Eomes identifies thymic precursors of self-specific memory-phenotype CD8⁺ T cells

The CD8-MP TCR repertoire is distinct and recurrent

Despite comprising >5% of the CD8⁺ T cell repertoire in unprimed specific pathogen free (SPF) and germ-free (GF) mice (Extended Data 1), little is known about the array of TCRs expressed by CD8-MP cells. In particular, it is unknown whether the CD8-MP TCR repertoire has limited diversity, whether it varies stochastically from mouse to mouse, and whether it overlaps with the TCR repertoire expressed by naïve-phenotype CD8⁺ T cells. To comprehensively examine these questions, we sequenced the complete TCR repertoires of naïve-phenotype CD8⁺ T cells (CD8-Naïve cells, CD44^loCD122⁻) and memory-phenotype CD8⁺ T cells (CD8-MP cells, CD44^hiCD122⁺) using previously established methods^{14, 15}. We focused our analysis on CD8-MP cells from C57BL/6 mice, in which interleukin 4 (IL-4)-induced CD8⁺ memory cells^{1, 3, 16} make minimal contributions to the peripheral CD8-MP pool^{12, 16}. CD8-Naïve and CD8-MP cells from the pooled spleen and lymph nodes were purified by cell sorting from five different 9-week-old male mice expressing a fixed transgenic TCRβ chain on the C57BL/6J (B6) background. By fixing the TCRβ chain, a complete survey of the TCRαβ repertoire can be obtained by sequencing the endogenous Tcra chains using the iRepertoire platform^{14, 17}. This approach yielded approximately 6.5 × 10⁵ Tcra sequence reads per sample, providing a broad survey of TCR usage (Supplementary Table 1). Our data revealed the striking finding that the TCR repertoire of CD8-MP cells is largely distinct from that of CD8-Naïve cells, and was highly recurrent from mouse to mouse (Fig. 1). Comparative analysis of TCR frequency identified 493 recurrent Tcra chains that were significantly overrepresented in the CD8-MP subset relative to the CD8-Naïve subset (Fig. 1a, right arm of volcano plot). Cumulatively, these TCRs accounted for 65 +/− 1% of the CD8-MP TCR repertoire, demonstrating the broad extent of repertoire skewing. The recurrent nature of the CD8-MP repertoire was further illustrated using the Morisita-Horn (MH) similarity index, for which a value of 1 denotes identity and a value of 0 indicates complete dissimilarity. Pairwise comparison of the CD8-MP TCR repertoires from five different mice revealed a mean MH index of 0.92 +/− 0.03 (Fig. 1b), demonstrating that distinct T cell clones were reproducibly directed into the CD8-MP subset. By contrast, pairwise comparisons of CD8-Naïve vs. CD8-MP repertoires between the five mice revealed a mean MH index of 0.06 +/− 0.01, indicative of minimal overlap. The concepts that the CD8-MP TCR repertoires were highly recurrent and distinct from the CD8-Naïve TCR repertoire are also illustrated in Fig. 1c, which plots the frequency of the 10 most prevalent CD8-MP TCRs (left) and 10 most prevalent CD8-Naïve TCRs (right) within the CD8-Naïve and CD8-MP subsets in individual mice. Lastly, analysis of repertoire complexity using the Shannon diversity index demonstrated that the CD8-MP TCR repertoire was less diverse than the CD8-Naïve repertoire (Fig. 1d), with an average complexity of 6,979 TCR complementarity determining regions 3 (CDR3) segments for the CD8-MP subset and 24,443 CDR3s for the CD8-Naive subset. Analyses of CDR3α hydrophobicity and amino acid length revealed no consistent differences between the CD8-Naïve and CD8-MP TCRs (Extended Data 2). Collectively, these data suggest that CD8-MP differentiation is a robust TCR-directed process that occurs in a highly reproducible, orchestrated fashion.

