Affinity-based Selection of Regulatory T Cells Occurs Independent of Agonist-mediated Induction of Foxp3 Expression¹

Mice

The transgenic mice that express the N3.L2 TCR, the Hb(64-76) peptide or APL under control of the Eα promoter, or the sHEL/Hb construct were previously described (22, 23). Briefly, the N3.L2 TCR recognizes Hb(64-76)/I-E^k or its APLs with varying affinity. CCSP/Hb mice express a chimeric membrane protein containing both hen egg lysozyme (HEL) and hemoglobin (Hb) epitopes under the control of a 2.4 kb fragment of the Clara cell secretory protein (CCSP) promoter. This transgene is expressed in type II pneumocytes as well as in the thymus. When CCSP HEL/Hb mice are crossed with N3.L2 TCR transgenic mice, the lungs of bigenic mice develop peribronchial infiltrates and interstitial pneumonitis (manuscript in preparation).

N3.L2 mice were mated to the appropriate ligand-expressing mice to generate bigenic N3.L2 × Hb crosses that were heterozygote for each transgene. All mice were housed in the animal facility at the Medical College of Wisconsin and handled in accordance with the institutional guidelines.

Flow cytometry

Depending on the yield, single cell suspensions of up to 5 × 10⁶ cells/sample were placed in polypropylene culture tubes (12 × 75 mm; VWR, West Chester, PA) and washed with 3 ml of FACS buffer (PBS supplemented with 0.5% BSA and 0.1% NaN₃). Each sample was subsequently stained using the following protocol. Cells were treated at room temperature for 10 minutes with 0.5 μg Fc block/1 × 10⁶ cells in 100 μl of FACS buffer, then incubated for 30 minutes on ice after adding another 100 μl of FACS buffer containing 0.1 μg/5 × 10⁶ cells of biotinylated or directly labeled antibodies. Cells were washed with FACS buffer and incubated on ice for 20 minutes with a streptavidin-fluorochrome conjugate where appropriate. Cells were washed again, and fixed in PBS-azide (0.1%) plus 1% paraformaldehyde for analysis. For intracellular Foxp3 staining, fixed samples were washed in 1 ml of PBS then permeabilized with two washes in FACS buffer containing 0.1% Triton X-100, which was also used to carry out similar FcR blocking, primary antibody staining and two subsequent washes before resuspension in the fixative. Data were acquired using the LSR-II flow cytometer with FACSDiva software (BD Biosciences), then analyzed using FlowJo software (Tree Star, Inc.).

Peptides and Cytokines

The peptides were synthesized, purified and analyzed, as previously described (24). The peptide sequences, in single-letter amino acid code, are: GKKVITAFNEGLK (N72(wild-type)); GKKVITAFTEGLK (T72); GKKVITAFIEGLK (I72). TGFβ was acquired from R&D Systems.

Antibodies

The antibodies used were Pacific Blue-conjugated anti-mouse CD4 (Caltag); Allophycocyanin-conjugated anti-mouse CD5 (BD PharMingen); Pacific Orange-conjugated anti-mouse CD8a (Caltag); PE-conjugated anti-mouse CD24 (BD PharMingen); PE-TexasRed-conjugated anti-mouse CD62L (Caltag); PE-TexasRed-conjugated anti-mouse CD69 (Caltag); Allophycocyanin- or PE-conjugated anti-mouse TCRβ (BD PharMingen); FITC-conjugated anti-mouse Foxp3 (eBiosciences); Fc block (BD PharMingen); and biotinylated CAb. Cells stained with CAb were subsequently stained with Tricolor streptavidin (BD PharMingen).

