Mesenchymal stromal cell mitochondrial transfer to human induced T-regulatory cells mediates FOXP3 stability

The key obstacle to clinical application of human inducible regulatory T cells (iTreg) as an adoptive cell therapy in autoimmune disorders is loss of FOXP3 expression in an inflammatory milieu. Here we report human iTreg co-cultured with bone marrow-derived mesenchymal stromal cells (MSCs) during short-term ex vivo expansion enhances the stability of iTreg FOXP3 expression and suppressive function in vitro and in vivo, and further that a key mechanism of action is MSC mitochondrial (mt) transfer via tunneling nanotubules (TNT). MSC mt transfer is driven by mitochondrial metabolic function (CD39/CD73 signaling) in proliferating iTreg and promotes iTreg expression of FOXP3 stabilizing factors BACH2 and SENP3. These results elucidate cellular and molecular mechanisms underlying human MSC mt transfer to proliferating cells. MSC mt transfer stabilizes FOXP3 expression in iTregs, thereby enhancing and sustaining their suppressive function in inflammatory conditions in vitro and in vivo.

Human and murine CD4 T cells can be efficiently differentiated ex vivo in the presence of TGF-β to generate inducible regulatory T cells (iTregs) with equivalent suppressive capacity as thymus-derived T regulatory cells (tTreg) 1,2 . iTregs have potentially enormous clinical application in diverse inflammatory conditions, including autoimmune disorders and graft versus host disease 3 . However, the instability of FOXP3 expression in iTregs, resulting in loss of suppressive function in inflammatory conditions, has to date limited wide clinical application 4 .
A number of studies have shown tTreg cells prevent adverse inflammation in multiple disease settings including autoimmunity and graft vs. host disease 5,6 . Although ex vivo protocols to expand tTregs have been developed, the limited number and anergic nature of tTregs, the risk of effector T cell contamination during expansion [7][8][9] , and the instability of tTreg phenotype and suppressor function upon transfer to an inflammatory environment 10,11 , remain significant obstacles to clinical efficacy.
Treg suppressive function is tightly controlled by Foxp3 co-transcriptional proteins 12,13 . BACH2, a Broadcomplex-tramtrack-Bric-a-Brac and Cap'n' collar homology 2 bZip transcription factor, has been shown to stabilize Treg Foxp3 expression in recent murine studies 14,15 . BACH2 has also been shown to regulate FOXP3 expression in human Umbilical Cord Blood (UCB) iTreg 16 . Further, murine studies identify SUMO specific protease 3 (SENP3) to be a critical regulator of Treg function by regulating the SUMOylation and nuclear localization of BACH2 17 .
Expansion of T cells in IL-2 based standard suspension cultures in vitro is associated with cell differentiation and senescence 18 . We observed previously, the addition of feeder layer MSCs provides improved cytokine driven expansion of early UCB CD34(+)/CD38(−)/HLA-DR(−) hematopoietic progenitors and inhibits differentiation and rates of apoptosis during short-term cytokine driven in vitro expansion 19 . Further, previous work has shown that MSC improve recovery of injured neuronal cells via mitochondrial transfer 20,21 . MSC have been identified to favor iTreg formation 22 , yet the potential of MSC to promote iTreg functional stabilization has not been explored.
