Regulatory T Cells Require Mammalian Target of Rapamycin Signaling To Maintain Both Homeostasis and Alloantigen-Driven Proliferation in Lymphocyte-Replete Mice

Abstract

Rapamycin (Rapa), an immunosuppressive drug that acts through mammalian target of Rapa inhibition, broadly synergizes with tolerogenic agents in animal models of transplantation and autoimmunity. Rapa preferentially inhibits conventional CD4⁺ Foxp3⁻ T cells (Tconv) and promotes outgrowth of CD4⁺Foxp3⁺ regulatory T cells (Treg) during in vitro expansion. Moreover, Rapa is widely perceived as augmenting both expansion and conversion of Treg in vivo. However, most quantitative studies were performed in lymphopenic hosts or in graft-versus-host disease models. We show in this study that in replete wild-type mice, Rapa significantly inhibits both homeostatic and alloantigen-induced proliferation of Treg, and promotes their apoptosis. Together, these lead to significant Treg depletion. Tconv undergo depletion to a similar degree, resulting in no change in the percent of Treg among CD4 cells. Moreover, in this setting, there was no evidence of conversion of Tconv into Treg. However, after withdrawal of Rapa, Treg recover Ag-induced proliferation more quickly than Tconv, leading to recovery to baseline numbers and an increase in the percent of Treg compared with Tconv. These findings suggest that the effects of Rapa on Treg survival, homeostasis, and induction, depend heavily on the cellular milieu and degree of activation. In vivo, the resistance of Treg to mammalian target of Rapa inhibition is relative and results from lymphopenic and graft-versus-host disease models cannot be directly extrapolated to settings more typical of solid organ transplantation or autoimmunity. Moreover, these results have important implications for the timing of Rapa therapy with tolerogenic agents designed to increase the number of Treg in vivo.

CD4⁺Foxp3⁺ regulatory T cells (Treg) play an important role in prevention of autoimmunity and graft-versus-host disease (GVHD) and in the induction and maintenance of transplant tolerance (1, 2). The best characterized of these are naturally occurring Treg which are generated as a distinct lineage of CD4 cells in the thymus and express the Foxp3 transcription factor. Mutation of Foxp3 results in systemic autoimmunity in both humans and mice (1). In the periphery, natural Treg normally comprise 5–10% of CD4 cells. Under specialized circumstances, conventional T cells (Tconv) can be induced to express Foxp3 and exhibit regulatory activity both in vitro and in vivo (3–5). However, the exact role of induced Treg remains to be defined (1, 6, 7).

Given their key role in immunologic tolerance, a major effort has focused on ex vivo or in vivo expansion of Treg for therapeutic purposes (8–10). In this regard, a variety of agents that promote immunological tolerance in mice either directly or indirectly increase Treg (9–14). There has also been great interest in finding approved immunosuppressive drugs that can synergize with experimental agents to enhance Treg and promote tolerance. We and others have shown that in steady state, Treg undergo rapid turnover in vivo compared with Tconv (15–17). Therefore, immunosuppressive agents that interfere with homeostatic proliferation (HP) of Treg might reduce Treg numbers and interfere with tolerance. This may help explain why cyclosporin A, which inhibits IL-2 production and reduces Treg, interferes with tolerance in many allograft models (18–20).

Rapamycin (Rapa) is a potent immunosuppressive drug used clinically to prevent transplant rejection. Rapa inhibits proliferation of many cell types, including T cells, through inhibition of mammalian target of Rapa (mTOR), a serine/threonine kinase involved in regulation of cell proliferation, adhesion, and survival in response to activation and growth factor signaling (21–23). However, not all T cells are equally sensitive to Rapa. As initially shown by Battaglia et al. (24–26), addition of Rapa to in vitro cultures of murine or human CD25⁺ CD4 cells results in preferential expansion and increased recovery of Foxp3⁺ cells with regulatory activity. Moreover, in contrast to cyclosporin A, Rapa appears to maintain or induce Treg in vivo and promotes tolerance in both transplant and autoimmune models when combined with costimulatory blocking agents (19, 20, 27–29). Further, in vitro studies show that proliferating Treg are less dependent on mTOR pathways than Tconv, providing a biochemical basis for the resistance of Treg to Rapa (30, 31). In fact, direct examination of Treg homeostasis in a GVHD model revealed that Rapa blocked proliferation and expansion of Tconv but not Treg (30). Finally, in GVH and lymphopenic settings, Rapa was reported to directly increase Treg through de novo conversion of Foxp3⁻ Tconv cells (28). Taken together, the literature suggests that Treg are resistant to mTOR inhibition and that Rapa preserves Treg homeostasis and increases Treg number in both auto- and alloimmune models.

Although Rapa promotes outgrowth of Treg from in vitro cultures contaminated with Tconv cells, the in vivo requirements for mTOR in homeostatic and Ag-driven expansion of Treg may differ. In this regard, in vivo effects of Rapa have frequently been reported as the percentage and not actual number of Treg. Moreover, in some studies, the results may be confounded by assessments performed after Rapa therapy was stopped. Finally, quantitative assessment of Treg homeostasis or induction were performed in GVH or lymphopenic models, in which the degree of stimulation, competition for cytokines, and regulation of HP may differ from typical organ transplant models in which Ag exposure is more restricted, and lymphopenia is not pronounced. Indeed, an increase in the percent of Treg can occur in the face of significant but disproportionate depletion of both Treg and Tconv (32). Although an increased Treg/Tconv ratio might still promote allograft survival, this outcome is biologically distinct from direct induction and expansion of Treg.

