Dichotomy of short and long thymic stromal lymphopoietin isoforms in inflammatory disorders of the bowel and skin

Abstract

Background

Thymic stromal lymphopoietin (TSLP) is a cytokine with pleiotropic functions in the immune system. It has been associated with allergic reactions in the skin and lungs but also homeostatic tolerogenic responses in the thymus and gut.

Objective

In human subjects TSLP is present in 2 isoforms, short and long. Here we wanted to investigate the differential expression of the TSLP isoforms and discern their biological implications under homeostatic or inflammatory conditions.

Methods

We evaluated the expression of TSLPs in tissues from healthy subjects, patients with ulcerative colitis, patients with celiac disease, and patients with atopic dermatitis and on epithelial cells and keratinocytes under steady-state conditions or after stimulation. We then tested the immune activity of TSLP isoforms both in vitro and in vivo.

Results

We showed that TSLP isoforms are responsible for 2 opposite immune functions. The short isoform is expressed under steady-state conditions and exerts anti-inflammatory activities by affecting the capacity of PBMCs and dendritic cells to produce inflammatory cytokines. Moreover, the short isoform TSLP ameliorates experimental colitis in mice and prevents endotoxin shock. The long isoform of TSLP is proinflammatory and is only expressed during inflammation. The isoforms are differentially regulated by pathogenic bacteria, such as Salmonella species and adhesive-invasive Escherichia coli.

Conclusions

We have solved the dilemma of TSLP being both homeostatic and inflammatory. The TSLP isoform ratio is altered during several inflammatory disorders, with strong implications in disease treatment and prevention. Indeed, targeting of the long isoform of TSLP at the C-terminal portion, which is common to both isoforms, might lead to unwanted side effects caused by neutralization of the homeostatic short isoform.

Allergic and inflammatory diseases of the skin, airways, and gut affect millions of persons worldwide, and it is becoming clear that these diseases can be pathophysiologically connected. For instance, 70% of patients with atopic dermatitis (AD) undergo development of asthma through a process called the atopic march,¹ and similarly, it has been hypothesized that exposure to food allergens through a disrupted skin barrier early in life leads to antigen sensitization and food allergy.² Hence there is a link between skin, airway, and gut inflammatory disorders.

Thymic stromal lymphopoietin (TSLP) is a cytokine involved in several physiologic and pathologic immune activities.^3,4 TSLP has homeostatic activities because it is expressed by Hassal corpuscles in the thymus and regulates the capacity of dendritic cells (DCs)⁵ and plasmacytoid DCs⁶ to drive the development of natural regulatory T cells. TSLP can also promote the homeostatic polyclonal expansion of T cells in the absence of foreign antigen.⁷ In the gut we have shown that TSLP is expressed by intestinal epithelial cells and educates noninflammatory DCs that have reduced ability to produce IL-12p70⁸ and drive the differentiation of inducible regulatory T cells.⁹

Conversely, TSLP has been shown to play a pathogenic role in several immune disorders, including AD, asthma, and food allergy.⁴ Indeed, TSLP can drive the development of strong allergic T_H2 responses with release of IL-4, IL-5, IL-13, and TNF-α¹⁰ through upregulation of OX40 ligand expression on TSLP-treated DCs.¹¹ TSLP is overexpressed in the airways of asthmatic patients and is associated with T_H2 cytokines.^12,13 TSLP is directly involved in the itch response,¹⁴ in activating innate lymphoid cells to induce skin inflammation,¹⁵ and in the atopic march.¹⁶ Consistently, TSLP polymorphisms are linked to asthma susceptibility^17-19 and AD.^20,21 TSLP influences the differentiation of basophils, which are more prone to promote T_H2 inflammation²² and involved in eosinophilic esophagitis.²³

Hence it seems that TSLP has both physiologic and pathologic activities that are difficult to reconcile. Here we show that this dichotomy can be explained by the presence of 2 distinct isoforms of TSLP, a long and a short one, which are differentially regulated and involved in inflammatory and anti-inflammatory responses, respectively.

