Aire-dependent thymic development of tumor-associated regulatory T cells^*

Treg cells are critical for the prevention of autoimmunity, the maintenance of immune homeostasis, and the suppression of anti-tumor immune responses (1, 2). For many human cancers, the density of Tregs within tumor lesions is predictive of poor clinical outcome (3), suggesting that Tregs play a functional role in cancer progression. In this study, we set out to establish a tractable animal model in which a single specificity of naturally occurring tumor-associated Tregs could be studied in the context of a genetically driven mouse model of autochthonous cancer. In order to identify an endogenous tumor-associated Treg response, we analyzed the immune response in TRAMP mice, which develop prostatic adenocarcinoma due to the transgenic expression of the model oncogene SV40 T antigen in the prostate (4, 5). Unlike the prostates of tumor-free mice, which contain very few Treg cells (which are identified as CD4⁺Foxp3⁺), a substantial population of Tregs can be detected in the prostate tumors of TRAMP mice (fig. S1). We employed an experimental system involving T cell antigen receptor alpha chain (TCRα) repertoire analysis of T cell populations from TRAMP mice expressing the Foxp3^gfp reporter (6) and a fixed (transgenic) TCRβ chain. The fixed TCRβ used in this study was a TCRβ chain that was found to be recurrently expressed by CD4⁺Foxp3⁺ Tregs isolated from the prostates of TRAMP mice (fig. S2), and will be referred to hereafter as “TCRβtg”. TCR complementarity determining region 3 (CDR3) length distribution analysis of cDNA from purified CD4⁺Foxp3⁺ and CD4⁺Foxp3^neg T cells from the prostate tumors of TRAMP^+/− Foxp3^gfp TCRβtg males revealed a substantial overrepresentation of CD4⁺Foxp3⁺ T cells expressing a Vα2 (TRAV14) TCRα chain with a CDR3 of nine amino acids in length (as defined by IMGT, http://www.imgt.org) (Fig. 1A, denoted in red). Deep sequencing of these samples revealed that the identical TCRα chain, of CDR3 sequence LYYNQGKLI, was recurrently expressed by Foxp3⁺ Tregs (Fig. 1B), indicating that Tregs of a single specificity are recurrently enriched within TRAMP prostate tumors. Strikingly, in many prostate samples, the Vα2-LYYNQGKLI TCR chain was encoded by a single nucleotide sequence (fig. S3), suggesting that in many cases, tumor-infiltrating Tregs of this specificity may originate from a single clone.

A survey of different anatomical sites of TRAMP^+/− Foxp3^gfp TCRβtg mice using CDR3 length distribution analysis revealed that the overrepresentation of Tregs expressing a Vα2⁺ TCRα chain of nine amino acids in length was observed in the prostate tumor and prostate-draining periaortic lymph nodes, but was not detected over background in non-draining brachial lymph nodes (Fig. 1C). Thus, Tregs of this specificity are not expanded systemically in tumor-bearing mice, but are instead selectively enriched in the prostate tumor environment.

In order to gain insight into the antigen specificities of polyclonal tumor-infiltrating CD4⁺ T cells, we determined the extent of overlap of the Vα2⁺ TCR repertoire for T cell subsets isolated from the prostate tumors of TRAMP^+/− Foxp3^gfp TCRβtg mice. Repertoire overlap was assessed using the Morisita-Horn (MH) similarity index (7–11), for which a value of 1 indicates identity, and a value of 0 denotes complete dissimilarity (Fig. 1D). The analysis revealed that the TCR repertoire of CD4⁺Foxp3⁺ and CD4⁺Foxp3^neg populations isolated from a particular prostate tumor were largely distinct and non-overlapping (MH = 0.07 ± 0.10 SD, Fig. 1D, samples intersecting at red lines), implying that the antigens recognized by tumor-infiltrating Tregs are different from those recognized by conventional CD4⁺ T cells. Second, the TCR repertoire of CD4⁺Foxp3⁺ cells isolated from the prostates of different mice exhibited a high degree of similarity from mouse to mouse (MH = 0.38 ± 0.39 SD, Fig. 1D, upper left quadrant). While a substantial proportion of this similarity was due to the recurrent enrichment of the Vα2-LYYNQGKLI TCR, additional TCRα chains were identified that were recurrently expressed by prostatic Foxp3⁺ Tregs (fig. S4). This finding suggests that TRAMP prostate tumors do not recruit polyclonal Tregs of arbitrary specificity, but instead are associated with the reproducible enrichment of Tregs of distinct specificities.

