Treg Fragility: A Prerequisite for Effective Antitumor Immunity?

Building Up: How Is Treg Stability Maintained?

Treg stability has been discussed across many disease types. Treg instability was initially defined as the loss of Foxp3 expression in cells and subsequent loss of suppressive function. This is thought to be, in part, due to lack of demethylation or remethylation at certain sites within the Foxp3 locus (33). Demethylation in this locus was first described in 2007 when a region in the 5′ UTR of the Foxp3 locus containing a number of conserved demethylated CpG motifs, was identified (known as the TSDR/CNS2). This demethylation pattern is observed in both thymic and mature peripheral Tregs in mice and peripheral blood of humans. In contrast, the CNS2 is methylated in effector CD4⁺ T cells. Hypomethylation of the Foxp3 locus is thought to be required for Treg stability, but occurs independently of Foxp3 upregulation (34). It was shown that TGFβ-induced Tregs (iTregs) display a somewhat hypermethylated Foxp3 locus (34), and tumor-derived Tregs exhibit a primarily demethylated locus (35). However, variations within CNS2 methylation patterns (∼10%–60%) suggest that the intratumoral Treg pool may be heterogeneous. Previous studies have identified two subsets of Tregs (Foxp3^Hi and Foxp3^Lo) in the tumor and periphery of colorectal patients that display differences in suppressive capacity and CNS2 methylation status (36, 39). Another possibility that could contribute to the heterogeneity of Tregs within the tumor is the conversion of conventional T (T_conv) cells to Tregs. However, TCR repertoires between these two populations appear to be distinct (37, 38). Whether Foxp3⁺ Tregs with a partially methylated Foxp3 locus are a functionally unstable population or a heterogeneous population comprised of both stable and unstable Tregs remains unclear. The possibility that there are other factors that lead to the remethylation of the locus in the TME warrants further investigation.

In addition to epigenetic alterations of the Foxp3 locus, loss of Foxp3 expression is a hallmark of unstable Tregs. A number of factors involved in maintaining Foxp3 expression have been identified, including IL2/STAT5 and Foxo1/3a. STAT5 binds to the Foxp3 locus, and in its absence, Treg development is reduced (Fig. 1; ref. 40), while Foxo1 and Foxo3a translocate to the nucleus of Tregs and prevent effector functions (41). Induction of Foxp3 and subsequent Treg development can also be prevented by persistent TCR stimulation, leading to constitutive activation of the PI3K/Akt/mTOR pathway (42), or in the absence of the microRNA processing enzyme Dicer (43).

Loss of Foxp3 expression has also been reported in certain disease settings, such as lymphopenia and autoimmune diabetes (27). Fate-mapping mice were used to trace all cells that currently or previously expressed Foxp3, regardless of subsequent downregulation. “ExTregs,” which no longer express Foxp3, upregulated IL7R, secreted IFNγ and IL17, took on a pathogenic role, and worsened disease (29). Other groups report that Foxp3 expression is stable and that previously identified exTregs were likely T cells that transiently upregulated Foxp3 during differentiation or activation (31, 44). Very few exTregs have been observed in mouse models of cancer (32), suggesting Foxp3 is stable in these models. Other studies have highlighted a role for Helios in maintaining Foxp3 expression and showed that in its absence, Foxp3 was reduced, and Helios-deficient Tregs secreted IFNγ in the TME (45, 46). However, whether unstable Tregs exist in patients is unknown and remains difficult to assess without cell lineage tracing capabilities or definitive markers.

Breaking Down: How Is Treg Fragility Induced?

