Non-deletional CD8+ T cell self-tolerance permits responsiveness but limits tissue damage

Self-specific CD8+ T cells often escape clonal deletion, but the properties and capabilities of such cells in a physiological setting are unclear. We characterized polyclonal CD8+ T cells specific for the melanocyte antigen tyrosinase-related protein 2 (Trp2) in mice that express or lack this enzyme due to deficiency in Dct, which encodes Trp2. The size, phenotype, and gene expression profile of the pre-immune Trp2/Kb-specific pool were similar in wild-type (WT) and Dct-deficient (Dct-/-) mice. Despite comparable initial responses to Trp2 immunization, WT Trp2/Kb-specific cells showed blunted expansion, and scRNAseq revealed WT cells less readily differentiated into a CD25+ proliferative population. Functional self-tolerance clearly emerged when assessing immunopathology: adoptively transferred WT Trp2/Kb-specific cells mediated vitiligo much less efficiently. Hence, CD8+ T cell self-specificity is poorly predicted by precursor frequency, phenotype or even initial responsiveness, while deficient activation-induced CD25 expression and other gene expression characteristics may help to identify functionally tolerant cells.

Accurate discrimination between harmful (pathogens, toxins, cancerous cells) and non-42 harmful entities (self, innocuous environmental components, non-pathogenic microbes) 43 underlies effective functioning of the immune system. Understanding the mechanisms 44 that normally enforce immunological tolerance to self is a prerequisite for safely and 45 effectively manipulating the immune system to therapeutically induce or break self-46 tolerance.

48
Tolerance can be mediated by the clonal deletion of developing self-reactive T cells 49 (Hogquist et al., 2005; Kappler et al., 1987). Largely based on studies in transgenic 50 mouse models, this process has long been regarded as common and highly efficient 51 (Palmer, 2003). However, recent studies have revealed that thymic clonal deletion is 52 less effective than previously thought (Richards et al., 2016). Self-reactive CD8+ T cells 53 have been shown to escape negative selection in mice (Bouneaud et al., 2000;Zehn & 54 Bevan, 2006), with one group proposing that up to 4% of peripheral CD8+ T cells are 55 self-specific (Richards et al., 2015). Furthermore, studies in humans indicated that 56 precursor frequencies of blood CD8+ T cells specific for certain self-peptides were 57 comparable to those demonstrated for foreign peptides (Yu et al., 2015) and suggested 58 that such cells might be capable of overt autoreactivity if suitably stimulated (Maeda et 59 al., 2014). . It is also unclear how the presence and reactivity of self-specific CD8 + T cells 69 relates to their ability to drive immunopathology. The majority of commonly-used mouse 70 models of tolerance have the drawbacks of relying on T cell receptor (TCR) transgenic 71 animals that may not recapitulate normal physiology or utilizing in vitro analyses for 72 characterization of functionality.

74
Our studies are intended to provide a better understanding of non-deletional CD8+ T 75 cell tolerance by utilizing a more physiologic and translationally-relevant mouse model 76 in which an epitope from the melanocyte differentiation enzyme tyrosinase-related 77 protein 2 (Trp2) is recognized by CD8+ T cells as either self or foreign. Tyrosinase-78 related protein 2, an enzyme involved in melanin biosynthesis encoded by the 79 dopachrome tautomerase (Dct) gene, is normally expressed by melanocytes in the skin 80 in both humans and C57BL/6 mice and is overexpressed by many melanomas 81 (Avogadri et al., 2016;Wang et al., 1996). Using wild-type (WT) mice and a novel Dct-82 deficient (Dct -/-) strain, we compared responses to Trp2180-188/K b (Trp2/K b ) as a self-83 versus foreign antigen. This model is relevant to human health, as Trp2 is a common 84 129 130 In some systems, T cells bearing TCRs with low affinity for self-antigens avoid deletion; 131 low affinity TCRs can often be identified by reduced peptide/MHC tetramer binding to 132 these cells (Bouneaud et  and Dct -/mice. The average Trp2/K b tetramer staining was higher on Dct -/cells, but the 136 MFI largely overlapped between the two populations ( Figure 1B), suggesting that the 137 range of TCR avidities did not markedly differ between the Trp2/K b -specific pools. 138 Indeed, the tetramer staining differences we observed (a WT:Dct -/tetramer MFI ratio of 139 ~0.75) are more subtle than that noted in a previous study using a transgenic mouse 140 model, which reported a tetramer ratio of ~0.35 between mice with vs. without self-141 antigen expression (Bouneaud et al., 2000). As a control, we also assessed the avidity 142 of cells specific for an irrelevant foreign epitope-B8R/K b from vaccinia virus-in WT 143 and Dct -/mice; the tetramer MFI of B8R/K b -specific cells was comparable between the 144 strains ( Figure 1C). 145 146 We also examined the phenotype of Trp2/K b -specific cells in pre-immune WT and Dct -/-147 mice. No consistent differences in the expression of activation/memory markers (CD69, 148 CD44, CD122) or anergy/exhaustion markers (PD-1, LAG3, CTLA-4, TIM3) were 149 identified between Dct -/and WT Trp2/K b -specific cells (Figures 1C-F, S1B, S1C). The 150 majority of cells exhibited low expression of the memory markers CD44 and CD122, 151 and anergy/exhaustion marker expression was low in both populations. CD5 can 152 indicate self-antigen recognition (Azzam et al., 1998;Fulton et al., 2015), but we did not 153 detect significant differences in expression between the groups ( Figure 1G). RNAseq 154 analysis of Trp2/K b tetramer-binding cells isolated from pre-immune mice by 155 fluorescence-activated cell sorting (FACS) showed no consistent differences in gene 156 expression related to their derivation from WT versus Dct -/mice (Supplemental figure  157 1D), although this does not rule out the possibility of epigenetic differences between the 158 populations. 159 160 To ensure our findings were not unique to Trp2180/K b -specific cells, we used tetramer 161 enrichment to isolate CD8+ T cells specific for other skin antigens-a distinct Trp2 162 epitope (Trp2363/D b ) and a tyrosinase-related protein 1 epitope (Trp1455/D b )-in mice 163 expressing or lacking these antigens. We were able to identify cells with these 164 specificities present at numbers similar to slightly less in mice expressing antigen 165 relative to mutant mice (Supplemental figure 1E). This suggests that CD8+ T cells 166 specific for other melanocyte self-epitopes also largely escape clonal deletion.

