TCR-MHC Interaction Strength Defines Trafficking and Resident Memory Status of CD8 T cells in the Brain

T cell receptor-Major histocompatibility complex (TCR-MHC) affinities span a wide range in a polyclonal T cell response, yet it is undefined how affinity shapes long-term properties of CD8 T cells during chronic infection with persistent antigen. Here, we investigate how the affinity of the TCR-MHC interaction shapes the phenotype of memory CD8 T cells in the chronically Toxoplasma gondii-infected brain. We employed CD8 T cells from three lines of transnuclear (TN) mice that harbour in their endogenous loci different T cell receptors specific for the same Toxoplasma antigenic epitope ROP7. The three TN CD8 T cell clones span a wide range of affinities to MHCI-ROP7. These three CD8 T cell clones have a distinct and fixed hierarchy in terms of effector function in response to the antigen measured as proliferation capacity, trafficking, T cell maintenance and memory formation. In particular, the T cell clone of lowest affinity does not home to the brain. The two higher affinity T cell clones show differences in establishing resident memory populations (CD103+) in the brain with the higher affinity clone persisting longer in the host during chronic infection. Transcriptional profiling of naïve and activated ROP7-specific CD8 T cells revealed that Klf2 encoding a transcription factor that is known to be a negative marker for T cell trafficking is upregulated in the activated lowest affinity ROP7 clone. Our data thus suggest that TCR-MHC affinity dictates memory CD8 T cell fate at the site of infection.


52
CD8 T cells are a cornerstone of the adaptive immune defence to intracellular pathogens with 53 their capacity to operate as antigen-experienced effector and memory cells. Pathogen-specific 54 CD8 effector T cells rapidly expand and differentiate during the acute infection, followed by a 55 phase of contraction and development of long-lived memory T cells (1,2). Most of our 56 understanding of T cell responses to chronic infections is derived from models where pathogen 57 control is incomplete and T cell become functionally impaired or exhausted over time (3,4). 58 We thus lack knowledge of what drives long-lasting control of chronically persistent 59 pathogens. 60 61 The interaction of the TCR with the pathogen antigenic epitope loaded on the MHC is essential 62 in maintaining effective CD8 T cell control of persistent intracellular pathogens. The ab TCR 63 stochastically assembles and is selected during thymic development, and it is via this receptor 64 that the immune system tunes the breath and strength of its response (2,5). Efforts have been 65 made to elicit the effect of TCR-MHC affinity on the fate of the resulting T cells, however, 66 often this relied on varying the antigenic peptide rather the TCR (2,6). The simple question of 67 how T cells of different affinity to a given antigen fare during chronic infection remains 68 unresolved.

70
In order to model a persistent chronic infection, we deemed a resistant mouse strain infected 71 with the protozoan parasite Toxoplasma gondii to be most suitable. Toxoplasma is the most 72 common parasitic infection in man, whereby in immunocompetent hosts the acute phase of 73 infection is generally asymptomatic and proceeds to the chronic phase, which is incurable and 74 defined by tissue cyst formation preferably in the brain. The parasite poses a serious health 75 threat to immunocompromised individuals, especially AIDS patients. It is unclear how 76 Toxoplasma maintains the intricate balance between survival and host defence. CD8 T cells 77 and their ability to produce IFNg have been shown to secure the latency of the parasitic 78 infection (7,8).

80
Mice harbouring the MHCI allele H-2L d (e.g. BALB/c) control Toxoplasma infection due to 81 an immunodominant epitope derived from the GRA6 parasite protein (9-11). BALB/c mice 82 exhibit very few tissue brain cysts and the functionality of their CD8 T cells in the Toxoplasma-83 infected brain is defined by their capacity to produce IFNg and perforin (7,12,13). Recently, 84 using the murine BALB/c chronic Toxoplasma model, a T cell population (Tint) in an 85 intermediate state between effector and memory status was discovered, highlighting the value 86 of this model for defining the fate of CD8 T cells during chronic infection with persistent 87 antigen (14).

