Chemokine Signatures of Pathogen-Specific T Cells I: Effector T Cells

Broad survey of T cell–produced chemokines

To delineate the principal spectrum of chemokines synthesized by activated T cells, spleen cells obtained from unmanipulated wt mice were stimulated for 5 h with PMA/ionomycin and interrogated by FC for production of IFN-γ and 37 individual chemokines using a stringently validated panel of chemokine-specific Abs that reliably detect chemokine expression in appropriate positive control samples (33). In stimulated T cells, robust chemokine induction was observed for CCL3, CCL4, and XCL1 and to a lesser extent also CCL5; 1–2% of T cells synthesized CXCL2; and very small subsets produced CCL1 or CCL9/10 (Supplemental Fig. 1A–C). CCL3, CCL4, CCL5, and XCL1 production was particularly prominent in the IFN-γ⁺ T cell subset (40–80% coexpression), and despite their low frequency, CCL1- and CCL9/10–expressing T cells were also enriched in the IFN-γ⁺ population (3–4% coexpression); in contrast, no such enrichment was observed for CXCL2 (Supplemental Fig. 1A, 1C). Although T cells have previously been described as a source for these chemokines in various experimental scenarios, we note that the comprehensive nature of our screen, within the limits of specific experimental constraints and the sensitivity afforded by chemokine FC (33), can apparently rule out 30 other chemokines as potential products of highly activated T cells.

Defining the complete spectrum of chemokines produced by virus-specific CD8⁺T_E

To refine these analyses within the context of infectious diseases and to define the complete range of chemokines produced by pathogen-specific CD8⁺T_E, we first employed the established p14 chimera system to quantify chemokine mRNA and protein expression by virus-specific CD8⁺T_E with a combination of gene arrays and FC-based assays (33, 37). In brief, p14 chimeras were generated by transducing congenic B6 mice with a trace population of naive TCR-transgenic CD8⁺T cells (p14 T_N) specific for the dominant LCMV–GP_33–41 determinant; after challenge with LCMV, p14 T_E populations rapidly differentiate, expand and contribute to efficient virus control before contracting and developing into p14 T_M ∼6 wk later (35, 37, 51). At the peak of the effector phase (day 8), p14 T_E were purified, and RNA was extracted either immediately or after a 3-h in vitro TCR stimulation and processed for gene array hybridization. Overall, 10 chemokine mRNA species were detectable in p14 T_E evaluated ex vivo and/or after TCR stimulation, and their expression patterns could be allocated to three groups (Fig. 1A): 1) absence of ex vivo detectable mRNA but robust transcription after TCR stimulation (Ccl1 and Xcl1), 2) constitutive mRNA expression that significantly increased upon TCR engagement (Ccl3, Ccl4, Ccl9/10, and Cxcl10), and 3) chemokine mRNA species that were slightly downregulated by TCR activation (Ccl5, Ccl6, Ccl25, and Ccl27). A list of all murine chemokine genes and gene array IDs is found in Supplemental Fig. 2A. We also quantified chemokine mRNA expression for the known members of the related chemokine-like factor superfamily (CKLFSF) (56) (Supplemental Fig. 2B). Four out of ten Cklfsf mRNA species were detected in p14 T_E but none were increased or decreased upon TCR stimulation. Information about the biological function of CKLFSF members remains limited and is centered around the pleiotropic effects of CKLF1, which may be produced by human T cells after prolonged in vitro stimulation (57, 58). At the present stage, we have refrained from a further analysis of this gene family.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Chemokine mRNA and protein expression by virus-specific CD8⁺T_E. (A) p14 T_E (day 8) were obtained from spleens of LCMV-infected p14 chimeras, enriched to >99% purity, and processed for RNA extraction (either immediately or after 3-h anti-CD3/anti-CD28 stimulation) and gene array analysis (n = 3 individual mice). The bar diagrams display MAS5-normalized values of chemokine mRNA expression of p14 T_E analyzed ex vivo (gray bars) or after TCR stimulation (black bars); statistically significant differences are indicated by asterisks. The broken line indicates the detection threshold set at a MAS5 value of 40; coverage: 39/40 chemokines (Ccl26 not on chip). (B) p14 T_E (day 8) were analyzed for chemokine protein expression ex vivo or after 5-h stimulation with GP₃₃ peptide. Histograms are gated on splenic p14 cells (gray histograms: control stains; black tracings: indicated chemokine stains; red dots identify panels demonstrating detectable chemokine expression; representative data from three experiments with [n ≥ 3 individual mice per experiment]). (C) Summary of p14 T_E- and T_M-expressed chemokines and chemokine receptors; gray font indicates presence of mRNA in the absence of constitutive or induced protein expression. (D) Genomic organization of murine chemokine genes transcribed and translated by T cells (modified after Ref. 14). The genes for 4–6 chemokines produced by T cells (Ccl3/4/5 and Ccl9/10) are found in the MIP region on mouse chromosome 11; the Ccl1 gene is located in the MCP region but rather distantly related to other members of the MCP group, and the nonclustered Xcl1 gene is found on chromosome 1. Arrows indicate chemokine genes and their transcriptional orientation; colors identify homeostatic (green), inflammatory (red), and dual function (yellow) chemokine genes; gray arrows indicate pseudogenes. Based on our results reported here and in Ref. 33, we propose to classify the CCL3/4/5 and 9/10 as dual function chemokines rather than simply inflammatory (CCL3/4/5) or homeostatic (CCL9/10). *p < 0.05, **p < 0.01, ***p < 0.001.