MSC mt transfer requires direct cell-cell contact. Intercellular mt transfer by MSC has been previously described in neuronal injury and cancer models 36 . We measured total mt content in iTreg cells following MSC co-culture by flow cytometric analysis of iTregs labeled with the cardiolipin-binding dye nonyl acridine orange (NAO) 37 . MSC co-culture induced a ~ twofold increase in NAO mean fluorescence intensity in iTregs (Fig. 3a). To determine whether this resulted from increased endogenous iTreg mt biogenesis or MSC mt transfer, we labeled MSC mt with MitoTracker and enumerated MitoTracker positive iTreg in MSC co-cultures by direct visualization using epifluorescence microscopy. Intriguingly, robust mt transfer was observed (Fig. 3b), as the proportion of MitoTracker + iTregs reached ~ 16% within 16 h MSC co-culture ( Supplementary Fig. 3a). Expanded iTreg mt copy number was further increased in iTregs during short-term (21 day) expansion in MSC co-culture condition (Fig. 3c). Mitochondrial transfer was directly visualized via tunneling nanotubules (TNT) between MSCs and iTregs ( Fig. 3d and Supplementary Fig. 3b,c). To determine whether MSC mt transfer was reliant on TNT for transfer to proliferating iTreg during IL-2 driven expansion culture, Cytochalasin B (Cyto B), an inhibitor of TNT formation, was added and completely blocked the transfer of MitoTracker + mt from MSC to proliferating Tregs (Fig. 4a,b and Supplementary Fig. 3d). Of note, Cytochalasin B was utilized at a concentration (350 nM) known to inhibit TNT formation with minimal effects on cell viability in a range of cells including MSC, acute myeloid leukemia cells, and PC12 cells 38,39 . Nevertheless, one possibility is that reduced mt transfer was mediated by cyto B induced cytotoxicity in MSCs. Thus, we measured apoptosis by Annexin V (AV) staining. No differences in the proportion of AV + MSCs (Fig. 4c), or in AV + Tregs ( Supplementary Fig. 3e) between mock and cyto B treatment conditions was observed. Further, pre-treatment of MSCs, but not iTregs, with cyto B inhibited mt transfer, indicating MSC mt transfer occurs via TNT formation from MSCs ( Supplementary  Fig. 3f,g). We next tested the impact of MSC mt transfer on FOXP3 stability in iTregs. Cytochalasin B treatment significantly reduced FOXP3 expression in iTregs expanded on MSC (Fig. 4d).
CD39/CD73 signaling plays a critical role in MSC mt transfer and stabilization of iTreg suppressive function. Experiments were conducted to determine the mechanisms driving MSC mt transfer into proliferating iTregs. Treg express apyrases (CD39) and ecto-5'-nucleotidase (CD73) that contribute to inhibitory function by generating adenosine 41,42 . Additionally, CD73-generated adenosine induces cortical actin polymerization via adenosine A1 receptor (A1R) induction of a Rho GTPase CDC42-dependent conformational change of the actin-related proteins 2 and 3 (ARP2/3) actin polymerization complex member N-WASP 43 . This is noteworthy since we identified MSC mt transfer to UCB iTreg occurs via cytochalasin B-sensitive, actin-dependent transport via TNTs. To test whether CD73 contributes to MSC mt transfer, we added CD73 blocking Ab to MSC co-cultures and measured iTreg mt mass. mt mass was significantly reduced in proliferating iTregs after CD73 blocking Ab treatment (Fig. 5a). To confirm MSC transfer mitochondria to iTregs during ex vivo expansion, we generated MSC stably expressing mitochondrially targeted GFP. MSCs were transduced with mito-GFP lentivirus to generate stable mtGFP + MSCs ( Supplementary Fig. 6a) and co-cultured with iTregs expanded in IL-2/ media. Significantly less mtGFP + iTregs were detected during 21 day co-culture with mito-GFP lentivirus transduced MSCs when CD73 signaling was blocked or TNT formation was inhibited ( Fig. 5b and Supplementary . As CD39 and CD73 calibrate purinergic signals delivered to immune cells through the conversion of ADP/ATP to AMP and AMP to adenosine, respectively 44, 45 , we inhibited each pathway and observed reduced mtGFP + iTregs ( Fig. 5c and Supplementary Fig. 6d). As Miro1 has been shown to be a key regulator in mitochondrial intracellular transport 46 , we sought to determine if Miro1 played a role in MSC mt transfer to proliferating iTreg via TNT. We measured Miro1 expression in MSC after co-culture with iTregs. Co-cultivation with proliferating iTreg was associated with significantly enhanced expression of Miro1 in MSC, including both protein and RNA levels (Fig. 5d). Further, the enhanced MSC expression of Miro1 was inhibited by CD39 inhibition (Fig. 5e).
