Non-self antigens are the cognate specificities of Foxp3⁺ regulatory T cells.

Abstract

The majority of regulatory Foxp3⁺CD4⁺ T cells naturally arises in the thymus. It has been proposed that T cell receptors (TCRs) on these cells recognize self MHC class II-peptide complexes with high or higher affinity, and that their specificities mirror specificities of autoreactive T cells. Here we analyzed hundreds of TCRs derived from regulatory or non-regulatory T cells and found little evidence that the former population preferably recognizes self-antigens as agonists. Instead, these cells recognized foreign MHC-peptide complexes as often as non-regulatory T cells. Our results show that high affinity, autoreactive TCRs are rare on all CD4⁺ T cells and suggest that selecting self-peptide is different from the peptide that activates the same regulatory T cells in the periphery.

Introduction

Central tolerance eliminates thymocytes expressing T cell receptors (TCRs) with high affinity to self antigens, but routinely a small fraction of potentially pathogenic T cells escape this process. In the periphery, autoreactive T cells are controlled by active suppression mechanism carried out by regulatory CD25⁺CD4⁺ T (T_reg) cells that express transcription factor Foxp3 (Hori et al., 2003; Fontenot et al., 2003; Khattri et al., 2003; Shevach et al., 2006). Foxp3 was initially considered as a lineage specification factor that determines the fate of thymic precursors of T_reg cells (Fontenot et al., 2005a; Fontenot et al., 2005b). However, as recently shown in different mouse models with a fluorescent reporter knock-in gene, Foxp3 does not act as a master switch for T_reg cell lineage commitment (Gavin et al., 2007; Wan and Flavell, 2007). Instead, it stabilizes the phenotype and suppressive function of anergic thymocytes by altering the wide array of T_reg cell-specific genes (Gavin et al., 2007; Zheng et al., 2007). Thus, although Foxp3 and the suppressive functions of T_reg cells can now be correlated with its late expression in mature CD4⁺CD8⁻ thymocytes (Fontenot et al., 2005a), the signals inducing the lineage commitment of these cells and Foxp3 expression remain elusive. It has been proposed that some of the thymocytes that recognize self MHC-peptide complexes with high affinity, instead of dying during negative selection or becoming autoreactive T cells, would acquire Foxp3 and differentiate to T_reg cells (Coutinho et al., 2005; Fontenot and Rudensky, 2005; Liu, 2006; Cabarrocas et al., 2006).

The direct evidence implying that precursors of T_reg cells require high affinity, self-reactive TCRs to become T_reg cells came from the TCR transgenic mouse model that co-express cognate peptide in the thymus (Jordan et al., 2001). In this model, interactions between TCR and class II MHC bound with strong but not weak agonist peptide lead to the selection of Foxp3⁺ T_reg cells with regulatory properties (Caton et al., 2004; D'Cruz and Klein, 2005; Kim and Rudensky, 2006). It was also shown in the thymic medulla that expression of agonist peptides induces a non-deletional differentiation to Foxp3⁺ T_reg cells, and that in the periphery suboptimal dose of cognate antigen can convert naive, TCR transgenic cells to Foxp3⁺ T_reg cells (Apostolou et al., 2002; Watanabe et al., 2005; Liu, 2006; Aschenbrenner et al., 2007). However, another study that used different, intrathymically expressed agonist peptide could not reproduce selection of T_reg cells, suggesting that factors other than agonist peptides may induce natural differentiation of T_reg cells (van Santen et al., 2004).

If higher affinity TCR-MHC-peptide interactions control natural T_reg differentiation one should also expect that these cells abundantly express self-reactive TCRs. Indeed, experiments performed in TCRβ transgenic mice, where conventional CD4⁺ T cells were transduced with TCRs derived from T_reg cells, showed that in lymphopenic hosts and in co-cultures with autologous APCs, these transductants proliferate faster than transductants expressing TCRs cloned from naive T cells (Hsieh et al., 2004). Moreover, when these TCRβ transgenic mice were crossed with Foxp3 deficient mice that are prone to autoimmunity, the T and autoreactive Foxp3⁻CD25⁺ cells but not the Foxp3⁻CD25⁻ cells expressed TCRs with overlapping antigenic specificities (Hsieh et al., 2006). These results imply that T_reg cells express TCRs with higher affinity to self and that these TCRs overlap with TCRs expressed by pathogenic, self-reactive T cells. However, this proposition remains uncertain, as the contribution of Foxp3-deficient T_reg cells in the pathogenesis of scurfy male mice is still unclear and transfer of these cells does not induce autoimmune diseases in lymphopenic hosts (Lin et al., 2007). In addition, the incidence of TCRs specific to foreign antigens on these cells has also been confusing. There are multiple experimental reports that T_reg cells recognize exogenous antigens derived from microbes, parasites, neo-antigens or allogeneic tissues, but it is unclear whether these responses are antigen specific or rely on inherent cross-reactivity, dual TCRs or a bystander effect (Belkaid et al., 2002; Suvas et al., 2003; Ochando et al., 2005; Tuovinen et al., 2006; Lerman et al., 2004; Larkin, III et al., 2007). Interestingly, analyses of TCRs in mice expressing a TCRβ transgene or miniature repertoire of TCRs found that dominant TCRs on naive T and regulatory T cells overlap to some extent (Hsieh et al., 2004; Pacholczyk et al., 2006; Wong et al., 2007) and there is an increased diversity of TCRs on T_reg cells, suggesting that this repertoire encodes multiple antigen specificities (Pacholczyk et al., 2006). Therefore here we wanted to comprehensively analyze antigenic specificities of Foxp3⁺ T_reg cells and to determine if their TCRs are inherently autoreactive.

