The NF-κB binding site located in the proximal region of the TSLP promoter is critical for TSLP modulation in human intestinal epithelial cells


Full correspondence: Dr. Hervé M. Blottière, INRA, UMR 1319 MICALIS, Domaine de Vilvert, F-78352 Jouy-en-Josas, France

Fax: +33-1-34-65-24-92



Thymic stromal lymphopoietin (TSLP) is constitutively secreted by intestinal epithelial cells. It regulates gut DCs, therefore, contributing to the maintenance of immune tolerance. In the present report, we describe the regulation of TSLP expression in intestinal epithelial cells and characterize the role of several NF-κB binding sites present on the TSLP promoter. TSLP expression can be stimulated by different compounds through activation of p38, protein kinase A, and finally the NF-κB pathway. We describe a new NF-κB binding element located at position –0.37 kb of the promoter that is crucial for the NF-κB-dependent regulation of TSLP. We showed that mutation of this proximal NF-κB site abrogates the IL-1β-mediated transcriptional activation of human TSLP in several epithelial cell lines. We also demonstrated that both p65 and p50 subunits are able to bind this new NF-κB binding site. The present work provides new insight into epithelial cell-specific TSLP regulation.


A single layer of columnar intestinal epithelial cells (IECs) physically separates the intestinal lumen from the underlying mucosal immune cells and defects in their barrier function are associated with inflammatory bowel diseases [1, 2]. Through the secretion of molecules including thymic stromal lymphopoietin (TSLP), retinoic acid, and TGF-β, IECs play a critical role in maintaining the immune tolerance toward food antigens and commensal bacteria contributing to gut homeostasis. TSLP is an IL-7-related cytokine mainly expressed by nonhematopoietic cells including epithelial cells and fibroblasts, originally shown to support β-cell development in mice [3, 4]. It was recently shown that TSLP acts on DCs resulting in their activation and induction of a TH2 type immune response [5]. Although sequence homology is weak (43% amino acid sequence identity), human and mice TSLP share similar biological functions [6]. TSLP exerts its activity by binding to a high-affinity heterodimeric receptor that consists of the IL-7 receptor alpha chain (IL-7Rα) and the TSLP receptor (TSLPR) chain and transmits signals via STAT5 activation [7-9]. TSLPR alone has low affinity for TSLP but together with IL-7Rα forms a high-affinity binding site for TSLP [8, 10]. It has been shown that the interaction TSLP-TSLPR is essential for promoting immune responses against the intestinal nematode pathogen Trichuris [11, 12]. TSLP is expressed at several mucosal surfaces such as skin, lungs, thymus, and gut, but most of the studies focused on its functions in allergic diseases such as asthma and skin atopic dermatitis where a positive correlation between increased TSLP expression and the aggravation of atopic dermatitis and lung inflammation has been shown [13, 14]. Previous works showed that TSLP expression is upregulated following exposure to different factors including inflammatory mediators, TLR activation and/or tissue damage by a NF-κB dependent mechanism [15, 16]. In addition, it has been demonstrated that the MAPK pathway is also involved in the regulation of TSLP expression in response to IL-1 and PMA-mediated signaling [17, 18]. This infers that both NF-κB and MAPK pathways cooperate in regulating TSLP expression. The role of TSLP in the gut is less extensively studied. Thus far, it has been shown that TSLP is constitutively expressed in IECs from healthy subjects, where it inhibits IL-12 production by DCs in response to bacteria, but not in cells from patients with chronic inflammation caused by active Crohn's disease [5]. The aim of this work was to investigate the transcriptional regulation of the TSLP gene in the gut using IEC lines, HT-29, and Caco-2. We examined a 4 kb region of the human TSLP promoter and identified a number of putative NF-κB and AP-1 binding sites. We demonstrated that the NF-κB site located at –370 bp from the ATG (isoform 1) is the key site for IL-1-mediated transcriptional activation of TSLP in the IECs. Further analysis of other epithelial cell models (A549, HEK293, HeLa) confirmed the absolute requirement of this proximal NF-κB binding site for the NF-κB-dependent activation of TSLP gene transcription in epithelial cells. This work has revealed an important cell-specific aspect in the regulation of TSLP in epithelial cells.


