Similarities and differences in the transcriptional control of expression of the mouse TSLP gene in skin epidermis and intestinal epithelium

Significance

Thymic stromal lymphopoietin (TSLP) is a critical immunoregulatory cytokine that plays important physiological functions in epithelial cells in skin and intestinal barriers. However, the molecular mechanisms controlling TSLP expression in vivo are still poorly understood. Using tissue-selective mutagenesis in mice, we have identified the involvement of multiple transcriptional factors, including several nuclear receptors and their cognate agonistic ligands, in the transcriptional regulation of TSLP in epidermal keratinocytes and intestinal epithelial cells. Importantly, this investigation also demonstrates that the retinoid X receptor (RXR) and retinoic acid receptor (RAR) isotypes are not functionally redundant in vivo. Taking our data together, the present study unveils the topological map and the combinatorial mechanisms involved in tissue-specific transcriptional regulation of TSLP expression in epidermal keratinocytes and intestinal epithelial cells.

Abstract

We previously reported that selective ablation of the nuclear receptors retinoid X receptor (RXR)-α and RXR-β in mouse epidermal keratinocytes (RXR-αβ^ep−/−) or a topical application of active vitamin D3 (VD3) and/or all-trans retinoic acid (RA) on wild-type mouse skin induces a human atopic dermatitis-like phenotype that is triggered by an increased expression of the thymic stromal lymphopoietin (TSLP) proinflammatory cytokine. We demonstrate here that in epidermal keratinocytes, unliganded heterodimers of vitamin D receptor (VDR)/RXR-α and retinoic acid receptor-γ (RAR-γ)/RXR-β are bound as repressing complexes to their cognate DNA-binding sequence(s) (DBS) in the TSLP promoter regulatory region. Treatments with either an agonistic VD3 analog or RA dissociate the repressing complexes and recruit coactivator complexes and RNA polymerase II, thereby inducing transcription. Furthermore, we identified several functional NF-κB, activator protein 1 (AP1), STAT, and Smad DBS in the TSLP promoter region. Interestingly, many of these transcription factors and DBS present in the TSLP promoter region are differentially used in intestinal epithelial cell(s) (IEC). Collectively, our study reveals that, in vivo within their heterodimers, the RXR and RAR isotypes are not functionally redundant, and it also unveils the combinatorial mechanisms involved in the tissue-selective regulation of TSLP transcription in epidermal keratinocytes and IEC.

We reported that keratinocyte-selective ablation of retinoid X receptor (RXR)-α and RXR-β in epidermal keratinocytes of the mouse (RXR-αβ^ep−/− mutants) results in a skin and systemic syndrome that mimics human atopic dermatitis (AD) and is preceded, in epidermal keratinocytes, by enhanced expression of the thymic stromal lymphopoietin (TSLP) cytokine (1). Moreover, several lines of evidence have revealed that TSLP expression is both necessary and sufficient to induce an atopic inflammation in the mouse (1–4). TSLP, which is also expressed in human AD skin lesions (5, 6), has been considered to be the master regulator of allergic inflammation (7).

Interestingly, the TSLP promoter region was found to contain several putative nuclear receptor (NR) DNA-binding sequence(s) (DBS) (2). Topical treatment with active vitamin D3 [1α, 25(OH)2 (VD3)] or its low-calcemic analog calcipotriol (also named MC903; hereafter, MC), all-trans retinoic acid (RA), and the retinoic acid receptor-γ (RAR-γ)–selective retinoid BMS961, which are agonistic ligands for vitamin D receptor (VDR), all three RARs, and RAR-γ, respectively, could induce TSLP expression in mouse keratinocytes on their own or synergistically (2). However, MC was more efficient than BMS961 at inducing TSLP expression in these cells, and long-term treatment with MC resulted in an AD-like syndrome similar to the syndrome observed in RXR-αβ^ep−/− mice. To reveal how both the keratinocyte-selective ablation of RXR-α and RXR-β (RXR-γ is not expressed in keratinocytes) or MC and/or BMS961 treatments could induce TSLP expression, we posited that because there is no RA and very little, if any, VD3 in keratinocytes (2) under in vivo homeostatic conditions, the activity of the TSLP promoter could be repressed by corepressor-bound unliganded RXR-α (or RXR-β)/VDR and RXR-α (or RXR-β)/RAR-γ heterodimers bound to VDR and RAR response elements (VDRE and RARE, respectively). Thus, RXR-α and RXR-β ablations, which release both heterodimers from their DBS, might abolish repression and allow other promoter-bound transcription factors (TFs) to stimulate TSLP transcription. Moreover, MC application would generate RXR/VDR-coactivator complexes, the transcriptional activity of which would be sufficient to overcome the repression exerted by RXR/RAR-γ corepressor complexes, thereby enhancing the basal promoter activity, whereas RXR/RAR-γ coactivator complexes formed upon application of BMS961, would be less efficient at relieving the repression exerted by RXR/VDR-corepressor complex (figure 5 of ref. 2).

We investigated here the validity of such a mechanism through which RXR/VDR and RXR/RAR-γ heterodimers regulate TSLP expression. We also characterized DBS for additional TFs, which can function independent of RAR-γ and VDR. Furthermore, because TSLP is also expressed in intestinal epithelial cell(s) (IEC), we compared the binding pattern of TFs associated with the TSLP promoter region in epidermis and IEC and revealed striking differences between these tissues. Taken together, our studies unveil the complex organization of the TSLP promoter, and demonstrate how TSLP expression is controlled at the transcriptional level in mouse epidermis and IEC.

Results

VDR, RARs, and RXRs Control TSLP Expression at the Transcriptional Level in Mouse Epidermal Keratinocytes.

To investigate whether MC- and RA-induced increases in TSLP RNA level in epidermis of wild-type (WT) mice (2) are due to increased transcription, nuclear run-on assays were performed using epidermis of WT mice treated with vehicle (ethanol), MC, or RA. TSLP-specific run-on transcript signals were exclusively detected in MC- and RA-treated samples (Fig. 1A, lanes 1–3). Moreover, run-on transcripts after MC treatment produced a stronger signal than the signal generated by RA, indicating that a higher TSLP RNA level achieved with MC versus RA (2) was due to a higher rate of transcription. Similarly, cotreatment with MC and RA generated a stronger TSLP signal (Fig. 1A, lane 4), in agreement with the synergistic increase observed at the RNA level (2). Epidermis from RXR-αβ^ep−/−, RAR-γ/VDR^ep−/−, and RAR-αγ/VDR^ep−/− mice (Fig. 1A, lanes 8–10) was subjected to nuclear run-on assays, whereas β-actin transcription was measured to ensure that an equal amount of nuclei was used across the various samples and the pSK⁺ vector served as a negative control (Fig. 1A). In all cases, increases in TSLP RNA (Fig. 2 B–E) and in protein levels (Fig. 2 F and G) were correlated with increased transcription from the TSLP promoter (Fig. 1A).

Fig. 1.

Multiple NRs regulate TSLP expression in the mouse epidermal keratinocytes and in IEC. (A) Nuclear run-on assay using epidermis from WT mice treated as indicated (panels 1–5 and panel 7) or from different keratinocyte-selective mutants (panels 8–10). Autoradiograms of labeled transcripts hybridized with TSLP, β-actin, and control vector (pBSK⁺) DNA are displayed in lanes 1–5 and 7–10. Lanes 6 and 11 show ethidium bromide (EtBr) staining of DNA probes. Veh., vehicle. (B) Schematic representation of NR DBS present in the mouse TSLP gene. The A base of the translation initiation codon (ATG) was taken as +1. Every indicated DBS recruited its cognate NRs in vivo in at least one of the examined tissues. (C) ChIP assays using WT mouse epidermis, treated as indicated, show the binding of RXR-α and VDR to TSLP DR3 DBS. Ab IP, antibodies used for immunoprecipitation. (D) ChIP assays using epidermis from WT and VDR^ep−/− mice to reveal VDR and RXR-α recruitment to the TSLP DR3d region. (E) ChIP assays using epidermis from WT and RAR-γ^−/− mice to probe RAR and RXR-α recruitment to the TSLP DR2b DBS. (F) ChIP assays using Veh. and RA-treated epidermis to reveal RAR and RXR binding to the indicated DBS. (G) Quantitative RT-PCR (Q-RT-PCR) of genes, as indicated, from epidermis of WT and RAR-γ^−/− mice topically treated for 16 h. (H) ChIP assays using epidermis from WT and various keratinocyte-selective NR mutant (Mut) mice to detect VDR and RAR binding to indicated DBS. (I) ChIP assays using epidermis and ileal and colonic epithelium from WT mice to reveal RAR and RXR-α binding to different DR2 DBS, as indicated. (J) ChIP assays of WT mouse epidermis to reveal RXR-α binding to TSLP DR1 elements, as indicated. (K) Q-RT-PCR of TSLP and PPAR-γ RNA from ileal and colonic epithelium of WT and PPAR-γ^iec−/− mice injected i.p. with rosiglitazone (Rosi.). (L) ChIP assays using ileum and colon epithelium isolated from WT mice injected i.p. with Veh. or Rosi. to detect PPAR-γ and RXR-α binding to indicated TSLP DR1 DBS. All Q-RT-PCR values are mean ± SEM.

Fig. 2.

Repressing and activating complexes are assembled by association of RXR-β/RAR-γ and RXR-α/VDR on TSLP DR2 and DR3 DBS, as well as on the PP region, in the absence and presence of their cognate ligands, respectively. (A) ChIP assays performed from WT mouse dorsal epidermis, treated as indicated. (B) TSLP Q-RT-PCR from WT and RAR-γ^−/− mice treated with 0.25 nmol MC. (C) TSLP Q-RT-PCR from WT and VDR^ep−/− mice treated with 1 nmol RA per ear. (D) TSLP Q-RT-PCR from WT and RXR-β^ep−/− mice treated with 4 nmol MC per ear. (E) TSLP Q-RT-PCR from ears of WT and mutant mice, as indicated, 2 wk after the first Tam injection. (F) Serum TSLP level in the mice used in Fig. 2E. n.d., not detected. (G) TSLP immunohistochemistry (Upper) and hematoxylin/eosin (HE) staining (Lower) images of ear sections of WT and mutant mice, as indicated. The dotted line indicates the epidermis/dermis junction. (Upper) Yellow and cyan colors reveal the staining of TSLP and nuclei, respectively. Note the different degrees of epidermal hyperproliferation and dermal cell infiltration in HE-stained ear sections of mutant mice. (Scale bars: 25 μm, Upper; 100 μm, Lower.) In B–D, all Q-RT-PCR values are mean ± SEM. All topical treatments to the ear were performed for 16 h.

Unliganded RXR-α/VDR and RXR-β/RAR-γ Heterodimers Bind to Their TSLP Cognate Response Elements, but Activation of Transcription Requires the Presence of Agonistic Ligands.