168
Hence, although we identified some minor differences between Trp2/K b -specific cells 169 from WT versus Dct -/mice, these pre-immune populations generally resembled each 170 other in number, phenotype, and gene expression, arguing against a major role for 171 clonal deletion or overt steady-state anergy induction as tolerance mechanisms to this 172 antigen. These findings resonate with studies in humans, which have shown that the 173 precursor frequency and peptide/MHC tetramer staining intensity were only modestly 174 reduced for a self-versus non-self antigen (Yu et al., 2015) and that self-specific cells 175 can be phenotypically naïve (Maeda et al., 2014;Yu et al., 2015). Accordingly, these 176 data suggested that analysis of the Trp2/K b -specific responses in mice could serve as a 177 useful model to investigate the basis for and limits of non-deletional self-tolerance. 178 179 WT Trp2/K b -specific CD8+ T cells display impaired responses to Trp2 immunization 180 181 It was possible that the lack of substantial clonal deletion or signs of prior activation in 182 WT Trp2/K b -specific cells indicated "ignorance" of Trp2 and/or that the Trp2180 epitope, 183 which shows suboptimal binding to K b (McWilliams et al., 2006), was unable to prime a 184 vigorous immune response. To investigate this, we challenged WT and Dct -/mice with 185 Trp2 in an immunogenic context using TriVax, a subunit immunization strategy 186 comprising peptide, agonist anti-CD40 antibody, and poly(I:C) (Cho & Celis, 2009). It 187 should be noted that the TriVax approach uses the minimal peptide for priming, which 188 likely excludes antigen-specific Treg involvement. We included B8R peptide in addition 189 to Trp2 peptide in these experiments as an internal control. While WT and Dct -/mice 190 responded similarly to B8R, WT mice showed a more limited response to Trp2 at an 191 effector time point (day 7) relative to Dct -/mice ( Figure 2A, B), ruling out ignorance as 192 the dominant tolerance mechanism. We observed a significantly larger number and 193 frequency of Trp2/K b -specific cells in Dct -/mice, and the Dct -/cells exhibited higher 194 apparent Trp2/K b avidity (as measured by tetramer MFI; Figure 2C-E). Still, WT 195 Trp2/K b -specific cells expanded > 1000-fold (Dct -/cells expanded ~4000-fold). The 196 WT:Dct -/tetramer ratio was little changed relative to the pre-immune populations, 197 suggesting the difference in avidity between WT and Dct -/cells had not been amplified 198 by activation. The frequency of PD-1+ cells was comparable between WT and Dct -/-199 Trp2/K b -specific populations at this time point ( Figure 2F), suggesting similar exposure 200 to antigen.

