RDH10, RALDH2, and CRABP2 are required components of PPARγ-directed ATRA synthesis and signaling in human dendritic cells.

All-trans retinoic acid (ATRA) has a key role in dendritic cells (DCs) and affects T cell subtype specification and gut homing. However, the identity of the permissive cell types and the required steps of conversion of vitamin A to biologically active ATRA bringing about retinoic acid receptor-regulated signaling remains elusive. Here we present that only a subset of murine and human DCs express the necessary enzymes, including RDH10, RALDH2, and transporter cellular retinoic acid binding protein (CRABP)2, to produce ATRA and efficient signaling. These permissive cell types include CD103+ DCs, granulocyte-macrophage colony-stimulating factor, and interleukin-4-treated bone marrow-derived murine DCs and human monocyte-derived DCs (mo-DCs). Importantly, in addition to RDH10 and RALDH2, CRABP2 also appears to be regulated by the fatty acid-sensing nuclear receptor peroxisome proliferator-activated receptor γ (PPARγ) and colocalize in human gut-associated lymphoid tissue DCs. In our model of human mo-DCs, all three proteins (RDH10, RALDH2, and CRABP2) appeared to be required for ATRA production induced by activation of PPARγ and therefore form a linear pathway. This now functionally validated PPARγ-regulated ATRA producing and signaling axis equips the cells with the capacity to convert precursors to active retinoids in response to receptor-activating fatty acids and is potentially amenable to intervention in diseases involving or affecting mucosal immunity.

surface protein expression and triggers intense iNKT expansion ( 24 ). However, the requirement of any of these enzymes for ATRA synthesis in DCs has not been proven yet.
The work presented here has been based on our previous results and it is a direct extension of it ( 24 ). Here we decided to comprehensively characterize the key components that are required for ATRA synthesis and transport in human mo-DCs using genetic as well as pharmacological tools to gain insights into their functional requirements. We have identifi ed CRABP2 as a PPAR ␥ -regulated transcript and characterized the functional contribution of RDH10, RALDH2, and CRABP2 to the enhanced transcriptional activation of ATRA target genes and also to the lipid antigen presentation capacity of mo-DCs. Using a small interfering RNA (siRNA) strategy targeting RDH10 , RALDH2 , or CRABP2 , we showed that attenuated expression of each of these proteins downmodulated the PPAR ␥ -induced CD1D and transglutaminase 2 ( TGM2 ) expression. Furthermore, down-modulated CD1d protein level reduced ␣ -GalCer-activated iNKT expansion in DC-T cell coculture experiments.
Furthermore, an immunohistochemistry (IHC) survey of human tissues has revealed that DCs in gut-associated lymphoid tissue (GALT) coexpress PPAR ␥ with RDH10, RALDH2, and CRABP2 in vivo.
These data collectively suggest that RDH10 and RALDH2 along with CRABP2 represent a linear pathway, and these genes are functionally required for PPAR ␥ -induced ATRA production and signaling in DCs ex vivo, and likely to play a similar role in vivo as well.

Generation of bone marrow-derived DCs
Bone marrow (BM) cells were isolated from the femur of C57BL/6 mice. Animals were housed under specifi c pathogen free conditions and the experiments were carried out under Committee of Animal Research of the University of Debrecen institutional ethical guidelines and licenses (license number: 21/2011/ DEMÁB). BM cells were differentiated to bone marrow-derived DCs (BM-DCs) in RPMI 1640 culturing medium supplemented with 10% FBS (Invitrogen), 500 U/ml penicillin/streptomycin (Invitrogen), 2 nM L -glutamine (Invitrogen), 20 ng/ml granulocytemacrophage colony-stimulating factor (GM-CSF), and 20 ng/ml interleukin-4 (IL-4) or 20 ng/ml GM-CSF alone for 9 days. Cytokine treatment was repeated at days 3 and 6. After a 9 day culturing period cells were harvested in Trizol reagent (Invitrogen) for RNA isolation .

Splenic and MLN DC separation
Pooled spleens and MLNs of male C57BL/6 mice were cut into small fragments and digested with collagenase D (Roche) for 40 min at 37°C. Solutions were fi ltered through a nylon mesh DCs have been reported to be important sentinels that play fundamental roles by capturing, processing, and presenting antigens to naive T cells in draining lymph nodes, which elicit antigen-specifi c immune responses. Under steady-state conditions, the CX 3 CR1 Ϫ /CD103 + subset continuously migrates to mesenteric lymph nodes (MLNs) and these DCs are involved in maintaining gut tolerance and homeostasis ( 1,2 ). They promote conversion of Foxp3 + regulatory T cells, induce gut-homing receptors CCR9 and ␣ 4 ␤ 7 expression on T and B cells, and provoke T cellindependent IgA switch in naive B cells (3)(4)(5)(6). These intestinal immune responses are established by transforming growth factor ␤ along with ATRA, a key cofactor for these processes (7)(8)(9)(10). However, the mechanism of ATRA generation in DCs is not fully understood. Retinol, acquired from the diet, can be metabolized to retinal by either the members of the medium-chain dehydrogenase/reductase (MDR) superfamily or by retinol dehydrogenases from the short-chain dehydrogenase/reductase (SDR) superfamily ( 11 ). The role of RDH10 for embryonic ATRA synthesis was identifi ed by N-ethyl-N-nitrosourea-induced forward genetic screen. These trex mutant/ Rdh10 -defi cient animals have organ abnormalities and the missense point mutation causes an embryonic lethal phenotype at embryonic day 13.5 ( 12,13 ).
In the second irreversible oxidative step, retinal is further converted to ATRA, catalyzed by RALDH1/ALDH1A1, RALDH2/ALDH1A2, or RALDH3/ALDH1A3 ( 6,7,14 ). Genetic deletion experiments have proved the physiological contribution of RALDH isoenzymes to ATRA production ( 15 ). Raldh2 Ϫ / Ϫ mice show early lethality, suggesting that this enzyme plays indispensible roles in ATRA production in embryos (16)(17)(18). Interestingly, these studies also revealed that the sites of expression of RDH10 overlap with those of RALDH2, which suggested that coexpression of both metabolic enzymes were required for ATRA generation. Finally, ATRA binds to the cellular retinoic acid binding proteins (CRABPs), which transport ATRA to the nucleus allowing the activation of retinoic acid receptors (RARs). Upon ligand activation, RARs form heterodimers with retinoid X receptors (RXRs), bind to the retinoic acid response element (RARE), and regulate the transcription of target genes (19)(20)(21).
In addition, the activation of another nuclear hormone receptor, peroxisome proliferator-activated receptor ␥ (PPAR ␥ ), can also modulate the immunophenotype of DCs ( 22 ). DCs treated with the PPAR ␥ -specifi c ligand rosiglitazone (RSG) have enhanced lipid antigen presentation capacity ( 23,24 ). DCs have the ability to present lipid antigens such as ␣ -galactosylceramide ( ␣ -GalCer) via the cell surface-expressed CD1d to stimulate invariant natural-killer T cell (iNKT) cells that represent a specifi c T cell population with expressed Natural Killer (NK) cell lineage receptors and semi-invariant CD1d-restricted V ␣ 24J ␣ Q T cell receptor-␣ chain paired with V ␤ 11 in humans ( 25,26 ). We have previously shown that PPAR ␥ activation in monocytederived DCs (mo-DCs) primes them for ATRA production by the upregulation of RDH10 and RALDH2 expression. The endogenously produced ATRA then induces CD1d cell

