Immunomodulatory Function of Bone Marrow-Derived Mesenchymal Stem Cells in Experimental Autoimmune Type 1 Diabetes¹

Mesenchymal stem cells (MSC)⁴ are multipotent progenitor cells that can be isolated from a number of sources, including bone marrow (BM). MSC have been noted for their ability to give rise to cells of various lineages, including bone, cartilage, and adipose tissues (1). Isolation of MSC is commonly performed from BM, based on procedures involving the adherence of fibroblast-like cells to the plastic substrate of cell culture plates, together with the exclusion of marrow-derived hematopoietic cells (2, 3). MSC have been characterized in humans and several animal models using a battery of negative and positive cellular markers (3). Although subject to some discrepancy, a consensus opinion suggests they lack specific cell surface markers of hematopoietic cells (CD34), monocytes/macrophages (CD14), lymphocytes (CD11a/LFA-1), leukocytes (CD45), RBC (glycophorin A), and endothelial cells (CD31), but express SH2 (CD105 or endoglin), SH3, SH4, CD44, and CD29 (3, 4, 5, 6). The primary impetus behind MSC research has been their developmental plasticity and ability to replace injured tissues, yet MSC have also been noted for their profound immunomodulatory effects in vivo (7, 8). Indeed, MSC have been used to reduce the burden of a variety of immune-mediated diseases, including graft rejection, graft-vs-host disease, collagen-induced arthritis, and myelin oligodendrocyte glycoprotein-induced experimental autoimmune encephalomyelitis (8, 9, 10, 11). A significant number of trials have also been designed to assess the safety, feasibility, and efficacy of MSC therapy for a multitude of disorders in humans (12, 13, 14). Among these, Le Blanc et al. have shown promising results in using MSC for the treatment of steroid-refractory graft-vs-host disease in humans (15, 16). Clinical studies have also yielded success in treating common diseases such as cancer, heart failure from massive myocardial infarction, and neurological diseases (17, 18, 19, 20, 21, 22, 23). Taken collectively, these reports indicate that MSC may be of considerable therapeutic benefit and provide a basis for the development of efficacious and safe stem cell therapies (24). At least two clinical trials using allogeneic (Osiris) and autologous (European Consortium) MSC for therapy of recent onset type 1 diabetes (T1D) have recently been formed, without preclinical data (14). In this report, we characterize MSC obtained from diabetes-prone as well as -resistant mice, and evaluate their immunomodulatory effects in autoimmune T1D by assessing their ability to prevent the onset of diabetes, or to reverse overt hyperglycemia, in NOD mice.

BALB/c ByJ, NOD/LtJ, nonobese resistant mice (NOR)/LtJ, BDC2.5, CBA/J, and C57BL/6 mice were purchased from The Jackson Laboratory. Animals were cared for and/or bred under specific pathogen-free conditions at the Harvard Medical School Facilities for Animal Care and Housing. Protocols were approved by the Institutional Animal Care and Use Committee.

To generate MSC, BM mononuclear cells were isolated from the femurs and tibiae of at least five mice to minimize cell variability. Cells were seeded in flasks at a concentration of 10 × 10⁶/25 cm² in M10 medium (DMEM medium [Cambrex] containing 10% FCS [HyClone], 1% penicillin-streptomycin, and 1% l-glutamine [both from Cambrex]). To examine MSC in an inflammatory setting, 7.5 × 10⁵ NOD- or BALB/c-MSC/well were cultured for 48 h in 6-well plates in M10 medium containing 10 ng/ml recombinant murine IL-1β (PeproTech).

MSC were analyzed for surface expression of a battery of markers at passage 4 (P4). Anti-mouse Abs against CD45, Sca-1, CD44, CD90.2, CD73, CD80, CD86, CD40, CD40L, Fas, and FasL were purchased from BD Biosciences. Abs purchased from eBioscience were CD105, CD29, ICOS, OX40, PD-1, PD-L1, and PD-L2. T regulatory cells (Tregs) were assessed using CD4, CD25, and FoxP3 (eBioscience).

