Differential Roles of IL-2 Signaling in Developing versus Mature Tregs

Generation and Validation of CD25^fl/fl × Rosa-RFP × Foxp3^{EGFP-Cre-ERT2} “CD25-iΔTreg” mice

The gene for CD25 (Il2ra) consists of 8 exons encoding 2 extra-cellular domains and 1 transmembrane domain (Malek, 2008). We generated chimeric mice using embryonic stem cells in which exon 4 (encoding one of two extracellular domains) is flanked by loxP sites (Figure S1A). Loss of this exon is known to abolish binding to IL-2 (Leonard et al., 1984). Resultant “CD25^fl/fl” mice were bred with Foxp3^YFP-Cre mice (Rubtsov et al., 2008) to confirm that exon 4 deletion yielded functional effects consistent with known knockout phenotypes. By 2.5 weeks after birth, CD25^fl/fl × Foxp3^YFP-Cre mice developed enlarged lymph nodes and thymic involution (Figure S2A) with a reduced percentage of splenic Foxp3⁺ cells (Figure S2B) and increased percentages of activated CD4⁺ CD44⁺ CD69⁺ conventional T cells (Tconv) (Figure S2C). No CD25^fl/fl × Foxp3^YFP-Cre mice survived beyond 3.5 weeks. Nearly identical phenotypes have been described in mice lacking CD25 or IL-2 entirely (Willerford et al., 1995; Sadlack et al., 1993) and in a recent study in which an independently derived CD25^fl/fl line was crossed with Foxp3^YFP-Cre (Chinen et al., 2016), confirming that deletion in the CD25^fl allele recapitulates functional loss of CD25.

To study the role of IL-2 in post-developmental Tregs, we obtained mice bearing a tamoxifen-inducible Foxp3^{EGFP-Cre-ERT2} construct (Rubtsov et al., 2010), which causes Foxp3⁺ cells to express EGFP and a Cre recombinase that only enters the nucleus after tamoxifen exposure. We also obtained a Rosa-RFP recombination reporter line (Luche et al., 2007) to label Foxp3⁺ cells present at the time of tamoxifen exposure and track potential loss of Foxp3 expression. Thus, two strains of mice were generated: (1) a CD25^fl/fl × Rosa-RFP × Foxp3^{EGFP-Cre-ERT2} strain featuring CD25-inducible deletion among Tregs, which we refer to as CD25-iΔTreg mice; and (2) a CD25^+/+ × Rosa-RFP × Foxp3^{EGFP-Cre-ERT2} control strain, which we refer to as WT mice.

By 4 days after the first tamoxifen injection, the majority of all Foxp3⁺ cells from CD25-iΔTreg mice lose CD25 (Figure S3A), with surface expression comparable to that of unstimulated Foxp3⁻ CD4⁺ Tconv (Figures S3B and S3C). Deletion was restricted to Cre-expressing cells, as Foxp3⁻ CD4⁺ and CD8⁺ T cells from CD25-iΔTreg mice expressed CD25 upon stimulation (Figure S3D). The Rosa-RFP reporter was rapidly activated, with a distinct RFP⁺ population appearing 4 days after the first tamoxifen injection (Figure S3E). 90% of RFP⁺ Foxp3⁺ cells in CD25-iΔTreg mice had deleted CD25 (Figure S3G), and we refer to these CD25⁻ RFP⁺ Foxp3⁺ cells as CD25-iΔTregs (Figure S3F). We also refer to the RFP⁺ Foxp3⁺ population in wild-type (WT) mice as WT-Tregs and further stratify these cells into WT CD25^hi and WT CD25^lo subsets. Deletion of CD25 was not always reflected by activation of the RFP reporter, as many CD25⁻ Tregs failed to express RFP (Figure S3H). However, because thymic output continues beyond the course of tamoxifen treatment, the RFP⁻ population contains an indeterminate mixture of CD25 expressing and CD25-deleted Tregs. We therefore refer to the collective RFP⁻ population as “RFP⁻ CD25^fl/fl” or “RFP⁻ CD25^+/+” Tregs.

