The cellular microenvironment regulates CX3CR1 expression on CD8+ T cells and the maintenance of CX3CR1+ CD8+ T cells

Expression levels of the chemokine receptor CX3CR1 serve as high‐resolution marker delineating functionally distinct antigen‐experienced T‐cell states. The factors that influence CX3CR1 expression in T cells are, however, incompletely understood. Here, we show that in vitro priming of naïve CD8+ T cells failed to robustly induce CX3CR1, which highlights the shortcomings of in vitro priming settings in recapitulating in vivo T‐cell differentiation. Nevertheless, in vivo generated memory CD8+ T cells maintained CX3CR1 expression during culture. This allowed us to investigate whether T‐cell receptor ligation, cell death, and CX3CL1 binding influence CX3CR1 expression. T‐cell receptor stimulation led to downregulation of CX3CR1. Without stimulation, CX3CR1+ CD8+ T cells had a selective survival disadvantage, which was enhanced by factors released from necrotic but not apoptotic cells. Exposure to CX3CL1 did not rescue their survival and resulted in a dose‐dependent loss of CX3CR1 surface expression. At physiological concentrations of CX3CL1, CX3CR1 surface expression was only minimally reduced, which did not hamper the interpretability of T‐cell differentiation states delineated by CX3CR1. Our data further support the broad utility of CX3CR1 surface levels as T‐cell differentiation marker and identify factors that influence CX3CR1 expression and the maintenance of CX3CR1 expressing CD8+ T cells.


Introduction
Upon priming, CD8 + T cells differentiate into a heterogeneous population.Primed CD8 + T cells that retain stemness-associated transcription factors such as TCF1 remain in a minimally differentiated state, indicative of longevity and the ability to mount large recall responses, to form resident memory T cells, and to survey resting lymph nodes and splenic white pulp [1][2][3][4][5][6][7].The more differentiation-promoting transcription factors (e.g.TBET, ZEB2, ID2) primed CD8 + T cells express, the further they differentiate and the more molecules involved in granule-mediated cytotoxicity they express [8][9][10][11][12].The furthest differentiated CD8 + T cell populations are furthermore relatively short-lived, restricted to blood and splenic red pulp, and do not expand well upon antigen reencounter [2,13,14], but nevertheless contribute to pathogen clearance [15].Intermediately differentiated CD8 + T cells circulate between blood, lymphoid, and nonlymphoid tissues [13] and are the most potent producers of the effector cytokines IFNγ and TNF upon antigen recognition [12].
Given the link between T-cell differentiation state and function, T-cell differentiation markers serve as valuable proxy for T-cell function when functional assays are not feasible.While the classical T-cell differentiation markers are distinct for murine and human CD8 + T cells, we have recently identified the chemokine receptor CX3CR1 to be a cross-species T-cell differentiation marker whose fine-graded expression levels capture the T-cell differentiation state gradient in both murine and human T cells [12].CX3CR1 expression levels identify CD8 + T cell states with comparable functional properties in both species [12].
The factors that regulate CX3CR1 expression in T cells are, however, incompletely understood.Naïve T cells do not express CX3CR1 [13,16] and its induction requires T-cell receptor (TCR) triggering [13].Development of a CX3CR1 + T-cell population was furthermore shown to depend on T-cell sensing of the differentiation-promoting cytokine IL-12 [16], which in turn enhances the expression of TBET [8].In line with this, robust CX3CR1 expression on T cells depends on TBET [13] and ZEB2 [16].It is, however, unclear whether additional external factors than IL-12 influence CX3CR1 induction upon T-cell priming.Likewise, there is incomplete understanding of external factors that potentially influence CX3CR1 expression on memory CD8 + T cells.Consequently, it remains unclear if there are situations during which differentiation-unrelated factors influence CX3CR1 expression and thereby hamper the interpretability of CX3CR1 expression as T-cell differentiation marker.
CX3CR1, previously designated RBS11 or V28, is a chemokine receptor [17][18][19][20] and CX3CL1, also known as fractalkine or neurotactin [21,22] is its ligand in mice and humans.CCL26, also known as Eotaxin-3 was identified as an additional ligand for human CX3CR1 [23] but in mice, Ccl26 is a pseudogene [24].Chemokine receptors are generally internalized upon ligand binding [25,26], and accordingly, it has been suggested that CX3CL1 binding leads to endocytosis of CX3CR1 on CD8 + T cells [27,28].However, the current data leave room for alternative explanations.Consequently, it has not yet been established unambigu-ously whether CD8 + T cells lose CX3CR1 surface expression upon exposure to CX3CL1, and if so, whether this affects the interpretability of the CX3CR1 gradient as differentiation marker.
In this study, we investigated the influence of a range of different factors on CX3CR1 expression by CD8 + T cells.We report that in contrast to in vivo priming, in vitro priming of naïve CD8 + T cells failed to robustly induce CX3CR1.Nevertheless, in vivo generated memory CD8 + T cells maintained their CX3CR1 expression levels during culture.TCR stimulation led to downregulation of CX3CR1 on memory CD8 + T cells.In the absence of TCR stimulation, CX3CR1 hi CD8 + T cells had a selective survival disadvantage, which was enhanced by factors released from necrotic but not apoptotic cells.CX3CL1 exposure did not rescue their survival and resulted in a dose-dependent loss of CX3CR1 surface expression.However, in vivo exposure to physiological doses of CX3CL1 reduced CX3CR1 surface expression only minimally and did not hamper the interpretability of CD8 + T-cell differentiation states delineated by CX3CR1.Our data further support the utility of CX3CR1 levels as T-cell differentiation marker for in vivo primed CD8 + T cells and identify factors that influence CX3CR1 expression and the maintenance of CX3CR1 expressing CD8 + T cells.
At day 5, all CD3/CD28 dynabeads-exposed CD8 + T cells had divided (Fig. 1A, Fig. S1B,C), which confirmed their activation.The divided cells had furthermore upregulated the cytolytic molecules granzyme A (GranA) and B (GranB) (Fig. 1B).The expression patterns of cell surface T-cell differentiation markers confirmed that dividing CD8 + T cells underwent differentiation (Fig. 1C).Expression of CCR7 and CD45RA gradually decreased, while CD45RO increased with divisions (Fig. 1C), which indicates ongoing CD8 + T cell differentiation in our primed cultures [60,61].The chemokine receptor CXCR3 increased with divisions in all donors except one.Expression of the costimulatory molecule CD27 increased over the first few cell divisions, and CD62L was transiently reduced in early division cells, likely due to T-cell activation-induced cleavage of CD62L [62].
Despite the efficient CD8 + T-cell activation and ongoing differentiation, we observed only minimal induction of CX3CR1 in the later divisions at day 5 (Fig 1D).While CX3CR1 expression was slightly higher after 8 days (Fig. 1E), the expression did not reach the high levels as on in vivo primed CD8 + T cells from the same donor (Fig. 1F).This lack of robust CX3CR1 induction on in vitro primed CD8 + T cells was also observed for murine cells (data not shown).Our data indicated that even when CD8 + T cell were in vitro primed through CD3 and CD28 stimulation in the presence of pro-inflammatory cytokines and additional costimulation, the resulting in vitro CD8 + T-cell differentiation only partially mimics in vivo CD8 + T-cell differentiation.