202
To ensure that our results were not specific to the TriVax system, we also infected mice 203 with a recombinant Listeria monocytogenes strain expressing Trp2 (LmTrp2) (Bruhn et 204 al., 2005) and sacrificed the mice at effector (day 7) and memory (day 45) time points, 205 assessing the percentage and number of Trp2/K b -specific CD8+ T cells and cytokine 206 production in response to ex vivo Trp2 stimulation. Again, the Trp2/K b -specific response 207 was greater in Dct -/mice at both effector and memory time points (Supplemental figures  208 2A-C). Similarly, the frequency of all CD8+ T cells responding to ex vivo Trp2 209 stimulation with cytokine production (IFN-γ, TNF-α) was larger in Dct -/mice; the percent 210 producing IFN-γ approximated the tetramer-positive population, suggesting that the 211 majority of Trp2/K b -specific cells were able to produce this cytokine in both strains of 212 mice (Supplemental figure 2D). Among IFN-γ-producing cells, those from Dct -/mice 213 tended to produce increased amounts of cytokine on a per-cell basis (as assessed by 214 IFN-γ MFI). 215 216

218
With both the TriVax and LmTrp2 approaches, polyclonal Trp2/K b -specific cells from WT 219 mice showed evidence of tolerance, i.e., submaximal responsiveness to Trp2, while 220 those from Dct -/mice mounted a stronger response consistent with a typical response 221 to a foreign antigen. Because cells from Dct -/mice were able to respond robustly in 222 these experiments, the poor responsiveness of WT Trp2/K b -specific cells could not be 223 attributed solely to ignorance or ineffective immunization. 224 225 226 Tolerance to Trp2/K b is CD8 + T cell-intrinsic 227 Both cell-intrinsic and cell-extrinsic mechanisms of CD8+ T cell tolerance have been 228 previously described. Sakaguchi's group (Maeda et al., 2014) identified anergic CD8+ T 229 cells specific for melanocyte antigens in healthy human donors and concluded that 230  found human self-antigen-specific T cells to be poorly responsive to antigenic 233 stimulation even in the absence of Tregs, and Greenberg and colleagues (Schietinger et 234 al., 2012) showed that tolerant self-reactive murine CD8+ T cells remained tolerant 235 when transferred into new hosts that lacked antigen expression.

237
Accordingly, we investigated whether cell-intrinsic or -extrinsic mechanisms were active 238 in restraining Trp2/K b -specific CD8+ T cells in WT mice. To assess this, we transferred 239 bulk WT polyclonal CD8+ T cells to both WT and Dct -/recipients, then primed the mice 240 with TriVax and examined the effector response at day 7 post-immunization ( Figure 3A, 241 3B, 3D). Transferred WT Trp2/K b -specific cells did proliferate (~100-fold expansion), 242 albeit to a much lesser degree than endogenous Dct -/-Trp2/K b -specific cells.

243
Importantly, their expansion was comparable in both WT and Dct -/recipients ( Figure  244  3B, D), suggesting that the WT CD8 + T cells remained hyporesponsive even in an 245 environment where endogenous cells were not tolerant to Trp2, supporting a cell-246 intrinsic basis for the impaired reactivity of WT Trp2/K b -specific CD8 + T cells.

248
We also assessed the performance of Dct -/-CD8 + T cells when transferred into Dct -/-249 and WT hosts prior to priming to determine whether they would acquire tolerance in the 250 WT environment ( Figure 3A). These Trp2/K b -specific donor cells were able to expand 251 robustly in both Dct -/and WT recipients ( Figure 3C, E), further demonstrating a lack of 252 extrinsic regulation in the WT environment. Dct -/cells actually performed better in WT 253 recipients than in Dct -/recipients; the basis for this outcome is not clear but could be 254 due to reduced competition by endogenous Trp2/K b -specific cells in WT hosts. 255 Preliminary studies indicated that Dct -/cells still showed strong expansion when the 256 interval between cell transfer and TriVax was extended from one day to one week, 257 suggesting that these cells did not acquire tolerance characteristics within this 258 timeframe (data not shown).

