Off-target effects of Cre recombinase reveal limits of adoptive T-cell transfers and persistent proliferation of effector CD8 T-cells

Effector-memory T-cells (TEM) are assumed to be short-lived cells that poorly proliferate upon antigenic restimulation, thus depending on central-memory T-cells (TCM) to replenish their numbers during homeostasis, largely depending on adoptive transfer evidence. Here we analyzed T cells in their natural environment and observed robust long-term in vivo cycling within the TEM subset that was stronger than the one in the TCM subset. Murine Cytomegalovirus (MCMV) induces inflationary TEM responses that remain high during latency. We analyzed Ki67 expression during acute and latent MCMV infection and found Ki67hiBcl2lo TEM in latently infected mice, arguing for antigen-driven TEM proliferation. TEM acquired deuterium more rapidly than TCM in an in vivo labeling experiment, and were replenished more rapidly than TCM after memory depletion, suggesting that TEM cycle faster than TCM. We depleted selectively the proliferating T-cells by Cre-overinduction, which resulted in a selective loss of Ki67hiBCl2lo effector T-cells, and an increase in the death of TEM in the spleen, while it hardly affected the TCM subset, arguing for robust proliferation of TEM in the spleen. On the other hand, TEM homing to the spleen upon adoptive transfer was substantially poorer than TCM, explaining the previously reported expansions of TCM, but not TEM, upon transfer. In conclusion, our data suggest that memory inflation is maintained by proliferation of antigen-specific TEM, rather than by continued expansion and differentiation of TCM. Author Summary The naïve T cell population consists of T cells that have the potential to recognize millions of different pathogens. Upon infection, naïve T cells that recognize the pathogen expand, and differentiate into effector T cells that eliminate infected cells. Once the infection is contained, the T cell pool contracts and only a small population of central memory T cells remains that can expand quickly upon re-infection. Cytomegaloviruses cause persistent infections that are not cleared from the organism after the initial immune response. In infected individuals a pool of CMV-specific effector memory T cells dominates the immune system in a phenomenon called memory inflation. Previous research using the transfer of central memory or effector memory T cells from CMV-infected mice into mice with a matching infection, showed expansion of central memory T cells but not effector memory T cells. Here we show that effector memory T cells have a reduced capacity to home into lymphoid organs, where T cell activation takes place, compared to central memory T cells. Using methods that do not interfere with T cell differentiation and homing, we show that effector memory T cells are proliferating during the persistent phase of CMV infection, significantly contributing to the upkeep of the inflationary population.


Introduction
T-lymphocytes play a unique role in the control of intracellular pathogens. T-cell receptor (TCR) 55 recognition of antigenic epitopes presented on MHC molecules results in clonal expansion of activated 56 cells and long-term maintenance of antigen-specific memory cells. This results in natural selection of 57 oligoclonal memory T-cells recognizing previously encountered pathogens, which must be balanced 58 with the need to maintain a broad repertoire of naïve T-cells to allow the recognition of a broad swath 59 of potential infections. Therefore, the proliferation and renewal of the peripheral T-cell compartments 60 is marked by exquisite complexity (1).

61
The pool of peripheral naïve T-cells is mainly quiescent (2), but ongoing maturation of T-cells from 62 the thymus replenishes it and maintains TCR diversity. The homeostatic turnover of memory CD8 T-153 154 We considered it likely that differences in Bcl2 expression reflect antigen-driven proliferation 162 Consistent with previous reports (11), effector T-cells were more frequent in MCMV than in VACV 163 infection at 120 dpi and displayed a Bcl2 lo phenotype (data not shown). Ki67cells showed a 164 predominantly naïve phenotype in both groups (data not shown), and were not analyzed further. On 165 the other hand, both the Ki67 hi and the Ki67 int subset displayed a significantly higher frequency of TE 166 cells in MCMV-infected mice, and conversely a higher frequency of TM cells in the VACV infected group 167 (Fig. 2B). Therefore, we concluded that the significantly higher frequency of Ki67 hi and Ki67 int CD8 T-168 cells in MCMV-infected mice at day 120 was mainly due to the high frequencies of Bcl2 lo effector T-169 cells found in long-term MCMV infection. 171 We independently quantified CD8 T-cell proliferation in latently MCMV-infected mice by in 172 vivo stable isotope labeling, a method that is not toxic and does not interfere with cell dynamics. Mice 173 were given deuterated water (D 2 O) at 120 dpi over a course of 4 weeks. 174 We measured deuterium levels in the DNA of different CD8 T-cell subsets during the labeling period 175 and for additional 16 weeks thereupon. We sorted CD8 + splenocytes into CD44 + CD62L + (central- T-cells had a relatively low turnover rate of 0.8% per day, TCM cells an almost two-fold higher turnover 185 rate of 1.4% per day, but TEM cells displayed the highest turnover rate of 2.0% per day (see Table 1).