We previously noted the presence of several putative NR DBS in the mouse TSLP gene upstream promoter region (2). Allowing for two base mismatches in NR consensus DBS, a thorough “ocular” and bioinformatics analysis of 100-kb upstream and 20-kb downstream DNA sequences from the mouse TSLP +1 position (considering “A” of the translation initiation codon ATG as position +1) revealed seven putative VDREs (DR3a–DR3g), two putative RAREs (DR2a–DR2b), and three putative DR1 DBS (DR1a–DR1c) known to bind RXRs heterodimerized with either peroxisome proliferator-activated receptor (PPAR), liver X receptor (LXR), or farnesoid X receptor (FXR), among others (Figs. 1B and 5A and Table S1). Note that DR3b, DR3c, and DR3e DBS, as well as DR3d, DR3f, and DR3g DBS, contain identical sequences (Table S1), whereas DR3f and DR3g DBS are present within a 2.13-kb-long repeated sequence spanning positions −32,824 to −30,694 (encompassing DR3f) and −44,655 to −42,526 (encompassing DR3g).

Fig. 5.

Schematic representation of the mouse TSLP promoter-enhancer region. (A) In vivo DNA binding sites (DBS; indicated in red) for TFs and NRs located in the proximal and distal regions in the mouse TSLP gene. The A base of the translation initiation codon (ATG) was taken as +1 (also Table S1). (B) In epidermal keratinocytes of untreated WT mice, none of the TFs (except Smad2, Smad3, and Smad4) are recruited to their cognate DBS. Topical skin treatment with TPA or IL-1β to cells in culture induces NF-κB (p65/p50), AP1 (c-Fos/c-Jun), and STAT (STAT5/STAT6) binding, as well as TSLP transcription. Note that the indicated NR DBS are permanently occupied in vivo by their cognate NRs. All of the DBS are indicated in red. (C) In IEC of WT mice, the indicated (*) sites are permanently occupied. Note that in IEC, the DR2a DBS is functional and recruits RAR-γ/RXR-α heterodimer (indicated in blue), whereas the DR2b DBS is used in epidermis, in which RAR-γ/RXR-β is recruited. All of the DBS are indicated in red.

Table S1.

Position, sequence, and functional characteristic of various TF DNA binding sites (DBS) present in mouse TSLP gene

Name of the DBS	Position from ATG (A = +1)	DBS	TF binding in mouse epidermis and/or IEC
			In vitro EMSA + antibody shifting	In vivo ChIP (IP antibody)
Consensus DR1		RggtcanRggtca
DR1a	−2,176 to −2,164	gggtcaggggaca	RXR-α, PPAR-α, PPAR-γ	RXR-α, PPAR-γ
DR1b	−5,727 to −5,715	atggcaaaggtca	RXR-α, PPAR-α, PPAR-γ	RXR-α, PPAR-γ
DR1c	−14,702 to −14,690	aggtaagatgtca	—	RXR-α, PPAR-γ
Consensus DR2		RggtcannRggtca
DR2a	−1,063 to −1,050	agctcaacaggtca	RXR-α, RAR-γ	Pan-RAR
DR2b	−13,892 to −13,879	aggtcatgagttca	RXR-α, RAR-γ	Pan-RAR
Consensus DR3		RggtcannnRggtca
DR3a	−4,038 to −4,024	agttcattccggtca	—	—
DR3b	−4,121 to −4,107	aggacagccagggct	—	—
DR3c	−6,506 to −6,492	aggacagccagggct	—	—
DR3d	−7,369 to −7,355	gagccagaggggtca	RXR-α, VDR	RXR-α, VDR
DR3e	−13,658 to −13,644	aggacagccagggct	—	—
DR3f	−32,704 to −32,690	gagccagaggggtca	RXR-α, VDR	RXR-α, VDR
DR3g	−44,535 to −44,521	gagccagaggggtca	RXR-α, VDR	RXR-α, VDR
Consensus NF-κB		gggRnnYYcc
NF-κBa	−237 to −218	gggaaattcc	p65, p50*	p65, p50*
NF-κBb	−1,447 to −1,438	aggaacttcc	—	—
NF-κBc	−3,587 to −3,578	caggaatttc	p65, p50*	p65, p50*
NF-κBd	−3,771 to −3,762	gaaaattacc	—	—
NF-κBe	−3,997 to −3,988	gagaatcccc	—	—
NF-κBf	−4,612 to −4,601	agggactctc	—	—
Consensus AP1		tgagtca
AP1a	−1,235 to −1,228	tgactca	c-Fos, c-Jun*	—
AP1b	−16,647 to −16,640	tgagtca	c-Fos, c-Jun*	c-Fos, c-Jun*
AP1c	−29,950 to −29,943	tgactca	c-Fos, c-Jun*	c-Fos, c-Jun*
AP1d	−41,783 to −41,776	tgactca	c-Fos, c-Jun*	c-Fos, c-Jun*
AP1e	−46,283 to −46,276	tgagtca	c-Fos, c-Jun*	—
AP1f	−71999 to −71992	tgagtca	c-Fos, c-Jun*	c-Fos, c-Jun*
AP1g	−77,688 to −77,681	tgagtca	c-Fos, c-Jun*	c-Fos, c-Jun*
Consensus Smad		agccagaca
Smad a	−27,191 to −27,183	agacagaca	Smad3, Smad4	Smad2, Smad3, Smad4
Smad b	−38,999 to −38,991	agacagaca	Smad3, Smad4	Smad2, Smad3, Smad4
Consensus STAT5		ttcnnngaa
STAT5a	−6,967 to −6,959	ttcggggaa	nd	Stat5*
Consensus STAT6		ttcnnnngaa
STAT6a	+2,918 to +2,927	ttcgtgtgaa	nd	Stat5/6*
STAT6b	−623 to −614	ttccagtgaa	nd	Stat5/6*
STAT6c	−20,262 to −20,253	ttccctagaa	nd	Stat5/6*
STAT6d	−48,832 to −48,823	ttcctctgaa	nd	Stat5/6*
STAT6e	−61,255 to −61,246	ttcccaagaa	nd	Stat5/6*
STAT6f	−72,700 to −72,691	ttctcaagaa	nd	Stat5*
STAT6g	−84,196 to −84,187	ttcctgagaa	nd	Stat5*

List of various TF binding sites present in the mouse TSLP gene. Functionality of the binding sites was determined in vitro by EMSA and antibody shifting and in vivo by ChIP assays using either mouse epidermal extracts or IEC, as described in Results. Consensus sequence of the binding sites is denoted in bold letters. IP, immunoprecipitation; n, a/c/g/t; nd, not done; R, purine; Y, pyrimidine.

^*Strongly stimulated by topical TPA treatment or IL-1β.

Binding of their cognate NRs to these DBS was analyzed in vitro by electrophoretic mobility shift assay (EMSA) and supershift assay, using epidermal nuclear extracts (NEs) and the respective antibodies. Only DR3d, DR3f, and DR3g formed a complex that was supershifted with RXR-α and VDR antibodies. Because DR3d, DR3f, and DR3g have identical sequences (Table S1), only DR3d was tested. Importantly, this complex was absent in NE from VDR^ep−/− mice (Fig. S1A). Both DR2a and DR2b complexes were shifted by RXR-α and RAR-γ antibodies (Fig. S1B), whereas only DR1a- and DR1b-bound complexes were supershifted by RXR-α, PPAR-α, and PPAR-γ antibodies (Fig. S1C; no efficient PPAR-β antibody was available for supershift). MC, RA, fenofibrate (PPAR-α agonist), or rosiglitazone (PPAR-γ agonist) treatment did not affect VDR, RAR, and PPAR binding to their response elements in vitro. None of these NR DBS were perfect consensus elements, and only those NR DBS containing at least one consensus repeated motif associated with the corresponding NR (Table S1).

Fig. S1.

NRs control the TSLP expression in mice. (A) EMSA and antibody supershift assay using epidermal NE from WT and VDR^−/− mice show the binding of VDR and RXR-α to various DR3 elements, as indicated. (B) EMSA and antibody supershift assay using epidermal NE from WT and RAR-γ^−/− mice show the binding of RXR-α and RAR-γ to various DR2 elements, as indicated. (C) EMSA and antibody supershift assay using WT mouse epidermal NE show the binding of RXR-α, PPAR-α, and PPAR-γ to various DR1 elements, as indicated. (D) ChIP assays using dorsal epidermis and ileal and colonic epithelium from WT mice to detect VDR and RXR-α binding to DR3d and DR3f/g elements, as indicated. (E) ChIP assays using ileum and colon epithelium from WT mice. Immunoprecipitation (IP) was done using different antibodies, as indicated (Ab IP), followed by PCR with DR3d and DR3f/g primers, as indicated. In A–C, arrows indicate supershifted bands. ab., antibody; Cons., consensus DBS.

We then tested whether DR3, DR2, and DR1 DBS were associated with their corresponding NR partners in epidermis. Using RXR-α and VDR antibodies, and irrespective of MC treatment, chromatin immunoprecipitation (ChIP) assays from vehicle-treated epidermis revealed the association of VDR and RXR-α with DR3d and DR3f/g DBS (Fig. 1C) but, as expected, not with DR3a and DR3b. Because DR3f and DR3g are present within a repeat sequence, unique primers could not be designed specifically to assess VDR and RXR-α binding to these regions. The specificity of the ChIP assay was confirmed by the lack of VDR and RXR-α binding to the DR3d DBS in epidermis of VDR^ep−/− mice (Fig. 1D). A similar pattern was observed for VDR and RXR-α binding to the DR3d and DR3f/g DBS in WT mouse IEC (Fig. 5 A–C and Fig. S1D).

Because no “ChIP-grade” antibody specific against RAR isotypes was available, a pan-RAR antibody (reacting with all three RAR isotypes) was used to investigate whether a RAR could associate with TSLP DR2 DBS. Irrespective of RA treatment, neither a RAR nor RXR-α was associated with the DR2a DBS in epidermis, whereas under identical conditions, a RAR was bound to the DR2b DBS (Fig. 1F). To identify the RAR isotype associated with the TSLP DR2b DBS, we performed ChIP assays using epidermis from RAR-γ^−/− mice. No binding of RAR with DR2b was detectable, indicating that RAR-γ selectively associated with DR2b (Fig. 1E). The specific involvement of the RAR-γ isotype in regulation of TSLP transcription was confirmed by using the isotype-specific RAR agonistic ligands BMS961 and BMS753 (8), which are selective for RAR-γ and RAR-α, respectively. Measurement of TSLP RNA level in WT and RAR-γ^−/− mice treated with these ligands revealed that both RA and BMS961, but not BMS753, enhanced TSLP levels in WT animals, whereas none of these ligands increased TSLP levels in RAR-γ^−/− mice (Fig. 1G). As a control, the above ligands were tested for the induction of CRABPII RNA [a known RA target gene (9)]. In WT mice, RA, BMS961, and BMS753 could increase CRABPII RNA level, but RA and BMS961 were more efficient than BMS753. As expected, in RAR-γ^−/− mice, only RA and BMS753 could weakly induce CRABPII RNA (Fig. 1G).