260
We conducted similar transfers utilizing LmTrp2 instead of TriVax, again finding 261 evidence of cell-intrinsic tolerance. The transferred cells behaved in accordance with 262 the donors' Trp2 expression rather than that of the recipients: WT cells remained 263 tolerant when primed in Dct -/recipients, while Dct -/cells retained the ability to expand 264 when primed in WT recipients (Supplemental figure 2E, 2F). Collectively, these data 265 indicate that cell-intrinsic mechanism(s) enforce tolerance among WT Trp2/K b -specific 266 cells. 267 268 WT Trp2/K b -specific cells are capable of an acute response to Trp2 269 270 Although the response to Trp2 immunization was weaker in WT versus Dct -/mice, the 271 WT response was still substantial ( 2009). Alternatively, it was possible that fewer clones would be recruited into the Trp2 275 response in WT mice, leading to decreased expansion relative to Dct -/animals from the 276 initiation of an immune response. To distinguish between these possibilities, we studied 277 the expansion kinetics of the Trp2/K b -specific response in WT and Dct -/mice. In order 278 to track early polyclonal responses, TriVax with a higher dose of Trp2 peptide was used 279 in these studies, and tetramer enrichment was used to isolate Trp2/K b -specific cells. 280 Interestingly, WT Trp2/K b -specific cells were capable of an initial response that largely 281 paralleled that shown by their Dct -/counterparts ( Figure 4A). One day after TriVax 282 immunization, few cells were isolated, likely due to either trapping within the tissues 283 (Weninger et al., 2001) or TCR downregulation (Cai et al., 1997). Slightly more Trp2/K b -284 specific cells were identified in WT mice on day two, while increased numbers of 285 Trp2/K b -specific cells were seen in Dct -/mice on days three through five. By days six 286 and seven after high dose TriVax immunization, Trp2/K b -specific cells in Dct -/mice 287 outnumbered those in WT mice by an average ratio of 4:1. Although significant, these 288 differences in expansion were modest in comparison with the > 1000-fold expansion of 289 Trp2/K b -specific cells in both strains ( Figure 4A). Preliminary assessment of apoptosis 290 induction (annexin V staining) showed no differences between the strains at one or 291 three days after TriVax (data not shown). 292 293 Figure 4. WT Trp2/K b -specific cells are capable of an initial response to Trp2 similar to that of Dct -/cells WT and Dct -/mice received intravenous injections of TriVax with 200 ug Trp2 peptide. Tetramer enrichment was used to enumerate Trp2/K b -specific cells and assess their phenotype at the indicated time points following immunization (A-C, E). The ratio between the mean experimental PE MFI of Trp2/K b -specific cells in WT mice relative to Dct -/mice is plotted in D, with each symbol representing one experiment. Data are compiled from three or more experiments in A, D, and D. Representative flow plots from one day four experiment are shown in B, and the same representative day four experiment is shown in E. Squares indicate male animals; the dotted line indicates the average naïve precursor frequency from the spleen and lymph nodes. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001 by one-way ANOVA with Sidak's multiple comparisons test (performed on logtransformed data in [a]).
We also assessed the phenotype of responding Trp2/K b -specific cells acutely after 294 TriVax. With this approach, CD69 did not serve as a reliable indicator of activation due 295 to the type I interferon response induced by poly(I:C) leading to CD69 upregulation 296 (Shiow et al., 2006), and widespread CD44 expression was seen in both tetramer 297 positive and negative cells because of the potent inflammatory response unleashed by 298 this method of immunization. Accordingly, we tracked CD25 expression as an indicator 299 of activation. CD25, the high affinity alpha component of the IL-2 receptor, is 300 upregulated with activation in certain situations (Valenzuela et al., 2002) and may 301 enable a stronger effector response by cells expressing it (Obar et al., 2010). The 302 proportion of Trp2/K b -specific cells expressing CD25 was significantly greater in Dct -/-303 mice on day four, and trended higher on days two and three ( Figure 4B, 4C). The CD25 304 MFI of CD25+ cells was also higher on Dct -/-Trp2/K b -specific cells on day four ( Figure  305 4B), suggesting that Dct -/cells expressed more CD25 on a per-cell basis. Once again, 306 Dct -/-Trp2/K b -specific cells displayed significantly higher tetramer MFI than WT cells on 307 days two through seven, but the avidity differences detected by tetramer staining did not 308 demonstrate a progressive increase with time; the ratio between the WT and Dct -/-309 tetramer MFI transiently dropped at days 4-6, but the ratio at day 7 was similar to that 310 of pre-immune cells ( Figure 4D, S3A). Tetramer MFI was highest among the CD25+ 311 subset for both WT and Dct -/cells; the tetramer MFI of CD25+ WT cells was similar to 312 the MFI of the overall tetramer-binding Dct -/population on day four ( Figure 4E).

314
We also assessed the early response following peptide stimulation alone, since this 315 would be analogous to encountering Trp2 in a non-inflammatory context. We again 316 found the early response to be similar between WT and Dct -/-Trp2/K b -specific cells. The 317 number (Supplemental figure 3B) and phenotype of WT Trp2/K b -specific cells was 318 comparable to that of Dct -/cells on day one post-peptide. The activation markers CD44 319 and CD69 were similarly upregulated in both (Supplemental figure 3C, 3D), however, 320 responses began to diverge by day two after peptide stimulation, and by day three 321 significant differences in number and CD44 expression had emerged, with Dct -/cells 322 clearly outperforming WT cells (Supplemental figure 3B, 3C).

324
These findings indicate that the response to Trp2/K b in WT and Dct -/mice follows 325 similar kinetics and magnitude but that expansion in WT animals terminates 326 prematurely. Interestingly, we did not observe a progressive increase in apparent TCR 327 avidity over time among the Dct -/responder pool relative to WT cells, as might be 328 expected if the subset of Dct -/-T cells with higher avidity TCRs became more dominant 329 in the response to Trp2/K b .