186
Although part of the label incorporation in the TEM population may have been obtained during T-cell 187 proliferation at the TCM stage after which these cells differentiated into TEM cells (25)

226
On the other hand, we observed no significant loss of Ki67 int cells, and a mild increase in the percentage 227 of Ki67cells (Fig. 3B). Since Tamoxifen treatment reduced specifically the CD8 T-cell subsets with Ki67 hi   228 expression, but had no effect on Ki67 int cells, our data suggested active cycling of Ki67 hi CD8 T-cells and 229 quiescence in most of the Ki67 int population (Fig. 3B). 231 Both the Ki67 staining and the deuterium labeling results were consistent with robust proliferation 232 of effector (CD127 -) or effector-memory (CD62L -CD127 +/-) T-cells. Tam treatment affected the relative 233 size of blood and splenic T-cell subsets, decreasing the size of the effector subset (CD127 -) and

Tamoxifen targets splenic effector T-cells for depletion
234 increasing relatively all the other ones (Suppl. Fig. 2A

253
The reduction of Ki67 hi subsets in blood and spleen indicated that effector CD8 T-cells may cycle in 254 either compartment, but did not allow us to define if cycling is restricted to one compartment over the 255 other. We reasoned that a more immediate assay to define the cycling activity would be to measure 256 cell death upon Tam treatment. Therefore, we stained the CD8 T-cells of latently infected R26 CreER T2 257 mice with antibodies against CD62L and CD44 and determined the fraction of dead cells via 7AAD. We  276 secondary lymphoid organs 277 Our deuterium labeling and depletion data argued for robust cycling of effector T-cells, which was 278 unexpected because studies using adoptive transfer suggested that inflationary effector CD8 T-cells 279 can only be replenished from a pool of central-memory T-cells (19, 20). We speculated that the 280 discrepancy might be due to differences in the experimental design. Since we recently observed an 281 association between the latent MCMV load in the spleen and the size of memory inflation (28) Fig. 3). In consequence, adoptive transfer skews experimental results towards the 290 CD62L + cell subset, which efficiently homes to the spleen.

291
The effector-memory compartment is restored more rapidly than the central-

memory after depletion of primed CD8 T-cells with anti-asialo GM1
293 While all previous experiments suggested ongoing proliferation of effector T-cells in latently 294 infected mice, they did not allow a head-to-head comparison of proliferation in the effector-memory 295 and the central-memory compartment. Adoptive transfer of cell subsets is a method of choice to 296 measure such activity, but it was not suitable due to the homing bias of the transferred subsets.
297 Therefore, to validate if TEM or TCM would replicate more rapidly, we developed an assay where most 298 of the primed CD8 T-cell compartment is depleted upon which we monitored the ability of T-cells to 299 repopulate this compartment. Asialo-GM1 antibodies are widely used for the depletion of NK cells (29, 300 30), but were shown to also target virus-specific CD8 T-cells (31). We used repeated weekly i. p. injections of asialo-GM1 antibodies starting at 60 dpi and proceeding for four consecutive weeks to 302 deplete the entire CD44 + CD62L + (TCM) and CD44 + CD62L -(TEM) compartment, but leave the naïve 303 subset (CD44 -CD62L + ) intact (Fig. 4B). Thereupon, we analyzed the fraction of total primed CD8 T-cells 304 at 4 days and 2 months after the last depletion. Antibody administration severely reduced the fraction 305 of primed cells in MCMV and VACV infection, but the cells rebounded at 2 months post depletion, and 306 this was more pronounced in the group infected with MCMV ( Fig. 4C). We next defined if the rebound 307 at 2 months was evenly distributed among TCM and TEM cells or more pronounced in any of the 308 subsets, and observed a prominent restoration of TEM cells in MCMV and VACV infection, but a very 309 weak rebound of TCM cells (Fig. 4D). To understand if this effect may be observed in long-term latency 310 and in other latent infections, mice were infected with MCMV, murine herpesvirus clone 68 (MHV-68), 311 or herpes simplex virus type 1 (HSV-1), allowed to establish latency, and treated with asialo-GM1 at 10 312 months post infection. As in the previous experiment, the depletion strongly reduced the number of 313 primed cells, but their numbers were substantially increased 2 months later. Even more interestingly, 314 the rebound of primed CD8 T-cells was almost exclusively due to the TEM fraction, and essentially no 315 rebound was seen among TCM cells (Fig. 4E). In conclusion, our data showed that in conditions of 316 competition, the TEM cells proliferate more rapidly than TCM cells upon in vivo depletion, and that 317 this is a feature that can be observed in other herpesviral infections as well. 324 Consistent with prior reports (32), M38-specific cells expanded from 4.6% of the CD8 T-cell pool at 325 7 dpi to 14% at 120 dpi (Fig. 5B). Very few M38-specific cells were low in Ki67 expression both at 7 and at 120 dpi. The percentage of Ki67 int T-cells increased from a mean of 26% at 7 dpi to 86% at 120 dpi, 327 while the percentage of Ki67 hi T-cells decreased from 68% on average at 7 dpi to 12% at 120 dpi (Fig.   328 5C). Similarly, M38-specific CD8 T-cells were mainly Ki67 hi Bcl2 lo (72%) at 7 dpi, whereas at 120 dpi most 329 T-cells were Ki67 int Bcl2 lo (83%). 12% retained the Ki67 hi Bcl2 lo phenotype (Fig. 5D). Thus, while the data 330 resembled the time-associated pattern that was observed in the total CD8 T-cell pool, a larger fraction 331 of M38-specific cells were Ki67 hi at 120 dpi, than the data observed in the total CD8 T-cell pool (median 332 of 6.2%, Fig. 1B). The immunodominant peptide (RALEYKNL) derived from the ie3 MCMV gene (33) 333 showed similar phenotypes (data not shown). Taken together, the results suggest that CD8 T-cells 334 recognizing immunodominant inflationary antigens may vigorously proliferate during MCMV latency. whereas VACV-infected mice showed essentially no recovery (Fig. 5G). Taken together, the results 357 confirmed that the proliferative potential of CD8 T-cells in persistent infections is likely due to antigen-358 driven proliferation that cannot be observed in infections that are cleared.

