Thymic self-recognition-mediated TCR signal strength modulates antigen- specific CD8⁺ T cell pathogenicity in non-obese diabetic mice

Reviewer #1 (Public review):

Summary

In their manuscript, Ho and colleagues investigate the importance of thymically-imprinted self-reactivity in determining CD8 T cell pathogenicity in non-obese diabetic (NOD) mice. The authors describe pre-existing functional biases associated with naive CD8 T cell self-reactivity based on CD5 levels, a well characterized proxy for T cell affinity to self-peptide. They find that naive CD5hi CD8 T cells are poised to respond to antigen challenge; these findings are largely consistent with previously published data on the C57Bl/6 background. The authors go on to suggest that naive CD5hi CD8 T cells are more diabetogenic as 1) the CD5hi naive CD8 T cell receptor repertoire has features associated with autoreactivity and contains a larger population of islet-specific T cells, and 2) the autoreactivity of "CD5hi" monoclonal islet-specific TCR transgenic T cells cannot be controlled by phosphatase over-expression. Thus, they implicate CD8 T cells with relatively higher levels of basal self-reactivity in autoimmunity. The data presented offers valuable insights and sets the foundation for future studies, but some conclusions are not yet fully supported.

Specific comments

There is value in presenting phenotypic differences between naive CD5lo and CD5hi CD8 T cells in the NOD background as most previous studies have used T cells harvested from C57Bl/6 mice or peripheral blood from healthy human donors.

The comparison of a marker of self-reactivity, CD5 in this case, on broad thymocyte populations (DN/DP/CD8SP) is cautioned. CD5 is upregulated with signals associated with b-selection and positive selection; CD5 levels will thus vary even among subsets within these broad developmental intermediates. This is a particularly important consideration when comparing CD5 across thymic intermediates in polyclonal versus TCR transgenic thymocytes due to the striking differences in thymic selection efficiency, resulting in different developmental population profiles. The higher levels of CD5 noted in the DN population of NOD8.3 mice, for example, is likely due to the shift towards more mature DN4 post-b-selection cells. Similarly, in the DP population, the larger population of post-positive selection cells in the NOD8.3 transgenic thymus may also skew CD5 levels significantly. Overall, the reported differences between NOD and NOD8.3 thymocyte subsets could be due largely to differences in differentiation/maturation stage rather than affinity for self-antigen during T cell development. The authors have added some additional text to the revised manuscript that acknowledges some of these limitations.

The lack of differences in CD5 levels of post-positive selection DP thymocytes, CD8 SP thymocytes, and CD8 T cells in the pancreas draining lymph nodes from NOD vs NOD8.3 mice also raises questions about the relevance of this model to address the question of basal self-reactivity and diabetogenicity and the authors' conclusion that "that intrinsic high CD5-associated self-reactivity in NOD8.3 T cells overrides the transgenic Pep-mediated protection observed in dLPC/NOD mice"; the phenotype of the polyclonal and NOD8.3 TCR transgenic CD8 T cells that were analyzed in the (spleen and) pancreas draining lymph nodes is not clear (i.e., are these gated on naive T cells?). Furthermore, the rationale for the comparison with NOD-BDC2.5 mice that carry an MHC II-restricted TCR is unclear.

In reference to the conclusion that transgenic Pep phosphatase does not inhibit the diabetogenic potential of "CD5hi" CD8 T cells, there is some concern that comparing diabetes development in mice receiving polyclonal versus TCR transgenic T cells specific for an islet antigen is not appropriate. The increased frequency and number of antigen specific T cells in the NOD8.3 mice may be responsible for some of the observed differences. Further justification for the comparison is suggested.

The manuscript presents an interesting observation that TCR sequences from CD5hi CD8 T cells may share certain characteristics with diabetogenic T cells found in patients (e.g., CDR3 length), and that autoantigen-specific T cells may be enriched within the CD5hi naive CD8 T cell population. However, the percentage of tetramer-positive cells among naive CD8 T cells appears unusually high in the data presented, and caution is warranted when comparing additional T cell receptor features of self-reactivity/auto-reactivity between CD4 and CD8 T cells.

The counts for the KEGG enrichment pathways presented are relatively low, and the robustness of the analysis should be carefully considered, particularly given that several significance values appear borderline. That said, the differentially expressed genes among CD5lo and CD5hi CD8 T cells are generally consistent with previously published datasets.