331
Single-cell sequencing reveals an impaired ability to differentiate into a highly 332 proliferative population among WT Trp2/K b -specific cells 333 334 To better understand the defects in expansion and functionality observed among WT 335 Trp2/K b -specific cells and assess the heterogeneity within this population, we performed 336 single-cell RNA sequencing on Trp2/K b -specific cells from WT and Dct -/mice at day 337 three after TriVax priming. After initial data processing, the WT and Dct -/datasets were 338 merged (Stuart et al., 2019); clusters based on the cells' transcriptomes were generated 339 in an unbiased manner and visualized using uniform manifold approximation and 340 projection (UMAP). Cells clustered into two major groups separated along the x-axis, 341 each comprised of smaller clusters ( Figure 5A). Interestingly, over half of the cells 342 (58%) from Dct -/mice were localized in cluster 0, but this cluster was nearly devoid of 343 WT cells, representing only a small subset (13%) of WT cells ( Figure 5B). Suspecting 344 that this cluster might contain a more functional subset poorly represented in the WT 345 population, we assessed its characteristics in more detail. Because this cluster made up 346 the majority of Group A (left group), we performed differential gene expression analysis 347 between the two major groups in the merged dataset: A and B (right group).

349
Histone genes (e.g., Hist1h1b, Hist1h1e, Hist1h1d, Hist2h2ac) were among the most 350 upregulated in Group A compared to Group B; these genes are commonly induced in 351 association with cellular replication (Mei et al., 2017). Other genes associated with 352 proliferation, such as Myc, Nolc1, Npm1, and Ccne2, were also upregulated in Group A 353 ( Figure 5C), and cell cycle analysis revealed that the majority of cells in Group A were in 354 stages G2/M or S of the cell cycle ( Figure 5D). Among cells in cluster 0 (Group A), 76% 355 of Dct -/cells were in G2/M or S versus 42% of WT cells ( Figure 5E). Gene set 356 enrichment analysis revealed a strong enrichment of gene sets comprising Myc targets, 357 E2F targets, and genes related to mTORC1 signaling and the G2/M checkpoint 358 (Supplemental figure 4A).

360
Many of the cells in Group A expressed CD25, with the majority of the remainder 361 located in cluster 2 of Group B ( Figure 5F); cells in these clusters also showed 362 enrichment for a gene signature associated with IL-2 receptor signaling (Supplemental 363 figure 4B). This aligns with our finding that the frequency of cells expressing CD25 was 364 greater in Dct -/than WT Trp2/K b -specific effectors ( Figure 4C). Indeed, group A cells 365 showed significantly higher expression of certain genes relevant to the IL-2 signaling 366 pathway with known impacts on T cell function, such as Irf4 and Myc; previous work has 367 demonstrated that signaling through the IL-2 receptor is important for sustained Myc 368 expression (Preston et al., 2015). Based on these data, we administered IL-2 complex 369 (IL-2 + anti-IL-2 S4B6 antibody) treatment to WT and Dct -/mice previously primed with 370 TriVax to determine whether this would correct the defective proliferation of the WT 371 Trp2/K b -specific cells. IL-2 complex acts through the b and g components of the IL-2 372 receptor, negating the impact of differential CD25 expression. IL-2 complex treatment 373 on day five after TriVax or LmTrp2 improved the expansion of both WT and Dct -/-374 Trp2 Taken as a whole, the RNA-sequencing data suggest that WT Trp2/K b -specific cells are 382 deficient in their ability to form the more proliferative subpopulation that comprises a 383 majority of the Dct -/-Trp2/K b -specific population on day three after priming. 384 Nevertheless, proliferation of the WT population was not entirely constrained, since this 385 pool continued to expand over successive days (Figure 4). vehicle (acetone/olive oil) was applied to the right flank ( Figure 6A). Analysis of the 406 blood six days after transfer and boosting revealed expansion of both types of donor 407 cells; although there was a trend for transferred Dct -/cells to expand to a greater 408 degree than donor WT cells, the difference was not statistically significant ( Figure 6B). 409 Recipient mice were subsequently monitored for vitiligo development on a weekly basis 410 and scored using a numeric metric (Supplemental figure 5A). nevertheless, preliminary studies using irrelevant TCR transgenic mice (P14) as 427 recipients showed that Trp2/K b -specific donor cells from Dct -/mice were still able to 428 induce vitiligo in this setting (data not shown).

430
Vitiligo severity (average vitiligo score) was positively correlated with the number of 431 donor Trp2/K b -specific cells in the blood on day 6 after transfer and TriVax boost 432 (Supplemental figure 5D). This suggests that the enhanced proliferative capacity of Dct -433 /cells was a factor in their superior ability to induce vitiligo, although it does not rule out 434 additional qualitative differences between the WT and Dct -/populations.

