The arachidonic acid metabolite 11β-ProstaglandinF2α controls intestinal epithelial healing: deficiency in patients with Crohn’s disease

In healthy gut enteric glial cells (EGC) are essential to intestinal epithelial barrier (IEB) functions. In Crohn’s Disease (CD), both EGC phenotype and IEB functions are altered, but putative involvement of EGC in CD pathogenesis remains unknown and study of human EGC are lacking. EGC isolated from CD and control patients showed similar expression of glial markers and EGC-derived soluble factors (IL6, TGF-β, proEGF, GSH) but CD EGC failed to increase IEB resistance and healing. Lipid profiling showed that CD EGC produced decreased amounts of 15-HETE, 18-HEPE, 15dPGJ2 and 11βPGF2α as compared to healthy EGC. They also had reduced expression of the L-PGDS and AKR1C3 enzymes. Produced by healthy EGC, the 11βPGF2 activated PPARγ receptor of intestinal epithelial cells to induce cell spreading and IEB wound repair. In addition to this novel healing mechanism our data show that CD EGC presented impaired ability to promote IEB functions through defect in L-PGDS-AKR1C3-11βPGF2α dependent pathway.

Compelling evidence has demonstrated that defects in mucosal healing are central to Crohn's Disease (CD) pathogenesis and prognosis. In particular several studies have reported that intestinal mucosal lesions precede inflammation and are considered as a predictive factor of relapse [1][2][3] . Mucosal healing has further been suggested to represent a treatment goal and a predictive factor for sustained clinical remission in CD [4][5][6] .
The intestinal epithelium is a dynamic interface between the environment and the organism that must constantly preserve its integrity to maintain digestive and barrier functions. After injury, mucosal repair is a key process to restore epithelium lining and functions such as permeability control leading ultimately to intestinal homeostasis. Three main concomitant regenerative processes participate to mucosal healing and include epithelial restitution, that involves cell spreading and migration into the wound, cell proliferation and differentiation 7 .
It has now been well demonstrated that IEB functions, including intestinal healing, are regulated by neighboring cells, the so-called microenvironment, and in particular the enteric nervous system (ENS) 8 .
The ENS is an integrative neuronal network localized along the gastrointestinal tract that regulates key digestive functions such as gut motility and mucosal secretion 9,10 . ENS is composed of enteric neurons and enteric glial cells (EGC) that outnumber enteric neurons by a factor of 4 to 10 11 and share common markers and functional properties with central nervous system astrocytes [12][13][14][15][16] . EGC form a dense network that surrounds intestinal crypts, and are located at less than 2 μ m from intestinal epithelial cells (IECs). A large number of studies from our group and others have now well demonstrated that EGC are key regulators of IEB homeostasis and functions 17,18 . In particular EGC impact IEC major functions via paracrine signaling. For instance they inhibit IEC Primary cultures of EGC from CD patients have lost their ability to speed up IEB healing. In previous studies we have shown that EGC stimulate intestinal epithelial cell monolayer healing in part through spreading acceleration. These effects are associated with an increase in Caco-2 cell transepithelial electrical resistance (TER), indicative of increased sealing of the monolayer 21 . To determine whether CD EGC have impaired functional impact on IEB, we compared effects of EGC isolated from CD patients vs. control patients on Caco-2 cell TER, spreading and healing in a non-contact co-culture model. Using an in vitro mechanically induced wound-healing model, we showed that while control EGC enhanced wound closure, CD EGC had no significant impact (Fig. 2a,b). Consistent with previous work 29 , human control EGC significantly increased Caco-2 cell spreading and monolayer resistance (Fig. 2c-e). In contrast, CD EGC did not impact spreading (Fig. 2c,d) or TER (Fig. 2e) as compared to Caco-2 monolayer cultured alone. These results demonstrate that EGC from CD patients have impaired effects on IEB healing. CD EGC express the same level of IL-6, TGF-β1, EGF and GCLc than control EGC. We next hypothesized that CD-associated impairment of EGC functions was due to altered soluble factor production. Thus, we first quantified four EGC-derived soluble factors known to regulate IEB functions. Those included Interleukin 6 (IL6), TGF-β 1, EGF and glutathione. Control EGC and CD EGC expressed similar levels of IL6 (Fig. 3a) and TGF-β1 (Fig. 3c) mRNA (p = 0.21 and p = 0.06 respectively), and released similar concentration of IL6 (Fig. 3b) and TGF-β 1 (Fig. 3d) (p = 0.74 and p = 0.48 respectively). EGF (Fig. 3e) and glutamate cysteine ligase catalytic subunit (GCLc; Fig. 3g) mRNA were expressed at similar levels in control and CD EGC. Consistently proEGF (Fig. 3f) and glutathione (Fig. 3h) proteins were produced at equivalent levels by control and CD EGC.