For this purpose we characterized hundreds of TCRs originally expressed by CD4⁺Foxp3⁺ T_reg or CD4⁺Foxp3⁻ naive T cells and concluded that the frequency of TCRs with high-affinity for self versus non-self antigens is similar within the two CD4⁺ populations. Our results also demonstrate that T_reg cells maintain a highly diverse TCR repertoire that included most of the TCRs dominantly expressed by naive T cells. In contrast, many TCRs expressed by T_reg cells were not found on naive CD4⁺ T cells. Finally, we showed that an outbreak of immune wasting disease in lymphopenic mice relied on the hosts class II MHC ability to present foreign and not self antigens. Therefore, when wasting disease is induced by T cells bearing T_reg-derived TCRs (Hsieh et al., 2004), these TCRs recognize non-self antigens at least as efficiently as TCRs expressed on naive T cells.

Results

Naïve T and T_reg cells express many identical TCRs that are asymmetrically distributed

It has been reported that naive T and T_reg cells express largely different repertoires of TCRs, suggesting that these populations are separated based on the affinity of individual TCRs to self MHC-peptide ligands. However, naive T and T_reg cells share a portion of TCRs that, depending on the experimental model and evaluation method, varies from 10% to 42% (Hsieh et al., 2006; Pacholczyk et al., 2006; Wong et al., 2007). These analyses included many infrequent TCRs, possibly underestimating the overlap between compared regulatory and non-regulatory populations. To determine the nature of this overlap we took advantage of the recently described TCR^mini mouse model, where sequences of CDR3 regions of TCRα chains are utilized to track individual T cell clones. In these animals, T cells express one transgenic TCRβ chain and many TCRα chains derived from a mini-locus that can rearrange one Vα2 segment and two Jα26 or Jα2 segments, during T cell development in the thymus. In the TCR^mini mice used here Foxp3⁺ T_reg and Foxp3⁻ naive T cells are naturally selected in the thymus on A^b molecules bound with many endogenous peptides (TCR^mini mice) or a single covalently linked Ea(52−68) peptide (TCR^miniEp mice). First, we looked at how many of the “most frequent” TCRα transcripts from naive T cell clones were present at least once in the Foxp3⁺ T_reg population. As a cut-off value for determination of frequent TCR clones, we calculated the maximum frequency φ_max (with a 95% confidence) of a TCR clone whose sequence could not be found after accumulation of n TCR sequences in a sample (Fig. 1A) (Baron et al., 2003). Based on this evaluation we could assume that we found all TCR clones in the tested repertoire with a frequency above 0.62% (0.91% for TCR^miniEp). As shown in Fig. 1B we found 41 different TCRs from TCR^mini mice (32 from TCR^miniEp) that had a frequency above φ_max value, and then compared their frequencies within the naive T and T_reg repertoires. 68% of TCRs that were dominantly used by naive T cells were also found on T_reg cells, although with very different frequencies. In contrast, 39% of the dominant TCRs for T_reg cells were identified on naive T cells. Furthermore, when a similar comparison was performed for TCRs retrieved from TCR^miniA^bEp mice, 77% of the dominant TCRs from naive T cells were also found on T_reg cells, whereas 42% of TCRs from T_reg cells were present on naive T cells (Fig. 1B). Thus, when we concentrated TCR analysis on T cell clones with high frequency, the overlap between T_reg and naive T populations became substantially greater than initially reported (Pacholczyk et al., 2006). In addition, more TCRs from naive T cells were found on T_reg cells, suggesting that TCR repertoire expressed by T_reg cells may span over a large portion of the naive T cell repertoire. This result corroborates with our previous finding that T_reg cells have higher diversity of TCRs, suggesting that observed variation of dominant TCRs expressed on naive T and T_reg cells can arise from asymmetric distribution of the same TCRs not separate selection of different TCRs. The overlap between these two TCR repertoires appeared to be even more significant in TCR^miniEp mice, where the diversity of different self class II MHC-peptide complexes is minimal and positive selection of CD4⁺ T cells proceeds on one peptide covalently bound to A^b molecules. This expression of identical TCRs on T_reg and naive T cells argued against the hypothesis that naive T and T_reg lineages separate based on the different affinities of TCRs for the selecting MHC-peptide ligands.

Figure 1. Overlap of TCRs between naïve T and T_reg populations.

(A) The maximum frequency (ϕ_max) of a CDR3α transcript whose sequence did not appear after accumulation of n sequences, with 95% confidence, was computed based on equation 1. Upper and lower limits of the 95% confidence interval for a clone with ϕ_max frequency were calculated based on equation 2. Both equations are shown in Experimental Procedures. (B) Observed frequency of most dominant naive and T_reg clones found in TCR^mini or TCR^miniEp mice. Percentage of overlap was calculated by dividing shared clones by a total number of dominant clones of a given population. Shown amino acid sequences represent fragments of CDR3 regions including amino acids beginning with the third amino acid after the invariant C residue in all TCRAV genes (Y-L/F-C-A-X-1) and spanning the amino acid immediately preceding the TCRAJ motif (2-F/W-G-X-F-G-T). Clones shared between populations are highlighted by grey box. Table below the graphs shows at what frequency dominant clones of one population were found at least once in second population (for both type of mice).