Based on these concerns, we readdress in this article the in vivo effects of Rapa on homeostatic and alloantigen-driven proliferation, survival, number, and induction of Treg in replete wild-type (wt) mice in the presence or absence of a skin allograft. We found that unlike the lymphopenic or GVH setting, in replete mice, Rapa therapy dramatically reduces both Treg homeostatic and alloantigen-driven proliferation and promotes apoptosis. Moreover, in this setting, Rapa does not induce conversion of Tconv into Treg. As a result, Rapa significantly reduces Treg number. Interestingly, Treg do recover Ag responsiveness more rapidly than Tconv once Rapa is withdrawn. Despite this partial resistance to Rapa, mTOR signaling plays a critical role in Treg homeostasis in lymphocyte-replete recipients. These results have important implications for the use of Rapa with other agents that are intended to expand Treg number in vivo.

Materials and Methods

Mice and treatment regimens

Sex-matched 6–12-wk-old C57BL/6 (B6; H-2b) were from the National Cancer Institute, and B6.PL-Thy1a/Cy (B6 Thy1.1; H-2b) and BALB/c (H-2d) were from The Jackson Laboratory. Foxp3-red fluorescent protein (RFP) knockin mice (FIR mice; B6, H-2b) were as previously described (3). Mice were fed ad libitum and maintained according to the guidelines of the Yale Animal Research Committee and all procedures were approved by the Institutional Animal Care and Use Committee. Rapamycin (Alexis Biochemicals) was dissolved in DMSO and diluted in PBS. Rapa was administered according to similar regimens in the literature (28, 30, 32) as follows: daily Rapa (1.25 mg/kg/d i.p., days 0–9, analysis day 10); or alternate day Rapa (3 mg/kg/d i.p., days 0, 1, 2, 3, 5, and 7, analysis day 11). As indicated, the alternate day protocol was extended to provide an extra dose on day 9.

T cell transfer and CFSE-based cell proliferation

Splenocytes isolated from congenic C57BL/6 (Thy1.1) mice (1 × 10⁷ /ml) in PBS were stained with CFSE (2.5 μm; Invitrogen Life Technologies) for 10 min at 37°C. CD4⁺ cells were then enriched (~85% purity) by negative selection with BioMag immunomagnetic beads (Qiagen), as we previously described (33). A total of 10⁷ cells were adoptively transferred to gender-matched C57BL/6 (Thy1.2) recipients by i.v. injection. The proliferation of transferred CD4⁺Foxp3⁺ and CD4⁺Foxp3⁻ was calculated by the proportion of cells that exhibited dilution in CFSE fluorescence intensity in the respective gates.

In vivo Treg conversion

CD4⁺RFP⁻ (Foxp3⁻) cells (< 0.1% CD4⁺RFP⁺) were FACS-sorted from naive FIR mice (Thy1.1). A total of 1 × 10⁷ Foxp3⁻ cells was injected i.v. into C57BL/6 (Thy1.2) hosts. After cell transfer, C57BL/6 mice were either left untreated or received alternate day Rapa. On day 11, spleen and lymph node (LN) cells from C57BL/6 hosts were analyzed by FACS. Conversion was examined by percentage and number of transferred (Thy 1.1⁺) CD4 cells expressing Foxp3.

Skin transplantation and thymectomy

FACS-sorted Thy 1.1⁺ CD4⁺RFP⁻ (Foxp3⁻) or magnetic bead-enriched Thy 1.1⁺ CD4 cells were transferred by tail vein injection into C57BL/6 wt mice on day −1. One day later (day 0), mice were transplanted with allogeneic tail skin from BALB/c donors. Mice underwent thymectomy as described (12). Briefly, a Pasteur pipette was introduced into the mediastinum through a sternotomy, and the thymus was removed by suction.

Harvesting lymphocytes from nonlymphoid organs

Blood was collected from the retro-orbital sinus. Lungs were removed following cardiac perfusion with PBS, minced, digested in RPMI 1640 containing 1 mg/ml collagenase IV (Sigma-Aldrich) and 0.02 mg/ml DNASE I (Roche), and stirred at 37°C for 60 min. Livers were mechanically disrupted in PBS. Cell suspensions from liver and lung were passed through a 70-μm cell strainer. Lymphocytes in blood, lung, and liver were isolated using Lympholyte M (Cedarlane Labs).