Aldefl uor assay
Aldehyde dehydrogenase activity of mo-DC was assessed using an ALDEFLUOR kit (StemCell Technologies). Cells were incubated at the density of 1 × 10 6 cells/ml in ALDEFLUOR assay buffer containing activated ALDEFLUOR substrate with or without DEAB for 40 min at 37°C. ALDEFLUOR reactive cells were monitored in FL1 channel of FACSCalibur compared with DEABtreated control samples.

RT-qPCR
Total RNA was isolated from cells using Trizol reagent (Invitrogen). One hundred nanograms of total RNA were reverse transcribed with Superscript reverse transcriptase (Invitrogen) and random primers (Invitrogen). This was performed at 42°C for 2 h. Quantitative PCR was performed on LC480 platform (Roche), 40 cycles of 95°C for 10 s and 60°C for 30 s. Gene expression was quantifi ed by the comparative threshold cycle method and normalized to human or mouse Cyclophilin A ( PPIA and Ppia ) expression as housekeeping gene. All PCR reactions were performed in triplicate. Values are expressed as means ± SD.

Western blot analysis
Twenty micrograms of protein, whole cell lysate, was separated by 12.5% polyacrylamide gel and transferred to PVDF membranes (Millipore). Membranes were probed with anti-CRABP2 (208) antibody, kindly provided by Cecile R. Egly (IGBMC, INSERM, France), and then the membranes were stripped and reprobed with anti-GAPDH antibody (ab8245-100; Abcam) according to the manufacturer's recommendations.

Immunoperoxidase staining
For IHC, monocytes, vehicle-treated DCs, or RSG-treated DCs (6 × 10 6 cells/group) were pelleted and fi xed in 4% paraformaldehyde (pH 7.3) for 24 h at 4°C. Cell blocks were then embedded in paraffi n, followed by serial sectionings (4 m). After deparaffi nization and dehydration, sections from each cell group were mounted on the same glass slides and were used for peroxidase-based indirect IHC. In brief, sections were treated with 3% H 2 O 2 in methanol for 15 min at room temperature to block the endogenous peroxidase. For antigen unmasking, sections were heated in antigen-retrieving citrate buffer (pH 6.0, Dako) for 2 min at 120°C using a pressure cooker. Immunostaining of cells for CRABP2 was carried out using the standard ABC technique utilizing the primary antibody-specifi c biotinylated secondary antibodies (Vectastain kits, Vector Laboratories). After blocking the nonspecifi c binding sites, sections were incubated with the primary rabbit antibody to CRABP2 (208) at dilutions of 1 × 1/50 for 1 h at room temperature prior to use of the biotinylated secondary antibodies . The peroxidase-mediated color development was set up for 5 min using the VIP substrate (Vector Labs). Finally, the sections were counterstained with methyl green. and washed. Cell suspension was preincubated for 10 min at 4°C with anti-mouse CD16/CD32 mouse BD Fc Block antibody (BD Pharmingen). CD11c + cells were obtained followed by anti-CD11c MACS bead (Miltenyi Biotec) separation. CD103 + and CD103 Ϫ DCs were separated by labeling the cells with anti-CD11c-APC and anti-CD103-PE (BD Pharmingen) antibodies and subsequent sorting on FACSVantage (BD Bioscience). Cells were harvested in Trizol reagent (Invitrogen).

DC/Splenocyte coculture experiment
Pooled MLN CD103 + DCs were obtained as described above. We purifi ed splenocytes from pooled spleen of BALB/c mice. Spleens were placed in a Petri dish containing RPMI 1640 medium supplemented with 10% FBS (Invitrogen). Cells were squeezed out with a glass plunger. After washing, we applied Lysing buffer (BD Pharm Lyse, BD Biosciences) against red blood cells. The cell suspension was plated in Petri dishes for 12 h to attach splenic DCs (Sp-DCs). We set the coculture experiment in 12-well plates, DC/splenocyte ratio was 1:20, corresponding to 1:10 DC:T cell ratio in 2 ml per well. After 72 h incubation at 37°C, MLN CD103 + DCs were separated by labeling cells with anti-CD11c-APC and anti-CD103-PE (BD Pharmingen) antibodies and subsequent sorting on FACSVantage (BD Bioscience). Cells were harvested in Trizol reagent (Invitrogen).

