Highly Angiogenic, Nonthrombogenic Bone Marrow Mononuclear Cell–Derived Spheroids in Intraportal Islet Transplantation

Despite the encouraging results of the first multicenter phase 3 clinical trial in patients with type 1 diabetes complicated by severe hypoglycemia (1), the long-term metabolic outcome of clinical allogeneic islet transplantation is still not curative, even with potent immunosuppression (1–3). Although impaired revascularization of islet grafts in the portal sinusoid has long been considered one of the most important factors in the destruction of islet grafts in the early posttransplant period (4,5), no definite solution is clinically available. In this context, cotransplantation of bone marrow–derived mononuclear cells (BM-MNCs) and early or late endothelial progenitor cells with islets has been attempted and has shown promising results in small animal models (6–9), although the scarcity of the therapeutic cell population has been an obstacle to this approach. Recently, we tested a strategy to expand an autologous therapeutic cell population that possesses higher angiogenic or paracrine activities (10,11), namely, cotransplantation of islets with spheroids formed by selective ex vivo proliferation of the highly angiogenic CD31⁺CD14⁺CD34⁺ monocyte subset via a three-dimensional (3D) culture technique that uses a high cell density and nonattachable culture system (12). In our previous study, highly angiogenic bone marrow–derived spheroids (BM-spheroids) not only had robust angiogenic capacities, and actively participated in vessel formation in vivo, but also improved the survival and function of islet grafts in a mouse syngeneic marginal mass renal subcapsular islet transplantation model (12).

However, clinical application of this approach should be preceded by proof of its efficacy and safety in a clinically relevant intraportal islet transplantation model, in which loss of up to 60% of islets occurs in the first 2–3 days after transplantation, primarily because of the nonspecific inflammatory and thrombotic reaction called the instant blood-mediated inflammatory reaction (IBMIR) (13–15). The colocalization of islets with other therapeutic cells in portal sinusoids is another issue for the clinical application of this approach to intraportal islet transplantation because cotransplanted therapeutic cells could circulate throughout the whole body after intraportal injection, whereas islets are entrapped in the portal sinusoids. For reproduction of the benefits of their possible paracrine effects and direct incorporation into vessel formation in the renal subcapsular islet transplantation model in which islets and BM-spheroids are physically colocalized, the colocalization of BM-spheroids and islets in portal sinusoids should be secured in large-animal intraportal islet transplantation models. In addition, the safety of the therapeutic cells cotransplanted with islets should be documented in an intraportal islet transplantation model, with attention to previous studies that have reported that mesenchymal stem cells (MSCs) might not be compatible with human blood and might trigger IBMIR, depending on the cell dose and passage number (16). Thus, compatibility of BM-spheroids with recipient blood in the portal sinusoid should be confirmed before any human trial of this approach.

In this study, we wondered whether the robust angiogenic efficacy of BM-spheroids could be demonstrated in rodent and nonhuman primate (NHP) intraportal islet transplantation models without increasing the risk of portal thrombosis.

C57BL/6J or green fluorescent protein-transgenic (GFP-Tg) mice from the C57BL/6J background were obtained from The Jackson Laboratory (Bar Harbor, ME). Male GFP-Tg and wild-type C57BL/6J mice aged 10–12 weeks were used as donors and recipients, depending on the experiment. Diabetes was induced with streptozotocin (STZ) (Sigma-Aldrich, St. Louis, MO), administered intraperitoneally at 180 mg/kg in a citrate buffer solution (Sigma-Aldrich). Blood glucose concentration was measured with Accu-Chek glucose monitors (Roche, Mannheim, Germany) using blood from tail pinpricks. Diabetic mice with nonfasting blood glucose level ≥400 mg/dL for more than two consecutive days were used as recipients for islet grafts.

Cynomolgus monkeys (Macaca fascicularis) were 4–5 years of age and weighed 3–4 kg (Orient Bio Co. Ltd, Seongnam, Korea). Diabetes was induced with subtotal pancreatectomy followed by intravenous administration of 60–80 mg/kg STZ as previously described (17) (Supplementary Data). After induction of diabetes, all four NHPs satisfied all three of the following criteria (18): 1) sustained hyperglycemia (blood glucose level >250 mg/dL), 2) fasting NHP C-peptide level consistently <1.0 ng/mL or less than one-third of the preinduction level, and 3) absence of stimulated C-peptide response with intravenous glucose tolerance test (IVGTT).

