Presentation of Human Neural Stem Cell Antigens Drives Regulatory T Cell Induction

Multiple sclerosis (MS) is a debilitating autoimmune disease caused by autoreactive T cells that results in progressive damage to the CNS (1). Current Food and Drug Administration–approved therapies for the relapsing-remitting form of MS function by suppressing autoreactive immune cells or by limiting access to the CNS, but because they do not promote either substantial remyelination or preservation of axons, these treatments do not provide long-term relief and are not effective for progressive forms of MS that lack spontaneous remyelination. An important unmet clinical need for MS patients is an effective method for promoting remyelination that can ameliorate clinical symptoms associated with demyelination and restore motor function while limiting immune cell infiltration into the CNS (2, 3).

Neural stem cell (NSC) transplantation has long been considered a potential therapeutic approach to repairing and replacing damaged cells in the CNS (4, 5). In murine models of MS, syngeneic mouse NSC transplants into the spinal cord have been shown to effectively engraft, differentiate, and repair and replace damaged neurons and oligodendrocytes (6–10). Due to the autoimmune nature of these models, it is, however, difficult to assess the capacity of human NSCs (hNSCs) to repair damage in the context of a xenogeneic transplant because the murine immune system causes transplanted hNSCs to be rapidly rejected (6, 11). Interestingly, despite this rejection, hNSC transplantation leads to a reduction in the number of activated inflammatory CD4⁺ T conventional (Tconv) cells in the CNS and periphery, as well as a large, localized increase in CD4⁺CD25⁺Foxp3⁺ T regulatory cells (Tregs) at the site of hNSC transplantation in the spinal cord (6, 11–14). Of note, this Treg influx also results in localized remyelination and repair in the CNS, and depletion of such Tregs abrogates the repair induced by hNSC transplantation (6, 11). Therefore, it is critical to understand how hNSC transplantation drives this localized increase in Tregs to gain insight into this mechanism and potentially employ this process as a future therapeutic modality.

Tregs are the primary suppressive cells of the adaptive immune system and function to maintain immune homeostasis. They are characterized by their expression of the master transcription factor Foxp3 and the high-affinity IL-2Rα-chain, CD25 (15, 16). Tregs are known for their capacity to actively suppress immune responses against self-antigens and their absence results in massive systemic autoimmunity (17, 18). Current models of autoimmunity hypothesize that Treg dysfunction is one of the root causes of autoimmune disease, including MS and type I diabetes (19–22). Recent work has also shown that Tregs are capable of directly interacting with a variety of distinct cell types and tissues to instigate repair as well as serving homeostatic functions (13, 16, 23–31).

This knowledge paired with our findings of a localized Treg response after hNSC transplantation has led us to explore the possibility that these Tregs are responsible for both the suppression of Tconv cells in the CNS as wells as their ability to promote tissue repair within the CNS in the form of remyelination (6, 11, 12). In this study we investigated the mechanisms by which hNSCs generate a Treg response. Our data provide evidence that hNSCs indirectly promote Treg induction through increasing the availability of CNS-specific Ags, resulting in the generation of peripheral Tregs (pTregs) from the Tconv cell population.

All experiments were approved by the University of California, Irvine Institutional Animal Care and Use Committee. C57BL/6 mice (strain 027) were obtained from Charles River Laboratories. Foxp3-GFP⁺ mice (B6.Cg-Foxp3^tm2Tch/J, stock no. 006772) were obtained from The Jackson Laboratory (32, 33). RAG2^−/−2D2⁺ mice were generated by breeding RAG2^−/− (B6.129S6-Rag2^tm1FwaN12; Taconic, model no. RAGN12) (34) and 2D2 TCR-transgenic (tg) mice recognizing myelin oligodendrocyte glycoprotein (MOG)_35–55 (C57BL/6-Tg(Tcra2D2,Tcrb2D2)1Kuch/J; The Jackson Laboratory, stock no. 006912) (35). RAG2^−/−OT-II⁺ TCR-tg mice recognizing OVA_323–339 were obtained from Taconic (B6.129S6-Rag2^tm1FwaTg(TcreTcrb)425Cbn, model no. 11490) (34, 36). Splenocytes from RAG1^−/−OT-II⁺ TCR-tg mice recognizing OVA_323–339 were provided by Stephen Schoenberger at the La Jolla Institute for Allergy and Immunology.