Human mo-DC culture
Human monocytes (98% CD14 + ) were isolated from buffy coats of healthy volunteers, obtained with the Regional Ethical Board permit from the Regional Blood Bank, by Ficoll gradient centrifugation, followed by magnetic bead separation using anti-CD14-conjugated microbeads (Miltenyi Biotec). Monocytes were differentiated to DCs at the density of 1.5 × 10 6 cells/ml in RPMI 1640 medium supplemented with 10% FBS (Invitrogen), 500 U/ml penicillin/streptomycin (Invitrogen), 2 nM L -glutamine (Invitrogen), 800 U/ml GM-CSF (Gentaur Ltd.), and 500 U/ml IL-4 (Peprotech). Cells were cultured for 5 days. Ligands or vehicle control were added to the cell culture at day 0 and at day 3.
All siRNAs were effi cient. The average siRDH10 effi ciency was 48.58 ± 8.44%, in the case of siRALDH2 the effi ciency was 39.22 ± 10.81%, and the average siCRABP2 effi ciency was 44.22 ± 9.25%. Next, we also analyzed other components of the downstream retinoid signaling pathway: we evaluated the gene expression of Rar and Rxr isoforms (supplementary Fig. IIA). RAR/RXR heterodimers regulate retinoid target genes such as cytochrome p450 26a1 ( Cyp26a1) , which had a similar transcription pattern to the Raldh2 gene ( 29 ). Cyp26a1 expression itself is regulated by ATRA through two identifi ed RAREs in the promoter of the gene, suggesting a negative feedback mechanism to control the retinoic acid concentration and active retinoid signaling in cells ( 30,31 ). We further analyzed retinoid signaling by measuring the transcript levels of Tgm2 and Cd1d1 , two well-known ATRA target genes ( 24,32 ). CD103 + and CD103 Ϫ MLN DC populations both expressed Tgm2 and Cd1d1 , but the normalized mRNA levels of the genes did not correlate with either Raldh2 expression or the ATRA production capacity of the cells ( Fig. 1C ).
In the search for a more suitable mouse DC model for mechanistic studies, we examined ex vivo differentiated cells using TaqMan low density arrays (TLDAs) in additional gene expression analyses. We differentiated GM-CSF-DC or GM-CSF+IL-4-DC from BM ( Fig. 1B ). To generate a negative control DC population that has no capacity for ATRA production ( 14 ), we isolated Sp-DCs from the same mouse strain (C57BL/6). We validated our BM-DC differentiation method by analyzing CD11c DC and F4/80 macrophage marker surface expression and CD11c on Sp-DCs by fl ow cytometry (supplementary Fig. IB, C). GM-CSF could elicit Raldh2 transcription, and the synergistic effect of the two cytokines was confi rmed, while Raldh2 expression in Sp-DCs was barely detectable as it was reported in case of BM-DCs ( 14 ). Next we focused on Rdh10 in in vivoand ex vivo-generated cells and found that all DCs expressed this gene ( Fig. 1D ). A similar expression pattern was observed regarding Cyp26a1 .
We compared the gene expression of all Rar ( a , b , and g ) and Rxr ( a or b ) expressed in mouse DCs (supplementary Fig. IIB). We also analyzed the gene expression of Tgm2 and Cd1d1 in these DC populations. All DC subsets expressed Cd1d1 , but it did not show any correlation with retinoid signaling. Tgm2 showed a similar gene expression pattern to Raldh2 , indicating that it could be a reliable marker of active retinoid signaling.
Next we assessed the expression of genes involved in retinol uptake and transport in DCs. However, the only known high-affi nity receptor for retinol uptake is STRA6 (stimulated by retinoic acid-6) ( 33 ) ( 34 ). We could not detect Stra6 gene expression in DC subsets derived from mice (data not shown). The structure of the recently identifi ed RBPR2 is related to human and murine STRA6 ( 35 ). Therefore, we hypothesized that RBPR2 could be an alternative receptor in murine and human DCs for retinol uptake. We detected the expression of the Rbpr2 gene in all DC subsets indicating the possibility of retinol uptake to DCs via this alternative receptor ( Fig. 1E ).
Unexpectedly, the transcription of the Crabp2 gene was not detectable in in vivo subsets ( Fig. 1E ). The detection of Crabp2 was successful in the case of GM-CSF or GM-CSF+IL-4 DCs in Fig. 1E ; therefore, we used the measured Crabp2

Double immunofl uorescence
Double immunofl uorescence (DI) was performed on formalin-fi xed paraffi n embedded intestinal tissue sections obtained from the archives of surgical specimens of the Department of Pathology, University of Debrecen as described earlier ( 28 ). Briefl y, following antigen-retrieving and peroxidase block (see previous section), the fi rst primary antibody was visualized with antibody-matched peroxidase-conjugated IgG followed by a tetramethyl-rhodamine-tagged tyramide (Perkin-Elmer) treatment (red fl uorescence). After washing and blocking the nonspecifi c binding sites, sections were incubated with the second primary antibody which was then developed with the use of matched biotinylated secondary antibody (IgG[Fab] 2 ) and streptavidin-FITC (Vector Labs, fl uorescent isothiocyanate) resulting in green fl uor escence. After thorough washings, nuclear counterstaining was made with 4 ′ ,6-diamidino-2-phenylindole (DAPI) containing the mounting medium (Vector Labs). To check the staining specifi cities, positive and negative controls were included for each immuno fl uorescent (IF) reaction as described earlier ( 28 ) and as indicated in the Results section.