CX3CR1 expression is gradually lost from the memory CD8 + T-cell population during in vitro culture
Next, we aimed to study what factors influence CX3CR1 expression on memory CD8 + T cells.Having observed that in vitro primed CD8 + T cells only express minimal levels of CX3CR1 (Fig. 1), we generated in vivo a well-defined pool of murine virusspecific CX3CR1 expressing memory CD8 + T cells.We achieved this by transferring naïve OT-I CD45.1 + CD8 + T cells, which bear a transgenic TCR specific for the SIINFEKL peptide of ovalbumin into congenic (CD45.1 -) C57Bl/six mice (Fig. 2A).One day later, we infected these mice with recombinant ovalbumin-expressing lymphocytic choriomeningitis virus (LCMV-ova) of the Armstrong strain, which results in acute infection and an OT-I T-cell response with typical effector and memory phase kinetics [13].At >40 days after infection, we therefore refer to the OT-I T cells as "OT-I memory T cells" throughout this study.
In line with previous studies [12,13], LCMV-ova infection led to the development of CX3CR1 expressing OT-I memory T cells (gating in Fig. S2A), of which cells with the highest CX3CR1 expression levels had downregulated CD27 (Fig. 2B input).Note that while the OT-I T cells expressed a CX3CR1-GFP reporter (CX3CR1 +/gfp ) [63], we evaluated CX3CR1 expression -unless stated otherwise -by CX3CR1 antibody staining since we were interested in CX3CR1 cell surface expression.Having observed that in vitro primed CD8 + T cells only expressed minimal levels of CX3CR1 (Fig. 1D,E), we wondered whether in vitro culture impacted CX3CR1 expression in an in vivo generated CD8 + memory T-cell population.To this end, we evaluated CX3CR1 expression on OT-I memory T cells over time in culture.Over the course of 24 h in culture (37°C), we observed a gradual loss of CX3CR1 in the OT-I memory population (Fig. 2B, Fig. S2B).Additionally, the proportion of CX3CR1 + OT-I memory cells decreased significantly in a time-dependent manner during culture at 37°C (Fig. 2B, Fig. S2B).In contrast, when the OT-I memory cells were kept at 4°C, CX3CR1 expression remained stable (Fig. 2B), which suggested that the loss of CX3CR1 from the cultured OT-I memory population was caused by an active biological process.Notably, this process did not affect all cell surface molecules, since CD8 and CD45.1 showed no consistent time-dependent decrease in expression with culture (Fig. 2C).The fraction of CX3CR1 expressing CD8 + T cells decreased also among cultured human CD8 + T cells, albeit with slower kinetics (Fig. S2C,D).
In summary, we observed that CX3CR1 was gradually lost from the CD8 + memory T-cell population during in vitro culture.Our in vitro culture setup thus provided a platform to investigate the factors responsible for this loss of CX3CR1.