Thymic organ culture

Fetal (E16.5), neonatal (1-day old) or reaggregated thymic lobes were each similarly incubated at 37°C, 5% CO₂ for three to seven days on a 0.8-μm filter (Whatman) that was resting on a gelfoam sponge (Pfizer Inc.) in a well of a 12-well plate containing 1 ml RPMI 1640 supplemented with 10% FCS (Hyclone), 2 mM Glutamax (Invitrogen), 50 μg/ml gentamicin (Invitrogen), 100 U/ml penicillin (Invitrogen), and 100 μg/ml streptomycin (Invitrogen). Any peptide and/or cytokine treatments were included throughout the course of the experiment, while media was exchanged after four days for longer cultures. Single-cell suspensions were then prepared for analysis. For reaggregated samples, neonatal lobes were digested at 37°C for 45 minutes in 0.05% trypsin-EDTA (Invitrogen), washed once with media, then resuspended in 4 μl of media to form a slurry that was applied as a droplet on the filter. Reaggregates were either composed of individually reconstituted neonatal thymi or of 2-3 neonatal thymic lobes mixed together with 5 × 10⁶ thymocytes or 1 × 10⁵ peripheral lymphocytes from an adult N3.L2 × Rag^-/- mouse.

Statistics

Two-tailed Student’s t-tests were used to determine significance between two given groups. For analyses containing more than two groups, one-way analysis of variance and Tukey test were applied.

nT_reg Cells in N3.L2 × Ligand Mice

We used the N3.L2 system to study the in vivo development of peptide-specific nT_reg cells over a wide range of ligand affinities. Based on current models, we postulated that N3.L2 T cells could become nT_reg cells if provided a ligand of appropriate affinity in the thymus. Developing thymocytes from N3.L2 mice that coexpressed either the wild-type Hb(64-76) peptide or one of several well characterized APLs (Table I) that span a broad spectrum of affinities were analyzed using flow cytometry (22, 25, 26). The APLs involved amino acid substitutions at position 72, the TCR contact residue in the middle of the peptide/MHC complex (P5 position). Individual substitutions replaced asparagine (N72) to form a weak agonist (T72), a strong antagonist (I72), a moderate antagonist (A72), a null ligand with charge conservation (Q72), or a null ligand with an opposite charge (E72). All Hb(64-76)-related epitopes were expressed as chimeric membrane proteins to target the MHC class II antigen presenting pathway (22).

Analysis of CD4⁺CD8⁻ (CD4SP) thymocytes in N3.L2 mice may be subdivided into cell populations based on the amount of the transgenic TCR on the cell surface, as determined by the level of staining with a clonotypic antibody (CAb^neg, CAb^int, CAb^high). CAb is specific for the unique combination of the transgenic TCRα and TCRβ chains together (27). As CD4SP thymocytes all express a similarly high level of TCRs (data not shown), the amount of CAb staining within the CD4SP-CAb^int gate is inversely related to the quantity of endogenous TCRα chains on the cell surface. Flow cytometric analysis of non-transgenic TCRα chains confirms their presence on CD4SP-CAb^int thymocytes (manuscript in preparation). Consequently, the CD4SP-CAb^int population consists of dual-receptor thymocytes. CAb^highFoxp3⁺ cells do not exist in N3.L2 mice on either the B6.AKR or the Rag^-/- backgrounds (Figure 1A), which is consistent with other models (3, 18). The Eα promoter drives transgene expression in both the cortex and medulla, including the APC populations necessary for the positive and negative selection of T_conv cells (22). Based on published studies, we predicted that expression of the cognate antigen N72 in the thymus of N3.L2 mice would result in the development of nT_reg cells with high levels of the transgenic TCR (18). However, when N72 was expressed under the control of the Eα promoter, all single-receptor CD4SP CAb^high cells were deleted, regardless of their Foxp3 status (Figure 1B). There were several possibilities that could explain these results: 1) the distribution of ligand expression among the various APC populations did not permit N3.L2 nT_reg cells to develop, 2) the interaction between the N72 ligand and the N3.L2 TCR may exceed the affinity threshold for nT_reg cell development, 3) early exposure of N3.L2 thymocytes to the cognate antigen might result in their elimination via negative selection, and 4) the N3.L2 TCR might not be compatible with nT_reg cell development. We designed experiments to investigate these possibilities.