Homeostatic expansion of T cells in vivo

Naive T cells that express TCRs with higher affinity to self-ligands have improved homeostatic expansion (Ge et. al., 2001). It was also shown that after adoptive transfer to lymphopenic hosts, naive T cells transduced with TCRs cloned from T_reg cells expand faster than ones transduced with TCRs derived from naive T cells (Hsieh et al., 2004). This observation has been interpreted in favor of the hypothesis that T_reg cells frequently recognize self MHC-peptide complexes with higher affinity and that this feature of T_reg-derived TCRs enhances their homeostatic fitness. Because natural T_reg cells require self-peptide-MHC complexes for homeostatic expansion and proliferate in vivo in the absence of exogenous antigen, we examined whether the reported higher affinity of T_reg cells to self-peptides will enhance the peripheral expansion of these cells in a non-lymphopenic host.

To investigate this issue, we chose the most frequent TCRs for the naive T and T_reg clones and followed their expansion ratios between the thymus and the periphery. If increased self-reactivity would help T_reg cells expand better, one should expect a higher proportion of T_reg than naive T clones expanding after leaving the thymus. The same fraction of both naive T and T_reg populations, roughly one third, expanded, contracted, or did not significantly changed their frequency after migration from the thymus to the periphery (Fig. 2). It suggested that the overall fitness required for the survival of naive T and T_reg cells in non-lymphopenic mileu is very similar.

Figure 2. Relative extrathymic proliferation of dominant T cell clones from TCR^mini mice.

The proportions of peripheral proliferation of most frequent T cell clones (shown in Fig. 1) was calculated by dividing the peripheral frequency of a given clone by its thymic frequency. CDR3α sequences are aligned from the most expanded to the least proliferated clones. Clones shared between populations are highlighted by grey box.

It is also possible that T_reg cells, due to their anergic nature, cannot proliferate beyond a particular threshold level. To investigate this possibility, we followed the fate of naive T cell clones expressing TCRs found also on T_reg cells (Fig. 2). If T_reg cells were selected based on elevated affinity to self MHC/peptide complexes, the naive clones expressing these TCRs should expand better than naive T cell clones bearing TCRs not expressed by T_reg cells. We found similar number of naive Tcell clones increased or decreased regardless of whether these cells had “shared” or “non-shared” TCRs, demonstrating that naive T cell clones bearing “regulatory” TCRs do not proliferate faster (Fig. 2). Altogether, these results implied that expression of T_reg-like TCRs on naive T or T_reg cells does not grant a T cell a homeostatic advantage in a non-lymphopenic environment.

Generation of CD4⁺ T cells with TCRs derived from T_reg cells

The inherent anergy of T_reg cells in vitro impedes the study of their antigenic specificities for self and non-self antigens. So far only a few TCRs derived from T_reg cells have been analyzed by cloning and individually transducing them into non-regulatory T cells (Hsieh et al., 2004). This elegant approach disconnects the TCR from an anergic parental T_reg cells, allowing the TCR specificity to be tested in vivo and in vitro, but it is not suitable to examine many different TCRs on per-cell basis. Furthermore, most of the TCRs analyzed so far were cloned from CD25⁺ T cells with unknown status of Foxp3, permitting unintentional contamination with TCRs derived from Foxp3⁻CD25⁺ autoreactive T cells. To overcome these difficulties, we generated T cell hybridomas from T_reg cells, and confirmed their origin by comparing their TCRs with TCRs found on single-cell sorted Foxp3⁺ T_reg cells. CD4⁺ T cell hybridomas are extensively used to identify low affinity agonists, reveal self antigens responsible for survival and thymic selection or to detect low-abundant, endogenous peptides, demonstrating that it is an established method to study the TCRs specificity (Grubin et al., 1997; Grakoui et al., 1999; Felix et al., 2007).

To produce hybridomas from T_reg cells, we sorted CD4⁺CD25⁺ T cells and cultured them using a protocol optimized to propagate Foxp3⁺ T cells in vitro. In brief, sorted CD4⁺CD25⁺ T cells were cultured in vitro for 7−10 days without loosing expression of Foxp3 and CD25 molecules (Fig. 3A). Importantly, after the expansion phase, the distribution of CDR3 regions of TCRα chains between sorted and expanded T_reg cells remained very similar (82% overlap), demonstrating that the expansion did not alter the original distribution of TCRs expressed on T_reg cells (Fig. 3B). Expanded T_reg cells could be efficiently immortalized by fusion with thymoma BW5147α⁻β⁻. After fusion, T_reg-derived hybridomas lost Foxp3 expression and other surface markers expressed by dividing T_reg blasts but acquired ability to secrete IL-2 upon TCR stimulation - a common feature of conventional CD4⁺ T cell hybridomas. To further ensure that T_reg-derived hybridomas express TCRs from CD25⁺Foxp3⁺ T_reg cells, we compared their TCR sequences with sequences obtained from a database of single cell-sorted T_reg cells. As shown in Fig 3C, after sequencing only forty TCRs, we found the most dominant TCRs originally expressed by T_reg cells, which demonstrated that the hybridomas expressed a true representation of TCRs found on T_reg cells. Finally, to examine if “regulatory origin” does not limit IL-2 secretion upon TCR stimulation, we directly compared the amounts of IL-2 produced by naive T- and T_reg- derived hybridomas. Regardless of origin, all T cell hybridomas produced comparable amounts of IL-2, demonstrating that this assay is appropriate to screen antigenic specificities of TCRs derived from T_reg cells (Fig. 3D). To ensure maximum sensitivity, we assayed only CD4^hi TCR^hi hybridomas and used either bone marrow-derived CD11c^hi A^{b hi} DCs or freshly prepared, non-irradiated splenocytes as the APCs. So far, we established over a thousand of T cell hybridomas, of which roughly half was derived from T_reg cells and another half was derived from naive T cells. These hybridomas were obtained from three different types of mice: the TCR^miniEp, TCR^mini and conventional C57BL/6. Use of TCR^mini repertoires allowed us to identify the origin of the TCRs sequenced from the T cell hybridomas by comparing them to the database of TCR sequences obtained from single-cell sorted T_reg cells (Pacholczyk et al., 2006).