Since CD39/CD73 signaling play significant roles in mitochondrial transfer, we hypothesized these signaling pathways are critical to MSC-mediated enhancement of iTreg suppressive function. Inhibition of MSC mt into iTregs resulted in significantly reduced suppressive function vs. control (Fig. 5f, Supplementary Fig. 6d). These data indicate MSC enhance and stabilize iTreg FOXP3 expression and iTreg suppressive function via CD39/ CD73-induced mitochondrial transfer to proliferating iTregs during IL-2 driven expansion.
MSC mt transfer augments iTreg expression of FOXP3 stabilizing factors BACH2 and SENP3. BACH2 maintains stability and function of murine Treg 15,47 . We previously identified BACH2 is highly expressed in human UCB-derived iTreg and increases FOXP3 expression 16 . Additional studies have shown that SENP3 modulates BACH2 SUMOylation of and enhances iTreg stability in response to changing environmental conditions, particularly intracellular ROS 17 . We postulated mt transfer from MSC modulates BACH2/SENP3 expression and/or function, given our observations MSC mt transfer exerts a positive effect on iTreg Foxp3 stability and suppressive function 17,48,49 . We measured iTreg BACH2 and SENP3 protein expression, comparing MSC co-culture and standard IL-2/media suspension conditions during short-term (day 14 and 21) in vitro expansion. iTreg FOXP3, BACH2 and SENP3 expression was significantly increased in MSC co-culture conditions (Fig. 6a). We next examined iTreg BACH2 and SENP3 protein levels after cytochalasin B treatment to block mt transfer and observed loss of BACH2 protein expression (Fig. 6b). Cytochalasin B treatment had no www.nature.com/scientificreports/ significant effect on iTreg SENP3 protein expression, as SENP3 activation and/or function may be altered in this setting (Fig. 6b).
Previous studies have shown that mtROS induces BACH2 expression in B cells 49 . We measured mtROS expression in iTregs and found MSC co-culture induced higher expression of total and mtROS compared to standard IL-2/media suspension culture conditions (Fig. 6c,d). To investigate whether mtROS is a critical factor for induction of BACH2 and SENP3, we utilized a ROS inhibitor in iTreg/MSC co-cultures. Notably, iTreg expression of BACH2 and SENP3 in MSC co-cultures were significantly reduced in the presence of the ROS inhibitor (Supplementary Fig. 7). Overall, these data are consistent with MSC mt transfer enhancing iTreg BACH2 expression and subsequent enhanced FOXP3 expression and stability.
iTregs expanded on MSC feeder cells suppress effector CD4 and CD8 T cell responses in a xenogeneic GVHD NSG murine model. Inhibition of CD39 and CD73 signaling blocked MSC mt transfer to iTregs, and resulted in significantly reduced suppressive function vs. control (Fig. 5f). For assessment of the in vivo suppressive functions of iTreg expanded in MSC co-culture vs. iTreg expanded in standard IL-2/media suspension condition alone, iTregs were adoptively transferred into an NSG xenogeneic GVHD mouse model. iTregs expanded short-term (21 days) were injected 7 days after adult human PBMC were injected to induce GVHD. We observed enhanced survival and reduced lymphocyte infiltration into liver, skin, tongue and intestinal tissues in mice treated with iTregs from MSC co-cultures ( Supplementary Fig. 8). RNA levels of FOXP3 and FOXP3 + CD4 T cells in harvested spleen were significantly increased in mice treated with MSC co-cultured iTregs at 2 weeks post-GVHD induction (Fig. 7a,b). IFNγ producing CD8 + and CD4 + T cells were dramatically reduced in spleen cells harvested from mice treated with MSC co-cultured iTregs (Fig. 7d). Consistent with the cytokine staining results, serum and ex vivo levels of pro-inflammatory cytokines IFNγ and TNF were significantly reduced in animals treated with MSC co-cultured iTregs ( Supplementary Fig. 9a,b). www.nature.com/scientificreports/ To explore the effect of MSC mt transfer on iTreg phenotype and function in vivo, we co-cultured iTregs with MSC in the presence of a CD39 inhibitor 50,51 . CD39 inhibitor treatment reduced the percentage of FOXP3 + iTreg in PBL induced GVHD NSG mice (Fig. 7c). IFNγ producing CD8 + and CD4 + T cells were significantly increased in mice injected with CD39 inhibitor-treated iTreg (Fig. 7d). IFNγ and TNF were also significantly increased in the serum of these mice (Supplementary Fig. 9c) and ex vivo (Supplementary Fig. 9d). Collectively, these results reveal suppressive iTregs are maintained by MSC mt transfer and can suppress effector T cell immune responses against human inflammatory xeno-immune disease in vivo.