To ensure that our cultured cells had multipotent potential, we tested MSC P4 cultures for their ability to undergo differentiation into chondrocytes, osteocytes, and adipocytes as previously published (10). Chondrogenic differentiation was induced by 50 μg/ml ascorbic acid and 1 ng/ml TGFβ-1 (PeproTech); osteogenic differentiation was induced by 50 μg/ml ascorbic acid, 10 mM sodium α-glycerophosphate, and 10⁻⁸ M dexamethasone; adipogenic differentiation was induced by 10⁻⁷ M dexamethasone and 6 ng/ml insulin.

Following fixation of adherent BALB/c- and NOD-MSC on slides using Merkofix (EMD Chemicals), MSC were incubated with primary Ab at 4°C overnight followed by the avidin-biotin complex procedure (25). Immunoreactions were employed using CD34, CD44, and CD105 (eBioscience) and Fas (Santa Cruz Biotechnology) abs. For electron microscopy, samples were fixed for 2 h in a solution of 2% paraformaldehyde and 2% glutaraldehyde in 0.05 M pH 7.3 cacodylate buffer at 4°C.

Immunohistology was performed as described previously (26).

T cells were extracted from spleens of BDC2.5 mice with CD4 magnetic beads (Miltenyi Biotec). Cells were then labeled using the Vybrant CFDA SE Cell Tracer kit (Invitrogen). Irradiated dendritic cells (DC; 5 × 10⁴) from NOD mice were incubated with 1 × 10⁵ CFSE-labeled CD4⁺ BDC2.5 cells and 100 ng/ml BDC2.5 islet peptide as previously described (26). BALB/c- and NOD-MSC (2 × 10⁴) irradiated with 3000 rad were added to assess the immunomodulatory effect on proliferation compared with control wells. At 72 h, cells were subjected to flow cytometric analysis to assess CFSE dilution.

DC (5 × 10⁴) generated from C57BL/6 BM were harvested at day 8 and used to stimulate 1 × 10⁵ CD4⁺ BALB/c cells isolated using CD4 microbeads (Miltenyi Biotec). Irradiated BALB/c- and NOD-MSC (2 × 10⁴) were added to determine their effect on T cell proliferation, in which cell counts were measured at day 3 following pulsing for 3 h with TdR (PerkinElmer) with a liquid scintillation counter.

The effect of MSC on cytokine production of autoreactive T cells was examined with the ELISPOT assay using 1 × 10⁵ CD4⁺ BDC2.5 cells, 5 × 10⁴ irradiated NOD DC, and 100 ng/ml BDC2.5 islet peptide for in vitro studies and 1 × 10⁶ splenocytes with 100 μg/ml BDC2.5 islet peptide for ex vivo prevention studies to assess IFN-γ, IL-4, IL-6, and IL-10 as described previously (26).

To assess cytokine production of murine serum and culture supernatant samples, a 21-plex cytokine kit (Millipore) was used according to the manufacturer’s instructions to analyze production of indicated cytokines.

To determine the effect of MSC administration on the prevention of diabetes, we injected 5 × 10⁵ MSC (BALB/c or NOD) i.v. once a week for 4 wk in NOD mice beginning at 10 wk of age. Blood glucose measurements were performed by tail bleeding according to National Institutes of Health guidelines. For prevention studies, >250 mg glucose/dL for more than 3 days was indicative of greater than 80% loss of islet mass and onset of diabetes. For reversal studies, we placed a s.c. sustained-release insulin pellet (LinBit; LinShin Canada) on day 2 of hyperglycemia. Within 24 h of pellet placement, we performed the first MSC injection (1 × 10⁶ cells i.v.). After all treatments, mice were monitored daily by blood glucose measurement until the time of sacrifice.

RNA was isolated from MSC at P4 with the RNeasy kit (Qiagen) and was analyzed as previously described (26).

BALB/c- and NOD-MSC were harvested and labeled with CFSE as above. Cells were then injected into the tail vein of 10-wk-old female NOD mice. Age-matched, untreated NOD female mice were used to establish basal tissue autofluorescence. Animals were killed at day 3 postinjection. Spleen, pancreatic lymph nodes (PLN), and pancreata were removed, frozen in OCT and cut; 4-μm sections of the samples were examined to discern CFSE-labeled cells.