In other systems, ‘tissue specific’ promoters have been used to direct antigen expression, resulting in the negative selection of T_conv cells together with the development of nT_reg cells (18, 28, 29). This result may be due to low-level ligand expression that is limited to certain thymic APC subsets, such as medullary thymic epithelial cells (mTECs) (30). We used the CCSP promoter to drive expression of the ligand in transgenic mice (CCSP/N72). In these mice, expression of the transgene was detectable by RT-PCR only in lung tissue (data not shown). However, CAb^high T_conv cells were deleted in N3.L2 × CCSP/N72 mice, demonstrating low-level thymic expression. In contrast to other models that used a similar strategy (29), CAb^high nT_reg cells were not detected (Figure 1B). We also examined sHEL/N72 mice that co-expressed a secreted form of the antigen under the control of a minimal thymidine kinase promoter (23). In these mice, the antigen circulates at a concentration of ≤0.1 ng/ml, which is sufficient to induce the negative selection of CAb^high T_conv cells in N3.L2 × sHEL/N72 mice (Figure 1B). Despite low level thymic expression and the different route of antigen delivery, no CAb^high nT_reg cells were produced.

If the clonal deletion of all CAb^high cells was a result of inappropriately strong avidity engagements during development, then attenuating the TCR signaling by introducing APLs of lower affinity could provide a thymic environment in which negative selection of these cells is limited to the T_conv subset. To examine this possibility, we crossed N3.L2 mice with a broad spectrum of APL mice and examined nT_reg and T_conv cells in the thymus and periphery. The thymus size correlated with ligand affinity (Figure 1C), indicating progressively less negative selection with coexpression of the lower affinity ligands. However, no significant CAb^highFoxp3⁺ T cell development was observed in any N3.L2 × APL combination, including T72 and I72, which efficiently deleted CAb^high N3.L2 T_conv cells (Figure 1D). Comparable results were observed in the lymph nodes of these mice (data not shown), confirming a lack of rare CAb^highFoxp3⁺ thymocytes that might later become evident by expanding in the periphery. Thus, reducing ligand affinity did not result in nT_reg cell development.

The Influence of Ligand Affinity on nT_reg Cell Selection

When considering all CD4SP thymocytes regardless of the CAb level, we found an increased proportion of Foxp3⁺ cells (Figure 1E, upper left panel) with coexpression of ligands that have high affinity for the N3.L2 TCR (N72, T72, I72). Among these ligands, the total number of nT_reg cells was inversely related to the TCR-peptide-MHC half-life (25) (Figure 1E, lower left panel). In N3.L2 × N72 mice, the total number of nT_reg cells was decreased, consistent with their negative selection. When T72 was coexpressed, the level of nT_reg cells became comparable to that found in the N3.L2 controls. Notably, with I72 coexpression, the total number of nT_reg cells was significantly higher. This increase in cell number was restricted to the CAb^neg compartment. Since CAb^neg cells express the N3.L2 TCRβ chain (Vβ8.3), they have some tendency toward Hb(64-76) peptide reactivity (unpublished data) and hence could be influenced by this ligand or its variants. In agreement with this interpretation, ligand-specific effects were not restricted to the Foxp3⁺ cell populations and were similar in CD4SP Foxp3⁻ thymocytes (data not shown). Reversing the charge of the amino acid residue in the P5 position (E72) also influenced nT_reg cell development, possibly reflecting a global influence on the development of N3.L2β⁺ thymocytes, as evidenced by the significant reductions in the CAb^neg and CAb^int populations (Figure 1E, lower middle panels).

Expression of APLs under the control of the Eα promoter results in both positive and negative selection of N3.L2 T_conv thymocytes, depending upon the affinity of the interaction (22). A priori, there is no reason why this should not also be true for the nT_reg population. In this system of antigen delivery, high affinity interactions uniformly result in negative selection, which should occur even among cells with N72 reactivity derived by expression of the N3.L2 TCRβ chain combined with endogenous TCRα chains. Arguably, the increase in nT_reg and T_conv cell numbers seen in N3.L2 × I72 mice might therefore be the result of positive selection based on lower affinity interactions.

