intestine tissues and MLNs, GM-CSF is produced by various types of cells, including epithelial cells, Paneth cells, macrophages, and T cells, while RA may be produced by some types of cells including RALDH1 + intestinal epithelial cells, RALDH2 + cDCs, and Aldh1a1 + Aldh1a2 + Aldh1a3 + MLN stromal cells [1,19,[40][41][42][43].Subsets of mature CD103 + cDCs in MLNs have high Aldh1a2 expression levels and RALDH2 activity under steady-state conditions.A major population of RA-producing CD103 + MLN-DCs appears to be derived from the small intestinal LP  [44], and most of CD103 + LP-DCs express weak RALDH2 activity [12].Thus, the early conditioning of these DCs for RALDH2 expression may occur in the small intestinal LP or just after their arrival at MLNs.RA or an RAR agonist alone induces only low Aldh1a2 expression levels in Flt3L-generated BM-DCs and SPL-DCs [11,17]; however, we found that RA markedly enhanced GM-CSF-induced Aldh1a2 expression in BM-DCs.Thus, RA is likely to contribute to this early conditioning.
It was reported that p38a deficiency or the oral administration of a MEK1/2-ERK inhibitor significantly inhibited Aldh1a2 expression or RALDH2 activity in MLN-DCs [17,30].We found that the activation of ERK and p38 MAPK signaling pathways was also essential for GM-CSF-induced Aldh1a2 expression and Sp1 translocation to the nuclei in BM-DCs.Immature DCs undergo a maturation process after stimulation with cytokines or TLR ligands via MAPK activation [45][46][47][48].Accordingly, DC maturation is required for RALDH2 expression [11].AU-rich elements (AREs) in the 39-untranslated regions of cytokine genes are mRNA destabilizing elements [49].p38 MAPK contributes to the expression of these cytokines by interfering with mRNA degradation through an ARE-targeted mechanism [50].We also identified several AREs in the 39-untranslated region of Aldh1a2 (data not shown).Thus, p38 MAPK may also contribute to GM-CSF-induced Aldh1a2 expression by stabilizing its mRNA through an AREs-dependent mechanism.
GM-CSF also activates the JAK2-STAT5 and STAT3 signaling pathways [31].We found putative STAT binding sites in the 59flanking region of Aldh1a2.However, the expression of a constitutively active form of STAT5 was insufficient to directly activate or enhance Aldh1a2 transcription (data not shown).Thus, STAT5 signaling may indirectly contribute to Aldh1a2 expression by inducing an early response gene.
GM-CSF also activates NF-kB [31].There are NF-kB-binding sites and a SREBP-binding site adjacent to the core promoter region of Aldh1a2, although we found that deleting these sites did not significantly reduce Sp1-dependent promoter activity.By upregulating SREBP-1c, LXR ligands induce Aldh1a1 and Aldh1a2 expression in some tissues and cell lines [51], but fail to induce Aldh1a2 expression in BM-DCs [11].
The GC-rich region in the Aldh1a2 promoter was mostly unmethylated not only in Flt3L-generated BM-cDCs and SPL-DCs but also in BM-pDCs, macrophages, and naı ¨ve CD4 + T cells that did not have significant Aldh1a2 expression.Thus, a DNA methylation-independent mechanism can also contribute to the regulation of Aldh1a2 expression.pDCs express lower levels of GM-CSF receptors than cDCs in mice [52].This may partly contribute to the differences in their Aldh1a2 expression.Sp1 acetylation may also contribute to the regulation of Aldh1a2 expression, as it has been reported that Sp1 acetylation is associated with loss of DNA binding at some promoters [53].Accordingly, we found that histone deacetylase inhibitors inhibited GM-CSF-induced Aldh1a2 expression in DCs (data not shown).
We revealed that Sp1-and RARa/RXRa-mediated signals were likely to be integrated through the Aldh1a2 promoter covered by an unmethylated CpG island.Analyses of epigenetic modifications, such as histone acetylation, and possible involvement of additional transcription factors may clarify the mechanism underlying the selective regulation of Aldh1a2 expression in specific DC subsets.