Methods

Tissue sampling

Intestinal mucosa was excised from the intestines of patients with ulcerative colitis (UC) at the time of surgery (n = 13). As a control, intestinal samples were obtained from healthy tissue (at least 7 cm away from neoplastic tissue) of patients undergoing surgery for colon cancer (n = 26). The mucosal layer was separated from the rest of the tissue by a pathologist and directly transferred to our laboratory. Transfer was carried out in both cases in HBSS buffer at 4°C supplemented with bacteriostatic antibiotics. Mucosa from patients with celiac disease (CD) was obtained by means of biopsy during routine colonoscopy (patients with treated CD, n = 15; patients with untreated CD, n = 13; healthy subjects, n = 13). Skin biopsy specimens were taken from patients with AD; lesional, nonlesional tissue, and healthy tissue (from the back or the thigh) was collected (n = 24). The clinical diagnosis was confirmed by means of dermatopathologic evaluation. All tissues were obtained from patients who signed an informed consent form approved by the institutional review board allowing material not required for diagnosis to be used for research purposes. The samples were handled in a completely anonymous manner and assigned a serial number by the pathologist for cross-checking histologic reference.

On the basis of a series of preliminary experiments, the size of samples for each disease group has been calculated to estimate the statistical differences with a power of 0.80 and an α value of .05.

Mice

Female C57Bl/6J mice (8-10 weeks of age) were obtained from Charles River Laboratories (Milan, Italy). Mice were bred and maintained at the IFOM-IEO Campus Animal Facility under specific pathogen-free conditions. All experiments were performed in accordance with the guidelines established in the Principle of Laboratory Animal Care (directive 86/609/EEC).

LPS-induced endotoxic shock

Mice were treated intraperitoneally with short isoform TSLP (shTSLP) at 50, 100, and 200 μg per mouse in 200 μL of injectable water twice (18 hours and 2 hours) before LPS administration (n = 8 per group). Control mice received water. LPS (Escherichia coli serotype 026:B6; Sigma-Aldrich, St Louis, Mo) was injected intraperitoneally at 200 μg per mouse in 200 μL of injectable water. After 6 hours, mice were killed by means of exsanguination after achievement of anesthesia, and blood was collected. IFN-γ, IL-6, and IL-12p40 levels were detected in serum by means of ELISA (R&D Systems, Minneapolis, Minn).

Dextran sodium sulfate–induced colitis

Colitis was induced by adding 3% (wt/vol) dextran sodium sulfate (DSS; TdB Consultancy AB, Uppsala, Sweden) to drinking water for 9 days. Mice were treated intraperitoneally with shTSLP (200 μg in 200 μL of injectable water) the day before DSS administration and every other day during the entire study (n = 12 per group). Mice were monitored daily for weight loss.

Bacterial strains

Salmonella enterica serovar typhimurium SL1344 strain FB62 (hereafter referred to as Salmonella typhimurium) and E coli strains LF82 and MG1655 (a kind gift from Arlette Darfeuille-Michaud) were cultured in TB broth. For infection, bacteria were grown at 37°C under agitation until reaching the exponential growth phase (OD between 0.55 and 0.65). Colony-forming unit numbers were controlled by means of plating.

Statistical analysis

All statistics were performed with a 2-tailed Student t test or 1-way ANOVA with GraphPad Prism software (GraphPad Software, La Jolla, Calif). Statistical significance was retained for P values of less than .05. When results are reported in terms of fold increase, statistical tests were applied to log-transformed data.