In our study, intermediate N3.L2 TCR levels influenced both nT_reg and T_conv cell development. Recent studies support the idea that nT_reg cell lineage commitment begins prior to Foxp3 expression (15, 16). We considered this possibility by directly comparing the effects of TCR ligation on total and N3.L2 TCR levels between the nT_reg and T_conv cell compartments. First, fold differences respective to N3.L2 control values for each N3.L2 × APL group (Figure 2A) were established, both in terms of the percent (upper panels) and the absolute number (lower panels) of cells within the CAb^neg and CAb^int gates. We then compared the Foxp3⁺ and Foxp3⁻ populations in terms of those fold differences. In the CAb^int gate, the magnitude of the fold increase in the proportion of Foxp3⁺ cells was significantly larger than those found for the Foxp3⁻ population when N72, T72 or I72 was coexpressed. Both Foxp3⁺ and Foxp3⁻ T cells underwent negative selection, as represented by their decreased numbers with higher affinity ligands. However, the extent of this deletion was significantly less for Foxp3⁺ thymocytes, which was consistent with their increased proportions. These data show the two lineages to be intrinsically different in terms of their responses to a given ligand and indicate that developing nT_reg precursors are more resistant to negative selection than their T_conv counterparts.

Interestingly, the pattern of positive fold differences in the proportion of CAb^intFoxp3⁺ T cells at higher affinity was inverted at lower affinity (A72, Q72 and E72). These negative fold differences were accompanied by a significantly decreased number of CAb^intFoxp3⁺ T cells at the lowest tested affinity (E72). In other words, in the presence of lower affinity ligands, forced expression of the N3.L2 transgenic TCR in the CAb^int population resulted in more negative fold differences in the proportion (A: −1.19 vs. +1.02, Q: −1.08 vs. 1.01, E: −1.69 vs. 1.02) and number (A: 1.15 vs. 1.43, Q: −1.01 vs. 1.10, E: −1.57 vs. 1.10) of nT_reg cells, when compared to their Foxp3⁻ counterparts. This indicated that the endogenous TCR-dependent development of nT_reg cells was more easily disrupted by the low-affinity interactions provided by the transgenic receptor compared to T_conv cells. Therefore, nT_reg precursors may also be less sensitive to positive selection in that they require stronger signals to undergo this process. The CAb^neg population confirmed the differential effects of affinity on the nT_reg and T_conv cell populations (Figure 2A, left panel). These trends were also maintained in the periphery (data not shown).

One measure of a thymocyte’s capacity to initiate downstream signaling is the level of TCR on its cell surface (31). For dual-receptor T cells, such as in the CAb^int gate, both TCRs can mediate signaling events, which may influence the outcome of thymocyte selection (32-34). In the CD4SP-CAb^int gate, we used the geometric mean fluorescent intensity of anti-TCRβ and CAb staining as a measure of the overall TCR levels (Figure 2B, upper panel) and N3.L2 TCR levels (lower panel), respectively. When stronger ligands (N72, T72, or I72) were coexpressed, nT_reg cells expressed lower levels of TCRβ compared to their Foxp3⁻ counterparts, while the N3.L2 TCR levels on nT_reg cells were proportionally higher. With lower affinity ligands (A72 or Q72), this bias was reversed in favor of endogenous receptors. The N3.L2 TCR may be expressed as early as the CD4⁻CD8⁻, double negative (DN) stage (data not shown), before positive and negative selection. Therefore, forced expression of the clonotypic receptor early in development has a differential effect on the regulatory and conventional populations, both in terms of the ultimate level of cell-surface TCR and the proportional composition of the transgenic and endogenous receptors.

In addition to influencing thymic development, the N3.L2/APL system creates distinct antigenic environments that impact the peripheral survival of T cells expressing the transgenic TCR. In the lymph nodes, across the spectrum of affinities, the level of CAb staining on CAb^intFoxp3⁺ T cells was generally higher than that found on CAb^intFoxp3⁻ T cells (Figure 2C). This preferential expression of the N3.L2 TCR in the nT_reg population suggested that their affinity requirement for homeostatic survival, which is mediated by continuous TCR ligation (35), is intrinsically different from that of the T_conv population. The relative ‘fitness’ of any T cell subset in a given population can be measured by changes in its proportional representation in that population over time, such as during the maturation process from the thymus to the periphery (36, 37). This comparison assumes that exposures of each population to commensal organisms and other factors that influence T cell homeostasis are identical within each mouse. Therefore, the effect of self-ligand affinity on nT_reg cell fitness may be measured by comparing the nT_reg:T_conv peripheral ratio to its respective thymic ratio within the CAb^int gate of each N3.L2 × APL bigenic mouse. Under normal circumstances, nT_reg cells comprise a greater proportion of CD4⁺ T cells in the periphery than in the thymus. However, in N3.L2 × N72 or T72 bigenic mice, the ratio of CAb^int nT_reg cells to CAb^int T_conv cells in the periphery was smaller than its respective thymic value (Figure 2D). These peripheral versus thymic ratios did not differ from control values when lower affinity ligands were coexpressed. These data indicate that the nT_reg and T_conv cells respond differently to an agonist self-peptide in the periphery. Although the overall numbers were lower relative to control mice, the agonist environment favors the T_conv subset over the nT_reg cells. This was the opposite of what was observed in the thymus and supports the notion that nT_reg cells have relatively blunted responses compared to their Foxp3⁻ counterparts.