Multiple factors have been reported previously to be important for preventing Treg fragility, including Nrp1, Foxo1, and Eos (14, 32, 46–49). Although Nrp1 can be expressed on a number of cell types, it is upregulated on Tregs and supports Treg function through binding of semaphorin-4a (Sema4a). In the absence of Nrp1 signaling, either through Nrp1 blockade or genetic deletion (using Nrp1^L/LFoxp3^Cre-YFP mice), intratumoral Tregs show reduced cell survival, reduced expression of suppressive markers (such as CD73, IL10, and IL35), and significantly impaired suppressive function (Fig. 1). Surprisingly, this did not result in loss of Foxp3 expression. Mice cleared tumors similarly to mice lacking Tregs_, but they displayed no signs of autoimmunity. Nrp1⁺ Tregs are also increased in metastatic melanoma and head and neck squamous cell carcinoma patients compared with healthy donors (32). Patients with a larger population of Nrp1⁺ Tregs correlated with reduced disease-free survival. It was later shown that Nrp1⁻ Tregs produce IFNγ and were less suppressive than wild-type (WT) Tregs but maintained Foxp3 expression. Secretion of IFNγ only occurred within the TME, likely as a result of sustained or increased Hif1α expression due to hypoxia and heightened Akt activity (14, 32, 50, 51). It is possible that the unique metabolic environment within the TME contributes to Treg stability, as they are more adept at tolerating this “harsh environment” compared with effector T cells. Distinct metabolic differences between Tregs and other T-cell subsets have been found (52, 53), which may also underlie the preferential restriction of a fragility Treg phenotype to the TME following Nrp1 deletion/blockade or immunotherapy.

Previous reports have shown that although Foxo1 is key for Foxp3 upregulation during Treg development, the loss or mutation of Foxo1 from mature Tregs leads to a fragile phenotype rather than an unstable one. Specifically, Foxo1-deficient Tregs are less suppressive and secrete IFNγ, but Foxp3 expression is maintained and the percentage of Tregs increases in vivo. Foxo1 deletion in Tregs leads to an IFNγ-dependent lethal inflammatory phenotype (54). However, unlike the phenotype observed in the absence of Nrp1, wherein fragility appears to be restricted to the TME (14, 32), the loss of Foxo1 results in a fragile phenotype that is systemic, leading to the inflammatory phenotype. Similarly, Eos is upregulated in Tregs, and when removed, Treg suppression is reduced, IFNγ and IL2 are upregulated, but Foxp3 expression remains unchanged (47). It has been reported that when LAG3 is deleted on Tregs, Eos is increased and leads to better suppressive function in an autoimmune diabetes setting (55). It is possible that LAG3 limits Eos expression in Tregs, thereby promoting fragility, but further studies are required.

Is Response to Immunotherapy Dependent on Treg Fragility?

Intratumoral Tregs display a distinct profile, suggesting specific markers could be targeted within the TME that might lead to Treg instability or Treg fragility. However, whether this is observed in cancer patient intratumoral Tregs and predicts patient responsiveness remains unknown. We hypothesize that Tregs within the TME upregulate stabilizing molecules, such as NRP1, and that patients who respond to immunotherapy exhibit a more fragile intratumoral Treg phenotype.

Treg fragility appears to be required for response to anti–PD-1 in murine tumor models. In an adenocarcinoma mouse model that is sensitive to PD-1 blockade, treatment of WT mice with anti–PD-1 led to the upregulation of IFNγ⁺ Tregs, consistent with an increased fragile phenotype. When Tregs were insensitive to IFNγ (through the use of an Ifngr1^L/LFoxp3^Cre-YFP mouse), mice were completely resistant to PD-1 blockade in comparison with ∼40% response in WT mice (32), suggesting a role for Treg fragility in responsiveness to immunotherapy. Similarly, reduction of tumor burden through the use of a GITR agonist antibody (DTA-1) was due to an increase in IFNγ⁺ Tregs and reduction in Helios expression (45, 56). As mentioned, CD8:Treg ratios have been shown to be indicative of patient response to therapy. However, the idea that Treg fragility is the key component to determining response to immunotherapy was previously unappreciated.

Similarly, previous studies have identified IFNγ⁺ Tregs in human samples in autoimmune diseases, such as multiple sclerosis and type 1 diabetes, where these cells have reduced suppression and altered methylation while maintaining Foxp3 expression, suggesting a fragile phenotype (34, 57, 58). Patients with malignant glioma (GBM) exhibit a higher percentage of circulating PD-1^Hi Tregs that are less suppressive and express IFNγ. PD-1^Hi Tregs bear a distinct transcriptional profile, are phenotypically exhausted as defined by upregulation of LAG3 and Tim3, and show a slight reduction in CNS2 demethylation. When GBM patients were treated with anti–PD-1 (nivolumab), the exhausted PD-1⁺IFNγ⁺ Treg population increased (59). This population has been observed in other tumor types, such as late advanced rectal cancer, where it correlated with poorer patient response (60). These data suggest that PD-1 blockade, as well as other immunotherapies, may act, in part, through inducing a fragile Treg phenotype in patients. Whether this is a direct effect of anti–PD-1 on Tregs or an indirect effect of increased IFNγ in the TME acting on Tregs to drive fragility remains to be determined.