CD EGC express less L-PGDS and AKR1C3 than control EGC. Both 15dPGJ 2 and 11β PGF 2 α derive from transformation of PGD 2 , which is synthetized by lipocalin like prostaglandin D synthase (L-PGDS) and cyclooxygenases (COX) enzymes. The aldo-keto reductase family 1 member C3 (AKR1C3) converts PGD 2 to 11β PGF 2 α . To assess if this pathway was altered in CD EGC, we measured Cox-1, Cox-2 and L-PGDS but also AKR1C3 mRNA expression. In CD EGC, while Cox-1 (Fig. 4a) and Cox-2 ( Fig. 4b) mRNA levels were unchanged, L-PGDS (Fig. 4c) and AKR1C3 (Fig. 4d) mRNA expressions were significantly reduced in CD EGC as compared to control EGC. We further demonstrated that the protein expression of AKR1C3 was significantly reduced in CD-EGC as compared to control EGC (Fig. 4e,f). These results show for the first time that human EGC expressed the enzyme AKR1C3 but especially indicate that L-PGDS and AKR1C3 pathway is down regulated in EGC derived from CD patients that are thereby defective for 11β PGF 2 α production. 11βPGF 2 α reproduces control EGC impact on IEC healing and spreading. The impact of glial 15dPGJ 2 on IEC was already described 19 , therefore we focused on 11β PGF 2 α effects on IEB wound closure. IEB healing was significantly accelerated by 11β PGF 2 α (Fig. 5a,b), and IEC spreading was also strongly enhanced by 11β PGF 2 α (Fig. 5c). To assess whether 11β PGF 2 α could restore CD EGC function, we have added 11β PGF 2 α in the culture medium of EGC co-cultivated with Caco-2 monolayer.11β PGF 2 α supplementation in CD EGC -Caco-2 co-cultures significantly increased IEB healing but has no additional effect when added in control EGC -Caco-2 co-cultures (Fig. 5d). These data show that the functional defect presented by CD EGC could be fixed by 11β PGF 2 α addition.
As it has been proposed that AKR1C3 or the prostaglandin D2 and its derivatives may regulate ligand access to the orphan nuclear receptor PPARγ 31 , we have analyzed whether 11β PGF 2 α could activate PPARγ . Immunocytochemistry showed a transient increase of PPARγ nuclear localization after 5min of IEC treatment with 11β PGF 2 α (Fig. 5e). PPARγ expression was also significantly increased after 24 hours of IEC treatment with 11β PGF 2 α (Fig. 5f,g). These data demonstrate that 11β PGF 2 α reproduced control EGC effects on IEB, namely IEC spreading and IEB healing promotion, and activate the orphan nuclear receptor PPARγ .
PPARγ agonist induces IEC spreading and IEB healing. We then assessed the impact of PPARγ agonist on IEB healing and TER, and on IEC spreading. The PPARγ agonist, rosiglitazone, resulted in an accelerated healing (Fig. 6a,b) and an increased cell spreading (Fig. 6c,e) and TER (Fig. 6d). These results show that as 11β PGF 2 α , PPARγ agonist enhances IEC spreading and IEB healing.

Functional defect of CD EGC on IEB properties involved DP2 and PPARγ-dependent pathways.