Traditional gene array analyses are fraught with several limitations, including measurement of mRNA levels across entire albeit purified cell populations rather than individual cells, difficulties in directly comparing mRNA expression between different mRNA species, and the impossibility of predicting whether mRNA is in fact translated. We therefore deployed chemokine FC to the interrogation of p14 T_E and found that specific TCR stimulation with GP₃₃ peptide induced CCL3, CCL4, and CCL5 expression in practically all p14 T_E, whereas CCL1, CCL9/10, and XCL1 production was restricted to p14 T_E subsets; neither Ccl6, Ccl25, Ccl27, Cxcl2, nor Cxcl10 or any other chemokine RNA was translated. Of note, and to our knowledge not previously reported, CCL5 was also detectable directly ex vivo and is thus the only chemokine expressed in the absence of TCR activation (Fig. 1B). The robust induction of CCL1 and CCL9/10 in p14 T_E contrasts with their more limited expression in our initial T cell survey (Supplemental Fig. 1A–C). To resolve these discrepancies, we quantified CCL1 and CCL9/10 production by CD8⁺T cells from LCMV-immune mice in response to stimulation with peptide anti-CD3/anti-CD28 or PMA/ionomycin. Interestingly, >20% of IFN-γ⁺ CD8⁺T cells also synthesized CCL1 or CCL9/10 after activation with peptide or anti-CD3/anti-CD28, but fewer than 5% produced these chemokines in response to PMA/ionomycin stimulation (Supplemental Fig. 1D). The reason for the only sparse CCL1 and CCL9/10 induction after PMA/ionomycin treatment remains unclear but emphasizes important limitations associated with this widely used T cell–stimulation protocol. In summary, six mRNA species found ex vivo and/or after brief TCR engagement in virus-specific p14 T_E serve as templates for induced protein synthesis and, in the case of Ccl5, also for effective constitutive translation (Fig. 1B, 1C). The genes of four of these six chemokines (Ccl3/4/5 and 9/10) are clustered in the “MIP region” on murine chromosome 11 with an additional gene (Ccl1) immediately adjacent in the “MCP region;” the Xcl1 gene is unclustered and located on chromosome 1 (Fig. 1D).

Finally, transcriptional profiling of p14 T_E–expressed chemokine receptors revealed a prominent presence of Ccr2, Cxcr3, Cxcr4, Cx3cr1, and Ccrl1 (all of which were significantly downregulated upon activation); Ccr5 and Cxcr6 (slightly increased after stimulation), and low levels of Ccr7 that remained unaffected by TCR engagement (Fig. 1C, data not shown). It therefore appears that CCR5 is the only receptor that may sensitize p14 T_E to potential auto- or paracrine actions of T cell–produced chemokines themselves (i.e., CCL3/4/5).

Constitutive and induced chemokine expression profiles of endogenously generated LCMV-specific CD8⁺ and CD4⁺T_E