Given that the impact of Treg fragility in immunotherapy has not been fully elucidated in the clinic, assessing the extent to which patient Tregs develop a fragile phenotype following immunotherapy could aid in both prediction of patient susceptibility to anti–PD-1, as well as provide a rationale for patient responsiveness to immunotherapy. Sensitizing Tregs to become fragile may be an effective strategy to utilize alongside PD-1 blockade. Although PD-1 blockade has been shown to upregulate IFNγ in CD8⁺ T cells, whether this directly affects Tregs remains unclear in the clinic. However, IFNγ-sensitive Tregs have been observed in patient samples and were found to be less suppressive following IFNγ treatment in vitro (32). Although PD-1 blockade has been the primary focus thus far, it is possible that Treg fragility plays a key role in responsiveness to other immunotherapies in the clinic or perhaps the efficacy of any immunotherapy. One possibility could be to target a known driver of Treg fragility prevention, such as NRP1, through antibody blockade. Although Tregs become fragile in mice upon loss of Nrp1 through either genetic deletion or antibody blockade, whether this is conserved in human Tregs after Nrp1 depletion remains to be further investigated. Although a higher percentage of Nrp1⁺ Tregs in patients appears to correlate with reduced disease-free survival, it remains unclear whether this is due to enhanced Treg stability. We would argue that inducing Treg fragility may be a preferred therapeutic strategy compared with Treg depletion or destabilizing Tregs because the effect on Tregs seems to be restricted to the TME, thereby preventing autoimmune side effects.

Identifying ways to target Treg fragility while leaving Treg stability intact may be critical, given the previously identified pathogenic nature of unstable Tregs or exTregs in various diseases (Fig. 1; refs. 27–29). It is possible that local Treg destabilization strategies may be efficacious. However, the potential systemic autoimmune effects of this are unknown, and distinguishing between the two Treg subsets can be challenging. Although some clear markers of fragile Tregs have been identified, including Nrp1, PD-1, and IFNγR1, specific markers do not exist for unstable Tregs that distinguish them from Th-like cells, and tracking the presence of exTregs in patient samples is not yet feasible. There may be more unappreciated markers of Treg fragility that warrant further investigation.

Although targeting molecules that prevent Treg fragility in the clinic may represent the clearest step forward, many interrelated questions remain: (i) What are the markers of unstable or exTregs in patients? (ii) Are there other drivers of Treg fragility? Although IFNγ has been shown to drive Treg fragility, it is possible that other cytokines or soluble factors could lead to a similar phenotype. (iii) Do fragile Tregs display hypermethylation at the Foxp3 CNS2 locus, and does this lead to reduced suppressive function? (iv) What is the level of Treg fragility and instability in checkpoint blockade responders and nonresponders, and do they correlate? (v) Is Treg fragility a biomarker of patient response to immunotherapy? (vi) Does patient response to immunotherapy depend on Treg fragility? Although loss of Nrp1 and increased IFNγ sensitivity have been identified as drivers of Treg fragility, other molecules that contribute to this phenotype have yet to be defined. We propose that the development of combinatorial immunotherapies that maximize Treg fragility may maximize efficacy and improve patient response to immunotherapy.

Disclosure of Potential Conflicts of Interest

D.A.A. Vignali reports receiving commercial research funding from Potenza, Tizona, and Bristol-Myers Squibb; has ownership interest in Bristol-Myers Squibb, Potenza, Tizona, Oncorus, and Merck; and is a consultant/advisory board member for Potenza, Tizona, Oncorus, FStar, and Pieris. No potential conflicts of interest were disclosed by the other author.