To assess if 11β PGF 2 α is the mediator of EGC regulation of IEB, and to unravel how it does, we blocked 11β PGF 2 α dependent pathways in our co-culture model of IEB and control or CD EGC. As 11β PGF 2 α can act through PPARγ but also through FP or DP2 receptor, we used the selective FP antagonist (AL8810), DP2 antagonist (CAY10595) as well as PPARγ antagonist (GW9662). CAY10595 and GW9662 significantly blocked IEB healing induced by control EGC but had no effect on IEB healing when Caco-2 monolayer was cultivated in the presence of CD EGC (Fig. 7a). The AL8810 had no significant effect upon wound healing in any condition tested (not shown). Both CAY10595 and GW9662 also blocked the increase of IEB healing induced by 11β PGF 2 α (Fig. 7b). These data show that DP2 and PPARγ -dependent pathways are responsible for IEB healing induced by 11β PGF 2 α glial production, and that CD EGC lack these regulations.

Discussion
The main goal of this work was to assess whether primary cultures of EGC from CD patients have altered impact on IEB healing, compared to EGC from control patients. A second goal of this study was to better characterize human EGC lipidic secretome and to study its possible impact on IEB repair. Our data unravel that human EGC from control patients produce the prostaglandin 11β PGF 2 α that induces IEC spreading and accelerates IEB wound healing. We demonstrate that 11β PGF 2 α acts through DP2-and PPARγ -dependent pathways. In addition we demonstrate that in contrast to control EGC, CD EGC do not enhance TER, spreading or wound closure, indicating that CD EGC have lost their ability to accelerate sealing and healing. We show that this defect is due to a down-regulation of L-PGDS and AKR1C3 expression and a reduced production of 11β PGF 2 α (Fig. 8).
One of the main findings of this work is that CD EGC did not accelerate sealing of epithelial monolayer, did not stimulate cell spreading and as a result, did not promote wound closure, as opposed to their healthy homologues. These data strongly implies that defects in mucosal healing observed in CD could be, at least in part, due to EGC that have lost their abilities to promote cell restitution. Altogether our ex vivo functional studies indicate that primary cultures of EGC isolated from CD patients exhibit loss of function as compared to EGC isolated from control patients. These findings suggest that CD EGC have altered intrinsic functional phenotype. Regarding the recent work of Pochard et al., that shows that CD EGC have decreased properties to control IEB permeability 28 , it is thus tempting to speculate that impaired intrinsic functions of EGC from CD patients predispose and/or directly participate to the disease onset or favor relapse. Another important result of this work is the description of human EGC lipidic signature. In previous studies we have already described the rat EGC lipidic signature 28 and have shown that 15dPGJ2 was produced by rat EGC 19 , and the prostaglandin PGE 2 has already been shown to be produced by enteric glia 32 as well as by central glial cells [33][34][35][36] . In the present study, among 31 PUFA metabolites measured in human EGC conditioned medium, we detected 19 other PUFA derivatives, in addition to these two prostaglandins previously described. These derivatives are produced by the activation of three major AA metabolic pathways, i.e. COX; LOX and p450 signaling pathways, which are active in EGC. According to the low stability of PUFA catabolic products, we cannot rule out that the PUFA derivatives that we did not detect are not produced in EGC. Concerning the role of these metabolites, little is known about their specific regulation of the neuro-glio-epithelial unit. We have already shown that glial production of 15dPGJ 2 exerts anti-proliferative and pro-differentiating effects on IEC 19 . Beside its impact on IEC, 15dPGJ 2 can regulate other cells of the glial microenvironment. For instance, 15dPGJ 2 is a neuroprotective agent [37][38][39] and is believed to be responsible for neutrophil recruitment 40 . On another hand, the prostaglandin PGD 2 and the COX pathway have been described to regulate enteric nervous system excitability 41,42 , and more specifically glial production of PGE 2 has been shown to potentiate neuronal response to bradykinin 32 . In this study we concentrated on determining the role of 11β PGF 2 α , the primary plasma metabolite of PGD 2 in vivo 43 . Up to now 11β PGF 2 α had no specific intestinal function. It inhibits ADP-or thrombin-induced human platelet aggregation 44 , induces human bronchial smooth muscle contractions 45 , inhibits adipocyte differentiation 46 and promotes prostate cell proliferation 47 . To the best of our knowledge this is the first study that demonstrates that 11β PGF 2 α is produced in the intestine by EGC and that it massively increases IEC spreading, enhances cell restitution and thereby wound closure.