Extending our findings from the p14 chimera system to endogenously generated T_E, direct ex vivo analyses of the dominant D^bNP₃₉₆⁺ CD8⁺T_E population in LCMV-infected B6 mice demonstrated patterns comparable to p14 T_E in that constitutive chemokine expression was largely limited to CCL5. The small subsets of specific CD8⁺T_E showing weak CCL1/3/4 (but no CCL9/10 or XCL1) staining suggest that their expression, in contrast to IFN-γ and other cytokines (59), can be maintained somewhat longer after cessation of TCR activation (Fig. 2A). Constitutive CCL5 expression, in contrast, is a general feature of LCMV-specific CD8⁺T_E, as based on analyses of additional epitope-specific CD8⁺T_E and the inclusion of LCMV-infected Ccl5-deficient mice as a negative control (we also observed a slight reduction of ex vivo–detectable CCL5 in Ccl5-heterozygous mice, indicative of a modest gene-dose effect) (Fig. 2A), and appears to be in fact unique for cytokines at large because no other effector molecules readily detected in restimulated CD8⁺T_E (IFN-γ, TNF-α, IL-2, GM-CSF, CD40L) are expressed in a constitutive fashion (Supplemental Fig. 2C, 2D). Rather, CCL5 expression resembles that of constituents of the granzyme/perforin pathway (60–63) (Supplemental Fig. 2C, 2E). Similar to CD8⁺T_E, LCMV-specific CD4⁺T_E cells also contained ex vivo detectable CCL5 albeit only in a subset (∼60%) and at lower levels (Fig. 2B).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Constitutive and induced chemokine expression by LCMV-specific CD8⁺ and CD4⁺T_E. (A) Top row, Endogenously generated LCMV-specific CD8⁺T_E (day 8) were analyzed directly ex vivo by chemokine FC (plots gated on CD8⁺T cells); values indicate SEM of chemokine⁺ subsets among NP₃₉₆-specific CD8⁺T_E. Middle and bottom rows, Constitutive CCL5 expression by LCMV-specific CD8⁺T_E subsets generated by LCMV-infected B6, B6.CCL5^+/−, and B6.CCL5^−/− mice. (B) Ex vivo–detectable CCL5 expression by LCMV-specific CD4⁺T_E. The adjacent bar diagram compares the fractions (percentage) and CCL5 expression levels in geometric mean of fluorescence intensity (GMFI) of CCL5⁺-specific CD8⁺ and CD4⁺T cells; statistical differences are indicated by asterisks. For the purpose of this direct comparison, MHC class I and II tetramer stains were performed under the same experimental conditions (90-min incubation at 37°C). (C) Induced chemokine production by NP₃₉₆-specific CD8⁺ (top row) and GP₆₄-specific CD4⁺ (bottom row) T cells as determined after 5-h in vitro peptide stimulation culture. (D) Left, Induced CCL1 expression following NP₃₉₆ peptide stimulation of day 8 spleen cells from LCMV-infected B6 and B6.CCL1^−/− mice. Right, Chemokine coexpression by NP₃₉₆-specific CD8⁺T_E (plots gated on IFN-γ⁺ CD8⁺T_E). (E) Summary of induced chemokine expression by LCMV-specific T_E subsets stratified according to epitope specificity; their restriction elements, relative size (immunodominance), and functional avidities (peptide concentration required to induce IFN-γ production in 50% of a given epitope-specific population) are indicated. Significant differences between chemokine-expressing CD8⁺ and CD4⁺T_E subsets are indicated by asterisks. (F) The composition of the NP₃₉₆-specific CD8⁺T_E response (day 8) was assessed by quantification of subsets expressing individual constitutive (GzmA/B and perforin) or inducible (all other, including CCL5) effector activities, and the pie chart depicts the sum and relative distribution thereof. Unless noted otherwise, data (SEM) were generated with splenic T cell populations and are representative of two to five experiments comprising groups of three to five individual mice. *p < 0.05, **p < 0.01, ***p < 0.001.

Upon in vitro restimulation, endogenously generated CD8⁺T_E rapidly synthesized the same six chemokines induced in p14 T_E but no other chemokines (Fig. 2C, data not shown). Accordingly, CCL1, CCL9/10, and XCL1 production by specific CD8⁺T_E was mostly restricted to a subset of IFN-γ⁺ cells, whereas CCL3/4/5 were produced by virtually all epitope-specific CD8⁺T_E. The patterns of induced chemokine synthesis further indicated the existence of particularly potent CD8⁺T_E populations, as demonstrated by the coexpression of high CCL3/4/5 levels and the relative restriction of CCL1 production to a subset of XCL1⁺ CD8⁺T_E (Fig. 2D). To ascertain whether “polyfunctionality” at the level of inducible chemokine production is a property of defined CD8⁺T_E subsets, we differentiated specific CD8⁺T_E according to CD127 and KLRG1 expression, a distinction that captures CD127⁻KLRG1⁺ short-lived effector cells and CD127⁺KLRG1⁻ memory precursor effector cells (MPECs) (38). As shown in Supplemental Fig. 2F, 2G, MPECs in comparison with short-lived effector cells were enriched for inducible CCL1, CCL9/10, and, in particular, XCL1 expression. This is broadly consistent with the chemokine profiles of specific CD8⁺T_M that overall exhibit greater CCL9/10 and XCL1 production capacity (37, 47). In regard to LCMV-specific CD4⁺T_E, their chemokine signatures were qualitatively similar, but induced CCL3/4/5 production was confined to a subset (∼60%), and very few cells produced CCL9/10 or XCL1 (Fig. 2C). Fig. 2E summarizes the above observations by displaying the fraction of endogenously generated chemokine⁺ LCMV-specific T_E stratified according to MHC restriction elements. Because the individual epitope-specific T_E populations not only differed according to immunodominance but also activation threshold (Fig. 2E), our findings establish that induced chemokine production is independent of mouse strain, immunodominant determinants, and functional avidities but quantitatively different in specific CD8⁺ and CD4⁺T_E.

To provide a rough estimate for the relative contribution of chemokine production to the totality of quantifiable CD8⁺T_E functionalities, we determined the respective fractions of NP₃₉₆-specific CD8⁺T_E capable of individual chemokine, cytokine, and TNFSF ligand synthesis, constitutive GzmA/B and perforin expression, and degranulation (see Materials and Methods for details); according to this estimate, >40% of the CD8⁺T_E response is in fact dedicated to chemokine production (Fig. 2F).