Identification of putative NF-κB and AP-1 binding sites on human TSLP promoter

In silico analysis of a 4 kb-long region of the human TSLP promoter revealed four potential NF-κB consensus binding sites (referred as NF1, NF2, NF3, and NF4) and three potential AP-1 binding sites (referred as AP1–1, AP1–2, and AP1–3) (Fig. 1A). NF1 site is located between positions -3992/–3982 from the ATG (A corresponding to position +1 of isoform 1). This site has been previously described and characterized in human airway epithelial cells [16]. NF2 site is located between positions –369/–359 of the human TSLP promoter. Two additional putative NF-κB sites, named NF3 and NF4 are located, respectively, at positions –1528 and –3421 of TSLP promoter. A search of the relevant vertebrate databases revealed that the region of human TSLP promoter containing the NF2 site, is conserved in numerous mammals namely Pongo abelii, Pan troglodytes, Mus musculus, Rattus norvegicus, Equus caballus, and Bos taurus (Fig. 1B). Within these species, no sequence corresponding to human NF1 was found in M. musculus and R. norvegicus. A sequence similar but not identical to human NF1 was found in E. caballus and B. taurus. As expected, both NF1 and NF2 sites, as well as NF3 and NF4 sites, were conserved in primates. The latters were also found in E. caballus and M. musculus but not in B. Taurus or R. norvegicus. Three putative AP-1 binding sites (AP1–1, AP1–2, and AP1–3), are located at positions –3942, –1255, and –263, respectively (Fig. 1). Like NF1 binding site, AP1–1 site has been described in human airway epithelial cells [16]. Moreover, AP1–2 and AP1–3 are conserved between human and mice but not AP1–1.

Figure 1.

(A) Putative NF-κB and AP1 binding sites located upstream of the human TSLP promoter. The distance of each site from the ATG (black arrow) is indicated. The critical NF2 site is indicated in bold letters. (B) Alignment of DNA region containing the TSLP NF2 site in several mammals. Identical nucleotides are indicated with asterisks. NF2 site is indicated in bold letters. Distances from ATG are shown.

Regulation of TSLP expression in human IECs

Since NF-κB and AP-1 are key transcription factors involved in various inflammatory pathologies in both humans and mice and several reports suggest TSLP to be regulated by NF-κB [16, 19] we focused our work on a number of inflammatory agonists including IL-1, TNF-α, and PMA as well as TLRs ligands to evaluate, at the transcriptional level, TSLP regulation in human IECs. For this purpose, we used a luciferase reporter assay where the luciferase gene was cloned under the control of a 4-kb-long fragment of TSLP promoter. Among the TLRs ligands used Flagellin and FSL1 were able to stimulate the reporter gene activity in HT-29 and Caco-2 cells, respectively (Supporting Information Fig. 1). When Caco-2 cells were stimulated with IL-1, a 12-fold increase in luciferase activity was measured at 24-h poststimulation, whereas a weaker activation was observed in cells stimulated with TNF (twofold) (Fig. 2A). PMA, a diacyglycerol analog that activates PKC and butyric acid, is an end-product of bacterial fermentation, that strongly regulates gene expression in IECs [20-22]. We found that PMA also strongly induced TSLP-dependent luciferase activity (ninefold). Moreover, when Caco-2 cells were co-incubated with PMA and butyric acid a dramatic stimulation (100-fold) of luciferase activity was noted (Fig. 2A). Similar results were obtained with HT-29 cells, however as expected, HT-29 cells were less sensitive to PMA and much more to TNF (data not shown). These results were also confirmed using real-time qRT-PCR in Caco-2 cells exposed to the same stimuli for 2 to 6 h (data not shown). A kinetic study of Caco-2 response to the various agonists revealed a peak in luciferase activity 12 h after stimulation with IL-1 and PMA. Later, the IL-1-induced activity slowly decreased but never below the 50% of the maximum activity. In the presence of butyric acid, however, luciferase values continued to increase after 48 h (Fig. 2B). Moreover, a combination of butyric acid and PMA induced a strong synergistic stimulation of luciferase activity similar to that induced by IL-1 following 12 h stimulation. This effect was still apparent at 48 h (Fig. 2B). Combining butyric acid and IL-1 did not result in any synergistic effect, but only an additive one (Supporting Information Fig. 2). This effect is also observed using trichostatin A (TSA), an histone deacetylase inhibitor.

Figure 2.