Flow cytometry

After sacrifice, single-cell suspensions of cells from thymus, spleen, LN, as well as lymphocytes isolated from blood, lung, and liver were subjected to multicolor immunofluorescence phenotypic analysis using biotin or fluorochrome-conjugated mAbs against CD4, CD25, Foxp3, CD90.1, CD90.2, B220, CD11b, CD11c, CD16/32, CD8, NK1.1, and Ig control Abs (from BD Biosciences or eBioscience). For secondary staining of biotin-conjugated Abs, streptavidin-Pacific blue (Invitrogen) was used. Foxp3 was determined by intracellular analysis after fixation and permeabilization using Alexa 647-conjugated anti-mouse Foxp3 (eBioscience). Ethidium monoazide bromide (Invitrogen) was used to exclude dead cells. Incorporation of the nucleoside analog 5-ethynyl-2′-deoxyuridine (EdU; Invitrogen; 100 μg i.p. for 1 or 3 d) was used to detect proliferation of endogenous cells on day 5 or 10, respectively. Anti-activated caspase 3 (BD Biosciences) was used to detect cells undergoing apoptosis. Immunofluorescence analysis was performed on an LSRII (BD Biosciences), and 1–2.5 × 10⁶ events were acquired per sample. As indicated, endogenous CD4 cells and adoptively transferred congenic CD4 cells in spleen and LN were assessed by flow cytometry for Foxp3 expression after exclusion of non-lymphocytes (using a mixture of mAbs reactive to CD11c, CD 11b, B220, CD16/32, and NK1.1). Data were analyzed using FlowJo software (Tree Star). Quantitative cell numbers were calculated according to total cell counts recovered from individual lymphoid compartments in each mouse.

Statistics

All data were presented as the mean ± SE. Student t test for comparison of means was used to compare groups. p < 0.05 was considered statistically significant.

Results

Daily Rapa significantly depletes both Treg and Tconv in naive and transplanted wt mice

Rapa has been examined in a variety of allograft models employing doses ranging from 0.2 mg/kg daily to 3.0 mg/kg on alternating days (19, 28). Dosed at 1.5 mg/kg/d, Rapa had no effect on Treg number or expansion in a bone marrow transplant/GVHD setting (30). We sought to determine the effect of Rapa on Treg homeostasis in fully replete wt recipients, a setting relevant to typical organ transplantation. wt C57BL/6 mice, in the presence or absence of an allograft, received daily Rapa (1.25 mg/kg, days 0–9) and were compared on day 10 to untreated controls. Treatment of naive (untransplanted) mice with daily Rapa resulted in a reduction in the size and weight of LNs and spleen, and the total number of cells isolated from either site was decreased by 40–60% (data not shown). As seen in Fig. 1A (no allograft), Rapa has no significant effect on the percent of endogenous Treg in the CD4 population compared with untreated controls. However, examination of cell number reveals that Rapa depletes both Treg and Tconv cells in spleen by ~60% (Fig. 1B). A similar average reduction in both CD4 subpopulations occurred in LN, but was not statistically significant because of variability in the control mice. Thus, in untransplanted mice, Rapa significantly depletes both Treg and Tconv populations to a similar degree. To determine the kinetics of T cell depletion, wt C57BL/6 mice were treated with daily Rapa and examined at various time points. As shown, Rapa progressively depletes both Treg and Tconv in spleen with time, starting from day 5 (Fig. 1C). Similar results were obtained in LN (not shown).

FIGURE 1.

Daily Rapa similarly depletes endogenous Tconv and Treg. wt C57BL/6 mice were untransplanted or received BALB/c skin allografts (day 0). Mice were untreated or received Rapa (1.25 mg) on days 0–9. A, Representative fluorescence histograms of Foxp3 expression in CD4⁺ cells (day 10). The numbers depict the average percent (± SEM). B, Number of Treg (CD4⁺Foxp3⁺) and Tconv (CD4⁺Foxp3⁻), relative to untreated controls. n = 4–6 mice per group in two independent experiments. C, Kinetics of T cell depletion in spleen of Rapa-treated wt C57BL/6 mice. Cell numbers of Treg and Tconv relative to untreated controls are shown. n = 3–7 mice/group in two independent experiments. D, Effect of Rapa on number of Treg and Tconv in peripheral blood, thymus (single-positive CD4⁺CD8⁻ cells), and nonlymphoid organs of wt C57BL/6 mice (day 10). n = 5 mice/group in two independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001 versus untreated.

To determine whether the presence of alloantigen alters the Treg depletion, control and Rapa-treated mice received skin allografts. The presence of a skin allograft did not meaningfully alter the results from those obtained in untransplanted mice. Rapa induced a significant reduction of both Treg and Tconv numbers in spleen (Fig. 1A, 1B). Thus, daily Rapa depletes both Tconv cells and Treg to a similar degree in wt mice and does not confer Treg with an absolute or relative benefit.

Daily Rapa significantly depletes Treg and Tconv in nonlymphoid as well as lymphoid tissues

Like effector/memory T cells, Treg can be detected in various tissues (34, 35). To determine whether Rapa alters cell migration or the distribution between nonlymphoid and lymphoid tissues, we also examined Treg and Tconv in nonlymphoid organs, such as lung and liver. As shown, daily Rapa similarly depletes the number of both Treg and Tconv in lung and liver as well as blood and thymus (Fig. 1D). Thus, Rapa broadly depletes Treg rather than causing a redistribution of Treg between lymphoid and nonlymphoid tissues.

