Research ArticleGraft-Versus-Host Disease

Massive ex Vivo Expansion of Human Natural Regulatory T Cells (Tregs) with Minimal Loss of in Vivo Functional Activity

See allHide authors and affiliations

Science Translational Medicine  18 May 2011:
Vol. 3, Issue 83, pp. 83ra41
DOI: 10.1126/scitranslmed.3001809


Graft-versus-host disease (GVHD) is a frequent and severe complication after hematopoietic cell transplantation. Natural CD4+CD25+ regulatory T cells (nTregs) have proven highly effective in preventing GVHD and autoimmunity in murine models. Yet, clinical application of nTregs has been severely hampered by their low frequency and unfavorable ex vivo expansion properties. Previously, we demonstrated that umbilical cord blood (UCB) nTregs could be purified and expanded in vitro using good manufacturing practice (GMP) reagents; however, the initial number of nTregs in UCB units is limited, and average yield after expansion was only 1 × 109 nTregs. Therefore, we asked whether yield could be increased by using peripheral blood (PB), which contains far larger quantities of nTregs. PB nTregs were purified under GMP conditions and expanded 80-fold to yield 19 × 109 cells using anti-CD3 antibody–loaded, cell-based artificial antigen-presenting cells (aAPCs) that expressed the high-affinity Fc receptor and CD86. A single restimulation increased expansion to ~3000-fold and yield to >600 × 109 cells while maintaining Foxp3 expression and suppressor function. nTreg expansion was ~50 million–fold when flow sort–purified nTregs were restimulated four times with aAPCs. Indeed, cryopreserved donor nTregs restimulated four times significantly reduced GVHD lethality induced by the infusion of human T cells into immune-deficient mice. The capability to efficiently produce donor cell banks of functional nTregs could transform the treatment of GVHD and autoimmunity by providing an off-the-shelf, cost-effective, and proven cellular therapy.


Acute graft-versus-host disease (GVHD) is a major cause of morbidity and mortality after hematopoietic cell transplantation (1). Natural regulatory T cells (nTregs) express the transcription factor Foxp3 and are required for immune self-tolerance (2). In murine models, adoptive transfer of nTregs prevents GVHD and donor bone marrow graft rejection, as well as speeds immune recovery in GVHD-prone animals (35), making Tregs an attractive therapeutic tool for preventing and/or treating disease in humans (69). However, clinical testing has been hampered by low nTreg frequency (1 to 2%) in peripheral blood (PB) (10), contamination with non-Tregs, such as CD25+ T-effector or T-memory cells (7, 11), and the lack of availability of good manufacturing practice (GMP)–compatible procedures for nTreg purification. Maximizing yield is also critical, because murine studies find that high Treg doses (~1:1 with donor T cells) are required to efficiently and reproducibly suppress GVHD (5).

Previously, we found that nTregs were more readily purified from umbilical cord blood (UCB) than PB because of the relative paucity of CD25+ non-Tregs in UCB; these cells could be expanded several hundred-fold ex vivo using anti-CD3/CD28 monoclonal antibody (mAb)–coated microbeads and interleukin-2 (IL-2) (11, 12). These studies allowed us to initiate the world’s first clinical trial to study the safety of ex vivo–expanded nTregs. Transferred nTregs remained Foxp3+ and could be tracked in blood for up to 14 days. No adverse effects were observed, and a trend toward a lower incidence of acute grade II to IV GVHD was observed, but the maximum cell dose was limited by insufficient and variable nTreg expansion rates for some UCB units (13). In other studies, we have shown that stimulation of UCB nTregs with cell-based artificial antigen-presenting cells (aAPCs) increases expansion (about fourfold) over bead-based aAPCs, but this increase alone would not have much effect on clinical nTreg dose. Because the nTreg number in UCB is limited and the dose-limiting toxicity was not reached, other nTreg sources need to be explored to determine the maximal efficacy of single- or multiple-dose nTreg therapy.

Despite non-Treg contaminants, isolation of PB nTregs offers several advantages over UCB nTregs, including increased nTreg number, continued donor availability for additional isolations, and use of autologous cells. PB nTregs can be successfully purified with cell sorting (14, 15) and expanded ~80-fold in vitro. However, cell sorting is a challenging GMP procedure, and overall nTreg yield from PB obtained with this isolation and expansion approach is not greatly increased over that from UCB. Restimulation increased total expansion to ~1000-fold, but cultures frequently lost Foxp3 expression and suppressive function concomitant with the appearance of effector T cells secreting IL-2 and interferon-γ (IFN-γ) (16, 17). Although nTregs can also be purified with mAb-coated magnetic beads, and ~30-fold more CD25high cells can be isolated from PB than from UCB (~150 × 106 compared with ~5 × 106, respectively), bead-purified nTregs contain higher numbers of CD25lo cells and are less pure than those obtained by flow cytometry sorting (18, 19). Thus, rapamycin, which preferentially inhibits cytokine responses in and survival of effector and memory T cells when compared with nTregs, is often added to bead-purified expansion cultures, albeit at the expense of a 5- to 10-fold reduction in nTreg expansion (2023).