Response of TSLP reporter clone (Caco-2) to IL-1 (10 ng/mL), TNF (10 ng/mL), PMA (1 μM), Butyric acid (2 mM), and PMA + Butyric acid. (A) Reporter gene activity was measured after 24 h stimulation. **p < 0.01, ***p < 0.001; Student's t-test. (B) Kinetic of the response. (A–B) Results are shown as mean ± SD of triplicate measurements of a representative of three independent experiments.

Consistent with the gene transcription data, TSLP protein was released in the supernatants 8 h after Caco-2 cells stimulation, with a maximum effect at 24 h in response to IL-1, TNF, PMA, and butyrate. Interestingly, TSLP concentration in the supernatant decreased by 48 h except when the cells were treated with a combination of PMA and butyrate (Fig. 3).

Figure 3.

Secretion of TSLP protein (pg/mL) from 1-week differenced Caco-2 cells exposed to IL-1 (10 ng/mL), TNF (10 ng/mL), PMA (1 μM), butyric acid (2 mM), and PMA + butyric acid for 8, 24, and 48 h. White bars represent cytokine production in nonstimulated cells. Results are shown as mean ± SD of triplicate measurements of a representative of three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001; Student's t-test.

TSLP is controlled by NF-κB and MAPK

To further decipher the mechanism of TSLP regulation by both IL-1 and PMA, several signaling pathways inhibitors were selected.

First, we used BAY 11–7082, a well-characterized inhibitor of the NF-κB signaling. At a 20 μM concentration, it inhibited TSLP promoter-driven luciferase activity by about 60% in cells stimulated with IL-1 (Fig. 4). We then tested several kinase inhibitors and found that the p38 inhibitor, SB203580, and the protein kinase A (PKA) inhibitor, H-89, were able to inhibit up to 50% of the IL-1-stimulated luciferase activity in Caco-2 reporter cells, while the MEK 1/2 inhibitor, UO126 had a lower but still statistically significant effect (Fig. 4). These results indicate that both the NF-κB and the AP-1 pathways are involved in the IL-1-dependent induction of TSLP. As expected, the PKC inhibitor, bisindolylmaleimide (BIM), had no significant effect on the IL-1-induced reporter gene activity but almost completely abolished the PMA-induced activity (Fig. 4).

Figure 4.

Effect of several kinase inhibitors on the IL-1 (10 ng/mL) and PMA (1 μM) stimulated luciferase activity of Caco-2 reporter cells. Bay11–7082 (20 μM), SB203580 (10 μM), UO126 (10 μM), H-89 (10 μM), bisindolylmaleimide (BIM – 2 μM). Reporter gene activity was measured after 24 h stimulation. Results are shown as mean + SD of triplicate measurements of a representative of three independent experiments. Different letters indicate statistically different results with a p value <0.05; Student's t-test.

We then investigated whether the remaining activity induced by IL-1 after BAY 11–7082 treatment was dependent on other kinases. The combined action of BAY 11–7082 with SB203580 or H-89 drastically reduced the remaining IL-1-induced activity, thus corroborating the hypothesis of cooperation between the NF-κB and AP-1 sites on the IL-1-induced TSLP promoter activity. Finally, we investigated the role of several kinases in the PMA-induced TSLP expression. Besides its expected inhibition by BIM, the other inhibitors did not affect the PMA-induced TSLP luciferase activity.

The proximal NF-κB site is crucial for IL-1-induced TSLP expression in human epithelial cells

As shown in Figure 1 using an in silico analysis of a 4-kb-long region of TSLP promoter, we identified several putative NF-κB and AP-1 binding sites. To further investigate the impact of these presumed sites on the IL-1 or PMA-dependent modulation of TSLP expression, we generated luciferase reporter plasmids containing fragments of different size of TSLP promoter. We used these constructs to transiently transfect both HT-29 and Caco-2 cells. The luciferase activities were normalized to those of the secreted alkaline phosphatase (SEAP) in which the SEAP gene was under the control of a constitutive promoter.

