PGE2 maintains self-renewal of human adult stem cells via EP2-mediated autocrine signaling and its production is regulated by cell-to-cell contact

Mesenchymal stem cells (MSCs) possess unique immunomodulatory abilities. Many studies have elucidated the clinical efficacy and underlying mechanisms of MSCs in immune disorders. Although immunoregulatory factors, such as Prostaglandin E2 (PGE2), and their mechanisms of action on immune cells have been revealed, their effects on MSCs and regulation of their production by the culture environment are less clear. Therefore, we investigated the autocrine effect of PGE2 on human adult stem cells from cord blood or adipose tissue, and the regulation of its production by cell-to-cell contact, followed by the determination of its immunomodulatory properties. MSCs were treated with specific inhibitors to suppress PGE2 secretion, and proliferation was assessed. PGE2 exerted an autocrine regulatory function in MSCs by triggering E-Prostanoid (EP) 2 receptor. Inhibiting PGE2 production led to growth arrest, whereas addition of MSC-derived PGE2 restored proliferation. The level of PGE2 production from an equivalent number of MSCs was down-regulated via gap junctional intercellular communication. This cell contact-mediated decrease in PGE2 secretion down-regulated the suppressive effect of MSCs on immune cells. In conclusion, PGE2 produced by MSCs contributes to maintenance of self-renewal capacity through EP2 in an autocrine manner, and PGE2 secretion is down-regulated by cell-to-cell contact, attenuating its immunomodulatory potency.


Results
Proliferation of human adult stem cells is decreased by the inhibition of COX-2 or mPGES-1 via G 1 cell cycle arrest. Indomethacin, the inhibitor for both COX-1 and COX-2, interferes with epithelial and tumour cell growth 21,22 . Therefore, we first investigated whether COX inhibition affected the proliferation of hUCB-MSCs and found that indomethacin treatment significantly decreased hUCB-MSC proliferation (Suppl. Fig. S1). In previous studies, similar effects were observed when COX-2 was selectively inhibited in other cell types 23,24 . We next investigated whether selective inhibition of COX-2 using celecoxib or inhibition of membrane associated PGE synthase 1 (mPGES-1), a PGE 2 synthesizing enzyme downstream of COX-2 signalling, using cay10526 affected the proliferative phenotype of hUCB-MSCs and hAD-MSCs. Treatment down-regulated the expression of COX-2 or mPGES-1 at the protein level in a dose-dependent manner (Fig. 1a) Consistent with the decrease in the protein level, inhibition of PGE 2 producing enzymes resulted in the remarkable dose-dependent decrease in the proliferation of both hUCB-MSCs and hAD-MSCs (Fig. 2b). The cumulative proliferative phenotype in MSCs in response to chemical inhibitors was further confirmed by evaluating the cumulative population doubling level (Fig. 1c). When COX-2 expression was inhibited by small interfering RNA (siRNA), the same results were observed (Suppl. Fig. S1). In addition, we showed that this suppression of self-renewal influenced cell confluency and cellular morphology. Compared with non-treated cells, celecoxib or cay10526-treated cells exhibited flattened or spread out cell bodies with low confluency (Fig. 1d, marked as ▼ ). To determine whether these changes in proliferation resulted from the lack of PGE 2 , the PGE 2 secretion level was measured by enzyme-linked immunosorbent assay (ELISA). Conditioned media (CM) from COX-2-suppressed hUCB-MSCs contained less than 20% of the PGE 2 concentration of the naïve MSC group (Suppl. Fig. S1). Inhibition of PGE 2 production led to lower proliferation via G 1 cell cycle arrest. The proportion of cells in G 1 phase gradually increased dose-dependently, whereas the proportion of cells in S phase decreased (Fig. 1e). The apoptotic rate of hUCB-MSCs was not affected by the suppression of PGE 2 synthesis ( Fig. 1f and Suppl. Fig. S1). We next investigated whether the suppression of COX-2/PGE 2 axis influence the other properties of MSCs. The expression pattern of surface antigens on hUCB-MSC was not altered after the treatment of celecoxib or cay10526 (Suppl. Fig. S2). Moreover, the expression levels of pluripotency marker genes in hUCB-MSCs were not significantly changed by the inhibition of the COX-2 (Suppl. Fig. S2). In addition, we found that up-regulation of COX-2/ PGE 2 signalling enhanced osteogenesis of hUCB-MSCs, in contrast, suppressed adipogenesis using RT-PCR and Scientific RepoRts | 6:26298 | DOI: 10.1038/srep26298 specific staining after the induction of differentiation. There was no significant change in chondrogenic differentiation (Suppl. Fig. S2). Taken together, these findings indicate that the COX-2/PGE 2 axis has a critical role in the maintenance of hMSC self-renewal and that down-regulation of this axis leads to cell cycle arrest in G1 phase without affecting cell apoptosis.