Here, we show, using GMP-grade reagents, that repetitive nTreg stimulation with cell-based aAPCs massively increases nTreg yield while maintaining Foxp3 and suppressive function. Expanded cells expressed nTreg-specific markers [Foxp3 and LAP (latency-associated peptide)], displayed Treg-specific demethylation in the Foxp3 gene, and contained very few IL-2–, IFN-γ–, or IL-17–secreting cells. Despite four restimulations and expansion of >50 million–fold, fresh and cryopreserved nTregs each were capable of suppressing lethality in a xenogeneic model of GVHD. These findings advance the clinical utility of expanded nTregs for the prevention and treatment of GVHD after blood and marrow transplantation, solid organ rejection, and autoimmune disease.


PB nTregs can be purified and expanded with GMP reagents and protocols

Although our Phase I studies showed UCB nTreg cellular therapy to be well tolerated, a dose-limiting toxicity of nTregs was not reached, possibly because of limitations in nTreg expansion rates. Moreover, GVHD was significantly reduced but not eliminated compared to historical controls (13). To determine whether nTreg yield could be increased if the source was changed from UCB to PB, we purified cells from leukapheresis products with a two-step protocol using GMP antibody-coated magnetic beads, whereby CD4+ cells were enriched by depleting cells expressing CD8, CD14, and CD19, followed by positive selection of CD25high cells (Fig. 1A). The starting purity of PB nTregs was assessed by flow cytometry for a phenotype that displays potent suppressive capacity (CD4+127Foxp3+; fig. S1A) (24) and was comparable to previous observations for UCB nTregs (95 ± 1% CD4+, of which 66 ± 2% were CD127Foxp3+). Of the non-CD4+ cells in either cellular preparation, <1% were positive for CD8, CD14, and CD19 (fig. S1A). The average yield of PB nTregs after expansion (233 ± 31 × 106 cells) was ~40-fold higher than with UCB nTregs (13).

Fig. 1

Restimulation greatly increases PB nTreg expansion, and cell-based aAPCs are more effective than bead-based aAPCs. nTregs were purified from PB leukapheresis products and expanded with GMP anti-CD3/CD28 mAb-coated beads or an anti-CD3 mAb-loaded cell line (KT64/86). (A) GMP purification schema. (B) Schema showing time course of experiment and ranges for size-based restimulation (R0 = no restimulation, R1 = one restimulation, etc.). (C and D) Fold nTreg expansion (average ± SEM); total (C) or after each stimulation (D). (E) Percentage of cultured cells (CD4-gated) that are CD127Foxp3+ after each stimulation. (F) Percent suppression of in vitro, anti-CD3–mediated CD8+ T cell proliferation at 1:4 (nTreg/PBMNCs) as determined by CFSE dye dilution. (G) nTregs from each stimulation were restimulated with PMA and ionomycin for 4 hours in the presence of brefeldin A, and the percentage of cells secreting IL-2 or IFN-γ was determined by flow cytometry. (H) Bead-purified PB nTregs restimulated three or four times (black and gray symbols, respectively) with anti-CD3 mAb-loaded KT64/86 cells were harvested and genomic DNA was purified. Foxp3 TSDR demethylation status was assessed with bisulfite sequencing and is compared to nTreg purity (percentage of CD4+ cells that are CD127Foxp3+) or percent suppression at a 1:4 ratio of nTregs/PBMNCs. Averages are for three independent experiments. Individual symbols in (E) and (F) represent independent experiments. Brackets indicate the range of days for each stimulus. *P < 0.05.

Purified cells were stimulated with clinical-grade anti-CD3/CD28 mAb-coated beads or KT64/86 cells, a recently GMP-licensed, cell-based, aAPC-expressing CD86 (a CD28 ligand) and CD64 (the high-affinity Fc receptor) because of lentiviral gene transfer. KT64/86 cells were loaded with anti-CD3 mAb. IL-2 (300 U/ml) and rapamycin (109 nM) were added to all cultures (Fig. 1B). As reported (25), stimulation with either KT64/86 cells or anti-CD3/CD28 mAb-coated beads increased PB nTreg expansion by ~5-fold (82 ± 11–fold; 18 ± 5–fold, respectively) (Fig. 1, C and D). More robust expansion observed with KT64/86 cells and anti-CD3/CD28 mAb-coated beads was associated with increased overall viability (94.3 ± 0.5%; 90.0 ± 0.3%, respectively; P ≤ 0.001) and decreased granzyme B production (P < 0.02) (fig. S1, B to D). nTreg cultures stimulated once with anti-CD3/CD28 mAb-coated beads or KT64/86 maintained an nTreg phenotype (97 ± 2% or 99.5 ± 0.3% CD4+, of which 81 ± 5% or 84 ± 7% were CD127Foxp3+, respectively) and in vitro function [84 ± 12% or 83 ± 6% inhibition of carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled CD8+ T cell proliferation after anti-CD3 stimulation at a 1:4 ratio of nTreg/PB mononuclear cells (PBMNCs)] (Fig. 1, E and F). These data demonstrate that suppressive, Foxp3+ nTregs can be expanded from PB with GMP procedures and show that stimulation with cell-based aAPC is superior to anti-CD3/CD28 mAb-coated beads. However, because PB nTreg expanded 5- to 10-fold less than UCB nTregs expanded without rapamycin, overall nTreg yield was not substantially increased.