Surprisingly, RXR-α could not be detected on the DR2b DBS (Fig. 1F). We therefore tested whether RXR-α/RAR heterodimers could associate with the DR2 DBS present in the CRABPII gene. Both RXR-α and RAR were detected on the CRABPII DR2 (Fig. 1F), indicating that the lack of RXR-α association with the DR2b DBS was specific to the TSLP gene. Whether RAR-γ could be selectively heterodimerized with RXR-β instead of RXR-α on the TSLP DR2b region was investigated using ChIP assays carried out with epidermis from RXR-α^ep−/−, RXR-β^ep−/−, and RXR-αβ^ep−/− mice. As expected, VDR binding was unaffected in RXR-β^ep−/− mice, whereas RAR-γ association decreased by ∼90% in these mice (Fig. 1H), thus indicating that VDR was heterodimerized with RXR-α, whereas RAR-γ was mostly associated with RXR-β. Accordingly, in RXR-α^ep−/− mice, VDR did not associate with DR3d DBS (Fig. 1H), whereas RAR-γ binding to the DR2b DBS was decreased by ∼30% (Fig. 1H). As expected, neither VDR nor RAR-γ associated with its cognate DBS in RXR-αβ^ep−/− mice. Interestingly, RAR bound to the CRABPII DR2 with equal efficiency in both WT and RXR-β^ep−/− mice, whereas RAR binding was decreased by more than 80% in RXR-α^ep−/− mice and no significant binding was detected in RXR-αβ^ep−/− mice (Fig. 1H), thus indicating that RAR-γ was mostly heterodimerized with RXR-α on the CRABPII DR2 RARE.

In contrast to epidermis, ChIP with WT mouse IEC from ileum and colon revealed RAR-γ and RXR-α associations with the TSLP DR2a DBS, whereas neither RAR nor RXR-α was detected at the DR2b DBS (Figs. 1I and 5 A–C). The CRABP II DR2 DBS in IEC showed a similar NR binding pattern as in epidermis (Fig. 1I).

To identify NRs bound to the TSLP DR1 DBS (Fig. 5A), ChIP was performed using RXR-α, PPAR-α, and PPAR-γ antibodies. Only RXR-α binding to the DR1a DBS was detected in epidermis (Fig. 1J). Skin topical treatment with PPAR agonists (fenofibrate and rosiglitazone) could not induce TSLP expression. In contrast, rosiglitazone increased the TSLP RNA level in the IEC of WT mouse and IEC-selective ablation of PPAR-γ (PPAR-γ^iec−/−) abolished this increase (Fig. 1K). ChIP assays using colonic cells further demonstrated the constitutive association of PPAR-γ and RXR-α to the DR1b and DR1c DBS, and rosiglitazone induced the binding of PPAR-γ and RXR-α to the DR1a region (Fig. 1L). In contrast, in WT ileal IEC, constitutive binding of PPAR-γ and RXR-α occurred only on the DR1c region, whereas DR1a, DR1b, and DR1c all bound PPAR-γ and RXR-α upon rosiglitazone treatment (Figs. 1L and 5C). Irrespective of treatments with cognate agonists, PPAR-α and PPAR-β binding could not be detected with any of the TSLP DR1-DBS in IEC.

Taken together, these results demonstrated that (i) RXR-α selectively heterodimerizes with VDR on TSLP DR3 DBS; (ii) RXR-α is also the predominant partner of RAR on the CRABPII DR2 DBS; (iii) RXR-β is the predominant partner of RAR-γ on the TSLP DR2b DBS in epidermis; (iv) RXR-β may partially substitute for RXR-α binding on the CRABPII DR2 element in epidermis of RXR-α^ep−/− mice; (v) RXR-α could be weakly redundant with RXR-β on the TSLP DR2b DBS; and (vi) there are striking differences in patterns of RAR/RXR-heterodimer binding to TSLP RAREs in IEC and epidermis, as well as differential rosiglitazone induction patterns of TSLP expression mediated by DR1 PPRE in IEC, thus illustrating the tissue-specific control of TSLP transcription by NRs (Fig. 5 A–C).

In the Absence and Presence of Their Cognate Ligands, VDR/RXR-α and RAR-γ/RXR-β Assemble Repressing and Activating Complexes, Respectively, on TSLP VDRE and RARE.

Because unliganded RXR-α/VDR and RXR-β/RAR-γ heterodimers are “constitutively” bound to their DBS on the TSLP gene in the absence of any treatment of epidermis, we investigated whether these heterodimers could be functionally active in repressing TSLP transcription. It is known that in the absence of agonistic ligands, DNA-bound RXR/VDRs and RXR/RARs assemble repressing complexes containing corepressors and histone deacetylases (HDACs), thereby establishing a transcriptionally inactive state (10). ChIP with WT epidermis revealed the presence of SMRT and HDAC2, (but not HDAC1, HDAC3, and HDAC7) on DR3d, DR3f, and/or DR3g and DR2b DBS (Fig. 2A). On the other hand, following MC treatment, the corepressors SMRT and HDAC2 disappeared from DR3d DBS, and we observed the appearance of the SRC2 (TIF2) and SRC3 coactivators, as well as p300, pCAF, and CDK7 TFs and RNA polymerase II (Pol II). In contrast, the DR3f and/or DR3g DBS, as well as the DR2b DBS, were still associated with corepressors. On the other hand, RA treatment resulted in loss of SMRT and HDAC2 and recruitment of SRC2 (but not of SRC3), as well as p300, pCAF, CDK7, and Pol II specifically to the DR2b DBS (Fig. 2A). To assess the overall transcription status of the TSLP promoter, we amplified the proximal promoter region (−318 to −8 bp from position +1; hereafter, PP). In untreated epidermis, the PP region was associated with SMRT, HDAC2, and pCAF, as well as VDR to a much lesser extent. Treatment with MC or RA resulted in dissociation of SMRT and HDAC2 and in association of SRC2 and SRC3 (SRC3 was seen only in MC-treated samples), as well as p300, pCAF, CDK7, and Pol II, on the PP region (Fig. 2A). In the ileum and colon, irrespective of VD3 treatment, the DR3d DBS was constitutively associated with SRC2, SRC3, and Pol II, along with RXR-α and VDR, whereas the DR3f/DR3g region recruited SMRT with RXR-α and VDR (Fig. S1 D and E).

Because both VDR/RXR-α and RAR-γ/RXR-β assembled repressing complexes in untreated epidermis, we aimed at identifying which of the two complexes was the dominant one. Ears of WT and RAR-γ^−/− mice were treated with a limiting dose of MC (0.25 nmol per ear). No significant increase in TSLP RNA was observed in WT mice, whereas an approximately fivefold increase was observed in RAR-γ^−/− mice (Fig. 2B). On the other hand, treatment of WT and VDR^ep−/− mice with RA at a dose of 1 nmol per ear resulted in an ∼1.5-fold increase and a twofold increase in TSLP RNA level, respectively (Fig. 2C), suggesting that the RAR-γ–associated repressor complex played a dominant role in repressing TSLP transcription. In agreement with these results, treatment of RXR-β^ep−/− mice with 4 nmol of MC per ear resulted in an ∼150-fold increase in TSLP RNA level, in comparison to only an ∼50-fold increase in WT mice (Fig. 2D).

We then investigated whether RAR-α was involved in the repression of TSLP expression. Because unliganded RXR-α/VDR and RXR-β/RAR-γ heterodimer complexes appeared to be responsible for the repression of TSLP transcription in WT mice epidermis, we assumed that selective ablation of VDR and RAR-γ in keratinocytes should result in increased TSLP expression, similar to the increased TSLP expression observed in RXR-αβ^ep−/− mice. Therefore, we generated conditional knockout mice lacking both RAR-γ and VDR in keratinocytes (RAR-γ/VDR^ep−/− mice). Although increases in TSLP RNA and protein levels were observed in these mice, they were lower than the levels observed in RXR-αβ^ep−/− mice (Fig. 2 E and F, also Fig. 1A). Moreover, these mice did not develop the pathological phenotype typical of RXR-αβ^ep−/− mice, as judged by the external ear phenotype (2) and histological analysis (epidermal hyperproliferation and dermal immune cell infiltrate; Fig. 2G). Interestingly, mice selectively lacking RAR-α in addition to RAR-γ and VDR in keratinocytes (RAR-αγ/VDR^ep−/−) developed a pathological phenotype closer to RXR-αβ^ep−/− mice (Fig. 2G). As expected, the levels of TSLP RNA and protein in the RAR-αγ/VDR^ep−/− triple mutants were comparable to the levels in RXR-αβ^ep−/− mice (Fig. 2 E and F, also Fig. 1A). That RAR-γ/VDR^ep−/− mutants exhibited a milder pathological phenotype than RXR-αβ^ep−/− mutant mice may reflect a redundancy between RAR-α and RAR-γ isotypes such that when RAR-γ is ablated, RAR-α could substitute for some of its repressor activity. However, treatment of RAR-γ/VDR^ep−/− mice with the RAR-α–selective BMS753 ligand did not result in any further increase in TSLP RNA, and we could not detect any RAR-α bound at the DR2b DBS in these mutants, which suggests the alternate possibility that ablation of RAR-αγ/VDR or RXR-αβ in the keratinocytes could activate additional TFs, which would, in turn, induce TSLP transcription.

Taken together, these ChIP assays and the nuclear run-on assays (within their limit of sensitivities) demonstrate that in WT epidermis, heterodimers of VDR/RXR-α and RAR-γ/RXR-β are bound to their respective DBS on the TSLP gene, along with the corepressors SMRT and HDAC2, thereby maintaining the TSLP gene in a transcriptionally inactive state. Upon MC or RA treatment, SMRT and HDAC2 are released from the DNA-bound NR complexes, followed by recruitment of SRC2- and/or SRC3-bearing coactivator complexes, leading to TSLP expression. Importantly, although all three DR3 DBS (DR3d, DR3f, and DR3g) could associate with RXR-α, VDR, SMRT, and HDAC2, only the DR3d DBS could recruit coactivators upon VD3 treatment (Fig. 2A), indicating that not all VDRE-bound RXR-α/VDR complexes are able to function as transcriptional activators even in the presence of an agonistic ligand. Interestingly, the constitutive assembly of an activating complex of VDR, RXR-α, SRC2, SRC3, and Pol II at the DR3d DBS in IEC suggested that VD3 could be instrumental in TSLP expression in these cells (11).

Multiple Functional Smad, STAT, NF-κB, and Activator Protein 1 DBS Are Present in the TSLP Promoter Region.

A bioinformatics analysis across 100 kb upstream and 20 kb downstream from the TSLP +1 translation initiation codon revealed the presence of several putative Smad, NF-κB, activator protein 1 (AP1), and STAT DBS (Fig. 5A). The TSLP upstream promoter region contains two putative Smad DBS (Fig. 3A and Table S1). In EMSAs, both Smad a and Smad b DBS could bind Smad3 and Smad4 (Fig. S2A). ChIP assays with WT epidermis, as well as with mouse lung epithelial 12 (MLE12) cells, revealed that Smad2, Smad3, and Smad4 could bind to Smad a and Smad b DBS (Fig. S2B). Additionally, pCAF was constitutively bound to these regions. However, no Smad binding was detected at the PP region in mouse epidermis, whereas it was readily detected at the TSLP PP region in MLE12 cells (Fig. S2B). Smad binding to its cognate DBS and PP was further enhanced by TGF-β treatment of the MLE12 cells (Fig. S2B). Smad2 or Smad3 binding to Smad DBS was not detected in Smad2^ep−/− or Smad3^ep−/− mice, respectively, confirming the specificity of ChIP antibodies (Fig. S2C). The functionality of Smads in regulating TSLP transcription was investigated by treating MLE12 cells with TGF-β and either a Smad3-specific inhibitor (SIS3) (12) or a TGF-β receptor–specific inhibitor (TGFR In) (13). Treatment with either SIS3 or TGFR In repressed the basal level of TSLP transcript by 25%, whereas TGF-β treatment stimulated TSLP transcript level by ∼30%, which was prevented by the inhibitors (Fig. 3B). The specificity of these inhibitors was ensured by determining Smad7 expression [a TGF-β target (14)], which showed the expected decrease (Fig. 3B).