89
In addition to memory T-cell populations, distinct memory T-cell population that persist long 90 term within non-lymphoid tissue have recently been documented and are resident in nature, 91 self-renewing, and highly protective against subsequent infections (15,16 Rather than varying the antigenic peptide, we sought to use distinct clonal T cells . In order to  102  answer how TCR-MHC affinity dictates trafficking and phenotype of memory CD8 T cells in  103  the brain during chronic infection, we employed three distinct clonal CD8 T cells, each  104  expressing a natural TCR recognizing the Toxoplasma antigen ROP7 (24,25). These cells were 105 obtained from transnuclear (TN) mice generated by somatic cell nuclear transfer from a nucleus 106 of a Toxoplasma antigen-specific CD8 T cell and have different affinity for MHC class I loaded 107 with the same ROP7 peptide (24,25). 108 109 Here we report that TCR-MHC affinity dictates the potential of a CD8 T cells to home to the 110 Toxoplasma-infected brain. We employed three natural CD8 T cell clones derived from a 111 resolving Toxoplasma infection by somatic cell nuclear transfer, defined to possess different 112 affinities for the same Toxoplasma antigen ROP7 (24,25). The two T cell clones with higher 113 affinity, R7-I and R7-III were found in the brain during chronic infection, while the lowest 114 affinity clone R7-II was not, despite all three clones being activated during the acute phase of 115 infection. As possible causes for this divergent homing we observed high expression of the 116 negative regulator of T cell activation Klf2 and its regulated genes in peptide-activated R7-II 117 T cells. Additionally, Ctla4, a negative regulator of T cell responses was also upregulated on 118 R7-II T cells. The highest affinity clone, R7-I, persisted longer during the chronic phase of 119 infection than R7-III and was able to generate more TRM cells in the brain. Thus, our results 120 indicate that higher affinity of the TCR-MHC interaction is better for trafficking and persisting 121 of the specific CD8 T cells at the site of chronic infection, here brain. 122

249
Quantitative real-time PCR was performed using Maxima SYBR Green/Rox qPCR master mix 250 (Thermo Scientific).

Statistical analyses 270
GraphPad software (Prism) was used to perform statistical tests. Comparisons between two 271 groups were made using Student's t test. Comparisons between multiple groups were made 272 using one-way analysis of variance (ANOVA) test. Levels of significance are denoted as 273 follows: * p ≤ 0.05, ** p ≤ 0.01, *** p≤ 0.001, **** p ≤ 0.0001. Non-significant results are 274 either not marked or indicated as NS. 275

277
Three CD8 T cell clones specific for the same peptide respond differently to in vitro TCR 278 stimulation with cognate antigen 279 We previously described transnuclear mouse lines (24), which we used herein as a source of 280 three CD8 T cell clones specific for the same peptide (IPAAAGRFF) derived from the ROP7 281 protein of Toxoplasma gondii. We refer to these CD8 T cell clones as R7-I, R7-II and R7-III 282 CD8 T cells (24). We previously showed these CD8 T cell clones to differ in their TCR 3D 283 affinity to cognate ROP7 peptide, with R7-I being the strongest binder at 4 µM and R7-II the 284 weakest at 109 µM. R7-III has a binding affinity of 24 µM (25). To further define the kinetics 285 of TCR signalling after stimulation with ROP7 peptide, we measured ER-driven calcium 286 release, phosphorylation of ERK1/2 kinase and cell proliferation as determinants of TCR 287 reactivity. Both the calcium release and phosphorylation assays reflected the hierarchy of the 288 TCR-MHC binding 3D affinity and were the fastest and strongest in R7-I CD8 T cells, while 289 the R7-II CD8 T cell response was lowest (Fig 1A and B). Additionally, we noted that R7-II 290 CD8 T cells had a basal level of free intracellular calcium that was higher than that of R7-I and 291 R7-III CD8 T cells ( Fig 1A). In the in vitro proliferation assay R7-II CD8 T cells were not able 292 to proliferate efficiently even at the highest (500 µM) concentration of ROP7 peptide loaded 293 onto splenocytes while R7-I and R7-III CD8 T cells reached the highest division index at 294 concentrations 5 µM and 0.5 µM respectively (Fig. 1C). These in vitro experiments suggest 295 that the 3D surface plasmon resonance affinity of the TCR-MHC binding reflects the strength 296 of downstream signalling and partially translates to proliferation capacity in vitro.