Restimulation greatly increases nTreg expansion

To maximize yield, we restimulated GMP bead-purified nTregs grown in rapamycin after they had returned to resting size (≤8.5 μm; fig. S1E), which we have shown maximizes CD4+ T cell expansion (26). nTregs stimulated with KT64/86 cells were found to have a higher peak cell size compared to anti-CD3/CD28 mAb-coated beads (fig. S1F). Restimulation with anti-CD3/CD28 mAb-coated beads or KT64/86 cells increased expansion 18- and 36-fold, respectively, to a total of 330- or 3000-fold over input cell number (Fig. 1, C and D). Cultures remained >65% Foxp3+ and suppressed in vitro T cell proliferation >50% at a ratio of 1:4 (nTregs/PBMNCs) (Fig. 1, E and F). Of the non-CD4+ cells expanded with either stimulus, <1% were positive for CD8, CD14, CD19, or CD56 (fig. S1G).

Others have shown that nTreg restimulation in the absence of rapamycin results in up to 30% of cells that secrete IL-2 and/or IFN-γ, two cytokines that could potentially exacerbate GVHD (16, 17). Therefore, we quantified the number of IL-2– and IFN-γ–secreting cells by intracellular cytokine staining after phorbol 12-myristate 13-acetate (PMA)/ionomycin stimulation of bead-purified nTreg cultured with anti-CD3/CD28 mAb-coated beads or KT64/86 cells and rapamycin. As shown in Fig. 1G, <1% of cells expanded with either anti-CD3/CD28 mAb-coated beads or KT64/86 cells secreted IL-2. Although less than 6% of cells in any culture expanded with either aAPCs or rapamycin were IFN-γ+, significantly fewer IFN-γ+ cells were found in nTreg cultures expanded with KT64/86 than CD3/CD28 beads (1% for KT64/86 restimulation; 4% for R0 bead restimulation; 6% for R1 bead restimulation).

Multiple restimulations lead to reduced suppressive function despite the presence of Foxp3

To determine whether bead-purified nTregs could be expanded even further, we stimulated the above cultures another three times (four restimulations total) (Fig. 1C). In contrast to anti-CD3/CD28 mAb-coated bead-expanded cultures, whose peak size declined after each stimulation and was <9.0 μm after the fourth restimulation, peak size after KT64/86 cell stimulation remained high at ~9.5 μm (fig. S1, E and F). In addition, the fold expansion induced by successive stimulations with anti-CD3/CD28 mAb-coated beads decreased more rapidly than with KT64/86 cells, ultimately resulting in 200-fold lower total expansion than with KT64/86 cells (25,000-fold versus ~5 million–fold, respectively) (Fig. 1, C and D). nTregs restimulated with anti-CD3/CD28 mAb-coated beads remained >80% Foxp3+, and although expression gradually decreased in KT64/86 cell–expanded nTregs, Foxp3 remained in >60% of cells after the fourth restimulation (Fig. 1E). nTregs expanded with anti-CD3/CD28 mAb-coated beads also had higher amounts of Foxp3 on a per cell basis than those expanded with KT64/86 cells (fig. S1, H and I). Despite achieving Foxp3 levels previously associated with significant suppressive function by expanded UCB nTregs (13), <50% suppression of anti-CD3 mAb-driven CD8+ T cell proliferation at a 1:4 nTreg/PBMNC ratio was observed in two of two and one of three cultures restimulated three or four times with anti-CD3/CD28 mAb-coated beads, respectively, as well as in two of three cultures restimulated with KT64/86 cells either three or four times (Fig. 1F).

Stable expression of Foxp3 is a trait of natural, but not induced, Tregs and is conferred through epigenetic modification of the Foxp3 gene at the Treg-specific demethylated region (TSDR) (27). To assess the methylation status of the Foxp3 gene in restimulated nTreg, we purified, bisulfite-modified, and sequenced DNA from cultures receiving three or four restimulations, and we determined the average percent methylation of 11 CpG sites contained in the TSDR. Because Foxp3 is on the X chromosome and becomes hypermethylated during X-inactivation, the data shown are restricted to male samples. An evaluation of two informative samples, Fig. 1H suggests that TSDR demethylation status is proportional to Foxp3 and slightly decreases between the third and the fourth restimulation, although the effects were not significant (r = 0.65, P = 0.35). As observed for Foxp3, TSDR demethylation is not directly proportional to suppressive function (r = 0.75, P = 0.25).

Sort-purified nTregs maintain Foxp3 and suppressive function after multiple stimulations

Decreased suppressive function could be caused by contaminating cells that become amplified after restimulation and acquire Foxp3 during the process of massive cell expansion. Therefore, PB nTregs were purified by flow cytometry sorting and restimulated with KT64/86 cells in the presence or absence of rapamycin (Fig. 2A). To enable more meaningful comparisons of the various restimulations, we also developed freeze/thaw conditions that allow expanded nTregs to maintain phenotype and suppressive function so that all samples are assayed simultaneously (fig. S2). The most common strategy for sorting nTreg is to first purify CD4+ cells and then gate on the 2% of cells with the highest expression of CD25. Although cells purified in this manner are regularly >90% Foxp3+; this method is relatively inefficient and only captures ~20% of the total Foxp3+ cells. To maximize yield, we performed an initial purification with magnetic anti-CD25 mAb-coated beads and then sorted for CD4+25hi127 cells, which allowed >25% of sorted cells to be positively selected. This method increased initial nTreg purity from 66 ± 2% CD127Foxp3+ for bead-based purification to 84 ± 3 (P ≤ 0.003), and resulted in a routine yield of 15 × 106 to 30 × 106 nTregs from 2 × 109 PBMNCs (fig. S3, A and B).