ATRA biosynthesis in mouse intestinal DCs
As the fi rst step in our studies, we have characterized the expression of the essential enzymes ( Fig. 1A ) of ATRA biosynthesis and signaling in cells derived from mouse in vivo or ex vivo models. Retinol is transported by retinol-binding protein (RBP), taken up by DCs via retinol binding protein receptor(s) [STRA6 and RBP4 receptor-2 (RBPR2)], and bound to the cellular retinol-binding protein (CRBP) and then serves as substrate for the members of MDR or the SDR superfamily and becomes converted to retinal ( Fig. 1A ). The second enzymatic step is catalyzed either by RALDH1 or RALDH2. The active metabolite is delivered to the nucleus by CRABP2 and induces the transcription of several retinoid target genes through activating the RAR/RXR heterodimers in mo-DCs. Despite extensive investigation, the components and the exact molecular regulation of this pathway are not completely characterized in either murine or human DCs.
We hypothesized that RDH10 might be the primary enzyme that initiates retinal generation in intestinal DCs and the coexpression of RDH10 and RALDH2 determines the ability of ATRA production in mucosal DCs.
To test this hypothesis, we set out to assess the steps required for ATRA production and signaling in different in vivo-and in vitro-generated DC subtypes ( Fig. 1B ). First we generated MLN DCs described in the Materials and Methods. The purity of the sorted population was assessed by postsort fl ow cytometric analysis (supplementary Fig. IA). We measured the expression level of genes involved in ATRA synthesis and compared the gene expression levels of mRNAs between sorted populations by RT-qPCR. As expected, we could detect Raldh2 only in CD103 + DCs. Rdh10 was expressed in both populations of the MLN DCs, but at an even higher level in the CD103 Ϫ cells ( Fig. 1C ).
Next, we examined the gene expression profi le of a group of select genes involved specifi cally in ATRA biosynthesis and signaling ( Fig. 2B ). RDH10 , RDH11 , and DHRS9 are expressed in mo-DCs ( 24 ). Both RALDH1/ALDH1A1 and RALDH2/ALDH1A2 are expressed at high levels in mo-DCs, while among the in vivo-generated cell types a moderate level of transcription was observed in dermal DCs only, suggesting the possibility of ATRA synthesis in that DC subtype. RALDH3/ALDH1A3 was not expressed in these DC subsets. Among the genes encoding ATRA-transporting proteins, CRABP1 was not expressed, while CRABP2 was expressed ubiquitously. We also analyzed the retinoid signaling by measuring the expression of target genes: TGM2 and CD1D were expressed in tonsillar CD1c + , blood CD1c + , pDC, and in mo-DCs.
Following these analyses we focused on PPARG and FABP4 expression, the latter is a known marker gene of activated PPAR ␥ signaling ( 45 ). We have shown earlier that PPAR ␥ enhances ATRA production by inducing RDH10 and RALDH2 gene expression in mo-DCs differentiated in the presence of the synthetic PPAR ␥ ligand RSG ( 24 ). In line with this observation, our microarray data showed that the PPARG gene is expressed in mo-DCs. The detectable level of FABP4 is likely to indicate either the presence of exogenous PPAR ␥ ligand in the serum or the presence of possible endogenous activators inside cells. Other in vivo-derived DC types failed to express PPARG or FABP4 . This systematic analysis suggested that retinoid signaling is only active in mo-DCs that coexpress RDH10 and RALDH2 , and these mo-DCs have PPAR ␥ ligand activation connected to the retinoid signaling pathway.
Next we validated the transcriptional changes of the genes involved in ATRA production during the full differentiation period by RT-qPCR. For this, we differentiated monocytes in the presence of RSG, RSG and GW9662 (a PPAR ␥ antagonist), or GW9662 alone to estimate possible roles of PPAR ␥ in gene expression. We quantifi ed the normalized mRNA level of RALDH genes at an early mRNA level in the GM-CSF+IL-4 DC sample, presented in Fig. 1D , as a technical control of the qPCR assay (labeled with Experiment II/black square in Fig. 1E ; the qPCR results of the CD103 + and CD103 Ϫ MLN-DC samples are labeled with Experiment I/white square). We assessed the role of cellular interactions (T cells) using allogenic splenocytes on Crabp2 and Rbpr2 gene expression coculturing with ex vivo DCs. In the CD103 + /splenocyte coculture experiment, the expression of both Crabp2 and Rbpr2 was induced in CD103 + DCs, suggesting an enhanced retinol uptake and ATRA delivery as a result of cellular, most likely T cell, interactions ( Fig. 1F ). Summarizing these data, three out of fi ve DC types (CD103 + MNL DCs, and GM-CSF-DCs or GM-CSF+IL-4-DCs) have a gene expression signature consistent with active ATRA biosynthesis in line with earlier published data ( 14 ). Cyp26a1 expression with Raldh2 appears to indicate active retinoid signaling CD103 + MNL DCs, and GM-CSF-DCs or GM-CSF+IL-4-DCs. We found that all DCs express the Rdh10 gene. Based on these gene expression results, we concluded that ATRA biosynthesis is not a universal feature of DCs, and that, in line with our hypothesis, Rdh10 expression overlaps with Raldh2 expression suggesting that DCs expressing both enzymes are likely to have active ATRA synthesis and signaling.