Similar chemokine expression profiles of LCMV-, VSV-, L. monocytogenes–, and vaccine-specific CD8⁺ and CD4⁺T_E

The regulation of pathogen-specific CD8⁺ and CD4⁺T immunity generated in response to VSV or L. monocytogenes shares many cardinal properties with the LCMV system (41, 44, 46, 64–67), yet the distinct biology of these pathogens may have an impact on aspects of the T cell chemokine response. In contrast to the noncytopathic arenavirus and natural murine pathogen LCMV, VSV is an abortively replicating cytopathic virus that causes a polio- or rabies-like neurotropic infection in immunodeficient mice (68). Similar to other pathogenic bacteria, such as Mycobacteria, Salmonella, Rickettsia, and Chlamydia, L. monocytogenes is a facultative intracellular bacterium, and the model of murine listeriosis constitutes one of the best-characterized experimental systems for bacterial infection (69). Acute infection with L. monocytogenes (rLM-OVA for induction of a traceable specific CD8⁺T_E population) or VSV generated specific CD8⁺T_E with constitutive CCL5 and inducible CCL3/4/5/9/10 and XCL1 expression akin to those found in LCMV-specific CD8⁺T_E; similarly, L. monocytogenes– and VSV-specific CD4⁺T_E presented with chemokine production profile resembling LCMV-specific CD4⁺T_E (Fig. 3A, 3B, Supplemental Fig. 3A). The considerable uniformity of chemokine signatures by T cells specific for three disparate pathogens therefore identifies fundamental functional attributes of the pathogen-specific T cell response at large (Figs. 2E, 3C).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Constitutive and induced chemokine expression by L. monocytogenes–, VSV-, and vaccine-specific CD8⁺ and CD4⁺T_E. (A) Induced chemokine expression by specific CD8⁺ and CD4⁺T_E (day 8) following challenge with rLM-OVA (data displayed as in Fig. 2C). (B) Constitutive CCL5 expression by rLM-OVA₂₅₇–specific CD8⁺T_E (day 8). (C) Summary of induced chemokine production by rLM-OVA– and VSV-specific CD8⁺ and CD4⁺T_E (restriction elements are indicated); asterisks denote significant differences between CD8⁺ and CD4⁺T_E specific for the same pathogen. (D) Ex vivo CCL5 expression by vaccine-specific CD8⁺ and CD4⁺T_E (day 7 after vaccination as explained below); dot plots are gated on blood-borne CD8⁺ or CD4⁺T cells as indicated. The two-tone dot plot is gated on both total CD4⁺T cells (gray) and I-A^b2WS1⁺ CD4⁺T_E (black). (E and F) Induced chemokine profiles of specific CD8⁺ and CD4⁺T_E generated by combined TLR/CD40 vaccination. Mice were challenged with OVA/anti-CD40/poly(I:C) (CD8⁺ vaccination) or 2WS1 peptide/anti-CD40/poly(I:C) (CD4⁺ vaccination) as detailed in Materials and Methods and analyzed 7 d later; asterisks indicate differences between CD8⁺ and CD4⁺T_E. All data (SEM) were generated with splenic T cell populations unless noted otherwise and are representative of two to five experiments comprising groups of three to five individual mice. *p < 0.05, **p < 0.01, ***p < 0.001.

Nevertheless, we noted some quantitative differences associated with the use of different infection protocols, and to ascertain whether the degree of infection-associated inflammation could modulate T_E chemokine production profiles in a given model system, we infected B6 mice with escalating doses of rLM-OVA (3 × 10²–3 × 0⁴ CFU). As expected, an increase of bacterial dose heightened early inflammation as determined by serum IFN-γ levels, but the numbers of OVA₂₅₇-specific CD8⁺T_E as well as their ex vivo CCL5 expression levels peaked postinfection with 3 × 10³ CFU rLM-OVA and declined at higher infection doses; in contrast, LLO₁₉₀-specific CD4⁺T_E numbers steadily rose with escalating challenge doses (Supplemental Fig. 3B). Although the fraction of IL-2–, TNF-α–, and CD40L-producing specific CD8⁺ and CD4⁺T_E remained impervious to bacterial challenge dose, induced chemokine synthesis by OVA₂₅₇-specific CD8⁺T_E was compromised by infection with higher rLM-OVA titers, especially the production of CCL1 and XCL1 and to a lesser extent also CCL3/4/5; the chemokine expression profiles of LLO₁₉₀-specific CD4⁺T_E, however, remained largely unaffected by the different challenge protocols (Supplemental Fig. 3C, 3D). These observations suggest that chemokine production by CD8⁺ but not CD4⁺T_E may be partially impaired under conditions of chronic infection, and we have further pursued this question in the accompanying report by Davenport et al. (47).

We also extended our delineation of chemokine expression profiles to vaccine-specific CD8⁺ and CD4⁺T_E. Using a strategy for induction of protective T cell immunity by combined TLR/CD40 vaccination [i.e., the immunization with whole proteins or peptides in conjunction with poly(I:C) and anti-CD40 administration (43, 70, 71)], we found that vaccine-induced CD8⁺ and CD4⁺T_E were remarkably similar to the respective pathogen-specific T_E populations at the level of constitutive (CCL5) and induced chemokine production capacity (Fig. 3D–F). Both effective vaccines and different infectious pathogens therefore elicit essentially the same T_E chemokine response that is quantitatively adjusted according to the particular conditions of T cell priming.

