Chemokine Signatures Of Pathogen-Specific T Cells II: Memory T Cells In Acute And Chronic Infection

Abstract

Pathogen-specific memory T cells (T_M) contribute to enhanced immune protection under conditions of re-infection, and their effective recruitment into a recall response relies in part on cues imparted by chemokines that coordinate their spatiotemporal positioning. An integrated perspective, however, needs to consider T_M as a potentially relevant chemokine source themselves. Here, we employed a comprehensive transcriptional/translational profiling strategy to delineate the identities, expression patterns and dynamic regulation of chemokines produced by murine pathogen-specific T_M. CD8⁺T_M, and to a lesser extent CD4⁺T_M, are a prodigious source for six select chemokines (CCL1/3/4/5, CCL9/10, XCL1) that collectively constitute a prominent and largely invariant signature across acute and chronic infections. Notably, constitutive CCL5 expression by CD8⁺T_M serves as a unique functional imprint of prior antigenic experience; induced CCL1 production identifies highly polyfunctional CD8⁺ and CD4⁺T_M subsets; long-term CD8⁺T_M maintenance is associated with a pronounced increase of XCL1 production capacity; chemokines dominate the earliest stages of the CD8⁺T_M recall response due to expeditious synthesis/secretion kinetics (CCL3/4/5) and low activation thresholds (CCL1/3/4/5/XCL1); and T_M chemokine profiles modulated by persisting viral antigens exhibit both discrete functional deficits and a notable surplus. Nevertheless, recall responses and partial virus control in chronic infection appear little affected by the absence of major T_M chemokines. While specific contributions of T_M-derived chemokines to enhanced immune protection therefore remain to be elucidated in other experimental scenarios, the ready visualization of T_M chemokine expression patterns permits a detailed stratification of T_M functionalities that may be correlated with differentiation status, protective capacities and potential fates.

INTRODUCTION

Pathogen-specific memory T cells (T_M) are an integral component of the anamnestic immune response and can provide immune protection by curtailing secondary (II°) infections, limiting morbidity and forestalling potential host death (1–5). These clinical outcomes are the net result of highly complex and coordinated interactions between multiple organ systems, tissues, cell types and extracellular factors that are marshaled into action following pathogen detection, and the relevant contributions of specific T_M to these processes are grounded in three fundamental determinants: their numbers, their location, and their differentiation status, i.e. the particular phenotypic, molecular and epigenetic makeup that permits T_M populations to respond with the elaboration of rapid effector activities as well as cooperative cellular interactions, local and systemic mobilization, II° effector T cell (T_E) differentiation, and proliferative expansion. The choreography of these events is in part governed by chemokines, a large family of mostly secreted small molecules that regulates the spatiotemporal positioning of motile cells (6–8).

Pathogen-specific T_M, by virtue of their distinct chemokine receptor expression patterns, are acutely attuned to varied chemokine cues as demonstrated in numerous in vitro and in vivo studies (6–11); however, as has been known for over two decades (12), T cells are also a relevant source for certain chemokines themselves, notably for CCL3, CCL4 and CCL5 which, beyond their chemotactic functions, can also act as competitive inhibitors of HIV binding to its co-receptor CCR5 (13–15); at micromolar concentrations, CCL5 may exert receptor-independent cellular activation, apoptosis and even antimicrobial activity, though some of the evidence is contradictory and the in vivo relevance unclear (16–21). Several other chemokines, including CCL1, CCL9/10 and XCL1, have further been reported as products of pathogen-specific T_M (22–26) but to date, experimental evidence in support of pathogen-specific T_M-derived chemokines as non-redundant contributors to effective immune protection at the level of II° T_E expansion, pathogen control and/or host survival remains limited to CCL3 and possibly XCL1 in some but not other murine model systems (27, 28).

Chemokine synthesis and secretion by T_M, similar to other effector functions such as cytokine and TNFSF ligand production, typically require a brief period of TCR activation, a prerequisite that may provide a safeguard against inappropriate T_M activation and immunopathology (29). It is therefore of interest that CCL5 is expressed in a constitutive fashion by human CD8⁺T cell subsets (22, 30, 31) where it is confined to a unique subcellular compartment and released with near instantaneous kinetics upon TCR engagement (30). Other studies, however, have demonstrated a preferential association of constitutively expressed CCL5 with cytolytic granules in HIV-specific CD8⁺T cell clones or primary human CD8⁺T cells (13, 31), but murine “memory-phenotype” CD8⁺T cells (CD8⁺T_MP), which contain abundant Ccl5 in addition to Ccl3 and Ccl4 transcripts, apparently do not express the corresponding proteins, the synthesis of which requires TCR stimulation (32, 33). Our recent work on chemokine signatures of pathogen-specific CD8⁺T_E may provide clues for a reconciliation of these discrepancies (34): at the peak of the effector response, CD8⁺T_E express constitutive CCL5 in a subcellular compartment distinct from granzyme- (GZM-) containing cytolytic granules but upon TCR stimulation, these structures partly coalesce just prior to secretion; in fact, release of pre-stored CCL5 proceeds so rapidly that temporarily, and within the confines of the immunological synapse, CCL5 concentrations well in excess of 1μM may be achieved (34), a potential foundation for in vivo CCL5 action at “supraphysiological” levels (16, 17).

Importantly, in the same study we provide detailed evidence that the induced production of all of the above (CCL1/3/4/5/9/10, XCL1) but no other chemokines constitutes a largely invariant functional signature of specific CD8⁺ and CD4⁺T_E generated in response to acute viral and bacterial infections as well as protective immunization (34). Building on this work, we have now extended our investigations to an interrogation of the identities, patterns, regulation and relevance of chemokines expressed by pathogen-specific T_M populations maintained after resolution of acute and under conditions of chronic infections. To be sure, due to the fundamentally different disease courses in these experimental scenarios, such T_M populations present with both distinctive and shared properties (35). If those differences warrant a terminological distinction may remain a matter of debate, but we favor an emphasis on relevant commonalities as argued by Jameson and Masopust (5). Hence, regardless of the acute or chronic course of an infection, we here refer to populations generated in the immediate wake of a pathogen challenge (~1 week) as T_E, and to populations present at later stages (>1 month) as T_M.

MATERIALS AND METHODS

Ethics statement

All procedures involving laboratory animals were conducted in accordance with recommendations in the “Guide for the Care and Use of Laboratory Animals of the National Institutes of Health”, the protocols were approved by the Institutional Animal Care and Use Committees (IACUC) of the University of Colorado (permit numbers 70205604[05]1F, 70205607[05]4F and B-70210[05]1E) and the Icahn School of Medicine at Mount Sinai (IACUC-2014–0170), and all efforts were made to minimize suffering of animals.

Mice, pathogens & challenge/vaccination protocols

C57BL6/J (B6), congenic B6.CD90.1 (B6.PL-Thy1^a/CyJ), congenic B6.CD45.1 (B6.SJL-Ptprc^a Pepc^b/BoyJ) and B6.CCL3^−/− (B6.129P2-Ccl3^tm1Unc/J) mice on a B6 background as well as Balb/c mice were purchased from The Jackson Laboratory; p14 TCRtg mice were obtained on a B6.CD90.1 background from Dr. M. Oldstone (CD8⁺T cells from these mice [“p14 cells”] are specific for the dominant LCMV-GP_33–41 determinant restricted by D^b). B6.CCL5^−/− mice (36) were obtained from Dr. M. von Herrath (these mice are identical to the commercially available B6.129P2-Ccl5^tm1Hso/J strain), and B6.CCL1^−/− mice were a gift from Dr. S. Manes (37). Lymphocytic choriomeningitis virus (LCMV) Armstrong (Arm; clone 53b) and clone13 (cl13) were obtained from Dr. M. Oldstone, stocks were prepared by a single passage on BHK-21 cells, and plaque assays for determination of virus titers were performed as described (38). Recombinant L. monocytogenes (LM) expressing full-length ovalbumin (rLM-OVA) (39) was grown and titered as described (34, 40). For acute infections, 8–10 week old mice were inoculated with a single intraperitoneal (i.p.) dose of 2×10⁵ plaque-forming units (pfu) LCMV Arm or 2×10³ cfu rLM-OVA i.v.; for persistent infections, mice were challenged with 2×10⁶ pfu LCMV cl13 i.v. (control cohorts were acutely infected with 2×10⁶ pfu LCMV Arm i.v.) or 4×10⁵ pfu γHV68 i.n. (intranasally). Combined TLR/CD40 vaccinations were performed as described (34, 41) by immunizing mice i.p. with 500μg ovalbumin (Sigma) in combination with 50μg αCD40 (FGK4.5, BioXCell) and 50μg polyinosinic:polycytidylic acid (poly[I:C], Amersham/GE Healthcare); all vaccinations were performed by mixing each component in PBS and injection in a volume of 200μl.

Lymphocyte isolation, T cell purification & stimulation cultures

Our procedures for isolation of lymphocytes from lymphatic and nonlymphoid tissues including total body perfusion with PBS are detailed elsewhere (42, 43). To generate p14 chimeras, naïve CD90.1-congenic p14 T cells were enriched by negative selection and ~5×10⁴ cells were transferred i.v. into sex-matched B6 recipients that were challenged 24h later with LCMV (25). For microarray analyses, splenic p14 T_M were positively selected using αCD90.1-PE antibody and PE-specific magnetic beads (StemCell Technologies) and further enriched to >99% purity by FACS sorting (BDBiosciences FACS Aria) as described (25). Adoptive transfer/re-challenge experiments (Fig.S3) were conducted with spleen cells obtained from LCMV-immune B6 mice and depleted of CD19⁺B cells or CD19⁺B and CD4⁺T cells using magnetic beads prior to i.v. transfer into B6 congenic recipients and LCMV challenge (25). For specific T cell stimulation, spleen cells were cultured for 5h with MHC-I- (1μg/ml) or MHC-II- (5μg/ml) restricted peptides in the presence (for FC analyses) or absence (for ELISA assays) of 1μg/ml brefeldin A (BFA, Sigma) (43). In some cases, highly purified p14 T_M were stimulated for 3h with plate-bound αCD3 (10μg/ml) and soluble αCD28 (2μg/ml) prior to processing for microarray hybridization.

