CD5 expression correlates with increased TCR signaling and effector potential of islet infiltrating T cells

In order to observe the distribution of effector functional potential in polyclonal islet infiltrating CD4⁺ and CD8⁺ T cells, we performed multiparameter spectral cytometry analysis of T cells isolated from infiltrated pancreatic islets of prediabetic NOD female mice that were restimulated in vitro to elicit the full spectrum of effector functions. Unbiased UMAP analysis of flow cytometric data based on a combination of 22 T-cell lineage markers and effector molecules revealed 10 major cell groups within TCR⁺ cells (Fig. 1A and B, Supporting Information Fig. S1). The populations were broadly distinguished based on CD4 or CD8 expression, and more specifically based on their effector profile, such as cell surface markers of activation and cytokine production. These populations included CD4 and CD8 effector T cells characterized in part by the expression of Tbet and production of IFN-γ (p2 at ∼16% and p9 at ∼8%, respectively, Fig. 1A–E). Expression of Foxp3 transcription factor was used to identify CD4⁺ Treg populations (p4, p5, and p6, Fig. 1C). The three CD4⁺Foxp3⁺ populations were distinguished based on markers of central Tregs (cTreg – CD62L, CCR7), activated effector Tregs (eTreg – CD62L^–CD11a^hi), and a small subpopulation of terminally differentiated or exhausted Tregs (exTreg – KLRG1⁺) (Supporting Information Fig. S1B). Interestingly, population 3 consisted of Foxp3 negative CD4⁺ T cells with some characteristics of type 1 regulatory (Tr1) cells, including production of IL-10 and expression of Tim3 and Lag3 (Population 3, Fig. 1D and F) [19]. CD8⁺ T cells also included a regulatory subpopulation characterized by the expression of Foxp3 and IL-10 (Population 10, Fig. 1C and D, Supporting Information Fig. S1B). Perhaps most striking, was the presence of nonactivated or naïve T cells (CD4⁺ p1, CD8⁺ p7 and p8). These represented a high frequency of all CD4⁺Foxp3^– T cells (35% of CD4⁺Foxp3^–) and CD8⁺ T cells (50% of CD8 T cells). The naïve-like populations were negative for most of the markers associated with T-cell activation and differentiation, including IL-21, perforin, PD-1, Lag3, TIGIT, Tim3, and Bcl6. Though these cells upregulated CD69 in response to in vitro stimulation, a marker of early T-cell activation, suggesting that they are functional (Fig. 1H). Given the observed heterogeneity of T-cell differentiation and effector function in the infiltrated T cells, we wondered how these phenotypes reflected the level of T-cell self-reactivity for pancreatic antigens and the downstream TCR signaling. CD5 is upregulated during thymic selection in direct proportion to the strength of TCR activation by self-ligands and serves as a relatively stable marker of T-cell reactivity to self-antigens after T-cell exit from the thymus [16]. We have previously shown that insulin B:9-23 reactive T cells in single TCR mice maintain levels of CD5 set during thymic selection after activation in the draining lymph nodes and in the pancreatic tissue, further supporting CD5 as a useful stable surrogate of TCR self-reactivity [2]. When we overlaid CD5 expression onto the population distribution plot, we observed that CD5^low cells were preferentially located within naïve-like populations 1 and 8, while CD5^high T cells populated highly activated effector populations 2 and 9 (Fig. 1I). To directly assess how effector function is correlated with the level of CD5 expression, we performed a standard gating analysis to compare effector molecule expression in CD5^high (top 35%) and CD5^low (bottom 35%) CD4⁺ and CD8⁺ T cells (Supporting Information Fig. S2A and B). In both T-cell compartments, effector molecule expression (IFN-γ, IL-21, and perforin) was associated with higher level of CD5 expression, suggesting that CD5^high T-cell compartment encompasses pathogenic effectors (Fig. 2A–H). However, further analysis revealed similar high-level expression of FasL on both CD5^high and CD5^low CD8⁺ T cells, suggesting that CD5^low T cells are not completely void of effector function (Fig. 2J). Regulatory cytokine IL-10 was increased in CD5^high CD4⁺ T cells (Fig. 2I), while TGF-β was expressed at equally low levels in both CD5^high and CD5^low T cells compared to high expression in Foxp3⁺ Tregs (Supporting Information Fig. S2C), excluding the possibility that CD4⁺CD5^low T cells are primarily of a regulatory phenotype. The effector functions correlated with the increase in markers of differentiated T cells including Tbet, CXCR3, and CXCR5 in CD5^high cells, while naïve- or central-memory-associated CCR7 was primarily expressed in CD5^low cells (Fig. 2K–R). Analysis of integrins CD103 and α4β7 associated with tissue infiltration showed an overall higher expression of CD103 on CD8⁺ T cells and partial preference for α4β7 expression on CD4⁺ T cells (Fig. 2S–V). Interestingly, CD4⁺CD5^high T cells maintained higher level of both integrins compared with CD5^low T cells, suggesting their competitive advantage in tissue accumulation and residence compared with CD5^low T cells that preferentially expressed LN homing marker CCR7. Interestingly, CD8⁺CD5^low T cells expressed higher levels of CD103 compared with CD5^high T cells (Fig. 2T). CD103 has been used as a marker of tissue-resident memory T cells and could be an indicator of resident cells within the CD8⁺CD5^low compartment.