Figure 3. TCRs on T cell hybridomas derived from expanded T_reg blasts accurately represent TCRs found on T_reg cells.

(A) Expression of Foxp3 by T_reg cells isolated from lymph nodes of TCR^miniEp mice and 7−10 days after in vitro culture of these cells. Cells were surface-stained with antibodies against CD4, CD8, and CD25 molecules, followed by intracellular staining with Foxp3 antibody. (B) 2D-F-SSCP revealed extensive (82%) overlap of CDR3 regions of TCRα chains from freshly sorted and expanded T_reg cells. cDNA was synthesized from isolated RNA of both types of cells. Run-off PCR was done with primer specific for TCRVα2, coupled with Cy3 (fluorescent dye). First, fluorescent PCR products were run in the first dimension using capillary electrophoresis in denaturing PAGE. The second dimension was run in a non-denaturing SSCP slab gel. Resulting 2D gels show diversity, distribution, and frequency of dominant CDR3α regions from indicated populations. The intensity of individual dots corresponds to the real frequency of a particular rearrangement in the original sample. Spots migrating with different speed on the second dimension SSCP gel represent different VαJα rearrangements. (C) TCR CDR3α sequences retrieved from T_reg-derived hybridomas (Hyb) were compared with the most abundant CDR3α sequences found in T_reg cells in TCR^miniEp mice (T_reg). (D) Hybridomas derived from naïve T or T_reg cells of TCR^miniEp mice, were co-cultured with splenocytes from C57BL/6 mice. Responding hybridomas were compared for the ability to produce IL-2. Production of IL-2 secreted by activated hybridomas was measured by HT-2 proliferation using MTT assay. Results are shown as arithmetical mean ± SD.

Hybridomas derived from T_reg cells of TCR^miniEp mice recognize non-self antigens

It has been suggested that TCRs expressed by T_reg cells may recognize abundant peripheral self-antigens with an affinity that is above that required for homeostatic expansion (Hsieh et al., 2004; Hsieh et al., 2006). To determine the fraction of T_reg cells that recognizes ubiquitously expressed self-antigens, we first examined the antigen specificities of T cell clones derived from naive T and T_reg cells from TCR^miniEp mice. The advantage of this model is that all self MHC class II molecules in the thymus and periphery remain bound with a single peptide and that the resulting repertoire of TCRs is tolerant to A^bEp complex but lacks tolerance to natural, endogenously processed peptides. Also, more than two-thirds of T cell hybridomas obtained from CD4⁺ T cells produce IL-2 when co-cultured with APCs derived from C57BL/6 mice (Ignatowicz et al., 1996). This feature of TCRs selected by A^bEp complex allowed us to precisely test the specificity for self (A^bEp) and non-self (A^bWT) MHC-peptide complexes. As shown in Fig. 4A, 68% and 87% of hybridomas derived from naive T or T_reg cells respectively, become activated after overnight co-culture with APCs expressing non-self A^bWT-peptide complexes, but none of these hybridomas produced IL-2 after incubation with self APCs. Furthermore, the same TCR^miniA^bEp-derived hybridomas were incubated with APCs derived from invariant chain-deficient (A^bIi⁻) mice that express a lower number of A^b molecules loaded with a reduced number of Ii-independent, endogenous peptides. Here, hybridomas derived from naive T and T_reg cells also responded to non-self APCs with similar, though expectedly lower, frequency. Importantly, recognition of non-self antigens was observed regardless of whether TCRs on T cell hybridomas represented a pool of “shared” or those only found on T_reg cells TCRs (Fig. 4B). In summary, these experiments showed that irrespective of the origin of the antigen(s), the frequency of TCRs directed to non-self antigens is similar between naive T and T_reg cells, and that autoreactive TCR specific to abundant self MHC-peptide complexes are rare on all CD4⁺ T cells.

Figure 4. T_reg-derived hybridomas from TCR^miniEp mice frequently recognize foreign but not self MHC-peptide complexes.

(A) Hybridomas were co-cultured overnight with splenocytes or bone marrow derived dendritic cells (BMDC's) from C57BL/6 (A^b), Ii-, or A^bEpIi⁻ (AEp) mice. Production of IL-2 secreted by activated hybridomas was measured by HT-2/MTT assay. In each experiment, all hybridomas were separately tested for response to anti-CD3 stimulation in the presence of autologous APCs, to determine the percentages of hybridomas able to respond to activation. The box beneath the bar graph shows number of hybridomas (n) used in each test. (B) Twelve dominant T_reg TCRs identified by single-cell RT-PCR were compared with TCRs found on T cell hybridomas. Based on reactivities of these hybridomas, antigenic specificity was assigned to the particular CDR3α sequence.