Fig. 3.

TSLP transcription is controlled by multiple Smad, STAT, NF-κB, and AP1 DBS. (A) Schematic location of TF DBS present in the mouse TSLP gene. The underlined DBS bind their cognate TFs in ChIP assays using epidermis isolated from WT mice. (B) Q-RT-PCR of indicated transcripts from MLE12 cells treated for 6 h, as indicated. (C) ChIP assays using WT mouse epidermis, treated as indicated, to detect p65 binding to putative NF-κB DBS. (D) TSLP Q-RT-PCR from ears of WT mice treated for 6 h, as indicated. (E) ChIP assays using TPA-treated dorsal epidermis of WT and NF-κBa^−/− (Mut) mice to detect p65 binding to TSLP NF-κBa and NF-κBc DBS. (F) ChIP assays using WT mouse dorsal epidermis, treated as indicated, to detect c-Fos and c-Jun binding to AP1a and AP1b DBS. (G) ChIP assays using IL-1β– or TPA-treated MLE12 cells to reveal c-Fos and c-Jun binding to TSLP AP1b and AP1g DBS. (H) ChIP assays from TPA-treated WT mouse dorsal epidermis using indicated antibodies. (I) ChIP assays from WT mouse ileal and colonic epithelium using antibodies, as indicated, for TSLP NF-κBa and NF-κBc DBS. (J) ChIP assays from WT mouse ileal and colonic epithelium using antibodies, as indicated, for TSLP AP1b and AP1g DBS. All Q-RT-PCR values are mean ± SEM.

Fig. S2.

TSLP transcription is controlled by multiple Smad and STAT binding elements. (A) EMSA and antibody supershift analysis using WT mouse epidermis NE show the binding of Smad3 and Smad4 to Smad a and Smad b elements present in the TSLP gene. (B) ChIP assays using WT mouse dorsal epidermis and mouse MLE12 cells treated as indicated, to reveal binding of Smad2, Smad3, and Smad4 to Smad a, Smad b, and PP regions. (C) ChIP assays using dorsal epidermis of WT, Smad2^ep−/− and Smad3^ep−/− mutant mice to reveal Smad2, Smad3, and Smad4 binding to Smad a and Smad b elements. (D) ChIP assays using WT mouse dorsal epidermis, treated as indicated, to reveal STAT5 binding to TSLP STAT6 DBS. (E) ChIP assays using MLE12 cells, treated as indicated, to reveal STAT3, STAT5, and STAT6 binding to TSLP STAT DBS.

Seven consensus STAT6 DBS (STAT6a–STATg) and one consensus STAT5 DBS [STAT5a (15, 16)] are located in the TSLP gene (Fig. 5A and Table S1). Interestingly, ChIP assays revealed that STAT5 was bound in epidermis to STAT6 and STAT5a DBS only after 12-O-tetradecanoylphorbol-13-acetate (TPA) treatment of mouse skin (Fig. S2D), whereas no STAT6 binding was detectable at any of these DBS. Overnight MC treatment did not induce STAT5 or STAT6 binding; however, 3 d of MC treatment (once daily) resulted in STAT5 binding to several STAT DBS (Fig. S2D). Neither STAT5 nor STAT6 binding was detected after 3 d of RA treatment (Fig. S2D). These observations were further verified in MLE12 cells treated with VD3, IL-1β, or TPA, which revealed a constitutive association of STAT5 and STAT6 to STAT6a–STATd DBS (Fig. S2E). VD3 treatment did not alter this pattern. However, IL-1β treatment resulted in the disappearance of STAT5 and increased STAT6 binding to the same DBS (Fig. S2E). Taken together, these data indicate a redundancy among STAT6 DBS present on the TSLP gene for STAT5 and STAT6 binding, and also suggest a (likely) contribution of STAT5 and/or STAT6 in mediating TPA- and/or MC-induced TSLP transcription in mouse epidermis and MLE12 cells.

One consensus NF-κB and five imperfect NF-κB DBS are present in the mouse TSLP gene (Fig. 5A and Table S1). Lee and Ziegler (17) reported that the “upstream” NF-κBc DBS were instrumental in NF-κB–mediated expression of the mouse TSLP gene. In EMSA, using TPA-treated epidermal NE, we found that the mouse NF-κBa element bound the p65 and p50 NF-κB components much more efficiently than the NF-κBc element (Fig. S3A). The in vivo association of “NF-κBa and NF-κBc” DBS with the p65 subunit was tested by ChIP assays. In vehicle and MC- or RA-treated epidermis, p65 was not bound to any of the TSLP NF-κB DBS (Fig. 3C). This lack of binding was due to a lack of NF-κB activation; following TPA treatment, the p65 subunit was recruited to NF-κBa and NF-κBc DBS (Fig. 3C), with the binding to NF-κBa being much more efficient than the binding to NF-κBc (Fig. 3C). A functional role of NF-κB in activating TSLP expression in epidermis was revealed by cotreatment with TPA and (E)3-[(4-methylphenyl)-sulfonyl]-2-propenenitrile (BAY 11-7082; BAY) (a specific inhibitor of IKK-β activity; 18), which resulted in an ∼70% decrease in TSLP RNA level (Fig. 3D), in keeping with nuclear run-on assays showing that topical TPA treatment enhances the rate of TSLP transcription (Fig. 1A, lane 5). To determine whether NF-κB stimulated the TSLP promoter when activated by a physiologically relevant inducer, MLE12 cells were treated with IL-1β or TPA. The level of TSLP RNA was indeed increased (Fig. S3C), whereas p65 and p50 proteins did bind to NF-κBa DBS, and bound much less efficiently to NF-κBc DBS (Fig. S3D). A null mouse mutant (NF-κBa^−/− null, in which the NF-κBa DBS were deleted) was engineered to evaluate the function of these DBS in vivo. As expected, upon topical TPA treatment, no p65 binding was observed on the NF-κBa DBS of NF-κBa^−/− null mice (Fig. 3E), whereas the weak binding on the NF-κBc DBS was unaffected. However, there was no significant decrease in TPA-induced TSLP RNA synthesis in NF-κBa^−/− mice, most likely due to a concomitant TPA induction of AP1 activity (discussed below). Interestingly, Cultrone et al. (19) reported that the NF-κBa DBS are conserved in humans and play a crucial role in TSLP expression, whereas the NF-κBc DBS have a minor role. This preeminent role of the NF-κBa DBS has also been confirmed in the mouse by Negishi et al. (20), who also unveiled an interesting synergy in activation of TSLP expression by the IFN regulatory factor IRF3 and NF-κB via IRF DBS and NF-κB DBS located in close vicinity.

Fig. S3.

Regulation of TSLP transcription by multiple NF-κB and AP1 binding elements (DBS). (A) EMSA and antibody shift analysis using untreated and TPA-treated mouse epidermal NE to reveal the binding of p50 and p65 to various NF-κB elements, as indicated. (B) EMSA and antibody shift analysis using untreated and TPA-treated mouse epidermis NE show the binding of c-Fos and c-Jun to various AP1 elements, as indicated. (C) TSLP quantitative RT-PCR (Q-RT-PCR) using RNA from MLE12 cells, treated for 6 h, as indicated (mean ± SEM). (D) ChIP assays using MLE12 cells, treated as indicated, to detect p65 and p50 binding to the TSLP NF-κBa DBS (Upper, 30 cycles of PCR amplification) or to the NF-κBc DBS (Middle and Lower, 30 and 35 cycles of PCR amplification, respectively). (E) Aliquots of TPA-treated samples from D were immunoprecipitated with c-Jun and c-Fos antibodies and PCR-amplified using primers flanking the different AP1 elements present in the TSLP gene, as indicated. (F) ChIP assays using dorsal epidermis and ileal and colonic epithelium from WT mice to detect p65 and p50 binding to NF-κBa and NF-κBc elements, as indicated. (G) ChIP assays using dorsal epidermis and ileal and colonic epithelium from WT mice to detect c-Fos and c-Jun binding to AP1b and AP1g elements, as indicated.

The mouse TSLP gene contains four consensus (b, e, f, and g) and three imperfect (a, c, and d) AP1 DBS (Fig. 3A and Table S1). In EMSA and supershift assays with TPA-treated epidermal NE, both the consensus and imperfect AP1 elements equally bound c-Fos and c-Jun (Fig. S3B). In ChIP assays using epidermis from vehicle, MC-treated mice, or RA-treated mice, c-Fos and c-Jun were not bound to any of the TSLP AP1 DBS (Fig. 3F). However, upon TPA treatment, which is known to induce AP1 activity (11), ChIP assays using epidermal extract revealed that both c-Fos and c-Jun could bind to AP1 (b–d, f, and g) DBS (Fig. S3E). As expected, cotreatment of skin with TPA and the extracellular regulated kinase inhibitor U0126 or the Jun kinase inhibitor resulted in an ∼50% and 70% decrease in the level of TSLP transcript, respectively (Fig. 3D). Similar association of c-Fos and c-Jun with the TSLP AP1 DBS was also revealed by ChIP assays using MLE12 cells treated with IL-1β or TPA to activate c-Fos and c-Jun (Fig. 3G).

ChIP assays were performed using TPA-treated epidermis to identify coactivators present at NF-κB and AP1 DBS. Owing to the close proximity of the NF-κBa DBS (−237 bp) with the PP region, PCR with PP primers was used to detect factors present at both the NF-κBa and PP regions. TPA treatment resulted in the binding of p300, CDK7, Pol II, p65, and c-Jun to the PP region (Fig. 3H), in addition to pCAF, which was also detected in the absence of TPA (Fig. 2A). None of these factors except p65 (discussed above) were detected at the NF-κBc DBS. AP1b and AP1g DBS were also associated with the same factors, except that, in addition, HDAC2 was present at AP1g (Fig. 3H). No cofactors were detected at the AP1c, AP1d, and AP1f DBS, although they recruited c-Fos and c-Jun (Fig. S3E). Note that both DR3d and DR2b DBS were associated with their respective NRs and corepressors in the presence of TPA (Fig. 3H), indicating that dissociation of corepressors from these DBS is not a prerequisite for transcriptional activation of the TSLP promoter by NF-κB and AP1.

Most notably, both NF-κBa and NF-κBc DBS, as well as AP1b and AP1g DBS, were found to be constitutively associated with p65, p50 (Fig. S3F), and c-Fos and c-Jun (Fig. S3G; also Fig. 5C), respectively, in the IEC of WT mouse. Moreover, NF-κBa-, AP1b-, and AP1g-containing regions were also constitutively associated with coactivator p300 and Pol II (Fig. 3 I and J), indicating that these regions are transcriptionally active.

Treatments with VD3, RA, and TPA Induce Chromatin Loops Between the Regions Containing Their Respective Response Elements and the PP Region of the TSLP Gene.

Direct interactions between regions containing DBS for activating complexes and PP regions are known to be instrumental in initiation of transcription (21, 22). Because DR3d and PP; DR2b and PP (Fig. 2A); and AP1b, AP1g, and PP (Fig. 3H) regions displayed similar cofactor dynamics in the presence of their respective agonists, we hypothesized that the DR3d, DR2b, AP1b, and AP1g DBS located in the upstream regulatory region may be interacting through a chromatin loop with the PP region in an activation-dependent manner. A chromatin conformation capture (3C) assay was performed on epidermal chromatin of mice topically treated with various activating compounds to test this possibility. Chromatin was digested with the AluI enzyme to separate fragments encompassing regions of interest (Fig. 4A), which were then ligated to reveal possible interactions between them.