Fig. 2

Sort-purified nTregs maintain Foxp3 and suppressive function after multiple stimulations. PB nTregs were sort-purified (CD4+25hi127) and expanded with anti-CD3 mAb-loaded KT64/86 in the presence or absence of rapamycin using four or two restimulations, respectively. (A) Schema showing time course of experiment and time ranges for size-based restimulation (R0 = no restimulation, R1 = one restimulation, etc.). Brackets indicate the range of days for each stimulus. (B) Average cell size (±SEM) over time for PB nTreg cultures restimulated ± rapamycin. (C and D) Representative examples (C) and average (D) expansion of nTregs ± rapamycin, respectively. Arrows in (C) on days 25 and 55 mark two distinct phases (plateau and growth phase) seen after first restimulation of nTreg cultures grown without rapamycin. (E to H) Average percent CD127Foxp3+ (CD4-gated) or percent suppression of in vitro T cell proliferation at a 1:4 ratio of nTregs/PBMNCs for cultures expanded in the absence (E and F) or presence (G and H) of rapamycin, respectively. Bars represent average; other symbols represent individual experiments.

nTregs stimulated with KT64/86 cells in the absence of rapamycin, which is known to affect size and proliferation (28), had both a larger peak size (Fig. 2B; 10.4 without rapamycin; 9.9 μM with rapamycin; P < 0.01) and increased expansion (Fig. 2, C and D; 290- and 55-fold, respectively). nTregs cultured in the presence or absence of rapamycin were restimulated at 8.5 μm (day 13 ± 1), and after 4 days of expansion and size increase, cultures started to decrease in size and stop expanding. However, after day 25, without additional stimulation, cultures grown without rapamycin increased in size to ~9.3 μM and started proliferating, impressively expanding >5 × 1011–fold throughout the 55-day observation period. Additional restimulation did not increase either cell size or maximal expansion. Day 55 cultures contained few CD127Foxp3+ nTregs and were not suppressive, whereas those harvested on day 25 had expanded 11,000 ± 2000–fold, were ≥60% CD127Foxp3+, and conferred ≥60% suppression of CD8+ T cell proliferation (Fig. 2, E and F).

In contrast to cultures established in the absence of rapamycin, sort-purified nTregs expanded with KT64/86 cells in the presence of rapamycin returned to resting size and ceased proliferating after each restimulation (Fig. 2, B and C). Cumulative expansion after restimulation of sort-purified nTregs + rapamycin was >6-fold higher than mAb-coated bead-purified nTregs (31 ± 14 × 106–fold and 4.7 ± 0.7 × 106–fold expansion, respectively, P < 0.05), due mainly to the fact that the fold expansion did not decline after each restimulation (fig. S3C). Repeated stimulation caused a gradual decrease in Foxp3 such that after the fourth restimulation, 63 ± 12% of cells were CD127Foxp3+ (Fig. 2G). However, unlike bead-purified nTregs, sort-purified nTregs expanded after four repetitive stimulations maintained >50% suppression of T cell responses for all restimulations (Fig. 2H). Table S1 summarizes the in vitro expansion characteristics (yield, Foxp3 expression, T-effector cell contamination, and suppressive function) for nTreg derived from UCB and PB with varying culture and restimulation conditions.

To determine whether restimulation affects the Treg phenotype, we assessed the level of several Treg-associated markers (including LAP, CD62L, CD27, CCR7, and CD45RA) on cells receiving either single or multiple stimulations. Although LAP, derived from the N-terminal region of transforming growth factor–β (TGFβ), was expressed on Foxp3+ cells after all four restimulations (fig. S4A), CD62L and CD27 staining was lost after two and four restimulations, respectively (fig. S4B). CCR7 behaved like an activation marker; it was more highly expressed at day 7 than at resting size (fig. S4C). However, if nTregs were maintained in culture after returning to basal size, a subpopulation of nTregs spontaneously regained CCR7 staining (fig. S4D). Although it is not surprising that restimulation decreased CD45RA expressed on naïve, resting T cells and Tregs, the finding that cells regained staining after returning to resting size was unanticipated (fig. S4E, especially restimulations 2 and 3).