In vivo T cell activation & chemokine blockade

In vivo activation of CD8⁺T_M (Fig.5D) was performed as described (44), i.e. vaccine-immune mice were injected with 100μg OVA₂₅₇ peptide i.v. followed by 250μg BFA/PBS i.p. 30min later; spleens were harvested 2h after peptide injection, processed and analyzed by FC. For in vivo chemokine neutralization in the context of CD8⁺T_M recall responses, we employed experimental designs, antibody dosages and treatment schedules as detailed in the legend to Fig.S3E/F using the following antibodies for CCL5 blockade: αCCL5 clone R6G9 (mIgG₁ (45)) or mIgG1 isotype clone MOPC-21 (Sigma); combined CCL3/4/5 blockade: αCCL3 clone 756605 (rIgG1), αCCL4 clone 46907 (rIgG2a), αCCL5 clone 53405 (rIgG2a) (RnDSystems) or rIgG control; and XCL1 blockade: polyclonal αXCL1 AF486 or goat IgG control AB-108-C (RnDSystems).

Microarray analyses

Microarray experiments were conducted with highly purified p14 T_M populations (directly ex vivo or after 3h in vitro stimulation with αCD3/αCD28) and Affymetrix M430.2 arrays as detailed in refs.(25, 34). Selected data are shown in Figs.1A, 2C, S1A & S2A/C, and the entire data sets, including those for ex vivo and αCD3/αCD28-stimulated p14 T_E, can be retrieved from the GEO repository accession number https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE143632; MAS5, RMA and GC-RMA normalization were performed and yielded essentially similar results (not shown).

Figure 1. Chemokine mRNA and protein expression by virus-specific CD8⁺T_M.

A., p14 T_M were obtained from spleens of LCMV-immune p14 chimeras (d46), enriched to >99% purity and processed for RNA extraction (either immediately or after 3h αCD3/αCD28 stimulation) and gene array analysis (n=4 individual mice). The bar diagrams display MAS5-normalized values of chemokine mRNA expression of p14 T_M analyzed ex vivo (gray bars) or after TCR stimulation (black bars). Statistically significant differences are indicated by asterisks, and the broken line indicates the detection threshold set at a MAS5 value of 40; coverage: 39/40 chemokines (Ccl26 not on chip). B., left: p14 T_M (d68) were analyzed for chemokine protein after 5h culture in the absence (gray histograms) or presence of GP₃₃ peptide (black tracings); all histograms are gated on p14 T_M (the broken line histogram in the CCL5 panel indicates a negative control stain). Right: summary of constitutive (no peptide) and induced (GP₃₃ peptide stimulation) chemokine expression by p14 T_M (SEM, n=3 individual mice, data from one of three similar experiments, asterisks indicate significant expression differences between unstimulated and stimulated p14 T_M).

Figure 2. Constitutive CCL5 expression by pathogen-specific T_M.

A., Constitutive chemokine expression by LCMV-specific CD8⁺T_M cells was analyzed in different tissues on d175 (top row, spleen) or d55 (middle and bottom rows) after acute LCMV infection. Bottom row, middle: PBMCs from rLM-OVA-immune mice were analyzed on d83. Two additional contour plots (right) gated on epitope-specific CD8⁺T_M recovered from LCMV-immune B6 (D^bNP₃₉₆⁺, d175, spleen) or Balb/c (L^dNP₁₁₈⁺, d174, PBMC) mice demonstrate a slight but significant reduction of constitutive CCL5 expression levels by CD62L^hi CD8⁺T_CM vs. CD62L^lo T_EM subsets. B., summary of constitutive CCL5 expression by LCMV-specific CD8⁺T_M recovered from lymphatic and nonlymphoid organs. C., temporal regulation of ex vivo detectable Ccl5 mRNA and CCL5 protein expression by aging p14 T_M (the gray shaded area demarcates the transitional period from effector [d8] to early memory [d42] stage, and asterisks identify significant differences comparing young [~d40–50] to older p14 T_M). D., complementary ex vivo CCL5 expression data by endogenously generated D^bNP₃₉₆⁺ CD8⁺T_M; the dot plot insert shows constitutive CCL5 vs. GzmA expression in blood-borne d53 D^bNP₃₉₆⁺ CD8⁺T_M. E., ex vivo CCL5 expression by CD8⁺T cells from age-matched naïve vs. LCMV-immune (d45) mice; the corresponding bar diagram quantifies the fraction (%) and expression levels (GMFI) of CCL5⁺ CD8⁺T_MP in naïve vs. young and aged LCMV-immune mice (statistical significance indicated by asterisks, no difference between d45 and d206 LCMV-immune mice). F., serum CCL5 levels as a function of time after LCMV infection of B6 mice (top) and age-matched uninfected B6 mice (bottom); the broken line indicates the detection threshold as determined in LCMV-immune B6.CCL5^−/− mice. All summary data represent SEM with 3–4 individual mice analyzed per group or time point in multiple independent experiments (panels A/B/E: 2–4 experiments; panels C/D/F: 1–2 experiments).

Peptides and MHC tetramers

Peptides corresponding to the indicated pathogen epitopes were obtained from Peptidogenic, the NJH Molecular Core Facility or GenScript at purities of >95%; their MHC-restriction and amino acid sequences are indicated. LCMV epitopes: GP_33–41 (D^b/KAVYNFATC), GP_276–286 (D^b/SGVENPGGYCL), NP_396–404 (D^b/FQPQNGQFI), GP_64–80 (IA^b/GPDIYKGVYQFKSVEFD), NP_118–126 (L^d/RPQASGVYM); and rLM-OVA epitopes: OVA_257–264 (K^b/SIINFEKL), LLO_190–201 (IA^b/NEKYAQAYPNVS). Note that the LCMV-GP₃₃ peptide also stimulates K^bGP₃₄⁺ CD8⁺T_E/M not captured by D^bGP₃₃ tetramer stains; we therefore refer to the T cells activated in GP₃₃ stimulation cultures as “GP_33/4-specific” CD8⁺T_E/M. D^bNP₃₉₆, D^bGP₃₃, D^bGP₂₇₆, L^dNP₁₁₈, K^bOVA₂₅₇ and IA^bGP₆₆ complexes were obtained from the NIH tetramer core facility as APC or PE conjugates and/or biotinylated monomers; and CD1/αGalCer tetramers were a gift from Dr. L. Gapin. Note that the IA^bGP₆₆ tetramer identifies the same population of CD4⁺T_E/M responsive to stimulation with the longer GP_64–80 peptide (46). Tetramer staining was performed as described, i.e. for 45min at 4°C or RT for MHC-I tetramers and for 90min at 37°C in the presence of sodium azide for MHC-II tetramers (47).

Antibodies, staining protocols and flow cytometry (FC)

All antibodies and FC staining protocols used in the present study are described elsewhere. Specifically, this includes antibodies/protocols for standard cell surface and intracellular staining as well as antibody conjugation (25), our detailed antibody characterization and staining strategies for detection practically all murine chemokines (22, 34), and the procedure for visualization of LCMV-NP expression in infected primary cells (42, 48); perforin stains were performed with antibody clone S16009B conjugated to PE (Biolegend). All samples were acquired on FACS Calibur, LSRII, LSRFortessa (BDBiosciences) or Attune NxT (ThermoFisher) flow cytometers and analyzed with DIVA (BDBiosciences) and/or FlowJo (TreeStar) software. To estimate the relative magnitude of the CD8⁺T_M chemokine response in the context of other CD8⁺T_M activities (Fig.3D), we determined the percentage of individual NP₃₉₆-specific CD8⁺T_M (~d42) subsets expressing constitutive (GzmA/B, perforin) or inducible (CCL1/3/4/5/9/10, XCL1; IFNγ, IL-2, IL-3, GM-CSF; TNFα, FasL, CD40L; degranulation) effector functionalities as shown/described here (Figs.3A/B & 4A/B) or in ref.(25); details about our calculations to estimate the relative magnitude and distribution of individual functional activities as displayed in Fig.3D are provided in ref.(34).

Figure 3. Induced chemokine expression by pathogen-specific T_M.

A., induced chemokine expression by splenic GP_33/4-specific CD8⁺ (top row) and GP₆₄-specific CD4⁺T_M (bottom row) evaluated on d55 after acute LCMV challenge of B6 mice. B., left: summary of induced chemokine expression by LCMV epitope-specific (IFNγ⁺) CD8⁺ and CD4⁺T_M (SEM, 3 mice/group, d42). Right: induced chemokine profiles of specific T_M recovered from the spleens of rLM-OVA-immune mice (SEM, 3 mice/group, d83). Data are representative for 2–3 independent experiments conducted at various time points after LCMV or rLM-OVA infection, and asterisks indicate significant differences between respective pathogen-specific CD8⁺ and CD4⁺T_M. C., induced chemokine production by NP₃₉₆-specific CD8⁺T_M (black) and GP₆₄-specific CD4⁺T_M (gray) as a function of time after LCMV challenge; details for data display and statistics as in legend to Fig.2C (SEM of 3–6 mice/time point; all data were generated with splenic T_M with the exception of XCL1 production for which blood-borne NP₃₉₆-specific CD8⁺T_M were analyzed, and were combined from four independent experiments). D., composition of the NP₃₉₆-specific CD8⁺T_M response (~d42) as reflected by the relative magnitude of individual subsets expressing constitutive (GzmA/B and perforin) or inducible (all other including CCL5) effector activities (data for the pie chart represent averages of ≥3 mice analyzed in up to five separate experiments).

Figure 4. Induced CCL1 expression as a determinant for highly polyfunctional CD8⁺ and CD4⁺T_M subsets.