Finally, small subsets of both LCMV- and L. monocytogenes–specific CD4⁺T_E have been described to exhibit a T_H2 phenotype (72, 73). Although the existence of specific IL-4–producing CD4⁺T_E in these model systems has been contested by others (67, 74), the description of CXCL2 production as a characteristic for in vitro–generated T_H2 cells (24, 75) permits an analysis of T_H2 functionality at the chemokine level. Indeed, primary murine T cells expressed CXCL2 after polyclonal activation preferentially under exclusion of IFN-γ (Supplemental Fig. 1A), and a very small subset of LCMV- but not rLM-OVA–specific CD4⁺T_E produced CXCL2 (Supplemental Fig. 3E). However, given the clearly predominant T_H1 phenotype of LCMV- and L. monocytogenes–specific CD4⁺T cells (46, 67, 74), we have not pursued a further characterization of T_H2 chemokines in these model systems.

Delayed acquisition of CCL5 production capacity by CD8⁺T_E

The acquisition of defined effector functions constitutes a hallmark of primary T_E differentiation, and the detailed work by A. Krensky’s (76–78) group has identified an unusual property of T cell–produced CCL5, namely its comparatively late synthesis only after 3–5 d of TCR stimulation as a consequence of regulation through the transcription factor KLF13. To elucidate these dynamics for pathogen-specific T cells, we compared the regulation of CCL5 expression with that of principal effector molecules (GzmB and IFN-γ) during the transition from naive to early effector stage of developing p14 T_E. Assayed over a 72-h period in vitro, the rapid and progressive induction of IFN-γ and slightly delayed GzmB synthesis as a function of cell division contrasted with a lack of constitutive and only minimal inducible CCL5 expression (Fig. 4A, 4B). Similarly, within the first 60 h after in vivo challenge, p14 T_E remained CCL5-negative and constitutive CCL5 expression by ∼50% of p14 T_E or endogenously generated CD8⁺T_E became discernible only on day 5 after LCMV infection; by day 8, however, practically all LCMV-specific CD8⁺T_E had acquired a CCL5⁺/GzmB⁺ phenotype (Fig. 4C, 4D). The protracted dynamics of CCL5 expression are indeed unique, as they differed not only from GzmB and IFN-γ but also GzmA and all other inducible CD8⁺T_E functionalities evaluated (CCL1/3/4, XCL1, IL-2, GM-SF, TNF-α, and CD40L; data not shown); accordingly, constitutive CCL5 protein expression may serve as a novel functional marker for mature Ag-experienced CD8⁺T_E.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

Expression and acquisition kinetics of CD8⁺T_E effector molecules. (A) Constitutive (gray histograms) and induced (black tracings) CCL5, GzmB, and IFN-γ expression levels by naive CD44^lo p14 cells (p14 T_N; note that the functionality of p14 T_N is restricted in this study to limited IFN-γ production). (B) CCL5, GzmB, and IFN-γ expression as a function of early in vitro p14 T_E proliferation. Dot plots are gated on p14 T cells analyzed directly after 72-h stimulation culture (no restimulation) or following GP₃₃ peptide restimulation in the presence of BFA during the final 5 h of culture as indicated. The diagrams on the right summarize the individual expression patterns as a function of p14 CFSE dilution (generation #0, no division). (C) Acquisition of constitutive CCL5 expression by p14 CD8⁺T_E in vivo was analyzed 60 h after adoptive transfer of CFSE-labeled p14 cells and LCMV challenge (top dot plot) or in p14 chimeras on day 5 postinfection (bottom plot gated on blood-borne CCL5^−/− [gray] and wt [black] p14 T_E); the adjacent diagrams depict the emergence of constitutive CCL5 expression by developing CD8⁺T_E in the p14 chimera system (middle) and LCMV-infected B6 mice (right diagram, fraction of CCL5⁺ T cells among total [gray] and D^bNP₃₉₆⁺ [black] CD8⁺T cells). (D) Left, GzmA and GzmB expression by specific CD8⁺T_E (day 8) analyzed ex vivo (top) or after 5-h restimulation culture (bottom); all dot plots were gated on splenic CD8⁺T cells. Right, Constitutive CCL5, GzmA, and GzmB expression by LCMV-specific CD8⁺T_E (the small subset of CCL5- and GzmA/B–negative CD8⁺T cell subset corresponds to the CD44^lo naive CD8⁺T cell fraction, data not shown). All data are representative of two to three experiments comprising groups of three or more individual mice (unless noted otherwise, all data obtained with splenic cell populations).

Constitutive coexpression and subcellular localization of CCL5 and granzymes in antiviral CD8⁺T_E