RAG2^−/−2D2⁺ mice were immunized by s.c. injections with 100 μl of emulsion containing 100 μg of MOG_35–55 (MEVGWYRSPFSRVVHLYRNGK-COOH; Pierce), neurofilament medium (NFM)_18–30 (TETRSSFSRVSGS; GenScript), or OVA_323–339 (ISQAVHAAHAEINEAGR; Neta Scientific) in PBS, with CFA containing 200 μg of Mycobacterium tuberculosis H37Ra (Difco Laboratories). Mice received i.p. injections with 200 ng of Bordetella pertussis toxin (List Biological Laboratories) day 0 and day 2 postimmunization. Mice were sacrificed via inhalation of a lethal dose of isoflurane, and cardiac perfusion with PBS was performed at defined time points postimmunization for tissue harvesting and analysis.

Cervical lymph nodes, spleens, brains, or spinal cords were dissociated into single-cell suspension, depleted of RBCs using Tris-acetic acid-chloride (TAC) as previously described (37). Cells from brain and spinal cord samples were further purified using a 23% Percoll gradient (38). Cells were filtered, washed, and counted before being stained with a Zombie Violet or Zombie Aqua fixable viability kit (BioLegend), blocked with CD16/32 (2.4G2), and stained with surface Abs against CD3 (145-2C11), CD4 (RM4-5), and CD25 (PC61.5). Intracellular staining for Foxp3 (FJK-16S) was performed using the eBioscience Foxp3/transcription factor staining buffer set (Thermo Fisher Scientific). All Abs were purchased from BD Biosciences, BioLegend, or eBioscience. Single-stain samples and fluorescence minus one controls were used to establish photomultiplier tube voltages, gating, and compensation parameters. Cells were processed using a BD LSR II or BD LSRFortessa flow cytometer and analyzed using FlowJo software (Tree Star).

Human embryonic stem cells were derived from WA09 human embryonic stem cells and differentiated into embryonic body NSCs and sorted according to established methods as previously described (39, 40). hNSCs were maintained in hNSC maintenance medium (DMEM/F12⁺ GlutaMAX, 0.5× N2, 0.5× B27 without vitamin A, 20 ng/ml basic fibroblast growth factor; all from Thermo Fisher Scientific) on Geltrex-coated dishes using Accutase to split cells when the cell density reached 80–90% confluence.

Spleens dissected from naive, age-matched (6- to 8-wk-old) RAG2^−/−OT-II⁺, or RAG2^−/−2D2⁺ mice were isolated into single-cell suspensions, after which 1.5 × 10⁵ splenocytes per well were cultured in round-bottom 96-well plates and incubated at 37°C, 5% CO₂ for 4 d in 200 μl final volume of complete RPMI 1640 medium (RPMI 1640, 10% FBS [Atlanta Biologicals], 1× nonessential amino acids, 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mM sodium pyruvate, 55 μM 2-ME; all from Thermo Fisher Scientific). Splenocytes were cultured alone; activated with plate-bound anti-Armenian hamster IgG (30 μg/ml; Vector Laboratories or Jackson ImmunoResearch Laboratories), Armenian hamster, anti-mouse CD3 (1 μg/ml, 2C-11, BioLegend), and soluble anti-mouse CD28 (1 μg/ml, BioLegend or Tonbo); with 20 μg/ml MOG_35–55 peptide (Pierce); with 20 μg/ml NFM_18–30 peptide (GenScript) with 20 μg/ml OVA peptide (Neta Scientific); or with 3 μg/ml high–molecular mass polyinosinic-polycytidylic acid (poly(I:C); InvivoGen) (41). After 4 d, cells were filtered and washed before being stained with a Zombie Aqua fixable viability kit (BioLegend), blocked with CD16/32 (2.4G2) and surface Abs against CD4 (RM4-5) and CD25 (PC61.5). Intracellular staining for Foxp3 (FJK-16S) was performed using the eBioscience Foxp3/transcription factor staining buffer set (Thermo Fisher Scientific). All Abs were purchased from BD Biosciences, BioLegend, or eBioscience. Single-stain samples and fluorescence minus one controls were used to establish photomultiplier tube voltages, gating, and compensation parameters. Cells were processed using a BD LSR II or BD LSRFortessa flow cytometer and analyzed using FlowJo software (Tree Star).