CX3CR1 expression is stable over time on individual CD8 + memory T cells
The reduction of CX3CR1 expression from the CD8 + memory T cell population during in vitro culture can be explained by two distinct mechanisms.First, individual CX3CR1 + cells lose their CX3CR1 expression through time.Alternatively, the expression of CX3CR1 on individual cells is stable, but the relative abundance of CX3CR1 + and CX3CR1 -cells changes over time, for example, due to the selective death of CX3CR1 + cells.
We first investigated whether individual human or murine CX3CR1 + memory CD8 + T cells lose their CX3CR1 expression during culture.If this was the case, this could be a sign of dedifferentiation of the T cells in vitro.Alternatively, if CD8 + T cells were to lose CX3CR1 expression in vitro without changing their differentiation state, this would hamper the usability of CX3CR1 as differentiation marker for CD8 + T cells in vitro.To determine whether CX3CR1 + CD8 + T cells lose CX3CR1 expression in vitro, we sorted antigen-experienced human CD8 + T cells into four populations by their CX3CR1 level, labeled them with CTV, and evaluated their CX3CR1 expression after 8 days in culture in the presence of IL-7 and IL-12 (Fig. 2D).Depending on the donor, some CD8 + T cells had divided within this timeframe (Fig. S3A), potentially due to subclinical immune responses in the donor, and/or the proinflammatory environment provided by IL-12.To avoid potential confounding effects related to differences in proliferation of the sorted populations (Fig. S3A), we restricted CX3CR1 analysis to undivided and thus resting memory T cells (day 8 div 0).Resting memory CD8 + T cells sorted as CX3CR1 - remained CX3CR1 -, and CX3CR1 hi remained CX3CR1 hi throughout culture (Fig. 2D, S3B).CX3CR1 expression on the populations sorted as CX3CR1 int1 or CX3CR1 int2 was variable between donors (Fig. S3B), potentially due to lower cell numbers.Overall, however, CX3CR1 levels remained stable on undivided CD8 + memory T cells over the 8-day culture (Fig. 2D).
Next, we studied whether CX3CR1 expression was maintained on murine memory T cells.We sorted in vivo generated OT-I memory CD8 + T cells into three populations by their CX3CR1-GFP levels and cultured the sorted populations for 3 days in the absence of exogenous cytokines to approximate an immunological steady-state (Fig. 2E).While we used CX3CR1-GFP for sorting, in all other settings, CX3CR1 expression was evaluated by CX3CR1 antibody to determine CX3CR1 cell surface expres-sion.We consistently observed a decrease in CX3CR1 levels from day 0 to day 3 in all sorted populations, including the population sorted as CX3CR1 -(Fig.2E bottom, gating in S3C), which serves as an internal control for the stability of the CX3CR1 signal over time.Given that our internal technical control indicated that the CX3CR1 levels could not be compared directly between days 0 and 3, we instead evaluated CX3CR1 expression from the sorted CX3CR1 + populations relative to the sorted CX3CR1 - population per day per experiment (Fig. 2F).This demonstrated that CX3CR1 expression by the sorted CX3CR1 + murine memory T cells remained relatively stable in culture (Fig. 2F).The sorted populations furthermore maintained their CD27 expression status throughout the culture (Fig. 2G), which further supported that the sorted cells maintained their differentiation state during culture.
Overall, our data demonstrate that at least high levels of CX3CR1 are stably maintained on human and murine memory CD8 + T cells during 3-8 days of culture.This is consistent with in vivo data demonstrating that murine CX3CR1 -memory CD8 + T cells remain CX3CR1 -, and CX3CR1 hi memory CD8 + T cells remain CX3CR1 hi for at least two months at steady-state [13].It is possible that some CX3CR1 int memory cells lost their CX3CR1 expression during the 8-day culture, which is consistent with the reported loss of CX3CR1 by a fraction of CX3CR1 int memory cells after one month in vivo [13].However, this would only account for a minimal reduction of CX3CR1 expression in our unsorted memory T-cell cultures (Fig. 2B) given the numerical underrepresentation of CX3CR1 int memory cells in the OT-I memory T-cell population.Consequently, our data from the sorting experiments indicated that the reduction in CX3CR1 expression we observed by 24 h of culture in the CD8 + memory T-cell population (Fig. 2B) resulted predominantly from a change in the relative abundance of CX3CR1 + and CX3CR1 -CD8 + T cells.