Described as equipotent to PGF 2 α , 11β PGF 2 α is believed to act through FP receptors, although this still needs to be clearly demonstrated. We did not observed significant effect of FP antagonist on IEB healing but have demonstrated that 11β PGF 2 α , as well as EGC, induced healing through type 2 PGD 2 receptor (DP2)-and peroxysome-proliferator activated receptor (PPARγ )-dependent pathways (Fig. 7). Besides canonical pathways such as Wnt/β -catenin or Notch, recent data have shown that PPARγ signaling is also a key pathway in the control of IEC functions 48 . For instance, PPARγ activation inhibits IEC proliferation and promotes cell differentiation 49,50 .
In the present study we demonstrate that PPARγ activation, induced by EGC or not, can promote IEC spreading and IEB wound closure. The barrier-promoting impact of the 11β PGF 2 α -PPARγ -pathway that we have observed in the intestine, could also be of interest in other epithelial tissues, and could explains protective effects of this pathway against lethal influenza infection with lung viral load reduction for example 51 .
In addition we have compared human EGC lipidic secretome from control versus CD patients. Among the 21 PUFA metabolites measured in control EGC conditioned media, four of them are significantly less present in CD EGC conditioned media: the 15-HETE, the 18-HEPE, the 15dPGJ2 and the 11β PGF 2 α . In this study we have demonstrated that 11β -PGF 2 α promotes wound healing. As mentioned above, 15dPGJ2 has neuroprotective effects 37-39 and is already described to regulate IEC proliferation and differentiation 19 . 15-HETE regulates different vascular functions 52-57 and concerning effects on epithelium, 15-HETE has been shown to induce cell growth of pre-confluent non-differentiated intestinal epithelial cells 58 and to increase Caco-2 TER and decrease permeability 59 . We have very recently demonstrated that the glial production of 15-HETE is responsible for the control of ZO-1 expression and IEB permeability, and that this regulation is lost in CD 28 . Very little is known about 18-HEPE role, but one paper suggests that 18-HEPE could play a role in mucosal repair 60 . Thus, it is likely that CD EGC deficit in 15-HETE, 18-HEPE, 15dPGJ2 and 11β PGF 2 α production overall participates to abnormalities/alterations observed during CD 1-3,6,26,61-63 .
Finally, our study suggest that 11β PGF 2 α reduced production by CD EGC results from reduced L-PGDS and AKR1C3 expression in CD-EGC as compared to control EGC. Indeed 15dPGJ 2 and 11β PGF 2 α derive from prostaglandin D2 (PGD2), which results from prostaglandin H2 (PGH2) isomerization by prostaglandin D synthase (PGDS). While IEC, mast cells and fibroblasts express H-PGDS, the most representative form of PGDS in the gut 64,65 EGC express the lipocalin-PGDS (L-PGDS) 19 We have shown that EGC also express the aldo-keto reductase AKR1C3 that converts PGD2 to 11β -PGF 2 α . First identified as an enzyme involved in steroid metabolism, AKR1C3 could metabolize a broad spectrum of carbonyl compounds and has broad role in the development of tumor diseases 66 . We could speculate that the role of glial AKRIC3 is wider than the production of 11β -PGF 2 α and the consequent induction of IEC spreading. The function and regulation of AKRIC3, as the ones of 15-HETE and 18-HEPE in EGC deserve further investigation.