359
It is important to note that adoptive transfer of M38-or m139-specific CD8 T-cells sorted for the 360 CD62L + or CD62Lphenotype from spleens of latently infected Ly5.1 C57BL/6J mice into naïve Ly5.2 361 C57BL/6J mice also showed significantly more homing of CD62L + CD8 T-cells to the spleen compared 362 to CD62L -CD8 T-cells ( Supplementary Fig. 3). Hence, we avoided testing the proliferative capacity of 363 antigen-specific cells by adoptive transfer. in the untreated control group (Fig. 6A). The reduction of the ie3-specific CD8 T-cell fraction was even 377 more pronounced (Fig. 6B). Once the Tam-spiked food pellets were replaced with standard diet, the 378 analyzed CD8 subsets rebounded (Fig. 6A, B). Taken together, the data confirmed that Tam treatment 379 reduced the effector subset of T-cells in general and antigen-specific inflationary cells in particular, 380 arguing strongly for continuous cycling of these subsets. Tam administration in latently infected 381 C57BL/6 mice resulted in no loss of TEM or antigen-specific CD8 T-cells ( Supplementary Fig. 4A-D),

382
arguing that the effects were specific for Tam effects in R26 CreER T2 mice. In an independent 383 experiment, we administered Tamoxifen at 60 dpi to mice infected with the SSIEFARL-expressing 384 MCMV (see Fig. 5F), and we observed a reduction of T-cells specific for the inflationary SSIEFARL and 385 the M38 epitope ( Supplementary Fig. 5A, B). for two inflationary epitopes derived from the ie3 and the M38 epitope. As in the previous experiment,

392
Tam feeding decreased the relative fraction of antigen-specific and effector cells (data not shown), and 393 this was reflected in a decrease of absolute counts of antigen-specific CD8 T-cells (Fig. 6C), as well as 394 effector CD8 T cells (Fig. 6D). Importantly, the drop in absolute counts of TE cells was stronger than the 395 loss of the TEM subset (Fig. 6E), consistent with the high percentage of Ki67 hi cells (Fig. 2B)    Comparisons between two groups were performed using the Mann-Whitney U test (two-tailed).

582
Statistical analysis was performed using the GraphPad Prism program.

583
Mathematical modeling of deuterium labeling data 584 We used a mathematical model to deduce the dynamics of naïve and memory T-cells from the 585 deuterium labeling data. To control for changing levels of deuterium ( 2 H) in body water over the course 586 of the experiment, a simple label enrichment/decay curve was fitted to 2 H enrichment in plasma:   Tamoxifen by oral gavage for 5 consecutive days, followed by a three day break, and one more day of