CX3CR1 + memory CD8 + T cells have a selective survival disadvantage that is enhanced by factors released from necrotic but not apoptotic cells
Having established that our in vitro culture of memory CD8 + T cells recapitulated the stability of CX3CR1 expression on individual CD8 + T cells (Fig. 2D-G), we used our culture setup to gain mechanistic insights into the regulation of CX3CR1 on memory CD8 + T cells, and the maintenance of CX3CR1 + T cells over time.First, we asked whether the change in the relative abundance of CX3CR1 + and CX3CR1 -memory CD8 + T cell populations over time resulted from preferential death of CX3CR1 + cells.We quantified the fraction of dead cells, using LIVE/DEAD dye, which stains free amines that are present at low levels on the cell surface, but highly abundant in the intracellular compartment that becomes accessible in dying cells.The frequency of dead OT-I memory T cells increased over time in culture to the extent that at 24 h, 15% of OT-I memory T cells were dead (Fig. 3A and Fig. S4A).In line with this, absolute numbers of viable OT-I memory T cells were significantly decreased after 24 h in culture (Fig. 3B).Given that the increase in CD8 + T-cell death (Fig. 3A) temporally coincided with the disappearance of CX3CR1 + memory T cells (Fig. 2B), we examined the frequency of dead cells among CX3CR1 + and CX3CR1 -OT-I memory T cells.While cell death increased over time within both CX3CR1 + and CX3CR1 - OT-I memory T-cell populations (Fig. 3C), the fraction of dead cells was higher among CX3CR1 + compared with CX3CR1 -memory cells at 12 and 24 h (Fig. 3D).Likewise, dead cells were more frequent among human CX3CR1 + than CX3CR1 -CD8 + T cells late during culture (Fig. S2E).This indicated that CX3CR1 + CD8 + T cells were more susceptible to cell death than CX3CR1 -CD8 + T cells during steady-state.
One to two months after LCMV infection, the majority of CX3CR1 + memory CD8 + T cells belong to the highly differentiated CX3CR1 hi CD27 -population (Fig. 2B, [13]).CX3CR1 int and CX3CR1 -populations, instead, both express CD27 (Fig. 2, [13]).Consistent with preferential death of the most highly differentiated memory T cells, the fraction of CD27 -memory OT-I T cells decreased over time in culture (Fig. 3E, S4B), while cell death increased (Fig. 3A).Of note, in three out of four experiments, we noticed the appearance of OT-I memory T cells with extremely bright GFP levels (CX3CR1-GFP super ) (Fig. S4C,D) that we do not usually observe (Fig. 2E, [13]).This population was most prominent at 24 h (Fig. S4E), when cell death was maximal (Fig. 3A,B).In addition, we occasionally observed this population directly ex vivo (not shown), and in those cases, we suspect its appearance is associated with suboptimal sample processing leading to more than usual cell death.
Taken together, our data indicate that highly differentiated CX3CR1 + /CX3CR1 hi memory CD8 + T cells are more susceptible to cell death during steady-state culture conditions than less differentiated CX3CR1 -memory T cells.
We were surprised at how rapidly the murine CX3CR1 + memory T-cell population was lost.It reduced from ˜75% to ˜30% of total OT-I memory T cells during 24 h of culture (Fig. S2B).Although CX3CR1 + memory CD8 + T cells also decline over time in vivo, they are still detectable >1 year after LCMV infection [13].Given that memory T cells are unlikely to encounter high concentrations of dying cells in vivo in the absence of infection or disease, while cell death is prevalent during in vitro culture of murine T cells, we wondered whether factors released by dying cells, e.g.damage-associated molecular patterns (DAMP) enhance the loss of CX3CR1 + memory T cells.Analysis of RNA sequencing data from the Immunological Genome Project (ImmGen) [64] revealed that several selected pattern recognition receptors that are known to recognize DAMP [65] are expressed by CD8 + T cells (Fig. 3F).To directly investigate whether factors released from dead cells contributed to the preferential loss of the CX3CR1 + memory T-cell population, we cultured OT-I memory T cells in fresh medium supplemented with supernatant from murine blood-derived leukocytes that were killed by harsh vortexing followed by ∼12 h rest in PBS at room temperature.After 2 h of culture in the presence of supernatant from dead leukocytes, CX3CR1 + memory T cells were modestly but significantly reduced as compared with culture in medium alone (Fig. 3G and Fig. S4F).Consistent with Fig. 2, between 2 and 12 h of culture, the percentage of CX3CR1 + CD8 + T cells declined even in the medium alone.After 12 h, the supernatant from dead leukocytes had only a modest additive effect on the reduction of CX3CR1 + T cells.Since cell death is substantially higher at 12 h than 2 h (Fig. 3A), it is conceivable that the concentration of factors released from dying cells is already so high at 12 h that the addition of the supernatant from dying cells had no further additive effect.Our data suggest that exposure to factors released from dead cells can accelerate the reduction of the CX3CR1 + memory CD8 + T-cell population.
DAMP release from dying cells depends on the type of cell death.Apoptosis is regarded as a non-inflammatory cell death mode, while necrosis leads to extensive DAMP release and is highly inflammatory [65].To investigate whether the disappearance of the CX3CR1 + memory CD8 + T-cell population was influenced by the type of cell death in the surroundings, we cultured OT-I memory T cells in the presence of supernatant from either apoptotic or necrotic cells.To obtain supernatant from apoptotic cells, we cultured mouse embryonic fibroblasts (MEF) in DMSOdissolved etoposide to induce apoptosis and thereafter collected the supernatant.Supernatant from necrotic cells was obtained by subjecting MEF to six freeze-thaw cycles.We also collected supernatant from MEFs before inducing necrosis (alive cell sup) to rule out potential effects of factors released from alive MEFs on  Antigen-experienced blood CD8 + T cells were sorted by CX3CR1 level from n = 5 healthy human donors, labeled with CTV, and either stimulated or not with CD3 and CD28 antibodies for 3 days.All populations were cultured with IL-7 and IL-12p70 for 8 days.(B) CX3CR1 expression level before (day 0 division 0) and after cell culture (day 8).Day 0 and day 8 division 0 groups without CD3 and CD28 stimulation are also shown in Fig. 3, and serve as controls here.Flow cytometry plots from representative donors.Red lines: average.Symbols: mean fluorescent intensity of CX3CR1 of individual donors.*p < 0.05, **p < 0.01 (one-way ANOVA+Tukey's).
CX3CR1 expression.Interestingly, exposure to supernatant from apoptotic cells did not influence CX3CR1 expression of CD8 + memory T cells (Fig. 3H).In contrast, culture in medium supplemented with supernatant from necrotic cells resulted in a significant acceleration of the loss of CX3CR1 + memory T cells (Fig. 3I).In agreement with our data in Fig. 3G, the supernatant from necrotic cells only affected the reduction of CX3CR1 + T cells early during the culture (2 h), when cell death in the culture itself is still minimal (Fig. 3A).Expression of CD8 and CD45.1 remained unaffected by the different supernatants (Fig. S4G,H), which confirmed that the reduction in CX3CR1 expressing cells was not the result of a technical staining or fluorescence detection issue.
Together, our findings suggest that factors released from necrotic but not apoptotic cells contribute to the preferential death of CX3CR1 + memory CD8 + T cells.

TCR stimulation of CX3CR1 + memory T cells leads to, at least temporary, loss of CX3CR1 surface expression
To gain further mechanistic insights into the factors that influence CX3CR1 expression on memory CD8 + T cells, we investigated whether TCR triggering affects surface CX3CR1 expression.We previously observed a transient drop in CX3CR1-GFP reporter expression by memory CD8 + T cells in vivo after secondary infection [13].However, since it was possible that GFP was transiently diluted in the rapidly dividing cells [13], that CX3CR1 + T cells had preferentially migrated to other body sites, or that GFP not fully recapitulates CX3CR1 surface expression, e.g.due to different protein half-lives, it had remained unclear whether T-cell activation affects CX3CR1 surface expression on memory CD8 + T cells.To directly test whether TCR stimulation influences CX3CR1 surface expression on in vivo primed CX3CR1 + CD8 + T cells, we sorted antigen-experienced CD8 + T cells from healthy human donors into four populations by their CX3CR1 levels, and stimulated these for 3 days with plate-bound CD3 and CD28 antibodies or left them unstimulated (Fig. 4A).All sorted populations were labeled with CTV before the start of the culture and were kept in IL-7 and IL-12 throughout the 8-day culture (Fig. 4A).
Akin to the TCR stimulation of naïve CD8 + T cells (Fig. 1), we did not detect robust CX3CR1 induction on antigen-experienced CX3CR1 neg CD8 + T cells upon stimulation through CD3 and CD28 (Fig. 4B and Fig. S5).The CX3CR1 + populations generally kept their CX3CR1 expression throughout culture in the absence of CD3 and CD28 stimulation (Figs.2D-F and 4B).After stimulation, however, the CX3CR1 + populations rapidly -within 2-3 divisions -lost their CX3CR1 surface expression (Fig. 4B and Fig. S5).In donors 1,3 and 5, it seems that CX3CR1 re-appears on the surface of some cells after 6+ divisions (Fig. 4 and Fig. S5), which is consistent with the CX3CR1 loss being transient, as observed in vivo with the CX3CR1-GFP reporter in CD8 + T cells [13].Since memory CD8 + T cells express low levels of Cx3cl1 mRNA according to ImmGen, we cannot formally exclude at this point a contribution of possible T-cell-derived CX3CL1 to the TCR-stimulationinduced CX3CR1 reduction on CD8 + T cells.Our data in Fig. 5, however, suggest that low levels of CX3CL1 do not lead to loss of CX3CR1 expression on CD8 + T cells.
Taken together, our data demonstrate that TCR stimulation in the presence of CD28 costimulation and the cytokines IL-7 and IL-12 leads to, at least temporary, loss of CX3CR1 surface expression on antigen-experienced CD8 + T cells.