Decreased cell proliferation by COX-2 inhibition is restored by soluble factors from naive hMSCs. We next examined whether PGE 2 produced by naive hMSCs could restore the proliferation of COX-2-suppressed hMSCs. To determine the effect of soluble factors from naïve hMSCs on COX-2-inhibited cells without cell-to-cell contact, hMSCs were treated with celecoxib for 3 days and subsequently co-cultured with naive cells for 24 hours using the transwell system. Interestingly, co-culture with intact cells rescued the cell growth rate of both celecoxib-treated hUCB-MSCs and hAD-MSCs (Fig. 2a). However, celecoxib-mediated COX-2 inhibition was restored as early as day 3 after treatment (Suppl. Fig. S3). To minimize the influence of the short duration of COX-2 inhibition, COX-2 was further inhibited by siRNA transfection. Specific siRNA for COX-2 (siCOX-2) stably inhibited the expression level of COX-2 until day 3 (Fig. 2b). Therefore, we further investigated whether the soluble factor, assumed to be PGE 2 , from naïve hMSCs restored the proliferation of COX-2-supressed hMSCs using siCOX-2 to induce relatively persistent inhibition. As expected, soluble factors from intact hUCB-MSCs significantly rescued the proliferation of COX-2-inhibited hUCB-MSCs, whereas secretory factors from siCOX-2-treated hUCB-MSCs did not (Fig. 2c). Moreover, direct PGE 2 treatment increased the proliferation of COX-2-inhibited hUCB-MSCs in a dose-dependent manner (Fig. 2c). These results suggest that PGE 2 exerts an autocrine regulatory function in the self-renewal of hMSCs. EP2 receptor is involved in autocrine PGE 2 signalling to regulate hMSC proliferation. PGE 2 has a regulatory role in the self-renewal of hMSCs, and receptor-mediated signalling is involved in this regulation, including through EP receptors 25,26 . Therefore, we examined which receptors are expressed in hMSCs. Because microglia express all EP receptor subtypes, the human microglia cell line HMO6 was used as a positive control. hUCB-MSCs and hAD-MSCs expressed four EP receptor subtypes (Fig. 3a). We next explored the crucial receptors involved in PGE 2 -mediated cell growth regulation by blocking each receptor with selective antagonists. Remarkably, hUCB-MSC proliferation decreased only when the EP2 receptor was blocked with its antagonist (AH-6809) (Fig. 3b,c). The other antagonists did not affect proliferation significantly and EP3 antagonist, L-798106, slightly increased the proliferation (Fig. 3b,c). To confirm these findings, hMSCs were treated with butaprost and sulprostone (agonists for EP2 and EP3 receptors) in the presence of celecoxib to determine the effect of specific receptor triggering. While EP2 receptor activation significantly enhanced hUCB-MSC proliferation, EP3 receptor activation did not (Fig. 3d). Taken together, these results indicate that the EP2 prostanoid receptor is the pivotal signalling pathway in PGE 2 -mediated regulation of hMSC self-renewal. PGE 2 secretion by hMSCs is regulated by cell contact. Direct cell-to-cell contact between MSCs and immune cells is crucially involved in regulating the proliferation and activation of immune cells 15,27 . Although these cell contact-dependent regulatory mechanisms in MSC function have been reported by a number of groups, few mechanistic studies elucidate the effect of cell contact among the hMSCs themselves on their function. Therefore, we investigated whether cell contact among hMSCs can affect COX-2 and mPGES-1 protein expression as well as PGE 2 secretion, which we proved to have a pivotal role in hMSC function. The COX-2 and mPGES-1 expression levels in both hUCB-MSCs and hAD-MSCs drastically decreased when the confluency of the same number of hMSCs was elevated by regulating the attachment area to allow cell-to-cell contact (Fig. 4a). These results were visually confirmed by the immunocytochemical staining of COX-2 and mPGES-1 in hUCB-MSCs plated at different confluencies. hUCB-MSCs plated at high density showed reduced expression levels of PGE 2 synthesizing enzymes compared to cells with low density, the non-contact group (Fig. 4b).