Results obtained from transfection experiments with reporter plasmids containing 1, 0.5, or 0.37 kb of the TSLP promoter showed equal reduction in luciferase activity in response to IL-1 stimulation (about 30%) when compared with the activity observed using the full length TSLP promoter construct (Fig. 5A). We first assumed that this reduction was due to the absence of the published NF1 and AP1–1 sites in these regions [16]. Surprisingly, TSLP-dependent luciferase activity was not affected in cells transfected with constructs lacking either NF1 site alone (3957 bp construct) or both the NF1 and the AP1–1 binding sites (3903 bp construct) suggesting an additional NF-κB site involved in TSLP expression. The in silico analysis revealed two putative NF-κB binding sites (NF4 and NF3) and one AP1 (AP1–2). The results obtained using a 3 kb-long promoter construct that lacks the NF4 site suggested that it might play a functional role in TSLP expression since a similar 30% reduction was noted (Supporting Information Fig. 3). A further significant reduction in luciferase activity was observed however, when a construct that lacked the NF2 site (0.29 kb construct), was assessed in response to IL-1 stimulation (Fig. 5A). These results pointed to the functional importance of NF2 site, located between positions –0.37 and –0.29 kb, in IL-1-induced TSLP expression. To confirm our hypothesis, site-directed mutagenesis targeting either NF1 or NF2 or both in the context of the full length 4 kb-long promoter region were performed. Mutation of NF1 did not modify the IL-1-induced luciferase activity.

Figure 5.

Effect of IL-1 on TSLP promoter-driven luciferase activity. Various 5′deletions of the TSLP promoter (A) and site directed mutagenesis of NF1, NF2, and NF1–2 (B) were cloned in pCDNA3.1-Luc and transiently transfected in Caco-2 cells. Length of each constructs as well the positions of NF1 and NF2 binding sites are indicated. Cells were stimulated 24 h after transfection and reporter gene activity was assayed 24 h after stimulation. Results are shown as mean ± SD of triplicate measurements of a representative of three independent experiments. Different letters indicate statistically different results with a p value <0.05; Student's t-test.

On the contrary, mutation of the NF2 site completely abrogated the reporter gene activity in IL-1 stimulated Caco-2 (Fig. 5B) as well as in HT-29 cells (not shown). The same results were obtained when Flagellin was used to stimulate the reporter system activity, indicating that TLR regulation is mediated by the same mechanism than IL-1 (Supporting Information Fig. 4).

To confirm that NF2 was a critical NF-κB binding site for TSLP modulation and that it was not restricted to epithelial cells of the intestine, lung (A549), cervical (HeLa), and kidney (HEK 293) epithelial cell lines were used. Again, we observed that mutation of NF1 did not alter the IL-1-mediated TSLP promoter activity whereas mutation of NF2 completely abolished the activity (Supporting Information Fig. 5). These data strongly support the absolute requirement for NF2 in the NF-κB-mediated regulation of TSLP in several epithelial cell lines.

Using transient transfection experiments (Supporting Information Fig. 6), we observed that at least four regulatory elements may be involved in the PMA-mediated effect in association or not with butyrate. It is noteworthy that, in those transient experiments, butyrate had no significant effect (Supporting Information Fig. 6B); however it strongly enhanced the effect of PMA (Supporting Information Fig. 6C). We therefore extended our strategy to analyze the putative role of AP-1 sites in the PMA effect on TSLP promoter. As the in silico analysis predicted an AP-1 binding site at position –1255 (AP1–2) and another one at position –263 (AP1–3), we generated two constructs containing 1256 bp and 250 bp, respectively, of the TSLP promoter region. By comparing the 1256 bp and the 1000 bp constructs, we observed no significant reduced activity on cells transfected with these plasmids and exposed to PMA. Similarly, a comparison between the 290 and the 250 bp ruled out the involvement of the other AP-1 binding site (data not shown). Finally, site-directed mutagenesis targeting AP1–1, AP1–2, or AP1–3 sites alone or in association with NF1 and NF2 mutations did not lead to any reduced luciferase activity on Caco-2 cells exposed to PMA (data not shown), suggesting that additional AP-1 sites or other transcription factors may be involved in PMA signaling.

NF-κB binds to both NF1 and NF2 sites on TSLP promoter

To further confirm the role of NF2 in the expression of TSLP, we prepared nuclear extracts from IL-1, TNF, and PMA-activated Caco-2 and HT-29 cells as well as from unstimulated cells and performed electrophoretic mobility shift assays.