Exposure to CX3CL1 reduces CX3CR1 surface expression on CD8 + memory T cells in a concentration-dependent manner
Chemokine receptors generally undergo ligand-dependent internalization [25,26], which may lead to a temporary reduction in chemokine receptor cell surface expression.In line with this, human blood CD8 + T cells lost CX3CR1 surface staining with increasing duration of CX3CL1 exposure during up to 24 h culture, which was interpreted as evidence for ligand-induced loss of CX3CR1 on memory CD8 + T cells [28].However, our finding that even in the absence of exogenous CX3CL1, CX3CR1 expression is reduced significantly in CD8 + memory T-cell populations after 24 h of culture (Fig. 2B) shed new light on the earlier observations.Our findings raised the question of to what degree the previously observed loss of CX3CR1 surface expression was ligand-induced or culture time-dependent.
We therefore investigated the effect of CX3CL1 on CX3CR1 cell surface expression by CD8 + memory T cells, while controlling for the culture time.We cultured OT-I memory T cells in the absence or presence of different concentrations of soluble CX3CL1 and analyzed CX3CR1 surface expression after 15 min, 2 h, 12 h, or 24 h.Without and with the addition of CX3CL1 to the cultures, CX3CR1 surface expression and the percentage of CX3CR1 + cells decreased over time in the OT-I memory T-cell population (Fig. 5A,B and Fig. S6A).This time-dependent decrease in CX3CR1 surface staining occurred irrespective of CX3CL1 concentration (Fig. 5A,B).CX3CL1 exposure, however, enhanced and accelerated the loss of CX3CR1 from the OT-I memory population in a concentration-dependent manner (Fig. 5A,C).Expression of CD8 and CD45.1 remained relatively unaffected by CX3CL1 exposure (Fig. S6C), consistent with CX3CL1 specifically reducing CX3CR1 surface expression.
To ascertain that the reduced CX3CR1 staining after CX3CL1 exposure did not result from CX3CL1 interfering with CX3CR1 antibody binding, we stained OT-I memory T cells that were cultured for 24 h in the absence of exogenous CX3CL1 with CX3CR1 antibody in two competitive settings.As positive control for binding competition, cells were stained with the standard antibody cocktail containing CX3CR1-BV421 spiked with a high concentration of the same CX3CR1 antibody clone conjugated to a different fluorophore (CX3CR1-PE/Dazzle).Addition of this competitive antibody indeed led to reduced staining with CX3CR1-BV421 (Fig. S6B).In contrast, staining with the standard antibody cocktail in the presence of 4096 ng/ml CX3CL1 did not reduce the CX3CR1-BV421 signal.This demonstrated that the reduced CX3CR1 antibody staining we observed after culture with CX3CL1 resulted indeed from the loss of CX3CR1 surface expression from the CD8 + memory T-cell population.
As shown in Figs. 3 and 4, the time-dependent reduction of CX3CR1 expression in culture without exogenous CX3CL1 resulted from preferential loss/death of the CX3CR1 + memory Tcell population.Since most of the CX3CR1 + memory T cells at the sampled timepoints after infection are CD27 -CX3CR1 hi (Figs.1-5, [13]), this led to a decline in the fraction of highly differentiated CD27 -memory T cells after 24 h (Fig. 3E) that can clearly be appreciated also in the flow cytometry plots in Fig. 5A.In contrast, the CX3CL1-dependent reduction in CX3CR1 surface expression was not accompanied by an obvious change in the fraction of CD27 -memory T cells (Fig. S6D).Instead, the CX3CR1 expression levels on the highly differentiated CD27 -memory T-cell population gradually shifted from high expression to no expression (Fig. 5A, flow cytometry plots).This suggested that exposure to soluble CX3CL1 led to a concentration-dependent reduction of CX3CR1 surface expression on individual memory T cells, consistent with receptor internalization.
While our in vitro data demonstrated that CX3CR1 surface expression can be regulated by soluble ligand exposure, it remained unclear to what degree the exogenously provided CX3CL1 reflected the in vivo availability of CX3CL1 to CD8 + T cells.To explore the effect of in vivo exposure to physiological levels of CX3CL1 on CX3CR1 surface expression by memory CD8 + T cells, we evaluated CX3CR1 expression by OT-I memory T cells directly ex vivo.OT-I memory T cells were isolated from different tissues and at different timepoints, including the effector phase of the T cell response (days 5-7), when endogenous CX3CL1 levels are expected to be highest, given that CX3CL1 expression is induced by inflammation [22,[34][35][36][37][38][39][40].Fig. 5A revealed that the appearance of a CD27 -CX3CR1 - population among pathogen-specific CD8 + T cells is indicative of highly differentiated CD8 + T cells having lost their CX3CR1 expression due to ligand-exposure.We therefore assessed whether pathogen-specific CD8 + T cells comprised a clear population of CD27 -CX3CR1 -cells over the course of an in vivo infection.When evaluated directly ex vivo, CD27 -CX3CR1 -cells comprised on average no more than 1% of blood OT-I T cells after LCMV-ova infection, not even during the effector phase (Fig. S6E).At the peak of the effector phase, splenic, lymph node, and lung-derived OT-I T cells in perfused mice contained 2-6% CD27 -CX3CR1 - cells after infection with ovalbumin recombinant Listeria monocytogenes (Lm-ova) (Fig. S6F).Although the frequency of the CD27 -CX3CR1 -population was slightly higher in tissues than in blood, it constituted a minority of cells.Furthermore, the CD27 - pathogen-specific (OT-I) T cells in the tissues were a well-defined population of cells with predominantly high CX3CR1 surface levels (Fig. S6F).Consequently, the level of CX3CR1 internalization that may occur in response to CX3CL1 encounter in vivo is likely so minimal or occurs in so few cells that it does not hamper clas-  Together, our data indicate that while ligand-induced CX3CR1 loss from the CD8 + T cell surface can occur, the endogenous CX3CL1 levels that CD8 + T cells are exposed to in vivo do not influence CX3CR1 surface expression to such a degree that it would hamper the assessment of CD8 + T cell differentiation states when using CX3CR1 levels as differentiation marker.