Discussion
Intrinsic defects in Tregs as observed in autoimmune disease may hamper the success of autologous tTreg adoptive cell therapies [52][53][54] . Previous studies have shown ex vivo-expanded, partially HLA-matched tTregs from allogeneic UCB sourcing are well tolerated in humans 11 . UCB-derived iTregs exhibit enhanced stability and www.nature.com/scientificreports/ suppressive function compared to adult blood-derived iTregs in large part due to enhanced expression of the FOXP3-stabilizing transcription factor BACH2 16 . Nevertheless, a major obstacle to the clinical implementation of iTregs is concern over the instability of iTreg FOXP3 expression and loss of suppressor function upon transfer to an inflammatory environment 55 . The results reported in this study demonstrate that MSC co-culture provides enhanced and sustained FOXP3 expression and iTreg suppressive function in inflammatory in vitro and in vivo environments. Importantly, our study provides insight into normal physiologic mechanisms to sustain human proliferating T-cells including mt transfer from adjacent MSC. Co-culture of iTregs with MSC feeder cells significantly improved the number and suppressive function of iTreg during 21 day IL-2 based ex vivo expansion. MSC co-culture increased iTreg expression of CD25, CTLA-4, and ICOS, while expression of LAG-3 and TIM-3, associated with exhaustion of activated T cells [56][57][58][59][60][61][62] , was decreased. We observed higher proportions of CD62L + and absolute numbers of CD45RA + iTreg cells after 21 day IL-2 expansion co-culture on MSC feeder cells. Inflammatory environments exert rapid loss of FOXP3 expression and reduced suppressive activity of iTregs 63 . Our data show significantly enhanced suppressive function, including inhibition of effector cell activation and maintained FOXP3 stability in MSC co-culture expanded iTregs exposed to inflammatory conditions in vitro and in vivo. It will be important in the future to examine whether epigenetic modification of the Treg-specific demethylated region in the FOXP3 gene occurs, in order to gain further insights into the mechanisms of MSC-enhanced iTreg stability 64 .
Although prior studies point to MSC enhancement of Treg function to be primarily via paracrine mechanisms [65][66][67] , our study shows MSC support of iTreg FOXP3 expression and sustained suppressive function requires cell-cell contact mt transfer. Recent work showed umbilical cord-derived MSC mt can promote adult PBMC CD4 T cell differentiation into iTreg 68 . Intriguingly, purified MSC mt could recapitulate the effects of MSC mt transfer to promote Treg differentiation. However, this study did not examine stabilization of iTreg FOXP3 expression and maintenance of naïve phenotype mediated by MSC mt transfer during ex vivo expansion. Our findings suggest direct cell-cell contact and mt transfer via TNTs play an important role in stabilization of iTreg FOXP3 expression, maintenance of naïve phenotype, and maintained iTreg suppressive function in adverse inflammatory conditions. MSC co-culture was also observed to support maintenance of tTreg phenotype. Notably however, the proportion of FOXP3 + tTregs was less sustained compared to iTregs during expansion (57% vs 96%, respectively at Day 21 in MSC co-culture) potentially reflecting the presence of contaminating effector cells and the more stable phenotype of iTregs generally.