Fig. 4.

MC, RA, or TPA treatment induces an interaction between the different regions that contain the respective cognate factor DBS (VDR, RARs, or AP1) and the PP region of the TSLP gene. (A) Mouse TSLP promoter and upstream region. The A base of the translation initiation codon (ATG) is represented by +1, and a–j are AluI-digested DNA fragments. The numbers 1–20 denote AluI sites. Boxes represent DBS for the indicated TFs with their coordinates. (B) Using skin epidermis from WT and VDR^ep−/− mice treated as indicated, the 3C assays reveal the interaction between the regions containing DR3d and PP and DR3f/DR3g and PP. (C) Using dorsal skin epidermis of WT and RAR-γ^−/− mice, treated as indicated, the 3C assays reveal the interaction between the DR2b and PP regions. (D) Using WT mouse dorsal epidermis, treated as indicated, the 3C assays detect a possible interaction between regions that contain different AP1 DBS and the PP regions, as indicated. D, Alu I digested; FPCR + D, final PCR product digested with AluI before Southern blotting; L, ligated; NC, not cross-linked; ND, not digested.

In the absence of MC, no interaction was observed between the DR3d and PP regions, whereas MC treatment resulted in a clear interaction between them in the WT mice, and, as expected, no interaction between the DR3d and PP regions was observed in VDR^ep−/− mice treated with MC (Fig. 4B). Moreover, DR3f and/or DR3g, which did not show any coactivator or Pol II binding (Fig. 2A), failed to interact with the PP region (Fig. 4B). Similarly, 3C assays using a DR2b region-specific probe revealed a selective interaction between the DR2b and PP regions in the presence of RA (Fig. 4C). Interestingly, 3C assays performed to determine the interaction between different AP1 element-containing regions with the PP region revealed that only the AP1b and AP1g elements were associated with the PP region upon TPA treatment (Fig. 4D).

Discussion

Several NRs Differentially Control the Expression of TSLP in Epidermal Keratinocytes and IEC.

TSLP is an important immune-regulatory cytokine that is expressed in several cell types, including the epithelial cells at barrier surfaces, such as the skin, lung, and intestine. It acts on cells of both myeloid and lymphoid lineages to promote T helper 2 (Th2) differentiation and Th2 cytokine-associated inflammation (23, 24). Constitutive TSLP expression in IEC confers tolerance against commensals (25). IEC-expressed TSLP also plays a critical role in mediating the recovery from colonic inflammation (26). In contrast, exacerbated expression of TSLP in keratinocytes, the lungs, and the esophagus correlates with the onset of various allergic diseases, such as AD, asthma, and food allergy-associated eosinophilic esophagitis (6, 27, 28). Collectively, these data illustrate various physiological and pathophysiological roles of TSLP throughout life, thus suggesting that strict spatiotemporal control of TSLP expression is essential for maintaining homeostasis. Our previous studies (1–3), as well as the studies of other groups (17, 29), have generated preliminary evidence regarding the mechanisms that control TSLP expression. In the present study, we have performed a detailed analysis of the cis-acting regulatory elements that control TSLP transcription in vivo in mouse epidermal keratinocytes and IEC. Our results unequivocally establish the role of the NRs VDR and RAR-γ in controlling TSLP transcription in keratinocytes of mouse epidermis. Unliganded heterodimers of RXR-α and VDR, together with unliganded heterodimers of RXR-β and RAR-γ, associated with the corepressors SMRT and HDAC2 are constitutively associated with their cognate DBS located in the TSLP upstream regulatory region (Fig. 2A). This mode of active repression by two different NR-associated repressor complexes ensures a tight control over TSLP expression, which is crucial for maintaining skin homeostasis.

Most interestingly, our study also reveals that the function of the multiple RXR and RAR isotypes is not redundant in vivo when bound as heterodimers to either a given DBS in different genes in a given tissue or the same DBS in different tissues (Fig. 5 B and C). In epidermal keratinocytes, and irrespective of the presence of their cognate ligands, VDR is heterodimerized with RXR-α on the DR3d, DR3f, and DR3g VDREs of the TSLP promoter, whereas RAR-γ is heterodimerized with RXR-β on the DR2b RARE of the TSLP gene and heterodimerized with RXR-α on the DR2b RARE of the CRABP II gene (Figs. 1 F and H and 2A). Derepression of TSLP expression via keratinocyte-selective ablation of either RXR-α and RXR-β (RXR-αβ^ep−/− mutants) or RAR-γ(α) and VDR (RAR-γ/VDR^ep−/− or RAR-γα/VDR^ep−/− mutants) results in the release of repressor complexes from their respective DBS and recruitment of the transcriptional machinery, thus initiating TSLP transcription (Fig. 1A). The role of RAR-α remains elusive, because we did not detect its binding to the DR2b DBS or an increase in TSLP transcript level upon application of a RAR-α–specific agonist (Fig. 1G).

Analysis of IEC revealed that the binding pattern of VDR and RXR-α on the DR3 VDREs is the same as in epidermis (Fig. S1D). However, in contrast to epidermis, no RAR-γ/RXR-α heterodimers are bound to the DR2b RARE; instead, they are bound to the DR2a RARE (Figs. 1I and 5 B and C). Along the same lines, although no PPAR isotypes in epidermis were found to be associated with any of the TSLP DR1 DBS, PPAR-γ was associated with all three DR1 DBS in IEC, where TSLP transcription was induced by rosiglitazone (Figs. 1 J–L and 5 B and C). These data demonstrate the complexity of TSLP transcriptional regulatory mechanisms and illustrate in vivo tissue-specific variations of the binding of a particular NR-isotype to cognate DBS. It would be of interest to explore through which epigenetic mechanism the accessibility of RAR-γ/RXR-α to the DR2a and DR2b sites, as well as the accessibility of RXR-α/PPAR-γ to DR1 DBS, is differentially controlled in epidermis and IEC.

Even though unliganded VDR/RXR-α–associated repressor complexes could be detected on all three DR3d, DR3f, and/or DR3g DBS, only DR3d recruited a coactivator complex upon MC treatment (Fig. 2A), which raises important questions regarding the physiological role of these DR3 DBS. Does only DR3d modulate TSLP transcription and do DR3f/g-bound VDRs behave only as repressors, or do they recruit coactivators under particular instances? It remains to be investigated whether epigenetic mechanisms (e.g., histone modifications, DNA methylations) prevent the recruitment of coactivators to the DR3f/g DBS. Note that the DR3f/g region also does not appear to interact with the PP region in the presence of MC (Fig. 4B), whereas fragments encompassing the DR3d and PP regions interact with each other (Fig. 4B), and, furthermore, identical sets of coactivators and Pol II are detected on both the DR3d and PP regions in VD3-treated samples (Fig. 2A), therefore suggesting the formation of an “activating” loop between the DR3d and PP regions (Fig. 4B). Note that, in IEC, where VD3 is synthesized (30), the DR3d DBS are constitutively functional (Fig. S1E). Similarly, topical treatment of mouse epidermis with RA induced the transcription of TSLP, which was accompanied by the formation of a loop resulting from interaction between the chromatin fragment encompassing the DR2b DBS and PP regions, and by the presence of an identical set of coactivators and Pol II at both regions (Figs. 2A and 4C). Interestingly, we have shown that both loops are destroyed upon binding of the liganded glucocorticoid (GC) receptor (GR) to the TSLP inverted repeat negative GR element (IRnGRE) (11). Finally, it is puzzling that even though both VD3 and RA induce TSLP synthesis in the epidermal suprabasal layers (2), only the SRC2 coactivator was recruited to the DR2b and PP regions upon RA treatment, whereas both SRC2 and SRC3 coactivators were recruited to the DR3d and PP regions in VD3-treated epidermis (Fig. 2A).

NF-κB, AP-1, STAT, and Smad Transactivators Are also Differentially Involved in the Control of TSLP Expression.

We have characterized multiple NF-κB, AP1, STAT, and Smad DBS in the sequences upstream and downstream of the mouse TSLP start site. We detected “constitutive” binding of Smads (Smad2–Smad4) to the Smad DBS in the TSLP gene in WT mice and in MLE12 cells (Fig. S2 B and C). However, their functional relevance in regulating TSLP transcription in vivo remains unclear. We have also shown that STAT5 and STAT6 are differentially recruited to STAT DBS depending on the stimulus (Fig. S2 D and E). Interestingly, it has been shown that TSLP induces STAT5 activity in cultured cells (31). Could TSLP be involved in regulating STAT5 activity at the TSLP promoter? Identification of multiple functional STAT DBS in the TSLP promoter region may point to the prevalence of a positive feedback loop that drives TSLP transcription in a STAT-dependent manner.

We have shown that ablation of the NF-κBa element within the TSLP promoter prevented the binding of the NF-κB p65 and p50 proteins. However, the actual contribution of this binding to the TPA-induced TSLP transcription could not be assessed because the NF-κBc and AP1 DBS remained “active” in the NF-κBa mutant. Analysis of the various AP1 DBS demonstrated that even though all of them are consensus DBS, only five of them (AP1b, AP1c, AP1d, AP1f, and AP1g) did associate with c-Fos and c-Jun in a TPA-dependent manner (Fig. S3E), and of these five, only the AP1b and AP1g DBS could be “functional” upon TPA treatment in epidermis through the formation of a chromatin loop with the PP region (Fig. 4D), thereby suggesting that only these two DBS are functional in epidermis. Interestingly, we have shown that in IEC, the microbiota-elicited Toll-like receptor signaling is the major determinant of both the NF-κB and AP-1 activity, and also that a reduction of this signaling leads to a decrease in TSLP expression in IEC (30). Additionally, microbiota-derived signaling is known to regulate RA synthesis in the intestine (32), and we have recently found that a reduction in the microbiota signaling also reduces the expression of 25-(OH)D3 1-α-hydrodxylase (Cyp27B1, the rate-limiting enzyme in VD3 synthesis pathway) in IEC, thereby decreasing TSLP expression [by “inactivating” the DR3d DBS (30)]. Altogether, it appears that microbe-derived signaling in the “gut” is the major determinant of TSLP expression.

Taken together with other recent reports (11, 17, 20, 30), our present study provides an overall topological and functional map of the TSLP gene cis-acting regulatory elements, which are targeted by numerous signaling pathways to fine-tune the spatiotemporal regulation of TSLP expression, which is differentially exerted in two important epithelial tissues throughout life.

Methods

Mice.

WT C57BL/6J mice, 6–8 wk old, were purchased from Charles River Laboratories. RXR-α^ep−/−, RXR-β^ep−/−, RXR-αβ^ep−/−, RAR-γ^−/−, and VDR^ep−/− are described (1, 2). RAR-γ/VDR^ep−/− and RAR-αγ/VDR^ep−/− were obtained by i.p. injection of tamoxifen (Tam) at a dose of 0.1 mg⋅d⁻¹, whereas PPAR-γ^iec−/− was obtained by i.p. injection of Tam (1 mg⋅d⁻¹). For each case, Tam was injected for consecutive 5 d. Floxed smad2 and smad3 mice have been described (33). These mice were crossed with the K14 CreER^T2 line to obtain floxed K14 CreER^T2 smad2 and floxed K14 cre ER^T2 smad3 animals. Smad2^ep−/− and Smad3^ep−/− mice were obtained by i.p. injection of Tam (1 mg⋅d⁻¹) for 5 d to the respective floxed animals. Age- and sex-matched littermates were used as WT controls. Breeding, maintenance, and experimental manipulations were approved by the Animal Care and Use Committee of the Institut de Génétique et de Biologie Moléculaire et Cellulaire/Institut Clinique de la Souris (ICS).