nT_reg Cell Development in Thymic Organ Culture

In the instructive model (16, 38), any developing thymocyte may be directed into the nT_reg lineage upon an appropriate TCR-peptide-MHC interaction. Indeed, the TCR repertoires of nT_reg and T_conv cells have some degree of overlap (39-42). It was therefore conceivable that CAb^high thymocytes in our system could upregulate Foxp3 if grown under permissive conditions. Factors that influence the selection of T cell precursors include the dose and timing of ligand presentation (43-50), which may be manipulated in organ culture systems. We investigated the effect of presenting the wild-type ligand to N3.L2 thymocytes at different stages of their development using fetal thymic organ culture (FTOC) and neonatal thymic organ culture (NTOC).

After FTOC of E16.5 thymi treated with 1μM N72 peptide, deletion of the CD4SP compartment was virtually complete. Of the few CD4SP cells that matured (Figure 3A), about two percent upregulated Foxp3. We also observed a ligand specific effect within the DN population, which showed a significant increase in both the proportion and number of Foxp3⁺ cells (Figure 3B). Although the proportion of Foxp3⁺ cells also increased in the other thymic subsets, the total number of these cells was not significantly different from control values. To allow more N3.L2 thymocytes to mature before peptide exposure, thymic lobes were harvested on the day of birth for NTOC. In contrast to FTOC, the neonatal thymi showed no significant reduction of the CD4SP compartment and significantly more CD4SP-CAb^highFoxp3⁺ cells were generated (Figure 3C, 27 vs. 1.82 × 10³, p=0.0058). Notably, the proportions of CD4SP nT_reg cells in the N3.L2 thymi closely matched those found in the B6.AKR thymi (Figure 3D, ~5% of the gate). Even when the system was biased toward one specific TCR-peptide interaction, the proportion of developing nT_reg cells was still limited. Harvesting neonatal thymi three days after birth had no effect on nT_reg cell yield (data not shown).

In order to determine the relationship between peptide concentration and N3.L2 nT_reg cell development, we performed NTOCs with N72 peptide treatments ranging from 0.01 μM to 10 μM. While the overall thymocyte yield was not significantly affected by the peptide dose (data not shown), the distribution of thymic subsets was changed as the peptide concentration increased (Figure 4A). The CD4⁺CD8⁺, double positive (DP) compartment was significantly depleted, which was accompanied by a significant increase in CD4⁻CD8⁺ (CD8SP) cells. These observations are consistent with previous data (51, 52). Although the size of the CD4SP compartment remained unchanged, the proportion and number of Foxp3⁺ cells increased in direct relation to the N72 peptide dose and reached a maximum at a concentration of 1 μM (Figure 4B). Significant increases in the DP-Foxp3⁺ proportions were also observed. NTOC with T72 peptide (weak agonist) treatment also induced Foxp3 expression, although it was markedly lower. In contrast, the antagonist I72 peptide, which is sufficient to induce negative selection of T_conv cells (Figure 1D, (22, 51)), did not result in the upregulation of Foxp3 (Figure 4C).