The manuscript includes some imprecise wording that may be misleading. For example (not exhaustive): The strength of TCR reactivity to foreign antigen is not "contributed by basal TCR signal" per se but rather correlates with sub-threshold TCR signals necessary for T cell development and survival, CD5 is not broadly expressed on all B cells as the text might suggest but is restricted to a specific subset of B cells, some of the proximal signaling molecules downstream of the preTCR are different than for the mature TCR, upregulation of CD127 at early timepoints post T cell activation is not directly suggestive of their "heightened capabilities in memory T cell homeostasis", etc. The statement "Our study exclusively examined female mice because the disease modeled is relevant in females" should be reconsidered. While the use of female NOD mice can be justified by their higher incidence of diabetes than their male counterparts, the current wording could be misleading.

For clarity and transparency, please consider while additional information is provided in the revised manuscript, gating strategies are not always clear (i.e., naive versus total CD8 T cells), and the age/status of the mice from which cells are harvested (i.e., prediabetic?) is not consistently provided as far as this reviewer noted.

Reviewer #2 (Public review):

Summary:

In this study Chia-Lo Ho et al. study the impact of CD5high CD8 T cells in the pathophysiology of type 1 diabetes (T1D) in NOD mice. The authors used high expression of CD5 as a surrogate of high TCR signaling and self-reactivity and compared the phenotype, transcriptome, TCR usage, function and pathogenic properties of CD5high vs. CD5low CD8 T cells extracted from the so-called naive T cell pool. The study shows that CD5high CD8 T cells resemble memory T cells poised for stronger response to TCR stimulation and that they exacerbate disease upon transfer in RAG-deficient NOD mice. The authors attempt to link these features to the thymic selection events of these CD5high CD8 T cells. Importantly, forced overexpression of the phosphatase PTPN22 in T cells attenuated TCR signaling and reduced pathogenicity of polyclonal CD8 T cells but not highly autoreactive 8.3-TCR CD8 T cells.

Strengths:

The study is nicely performed and the manuscript is clearly and well written. Interpretation of the data is careful and fair. The data are novel and likely important. However, some issues would need to be clarified through either text changes or addition of new data.

Weaknesses:

The definition of naïve T cells based solely on CD44low and CD62Lhigh staining may be oversimplistic. Indeed, even within this definition naïve CD5high CD8 T cells express much higher levels of CD44 than CD5low CD8 T cells.

Comments on revisions:

The authors addressed my previous comments thoughtfully and extensively.

Reviewer #3 (Public review):

Summary:

In this study, Ho et al. hypothesised that autoreactive T cells receiving enhanced TCR signals during positive selection in the thymus are primed for generating effector and memory T cells. They used CD5 as a marker for TCR signal strength during their selection at the double positive stage. Supporting their hypothesis, naïve T cells with high CD5 proliferated better and expressed markers of T cell activation compared to naïve T cells with lower levels of CD5. Furthermore, results showed that autoimmune diabetes can be efficiently induced after the transfer of naïve CD5 hi T cells compared to CD5 lo T cells. This provided solid evidence in support of their hypothesis that T cells receiving higher basal TCR signaling are primmed to develop into effector T cells. However, all functional characterisation was done on the cells in the periphery and CD5 hi cells in the peripheral lymphoid compartment can receive tonic TCR signaling. Hence, the function of CD5 hi T cells might not be related to development and programming in the thymus. This is a major hurdle in the interpretation of the results and justifying the title of the study. The evidence that transgenic PTPN22 expression could not regulate T cell activation in CD5 hi TCR transgenic autoreactive T cells was weak. Studying T cell development in TCR transgenic mice and looking at TCR downstream signaling could be misleading due to transgenic expression of TCR at all developmental stages.

Strengths:

(1) Demonstrating that CD5 hi cells in naïve CD8 T cell compartment express markers of T cell activation, proliferation and cytotoxicity at a higher level

(2) Using gene expression analysis, study showed CD5 hi cells among naïve CD8 T cells are transcriptionally poised to develop into effector or memory T cells.

(3) Study showed that CD5 hi cells have higher basal TCR signaling compared to CD5 lo T cells.

(4) Key evidence of pathogenicity of autoreactive CD5 hi T cells was provided by doing the adoptive transfer of CD5 hi and CD5 lo CD8 T cells into NOD Rag1-/- mice and comparing them.