To characterize our MSC colonies, MSC were generated from BM harvested from 10-wk-old female NOD and BALB/c mice and were evaluated morphologically by light and electron microscopy. Although BALB/c-MSC displayed a typical small spindle fibroblast-like pattern with eosinophilic cytoplasm and large clear nuclei showing dispersed chromatin and evident nucleoli, NOD-MSC formed large aggregates, primarily composed of packed, mid-sized to large cells, with rare spindle fibroblast-like cells (Fig. 1,A). At the ultrastructural level, BALB/c-MSC showed more abundant cytoplasm with rare short microvilli compared with NOD-MSC (Fig. 1, E and F). By immunohistochemistry, both BALB/c- and NOD-MSC were observed to be negative for CD34 (Fig. 1,B) and positive for CD44 (Fig. 1,C) and CD105 (Fig. 1,D), an expression pattern reported to identify MSC (14, 24). MSC cultured from NOD and BALB/c mice were able to differentiate into chondrocytes, osteocytes, and adipocytes (Fig. 1, G–I).

We also evaluated the expression of stem cell and immunoregulatory markers by flow cytometry. NOD- and BALB/c-MSC were characterized at P4 with the following panel of Abs: 1) stem cell markers (CD29, CD44, CD73, CD105, CD166, and Sca-1); 2) hematopoietic/lineage markers (CD45 and CD90.2); 3) costimulatory molecules (positive: CD40, CD80, CD86, and OX40; negative: PD-1, PD-L1, and PD-L2); and 4) apoptotic factors (Fas and FasL). Stem cell marker expression was positive for CD29, CD44, CD73, CD105, CD166, and Sca-1 in both strains (Fig. 2, A, B, D, F, G, and H, respectively), although CD105 expression was lesser in NOD- than in BALB/c-MSC (p = 0.001; Fig. 2,F). Neither BALB/c- nor NOD-MSC expressed lymphoid or myeloid markers (CD45 and CD90.2; Fig. 2, C and E). Positive costimulatory molecules appeared to be negative in both strains (Fig. 2, I–L), and the negative costimulatory molecules PD-1 and PD-L2 were not expressed in either strain (Fig. 2, M and O). However, PD-L1 expression was found to be 4-fold greater in BALB/c-MSC compared with NOD-MSC (p = 0.01; Fig. 2,N). Although FasL was negative in both NOD- and BALB/c-MSC mice (Fig. 2,Q), Fas (a receptor with proapoptotic properties) is expressed by both strains (Fig. 2,P). We and others have previously shown that diabetes development is accelerated in NOD mice by treatment with a blocking PD-L1 Ab, as well as in PD-L1-deficient NOD mice (27, 28). Furthermore, Augello et al. demonstrated that MSC regulate the immune response in vitro through cell-to-cell contact, which appears to be largely dependent on the PD-1 pathway (29). We therefore further investigated the dynamic changes that occur in the expression of PD-L1 following MSC stimulation with IL-1β, thereby mimicking an in vivo inflammatory setting relative to T1D and based on the notion that the immunoregulatory characteristics of MSC most likely change in an inflammatory environment. Following challenge with IL-1β for 48 h, the difference in the expression of PD-L1 became more distinct in BALB/c-MSC (BALB/c = 21.0 ± 4.4 vs NOD = 5.4 ± 1.0%, p = 0.024; Fig. 2 R). Altogether, a greater level of PD-L1 on BALB/c-MSC vs NOD-MSC may enable BALB/c-MSC to deliver immunosuppressive signals and to maintain quiescence of autoreactive T cells.