Results

shTSLP is the homeostatic isoform of TSLP and is constitutively expressed in healthy skin and gut

We carried out an analysis on the UCSC Genome Browser of human TSLP isoforms. There are 3 transcript variants (RefSeq database), but only 2 give rise to coding RNAs: the canonical TSLP transcript variant 1 (NM_033035_hg19 160 chr5:110407589-110411772) and a transcript variant 2 (NM_138551_hg19 64 chr5:110409281-110411772). The 2 coding transcripts code for the long isoform of thymic stromal lymphopoietin (lTSLP) of 159 amino acids (variant 1) and for shTSLP (variant 2), which encompasses the last 63 residues of lTSLP and is identical to its C-terminal portion (Fig 1, A). They are not alternatively spliced isoforms but derive from the activity of 2 putative promoter regions (Fig 1, A). In Fig 1, A, we portrayed the ENCODE track profiles showing active enhancers (H3K4me1 high, H3K4me3 low, and H3K27ac high) or promoters (H3K4me1 low, H3K4me3 high, and H3K27ac high) in the middle, and at the bottom, we portrayed the ENCODE DNaseI hypersensitive site clusters derived from 125 cell types and transcription factor binding regions assessed by using ChIP-seq experiments. This analysis highlights that although the region upstream from the long isoform under steady-state conditions is hardly used in most of the cell lines present in the UCSC database, the short isoform promoter shows a high capacity to bind a number of different transcription factors (Fig 1, A). Hence we assessed the expression of the 2 isoforms in healthy skin and gut compartments. We designed primers specific for either lTSLP or shTSLP (see Table E1 in this article's Online Repository at www.jacionline.org). Consistently with the data from the genome browser, we detected much higher levels of shTSLP expression in steady-state conditions in both tissues (Fig 1, B and C). Furthermore, we separated intestinal epithelial cells and lamina propria cells from healthy colonic mucosa and found that shTSLP was predominant in both cases, indicating that neither of these 2 cellular populations preferentially produces the long isoform under steady-state conditions (Fig 1, B). Thus we describe that shTSLP is the homeostatic isoform of TSLP present under steady-state conditions in the gut and skin.

Fig 1.

shTSLP is the predominant isoform expressed on human intestinal and skin tissue. A, Schematic representation of the human TSLP locus from the UCSC Genome Browser (hg19). B and C, Quantitative real-time PCR analysis of shTSLP and lTSLP mRNA in intestinal total tissue, isolated intestinal epithelial cells (IECs), and lamina propria (LP) cells (Fig 1, B) and skin samples (Fig 1, C) from healthy subjects. ***P < .001.

shTSLP exerts anti-inflammatory effects in vitro

Cross-talk between epithelial cell and underlying antigen-presenting cell populations is a key interaction for the maintenance of tissue homeostasis. We described an important homeostatic role for intestinal epithelial cell–derived TSLP in controlling the inflammatory potential of DCs.^8,9 However, we always used tools that would not discriminate which isoform was the one responsible for this effect. Having shown that shTSLP is the primary isoform expressed under steady-state conditions, we postulated that shTSLP might be homeostatic and anti-inflammatory. To investigate this hypothesis, we cloned the 2 isoforms in insect cells using baculovirus technology and chemically synthesized the shTSLP (see Fig E1 in this article's Online Repository at www.jacionline.org and not shown). We then cocultured PBMCs from different donors in the presence of the indicated concentrations of the 2 isoforms (Fig 2, A). We used IFN-γ release in culture supernatants as the main readout for the assay. Contrary to lTSLP, the presence of shTSLP in coculture medium reduced IFN-γ release in a dose-dependent fashion (Fig 2, A). We then analyzed whether shTSLP had anti-inflammatory properties on DCs. We derived DCs from peripheral blood monocytes (monocyte-derived dendritic cells [moDCs]) and incubated them for 24 hours with shTSLP before challenge with Salmonella typhimurium. We found that shTSLP-conditioned moDCs were strongly inhibited in their capacity to produce cytokines (Fig 2, B). We found that shTSLP did not affect the phenotype of mature DCs in response to Salmonella typhimurium (see Fig E2 in this article's Online Repository at www.jacionline.org).