298
To provide additional insight into the functional response of the three R7 CD8 T cell clones 299 during Toxoplasma infection, we determined their 2D affinity for the ROP7 antigen. The 300 micropipette adhesion frequency assay provides 2D based measures of TCR affinity for pMHC 301 in a context that is membrane anchored. 2D affinity correlate more closely to functional 302 responses than do 3D affinity measurements whose measurements are based on purified 303 proteins. Over 40 T cells for each clone were analysed to reveal similar high affinities for R7-304 I and R7-III (geometric means being 1.38E-04 and 1.27E-04 respectively) and that R7-I whilst 305 having a similar affinity to R7-III ( Fig 1D) has a higher adhesion frequency than R7-III ( Fig  306  1E) that being 0.91 and 0.81 respectively. R7-II had a 3-fold lower 2D affinity (7.01E-05). The 307 3-fold difference in affinity is functionally relevant as previously, we have demonstrated that 308 during Polyoma infection CD8 T cells with the highest 2D affinity are found in the CNS and 309 eventually comprise the TRM population (32). In addition, we have reported CD4 T cells 310 mediating EAE carry a 2-fold higher affinity as compared to the peripheral T cells (33).

312
All three clones of R7 CD8 T cells are efficiently primed in the acute phase of Toxoplasma 313 infection 314 Next, we sought to verify if these differences observed in vitro would still hold true in vivo. R7 315 CD8 T cells were adoptively transferred into congenic naïve recipient mice (CD90.1 BALB/c). 316 Subsequently, mice were orally infected with ME49 Toxoplasma tissue cysts. The donor R7 317 CD8 T cells could then be followed during the acute and chronic phase of infection ( Fig. 2A). 318 We were able to observe proliferated cells for all three R7 clones in the mLN earliest 6 days 319 post infection (p.i.) (Fig. 2B). Limited number of R7-II donor cells could be found in the mLN. 320 However, those that were recovered from mLN had low CFSE level indicating that they had 321 proliferated similarly to R7-I and R7-III CD8 T cells. Additionally, the R7 donor CD8 T cells 322 were all activated to the same extent based on CD69 expression (Fig. 2B). We conclude that 323 all three R7 clones are responsive to a Toxoplasma infection in vivo, as measured by 324 proliferation and activation status. 325 326

R7-II CD8 T cells do not persist and do not reach the brain of recipient mice during 327
Toxoplasma infection 328 Cysts in the brain characterise the chronic phase of Toxoplasma infection. IFNg produced by 329 CD8 T cells is crucial for the maintenance of the quiescent cyst form of Toxoplasma (8,13). 330 We showed that the three R7 CD8 T cell clones could be primed and proliferated in the acute 331 phase of Toxoplasma infection. Next, we investigated if R7 CD8 T cells could be found in the 332 brain in the chronic phase of Toxoplasma infection. We analysed brain, spleen, mLN and non-333 draining LN for the presence of transferred R7 CD8 T cells 3 weeks p.i.. R7-I and R7-III CD8 334 T cells were found in significant numbers in the brain at 3 weeks p.i. (Fig. 3A). Percentages 335 and absolute numbers of donor R7 CD8 T cells of all CD8 T cells in a given organ were 336 different depending on the clone (Fig. 3A). The R7-II clone was found in insignificant 337 percentages and numbers in all tested organs. R7-I and R7-III clones were present in higher 338 percentages from 2-15% of all CD8 T cells depending on the organ. There was no significant 339 difference between percentages of the R7-I and R7-III clone in the brain. In the spleen and 340 mLN we observed significantly higher percentages of the R7-III CD8 T cells. We also 341 determined absolute cell number for each clone at each site and did not observe significant 342 differences between the R7-I and R7-III clone (Fig. 3A right panel).