CD8-MP differentiation is a TCR-directed process

To study representative CD8-MP-biased specificities at the clonal level, we generated a series of monoclonal TCR “retrogenic” (TCRrg) mice expressing a single TCR of interest, and used T cells from these mice to perform both in vivo and ex vivo studies. In this study, we use the term “CD8-MP TCR” to define a TCR that is preferentially expressed by CD8-MP cells, and “CD8-Naïve TCR” for receptors skewed to the CD8-Naïve subset. For TCR nomenclature, we depict the amino acid sequence of the CDR3 segment of a given TCRα chain (Fig. 1c), and denote a TCR clone using a three-letter code reflecting the amino acids at positions 3-5 of the CDR3α. For example, TCRrg cells with CDR3α sequence AASMNYNQGKLI are denoted “SMNrg” cells. Our work focused on the study of four recurrent CD8-MP and four recurrent CD8-Naïve TCRs identified in Fig. 1, which were chosen because they span a range of frequencies and utilize distinct V and J-region segments (Supplementary Table 2). TCRrg mice were generated as previously described¹⁸. Briefly, bone marrow (BM) cells from Tcra^−/− Cd4-Cre⁺ TCRβtg⁺ CD45^.2/.2 mice were retrovirally transduced with a vector in which the genes encoding a TCRα chain of interest and an IRES-Thy1.1 reporter are preceded by a loxP-flanked stop cassette¹⁹. Transduced BM was engrafted along with wild-type filler BM cells into lethally irradiated 6-8-week-old CD45^.1/.1 B6.SJL hosts, and mice were analyzed >6 weeks post-engraftment. Expression of Cd4-Cre at the CD4⁺CD8⁺ stage of thymic maturation induces TCRα expression by TCRrg cells, recapitulating the natural kinetics of TCR expression during development. Resulting TCRrg T cells were identified using the Thy1.1 and CD45.2 markers.

Using this approach, we found that a substantial fraction of T cells expressing CD8-MP TCRs adopted the CD44^hiCD122⁺ memory phenotype in the periphery of TCRrg hosts, whereas very few TCRrg cells expressing CD8-Naïve TCRs acquired this phenotype (Fig. 2a-c, Extended Data 3). These findings validate the TCR sequencing and confirm the notion that CD8-MP differentiation is a TCR-instructed process. Notably, in TCRrg mice expressing CD8-MP TCRs, we found that not all peripheral TCRrg cells adopted a CD44^hiCD122⁺ phenotype at the time of analysis, and that the fraction of TCRrg CD44^hiCD122⁺ cells varied from mouse to mouse (Fig. 2a-c). A potential explanation for the lack of complete skewing to the CD8-MP phenotype is the existence of limited niches supporting CD8-MP differentiation, which may be overloaded in TCRrg mice harboring large numbers of monoclonal T cells. To define the phenotype and anatomical distribution of select CD8-MP clones, we generated TCRrg mice expressing CD8-MP TCRs at low clonal frequency (“low frequency” TCRrg mice; see Methods). In such mice, when TCRrg cells were present at frequencies of <15% of CD8⁺ T cells, the majority (>65%) of TCRrg cells expressing a CD8-MP TCR exhibited the CD44^hiCD122⁺ phenotype (with nearly all cells shifted away from a naive phenotype), whereas TCRrg cells expressing CD8-Naive TCRs did not (Extended Data 4). These results are consistent with the existence of limited niches supporting the differentiation of CD8-MP clones. Furthermore, peripheral CD8-MP TCRrg cells from these mice adopted a central memory-like phenotype characteristic of polyclonal CD8-MP cells, including high expression of CD44, CD122, CD62L, CD127, and Eomes, and low expression of CD49d (Fig. 2d). These collective data demonstrate that the expression of CD8-MP TCRs in TCRrg hosts recapitulated differentiation into the CD8-MP subset, validating the TCRrg approach and providing further evidence that CD8-MP differentiation is a TCR-directed process.