T_reg cells selected on A^bWT complexes rarely recognize self antigens

One could argue that ubiquitous expression of A^bEp complexes in TCR^miniEp mice enhances deletion of autoreactive TCRs, but in normal animals expressing A^bWT complexes, negative selection will be less efficient and incidence of autoreactive TCRs on T_reg cells could be higher. Therefore, we made another set of hybridomas from TCR^mini and C57BL/6 mice that express A^bWT molecules, and tested their ability to respond to APCs expressing self and non-self MHC-peptide complexes. When hybridomas from TCR^mini mice were co-cultured with autologous APCs, non of them produced IL-2 indicating that the frequency of T_reg cells with potentially autoreactive TCRs is very low (Fig. 5A). To detect the responses of these hybridomas to non-self antigens, we tested their ability to become activated by APCs expressing A^bm12 molecules. The A^bm12 molecule is similar to autologous A^b molecules, but due to the mutation in antigen binding groove, the A^bm12 molecules remain bound with a different set of endogenously processed peptides (Bill et al., 1989). As shown in Fig. 5A, hybridomas derived from T_reg and naive T cells recognized the A^bm12 APCs, but not other allogeneic APCs expressing A^k and E^k complexes. These results showed that TCRs expressed on T_reg cells recognize different non-self peptides bound to MHC, but are not overtly reactive to backbone of class II MHC complexes (Felix et al., 2007). Moreover a similar frequency of A^bm12 specific TCRs was observed on hybridomas derived from naive T and T_reg cells, confirming that these TCR repertoires have comparable ability to sense non-self APCs.

Figure 5. Hybridomas derived from naïve T and T_reg cells respond with the same frequency to stimulation with allogeneic APCs but not autologous APCs.

(A) Hybridomas derived from naive and T_reg cells from TCR^mini or C57BL/6 mice, were stimulated by splenocytes or BMDC's derived from autologous C57BL/6 (A^b), allogenic CBA Ca/J (A^kE^k) or bm12 (A^bm12) mice. All hybridomas were tested for response to anti-CD3 stimulation in the presence of autologous APCs. Production of IL-2 was measured by HT-2 proliferation using MTT assay. To determine percentages of hybridomas responding to APCs, only hybridomas that responded to the anti-CD3 stimulation were considered. The box beneath the bar graph shows number of hybridomas (n) used in each test. (B) The CDR3α regions of autoreactive TCRs identified in TCR^mini repertoire. TCRs on TCR^miniEp T_reg-derived hybridomas reactive to A^b were compared against database containing CDR3α sequences retrieved by single-cell RT-PCR from T cells in TCR^mini mice. Of the six CDR3 sequences identified so far, five were expressed on Foxp3⁺ T_reg cells and two were found on Foxp3⁻ T cells.

Because we have not found autoreactive TCRs in naïve T or T_reg populations, we decided to take advantage of the TCR^mini repertoire that allows for backtracking receptors to their occurrence in vivo. We sequenced TCRs on hybridomas derived from TCR^miniA^bEp that responded in vitro to A^b molecules bound with endogenous peptide, and then searched for identical TCRs in our database of almost 1500 TCRs sequenced from CD4⁺ T cells derived from TCR^mini mice. Although there are only a few TCRs shared between these mice, we identified six TCRs from TCR^miniEp mice that were highly reactive to A^bWT and were found in TCR^mini mice as well (Fig. 5B). Interestingly, when we examined on what type of T cells these TCRs have been found in TCR^mini mice, two TCRs were expressed on naive T cells and five on T_reg cells. Each of these autoreactive TCRs was rare in TCR^mini mice suggesting that although some of TCRs expressed by T_reg cells are potentially autoreactive, their frequency on naive T and T_reg cells was low.

To extrapolate our finding from TCR^mini mice to wild-type T_reg cells, we also tested antigenic specificities of T cell hybridomas obtained from C57BL/6 mice. As shown in Fig. 5A approximately 10% of the hybridomas responded to allogeneic APCs irrespectively of whether these TCRs originated from naive T or T_reg cells. However, none of the hybridomas tested here produced IL-2 following incubation with autologous APCs. This result confirmed our previous findings from TCR^mini mice that in general CD4⁺ T cells rarely express autoreactive TCRs that recognize self class II MHC-peptide complexes as agonists.

Wasting disease induced by adoptive transfer of effector T cells is caused by non-self antigens

Inflamatory bowel disease (IBD) is an immune mediated disease that manifests with inflammation on the colonic and rectal mucosa (ulcerative colitis), and Crohn's disease that affects both the small and large intestines. The genetic and environmental factors responsible for precipitation of the disease are linked with an aberrant inflammatory response toward commensal bacteria. There is an established experimental protocol to induce IBD in mice where transfer of CD4⁺CD45RB^hi T cells into immune-deficient mice, such as Rag2^−/− or SCID mice, leads to colitis, causing a wasting disease (reviewed in Coombes et al., 2005). It has also been shown that expansion of non-regulatory T cells engineered to express TCRs cloned from CD25⁺ T cells causes wasting disease, whereas expansion of T cells expressing TCRs cloned from CD25⁻ T cells does not (Hsieh et al., 2004). These results were interpreted as evidence that repertoire of TCRs expressed by T_reg cells is enriched in autoreactive specificities. However, so far there is no direct evidence that self-antigens are responsible for initiation of wasting disease in the transfer model.

To determine whether recognition of self-antigens can cause wasting disease, we sorted naive CD4⁺CD45RB^hi T cells from TCR^mini and TCR^miniEp mice and transferred them into lymphopenic A^bWT TCRα^−/− or A^bEp TCRα^−/− mice, respectively. The host mice that received CD4⁺CD45RB^hi T cells from TCR^mini mice started losing weight between 4−5 weeks after adoptive transfer (Fig. 6); this trend continued until mice had to be sacrificed. To ensure that the observed weight loss was a result of colitis, lesions in the intestine were confirmed macroscopically (data not shown). Interestingly, transfer of CD4⁺CD45RB^hi T cells from TCR^miniEp mice into A^bEp TCRα^−/− mice did not result in colitis. Because mice expressing covalent A^bEp complexes do not present other antigens our results support previous findings that recognition of bacterial antigens of commensal flora is required for initiation of wasting disease upon transfer of effector CD4⁺ T cells. Importantly, this experiment also implied that the previously reported precipitation of wasting disease by a transfer of T cells expressing T_reg-derived TCRs is caused not by a higher frequency of autoreactive TCRs but rather by sufficient frequency of non-self specificities on T_reg cells.