ChIP, Nuclear Run-On, and 3C Assays.

ChIP assays were performed using chromatin prepared from 1% formaldehyde cross-linked keratinocytes and IEC, using indicated antibodies. Regions of interest were PCR-amplified using specific primers. Nuclear run-on was performed from nuclei isolated from epidermis, which were incubated in run-on buffer containing α-[³²P] UTP; following the transcription reaction, the nascent RNA was extracted and hybridized with DNA probes. The 3C assays were performed from formaldehyde cross-linked NE in epidermis, which was digested with AluI, and subsequently ligated and PCR-amplified to reveal the possible interactions. Detailed procedures, all primers, and probes are listed in SI Materials and Methods.

Quantitative RT-PCR, Serum TSLP Determination, Hematoxylin/Eosin Staining, and TSLP Immunohistochemistry.

Quantitative RT-PCR was performed from the total RNA isolated from epidermis or IEC. Following isolation, RNA was reverse-transcribed to generate cDNA, which was used to detect indicated molecules. Paraformaldehyde-fixed ears were sectioned and stained with hematoxylin/eosin to reveal histopathological abnormalities. These sections were also immunoprobed with a specific antibody to detect TSLP protein in epidermis. Details are provided in SI Materials and Methods.

SI Materials and Methods

Mice and Treatments.

Transgenic mice lacking the NF-κBa DBS of TSLP promoter were generated by constitutive germline deletion of the corresponding sequence “GGGAAATTCC.” Mutant mice were genotyped by PCR using primers (forward: 5′ CCTAAAGTGACTGCGGGTGT 3′ and reverse: 5′ CCAGTGTTCTCTGCCTCCAT 3′). Deletion of the NF-κBa DBS in the mutant DNA was confirmed by sequencing of the genotyped PCR product. Age- and sex-matched littermates were used as WT controls. For topical treatment on ear and dorsal skin, 1 nmol/cm² of MC or TPA; 2 nmol/cm² of RA, BMS961, or BMS753; 500 nmol/cm² of BAY; 125 nmol/cm² of U0126; and 500 nmol/cm² of Jun kinase inhibitor (JNKI) were used. Rosiglitazone (50 mg/kg daily) was administered i.p. for 4 d.

Materials.

P65 (SC-372X), P50 (SC-1192X), c-Jun (SC-44X), c-Fos (SC-52X), STAT3 (SC-482X), STAT5 (SC-835X), STAT6 (SC-981X), Smad2 (ab71109), Smad3 (SC-8332X), Smad4 (SC-7154X), c-ski (SC-9140X), SnoN (SC-9141X), P300 (SC-584X), pan-RAR (SC-773X), HDAC3 (SC- 11417X) and SRC3 (SC-9119X) rabbit polyclonal antibodies were from Santa Cruz Biotechnology. Control rabbit IgG (ab46540), SRC1 (ab84) mouse monoclonal, HDAC1 (ab7028), and HDAC2 (ab7029) rabbit polyclonal antibodies were from Abcam. RNA Pol II (05-623) mouse monoclonal antibody was from Millipore Corporation. SMRT, NCoR, TIF2 (SRC2), pCAF, CDK7, and PPAR-γ mouse monoclonal antibodies, and VDR rabbit polyclonal antibodies were generated in-house. All antibodies generated in-house were tested by ELISA, immunoprecipitation, Western blotting, and absence of reaction in null mutant mice to ensure specificity. All other antibodies were tested by immunoprecipitation and Western blotting. Active VD3, RA, and BAY 11-7082 were from Sigma–Aldrich. JNKI II, TPA, TGF-β, SIS3, TGFR inhibitor, and fenofibrate (PPAR-α agonist) were from Calbiochem. Rosiglitazone (PPAR-γ agonist) was from Tocris. MC and recombinant human IL-1β were from Leo Pharma and R&D Systems, respectively. BMS961 and BMS753 were as described (8).

ChIP Assay.

ChIP assay was as carried out as described (11), with a few modifications. Epidermal cells from mouse dorsal skin or IEC prepared from ileum and colon (30) were cross-linked by adding formaldehyde to a final concentration of 1% and incubated on a flip-flop rocker for 10 min at 25 °C, followed by addition of 2 M glycine (0.125 M final concentration) and incubation for 5 min. Cells were pelleted (400 × g, 5 min) at 4 °C and washed twice in ice-cold PBS. Cross-linked mouse epidermal cells or A549 or MLE12 cells were resuspended in 300 μL of ice-cold lysis buffer [50 mM Tris⋅HCl (pH 8.1), 1% (wt/vol) SDS, 10 mM EDTA, 1× protease inhibitor mixture (Roche)], and all lysates were sonicated at 4 °C to generate 300- to 500-bp chromatin fragments (Bioruptor sonicator; Diagenode). Cellular debris was removed by spinning (10,000 × g, 10 min) at 4 °C, and the supernatant was precleared by incubating with 60 μL of protein G-agarose beads (preblocked with salmon sperm DNA and BSA) for 20 min at 4 °C over a flip-flop rocker. Beads were pelleted (100 × g, 1 min) at 4 °C and discarded. Equal amounts of lysates were used for immunoprecipitation. Ten percent of the lysate amount used for each immunoprecipitation reaction was saved to be used as input DNA in PCR amplification. Individual samples were diluted 1:8 (vol/vol) in ChIP dilution buffer [16.7 mM Tris⋅HCl (pH 8.1), 0.01% (wt/vol) SDS, 1.1% (vol/vol) Triton-X 100, 1.2 mM EDTA, 16.7 mM NaCl, 1× protease inhibitor mixture] and incubated overnight with different primary antibodies at 4 °C on a flip-flop rocker, followed by incubation with 60 μL of protein G-agarose beads (preblocked with salmon sperm DNA and BSA) for 90 min. Immunoabsorbed complexes were recovered by centrifugation (100 × g, 1 min) at 4 °C. Protein G beads were then washed once in low-salt buffer [20 mM Tris⋅HCl (pH 8.1), 0.1% (wt/vol) SDS, 1% (vol/vol) Triton X-100, 2 mM EDTA, 150 mM NaCl], once in high-salt buffer [20 mM Tris⋅HCl (pH 8.1), 0.1% (wt/vol) SDS, 1% (vol/vol) Triton T-100, 2 mM EDTA, 500 mM NaCl], once in LiCl buffer [10 mM Tris⋅HCl (pH 8.1), 250 mM LiCl, 1% (vol/vol) Nonidet P-40, 1% (wt/vol) sodium deoxycholate, 1 mM EDTA], and finally twice in 1 mL of TE buffer [10 mM Tris⋅HCl (pH 8.0), 1 mM EDTA]. Chromatin was released from the beads by incubation with 150 μL of elution buffer [1% (wt/vol) SDS, 100 mM NaHCO₃] for 15 min at room temperature with intermittent vortexing. The elution step was repeated, and both eluates were pooled. One microliter of 10 mg/mL RNase A and 5 M NaCl (200 mM final concentration) were added to the eluate and incubated at 65 °C overnight. One microliter of 30 mg/mL Proteinase K was then added to the samples and incubated at 50 °C for 1 h. DNA was subsequently purified from the eluate using a Qiagen PCR Purification Kit (Qiagen Corp.) in a final volume of 50 μL. Three microliters of the purified DNA was used as a template in different PCR amplifications (Applied Biosystems Thermocycler). The sequence of the various primers is listed below. The number of PCR cycles was optimized to maintain linear amplification in all experiments. PCR products were resolved by 2% (wt/vol) agarose gel electrophoresis. Images of ethidium bromide-stained DNA were acquired using an UV transilluminator equipped with a digital camera.

ChIP assay primers

Names of ChIP assay primers	Sequences of ChIP assay primers
mTSLP DR1a fp	5′TCTAAAAATCCCTGAAGGGGGCC
mTSLP DR1a rp	5′CCCCTGTAGGGAACCCACTTCTTT
mTSLP DR1b fp	5′TATGTCTCTTTGGGCGCATT
mTSLP DR1b rp	5′TTTCAGCTAGCAGGGGAATC
mTSLP DR1c fp	5′TTCCAGTCTTTCTTGGTTTGG
mTSLP DR1c rp	5′GCATGGTCATTGTCCTGAGA
mTSLP DR2a fp	5′CAACCAGGGGAGCGCAAATCTT
mTSLP DR2a rp	5′CAGCCGCCCTCGAAACTCATTA
mTSLP DR2b fp	5′TGAGGTATTTTATCAGAACAATGGAC
mTSLP DR2b rp	5′CCAAGTGCTGGGATTAAAGG
mTSLP DR3a fp	5′GAAAAATGAAGTTCAGGACCAA
mTSLP DR3a rp	5′TCTCCCTCCTCTTTCTCCTTC
mTSLP DR3b fp	5′CACCTTTAATCCCAGCACCt
mTSLP DR3b rp	5′TTGCTTCTTTAATTGGTCCTGAA
mTSLP DR3c fp	5′GCCTTTAATCCCAGCACTTG
mTSLP DR3c rp	5′CTTGACTCCCATGAAAGATGC
mTSLP DR3d fp	5′AGGTCCAGATACTGCATGCTC
mTSLP DR3d rp	5′CAGCAGCTGATGCAAACAGA
mTSLP DR3e fp	5′AAATACCTGGGCGGTGGT
mTSLP DR3e rp	5′GTTTCCAGTGTTAAATTACATTGTTT
mTSLP DR3f and DR3g fp	5′AGACCAAAGCAGACTCCATGA
mTSLP DR3f and DR3g rp	5′CGACCTAGGATGGAACCAAA
mTSLP PP/NF-κBa fp	5′GCATGTAGCAAGTGTTTAGGGCAGA
mTSLP PP/NF-κBa rp	5′GCCTGAACGTGGAGTCTTCCTGAT
mTSLP NF-κBb fp	5′CACCTCTGACTCCAGTCTGT
mTSLP NF-κBb rp	5′GTACTGGGAGCAGGAAAGAA
mTSLP NF-κBc/NF-κBd fp	5′AGGAGGGAGAAGTAAACAAAGC
mTSLP NF-κBc/NF-κBd rp	5′CTCACTGCTGCCACCTTCTT
mTSLP NF-κBe fp	5′AGAAGAGAAAGATGAGAGAA
mTSLP NF-κBe rp	5′CTAGAAAAACATTCACTTGC
mTSLP NF-κBf fp	5′ATGTAACTAGCAACTGAAGATC
mTSLP NF-κBf rp	5′TGAGACAGCCAAATTTCTAT
mTSLP AP1a fp	5′AGGAGAGTAGGGGTTGGGTTCGTGT
mTSLP AP1a rp	5′AGGGGAGGAACAGCTTCTGACCTGT
mTSLP AP1b fp	5′GTTACCACCTCCGTGGCTACT
mTSLP AP1b rp	5′TATCCACAGGTTATTTTGGGAAA
mTSLP AP1c fp	5′TTGCTTCCTTGTGATCTTGC
mTSLP AP1c rp	5′AAAGGAGAGTGGCCATGGTA
mTSLP AP1d fp	5′CATCCTTTCCCCTTCACTCA
mTSLP AP1d rp	5′AGCTCACACCCGAATAATCAA
mTSLP AP1e fp	5′GCCCATCCTTGTTGTGGATA
mTSLP AP1e rp	5′TGAGACCCCTTAGAACTTGTGG
mTSLP AP1f fp	5′TGATAAATCATCCCTTATTTTCCAA
mTSLP AP1f rp	5′AGAAGACATAGGCCAATTCACA
mTSLP AP1g fp	5′TGCCAAAGTTTCCCATTCTC
mTSLP AP1g rp	5′TGTGATAGTGGCCCAAACAT
mTSLP STATa fp	5′GGCAGGACTTTGCCTTCTTT
mTSLP STATa rp	5′CTGACGTTTTCTGGCTCACA
mTSLP STATb fp	5′CCTCTGAGTCCCAAGAGCTG
mTSLP STATb rp	5′CTGGCAGTGTTCCTCTTCGT
mTSLP STATc fp	5′TGGAAGTCCTGCCTGGTTAC
mTSLP STATc rp	5′AGAATCACATGGCTGCTGGT
mTSLP STATd fp	5′GACATGCATCTCTTCTGCTTTC
mTSLP STATd rp	5′GTCCCTTCACTTCCCTGTCA
mTSLP STATe fp	5′TCAAAGTCCCTTGGTACCTAACAT
mTSLP STATe rp	5′TCCTGTTTTCTCCTCCCAAA
mTSLP STATf fp	5′CCAAGAGCCACGACTTTACTG
mTSLP STATf rp	5′GGTGTGGCCTTGTTAAAGGA
mTSLP STATg fp	5′GCACTACTCCATTGCTGGTG
mTSLP STATg rp	5′TGGGGCATCTTTTGTTTTGT
mTSLP STATh fp	5′GGGTGGATCCCAACATAAGA
mTSLP STATh rp	5′TGTGCCAAGTTATTCCACCA
mTSLP Smad a fp	5′GGGTTTCATTGCTGTGAAGAG
mTSLP Smad a rp	5′TGGCACTATTAGGAGGTGTGG
mTSLP Smad b fp	5′CATTGCTGTGAAGAGACACCA
mTSLP Smad b rp	5′GGAGTGGCACTATTAGGAGGTG
mCRABPII DR2 fp	5′CCTTCCTCCCTCATCCTCTC
mCRABPII DR2 fp	5′AACAACGTTGCTCTGTGCTG