To determine whether the secretion profile of other soluble factors is affected by cell contact, levels of various cytokines were measured in the hUCB-MSC culture media. In contrast to PGE 2 , cell contact increased the

Figure 2. PGE 2 produced by hMSCs restores the COX-2-mediated inhibition of cell proliferation.
(a) COX-2 inhibited cells were co-cultured with naive cells for 24 hours, and the proliferation was determined by BrdU assay. After siRNA transfection, (b) sustained expression levels of COX-2 on day 1 and 3 were detected by Western blot analysis. (c) siCOX-2 transfected cells were treated with PGE 2 or co-cultured with naive and COX-2 suppressed cells, and proliferation was measured by direct cell counts and BrdU assay. Gel electrophoresis was conducted under the same experimental conditions, and images of blots were cropped. *P < 0.05, ***P < 0.001. Results are shown as the mean ± SEM. production of interleukin (IL)-6 and IL-8, representative cytokines in NF-κ B signalling (Fig. 4d,e). Secretion of prominent immunomodulatory factors from hMSCs, including transforming growth factor (TGF)-β 1, nitric oxide (NO) and indoleamine-2,3-dioxygenase (IDO)-1, was not affected by cell contact (Fig. 4f-h and Suppl. Fig. S4).
Furthermore, the cell contact status altered the expression pattern of EP receptor subtypes. Western blot analysis showed that hMSCs expressed higher levels of EP2 receptor under non-contact conditions, whereas EP3 receptor expression increased with cell contact ( Fig. 4i and Suppl. Fig. S4). These findings imply that cell-to-cell contact among hMSCs is important to regulate PGE 2 secretion and EP receptor expression.

Cell contact-dependent COX-2/PGE 2 axis suppression is mediated by gap junction intercellular communication (GJIC).
Gap junctions are formed when cells contact each other, and they regulate cellular function by allowing communication between adjacent cells. Therefore, we investigated whether gap junctions modulate the COX-2/PGE 2 pathway by treating cells with the gap junction decoupler, carbenoxolone (CBX). hMSCs under contact conditions were treated with 100 μ M carbenoxolone for 24 hours. Impaired expression of COX-2 and mPGES-1 in hMSCs with cell-to-cell contact was restored after CBX treatment to levels similar to those observed in the non-contact group (Fig. 5a,b). The restoration of the protein levels of these synthesis enzymes consequently led to increased PGE 2 secretion (Fig. 5c). These findings indicate that PGE 2 production is regulated by gap junction-mediated cell-to-cell interaction. PGE 2 production is critical for the immunomodulatory ability of hMSCs, and cell contactdependent inhibition of PGE 2 release leads to the decline in this ability. In our previous studies, we found that among soluble factors, PGE 2 is a key molecule for the immunomodulatory function of hMSCs 15,17 . Therefore, we explored the significance of various soluble factors from hUCB-MSCs to inhibit mitogen-induced proliferation of mononuclear cells (MNCs). Mitogen-activated proliferation of MNCs was suppressed when they were co-cultured with hUCB-MSCs, and this inhibitory effect was reduced by the inhibition of COX-2, IDO-1 or IL-10 (Fig. 6a). When culture media (CM) from target factor-inhibited hUCB-MSCs was used to culture mitogen-treated MNCs, the inhibition of MNC proliferation in the CM was restored by suppressing COX-2 and IL-10 (Fig. 6b). Moreover, COX-2 inhibition in hUCB-MSCs by celecoxib treatment showed a dose-dependent decline in the suppression of MNC proliferation in both co-culture conditions allowing cell-to-cell contact and using CM (Fig. 6c,d).
