Enhanced immunoregulation of mesenchymal stem cells by IL-10-producing type 1 regulatory T cells in collagen-induced arthritis

Mesenchymal stem cells (MSCs) possess immunomodulatory properties and have potential, however, there have been conflicting reports regarding their effects in rheumatoid arthritis (RA), which causes inflammation and destruction of the joints. Through a comparative analysis of regulatory T (Treg) and IL-10-producing type 1 regulatory T (Tr1) cells, we hypothesized that Tr1 cells enhance the immunoregulatory functions of MSCs, and that a combinatorial approach to cell therapy may exert synergistic immunomodulatory effects in an experimental animal model of rheumatoid arthritis (RA). A combination of MSCs and Tr1 cells prevented the development of destructive arthritis compared to single cell therapy. These therapeutic effects were associated with an increase in type II collagen (CII)-specific CD4+CD25+Foxp3+ Treg cells and inhibition of CII-specific CD4+IL-17+ T cells. We observed that Tr1 cells produce high levels of IL-10-dependent interferon (IFN)-β, which induces toll-like receptor (TLR) 3 expression in MSCs. Moreover, induction of indoleamine 2,3-dioxygenase (IDO) by TLR3 involved an autocrine IFN-β that was dependent on STAT1 signaling. Furthermore, we observed that production of IFN-β and IL-10 in Tr1 cells synergistically induces IDO in MSCs through the STAT1 pathway. These findings suggest co-administration of MSCs and Tr1 cells to be a novel therapeutic modality for clinical autoimmune diseases.


Results
Phenotypes of culture-expanded MSCs. MSCs were obtained from DBA1J mice bone marrow showing typical fibroblast-like morphology, as described previously 18 . Flow cytometric analysis demonstrated that MSCs did not express c-kit, CD11b, CD34, CD106, CD45, CD31, CD80, or CD86, but expressed MSC-associated antigens, such as Sca-1, CD44 and CD29 ( Supplementary Fig. 1A). Based on these results, we conclude that MSCs have the capacity to differentiate into osteocytes, chondrocytes and adipocytes in vitro (Supplemental Fig. 1B).
Comparative analysis of culture-expanded Treg and Tr1 cells characteristics. We and others previously reported that Treg cells can be generated in the presence of anti-IFN-γ, anti-IL-4, TGF-β and retinoic acid 24 . Approximately 70-80% of cells were CD4+Foxp3+ T cells, and low production of CD4+IL-10+ T cells from control and CIA mice on a DBA1J background was reported. On the other hand, the development of Tr1 cells was induced by Dex and Vit D3, and secretion of high IL-10 levels and minimal amounts of CD4+Foxp3+ T cells was observed (Fig. 1A). Tr1 cells showed decreased secretion of TGF-β and IL-17 and significantly increased IL-10 production compared to Treg cells. However, no significant difference in IL-6 levels was observed (Fig. 1B). To further explore this observation, we investigated the continuous production of IL-10 and IL-17 in re-cultured Treg and Tr1 cells supernatants by removing the stimulant at various time points. As a result, Tr1 cells maintained the production of IL-10 following removal of stimulated culture supernatants, whereas IL-17 production was barely detectable compared to Treg cells (Fig. 1C).
To perform a comparative analysis on the effects of Treg and Tr1 cells on disease activity and progression of CIA, DBA1J mice were immunized with CII and treated with Treg or Tr1 cells. The therapeutic effects of Tr1-treated CIA mice induced a reduction in the incidence (Fig. 1D, upper) and severity of clinical arthritis (Fig. 1D, lower) compared to Treg-treated CIA mice. Furthermore, Tr1-treated CIA mice produced increased levels of IFN-γ (control vs. Tr1, Treg vs. Tr1), but Foxp3 and IL-17 levels were similar in CII-stimulated T cells in spleen, as compared to Treg-treated CIA mice (Fig. 1E). Also, our results indicate that Tr1 cell-treated CIA mice showed increased serum IL-10 levels, compared to the control, by ELISA (Fig. 1F). These results suggest that the development of arthritis was regulated by IL-10 production in Tr1 cell-treated CIA mice.