Weaknesses:

(1) Although CD5 can be used as a marker for self-reactivity and T cell signal strength during thymic development, it can also be regulated in the periphery by tonic TCR signaling or when T cells are activated by its cognate antigen. Hence, TCR signals in the periphery could also prime the T cells towards effector/memory differentiation. That's why from the evidence presented here it cannot be concluded that this predisposition of T cells towards effector/memory differentiation is programmed due to higher reactivity towards self-MHC molecules in the thymus, as stated in the title.

(2) Flow cytometry data needs to be revisited for the gating strategy, biological controls and interpretation.

(3) Evidence linking CD5 hi cells to more effector phenotype using gene enrichment scores is very weak.

(4) Experiments done in this study did not address why CD5 hi T cells could be negatively regulated in NOD mice when PTPN22 is overexpressed resulting in protection from diabetes but the same cannot be achieved in NOD8.3 mice.

(5) Experimental evidence provided to show that PTPN22 overexpression does not regulate TCR signaling in NOD8.3 T cells is weak.

(6) TCR sequencing analysis does not conclusively show that CD5 hi population is linked with autoreactive T cells. Doing single-cell RNAseq and TCR seq analysis would have helped address this question.

(7) When analysing data from CD5 hi T cells from the pancreatic lymph node, it is difficult to discriminate if the phenotype is just because of T cells that would have just encountered the cognate antigen in the draining lymph node or if it is truly due to basal TCR signaling.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Review #1 (Public review):

Figures 1 through 4 contain data that largely recapitulate published findings (Fulton et al., 2015; Lee et al., 2024; Swee et al., 2016; Dong et al., 2021); it is noted that there is value in confirming phenotypic differences between naive CD5lo and CD5hi CD8 T cells in the NOD background. It is important to contextualize the data while being wary of making parallels with results obtained from CD5lo and CD5hi CD4 T cells. There should also be additional attention paid to the wording in the text describing the data (e.g., the authors assert that, in Figure 4C, the “CD5hi group exhibited higher percentages of CD8+ T cells producing TNF-α, IFN-γ and IL-2” though there is no difference in IL-2 nor consistent differences in TNF-α between the CD5lo and CD5hi population^hi CD8⁺ and CD5^loCD8⁺ T cells have been previously characterized in other genetic backgrounds. In our study, we aimed to confirm and extend these observations specifically in the autoimmune-prone NOD background, which had not been systematically addressed. Additionally, we carefully reviewed the text describing Figure 4C and revised the wording to accurately reflect the observed data (line 263-264). Specifically, we now state that the CD5^hi group exhibited higher levels of IFN-γ and a trend toward increased TNF-α, while IL-2 production did not show a significant difference.

The comparison of CD5 across thymocyte populations is cautioned due to variation in developmental stages, particularly in transgenic models. The reported differences may reflect maturation stages rather than self-reactivity.

We appreciate the reviewer’s important point regarding the interpretation of CD5 levels across thymocyte subsets. In our revised manuscript (lines 455–471), we have added clarification that CD5 expression in DN and DP subsets reflects pre-TCR and TCR signaling events during thymic development. We also acknowledge that differences in maturation stages, especially in the NOD8.3 transgenic model, may influence CD5 expression. We now discuss this caveat and interpret our results with caution, particularly emphasizing that our data support but do not sufficiently define their differential self-reactivity.

The conclusion that PTPN22 overexpression does not inhibit the diabetogenic potential of CD5^hiCD8⁺ T cells is potentially confounded by differences between polyclonal and TCR transgenic systems.

We thank the reviewer for raising this concern. We acknowledge that this system introduces confounders due to differences in precursor frequencies and clonal expansion compared to polyclonal repertoires. These differences may affect the responsiveness to phosphatase-mediated attenuation of signaling. Therefore, while our results support that high-affinity autoreactive CD8⁺ T cells may be less sensitive to PTPN22 overexpression, we do not claim that this finding generalizes to all autoreactive CD8⁺ T cells. Rather, it highlights a potential inability of peripheral tolerance in T cells with strong intrinsic self-reactivity.

TCR sequencing data shows variability; is this representative of the overall repertoire?