Spleens dissected from naive, age-matched (6- to 8-wk-old) C57BL/6, Foxp3-GFP⁺, RAG1^−/−OT-II⁺, or RAG2^−/−2D2⁺ mice were isolated into a single-cell suspension depleted of RBCs, as previously described (37). Splenocytes were combined with hNSCs at defined proportions (Supplemental Fig. 4) with 1.5 × 10⁵ cells total per well in round-bottom 96-well plates and incubated at 37°C, 5% CO₂ for 4 d in 200 μl final volume of complete RPMI 1640 medium. Splenocytes were activated with plate-bound anti-Armenian hamster IgG (30 μg/ml; Vector Laboratories or Jackson ImmunoResearch Laboratories), Armenian hamster, anti-mouse CD3 (1 μg/ml, 2C-11, BioLegend), and soluble anti-mouse CD28 (1 μg/ml, BioLegend or Tonbo). Addition of recombinant human IL-2 (100 U/ml) and recombinant human TGF-β (7 ng/ml) were added to cultures where indicated. After 4 d, cells were filtered and washed before being stained for flow cytometry analysis as previously mentioned. For quantification, cell counts were normalized to a starting population of 150,000 splenocytes.

For induced Treg (iTreg) and hNSC-Treg sorts, splenocytes from Foxp3-GFP⁺ mice were cultured with or without hNSCs as described above. After 4 d, cells were filtered, washed, and subjected to negative T cell isolation using magnetic separation and then sorted using FACS. Cells were counted and resuspended at 1 × 10⁸/ml in T cell isolation buffer (1× PBS without Ca²⁺ and Mg²⁺-containing 2% FBS and 1 mM EDTA) and 50 μl/ml normal rat serum (STEMCELL Technologies) with 10 μl/ml anti-mouse B220 biotin and 10 μl/ml anti-mouse CD11b biotin (BioLegend) for 10 min at room temperature (RT). MojoSort streptavidin nanobeads (50 μl/ml; BioLegend) were added and incubated for 2.5 min at RT. T cell isolation buffer was added up to 5 ml and samples were placed in an EasySep magnet (STEMCELL Technologies) for 2.5 min at RT, after which supernatants were collected. Cells were stained with propidium iodine viability dye, then blocked with CD16/32 (2.4G2) and surface Abs against CD4 (RM4-5) and CD25 (PC61.5). All Abs were from BioLegend or BD Biosciences. Cells were sorted using a BD FACSAria II into TRIzol LS (Thermo Fisher Scientific) and stored at −80°C until RNA extraction. For cocultures utilizing depleted splenocytes, splenocytes from Foxp3-GFP⁺ mice were sorted using a BD FACSAria II, then cultured with hNSCs.

iTregs or hNSC-Tregs were FACS purified as described above. RNA was extracted from cells by adding 140 μl of TET (10 mM Tris 8.0/0.01 mM EDTA/0.05% Tween 20) and then 140 μl of chloroform/isoamyl alcohol at 24:1 (Sigma-Aldrich). Samples were then centrifuged at 15,000 × g for 10 min at 4°C. The aqueous phase was collected and added to 1.5 μl of glycol blue (Thermo Fisher Scientific) and 10% vol 3 M sodium acetate and 1 vol isopropanol. Samples were mixed by inverting and stored at −20°C overnight. Samples were then spun at 15,000 × g for 30 min at 4°C. Supernatants were removed from pellets, and 500 μl of 75% ethanol was added to the pellet. Samples were then spun at 15,000 × g for 30 min at 4°C, and supernatants were removed from pellets. RNA pellets were resuspended in 30 μl of H₂O and used immediately or aliquoted and stored at −80°C. RNA concentrations were quantified using a Qubit fluorometer (Thermo Fisher Scientific), and RNA quality of each sample was determined using the 2100 Bioanalyzer (Agilent Technologies) to obtain an RNA integrity number. One hundred nanograms of RNA with acceptable RNA integrity numbers (9.0–10.0) was sent to iRepertoire for CDR3ε sequencing and analysis.