Characterization of retinoid signaling in human DCs
DCs also exist in the human small intestinal MLNs. Moreover, these DCs have similar functional properties compared to CD103 + MLN DCs in mice ( 36 ). These data indicated that specifi c DCs with the ability of de novo ATRA synthesis can be present in the human body. Despite much effort and previous work ( 37 ), the human DC phenotypes are not identical and easy to match up with the ATRAproducing murine DCs. Therefore we considered using human mo-DCs for mechanistic characterization of the components of retinoid signaling by functional assays. To prove that these ex vivo cells faithfully replicate the behavior of human in vivo DCs with typical DC morphology and characteristic surface markers ( 23,(38)(39)(40), we examined a microarray data set and compared the gene expression pattern of mo-DCs, Langerhans cells, dermal DCs from the skin, CD1c + DCs from the tonsil, and CD1c + and plasmacytoid DCs (pDCs) from healthy donors (41)(42)(43)(44). Our analysis supported the notion that mo-DCs expressed common DC markers as CD83 , CD1A , CD14 , CD86 , CD209 , induced by PPAR ␥ in RSG-treated mo-DCs, we further examined and characterized its regulation ( Fig. 2D ). CRABP2 delivers ATRA to the nucleus, thus enhanced expression of CRABP2 should increase the transcriptional activity of RAR ( 46,47 ). Simplistically, CRABP2 would act as a coactivator-like molecule; when it is present, retinoid signaling is more effi cient . As shown in Fig. 3A , PPAR ␥ activation profoundly induced the transcript levels of CRABP2 . Next we examined if the gene expression changes are manifested at the protein level in mo-DCs. We found that monocytes do not express CRABP2, while control-treated mo-DCs have a detectable level of the protein ( Fig. 3B ). RSG-treated mo-DCs have a highly enhanced CRABP2 protein level compared with control cells. This appears to be a DC-specifi c regulation, because monocyte-derived macrophages (M s) failed to express any CRABP2 ( Fig. 3B ). We postulated that the elevated CRABP2 expression in PPAR ␥ -instructed DCs might contribute to the enhanced ATRA response. By immunocytochemistry, we further confi rmed the elevated CRABP2 expression at the expression site of the delivery protein within mo-DCs. We observed elevated CRABP2 protein expression in DCs as compared with monocytes and a strong upregulation of CRABP2 upon RSG-treatment ( Fig. 3C ). Based on our immunocytochemistry and Western blot results, we concluded that PPAR ␥ -activated mo-DCs represent a relevant ex vivo model system that appears to be suitable to mechanistically dissect the ATRA biosynthesis and signaling pathway composed of RDH10, RALDH2, and CRABP2 proteins that are coordinately upregulated by PPAR ␥ . We also realized that further investigations are needed to provide direct evidence for CRABP-mediated ATRA transport to the nucleus in these cells.
PPAR ␥ , RDH10, RALDH2, CRABP2, and the ATRAregulated TGM2 colocalize in DCs of human GALT In order to provide in vivo relevance to our fi ndings, we systematically surveyed the expression of the components of ATRA biosynthesis and signaling in human tissues. We tested the expression of PPAR ␥ , RDH10, RALDH2, CRABP2, and TGM2 in resting human GALT with no associated intestinal infl ammation using in situ IF staining/DI. We chose GALT (for hematoxylin eosin section see Fig. 4B ), as this is the most likely place where lipid signaling could contribute to DC differentiation and subtype specifi cation in the gut . As shown in Fig. 4A (representing an IF image), we observed that PPAR ␥ could be readily detected in white adipose tissue (WAT; positive control to ensure the specifi city of the antibody during IF staining). DI of resting GALT for PPAR ␥ (red) demonstrates that PPAR ␥ is in part timepoint (6 h), and as indicated at later time points (24,72, and 120 h ) ( Fig. 2C ). In monocytes we detected a high level of RALDH1 that was rapidly downregulated and expressed at elevated level again on the fi fth day of differentiation, indicating that RALDH1 may be involved in retinal oxidation in fully differentiated cells. Unlike this gene, the human RALDH2 was barely measurable at the monocyte state. After 6 h, cells expressed RALDH2 at comparable levels to RALDH1 . At all later time points, RALDH2 was upregulated in the RSG-treated samples and, except for day 5, ligand treatment induced RALDH2 transcription at similar levels as compared with RALDH1 , suggesting that RALDH2 not only has a dominant role in developing DCs but also acts as a metabolizing enzyme in differentiated cells. The RALDH3 isoform was not detectable irrespective of treatments or time points. As expected, ligand treatment induced the normalized mRNA level of RDH10 after 6 h, indicating that PPAR ␥ activates this gene probably via direct molecular interactions. The expression of RDH10 in RSG-treated samples continuously increased during the entire differentiation period. We detected a similar expression pattern of FABP4 . Interestingly, both CRABP2 and TGM2 genes were upregulated in RSG-treated DCs after 24 h in accordance with earlier results ( 24 ), CD1D was expressed at a high level in monocytes, but rapidly decreased in cultured cells. Consistent with our previous results, increased CD1D transcription was observed at later time points ( 23 ) in RSG-treated samples. PPARG was immediately induced in differentiating cells, the highest expression was detected at 6 h, and the normalized mRNA level of the gene was detectable at a somewhat lower level in DCs ( Fig. 2D ).
In summary, ATRA production and signaling is not a universal feature of DCs and it appears to be tightly regulated. We found evidence that ex vivo-differentiated mo-DCs express all components required for retinol to ATRA conversion and delivery, suggesting that mo-DCs have the ability for de novo ATRA synthesis and signaling. This ATRA producing capacity can be induced by the coordinate upregulation of RDH10 , RALDH2 , and CRABP2 .

Transport of ATRA via CRABP2 to the nucleus is PPAR ␥ regulated
As a result of PPAR ␥ activation, we could detect elevated RDH10 and RALDH2 in the cytoplasm of mo-DCs, suggesting that these cells have an enhanced ability to produce retinoic acid ( 24 ).
In this part of the study, we aimed to examine whether intracellular ATRA transport could also be regulated by PPAR ␥ . Due to the fact that CRABP2 was also highly Next, we assessed RALDH activity in mo-DCs electroporated at monocyte stage using specifi c siRNAs against siRDH10, siRALDH2, and siCRABP2 and NS-scrambled control siRNA. We treated cells with DMSO/ethanol (C), RSG, and RSG+GW9662. At day fi ve ALDEFLOUR activity was quantifi ed. We measured lower RALDH activity only in the siRALDH2 electroporated sample (supplementary Fig. IV).
These results suggest that RALDHs are active in mo-DCs, and the enhanced ATRA production capacity of mo-DCs is PPAR ␥ dependent. Interestingly, we could detect heterogeneity in this respect in the RSG-mo-DC population, but the generated endogenous ATRA level is at the range of RAR activation.