A., patterns of induced cytokine and TNFSF ligand expression by GP_33/4–specific CD8⁺T_M (d67, spleen). Middle and bottom rows: red dots designate specific CD8⁺T cells producing a third cytokine (identified in red font) in addition to IFNγ/TNFα (or CD40L/IL-2) expression displayed in conventional dot plots. B., top row left: CCL3/4 co-expression by all GP_33/4–specific CD8⁺T_M; top row right: specific CD8⁺T_M subsets producing CCL1, CCL9/10 or XCL1 also express higher levels of CCL3 (the color-coded CCL3 GMFI values refer to cells identified with the same color; statistical significance is indicted by asterisks). Middle row: induced CCL1/XCL1 expression and corresponding CCL3 and IFNγ expression levels in subpopulations identified by color-coded histograms and corresponding GMFI values. In addition, IL-2 production by CCL1⁻ (black tracing) vs. CCL1⁺ (red tracing) subsets is shown. Bottom row, left: ~80% of CCL1⁺ cells fail to induce CCL9/10 expression suggesting that XCL1 and CCL9/10 are expressed in a reciprocal fashion in CCL1⁺ subsets (compare to above CCL1/XCL1 plot). Lack of CCL9/10 production is associated with higher CCL3 (as well as IFNγ and IL-2) expression. Bottom row, right: CCL4 GMFI values and IL-2/CD40L expression patterns by CCL1⁻ (blue) vs. CCL1⁺ (red) subsets. All values listed in panels A & B are the average of 3 mice analyzed. C., complementary analyses of cytokine and TNFSF ligand production by GP₆₄-specific CD4⁺T_M (data display as in panel A; note that IL-2⁺ cells are CD40L⁺, IFNγ^hi and TNFα^hi). D., top row: co-production of CCL3/4 by a major CD4⁺T_M subset and enrichment of CCL1, CCL9/10 and XCL1 expression in the CCL3/4⁺ subpopulation. Middle row: progressive enrichment of CD40L⁺ IL-2 producers in CCL1⁻/4⁻ → CCL1⁻/4⁺ → CCL1⁺/4⁺ subsets (color-coded gating strategy). Bottom row: no clear association of induced CCL9/10 and XCL1 production with CCL1 expression. However, CCL1⁺ CD4⁺T_M not only contain more IL-2 producers, the latter cells are also enriched for a pronounced IFNγ^hi phenotype. Representative data from two independent experiments are shown, and the text boxes below panels A/B and C/D summarize the functional characteristics of LCMV-specific CD8⁺T_M and CD4⁺T_M subsets identified by induced CCL1 expression.

ELISA

Quantitation of CCL3, CCL4 and CCL5 in sera and/or peptide-stimulated tissue culture supernatants (Figs.2F & 5D) was performed using respective Quantikine ELISA kits and protocols provided by the manufacturer (RnD Systems). To determine secretion kinetics of pre-stored CCL5 by specific CD8⁺T_M (Fig.5E), spleen cells from LCMV-immune mice were pre-incubated for 30min at 37°C with 10μg/ml cycloheximide (CHX, Sigma) to prevent translation, stimulated with NP₃₉₆ peptide, and CCL5 in the supernatant was quantified by ELISA (the amount of CCL5 secreted in the absence of peptide stimulation was subtracted from stimulated samples at all time points). To calculate CCL5 release on a per cell basis, FC analyses were performed in parallel to calculate the exact numbers of D^bNP₃₉₆⁺ CD8⁺T_M in the stimulation culture.

Figure 5. Kinetics and activation thresholds for induced chemokine production by specific CD8⁺T_M.

A. & B., spleen cells from LCMV-immune p14 chimeras (~10 weeks) were stimulated for 0–5h with GP₃₃ peptide and expression of indicated chemokines, cytokines and TNFSF ligands was determined as a function of stimulation time. The dot plots (gated on p14 T_M) show representative data for the temporal regulation of induced CCL1/3/4, XCL1 and IFNγ synthesis, and the bar diagram ranks inducible chemokine (black) and other effector molecule (gray) production according to ET₅₀ values (time required to induce effector functions in 50% of the population capable of producing a given effector molecule); significant differences are indicated by asterisks. Overall, induced CCL3/4 and IFNγ synthesis proceeds with significantly faster kinetics as compared to other effector functions (p<0.003) with the exception of TNFα. C., ET₅₀ values for induced XCL1 expression by young (d64) vs. aged (d637) p14 T_M were determined as above. D., ~2 months after ovalbumin/αCD40/polyI:C vaccination, B6 mice were injected with BFA only or OVA₂₅₇ peptide & BFA as indicated, and activation status (CD69) and in vivo IFNγ, TNFα and chemokine production by K^bOVA₂₅₇⁺ CD8⁺T_M was determined as described in Methods; values are the percentage of K^bOVA₂₅₇⁺ CD8⁺T_M in respective quadrants synthesizing indicated cytokines/chemokines. E., kinetics of pre-stored CCL5 release by individual NP₃₉₆-specific CD8⁺T_M (d45) were determined in stimulation cultures supplemented with CHX to prevent protein neosynthesis; for comparison, CCL5 secretion in the absence of CHX is featured for the 5h time point (gray symbol). F., intracellular IFNγ and CCL3/4 content of p14 T_M (~9 weeks) determined after 5h stimulation with graded doses of GP₃₃ peptide (plots gated on p14 T_M). The adjacent diagrams summarize the emergence of p14 T_M expressing effector functions as a function of GP₃₃ peptide concentration. G., summary of activation thresholds (EC₅₀ values) determined for individual chemokines (black bars) and other effector functions (gray bars) elaborated by splenic p14 T_M stimulated with graded doses of GP₃₃ peptide (10⁻¹¹-10⁻⁶M; data are combined results from 3 similar experiments evaluating 3 p14 chimeras each; to account for inter-experimental variability, all samples were co-stained for IFNγ, the average IFNγ EC₅₀ value was set at 1.0 [dotted line] and the EC₅₀ values of other effector functions were calculated accordingly). Statistically significant differences comparing the IFNγ activation threshold to CCL1, CCL4, CD40L and IL-2 are indicated by asterisks. H., CCL3/4/5 secreted by p14 CD8⁺T_M as a function of GP₃₃ peptide concentration was determined by ELISA (5h stimulation) and EC₅₀ values are summarized in the adjacent bar diagram. All data in panels B, C & E-H are representative for individual experiments conducted 2–4 times (SEM, n=3–4 mice per experiment).

Statistical analyses

Data handling, analysis and graphic representation was performed using Prism 4.0 and 6.0c (GraphPad Software, San Diego, CA). All data summarized in bar and line diagrams are expressed as mean ± 1 SE. Asterisks indicate statistical differences calculated by unpaired or paired Student’s t-test and adopt the following convention: *: p<0.05, **: p<0.01 and ***: p<0.001. ET₅₀ values (response kinetics; Fig.5A–C) and EC₅₀ values (activation thresholds; Fig.5F–H) were calculated by plotting the fraction of specific T_M demonstrating detectable chemokine/cytokine staining (FC) or the amount of secreted chemokines (ELISA) as a function of stimulation time (10⁻⁶M peptide for 0–5h) or peptide concentration (10⁻⁶-10⁻¹¹M peptide for 5h) followed by non-linear regression analysis using appropriate data format and analysis functions in the Prism software. Note that ET₅₀ and EC₅₀ values are independent of the fact that not all effector functions are induced in all T_M of a given specificity.

RESULTS

Chemokine signatures of virus-specific CD8⁺T_M.

In order to define the range of chemokines produced by virus-specific CD8⁺T_M, we employed a comprehensive transcriptional/translational profiling strategy previously used for the characterization of CD8⁺ effector T cells (CD8⁺T_E) generated in response to an acute infection with lymphocytic choriomeningitis virus (LCMV) (34). In brief, so-called “p14 chimeras” were generated by transfusing congenic B6 mice with a trace population of naïve TCR transgenic CD8⁺T cells (p14 T_N) specific for the dominant LCMV-GP_33–41 determinant; following infection with LCMV Armstrong (Arm; 2×10⁵ pfu i.p.), p14 T_E rapidly differentiate, expand and contribute to efficient virus control before contracting and developing into p14 T_M populations ~6 weeks later (25, 49, 50); highly purified p14 T_M were then subjected to gene array analyses conducted directly ex vivo or after a 3h in vitro TCR stimulation as detailed in Methods and refs.(25, 34).

In these analyses, 11 distinct chemokine mRNA species scored as present in p14 T_M and, according to their expression patterns, could be clustered into three groups (Fig.1A): 1., absence of ex vivo detectable mRNA but robust transcription after TCR stimulation (Ccl1 and Ccl17); 2., constitutive mRNA expression that significantly increased upon TCR engagement (Ccl3, Ccl4, Ccl9/10, Ccl27, Cxcl10 and Xcl1); and 3., chemokine mRNA species that remained unaffected or were downregulated by TCR activation (Ccl5, Ccl6 and Ccl25; this group also contains Cklf and Cklfsf3/6/7 [Fig.S1A], members of the related chemokine-like factor superfamily about which relatively little remains known to date (51)). Altogether, the chemokine mRNA expression patterns of p14 T_M are notably similar to the corresponding transcriptional profiles of p14 T_E (34).

Complementary protein expression analyses were conducted with chemokine flow cytometry (FC) using a portfolio of extensively characterized antibodies that permit detection of practically all murine chemokines at the single-cell level (22, 34). Upon TCR stimulation with GP₃₃ peptide, nearly all p14 T_M produced CCL3, CCL4 and CCL5; a large subset made XCL1; and smaller fractions expressed CCL1 and CCL9/10. Of note, neither Ccl6, Ccl17, Ccl25, Ccl27, Cxcl10 nor any other chemokine mRNA was translated by p14 T_M (Fig.1B and not shown). Remarkably, most p14 T_M expressed abundant CCL5 even in the absence of TCR activation (Fig.1B) establishing an expression pattern that is unique not only for CD8⁺T_M chemokines but for cytokines at large (25). With only six chemokine mRNA species serving as templates for induced and in the case of Ccl5 also constitutive translation, the chemokine signatures of p14 T_M are therefore equivalent to those of p14 T_E (34).

Constitutive CCL5 expression by endogenously generated pathogen-specific CD8⁺T_M.

We next extended our investigations to endogenously generated LCMV-specific CD8⁺T_M. Just like p14 T_M, these populations, regardless of epitope specificity, constitutively expressed CCL5 but no other chemokine, cytokine or TNFSF ligand (Fig.2A, ref.(25) and not shown). Moreover, this expression pattern was largely identical in CD8⁺T_M recovered from lymphatic and nonlymphoid tissues, was independent of mouse strain, and also pertained to specific CD8⁺T_M generated in response to infection with the bacterium L. monocytogenes (LM) (Fig.2A/B). We note, however, that constitutive CCL5 expression by CD8⁺T_M in peripheral lymph nodes (LNs) was somewhat reduced in comparison to all other tissues (Fig.2A/B). Since LNs are enriched for CD62L-expressing “central memory” T cells (T_CM), we evaluated the possibility that T_CM in the periphery express lower CCL5 levels than “effector memory” T cells (T_EM). While this was indeed the case (Fig.2A), the differences were small and the progressive T_EM→T_CM conversion in aging pathogen-immune mice (25, 52) was associated with only a slight reduction of Ccl5 message and overall CCL5 content in splenic CD8⁺T_M (Fig.2C/D). Thus, constitutive CCL5 expression, in contrast to the rapid (GzmB) or gradual (GzmA) decline of constitutively expressed components within the cytolytic effector pathway, is a distinctive hallmark of most peripheral CD8⁺T_M populations maintained in long-term memory (Fig.2D and ref.(25)).