A SuperScript III reverse transcriptase kit (Thermo Fisher Scientific) was used to generate cDNA from RNA collected from samples. Quantitative RT-PCR was performed on the cDNAs using TaqMan gene expression master mix (Thermo Fisher Scientific) to quantify transcript levels using TaqMan expression probes Mog and Nfm (Thermo Fisher Scientific) in an ABI ViiA7 thermocycler (Applied Biosystems). Human brain tissue samples were obtained from the University of California, Irvine Alzheimer’s Disease Research Center and used as controls for Mog and Nfm expression. Data were analyzed using the comparative Ct method (42).

Data were analyzed using Prism software (GraphPad Software). Comparisons were performed using a two-tailed t test and two-way ANOVA, where indicated. For all statistical models and tests described above, the significance is displayed as follows: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

We have previously reported that transplantation of hNSCs into the spinal cords of mice with immune-mediated demyelination results in an increase in CD4⁺CD25⁺Foxp3⁺ Tregs (11, 12). To investigate how hNSCs contributed to the observed increase in Tregs in vivo, we developed an in vitro coculture system in which naive mouse splenocytes are mixed with different ratios of H9 human embryonic stem cell–derived hNSCs (Fig. 1A, Supplemental Fig. 1A, 1B). Coculture of splenocytes with hNSCs resulted in a small but significant increase in the frequency of Tregs at a 1:5 ratio compared with splenocytes alone (Supplemental Fig. 1C). In subsequent experiments, naive splenocytes were cultured in conditions that support T cell activation and survival, with splenocytes stimulated with anti-CD3 and anti-CD28, as well as TGF-β and IL-2 in the presence or absence of hNSCs. TGF-β and IL-2 were added to cocultures to better mimic physiologic conditions of Ag presentation in the CNS, as microglia secrete TGF-β and IL-2 (43), factors that are essential for Treg generation and function (44, 45). Splenocytes cultured with hNSCs yielded 2-fold higher proportions of Tregs (hNSC-Tregs) when compared with iTreg controls generated by stimulation under Treg-skewing conditions alone (Fig. 1B–D). Additionally, Tregs cultured with hNSCs displayed 1.9- and 1.2-fold higher expression of Foxp3 (Fig. 1E) and CD25 (Fig. 1F), respectively, as measured by mean fluorescence intensity when compared with iTregs. hNSC-Treg generation was consistent among various ratios of C57BL/6 splenocytes to hNSCs (Supplemental Fig. 1C, 1D) and specific to hNSCs; cocultures with xenogeneic human dermal fibroblasts (Supplemental Fig. 1E, 1F) did not yield an expanded fraction of CD25⁺Foxp3⁺ Tregs. These data suggest that hNSCs directly contribute to the expansion of Tregs in vitro.

Foxp3⁺ Tregs can be derived from the thymus (tTregs) during thymic selection or in pTregs through the conversion of Tconv cells into Tregs (31, 43). To determine whether hNSCs expanded Tregs from the tTreg pool or by conversion of Tconv cells into pTregs, unsorted splenocytes and splenocytes depleted of Foxp3-GFP⁺ cells from Foxp3-GFP reporter mice were cocultured with hNSCs (Fig. 2A). Increases in the frequency (Fig. 2B) and number (Fig. 2C) of CD25⁺Foxp3⁺ hNSC Tregs were observed regardless of the presence of Foxp3⁺ cells in the starting splenocyte population, demonstrating that hNSCs are actively involved in converting Tconv cells into CD25⁺Foxp3⁺ Tregs.