To elucidate the regulatory function of MSC, we evaluated their production of inflammatory cytokines that may play a role in the pathogenesis of T1D, using the Luminex assay to analyze the supernatants of BALB/c- and NOD-MSC cultures at P4. Interestingly, NOD-MSC appeared to produce more proinflammatory cytokines and chemokines, including IL-1α (p = 0.01), IL-1β (p = 0.04), IP-10 (p = 0.003), and MCP-1 (p = 0.007) (Fig. 3,A) than BALB/C-MSC did. No difference was noted in the levels of IFN-γ, IL-6, IL-9, IL-10, IL-12p40, IL-12p70, IL-15, MIG, or RANTES production. The immunosuppressive abilities of BALB/c- and NOD-MSC were also studied in an MLR assay. BALB/c- or NOD-MSC (2 × 10⁴) were added, following irradiation with 3000 rad to prevent their proliferation in culture, to a fully mismatched MLR assay in which 1 × 10⁵ C57BL/6 CD4⁺ cells isolated from splenocytes were used as responders, with irradiated DC generated from CBA BM used as stimulators. MSC were therefore added as third-party cells. BALB/c- and NOD-MSC were both found to inhibit the MLR response in comparison to controls (p < 0.001), although BALB/c-MSC were noted to be more immunosuppressive (p < 0.001) in comparison to NOD-MSC (Fig. 3,B). These findings confirm that both strains have similar immunosuppressive properties in an alloimmune setting in vitro. To evaluate the immunomodulatory abilities of MSC in an autoimmune setting in vitro, we used an autoreactive T cell proliferation assay shown to provide useful information with respect to T cell immunity to islet-cell Ags in T1D (26). The immunosuppressive properties of MSC were thus examined by assessing their effect on the proliferation of autoreactive T cells extracted from transgenic TCR BDC2.5 mice in response to autoantigen by CFSE dilution. CFSE-labeled CD4⁺ T cells purified from splenocytes of BDC2.5 mice were cultured for 3 days in the presence of BM-derived NOD DC, BALB/c- or NOD-MSC, and 100 ng/ml BDC2.5 peptide. As shown in Fig. 3 C, both BALB/c- and NOD-MSC comparably inhibited the proliferative response of CD4⁺ T cells (p = 0.04 compared with untreated control), demonstrating that both strains of MSC have similar immunomodulatory properties against autoreactive T cells in vitro.

We next tested the ability of P4 NOD- or BALB/c-MSC to prevent diabetes onset when injected into prediabetic NOD mice. Specifically, we injected 5 × 10⁵ BALB/c- or NOD-MSC i.v. once a week for 4 wk into prediabetic 10-wk-old NOD mice (n = 35 for BALB/c-MSC, n = 29 for NOD-MSC, n = 29 for controls, n = 3 experiments). As shown in Fig. 4,A, BALB/c-MSC (p = 0.04 compared with controls), but not NOD-MSC, significantly delayed the onset of T1D in 37% of treated NOD mice in the long-term (up to 40 wk of age) as compared with 20% of control NOD mice. In conjunction with the survival data, we also assessed the pathological finding of pancreatic islets in treated and control animals. Pancreatic islet morphology and infiltration were evaluated in MSC-treated or untreated NOD mice. Pancreata were harvested from normoglycemic NOD mice from the three groups (BALB/c-MSC-treated, NOD-MSC-treated, and untreated) at three different time points and were analyzed for infiltration and morphology (n = 3 mice/group). At 14 wk of age, islets appeared well-preserved in normoglycemic BALB/c-MSC-treated mice, while in normoglycemic NOD-MSC-treated or in untreated mice, a slightly more pronounced islet infiltration was observed (data not shown). At 22 wk of age, normoglycemic BALB/c-MSC-treated mice showed a scanty infiltrate, reduced in quantity to age-matched normoglycemic NOD-MSC-treated or untreated control mice (Fig. 4,B, 1–3, respectively). With respect to insulitis score, while BALB/c-MSC-treated mice showed a pattern similar to untreated 10-wk-old normoglycemic NOD mice, the insulitis score of BALB/c-MSC-treated mice was lower than that of age-matched NOD-MSC-treated and untreated NOD mice (Fig. 4,C). Further investigation of the cell infiltrate subsets revealed fewer CD3⁺ (Fig. 4,B, 4–6) and B220⁺ cells (Fig. 4,B, 10–12) in BALB/c-MSC-treated mice compared with NOD-MSC-treated and untreated control mice. Very few FoxP3⁺ cells were detected within the islets of all three groups of mice (Fig. 4,B, 7–9). Insulin (Fig. 4,B, 13–15) and glucagon (Fig. 4 B, 16–18) staining showed well-preserved islets, particularly in the BALB/c-MSC-treated group, whereas a loss of insulin staining and of islet morphology was evident in NOD-MSC-treated and untreated control mice. At 40 wk of age, normoglycemic BALB/c-MSC-treated mice maintained less infiltrated islets than NOD-MSC-treated and untreated control mice (data not shown). However, insulin staining appeared weak in BALB/c-MSC-treated mice and was negative in NOD-MSC-treated and untreated control mice, whereas glucagon was preserved in BALB/c-MSC-treated mice only (data not shown). Altogether, our data support the fact that BALB/c-MSC are capable of delaying the onset of diabetes. These are vital preclinical data to support the concept of translating MSC-based cell therapy to individuals with T1D.