PPAR ␥ activation induces RAR signaling/gene expression via RDH10, RALDH2, and CRABP2
Based on these data, one can hypothesize that RDH10, RALDH2, and CRABP2 might be required for PPAR ␥regulated ATRA synthesis and gene expression. Despite the murine DC results ( 37 ), the model one can test this in is the human mo-DC. PPAR ␥ activation leads to transcriptional activation of several RAR target genes in human mo-DCs ( 24 ). Pharmacological analysis revealed that administration of the RALDH inhibitor DEAB reduced gene expression of CD1D and TGM2 upon RSG treatment, suggesting the importance of RALDH2 in PPAR ␥ -enhanced retinoid signaling ( 24 ). We have extended our studies by testing to determine whether the oxidizing enzymes and CRABP2 are indeed mechanistically indispensable for retinoid-regulated gene expression induced by PPAR ␥ . For this, we decided to use a siRNA-based approach. We delivered siRNA against RDH10 , RALDH2 , or CRABP2 and FABP4 (as a control) to CD14 + monocytes after cell separation via electroporation, and differentiated cells as described in the Materials and Methods. RSG was administered as indicated in Fig. 6 . After 24 or 48 h of RSG treatment, we assessed the transcript level of CD1D and TGM2 using RT-qPCR. We observed that PPAR ␥ -induced CD1D expression was down-modulated by all except FABP4 -specifi c siRNA at both indicated time points (24 and 48 h). TGM2 expression changes were similar at 24 h, but only RDH10specifi c siRNA inhibited it signifi cantly at 48 h as compared with nonsilencing control treated (NS) DCs ( Fig. 6A ). We quantifi ed the gene expression of RALDH2 by RT-qPCR. Only siRALDH2 could signifi cantly reduce the normalized mRNA level of RALDH2 as expected (supplementary Fig. V).
These data suggest that RDH10 is a key component of retinol conversion, and PPAR ␥ -mediated retinal oxidation is catalyzed by RALDH2.
In the next set of experiments, we electroporated monocytes with RDH10 -specifi c siRNA and treated as described in Fig. 6B ; then we measured CD1d cell surface protein expression by fl ow cytometry. Transient RDH10 siRNA transfection reduced CD1d levels on DCs and it was still downregulated at day 5 post-electroporation ( Fig. 6B ). These results strongly suggested that PPAR ␥ -mediated signaling induced retinol conversion by induced RDH10 in mo-DCs. The produced retinal is oxidized to ATRA by RALDH2. coexpressed with Dendritic Cell-Specifi c Intercellular adhesion molecule-3-Grabbing Non-integrin (CD-SIGN) (green cytoplasmic, arrows) in mucosal lymphoid tissue cells that have cytoplasmic projections in a network pattern indicating DC phenotype ( Fig. 4C ). Interestingly, nuclear PPAR ␥ (red) and the cytoplasmic TGM2 proteins (green) show coexpression in similar cells of GALT exhibiting cytoplasmic green projection characteristic of DC elements, comparable with the staining pattern as seen for PPAR ␥ -DC-SIGN of image C. Therefore, these cells coexpressing PPAR ␥ / TGM2 should represent the DC population of GALT, similarly to PPAR ␥ /DC-SIGN positive cells ( Fig. 4D ). These data indicate that in resting lymphoid tissues some of the PPAR ␥ -positive DCs express TGM2 simultaneously, suggesting that PPAR ␥ might regulate ATRA-dependent transcription in vivo as well. On the other hand, in GALT we showed that some PPAR ␥ (red) positive cells coexpressed RDH10 (green cytoplasm) ( Fig. 4E ). Similarly, we observed few PPAR ␥ -expressing DCs with RALDH2 and CRABP2 coexpression, respectively ( Fig. 4F, G ). RDH10, RALDH2, and CRABP2 (green cytoplasmic and/or nucleic) also colocalized with DC-SIGN (red cytoplasmic projections) in some mucosal DCs ( Fig. 4H-J ). Note that the number of RDH10-, RALDH2-,and CRABP2-positive cells are roughly the same when Fig. 4E-mpared to Fig. 4Hecting non-activated DC state in resting GALT. However, the number of PPAR ␥ positive cells are increased in casesof infl ammatory bowel diseases (IBDs) (data not shown).
These data collectively strongly suggest that the key components of ATRA synthesis and the PPAR ␥ are expressed together in some of the antigen-presenting cells (APCs) of the mucosal lymphoid tissues, consistent with a previous report which demonstrated that murine intestinal DCs expressed RALDH2 ( 6 ).

Increased RALDH activity in PPAR ␥ -activated mo-DCs
We wanted to provide functional evidence that indeed retinoic acid biosynthesis takes place in mo-DCs. Using a sensitive and quantitative liquid chromatography-mass spectrometry (LC-MS) method, we previously reported that mo-DCs have the ability to produce ATRA by oxidation of retinal in a PPAR ␥ -dependent manner ( 24 ). We aimed to further investigate this result and the function of cellular RALDHs using the ALDEFLUOR staining assay that is suitable to detect enzymatic activity of RALDHs inside the cells. Cells were differentiated (as described in the Materials and Methods) in the presence of DMSO/ethanol (C=Control) RSG, or RSG+GW9662. At 120 h, the cells were divided and incubated with fl uorescent ALDE-FLUOR, a substrate for RALDHs, either in the absence or the presence of DEAB, a specifi c RALDH inhibitor. After 40 min incubation, RALDH activity was measured by fl ow cytometry. There were 8% RALDH positive cells among vehicle-treated DCs ( Fig. 5 ). In the presence of RSG, the number of RALDH positive cells was increased to 40%. We noted that a much larger enzyme activity was displayed in these treated DCs than even in the positive ones in control cells. In the RSG and GW9662 cotreated sample, the RALDH activity was similar to vehicle-treated control DCs.
Based on these functional results, we could conclude that PPAR ␥ is acutely involved in retinoid signaling via inducing endogenous ATRA production. The primary enzyme for retinol oxidation is RDH10. The RSG-treated cells have the ability to synthesize ATRA because the coexpression of RDH10 and RALDH2 in the cells allows it. Moreover, retinoid signaling is more effective in the presence of PPAR ␥ -induced CRABP2 that transports ATRA to the nucleus.
The enhanced retinoid signaling was more effective in the presence of the CRABP2 ATRA transporter. In the nucleus, ATRA activates regulated target genes via RAR/RXR heterodimers due to integrated PPAR ␥ -RAR signaling .