To further investigate the differences between hNSC-Tregs and iTregs, we compared their TCRβ sequences using high-throughput TCRβ repertoire analyses. These analyses identified 7427 unique CDR3ε sequences for iTregs and hNSC-Tregs. To determine the clonal dominance in each population, we used the clonality index (inverse of Shannon’s entropy) (46), with 0 indicating that each clone only occurs once and 1 being a monoclonal population. Although both populations of Tregs were polyclonal, with a diversity index of <0.2, only 57 TCRβ sequences were common to both Treg populations whereas 5549 TCRβ sequences were unique to hNSC-Tregs (Fig. 3A). Additionally, using the diversity index where 0 is a monoclonal population and 50 is a diverse population, hNSC-Tregs had a lower diversity score (5.3–9) compared with iTregs (7.5–15.5). Although both populations of Tregs were diverse, clonal dominance can be estimated by the contribution of the top 10 most abundant clones. Many of the top 10 hNSC Treg clones were distinct when compared with the iTreg repertoire. In addition, hNSC-Treg clones displayed similar peptide sequences (Supplemental Fig. 2A, 2B). Vβ13 was one of the most predominant clones in the iTreg sample, which encodes the protein Vβ8.1 and has been implicated in conferring protection in a viral murine model of demyelination (47, 48). Vβ31, encoding for the protein Vβ14, was one of the most predominant clones in hNSC-Tregs. Vβ14 is known to recognize part of MOG, but it is not the predominant myelin-reactive clone Vβ11 (49). The gene Vβ16 encodes for Vβ11 and was absent in the top 10 clones of both iTregs and hNSC-Tregs. Furthermore, when comparing CDR3ε sequences of iTregs and hNSC-Tregs to human dermal fibroblast-induced Tregs, the latter were the least diverse (diversity index = 3) with only 1127 unique clones (data not shown). These results suggest that although there is high diversity within the TCR sequence of the hNSC-Treg group, many of these clones bear a striking similarity to one another and may recognize different epitopes of the same Ag. Taken together, these TCR repertoire studies provide evidence that hNSCs elicit a unique TCR repertoire within the expanded Treg population in these cocultures, supporting the concept that the response requires Ag presentation.

To determine whether the mechanism through which hNSCs promote Treg expansion occurs via passive Ag expression or via active expression of factors such as cytokines that promote Treg differentiation, we tested the ability of killed hNSCs to elicit Treg expansion. Because hNSCs do not survive longer than 8 d posttransplantation (12), we hypothesized that hNSCs were not required to be viable for the observed increase in Tregs, and that their Ags alone may be sufficient for the observed increase in Tregs. To test this hypothesis, we characterized cocultures of splenocytes mixed with hNSCs cultured under normal conditions (live) or hNSCs cultured overnight in PBS to induce cell death (PBS killed). Both live and PBS-killed hNSCs generated more Tregs in cocultures compared with iTreg controls (Fig. 3B, 3C). Although there was not a detectible difference in the frequency of CD25⁺Foxp3⁺ cells with live or PBS-killed hNSCs compared with one another (Fig. 3B), there was an increase in the number of CD25⁺Foxp3⁺ cells in conditions with PBS-killed hNSCs compared with live cells (Fig. 3C). These data suggest that direct T cell–hNSC interactions or hNSC-secreted cytokines are not responsible for the observed increase in Tregs. That hNSCs did not need to be alive to expand Tregs led us to investigate whether TCR antigenic recognition is responsible for this Treg expansion.

To confirm that hNSC Ags drive the observed increases in hNSC-Tregs, we characterized cocultures employing splenocytes from two different TCR-tg mouse lines that bear restricted, monoclonal TCR repertoires. We used RAG2^−/−2D2⁺ mice, which have T cells that recognize MOG and NFM self-antigens, and RAG1^−/−OT-II⁺ mice with T cells that recognize the non–self-antigen chicken OVA. If hNSCs Ags are necessary for Treg expansion, we would anticipate an increase in Tregs in RAG2^−/−2D2⁺ splenocyte cultures, but not in RAG1^−/−OT-II⁺ splenocyte cultures because hNSCs express NFM (Supplemental Fig. 2C, 2D). Supporting the hypothesis that hNSC Ag recognition drives an increase in Tregs, 4-d cocultures of RAG2^−/−2D2⁺ splenocytes and hNSCs had a significant, dose-dependent expansion of CD25⁺Foxp3⁺ Tregs (Fig. 4A, 4B). In contrast, cocultures of RAG1^−/−OT-II⁺ splenocytes and hNSCs did not result in an increase in CD25⁺Foxp3⁺ Tregs (Fig. 4C, 4D). Importantly, RAG2^−/−2D2⁺ mice lack Tregs in the periphery, suggesting that exposure to hNSC Ags promotes the conversion of naive Tconv cells into CD25⁺Foxp3⁺ Tregs, similar to the results seen when hNSC cocultures were performed with splenocytes depleted of tTregs.