We appreciate the reviewer’s comment. We acknowledge that data from bulk TCR sequencing has potential limitations, including variability across experiments and limited resolution at the clonotype level. To improve representativeness and reduce sampling bias, we performed TCR repertoire analysis in two independent experiments. In each experiment, naïve CD5^hi CD8⁺ and CD5^loCD8⁺ T cells were sorted from pooled peripheral lymph nodes of at least 20 individual NOD mice per group. This approach allowed us to capture a broader range of clonotypes and ensured that the resulting repertoire profiles reflect the characteristics of the overall CD5^hi and CD5^lo populations, rather than isolated outliers. Despite some variability, we observed consistent trends in key features, such as shorter CDR3β length, altered TRAV/TRBV usage and reduced diversity in the CD5^hi subset across both experiments. To enhance resolution and directly assess clonotype-specific reactivity, we plan to perform single-cell RNA and TCR sequencing in future studies, as noted in the revised Discussion (lines 466–471).

Clarifications are requested regarding naive gating, controls, gMFI reporting, and missing methods.

We thank the reviewer for these specific suggestions. We have revised figure legends to better describe gating strategies and included appropriate controls in Figures or Supplementary Figures. Regarding gMFI reporting, we have now shown in the figure legends whether values are reported as gMFI. Additionally, we have added the missing methods for cytokine staining, EdU incorporation, overlapped count matrix construction and TCR repertoire diversity metrics.

Review #2 (Public review):

Summary Comment:

The study is nicely performed, but the definition of naive T cells using only CD44 and CD62L may be oversimplified. CD5hi naive T cells express higher CD44 than CD5lo cells.

We thank the reviewer for the critical evaluation and thoughtful comment. As noted, we defined naïve CD8⁺ T cells using a well-established gating strategy based on CD44^lo and CD62L^hi expression, consistent with previous studies (Immunity. 2010; 32(2):214–26; Nat Immunol. 2015; 16(1):107–17). We acknowledge that CD44 is expressed along a continuum, and indeed, within the naïve gate, CD5^hi CD8⁺ T cells exhibited slightly higher CD44 levels compared to their CD5^lo counterparts. However, both subsets remained well below the CD44 expression observed in conventional effector/memory CD8⁺ T cells, supporting their classification as naïve. To further validate this, we assessed additional markers associated with activation and memory differentiation, including CD69, PD-1, KLRG1 and CD25. These analyses confirmed that the sorted CD5^hi and CD5^lo populations retained a phenotypically naïve profile while exhibiting meaningful differences in baseline activation readiness (Figure 1F).

Review #3 (Public review):

CD5 can be regulated by peripheral signals. Therefore, it cannot be concluded that predisposition to effector/memory differentiation is solely programmed in the thymus.

We thank the reviewer for this important point. We agree that CD5 expression can be dynamically regulated in the periphery by tonic TCR signals and antigen encounter, as also reflected in our own data that cells with high CD5 level display elevated activation potential upon encountering antigen (e.g., Figure 3L). To minimize the confounding effects of pre-existing peripheral activation, we performed an adoptive T cell transfer experiment (Figure 4). In this experiment, naïve CD5^hiCD⁺and CD5^loCD8⁺T cells were sorted from the peripheral lymph nodes of young (6–8-week-old) prediabetic NOD mice and transferred into NOD Rag1^–/– recipients. After 4 weeks, we compared the disease phenotypes and functional profiles of CD8⁺ T cells from these two groups. This approach allowed us to evaluate the stability and differentiation capacity of CD5^hi versus CD5^lo cells in a lymphopenic environment, while excluding the possibility that the observed differences were due to already activated CD8⁺T cells at the time of isolation. We have revised the Discussion (lines 440–450) to acknowledge these experimental limitations and clarify that, while our findings demonstrate functional differences between CD5^hiCD8⁺ and CD5^loCD8⁺T cells, we cannot fully exclude contributions from peripheral influences.

Experiments do not explain why PTPN22 overexpression protects in polyclonal T cells but not in NOD8.3 mice.

We appreciate this critical comment. Our findings support that autoreactive T cells with high-affinity TCRs as in NOD8.3 mice receive strong signaling that even PTPN22 overexpression is insufficient to attenuate their activation and effector function. We acknowledge that further mechanistic studies are needed to fully elucidate the differential effects of PTPN22 in polyclonal versus TCR-transgenic settings.

Evidence that PTPN22 does not regulate TCR signaling in NOD8.3 T cells is weak.