344
We analysed earlier time points to estimate the time when donor cells reached the brain and to 345 assess if there is a difference in donor cell number or percentages in the acute phase of 346 infection. R7-I and R7-III CD8 T cells were observed in the brain as early as 10 days p.i., 347 however, we failed to detect a distinct R7-II CD8 T cell population in the brain (Fig. 3B). R7-348 II CD8 T cells could not be observed in prominent numbers in any of the tested primary and 349 secondary lymphoid organs suggesting a lack of proper expansion and homing of the 350 transferred population to the brain ( Fig. 3B and Fig. S1A). R7-I and R7-III CD8 T cells were 351 a major part (60-80%) of the total CD8 T cell population in the brain at day 10 p.i.. Their 352 percentages decreased throughout time while host CD8 T cells reached the brain suggesting 353 that the transferred T cell clones had a head start compared to newly formed Toxoplasma-354 specific CD8 T cells of the host. There were no significant differences in percentages or 355 numbers of donor population between R7-I and R7-III CD8 T cells at day 10 or 2 weeks p.i. 356 ( Fig. 3B and Fig. S1A).

358
To dissect the reasons for the poor expansion and lack of R7-II cells in the brain we compared 359 the transcriptional profiles of in vitro ROP7-activated R7-II vs R7-I or R7-III cells. In the list 360 of the top 10 genes upregulated in R7-II (Table 1) we found Klf2 encoding a transcription factor 361 that is known to be important for T cell trafficking between the blood and lymphoid organs 362 (34-36). Klf2 is highly expressed in naïve and memory T cells but downregulated in effector 363 T cells upon binding of the TCR to its cognate peptide (37). The stronger the binding affinity, 364 the lower the expression of Klf2 and the better the activation of the T cells (38). Additionally, 365 S1pr1 which is regulated by Klf2 was also in the top 10 upregulated genes. We analysed the 366 expression of Klf2 across the unstimulated and activated samples of the RNASeq experiment 367 (Fig. 3C, left graph) as well as at the genes known to be regulated by Klf2 such as CXCR3, Sell 368 (CD62L), S1pr1, Itga4, CXCR4, Il7ra, Il6ra (Fig. 3C, right graph). Klf2 expression was at 369 similar level in all three naïve CD8 T cell clones. After activation, all three clones 370 downregulated Klf2 however downregulation in R7-II was the weakest reflecting the lowest 371 binding affinity of Rop7 peptide to R7-II. Expression of Sell (CD62L), S1pr1, Itga4, CXCR4, 372 Il7ra, Il6ra mirrored Klf2 expression being highest in R7-II and was comparable between R7-373 I and R7-III. On the other hand, CXCR3 expression was lowest in R7-II. We performed an 374 independent in vitro activation experiment and confirmed by qRT-PCR the RNAseq results 375 observed (Fig. 3D). mRNA expression detected by qRT-PCR correlated with the RNAseq data 376 for all tested genes besides the CXCR3 gene. 377 378 CTLA4 is known as a negative regulator of T cell responses. We analysed the expression of 379 the Ctla4 transcript in the RNASeq of activated ROP7 CD8 T cells as its high expression in 380 R7-II cells could explain their poor performance. As expected, we observed the highest 381 expression of Ctla4 in R7-II CD8 T cells in comparison with the two other R7 clones (Fig. 3E,  382 left graph). We could confirm the RNAseq data in an in vivo experiment, where R7 T cells 383 were adoptively transferred to congenic mice and analysed in LNs 5 days after infection with 384 Toxoplasma. CTLA4 surface expression determined by FACS in R7-II cells was highest in 385 comparison to R7-I and R7-III CD8 T cell clones (Fig. 3D, right graph). 386 387 R7-III has a higher contraction rate and does not persist in the late phase of chronic 388 infection 389 At 5 weeks p.i. compared to 3 weeks p.i., we observed a dramatic decrease in the R7-III CD8 390 T cell population in all of tested lymphoid and non-lymphoid organs while the R7-I CD8 T cell 391 population decreased more subtly, which could be explained by natural contraction of the 392 population after an initial expansion phase ( Fig. 4A and Fig. S1B).