As reported previously, IFN-γ-licensed MSCs exhibit upregulated IDO activity 25 . These results demonstrate a role for IFN-γ in MSC-mediated immune suppression in vivo 26 . Moreover, IFN-γ-induced IDO suppresses inflammation by suppressing Th17 cells, which are induced during the pathogenesis of CIA 27 . We hypothesized that IFN-γ production following Tr1 cell therapy is a potent inducer of IDO in vivo; the expression of IDO could be increased by stimulating MSCs with IFN-γ, which is associated with the regulation of Th17 and Treg cells 28,29 . Effects of MSCs and Tr1 cells co-culture on CII-specific CD4 T cell proliferative responses. As expected, very high 3 H-thymidine incorporation was observed in the CII-specific T cell (1 × 10 5 ) response, but co-culture of MSCs plus Treg or MSCs plus Tr1 cells resulted in a marked reduction in splenic T cell proliferation compared with single cells. Although there was no difference in CII-specific T cell proliferation between MSCs plus Treg (1 × 10 4 :1 × 10 4 = 1:1) and MSCs plus Tr1 (1:1) cells, the suppressive effect was more prominent in splenic T cells from MSCs plus Tr1 cells ratio of 1:2 (1 × 10 4 :2 × 10 4 ) compared to MSCs plus Tr1 cells ratio of 1:1 (1 × 10 4 :1 × 10 4 ) following CII-specific stimulation ( Fig. 2A). These data suggest that the combination of MSCs plus Tr1 cells more effectively regulates CII-specific stimulation.
Scientific RepoRts | 6:26851 | DOI: 10.1038/srep26851 Therapeutic effect of MSCs and Tr1 cells on CIA development and severity. To explore the effects of co-infusion of MSCs and Tr1 cells on CIA disease activity and progression, DBA1J mice were immunized with CII and treated with MSCs in the absence or presence of Tr1 cells. Intraperitoneal (i.p.) injection of MSCs and intravenous (i.v.) injection of Tr1 cells reduced the clinical arthritis score (Tr1 vs. MSC+Tr1 (from 5.5 to 9.5 weeks), MSC vs. MSC+Tr1 (from 3.5 to 9.5 weeks)) and incidence of arthritis (Tr1 vs. MSC+Tr1 [from 7 to 9.5 weeks], MSC vs. MSC+Tr1 [from 4 to 9.5 weeks]) compared with single cell therapy (Fig. 2B). There was no significant difference between MSCs alone and Tr1 cells alone. Although a single infusion of Tr1 cells prevented the development of destructive arthritis in the early post-therapeutic period, these effects were generally mild and not prolonged. MSC therapy had no inhibitory effect at the early stage; these cells were in a worse condition (Fig. 2B). Co-infusion of MSCs plus Tr1 cells prevented the development of swelling or redness of the front or hind paws at five weeks following the first immunization (Fig. 2C). Regarding the histopathological features of the ankle  Treg and Tr1 cells (differentiated with anti-IFN-γ, anti-IL-4, TGF-β, retinoic acid, anti-CD3, anti-CD28 and Dex, Vit D3, anti-CD3, anti-CD28, respectively) culture media were measured by ELISA. (C) Supernatant levels (n = 5) of IL-10 and IL-17 in the absence of stimulant at 0, 24, 48, and 72 h following differentiation for two days were measured by ELISA. (D) Mice were injected intravenously with 5 × 10 5 Treg or Tr1 cells three weeks following primary immunization. Control mice received intravenous injections of an equal volume of PBS at the same time points. Clinical incidence (upper) and arthritis score (lower) in treated mice. Data represent the means ± standard error of the mean (SEM) of the arthritis scores of six mice per group. Incidence of arthritis with a score >4. Data are representative of two independent experiments (n = 6 per group) with similar results. (E) Four weeks after primary immunization, intracellular flow cytometry was performed to analyze IFN-γ, IL-17 and Foxp3 expression in the gated CD4+ T cell populations from CIA mice. Values represent the percentages of positive cells. (F) On day 7 following injection of PBS, Treg, or Tr1 cells in CIA mice, serum IL-10 were measured by ELISA. All graphs display the means ± SEM; results are representative of two independent experiments and statistical significance was determined by Student's two-tailed t-test and ANOVA with Bonferroni correction for multiple comparisons (*p < 0.05; **p < 0.01).