PPAR ␥ -induced iNKT expansion is attenuated by RDH10, RALDH2, or CRABP2 knockdown
Next we aimed to assess the functional consequence of RDH10, RALDH2, and CRAB2 in in vitro functional assay. The lipid antigen-presenting capacity of PPAR ␥ -activated DCs is induced and mediated through the upregulated cell surface protein expression of CD1d protein ( 23,24,48 ). Human iNKT cells respond to ␣ -GalCer, a lipid antigen restricted exclusively to CD1d molecules. First, we sought to investigate whether RDH10 can infl uence the PPAR ␥ -mediated iNKT expansion capacity of the APCs. For this, we silenced the RDH10 gene in monocytes with RDH10 -specifi c siRNA or we used a nonsilencing control siRNA (NS siRNA) as described above for Fig. 5A . Cells were differentiated to mo-DCs in the presence of DMSO/ ethanol for the control-treated sample or RSG for PPAR ␥ activation and cells were pulsed with or without ␣ -GalCer for 48 h and then cocultured with autologous PBMCs at a 1:10 ratio. The iNKT proliferation capacity was monitored by V ␣ 24/V ␤ 11 double staining. As expected, enhanced iNKT expansion was detected in RSG-treated and NS siRNAtransfected samples, while reduced iNKT cell numbers were observed in RDH10 siRNA-treated cells ( Fig. 7A ).   Fig. 3. Regulation of CRABP2 expression in mo-DCs . A: Expression of human CRABP2 gene was quantifi ed by RT-qPCR. mo-DCs were obtained from three healthy donors and differentiated with IL-4 and GM-CSF for 5 days. Cells were treated at monocyte state with DMSO/ethanol (C=Control) or 2.5 M RSG. Means normalized to Cyclophilin A ± SD. B: CRABP2 protein level was determined by Western blot. Immune cell specifi city was shown as monocyte-derived macrophages (M s) failed to express CRABP2. GAPDH was used as loading control. C: CRABP2 expression was determined by IHC in monocytes (mo), control-treated mo-DC (mo-DC C), and RSG-treated mo-DC (mo-DC RSG) samples. CRABP2 was located in the cytoplasm of differentiated mo-DCs. Note that highest expression level for CRABP2 was observed in differentiated mo-DCs treated with RSG predominantly located in the cytoplasm. (Immune peroxidase reaction with methyl-green nuclear counterstaining; original magnifi cation: 40×.) DCs to generate de novo ATRA. We tested this hypothesis fi rst by using MLN DCs isolated from C57BL/6 mice, Sp-DCs from untreated mice, and ex vivo-differentiated BM-DCs in the presence of GM-CSF or GM-CSF and IL-4 ( 14 ). Using RT-qPCR and TLDA methods, we assessed that the transcription of the Rdh10 gene was detectable in all examined DC subtypes and that the expression of Rdh10 coincides/overlaps with Raldh2 . We could show clear correlation between the ATRA-generating capacity of in vivoand in vitro-produced DCs and the expression of the Cyp26a1 gene as a reliable marker for active retinoid signaling in intestinal DCs.
We also assumed that ATRA would infl uence the transcription of some of the known target genes identifi ed in human DCs. Interestingly, we did not observe correlations between Tgm2 and Cd1d genes and ATRA-producing capacity in in vivo subsets, particularly in the most widely characterized CD103 + cells.
Moreover, the gene encoding ATRA transporter Crabp2 was also barely or not detectable in DCs of different origin. An important question is how ATRA is transported in DCs. These transport mechanisms in MLN DCs are not characterized. ATRA can be delivered to the nucleus by CRABP2 that activates retinoid signaling via RAR/RXR heterodimers or by FABP5 molecules that enhance the ligand-induced transcriptional activities of PPAR ␦ /RXR heterodimers ( 47,53 ). In MLN CD103 + DCs CRABP2 is not expressed, but under the coculture condition the expression of Crabp2 mRNA was induced in sorted MLN DC103 + DCs suggesting that T cells most likely activate Crabp2 transcription in DCs. The signaling details of this induction require further investigations. We assessed the expression of Crabp1 in in vivo and in vitro DC subsets as well. According to our RT-qPCR measurement, Crabp1 is not expressed in these DCs. Based on these results, one might conclude that there might be other ATRA-delivering proteins in some of the mouse DCs.
The difference between human and murine DCs in this respect is apparent. Nuclear receptor-mediated signals might be involved in RALDH expression in intestinal DCs, leading to a specifi c subtype that not only stores and carries, but also produces a signifi cant level of ATRA . However, agonists of PPAR ␥ did not signifi cantly induce Raldh2 expression in Flt3L-generated BM-DCs and purifi ed Sp-DCs compared with the effect of IL-4 or IL-13 ( 14 ). These apparent discrepancies between the mouse and human DCs remain to be clarifi ed. We remain of the opinion that there is not suffi cient evidence to suggest that PPAR ␥ is able to regulate retinoid biosynthesis in murine DCs. We proved earlier that human monocytes expressed PPAR ␥ receptors and that the receptors are active under ex vivo culture conditions ( 23 ). Systematic microarray analyses also revealed that the activation of PPAR ␥ by RSG regulated the expression of genes contributing primarily to fatty acid uptake, transport, lipid storage, and attenuated lipid metabolism ( 43 ). Moreover, the PPAR ␥ -regulated lipid metabolic pathways could be associated with the altered immune response of DCs ( 23,24,48 ). PPAR ␥ enhanced the lipid antigen presentation capacity of DCs via CD1d DISCUSSION It is increasingly accepted that several links connect immune functions of the whole body to cellular metabolism . One of the most striking examples of such a regulatory role is the involvement of vitamin A/retinoid metabolism and ATRA production in the context of cell type specifi cation and immune regulation. ATRA's role is primarily linked to the gut and mucosal immunity.
Intestinal homeostasis is critically controlled by the interaction between immune cells, epithelial cells, and strains of commensal bacteria. The balance between immune reactivity and tolerance under steady-state and pathogenic conditions is primarily regulated by the immunoregulatory properties of GALT-associated DCs. The main goal of our study was to systematically survey mouse and human DCs for ATRA production and signaling, and to identify and validate additional key regulatory components of endogenous ATRA synthesis in human and mouse DCs.
The endogenous ATRA generation capacity in intestinal DCs was related to the expression of intracellular RALDHs. Although previous works have demonstrated that murine DCs express several Raldh isoforms ( 7,10,14,24,(49)(50)(51)(52), the functional contribution of these isoforms to ATRA synthesis remain to be determined and the prior enzymatic step has not been evaluated in these cell types at all.