Our findings are in apparent contrast to the reported presence of Ccl5 mRNA but absence of protein in murine CD44^hi CD8⁺T_MP (32, 33). To reconcile these discrepancies, we evaluated constitutive CCL5 expression in major hematopoietic cell subsets isolated from lymphatic and nonlymphoid organs of naïve mice (Figs.2E & S2). Here, ex vivo detectable CCL5 was restricted in all tissues to two hematopoietic lineages, NK and CD3ε⁺T cells. Moreover, substantial CCL5 expression by NK cells as reported previously (22) contrasted with the weak CCL5 expression by CD8⁺T_MP that was slightly more pronounced in the CD122⁺ subset (Fig.S1B–E); additional minor CCL5⁺ T cell subsets included γδTCR⁺T cells as well as CD4⁺T_MP and NKT cells that harbored constitutive CCL5 at marginal levels (Fig.S1C/F/G). In no case did we find ex vivo CCL5 expression by B cells, monocytes, macrophages, DC subsets or granulocytes (Fig.S2B).

The low-level CCL5 content of CD8⁺T_MP in naïve mice provides an explanation for the seeming absence of CCL5 in these cells (32, 33) and stands in stark contrast to the robust CCL5 expression pattern of CD8⁺T_MP in age-matched, LCMV-immune mice (Fig.2E). Thus, a viral challenge can leave an imprint of prior infection by shifting the relative distribution of constitutive CCL5 expressors from both NK cells and CD8⁺T_MP (each contributing ~50% to the CCL5⁺ population in naive mice) to CD8⁺T_MP as the predominant constituent of hematopoietic CCL5⁺ cells (>90%) (Figs.S2H/I). These observations, together with the presence of CCL5 in the sera of naïve mice at levels of ~1ng/ml (Fig.2F), may warrant a re-classification of CCL5 as a “dual function” rather than “inflammatory” chemokine. Although CD8⁺T_M would be an obvious source for the serum CCL5, levels were not increased in LCMV-immune mice (Fig.2F), and the absence of “spontaneous” CCL5 release by CD8⁺T_MP confirms earlier reports about the need for TCR stimulation to induce CCL5 secretion (32, 33).

Induced chemokine synthesis by pathogen-specific CD8⁺ and CD4⁺T_M.

To delineate the principal chemokine production capacity of endogenously generated pathogen-specific CD8⁺T_M, we visualized induced chemokine expression by peptide re-stimulation of CD8⁺T_M obtained from LCMV- and rLM-OVA-immune mice. Similar to p14 T_M, nearly all specific CD8⁺T_M readily made CCL3/4/5, large subsets produced XCL1, and 15–25% expressed CCL1 and CCL9/10; no other chemokines were synthesized in response to TCR activation (Fig.3A/B and not shown). Since these expression patterns were further consistent across different epitope specificities (and therefore also independent of immunodominant determinants and functional avidities (25, 34)), they apparently constitute a largely invariant functional property of pathogen-specific CD8⁺T_M. Similar analyses conducted with LCMV- and rLM-OVA-specific CD4⁺T_M also revealed subsets producing CCL3/4/5 (40–50%), CCL1 (20–30%), XCL1 (5–15%) or CCL9/10 (<5%) (Fig.3A/B); these expression patterns essentially recapitulate the chemokine response of the respective pathogen-specific CD4⁺T_E (34).

The long-term maintenance of CD8⁺T_M populations is associated with a gradual remodeling process that promotes their functional diversification as reflected in an increasing capacity for IFNγ, IL-2, CD40L and FasL synthesis (25). We therefore monitored the temporal modulation of chemokine mRNA species in aging p14 T_M and observed four distinct patterns of longitudinal mRNA expression: enduring absence (Ccl1); a continuous decrease (Ccl5, Ccl9); a progressive increase (Xcl1); and an initial ~5 month decline followed by a subsequent rise (Ccl3, Ccl4) (Fig.S2A), a somewhat unusual dynamic comparable to that previously reported for Ifng, Tnf, Prf1 and Gzmb/k/m (25). The relevance of the latter kinetics, however, remains unclear since inducible CCL3/4 production was not altered in aging NP₃₉₆- or GP_33/4-specific CD8⁺T_M (Figs.3C & S2B). Similarly, and notwithstanding the slight decline of constitutive CCL5 expression (Fig.2C/D), the capacity for stimulated CCL5 synthesis remained a stable property of aging CD8⁺T_M as did the CCL1 production potential by a major subset (Figs.3C & S2B). In contrast, inducible CCL9/10 and XCL1 expression tracked with the modulation of the respective mRNA patterns in the memory phase, i.e. aging CD8⁺T_M populations featured a decreasing potential for CCL9/10 but increasing proficiency for XCL1 production (Figs.S2A/B & 3C). In fact, within the temporal context of increasingly diversified CD8⁺T_M activities (25), the rising XCL1-production competence constitutes the most pronounced functional gain for aging CD8⁺T_M prompting further investigations as discussed below. Lastly, the relative proportions of chemokine-producing CD4⁺T_M subsets, despite the gradual decline of total specific CD4⁺T_M numbers (47), did not significantly change over time (Fig.3C).

CD8⁺T_M-produced chemokines in context.

The full spectrum of rapid effector functions elaborated by pathogen-specific CD8⁺T_M remains at present unknown. While the induction of cytokines (IFNγ, IL-2, IL-3, GM-CSF), TNFSF ligands (TNFα, CD40L, FasL) and degranulation are readily observed in short-term re-stimulation cultures (25, 43, 53), virus-specific CD8⁺T_M producing IL-4, IL-5, IL-10, IL-13 or IL-17 are rare if present at all in our experimental systems. To provide a provisional estimate for the relative contribution of chemokine production to the gamut of established CD8⁺T_M functionalities, we quantified the fractions of NP₃₉₆-specific CD8⁺T_M capable of individual chemokine, cytokine and TNFSF ligand synthesis; constitutive GzmA/B and perforin expression; and degranulation (Fig.3D); while at best a rough approximation, our calculations indicate that chemokine synthesis may account for >40% of CD8⁺T_M activities and thus suggest a potential importance of CD8⁺T_M-derived chemokines in the context of recall responses.

Induced CCL1 expression as a determinant for “polyfunctional” CD8⁺ and CD4⁺T_M subsets.

The extent to which individual effector molecules contribute to primary (I°) T cell-mediated pathogen control has been delineated in multiple experimental systems, yet comparatively little information is available about their precise role in immune protection afforded by recall responses. Rather, the concept of “T cell polyfunctionality”, typically measured in IFNγ/TNFα/IL-2 co-expression analyses, argues that the presence of T_M with a diversified and robust functional repertoire correlates with better pathogen control irrespective of the precise mechanisms by which the analyzed T cell functions may in fact contribute to this task (53, 54). Here, we demonstrate that induced CCL1 expression by virus-specific CD8⁺ and CD4⁺T_M identifies subsets characterized by highly diversified effector functions and increased production of multiple effector molecules (since we were unable to concurrently visualize more than 2–3 chemokines, we relied on a series of successive and complementary stains to define complex functional T_M profiles according to 11 parameters) (Fig.4).

In addition to IFNγ, LCMV-specific CD8⁺T_M subsets rapidly synthesize TNFα, IL-2, GM-CSF and CD40L (25, 43), and the level of IFNγ expression in TNFα⁺, IL-2⁺, GM-CSF⁺ and CD40L⁺ subsets was consistently and significantly higher than in CD8⁺T_M subsets producing only IFNγ (p<0.05, analysis of NP₃₉₆-, GP_33/4- and GP₂₇₆-specific CD8⁺T_M, not shown). Combinatorial analysis of these functions demonstrated that IL-2⁺ and CD40L⁺ but not GM-CSF⁺ CD8⁺T_M belong to subsets that also produce IFNγ and TNFα at higher levels. Among the IL-2 and CD40L producers, however, only ~25% exhibited co-expression while another ~25% were IL-2⁺/CD40L⁻ and ~50% expressed CD40L in the absence of IL-2 (Fig.4A). Similar to CD8⁺T_E, CD8⁺T_M co-produced CCL3/4 (Fig.4B, upper left panel) the particularly “tight” association of which may in part be due to the formation of intracellular CCL3/4 hetero-oligomers (34). Further comparing the extent of induced CCL3 production in subsets expressing additional chemokines with those that do not, CCL3 was significantly increased in CCL1⁺ and XCL1⁺ and to a lesser extent also CCL9/10⁺ populations (Fig.4B, upper panels). In fact, the CCL1⁺ population is a subset of XCL1⁺ cells that produced significantly more CCL3 and IFNγ compared to the XCL1⁺/CCL1⁻ CD8⁺T_M that in turn exhibited higher CCL3/IFNγ levels than the XCL1⁻/CCL1⁻ population. In addition, while IL-2 producers were found in both CCL1⁻ and CCL1⁺ populations, the latter subset was significantly enriched for IL-2⁺ cells (~45% vs. 23%) which also produced IL-2 at higher levels (p=0.0206) (Fig.4B, middle panels). CCL1⁻ and CCL1⁺ populations were also stratified according CD40L expression: here, the CCL1⁺ subset was clearly enriched for both IL-2⁺/CD40L⁺ and IL-2⁻/CD40L⁺ subsets (Fig.4B, lower panels). Finally, the majority (~3/4) of CCL9/10 expressors were found among CCL1⁻ cells and, expectedly, expressed significantly lower amounts of CCL3 (Fig.4B, lower panels). Therefore, induced CCL1 expression defines a highly polyfunctional virus-specific CD8⁺T_M subset comprising ~25% of epitope-specific cells that are characterized by a CCL3^hi/CCL4^hi/CCL5⁺/XCL1⁺/IFNγ^hi/TNFα^hi phenotype, are enriched for IL-2^hi and/or CD40L⁺ cells, and exhibit variable GM-CSF and/or CCL9/10 expression (CCL5 analyses are not displayed in Fig.4B since all specific CD8⁺T_M are capable of induced CCL5 production shown in Fig.3A–C).