The precise subcellular localization of CCL5 remains a matter of controversy. In humans, a reported preferential association with the content of cytolytic granules (GzmA, perforin, granulysin) (17, 79) contrasts with the identification of a unique subcellular CCL5 compartment (80), and the frequent use of T cell clones or blasts, the differential regulation of cytolytic effector gene and protein expression in primary murine CTL (63, 81–83), and the previously reported absence of constitutive CCL5 expression by mouse CD8⁺ memory phenotype T cells, in particular (84, 85), further complicate resolution of this issue. Our direct ex vivo FC analyses of LCMV-specific CD8⁺T_E now demonstrate a clear association of CCL5 and GzmB expression, whereas GzmA, as reported previously by us [and also similar to influenza-specific CD8⁺T_E (37, 63)], was expressed by only ∼60% of the CD8⁺T_E population (Fig. 4D). These observations indicate that CCL5 colocalization studies in murine CD8⁺T_E should be extended beyond the visualization of GzmA. Accordingly, confocal microscopy revealed the existence of multiple discrete vesicles that contained either GzmA, GzmB, or both (Fig. 5, rows 1, 2); in contrast, CCL5⁺ vesicles appeared mostly devoid of GzmA/B (Fig. 5, rows 3–5) and thus presented with an expression pattern reminiscent of the subcellular CCL5 localization in primary human CD8⁺ memory phenotype T cells (80). Nevertheless, we did observe polarization and coalescence of GzmA/B⁺ and CCL5⁺ vesicles in some cells, perhaps a result of recent CD8⁺T_E activation and an indication that these effector molecules are likely cosecreted (Fig. 5, row 6, 7). The overall distribution of granzymes and CCL5 across individual vesicles in CD8⁺T_E is therefore somewhat heterogeneous, a conclusion also reported for the subcellular expression patterns of CCL5, perforin, and granulysin in human CD8⁺T cells (79).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Subcellular localization of GzmA, GzmB, and CCL5 in CD8⁺T_E. Blood-borne CD8⁺T_E (day 8) were stained with anti-GzmA–FITC, anti-GzmB–allophycocyanin, and anti-CCL5/anti-goat–Cy3, and sorted GzmA⁺ subsets were analyzed by confocal microscopy as detailed in Materials and Methods (GzmA⁻ subsets were used as a negative staining control). (Rows 1 and 2) Subcellular GzmA and GzmB localization in two different cells; (rows 2–5) same cell analyzed for GzmA, GzmB, and/or CCL5 colocalization; (row 6) GzmB and CCL5 expression in the sorted GzmA⁻ subset; (row 7) example for partial polarization and coalescence of intracellular GzmA/B and CCL5 stores in another CD8⁺T_E (representative images are taken from one of three experiments conducted with two to three mice each). Original magnification ×63 with 1.9 or 5.95 digital zoom.

The precise molecular mechanisms underpinning constitutive CCL5 expression by CD8⁺T_E, unique among the cytokines/chemokines, remain unclear. Given the general similarity between CD8⁺T and NK cells (86) and, more specifically, the chemokine production profiles largely shared between CD8⁺T_E and NK cells (33), we considered the proposal that constitutive CCL5 expression by human NK cells is dependent on the JNK pathway (87). As shown in Supplemental Fig. 3F, 3G, however, Jnk1^−/− and Jnk2^−/− mice, both of which readily control an acute LCMV infection (88), did not present with any abnormalities at the level of constitutively expressed CCL5 in NK cells or virus-specific CD8⁺ or CD4⁺T_E.

Kinetics of chemokine synthesis by virus-specific CD8⁺T_E

The elaboration of diverse T cell effector functions is a coordinated event that integrates spatial and temporal constraints with potentially different activation requirements. To determine the velocity of chemokine production by specific CD8⁺T_E, p14 T_E were restimulated for 0–5 h with GP₃₃ peptide in the presence or absence of transcriptional (ActD), translational (CHX), and/or protein secretion inhibitors (BFA) and analyzed for intracellular chemokine content by FC using IFN-γ production as a reference. Induced IFN-γ, CCL3, and CCL4 expression became detectable after as little as 30 min of stimulation and reached a maximum after 4–5 h. The synthesis of these proteins was sufficiently robust to allow detection of intracellular IFN-γ and CCL3/4, even in the absence of BFA (Fig. 6A–C, panels 1, 2). Given the presence of ex vivo–detectable mRNA species for CCL3/4 and IFN-γ (Fig.1A, Supplemental Fig. 2C), protein synthesis, although reduced, was still observed under conditions of transcriptional blockade in the presence but not absence of BFA (Fig. 6A–C, panels 3, 4). The fact that protein synthesis increased over time in a homogeneous fashion in all p14 T_E (data not shown) suggests that the constitutive IFN-γ and CCL3/4 message is evenly distributed among individual cells rather than preferentially allocated to a particular p14 subset. As expected, inhibition of translation or combined transcriptional/translational blockade completely prevented the accumulation of IFN-γ and CCL3/4 proteins (Fig. 6A–C, panels 5–8).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

Regulation and kinetics of chemokine production by p14 CD8⁺T_E. (A–D) Spleen cells from LCMV-infected p14 chimeras (day 8) were cultured for indicated time periods in the presence (closed symbols) or absence (open symbols) of GP₃₃ peptide and indicated transcriptional (ActD), translational (CHX), and/or protein transport (BFA) inhibitors. Graphs depict the geometric mean of fluorescence intensity (GMFI) of IFN-γ, CCL3, CCL4, and CCL5 expression by p14 T_E as a function of culture time and inhibitor presence or absence (panels depicting traditional stimulation conditions [i.e., peptide plus BFA] are shaded in gray). To better compare the kinetic regulation of cytokine and chemokine production, the respective GMFI values were normalized (IFN-γ: the GMFI of rat isotype control stains was subtracted from all corresponding IFN-γ GMFI values and the resulting values of BFA/GP₃₃ cultures for the t = 5.0 h time point [A.1] were set at 100%; CCL3 and CCL4: similar normalization performed by subtraction GMFI values of goat IgG stains from corresponding CCL3 or CCL4 GMFI values; CCL5: ex vivo goat IgG control stain GMFI was subtracted from all CCL5 GMFI values, and the resulting normalized ex vivo CCL5 values [B.1] were set at 100%). The sigmoidal curve fit is based on optimal fits determined by nonlinear regression analyses of samples containing additional time points (n = 3 mice per group); data were from three similar experiments.