Figure 6. The CD4⁺CD45RB^hi T cells isolated from TCR^mini but not TCR^miniEp mice cause wasting disease upon adoptive transfer into autologous lymphopenic hosts.

T cells isolated from lymph nodes were FACS sorted using CD4 and CD45RB antibodies. 4×10⁵ of CD4⁺CD45RB^hi T cells (purity above 99%) from TCR^mini or TCR^miniEp mice were injected intraperitoneally to A^bWT TCRα^−/− or A^bEp TCRα^−/− lymphopenic hosts, respectively. The ratio of the starting weight of recipient mice is shown. Data are representative of two independent experiments with 2 or 3 mice per group. Weight loss was associated with colon inflammation, prolapse, bowel wall thickening, and typical hunched-back appearance (data not shown).

Discussion

We report that contrary to the general belief, non-self antigens constitute a primary source of peptides recognized by TCRs expressed on T_reg cells and with an affinity high enough to induce an antigenic response. Moreover these TCRs do not show a major enrichment in high-affinity, autoreactive specificities. This conclusion was obtained based on two approaches: an analysis of TCR usage and homeostatic fitness of naive T and T_reg clones in vivo, and an analysis of antigenic specificities of TCRs derived from Foxp3⁺ T_reg cells.

Our comparative analysis showed that 70 percent of the most frequent TCRs on naive T cells were also found on T_reg cells, but with very different frequencies. This disparity in the allocation of shared TCRs was already set in the thymus, suggesting that thymic, not peripheral, events enforce this asymmetric distribution. A similar observation of shared TCRs between regulatory and non-regulatory populations was also reported elsewhere, where seven out of ten most frequent TCRs on CD25- and CD25⁺ CD4⁺ T cells was shared between both analyzed populations (Hsieh et al., 2006). This substantial fraction of shared TCRs suggests that naive T and T_reg cells have, in fact, many shared affinities. In addition, 40 percent of dominant TCRs on T_reg cells were also found on naive T cells. Possibly, this overlap could be larger but remained partially hidden due to vastly different frequencies of the same TCRs and likely, a higher diversity of TCRs on T_reg cells (Pacholczyk et al., 2006). This could also indicate that TCRs used by naive T cells constitute a subset of TCRs expressed on T_reg cells.

The main role of T_reg cells is to suppress autoreactive T cells, which suggests that TCRs on T_reg cells may recognize self antigens as agonists. However, we found that T cells expressing TCRs derived from a T_reg population did not become activated and produced IL-2 upon exposure to self MHC-peptide ligands. Moreover, the majority of these T cell clones responded to non-self antigens that represent endogenously processed peptides and allogenic MHC-peptide complexes. Regardless of whether TCRs were derived from T_reg or naive T cells subpopulations, the frequency of TCRs recognizing non-self peptides was similar. Our data corroborate previous findings that partial reversal of the anergic state of T_reg cells by co-culture in the presence of IL-2 and/or IL-15 allows them to respond to allogeneic, but not syngeneic, APCs in both mice and humans (Dieckmann et al., 2001; Pacholczyk et al., 2002). Here, the only way we were able to identify potentially autoreactive TCRs specific for A^bWT complexes in TCR^mini mice was by cross-referencing A^bWT-reactive TCRs originally found on hybridomas from TCR^miniEp to the database of almost 1500 TCRs previously cloned from TCR^mini mice. This result demonstrated that the frequency of high-affinity, self-reactive TCRs in the T_reg population is very low, although it may be marginally higher than on naive T cells. One could argue whether T cell hybridomas are an appropriate tool to detect autoreactive TCRs. However, T cell hybridomas made from autoreactive T cells continued to manifest the original self-reactivity in vitro and for years T cell hybridomas were used to detect weak and strong agonists, sometimes at very low levels (Hampl et al., 1997; Grubin et al., 1997). In general, the self-reactive hybridomas are produced either after antigen priming or from autoimmune prone strains, but not directly from naïve TCR repertoire, which is expected to have a very low frequency of autoreactive T cells (Yanoma et al., 1988; Buzas et al., 2003).