Daily Rapa inhibits both homeostatic and Ag-driven proliferation of Treg in vivo

Whereas Tconv cells undergo limited HP in replete wt mice (5–10% in 10 d), Treg undergo a much higher rate of HP (35–50% in 10 d) to maintain steady-state levels in the periphery (17). The implication is that higher proliferation rates are balanced by higher rates of cell death. Moreover, both Treg and Tconv cells undergo Ag-mediated proliferation (1). To better understand the consequences of Rapa treatment on CD4 cell populations, we examined the effect of Rapa on the proliferation of both Treg and Tconv in the absence of exogenous Ag (HP) and in the presence of alloantigen (Ag-induced proliferation).

wt C57BL/6 mice received an adoptive transfer of CFSE-stained congenic (Thy1.1) CD4 cells followed by treatment with Rapa (1.25 mg/kg/d, as above). As seen in Fig. 2A, Rapa significantly inhibits HP of Treg (from 34.1% down to 19.1%; p < 0.01) and Tconv (10.6% to 3.1%; p < 0.001). Whereas Rapa reduces HP of Tconv cells by 71%, it reduces HP in Treg by 44%, indicating that Treg do exhibit partial resistance. However, because of the greater reliance of Treg than Tconv on HP, Rapa has a somewhat greater effect on reducing the number of transferred Treg recovered from spleen (comparing white bars, there is a 3.3-fold reduction of Treg versus 2.5-fold reduction of Tconv; Fig. 2B). Similar findings were observed in LN (not shown). Although the relative effect of Rapa is greater on proliferating cells, Rapa also has a potent effect, depleting both Treg and Tconv that did not undergo proliferation.

FIGURE 2.

Daily Rapa inhibits homeostatic and Ag-driven proliferation and induces apoptosis of both Treg and Tconv. wt C57BL/6 mice received an adoptive transfer of 10⁷ CFSE-stained congenic (Thy 1.1⁺) CD4 cells on day −1. Mice with or without a BALB/c skin allograft (day 0) were untreated or treated with Rapa (1.25 mg) on days 0–9. Spleen cells were assessed on day 10. A, CFSE-fluorescence profiles of adoptively transferred Treg and Tconv in the presence and absence of Rapa. The numbers in each histogram reflect average percent of cells proliferating (± SEM) for each experimental group, as determined by the proportion of gated cells exhibiting decreased CFSE fluorescence. B, Number of total transferred and proliferating Treg and Tconv in the presence and absence of Rapa. Numbers in each group were calculated based on the number of adoptively transferred CD4 cells recovered and the percent undergoing proliferation above. n = 4–6 mice per group in two independent experiments. C, Effect of Rapa on proliferation of endogenous Treg and Tconv in the presence or absence of Rapa (1.25 mg/d for 5 d) in spleen was assessed by EdU incorporation. n = 3 mice per group. Similar results were obtained on day 3; not shown). D, Effect of Rapa on apoptosis in Treg and Tconv. wt C57BL/6 mice were untreated or treated with Rapa (1.25 mg on days 0–4 or 3 mg on days 0–3), and apoptosis of CD4 cells in spleen (day 5) was assessed by detection of active caspase 3 staining. n = 3 mice per group. *p < 0.05, **p < 0.01, ***p < 0.001 versus untreated (similar results were obtained on day 3; not shown).

Exposure of mice to alloantigen increased proliferation of both Treg and Tconv (Fig. 2A). Approximately 50% of Treg underwent proliferation in the presence of a skin allograft, representing nearly a 1.5-fold increase from baseline HP. By comparison, Tconv cells underwent a 2-fold increase in proliferation (from 10.6% to 22.3%; Fig. 2A). Rapa significantly inhibits proliferation of both CD4 subsets after transplantation, but inhibits proliferating Tconv to a greater extent than Treg (73% versus 36% reduction, respectively). However, as in untransplanted mice, Rapa markedly reduces the recovery of adoptively transferred Treg as well as Tconv (Fig. 2B). Similarly, Rapa inhibits proliferation of endogenous Treg and Tconv cells (Fig. 2C). Thus, unlike irradiated recipients of allogeneic bone marrow (30), administration of daily Rapa to wt mice significantly depletes Treg and Tconv cells through inhibition of both homeostatic and Ag-driven proliferation. The modest resistance of Treg to the antiproliferative effects of Rapa does not protect them on a population basis from in vivo depletion because they rely so heavily on proliferation to maintain the peripheral T cell pool.

Daily Rapa depletes Treg and Tconv cells through apoptosis

The studies above reveal that in addition to a decrease in cell proliferation, Rapa causes a significant depletion in nonprolife-rating Tconv and Treg. One explanation for this finding is the induction of cell death by Rapa. Of note, cells that have undergone apoptosis and clearance may appear as a decrease in either proliferating and/or nonproliferating cells in the assessment above. To determine whether Rapa promotes apoptosis of Treg and Tconv, wt mice were treated with Rapa followed by assessment of activated caspase 3 on day 5. Rapa administered either at 1.25 mg/kg/d or at 3.0 mg/kg (days 0–3) significantly increased the percentage of Treg and Tconv cells undergoing apoptosis (Fig. 2D). EdU incorporation indicated that apoptosis occurred in both proliferating and nonproliferating Treg (35% and 65%, respectively, independent of treatment), whereas it was detected only in nonprolife-rating Tconv (not shown). Importantly, in this study, we directly demonstrate that Treg exhibit a >10-fold higher basal rate of apoptosis than Tconv cells. This is consistent with their higher basal rate of HP and cell turnover at steady state.