Given that COX-2 signalling is pivotal in the immunomodulatory effect of hUCB-MSCs and that signalling can be altered by cell-to-cell contact, we further assessed whether cell contact status can modulate the immunosuppressive property of hUCB-MSCs. CM from the non-contact group inhibited the proliferation of MNCs to a greater extent than CM from the contact group (Fig. 6e). Moreover, the level of IL-10, a prominent anti-inflammatory cytokine, was elevated in the co-culture media, and hUCB-MSCs without cell contact exhibited more potent IL-10 production than cells with cell contact (Fig. 6f). Taken together, our findings suggest that

Discussion
Although recent studies have demonstrated that a number of autocrine signalling events are involved in the induction or maintenance of MSC functions such as proliferation, differentiation, migration and immunoregulation [28][29][30][31][32] , most of these studies focused on the elucidation of differentiation-related mechanisms. In the present study, we investigated the autocrine effect of PGE 2 on MSCs proliferation, a major characteristic of stem cells that contributes to stemness. A few studies reported that hMSC proliferation is regulated by PGE 2 25,26 . In these studies, hMSCs derived from bone marrow or umbilical cord blood were treated with various doses of PGE 2 . MSC proliferation consistently increased in response to PGE 2 treatment via protein kinase A signalling in both studies. In the present study, we found that PGE 2 produced by hUCB-MSCs and hAD-MSCs plays a crucial role in the maintenance of their proliferative function by regulating COX-2 signalling using chemical inhibitors or siRNA for COX-2. COX-2 inhibition by indomethacin or celecoxib treatment resulted in a consistent decrease in hMSC proliferation. This phenotype might result from the inhibition of potent signalling pathways in hMSCs via COX-2 signalling rather than PGE 2 , as COX-2/PGE 2 signalling is involved in the several growth factor signalling pathways, including vascular endothelial growth factor and basic fibroblast growth factor 33,34 . We achieved similar effects by suppressing mPGES-1, an enzyme for PGE 2 synthesis downstream of COX-2 signalling. Moreover, PGE 2 treatment of COX-2-inhibited hMSCs rescued their proliferation in a dose-dependent manner. More importantly, we proved that secreted soluble factors from naïve hMSCs restored the proliferation of COX-2-inhibited hMSCs when they were co-cultured using a transwell system that prevented cell-to-cell contact, indicating that soluble factors from hMSCs themselves can contribute to their proliferation, presumably including PGE 2 . The previous study by Jang et al. showed that among the four major sub-types of E-type prostaglandin (EP) receptors, EP2 receptor has a pivotal role in PGE 2 -stimulated hUCB-MSC proliferation 25 . In this study, we demonstrated that only the selective antagonist for EP2 receptor down-regulated basal hMSC proliferation, implying that autocrine stimulation of PGE 2 on hMSC proliferation is mediated by EP2 receptor. In addition, treatment with a selective agonist for EP2 receptor, butaprost, restored the proliferation of celecoxib-treated hMSCs.
Although a number of previous studies have shown that cell-to-cell contact between MSCs and immune cells or cancer cells is an important factor in the immunomodulatory or anti-tumour effect of MSCs, a few groups have focused on the cell-to-cell contacts between MSCs themselves and subsequent alterations in the secretion profile. Schajnovitz et al. reported that MSCs derived from bone marrow (BM-MSCs) possess functional gap junctions, and CXCL12 secretion by BM-MSCs is regulated by cell contact, leading to functional changes in MSCs to maintain the homeostasis of haematopoietic stem cells 35 . We show here that PGE 2 secretion from the same number of MSCs was down-regulated when cell-to-cell contact was allowed, whereas production of IL-6 and IL-8 was increased, and TGF-β 1 or NO production was not affected by confluent culture conditions that allow cell-to-cell contact. Decreased production of PGE 2 exerted by cell contact was restored by the blockage of cellular communication using CBX, a well-known GJIC inhibitor, indicating that MSCs form functional syncytia via connexin gap junctions, leading to alterations in the secretion profile. In addition, previous studies from the Prockop group reported that compaction of hMSCs into spheroids self-activates signalling to enhance secretion of anti-inflammatory modulators such as PGE 2 , tumour necrosis factor α -induced protein 6 and stanniocalcin 1 36,37 . In these studies, hMSCs cultured as spheres using hanging drop culture produced markedly elevated levels of PGE 2 . This discrepancy in PGE 2 regulation by cell contact might result from the differences in culture conditions for MSCs. In the present study, the general plastic