We next examined changes in surface phenotype and T cell receptor (TCR) repertoire usage of the Tregs expanded with one (R0) or a total of five stimulations (R4). After multiple rounds of stimulation, the nTreg phenotype changed from CD27+CD45RACD57 to CD27CD45RACD57, suggesting that these cells were undergoing differentiation to a more mature state (fig. S5). However, an increase in CD57 expression was not noted after expansion, suggesting that the cells did not become terminally differentiated or senescent (29). Finally, TCR Vβ usage was essentially unchanged between R0 and R4, suggesting that particular TCR Vβ families were not preferentially expanded despite a massive increase in nTreg number during the course of cell culture (fig. S6).

nTreg cultures restimulated with KT64/86 cells in the presence of rapamycin do not secrete IL-2 or effector cytokines

Repetitive stimulation of TH cells in the absence of rapamycin generates effector cells, which secrete cytokines that could exacerbate GVHD. To determine the extent of effector T cell contamination in our cultures, we stimulated samples of each restimulation from KT64/86-expanded cultures grown with or without rapamycin with PMA/ionomycin and assayed them for IL-2, IL-4, IL-17, and IFN-γ (Fig. 3A) using intracellular cytokine staining. To make comparisons between various restimulations more valid, we assayed frozen nTregs representing all conditions simultaneously and co-stained them for Foxp3 to differentiate secretion by nTregs and non-Treg cells. Adding rapamycin suppressed effector cell generation such that ≤3% of PMA/ionomycin-stimulated cells secreted IL-2 and ≤2% secreted IFN-γ, compared to ≥17% and ≥6% for cultures without rapamycin (Fig. 3, B and C). In contrast, rapamycin was less effective at inhibiting IL-4 production in Foxp3+ or Foxp3 cells, and the percentage of IL-4+ cells increased with each successive restimulation from 8 ± 2% to 58 ± 17% (Fig. 3D). Finally, the total number of IL-17–secreting cells present in cultures of sorted nTregs was consistently low (<3.1%) for all stimulations with or without rapamycin (Fig. 3E).

Fig. 3

nTreg cultures restimulated with KT64/86 cells in the presence of rapamycin do not secrete IL-2 or effector cytokines. PB nTregs were sort-purified and expanded with multiple rounds of stimulation with anti-CD3 mAb-loaded KT64/86 cells in the presence or absence of rapamycin. The R1 without rapamycin sample corresponds to the day 25 time point with high Foxp3 staining. (A) Representative example of cytokine production by Foxp3+ and Foxp3 cells (CD4-gated). (B to E) Average (±SEM) percent of cells secreting IL-2 (B), IFN-γ (C), IL-4 (D), or IL-17 (E). Averages are for three independent experiments.

nTregs expanded with multiple rounds of stimulation ameliorate disease in a xenogeneic model of GVHD

Several groups have reported that nTregs are not terminally differentiated and can be reprogrammed into helper T cells (30) and T-effector cells (31, 32), which are capable of inducing proinflammatory responses and disease (32). Therefore, we used a xenogeneic model of GVHD in which nTregs are co-transferred at a 1:1 ratio with allogeneic PBMNCs (30 × 106 each) into NOD/Scid/γc−/− recipients to compare the stability and safety of in vitro–expanded nTregs and CD4+CD25 cells that were restimulated four times cultured in the absence or presence of TGFβ, which was used to induce Foxp3 (Fig. 4A). Adoptive transfer of nTregs increased median survival from 39 to 55 days (P < 0.01). Transfer of non-Tregs appeared to exacerbate GVHD, even if Foxp3 was induced with TGFβ (Fig. 4B), whereas Foxp3 cells present in nTreg cultures did not expand or persist long-term and, in contrast to cultures expanded from CD4+CD25 cells, did not exacerbate GVHD. We also tested the in vivo potency, stability, and safety of nTregs expanded 50 million–fold with four restimulations using KT64/86 cells. Although recipients of PBMNCs rapidly and uniformly succumbed to GVHD, mice given nTregs had a significantly prolonged survival, with 25% of mice surviving to day 55 (Fig. 4D; n = 8 to 10 per group; P < 0.05). GVHD amelioration was also indicated by a significant decrease in weight loss between days 14 and 21 (Fig. 4E). The partial protection seen using nTregs in this xenogeneic GVHD model has also been observed using UCB nTregs obtained after a single stimulation with anti-CD3/CD28 mAb-coated beads, which we have shown to rescue 50% of macrophage-depleted, sublethally irradiated Rag2−/− γc−/− recipients when infused at a 1:1 ratio with PBMNCs (12).

Fig. 4

PB nTregs expanded >50 million–fold can still ameliorate disease in a xenogeneic model of GVHD even after freezing and thawing. (A) Summary of purity (percentage of CD4+ cells that are CD127Foxp3+) and in vitro–suppressive function for in vitro–expanded nTregs or CD4+CD25 cells (grown ± TGFβ) after a single stimulation with KT64/86 cells. (B) Kaplan-Meier survival curve comparing NOD/Scid/γc−/− mice that received human PBMNCs only or co-transferred with nTregs, CD4+CD25 cells, or CD4+CD25 cells expanded in TGFβ co-transferred at 1:1 (for example, 30 × 106 PBMNCs and 30 × 106 nTregs). (C) Summary of fold expansion, purity (percentage of CD4+ cells that are CD127Foxp3+), and in vitro–suppressive function for nTregs expanded with four restimulations (R4). (D) Kaplan-Meier survival curve comparing NOD/Scid/γc−/− mice receiving human PBMNCs ± fresh nTregs restimulated four times (R4) co-transferred at 1:1. (E) Average weight (percentage of initial) for mice surviving on a given day for different groups of mice. P ≤ 0.05 for fresh nTregs from days 14 to 21. (F) Summary of fold expansion, purity (percentage of CD4+ cells that are CD127Foxp3+), and in vitro–suppressive function for expanded nTregs restimulated three or four times (R3 and R4, respectively). (G) Kaplan-Meier survival curve showing survival of mice receiving human PBMNCs ± cryopreserved and thawed R3 or R4 nTregs (HLA-A2+) co-transferred at 1:2 (that is, 15 × 106 nTregs and 30 × 106 PBMNCs). n = 10, 8, and 7 for groups PBMNCs, R3 nTregs, and R4 nTregs, respectively. (H) Average number (±SEM) of human CD4+HLA-A2, CD8+HLA-A2, or total CD4+/CD8+/HLA-A2 cells per microliter of blood on day 30 for animals in (G).