Mice experiments were approved by the institutional animal care and use committees (IACUCs) of Samsung Medical Center and the Korea Advanced Institute of Science and Technology. Monkey experiments were approved by the IACUCs of Samsung Medical Center and Orient Bio Laboratories.

We performed 3D culture of mice and NHP BM-MNCs as previously described (12) (Supplementary Data). The experimental scheme for the intraportal cotransplantation of allogeneic islets and autologous BM-spheroids in the NHP model is shown in Supplementary Fig. 1A. Autologous BM-MNCs derived from the NHP recipients were counted and stored in liquid nitrogen until the diabetic NHP recipients received allogeneic NHP islet transplants. Frozen BM-MNCs were thawed and cultured for formation of spheroids 4–5 days before islet transplantation. The average size of the spheroids was 108 ± 4.3 μm (Supplementary Fig. 1B and C), and the average number of single cells per spheroid was 2,063 ± 831 cells. The viability of the thawed BM-MNCs from three NHP recipients was ∼90%.

We tracked BM-spheroids labeled with ferucarbotran (Resovist; Bayer Schering Pharma AG, Berlin, Germany), a superparamagnetic iron oxide (SPIO), by MRI in the mouse and NHP models. GFP BM-spheroids derived from GFP-Tg mice were tracked by ex vivo bioluminescent imaging. To confirm the presence of the intraportally transplanted single cells dissociated from the BM-spheroids and intact BM-spheroids in the liver during the early posttransplant period, we used intravital microscopy (IVM) using a custom-built video-rate confocal microscopy platform as previously described (19,20). Detailed methods of SPIO labeling and MRI (21), ex vivo bioluminescent imaging, and IVM imaging are available in the Supplementary Data.

Mouse pancreatic-derived MSCs (mPMSCs) were isolated from male GFP-Tg and wild-type C57BL/6J mice aged 10–12 weeks. Pancreatic cells left at the bottom after islet isolation by density gradient separation were washed in PBS and cultured with α-minimum essential medium supplemented with 10% FBS and 1% penicillin/streptomycin. Medium was changed after 24–48 h to remove nonadherent cells. When cultures reached confluence, cells were trypsinized and subcultured. In the current study, we used mPMSCs at passage 3. The cell surface marker pattern of the mPMSCs was analyzed by flow cytometry. The expression levels of genes with thrombogenic and proinflammatory properties in MSCs, MSC-derived spheroids (MSC-spheroids), and BM-spheroids were compared by microarrays and Western blot analysis (Supplementary Data).

Mice and NHP islets were isolated using a modified Ricordi method (22,23). In the murine models, cotransplantation of islets and BM-spheroids or dissociated BM-spheroid–derived single cells (dissociated BM-spheroids) and analysis of posttransplant outcome were performed in a syngeneic marginal mass mouse intraportal islet transplantation model.

In the NHP model, allogeneic marginal mass (20,000 islet equivalents [IEQ]/kg) of NHP islets (control group, n = 2) and a mixture of allogeneic marginal mass (20,000 IEQ/kg) of NHP islets and NHP recipient autologous BM-spheroids (BM-spheroid group, n = 2: 35,750 spheroids in the BM-spheroid-1 recipient and 45,000 spheroids in the BM-spheroid-2 recipient) were transplanted with an immunosuppressive regimen used in a previous clinical study (24). Detailed methods of islet isolation, assessment of islet graft function, and immunohistochemistry of islet grafts are described in the Supplementary Data.

Values are expressed as the mean ± SEM. The two groups were evaluated using an unpaired t test, and multiple comparisons between data were analyzed with one-way ANOVA. Statistical analyses were conducted using GraphPad Prism 4 software (GraphPad Software, La Jolla, CA), and a P value <0.05 was considered statistically significant.

We first examined whether intraportal infusion of monolayer-cultured single MSCs (MSC-monolayers) and MSC-spheroids with various cell numbers per infusion evoked intraportal thrombosis. When cell-surface antigen expression was analyzed by flow cytometry, MSCs were positive for CD44, CD90.2, and Sca-1 but negative for CD31, CD34, CD45, CD117, and Flk-1 (Supplementary Fig. 2A). After 3 days of hanging drop culture, MSC-spheroids derived from different numbers (1 × 10³, 2.5 × 10³, 5 × 10³, and 1 × 10⁴) of MSCs had diameters that varied according to the number of source MSCs (116 ± 11.03 μm, 170 ± 13.73 μm, 210 ± 12.61 μm, and 272 ± 21.19 μm in diameter, respectively) (Supplementary Fig. 2B and C). The real-time-PCR results indicated that the expression levels of angiogenic factors Angpt1, Angpt2, Vegf, Fgf2, Igf1, Tgfb, and Plgf are higher in MSC-spheroids than in MSC-monolayers (Supplementary Fig. 2D).