Since CD5 is correlated with self-reactivity, we postulated that the CD5^low population did not accumulate sufficient TCR signal in response to self-antigens in the pancreas to elicit a functional response. We have previously observed a direct correlation between CD5 expression and TCR signaling at the pancreatic tissue site, as measured by Nur77^GFP reporter [20]. Using insulin tetramer⁺ CD4⁺ T cells as a positive control for tissue specificity, we compared the proportion of activated T cells in islet infiltrating CD5^high, CD5^low, and nondraining-LN-derived CD4⁺Foxp3⁻ T cells. As expected, islet infiltrating tetramer⁺ T cells exhibited high levels of Nur77^GFP, CD69, and Ki67 compared with cells in nondraining LN (Fig. 3A and B). While CD5^high T-cell population contained T cells with high levels of Nur77^GFP and CD69 expression, CD5^low cells exhibited a low level of TCR signaling on par with nondraining LN T cells. However, Ki67 was increased in CD5^high and to some level in CD5^low cells, although the difference did not reach significance for CD5^low cells. These observations suggest that even though TCR signaling was not detectable in CD5^low cells using Nur77^GFP reporter, there remained a possibility that CD5^low T cells were in their early stages of activation or partially activated due to lower TCR affinity. To assess whether equivalent high level of TCR stimulation would reveal reserved functional capacity of CD5^low cells, we stimulated islet infiltrating T cells with plate-bound anti-CD3. A comparable frequency of CD5^high and CD5^low T cells from infiltrated islets of NOD.Nur77^GFP mice upregulated GFP in response to TCR cross-linking (Fig. 3C and D). However, within the activated (GFP⁺) T cells, CD5^low cells again exhibited inferior effector potential as defined by their reduced ability to produce IFN-γ (Fig. 3E). In order to confirm that CD5^low population was indeed functional, we assessed TNF-α expression after TCR stimulation. TNF-α is expressed by naïve T cells and does not require full differentiation into Th1-type effectors, unlike IFN-γ [21]. Although CD5^high T cells exhibited significantly higher levels of TNF-α production, TNF-α was still expressed by a high frequency of CD5^low cells (38% vs. 32%, Fig. 3F), supporting functional potential and competency of CD5^low cells. Overall, these data suggest that CD5^low T cells possess reduced levels of effector function and might represent a naïve or bystander T-cell population in infiltrated pancreatic islets.

Level of CD5 expression distinguishes islet infiltrating populations with unique TCR repertoires

In order to assess whether CD5^high and CD5^low expressing T-cell populations were composed of distinct clonotypes, we compared their TCR repertoires. Since CD4⁺ T cells are more heterogeneous in their TCR repertoire and antigen specificity in autoimmune diabetes compared with CD8⁺ T cells, a large proportion of which respond to a single IGRP epitope in NOD islets [22, 23], we focused our analysis on CD4⁺ T cells to identify potential divergent clonotypes in CD5^high and CD5^low T-cell responses. We have previously shown that high expression of CD5 can identify functionality superior Foxp3⁺ islet infiltrating Tregs [20], which provided us with an internal control of two functionally distinct populations (CD5^high vs. CD5^low Tregs) where we expected limited TCR overlap. To that end, we evaluated TCR repertoires of CD5^high and CD5^low conventional (Foxp3^–) and regulatory (Foxp3⁺) CD4 T cells in pancreatic islets of prediabetic NOD.Foxp3^GFP mice. High-throughput TCR-beta sequencing revealed few shared TCR sequences between CD5^high and CD5^low CD4⁺Foxp3^GFP– conventional (Tconv) and CD4⁺Foxp3^GFP+ regulatory (Treg) cells (Tconv, 2 ± 0.5%, MH = 0.0778; Treg 3 ± 1.7%, MH = 0.1198, Fig. 4A and B, Supporting Information Fig. S3A and B). This observation confirmed that the level of CD5 expression is set in individual T-cell clones during development and is maintained in periphery [2, 16]. Moreover, the data supported previous observations of negligible conversion between effector and regulatory T-cell populations in autoimmune pancreatic infiltration [24]. Importantly, the minimal overlap between CD5^high and CD5^low populations indicated that CD5 level separates unique T-cell clones.