Alternate day Rapa selectively preserves Treg in the presence of an allograft

We next asked whether the effect of Rapa on Treg HP and Ag-driven proliferation was dependent on dose and timing. In this regard, Gao et al. (28) reported that Rapa 3.0 mg/kg for 3 d, and then on alternating days until day 14, resulted in de novo induction of Treg in lymphopenic recipients of skin allografts examined 4 d after cessation of therapy. Because control wt recipients used in our study completely reject skin allografts by day 14, the protocol was altered so that mice could be examined on day 11 after a similar 4-d withdrawal from Rapa. wt C57BL/6 mice, with or without BALB/c skin allografts, were treated with alternate day Rapa (3 mg/kg on days 0, 1, 2, 3, 5, and 7) and analyzed on day 11.

Alternate day Rapa significantly increased the percentage of Treg among CD4 cells in both spleen (21.5 versus 14.2%, p <0.01) and LN (21.3 versus 15.4%, p < 0.05) of allograft recipients (Fig. 3A). Although the number of endogenous Treg in both spleen and LN was similar to untreated control recipients, Rapa significantly reduced Tconv cells, giving rise to the increased Treg percent noted above (Fig. 3A, 3B). However, the results were quite different in the absence of a skin allograft. In this setting, alternate day Rapa had no effect on the percentage of Treg in either spleen or LN and depleted both Treg and Tconv in spleen and LN (Fig. 3A, 3B). Thus, in this alternate dosage regimen, Rapa preserves Treg, but only in the presence of an allograft. Tconv cells were depleted by Rapa either in the presence or absence of an allograft.

FIGURE 3.

Alternate day Rapa maintains Treg only in the presence of an allograft. wt C57BL/6 mice were untransplanted or received BALB/c skin allografts (day 0). Mice were untreated or received alternate day Rapa (3 mg/) on days 0, 1, 2, 3, 5, and 7. Spleen and LN cells were assessed on day 11. A, Foxp3 expression in CD4 cells in the presence and absence of Rapa. Representative fluorescence histograms are shown. The numbers depict the average percent (± SEM) Foxp3 expression in CD4⁺ cell for each experimental group. B, Number of Treg and Tconv relative to untreated controls. n = 5 to 6 mice per group in two independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001 versus untreated.

Treg proliferation after alternate day Rapa is rescued by the presence of an allograft

To better understand homeostasis of Treg and Tconv cells after alternate day Rapa, wt C57BL/6 mice received adoptively transferred CFSE-stained congenic CD4 cells 1 d before transplantation and initiation of treatment. In the absence of skin graft, alternate day Rapa reduced the percent of both Treg and Tconv undergoing HP (Fig. 4A). In the presence of a skin graft, alternate day Rapa followed by the 4-d withdrawal restored the proliferative capacity of Treg to Ag (42.9 versus 46.0%, NS; Fig. 4A). In contrast, the proliferative response of Tconv cells to alloantigen remained significantly reduced (46% inhibition). Consistent with the effect on endogenous cells (Fig. 3) and the proliferation data above, the recovery of adoptively transferred Treg, and specifically those undergoing proliferation, was restored in Rapa-treated mice by the presence of an allograft (Fig. 4B). In contrast, this Rapa regimen reduced both proliferating and overall recovery of Tconv in either the presence or absence of an allograft. Similar findings were observed in LN (not shown). Thus, alternate day Rapa followed by a 4-d hiatus preserves Treg, largely due to Ag-driven proliferation in the presence of an allograft. Despite the presence of an allograft, this regimen preferentially reduces both percentage and number of Tconv cells.

FIGURE 4.

The presence of an allograft rescues Ag-driven proliferation of Treg in mice treated with alternate day Rapa. wt C57BL/6 mice received an adoptive transfer of 10⁷ CFSE-stained congenic (Thy 1.1⁺) CD4 cells (day −1). Mice with or without a BALB/c skin allograft were untreated or received alternate day Rapa (3 mg/) on days 0, 1, 2, 3, 5, and 7. Spleen cells were assessed on day 11. A, CFSE-fluorescence profiles of adoptively transferred Treg and Tconv in the presence and absence of Rapa. The numbers in each histogram reflect average percent of cells proliferating (± SEM) for each experimental group, as determined by the proportion of gated cells exhibiting decreased CFSE fluorescence. B, Number of Treg and Tconv cells undergoing proliferation in the presence and absence of Rapa. The number of proliferating and nonproliferating cells in each group was calculated based on the number of adoptively transferred CD4 cells recovered and the percent undergoing proliferation above. *p < 0.05, **p < 0.01, ***p < 0.001 versus untreated. n = 4–6 mice/group in two independent experiments.

Treg recovery is due to Rapa withdrawal and is not a feature of alternate day Rapa dosing

In lymphopenic recipients of skin grafts, alternate day Rapa followed by a 4-d hiatus led to de novo induction and increased Treg (28). In replete mice, we observed that a similar Rapa regimen augmented the percent and maintained baseline numbers of Treg, but only in the presence of an allograft. Indeed, this Rapa dosing regimen allowed for selective expansion of Treg in response to alloantigen, not observed when mice were treated with Rapa at 1.5 mg/kg daily until the day of analysis. This suggests either differential effects on Treg of daily versus alternate day Rapa or, more likely, that Treg exhibit a more rapid functional recovery of proliferative capacity than Tconv, once Rapa has been discontinued. To test this hypothesis, wt allograft recipients were treated with an alternate day Rapa regimen that included one extra dose, ending treatment on day 9 rather than day 7. Analysis was performed on day 11, as before. The percent and number of Treg obtained in spleen from mice receiving alternate day Rapa with 4 versus 2 d since their last dose were compared.