436
In summary, in contrast to the relatively modest differences in the expansion of Trp2/K b 437 responders in WT and Dct -/mice, the ability of these populations to mediate 438 autoimmune damage-melanocyte destruction-was strikingly different. knowledge is critical to designing effective therapies to restrain these cells (e.g., to 457 control autoimmune disease) or induce their responses (e.g., for cancer 458 immunotherapy). As described in this report, we developed a polyclonal mouse model 459 for non-deletional self-tolerance, enabling us to define the characteristics of these cells 460 and their reactivity in a physiological setting.

462
We found that the pre-immune population of Trp2/K b -specific cells in WT and Dct -/-463 strains were qualitatively similar, sharing a naïve phenotype and indistinguishable gene 464 expression profile; there were no clear signs of prior antigen exposure among the WT 465 cells. The size of the Trp2/K b -specific precursor pool was only slightly (although 466 significantly) smaller in WT mice and the response to Trp2 immunization was 467 substantial in both strains, leading to a > 1000-fold expansion of Trp2/K b -specific cells in 468 both WT and Dct -/mice. Despite these commonalities, cells primed in Dct -/mice 469 proliferated more extensively and were capable of much more rapid and widespread 470 tissue destruction, read out as vitiligo, after adoptive transfer.

472
Our adoptive transfer studies showed that the observed restraint in the WT Trp2/K b -473 specific response did not depend on extrinsic factors but was a cell-intrinsic feature of induction was not a focus of the current study, but it will be interesting to determine 487 whether instances of non-deletional CD8 + T cell self-tolerance correlate with self-488 antigen expression patterns in the thymus (e.g., AIRE-regulated tissue-specific antigen 489 expression).

491
Many of the characteristics we report for Trp2/K b -specific CD8 + T cells in WT mice are 492 strongly reminiscent of T cells with low affinity/avidity for antigen ( induced by non-tolerant and/or higher affinity cells. Our data indicate that the opposite 508 can also occur: despite largely overlapping tetramer staining profiles, WT and Dct -/-509 Trp2/K b -specific cells exhibit markedly different abilities to mediate widespread vitiligo. 510 Hence, the impact of CD8 + T cell tolerance toward some self-antigens only partially 511 limits expansion but can prevent the generation of cells readily capable of potent tissue 512 destruction: tolerance is not a binary state.

514
A goal of this work was to define which characteristics, if any, could be associated with 515 self-tolerant CD8 + T cells that escape clonal deletion. Unexpectedly, we found that 516 multiple measures-the enumeration, peptide/MHC tetramer staining intensity, 517 phenotype, and gene expression of pre-immune cells as well as the ability of these cells 518 to proliferate following immunogenic exposure to self-antigens-failed to provide a 519 robust metric for distinguishing CD8 + T cells that are functionally tolerant (i.e., unlikely to 520 cause autoimmune pathology) versus non-tolerant (i.e., have a high propensity to 521 induce autoimmunity). This finding highlights the limitations of currently available assays 522 for accurately predicting responsiveness to self-antigens. 523 524 However, we were able to delineate an inflection point following priming at which the 525 responses of tolerant and non-tolerant cells diverged. Flow cytometry and single-cell 526 RNA sequencing of Trp2/K b -specific CD8 + T cells soon after priming demonstrated that 527 WT responders failed to differentiate into a CD25+, IRF4+ population (a characteristic of 528 most Dct -/responder cells) and indicated that WT cells showed poor commitment to 529 sustained proliferation. These combined features may be useful for further defining the 530 responses by self-antigen specific cells that are or are not capable of overt tissue 531 destruction. 532 533 20 The early effector population of Trp2/K b -specific cells demonstrates heterogeneity on a 534 transcriptomic level in both strains. Whereas the majority of Dct -/cells show a highly 535 proliferative phenotype characterized by active cell cycling and responsiveness to 536 mTOR and Myc, few WT Trp2/K b -specific cells fall into this group. The reason(s) 537 underlying the inability of WT cells to optimally engage these important pathways and 538 proliferate efficiently requires further investigation but may relate to impaired sensitivity 539 to endogenous IL-2 or other cytokines, the composition of the TCR repertoire, and/or 540 altered TCR signaling. How these or other factors relate to the relative inability of 541 primed WT Trp2/K b -specific cells to mediate overt tissue damage is currently unclear, 542 but it will be critical to identify the cellular and molecular mechanisms involved in future 543 studies. Ongoing and future work will address the TCR repertoire and downstream 544 signaling, metabolic characteristics, and epigenetic landscape of WT versus Dct -/-545 Trp2 We were able to identify CD8+ T cells specific for other melanocyte epitopes/antigens in 552 pre-immune mice; these cells had a similar phenotype to WT Trp2/K b -specific cells. 553 Accordingly, we predict that our results will apply to other populations of CD8+ T cells 554 specific for melanocyte and potentially other tissue-restricted antigens. Similar 555 populations of self-specific CD8 + T cells may exist in humans, and the ability of such 556 cells to respond to self-antigen immunization while not causing autoimmune damage is 557 relevant for understanding the limits of "breaking" tolerance, e.g., for cancer 558 immunotherapy. Indeed, our results align with work examining polyclonal self-antigen-559 specific cells in human adults (Yu et al., 2015) with regard to the phenotype (modestly 560 lower tetramer MFI, lower CD25 expression) and response to cognate peptide 561 (diminished) observed among tolerant cells. Another study examining self-specific CD8 + 562 T cells (Maeda et al., 2014) attributed their restrained responsiveness to Treg-mediated 563 suppression; while we did not detect a cell-extrinsic regulatory mechanism in our 564 studies, it is certainly possible that this mechanism limits the response to some self-565 antigens.