Fig 2.

shTSLP has anti-inflammatory properties. A, IFN-γ production in a bidirectional mixed lymphocyte reaction in the presence of shTSLP and lTSLP. Results of 3 independent experiments are shown. B, Inflammatory cytokine secretion of moDCs conditioned with shTSLP and then infected with Salmonella typhimurium. Results of 4 independent experiments are shown. Data are expressed as a “fold change” considering untreated PBMCs or cells treated only with Salmonella typhimurium as a control (value = 1). Significant differences were determined by means of 1-way ANOVA with the Dunnett test. *P < .05 and **P < .01.

The molecular pathway of lTSLP signaling is well characterized; it mediates its effects after binding to its own receptor, a complex formed by thymic stromal lymphopoietin receptor (TSLPR) and the IL-7 receptor α chain,^24,25 and phosphorylation of signal transducer and activator of transcription 5 (STAT5) is a key downstream event.²⁶ To elucidate the pathways of shTSLP signaling, we used a human phosphokinase array that analyzes phosphorylation of 46 intracellular kinases (Human Phospho-Kinase Array, R&D Systems) after 5 and 20 minutes of stimulation (Fig 3, A). shTSLP induced phosphorylation of a limited number of molecules, including p38α, extracellular signal-regulated kinase 1/2, and Lyn, but had only a mild effect on STAT5 phosphorylation (Fig 3, A). The capacity of shTSLP to activate p38α was confirmed by means of Western blotting in 4 different moDC donors (Fig 3, B). We then performed the same kinase array on LPS-stimulated moDCs previously conditioned or not with shTSLP to assess whether it was modulating Toll-like receptor (TLR)–mediated signaling. Interestingly, the level of p38α and extracellular signal-regulated kinase 1/2 phosphorylation was decreased when DCs were preconditioned with shTSLP (see Fig E3 in this article's Online Repository at www.jacionline.org), suggesting that it affects LPS-mediated mitogen-activated protein kinase (MAPK) activation. We could speculate that the slight phosphorylation of p38 MAPK induced by shTSLP might either desensitize the cells against further activation or raise the threshold of moDC activation, thus resulting in decreased cytokine production after TLR engagement.

Fig 3.

shTSLP increases p38α phosphorylation in moDCs. A, Human phosphokinase array assay of moDCs left untreated or conditioned for 5 and 20 minutes with 100 ng/mL short TSLP. B, Quantification of p38α phosphorylation in untreated and shTSLP-stimulated moDCs by means of Western blotting: representative membranes from 2 donors (upper panel) and quantification analysis of 4 independent donors (lower panel). Values are normalized against total p38α protein levels and compared with untreated cells.

moDCs do not express TSLPR, and thus to test the capacity of both isoforms to induce TSLPR-mediated signaling, we switched to myeloid blood-isolated CD1c⁺ DCs that are TSLPR⁺. We found that only lTSLP induced STAT5 phosphorylation (Fig 4, A). This did not exclude the possibility of shTSLP binding to TSLPR and acting as an antagonist, blocking signal from the lTSLP. However, when myeloid dendritic cells (mDCs) were stimulated with the 2 isoforms, either alone or together in an equimolar ratio, shTSLP had no capacity to block or inhibit lTSLP-dependent STAT5 phosphorylation (Fig 4, A). This suggests that shTSLP does not bind to TSLPR and is not dominant negative. Indeed, only lTSLP significantly upregulated CCL17 and CCL22 expression, as expected from TSLP activation,¹⁰ and this was not affected by the presence of shTSLP (Fig 4, B). Consistently, lTSLP-conditioned mDCs were inflammatory and significantly upregulated secretion of TNF-α from naive T cells in an allogeneic mixed lymphocyte reaction, both in the presence and absence of shTSLP (Fig 4, C). However, shTSLP is capable of exerting an anti-inflammatory activity on mDCs because shTSLP-conditioned mDCs were impaired in inducing IFN-γ secretion by T cells during the mixed lymphocyte reaction (Fig 4, D), a property not shared by lTSLP (data not shown), indicating that mDCs respond to shTSLP in an anti-inflammatory way.

Fig 4.