Several mechanisms have been proposed to explain how mt may promote Treg formation. For instance, it was proposed retrograde mitochondria to nuclear signaling alters the expression of nuclear genes involved in Treg phenotype via epigenetic mechanisms 68 . However, prior to our studies, the cellular and molecular mechanisms driving MSC mt transfer to proliferating cells and signaling pathways leading to altered phenotype of recipient cells have not been elucidated. Our studies identify iTreg CD39/CD73 signaling drives MSC mt transfer and further mt transfer results in augmented iTreg BACH2 and SENP3 expression. BACH2 regulates human UCB iTreg development via direct transcriptional activity at the FOXP3 promoter, and by suppressing TCR induced gene transcription 16,69 . SENP3 is a SUMO specific protease known to maintain Treg stability 17 . SENP3 activity can be regulated by a variety of cell stressors including but not limited to ROS 70 . In addition, SENP3 has been linked to mitochondrial signaling mechanisms and function. SENP3 can deSUMOylate and modulate the localization and function of Drp1, a key regulator of mitochondrial fission in neurons. Prior studies have also identified a role for mitochondrial regulation of BACH2 function and cell fate decisions in B cells 49 . Thus, our findings suggest a potential mechanism whereby mitochondrial transfer stabilizes iTreg suppressor function via altered expression of SENP3 and BACH2.
CD39 and CD73 together play strategic roles in immune responses 44,45 . CD39 and CD73 degrade extracellular ATP to yield AMP and anti-inflammatory adenosine 71 . CD73−/− mice show enhanced antitumor immunity 72 and more severe gastritis compared to functional CD73 controls, and adoptive injection of WT Tregs reversed these immune responses 42 . Prior studies 73,74 identify CD73 signaling on Tregs is critical to maintain Treg suppressive function. Our study identifies MSC mt transfer to iTregs contributes to enhanced iTreg phenotypic stabilization. In addition, the signaling mechanisms involved in mt transfer to Tregs have not been previously examined. We identify CD39 and CD73 signaling mediates MSC mitochondrial transfer to proliferating iTregs during IL-2 driven ex vivo expansion. Further studies on CD39 and CD73 mediated mt transfer may elucidate mechanisms that may be exploited for Treg functional stabilization.
MSCs mt transfer has been shown in several cell types, including proliferating acute leukemia blasts in the marrow microenvironment 39 . Molecular pathways of mitochondrial transfer have been identified, including the mitochondrial Rho-GTPase 1 Miro1 40 . Miro1 regulates intercellular mt transfer by attaching mitochondria to KLF5 the kinesin motor protein 75,76 and has been implicated in enhanced injured cell recovery 40 . In one study, injured astrocytes induced increased levels of Miro1 expression in MSC and this was correlated with mt transfer 20 . However, MSC mt transfer to Treg cells and the cellular and molecular mechanisms underlying mt transfer has not been previously identified. Our data show mitochondrial metabolic function through CD39/ CD73 signaling, in proliferating iTregs induced Miro1 expression in MSC during co-culture ex vivo. MSC mt transfer occurs via TNT rather than exosome mediated transfer. These findings provide impetus to study the cell signaling mechanisms in proliferating iTreg serving to induce MSC Miro1 expression and facilitate TNT formation for mt transfer.
Mitochondria are an important source of ROS 77 and regulate cell cycle, modulate signal transduction 78,79 , and are critical for cancer cell tumorigenicity 80 . This is particularly relevant considering ROS is a major driving force for mt transfer via TNTs from bone marrow stromal cells to leukemic blasts 39  www.nature.com/scientificreports/ not a critical mechanism driving this process. Nonetheless, cytochalasin B treatment led to diminished number of FOXP3 + iTregs, suggesting TNT transfer of mt from MSC play an important role in maintenance of FOXP3 + iTreg stability. Regardless of cell type, mt are now recognized to play roles extending well beyond the production of energy. T cells undergoing changes in cellular metabolism during activation is a central idea emerging from many recent studies 81 . In contrast to quiescent T cells, with limited metabolic needs mainly encompassing cellular trafficking and housekeeping functions, actively proliferating cells must generate additional ATP for functions including the generation of intermediates required in various biosynthetic pathways and signaling molecules for anabolic metabolism. Future work is needed to identify how signaling mechanisms can be modulated to modulate mt transfer. Importantly, several lines of evidence support mt transfer is a key mechanism of MSC supportive function. Our studies identify iTreg mt membrane potential is enhanced by culture over MSC feeder cells, and further pharmacologic inhibition of mt transfer nearly completely negates the benefit of MSCs on iTreg FOXP3 expression and suppressive function.