567
Our finding that polyclonal melanocyte-specific cells exhibit covert cell-intrinsic tolerance 568 characterized by a partial defect in proliferation and a profound defect in tissue damage 569 has implications for utilizing such cells therapeutically. This model has clear relevance 570 to human physiology and will be useful in exploring methods of correcting the 571 proliferative defects of tolerant cells to more effectively mobilize them in cancer 572 immunotherapy approaches targeting tumor antigens shared with self. 573

Materials and methods 574 575
Mice 576 C57BL/6 (WT) mice were obtained from Charles River laboratories and housed in 577 specific pathogen-free conditions at the University of Minnesota. Dct -/mice on a 578 C57BL/6 background were developed by A. Andy Hurwitz when at the NCI; the mice 579 were subsequently bred in-house on different congenic backgrounds and housed in 580 specific pathogen-free conditions. Animals were used at 6-14 weeks of age. All animal 581 experiments were approved by the Institutional Animal Care and Use Committee at the 582 University of Minnesota. In accordance with NIH guidelines, both male and female 583 animals were used in experiments; males are indicated by square symbols in the 584 figures.

586
Tetramer enrichment 587 Tetramer enrichment was used to isolate antigen-specific cells from pre-immune or 588 acutely challenged mice. A modification of the method used by Obar et al. (Obar et al.,589 2008) was employed. Following digestion with collagenase D, single-cell suspensions 590 were prepared from the spleens (acutely challenged mice) or spleen and 591 peripheral/mesenteric lymph nodes (pre-immune mice). When possible, the same 592 tetramer (Trp2180-188/K b ) was used in both APC and PE to ensure specificity. Anti-PE 593 and anti-APC beads and magnetized columns (both from Miltenyi Biotec) were used to 594 enrich for tetramer-bound cells. Samples were stained and analyzed by flow cytometry; 595 CountBright counting beads (Invitrogen) were used for enumeration. 596 597 In vivo priming with Trp2 598 TriVax immunization was used as previously described (Cho & Celis, 2009); mice were 599 immunized intravenously (via tail vein injections) with Trp2180-188 peptide or Trp2 and 600 B8R20-27 peptides, agonist-anti CD40 antibody (BioXCell), and vaccine-grade poly(I:C), 601 a toll-like receptor 3 agonist (InvivoGen). Peptide doses of 50, 100, and 200 µg were 602 used for effector timepoints, transfer experiments, and acute timepoints, respectively. 603 Animals that received TriVax immunization via intraperitoneal instead of intravenous 604 injection were removed from the analysis.

613
Ex vivo stimulation 614 In some experiments, splenocytes were stimulated ex vivo after isolation from infected 615 mice. Splenocytes were incubated with Trp2 peptide (10 -6 M) and Golgiplug (BD 616 Biosciences) for 4-6 hours at 37°C; parallel wells with no peptide were used as a 617 control. Cells were washed and stained with surface antibodies, followed by fixation and 618 permeabilization with a FoxP3 Fix/Perm kit (eBioscience) or FoxP3/transcription factor 619 staining buffer kit (Tonbo Biosciences) and staining with intracellular antibodies. 620 621 622 Adoptive transfer experiments 623 Bulk polyclonal CD8+ T cells were isolated from the spleen and lymph nodes of WT or 624 Dct -/mice using negative magnetic enrichment (CD8a+ T cell Isolation Kit; Miltenyi 625 Biotec). Enriched CD8+ T cells (typically ~75-90% pure) were resuspended in sterile 626 PBS and 2-2.5 x 10 6 CD8+ T cells were injected intravenously per recipient mouse; 627 recipient mice were congenically distinct. One day later, the recipient mice were 628 immunized with TriVax or LmTrp2 intravenously or intraperitoneally. Mice were 629 sacrificed for analysis seven days later.