Effects of shTSLP on mDCs in vitro. A, Fluorescence-activated cell sorting analysis of phospho-STAT5 on mDCs left untreated or incubated with shTSLP, lTSLP, or both. B, ELISA of CCL17 and CCL22 levels in supernatants of mDCs treated as in Fig 4, A. Results of 2 independent experiments are shown. The same cells were washed and cultured with naive CD4⁺ T cells. C and D, Intracellular staining for TNF (Fig 4, C) and detection of IFN-γ in supernatants (Fig 4, D) were performed after 6 days of coculture. Fig 4, C, Results of 4 independent experiments are shown. Fig 4, D, Data show a representative of 3 independent experiments. Significant differences were determined by using 1-way ANOVA with the Tukey test. In Fig 4, A and C, a significant increase over untreated cells is indicated by asterisks. *P < .05 and ***P < .001.

shTSLP prevents endotoxin shock and ameliorates experimental colitis in vivo

In the mouse a short TSLP isoform has not been described or annotated in RefSeq. However, the C-terminal region of mouse TSLP shares 40% identity with human shTSLP (see Fig E4 in this article's Online Repository at www.jacionline.org). Thus we analyzed whether human shTSLP might have anti-inflammatory properties also in the mouse in vivo. To this end, C57/BL6 mice were injected intraperitoneally with shTSLP 18 and 2 hours before injection with LPS. Six hours after LPS injection, mice were killed, and cytokines were measured in the sera. Prophylactic injection with shTSLP led to a significant decrease of IL-6, IL-12/23p40, and IFN-γ levels in a dose-dependent fashion (Fig 5, A). Because we found that in the human gut TSLP is predominant at steady-state conditions, we assessed whether it might exert a protective role during gut perturbations, such as during experimental colitis. We administered shTSLP to mice every other day to evaluate whether it could protect against DSS-induced colitis. Mice treated with 200 μg of shTSLP displayed significantly reduced weight loss and faster recovery (Fig 5, B), suggesting a protective role during gut-induced inflammation.

Fig 5.

shTSLP has anti-inflammatory effects in vivo. A, Cytokine levels in the sera of mice treated intraperitoneally with shTSLP 18 hours and 2 hours before inducing LPS shock. Blood was collected 6 hours later. Results of 2 independent experiments are shown. Significant differences were determined by using 1-way ANOVA with the Dunnett test. B, Body weight changes in mice with DSS-induced colitis treated with shTSLP or vehicle intraperitoneally. The graph shows the percentage of body weight relative to initial body weight. *P < .05, **P < .01, and ***P < .001.

Taken together, these data support a homeostatic/anti-inflammatory role of shTSLP in vitro and in vivo.

Short/long TSLP ratios are affected in gut disorders

We previously described that intestinal epithelial cells isolated from patients with Crohn disease have lost the ability to express TSLP by using primers that do not distinguish between the 2 isoforms.^8,9 Knowing that lTSLP is not expressed under steady-state conditions (Fig 1, B), we now know that the reduction in TSLP expression was attributable to the lack of shTSLP expression. Hence we evaluated whether there was a deregulation of TSLP expression in patients with other inflammatory bowel disorders. We analyzed the expression of TSLP isoforms in cells isolated from patients with UC and in total tissue biopsy specimens from patients with CD (see Table E2 in this article's Online Repository at www.jacionline.org). We found that lTSLP and TSLPR were significantly upregulated in tissues from patients with UC compared with levels seen in healthy colonic mucosa, whereas shTSLP expression was unchanged (Fig 6, A, and see Fig E5, A, in this article's Online Repository at www.jacionline.org). These data were confirmed at the protein level through immunofluorescence staining and Western blotting with an antibody directed specifically to lTSLP. We generated a polyclonal antibody targeted to the N-terminal region of lTSLP that is not present in shTSLP (see Fig E1). This antibody did not detect any protein in healthy tissue sections, whereas it stained positive on biopsy specimens from patients with UC (Fig 6, B) both in the isolated epithelial cell and lamina propria compartments. By contrast, an antibody that recognizes both isoforms stained positive on healthy tissues, indicating the presence of shTSLP (Fig 6, B). We used the same antibodies for Western blot analysis of protein lysates from biopsy specimens of healthy subjects (n = 9) and patients with UC (n = 9). The antibody that recognizes both isoforms can be used to identify shTSLP based on its molecular weight. We confirmed that lTSLP expression is greatly upregulated in patients with UC. shTSLP is present in tissues from both healthy subjects and patients with UC, and interestingly, it shows one additional band (about 6 KDa), probably because of cleavage, degradation, or both, particularly in UC samples (see Fig E6 in this article's Online Repository at www.jacionline.org). Hence in healthy conditions the only isoform expressed is the shTSLP, whereas under inflammation, the lTSLP appears.