In summary, our findings highlight the importance of mt function in maintaining iTreg stability in MSC coculture expanded iTreg and provide insight into cellular and molecular mechanisms driving MSC mt transfer to proliferating iTreg. This process inhibits differentiation, cell senescence, maintains robust FOXP3 expression, as well as suppressive function despite an adverse inflammatory milieu in vitro and in vivo. Although several factors modulate iTreg stability and function in disease models, our results show the mt BACH2 pathway is a prominent mechanism of MSC-enhanced iTreg stability and function. Additionally our work highlights the important role played by CD73-regulated TNT formation that serves to facilitate mt transfer into proliferating iTreg during expansion.

Methods
Mice. NSG mice were ordered from Case Western Reserve University (CWRU) a thymic facility. Mice were maintained in the SPF animal facility at CWRU. Mice were ear-punched, weighed, placed in 12-pie irradiator holder and exposed to 240 rads from Gammacel 137 Cs source.
Human UCB and adult peripheral blood samples. Acquisition of de-identified volunteer young adult peripheral blood and UCB were obtained with written informed consent per Case Western Reserve University/Cleveland Cord Blood Center IRB-approved protocols. All methods in these studies were carried out in accordance with relevant guidelines and regulations. Mononuclear cells were isolated by Ficoll-Paque PLUS (GE Healthcare Life Sciences, Piscataway, NJ) density gradient centrifugation with SepMate-50 tubes (STEMCELL Technologies, Vancouver, BC, Canada). FOXP3 + iTreg generation and ex vivo expansion. UCB derived FOXP3 + iTregs cells were generated as previous described 82 . Ex vivo expansion of FOXP3 + iTregs was performed with CD25 + cells MACS purified from day 4 TGFβ-induced UCB iTreg. On Day 4 of TGF-β conditioning, CD25 + iTregs were MACS purified, rested for 48 h, and on Day 6, 5 × 10 5 purified cells were split into expansion cultures with either media containing 100 U/ml IL-2 in suspension or identical media/IL-2 over MSC feeder cells. Cultures were fed every other day with fresh media/IL-2 and MSC during short-term culture (21-28 days). No re-stimulation was performed during the 21-28 day expansion following the initial 4 day stimulation consistent with a previously described UCB Treg expansion protocol 83 . For isolation of UCB-derived tTregs, EasySep Human CD4 + CD127lowCD25 + Regulatory T Cell Isolation Kit (Stem Cell Technologies) was used. tTregs were maintained under identical conditions as iTregs. Absolute number of FOXP3 + iTreg cells were calculated from the percentage of CD4 + cells based on the total viable cell count obtained by trypan blue dye exclusion at each indicated time point. For iTreg stability assays, harvested iTregs were stimulated with CD3 and CD28 antibodies (1 μg ml −1 ) with IFNγ (10 μg ml −1 ), IL-6 (10 μg ml −1 ), and TNF (10 μg ml −1 ) at concentration of 5 × 10 5 cells ml −1 for 72 h and analyzed. Suppression assay. In vitro suppression assays were performed as previously described 82  Western blot analysis. Collected cells were lysed using radio-immune precipitation assay (RIPA) buffer and protein was quantified using the Bradford assay (BioRad, Hercules, CA), with 40 μg of protein separated by 4-20% SDS-PAGE (BioRad). Proteins were electro-transferred onto Immobilon membrane (Millipore) and were probed with polyclonal rabbit anti-BACH2 (D3T3G; Cell signaling technology, Beverly, MA) anti-SENP3 (D20A10; Cell signaling technology) antibody, monoclonal rabbit anti-FOXP3 (D25D4; Cell signaling technology), and monoclonal mouse/human/rat anti-β-actin antibody (R&D systems, Minneapolis, MN). Applicable HRP-conjugated secondary antibodies were used with ECL chemi-luminescence to observe relevant bands using ChemiDoc XRS + image system (BioRad). The band intensity was quantified using Image J.