Since we have previously observed that CD5^high Tregs exhibited increased activation and proliferation in the pancreas of NOD mice [20], we expected increased clonal expansion within the CD5^high Treg population. As expected, there was a trend toward increased clonality and reduced diversity within the CD5^high Tregs, consistent with our hypothesis (Fig. 4C, Supporting Information Fig. S3C). Surprisingly, CD5^high and CD5^low Tconv populations showed overall similar levels of clonality, with some mice exhibiting increased clonal expansion within CD5^high population and others in CD5^low population (Supporting Information Fig. S3D). Moreover, several highly expanded clones within the CD5^low Tconv compartment shared CDR3 sequence motifs with previously characterized insulin-specific T-cell clones (Fig. 4D) [2]. Although, we did not find CDR3 motifs associated with specific beta-cell antigen specificities within the top Treg TCR sequences, one of the highest frequency clones was shared among CD5^low Treg populations of three separate mice, suggesting similar antigenic targets and public TCR usage (Supporting Information Fig. S3F). In general, Tconv cells displayed both decreased clonality and increased repertoire diversity compared with Tregs (Fig. 4C, Supporting Information Fig. S3E), suggesting that Tregs possess a more restricted TCR repertoire in the pancreatic tissue, and their expansion at the tissue site is less clonally focused than Tconvs. In the case of Tconv cells, these results demonstrate that clonal expansion can be observed within either high- or low-affinity tissue infiltrating effector populations.

Transcriptional characteristic of CD5^low islet infiltrating T cells

In order to assess whether functional differences observed in restimulated CD5^high and CD5^low CD4⁺ islet infiltrating T cells could be observed directly ex vivo, we analyzed the transcriptional profile of CD5^high and CD5^low CD4⁺Foxp3^– Tconv cells sorted from the islets of NOD female mice. A total of 55 genes were significantly different (q < 0.05) between the two populations (Fig. 5A and B). Genes involved in proliferation and activation of T cells were expressed at a higher level in CD5^high subset (Ki67, Csf1, Tigit, Tnfrsf4 (OX40), Ctla4, Tnfrsf18 (GITR)), which showed that TCR reactivity for tissue antigen correlated with an overall higher level of activation. In order to confirm that TCR affinity for self-antigens was the determinant factor in driving transcriptional differences between CD5^high and CD5^low cells, we compared the genes with a previously obtained data set from T cells expressing TCRs with defined affinities. The dataset was derived from a two-TCR chimera from which we obtained transcriptional profiles of islet infiltrating high-affinity T cells expressing insulin B:9-23-specific TCR 4–8 and low-affinity InsB:9-23-specific 12–4.4m1 T cells sorted from the same pancreatic islets [25]. Importantly, unlike the majority of insulin-reactive T cells that show direct correlation between CD5 and TCR affinity for insulin epitope, the low affinity 12–4.4m1 T cells developed with a high level of CD5 expression that did not correlate with TCR reactivity for insulin [2]. The elevated levels of CD5 in the low affinity 12–4.4m1 T cells were likely due to strong reactivity for an unrelated self-antigen during thymic selection. Using 12–4.4m1 TCR allowed us to exclude the possibility that the transcriptional differences were induced by different levels of CD5 expression, rather than TCR activation. We observed similar genes up- or downregulated in high- and low-affinity cells in the two-TCR system and between polyclonal populations sorted on CD5 high or low expression (Fig. 5B and C). Transcripts upregulated in both CD5^high and high-affinity TCR-expressing cells included Tigit, Lag3, Csf1 genes associated with T-cell activation. When we focused our analysis on transcripts that were significantly upregulated in high-affinity 4–8 T cells from the two-TCR comparison (q < 0.05, total 86), we observed that many of the genes (73 out of 86) were preferentially associated with CD5^high cells (Fig. 5D). Similarly, transcripts significantly increased in T cells expressing low-affinity insulin 12–4.4 mL TCR (q < 0.05, total 61) were primarily associated with CD5^low T cells (43 out of 61; Fig. 5E). These observations confirmed that TCR affinity for antigen is an important factor in driving distinct T-cell phenotype of CD5^high and CD5^low cells.