Loss of Tregs and Maintenance of Foxp3 following CD25 Deletion

To assess how CD25 loss impacts Treg homeostasis, mice were given intraperitoneal tamoxifen to induce CD25 deletion (Figure 1A), and blood samples were obtained over a period of 10 weeks. We found that following tamoxifen injection, CD25-iΔTregs were lost with a half-life of ~2 weeks (Figure 1B). As with the above characterization studies, 90% of RFP⁺ Foxp3⁺ cells lost CD25 immediately after tamoxifen administration; the remaining 10% steadily comprise a greater percentage of the RFP⁺ Foxp3⁺ population over time, likely due to a survival advantage over RFP⁺ CD25-deleted Tregs (Figure 1C). No differences were observed between male and female mice (data not shown). In both WT and CD25-iΔTreg mice, the total percentage of Foxp3⁺, CD4⁺, and CD8⁺ T cells remained the same (Figures 1D, S4A, and S4B). Total Treg numbers were maintained in CD25-iΔTreg mice by new thymic output as demonstrated by increased expression of CD31, a marker of recent thymic emigrants, among RFP⁻ CD25^fl/fl Tregs (Figure 1E). These newly generated cells have not been exposed to tamoxifen, are mostly CD25^hi, and rapidly comprise the majority of the RFP⁻ CD25^fl/fl Treg population (Figure S4C).

While no differences were observed in BrdU uptake between CD25-iΔTregs, WT CD25^lo Tregs, and WT CD25^hi Tregs (Figure 1F), we found an increase in the percentage of apoptotic CD25-iΔTregs as measured by Annexin-V × 7-AAD staining (Figure 1G). RFP⁻ CD25^fl/fl Tregs, in contrast, showed increases in bromodeoxyuridine (BrdU) uptake, but not in cell death (Figures S4D and S5E). This likely reflects increased proliferation to compensate for the death of CD25-iΔTregs to maintain total Treg percentages in CD25-iΔTreg mice. Preferential sequestration could not account for the loss of CD25-iΔTregs from the blood, as all major lymphoid organs showed similar losses of CD25-iΔTregs 2 weeks following tamoxifen injection (Figures S4F-S4J). No RFP⁺ cells were observed in any organs prior to tamoxifen administration, confirming no “leakiness” of the Rosa-RFP reporter. These results implicate cell death as the primary cause for the loss of Tregs following CD25 depletion.

Foxp3⁻ CD4⁺ and CD8⁺ T cells constitute the primary sources of IL-2 in vivo (Malek, 2008; Liu et al., 2015) and are more sparsely distributed in peripheral tissues than in lymphoid organs. Since Tregs in these environments might have adapted to reduced IL-2 availability, we examined whether they were less susceptible to death following CD25 deletion. We selected the liver as one such peripheral tissue site and used another, visceral adipose tissue (VAT), for comparison due to ready availability of IL-2 from VAT-resident invariant natural killer T cell (iNKT) cells (Lynch et al., 2015). In both tissues, we observed losses of Tregs to a similar magnitude as seen in lymphoid organs following CD25 deletion (Figures S4K and S4L).

In prior studies, blocking antibodies against IL-2 or CD25 has led to reductions in Foxp3 expression among Tregs, suggesting that continuous IL-2 signaling in post-thymic Tregs is necessary to maintain Foxp3 expression (Kohm et al., 2006; Rubtsov et al., 2010). Since antibodies only temporarily block IL-2 signaling, it has not been certain whether continued IL-2 signaling is needed to maintain long-term Treg lineage stability. We found that although Foxp3 levels were reduced shortly following CD25 deletion, Foxp3 expression was maintained without further losses for the next 10 weeks (Figure 2A). Gating on RFP⁺ Tregs revealed a unimodal Foxp3 histogram in both genotypes, with no evidence for a distinct RFP⁺ FoxP3⁻ “ex-Treg” population as far as 10 weeks after tamoxifen injection (Figure 2B). Interestingly, although average FoxP3 expression among CD25-iΔTregs was ~20% lower than that of WT CD25^hi Tregs, it was similar to that of WT CD25^lo Tregs (Figures 2A and 2C).