Mature T cells may be induced to express Foxp3 upon TCR stimulation in the presence of TGFβ (53). To explore whether or not the Foxp3⁺ cells generated during NTOC were being converted from CD4SP-Foxp3⁻ cells that were already present at birth, we set up similar cultures with or without 5 ng/ml TGFβ and analyzed the results after three or six days. While TGFβ treatment had no effect on the overall thymus size (data not shown), the numbers of CD4SP cells significantly increased after six days of culture (Figure 5A, right panel). Within this gate, TGFβ enhanced the percent and number of Foxp3⁺ cells (Figure 5B). This TGFβ-mediated, selective increase in Foxp3⁺ cells was also seen in all N3.L2 thymic subsets (data not shown). While conversion of Foxp3⁻ cells should have been noticeable after three days, we employed reaggregate thymic organ culture (RTOC) techniques to rule out any contribution by conversion to the enhancement of Foxp3⁺ cells seen after six days of culture. The generation of clonotypic nT_reg cells may be recapitulated using RTOC when N3.L2 × Rag^-/- thymocytes are introduced into a neonatal B6.AKR thymic environment in the presence of 1 μM N72 peptide (Figure 5C, second panel). However, incorporating mature lymphocytes in parallel RTOC experiments failed to show conversion regardless of TGFβ treatment (Figure 5C, right two panels). These data collectively showed that thymic APCs do not mediate conversion of mature T_conv cells and that TGFβ enhanced nT_reg cell recovery without inducing Foxp3 expression in mature CD4⁺ cells.

In the RTOC experiments, we observed an increased proportion of Foxp3⁺ cells in both the wild-type and transgenic CD4SP gates. This suggested that disrupting the thymic architecture might enhance nT_reg cell development by increasing access to mTECs or other appropriate APCs (30). Next, we examined the effect of reaggregation per se on nT_reg cell development by trypsinizing and reconstituting individual neonatal thymi before culturing for six days. The results were compared to those obtained in NTOC. The total recovery of thymocytes after RTOC was significantly reduced (Figure 6A). This was predominately due to depletion of the DP subset (Figure 6B). The yield of cells in the other subsets was not affected, except for the CD8SP cells in N3.L2 mice. When we considered only those cells that expressed Foxp3, B6.AKR cultures showed a significant increase in the proportion of these cells while they also trended toward an increase in their total numbers (Figure 6C). This was not the case for the N3.L2 thymi. In B6.AKR thymi, the percent and number of Foxp3⁺ cells in all thymic subsets increased, reaching significance in both the DN and DP gates. DN-Foxp3⁺ thymocytes were detectable only at the DN4 (CD25⁻CD44⁻) stage of development (data not shown). In N3.L2 thymi, peptide-induced Foxp3 expression in the DN population was not apparent in either NTOC or RTOC. Early expression of the N3.L2 TCR bypasses normal β-selection, resulting in fewer DN2 (CD25⁺CD44⁺) and DN3 (CD25⁺CD44⁻) thymocytes and an abnormal population of DN4 cells that are reportedly inefficient at this process (27). Transgenic TCR-mediated β-selection may therefore limit the number of DN4 T cells capable of Foxp3 expression.

To further dissect out the differences between the nT_reg and T_conv populations derived from RTOC, we examined a series of phenotypic markers (TCRβ, CD5, CD24, CD62L, and CD69) at the various stages of thymocyte development (Figure 6D). Most differential expression of these molecules was observed within the DP gate. The DP-Foxp3⁺ cells in both B6.AKR and N3.L2 mice followed a similar phenotypic pattern, which generally expressed higher levels of TCRβ, CD62L, CD69, and lower levels of CD24 when compared to their Foxp3⁻ counterparts. These data are consistent with TCR signaling in DP-Foxp3⁺ cells. In contrast, the relative TCRβ levels on Foxp3⁺ cells within the less mature thymic subsets (DN4 and DP) of B6.AKR mice were consistently higher. This trend was reversed among CD4SP cells where the nT_reg cells expressed lower levels of TCRβ when compared to their T_conv counterparts. Thus, in reaggregate cultures, Foxp3 expression could be induced at the DN4 stage. These data suggested that nT_reg precursors may be capable of Foxp3 induction much earlier than expected and that the normal thymic architecture plays a role in controlling access of developing thymocytes to the relevant APC population(s). Even under the permissive reaggregate culture conditions, only about 10% of CD4SP cells expressed Foxp3. These data further support the conclusion that nT_reg cells make up a distinct lineage that differs from the T_conv lineage and that nT_reg precursors are competent to express Foxp3 immediately following β-selection.