MSC-monolayers (1 × 10⁵, 2.5 × 10⁵, and 5 × 10⁵ cells), ∼300 MSC-spheroids containing 3 × 10⁵ cells (∼1 × 10³ cells per spheroid [Sph 1 × 10³]), and ∼100 MSC-spheroids containing 1 × 10⁶ cells (∼1 × 10⁴ cells per spheroid [Sph 1 × 10⁴]) were intraportally injected. After 24 h, we grossly observed the infarcted regions in the peripheral areas of livers that received MSC-monolayers and Sph 1 × 10⁴ (Fig. 1A). The livers that received Sph 1 × 10³ did not grossly show an infarction region. However, hematoxylin-eosin (H-E) staining revealed areas of necrosis accompanied by infiltration of neutrophils in all livers that received Sph 1 × 10³, Sph 1 × 10⁴, and MSC-monolayers (Fig. 1C). The infused GFP MSC-spheroids were tightly trapped in the CD31-stained blood vessels and observed in the proximity of the necrotic regions (Fig. 1D). Both H-E staining and GFP fluorescence revealed that the MSC-spheroids were responsible for the thrombosis (Fig. 1C and D). The GFP MSC-spheroids did not scatter throughout the whole liver (Supplementary Fig. 3). In contrast, the intraportally administrated GFP MSC-monolayers, which caused most severe hepatic infarction, scattered throughout the whole liver (Supplementary Fig. 3).

We then examined whether thrombosis can be evoked by single cells dissociated from BM-spheroids (dissociated BM-spheroids) and intact BM-spheroids. Different numbers of dissociated BM-spheroids (5 × 10⁵ and 1 × 10⁶ cells) or intact BM-spheroids (containing 5 × 10⁵, 1 × 10⁶, and 2 × 10⁶ cells) were intraportally transplanted. Regardless of the infused cell number, no infarcted regions were grossly observed in the livers that received either dissociated BM-spheroids or intact BM-spheroids (Fig. 1B). Histologically, no infarcted regions were observed (Fig. 1C).

Next, we performed microarray analysis to investigate the differences in the genes involved in IBMIR between BM-spheroids and MSC-spheroids. Higher mRNA expression levels of genes with thrombogenic and proinflammatory properties (TF, FBLN1, SERPINE1, CCL2, CCL8, CXCL12, and IL6) were observed in MSC-spheroids than in BM-spheroids (Fig. 2A). Because TF (also known as coagulation factor III and F3) and CCL2 are major coagulation and proinflammatory cytokines, we used Western blotting to examine difference in the protein levels of MSC-monolayers, MSC-spheroids, and BM-spheroids. The MSC-monolayers showed higher TF protein levels than the MSC-spheroids (up to 5 × 10³ cells/spheroid). The TF and CCL2 protein levels were significantly lower in the BM-MNCs and BM-spheroids than in the MSC-monolayers and MSC-spheroids (Fig. 2C and D). Immunofluorescence also revealed that the infused MSC-monolayers and MSC-spheroids expressed TF, but the BM-spheroids did not (Fig. 2B).

To examine whether cotransplantation of BM-spheroids and islets could enhance the outcomes from intraportal islet transplantation, we evaluated in vivo islet function using a syngeneic intraportal marginal mass (300 handpicked islets) islet transplantation model. As shown in Fig. 3A, diabetic mice received islets alone (islets alone group), islets plus ∼150 BM-spheroids (containing 5 × 10⁵ cells [BM-spheroid group]), or islets plus single cells dissociated from the same number of BM-spheroids (dissociated BM-spheroid group) in the liver through the portal vein. Posttransplant blood glucose levels were lower in the BM-spheroid group than in the islets alone and dissociated BM-spheroid groups. The average blood glucose level of the BM-spheroid group was significantly lower than that of the islets alone (P < 0.01) and dissociated BM-spheroid groups (P < 0.05) at day posttransplantation (DPT) 28 (Fig. 3B–D). The diabetes reversal rate was also higher in the BM-spheroid group than in the islets alone and dissociated BM-spheroid groups (Fig. 3E). Intraperitoneal glucose tolerance tests at DPT 28 revealed that glucose tolerance in the BM-spheroid group was significantly better than in the islets alone and dissociated BM-spheroid groups (Fig. 3F). The area under the curve for glucose was smaller in the BM-spheroid group than in the other groups (Fig. 3G). Serum insulin concentrations before and 30 min after glucose injection were also higher in the BM-spheroid group than in the other groups (Fig. 3H).