Intriguingly, CD5^low subset showed higher expression of genes associated with undifferentiated naïve- or central-memory T cells, while CD5^high population exhibited a signature of terminal differentiation or exhaustion. CD5^low T cells expressed higher levels of Satb1, a transcriptional regulator involved in repression of PD-1 locus and maintaining T-cell responsiveness to antigen in autoimmunity [26, 27]. Consistent with this observation, CD5^low cells had reduced levels of PD-1 expression (Supporting Information Fig. S4). In addition, CD5^high T cells exhibited increased expression of inhibitory receptors Lag3, Tigit, and the marker of anergy Folr4, characteristic of terminally differentiated T cells (Fig. 5A and B). This transcriptional increase was confirmed at the protein level (Supporting Information Fig. S4). In contrast, CD5^low cells expressed a number of genes reminiscent of naïve- or central-memory T-cell phenotype, including higher levels of Ccr7, Sell (CD62L), and Il7r (Fig. 5B and E). This observation was also confirmed at the protein level (Fig. 2, Supporting Information Figs. S1 and S4). Overall, our data showed that CD4⁺ T cells infiltrating pancreatic islets of prediabetic NOD mice are heterogeneous in their level of differentiation, effector function, and TCR repertoire, but it was unclear how these differences affect their capacity to cause disease.

CD4⁺ CD5^low islet-infiltrating T cells induce diabetes upon transfer

Since CD5^high cells exhibited all the characteristics of activated effector T cells, while CD5^low cells had a nonactivated, potentially bystander, phenotype we wanted to compare intrinsic autoimmune potential of the two populations in a transfer model of diabetes. Given that CD5^low cells exhibited reduced TCR activation at the tissue site indicating their low reactivity for tissue antigen, we predicted that this cell population would have a reduced ability to home to the pancreas and cause beta-cell damage. We sorted CD5^high and CD5^low CD4⁺ Tconvs from infiltrated islets of prediabetic female NOD.Foxp3^GFP mice, and transferred cells into NOD.TCRα⁻/⁻ recipients. Mice that received CD5^high subset exhibited accelerated diabetes development starting at 5 weeks after transfer, compared with the group that received CD5^low T cells, in which diabetes onset was delayed by 3 weeks (Fig. 6A). Strikingly, both of the T-cell subsets induced diabetes in 100% of mice within 18 weeks post-transfer. Histological analysis confirmed that both CD5^high and CD5^low cells infiltrated the pancreatic islets and induced insulitis, which could be observed as early as 6 weeks post-transfer (Fig. 6B–D). Consistent with the delayed diabetes induction, insulitis was reduced in CD5^low recipient mice at this early time point. Nevertheless, about 40% of islets in CD5^low-recipient mice exhibited some level of insulitis, and about 18% of islets had lymphocytic penetration of more than 75% of the islet (Fig. 6D). These observations further confirm the ability of CD5^low cells to infiltrate the pancreatic islets and induce islet damage. Importantly, the differences in the levels of CD5 expression of the two sorted populations were maintained post-transfer and were consistent among the examined organs irrespective of antigen availability (Fig. 6E and F). These results show that although CD5^low population seemed “dormant”, it had a strong pathogenic potential.

We next assessed whether the delay in disease development in our transfer model was due to reduced infiltration, expansion, or effector function of CD5^low cells. There was no difference in the number of CD4⁺ T cells in the spleens or the islets of the two groups at the endpoint, suggesting that the two cell subsets survive, expand, and can reach similar levels of islet infiltration (Supporting Information Fig. S5A). CD69, PD-1, and Lag3 activation markers were equivalently upregulated on both CD5^high and CD5^low cells post-transfer (Supporting Information Fig. S5B–D). Previous studies have shown that CD5^high T cells were more equipped at differentiating into Foxp3⁺ cells [28]. Although the proportion of Foxp3⁺ regulatory T cells in CD5^high and CD5^low subsets was variable (0–12%), there was no significant difference between the two populations in either spleens or islets (Supporting Information Fig. S5E). Intriguingly, we observed that CD5^low cells acquired markers of anergy (FR4 and CD73) in the draining pancreatic draining LNs; however, the anergic phenotype was lower at the tissue site (Supporting Information Fig. S5F). Thus, CD5^high and a significant proportion of CD5^low Tconvs were able to avoid anergy and ultimately cause diabetes.