Using specific 32P-labeled oligonucleotides containing NF1 or NF2 binding sites, we were able to detect protein binding (shift) to both sites upon cells stimulation with all the agonists tested, while no shift was observed in the case of nonstimulated cells (Fig. 6A–C). We confirmed the specificity of NF-κB binding by incubating nuclear extracts from stimulated cells with antibodies against p50 or p65 subunits. A strong supershift was observed for both NF1 and NF2 sites in the case of p65 subunit, while a weaker, but still clear, signal was detected with p50 specific antibody (Fig. 6A–C). Mutation of either NF1 or NF2 core sequences or incubation of nuclear extracts with an excess of the unlabeled oligonucleotides abrogated the binding capacity of the probes (Fig. 6B–D). Thus, our results clearly demonstrate that NF-κB complex was able to bind to NF1 and potentially more importantly, the NF2 site.

Figure 6.

NF-κB binds to the NF-κB motif of the human TSLP promoter. Nuclear extracts prepared from HT-29 cells. (A–C) (1) Negative control (2) unstimulated cells (3) cells stimulated with IL-1 (4) cells stimulated with TNF (5) cells stimulated with PMA and incubated with labeled oligonucleotide probes and subjected to EMSA and supershift with p50 and p65 antibodies. (B–D) (1) Negative control (2) unstimulated cells (3) cells stimulated with IL-1 (4) cells stimulated with TNF (5) cells stimulated with PMA and incubated with labeled oligonucleotide probes and subjected to EMSA and supershift with p50 and p65 antibodies. For competition the reaction was incubated in the presence of a molar excess of unlabeled oligonucleotide and for specificity of binding reaction was incubated with a mutated oligonucleotide probe. Data shown are representative of one out of three experiments.


During the last decade, TSLP has been the subject of intense studies because of its involvement in the maintenance of immune homeostasis [11, 23, 24]. TSLP, a cytokine mainly released from the basolateral side of IECs, contributes to DC maturation and stimulates a TH2-like inflammatory response characterized by IL-4, IL-5, IL-13, and TNF upregulation and IL-10, and IFN-γ downregulation [25-27]. TSLP is constitutively expressed in both the small and large intestine and it plays a key role in gut homeostasis as highlighted in mouse models [28, 29] and in human cell models [5]. As shown previously [16], we observed that TSLP expression was strongly regulated by proinflammatory cytokines such as TNF and IL-1. In addition, we observed that the PKC activator, PMA, as well as a bacterial fermentation end product, butyrate, also regulated TSLP expression both at the mRNA and protein level. Moreover, a strong synergistic effect between PMA and butyrate was observed. The latter effect may be physiologically relevant given the major biological function of butyrate as an energy source in the colon [30] as well as its function as an epigenetic regulator [31].

As expected, stimulation of IECs by IL-1 induced NF-κB translocation into the nucleus and TSLP transcription involving IKK-β activity as revealed by the specific inhibition induced by Bay 11–7082. Clearly, the functional importance of both p38 and PKA was also identified using SB203580 and H-89, respectively. Conversely, extracellular signal-regulated kinase (ERK) had little effect since UO126 barely inhibited TSLP transcription. We first postulated that both p38 and PKA may act independently of IKK-β involvement since their specific pathway inhibitors were effective in the presence of Bay 11–7082, whereas UO126 had no effect. However, when transient transfections were performed with a 4 kb TSLP-promoter region, mutated for NF2 binding site, the stimulatory effect of IL-1 was completely abolished; thus arguing for a NF-κB only dependent regulation. We present in Figure 7 our working hypothesis that can explain the overall results obtained in IL-1-dependent TSLP regulation. Considering that, in the presence of BAY 11–7082, the effects of both p38 and PKA inhibitors are still apparent, we can argue that, since BAY 11–7082 has an IC50 of 10 μM [32], at the concentration 20 μM used in the current study, IKK-β may only be partly inhibited and that the remaining TSLP transcription activity is still mediated by IKK-β. This has been verified using a NF-κB-dependent SEAP reporter system [33]. In fact, at the 20 μM concentration, BAY 11–7082 only inhibited IL-1-dependent NF-κB activation by about 60% in Caco-2 cells. To explain the effects of the p38 inhibitors, our hypothesis is that IL-1 is activating IKK-β by two separate modes; first via the classical IL-1 receptor associated kinase/TGF-β activated kinase (IRAK/TAK) dependent pathway and second via a MKK/p38-dependent pathway as revealed previously for IL-6 [34]. Thus, inhibition of p38 resulted in a decreased TSLP expression due to a reduced activation of IKK-β, and enhances BAY 11–7082 direct inhibitory activity. Considering the involvement of PKA, it has been shown that PKA can also interfere with the NF-κB pathway; indeed PKA was revealed to phosphorylate p65 in a cAMP-independent manner therefore increasing transcriptional activity [35]. Our results argue for a similar regulation of TSLP transcription in human IECs.