CX3CL1 does not rescue the time-dependent death of cultured memory CD8 + T cells
CX3CL1 was shown to rescue human blood monocytes and murine CD4 + Th1 and Th2 cells in culture from serum starvationand chemical-induced cell death [49,50], and to promote the survival of murine CD4 + Th1 and Th2 cells in the airways during allergen-induced airway inflammation, but not during homeostasis [49], nor in lymph nodes after epicutaneous antigen delivery [66].Whether CX3CL1 sensing provides a survival advantage to CD8 + T cells has not been investigated.
Having shown that CX3CR1 + memory CD8 + T cells preferentially die in culture (Fig. 3A-D), we investigated whether exposure to CX3CL1 rescued the time-dependent cell death of cultured memory CD8 + T cells.We seeded equal numbers of OT-I memory T cells per well, cultured them in the absence or presence of different concentrations of soluble exogenous CX3CL1, and quantified the absolute number of living OT-I memory T cells after 15 min, 2 h, 12 h, and 24 h.The addition of CX3CL1 did not significantly influence the number of viable OT-I memory T cells in culture (Fig. 5D).This indicated that CX3CL1 does not act as a survival factor for memory CD8 + T cells during time-in-culture dependent cell death.

Discussion
Our data highlight the importance of considering divergent lifespans of distinct CD8 + T-cell populations when interpreting CD8 + T-cell differentiation in vitro.We observed preferential death of the more extensively differentiated memory CD8 + T cells resulting in a significant decrease in the representation of CX3CR1 + T cells in cultures without TCR stimulation.In vivo, highly differentiated T cells disappear more gradually [13,67].It is possible that prosurvival cytokines can prolong the maintenance of the highly differentiated population, allowing their study over prolonged periods of time in vitro.Exposure to soluble CX3CL1 did not alter the viability of memory CD8 + T cells in our setup.This suggests that CX3CL1 does not act as a general survival factor for CD8 + T cells, even though it promotes the survival of CD4 + T cells in some conditions [49,50,66].Given that monocyte survival was enhanced by membrane-anchored but not secreted CX3CL1 [29] it will be interesting to compare the effect of both CX3CL1 forms on CD8 + T cells in the future.
We furthermore revealed that extensively differentiated CX3CR1 hi CD8 + T cells, which are highly abundant after in vivo priming [12,13,17,68,69], were absent 8 days after in vitro priming.While it is possible that development of highly differentiated CD8 + T cells requires more time in vitro than after in vivo infection, our observation that these cells preferentially die in vitro provides an additional possible explanation.Another possibility is that in vitro priming through CD3 and CD28, even in the presence of inflammatory stimuli is insufficient to drive CD8 + T-cell differentiation toward the development of the most highly differentiated T cells.This is consistent with a recent report showing that CD8 + T-cell priming by CD3 and CD28 antibodies is weaker than by dendritic cells presenting high-affinity peptides, and leads to biologically distinct responses [70,71].It is also consistent with our observation that CD62L is only transiently downregulated upon priming through CD3 and CD28 stimulation, and only a few primed cells remain CD62L -through later divisions.Altogether, our data highlighted the limitations of standard in vitro priming setups in mimicking in vivo T-cell differentiation.
Our observation that the preferential disappearance of the CX3CR1 + memory CD8 + T-cell population was enhanced by exposure to supernatant from necrotic cells is consistent with an effect of DAMP on CD8 + T cells and may result from DAMP-sensing enhancing the preferential cell death of CX3CR1 + memory T cells.This could be relevant at sites of infection, where necrosis occurs, to limit further tissue damage by the highly differentiated and cytotoxic [12,13,17,68] CX3CR1 + CD8 + T cells.
We furthermore established that TCR stimulation leads to a reduction in CX3CR1 surface expression in early dividing memory CD8 + T cells.Our in vitro restimulation data do not provide clear evidence for the loss of CX3CR1 expression being temporary, but it is likely that the same mechanisms that hampered robust induction of CX3CR1 expression on in vitro primed CD8 + T cells acted on the in vitro restimulated memory T cells.Together with the previous in vivo data [13], our data from this study support the notion that TCR triggering induces temporary downregulation of CX3CR1 surface expression on memory CD8 + T cells.Consequently, CX3CR1 surface expression does not reliably report the differentiation state of memory CD8 + T cells early after TCRmediated restimulation.
Previous studies had suggested that CD8 + T cells undergo ligand-induced CX3CR1 internalization, but the experimental setups left room for alternative explanations, in particular in light of our observation that the fraction of CX3CR1 + CD8 + T cells decreases with time in culture.Here, we unambiguously demonstrated that memory CD8 + T cells lose CX3CR1 surface expression upon exposure to the CX3CR1 ligand CX3CL1 in a concentrationdependent manner.While we did not formally address whether this loss of CX3CR1 surface expression is due to receptor internalization, the kinetics of the ligand-induced CX3CR1 loss by the CD27 -memory population suggest that this is likely the case.Our data established that memory CD8 + T cells can lose CX3CR1 surface expression upon exposure to soluble CX3CL1.However, when we examined CX3CR1 surface expression on LCMV-and Listeria monocytogenes-specific CD8 + T cells in the blood, spleen, lymph nodes, and lungs during acute infection, we found that the endogenous CX3CL1 levels that CD8 + T cells are exposed to in vivo do not reduce CX3CR1 surface expression to a degree that hampers the assessment of CD8 + T-cell differentiation states when using CX3CR1 levels as differentiation marker.
Altogether, this study supports the broad usability of CX3CR1 levels as T cell differentiation marker, except early after TCR triggering, and identifies factors that influence CX3CR1 expression, and the maintenance of CX3CR1 expressing CD8 + T cells.