fp, forward primer; rp, reverse primer.

Nuclear Run-On Assay.

The pSK⁺ empty vector and pSK⁺ β-actin (2 kb of β-actin cDNA cloned into pSK⁺ vector at EcoRI and BamHI sites) plasmids were linearized by digestion with BamHI. A 4-kb region of the mouse TSLP gene-coding sequence was amplified by PCR from a BAC construct (Invitrogen) [forward primer: 5′-AGTTCTTCTCAGGAGCCTCTTCATC-3′ (+165 to +190 with respect to “A” of the translation initiation codon ATG as +1), reverse primer: 5′-CCAGATTCCACAATCTTCTTTCAGA-3′ (+4,157 to +4,181 with respect to A of ATG as +1)]. Approximately 10 μg each of pSK⁺ empty vector, pSK⁺ β-actin, and TSLP DNA was separated by 0.8% agarose gel electrophoresis, and transferred to 0.45-μm nitrocellulose membranes. Membranes containing the transferred DNA probes were baked for 2 h at 80 °C in a vacuum oven, followed by washing in 2× SSC.

The run-on assay was as described (11). Approximately 10⁸ epidermal cells were resuspended in 5 mL of ice-cold Nonidet P-40 lysis buffer [10 mM Tris⋅HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl₂, 0.5% (vol/vol) Nonidet P-40] and incubated for 5 min on ice. Nuclei were pelleted (100 × g, 10 min) at 4 °C. The nuclear pellet was resuspended in 250 μL of nuclear freezing buffer [50 mM Tris⋅HCl (pH 8.3), 40% (vol/vol) glycerol, 5 mM MgCl₂, 0.1 mM EDTA]. Two hundred twenty-five microliters of the nuclei was added to 60 μL of 5× run-on buffer [25 mM Tris⋅HCl (pH 8.0); 12.5 mM MgCl₂; 750 mM KCl; 12.5 mM each of ATP, GTP, and CTP]. Then, 150 µCi of α-[³²P] UTP was added and the reaction was incubated at 37 °C for 30 min. Twenty microliters of 10 mM CaCl₂ and 10 μL of 1 mg/mL RNase-free DNase were added and incubated at 30 °C for 5 min, followed by addition of 35 μL of proteinase K buffer [10% (wt/vol) SDS, 50 mM EDTA, 10 mM Tris⋅HCl (pH 7.4), 3 mg/mL proteinase K] and incubation at 37 °C for 45 min. The reaction mixture was extracted twice with phenol-chloroform and once with chloroform, followed by 1:1 dilution with 5 M ammonium acetate and precipitation with isopropanol. The precipitated pellet was resuspended in 100 μL of TE (pH 8.0), followed by addition of 100 μL of solution containing 10 mM MgCl₂ and 5 mM CaCl₂. Ten microliters of 1 mg/mL RNase-free DNase was added, and the reaction mixture was incubated at 37 °C for 5 min, followed by chilling on ice for 5 min. Fifty microliters of 1 M NaOH was added, and the reaction mixture was incubated on ice for another 2 min. Alkali were neutralized by addition of 77 μL of 1 M Hepes (free acid), followed by precipitation of RNA in isopropanol and resuspension in 100 μL of TE. Ninety microliters of the radiolabeled RNA was hybridized to the nitrocellulose filter containing the above DNA probes for 24 h at 65 °C in 2 mL of hybridization buffer [6× SSC, 5× Denhardt’s reagent, 50% formamide, 0.5% (wt/vol) SDS, 200 μg of salmon sperm DNA, 2 μg of polyA]. Membranes were washed four times in 2× SSC, and excess liquid was removed; membranes were then wrapped in Saran wrap and exposed to X-ray films for 48 h in the presence of an intensifying screen at −70 °C.

EMSA.

Synthetic oligonucleotides encompassing different DNA elements were annealed, and double-stranded oligonucleotides were purified by 15% polyacrylamide gel electrophoresis. Fifty nanograms of purified oligonucleotides was 5′-end-labeled using γ-[³²P] ATP, which was removed by purifying the probe through a Sephadex G-50 column. NE was prepared as described (11). Eight femtomoles of radiolabeled probe was mixed with 10 μg of WT mouse epidermal NE in EMSA buffer [20 mM Hepes (pH 7.9), 150 mM KCL, 5 mM MgCl₂, 2 mM DTT, 10% (vol/vol) glycerol, 2 μg of poly(deoxyinosinic-deoxycytidylic) acid (poly dI-dC), 10 μg BSA] and incubated for 20 min at room temperature. For the supershift assay, specific antibody or rabbit IgG was added to the reaction mixture, and incubation was continued further at room temperature for 20 min. Samples were electrophoresed on a 4.5% nondenaturing polyacrylamide gel in TBE buffer [90 mM Tris⋅borate, 2.5 mM EDTA (pH 8.3)]. Vacuum-dried gels were exposed to X-ray film overnight at −70 °C in the presence of an intensifying screen.

EMSA oligonucleotides

Names of EMSA oligonucleotides	Sequences of EMSA oligonucleotides
DR1 cons. U	5′TTTCAGGTCATAGGTCAAAAC
DR1 cons. L	5′GTTTTGACCTATGACCTGAAA
DR1a U	5′ATGGGGGTCAGGGGACAGGTT
DR1a L	5′AACCTGTCCCCTGACCCCCAT
DR1b U	5′TCAGGATGGCAAAGGTCAGAGAG
DR1b L	5′CTCTCTGACCTTTGCCATCCTGA
DR1c U	5′TGTGTAGGTAAGATGTCACGGTT
DR1c L	5′AACCGTGACATCTTACCTACACA
DR2 cons. U	5′TTTCAGGTCAATAGGTCAAAAC
DR2 cons. L	5′GTTTTGACCTATTGACCTGAAA
DR2a U	5′ACACAGCTCAACAGGTCAGAA
DR2a L	5′TTCTGACCTGTTGAGCTGTGT
DR2b U	5′TTCCGAAGGTCATGAGTTCAAATCCC
DR2b L	5′GGGATTTGAACTCATGACCTTCGGAA
DR3 cons. U	5′AGCTTCAGGTCAAGGAGGTCAGAGAGC
DR3 cons. L	5′GCTCTCTGACCTCCTTGACCTGAAGCT
DR3a U	5′ACATGTAGTTCATTCCGGTCATCCCAG
DR3a L	5′CTGGGATGACCGGAATGAACTACATGT
DR3b U	5′AGTTCCAGGACAGCCAGGGCTACACAG
DR3b L	5′CTGTGTAGCCCTGGCTGTCCTGGAACT
DR3c U	5′AGTTTCAGGACAGCCAGGGCTACACAG
DR3c L	5′CTGTGTAGCCCTGGCTGTCCTGAAACT
DR3d U	5′AGTTCCAGGACAGCCAGGGCTATACAG
DR3d L	5′CTGTATAGCCCTGGCTGTCCTGGAACT
DR3e U	5′AGTTCCAGGACAGCCAGGGCTATACAG
DR3e L	5′CTGTATAGCCCTGGCTGTCCTGGAACT
DR3f/g U	5′TGTTCTTGACCCCTCTGGCTCCTACAA
DR3f/g L	5′TTGTAGGAGCCAGAGGGGTCAAGAACA
NF-κB cons. U	5′AGTTGAGGGGACTTTCCCAGGC
NF-κB cons. L	5′GCCTGGGAAAGTCCCCTCAACT
NF-κB mut. U	5′AGTTGAGGCGACTTTCCCAGGC
NF-κB mut. L	5′GCCTGGGAAAGTCGCCTCAACT
NF-κBa U	5′AAAAAGGTGAGGGAAATTCCTGATGATTTTGC
NF-κBa L	5′GCAAAATCATCAGGAATTTCCCTCACCTTTTT
NF-κBb U	5′GCAGTGACTCCTAGGAACTTCCCGAAGACGAGAC
NF-κBb L	5′GTCTCGTCTTCGGGAAGTTCCTAGGAGTCACTGC
NF-κBc U	5′CTCTTCAGCCCACAGGAATTTCCAAAAATATACC
NF-κBc L	5′GGTATATTTTTGGAAATTCCTGTGGGCTGAAGAG
NF-κBb U	5′CTCCCTGGGTAAGAAAATTACCAAACTCCCTG
NF-κBd L	5′CAGGGAGTTTGGTAATTTTCTTACCCAGGGAG
NF-κBe U	5′AGGATGTGACAGGAGAATCCCCACAACCAGGAGG
NF-κBe L	5′AGGATGTGACAGGAGAATCCCCACAACCAGGAGG
NF-κBf U	5′TTACTACAAAGAAGGGACTCTCCACCCATTTCTTC
NF-κBf L	5′GAAGAAATGGGTGGAGAGTCCCTTCTTTGTAGTAA
AP1 cons. U	5′CGCTTGATGACTCAGCCGGAA
AP1 cons. L	5′TTCCGGCTGAGTCATCAAGCG
AP1 mut. U	5′CGCTTGATGACTTGGCCGGAA
AP1 mut. L	5′TTCCGGCCAAGTCATCAAGCGAGAAAGAATGACATCAAGGAAAAT
AP1a U	5′AGAAAGAATGACATCAAGGAAAAT
AP1a L	5′ATTTTCCTTGATGTCATTCTTTCT
AP1b U	5′AGTTTTCTTTAATCAAGCTTAGGC
AP1b L	5′GCCTAAGCTTGATTAAAGAAAACT
AP1c U	5′GATTGATGACTCAGCAGTTA
AP1c L	5′TAACTGCTGAGTCATCAATC
AP1d U	5′CAAACTTGACTCAGACACTT
AP1d L	5′AAGTGTCTGAGTCAAGTTTG
AP1e U	5′ATCTCATGACTCAGTGGAAA
AP1e L	5′TTTCCACTGAGTCATGAGAT
AP1f U	5′ACACGGTGACTCACAGCCAT
AP1f L	5′ATGGCTGTGAGTCACCGTGT
AP1g U	5′GATGGCTGAGTCAGACACCA
AP1g L	5′TGGTGTCTGACTCAGCCATC
Smad cons. U	5′GGGTGTCTAGACGGCC
Smad cons. L	5′GGCCGTCTAGACACCC
Smad a U	5′TAGAGGCCCACAGCCAGACATTAGGTGGAG
Smad a L	5′CTCCACCTAATGTCTGGCTGTGGGCCTCTA
Smad b U	5′TTAGAGGCCCACAGCCAGACATTAGGTGGA
Smad b L	5′TCCACCTAATGTCTGGCTGTGGGCCTCATA