It is clear that TCR engagement with presented self-antigen is critical in promoting an increase in hNSC-Tregs; however, it remained unclear whether peptide presentation is sufficient to upregulate Foxp3 expression in the absence of hNSCs. To examine this further, splenocytes from RAG2^−/−2D2⁺ and RAG2^−/−OT-II⁺ mice were cultured for 4 d ex vivo in the presence of various stimuli: alone, with anti-CD3 + anti-CD28, MOG peptide, NFM peptide, OVA peptide, or poly(I:C), with the latter serving as a nonspecific inducer of inflammation. T cells from RAG2^−/−2D2⁺ splenocytes cultured in the presence of anti-CD3 + anti-CD28, MOG, and NFM all became activated and upregulated CD25, although to a lesser extent than in the MOG and NFM conditions (Fig. 5A). Strikingly, only RAG2^−/−2D2⁺ splenocytes cultured with MOG or NFM upregulated Foxp3 (1.47 ± 0.34 and 3.43 ± 0.57%, respectively) (Fig. 5A, 5C, 5D). Addition of TGF-β and IL-2 to MOG and NFM cultures greatly enhanced the generation of CD25⁺Foxp3⁺ Tregs (26.26 ± 5.10 and 30.3 ± 1.70%, respectively), even when compared with anti-CD3 + anti-CD28 iTreg controls (6.97 ± 1.89%) (Supplemental Fig. 3A, 3C, 3D). Although all conditions containing MOG and NFM resulted in the generation of CD25⁺Foxp3⁺ RAG2^−/−2D2⁺ cells, there was a greater frequency of CD25⁺Foxp3⁺ cells under conditions containing NFM compared with MOG. Of note, RAG2^−/−2D2⁺ splenocytes cultured in the presence of poly(I:C) did not upregulate Foxp3, suggesting that Foxp3 expression was not due to bystander inflammation (41), but instead was mediated via TCR antigenic stimulation. Thus, exposure to self–peptide-MHC (pMHC) is sufficient to upregulate Foxp3 in vitro. RAG2^−/−OT-II⁺ splenocytes failed to develop CD25⁺Foxp3⁺ Tregs under any of these stimulation conditions, although cells did upregulate CD25⁺ when cultured in the presence of anti-CD3 + anti-CD28 and OVA, suggesting that these CD4⁺ cells are capable of recognizing peptide and becoming activated, but ultimately do not differentiate into Tregs (Fig. 5B, 5E, 5F). RAG2^−/−OT-II⁺ splenocytes were capable of generating CD25⁺Foxp3⁺ iTregs when TGF-β and IL-2 were added to anti-CD3 + anti-CD28 and OVA cultures (34.53 ± 7.12 and 3.05 ± 0.82%, respectively) (Supplemental Fig. 3B, 3E, 3F).

Next, we sought to address whether self-pMHC exposure was sufficient to upregulate Foxp3 expression in RAG2^−/−2D2⁺ T cells in vivo. Immunization of RAG2^−/−2D2⁺ mice with MOG peptide has been used as a model for autoimmune-mediated demyelination, mimicking certain features of MS (35). However, previous studies have not examined Foxp3 expression following immunization, possibly because these mice do not possess CD25⁺Foxp3⁺ cells under homeostatic conditions. To investigate this further, we immunized RAG2^−/−2D2⁺ mice with their cognate self-peptides MOG_35–55 or NFM_15–35, or OVA_323–339 as an irrelevant Ag control. Mice immunized with MOG and NFM peptides developed Foxp3⁺ Tregs within the brain and spinal cord (Fig. 6A) at 10 d postimmunization, displaying an increase in both frequency (Fig. 6B) and number (Fig. 6C) of CD25⁺Foxp3⁺ Tregs as compared with OVA-immunized controls. We also observed a small population of CD25⁺Foxp3⁺ cells in the cervical lymph nodes of MOG immunized mice, but not within the spleen under any immunization conditions (Fig. 6A), indicating that CD25⁺Foxp3⁺ Tregs are found primarily in tissues were cognate Ag is expressed. Taken together, these data suggest that exposure to self-pMHC is sufficient to upregulate Foxp3 in vivo.