For ATRA generation, retinol is believed to be converted to retinal by members of the aldehyde dehydrogenases. Alternatively, enzymes of the SDR superfamily may also participate in this process ( 6 ). Genetic evidence that RDH10 is indispensable for embryonic retinol metabolism has been provided ( 12,13 ) and pinpointed to this enzyme as the key mediator of this conversion. Furthermore, intracellular localization of RDH10 and RALDH2 could be correlated (16)(17)(18). These results suggested that RDH10 might be the key enzyme that could initiate retinol metabolism in intestinal DCs, and coexpression of RDH10 and RALDH2 is essentially required for the ability of mucosal CD103 +   earlier that RDH10 and RALDH2 proteins were expressed in mo-DCs and were upregulated upon RSG-treatment ( 24 ), and by using CRABP2-detecting IHC analysis, we confi rmed that mo-DCs are a relevant ex vivo model for investigating DC ATRA producing capacity and transport . Direct evidence for CRABP-mediated ATRA transport to the nucleus requires further experimentation.
To date our knowledge is still limited about the nature of in vivo ATRA-producing APCs in the human intestinal system. GALT is the most likely place where intensive lipid absorption occurs and PPAR ␥ activators can be generated. Therefore it is reasonable to assume that lipid signaling could participate in DC subtype specifi cation in these lymphoid tissues. Our IHC/DI results clearly suggested that DC-SIGN positive, GALT-associated immune cells readily express the key components of ATRA producing and signaling. In addition, PPAR ␥ -positive DCs coexpressed TGM2, strongly suggesting that these cells have an active retinoid signaling system and represent an in vivo relevant ATRA-producing cell type. Presumably our ex vivo mo-DCs may correspond to these in vivo DC/macrophage-like APCs.
The question of whether PPAR ␥ -induced ATRA generated by mucosal DCs is physiologically relevant remains to be answered. However, a link to the prevention of human IBD can be established. PPAR ␥ was shown to be associated with IBD in a mouse model of experimental colitis (56)(57)(58). Genetic evidence of PPAR ␥ -mediated protection against colon infl ammation was shown in PPAR ␥ heterozygous mice ( Pparg +/ Ϫ ), and targeted disruption of the Pparg gene in intestinal epithelial cells enhanced the susceptibility to dextran sodium sulfate (DSS)-induced colitis ( 59 ). RSG treatment and other PPAR ␥ ligands attenuate the severity of colitis in both intestinal epithelial cell and macrophagespecifi c PPAR ␥ mutant mouse strains ( 57,60 ). Due to potential side effects, the application of RSG in the treatment of IBD is not likely ( 61,62 ). Efforts toward the discovery of a new class of PPAR ␥ agonist that elicits therapeutic effects against IBD with limited or no adverse side effects revealed that conjugated linoleic acid (CLA) is a safer alternative to RSG in the model of spontaneous pan-enteritis and DSS-induced colitis (63)(64)(65).
Future studies should assess the role of DC-expressed PPAR ␥ in IBD and potentially in other diseases. Our experimental results can be useful to design an even more effective treatment in the prevention or amelioration of ulcerative colitis and Crohn's disease.
In summary, we have established a linear relationship between RDH10, RALDH2, and CRABP2 in PPAR ␥ -induced ATRA synthesis and signaling that showed the requirement of these proteins. Further studies are needed to dissect the regulation of these pathways and ways to modulate their expression and activity. These new pathways could provide insight into how ATRA production could shape the immune response, and could potentially be used to treat diseases affecting mucosal immunity.
The authors would like to acknowledge the excellent technical help of Ms. Ibolya Furtos and Marta Beladi. The authors are molecules ( 23,24,48 ), which is essential for ligand presentation and recognition by the iNKTs ( 26,54,55 ).
In this cellular context, we have found that activation of PPAR ␥ initiates a transcriptional program of altered lipid metabolism in mo-DCs via induced expression of genes that are required for endogenous ATRA production ( 24 ). Importantly, we assumed that retinol is metabolized to retinal primarily by RDH10 in this process. The mRNA level of RDH10 was increased by PPAR ␥ after 6 h, suggesting that this gene, like FABP4 , could be regulated directly by PPAR ␥ , while the direct regulatory function and binding of PPAR ␥ /RXR via a response element in the RDH10 promoter still needs to be investigated. mo-DCs express the key enzymes (RALDHs) of ATRA production. ATRA activated RAR ␣ receptors and induced CD1d expression at mRNA and protein levels. These results suggested that transcriptional events in human mo-DCs that upregulate the CD1D gene are coordinately mediated by PPAR ␥ and RAR ␣ receptors ( 23,24 ).
We characterized the expression of genes responsible for retinol uptake into murine and human DCs. We could not detect STRA6 expression in mo-DCs, the only known high-affi nity receptor for RBP4 ( 33,34 ), as we expected based on our microarray results (data not shown). We have similar results in the case of murine DCs. The structure of RBPR2, a novel retinol transporter, is related and highly conserved between human and murine STRA6 ( 35 ). Therefore, we hypothesized that RBPR2 could be an alternative receptor in murine and human DCs for retinol uptake. We detected Rbpr2 gene expression in a different subset of murine DCs. This gene was upregulated in MLN CD103 + DC under coculture conditions. The human ortholog of RBPR2 is localized at two separate sites ( RBPR2A and RBRP2B ) of chromosome 9's p and q arms ( 35 ). The homology between murine Rbpr2 and the two human ortholog Open Reading Frames (ORFs) suggests the possibility that if these ORFs are transcribed, they can form functional protein on the DC surface. The expressions of human RBPR2A and RBRP2B were assessed and presented in supplementary Fig. VI. The expression of RBPR2A could be detected; however, this was highly donor-dependent and not altered by PPAR ␥ . RBRP2B was also expressed by DCs, but the normalized mRNA level of this fragment was much lower compared to RBPR2A . Additional RBP-binding assays would be required to confi rm whether RBPR2A and RBRP2B can form a functional receptor on DCs.
Based on these and our previous results we chose mo-DCs as our model system to test the functional contribution of the above enzymes for ATRA synthesis and signaling. This approach proved to be successful and validated the functional requirement suggested by the gene expression studies not only at the level of gene expression, but also in T cell activating function.
An additional fi nding of this work is that not only ATRA synthesis can be regulated by PPAR ␥ but also the delivery of the active metabolite to the nucleus. We found that CRABP2 transporter protein is upregulated upon RSG-treatment, suggesting that DCs with activated PPAR ␥ have enhanced ATRA transport to the nucleus. We showed