The kinetics of intracellular CCL5 accumulation were predictably more complex because TCR-induced release of prestored CCL5 and initiation of protein neosynthesis occurred in parallel. In fact, a modest loss of CCL5 observed 30 min after TCR stimulation was quickly compensated by a pronounced increase of intracellular CCL5 in cultures containing BFA, a pattern that contrasted with rapid if only partial CCL5 depletion in the absence of BFA (Fig. 6D, panels 1, 2). It is therefore worth mentioning that the release of newly synthesized CCL5, in contrast to the release of prestored CCL5 (80), was largely inhibited by BFA. Furthermore, the kinetics of intracellular CCL5 accumulation were comparable in the presence and absence of transcriptional inhibition (Fig. 6D, panels 1, 3), consistent with our observation that TCR stimulation does not increase the level of CCL5 mRNA (Fig. 1A), and an early increase of intracellular CCL5 in cultures without TCR stimulation and translational blockade emphasizes that maintenance of constitutive CCL5 expression by T cells is an active process; the eventual decline of CCL5 expression at later time points is likely due to degradation because we did not observe CCL5 secretion by unstimulated T cells (Fig. 6D, panels 2, 4, data not shown). Interestingly, upon TCR stimulation in the presence of translational or combined transcriptional/translational blockade, ∼2/3 of prestored CCL5 were released within 30–60 min; additional depletion of CCL5 stores occurred with slower kinetics and was inhibited by BFA (Figs. 6D, panels 5–8, 7A).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 7.

Rapid surface translocation and secretion of prestored CCL5 by antiviral CD8⁺T_E. (A) Spleen cells from day 8 p14 chimeras were preincubated with CHX prior to initiation of TCR stimulation by addition of GP₃₃ peptide and subsequent analysis of p14 GzmB and CCL5 content 0–5 h later. GzmB and CCL5 expression levels in geometric mean of fluorescence intensity (GMFI) were normalized such that t = 0 h levels correspond to 100% and the GMFI of respective control stains are set at to 0%. Left, Kinetics of prestored CCL5 release in the presence versus absence of BFA. Right, Immediate depletion of CCL5 stores versus delayed GzmB release. At t = 0.5 h, ∼2/3 of CCL5 but <10% of GzmB stores are emptied (n = 3 mice) for one to three independent experiments. (B) Conjugate formation between purified splenic p14 T_E (CD90.1) and GP₃₃ peptide-coated or uncoated EL4 cells (CD90.2) as well as CCL5 surface expression (sCCL5) were assessed 20 min after initiation of coculture as detailed in Materials and Methods. Four populations were distinguished according to CD90.1/2 expression levels and forward scatter (FSC) properties (cell size) as follows: (B1) EL4 cells; (B2) EL4:p14 T_E conjugates; (B3) p14 T_E expressing low levels of CD90.2, likely acquired by trogocytosis; and (B4) p14 T_E. Note the weak but distinctive sCCL5 staining detectable among specific (black tracings) but not unspecific (gray histograms) EL4:p14 T_E conjugates (population 2). (C) Conjugation assays were performed as above and analyzed by confocal microscopy to visualize sCCL5 (green) and CD90.2 (red) expression (panels C1 and C3 and panels C2 and C4 are identical with CD90.2 signals removed from panels 1 and 2 to better visualize sCCL5 expression). Note the “blebbing” of the EL4 cell in panel C5 consistent with the induction of apoptosis; panel C6 features a magnification of the p14 T_E in panel 5 to demonstrate IS (white arrow) and antipolar (gray arrow) localization of sCCL5. (D) Spleen cells from LCMV-infected B6 mice (day 8) were preincubated with CHX and stimulated with NP₃₉₆ peptide (no BFA), and CCL5 in the supernatant was quantified by ELISA. To calculate the amount of prestored CCL5 secreted by individual NP₃₉₆-specific CD8⁺T_E, complementary FACS analyses were performed to determine the absolute numbers of cultured specific CD8⁺T_E. Further, the amount of CCL5 secreted in the absence of TCR stimulation was subtracted from stimulated samples at all time points. Note that after 30 min, ∼60% of total CCL5 is already secreted; at 1 h, it is ∼90%. Data are representative for two to four experiments conducted with three individual mice each. *p < 0.05, **p < 0.01.