Similar functional stratifications were also performed for LCMV-specific CD4⁺T_M taking into account that their relative IL-2⁺ and CD40L⁺ subsets are significantly larger than the corresponding CD8⁺T_M subpopulations (43) (Fig.4C/D). Collectively, these analyses demonstrate that induced CCL1 production is also useful to delineate a polyfunctional CD4⁺T_M subset that accounts for ~25% of the specific CD4⁺T_M compartment and is defined by a CCL3^hi/CCL4^hi/IFNγ^hi/TNFα^hi/IL-2⁺/CD40L^hi expression pattern with variable induction of XCL1, GM-CSF, and/or CCL9/10.

Orchestration of the CD8⁺T_M chemokine response (I): kinetics of chemokine production.

The presence of chemokine transcripts in T_M has been suggested to confer a kinetic advantage that permits more rapid protein synthesis (32, 55), but this hypothesis has to our knowledge not yet been experimentally tested. The distinct expression patterns of ex vivo detectable chemokine mRNA species in LCMV-specific CD8⁺T_M (absence of Ccl1 mRNA, abundant Ccl3/4/5 mRNA and progressively increasing Xcl1 mRNA; Fig.S2A) would therefore suggest that early synthesis of CCL3/4 is followed by that of XCL1 and finally CCL1 protein. Experiments performed with LCMV-immune p14 chimeras indeed confirmed this prediction: by calculating the time point at which p14 T_M demonstrate detectable chemokine expression in half of the responder population (“effective time 50”: ET₅₀), we found ET₅₀ values of ~1h10min for induced CCL3/4 and IFNγ production while XCL1 synthesis was significantly delayed (~2.0h, p<0.0002) yet faster than the eventual induction of CCL1 (~3.5h, p<0.0001) (Fig.5A/B). In fact, a direct comparison to other effector functions demonstrated that induced chemokine production dominates the earliest (CCL3/4) and later (CCL1, CCL9/10) stages of the CD8⁺T_M response (Fig.5B) and also illustrates that the presence/absence of preexisting mRNA species is not the sole predictor for the temporal elaboration of effector functions: although absent at the mRNA level in resting p14 T_M, Il2, Csf2/Gmcsf and Cd40lg mRNA species were rapidly transcribed following TCR stimulation (Fig.S2C) but translation of CD40L, IL-2 and GM-CSF proteins occurred with significantly different kinetics (Fig.5B). Conversely, despite the presence of Ccl9 mRNA (Fig.S2A), synthesis of the corresponding protein was a relatively late event (Fig.5B). Since differential mRNA stability may shape the temporal order in which genes encoding inflammatory mediators are expressed (56), it is conceivable that this phenomenon also contributes to the observed differences in translation kinetics. Nevertheless, the significant increase of Xcl1 message in aging p14 T_M (Fig.S2A) allowed for an analysis of the kinetics with which an individual chemokine is synthesized as a function of changing mRNA levels. As shown in Fig.5C, not only did more aged p14 T_M (~21 months) produce XCL1 but they did so ~20min faster than younger p14 T_M (~9 weeks). In contrast, the kinetics of CCL1/3/4 and IFNγ production were not significantly different in young and old CD8⁺T_M (not shown).

To relate the preceding in vitro data to a specific in vivo CD8⁺T_M response, B6 mice previously immunized with combined TLR/CD40 vaccination were subjected to a peptide challenge and co-administration of BFA to allow for the intracellular accumulation of chemokines and other effector molecules in vivo (34, 44). Within 2h after challenge, ~2/3 of the vaccine-specific CD8⁺T_M became activated as determined by CD69 expression, and the majority of these cells readily synthesized CCL3/4 and XCL1 (Fig.5D) demonstrating that the in vivo CD8⁺T_M response features rapid and abundant chemokine production. Lastly, we determined the secretion kinetics and quantities of CCL5 constitutively expressed by LCMV-specific CD8⁺T_M (d45) in an ELISA format. Interestingly, both the rapid speed of CCL5 release and the amount of pre-stored CCL5 secreted by individual CD8⁺T_M (Fig.5E) was equivalent to CD8⁺T_E (34). Thus, the CD8⁺T_M chemokine response proceeds through an ordered sequence of overlapping events that are in part dictated by the extent of pre-existing protein and/or mRNA species: CCL5 → CCL3/4 & IFNγ → TNFα → CD40L → XCL1 → IL-2 → GM-CSF → CCL1 → CCL9/10. Given the particularly fast release of CCL3/4/5, stimulated CD8⁺T_M rather than DCs or other APCs (57) likely constitute the principal source for these chemokines in a recall response.

Orchestration of the CD8⁺T_M chemokine response (II): activation thresholds for chemokine production and secretion.

In addition to the temporal regulation of the T cell chemokine response, production of individual chemokines may be controlled by distinct activation thresholds. The “functional avidities” of specific T cells are often determined by measuring induced IFNγ synthesis in response to graded concentrations of antigenic peptide (34, 58). Although the interpretation of such data provides insights into properties intrinsic to the interaction between TCRs and peptide/MHC complexes it should be noted that the experimental readout is modulated by complex signaling events that precede the induction of a given T cell function. For example, functional avidities for induced IFNγ production were reported to be lower than those required for IL-2 production by the same cells suggesting that limiting antigen concentrations may permit the elaboration of direct T cell effector functions (IFNγ) whereas induction of proliferative T cell responses, mediated by T cell-produced IL-2, preferentially occurs at higher antigen loads (53).

We therefore compared the functional avidities/activation thresholds for induced chemokine and cytokine/TNFSF expression by LCMV-specific CD8⁺T_M (Fig.5F–H). Although the activation thresholds for 10 distinct CD8⁺T_M effector functions varied at most by a factor of only ~3.0 (Fig.5G), several findings are noteworthy: 1., the activation thresholds for CCL3/4 production were only slightly lower than for IFNγ induction yet we consistently observed the emergence of a small CCL3⁺ or CCL4⁺T cell subset in the absence of detectable IFNγ at limiting peptide concentrations. In fact, at a peptide concentration of 10⁻¹⁰M, the very small subset of responding p14 T_M consisted to >50% of CCL3/4⁺ cells that were IFNγ⁻. In contrast, no IFNγ induction was observed in the absence of CCL3/4 expression (Fig.5F). Parallel analyses of aged p14 T_M (~21 months) yielded identical results and indicate that subtle differences pertaining to the thresholds for defined effector functions are preserved in long-term memory (not shown). 2., although production of CCL1 and XCL1 was initiated with delayed kinetics (Fig.5B), the activation thresholds for these chemokines were relatively low and comparable to CCL3/4 (Fig.5G). 3., the highest activation thresholds were recorded for CD40L and, in agreement with earlier observations (53), IL-2 (Fig.5G). To directly compare the thresholds for CCL3/4 synthesis with CCL5 release and production, we determined the respective functional avidities by ELISA (Fig.5H). The activation thresholds for CCL3 and CCL4 secretion were expectedly comparable but importantly, significantly higher than that for induced CCL5 release. In summary, at limiting antigen concentration, the IFNγ response is preceded by secretion of the chemokines CCL5 and to a certain extent also CCL1/3/4 and XCL1. These chemokines may stabilize interactions between T cells and target cells or APCs (59), may provide co-stimulatory signals (60), exert direct effector functions (27) or mediate additional functions currently under investigation.

II° CD8⁺ and CD4⁺T_E/M immunity: expression profiles and functional roles of T cell chemokines.

Given the prominence of the T_M chemokine response, we next assessed the potential modulation of chemokine expression profiles in the context of a recall response. To this end, we employed an adoptive transfer system where T cell-enriched populations from LCMV-immune mice are transferred into naïve congenic recipients that are subsequently challenged with LCMV; here, II° T_E responses derived from the transferred population can be concurrently visualized with the I° T_E response generated by the host (25). At the level of II° CD8⁺T_E responses, we observed a distinct reduction of functional diversities, i.e. in comparison to I° CD8⁺T_E, the former cells produced less IFNγ, TNFα and CD40L as well as CCL1 and XCL1; at the same time, constitutive CCL5 and induced CCL3/4/5/9/10 expression were undistinguishable (Fig.S3A/B). In contrast, II° CD4⁺T_E presented with a comparatively enhanced functional diversity as reflected in substantially larger fractions of TNFα⁺, CCL1⁺, CCL3⁺, CCL4⁺, CCL5⁺ and XCL1⁺ subsets, a chemokine profile that in fact resembles the response of I° CD8⁺T_E (Fig.S3B). We also extended these analyses to the formation of II° memory, a stage at which II° CD8⁺T_M typically display a less mature phenotype than I° CD8⁺T_M (ref.(25) and Fig.S3C). In agreement with this notion, II° CD8⁺T_M produced more CCL9/10 and less XCL1 (Fig.S3D), the only two chemokines subject to CD8⁺T_M maturation-associated modulations (cf. Fig.3C); induced CCL1/3/4/5 expression, however, was largely similar between I° and II° CD8⁺T_M. II° CD4⁺T_M, on the other hand, mostly preserved the “functional gains” accrued in the II° effector phase and exhibited enhanced CCL1/3/4/5 and XCL1 production capacity in comparison to I° CD4⁺T_M (Fig.S3D).

To determine if the major T_M-derived chemokines contribute to the regulation of II° CD8⁺T_E responses, we conducted adoptive transfer/re-challenge experiments with CD8⁺T cell-enriched donor populations under conditions of systemic chemokine neutralization. We first focused on the potential role of CCL5 and observed that its blockade resulted only in a very minor impairment of II° CD8⁺T_E responses (Fig.S3E). A similarly negligible impact on the concurrent I° CD8⁺T_E expansions (Fig.S3E) is in general agreement with the unperturbed I° CD8⁺T_E responses generated by B6.CCL5^−/− mice in the wake of an acute LCMV challenge (34, 61). Our conclusion that CCL5 is dispensable for CD8⁺T_M recall responses is further supported by experiments that demonstrated commensurate II° CD8⁺T_E reactivities of wild-type and CCL5-deficient CD8⁺T_M (not shown and ref.(61)). Somewhat surprisingly, even the combined neutralization of CCL3/4/5 or XCL1 failed to alter II° CD8⁺T_E responses (Fig.S3F). Thus, at least in the context of an acute LCMV re-challenge, none of the major CD8⁺T_M chemokines appear to contribute to the regulation of II° CD8⁺T_E responses.