In this study, we demonstrate that hNSCs drive the Ag-specific conversion of Tconv cells into a CD25⁺Foxp3⁺ Treg phenotype in vitro. Coculturing of splenocytes (containing peripheral T cells and APCs) with hNSCs resulted in a doubling in both the percentage and number of Tregs. This doubling occurred when tTregs were excluded from culture, suggesting that the observed increase in Tregs is the result of Tconv cell conversion as opposed to expansion of tTregs (Fig. 2). Furthermore, Treg conversion is unlikely a result of direct signaling or secretion of inductive factors by hNSCs, as Treg conversion occurred in the absence of viable hNSCs (Fig. 3A, 3B). Additionally, hNSCs failed to drive this observed increase in the absence of TCR signaling by cognate Ag, as was seen in hNSC RAG1^−/−OT-II⁺ cocultures (Fig. 4A, 4C), but it did occur when Ag was presented as seen in the hNSC RAG2^−/−2D2⁺ cocultures (Fig. 4B, 4D). Furthermore, in the absence of hNSCs, cognate self-antigen was shown to be sufficient to drive Treg conversion when RAG2^−/−2D2⁺ splenocytes were cultured in the presence of their cognate self-peptide MOG and NFM. These conversions failed to occur with polyclonal stimulation of the TCR via anti-CD3 and anti-CD28 and were not the result of bystander inflammation because poly(I:C) failed to induce a similar response (Fig. 5). Significantly, RAG^−/−OT-II⁺ T cells failed to generate a Treg response when stimulated with their cognate Ag OVA. These results strongly support the hypothesis that Treg conversion of naive Tconv cells in the periphery is driven through recognition of self-antigens that are presented in the thymus during thymic selection. These findings were also confirmed in vivo where RAG2^−/−2D2⁺ mice immunized with MOG and NFM generated strong Treg responses despite lacking a pool of tTregs in the periphery (Fig. 6). This response was not the result of general inflammation because OVA immunization of RAG2^−/−2D2⁺ mice failed to illicit a Treg response. Although these findings may seem narrow in their impact, they have broader immunological implications and impacts on therapies.

The activation of cross-reactive TCRs is one of many hypotheses to account for the genesis of autoimmunity; however, the importance of cross-reactive TCRs in tolerance is not well understood (50). The cross-reactivity of the 2D2 TCR to MOG and NFM is well understood and has been thoroughly examined through the lens of experimental autoimmune encephalomyelitis (EAE) (51–53). Despite their nearly identical peptide sequence, immunization with NFM fails to generate a sufficient MOG-reactive T cell response to induce EAE (53). Furthermore, knockout of NFM, but not MOG, results in fulminant EAE, suggesting that NFM may play a tolerizing role in the thymus or periphery (53). Additionally, previous work has demonstrated that NFM does not prophylactically tolerize against MOG-mediated EAE, but it suppresses disease progression when administered at the peak of disease. These findings suggest that NFM may not tolerize against MOG reactivity in the peripheral lymph nodes, but instead broadly suppresses MOG-reactive T cells in the CNS during disease progression (54). However, these studies did not examine the role of Tregs in their findings. The present study demonstrates that NFM peptide is capable of generating a pTreg response, which may add to our understanding of previous findings in the literature. Critically, we also recognize the limitations of the TCRs and peptides we examined in our efforts to understand how hNSC Ags drive pTreg generation and acknowledge that there are likely additional antigenic hNSC self-peptides that drive this conversion in wild-type mice. Additionally, tolerization by presentation of apoptotic hNSCs may further contribute to the strength of the Treg response in vivo and in vitro.