CD8-MP T cell clones exhibit self-reactivity

Our TCR repertoire and TCRrg mouse analyses suggest that CD8-MP differentiation is antigen driven. Since CD8-MP cells are abundant in naïve mice that have never been exposed to known foreign pathogens, we hypothesized that CD8-MP cells may be reactive to endogenous self-ligands. In TCRrg mice, we found that CD8-MP clones were equally distributed in all secondary lymphoid organs examined (Extended Data 5) and exhibited elevated percentages of proliferative cells as measured by Ki67 staining, suggestive of active sensing of ligands at steady state (Fig. 3a-c). To gain insight into the antigen specificity of CD8-MP clones, we assessed the ex vivo reactivity of purified naïve-phenotype (CD44^loCD122⁻) CD8⁺ T cells isolated from TCRrg mice expressing CD8-MP or CD8-Naïve TCRs. We purified naïve-phenotype CD44^loCD122⁻ from TCRrg mice expressing each of the eight TCRs, and used these cells as a probe for antigen in in vitro stimulation assays. We found that the four CD8-MP clones underwent proliferation upon co-culture with splenic dendritic cells (DCs) and recombinant IL-2, whereas the four CD8-Naïve clones did not (Fig. 3d,e). This reactivity was abolished by the addition of anti-MHC-I blocking antibodies (Fig. 3d,e), indicative of reactivity to ligands displayed by classical MHC-I molecules. Notably, the reactivity of CD8-MP clones was not impaired using splenic DCs isolated from germ-free mice (Fig. 3d,e). Thus, our data provide direct evidence that four canonical CD8-MP TCRs examined confer overt reactivity to endogenous self-ligands presented by splenic DCs in the context of classical MHC-I molecules.

Our cumulative data suggest that CD8-MP differentiation is an orchestrated TCR-dependent process driven by reactivity to self-ligands. We next set out to define the stage at which CD8-MP differentiation is triggered. Previous studies report that CD44^hiCD122⁺ CD8⁺ T cells first appear in the periphery of neonatal B6 mice¹², suggesting that CD8-MP cells differentiate in the periphery^{1, 3}. Congruent with this idea, we found that CD44^hiCD122⁺ cells were not detected amongst polyclonal GFP⁺ CD8⁺CD4⁻ (“CD8 single-positive”, hereafter referred to as “CD8SP”) thymocytes from adult Rag2-green fluorescent protein (GFP) reporter mice (Fig. 4a), in which GFP expression marks newly developing thymocytes that have recently undergone Rag2-dependent TCR rearrangement²⁰.

To address this question at the clonal level, we utilized our TCRrg approach to track the developmental trajectories of distinct CD8⁺ T cell clones that are destined to adopt a CD8-MP phenotype. For our thymic analyses of TCRrg mice, we used CD73-negativity as a surrogate marker for Rag2-GFP⁺ cells, as >97% of CD73⁻ cells are Rag2-GFP⁺ in the thymus (²¹ and Extended Data 6a). Consistent with observations for polyclonal cells, we found that TCRrg cells expressing CD8-MP TCRs did not adopt a CD44^hiCD122⁺ phenotype in the thymus (Extended Data 6b). However, CD8-MP TCRrg thymocytes exhibited elevated percentages of Ki67⁺ cells and increased densities of CD5 when compared to CD8-Naïve TCRrg thymocytes, suggestive of elevated TCR signaling in the thymus (Extended Data 6c-f). Consistent with this observation, we found that CD11c⁺ DCs isolated from the thymus stimulated the in vitro proliferation of TCRrg cells expressing CD8-MP, but not CD8-Naïve TCRs (Fig. 4b), indicating that ligands recognized by CD8-MP clones are displayed by DCs in the thymus. Comparative analysis of TCRrg thymocytes revealed that CD8-MP clones at the CD4⁺CD8⁺ stage displayed reduced densities of the CD4 and CD8 co-receptors (Extended Data 6g), a potential indicator of clonal deletion²². However, the findings that CD8-MP clones exhibited negligible staining for cleaved Caspase 3 (a marker of ongoing apoptosis, Extended Data 6g) and were readily detected within polyclonal and monoclonal repertoires suggests that CD8-MP clones are not substantially impacted by clonal deletion.