It has been hypothesized that TCRs on T_reg cells “draw from the same pool of TCRs as autoreactive T cells” and a majority of TCRs on T_reg cells have pathogenic potential (Hsieh et al., 2006). This conclusion was based on the observation that CD25⁺ T_reg cells from wild-type mice express TCRs similar to the receptors expressed on pathogenic autoreactive CD4⁺CD25⁺ T cells from Foxp3⁻ mice. Indeed, in the recently produced Foxp3-GFP knock-in mice the “disabled” T_reg cells expressing GFP left the thymus and partially accumulated in non-lymphatic tissues suggesting direct involvement of these cells in pathogenesis (Gavin et al., 2007; Lin et al., 2007). However, the role of “disabled” T_reg cells in the autoimmune disease observed in Foxp3-deficient mice is unclear, since these cells remain anergic, cannot provoke an autoimmune disease upon adoptive transfer to a lymphopenic host, and deteriorate in the absence of other T cells (Lin et al., 2007). Furthermore, when the same questions were addressed in scurfy mice expressing transgenic GFP under the control of Foxp3 regulatory sequences, the proportion of GFP⁺ in CD4⁺ T cells in sick scurfy and healthy B6 males were similar and the GFP⁺ T cells were activated no more than GFP⁻ T cells (Kuczma et al., submitted). Concurrently, the depletion of Foxp3-expressing T cells in adult animals leads to autoimmune disease with a severity similar to the disease observed in the scurfy mice that is entirely driven by autoreactive T cells from the pool of Foxp3⁻ T cells (Kim et al., 2007). Other evidence for pathogenic autoreactive specificities of T_reg cells is based on an adoptive transfer of naive T cells that are retrovirally transduced with T_reg-derived TCRs and cause autoimmunity manifested by wasting disease (Hsieh et al., 2004). However, onset of colitis depends on an overly aggressive T cell response to a subset of commensal enteric bacteria, and the transfer of naive T cells into germ-free animals does not cause the disease (Gad, 2005; Sartor, 2006). The wasting disease described here was induced by adoptively transferred naive T cells only in animals where class II MHC was able to present foreign antigens, but not in animals where class II MHC was preoccupied with covalently bound self-peptide, confirming that in such an experimental set-up, non-self antigens are a necessity. Hence, the fact that described in other report naive T cells with “regulatory” TCRs cause wasting disease appears to suggest that specificities directed toward non-self (bacterial) antigens are well represented in the T_reg repertoire (Hsieh et al., 2004).

In general, conventional T cells require weak interactions with self MHC-peptide complexes to pass thymic selection and to survive in the periphery, and higher affinity interactions with non-self MHC-peptide ligands to induce activation and effector functions. In contrast, it has been proposed that a commitment of T_reg cells is driven by higher affinity interactions between TCR and self MHC/peptide complexes (Jordan et al., 2001; Apostolou et al., 2002; Hsieh et al., 2004) and that T_reg cells recognize ubiquitous self-ligands with an avidity above that required for selection and survival (Hsieh et al., 2006). These interactions induce T_reg phenotype and Foxp3 expression and skew the specificity of TCR repertoire toward self-antigens. Consequently naive T and T_reg cells express different sets of TCRs that recognize non-self or self-antigens, respectively. Therefore, under these circumstances T_reg cells may use the same self-ligand(s) for positive selection in the thymus and effector function in the periphery, which is fundamentally different from the conventional model of T cell selection and/or activation. Furthermore, if lineage commitment of T_reg cells depends on higher TCR affinity for self-ligand(s), the repertoire of TCRs on T_reg cells would be less diverse, than the repertoire on naive T cells. This is because for a given antigen there will be fewer TCRs with a higher affinity than TCRs with lower affinity. However, our data show that T_reg cells may not only have higher diversity of TCRs than conventional T cells, but also share a substantial portion of TCRs with that population (Fig. 1) (Pacholczyk et al., 2006).

Our findings question the direct role of TCR signaling in the linage commitment of T_reg cells, but support the view that recognition of self MHC-peptide complexes is required for these cells to survive in the thymus and periphery. Subsequently, we would like to offer an alternative model where commitment of T_reg precursors occurs prior to thymic selection by the TCR (Pacholczyk et al., 2002; van Santen et al., 2004; Pennington et al., 2006) and even weak interactions with self-ligands can induce survival of thymic T_reg cells. In this model precursors of T_reg cells (pre-T_reg) possess a unique physiological status and exist as a fraction of late CD4⁻CD8⁻ (DN) or cortical CD4⁺CD8⁺ (DP) thymocytes. This condition may be triggered by a “near-death experience,” different proliferation propensity prior to the DP stage, or a currently unknown signal(s) delivered upon the contacts of thymocyte with other thymocytes or thymic stromal cells. Different physiological status of pre-T_reg cells, not signaling through the TCR, will determine their lineage commitment and TCR distribution. Positive selection of both lineages proceeds on the same pool of self ligands generating unbiased TCRs distribution. Both naive T and T_reg repertoires are primarily directed to recognize non-self antigens with small fractions of potentially autoreactive TCRs. After positive selection, the TCR repertoires on T_reg cells is wider because physiological status of pre-T_reg thymocytes can make positive selection of T_reg cells more promiscuous. Some T_reg cells can express autoreactive TCRs, but in contrast to the current dogma, the self peptide(s) involved in positive selection of T_reg cells is not responsible for activation of their suppressor functions in the periphery. In support of a “pre-comittment” hypothesis, it has been recently reported that Foxp3 expression can occur in pre-selected CD4⁻CD8⁻ thymocytes, when these cells develop in the absence of trans-conditioning by CD4⁺CD8⁺ thymocytes (Pennington et al., 2006).

Experimental Procedures

Mice

TCR^miniA^bWT mice were obtained as previously described (Pacholczyk 2006). TCR^miniA^bEp mice were generated by crossing VαJα/β^Tg A^bWT mice with mice deficient for the invariant chain and endogenous Aβ^b chain, but expressing a single covalent A^bEp complex (Ignatowicz et al., 1996). To eliminate expression of endogenous TCRα chains, all TCR^mini mice used in this study were crossed with mice deficient in endogenous TCRα genes. All TCR^mini mice were 6 to 8 weeks old and were heterozygous for the TCRα mini locus to ensure expression of a single TCR.

All mice were housed under specific pathogen-free conditions in the animal care facility at the Medical College of Georgia. All work involving animals was conducted under protocols approved by the Animal Care and Use Committee at Medical College of Georgia.