Whereas alternate day Rapa with a 4-d hiatus does not change Treg number from the baseline in untreated controls, allograft recipients treated with one extra Rapa dose (2-d hiatus) exhibit a 50% decrease in Treg number (Fig. 5A). In contrast, Tconv cells remain reduced in number whether Rapa is withheld for 2 or 4 d. These data suggest that Treg recover more rapidly than Tconv from the suppressive effects of Rapa, resulting in an increase in the percentage of Treg after a 4-d withdrawal that is not seen when Rapa is continued on an alternate day basis (2 d withdrawal) in which Tconv and Treg remain similarly depleted (Fig. 5A). To confirm the more rapid recovery of Treg, we examined the proliferation of endogenous Treg and Tconv cells in the setting of Rapa withdrawal by administering EdU on days 8–10. EdU incorporation reveals a significant increase in proliferation of Treg after a 4-d compared with a 2-d hiatus from Rapa, whereas the proliferation of Tconv remained reduced even after a 4-d hiatus (Fig. 5B). Thus, although both daily and alternate day Rapa significantly deplete Treg in vivo, Treg exhibit a more rapid recovery of Ag responsiveness than Tconv after withdrawal of Rapa, promoting the recovery of Treg and a beneficial Treg/Tconv ratio.

FIGURE 5.

Baseline levels of Treg in allograft recipients treated with alternate day Rapa is due to Treg recovery during the 4-d withdrawal period. wt C57BL/6 recipients of BALB/c skin grafts were untreated or treated with Rapa (3 mg) according to the alternate day regimen (days 0, 1, 2, 3, 5, and 7) or an extended regimen (days 0, 1, 2, 3, 5, 7, and 9). Analysis of the number of Treg and Tconv in spleen was performed on day 11 after a 4-d or 2-d hiatus. A, Number of Treg and Tconv and percentage of cells expressing Foxp3⁺ among CD4 cells recovered from spleen on d 11 in each treatment group. n = 7–13 mice/group in three independent experiments. B, Number of proliferating Treg and Tconv based on EdU incorporation by CD4 cells recovered in each group. n = 4 mice/group. *p < 0.05; **p < 0.01; ***p < 0.001.

Alternate day Rapa does not induce de novo conversion of Tconv into Treg in replete wt mice

In addition to proliferation, another factor that could contribute to the maintenance of Treg numbers in transplanted mice treated with alternate day Rapa is de novo conversion from Tconv cells. In this regard, using a similar alternate day Rapa regimen in lymphopenic recipients of skin allografts, Gao et al. (28) showed induction of Treg from adoptively transferred Tconv cells. To determine whether alternate day Rapa induces conversion of Tconv to induced Treg in replete wt recipients, FACS-sorted CD4⁺Foxp3⁻ cells from naive Foxp3-reporter mice were transferred into congenic C57BL/6 hosts on day −1. BALB/c skin was transplanted the following day, and mice were treated with Rapa (as above). In the absence of Rapa, 1.3% of transferred CD4 cells isolated from spleen and 3.0% of those isolated from draining LN now expressed Foxp3 (Fig. 6). These cells are most likely derived from expansion of a miniscule contaminating population of Treg (<0.1%) present in the initial adoptive transfer or from low-level conversion from Foxp3⁻ cells. Treatment with alternate day Rapa followed by a 4-d withdrawal had no effect on either the percentage or actual number of Treg isolated from draining LN or spleen. The failure of this regimen to induce de novo conversion is also consistent with the failure of Rapa to augment the number of Treg among adoptively transferred CD4⁺ cells (Fig. 4).

FIGURE 6.

Rapa does not induce conversion of Treg from Tconv in skin graft recipients. Total of 10⁷ FACS-sorted CD4⁺Foxp3⁻ cells from naive congenic (Thy1.1⁺) Foxp3-reporter mice were adoptively transferred into wt C57BL/6 hosts on day −1. Mice received a BALB/c skin allograft on day 0. Mice were untreated or treated with alternate day Rapa (3 mg/) on days 0, 1, 2, 3, 5, and 7. Cells were harvested from spleen and draining LN for analysis on day 11. Representative dot plot of Foxp3 expression gated on transferred (Thy1.1+) CD4 cells are shown. The numbers in Foxp3⁺ gate represent the average percentage (± SEM) of transferred CD4 cells expressing Foxp3. The numbers in parentheses represent the average number (± SEM) of Foxp3⁺ cells recovered from each site. n = 5 mice per group in two independent experiments. *p < 0.05 versus untreated.

Thymic output does not contribute to the recovery of peripheral Treg after 4-d withdrawal from Rapa

Thymic output could also contribute to recovery of Treg observed after a 4-d withdrawal from Rapa, particularly because previous studies have shown that Rapa increases the percent of Treg in thymus (19, 32). In agreement, we found that alternate day Rapa does increase the percent of Treg in thymus (Fig. 7A). However, this occurs because Rapa dramatically depletes thymocytes, but depletes thymic Treg to a lesser extent than Tconv (Fig. 7B). To determine whether the increased frequency of Treg in thymus contributes to the increase in peripheral Treg after a 4-d withdrawal from Rapa, skin transplants were performed on thymectomized mice. As shown, thymectomy did not affect the recovery of Treg (or Tconv) in the spleens of mice treated with alternate day Rapa (4-d withdrawal) (Fig. 7C). Thus, the recovery of Treg after Rapa withdrawal appears to result from increased proliferation in the face of allogeneic stimulation.