Chemokine profiles of specific T cells in chronic LCMV infection (I): “effector stage”.

Challenge of naïve mice with the LCMV variant clone 13 (cl13; 2×10⁶ pfu i.v.) results in a chronic infection characterized by prolonged virus persistence and T cell exhaustion as reflected in a progressive functional deterioration of their IL-2, TNFα and IFNγ production capacity (48, 62, 63). We previously noted that an escalation of acute pathogen infection dosage depressed CCL1/3/4/5 and XCL1 synthesis potential of specific CD8⁺ but not CD4⁺T_E (34) and therefore speculated that a similar CD8⁺T cell incapacitation would also occur under conditions of chronic viral infection. Accordingly, we inoculated mice with LCMV cl13 to initiate a persistent infection and performed a first set of analyses at the height of the T_E stage (d8; control cohorts were infected with 2×10⁶ pfu LCMV Arm i.v. which does not result in viral persistence or T cell exhaustion). In agreement with its previously described tropism (64, 65), LCMV cl13 preferentially infected fibroblastic reticular cells and APC subsets but largely spared lymphocytes (Fig.6A and not shown); furthermore, CD8⁺T_E expansions after cl13 infection were diminished in the expected epitope-dependent fashion (i.e., NP₃₉₆ > GP_33/34 > GP₂₇₆), and the functional impairment of CD4⁺T_E appeared particularly pronounced (compare numbers of IA^bGP₆₆⁺ and corresponding IFNγ-producing GP₆₄-specific CD4⁺T_E) (Fig.6B/C; note that D^bGP₃₃ tetramer stains and GP₃₃ peptide stimulation cannot be directly compared since the latter method also activates K^bGP₃₄⁺ CD8⁺T_E). In regard to constitutive chemokine expression by specific CD8⁺T_E, CCL5 was the only ex vivo detectable chemokine after high-dose Arm i.v. infection (Fig.6D), presenting with a chemokine profile practically identical to that observed after lower dose i.p. Arm infection (34). In contrast, CD8⁺T_E generated in response to cl13 infection expressed, in addition to CCL5, low levels of all other T cell chemokines indicative of recent TCR activation in the presence of enhanced viral replication (Fig.6D). The latter pattern, i.e. low-level ex vivo detectable chemokine expression, was also observed for specific CD4⁺T_E yet its presence after both cl13 and Arm infection indicates that the CD4⁺T_E chemokine response is more impervious to chronic viral challenge. Indeed, this contention is supported by a comparison of induced chemokine production profiles where CD8⁺T_E in cl13-infected mice displayed impaired CCL1/3/4 and XCL1 (and elevated CCL9/10) expression in an epitope-dependent manner (NP₃₉₆ > GP_33/34 and GP₂₇₆) whereas the CD4⁺T_E chemokine response was not compromised and if anything, presented with greater CCL9/10 and XCL1 expression (Fig.6E).

Figure 6. Chemokine profiles of specific T cells in chronic LCMV infection (I): “effector stage”.

B6 mice were infected with 2×10⁶ pfu LCMV Arm or cl13 i.v. as indicated and subsequently analyzed on d8. A., LCMV-NP (nucleoprotein) expression in myeloid and lymphoid cell subsets (the APC dot plots are gated on CD3ε⁻CD19⁻NK1.1⁻ cells and feature LCMV-NP⁺ subsets in red; the histograms are gated on indicated lymphoid cell subsets and compare LCMV-NP expression in Armstrong [gray] vs. cl13 [black tracing] infection). B. & C., enumeration of specific T_E frequencies and numbers by tetramer and IFNγ stains; the arrows/values indicate the factor by which respective T_E frequencies or numbers differ between Arm- and cl13-infected mice. D., ex vivo chemokine expression by specific CD8⁺T_E (top) and CD4⁺T_E (bottom) in Arm and cl13 infection. E., quantification of induced chemokine production by epitope-specific CD8⁺ and CD4⁺T_E; asterisks indicate significant differences between Arm- and cl13-infected mice (all summary data are SEM with n=4 mice per group and are representative for results obtained in 2–3independent experiments).

Chemokine profiles of specific T cells in chronic LCMV infection (II): “memory stage”.

Extending our investigations to the d30 time point at which chronic LCMV infection is fully established, we found preferential infection of APC subsets and in particular lymphoid reticular cells identified as CD45⁻/podplanin⁺ cells (66) in cl13- but not Arm-challenged mice (Fig.7A), a numerical reduction of specific CD8⁺T_M compartments according to epitope-specificity (NP₃₉₆ > GP₃₃ > GP₂₇₆) but also a trend toward increased specific CD4⁺T_M numbers in comparison to the corresponding CD4⁺T_M pool in Arm-immune mice (Fig.7B/C). Although the effects of cl13 infection on compromised CD8⁺T_E expansions are well known, a direct enumeration of specific T_M numbers by ex vivo tetramer-staining vs. induced IFNγ production can help to clarify certain peculiarities pertaining to impaired functionalities (Fig.7D): reduction of the splenic NP₃₉₆-specific CD8⁺T_M pool by a factor of ~6 is captured in absolute numbers by both analysis modalities implying that despite the considerable pressure of the chronic infection (67), the IFNγ production capacity of surviving NP₃₉₆-specific CD8⁺T_M is reasonably well preserved. In contrast, a minor numerical decrease of D^bGP₂₇₆⁺ CD8⁺T_M is associated with 2.5-fold fewer IFNγ-producing GP₂₇₆-specific CD8⁺T_M demonstrating a considerable functional incapacitation (Fig.7D). Similar comparisons of GP₆₄-specific CD4⁺T_M further complicate this picture since in both LCMV cl13 and Arm infection, ~3.5-fold fewer IFNγ-producing than tetramer⁺ CD4⁺T_M were recovered from the spleen (Fig.7D). Thus, in the dynamic interplay between specific T_M and persisting virus that may constrain (CD8⁺T_M) or expand (CD4⁺T_M) reactive T cell pools, concurrent functional impairments, at least at the level of IFNγ production, may become “uncoupled” such that lesser numerical reduction can be associated with greater functional handicaps.

Figure 7. Chemokine profiles of specific T cells in chronic LCMV infection (II): “memory stage”.

Analyses of LCMV Arm- or cl13-infected B6 mice (2×10⁶ pfu i.v.) as conducted on d30. A., left: LCMV-NP expression (red) by APC and other cell subsets in cl13 but not Arm infection (plots gated on CD3ε⁻CD19⁻NK1.1⁻ cells). Right: LCMV-NP detection in fibroblastic reticular (CD45⁻podoplanin⁺) in cl13 but not Arm infection. B. & C., enumeration of specific T_M frequencies and numbers by tetramer and IFNγ stains; the arrows/values indicate the factor by which respective T_E frequencies or numbers differ between Arm- and cl13-infected mice (black font: significant differences, gray font: not significant). D., comparison of epitope-specific T_M numbers in the spleen as based on tetramer (ex vivo) vs. IFNγ (stimulated) stains; the arrows/values indicate significant differences emerging as a consequence of the two distinct analysis modalities used for T_M enumeration. E., cytokine and TNFSF ligand production by GP_33/4-specific CD8⁺T_M. F., ex vivo (top) and induced (bottom) chemokine expression by GP₂₇₆-specific CD8⁺T_M in Arm and cl13 infection (all plots gated on CD8⁺T cells). All summary data are SEM with n=4 mice per group, and represent results from two independent experiments.

We therefore sought to broaden the functional characterization of specific T_M populations in cl13 infection. At the level of inducible cytokines and TNFSF ligands, GP_33/4-specific CD8⁺T_M, and to a similar though lesser extent also GP₆₄-specific CD4⁺T_M, produced less TNFα and CD40L while no significant differences were noted for residual IL-2, GM-CSF, IL-3 or IL-10 expression (Fig.7E and not shown). Importantly, while constitutive chemokine expression by specific CD8⁺T_M in both infection models was restricted to CCL5, induced chemokine production in cl13-infected mice revealed the emergence of CCL3⁺, CCL4⁺ and XCL1⁺ subsets that lacked IFNγ expression; at the same time, very few CD8⁺T_M synthesized CCL1 or CCL9/10 (Fig.7F). We therefore extended and summarized these observations in two complementary ways: 1., traditional data display quantifying chemokine producing cells as a fraction of IFNγ⁺ T_M subsets demonstrated impaired chemokine responses (CCL1/3/4/5 and XCL1) in cl13 infection that were more pronounced in NP₃₉₆- as compared to GP_33/4- and GP₂₇₆-specific CD8⁺T_M; differences were les evident in the CD4⁺T_M compartment though similar to the T_E stage, GP₆₄-specific tended to make more CCL9/10 and XCL1 while their CCL3/4 expression was somewhat diminished (Fig.8A). 2., by including IFNγ-negative CCL3⁺, CCL4⁺ and XCL1⁺ subsets in our analyses and relating the functional distributions (IFNγ⁻/chemokine⁺, IFNγ⁺/chemokine⁺, IFNγ⁺/chemokine⁻) to the number of tetramer⁺ cells quantified in parallel assays, we found that the seeming numerical reduction of functional GP₂₇₆-specific CD8⁺T_M subsets as determined solely by IFNγ production largely disappears (Figs.7D & 8B, left). Thus, in the context of these complementary analyses, CD8⁺T_M in chronic LCMV infection are found to incur both functional deficits (reduced chemokine production by the IFNγ⁺ fraction) and functional gains (chemokine production in the absence of IFNγ expression). We also note that the limited expression of CCL1 including the lasting absence of a CCL1⁺/IFNγ⁻ subset (Fig.8B) may make this chemokine a particularly useful marker to ascertain the overall functional capacities of CD8⁺T_M in chronic infection. Lastly, analyses conducted on d107 after cl13 infection, i.e. at a time point when infectious virus is eliminated from many tissues and a phenotypic/functional “recovery” of CD8⁺T_M is underway (63, 68), revealed a slightly smaller functional GP₂₇₆-specific CD8⁺T_M pool with a relative reduction of the compartments expressing only CCL3/4 or XCL1 in favor of subsets co-producing IFNγ (Fig.8B, right). At the same time, the NP₃₉₆-specific CD8⁺T_M pool, likely as a consequence of its “protracted recovery”, exhibited particularly prominent fractions of IFNγ⁻/chemokine⁺ subsets (Fig.8B, right).