Figure 1 .
Figure 1.Sp1 participates in the Aldh1a2 expression.(A) The genomic organization of the mouse Aldh1a2 gene and its 59-flanking region is shown.A fragment containing exon 1 and its 59-flanking region from 22,600 to +182 was inserted into reporter vectors.DNA binding sites (STAT-binding sites, NF-kB binding sites, a SREBP binding site, and putative RARE half-sites (RARE-h)), the TATA box, and the GCrich region in the fragment are indicated.(B) Flt3L-generated BM-DCs were cultured with 10 ng/ml GM-CSF for 16 h in the presence or absence of 1 mM mithramycin A. After the culture, Aldh1a2 mRNA expression was assessed by real-time PCR (Left panel), and protein expression of RALDH2 (ALDH1A2) and a-tubulin was analyzed by Western blotting (Right panel).Relative mRNA expression levels were calculated by defining the Aldh1a2 mRNA expression level in the cells incubated with medium alone for 16 h was set to 1 (Left panel).Data are representative of three (Left panel) or two (Right panel) independent experiments.(C) Serial-deletion fragments derived from the 59-flanking region of the mouse Aldh1a2 gene were inserted in the reporter vector, pGL3-basic.COS-7 cells were transfected in triplicate with one of the deletion constructs (1.25 mg) or the pGL3-RALDH2 (22,600) reporter vector (1.25 mg) in combination with or without the 0.5 mg of pCMV-Myc-Sp1 expression vector.One day after the transfection, luciferase activity was measured.Relative promoter activities were calculated by arbitrarily defining the activity of pGL3-basic alone as 1.Data are presented as mean + SD of triplicate cultures.Statistical significance between two groups was determined by the Student's t test (**p,0.01,***p,0.001).Data are representative of three independent experiments.doi:10.1371/journal.pone.0096512.g001

Figure 2 .
Figure 2. The MEK1-ERK-signaling pathway and the p38 MAPKsignaling pathway are required for the GM-CSF-induced Aldh1a2 expression and nuclear translocation of Sp1 in BM-DCs.(A) Flt3L-generated BM-DCs were cultured with or without 10 ng/ ml GM-CSF for 16 h in the presence or absence of 50 mM PD98059 (PD) or 25 mM SB203580 (SB).After the culture, Aldh1a2 gene expression was assessed by real-time PCR.The Aldh1a2 mRNA expression level in the cells incubated with medium alone for 16 h was set to 1.The results are shown as the mean + SD of triplicate cultures.Statistical significance was determined by the Student's t test (***p,0.001).(B) BM-DCs were cultured with or without 10 ng/ml GM-CSF in the presence or absence of 50 mM PD98059 or 25 mM SB203580 for 16 h.After the culture, cytosolic and nuclear proteins were analyzed for the presence of Sp1, atubulin, and lamin B1 by Western blotting.(C) Structures of the deletion constructs used in this figure are shown.D1, D2, and D3 constructs were prepared by deleting the region spanning from the 256 to 241, 299 to 278, and 2139 to 2123, respectively, from the pGL3-RALDH2 (2373) reporter vector.COS-7 cells were transfected in triplicate with the 1.25 mg of pGL3-RALDH2 (2373) reporter vector or the deletion mutants in combination with 0.5 mg of the pCMV-Myc-Sp1 expression vector or control empty vector.One day after the transfection, cells

Figure 3 .
Figure 3. Sp1 binds to the Aldh1a2 promoter region.(A) The locations and nucleotide sequences corresponding to Probe A, Probe B, and Probe C in the 59-flanking region of the mouse Aldh1a2 gene are shown.COS-7 cells were transfected with the 0.5 mg of pCMV-Myc-Sp1 or control empty vector.One day after transfection, cell lysates were analyzed for DNA-binding activity by DNAP assay using the indicated biotinylated DNA probes and anti-Myc Ab. (B) Flt3L-generated BM-DCs were cultured in the presence or absence of 10 ng/ml GM-CSF.After 16 h, nuclear extracts were analyzed for the presence of lamin B1 and Sp1 by Western blotting using anti-lamin B1 and anti-Sp1 Abs (left panel), or assessed for DNA binding activity by DNAP assay using biotinylated DNA Probe C and anti-Sp1 Ab (right panel).Data are representative of at least three independent experiments.doi:10.1371/journal.pone.0096512.g003

Figure 4 .
Figure 4. RA enhances GM-CSF-induced Aldh1a2 expression via the RARa/RXRa heterodimer bound to the RARE half-site.(A) Localization of the putative RARE half-sites (underlined) in Probe C and their mutants, Probe C(RARE-h mt1) and Probe C(RARE-h mt2), are shown.COS-7 cells were transfected with the 0.5 mg of pSG5-RARa and/ or pSG5-RXRa.One day after transfection, cell lysates were subjected to DNAP assay using the biotinylated DNA probes.The precipitates were analyzed by Western blotting using anti-RARa(upper panel) and RXRa (lower panel) Abs.(B) Flt3L-generated BM-DCs were cultured with 10 ng/ml GM-CSF for 16 h in the presence or absence of 100 nM RA.LE540 (1 mM) was added to the indicated cultures.After the culture, Aldh1a2 mRNA expression was assessed by real-time PCR.The Aldh1a2 mRNA expression level in the cells incubated with medium alone for 16 h was set to 1. (C) BM-DCs were cultured with (closed circle) or without (open circle) 10 ng/ml GM-CSF for 16 h in the presence of graded concentrations of RA.After the culture, Aldh1a2 mRNA expression was assessed by real-time PCR.Relative expression levels were calculated by defining the Aldh1a2 mRNA expression level in the cells incubated with medium alone for 16 h was set to 1. Asterisks