Scientific RepoRts | 6:26851 | DOI: 10.1038/srep26851 joints, MSCs plus Tr1 cell-treated CIA mice showed a lower degree of inflammation, mononuclear cell infiltrate and pannus formation with superficial cartilage damage compared with controls and single cell therapy (Fig. 2D).
These data are in agreement with our hypothesis that the combination of MSCs plus Tr1 cells induces both CD4+IFN-γ+ and CD4+Foxp3+ T cells differentiation, and upregulation of CD4+TNF-α+ T cells are abolished following CII stimulation as compared to single cell therapy (Fig. 2E).
A dual role of TGF-β-induced anti-inflammatory Treg cells and pro-inflammatory Th17 cells from naïve T cells 30 is indicated by the coordinate signaling by IL-6 31 . Serum levels of TGF-β and IL-6 were measured using ELISA to determine whether treatment with MSCs and Tr1 cells was associated with alterations in the cytokine response to T cells. Mice treated with MSCs plus Tr1 cells showed slightly reduced IL-6 levels (not statistically significant) with a significant increase in TGF-β levels (Control vs. MSC+Tr1) in serum on day 7 (four weeks) following injection of MSCs with or without Tr1 cells in the CIA model (Fig. 2F).

Reciprocal regulation of CII-specific Treg and Th17 cells by MSCs and Tr1 cells in CIA mice. Next,
we examined whether co-infusion of MSCs and Tr1 cells contributes to Treg cells induction and inhibits the development of Th17 cells. Changes in Treg and Th17 cells related to transcription factors were investigated in splenic cells from CIA mice. RORγ-T plays an important regulatory role in thymopoiesis, by promoting thymocyte differentiation into pro-inflammatory Th17 cells 32,33 . Splenic cells from CIA mice exposed to MSCs plus Tr1 (E) Splenic cells were isolated from each group and analyzed using flow cytometry to examine IFNγ, Foxp3 and TNF-α expression in the gated CD4+ T cell populations. (F) On day 7 following injection of PBS, MSCs, Tr1 cells, or a co-infusion of MSCs and Tr1 cells in CIA mice, levels of TGF-β (n = 4) and IL-6 (n = 4) in serum were measured using ELISA. All graphs display the means ± SEM; results are representative of two independent experiments and statistical significance was determined by Student's two-tailed t-test and ANOVA with Bonferroni correction for multiple comparisons (*or # p < 0.05; **or ## p < 0.01; $$$ p < 0.001). cells showed markedly increased expression of Foxp3 (Control vs. MSC+Tr1), but slightly decreased expression of RORγ-T ( Fig. 3A; MSC vs MSC+Tr1).
To further increase our understanding of the signaling events involved in the reciprocal effects of co-infusion with MSCs and Tr1 cells in CIA mice, we analyzed the inhibition of STAT3, which limits Th17 differentiation, as well as the expression of STAT5, which induces Treg cells differentiation 34 . Two weeks following the second immunization, MSCs and Tr1 cell-treated CIA mice showed markedly increased expression of p-STAT5 compared with those treated with either MSCs or Tr1 cells alone, whereas p-STAT3 activity decreased compared with both the control and single cell therapy groups (Fig. 3B).