adherent 2D culture method was used, and only the plating area was controlled to regulate cell-to-cell contact, whereas the hanging drop method to generate 3D spheroids was used in the studies by Prockop group. PGE 2 is a potent immunomodulator produced by hMSCs. hMSCs suppress the differentiation and maturation of T lymphocytes and induce the generation of regulatory T cells via COX2-mediated PGE 2 secretion 27,38,39 . Moreover, PGE 2 secretion from MSCs in response to certain inflammatory milieu critically contributes to the immunoregulatory function of MSCs against several immune disorders, including arthritis and colitis 15,40 . PGE 2 produced by MSCs exerts anti-inflammatory effects through the regulation of immune cell activation and maturation, including CD4 + helper T cells, B cells, dendritic cells, natural killer cells, monocytes and macrophages 41 . In the present study, we have proven that cell contact-dependent regulation of PGE 2 secretion correlates with the functional phenotype of MSCs. Importantly, down-regulation of PGE 2 secretion by cell-to-cell contact led to a decreased immunomodulatory effect of MSCs in a mixed leukocyte reaction. Based on these findings, control of culture conditions regulating cell contact might be a crucial point in the development of therapeutics from MSCs, as a number of studies have been performed and are being conducted to produce therapeutics or cosmetics using conditioned media from MSCs 42-44 . Taken together, the present study revealed novel information indicating that PGE 2 secreted from adult stem cells exerts autocrine effects on MSC proliferation by triggering the EP2 receptor, and PGE 2 production is dependent on cell-to-cell contact mediated by GJIC, resulting in a decline in immunoregulatory ability.

Methods
Isolation and culture of hUCB-MSCs. All experiments involving human umbilical cord blood (UCB) or UCB-derived cells were carried out in accordance with the approved guidelines of the Boramae Hospital Institutional Review Board (IRB) and the Seoul National University IRB (IRB No. 0603/001-002-10C4). The UCB samples were provided immediately after birth with informed consent and parent approval. The UCB from a donor was mixed with HetaSep solution (Stem Cell Technologies, Vancouver, Canada) at a ratio of 5:1 and incubated at room temperature for approximately one hour to remove red blood cells. Then, supernatant was collected using Ficoll, and mononuclear cells were separated after centrifugation at 2,500 rpm for 20 min. The cells were washed twice in phosphate-buffered saline (PBS). Isolated cells were seeded in growth media consisting of D-media (Formula No. 78-5470EF, Gibco BRL, NY, USA) containing EGM-2 SingleQuot and 10% foetal bovine serum (Gibco BRL, NY, USA). After 3 days, unattached cells were washed out, and adherent cell colonies were cultured to consistently establish sharp and spindle-shaped hUCB-MSCs. For the expansion of cells, KSB-3 complete media (Kangstem Biotech, Seoul, Korea) was used.
Isolation and culture of hMNCs. The UCB samples were mixed with HetaSep solution (Stem Cell Technologies, Vancouver, Canada) at a ratio of 5:1 and incubated at room temperature for approximately one hour to remove red blood cells. Then, supernatant was collected with Ficoll, and mononuclear cells were separated after centrifugation at 2,500 rpm for 20 min. The cells were washed twice in PBS. Isolated cells were seeded in growth media consisting of RPMI 1640 (Gibco BRL, Grand Island, NY, USA) containing 10% foetal bovine serum.
Isolation and culture of hAD-MSCs. All procedures using human adipose tissue or adipose tissue-derived mesenchymal stem cells were conducted in accordance with guidelines approved by Seoul National University IRB (IRB No. 0611/001-001). Freshly excised human mammary fat tissue, the waste from reduction mammoplasty, was digested for 2 hours with 1 mg/mL of type Ι A collagenase (≥125 CDU/ BrdU assay. For this assay, the cell proliferation ELISA kit (Roche, Indianapolis, IN, USA) was used. After the indicated treatment, cells were washed twice in PBS and incubated in growth media containing 100 μ M bromodeoxyuridine (BrdU) labelling reagent for 2 hours at 37 °C in a humidified atmosphere with 5% CO 2 . After removing media and drying the cell surface, cells were fixed with the provided FixDenat solution for 30 minutes and incubated in peroxidase-conjugated anti-BrdU antibody (anti-BrdU-POD) solution for 90 minutes at room temperature. Cells were then washed three times in diluted washing solution and incubated with the provided substrate (tetramethyl-benzidine; TMB) solution for 5 to 30 minutes. After sufficient reaction and stop solution addition, the reaction products, which demonstrated cell proliferation levels, were quantified by measuring absorbance at the wavelength 450 nm and 690 nm (as a reference) using spectrophotometer.