Figure 7.

Hypothesis for IL-1-dependent TSLP regulation in IECs. IL-1β activates IKK through IRAK/TAK/TAB-dependent pathway, which will lead to Iκ−B phosphorylation and degradation, and then release of p50/p65 dimer to activate TSLP expression. PKA linked to NF-kB complex also controls TSLP expression. Activated TAK/TAB also controls MKK, which directly or indirectly leads to IKK activation via p38 activation. Specific inhibitors (gray bars) interfere with TSLP expression.

Recently, TSLP has been shown to be regulated by NF-κB in both human and mice airway epithelial cells [16]. A site located at –3.8 kb from the transcriptional start site of TSLP was found responsible for the IL-1-mediated transcriptional activity of this gene. Therefore, we wondered whether TSLP expression in human IECs was regulated in a similar fashion. Although we also observed that TSLP was regulated by NF-κB in Caco-2 and HT-29 cell lines in response to IL-1, we found contradictory results concerning the precise promoter site responsible for the NF-κB-dependent regulation of TSLP. The in silico analysis of a 4 kb-long region of human TSLP promoter allowed us to identify four potential NF-κB sites. Although human and murine TSLP promoters do not share significant sequence homology, one of these putative sites is conserved in mice TSLP promoter as well as in other mammals. Moreover, in mice a site corresponding to NF2 exerts the same biological function as that observed in human TSLP regulation and expression (P. Chambon, unpublished data and [36]). In our study, we used different strategies to demonstrate that NF2, a newly identified NF-κB responsive element located in the proximal region of TSLP promoter, is functionally important for the NF-κB-dependent regulation of human TSLP in IECs. We also demonstrated the functional importance of NF2 in regulating TSLP expression in other epithelial cells, including lung, cervical and kidney epithelial cells.

Despite the fact that both NF1 and NF2 sites showed similar binding capacities for p65 and p50 subunits of NF-κB, as revealed by EMSA experiments using nuclear extracts from IL-1-, TNF- or PMA- stimulated Caco-2 and HT-29 cells, they produced a different impact on TSLP modulation. First, we assumed that both NF1 and NF2 sites were necessary to support the full transcriptional activity of NF-κB complexes in response to the different ligands. However, TSLP promoter lacking a functional NF1 site was still able to respond to IL-1 in IECs as well as in other epithelial cells, including the lung cell line, A549, which has been used in the previously published paper [16]. By contrast, all the IL-1-induced activity was lost following NF2 site mutation, demonstrating the absolute requirement of NF2 for the NF-κB-dependent regulation of TSLP driven by IL-1. We speculate that the presence of two NF-κB sites, one of which fails to respond to inflammatory agonist IL-1, could be necessary for constitutive expression of TSLP, while the other responses to upregulate TSLP expression under specific conditions. Overall, our data did not reveal other regulatory elements, other than NF2, that are absolutely essential for the IL-1-induced expression of TSLP.

In accordance with previous studies [16, 17], we showed that TSLP promoter contains several putative AP-1 binding sites. These sites either cooperate with NF-κB sites to mediate the effects of IL-1 via ERK pathway or are involved in PKC signaling via PMA. Our results argue for the functional role of four binding elements, however, we found that the AP1–1 site that had previously been reported to be involved in TSLP modulation was not involved in IEC-specific TSLP expression. Similarly, the additional putative sites (AP1–2 and 3) identified in silico, appeared to be functionally irrelevant. We thus consider that other transcription factors may be involved in TSLP modulation via PMA. Indeed, we have identified two putative AP-2 binding sites in the proximal region of TSLP promoter. Our results, obtained using transfected cells with small fragments of TSLP promoter (212 and 74 bp, respectively) lacking these two putative sites, suggest that a presumed AP-2 site located at –85 bp from the ATG could be responsible for the residual PMA-depending activity of TSLP observed when NF2 is absent (Supporting Information Fig. 6A). Indeed, we have demonstrated that the IL-1 stimulated luciferase activity is completely lost in cells transfected with the 290 bp construct that lacks the NF2 site (Fig. 5A), while a lower but still significant activity is measured on cells exposed to PMA (Supporting Information Fig. 6A).