To facilitate the study of T cells expressing the canonical Vα2-LYYNQGKLI TCRα chain paired with the fixed TCRβ chain (hereafter referred to as “MJ23” T cells), we generated transgenic (tg) mice expressing the MJ23 TCRαβ heterodimer. In female MJ23tg Rag1^−/− mice, CD4⁺ T cells in the periphery were phenotypically naïve (Fig. 2A and fig. S5). Moreover, Foxp3⁺ MJ23tg cells were not detected above background in the thymus and periphery of these mice (Fig. 2, A and B, and fig. S5). In contrast, analysis of tumor-free male MJ23tg Rag1^−/− mice revealed spontaneous accumulation of activated CD44^hi CD4⁺ MJ23tg T cells in the prostate and prostate-draining periaortic lymph nodes (Fig. 2, A and B, and fig. S5), indicative of MJ23 reactivity to an autoantigen at these sites. In the prostate, T cell accumulation was observed in both the stroma and the epithelium, and was associated with disruption of the basement membrane of prostatic glands (Fig. 2C). The presence of activated CD4⁺ T cells was accompanied by the concomitant appearance of a percentage of CD4⁺Foxp3⁺ T cells (Fig. 2, A and B, and fig. S5). Foxp3⁺ cells exhibited in vitro suppressive activity (fig. S6) and phenotypic characteristics of Tregs (fig. S7), including expression of neuropilin-1 (supplementary text and figs. S7 and S8). Taken together, our data demonstrate that MJ23 Tregs, identified based on their enrichment in mouse prostate tumors, are not reactive to a unique tumor-specific antigen, but instead recognize a self antigen associated with the organ of cancer origin.

In other experiments, CD11c⁺ dendritic cells isolated from TRAMP prostate tumors induced robust proliferation of MJ23tg T cells ex vivo (Fig. 2D), but did not stimulate OT-II transgenic T cells expressing an irrelevant class II-restricted TCR (12). These data provide direct evidence that prostate tumors contain the antigen recognized by MJ23 T cells.

In order to study the development of MJ23tg T cells at physiological clonal frequencies, we generated chimeric animals in which bone marrow cells from MJ23tg Rag1^−/− donors were engrafted into host mice at low frequencies (<1%), and the development and distribution of MJ23tg T cells was assessed (Fig. 3 and fig. S9). Analysis of B6 male and tumor-bearing TRAMP male hosts revealed that Foxp3⁺ MJ23tg Tregs developed in the thymus (Fig. 3, A and B). The efficiency of development varied from mouse to mouse, and was inversely correlated with the frequency of MJ23tg precursors (Fig. 3B), a finding that is consistent with previously published studies (13, 14). In other experiments, OT-II Rag1^−/− precursors did not develop efficiently into Foxp3⁺ cells in the thymus, demonstrating the specificity of MJ23tg Treg development (fig. S10). In the periphery of male chimeras, Foxp3⁺ MJ23tg cells were distributed throughout the secondary lymphoid organs, but were preferentially enriched in the prostate-draining periaortic lymph nodes (Fig. 3C), a finding consistent with evidence of reactivity to a prostate-associated antigen (Fig. 2). Very few donor-derived MJ23tg Tregs were observed in the prostate (Fig. 3, A and C), likely reflecting competition with endogenous MJ23 Tregs (supplementary text and fig. S11). Upon intravenous transfer of naïve CD4⁺Foxp3^neg MJ23tg T cells into TRAMP males, the induction of Foxp3 expression by MJ23tg cells was negligible (fig. S12), suggesting that antigen exposure in the periphery does not favor the development of induced Tregs. Taken together, our results demonstrate that expression of the MJ23 TCR facilitates Treg development in the thymus of male hosts. Thus, T cells of this specificity encounter the antigen(s) driving Treg development during maturation in the thymus, prior to their exposure to the prostate or tumor environment.