Fig 6.

TSLP isoforms are differentially expressed in the intestines of patients with UC and those with CD compared with healthy subjects and differentially modulated in epithelial cells. A and C, Quantitative PCR analysis of shTSLP and lTSLP in epithelial and LP cells from tissues of healthy subjects (H; n = 26) patients with UC (n = 13; Fig 6, A) or biopsy specimens of patients with untreated CD (UCD; n = 13), patients with treated CD (TCD, n = 15), patients with CD, and healthy subjects (n = 13; Fig 6, C). B, Representative immunofluorescent stainings of lTSLP (upper panels, red), both TSLP isoforms (bottom panels, red), and CD14⁺ cells (green) in sections of colon. D, Quantification of TSLP isoforms protein expression in untreated or infected Caco-2 cells. *P < .05, **P < .01, and ***P < .001.

We then analyzed the expression of the 2 isoforms in tissue from patients with CD either untreated or under a gluten-free diet (treated). In this case both isoforms were significantly downregulated in patients with untreated CD compared with those with treated CD and biopsy specimen from healthy subjects (Fig 6, C). TSLPR expression instead was not modulated between the groups (see Fig E5, B). The data are referred to cytokeratin 18 expression to take into consideration a possible reduction in the number of epithelial cells in untreated patients. This indicates that intestinal inflammatory disorders are characterized by an alteration of the ratio between shTSLP and lTSLP.

TSLP isoform expression in epithelial cells is regulated by proinflammatory stimuli

Given the divergent properties of TSLP isoforms, we analyzed whether there were conditions that affected TSLP expression and that could be of relevance to the inflammatory gut disorders in which the ratio of the 2 isoforms is altered. We first assessed basal levels of expression of the 2 isoforms by the intestinal epithelial cell line Caco-2 using real-time PCR, and we noticed an identical trend to the one we observed in primary intestinal epithelial cells (see Fig E7, A, in this article's Online Repository at www.jacionline.org). We confirmed these data also at the protein level by means of Western blot analysis on noninfected Caco-2 cells (see Fig E7, B). However, this pattern was completely reversed when Caco-2 cells were challenged with Salmonella typhimurium. In this case a significant upregulation of lTSLP expression, accompanied by a downregulation of shTSLP expression, was observed (Fig 6, D). A similar trend of lTSLP expression upregulation was observed when we used the adherent-invasive E coli strain LF82, although this did not reach statistical significance (Fig 6, D). Interestingly, the E coli nonpathogenic strain MG1655 had no effect on both isoforms.

TSLP isoforms are deregulated in patients with AD and skin keratinocytes exposed to inflammatory stimuli

When we analyzed the expression of TSLP isoforms in skin inflammatory disorders, such as AD, we observed an even more complex picture. As expected from the literature,¹⁰ we found an upregulation of the lTSLP isoform in lesional as opposed to nonlesional biopsy specimens (Fig 7, A); unexpectedly, we found that shTSLP was significantly downregulated in lesional biopsy specimens (Fig 7, A). No difference was observed in TSLPR expression (see Fig E5, C). Immunofluorescence staining of tissue sections with the 2 antibodies confirmed the results obtained by means of qPCR, even though technically we could not demonstrate a reduction in the expression of shTSLP in lesional skin (Fig 7, B). This indicates that we have a further unbalance of the 2 isoforms in patients with AD because we find a concomitant downregulation of shTSLP and upregulation of lTSLP.