We thank the reviewer for this critical comment. Our data show that NOD8.3 T cells with an intrinsic high CD5-associated self-reactivity are more resistant to transgenic Pep-mediated change in the phosphorylation status of TCR signaling molecules CD3ζ and Erk and CD5 expression (Figure 6, B-D). However, we agree that additional functional assays would strengthen this conclusion.

TCR sequencing does not conclusively link CD5hi cells with autoreactivity; single-cell analysis is needed.

We agree with this critical comment. Bulk TCR sequencing revealed repertoire features associated with autoreactivity, but cannot definitively link specific TCRs to function. We have acknowledged this in the discussion (lines 466–471) and highlighted plans to perform single-cell analysis.

CD5hi cells in the PLNs may reflect antigen exposure rather than basal signaling.

We thank the reviewer for this insightful comment. As also noted in Figure 3L, CD5 expression can be influenced by peripheral tonic TCR signals and recent antigen exposure. To minimize the contribution of peripheral activation, we particularly characterized naïve CD8⁺T cells isolated from the peripheral lymph nodes of young (6–8-week-old) prediabetic NOD mice before the onset of overt autoimmunity. Furthermore, we performed an adoptive transfer experiment (Figure 4) using sorted naïve CD5^hiCD8⁺ and CD5^loCD8⁺T cells from these mice and characterized their disease phenotype after 4 weeks in lymphopenic NOD Rag1^–/– recipients and evaluated the effector function of CD8⁺T cells. This approach allowed us to compare the differentiation potential of these subsets in a controlled setting, independent of their activation status at the time of isolation. We have revised the Discussion (lines 440–450) to emphasize that, while our data support functional differences between CD5^hiCD8⁺ and CD5^loCD8⁺T cells, we cannot fully exclude the role of peripheral cues in shaping CD5 expression.

Provide proper gating controls and representative flow plots.

We thank the reviewer for this comment. We have revised figure legends to better describe gating strategies and included representative flow cytometry plots and appropriate gating controls in Figures or Supplementary Figures.

Recommendations for the authors:

Reviewer #1 (Recommendations For The authors):

(1) The figure presentation is inconsistent and the labels/font are often too small to read easily.

As Reviewer suggested, the figure presentation has been revised for consistency. Labels and fonts have been adjusted for improved readability. Specific figures that were difficult to read have been reformatted with larger fonts and clearer legends.

(2) A careful review of the text to ensure clarity of the content is suggested (e.g., “gratitude” at line 91, “were generally lied” at line 123).

Thanks for Reviewer’s comments. The text has been carefully reviewed for clarity and grammatical accuracy. Corrections have been made, including changing “gratitude” to “magnitude” (line 47) and “were generally lied” to “fell between” (line 79).

Reviewer #2 (Recommendations For The Authors):

(1) The definition of naïve T cells based solely on CD44low and CD62Lhigh staining may be oversimplistic. Indeed, even within this definition, naïve CD5high CD8 T cells express much higher levels of CD44 than CD5low CD8 T cells.

Thanks for Reviewer’s comments. We used a literature-supported gating strategy (Immunity. 2010; 32(2):214–26; Nat Immunol. 2015; 16(1):107–17) to define naïve T cells based on CD44^low and CD62L^high expression. It is important to note that CD44 expression exists along a continuum. While we were initially surprised to observe that CD5^loCD8⁺T cells expressed relatively higher levels of CD44 than CD5^loCD8⁺T cells within the naïve gate, both populations still exhibited significantly lower CD44 expression compared to conventional effector/memory CD8⁺T cells. To further validate the distinction between CD5^hi and CD5 subsets, we also examined additional markers such as CD69, PD1, KLRG1 and CD25, which supported their phenotypic differences within the naïve compartment (Figure 1F).

(2) Figure 1G should show the proportion of IGRP-tetramer+ in the three groups of CD8 T cells. Additionally, it would be useful to assess reactivity against a pool of other islet autoantigens using a similar strategy.

As suggested by the reviewer, the revised manuscript now includes additional data showing the proportion of IGRP-tetramer+ cells (Supplementary Figure 1D), as well as reactivity against another islet autoantigen, insulin-1/insulin-2 (Insulin B15–23) (Supplementary Figure 1E). The description of these results, including the proportions of IGRP-tetramer⁺ and Insulin B15–23⁺ CD8⁺Tcells, has been added to lines 126–129 of the revised manuscript.

(3) The resolution of Figure 2 is suboptimal and at places poorly visible. Figure 2D is stated to show “two significant pathways stand out.” In fact, the data are barely significant, and the authors may want to correct their statement.