Because there was a modest decrement in %CD127Foxp3+ and in vitro suppression of nTregs in the GVHD model, we used a suboptimal ratio of nTregs/PBMNCs (1:2) to help uncover potential differences between nTreg restimulated three or four times. Figure 4F shows that nTregs restimulated three or four times maintained their phenotype and in vitro–suppressive function after cryopreservation and thawing. Both expanded nTreg preparations significantly reduced GVHD-induced lethality when compared with PBMNC controls, and there was no difference in their relative potency (Fig. 4G; P < 0.003 and P < 0.001 compared with PBMNC controls for three or four restimulations, respectively). Expansion of PB-derived CD4+ and CD8+ T cells is predictive of GVHD severity, and Fig. 4H shows that, like UCB nTregs, co-transfer of restimulated PB nTregs significantly reduced the number of GVHD-causing T cells on day 30 after transfer.


The therapeutic potential of nTregs to prevent or cure multiple autoimmune diseases or GVHD in murine or xenogeneic models has been well documented (35). Two critical obstacles to overcome before implementing this therapy in humans are generating sufficient cell numbers and demonstrating their in vivo safety and stability. Here, we show that sort-purified nTregs could be expanded at least 50 million–fold by repetitive stimulation with cell-based aAPCs while maintaining suppressive function in vitro and in vivo. Addition of rapamycin minimized contamination with T helper 1 (TH1) inflammatory cytokine-secreting cells, but not TH2 cells, which skew immunity away from inflammatory responses. Restimulated nTregs differentiated from a CD27+ memory phenotype to CD27 memory phenotype but, importantly, did not adopt a senescent (CD57+) phenotype (33, 34). The lack of Vβ skewing in the TCR repertoire indicates that massively expanded nTregs retain a broad spectrum of reactivities and are not transformed.

Maximizing nTreg expansion, while minimizing loss of suppressive function and contamination with non-Tregs, is critical for establishing an nTreg cellular therapy. Three studies have shown that nTregs can be expanded >1000-fold if restimulated in the absence of rapamycin, but in each case, cultures contained high numbers of IL-2– and IFN-γ–secreting cells that were both Foxp3 and Foxp3+ (16, 17, 35). We confirmed these data and found that nTreg cultures eventually lost Foxp3 and suppressive function in the absence of rapamycin. Loss of Foxp3 correlated with an increased ratio of cycling (that is, Ki-67+) Foxp3 cells (fig. S7), suggesting that loss of purity is due to the outgrowth of Foxp3 cells as opposed to conversion of Foxp3+ cells as suggested by one report (16).

We previously demonstrated that the increased stimulatory capacity of cell-based aAPCs allowed PB nTregs to be expanded 1000-fold with a single restimulation, even in the presence of rapamycin, and nTregs expanded with aAPCs were equal to anti-CD3/CD28 mAb-coated bead-expanded cells at suppressing xenogeneic GVHD [fig. S1K and (25)]. For these initial studies, restimulation was performed at the growth plateau phase, but the high variability (days 8 to 12) and difficulty of determining this time point are not conducive to clinical production. Restimulation on a specific day is optimal for clinical trials. However, although studies without rapamycin showed that restimulation on day 7 increased expansion, we observed no increase in expansion of day 7 restimulation (n = 3) using bead-purified nTreg stimulated with anti-CD3/CD28 mAb-coated beads (25- and 18-fold for with or without day 7 restimulation, respectively). Although restimulation based on cell size resulted in more variability in the day of optimal restimulation than would be the case at a single time point, such an approach identified a time range (day 13 ± 1) more suitable for clinical restimulation.

All nTreg cultures contain some number of Foxp3 cells, which have the potential, especially after restimulation, to become effector T cells and exacerbate disease. Although nTreg cultures restimulated in the absence of rapamycin contained high numbers of IL-2– and IFN-γ–secreting cells, the number of these cells did not increase with restimulation in the presence of rapamycin. Furthermore, when transferred in vivo, Foxp3 cells present in nTreg cultures did not expand or persist long-term and, in contrast to cultures expanded from CD4+CD25 cells, did not exacerbate GVHD. In addition, studies show that rapamycin temporally imparts Foxp3 expression and Treg-like activity to effector T cells, which can reacquire T-effector cell function if rapamycin is removed (36). LAP expression differentiates activated nTregs from stimulated CD4+CD25 T cells expressing Foxp3 spontaneously or after exposure to TGFβ or rapamycin (17). Even after four restimulations, most Foxp3+ cells expressed LAP even 7 days after restimulation, showing that the cultures remain primarily nTregs. Furthermore, cultures expanded >1 million–fold maintained nTreg-specific demethylation in the Foxp3 gene. Murine T cells expanded in rapamycin are TH2-skewed, secrete IL-4 and IL-10, and, after adoptive transfer, decrease allospecific IFN-γ secretion and ameliorate disease in a murine model of GVHD (37). Although rapamycin almost completely inhibited the differentiation of IL-2– and IFN-γ–secreting cells in our cultures of human cells, the effect on IL-4 was not complete, and >50% of cells secreted IL-4 (both Foxp3+ and Foxp3 cells) after the fourth restimulation.