394
The formation of memory CD8 T cells in persistent infections is still controversial (39). In

BALB/c mice where persistent Toxoplasma infection is known to be controlled by CD8 T cells 396 (13), the presence of memory CD8 T cells is expected. We analysed if R7 CD8 T cells 397 differentiated into memory cells and if the ratio of differentiation into effector and memory 398
precursors was the same between R7-I and R7-III CD8 T cells. Short lived effector cells 399 (SLEC, CD127 -KLRG1 + ) were present in similar percentages in R7-I and R7-III CD8 T cell 400 populations in the brain and spleen at 3 and 4 weeks p.i. (Fig. 4B). Memory precursor effector 401 cells (MPEC, CD127 + KLRG1 -) where not present at 3 weeks p.i., however we detected 402 population of MPEC at 4 weeks p.i. in the spleen in similar percentages between R7-I and R7-403 III CD8 T cell (Fig. 4B, bottom panel).

405
In the chronic phase of Toxoplasma infection in C57BL/6 mice CD8 T cells in the brain are 406 exhausted and express high levels of the exhaustion marker PD1 (40). Blockade of the PD1-407 PDL1 pathway has been shown to rescue the exhaustion phenotype of CD8 T cells and prevent 408 mortality of chronically Toxoplasma infected animals (41). In contrast to C57BL/6 mice, 409 BALB/c mice are resistant to chronic Toxoplasma infection (9) and Toxoplasma GRA6-410 specific CD8 T cells in BALB/c mice lack PD1 expression during the chronic phase (14). We 411 investigated if R7 CD8 T cells in brain of BALB/c mice at 3 weeks p.i. were exhausted. Almost 412 90% of R7 CD8 T cells expressed PD1 (Fig. 4C). No significant difference was observed 413 between R7-I and R7-III CD8 T cell populations in the brain suggesting that it is not exhaustion 414 that leads to greater contraction of the R7-III CD8 T cell population. As PD1 can be also a 415 marker for recently activated cells we analysed the ability of PD1 + cells to produce IFNg, since 416 cytokine production is lost in a truly exhausted cell (42). Brain mononuclear cells from week 417 3 of infection were ex vivo re-stimulated with PMA and ionomycin and stained for IFNg (Fig.  418 4D). More than half of PD1 positive cells were able to produce IFNg and only 1/3 of R7 CD8 419 T cells in the brain were positive for PD1 and negative for IFNg. At the same time, only 5% of 420 R7 cells in spleen expressed PD1 and did not produce IFNg. No significant difference between 421 R7-I and R7-III CD8 T cells was observed indicating that the reason for the disappearance of 422 R7-III CD8 T cells is independent of their exhaustion state.

424
Next, we considered CXCR3 as a candidate molecule to unravel the differences in persistence 425 between R7-I and R7-III. The CXCR3 receptor is important in trafficking of CD8 T cells to 426 nonlymphoid tissues including the brain (43). We evaluated CXCR3 expression levels on R7-427 I and R7-III CD8 T cells during Toxoplasma infection. CXCR3 expression on R7-III CD8 T 428 cells was significantly lower at day 10 p.i. both in the brain and in the spleen (Fig. 4E). This 429 suggests that R7-III cells may have a lower ability to travel to the brain than R7-I thus 430 amounting a difference in their presence later in infection. To better understand the observed 431 decrease in the brain population of R7-III CD8 T cells we set up bone marrow (BM) chimeras. 432 We used the transplant conditioning drug busulfan to induce myeloablation and create a niche 433 for the R7-I or R7-III bone marrow (44).

435
In the brain and spleen, R7-I and R7-III BM chimeras have similar percentages of CD8 T cells 436 originating from the donor's BM at both week 3 and 5 p.i. (Fig. 4E). These results show that 437 constant replenishment from the periphery is necessary for the persistence of donor cells is the 438 brain at 5 weeks p.i. When R7-III CD8 T cells disappear from the periphery, they also disappear 439 from the brain. By creating chimeras, we showed that when we keep a constant R7-III 440 population in the periphery these cells also persist in the brain. 441 442