CD8-MP clones upregulate Eomes in the thymus

Collectively, the above results suggest that CD8-MP cells encounter endogenous self-ligands in the thymus. Thus, we considered the possibility that CD8-MP differentiation is triggered by the recognition of self-ligands in the thymus, and that the upregulation of the CD44^hiCD122⁺ phenotype is delayed until cells emigrate to the periphery. To test this idea, we looked for early hallmarks of CD8-MP differentiation that lie upstream of CD122 upregulation. Our analysis focused on Eomes, a transcription factor that is highly expressed by peripheral CD8-MP cells, is required for the differentiation and/or survival of CD8-MP cells²³, and promotes CD122 upregulation by direct binding to the Il2rb promoter²⁴. Strikingly, in TCRrg mice expressing CD8-MP TCRs, a fraction of TCRrg CD8SP thymocytes upregulated expression of Eomes, whereas Eomes upregulation was not observed for TCRrg thymocytes expressing CD8-Naïve TCRs (Fig. 4c-e). These findings suggest that the differentiation of many CD8-MP clones is triggered during T cell maturation in the thymus, prior to upregulation of the CD44^hiCD122⁺ phenotype in the periphery. To examine this phenomenon further, we determined whether there are saturable niches driving Eomes upregulation by select CD8-MP clones during thymic maturation, as has been described for Foxp3⁺ regulatory T cells^{25, 26}. To do this, we generated a series of low-frequency TCRrg mice expressing the SAV CD8-MP TCR, and calculated the extent of Eomes upregulation at varying clonal frequencies. This approach demonstrated that the fraction of TCRrg thymocytes expressing Eomes increased with decreasing clonal frequency (peaking at >90%, Fig. 4f,g), suggesting the existence of a saturable niche supporting Eomes upregulation by the SAV CD8-MP clone. In contrast, we did not observe this phenomenon with the control SMN CD8-Naïve TCR (Fig. 4f,g), further highlighting the TCR-dependency of Eomes upregulation.

Next, we performed additional phenotypic analyses to examine hallmarks of antigen sensing at different stages of thymic maturation. To do this, we used a staging approach defined previously²⁷, in which post-selection TCRβ^hiCCR7⁺ thymocytes are sub-divided into cells of progressing maturational stage, ranging from semi-mature (CD69⁺MHC-I⁻), mature-1 (CD69⁺MHC-I⁺), and mature-2 (CD69⁻MHC-I⁺) (Fig. 4h). Notably, we found that for developing polyclonal T cells and the CD8-MP SAVrg clone, Eomes⁺ cells primarily exhibited a CD69⁻MHC-I⁺ mature-2 phenotype (Fig. 4i), suggesting that Eomes expression is induced in the latest stages of thymic maturation, when cells are known to reside in the medulla.