The prevailing opinion in Treg biology has been that tTregs are generated following agonist selection against self-antigens in the thymus and, conversely, pTregs are induced in the periphery by TCR reactivity to non–self-antigens such as those produced by commensals on the skin and in the gut. However, his paradigm has begun to shift as more research has demonstrated that non–self-antigens are presented by peripheral dendritic cells that traffic into the thymus (55, 56). Moreover, it has been demonstrated that pTreg generation in the periphery is dependent on epigenetic modifications that occur earlier during thymic development (57, 58). Interestingly, although these studies revealed that pTregs were generated in the periphery based on these epigenetic changes, the influence of self-antigen versus non–self-antigen recognition was not examined. Our findings suggest that pTregs are generated in response to self-antigens from the CNS within the periphery, as an increase in Tregs was also observed in the CNS draining cervical lymph nodes. Of note, our in vitro data suggest that peripheral presentation and recognition of thymically presented cognate self-antigen is sufficient to induce expression of Foxp3 in Tconv cells (59, 60). This finding may lend further refinement of work presented by others suggesting that presentation of Ag via tolerizing dendritic cells is sufficient to induce Tregs, as RAG2^−/−OT-II⁺ mice failed to generate Tregs when presented with OVA, a non–self-antigen that is not expressed in the thymus of these mice (61). Instead, it may be the case that naive Tconv cells are poised to convert into pTregs when they are presented with self-antigens that they have previously encountered during thymic selection. Although the local inflammatory environment may contribute to Treg induction, inflammation is not sufficient to generate pTregs, because poly(I:C) failed to generate a Treg response. Additionally, this response requires antigenic recognition because polyclonal stimulation by anti-CD3 and anti-CD28 failed to induce Tregs. Evidence indicates that Treg induction is based on specific signaling that may be imparted in the thymus. To our knowledge, these findings shine new light on our understanding of autoreactivity in T cells leaving the thymus. Instead of being overtly autoreactive, T cells with affinity to self may offer a mechanism of tolerance under circumstances in which Ags fail to generate Tregs in the thymus or rarely presented Ags become available during tissue damage.

Stem cell transplantation remains a promising therapeutic approach to repair the damage caused by the recurrent inflammation within the CNS that occurs in MS. Although cell replacement therapy remains a primary goal for these treatments, there has been a shift in our understanding of how NSC transplantation drives repair in the CNS, particularly with evidence that NSCs induce repair through the secretion of neurotropic factors (62–64). These novel approaches have the advantage of eliminating the problems associated with stem cell transplantation. Adding to this promise, our findings that hNSC Ags drive Treg generation is particularly exciting. Previous studies in our laboratory demonstrated that Tregs promote strong remyelination responses in mouse models of MS following hNSC transplantation (6, 13), and although the mechanism by which Tregs induce remyelination remains unknown, these findings are part of a growing literature that demonstrates a reparative role for Tregs outside of their immune-suppressive activities (13, 27, 28, 30, 65). This makes the findings in this study particularly exciting, as it may in fact be possible to simply expose patients to boluses of CNS Ags to both reduce autoimmune infiltration as wells as induce repair in the CNS by generating pTregs. There are currently a number of clinical trials underway examining the potential of vaccination against myelin epitopes as a method for inducing tolerance against these Ags. Although these trials show promise, they target induction of peripheral tolerance through Ag presentation in the periphery as opposed to presentation within the CNS. More research is required to examine the importance of local versus peripheral presentation to determine whether peripheral immunization is capable of inducing Treg infiltration into the CNS.

This study provides, to our knowledge, new insight into the generation of peripheral Tregs and reveals novel interactions between transplanted hNSCs and the adaptive immune system. In contrast to studies that suggest a direct role for hNSCs in suppression of autoimmune responses (14, 66), we have found that presentation of self-antigens derived from hNSCs drives tolerance through the induction of pTregs. This study further emphasizes the importance of understanding how the immune system interacts with transplanted cells in models of stem cell transplantation in the CNS and highlights the importance of studying Tregs not only for their suppressive capacity but their capacity to generate repair responses in the CNS.

The authors have no financial conflicts of interest.

We thank the animal care staff at University of California, Irvine (UCI), Vanessa Scarfone and Pauline Nguyen at the Sue and Bill Gross Stem Cell Research Center, Jennifer Atwood at the UCI Institute for Immunology, the UCI Alzheimer’s Disease Research Center for human brain tissue samples, and Melanie Oakes at the UCI Genomics Core. We thank Stephen Schoenberger and Joey Lee for the contribution of RAG1^−/−OT-II⁺ mice.