Data limitations and perspectives
We showed that factors derived from necrotic cells contribute to the time-dependent reduction of CX3CR1 + CD8 + T cells in vitro.It remains to be explored whether this mechanism contributes to preferential death of highly differentiated CX3CR1 + CD8 + T cells in vivo, where necrotic material is constantly cleared by phagocytes.Furthermore, it is likely that additional mechanisms play a role in regulating the survival of highly differentiated CX3CR1 + CD8 + T cells.One additional possibility that cannot be assessed in a simple in vitro experimental setup is that highly differentiated memory T cells require for their survival specific cellular interactions and microenvironments that are present in vivo but not in vitro.In this study, we focused on CX3CR1 and did not measure the expression of other chemokine receptors.We therefore have not tested whether expression of other chemokine receptors is comparably stable on individual T cells as CX3CR1.We have neither tested if CD8 + T cells expressing any other chemokine receptor will preferentially die in culture.The latter is, in our opinion, unlikely, given that some chemokine receptors, like CX3CR1, are expressed by highly differentiated cells, and other chemokine receptors, like CCR7, are expressed by minimally differentiated cells, and we report a preferential death of highly differentiated CX3CR1 + T cells.
For TCR stimulation of naïve T cells, Dynabeads Human T-Activator CD3/CD28 (Gibco) were added (1:5 bead-to-cell ratio) for 3 days and thereafter magnetically removed.The supplemented medium was refreshed every few days.For TCR stimulation of antigen-experienced CD8 + T cells, 5 μg/mL plate-bound CD3 (OKT3, BioXcell) and 5 μg/mL plate-bound CD28 (9.3, BioXcell) were used.After 3 days, the stimulation was stopped by washing and transferring the cells into fresh plates.The supplemented culture medium was refreshed every few days.

Dead cell supernatants
To obtain supernatant from dead leukocytes, ˜40 μL murine blood was collected and erythrocytes lysed (ACK).PBS-washed leukocytes were vortexed harshly in 1 mL PBS and kept at room temperature for ∼12 h.The next day, cells were centrifuged at 500 g and the supernatant was collected.
To obtain supernatant from apoptotic cells, MEF were treated after the medium change with 2.5 μM etoposide (Merck) dissolved in dimethylsulfoxide (DMSO, PanReacAppliChem).Cells were kept in culture for 24 h after which apoptosis was confirmed by microscopy (apoptotic bodies present).The supernatant was stored at −80°C.As control for the potential effect of DMSO on memory T cells, we collected supernatant from MEF after culturing them for 24 h with the same concentration of DMSO we used to dissolve etoposide.As control for potential effect of etoposide remnants on memory T cells, we added the same concentration of DMSO-dissolved etoposide as was used to induce apoptosis in MEF (etoposide ctrl).To collect supernatant from necrotic cells, MEF were cultured for 24 h following medium change.Thereafter, MEF in the medium were subjected to six cycles of freezing (dry ice, approximately 5 mins) and thawing (incubator, 37°C, approximately 10 mins) to induce necrosis.We also collected supernatant from MEF before inducing necrosis ("alive cell sup").MEF supernatants were centrifuged and stored at −80°C until usage.Added supernatants made up 25% of the OT-I culture volume.
Samples were acquired on an LSR Fortessa (BD-Biosciences) or Cytek Aurora 5L (Cytek Biosciences, Fig. S6E) and analyzed using FlowJo_v10 (BD-Biosciences).Murine samples were visualized in FlowJo.For human samples, compensated and transformed signal intensities were exported from FlowJo, and gating information was maintained for further analysis and visualization using Python.Samples with over time unstable fluorescence intensity ranges were either excluded or the unstable timeframe was excluded.If a population of interest contained <50 cells, it was excluded from the analysis.Absolute cell numbers were quantified using counting beads.To calculate absolute cell numbers, the number of total cells was divided by the frequency of recovered counting beads (CountBright Absolute_Counting_Beads, Ther-moFisher Scientific).

Statistical analysis
Statistical analysis was performed with GraphPad Prism 7 or 9. Paired two-tailed t-test or unpaired t-test was used when comparing one variable between two groups.When more than two groups were compared, one-way ANOVA with Tukey's multiple comparisons test was used.