Next, we intraportally cotransplanted lower numbers of BM-spheroids (1 × 10⁵ cells [lower number BM-spheroid group]) with islets and compared that outcome with outcomes from intraportal islet transplantation with or without concurrent tail vein injection of higher numbers (0.5–1 × 10⁶ cells) of single cells dissociated from BM-spheroids (tail vein dissociated BM-spheroid group) (Supplementary Fig. 4A). The posttransplant blood glucose level, diabetes reversal rate, and glucose tolerance were markedly improved in the lower number BM-spheroid group compared with those in the islets alone and tail vein dissociated BM-spheroid groups (Supplementary Fig. 4B–G).

Islet morphology and vessel density were assessed in graft-bearing livers at DPT 28. Islets were detected within the liver parenchyma and around the portal venules in all groups. Insulin staining of intraportally transplanted islets indicated more intact islet morphology in the BM-spheroid group than in the other groups (Fig. 4A). The fractional β-cell area, islet size, and islet number/total area of the BM-spheroid group were significantly greater than those in the other groups (Fig. 4B–D). The relative distribution of the glucagon-positive cells did not differ between groups (Supplementary Fig. 5A). For identification of whether BM-spheroids remained within the recipient liver, BM-spheroids from GFP-Tg mice (GFP BM-spheroids) and dissociated single cells from GFP BM-spheroids (GFP dissociated BM-spheroids) were cotransplanted with islets via the portal vein. Transplanted GFP BM-spheroids were found throughout the recipient liver at DPT 14 and 28, whereas GFP dissociated BM-spheroids were rarely observed (Fig. 4E and Supplementary Fig. 5B). The vascular density of the insulin-positive area in the BM-spheroid group was significantly higher than that in the islets alone and dissociated BM-spheroid groups at DPT 14 and 28, even without incorporation of GFP BM-spheroids into the intraislet vasculatures (Fig. 4F–H).

When the islet grafts in the experiments depicted in Supplementary Fig. 4 were analyzed, the lower number BM-spheroid group showed higher fractional β-cell area, islet size, islet number/total area, and vascular density of insulin-positive area than the islets alone and tail vein dissociated BM-spheroid groups (Supplementary Fig. 6).

When the recipient pancreata in each experiment were analyzed, insulin-stained cells were rarely observed, and even if observed, the insulin-positive cells comprised only an extremely small fraction of pancreatic islets (Supplementary Fig. 7A).

After injecting GFP dissociated BM-spheroids or intact GFP BM-spheroids via the portal vein, we used bioluminescent imaging to trace the transplanted GFP cells in the spleen, lung, and liver at DPT 1. Most of the intact GFP BM-spheroids accumulated in the liver—not in the lung or spleen. However, GFP dissociated BM-spheroids showed lower fluorescence intensity in the liver than intact GFP BM-spheroids and high fluorescence intensity in the lung (Fig. 5A).

In parallel, we aimed to confirm the distribution of BM-spheroids in recipient livers using MRI of SPIOs. BM-MNCs were labeled with ferucarbotran for 24 h and then cultured for 4 days to form BM-spheroids. Ex vivo MRI of BM-spheroids labeled with ferucarbotran revealed visible hypointense spots representing the labeled BM-spheroids (Fig. 5B). The labeling agent did not compromise the viability or 3D formation of BM-spheroids during the culture period (data not shown). When we infused the ferucarbotran-labeled dissociated BM-spheroids or ferucarbotran-labeled intact BM-spheroids via the portal vein, the number of hypointense spots representing the labeled cells in the BM-spheroid group was greater than that in the dissociated BM-spheroid group at DPT 1 and 8 (Fig. 5C).