Cumulative population doubling level (CPDL) analysis. Cell proliferation was also measured by CPDL analysis. Estimated growth rates and proliferation levels were determined through the formula CPDL = ln (Nf/Ni) ln2, where Ni is the initial number of cells seeded, Nf is the final number of harvested cells, and ln is the natural log. First, 3 × 10 5 cells isolated from different donors were seeded with or without indicated treatments, and the number of cells was counted after 3 to 5 days. Then, 3 × 10 5 cells were seeded again with the same treatment. To determine the CPDL, population doublings for each passage were calculated and added.
MTT assay. To indirectly assess cell viability and proliferation, an MTT assay was conducted. After treatment, cells were incubated in fresh medium containing 200 μ g/ml of MTT reagent (Amresco, Solon, OH, USA) for Scientific RepoRts | 6:26298 | DOI: 10.1038/srep26298 were analysed by flow cytometry, which was performed on a FACScalibur using the Cell Quest software (BD Bioscience, San Jose, CA, USA).
Apoptosis assay. For the apoptosis assay, commercially available Apoptosis Detection Kits (BD Bioscience, San Jose, CA, USA) were used. After indicated treatments, the cells were washed twice in PBS and resuspended in 100 μ l of 1X binding buffer at a concentration of 1 × 10 5 . Then, 5 μ l of FITC annexin V and 5 μ l propidium iodide (PI) were added. The mixtures were gently vortexed and incubated for 15 min at room temperature in the dark. Then, 400 μ l of 1X binding buffer was added to the mixtures, and all samples were analysed by flow cytometry, which was performed on a FACScalibur using Cell Quest software (BD Bioscience, San Jose, CA, USA).
Immunocytochemistry. Cells at different confluencies were washed in PBS and fixed with 4% paraformaldehyde (PFA) at room temperature for 10 min. For permeabilization, the cells were incubated with 0.05% Triton X-100 solution at room temperature for 10 min and blocked with 5% normal goat serum (NGS) at room temperature for 1 hour. Then, the cells were stained with specific primary antibodies against COX-2 and mPGES-1 (Abcam, Cambridge, MA, USA) followed by 2 hours of incubation with Alexa 488-labelled secondary antibody (1:1,000; Molecular Probes, Eugene, OR, USA). The nuclei were stained with DAPI. The images were captured by a confocal microscope.
Cytokine production. To determine the secretion level of various cytokines, culture supernatants were collected from cells incubated for 24 hours in non-contact or contact conditions. To determinate each concentration, commercial ELISA kits for PGE 2 , TGF-β 1, IL-6, IL-8 (R&D Systems, Minneapolis, MN, USA) and NO (Cayman Chemical, Ann Arbor, MI, USA) were used according to the manufacturer's protocols.

Mixed leukocyte reaction.
To collect culture supernatants (hUCB-MSC conditioned media; UCM), cells were incubated in non-contact or contact conditions for 24 hours and treated for 3 days with celecoxib in RPMI 1640 (Gibco BRL, Grand Island, NY, USA). Then, media were harvested after centrifugation. hMNCs prepared as described above were treated with ConA in collected culture supernatants for 5 days, and hMNC proliferation was determined by cell proliferation ELISA, BrdU kit (Roche, Indianapolis, IN, USA).

Statistical analysis.
Mean values of all results were expressed as the mean ± SEM. Statistical analyses were conducted using Student's 2-tailed t-test or one-way ANOVA followed by Bonferroni post-hoc test for multigroup comparisons using GraphPad Prism version 5.0 (GraphPad Software, San Diego, CA, USA). Statistical significance is indicated in the figure legends.