Fig 7.

shTSLP and lTSLP are differentially expressed in skin tissue of patients with AD. A, Quantitative PCR analysis of shTSLP and lTSLP in skin biopsy specimens from nonlesional (NL) and lesional (L) skin of patients with AD (n = 24). B, Representative immunfluorescencent staining of lTSLP (upper panels, red) and both TSLP isoforms (bottom panels, red) in tissue sections of nonlesional and lesional skin. C and D, ChIP–quantitative PCR analysis of H3K27Ac-enriched regions in the promoters of the TSLP isoforms (Fig 7, C) and mRNA levels of TSLP isoforms (Fig 7, D) in untreated HEKa or those stimulated with vitamin D3 (10 μmol/L) and poly I:C (5 μg/mL). *P < .05, **P < .01, and ***P < .001.

We then examined the effect of proinflammatory and anti-inflammatory stimuli in the 2 isoforms' expression on human primary keratinocytes, HEKa cells. Vitamin D deficiency has been associated with AD severity in children^27,28 and adult Koreans.²⁹ Furthermore, putative vitamin D response elements have been found in the upstream region of lTSLP³⁰ in addition to the well-known nuclear factor κB (NF-κB)–responsive element that allow its expression in response to TLR ligands.³¹ Thus we challenged HEKa cells, which were previously starved for 24 hours, with either vitamin D3 (10 μmol/L) or polyinosinic:polycytidylic acid (poly I:C; 5 μg/mL), a TLR3 agonist, and we performed chromatin immunoprecipitation and quantitative PCR analysis of H3K27 acetylated long and short promoter regions. We observed that the promoter of shTSLP was more active than that of lTSLP in untreated cells (Fig 7, C). Interestingly, vitamin D3 induced enrichment in H3K27Ac in the promoter of both lTSLP and shTSLP, resulting in increased transcription primarily of the shTSLP isoform (Fig 7, C and D). On the other hand, poly I:C induced H3K27ac deposition only in the promoter of lTSLP, and this resulted in a stronger induction of lTSLP mRNA compared with vitamin D3 stimulation, even though the acetylation level was lower (Fig 7, C and D). This indicates that poly I:C and presumably NF-κB activate only the lTSLP promoter and result in strong lTSLP mRNA transcription. By contrast, vitamin D3 activates both regions and induces shTSLP transcription but is not sufficient for lTSLP transcription. A similar result was obtained with the keratinocyte cell line HaCaT (see Fig E8 in this article's Online Repository at www.jacionline.org) and is consistent with the literature on primary keratinocytes, where an upregulation of total TSLP mRNA, but not lTSLP mRNA, was shown in response to vitamin D3.³² Thus we can conclude that isolated epithelial cells and skin keratinocytes express lTSLP only in response to invasive bacteria or proinflammatory stimuli, whereas expression of shTSLP is constitutive but can be upregulated by anti-inflammatory mediators (vitamin D3) or downregulated by pathogenic bacteria (Salmonella typhimurium).

Discussion

Here we have shown that the dilemma of TSLP mediating proinflammatory and anti-inflammatory activities can be explained by the existence of 2 isoforms of TSLP that are not alternatively spliced, but each one is driven by a differentially regulated promoter region. At the protein level, shTSLP encompasses the last 63 residues of lTSLP. Consistent with a higher activity of shTSLP promoter in untreated cells, shTSLP is the primary isoform expressed under steady-state conditions and is anti-inflammatory and homeostatic, whereas lTSLP is pathogenic and induced by inflammatory stimuli. shTSLP does not seem to have antagonistic activity on lTSLP but can condition both mDCs and moDCs to limit their inflammatory potential, probably by inducing a suboptimal activation of MAPK, at least on moDCs. Consistently, shTSLP does not induce or inhibit signaling through the receptor of lTSLP, suggesting a signaling pathway that is independent on TSLPR, which is not expressed on moDCs at steady-state conditions.