Eomes identifies polyclonal thymic precursors of CD8-MP T cells

Next, we aimed to determine whether the principles observed using our clonal approach extended to polyclonal T cell populations in wild-type mice. Specifically, we hypothesized that Eomes-expressing CD8SP thymocytes represent thymic intermediates that are destined to upregulate the CD44^hiCD122⁺ CD8-MP phenotype following emigration to the periphery. To test this hypothesis, we purified Eomes-GFP⁺ or Eomes-GFP⁻ CD73⁻ CD69⁻ CD8SP thymocytes from 4-week-old Eomes-GFP reporter mice²⁸, and transferred these cells separately into the periphery of congenically disparate wild-type hosts via intravenous injection. 3 weeks post-transfer, we found that ~47% of the Eomes-GFP⁺ donor cells recovered from recipient mice exhibited a CD44^hiCD122⁺ phenotype, compared to ~7% of Eomes-GFP⁻ donor cells (Fig. 5a-c), indicating that Eomes-expressing CD8SP thymocytes are enriched for thymic intermediates with the potential to upregulate the CD8-MP phenotype in the periphery. To determine whether Eomes⁺ CD8SP thymocytes upregulate CD44 and CD122 when thymic egress is pharmacologically delayed, we utilized FTY720 to block sphingosine 1-phosphate receptor-1 (S1PR1) and prevent thymic egress. We treated B6 mice with FTY720 or PBS control for 5 days and assessed the phenotype of CD8SP thymocytes at day 6. This treatment induced a marked increase in the percentage and number of Eomes-expressing CD8SP thymocytes (Fig. 5d-f), but did not induce an increase in the fraction of Eomes⁺ thymocytes that exhibit a CD44^hiCD122⁺ phenotype (Fig. 5g-i). This finding is consistent with a model in which CD8-MP differentiation is triggered in the thymus, but requires a subsequent consolidation phase that can only be conferred in the periphery.

CD8-MP cells infiltrate murine prostate tumors

As introduced above, there remains a lack of available markers to distinguish CD8-MP cells from conventional CD8⁺ effector or memory T cells in the context of immune activation. Thus, the contribution of CD8-MP cells to the immune response to human and murine cancers is undefined. To examine this question from a unique perspective, we co-transferred congenically disparate polyclonal CD8-Naive and polyclonal CD8-MP cells into 2-month-old TRAMP males, in which transgenic expression of a model oncogene drives the development of prostatic adenocarcinoma with high penetrance²⁹. Four months later, when the mice developed advanced prostate tumors, we analyzed the fate of the transferred cells. We found that donor CD8-MP cells constituted a substantial fraction of the tumor-infiltrating CD8⁺ T cells, ranging from 2-17% of intratumoral CD8⁺ T cells (Fig. 6a-f). Strikingly, the majority of donor CD8-MP cells expressed high densities of the inhibitory receptor PD-1 (Fig. 6a-f).

To further address the role of CD8-MP cells in anti-tumor immunity, we utilized our TCR sequencing approach to determine the extent to which CD8-MP-skewed clones contribute to the pool of tumor-infiltrating lymphocytes (TILs) in TRAMP mice. To do this, we isolated CD8⁺ T cells from the prostate tumors of five 27-week-old TRAMP^+/− males expressing the fixed TCRβtg chain, and subjected these samples to deep Tcra sequencing (Supplementary Table 1). The survey identified numerous CD8⁺ T cell clones that are recurrently enriched in TRAMP prostate tumors (Fig. 6g, bottom). In addition, by examining the frequency of these intratumoral clones in the CD8-MP and CD8-Naive TCR data sets derived from the secondary lymphoid organs of tumor-free mice, we found that two of the ten most prevalent intratumoral clones were skewed to the CD8-MP subset (Fig. 6g, red boxes), and that the 8 remaining clones were rare within both the CD8-MP and CD8-Naive data sets. Notably, one of these CD8-MP-skewed clones was the “SAT” (AASATNAYKVI) clone examined elsewhere in this study. Thus, the use of comparative TCR profiling revealed that TRAMP prostate tumors drive the recurrent enrichment of self-specific “tumor-associated” CD8-MP clones that are uncommon in the periphery but are selectively enriched in prostate tumors. These collective findings demonstrate that recurrent CD8-MP clones make measurable contributions to the tumor-infiltrating T cell pool in TRAMP mice and express high densities of PD-1, suggesting that intratumoral CD8-MP cells may functionally impact anti-tumor immunity and may be directly impacted by anti-PD-1 or anti-PD-L1 checkpoint blockade antibodies.

Taken together, our findings demonstrate that CD8-MP differentiation is a robust TCR-directed process that is triggered by the recognition of self-ligands in the thymus, that Eomes identifies thymic precursors with a propensity to upregulate the CD8-MP phenotype in the periphery, and that CD8-MP cells make measurable contributions to the immune infiltrate of autochthonous prostate tumors.