These data are consistent with our hypothesis that CD4+ T cells activation in the presence of CII stimulation induces CD4+CD25+Foxp3+ Treg cells differentiation, and that the upregulation of CD4+IL-17+Th17 cells is abolished after CII stimulation in the draining lymph node by co-infusion of MSCs and Tr1 cells to CIA mice compared with controls ( Fig. 3C,D). Inhibition of IL-17 by co-infusion is associated with increased Foxp3 expression, suggesting that co-infusion reciprocally affects both Treg cells induction and Th17 cells differentiation. Thus, co-infusion of MSCs and Tr1 cells markedly alters the balance between Treg cells and Th17 cells in CIA mice, resulting in an altered disease course.

Tr1 cells enhance TLR3 expression in MSCs. Previous studies have indicated a role for TLR activation in
MSCs during the development of new therapeutic strategies for inflammatory or autoimmune diseases, because MSCs are likely activated through TLR ligands in sites of injury or inflammation 13 . Therefore, we hypothesized that TLR stimulation by Tr1 cells regulates MSC function. Changes in TLR expression in MSCs stimulated by T, Treg and Tr1 cells were investigated to explore the mechanisms involved in the Tr1-treated MSC-mediated immunosuppressive effects. To investigate the expression of TLRs in MSCs, which were sorted using flow cytometry from co-cultures of MSCs with T, Treg or Tr1 cells for 24 h, the expression of TLR2, TLR3, TLR4, and TLR9 were assessed using real-time PCR and flow cytometry. As a result, co-cultured Tr1 cells showed MSC-upregulated expression of TLR3 (T-treated MSC vs. Tr1-treated MSC, Treg-treated MSC vs. Tr1-treated MSC) (Fig. 4, Supplementary Fig. 2).

MSCs upregulate TLR3 by IL-10-dependent expression of IFN-β in Tr1 cells. Previous studies
reported that TLR3 mRNA levels are significantly increased following stimulation with IFN-γ and IFN-β, and to a lesser extent, IL-1β 35 . Therefore, we investigated the production of IL-10, as well as IFN-γ and IFN-β, in the presence of Dex and Vit D3 in Tr1 cells. Tr1 cells did not cause IFN-γ secretion (Fig. 5A) 36 . Interestingly, IFN-β expression was significantly higher in Tr1 cells than in Treg cells, according to real-time PCR and ELISA (Fig. 5B, Supplementary Fig. 3A). Utilizing anti-IFN-β-and anti-IL-10-targeting Tr1 cells, IFN-β expression significantly decreased in anti-IFN-β-treated Tr1 cells, as well as anti-IL-10-treated Tr1 cells compared to Tr1 cells. Therefore, IL-10-dependent IFN-β expression in Tr1 cells was higher than in Treg cells (Fig. 5B). On the other hand, anti-IFN-β-treated Tr1 cells showed no differences in IL-10 production compared to Tr1 cells (Fig. 5C). Therefore, IL-10 secretion is likely produced through a pathway independent of IFN-β in Tr1 cells.
Signaling required for IDO induction is associated with IFN induction and STAT1-dependent signaling 38 . Next, we investigated whether Tr1 cells induce STAT1 expression in MSCs. We observed an increase in STAT1 mRNA and protein levels at 24 h after co-culturing Tr1 cells with MSCs compared to co-culturing anti-IFN-β-treated Tr1 cells and/or anti-IL-10-treated Tr1 cells with MSCs (Fig. 5E, Supplementary Fig. 3C). Although STAT3 is a key mediator of the IL-10 response 39 , Tr1 cells did not induce STAT3 expression in MSCs (data not shown).
In a similar manner, Tr1-treated MSCs demonstrated significantly increased expression of IDO compared to Treg-treated MSCs (Fig. 5F, Supplementary Fig. 3D). These results suggest that IFN-β production by Tr1 cells resulted in significant induction of IDO expression, which is dependent on IL-10 signaling.