R7-I CD8 T cells of TRM (CD103 + ) phenotype are present in higher percentages in the 443 brain than R7-III CD8 T cells 444
Previous studies have shown that tissue resident memory T cells found in the brain can survive 445 without replenishment from the CD8 T cells circulating in the blood (15). Resident memory T 446 cells were observed in higher percentage in R7-I than R7-III CD8 T cell populations 3 weeks 447 p.i. in the brain (Fig. 5A). Bone marrow chimeras 3 weeks p.i. also exhibited lower percentages 448 of R7-III TRM cells in the brain (Fig. 5B) indicating that the difference to produce less TRM is 449 intrinsic to the R7-III clone independent of replenishment from periphery. 450

452
Affinity of TCR-MHC interaction influences the fate of the activated T cell (2). Immediate 453 effects of strong or weak interactions on activation and expansion of the T cells have been 454 studied broadly (2,45,46). However, it is unclear how affinity influences memory CD8 T cell 455 formation. Herein we studied three different clones of CD8 T cells (R7-I, II and III) during 456 Toxoplasma infection in BALB/c mice. These three CD8 T cell clones harbour TCRs specific 457 for the same peptide of the Toxoplasma protein ROP7, but differ in their sequence and affinity 458 for that peptide presented in MHC class I (24,25). The hierarchy of affinity and functional in 459 vitro responsiveness to ROP7-MHCI of the clones was R7-I>R7-III>R7-II (25). The lowest 460 affinity clone R7-II failed to traffic to the brain during the chronic phase of infection even 461 though we could show acute phase proliferation. R7-I outperformed R7-III in persistence in 462 lymphoid organs and the brain in chronic infection. Additionally, R7-I was able to form more 463 resident memory T cells in the brain than R7-III.

465
In order to test the affinity of the three R7 CD8 T cell clones in a more physiological setting, 466 in addition to our previously published 3D affinity measurements, we performed 2D affinity 467 measurements. Interestingly, we demonstrated little affinity difference between R7-I and R7-468 III. However, R7-I had a higher adhesion frequency than R7-III, possibly explaining the 469 functional differences we observed between these clones in the chronic phase of infection.

471
R7-II, the clone with the lowest affinity for ROP7 peptide, was not found in the chronic state 472 of Toxoplasma infection. Small number of the cells of that clone got activated and proliferated 473 in the acute phase of infection, but after the contraction phase that clone could not be found 474 neither in any analysed lymphoid organs nor in the brain. As one possible hypothesis, we 475 demonstrated by RNAseq and subsequent qPCR that R7-II did not cross the affinity threshold 476 required to downregulate Klf2 levels and consequently set the cells for homing to infected 477 tissues. Lack of homing and retention in the LN as a mechanism responsible for lack of R7-II 478 in brain was also supported by increased CTLA-4 expression. CTLA-4 expression on T cells 479 is known to be responsible for cells being stuck in the LN after antigen encounter and mark 480 anergic cells (47,48). It is considered as one of T cell-intrinsic function of CTLA-4 to control 481 self-reactive T cell motility in tissues (49). Signals from week TCR-MHC interaction may 482 prevent full activation of the T cell, but still enable it to receive partial signals (2). Little is 483 known about transcriptional regulation of Ctla-4 (50). It has not been investigated if Klf2 can 484 directly or indirectly regulate Ctla-4 expression. 485 486 R7-I and R7-III clones were of higher affinity than the R7-II clone and both were found in the 487 brain during the chronic phase of infection. The initially quite similar clones performed 488 differently in the later phase of infection. While both clones were able to form SLEC and 489 MPEG memory cells, the R7-III clone did not persist in the periphery and brain in the later 490 phase of chronic infection. Additionally, in the brain, more R7-I than R7-III cells showed a 491 phenotype of resident memory cells (CD69 + CD103 + ). We postulate that these differences can 492 be attributed to the increased adhesion frequency we observed in 2D measurements, as well as 493 the increased TCR-MHC binding affinity exhibited in 3D measurements (25). We were not 494 able to exactly pinpoint the reason for the disappearance of R7-III cells during the chronic 495 phase of infection. It could be attributed to slower replication, higher rate of death, or formation 496 of different types of cells that have different abilities to survive. 497 498 R7-III has been shown to be more proliferative than R7-I (more cell cycle terms in GO analysis) 499 (25). Also, we observed slightly higher percentages of R7-III than R7-I cells in the spleen at 3 500 weeks p.i. indicating that the initial slower replication rate of R7-III is not the reason for 501 differences between R7-I and R7-III observed in the later chronic phase of infection. SLEC 502 and MPEC percentages were not significantly different between two clones. Additionally, the 503 expression of the exhaustion marker PD-1 and the ability to produce IFN also did not differ 504 between R7-III and R7-I at 3 weeks p.i.