Collectively our results demonstrate that EGC from CD patients exhibit loss of function as compared to control EGC, due to defects in L-PGDS, AKR1C3 and 11β PGF 2 α expression/synthesis. Thus this work confirms and extends the emerging major role of EGC on intestinal homeostasis and mucosal repair after injury 17,18,67 . Enteric glia not only regulates epithelial cells, but also forms a cellular and molecular bridge between enteric nerves 68,69 , enteroendocrine cells 70,71 and immune cells 72 that lets suspect even broader impact of EGC upon intestinal Figure 7. Implication of DP receptors and PPARγ in IEB healing induced by EGC. DP (CAY10595) and PPARγ (GW9662) antagonists were applied directly on IEC cultivated alone, on IEC co-cultivated with Control EGC or on IEC co-cultivated with CD EGC during three days. Quantification of epithelial healing was calculated as the difference between the percentage of healing between t0 and 48 h for co-cultures and the percentage of healing between t0 and 48 h for Caco-2 cultivated alone (b) CAY10595 and GW996 were also added to IEC culture in presence of 11β PGF 2 α . Quantification of epithelial healing was calculated as a percentage of healing between t0 and 48 h and compared to 11β PGF 2 α healing considered as one. n = 4-7 independent experiments; Kruskal-Wallis test; *p < 0.05; **p < 0.01 as compared to without treatment.
Scientific RepoRts | 6:25203 | DOI: 10.1038/srep25203 functions. Our results not only imply that CD EGC could be actors of CD pathogenesis or development, but also establish a molecular basis for developing and testing therapeutic strategies by targeting 11β PGF 2 α production or DP2-PPARγ dependent pathways.

Enteric glial cells.
Cultures of human enteric glial cells (EGC) were obtained according to the procedure described by Soret et al. 29 . Briefly, human EGC originated from intestinal resections of control patients (patients having undergone surgery for colorectal cancer, 10cm from the tumor area) and from patients with an established diagnosis of CD according to international criteria. All intestinal resections were macroscopically healthy. 15 control (age 56-89 y; sex ratio 5men:10women) and 11 CD (age, 17-63 y; sex ratio 4men:7women) patients were included in this study. Patients gave their informed consent to take part in the study and all procedures were performed according to the guidelines of the French Ethics Committee for Research on Human and registered under the no. DC-2008-402. An immunocytochemistry study was performed to verify the purity of the EGC population as described by Soret et al. 29 . EGC cultures presenting more than 80% of GFAP-, Sox10-, and S100β positive cells were used for these experiments, at passage 3.  Human EGC can produce PUFA metabolites derived from n-3 or n-6. Among the 21 mediators secreted by human EGC, only two by-products of prostaglandin D2 (PGD 2 ), 15dPGJ 2 and 11β PGF 2 α were misproduced by CD EGC that express less L-PGDS and AKR1C3 but the same level of COX1 and COX2. 15dPGJ 2 and 11β PGF 2 α glial production regulates IEB healing through PPARg and DP2 dependent pathway.

Co
Scientific RepoRts | 6:25203 | DOI: 10.1038/srep25203 Corning, Avon, France) at 90 × 10 3 cells/cm 2 and in 24-well plates at 7,9 × 10 3 cells/cm 2 for three days. The IEC size was measured thanks to anti-zonula occludens-1 immunostaining (see Immunofluorescence staining). The transepithelial electrical resistance (TER) was measured every day with an epithelial volt-ohmmeter (EVOM, World Precision Instruments, Inc). For wound healing experiments, Caco-2 cells and EGC were respectively seeded onto 6-well filters at 122 × 10 3 cells/cm 2 and in 6-well plates at 2,1 × 10 3 cells/cm 2 for 15 days. Caco-2 monolayers were wounded with the 0.5-10 μ l tip of a pipette. Each hole was photographed at 0 and 48 h by using a microscope (Axiovert 200M; Zeiss; objective lens 5× ) with Axiovision software and surface was quantified with ImageJ software (National Institute of Health, Bethesda, Maryland, United States). Epithelial healing was calculated as the percentage of the healing after two days of treatments or EGC co-culture compared to t0.