Figure 8. Total T_M chemokine responses and viral persistence under conditions of systemic chemokine deficiency.

A., quantification of chemokine production by specific CD8⁺ and CD4⁺T_M on d30 after Arm or cl13 infection (significant differences are indicated by asterisks). As before, these analyses depict chemokine-producing T_M subsets as a fraction of the respective IFNγ⁺ epitope-specific T_M populations. B., quantification of chemokine-producing CD8⁺T_M subsets on d30 (left; Arm and cl13) and d107 (right; cl13 only). Here, the percentage of D^bGP₂₇₆⁺ and D^bNP₃₉₆⁺ CD8⁺T_M as revealed by tetramer staining in parallel experiments (not shown) was set at 100%, the relative fraction of IFNγ− and chemokine-producing CD8⁺T_M populations was calculated accordingly, and was further stratified into IFNγ⁺/chemokine⁻ (light gray), IFNγ⁺/chemokine⁺ (medium gray), and IFNγ⁻/chemokine⁺ (dark gray) subsets (the summary data in panels A & B are SEM with n=4 mice/group analyzed in two independent experiments). C., quantification of LCMV serum titers in cl13-infected B6, B6.CCL1^−/−, CCL3^−/− and CCL5^−/− mice (n=5 mice/group and time point; data summarized from individual time course experiments).

Chronic LCMV infection under condition of systemic chemokine deficiency.

Finally, to explore potential roles for major T cell-derived chemokines in control of a chronic LCMV infection, we challenged B6, B6.CCL1^−/−, B6.CCL3^−/− and B6.CCL5^−/− mice with cl13 and quantified viral titers in the serum in bi-weekly intervals over a period of ~10 weeks. In B6 mice, high titers of infectious virus in the serum are maintained for ~1 month before onset of a slow decline and eventual virus clearance from this compartment by ~3 months (48, 63). In the present experiments, the kinetics of virus control were identical in B6, B6.CCL1^−/− and B6.CCL3^−/− mice while B6.CCL5^−/− mice, as reported in an earlier study (61), presented with a modest impairment of late LCMV control (Fig.8C). Furthermore, virus titers in various tissues as well as specific T_M frequencies, numbers, phenotypes and functionalities were not significantly different in B6, B6.CCL1^−/− and B6.CCL3^−/− mice (Fig.S4 and not shown). While these experiments cannot directly address the role of T_M-derived CCL1 or CCL3, the absence of a phenotype under conditions of systemic chemokine deficiency strongly suggests that T_M-produced CCL1 or CCL3 are unlikely to contribute to the partial control of a chronic LCMV infection.

DISCUSSION

We previously profiled the chemokine response of pathogen- and vaccine-specific T_E and defined a series of distinctive properties pertaining to its constituents (CCL1/3/4/5/9/10 and XCL1 but no other chemokines), apparent magnitude, quantitative differences between CD8⁺ and CD4⁺T_E, consistency of expression patterns across different challenge models, and to unique aspects of individual chemokine synthesis, co-expression, cooperative regulation and/or secretion (34). We now demonstrate that most of these properties are carried forward into the memory stage, identify discrete traits subject to further temporal modulation, and delineate notable adjustments of the T_M chemokine response under conditions of pathogen persistence. Collectively, these characteristics establish T_M-derived chemokines as particularly prominent, robust, and rapidly mobilized components of the recall response; in addition, we demonstrate that the analytical visualization of T_M chemokine expression patterns permits a more detailed stratification of T_M functionalities that may be correlated with differentiation status, protective capacities and potential fates.

Similar to pathogen-specific CD8⁺T_E (34), specific CD8⁺T_M evaluated 6–12 weeks after acute challenge with LCMV or LM present with a spectrum of inducible chemokines that is both circumscribed and distributed across larger and smaller subsets (CCL3/4/5: 85–95%, XCL1: 60–70%, CCL1 and CCL9/10: 15–30%). The same select chemokines are also produced by pathogen-specific CD4⁺T_M, though in a resemblance to the corresponding effector populations (34), induced chemokine synthesis is restricted to somewhat smaller subsets (CCL3/4/5: 40–50%, XCL1: 5–15%, CCL1: 20–30% and CCL9/10: <5%). These CD8⁺ and CD4⁺T_M chemokine signatures are largely preserved throughout long-term memory but with two significant exceptions: a precipitous decline of Ccl9 mRNA in aging CD8⁺T_M diminishes their corresponding protein production capacity, and a progressive increase of Xcl1 message permits larger CD8⁺T_M subsets to synthesize XCL1 with faster kinetics. Escalating XCL1 production potential, part of an intriguing functional diversification associated with long-term CD8⁺T_M maintenance (25), is especially pronounced in blood-borne CD8⁺T_M and arguably constitutes one of the largest “functional gains” bestowed on aging CD8⁺T_M populations. At the same time, however, the repertoire of CD8⁺T_M-produced chemokines is not broadened with age and remains strictly confined to the same six chemokines distinctive for effector and early memory stage.

Our conclusion about the defined, restricted and mostly invariant nature of CD8⁺ and CD4⁺T_M chemokine responses shared across different infectious disease models is contingent on the sensitivity, specificity and breadth of our assay systems. As we have detailed in our earlier reports (22, 34), the analytical power of gene array and FC technology is delimited by a variety of factors that, together with the specifics of the experimental context, inform the degree of certainty with which statements about the presence/absence of individual gene products can be made. Accordingly, while we cannot categorically rule out the possibility that minor quantities of other chemokines are produced by pathogen-specific T_M under some experimental or naturally occurring conditions, we are confident that the visualization of induced CCL1/3/4/5/9/10 and XCL1 production accurately captures their distinctive chemokine production potential. Related caveats also pertain to our attempt to position the prima facie already considerable magnitude of the CD8⁺T_M chemokine response in a larger frame of reference: to estimate the relative contribution of chemokine production to the overall size of the inducible CD8⁺T_M “functionome”, we quantified a broad though likely incomplete spectrum of specific CD8⁺T_M activities and estimate that >40% of CD8⁺T_M effector functions are dedicated to chemokine production. Though clearly only a provisional approximate, it nonetheless suggests a remarkable prominence of chemokine synthesis as part of the immediate pathogen-specific CD8⁺T_M response.

By extending our investigations to the visualization of complex chemokine co-expression patterns, we further expand the notion of “T cell polyfunctionality” and at the same time propose a potential analytical simplification as based on inducible CCL1 expression. “T cell polyfunctionality” as a diagnostically relevant concept was first introduced by R. Seder’s group who identified a simple metric (composed of the number of IFNγ⁺/TNFα⁺/IL-2⁺ specific CD4⁺T_M [“triple producers”] as well as the amount of cytokines produced by these cells [measured by the mean fluorescence intensity of cytokine stains]) that can predict the extent of immune protection in a model for L. major infection (69). While it is important to note that “T cell polyfunctionality” is a “correlate for immune protection”, i.e. it does not make claims about the precise mechanisms by which specific T cell activities contribute to immune protection, the concept has been successfully applied to other pathophysiological scenarios, CD8⁺T cells, and may further incorporate cytotoxic capacities (54, 70). Our stratification of LCMV-specific CD8⁺T_M responses according to five cytokines/TNFSF ligands and six chemokines now reveals that visualization of induced CCL1 marks a highly polyfunctional CD8⁺T_M subset that produces larger quantities of IFNγ, TNFα, CCL3 and CCL4, readily expresses CCL5 and XCL1, and is enriched for IL-2^hi and/or CD40L⁺ cells; a corresponding dissection of LCMV-specific CD4⁺T_M populations demonstrates a similar utility of CCL1 expression analyses since this subset presents with a distinct IFNγ^hi/TNFα^hi/CCL3^hi/CCL4^hi/IL-2⁺/CD40L^hi phenotype. Furthermore, and in contrast to all other chemokines, CCL1 synthesis is strictly dependent on de novo mRNA transcription, and its induction in specific CD8⁺T_M is compromised under conditions of persistent viral infection. While the simple visualization of CCL1-expressing T_M subsets does not allow for an integration of T cell activities into a “polyfunctionality index” (54), it may nevertheless serve as a singular metric for the quantification of highly polyfunctional T cells (acute infection) or T cells with residual functional reserves (chronic infection). Mechanistically, however, CCL1 appears to be dispensable as shown by the normal virus control kinetics achieved by CCL1-deficient mice in acute (34) and chronic LCMV infection.

Perhaps the most striking aspect of the CD8⁺T_M chemokine response is the constitutive expression of CCL5 by specific CD8⁺T_M in lymphatic and nonlymphoid organs alike, throughout long-term memory, and even in chronic LCMV infection. Unique among CD8⁺T_M chemokines, cytokines and TNFSF ligands, the CCL5 expression pattern is reminiscent of that reported for constituents of the perforin/granzyme pathway. However, following acute infection, GzmB expression is quickly downregulated to low or undetectable levels in LCMV-, LM- and other pathogen-specific CD8⁺T_M recovered from lymphatic tissues (25, 52, 71–73), and constitutive GzmA expression by LCMV-specific CD8⁺T_M is subject to a protracted downmodulation (25) (though perforin is expressed by practically all CD8⁺T_M in the early memory stage, its subsequent temporal regulation remains at present unknown). Together, these dynamics provide further evidence for the differential regulation of cytotoxic effector mechanisms (74–76) and establish enduring constitutive CCL5 expression by pathogen-specific CD8⁺T_M in the absence of inflammatory stimuli as a distinctive “functional signature”.