On the basis of our data indicating that MJ23 T cells are reactive to a prostate-associated antigen, we anticipated that Foxp3⁺ MJ23tg Tregs would not develop in B6 female hosts. However, unexpectedly, Foxp3⁺ MJ23tg Tregs developed in the thymus of chimeric B6 females (Fig. 3, A and B, and fig. S9). In the periphery of female mice, Foxp3⁺ MJ23tg T cells were broadly distributed in the spleen and all lymph nodes examined, but selective enrichment in the periaortic lymph nodes was not observed in female hosts (Fig. 3C). In order to elucidate the mechanisms underlying MJ23tg Treg development in both male and female mice, we examined the role of Aire in development. Aire encodes a transcriptional regulator that drives the ectopic expression of peripheral tissue-specific antigens by medullary thymic epithelial cells, and is critical for the maintenance of immune tolerance (15–17). Analysis of MJ23tg development in chimeric mice in which MJ23tg precursors were engrafted into Aire^+/+ or Aire^−/− hosts revealed that Foxp3⁺ MJ23tg Tregs failed to develop in the thymus and periphery of Aire^−/− hosts, both male and female (Fig. 4, A and B). In addition, Aire^−/− hosts exhibited a reduction in the percentage of thymic Tregs relative to Aire^+/+ hosts, consistent with previous reports (15, 18), but were not characterized by a complete deficiency of polyclonal Tregs (Fig. 4, A and C). Despite the lack of MJ23tg Treg development, mature CD4⁺Foxp3^neg MJ23tg T cells developed in Aire^−/− hosts (Fig. 4A and fig. S13), indicating that T cells of this specificity are not dependent on Aire for their positive selection into the CD4⁺ lineage. This result also suggests that the thymic development of Tregs reactive to an Aire-dependent antigen may require both the recognition of self peptide/MHC ligands for positive selection, and the recognition of an additional Aire-dependent antigen for commitment to the Treg lineage. Importantly, similar experiments utilizing a second naturally occurring tumor-infiltrating Treg TCR, termed “RT83”, demonstrated that Aire is required for the thymic development of RT83tg Tregs (Fig. 4D and fig. S14). Thus, our data demonstrate that Aire is critical for the thymic development of multiple naturally occurring Treg specificities. These findings are consistent with a model in which the Aire-mediated expression of peripheral tissue-restricted antigens by thymic epithelial cells drives the development of a subset of tissue-specific or organ-specific Tregs (19).

While there is substantial evidence demonstrating that Aire plays a critical role in the deletion of autoreactive thymocytes reactive to peripheral tissue antigens (15, 20–22), a definitive role for Aire in the thymic selection of naturally occurring Tregs has not been previously established. Here, we provide direct evidence that Aire is critical for the thymic development of Tregs of naturally arising specificities. Thus, the integration of available evidence suggests a dual role for Aire in the maintenance of immune tolerance, in which Aire drives both the deletion of autoreactive T cells and the development of a subset of Foxp3⁺ Tregs.

In sum, our data support a model in which a tumor does not drive the de novo conversion of tumor-specific CD4⁺ effector T cells into induced Foxp3⁺ Tregs, but instead recruits pre-existing thymic-derived Tregs reactive to Aire-dependent self antigens associated with the organ of cancer origin. Thus, a developing neoplasm co-opts endogenous mechanisms that have evolved to preserve the integrity of the host by maintaining organ-specific immune tolerance.