Next, we compared the functional profile of the two subsets post-transfer. After 5 weeks post-transfer, we detected comparable levels of IFN-γ (Fig. 6G), which was contrary to our observations prior to transfer, where CD5^high cells expressed higher levels of IFN-γ (Fig. 2E). In addition, levels of CD107a and perforin did not differ between these two subsets post-transfer (Fig. 6H and I), demonstrating similar level of activation, Th1 differentiation, and effector mechanisms utilized by the two populations. Collectively, our results demonstrate that CD4⁺ Tconvs with high- and low-reactivity have comparable capacity for Th1 differentiation and IFN-γ production, indicating similar effector mechanisms in the pancreas.

Mice

NOD/ShiLtJ (NOD), NOD/ShiLt-Tg(Foxp3-EGFP/cre)1cJbs/J (NOD.Foxp3^GFP), NOD.129P2(C)-Tcratm1Mjo/DoiJ (NOD.TCRα^–/–), and NOD.CB17-Prkdc^scid/J (NOD.scid) mice were purchased from Jackson Laboratories and maintained at our facility. NOD.Nur77^GFP mice were previously backcrossed to NOD and maintained in our mouse facility [2, 20]. Mice were housed in specific pathogen-free conditions and maintained on a standard diet. Experimental mice were monitored weekly for diabetes development by urine glucose testing (Diastix; Bayer); positive readings were confirmed by blood glucose testing (Breeze2 glucometer; Bayer). Mice were considered diabetic after two consecutive blood glucose readings >300 mg/dL or a single reading >400 mg/dL. The protocols were approved by the Baylor College of Medicine and University of Utah Institutional Animal Care and Use Committees.

Isolation of islet infiltrating cells

T cells were isolated from pancreatic islets of NOD, NOD.Foxp3^GFP, or NOD.Nur77^GFP mice as described previously [20]. Briefly, pancreata were perfused by injecting 600 u/mL collagenase 4 (Worthington) 5%FBS/HBSS solution into the bile duct, excised, and digested at 37°C for 30 min. After tissue disruption and washing, single islets were picked using microdissection scope with a p10 pipette. Islets were dissociated into single-cell suspension using enzyme-free cell dissociation buffer (Gibco). Islet cells were directly analyzed by flow cytometry or sorted for further analysis.

T-cell stimulation

Dissociated islets were stimulated for 5 h with plate-bound anti-CD3 (1 μg/mL) or 10 ng/mL phorbol myristic acid (Sigma) and 1 μmol/L ionomycin (Sigma) in the presence of brefeldin A and monensin.

Flow cytometry and antibodies

Flow cytometry analyses were performed on Cytek Aurora or BD LSRFortessa II, and data were analyzed with FlowJo software (Tree Star Inc.). Antibodies used in this study are as follows: anti-Foxp3 (FJK-16s, BioLegend or eBioscience), anti-Bcl6 (7D1), anti-CD3 (145-2C11, BioLegend or BD), anti-CD4 (GK1.5 or RM4-5), anti-CD8a (53-6.7), anti-CD8b (H35-17.2, eBioscience), anti-CD5 (53-7.3), anti-Ki67 (B56), anti-CD11a (M17/4, BD), anti-CTLA-4 (UC10-4B9), anti-PD1 (29F.1A12), anti-LAG3 (C9B7W, BD), anti-TIGIT (1G9, BD or GIGD7, eBioscience), anti-CD73 (TY/11.8), anti-FR4 (12A5), anti-CD44 (IM7), anti-CD62L (MEL-14), anti-IFNγ (XMG1.2, BioLegend or BD), anti-perforin (S16009A), anti-Tbet (4B10 BioLegend or eBiosciences), anti-CD69 (H1.2F3, BD), anti-CCR7 (4B12, BD), anti-KLRG1 (2F1, BD), anti-IL-10 (JESS-16E3), anti-Tim3 (RMT3-23), anti-IL-2 (JES6-5H4, BD), anti-IL-21 (mhalx21, eBiosciences), anti-CD103 (2E7), anti-FasL (MFL3, eBioscience), anti-CXCR3 (CXCR3-173), anti-CXCR5 (L138D7), anti-integrinα4β7 (DATK32), anti-LAP (pro-TGFβ, TW7-16B4), anti-TNF-α (MP6-XT22), and anti-CD107a (1D4B). Antibodies were obtained from BioLegend, unless indicated otherwise. Double staining with insulin B:10-23/I-Ag7 tetramers (HVEALYLVCGGEG) conjugated to PE and APC was used to identify antigen-specific T cells. Flow cytometry and sorting studies adhere to the “Guidelines for the use of flow cytometry and cell sorting in immunological studies” [48].