To confirm the presence of the intraportally transplanted dissociated BM-spheroids and intact BM-spheroids in the liver during the early posttransplant period, we obtained IVM images. The number of fluorescent spots representing intact GFP BM-spheroids was greater than the number representing GFP dissociated BM-spheroids at DPT 0, 1, and 7. Variable sizes of GFP BM-spheroids were detectable at DPT 0, and their shapes were rarely spherical (Fig. 5D). GFP dissociated BM-spheroids rapidly disappeared over time, so that no fluorescent spots representing GFP dissociated BM-spheroids were observed at DPT 7 (Fig. 5D). Collectively, these results indicate that the BM-spheroids are trapped within the liver, whereas dissociated BM-spheroids drain into the systemic circulation.

To confirm the localization of NHP autologous BM-spheroids within the recipient liver, we first autologously transplanted NHP BM-spheroids labeled with an SPIO (ferucarbotran). At DPT 2, an MRI of the recipient liver revealed hypointense spots representing the labeled BM-spheroids (Fig. 6A). At DPT 17, the NHP was sacrificed, and ex vivo MRI of the liver and iron staining were performed. In this analysis, the hypointense spots were confirmed to represent Prussian blue–stained BM-spheroids (Fig. 6B). In addition, we found no evidence of thrombosis in the portal veins or venules (data not shown).

We then transplanted an allogeneic marginal mass (20,000 IEQ per recipient’s body weight in kilograms) of NHP islets with (BM-spheroid group) or without (control group) the NHP recipient’s autologous BM-spheroids into NHP recipients with STZ-induced diabetes (n = 2 per each group).

In the immunohistochemistry analysis of the recipient livers in the control group, the majority of the graft sites were infiltrated with many T cells and macrophages, and only a few islets showed successful engraftment (Fig. 6C), resulting in negligible fractional β-cell areas (Fig. 6D). In contrast, the immunohistochemistry analysis of the recipient livers in the BM-spheroid group revealed intact islet grafts with minimal infiltration of T cells and macrophages. CD31-positive vessels were observed in the periphery of the islets or within insulin-positive endocrine areas in the BM-spheroid group but not in the control group (Fig. 6C). The insulin-positive area in the BM-spheroid group was larger than that in the control group (Fig. 6D). There was no evidence of portal vein thrombosis or infarction of the liver in either group of recipients (data not shown).

The two recipients in the control group continued to receive exogenous insulin, with the doses reduced from those before islet transplantation, whereas the posttransplant blood glucose levels were maintained within the target range (random glucose <200 mg/dL) without administration of exogenous insulin in the two recipients in the BM-spheroid group (Fig. 7). In the recipient pancreata of control and BM-spheroid groups, insulin-stained cells were rarely observed, and even if observed, the insulin-positive cells comprised only an extremely small fraction of pancreatic islets (Supplementary Fig. 7B).

The 3D culture models allow various cells to behave more like in vivo cells, providing enhanced therapeutic efficacy (25) and local persistence of therapeutic cells (26). In the current study, the robust angiogenic efficacy of the BM-spheroids in our previous study (12) was reproduced in more clinically relevant intraportal islet transplantation models, and the BM-spheroids did not evoke portal thrombosis even when the infused cell number far exceeded the number of MSCs that caused liver infarction after intraportal injection in parallel experiments.

In this study, generating BM-spheroids by 3D culture prevented the rapid migration and disappearance of intraportally infused therapeutic cells. This was in contrast to the behavior of dissociated BM-spheroids in this study and MSC-monolayers in a previous study (27), which rapidly drained to the systemic circulation and became trapped in the lung after administration. The benefit of cotransplanted BM-spheroids likely depends on their colocalization with islets in portal sinusoids because dissociated BM-spheroids did not retain the benefit of islet engraftment, even with greater cell numbers than used for the intact BM-spheroids, regardless of the administration route. Given the extraislet location of GFP BM-spheroids suggesting paracrine mechanism and the rapid time course of revascularization and glycemic benefit, the main mechanism for the angiogenic efficacy of BM-spheroids is likely to be a paracrine effect rather than a direct incorporation or a systemic effect. In addition, the reduced infiltration of T cells in the islet grafts (Fig. 6C) could be another explanation for the angiogenic efficacy of BM-spheroids because the effector T-cell cytokine interferon-γ has been shown to be antiangiogenic in tumor (28) and lung ischemia models (29).