Murine and human nTregs are not terminally differentiated and can be reprogrammed to secrete IL-17 in vitro or in vivo when activated in the presence of IL-6 (31, 32, 38). Adoptive transfer of reprogrammed murine nTregs induced autoimmune diabetes but, unlike their human counterparts, these cells also produced IFN-γ and tumor necrosis factor–α (TNFα). It is not known whether reprogrammed human nTregs will cause disease, because only ~5% of nTregs become IL-17+ in vitro (38), and these retain suppressive function (31). Several findings from this study suggest that nTreg reprogramming may not be a grave issue in developing a cellular therapy for in vitro–expanded nTregs. First, IL-17 was undetectable in the supernatants of all restimulation samples cultured with rapamycin (limit of detection, 0.3 pg/ml). Second, the number of expanded cells that were IL-17+ cells was very low (<2% total and ≤0.5% Foxp3+IL-17+) and, even more important, did not increase significantly over the four restimulation cycles (fig. S1J). Although the likelihood for in vivo reprogramming of nTregs and especially expanded nTregs may be context-dependent, the high degree of TSDR demethylation of these cells may provide some degree of resistance to the reprogramming process.

In summary, the degree of nTreg expansion reported here could lead to the widespread application of nTreg cellular therapy for GVHD and graft rejection through the creation of an off-the-shelf therapy using nTreg banks generated from human leukocyte antigen (HLA)–typed donors with known safety and potency records. The massive expansion observed with repetitive polyclonal stimulation should also allow relatively rare, autoantigen-specific nTreg clones to be expanded to treat autoimmune diseases. Ultimately, this strategy could be applied to expansion of antigen-specific nTregs, which are more effective than polyclonal Tregs at suppressing disease. This strategy is potentially preferable to using Tregs induced in vitro by Foxp3 gene transfer or other conditions that favor Foxp3 expression. Furthermore, if increased purity and/or suppressive function is required, nTregs could be reisolated after expansion using a protocol described recently by Shevach’s group based upon LAP expression (17). Although GMP sorting can be challenging for many institutions, restimulation-driven expansion could produce sufficient numbers of cells in a small number of sorts to support the creation of a master cell bank that would contain matches for multiple patients. Finally, an nTreg master cell bank would be an effective treatment for multiple diseases because, as shown here, nTregs suppress third-party responses, are able to maintain suppressive function after freeze/thaw, and ameliorate disease without long-term persistence.

Materials and Methods

Treg isolation and culture

For all experiments, nonmobilized PB leukapheresis products were collected from normal adult volunteers with Food and Drug Administration (FDA)–approved/cleared apheresis instruments. Written informed consent was obtained from all subjects with approval from the University of Minnesota Institutional Review Board. nTregs were purified with GMP magnetic beads or by sorting and cultured as in the Supplementary Material. Where indicated, rapamycin (Rapammune, Wyeth-Ayerst) at 109 nM was added on day 0 and with subsequent media supplementation. Cell size and viability were determined by ViCell (Beckman Coulter).

For mAb bead-based nTreg purification, CD4+ T cells were enriched by MACS (all beads from Miltenyi Biotec) by depleting non-CD4+ cells with GMP-grade mAb-coated microbeads (cocktail of CD8, CD14, CD19 ± CD56, 7.5 ml each/apheresis product) in combination with a CliniMACS (Depletion 2.1, max TNC = 2.0 × 1010). Unbound cells were washed and CD25high Tregs were subsequently purified by positive selection with GMP-grade anti-CD25 mAb-coated microbeads (7.5 ml/apheresis product) and CliniMACS (Enrichment 3.2). CD8/CD14/CD19/CD25 cells were subsequently enriched for CD3+ feeder cells with GMP-grade anti-CD3 microbeads. All bead incubations were performed as specified by the manufacturer (that is, 30 min at room temperature for GMP-grade beads). All washes were performed at 300g for 10 min at room temperature.

nTregs were sort-purified from PBMNCs (Ficoll-Hypaque, Amersham Biosciences) in a two-step procedure in which CD25+ cells were initially enriched from PBMNCs by AutoMACS (PosselD2) with GMP-grade anti-CD25 microbeads (75 μl/2 × 108 cells). CD25high cells were stained with CD4, CD8, CD25, and CD127 and sorted via FACSAria as CD4+, CD8, CD25high, and CD127. Note that the bead-bound and fluorochome-conjugated anti-CD25 antibodies recognize different epitopes.