Cytoplasmic and nuclear protein isolation. Cytoplasmic and nuclear fractions were isolated from ~ 10-20 × 10 6 Treg cells from IL-2/media only cultures or Tregs from IL-2/media MSC co-cultures using the NE-PER Nuclear and Cytoplasmic extraction reagents kit (Invitrogen 78833) according to the manufacturer's instructions. In brief, Treg cells were removed from co-culture, washed with PBS, and ice cold CER1 reagent was added, and samples were pelleted by centrifugation. Samples were vortexed for 15 s and incubated on ice for 10 min. Ice cold CERII reagent was added and samples were vortexed for an additional 5 s and incubated on ice for 1 min. After additional 5 s vortex, samples were centrifuged for 5 min at maximum speed. The supernatant containing the cytoplasmic extract was transferred to a new tube. The insoluble pellet fraction containing the nuclei was suspended in ice-cold NER and vortexed for 15 s every 10 min for 40 min. Nuclear fraction was centrifuged at maximum speed for 10 min and supernatant containing the nuclear extract was transferred to a new tube. Isolated fractions were kept on ice. Protein concentration was determined by Bradford assay and equivalent amount of total protein for each sample (5-10 μg nuclear fraction, 20-40 μg cytoplasmic fraction) was loaded onto a 4-20% SDS-PAGE gel and processed for analysis of BACH2 and SENP3 expression by western blotting as described above.
ELISA assay. Levels of IFNγ and TNF were detected from serum and culture supernatant. IFNγ and TNF were determined by ELISA (R&D systems), according to the manufacturer's instructions.

GVHD model and tissue histology.
For establishment of an acute GVHD model, NSG mice received cord blood CD34 + cells. 5-6 weeks later, mice were injected with adult PBL (1 × 10e7). GVHD severity was assessed by an established scoring system based on weight loss, posture, activity, fur texture and skin texture 87 . Pathologic evaluation was performed for organs targeted by acute GVHD, including liver, skin, tongue and intestine 88 . Tissues were collected and hematoxylin/eosin (H&E) stained at day 21 after injection of PBL. Intestines were washed thoroughly prior to sample collection. All samples were fixed using 10% neutral-buffered formalin for 24 h and embedded in paraffin. Paraffin sections (1 μm) were cut, dewaxed and stained with H&E 89 .
Statistical analysis. Statistical comparative analyses were performed using the Paired t-test (Prism 7 software-GraphPad, San Diego, CA). Multiple comparisons analysis was performed using the Friedman test. Data are presented as the mean ± standard deviation, SD. A p value of < 0.05 was considered significant. *p < 0.05, **p < 0.01, ***p < 0.001.

Study approval.
This study was designed to determine whether MSCs can enhance the stability and suppressive capacity of iTregs during ex vivo expansion. MSC were obtained from healthy donor derived bone marrow and umbilical cord blood from at least 3 different donors to confirm our observation of MSC mitochondrial transfer mediated stabilization of iTregs. The study was carried out in accordance with the approved guidelines and regulations. De-identified human cells tested in this study were approved by Cleveland Clinic Foundation for volunteer umbilical cord blood collection with written informed consent from donating mothers. In addition, the collection of normal adult peripheral blood from de-identified human donors was approved by the Case Western Reserve University IRB for the Case Comprehensive Cancer Center Hematopoietic Stem Cell Facility (CASE 12Z05). For GVHD studies, iTregs expanded in either media/IL-2 suspension culture condition or media/IL-2 over MSC feeder cells, were harvested and injections were performed by a second investigator (DA) who was blinded to the identity of each sample. All animal experiments and methods were performed in accordance with the relevant guidelines and regulations approved by the Institutional Animal Care and Use Committee of Case Western Reserve University. In addition, mice studies were carried out in compliance with the ARRIVE guidelines. Mice were randomized on the basis of their age, sex and weight. Numbers of mice were chosen based on magnitude of observed effects and variability in the results and are indicated in figure legends.

Data availability
All data generated or analyzed during this study are included in this published article and its Supplementary Information files.