Cells preparation and flow cytometry analysis

Single-cell suspensions were prepared from lymph nodes by mechanical disruption through nylon mesh. The following antibodies were used for flow cytometry analyses: anti-CD4-APC, anti-CD8-PerCP, anti-CD25-FITC, anti-CD45RB-PE, anti-Vα2-PE, anti-Vβ14-FITC, anti-TCRβ-APC and anti-Foxp3-APC (all from BD PharMingen or eBioscience). Intracellular staining for Foxp3 expression was performed using intracellular staining kit, according to the manufacturer's protocol (eBioscience). FACS analysis was performed using BD FACSCanto and Diva software (BD PharMingen). Dead cells were excluded by gating of forward and side scatter.

Generation of T cell hybridomas

To generate conventional T cell hybridomas, CD4⁺ T cells sorted from lymph nodes were cultured for 3 days in culture medium in the presence of plate-bound aCD3 mAb and autologus irradiated splenocytes. On day 3, dead cells were removed by gradient centrifugation on Lymphocyte Separation Media (Cellgro, Herndon, VA). The recovered live cells were expanded in the presence of IL-2 for 4 days, and T cell blasts were converted into T cell hybridomas as previously described (Ignatowicz et al., 1996). To generate T cell hybridomas derived from regulatory T cells, we sorted CD4⁺CD25⁺ lymphocytes and induced their proliferation by aCD3 and aCD28 antibodies fixed on the solid surface in the presence of exogenous IL-2 for 10 days (Singh et al, manuscript in preparation). Expanded CD25⁺ T cell blasts were also converted into T cell hybridomas.

HT-2 assay

T cell hybridomas were tested for their responsiveness toward self and non-self APCs by HT-2 assay (Ignatowicz et al., 1996). Briefly, 10⁵ hybridoma cells were incubated with 5×10⁵ splenocytes or 10⁵ bone marrow dendritic cells, from different mice, as indicated. After 24 h, the amount of secreted IL-2 was measured using the detector HT-2 cell line. The proliferation of HT-2 cells in response to IL-2 was measured using 3-(4,5-dimethythiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; Sigma) assay.

MoFlo sorting of CD4⁺CD45^hi T cells and adoptive transfer to lymphopenic hosts

Lymphocytes were stained with appropriate antibodies and were subjected to double sorting on MoFlo sorter (Cytomation) (Pacholczyk et al., 2006). Purity of the CD4⁺CD25⁺ or CD4⁺CD45RB^hi populations was >99% TCRα^−/− mice were injected intraperitoneally with sorted CD4⁺ T cell subpopulations in PBS. Mice received 4 × 10⁵ CD45RB^hi CD4⁺ cells. Following adoptive transfer mice weight was monitored on a weakly basis.

RT-PCR and two-dimensional, fluorescent, single-stranded conformational polymorphism (2D-F-SSCP)

RNA was isolated from sorted populations using Trizol according to the manufacturer's protocol (Invitrogen). cDNA synthesis was done using M-MLV reverse transcriptase and Random Hexamers (all from Promega). cDNA was amplified in a standard PCR reaction with Vα2-specific primers (Va2_287s, CA_29r). 2 μl of PCR product was used as a template for a run-off reaction with a fluorescent primer Va2_287s labeled on the 5' end with Cy3 (synthesized by IDT Inc.). The denatured fluorescent products were then subjected to 2D electrophoresis as previously described (Pacholczyk et al., 2006) Fluorescent images were acquired by scanning the slab gel in a Typhoon 9410 imager (Amersham-Pharmacia). Images were analyzed using Image Master 5.0 Platinum software (Amersham-Pharmacia).

Estimation of most frequent TCRs

Maximum frequency ϕ_max of possibly yet unseen TCR clone was calculated based on equation published elsewhere (Baron et al., 2003):

$$n\left({\varphi }_{\max }-1.65\sqrt{\frac{{\varphi }_{\max }(1-{\varphi }_{\max })}{n-1}}\right)=1,$$ (1)

where n is a total number of accumulated sequences. To estimate boundaries of a real frequency (ϕ_clon) of a TCR clone with observed frequency f_clon = ϕ_max we used equation:

$${\varphi }_{clon}\in \left[f_{clon}-1.96\sqrt{\frac{f_{clon}(1-f_{clon})}{n-1}},f_{clon}+1.96\sqrt{\frac{f_{clon}(1-f_{clon})}{n-1}}\right],$$ (2)

Table 1.

Treg-derived CDR3α sequences from TCRminiEp mice	Used by cells:	Hybridomas reactive to:
		AbWT (Non-self)	AbEp (Self)
CAASAHSNYQLIWG	TN / Treg	+	−
CAADNYQLIWG	Treg	+	−
CAARNYQLIWG	TN / Treg	+	−
CAASNYQLIWG	TN / Treg	+	−
CAASSNSNYQLIWG	Treg	+	−
CAPSNYQLIWG	Treg	+	−
CAASDNSNYQLIWG	Treg	+	−
CAASHYQLIWG	TN / Treg	N/A	N/A
CAARSHSNYQLIWG	Treg	N/A	N/A
CAAFNYQLIWG	TN / Treg	+	−
CAARAHSNYQLIWG	TN / Treg	+	−
CAASGSNYQLIWG	TN / Treg	−	−

Table 2.

Sequences of TCRs reactive to AbWT found in TCRminiEp and TCRmini mice	Distribution on CD4+ T cells in TCRmini mice
	Treg	TN
CAASNYQLIWG	+
CAASAHSNYQLIWG	+
CAGSNYQLIWG		+
CAARSHSNYQLIWG	+	+
CAASDYQLIWG	+
CAASGWDSNYQLIWG	+