Tryptophan-degrading enzyme indoleamine-2,3-dioxygenase (IDO) has been implicated in potent immunosuppressive mechanisms. In our experiment, to further investigate whether IDO-mediated immunosuppressive effects were involved, we used the IDO inhibitor 1-methyl-D-tryptophan (1-MT), which inhibits mRNA expression of IDO through p38 mitogen-activated protein kinase (MAPK) and c-Jun N-terminal kinase (JNK) signaling 40 . As a result, Tr1-treated MSCs showed significantly decreased expression of IDO by 1-MT, suggesting that IDO is involved in the interaction between MSCs and Tr1 cells. To increase our understanding of the mechanisms associated with the immunuosuppressive effect of co-culture Tr1 cells with MSCs, we examined changes in IFN-β, STAT1 and IDO levels in Tr1-treated MSCs using IFN-β and IL-10-neutralizing antibodies. IFN-β and IL-10 share STAT1 as a signaling component. Therefore, we investigated whether these cytokines affect STAT1 activation and IFN-β production in MSCs. As a result, autocrine IFN-β production in Tr1-treated MSCs was inhibited compared to Tr1-treated MSCs, whereas IL-10 production did not affect autocrine IFN-β production in Tr1 cells (Fig. 6A).
In addition, STAT1 expression was decreased by use of anti-IFN-β and/or anti-IL-10 neutralizing antibodies in co-cultured Tr1-treated MSCs. When treated with both anti-IFN-β and anti-IL-10, a significant decrease in STAT1 expression was observed relative to incubation with anti-IFN-β or anti-IL-10 alone in co-cultured Tr1-treated MSCs (Fig. 6B, Supplementary Fig. 4A).
In a similar manner, anti-IFN-β and anti-IL-10 neutralizing antibodies significantly decreased IDO expression compared to treatment with a single neutralizing antibody in co-cultured Tr1-treated MSCs (Fig. 6C,  Supplementary Fig. 4B). Therefore, these results suggest that IFN-β and IL-10 production by Tr1 affects the expression of IDO by activating STAT1 through different pathways in Tr1-treated MSCs.

Discussion
The immunomodulatory properties of MSCs offer a potentially attractive therapeutic modality for autoimmune diseases 41 . Although MSCs are known to possess immunoregulatory functions, in a chronic inflammatory To compensate for the disadvantages of MSC therapy, we previously reported a combination of MSCs and Treg cell therapy as a novel method for synergistic immunomodulatory effects on allograft rejection 44 , acute GVHD 45 and induction of mixed chimerism 46 . Therefore, we speculated that a conditional microenvironment containing regulatory immune cells, such as Treg or Tr1 cells, might increase the immunoregulatory strength of MSCs in RA.
We first analyzed the effects of Treg and Tr1 cells as MSC partners in an experimental model of arthritis. As seen in Fig. 1A, the generation of Tr1 cells from naïve CD4+ T cells using both Dex and Vit D3 was first described by Barrat et al. 36 . These Tr1 cells stably produce IL-10, but show only marginal or no IL-2, IL-4 (data not shown) and IFN-γ production in stimulated and removed stimulated culture supernatants. Foxp3 is critical for both the development and function of Treg cells in mice (Fig. 1A). However, unlike CD4+ T cells from C57BL/6 mice, DBA1J-derived CD4+ T cells are resistant to TGF-β-and retinoic acid-induced Treg cells conversion 24 . It is possible that a small number of contaminating CD4+Foxp-T cells produce IL-17 in Treg cells after removing the stimulant (Fig. 1C). On the other hand, despite blocking IL-10 and IFN-β signaling, Tr1 cells did not produce IL-17 (data not shown), but stably produced IL-10 in vitro (Fig. 4C). Therefore, our study suggests that Treg cells may facilitate the conversion of CD4+Foxp3-T cells towards a CD4+IL-17+Th17 phenotype, whereas Tr1 cells do not induce the conversion of CD4+IL-10-Tr1 cells to CD4+IL-17+ T cells. Our results provide evidence that Tr1 cells are more stable and functional than Treg cells in mice with established autoimmunity.