Phenotypic Resemblance of CD25-iΔTregs to WT CD25^lo Tregs

Numerous subsets of Tregs have been described in recent years, including peripherally induced Tregs (pTregs), which derive from CD4⁺ CD25⁻ naive T cells, express lower levels of CD25 than thymically derived Tregs (tTregs), and may develop in the absence of IL-2 (D’Cruz and Klein, 2005). While it was possible pTregs might be less dependent than tTregs on IL-2 and possess a survival advantage following CD25 deletion (Malek, 2008), we did not observe any differences in Helios or Neuropilin-1 expression among CD25-iΔTregs, WT CD25^lo Tregs, and WT CD25^hi Tregs, suggesting no enrichment of pTregs among the CD25-iΔTreg population (Figures S5A and S5B). Tregs may also be divided into central (cTreg) and effector (eTreg) subsets, with the latter enriched among CD25^lo Tregs (Fisson et al., 2003; Smigiel et al., 2014). Consistent with published results, we found an increase in CD44^hi CD62L^lo cells among both WT CD25^lo and CD25-iΔTregs as compared to WT CD25^hi Tregs (Figure S5C). A greater percentage of CD25-iΔTregs expressed PD-1 and ICOS compared to WT CD25^hi Tregs (Figures 3A and 3B), while the expression of other major Treg surface molecules such as GITR and CTLA-4 was only slightly reduced in CD25-iΔTregs (Figures S5D-S5F). Similar observations were made when comparing WT CD25^lo to WT CD25^hi Tregs. For most surface molecules, CD25-iΔTregs strongly resembled WT CD25^lo Tregs (Figures 3 and S5).

Origins of Surviving CD25-iΔTregs

Because of the phenotypic similarity between CD25-iΔTregs and WT CD25^lo Tregs, we asked whether the surviving CD25-iΔTregs derived primarily from CD25^hi Tregs that had lost CD25 and adopted CD25^lo characteristics or whether they derived from preferential survival of the CD25^lo population. To test whether CD25^hi or CD25^lo Tregs could contribute to the CD25-iΔTreg population, we tracked the fate of each subset in an adoptive transfer model. CD25^hi and CD25^lo Foxp3⁺ cells were sorted from WT and CD25-iΔTreg mice not exposed to tamoxifen and then transferred in equal numbers into separate CD45.1⁺ congenic recipients. Mice received tamoxifen after 1 week (Fisson et al., 2003) and were euthanized 2 weeks later. All CD4⁺ splenocytes were collected to measure total splenic recovery of transferred CD45.2⁺ cells (Figures 4A and 4B). Nearly all cells from CD25-iΔTreg donors lost CD25 (Figures 4C and 4D). By gating out host CD45.1⁺ cells, we were able to exclude new thymic output from the CD45.2⁺ RFP⁻ population. We noted an equal ratio of RFP⁺ to RFP⁻ Tregs among cells from both CD25-iΔTreg and WT donors (Figure 4E), indicating that the expression of RFP had no effect on cell survival. Importantly, we observed that both Tregs that were initially CD25^hi and those that were initially CD25^lo could survive after CD25 deletion, indicating that both groups can contribute to the CD25-iΔTreg population (Figure 4F). However, because CD25^hi precursors showed increased survival following CD25 deletion, and because CD25^hi cells comprise the majority of all Tregs prior to tamoxifen treatment, we conclude that the majorityof CD25-iΔTregs derive from CD25 precursors based upon a survival advantage.

Reduced Suppressor Function and Glycolysis among CD25-iΔTregs

To test whether Treg suppressor function was affected by loss of IL-2 signaling, we used a rapid in vitro suppression assay in which the primary readout was expression of the activation markers CD69 and CD154 (CD40L) among target cells 7 hr after activation in co-culture (Canavan et al., 2012; Ruitenberg et al., 2011). We found that CD25-iΔTregs were less effective than WT Tregs at suppressing target cell upregulation of CD69 and CD154 (Figures 5A and 5B).

We further evaluated Treg function in vivo using an experimental autoimmune encephalomyelitis (EAE) model in which disease induction is Treg independent but disease stabilization and remission requires functional Tregs (Kohm et al., 2002). Mice were given tamoxifen both prior to EAE induction and again after 2 weeks to ensure a majority of Tregs lacked CD25 throughout the course of the experiment. Although no differences were observed in the onset or initial severity of EAE, CD25-iΔTreg mice failed to undergo the Treg-dependent disease remission phase (Figure 5C). These results indicate diminished in vivo suppressor function among Tregs in the absence of continuous IL-2 signaling.

Glycolytic metabolism is needed for full Treg suppressor function in vivo, as it promotes expression of key Treg effector molecules such as CTLA-4 and promotes Treg proliferation (Newton et al., 2016; Zeng et al., 2013; Gerriets et al., 2016). We therefore investigated the effects of IL-2 signaling loss on Treg metabolism using a Seahorse flux analyzer to measure extracellular acidification rate (ECAR), a marker of glycolytic flux. As shown in Figures 5D–5F, IL-2 enhances glycolysis in all Tregs (i.e., WT CD25^hi Tregs, WT CD25^lo Tregs, and CD25-iΔTregs), with by far the greatest increase being seen in the WT CD25^hi Treg subset. This is consistent with levels of high-affinity IL-2R expression, and we attribute the effects of IL-2 in WT CD25^lo Tregs and CD25-iΔTregs to effects mediated via the lower-affinity beta and gamma subunits of the IL-2 receptor. IL-7 did not increase glycolysis in WT CD25^hi Tregs but had small effects (i.e., similar to those seen with IL-2) in WT CD25^lo Tregs and CD25-iΔTregs. The same cytokine treatments had negligible effects on the glycolysis of Tconv (data not shown).