The existence of a parallel and molecularly independent mechanism underlying the anti-inflammatory functions of shTSLP is fully supported by structural analysis of the murine TSLP–TSLPR–IL-7 receptor α complex.³³ Mouse TSLP, which is the ortholog of human lTSLP, is 140 residues long and folds as a 4-helix bundle, stabilized by 3-disulphide bonds and several hydrophobic interactions contributed by the N-terminal portion of the protein. Because this region is not present in shTSLP, we predict that shTSLP adopts a significantly different conformation than lTSLP, which might explain why it uses a different receptor. The 3-dimensional structure of mouse and human lTSLP is likely conserved because most of the residues involved in the scaffold of TSLP are conserved³³ and there is a 40% amino acid identity in the sequences of the regions corresponding to shTSLP between mice and human subjects. Hence even though in the mouse there is no open reading frame for shTSLP, this protein might still be generated by protease cleavage from lTSLP. Hence it is not surprising that, when tested in mice, human shTSLP could protect them from endotoxin shock and DSS colitis, suggesting an anti-inflammatory role in vivo.

The 2 isoforms are modulated during inflammation. In particular, invasive pathogenic bacteria, such as Salmonella typhimurium or adherent-invasive E coli strain, can upregulate the expression of the lTSLP in intestinal epithelial cells, whereas Salmonella typhimurium is also capable of dowregulating shTSLP. Because E coli with adhesive and invasive capacities has been associated with inflammatory bowel disease,^34,35 this might explain why there is an upregulation of lTSLP in patients with UC. By contrast, vitamin D3 can activate the promoter and upregulate the expression of shTSLP in skin keratinocytes. This is an important observation because we found shTSLP expression to be downregulated in patients with AD, and a negative correlation between vitamin D3 levels and the severity of AD has been shown.^27-29 Hence one of the mechanisms of action of vitamin D3 supplementation, which has been shown to be beneficial in the treatment of AD,³⁶ might be through the re-establishment of a homeostatic condition through upregulation of the shTSLP. In patients with CD, we found a downregulation of TSLP levels that are normalized when patients are on a gluten-free diet, indicating that this is probably not a primary defect in these patients.

Interestingly, because we found that shTSLP is associated with the control of inflammatory mediator release both in vitro and in vivo and with a reduction of IFN-γ production, we can hypothesize that conditions leading to a reduction in shTSLP expression could lead to an uncontrolled T_H1 type of responses, such as in patients with CD. By contrast, when lTSLP is upregulated, such as in patients with AD and UC, presumably through NF-κB activation, which is a hallmark of inflammation in both inflammatory bowel disease and AD,^37-39 a T_H2 component is induced. In addition, 2 recent reports have shown an antimicrobial activity of peptides within the shTSLP sequence, indicating a possible additional homeostatic activity of shTSLP in vivo.^40,41 Hence in a personalized view of treatment for patients with inflammatory disorders, one should take these 2 aspects into account and might on the one hand block the inflammatory potential of lTSLP when this is upregulated and on the other hand re-establish immune homeostasis through administration of shTSLP or drugs involved in its upregulation, when shTSLP is downregulated. Additionally, caution should be taken when using in therapy antibodies or biologicals that do not distinguish between the 2 isoforms and that might elicit unwanted effects by targeting shTSLP.

Key messages.

• •Human TSLP exists in 2 distinct isoforms: shTSLP and lTSLP.
• •shTSLP is the only isoform expressed in healthy skin and gut tissue, whereas lTSLP is present in diseased inflammatory conditions.
• •The shTSLP/lTSLP ratio is altered in patients with UC, CD, and AD.
• •shTSLP, contrary to lTSLP, has an anti-inflammatory effect both in vitro and in vivo.
• •The dual role of TSLP isoforms could explain the literature paradox of human TSLP being both proinflammatory and tolerogenic.

Supplementary data

Fig E1.

Fig E2.

Fig E3.

Fig E4.

Fig E5.

Fig E6.

Fig E7.

Fig E8.