506
Creating bone marrow chimeras, we provided an artificial model where R7-specific CD8 T 507 cells are routed from the bone marrow via the thymus to the periphery also during infection. 508 This phenomenon has been described in persistent viral infections where host cells, but not 509 donor cells, can be resupplied through thymic output, and new, naive specific CD8 T cells are 510 being generated and subsequently primed during persistent infection (51). Newly generated T 511 cells preserve antiviral CD8 T cell populations during chronic infection (51). In the C67BL/6 512 mice model of Toxoplasma infection T cells are recruited from the periphery to the brain in the 513 chronic stage (52). 514 515 In our bone marrow chimeras model, even if cells get exhausted or/and stop dividing, new R7 516 specific cells are available to traffic to the brain. We concluded that constant replenishment 517 from the periphery is necessary to keep the population of R7-III in the brain in the later stages 518 of chronic infection. If the cells are available in the periphery they will traffic to the brain. 519 Thus, since the R7-I clone exhibits longer survival in the periphery and in the brain without 520 replenishment, we can conclude that cells of stronger affinity perform better in chronic 521 infections.

523
The proportion of the cells with TRM phenotype was different between R7-I and R7-III donor 524 population in the brain. R7-III cells persistently exhibited a lower percentage of TRM cells, no 525 matter if able to replenish from the periphery in bone marrow chimeras or not. Additionally, 526 R7-III cells exhibiting the phenotype of resident memory T cells (CD69 + CD103 + ) at 3 weeks 527 p.i. were absent from the brain at 5 weeks p.i.. This suggests that these were possibly not true 528 (classical) TRM cells that are long lasting and shown to persist for years after infection. 529 However, 'classical' TRM were defined in acute infection models with rechallenge (20,53). It 530 is thus conceivable that in chronic infection they may have a different characteristic. Constant 531 antigen stimulation during persistent infection may have a negative influence on TRM that are 532 considered to be antigen independent. Persistent antigen stimulation has been shown to lower 533 CD103 expression on TRM but it did not block their formation (54). It is possible that these 534 cells become exhausted when constantly stimulated and thus only newly formed CD103 535 positive cells contribute to the cells observed in the brain (23). It is possible that the strength 536 of the antigen stimulation influences this process and lower affinity leads to lower number of 537 R7-III cells expressing CD103, eventually leading to the elimination of these cells. Indeed, it 538 has been shown that TRM in the brain exhibit about 20-fold higher affinity as compared to 539 splenic memory cells (32). As in our model R7-I and R7-III CD8 T cell are different only in 540 their TCR receptor, we propose that observed differences are due to the affinity. One caveat is 541 that this difference may not only be derived from the affinity of the interaction with the cognate 542 ROP7 peptide, but different TCRs may already shape the fitness of the cells differently in the 543 thymus (25,55). Indeed, we previously showed that R7-I and R7-III cell respond differently to 544 CD3/CD28 stimulation with R7-III is being more proliferative. Thus, the self-reactivity of T 545 cells may also play a role in the later stages of an infection in shaping the memory CD8 T cell 546 phenotype.

548
Effector T cells in chronic infection are constantly exposed to antigen leading to exhaustion 549 (56,57). In the Toxoplasma infection model of C57BL/6 mice, it has been shown that chronic 550 infection leads to exhaustion of CD8 T cells in the brain (41,58

586
The authors declare no conflict of interests. 587 588  R7-III samples upon activation, where only genes differentially expressed in R7-II vs R7-I and 852

Figure Legends
R7-II vs R7-III comparisons but similar expression between R7-I and R7-III were included. 853 For each gene in the table fold change (FC) and false discovery rate (FDR) is shown form R7-854 II vs R7-III comparison. 855