Mice

The following mice were purchased from the Jackson Laboratory, and bred and maintained at the University of Chicago under specific-pathogen free conditions: C57BL/6J (B6) mice, CD45^.1/.1 B6.SJL-Ptprc^a Pepc^b/BoyJ mice, Rag1^−/− B6.129S7-Rag1^tm1Mom/J mice, Tcra^−/− B6.129S2-Tcra^tm1Mom/J mice, CD4-Cre B6.Cg-Tg(Cd4-cre)1Cwi/BfluJ mice and TRAMP C57BL/6-Tg(TRAMP)8247Ng/J mice. “TCRβtg” mice expressing a fixed TCRβ chain of sequence TRBV26-ASSLGSSYEQY were generated as described previously¹⁵. Rag2-GFP C57BL/6-Tg(Rag2-EGFP)1Mnz/J mice were received from the laboratory of M. Nussenzweig at Rockefeller University²⁰. Eomes-GFP C57BL/6-Eomes^tm1.1Twa/Cnbc mice were received from the laboratory of T. Walzer at Inserm²⁸. All mice were bred and maintained in accordance with the animal care and use regulations of the University of Chicago. Germ free C57BL/6 mice were housed at the University of Chicago gnotobiotic facility under strict germ-free conditions. Both male and female mice were used across individual experiments. Mice were not randomized for assignment to experimental group and experiments were not conducted in a blinded fashion.

Antibodies, Flow Cytometry, and FACs

All antibodies used were purchased from BioLegend or Fisher Scientific. Cells were stained with conjugated antibodies specific for the following proteins (with clone name in parentheses): Active caspase-3 (C92-605), CCR7 (4B12), CD4 (GK1.5, RM4-4, or RM4-5), CD5 (53-7.3), CD8β (YTS156.7.7), CD44 (IM7), CD45.1 (A20), CD45.2 (104), CD49d (R1-2), CD62L (MEL-14), CD69 (H1.2F3), CD73 (TY/11.8), CD122 (TM-β1), CD127 (A7R24), Eomes (Dan11mag), H-2K^b (AF6-88.5), Ki67 (SolA15), PD-1 (RMP1-30), TCRβ (H57-597), and Thy1.1 (OX-7). Cells were stained for 20 min on ice in staining buffer (phosphate-buffered-saline with 2% FCS, 0.1% NaN₃, 5% normal rat serum, 5% normal mouse serum, 5% normal rabbit serum (all sera from Jackson Labs), and 10 μg/mL 2.4G2 antibody). Intracellular staining for Eomes, Ki67, and active Caspase3 was performed using fixation and permeabilization buffers from eBioscience. Cells being stained for CCR7 were incubated for 30 min at 23°C with surface stain antibodies in 2% FCS in PBS. Flow cytometry was performed on an LSR Fortessa (BD Biosciences) and data was analyzed using FlowJo software (Tree Star). FACS was performed using a FACSAria (BD Biosciences). An example of the gating strategy is shown in Extended Data 7.

TCR sequence analysis

CD8-Naïve (CD44^loCD122⁻) and CD8-MP (CD44^hiCD122⁺) T cells were purified by FACS from 9-week-old TCRβTg males. To identify recurrent, tumor-infiltrating CD8⁺ T cell clones, CD8⁺ T cells were purified by FACs from the prostate tumors of 27-week-old TRAMP^+/− TCRβTg males. RNA was extracted from purified subsets and subjected to complete Tcra sequencing using the Amp2Seq service from iRepertoire, a platform based on semi-quantitative multiplex PCR coupled with Illumina sequencing. This approach allows analysis of the complete TCRα repertoire, regardless of variable-region usage. Typically, >6.5 × 10⁵ TCR sequence reads were obtained per sample. TCRs were analyzed solely based on the predicted CDR3 sequence, regardless of V-region usage. We first filtered for TCRs with CDR3 segments between 7 and 17 amino acids in length. In order to focus on recurrent TCRs, we then removed those TCRs with counts that were less than ten in N_s-1 samples where N_s was the size of the smaller group. The differential analysis was done using R/Bioconductions edgeR package. The standard edgeR normalization function calcNormFactors was called. For the differential testing the GLM method was used. Since the data consisted of paired samples from each of several mice a paired test was done with the following design:

• design ← model.matrix( ~ mouse + group )

where group is the “treatment” grouping variable and mouse is the mouse id. We then processed the data using the GLM version of the dispersion estimators:

• d ← estimateGLMCommonDisp(d,design)

• d ← estimateGLMTrendedDisp(d,design)

• d ← estimateGLMTagwiseDisp(d,design)

and then for differential testing:

• fit ← glmFit(d,design)

• results ← glmLRT(fit,coef=group.cont)

where group.cont is the contrast that selects for comparisons between groups. For the purposes of computing average expression in a natural scale, the normalized counts were rescaled to the original size of the dataset. This was done by setting a scaling factor that is equal to the geometric mean of the total sample counts for the samples. Significant CDR3s were filtered to an FDR less than 0.05. To analyze TCR repertoire similarity, we also analyzed CDR3 elements using the Morisita-Horn similarity index⁴¹. The grand average of hydropathy (GRAVY) value (http://www.gravy-calculator.de) for CDR3 regions was calculated by taking the sum of hydropathy values of all amino acids divided by the protein length.

Retrovirus production, infection, and generation of TCRrg mice

TCRrg mice were generated as described previously¹⁸. Briefly, Tcra sequences encoding TCRα chains of interest were cloned into a modified retroviral construct^{18, 19}. Plat-E cells⁴² were used to generate retrovirus. Tcra^−/− CD4-Cre⁺ TCRβtg⁺ mice on a C57BL/6 background were injected with 5-fluorouracil (APP Pharmaceuticals) 3 days prior to bone marrow harvest. Bone marrow cells were cultured for 2 days in X-Vivo 10 (Lonza) containing 15% FCS, 1% penicillin/streptomycin, mouse SCF, mouse IL-3 and mouse IL-6 (BioLegend). Cells were infected with retrovirus by spinfection in the presence of 6 μg/mL polybrene (EMD Millipore) and cultured for an additional 24 h (TCRrg bone marrow). TCRrg bone marrow from one mouse was then mixed with 5 × 10⁶ freshly harvested bone marrow "filler" cells from Rag1^−/− mice and injected intravenously into one lethally irradiated (800 rad) CD45^.1/.1 B6.SJL recipient mouse. "Low frequency" TCRrg mice were generated by mixing 12-30% cultured TCRrg bone marrow with 70-88% CD45^.1/.11 B6.SJL bone marrow. 5 × 10⁶ bone marrow cells were injected intravenously into one sub-lethally irradiated (500 rad) B6.SJL recipient mouse. TCRrg and “low frequency” TCRrg mice were analyzed 6-11 weeks after bone marrow reconstitution.