The Th1 vs Th2 cytokine shift has also been reported to affect the incidence of T1D (30). We assessed the Th1/Th2 cytokine patterns in response to MSC treatment by stimulating splenocytes recovered with BDC2.5 peptide from control and treated animals, as previously described (26). Splenocytes of normoglycemic BALB/c- or NOD-MSC-treated and normoglycemic untreated control mice from our prevention studies were isolated at 14 wk of age (i.e., after 4 wk of treatment) and were challenged with BDC2.5 peptide in an ELISPOT assay to evaluate IFN-γ, IL-4, and IL-10 production. As shown in Fig. 4,D, both BALB/c- and NOD-MSC-treated animals exhibited a lower frequency of IFN-γ-producing cells than untreated normoglycemic NOD mice (p = 0.01 and p = 0.005 respectively for BALB/c- and NOD-MSC-treated mice). However, a shift toward a Th2 profile, with an increase in production of IL-4 (p < 0.0001 compared with controls) and IL-10 (p = 0.01 compared with controls), was evident in splenocytes isolated from BALB/c-MSC-treated mice. Such a shift was not observed in NOD-MSC-treated mice compared with control during autoantigen stimulation (Fig. 4 D). Th2 cytokines have commonly been reported to confer protection against Th1-mediated destruction of β-islet cells (31).

Tregs have been shown to play a crucial role in regulating autoimmunity, (32, 33) and MSC have been shown to facilitate the generation of Tregs (34, 35). Upon examination of the effect of MSC treatment on Treg percentage, we noted a marginal but not substantial increase in the percentage of CD4⁺CD25⁺FoxP3⁺ cells at 14 wk of age in the PLN of normoglycemic BALB/c-MSC-treated, but not of NOD-MSC-treated mice, compared with untreated age-matched controls (BALB/c-MSC-treated = 12.6 ± 0.8, untreated control = 10.0 ± 0.8, NOD-MSC-treated = 10.3 ± 1.1%, p = 0.04 BALB/c vs NOD-MSC-treated). No difference was observed in the percentage of Tregs recovered from the spleens of these mice (data not shown).

Given the protective effect of BALB/c-MSC observed in our prevention studies, we also evaluated the ability of BALB/c-MSC to reverse diabetes in new-onset hyperglycemic NOD mice. After a second consecutive hyperglycemic measurement (>250 mg/dl) within 24 h, NOD mice were injected i.v. with 1 × 10⁶ BALB/c-MSC twice a week for 4 wk. An insulin pellet was placed under the skin of the dorsum of hyperglycemic NOD mice to improve glucose homeostasis. Although control NOD mice invariably returned to hyperglycemia (mean survival time = 22.5 days; Fig. 5, B and C), seven of eight BALB/c-MSC-treated mice remained normoglycemic for an average of 41.5 days (p = 0.0002 vs untreated mice; Fig. 5, A and C). Two BALB/c-MSC-treated mice remained normoglycemic for more than 60 days; one then reverted to hyperglycemia and the other died normoglycemic of unknown causes (Fig. 5, A and C). Despite the lack of therapeutic efficacy in our prevention studies using NOD-MSC, we tested the effect of NOD-MSC in reversing NOD mice with recent onset diabetes. In the three mice treated with NOD-MSC, none experienced reversal of their diabetes (Fig. 5 B). Moreover, we observed tumor formation in all three hosts at the site of injection in the tail as well as at the site of the resident insulin pellet.