Figure 2 .
Figure 2. CX3CR1 expression is gradually lost from the memory CD8 + T cell population during in vitro culture but is stable through time on individual CD8 + memory T cells.(A) Generation of murine memory CD8 + T cells.Splenocytes enriched for OT-I memory T cells (CD45.1 + ) were used as starting material in all in vitro mouse cell experiments in this study unless stated that sorted cells were cultured.(B, C) Cultures were carried out with murine splenocytes enriched for OT-I memory T cells, but only OT-I memory T cells were evaluated.CX3CR1 (B), CD8, and CD45.1 (C) expression on OT-I memory T cells.(B, Left) representative flow cytometry plots.Numbers: %CX3CR1 + cells.(B, C) Bar graphs: data from five experiments.Bars: average.Individual data points: individual wells.Symbol shapes: independent experiments.Each time point was included in two or five experiments.2-5 wells per timepoint per experiment.To account for variability between experiments, individual data points were normalized per experiment to the average of the 2 h 37°C group.*p < 0.05, **p < 0.01, ***p < 0.001 (one-way ANOVA+Tukey's).(D) Left: Experimental setup.Antigenexperienced (Ag-exp) blood CD8 + T cells of n = 5 healthy human donors were sorted by CX3CR1 level, labeled with CTV, and cultured with IL-7 and IL-12p70 for 8 days.Antigen-experienced CD8 + T cells are defined as non-naïve according to Fig.S1A.Flow cytometry plots from representative donors.Right: CX3CR1 expression level on undivided (div0) CD8 + T cells from the sorted populations.Undivided cells are defined as in Fig.S1B.Red lines: average.Symbol shapes: individual donors.ns: not significant (two-tailed paired t-test).(E) Top: Experimental setup.Per experiment, splenocytes were pooled from two to three mice from which OT-I memory T cells were sorted by CX3CR1 level.Sorted populations were cultured for 3 days in a medium.One well per experiment.Bottom: CX3CR1 expression by the sorted populations.Representative experiment of five independent experiments.(F) Left: Ratio of CX3CR1 expression level between sorted subsets.Bars: average.Symbol shapes: individual experiments.Data from 5 experiments.ns: non-significant (two-tailed paired t-test).Right: CX3CR1 expression by the sorted populations.Representative experiment of n = 5 independent experiments.(G) CD27 expression by the sorted populations.Representative experiment of n = 5 independent experiments.(C, F) gMFI, geometric mean fluorescent intensity.

Figure 3 .
Figure 3. CX3CR1 + memory CD8 + T cells have a selective survival disadvantage that is enhanced by factors released from necrotic but not apoptotic cells.In vitro experiments (A-E, G-I) were carried out with murine splenocytes enriched for OT-I memory T cells, but only OT-I memory T cells were evaluated.(A) Frequency of dead cells among OT-I memory T cells.(B) Absolute count of live OT-I memory T cells per well.(C, D) %dead cells within CX3CR1 + and CX3CR1 -OT-I memory T-cell populations.(E) %CD27 -cells within OT-I memory T-cell population.(F) mRNA expression of pattern recognition receptors known to sense DAMP.Expression on splenic murine CD8 + T cells at indicated days after LCMV infection.RNA sequencing data was obtained from ImmGen.Naïve:CD62L + CD44 -, TE:KLRG1 + CD127 -, MP:KLRG1 -CD127 + , Tcm:CD44 + CD62L + , Tem:CD44 + CD62L -.See methods for ImmGen annotations.(G-I) CX3CR1 expression by OT-I memory T cells.(G) Cells were cultured in medium (medium Ctrl) or in medium supplemented with supernatant from dead leukocytes (dead cell Sup).(H) Cells were cultured either in a medium containing supernatant from apoptotic MEF cells in which apoptosis was induced by the addition of DMSO-dissolved etoposide (apoptotic cell Sup), or supernatant from MEF treated with the equivalent concentration of DMSO only (DMSO Ctrl) or in DMSO-dissolved etoposide without MEF cells (etoposide Ctrl).(I) Cells were cultured in medium supplemented with supernatant from alive (alive cell Sup) or necrotic MEF cells, in which necrosis was induced by repeated freeze-thaw cycles (necrotic cell Sup).Bar/scatter plots: data from (A, D left) n = 2, (B, G-I) n = 3, or (C, D right, E) n = 4 independent experiments.Individual data points: individual wells.Symbol shapes: independent experiments.Each timepoint/condition was included in 1-4 experiments.1-5 wells per time point/condition per experiment.Bars and red lines: average.To account for variability between experiments, individual data points were normalized per experiment to the average of the (E) 2 h group, (G) 2 h medium ctrl, (H) 2 h DMSO Ctrl, or (I) 2 h alive cell sup control group.ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001 by (D) two-tailed paired t-test, (G, I) unpaired t-test or (A-C, E, H) one-way ANOVA+Tukey's multiple comparisons test.

Figure 4 .
Figure 4. TCR stimulation of CX3CR1 + memory T cells leads to, at least temporary, loss of CX3CR1 surface expression.(A) Experimental setup.Antigen-experienced blood CD8 + T cells were sorted by CX3CR1 level from n = 5 healthy human donors, labeled with CTV, and either stimulated or not with CD3 and CD28 antibodies for 3 days.All populations were cultured with IL-7 and IL-12p70 for 8 days.(B) CX3CR1 expression level before (day 0 division 0) and after cell culture (day 8).Day 0 and day 8 division 0 groups without CD3 and CD28 stimulation are also shown in Fig.3, and serve as controls here.Flow cytometry plots from representative donors.Red lines: average.Symbols: mean fluorescent intensity of CX3CR1 of individual donors.*p < 0.05, **p < 0.01 (one-way ANOVA+Tukey's).

Figure 5 .
Figure 5. CX3CL1 exposure reduces CX3CR1 surface expression on CD8 + memory T cells in a concentration-dependent manner but does not rescue their time-dependent death.Experiments were carried out with murine splenocytes enriched for OT-I memory T cells, but only OT-I memory T cells were evaluated.(A) Left: Representative flow cytometry plots of OT-I memory T cells.Numbers: %CX3CR1 + cells.Right: CX3CR1 expression levels and %CX3CR1 + cells within OT-I memory T-cell population.Data from four independent experiments.Error bars: S.D. Individual data points underlying the gMFI graph are shown in B and C. (B, C) Expression levels of indicated molecules on OT-I memory T cells.(D) Absolute numbers of viable OT-I memory T cells.(B-D) Bar plot: data from n = 4 (B, C) or n = 3 (D) independent experiments; 15 min and 12 h timepoints included in n = 2 (B, C) or n = 1 (D) experiments.Bars: average.Individual data points: individual wells.Symbol shapes: independent experiments.Two to eight wells per timepoint/condition per experiment.To account for variability between experiments, individual data points were normalized per experiment to the average of the 2 h,0 ng/mL CX3CL1 group.*p < 0.05, **p < 0.01, ***p < 0.001 (one-way ANOVA+Tukey's).gMFI, geometric mean fluorescent intensity.
sification of CD8 + T cells into distinct differentiation states based on CX3CR1 expression, at least in the assessed tissues.