Purified CD4+CD25+ cells were cultured either with GMP anti-CD3/CD28 mAb-coated Dynabeads (26) (3:1 bead/cell) or with K562 cell lines engineered to express CD86 and the high-affinity Fc receptor (CD64) (37) (2:1 nTreg/KT), which had been irradiated with 10,000 cGy and incubated with anti-CD3 (Orthoclone OKT3, Janssen-Cilag). In some experiments, nTregs were stimulated with KT64/86 cells that were preloaded, irradiated, and frozen (1:1 nTreg/KT). Irradiated feeder cells (26 Gy, CD8/CD14/CD19/CD25/CD3+) were added to CD3/CD28 bead cultures at 1:1 feeder/Treg. nTregs were cultured in X-Vivo 15 media (BioWhittaker) supplemented with 10% human AB serum (Valley Biomedical), GlutaMAX (Gibco), and N-acetylcysteine (USP). Recombinant IL-2 (300 IU/ml, Chiron) was added on day 2 and maintained for culture duration. Cultures were maintained at 0.3 × 106 to 0.5 × 106 viable nucleated cells/ml every 2 to 3 days.

Intracellular cytokine staining

Fresh or frozen nTregs were cultured in supplemented X-Vivo 15 for 4 hours ± PMA (2 pg/ml) and ionomycin (1 μg/ml) in the presence of brefeldin A (100 ng/ml) (all Sigma). Frozen/thawed samples were cultured for 1 hour at 37°C before restimulation. Cells were then harvested and stained for CD4, CD25, Foxp3, and cytokine (IL-2, IL-4, IL-17, and IFN-γ) or granzyme B with the standard Foxp3 intracellular staining kit.

Suppression assays

The in vitro–suppressive capacity of expanded nTregs was assessed with a CFSE inhibition assay as previously published. Briefly, PBMNCs were purified, labeled with CFSE (Invitrogen), and stimulated with anti-CD3 mAb-coated beads (Dynal) ± cultured nTreg (1:2 to 1:32 nTregs/PBMNCs). On day 4, cells were stained with antibodies to CD4 and CD8 and proliferation, data were analyzed with FlowJo (8.8.7), and suppression was determined from the Division Index (TreeStar). nTregs suppressed CD4+ and CD8+ T cell responses equivalently (fig. S8), and only CD8 data are presented. Xenogeneic GVHD experiments were performed as in (25) and are described in the Supplementary Material.

Statistical analysis

Data were analyzed by analysis of variance (ANOVA) or Student’s t test. Probability (P) values of ≤0.05 were considered statistically significant.

Supplementary Material

Materials and Methods

Fig. S1. nTregs stimulated with cell-based aAPCs have increased peak size but decreased Foxp3.

Fig. S2. Cultured nTregs maintain Foxp3-suppressive function after cryopreservation and thawing.

Fig. S3. Sort-purified nTregs expand more than bead-purified nTregs.

Fig. S4. Phenotype of restimulated nTregs.

Fig. S5. Massively expanded nTreg phenotype as memory T cells, but are not exhausted.

Fig. S6. No T cell receptor (TCR) Vβ skewing after massive expansion of Tregs.

Fig. S7. Loss of Foxp3+ cells is due to increased cycling of Foxp3 cells.

Fig. S8. nTregs stimulated once or a total of five times suppress CD4+ and CD8+ T cell responses equivalently.

Table S1. Summary of in vitro nTreg expansion comparing source, stimulation conditions, and restimulation.


  • Citation: K. L. Hippen, S. C. Merkel, D. K. Schirm, C. M. Sieben, D. Sumstad, D. M. Kadidlo, D. H. McKenna, J. S. Bromberg, B. L. Levine, J. L. Riley, C. H. June, P. Scheinberg, D. C. Douek, J. S. Miller, J. E. Wagner, B. R. Blazar, Massive ex Vivo Expansion of Human Natural Regulatory T Cells (Tregs) with Minimal Loss of in Vivo Functional Activity. Sci. Transl. Med. 3, 83ra41 (2011).

References and Notes

  1. Acknowledgments: We would like to thank C. Nelson for assistance with animal husbandry. Funding: This work was supported in part by research grants from the Children’s Cancer Research Fund and Blood and Marrow Transplant Research Fund to K.L.H.; Leukemia and Lymphoma Translational Research (grant R6029-07), R37 HL56067, and P01 AI056299 to B.R.B., P01 CA067493 and N01HB037164 to B.R.B., J.E.W., and J.S.M.; support from Miltenyi Biotec to B.R.B. and J.E.W.; a grant from Becton Dickinson to J.E.W.; and support from the JDRF Collaborative Centers for Cell Therapy and the JDRF Center on Cord Blood Therapies for Type 1 Diabetes to J.L.R. and C.H.J. Author contributions: K.L.H. designed the research, performed the experiments, interpreted the data, and wrote the paper. S.C.M., D.K.S., C.M.S., and P.S. performed the experiments, interpreted the data, and assisted with the paper. D.S. and D.M.K. performed the research. J.S.M., J.E.W., D.C.D., D.H.M., J.S.B., B.L.L., C.H.J., and J.L.R. designed the research and wrote the paper. B.R.B. designed the research, interpreted the data, and wrote the paper. Competing interests: J.L.R., C.H.J., and J.E.W. have research funding from Becton Dickinson and C.H.J. and B.R.B. were previously scientific consultants for Becton Dickinson, although this funding did not conflict with this manuscript. K.L.H., J.L.R., C.H.J., and B.R.B. are authors on U.S. provisional patent application number 61/322, 186, “Methods to expand a T regulatory cell master cell bank.” The other authors declare that they have no competing interests.
View Abstract

Navigate This Article