Immunofluorescence staining and western blotting. IEC transwell filters were fixed in PBS 4% paraformaldehyde for 30 min. The IEC size was measured thanks to anti-zonula-occludens-1 (ZO-1) immunostaining. Briefly, fixed filters were incubated for 30 min at RT with PBS containing 0.5% Triton X-100 (Sigma-Aldrich, Saint-Louis, MO, USA) and 10% horse serum (Merk Millipore, Billerica, MA, USA; PBS-TX-HS) and then incubated with a mouse monoclonal antibody anti-ZO-1 diluted in PBS-TX-HS (1:500; Life Technologies) for 2 h at RT. After washing with PBS, filters were incubated with an anti-mouse CY-3 (1:500; Jackson ImmunoResearch, West Grove, PA, USA) for 45 min at RT. After washing with PBS containing DAPI (Sigma-Aldrich) for the first wash, filters were mounted on slides for fluorescence microscopy analysis. Images were acquired with a digital camera (Olympus DP 50) coupled to a fluorescence microscope (Olympus IX 50). Cell surface area was measured with ImageJ software. An average of 149.9+ /− 8.6 IEC was analyzed for each experimental condition. PPAR-γ nuclear localisation was studied by the same immunostaining procedure using anti-PPAR-γ antibody (sc-7273 diluted 1/500). The same antibody diluted 1/500 was used for western blotting procedure.
Real-time quantitative PCR analysis. IEC grown on filters were lyzed in RA1 buffer (Macherey-Nagel, Düren, Germany) in order to study mRNA and protein expression. According to the manufacturer's recommendations, total RNA extraction from cells was performed with Nucleospin RNAII kit (Macherey-Nagel) and 1 μ g purified RNA was denatured and processed for reverse transcription using Superscript II reverse transcriptase (Invitrogen). PCR amplifications were performed using the Absolute Blue SYBR green fluorescein kit (Roche) and run on a Rotor-Gene (Qiagen, Venlo, The Netherlands). The following primers (Sigma-Aldrich) were used: AKR1C3 (Aldo-keto reductase family 1, member C3) # NM_003739. Immunoblotting. Lysates from RA1 extraction were precipitated and pellets were resuspended in PSB/TCEP (Macherey-Nagel). Samples were processed for electrophoresis using NuPAGE MES SDS buffer kit (Life technologies) and separated on 4-12% Bis-Tris gels (NuPAGE, Life Technologies). Proteins were transferred to nitrocellulose membranes with the iBlot ™ system (Life Technologies). After blocking with TBS/0.1% Tween20/5% nonfat dry milk for 1 hour, blots were incubated overnight at 4 °C with primary antibodies diluted in TBS/5% nonfat dry milk for rabbit anti-proEGF (1:500; Pierce, Thermo Scientific), and mouse anti-β -actin (1:10000; Sigma-Aldrich). Immunoblots were probed with the appropriate horseradish peroxidase-conjugated secondary antibodies (Life Technologies) and visualized by chemiluminescence (ECL, Biorad, Hercules, CA, USA). Quantitative analysis was performed using ImageJ software. The value of protein immunoreactivity was normalized to the amount of β -actin immunoreactivity.
Treatment of IEC monolayers. IEC cultivated on filters as described in our "Co-culture model" section were also treated with 11β PGF 2 α (5 μ M; Cayman Chemical, Ann Arbor, MI, USA) or the PPARγ agonist (Rosiglitazone; 5 μ M; Sigma-Aldrich) for three days, and spreading, TER, or healing were measured as described above. IEC co-cultivated with control or CD EGC were also treated with FP receptor antagonist (AL8810; 1 μ M; Cayman Chemical), DP2 receptor antagonist (CAY10595; 10 nM; Cayman Chemical), or PPARγ antagonist (GW9662; 10 μ M; Sigma-Aldrich) for two days, and healing was measured as described above.
Statistics. Data are expressed as the mean ± SEM of three to ten independent experiments. The significance of differences was determined using a Mann-Whitney test to compare two groups and a one-way analysis of variance (ANOVA), Kruskal-Wallis test followed by a Dunn's post-test, to compare three groups or more. Differences were considered statistically significant for p < 0.05.

Study approval.
Patients gave their informed consent to take part in the study and all procedures were performed according to the guidelines of the French Ethics Committee for Research on Human. All experimental protocols were approved by local Committee on Ethics and Human Research and the Inserm (Institut national de la santé et de la recherche médicale) and registered under the no. DC-2008-402.