After 14 days of hyperglycemia in untreated NOD mice, islets are extensively infiltrated by lymphoid cells (Fig. 5,D) with disrupted structure and a reduction in insulin and glucagon staining (data not shown). The lymphoid infiltrate was composed predominantly of B220⁺ cells, along with a smaller population of CD3⁺ cells (Fig. 5,D) and very few FoxP3⁺ cells (data not shown). In the BALB/c-MSC-treated mice 14 days after treatment, islets showed very mild infiltrates confined to the borders of β cells, with an almost complete absence of B220⁺ and CD3⁺ cells (Fig. 5,E), along with well-maintained glucagon staining and scarce insulin staining (data not shown). More infiltrating cells also appeared to be FoxP3⁺ cells in islets of the BALB/c-MSC-treated mice, which were absent in control mice (data not shown). In the long-term normoglycemic (i.e., 90 days after treatment) BALB/c-MSC-treated mice, islets still appeared preserved, despite a marked increase in infiltrate (B220⁺ and CD3⁺ cells) (Fig. 5 F).

A paramount observation noted during our prevention and reversal studies was the identification of tumors in mice treated with autologous MSC. In prevention and reversal groups, NOD mice treated with NOD-MSC developed neoplasia (∼20% and 100%, respectively), a finding not observed in the BALB/c-MSC-treated group. Most tumors detected formed nodular masses in the legs and tails of mice (Fig. 6,A). Tumors were also identified as numerous nodules of 0.1 to 0.3 cm in diameter in the lungs and liver of animals (Fig. 6, B and C). Optical microscopy revealed that the malignant tumor was made up of a homogeneous population of malignant spindle cells in sheets and fascicles (Fig. 6,D). The tumor was shown to invade muscle, nerve, skin, bone (Fig. 6,E), and peritoneum, and was located adjacent to the colon (Fig. 6,F). In the lung, the tumor formed nodular masses with alveolar wall infiltration (Fig. 6,G), and in the liver appeared as nodular masses (Fig. 6,H). The tumor lacked malignant or benign-looking epithelium. Electron microscopy of the hepatic tumor identified compact, intertwined processes covered by basal lamina, consistent with Schwann cell differentiation (Fig. 6 I). The diagnosis was compatible with a malignant peripheral nerve sheath tumor. Of note, following injection of BALB/c mice with BALB/c-MSC, no neoplasia was observed up to 5 wk after the initiation of treatment.

Following culture and expansion of MSC (as above), mRNA was extracted from MSC for use in gene expression profile analysis using microarray technology. We examined the gene expression profile of NOD-MSC (samples no. 1 to 4) and BALB/c-MSC (samples no. 5 to 8) obtained from normoglycemic NOD mice and age-matched BALB/c mice. A striking differential expression in the pattern of gene regulation across the entire screen was observed when BALB/c-MSC genes were compared with the NOD-MSC (Fig. 6,J). Notably, a considerable number of genes promoting cell cycle and proliferation were up-regulated in NOD-MSC (e.g., Plf2, Birc5, Ifgbp3, Sgol2, Btc, and Ccnb1); among these, Plf, Gfra1, and Ngef are related to neural tumor development, which is consistent with the pathogenesis behind the formation of neural tumors in our animals (Tables I and II, accession number GSE15516, http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc = GSE15516). Although the formation of tumors when using NOD-MSC could be due to an abnormality within the NOD strain similar to other genetic malformations reported in this strain, the data should raise awareness regarding the potential tumor formation in settings of autologous MSC in certain immune disorders (e.g., T1D).

MSC are able to selectively migrate into sites of injury (36). Furthermore, it has been shown that diabetogenic T cells are resident to the PLN, where T cells are introduced to β-cell autoantigens, giving rise to autoantigen-specific T cells (37). To study MSC trafficking in our experimental settings, BALB/c- as well as NOD-MSC were labeled with CFSE and were injected into the tail vein of 10-wk-old NOD mice. Pancreata, spleens, and PLN of treated and untreated animals were harvested at day 3 postinjection. As shown in Fig. 6, the frequency of trafficked cells was similar in both spleen and pancreas between BALB/c-MSC-treated (Fig. 6, L and O, respectively) and NOD-MSC-treated mice (Fig. 6, M and P, respectively); however, a significant increase in the number of trafficked BALB/c-MSC was observed in the PLN as compared with NOD-MSC (Fig. 6, R and S, respectively). These data indicate that the prevention of diabetes observed with treatment of BALB/c-MSC could be due in part to the preferential trafficking of BALB/c-MSC to the PLN, where they may interfere or deliver inhibitory signals to prevent the expansion of autoreactive T cells. Chemokines have increasingly become recognized as the key regulators of such preferential migration of cells, including MSC (38). It is likely that the disparity in the migration of MSC from these two strains could be due to the differential expression of chemokine receptors. We therefore assessed the expression of chemokine receptors CCR5, CCR7, CXCR3, and CXCR4 on MSC by FACS. The expression of CXCR3 and CCR7 was significantly higher on BALB/c-MSC in comparison to NOD-MSC (Fig. 6 T). Interestingly, CXCR3 has in particular been associated with the migration of autoreactive T cells into the PLN in the presence of high gradients of IP-10 (CXCL-10) in the PLN (39).