The resolution of Figure 2 has been improved. As Reviewer suggested, the text has been revised to state “two potential pathways stand out” (line 187) instead of “two significant pathways stand out”.

(4) Figure 3C-F and 3H, showing fold change over baseline values would be much easier for the reader to grasp the data.

As Reviewer suggested, data in Figures 3C-F and 3H now are shown in fold change over baseline values for clarity. Baseline gMFI is the mean of each group (total CD⁺ , CD5^hiCD8⁺ and CD5^loCD8⁺) at 0 μg/ml anti-CD3, with fold changes calculated for stimulation conditions (0.625-10 μg/ml anti-CD3). The figure legend has been updated accordingly.

(5) Figure 4A, it would be much more valuable to show the diabetes frequency upon transfer of CD25- CD4 T cells alone and upon transfer of CD5high CD8 T cells alone. The word “spontaneous” in the Figure 4A legend seems inappropriate.

Thanks for the Reviewer’s comment. We apologize for not including the data for the CD25 CD4⁺ T cell transfer group in the original manuscript. While this group was part of our initial experimental design, we had considered it a control group and unintentionally omitted it from the figure. The revised manuscript now includes this group in Figure 4A. In addition, the term “spontaneous” has been replaced with “diabetes incidence” in the Figure 4A legend and manuscript (line 248). Regarding the suggestion to assess CD5^hiCD8⁺T cells transfer alone, we appreciate the Reviewer’s point. However, previous studies have shown that CD8⁺ T cells alone are not effective and sufficient to induce diabetes in adoptive transfer models, and that effective β-cell destruction typically requires both CD4⁺ and CD8⁺ T cell subsets. For instance, Christianson et al. (1993) demonstrated that enriched CD8⁺ T cells from NOD mice fail to transfer diabetes on their own, while CD4⁺ T cells—particularly from diabetic donors—can induce disease only under specific conditions and are significantly potentiated by co-transfer of CD8⁺cells. These findings have contributed to the widely available standard of co-transferring both subsets when studying diabetogenic potential in NOD models (Diabetes. 1993;42(1):44–55).

(6) Line 257-258, please remove “indicating superior in vivo proliferation by the CD5hi subset.” Indeed, several other possibilities may explain the phenotype, including survival, migration, etc.

As Reviewer suggested, the phrase “indicating superior in vivo proliferation by the CD5^hi subset” has been replaced with “implying increased expansion and activation/effector potential” (line 261).

(7) Figure 5A, it is unclear to this referee what is the significance of CD5 and pCD3zeta expression on DN thymocytes. Do these cells express rearranged alpha/beta TCR? Is it signaling through pre-TCRalpha/TCRbeta pairs?

Thanks a lot for this important question. In the revised manuscript, we have expanded the discussion (line 455–471) to address the developmental significance of CD5 and pCD3ζ expression on DN thymocytes. CD5 expression at this stage reflects pre-TCR signaling strength during early selection, which occurs following successful TCRβ rearrangement. The associated phosphorylation of CD3ζ indicates activation of downstream signaling through the pre-TCRα/TCRβ complex. As discussed in the revised text, these early signals play a critical role in determining lineage progression and self-reactivity tuning. We now acknowledge that signaling at the DN stage occurs through the pre-TCRα/TCRβ heterodimer, not a fully rearranged αβ TCR, and that CD5 expression serves as a marker of the strength of these initial pre-selection signals (Sci Signal. 2022;15(736):eabj9842.). These developmental checkpoints are essential for calibrating TCR sensitivity and ensuring proper thymocyte maturation. This has been clarified in the revised discussion (line 455–471).

(8) Figure 5F, could the DP TCRbeta- CD69- thymocytes from 8.3-TCR NOD mice already express low levels of the self-reactive TCR at this stage to explain their high expression of CD5? Addressing the question experimentally would be useful.