The balance between Th17 and Th1 responses is important because any enhanced IL-17 activity could be secondary to robust Th-cell responses typical of RA. Moreover, Th1 cytokine IFN-γ is known to suppress Th17 development 47 . In addition, exposure to IFN-γ provides a 'licensing' step that is fundamental for the induction of MSC-mediated immunosuppression. Although Tr1 cells can generate IL-10 (but not IFN-γ) in the presence of Dex and Vit D3 when adoptively transferred in a CIA model, these cells prevent RA by slightly decreasing the expression of IL-17 (no statistical significance) via production of CD4+IFN-γ+ T cells compared to Treg-treated CIA mice (Fig. 1E). These data indicate that, because Tr1-treated CIA mice exhibit increased IFN-γ production in vivo, co-infusion of MSCs and Tr1 cells may induce an increase in IDO production in IFN-γ-licensed MSCs, and may therefore be applicable as a treatment for RA in a synergistic manner.
Previous reports suggest that MSCs inhibit Th17 cells, but they do not inhibit Treg cells; therefore, this effect must occur, at least in part, through IDO expression 29 . This study showed that the therapeutic effects of co-infusion of MSCs and Tr1 cells imply reciprocal regulation of CII-specific endogenous Treg and Th17 populations in draining lymph nodes and spleen cells (Fig. 3). Moreover, the secretion of TNF-α and IL-6 decreased while that of IFN-γ and TGF-β increased, in serum (Fig. 2F) and spleen cells (Fig. 2D), suggesting that co-administration of MSC-and Tr1 cell-mediated suppressive activity in vivo is partly attributable to soluble mediators. Furthermore, we believe that Treg-treated MSCs increase the production of TNF-α compared to Tr1-treated MSCs in vitro (data not shown), indicating that co-infusion of MSCs and Treg cells may reduce immune suppression during the development of RA compared to co-infusion of MSCs and Tr1 cells. For this reason, although there was no difference in CII-specific T cell proliferation between MSCs plus Treg (1: 1) or Tr1 (1: 1) cells, the suppressive effect was reduced from MSCs plus Treg (1: 2) compared to MSCs plus Tr1 (1: 2) cells and MSCs plus Treg (1: 1) cells following CII-specific stimulation ( Fig. 2A). Although we did not assess the effect of co-infusion of MSCs plus Treg cells in an experimental mouse model of arthritis, co-administration of MSCs plus Treg cells is also expected to ameliorate arthritis, according to the results of CII-specific T-cell responses in vitro. However, we suggest that MSCs plus Tr1 cell therapy would be more effective in RA than MSCs plus Treg cell therapy.
Our results showed that the combination of MSCs and Tr1 cells induced a significant reduction in the incidence and severity of clinical arthritis, while MSCs suppressed the CII-specific T cell response in vitro but failed to prevent severe joint swelling and joint inflammation due to mononuclear cell infiltration in vivo (Fig. 2B-D). Although MSCs possess an immunosuppressive ability, this has recently been found to be not constitutive, and depends strongly on the prevailing inflammatory conditions. Thus, MSCs must be administered under conditions that maintain their immunosuppressive function.
Of note, our results suggested reciprocal regulation of CII-specific Treg and Th17 cells by MSCs and Tr1 cells in CIA mice (Fig. 3C,D). IL-6 contributes to the development of Th17 cells, but suppresses Treg cells differentiation in the presence of TGF-β. We hypothesize that co-infusion of MSCs plus Tr1 cells to CIA mice resulted in significantly increased production of TGF-β, but that of IL-6 decreased in vivo. However, serum TGF-β and IL-6 levels did not differ significantly between the MSC and MSC plus Tr1 groups (Fig. 2F). This is likely because MSCs exert an immunosuppressive effect by secreting TGF-β and IL-6 48 .