Previous works showed that PMA significantly increases MCT1 expression in Caco-2 cells, a monocarboxylate transporter important for butyrate absorption in the human colon [37, 38]. Recently, Saksena et al. [39] demonstrated that the effect of PMA on MCT1 gene expression was mediated through a PKC-ζ-dependent pathway involving the AP-2 transcription factor. Although we cannot rule out this hypothesis, we observed that BIM used at 2 μM abolished the PMA-dependent TSLP transcription, while PKC-ζ is reported to require higher concentration of BIM (>5 μM) to be inhibited. Other transcription factors or binding elements seem to be involved in PMA-mediated TSLP transcription.

Finally, we showed that butyrate is a weak stimulator of TSLP expression when used alone, but strongly enhances the stimulatory effect of PMA. This effect is specific for PMA/butyrate association, since the combined action, IL-1/butyrate, produces only a weak synergy (Supporting Information Fig. 2). Moreover, we observed that butyrate alone was not able to directly activate luciferase when constructs with different size of TSLP promoter were transiently transfected in IECs (Supporting Information Fig. 6B). This suggests that the effect of butyrate may not depend on a specific butyrate binding site on TSLP promoter but involve the epigenetic modification properties of butyrate, i.e. its histone deacetylase (HDAC) inhibitory properties [21, 40]. The fact that TSA, another HDAC inhibitor, displays identical effects to butyrate alone or in conjunction with PMA strongly argues for this hypothesis (Supporting Information Fig. 2).

In conclusion, our work contributes to a better understanding of the mechanism of regulation of TSLP expression in epithelial cells. Moreover, it provides evidence for the critical transcriptional role of the proximal NF-κB binding site in human TSLP promoter in driving TSLP expression response to IL-1.

Materials and methods

Reagent and cell culture

The human epithelial cell lines, HT-29, Caco-2, Hela, A549, and HEK293 were obtained from the American Type Culture Collection (ATCC, Rockville, MD, USA). The cells were grown in RPMI 1640 (HT-29, A549, HeLa, HEK293) or DMEM (Caco-2) media (Lonza) supplemented with 2 mM L-glutamine, 50 IU/mL penicillin, 50 μg/mL streptomycin, and 10% (or 20% in the case of Caco-2) heat-inactivated fetal calf serum ( Lonza) in a 37°C humidified atmosphere of 5% (HT-29, A549, HeLa, HEK293) or 10% (Caco-2) CO2. For reporter cell line characterization, cells were seeded at 5.0 × 104 per well in 96-well plates. After overnight culture, cells were stimulated 24 h with recombinant human IL-1β (10 ng/mL, Peprotech and referred as IL-1 throughout the text), TNF-α (10 ng/mL, Peprotech and referred as TNF throughout the text), Phorbol myristate acetate (PMA, 1 μM), butyric acid (2 mM, SIGMA), TSA (0.5 – 1–10 μM).

The TLR response profile was determined using the TLR1–9 agonist kit (Invivogen) according to manufacturer's instruction. Ligands and working concentrations are for TLR1–2: Pam3CSK4 (1 mg/mL); TLR2: Heat-Killed Listeria monocytogenes (108 cells/mL); TLR3: Poly(I:C) (10 mg/mL); TLR4: Escherichia coli K12 LPS (10 mg/mL); TLR5: Salmonella typhimurium Flagellin (10 mg/mL); TLR6/2: FSL1 (1 mg/mL); TLR7: Imiquimod (1 mg/mL); TLR8: ssRNA40 (1 mg/mL); and TLR9: ODN2006 (5 mM). In transient transfection assays, Flagellin was used at working concentration of 1 μg/mL.

MAPK kinase inhibitors, U0126 and SB203580, and PKA inhibitor, H-89 were used at 10 μM; PKC inhibitor, BIM was used at 2 μM and NF-κB inhibitor, BAY 11–7082 ((E)3-((4-methylphenyl)sulfonyl)-2-propenenitrile) was used at 20 μM. All compounds were purchased from Calbiochem.