Several additional aspects pertaining to the constitutive CCL5 expression by CD8⁺T_M are noteworthy: 1., acute pathogen infections leave a conspicuous and lasting imprint that is readily discernible even without an analytical focus on specific CD8⁺T_M: in naïve mice, ~25% of CD8⁺T_MP express low levels of constitutive CCL5; in contrast, ~70% of CD8⁺T_MP in LCMV-immune mice present with a robust CCL5⁺ phenotype. In fact, in such mice CD8⁺T_MP rather than innate immune cells constitute the predominant hematopoietic source of pre-stored CCL5 (CD8⁺T_MP: ~90%, NK cells: ~10%, other [mostly CD4⁺T_MP]: <5%). 2., as shown earlier, constitutive CCL5 expression is a dynamic process that continuously replenishes intracellular CCL5 storage pools (31, 34); given the associated bioenergetics demands, the long-term preservation of a CCL5⁺ CD8⁺T_M phenotype may therefore provide crucial kinetic advantages for the recall response. 3., under steady-state conditions, however, pre-stored CCL5 is not released from CD8⁺T_M as evidenced by comparable CCL5 serum levels in LCMV-immune and naïve mice; rather, CCL5 secretion requires TCR stimulation as reported previously (32, 33). 4., yet very similar to CD8⁺T_E (34), the initial burst of TCR-triggered CCL5 secretion by CD8⁺T_M occurs independently of transcriptional/translational regulation and proceeds with extraordinary speed; we estimate that within 30min after TCR activation, individual CD8⁺T_M release as much as 1.0fg of pre-stored CCL5.

The latter kinetics also impinge on the precise coordination of the CD8⁺T_M chemokine response: due to a very low activation threshold and its rapid release, TCR-induced CCL5 secretion constitutes the first step in a cascade of CD8⁺T_M activities that otherwise require efficient protein synthesis. Here, a combination of pre-existing mRNA levels, functional avidities (EC₅₀ values) and protein production/secretion kinetics (ET₅₀ values) particular for individual effector functions shapes the coordinated elaboration of the CD8⁺T_M response. Accordingly, and in direct comparison to cytokine (IFNγ, IL-2, GM-CSF) and TNFSF ligand (TNFα, CD40L) synthesis, chemokines play a preeminent role since the production of CCL3/4 and IFNγ dominates the early stages of the CD8⁺T_M response and, together with CCL1 and XCL1, is preferentially induced under conditions of limited antigen availability. These conclusions are further reinforced by in vivo experiments in which just 2h after peptide injection, >55% of specific CD8⁺T_M produce CCL3/4 and XCL1 whereas ~45% express IFNγ and only ~25% TNFα.

Collectively, this work identifies a previously unappreciated and striking prominence of chemokine production as part of the CD8⁺T_M recall response. Nevertheless, the evidence for a decisive role of T_M-produced chemokines in the regulation of II° responses remains limited. Our own experiments, in agreement with a previous report (61), demonstrate that CCL5 is largely dispensable for the proliferative expansion of II° CD8⁺T_E, and even combined CCL3/4/5 or XCL1 blockade does not compromise the LCMV-specific recall response. To our knowledge, there are two reports documenting a pertinent role for CD8⁺T_M-derived chemokines under conditions of a II° pathogen challenge. In the first, II° LM-specific CD8⁺T_E were shown to provide immune protection through CCL3 secretion and induction of TNFα and reactive oxygen species in phagocytes (27). The evidence in the second report is more indirect. In their elegant study, Alexandre et al. conditionally depleted DCs bearing XCR1 (the sole XCL1 receptor) in pathogen-immune mice and found that II° CD8⁺T_E expansions generated in response to LM, vesicular stomatitis virus and vaccinia virus but not cytomegalovirus were compromised (28). Complementary experiments revealed that early IFNγ production by II° CD8⁺T_E was contingent on an NK cell/IFNγ-dependent induction of IL-12 and CXCL9 in XCR1⁺ DCs but while IL-12, CXCR3 or NK cell neutralization/depletion all depressed the emergence of the early IFNγ⁺ II° CD8⁺T_E phenotype, neither IL-12, CXCL9 nor NK cells were required for efficient II° CD8⁺T_E expansions (28). Other than NK cells, the only major hematopoietic XCL1 source are CD8⁺T_E/M (22, 23, 34) raising the possibility that in some model systems, II° CD8⁺T_E-derived XCL1 is the defining factor for productive engagement of XCR1⁺ DCs and the regulation of II° CD8⁺T_E proliferative responses.

In regard to the functional properties of II° CD8⁺T_E and in particular T_M generated in the wake of a recall response, earlier work has described discrete alterations in comparison to the respective I° CD8⁺T_E/M populations such as reduced IL-2 production or enhanced GzmB expression, killing capacities and TNFα induction (72, 73). Together with corresponding phenotypic alterations, these observations are consistent with a greater degree of II° CD8⁺T_E differentiation and delayed II° CD8⁺T_M maturation (25). In agreement with this interpretation, we demonstrate reduced IL-2, TNFα, CD40L, CCL1 and XCL1 production by II° CD8⁺T_E, as well as a somewhat greater induction of TNFα, CCL3/4 and CCL9/10 but decreased XCL1 expression by II° CD8⁺T_M. Interestingly, II° CD4⁺T_E avoid a corresponding attrition of “functional diversity” and display a substantial increase of CCL1/3/4/5 and XCL1 production that also extends into the II° CD4⁺T_M stage. Given the distinct CCL3/4 co-expression pattern of a major I° CD4⁺T_M subset, it might be tempting to postulate the existence of yet another “helper T cell subset” but we consider this option a distraction (5) and have focused ongoing investigations on the question if the functional skewing observed in II° CD4⁺T_E responses involves a conversion of I° CD4⁺T_M previously incapable of CCL1/3/4 production or the preferential recruitment of CCL1/3/4 expressors into the recall response (not shown).

Chronic viral infections are associated with molecular, phenotypic and functional alterations of the specific T cell compartments that are aptly described by the term “T cell exhaustion” because cardinal effector T cell activities are compromised to varying degrees (77). For our interrogation of the T_E/M chemokine signatures in chronic LCMV infection, not previously undertaken in a systematic fashion (61, 78), we therefore had to account for the fact that T_E/M dysfunctions also affect inducible IFNγ production the extent of which is furthermore regulated in an epitope-specific fashion for CD8⁺T_M (63, 67). Since concurrent analyses of MHC tetramer binding and peptide-induced functionalities are technically not feasible for murine T cells, we conducted careful complementary analyses to assess the extent of T_E/M abundance and dysfunction. While these experiments expectedly confirmed a hierarchical numerical deficit of CD8⁺T_E/M as a function of epitope-specificity in LCMV cl13- vs. high-dose Arm-infected control mice, we also noted a modest increase of both tetramer⁺ CD4⁺T cell frequencies/numbers and IFNγ producers in the “memory” but not “effector” stage. This unexpected “expansion” is reminiscent of the phenomenon of “memory inflation” described for some epitope-specific CD8⁺T_M populations in persistent infections with cytomegalovirus, herpes simplex virus-1, polyomavirus or parvovirus (79–81) and may be related to a more restricted exposure of CD4⁺T_M to persisting LCMV due to its preferential localization to fibroblastic reticular cells (FRCs) (64). Although FRCs may express MHC-II under certain inflammatory conditions (82), LCMV-infected FRCs do not (64) thus limiting the productive engagement of specific CD4⁺T_E/M populations to MHC-II-expressing LCMV-infected APCs (42). Accordingly, and despite the fact that IFNγ production by CD4⁺T_EM in both LCMV cl13 and high-dose Arm infection is partially reduced, we predicted and indeed observed an overall lesser functional impairment of CD4⁺T_E/M at the chemokine production level.

In contrast, although specific CD8⁺T_E generated immediately after a cl13 infection preserve a constitutive CCL5⁺ phenotype, their chemokine production capacity is modestly altered (reduced CCL1/3/4 and XCL1, increased CCL9/10), and the extent of impaired CCL1 induction tracks with CD8⁺T_E epitope specificity (i.e., NP₃₉₆ > GP_33/4 > GP₂₇₆). The compromised chemokine profiles, as a property of IFNγ producing subsets, are maintained if not somewhat exacerbated by the time chronic LCMV infection is fully established but importantly, are complemented by the emergence of CCL3⁺, CCL4⁺ and XCL1⁺ subsets that do not make IFNγ. In fact, inclusion of these subsets into a calculation of GP₂₇₆-specific CD8⁺T_M frequencies confirms an abundance that is roughly equivalent to the corresponding functional CD8⁺T_M subset in acute Arm infection. At later stages of cl13 infection (>3 months), a gradual “functional resurgence” of CD8⁺T_M occurs (68) and is reflected at the chemokine level by a contraction of the “CCL3/4/XCL1-only” compartment in favor of their co-expression with IFNγ. Of note and in further support of CCL1 as a marker for CD8⁺T_M with residual functionalities, CCL1-production is not observed in the absence of IFNγ co-expression in either early or late phases of cl13 infection. We therefore suggest that future functional profiling of T cells in chronic viral infection and related scenarios of prolonged antigen persistence such as cancer and autoimmunity should include, ideally in a highly multiplexed fashion, a visualization of the T cell chemokine response to account for the full presence and diversity of reactive T cell populations.

Finally, and in an echo of a seeming disconnect between the prominence of T_E and T_M chemokine responses and their limited relevance for the regulation of I° (34) and II° T_E expansions, systemic CCL1- or CCL3-deficiency also has no bearing on virus control or T cell exhaustion in chronic LCMV infection; a modest impact observed under conditions of CCL5-deficiency corroborates an earlier report (61) but leaves unclear the specific contribution of T cell-produced CCL5. As argued previously (34), an apparent absence of pronounced phenotypes in mice deficient for major T cell-produced chemokines may be grounded in redundancies of the chemokine system and thus emerge only in the context of compound chemokine mutations, or may be observed in other experimental or natural scenarios characterized by a distinctive dynamic interplay between specific T cells and persisting antigens.

In summary, the present work defines six chemokines (CCL1/3/4/5/9/10, XCL1) as a major component integral to the functional repertoire of pathogen-specific T cells in long-term memory and chronic viral infection. Unique expression patterns such as constitutive CCL5 and inducible CCL1 expression may serve targeted diagnostic purposes, and the relative dominance exerted by T cell-derived chemokines in the early recall response establishes memory T cell populations as a principal source for CCL1/3/4/5 and XCL1 in particular. How such prodigious chemokine production, which necessitates the mobilization of not inconsiderable bio-energetic resources, exactly contributes to enhanced immune protection, however, remains to be investigated in in greater detail.

KEY POINTS.

• Pathogen-specific memory T cells (T_M) are a prodigious source of chemokines.

• The chemokine expression patterns of T_M largely resemble those of effector T cells.

• Unique T_M chemokine signatures in acute/chronic infection are of diagnostic value.