Previous studies have indicated the important role of TLRs in MSC immunomodulating properties, which may modulate MSC therapeutic potential in vivo 13 . Opitz et al. showed that MSCs pre-stimulated with TLR3 and TLR4 agonists showed enhanced IDO-mediated T cell suppression. The investigators defined a novel IFN-γ-independent pathway involving an autocrine IFN-β signaling loop that requires protein kinase R and STAT1/IRF1 activation 37 . In addition, according to Waterman et al., TLR3 stimulation leads to the secretion of factors with mostly immune-suppressive properties, while stimulation of TLR4 with lipopolysaccharide (LPS) results in the secretion of more pro-inflammatory factors. Furthermore, when co-cultured with T cells, TLR4-primed MSCs augment T cell activation, whereas TLR3-primed MSCs suppress T cell activation 14 . We and other groups found that TLR3 is enhanced upon stimulation with IFN-β and/or IFN-γ 35 . Tr1 cells induced with Dex and Vit D3 express IFN-β (Fig. 5B), allowing for the induction of both endogenous (Fig. 5D) and exogenous IFN-β (Fig. 5B) in Tr1-treated MSCs. These results suggest that Tr1-treated MSCs induce TLR3 stimulation (Fig. 4), leading to alternate MSC2 polarization associated with anti-inflammatory resolution responses. According to this paradigm, MSC2 produces increased levels of IL-10, IDO and PGE 2 14 . A comparative analysis of interactions between MSCs and Treg or Tr1 cells with a focus on PGE 2 , TGF-β and IL-10 was performed. However, there was no difference between Treg-and Tr1-treated MSCs in MSC-mediated immunoregulation (data not shown).
Tr1 cells primarily generate type I IFN-dependent IL-10 49-51 . However, as shown in Fig. 5C, IL-10 secretion is produced through a pathway independent of IFN-β in Tr1 cells. Tr1 cells may induce increased IDO expression by activating STAT1 signaling in a different pathway in Tr1-treated MSCs, according to studies using anti-IFN-β and anti-IL-10 antibodies. When treated with anti-IL-10, a greater decrease in STAT1 expression was observed relative to co-cultured Tr1-treated MSCs treated with anti-IFN-β alone (Fig. 6B). These results suggest that, since Tr1 cells produce higher levels of IL-10 compared to IFN-β, IL-10 inhibition strongly decreases STAT1 mRNA levels compared to inhibition of IFN-β (Fig. 6B). In addition, treatment of MSCs with anti-IFN-β-treated Tr1 cells significantly increased the clinical arthritis score and incidence of arthritis compared with MSC and Tr1 cell therapy (data not shown), suggesting that IFN-β plays an important role in IDO induction in MSCs.
In summary, we showed that co-infusion of MSCs and Tr1 cells exerts a synergistic immunoregulatory effect through increased IDO expression in MSCs via IFN-β and IL-10 production by Tr1 cells for the alleviation of RA development (Fig. 7). Thus, co-infusion of MSCs and Tr1 cells, which include highly active components of inflammation modulation and tolerance induction, may represent a new method for the treatment of RA in clinical trials.
Induction and treatment of CIA. CIA was induced as described previously 53 . Briefly, DBA1J mice were injected in the tail base with 100 μg bovine type II collagen (CII, Chondrex, Redmond, WA, USA) emulsified in complete Freund's adjuvant (CFA). Two weeks later, 100 μg CII with incomplete Freund's adjuvant (IFA) was injected into the tail as a booster. Mice were injected intraperitoneally (i.p.) with 5 × 10 5 MSCs and/or intravenously (i.v.) with 5 × 10 5 Tr1 cells or intravenously (i.v.) with 5 × 10 5 Treg cells 3 and 3.5 weeks (21 and 24 days) after the primary immunization. Control mice received i.p. or i.v. injections of an equal volume of phosphate-buffered saline (PBS, Gibco) at the same time points. The onset and severity of arthritis were determined by three independent observers, based on a scoring system described previously 54 . Twice weekly, the mice were examined visually for joint arthritis, and two independent observers who were blind to the experimental and control groups scored disease severity in each joint according to the following scale: 0, no evidence of erythema or swelling; 1, erythema and mild swelling confined to the mid-foot (tarsals) or ankle joint; 2, erythema and mild swelling extending from the ankle to the mid-foot; 3, erythema and moderate swelling extending from the ankle to the metatarsal joints; and 4, erythema and severe swelling encompassing the ankle, foot, and digits. The maximum possible score per mouse was 16 (4 joints × 4).