In vitro T cell stimulation assay

CD8⁺ T cells from pooled spleen and lymph nodes (axillary, brachial, cervical, inguinal, mesenteric, and periaortic) from TCRrg mice were enriched for CD8⁺ T cells by MACS (Miltenyi Biotech), and Thy1.1⁺ CD44^loCD122⁻ CD8⁺ T cells were purified by FACs. Thy1.1⁺ CD44^loCD122⁻ CD8⁺ TCRrg cells were CellTrace-Violet (ThermoFisher) labeled per manufacturer instructions with slight modification. Briefly, cells were pelleted, resuspended in CellTrace-Violet (CTV) at 1:1000 dilution and incubated for 20 min at 37°C. The reaction was quenched by the addition of 13 mL of complete culture media. To isolate dendritic cells, spleens were isolated from 6-week-old SPF or GF C57BL/6 mice and thymi were isolated from 3 to 4-week-old SPF C57BL/6 mice. Spleens or thymi were injected and digested with Liberase TL (400 μg/mL, Roche) and DNase (800 μg/mL, Roche) in RPMI for 30 min at 37°C. For thymic APC isolation, EDTA (10 mM) was added to digests and enriched by layering digested thymocytes on top of a discontinuous Percoll gradient (GE Healthcare) at 1.115 g/mL in PBS, followed by centrifugation at 1,350 × g for 30 min and isolation of cells settling at the Percoll interface⁴³. APCs were enriched from spleens or thymi for CD11c⁺ cells by MACS-based (Miltenyi Biotech) positive selection. 1 × 10⁴ CTV-labeled T cells were co-cultured with 5 × 10⁴ CD11c⁺ APCs and 100 U/mL recombinant mouse interleukin-2 (IL-2). Where indicated, anti-MHC-I (H2) blocking antibody (clone M1/42.3.9.8, BioXCell) or rat IgG2a isotype control antibody (clone 2A3, BioXCell) was added to indicated cultures at a final concentration of 500 μg/mL. Cell cultures were set up in 384-well ultra-low attachment, round-bottom plates (Corning). Dilution of CTV was assessed by flow cytometry on day 5.

Thymocyte adoptive transfer experiments and FTY720 treatment

Thymi were isolated from 4-week-old CD45.2⁺ Eomes-GFP⁺ mice and CD4⁺ thymocytes cells were depleted by incubating thymi with anti-mouse CD4 antibody conjugated to Biotin (clone RM4-5, BioLegend) for 10 min at 23°C, followed by incubation with Streptavidin beads (StemCell) for 3 min at 23°C. Cells were then placed on the EasySep magnet (StemCell) for 2 min and the supernatant was isolated. CD4⁻ CD73⁻ CD69⁻ mature CD8β⁺ Eomes-GFP⁺ and Eomes-GFP⁻ thymocytes cells were purified by FACs and 1 × 10⁵ GFP⁺ or GFP⁻ cell suspensions were transferred intravenously into CD45.1⁺ B6.SJL hosts. 3 weeks later, CD8⁺ T cells from pooled spleen and lymph nodes (axillary, brachial, cervical, inguinal, and periaortic) from host mice were enriched for CD8⁺ T cells by MACS (Miltenyi Biotech) and the phenotype of transferred cells was analyzed by flow cytometry. To prevent thymic egress of developing T cells, B6 mice were injected I.P. every other day for 5 days with 140 μg FTY720 (Sigma) in 50 μL diH₂0 or PBS control. The phenotype of thymocytes was analyzed by flow cytometry on day 6.

Lymphocyte isolation from TRAMP prostate tumors

Whole male genitourinary tracts were isolated and prostate lobes (anterior, dorsolateral, and ventral) were separated by microdissection from 24-week-old TRAMP^+/+ or 27-week-old TRAMP^+/− TCRβTg males. Prostate lobes were injected and digested with Liberase TL (10 mg/mL, Roche) and DNase (20 mg/mL, Roche) in RPMI for 30 min at 37°C. Digested tissue was mechanically disrupted with frosted microscope slides and viable lymphocytes were enriched using Histopaque 1119 (Sigma). CD8⁺ T cells were analyzed by flow cytometry or purified by FACs.

Statistical Analysis

Data were analyzed using Prism software (GraphPad). Significance testing was performed using the nonparametric Mann-Whitney test (two-tailed) or one-way ANOVA with Bonferroni’s multiple comparisons test (two-tailed). Statistical analysis of TCR sequencing data was performing with R (The R Project for Statistical Computing) using EdgeR-based methods (see “TCR sequence analysis” section for script details). Statistical testing on CDR3 length and GRAVY distributions was performed using R. Significance testing was performed using both the paired Wilcoxon signed-rank test (two-tailed), which tests for distribution shift, and the Kolmolgorov-Smirnov test, which tests for differences in distribution shape.