631
Bulk RNA sequencing of pre-immune mice 632 Trp2/K b -specific cells were isolated from pre-immune WT and Dct -/mice using tetramer 633 enrichment followed by fluorescence-activated cell sorting on double tetramer positive 634 cells. Cells were isolated from three separate cohorts, with each cohort comprising eight 635 WT and eight Dct -/mice. The Clontech StrandedRNA Pico Mammalian workflow was 636 used for library preparation, and samples were sequenced using an Illumina NextSeq 637 instrument (2x75 bp paired end reads).

639
Bulk RNAseq analysis 640 Raw sequencing data were demultiplexed by sample into FASTQs (mean 24.6 million 641 reads/sample) and mapped against the mouse genome (Ensembl GRCm38 release 95) 642 using Hisat2 software (v 2.1.0). Gene level quantification was completed using Subread 643 featureCounts software (v 1.6.2) and the read counts table was processed in R (v 644 3.5.2). Differentially expressed genes were identified with DESeq2 software (v 1.22.2) 645 using a negative binomial model with effect size estimation completed by apeglm 646 algorithm via the lfcShrink function. Group comparison p-values were adjusted by the 647 Benjamini and Hochberg method to account for multiple hypothesis testing where genes 648 with a false discovery rate (FDR) q < 0.05 were investigated in downstream analyses.

650
Single-cell RNA sequencing 651 WT and Dct -/mice were primed with TriVax (200 µg Trp2), and Trp2/K b -specific cells 652 were isolated on day three using tetramer enrichment followed by fluorescence-653 activated cell sorting. Cells were submitted for barcoding and library preparation using Single-cell RNAseq analysis 659 Raw sequencing data were processed using Cell Ranger (v 3.0.2; 10x Genomics) 660 software programs "mkfastq" for demultiplexing the WT and Dct -/-Illumina libraries and 661 "count" for read alignment against the mouse genome (mm10, provided by 10x 662 Genomics, ver 3.0.0) and generation of the mRNA transcript count table. Raw count 663 data were loaded into R (v 3.6.1) and analyzed with the Seurat R package (v was performed using the normalized, mean-centered, and scaled SCT dataset 681 (RunPCA function  Fig. S1 shows additional analysis of the Trp2/K b -specific population and cells specific 752 for other melanocyte epitopes in pre-immune mice. Fig. S2 presents data using LmTrp2 753 as an alternative method of stimulating with Trp2. Fig. S3 shows overlays of tetramer 754 staining at various time points after TriVax immunization, along with the acute response 755 to in vivo stimulation with Trp2 peptide alone. Gene set enrichment analysis of single-756 cell data and the response of WT and KO Trp2/K b -specific cells to IL-2C are presented 757 in Fig. S4. Fig. S5 shows the vitiligo scoring metric and correlation analysis of the 758 average vitiligo score relative to the number of transferred       (A) Tetramer staining of pre-immune lymphocytes (enriched fraction, gated on live, dumpnegative CD8+ T cells). Dual tetramer staining was used to facilitate more accurate gating on antigen-specific cells. (B, C) CD44 and CD122 staining of pre-immune Trp2/K b -specific cells from a WT mouse (B) and a Dct -/mouse (C). (D) Bulk RNA sequencing of Trp2/K b -specific CD8+ T cells from pre-immune WT and Dct -/mice was performed; differentially expressed genes were not identified between WT and Dct -/samples as shown by the volcano plot. (E) Quantification of CD8+ T cells specific for a D b -restricted Trp2 epitope and an epitope from tyrosinase-related protein 1 (Trp1) in pre-immune mice reveals a similar or slightly lesser number of cells in mice expressing antigen (WT mice) relative to those that do not. Samples used for tetramer enrichment of Trp1/D b -specific cells were obtained from shipped samples; accordingly, these data likely underestimate the precursor frequency. PC, principal component. * p < .05 by one-way ANOVA with Sidak's multiple comparisons test.

Supplemental figure 2. Response to infection with LmTrp2 in primary and transfer settings
Mice were infected with a recombinant Listeria monocytogenes strain expressing Trp2 (LmTrp2; A-D). The percent (A) or number (B, C) of splenic Trp2/K b -specific cells was assessed at the indicated day. (D) Day 7 splenocytes were stimulated for 4-6 hours with Trp2 peptide and intracellular staining was performed to assess cytokine production. (E, F) CD8+ T cells from pre-immune WT or Dct -/donors were negatively enriched and bulk CD8+ T cells were transferred into congenically distinct WT or Dct -/recipients. One day later, mice were infected with LmTrp2. Donor and endogenous cells were collected from the blood of recipient mice on day 7 following infection and assessed for Trp2/K b tetramer binding. Data in A, B, and D are representative of 3 similar experiments; data in C, E, and F represent individual experiments with 2-5 mice per group. Squares indicate male animals. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001 by unpaired t test (A-D) or one-way ANOVA with Sidak's multiple comparisons test (E, F). Endog, endogenous; recip, recipients.