Antigen-specific T cell stimulation. Draining lymph node cells (5 × 10 5 ) from mice were incubated with Nil (unstimulated cell), CII (40 μg/ml) or ovalbumin (OVA, Sigma-Aldrich) (40 μg/ml)-stimulated cells for three days at the same time points. OVA was used as the negative control. The figure depicts intracellular staining for Foxp3 expression and IL-17 production and is representative of three independent experiments. Frequencies of CD4+IL-17+ T cells and CD25+Foxp3+ cells in the gated CD4+ T cells populations were assessed using intracellular flow cytometry.

Enzyme-linked immunosorbent assay (ELISA).
Concentrations of IL-10, TGF-β, IL-6, and IL-17 were measured using a sandwich ELISA as follows. Anti-mouse IL-10, TGF-β, IL-6, or IL-17 monoclonal antibodies (R&D Systems) were added to a 96-well plate (Nunc, Roskilde, Denmark) and incubated overnight at 4 °C. Wells were blocked with blocking solution (PBS containing 1% bovine serum albumin and 0.05% Tween 20) for 2 h at room temperature. Test samples and standard recombinant mouse IL-10, IL-6, IL-17 and, human TGF-β (R&D Systems) were added to separate wells and the plate was incubated at room temperature for 2 h, after which the plate was washed. Biotinylated IL-10, TGF-β, IL-6, and IL-17 polyclonal antibodies (R&D Systems) were added, and the reaction was allowed to proceed for 2 h at room temperature. The plate was washed, ExtrAvidin-alkaline phosphate (1:2,000 dilution) (Sigma-Aldrich) was added, and the reaction was allowed to proceed for an additional 2 h. The plate was washed and 50 μl p-nitrophenyl phosphate disodium salt (Pierce Chemical Company, Rockford, IL), diluted to 1 mg/ml in diethanolamine buffer, and was applied. Experiments were performed according to the manufacturer's instructions.
Western blot analysis. Spleen cells were collected from DBA1J mice one week following cell therapy. Spleen cells were prepared from 5 × 10 6 cells by homogenization in lysis buffer with a protease/phosphatase inhibitor cocktail (Cell Signaling, Danvers, MA) and centrifuged for 15 min at 14,000 revolutions per min. The protein concentration in the supernatant was measured by the Bradford method (Bio-Rad). Protein samples were separated by sodium dodecyl sulfate (SDS) gel electrophoresis and transferred to a nitrocellulose membrane (Amersham Pharmacia Biotech, Buckinghamshire, UK). Membranes were stained with primary antibodies specific to p-STAT3 (727), STAT3, p-STAT5, STAT5, or β-actin (Cell Signaling, Danvers, MA). HRP-conjugated secondary antibodies were then added. After washing with Tris-buffered saline and Tween 20 (TBST), hybridized bands were detected using an enhanced chemiluminescence (ECL) detection kit and Hyperfilm-ECL reagents (Amersham Pharmacia Biotech).
Statistical analysis. Statistical significance was determined using Student's two-tailed t-test and one-way analysis of variance (ANOVA) with Bonferroni correction for multiple comparisons. Differences between arthritis incidences at a given time point were evaluated by χ 2 contingency analysis. In all analyses, P values less than 0.05 were considered to indicate statistical significance.