It could be argued that the preventive effect we observed in the case of BALB/c-MSC was due to the activation of immune responses by allogeneic cells. To examine the effect of MSC from a strain more similar to NOD, we isolated MSC from NOR mice. NOR is a diabetes-resistant strain with ∼90% genomic identity with NOD mice. NOR share, for instance, the MHC class II molecule of NOD mice, including the diabetogenic H-2g7 complex (40). For this effort, 5 × 10⁵ MSC were injected into prediabetic female NOD mice (10 wk of age) once a week for 4 wk, a strategy identical with treatments using BALB/c- and NOD-MSC. As shown in Fig. 7,A, delivery of NOR-MSC to prediabetic NOD mice was able to prevent diabetes. Furthermore, histological analysis of the pancreas harvested from NOR-MSC-treated mice at 30 wk posttreatment showed preserved islets with substantial insulin staining and minimal cellular infiltrate (Fig. 7 B). Importantly, similar to BALB/c-MSC treated mice, we did not observe any tumor development in NOR-MSC-treated recipients.

The incidence of T1D continues to climb steadily, making it one of the most challenging health issues of the 21st century worldwide (41). In addition, the morbidity and mortality associated with diabetes are constantly rising and are anticipated to remain a problem for years to come (41, 42). The use of immunosuppressants to reverse or halt diabetes has received much attention, but the shortfall of monotherapy and serious morbidity from lifelong immunosuppression has prompted investigators to search for alternative approaches. Investigators have begun to use MSC much more quickly than other cell-based therapies in a variety of human diseases. Interestingly, administration of MSC in many of these settings has offered significant benefits with no major adverse effects. Although MSC plasticity is a propitious feature of the cell and has been used as a rationale to develop clinical trials (http://clinicaltrials.gov; see Refs. 17, 18, 19, 20, 21, 22, 23), the immunomodulatory effects of MSC have recently become of interest as well (11, 12, 14, 24).

Our data in this study show that allogeneic MSC are capable of protecting islet mass and of delaying the onset of disease when injected into prediabetic NOD mice. Administration of allogeneic BALB/c-MSC also showed promising results in temporarily reversing hyperglycemia in new-onset diabetic NOD mice. These novel data indicate that MSC could be a component of an innovative and successful cell-based regimen for T1D treatment on its own. An important lesson from intervention therapies in T1D is that monotherapy often fails to provide a durable favorable outcome, particularly for reversal of the disease. In theory, MSC therapy could allow investigators to use lesser amounts of immunosuppressive agents, which, as previously noted, have been shown to carry a high degree of toxicity. Although the hypoimmunogenicity of allogeneic MSC is certainly of value, MSC are reported to be efficiently recognized by the host immune response and eventually are rejected (43). The lack of a long-term effect of BALB/c-MSC may be related to this recognition and subsequent rejection. Therefore, future studies should focus on understanding the life span of MSC and the value of repeated administration for a longer period of time. Additional research is also needed to identify various synergistic immunosuppressive strategies and to optimize prevention and reversal of established diabetes. Given that this strategy requires the generation of large numbers of MSC, these studies will certainly represent a considerable undertaking. Although the use of MSC comprises an exciting therapeutic approach for T1D, translating our results to the clinic requires greater investigation to provide robust preclinical data to safely and effectively use MSC as a cell-based therapy for T1D.