The thrombogenicity of MSC-spheroids in portal sinusoids is another novel finding of this study. Although little attention has been given to it, MSCs provoke IBMIR (16), and that could contribute to their lack of persistence in target tissue. Even though we used short-term expanded MSCs (three passages) with less procoagulant activity than MSCs with higher passage numbers, the infused MSC-monolayers scattered throughout the whole liver and caused extensive portal thrombogenesis regardless of infused cell number. Interestingly, both thrombogenicity and the increase in the level of TF expression were attenuated in MSC-spheroids compared with MSC-monolayers. The GFP MSC-spheroids did not scatter throughout the whole liver, with much less portal thrombosis than that of MSC-monolayers (Supplementary Fig. 3). In contrast to the MSC-monolayers and MSC-spheroids, the nonthrombogenicity of BM-spheroids was histologically proven in this study and was further supported by low expression levels of TF and CCL2.

The results of our pilot NHP study, which used a readily available immunosuppressive regimen, supported the clinical feasibility of cotransplantation of allogeneic islets and autologous BM-spheroids. We also demonstrated the localization of BM-spheroids in the liver of NHP recipients using a 4.7T preclinical MRI of SPIO-labeled BM-spheroids. For clinical application, this strategy has several practical benefits. First, we derived the autologous BM-spheroids from precollected bone marrow from the recipient monkey. We froze the bone marrow and thawed it immediately before islet isolation. This is compatible with actual clinical settings for allogeneic islet transplantation, in which isolated islets are typically cultured for up to 3 days (1). Second, the procedure does not require surface modification of transplanted islets, which usually causes loss of therapeutic islet mass due to structural destruction of isolated islets. This is in contrast to other approaches, such as encapsulation of islets (30), islet surface camouflage (31,32), and coating the islet surface with endothelial cells (33) or endothelial progenitor cells (7). Third, our histological analysis of the recipient liver revealed small, elliptical BM-spheroids in more peripheral areas of the liver than the areas containing islet grafts. This morphological transformation of BM-spheroids might have contributed to the lack of hepatic infarction after infusion of the BM-spheroids into the portal sinusoids. In this context, it is also reassuring that cotransplantation of islets with a lower number of BM-spheroids (approximately one-third that of the cotransplanted islet cells) still produced the benefit.

This study had several limitations that should be discussed. Although portal thrombosis was evoked by MSCs in our experiments, the results of this study should not be interpreted as denial of the clinical feasibility of MSC-islet cotransplantation because various studies on the optimization of 3D culture methods and biomaterial-based modification are ongoing (34,35). In addition, we failed to prove the difference in C-peptide response after intravenous glucose challenge between the groups in NHP experiments. Although posttransplant body weight was stable in all the NHP recipients and all the recipient pancreata had minimal fraction of insulin-stained cells in this study, the findings of our NHP pilot study should be confirmed by further study with better control of possible model-induced confounding factors such as insufficient insulin secretion due to the marginal mass of islet graft, as well as malabsorptive caloric reduction due to the immunosuppressant-induced gastrointestinal toxicity (36,37).

In conclusion, cotransplantation of angiogenic monocyte subset spheroids derived from BM-MNCs enhances intraportal islet transplantation outcomes without portal thrombosis in mice and NHPs. In addition to their robust angiogenic capacity in vivo, 3D culture-formed BM-spheroids provided additional benefits in terms of localized persistence in portal sinusoids.

Funding. This research was supported by grants from the Samsung Biomedical Research Institute (GL1B30811), the National Research Foundation of Korea (NRF-2015R1A2A2A01003509, NRF-2015R1A2A1A1A5053895), and the Korea Health Technology R&D Project through the Korea Health Industry Development Institute, funded by the Ministry of Health and Welfare (A084065 and HI12C1853).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. B.J.O., S.-M.J., S.J.K., and J.H.K. conceived the work, designed the experiments, and wrote the manuscript. B.J.O. and J.M.C. performed intraportal transplantation using MSCs, MSC-spheroids, and BM-spheroids. H.-S.L., Ge.K., and H.J.P. conducted NHP experiments. Y.H. and P.K. performed IVM images and ex vivo bioluminescent imaging. B.J.O. and J.M.C. assisted with the microarray analysis and tissue staining. B.J.O., S.-M.J., and Gy.K. acquired data and analyzed data. S.J.K. and J.H.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.