Plasmid construction

The luciferase reporter gene was cloned at KpnI/XbaI sites in pCDNA3.1/Zeo(+) vector (Invitrogen) in which the pCMV (Cytomegalovirus) promoter was removed with a NruI/NheI digestion. A 4 kb-long region of the human TSLP promoter was amplified from human genomic DNA by PCR using the High Fidelity PCR Mix (Fermentas) and cloned as an NheI/KpnI fragment in pCDNA3.1-Luc plasmid (the resulting plasmid referred as pTSLP-Luc). The 4000-bp-cloned genomic region was used as template to amplify the other promoter fragments used in the present study. The Secreted Alcaline Phosphatase gene was extracted from pTal-SEAP plasmid (Clontech) by a HindIII/EcoRV digest and cloned in pCDNA3.1/Zeo(+). Site-directed mutagenesis of NF-κB binding sites was performed using the QuikChange Lightning Site-Directed Mutagenesis kit (Agilent Technologies). The mutation in the NF1 binding site was performed as described by Lee and Ziegler [16]. The NF2 binding site, GggaAATTCC, was mutated in GttcAATTCC and the mutation was verified by sequencing.

Transfection and luciferase assay

The stable HT-29 cl.23 (HT-29/tslp-23) and Caco-2 cl.6 (Caco-2/tslp-6) reporter clones were obtained by transfecting 2.5 × 105 cells with 1 μg of pTSLP-Luc plasmid using Amaxa Cell Line Nucleofector kits (Lonza) following the manufacturer's instructions. Cells were cultured for 3 weeks under Zeocine™ (50 μg/mL, Invitrogen) selection and cloned. The HT-29/tslp-23 and the Caco-2/tslp-6 were selected for their response to 10 ng/mL of IL-1β after 24 h stimulation.

In transient transfections assays, 1.0 × 106 cells (HT-29 and Caco-2) were transfected with 1 μg of the selected plasmid using the AmaxaR Nucleofector kits (Lonza). After transfection, cells were seeded at 9 × 104 cells/well and cultured for 18 h before stimulation with IL-1β (10 ng/mL). The empty pcDNA-Luc plasmid was used as control. Co-transfection with a plasmid harboring the SEAP driven by CMV promoter (pCMV-SEAP) was used for normalization.

Luciferase activity, quantified as relative luminescence units, was measured using the ONE-GloTM Luciferase Assay System (Promega) according to the manufacturer's instructions using a microplate reader (Infinite 200, Tecan).


Caco-2 cells were grown for 1 week in 24-well plates (100 000 cells/well) and media was changed every day. Supernatants from 8-, 24-, and 48-h-stimulated Caco-2 cells were collected, centrifuged at 1200 rpm for 5 min at 4°C and analyzed using the “Human TSLP ELISA Development Kit” (PeproTech) following the manufacturer's instructions.

Gel shift assay

Nuclear extracts were prepared as described in [41]. In brief, five microgram of nuclear extracts were incubated at room temperature for 20 min with 0.07 pmol (50–200 000 cpm) of double stranded (32P)-labeled oligonucleotide probes containing consensus binding sequences for NF1 and NF2 sites, then separated by electrophoresis and visualized by autoradiography. EMSA supershifts were performed using 1 μg of specific NF-κB antibodies against the p50 and p65 subunits (Santa Cruz Biotechnology). For competition assay, the reaction was pre-incubated with 1000-fold molar excess of unlabeled probe for 30 min at room temperature before the addition of labeled probe.

The oligonucleotides used as probes were as follows:


Statistical analysis

Presented results were representative of at least three independent experiments. Results were expressed as mean ± SD of triplicate measurements of a representative experiment. Data were analyzed by Student's t-test.


This work was supported by grants from the European Community's Seventh Framework Programme (FP7/2007–2013): MetaHIT, grant agreement HEALTH-F4-2007-201052. TdW, DK, JD, and HB are partners of the European Marie-Curie Initial Training Network Cross-Talk (grant agreement # 215553). TdW has been supported by the French National Research Agency (ANR) funded project, MicroObes.

We thank Pierre Chambon for sharing unpublished results, Ronan Legoffic for helpful discussion and Karine Le Roux for technical assistance.

Conflict of interest

The authors declare no commercial or industrial conflict of interest.




intestinal epithelial cell


protein kinase A


secreted alkaline phosphatase


thymic stromal lymphopoietin