Thanks a lot for this useful comment. According to a review by Huseby et al. (2022), expression of a functional TCRβ chain begins at the DN3 stage, initiating progression through the β-selection checkpoint. This is followed by TRAV locus recombination, resulting in the generation of αβ TCR-expressing double-positive 1 (DP-1) thymocytes. At the DP-1 stage, the quality of TCR signaling driven by self-pMHC interactions governs both positive and negative selection, as well as the development of nonconventional T cell lineages. We hypothesize that in transgenic NOD8.3 mice, which express pre-rearranged Tcra and Tcrb transgenes derived from the islet-reactive CD8⁺T cell clone NY8.3, thymocytes undergo allelic exclusion and lack the clonal diversity seen in non-transgenic mice. As a result, NOD8.3 thymocytes may receive strong TCR signals from early developmental stages (DN3 and DP-1) even without undergoing normal selection checkpoints. While the elevated TCR signal observed in NOD8.3 is indeed artificial, this model provides a unique system to test our hypothesis—namely, whether a strongly self-reactive TCR can generate high basal signaling during thymic development that overrides the negative regulatory effects of phosphatases like Pep. This possibility has been acknowledged in the revised Discussion section, along with a plan to validate the hypothesis experimentally (line 455–471).

(9) Figure 7, single-cell TCR-seq would be much more appropriate to tackle the question of self-reactivity of CD5hi vs. CD5low CD8 T cells.

Thanks a lot for this useful comment. The limitations of bulk TCR-seq are acknowledged, and single-cell TCR-seq is proposed as a future direction (line 455–471).

Note, for Reviewer #2 (Recommendations For The Authors) (7) (8) (9), the discussion paragraphs are included to address the reviewers’ questions (line 455–471).

Reviewer #3 (Recommendations For The Authors):

(1) Positive controls (activated T cells from PLN or spleen), gating controls (whole naïve T cells), and representative flow-cytometry plots are needed for T-bet, EOMES, GzmB, and cytokine staining in Figure 1.

As Reviewer suggested, we added representative gating controls for T-bet, EOMES, GzmB and cytokine staining in Supplementary Figure 1 of revised manuscript.

(2) For Figure 1F, MFI for activation markers for the CD44hiCD62Llo cells should be provided for the comparison of PLN data.

As Reviewer suggested, MFI data for these markers have been included in Figure 1F of revised manuscript.

(3) In many places and figure legends, it is not mentioned from which organ cells were collected, i.e., spleen or PLN.

As Reviewer suggested, the origin of cells for each experiment has been explicitly indicated in the figure legends or figure content to ensure clarity.

(4) In the pancreatic lymph node, autoreactive T cells might be upregulating CD5 because they are encountering antigens. This should be addressed in the discussion.

As Reviewer suggested, this issue has been included in the discussion of revised manuscript (line 440-450).

(5) It is not clear if T cells from the spleen and PLN were stimulated to detect the production of pro-inflammatory cytokines.

Thanks for the critical comment. The stimulation protocol and cytokine staining method have been added to the Supplementary material’s Supplementary methods section Cytokine staining in revised manuscript.

(6) Figure 4C-D: It is not clear if analysis was done on naïve T cells or if they were stimulated.

Thanks for the comment. Additionally, the stimulation and cytokine staining methods used in Figure 4C-D have been described in detail in the Supplementary Materials section Cytokine staining of revised manuscript.

(7) IGRP gating in Figure 4F should be revisited with negative controls.

Thanks for the critical comment. Negative controls have been added and used to adjust IGRP gating, and this is now mentioned in the figure legend of revised manuscript.

(8) Interpretation that only CD5hi cells form a central memory T cell population (Figure 4F) could be misleading.

Thanks for this valuable comment. We agree with that in conventional CD8⁺ T cell immune responses, both CD5^hi and CD5^lo subsets have the potential to differentiate into central memory T cells. In our experimental approach, we adoptively transferred sorted CD5^hiCD8⁺ or CD5^loCD8⁺cells into Rag1^-/- recipients and specifically analyzed PLNs four weeks after transfer. Using CD44 and CD62L expression as conventional markers for central memory T cells, we barely observed a CD44^hiCD62L^hi population in CD5^loCD8⁺transferred group. Based on these results, we stated: “This analysis underscores that the central memory T cell population and the frequency of islet autoantigen-specific CD8⁺T cells are higher in the CD5^hi transferred subset within the PLNs, implying more robust immune responses initiated by the CD5^hicells” (line 272–274). Importantly, we did not intend to imply that only CD5^hi cells can form central memory T cells, but rather that they were more enriched for this phenotype under the specific conditions and time point analyzed.

(9) IL-2 gating representative plot should be provided for Figure 5A.

As Reviewer suggested, a representative IL-2 gating plot has been included in the revised Supplementary Figure 3B.