Figure 5 .
Figure 5. Sp1 and RARa/RXRa enhance each other's binding to the Aldh1a2 promoter and cooperatively enhance its activity.(A) COS-7 cells were transfected with the 0.5 mg of pCMV-Myc-Sp1, the combination of pSG5-RARa and pSG5-RXRa, or the three.One day after transfection, cell lysates were subjected to DNAP assay using anti-Myc Ab, anti-RARa Ab, or anti-RXRa Ab, and biotinylated DNA Probe C whose sequence is shown in Figure 3. (B) COS-7 cells were transfected in triplicate with the 1.25 mg of pGL4-RALDH2 (2873) reporter vector and the 0.5 mg of expression vectors, pCMV-Myc-Sp1, pCMV-Myc-Sp1db, pSG5-RARa, and pSG5-RXRa, or control empty vectors.One day after transfection, cells were stimulated with or without 100 nM RA for 16 h.Then luciferase activities were measured.Relative promoter activities were calculated by arbitrarily defining the activity of pGL4-RALDH2 (2873) alone without RA as 1. (C) Flt3L-generated BM-DCs were cultured with or without 10 ng/ml GM-CSF or 10 nM RA.These cells were subjected to ChIP assay with anti-Sp1 or anti-RARa Ab or control IgG1.Binding of Sp1 and RARa proteins to the Aldh1a2 promoter site was estimated by real-time PCR.Data in (B and C) are presented as mean + SD of triplicate cultures.Statistical significance between two groups was determined by the Student's t test (*p,0.05,**p,0.01,***p,0.001).Data are representative of three independent experiments.doi:10.1371/journal.pone.0096512.g005

Figure 6 .
Figure 6.The short DNA regions containing the GC-rich region, the RARE half-site, and the TATA box in the 59-flanking region of the Aldh1a2 genes are well conserved among various species.The sequence of the 59-flanking region of the mouse (Mus musculus) Aldh1a2 gene was compared with that of the human (Homo sapiens) ALDH1A2 gene, that of rat (Rattus norvegicus) Aldh1a2 gene, that of cattle (Bos taurus) ALDH1A2 gene, that of chicken (Gallus gallus) ALDH1A2 gene, and that of zebrafish (Danio rerio) aldh1a2 gene.The sequence data were obtained using the NCBI MapViewer.The shaded regions indicate homology with the mouse sequence.The locations of the conserved RARE half-sites (RARE-h) and TATA boxes are indicated by boxes.doi:10.1371/journal.pone.0096512.g006

Figure 7 .
Figure 7. Methylation of the CpG island in the Aldh1a2 promoter prohibits Sp1 to activate the promoter, whereas the Aldh1a2 promoter is largely unmethylated in BM-pDCs as well as in BM-cDCs.(A) pCpGL-basic and pCpGL-RALDH2 (2873) reporter vectors were methylated with 1.25 mg of M.SssI.COS-7 cells were transfected with methylated or unmethylated pCpGL-basic or pCpGL-RALDH2 (2873) reporter vector in combination with or without the 0.5 mg of pCMV-Myc-Sp1 expression vector.One day after transfection, luciferase activity was measured.Relative promoter activities were calculated by arbitrarily defining the activity of pCpGLbasic alone as 1. (B) COS-7 cells were transfected with pCMV-Myc-Sp1.One day after transfection, cell lysates were analyzed for DNA binding activity by DNAP assay using DNA Probe B and Probe C methylated with M.SssI or left unmethylated.The bound proteins were analyzed by SDS-PAGE followed by Western blotting with anti-Myc Ab. (C) Flt3Lgenerated BM-DCs were stained with allophycocyanin-labeled anti-CD11c Ab and phycoerythrin-labeled anti-B220 Ab, and were sorted to cDC and pDC fractions with a FACSAria.(D) Sorted BM-pDCs and BM-cDCs were cultured for 16 h with or without 10 ng/ml GM-CSF.Expression of Aldh1a2 mRNA was analyzed by real-time PCR.Relative expression levels were calculated by defining the Aldh1a2 mRNA expression in the cells incubated with medium alone for 16 h was set to 1. Data in (A and D) are presented as mean + SD of triplicate cultures.Statistical significance between two groups was determined by the Student's t test (***p,0.001).Data in (A, B, and D) are representative of at least three independent experiments.(E) Genomic DNA was isolated from BM-pDCs and BM-cDCs, denatured, modified with sodium bisulfite, and used in nested PCR (2424 to 230) for bisulfite