IL-7 Is Less Effective than IL-2 at Maintaining Downstream Signals

STAT5 signaling is known to promote Treg survival and suppressor function (Chinen et al., 2016). While IL-2 constitutes the primary source of STAT5 signals in Tregs, other common gamma chain cytokines such as IL-7 and IL-15 may also fill this role. We found that among unstimulated cells, a greater percentage of CD25-iΔTregs phosphorylated STAT5 in response to IL-7 than either WT CD25^hi or WT CD25^lo Tregs, while no differences were observed in STAT5 responses to IL-15 (Figure S6A). This was consistent with surface expression data for the relevant cytokine receptors. CD127, the alpha subunit of the IL-7 receptor, was elevated in CD25-iΔTregs relative to WT-Tregs (Figure 3C), suggesting that CD25-iΔTregs might rely on IL-7 following CD25 deletion. CD25-iΔTregs showed no increases compared with WT-Tregs in CD122 or CD132, the beta and gamma subunits shared between the IL-2 and IL-15 receptors (Figures S5G and S5H), and failed to express CD215, the alpha subunit of the IL-15 receptor (Figure S5I).

Since processes such as cellular metabolism are dependent on PI(3)K rather than STAT5 signaling, we also examined how reliance on IL-7 or IL-15 might affect signaling along all IL-2 downstream pathways. WT CD25^hi Foxp3⁺ Tregs were stimulated overnight with plate-bound anti-CD3/CD28 and analyzed for activation of the STAT5, PI(3)K, and ERK pathways. As expected, nearly all cells had activated the STAT5, PI(3)K, and ERK signaling pathways after stimulation. We then transferred the cells to new wells lacking any TCR stimulus and incubated them for 24 hr with IL-2, IL-15, IL-7, or no cytokine. We found that IL-2 was more effective than either IL-7 or IL-15 in maintaining signaling along all downstream pathways, including PI(3)K signaling (Figures 6A–6C). Although neither cytokine could maintain downstream signals as well as IL-2, both were still effective compared to the no cytokine condition.

CD25-iΔTregs Rely on IL-7 for Survival

Finally, to evaluate the roles of IL-15 and IL-7 in the homeostasis of CD25-iΔTregs, we used an adoptive transfer approach to deprive CD25-iΔTregs of either cytokine. After administering tamoxifen and sorting CD25-iΔTregs or WT CD25^hi Tregs, we transferred the cells into WT, IL-15 knockout (KO), or IL-7 KO hosts with an equal number of CD45.1⁺ CD25^hi cells as a reference population. Recipient mice were sacrificed after 2 weeks, and recovery of RFP⁺ cells was measured (Figures 6D-6F). No survival differences were observed among WT-Tregs, indicating that IL-7 and IL-15 are dispensable for Treg survival in the presence of IL-2 (Figure 6G). Because of the similarity between the IL-2 and IL-15 receptors and their downstream gene expression profiles (Ring et al., 2012), we expected that IL-15 would be the primary cytokine compensating for IL-2. However, we found only a modest and nonstatistically significant decrease in the survival of CD25-iΔTregs transferred into IL-15 KO hosts. A larger and statistically significant exacerbation of cell death was instead observed among CD25-iΔTregs transferred to IL-7 KO hosts (Figure 6H). Foxp3 expression was not significantly affected following transfer into either KOt host (Figures S6B-S6D).

KEY RESOURCES TABLE

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Laurence A. Turka (lturka@partners.org).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

METHOD DETAILS

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analysis was performed with Prism 7 (Graphpad). Statistical details, including statistical tests used, number of mice analyzed (n) can be found in the legend for each figure. Significance was defined as p ≤ 0.05, with further stratification of p values reported as: not significant (ns) = p > 0.05, * = p ≤ 0.05, ** = p ≤ 0.01, *** = p ≤ 0.001, **** = p ≤ 0.0001. All error bars are represented as mean ± SD.