FIGURE 7.

Thymic output does not contribute to the recovery of Treg by allograft recipients 4 d after withdrawal of Rapa. wt C57BL/6 mice recipients of BALB/c skin allografts (day 0) were untreated or received alternate day Rapa (3 mg; days 0–3, 5, and 7). Thymocytes were assessed on day 11. A, Percentage of Foxp3⁺ cells in CD4⁺CD8⁺ and CD4⁺CD8⁻ populations in each group. B, Number of Foxp3⁻ and Foxp3⁺ cells in CD4⁺CD8⁺ and CD4⁺ CD8⁻ populations in each group. n = 5 mice/group in two independent experiments. C, Thymectomy (TXM) was performed on wt C57BL/6 mice 7 d before transplantation (day 0) with BALB/c skin allografts. Rapa was administered as above. The number of Treg and Tconv cells in spleen were assessed on day 11. Data are representative of two independent experiments. n = 3–5 mice/group. *p < 0.05, **p < 0.01, ***p < 0.001 versus untreated.

Discussion

mTOR is a serine-threonine kinase lying downstream of growth factor and receptor-mediated signaling pathways (21, 23, 36). mTOR complexed to Raptor (TORC1) lies downstream of PI3K and AKT and regulates cell proliferation by promoting protein translation through activation of ribosomal S6K and inhibition of the translational repressor 4EBP1. mTOR also complexes with Rictor (TORC2) to promote cell survival through feedback activation of AKT and its downstream survival pathways. Rapa inhibits mTOR/TORC1, reducing translation and cell proliferation. In contrast, inhibition of TORC2/AKT by Rapa is variable and cell-type dependent (36). Rapa has generated considerable interest because unlike other approved immunosuppressive drugs such as cyclosporin A, Rapa synergizes broadly with experimental tolerogenic agents (29). Initial studies attributed this synergy to Rapa-induced apoptosis of alloreactive T cells that were not inhibited by costimulatory blockade (37). Subsequently, it was discovered that ex vivo, Rapa preferentially inhibits proliferation of Tconv, promoting outgrowth of Treg (24, 25, 38). Further studies showed that Rapa treatment was also associated with an increase in Treg in vivo in both autoimmune and transplant models (19, 27). Finally, in the setting of GVH and lymphopenia, Rapa did not inhibit Treg expansion and could actually increase Treg through conversion of Tconv cells (28, 30).

Together, the studies above support a widely held view that Rapa promotes tolerance through its ability to maintain and even augment Treg numbers while inhibiting Tconv (29, 39). As a result, new therapeutic strategies designed to enhance the number of Treg in vivo frequently include Rapa. In contrast, we demonstrate in this article that in replete mice, Rapa inhibits both Ag-driven and HP of Treg and promotes Treg apoptosis, leading to their significant depletion. Although these results do not negate the ability of Rapa to prolong allograft survival, they require reinterpretation of the biological effects of Rapa on Treg in vivo and have important implications for use of Rapa in regimens aiming to promote tolerance through increased Treg.

How can our results be rectified with previous reports in the literature? First, ex vivo expansion of human Treg for therapeutic purposes has been limited by the inability to positively identify and expand pure populations of Treg in vitro (8). Rapa promotes outgrowth of Treg from such cultures, largely through preferential inhibition of contaminating Tconv (25, 26). Differential use of signaling pathways, at least in vitro, appears to underlie the partial resistance of Foxp3⁺ Treg to Rapa. Treg are heavily dependent on IL-2R signaling, but they rely more on downstream signals through STAT5 and are less dependent on AKT and mTOR for proliferation and survival than Tconv (26, 30, 31, 40, 41). Although Treg may exhibit resistance to mTOR inhibition in vitro, this differs substantially from the in vivo setting in terms of activation signal strength, costimulatory milieu, competition for cytokines, and dependence on homeostatic proliferation. Thus, it is not surprising that the degree of resistance to mTOR inhibition in vivo and in vitro differ. Similarly, our study in replete mice contrasts with studies performed in GVHD and lymphopenic models. Zeiser et al. (30) directly demonstrate that Rapa has no effect on expansion of Treg adoptively transferred into irradiated recipients of allogeneic bone marrow transplants. In this setting, allogeneic Treg respond to ubiquitous (GVH) Ag in a lymphopenic host. Irradiation promotes maturation of dendritic cells and may alter exposure to cytokines such as IL-7 and IL-15 (42, 43). Competition for specific niches, costimulation, and cytokines and intensity of stimulation dramatically differ from those in a replete organ allograft recipient and are likely to alter Treg responsiveness and sensitivity to Rapa.