Our data demonstrating the similarity in the ability of BALB/c- and NOD-MSC to suppress the T cell response in vitro is in accordance with data using MSC from healthy individuals and from patients with autoimmune diseases such as scleroderma and rheumatoid arthritis in which comparable suppression of T cell proliferation was achieved in vitro (44). Given the disparity we observed in the in vivo effects of NOD- and BALB/c-MSC in regulating diabetes, we further examined MSC from these two strains to explain the outcome differences noted in our studies. Although as previously stated, both strains of MSC were able to suppress alloimmune and autoimmune responses, several intrinsic differences were evident between them. First, an increased growth rate of NOD-MSC as compared with BALB/c-MSC was observed as well as a resultant production of a proinflammatory or diabetogenic cytokine profile. Gene arrays confirm the presence of two distinct gene profiles, with NOD-MSC expressing more proinflammatory genes compared with BALB/c-MSC. Furthermore, ultrastructural analysis with electron microscopy revealed differences in intracytoplasmic organelles and pseudopods, which may have an impact on cell-to-cell interactions as well as on cell trafficking. Interestingly, the expression of PD-L1, a major regulatory molecule in T1D, was much greater in BALB/c-MSC, which may serve as a potential mechanism by which BALB/c-MSC exert their suppressive effect on autoreactive T cells in vivo (45). Following injection of MSC, we did observe an increase in the percentage of Tregs and Th2 cytokines, both of which have been shown to be protective in T1D (32, 33, 46, 47). A key feature of MSC is their ability to selectively migrate into sites of injury. Moreover, it has been shown that diabetogenic T cells are generated in the PLN, where they are introduced to Ags by DC. The preferential recruitment of BALB/c-MSC to the lymph node, along with high levels of PD-L1 expression on BALB/c-MSC, supports the idea that these cells could directly suppress autoreactive T cells in vivo. Our chemokine expression studies of MSC suggest a preferential expression of specific chemokine receptors on BALB/c-MSC, which correlates with the postulation that by inhibiting or stimulating chemokines/receptors, it will be possible to promote the homing of MSC to specific anatomical sites to achieve the desired therapeutic effects (36). It is known that immunoregulatory effects of MSC are exerted through different putative mechanisms, including immune cell interaction, production of soluble factors, and promotion of Treg generation (14). These mechanisms are most likely not mutually exclusive, and future studies should examine their relative contributions to the immunomodulatory functions of MSC. It should be noted that given the plasticity of MSC, transdifferentiation of autologous MSC could occur; however, because NOD-MSC were observed to have no effect in our prevention studies, the possibility of MSC differentiation into islets for example seems unlikely. Nevertheless, this requires additional experiments beyond the scope of the present study.

We observed rapidly growing tumors in mice treated with autologous MSC. Tumors were detected in the liver and lungs of our injected mice, where a large number of MSC are retained after i.v. injection. The incidence of tumor formation was higher in the reversal group, most likely due to the more intense administration regimen of NOD-MSC in the reversal group compared with the prevention group. Inbred mice such as NOD have been shown to carry various genetic abnormalities and are prone to developing lymphoid tumors, therefore it remains to be elucidated whether autologous MSC indeed pose a risk, whether the observed tumor formation is related to NOD abnormalities, or whether it has any clinical implication for humans (48). At present, tumor formation following use of autologous MSC has not been demonstrated to be a major impediment to current clinical trials, although this has not been studied in the context of T1D. Nevertheless, it is important to impart this message to any studies aiming to use autologous MSC to treat the disease of interest (49). The lack of tumor formation in the case of allogeneic BALB/c-MSC could be due to their recognition by the host immunosurveillance, ensuring that their proliferation is controlled. Moreover, NOD-MSC were unable to exhibit any effect on the prevention of diabetes. Lack of tumor formation using our NOR-MSC, along with a simultaneous effect on suppressing diabetes, may argue in favor of using MSC from diabetes-resistant siblings. In summary, we believe our data provide a basis for the development of a cell-based therapy for which MSC from nondiabetic donors could be isolated and administered to patients for the prevention and reversal of T1D. Indeed, the translation of this knowledge into clinical practice could eventually transform the overall strategy we use to treat this disease.

Microarray studies were performed by the Molecular Genetics Core Facility at Children’s Hospital Boston.

The authors have no financial conflict of interest.