Sorting and transfer of islet infiltrating CD4⁺ T cells

Islet-derived CD4⁺ Foxp3^– conventional T cells (Tconvs) were sorted based on CD5 expression (CD4⁺Foxp3^GFP– CD5^high/low) from 11- to 20-week-old prediabetic NOD.Foxp3^GFP mice. Sorting was performed on a FACSAria Fusion flow cytometer (BD Biosciences). The sort purity was above 90% for CD5^high Tconvs, and 95% for CD5^low Tconvs. CD5^high and CD5^low CD4⁺ Tconvs were adoptively transferred (i.p.) into female NOD.TCRα^–/– recipient mice. Diabetes was induced with 20 000 cells per mouse. Insulitis and islet infiltration were assessed after the transfer of 10 000 cells.

Histological analysis of islet infiltration

Five-step sections from each mouse were stained with H&E. An average of 90–100 islets per mouse were scored in a blinded manner. Two methods of insulitis measurement were used. First, insulitis and peri-insulitis were determined based on the percentage of the islets that possessed leukocyte infiltrate using the following metric: no insulitis (score 0: normal islet and no infiltration), peri-insulitis (score 1: infiltration on edges of islet), and insulitis (score 2: up to 25% of islet, score 3: 25–75% of islet, score 4: >75% of islet). Second, a defined insulitis score was determined using the method outlined in Current Protocols in Immunology [49].

Immunofluorescence

Pancreatic sections were stained with anti-CD3 (CD3-12, Thermo Fisher), anti-BSRP-A (polyclonal sheep IgG, R&D), and DAPI. Images were obtained using Nikon Ti-E widefield microscope at 20× magnification. BSRP-A was used to identify endocrine cells [50, 51]. Nikon NIS-Elements scope/software was used to obtain the images. Background autofluorescence was removed using a previously published method in Fiji [52].

TCR sequence analysis

TCR sequence analysis was performed on individual mice. Genomic DNA was extracted from sorted Foxp3^GFP+ and Foxp3^GFP– CD4⁺TCR⁺ T cells isolated from pancreatic islets of prediabetic NOD.Foxp3^GFP mice at 14–16 weeks of age. Purification of gDNA was performed using QIAamp DNA Micro Kit. TCRβ repertoires of Tconv and Treg populations were assessed by ImmunoSEQ Survey assay (Adaptive Biotechnologies). Data were analyzed by ImmunoSEQ Analyzer version 2.0.

RNAseq analysis

RNA was isolated from CD4⁺ T cells sorted from single-cell suspensions of the pancreatic islets of NOD.Foxp3^GFP mice stained with anti-CD4 and anti-CD5 antibodies. All samples were sorted with >97% purity. cDNA was synthesized using the SMARTer Ultra Low Input RNA Kit (Clonetech). Library preparation was performed with the Illumina Nextera XT kit before paired-end RNA-sequencing using the Illumina NextSeq500 platform for 150 cycles (NextSeq500 Mid Output Kit). Sequencing reads were aligned to the mouse genome (RefSeq mm10) using TopHat Alignment (version 1.0.0; [53]) and gene expression was quantified by FPKM. Cufflinks Assembly & DE (version 1.1.0 [54]) was used to compute differential expression between groups, with differentially expressed genes defined by q < 0.05 with Benjamini–Hochberg correction for multiple testing. Gene ontology analysis was performed using Metascape. Two-TCR RNAseq data files were previously published in a study by Sprouse et al. [25]. Data resources: The accession numbers for the raw data reported in this paper are GSE102231, GSE106467, and GSE131345.

Statistical analysis

Statistical analyses were performed using Prism, except for TCR sequencing analysis, which was performed using ImmunoSEQ Analyzer.