In contrast to the studies above, in nonlymphopenic settings more typical of organ transplantation or autoimmune disease, Rapa significantly depletes Treg. Compared to Tconv cells, Rapa inhibits Treg HP and Ag-induced proliferation and augments apoptosis to a smaller degree than Tconv cells in the same mice. Thus, Treg do exhibit relative resistance to Rapa. However, the high basal rate of Treg HP implies that Treg also undergo a high rate of cell death, to our knowledge, documented in this study for the first time. Given the high rate of Treg turnover, anything that interferes with proliferation or augments cell death will disproportionately deplete Treg. Because the loss in Treg in our study is matched by similar depletion of Tconv, the percent of Treg is essentially unaltered. In studies in which the Treg were assessed on a percentage basis, the extent of Treg depletion was not appreciated. This includes not only some earlier murine studies, but also applies to human studies, in which the percent of Treg in peripheral blood may not directly reflect the actual number of Treg in secondary lymphoid organs or tissues (19, 27, 38, 44).

Interestingly, when Rapa is withdrawn, peripheral Treg recover their proliferative capacity to alloantigen more quickly than do Tconv. Thus, Treg numbers can reach or even exceed baseline levels, and, at least temporarily, Treg are increased on a percentage basis. This proliferation and recovery is strictly alloantigen dependent. This may help explain findings in which increased Treg were observed shortly after Rapa therapy was stopped (28). To the extent that Rapa depletes Tconv in excess of Treg, Rapa can be said to enhance tolerance through regulatory mechanisms. However, this is biologically distinct from induction and enhancement of Treg numbers.

Another notable finding is that (4-d) withdrawal of Rapa results in a significantly higher number of proliferating cells than that seen in the presence of ongoing alternate day Rapa (2-d hiatus), but proliferation is also higher than that seen in untreated allograft recipients (Fig. 5B). In fact, even in mice treated with alternate day Rapa (2-d hiatus), Treg proliferation was the same as that in untreated recipients. This appears to be at odds with CFSE data showing that Rapa has no effect on Treg proliferation in transplant recipients (4-d hiatus) and reduces Treg proliferation in other settings (Figs. 2B, 4B). We previously showed that Treg HP measured by CFSE dilution and BrdU labeling give similar results in naive mice (17). However, CFSE measures proliferation over the entire 11-d course of the experiment, whereas EdU labeling measures only the last 3 d of proliferation. Thus, Rapa may inhibit proliferation of Treg proliferation initially, contributing to decreased Treg number. Subsequently, Treg proliferation may rebound due to a lower number of Treg competing for IL-2 and Ag, ultimately establishing a new steady state. In support of this, pulsing allograft recipients with EdU reveals that Rapa inhibits Treg proliferation on days 3–5 by ~50% (Fig. 2C). Neither thymic output nor conversion of Tconv cells into Treg contributed to the recovery of Treg after withdrawal of Rapa. Indeed, in contrast to Gao et al. (28), we could not demonstrate induction of Foxp3⁺ expression in Foxp3⁻ cells transferred into Rapa-treated wt allograft recipients. Similar Rapa regimens were used in both studies, including a 4-d hiatus prior to analysis. Our findings suggest that although Rapa might enhance conversion of Foxp3⁻ into Foxp3⁺ cells in lymphopenic hosts, it does not do so in replete wt recipients. It is also possible that in the setting of lymphopenia and alloantigenic stimulation, relative resistance to Rapa allows outgrowth of a small number of contaminating Treg that contribute to the apparent conversion observed. Interestingly, Gao et al. (28) also observed dramatic induction of Foxp3 by Rapa, in a non-lymphopenic GVHD-like model, in which Foxp3⁻ cells were placed into haplo-mismatched F1 hosts. Unlike the long-lived Foxp3⁺ Treg observed on days 18 and 30 in lymphopenic skin graft recipients, in this GVHD model, Rapa-induced Foxp3⁺ cells peaked on day 4 and decreased to half that number (in spleen) by day 7. These marked differences in tempo and durability of Foxp3 induction support our hypothesis that both lymphopenia and the degree of stimulation/abundance of Ag and likely differences in cytokine availability have profound effects on the response of Treg to mTOR inhibition.

Our observations provide new insights into Treg biology and have important implications for use of Rapa in tolerogenic settings. First, the resistance of Treg to Rapa is both partial and conditional and may depend on competition for cytokines, cellular niches, and strength of activation. Second, we can no longer directly extrapolate results from in vitro to in vivo experiments or among GVH, lymphopenic, and replete allograft models. The latter setting better represents most organ allograft recipients and patients with auto-immune disease who are not profoundly lymphopenic. Third, in rodent models, Rapa is frequently given as a short-term agent that enhances tolerance induced by other agents. In primates, Rapa is usually given chronically, which could significantly affect Treg induction and homeostasis. Our findings have important practical implications for tolerogenic strategies aimed at increasing Treg. Perhaps delaying administration of induction agents until a short course of Rapa is completed might further augment Treg. Finally, further study will be required to better define the signals controlling Treg induction and homeostasis in different settings in vivo, and these are likely to provide additional insights into Treg biology and tolerance.

Abbreviations used in this article

EdU

5-ethynyl-2′-deoxyuridine

GVHD

graft-versus-host disease

HP

homeostatic proliferation

LN

lymph node

mTOR

mammalian target of rapamycin

Rapa

rapamycin

RFP

red fluorescent protein

Tconv

conventional T cell

TORC1

mammalian target of rapamycin complexed to Raptor

TORC2

mammalian target of rapamycin complexed to Rictor

Treg

CD4⁺Foxp3⁺ regulatory T cell

wt

wild-type