cons., consensus; L, lower strand; mut., mutant; U, upper strand.

The 3C Assay.

The 3C assay was carried out as described (22). Epidermal cells were cross-linked by adding formaldehyde to a final concentration of 2%, and incubation was performed on a flip-flop rocker for 10 min at 25 °C, followed by addition of 2 M glycine (0.125 M final concentration) and incubation for 5 min. Cells were pelleted (400 × g, 5 min) at 4 °C, washed twice in ice-cold PBS, and resuspended in lysis buffer [10 mM Tris⋅HCl (pH 8.0), 10 mM NaCl, 0.2% (vol/vol) Nonidet P-40, 1× protease inhibitor mixture] and incubated for 5 min on ice, followed by centrifugation (100 × g, 10 min) at 4 °C. The nuclear pellet was resuspended in 200 μL of digestion buffer [1× AluI restriction enzyme digestion buffer (New England Biolabs) + 0.3% (wt/vol) SDS] and incubated at 37 °C for 1 h on a flip-flop rocker. Then, 1.8% (vol/vol) Triton X-100 (final concentration) was added to the samples and incubated for 1 h under similar conditions. Precipitate was removed by short spinning. Four hundred units of AluI restriction enzyme (New England Biolabs) was added and incubated at 37 °C for 2 h, followed by addition of another 200 units of the respective enzymes and overnight incubation at 37 °C. Restriction enzyme was inactivated by addition of SDS (1.6% vol/vol, final concentration) and incubation at 70 °C for 20 min. SDS was quenched by addition of Triton X-100 (1% vol/vol, final concentration) and 1 h of incubation at 37 °C. Individual samples were then diluted 12-fold in ligase buffer (final DNA concentration of ∼3 ng/μL). Ligation was done for 4.5 h at 16 °C, followed by 30 min of incubation at room temperature using 400 units of T4 DNA ligase (New England Biolabs). For mock-ligated samples, no ligase was added. Next, 1 μL of 10 mg/mL RNase and 5 M NaCl (final concentration of 200 mM) were added to the samples, which were incubated overnight at 65 °C. Proteinase K (200 μg/mL) was added to the samples, which were incubated for 30 min at 50 °C, and DNA was purified by phenol-chloroform extraction and isopropanol precipitation. The amount of DNA was estimated by spectrophotometry, and equal amounts of DNA were used for PCR amplification (primer sequences are listed below). PCR-amplified DNA was separated by 2% agarose gel electrophoresis, followed by a hybridization to a Southern probe (32) designed to contain 20 nt on each side (5′ and 3′, probe length = 40 nt) of a possible junction between two AluI fragments of the TSLP gene region of interest. The probe was 5′–end-labeled using γ-[³²P] ATP and purified through a Sephadex G-50 column to remove free γ-[³²P] ATP. As a positive control for AluI digestion, ligation, PCR amplification, and Southern hybridization, a BAC DNA encompassing the region of interest was simultaneously processed (denoted as TSLP BAC).

Primers and probes used in the 3C assay

Junctions	PCR primers/Southern probe sequence	Product length, bp
DR2b and PP	Forward primer: −13,711 GAGGCAGGTGGATTTCTGAG −13,692
	Reverse primer: −200 AGCCAGCCAAATCTATGGAA −181
	Southern probe: 5′-CAGAACAACACAGTCTTAGCTAAAAAATAAATAAATAAATA-3′ (−13,521 to −13,501) (−294 to −272)	324
DR3d and PP	Forward primer: −7,365 CCTCTGGCTCCCTCAATTCT −7,346
	Reverse primer: −87 TCCCCTAGGCCAATGATTTT −106
	Southern probe: 5′-CCACATCTGTTTGCATCAGCTAAAAAATAAATAAATAAATA-3′ (−7,288 to −7,268) (−294 to −272)	301
DR3f/g and PP	Forward primer: −32,702 ACCCCTCTGGCTCCTACAAT −32,683
	Reverse primer: −53 CCAGTGTTCTCTGCCTCCAT −34
	Southern probe: 5′-CCTTCCTTAGGATCCCCTCAAGCTAAAAAATAAATAAAT-3′ (−32,672 to −32,652) (−294 to −268)	163
AP1b and PP	Forward primer: −16,618 CAGGGCCCAGAACATAGAAA −16,599
	Reverse primer: −250 CCCTCACCTTTTTCCTTTCC −231
	Southern probe: 5′-AGTTTCTCTGTGTAGCCCTAGCTAAAAAATAAATAAAT-3′ (−16,379 to −16,357) (−294 to −268)	374
AP1c and PP	Forward primer: −30,007 TTCCCATGGTGACAGAATTT −29,988
	Reverse primer: −113 GCCAATGATTTTCCTTGATG −94
	Southern probe: 5′-CCCCTCACCGAGGTTCTCAGCTAAAAAATAAATAAAT-3′ (−29,904 to −29,883) −294 to −268)	237
AP1d and PP	Forward primer: −41,912 CTAGAGAACCCCAGGATTGC −41,893
	Reverse primer: −200 AGCCAGCCAAATCTATGGAA −181
	Southern probe: 5′-ACCCCTCACGGAGGTTCTCAGCTAAAAAATAAATAAAT-3′ (−41,738 to −41,716) (−294 to −268)	309
AP1f and PP	Forward primer: −71,660 TCGATTAACCTATCCCCATGTC −71,639
	Reverse primer: −53 CCAGTGTTCTCTGCCTCCAT −34
	Southern probe: 5′-TCTGAAGAGTGACTGAAGAGCTAAAAAATAAATAAAT-3′ (−71,620 to −71,599) (−294 to −268)	174
AP1g and PP	Forward primer: −77,526 TGGGCCTTAAATCCAACAAG −77,507
	Reverse primer: −84 CTAAACGCCTACGGGCTCTC −65
	Southern probe: 5′-CACTATCACACTAGCAGCTAAAAAATAAATAAAT-3′ (−77,474 to −77,456) (−294 to −268)	183

Bold sequences represent the junctions between a DNA segment of a proximal promoter (PP) and the DNA sequence to which a given transcription factor is bound. The underlined sequence represents the restriction enzyme site for Alu I.

Histopathology.

Skin biopsies of age-matched mice were taken from similar body sites, fixed overnight with 4% paraformaldehyde (Sigma) at 4 °C, and embedded in paraffin. Five-micrometer sections were stained with hematoxylin/eosin (H&E).

Immunohistochemistry.

After fixation in 2% paraformaldehyde, 10-µm cryosections were blocked in 5% normal goat serum (Vector Laboratories). For TSLP labeling, 5-μm paraffin sections were treated with citric buffer (10 mM citric acid, pH 6) to retrieve antigen, incubated with 5% normal donkey serum (Vector Laboratories), treated with a streptavidin/biotin-blocking kit, and incubated with biotinylated goat polyclonal anti-TSLP (R&D Systems). FITC-conjugated streptavidin was used as a secondary antibody, and was mounted with Vectashield medium (Vector Laboratories) containing DAPI (Invitrogen).

Serum TSLP levels were determined with a Quantikine Mouse TSLP Immunoassay (R&D Systems).

Cell Culture Experiments.

MLE12 mouse lung epithelial cells (CRL-2110; American Type Culture Collection) were maintained in DMEM/Ham-F12 (1:1) medium containing 2% FCS, 5 μg/mL insulin, 10 μg/mL apo trans-bovine, 35 nM sodium selenite, 10 nM β-estradiol, 10 mM Hepes, and gentamicin. For RNA isolation, cells were seeded at 50% density in 60-mm plates. Twenty-four hours after seeding, complete medium was replaced with medium containing charcoal-treated serum. Twenty-four hours later, 1 μM active VD3, 100 ng/mL TPA, 1 μM RA, 5 ng/mL IL-1β (final concentration), or vehicle (ethanol) was added to the medium for 6 h. RNA was isolated using TRI reagent (MRC, Inc.) following the manufacturer’s protocol. For ChIP assay, cells were cross-linked by adding formaldehyde (1% final concentration) to the medium and incubation on a rotary shaker for 10 min at room temperature. Cross-linking was stopped by adding 2 M glycine (0.125 M final concentration) and incubation for 5 min under similar condition, followed by washing two times in ice-cold PBS. Cells were scraped using a rubber policeman instrument, pelleted (400 × g, 5 min) at 4 °C, washed once in ice-cold PBS, snap-frozen in liquid nitrogen, and stored at −70 °C before proceeding with the ChIP assay.

Quantitative RT-PCR primers

Forward and reverse primers	Sequence
mHPRT fp	5′ GTTGGATACAGGCCAGACTTTGTTG
mHPRT rp	5′ GATTCAACTTGCGCTCATCTTAGGC
mTSLP fp	5′ AGCTTGTCTCCTGAAAATCGAG
mTSLP rp	5′ AGGTTTGATTCAGGCAGATGTT
mCRABPII fp	5′ CCAGCAGTCGAGATCAAACA
mCRABPII rp	5′ TTCCACTCTCCCATTTCACC
mSmad7 fp	5′ TCGGACAGCTCAATTCGGAC
mSmad7 fp	5′ GGTAACTGCTGCGGTTGTAA

Statistical Analysis.

Data are represented as mean ± SD of at least three independent experiments, and were analyzed using SigmaStat (Systat Software) by the Student t test or the Mann–Whitney rank-sum nonparametric test depending on results from the Kolmogorov–Smirnov test (with Lilliefors’ correction) for normality and the Levene’s test (median) for equal variance. P < 0.05 was considered significant.