Rapid CCL5 surface translocation and secretion by virus-specific CD8⁺T_E

To interrogate the remarkably fast CCL5 release kinetics by CD8⁺T_E in more detail, we compared the concurrent depletion of prestored CCL5 and GzmB from TCR-stimulated p14 T_E in the presence of CHX. In this study, the near-instantaneous release of CCL5 contrasted with a ∼30-min lag period before intracellular GzmB began to decline; yet, the subsequent loss of GzmB proceeded so rapidly that the relative extent of CCL5 and GzmB depletion was comparable by 1 h after initiation of T cell stimulation (Fig. 7A). Overall, the kinetic differences between CCL5 and GzmB depletion, as well as the differential sensitivity of constitutive versus induced CCL5 expression/release to BFA, corresponds well with the heterogeneous distribution of CCL5 and GzmB across different subcellular compartments as shown in Fig. 5.

To better visualize the earliest events of TCR-induced CCL5 release, we performed an in vitro conjugation assay using purified p14 T_E (CD90.1) and congenic (CD90.2) GP₃₃ peptide-coated versus uncoated EL4 target cells (Fig. 7B, 7C). Within 20 min after TCR engagement and thus before initiation of CCL5 neosynthesis, CCL5 was translocated to the cell surface, deposited preferentially at the interface between p14 T_E:EL4 conjugates, and the engagement of EL4 cells was readily demonstrated by the focused redistribution of CD90.2 around the immunological synapse (IS) (89); formation of “unspecific” conjugates (i.e., in the absence of GP₃₃ peptide) was not associated with cell surface exposure of CCL5 or a clustering of EL4-expressed CD90.2 (Fig. 7B, 7C). Our data therefore support the notion that mobilization of intracellular CCL5 stores is primarily directed toward the IS, similar to the polarization reported for polyclonally activated human CD8⁺T cell blasts and clones (80, 90); yet, they apparently differ from results of another study in which the de novo synthesis of CCL5 by in vitro–generated murine CD4⁺T cell blasts resulted in multidirectional release of this chemokine (i.e., an early [2-h] association of intracellular chemokine stores with the IS followed by a later [4-h] distribution) in multiple compartments throughout the cytoplasm (91). In agreement with the latter report, however, we found that p14 T_E on occasion presented low amounts of antipolar surface CCL5 (Fig. 7C, panel 6), and it is tempting to speculate that the heterogenous subcellular distribution of prestored CCL5 is related to the reported association with distinct trafficking proteins that mediate a multidirectional versus focused chemokine release (90, 91).

The rapid accumulation of CCL5 within the IS, a defined space with an estimated volume of 0.5–5.0 × 10⁻¹⁶ l (91, 92), suggests that local CCL5 concentrations may temporarily reach “supraphysiological” levels. The latter term describes multiple observations that in vitro exposure of cells to oligomeric CCL5 in excess of ∼1 μM promotes receptor-independent binding to surface glycosaminoglycans and generalized activation that, depending on the cell type under study, results in cellular proliferation and differentiation as well as enhanced survival, CTL activity, and cytokine and chemokine release (93–99). Whether these effects have an in vivo correlate has remained doubtful, and to provide a more quantitative estimate, we combined FC and ELISA assays conducted in the presence of CHX to calculate the amount of prestored CCL5 that is secreted by an individual LCMV-specific CD8⁺T_E. The rapid increase of CCL5 in the ELISA culture supernatant (Fig. 7D) mirrors the loss of intracellular CCL5 in Fig. 7A and corresponds to ∼0.5 fg per CD8⁺T_E released in the first 30 min after TCR triggering, an amount that could in principle result in a CCL5 concentration of >100 μM within the confines of the IS. Even if these calculations constitute a gross overestimate due to incomplete CCL5 release, multiple CD8⁺T_E:target cell contact sites (100), limited spatial constraints, and/or rapid diffusion, it would appear likely that CCL5 concentrations of >1 μM could be achieved in a spatially and temporally confined fashion in vivo and therefore might contribute to target cell activation in a receptor-independent fashion.

Induced CCL3/4/5 colocalization and cosecretion by virus-specific CD8⁺T_E

The coproduction of CCL3/4/5 by stimulated CD8⁺T_E, as evidenced by the “diagonal” event distribution in FC plots (Fig. 2D), suggests a tight association and potential colocalization of these chemokines following CD8⁺T_E activation. When analyzed by confocal microscopy under these conditions, CCL3/4/5 as well as GzmB and IFN-γ indeed tended to cluster in a single defined location close to the plasma membrane and in immediate proximity of the IS (Supplemental Fig. 4A, data not shown). The colocalization of induced CCL3/4/5 expression, in particular, might provide a basis for the joint release of these chemokines bound to sulfated proteoglycans as described for HIV-specific CD8⁺T cell clones (17). Moreover, human PBMC stimulated with PMA secrete CCL3/4 as heterodimeric complexes (101), and up to half of the CCL3/4 content in medium conditioned with lymph node cells from recently immunized mice is in a state of hetero-oligomerization (29). To determine the extent of heterologous chemokine complex formation in the context of a virus-specific T cell response, splenic CD8⁺T_E were restimulated for 5 h, the supernatants were collected and preabsorbed with anti-CCL3, anti-CCL4, anti-CCL5, or control Abs prior to quantitation of CCL3/4/5 by ELISA (Supplemental Fig. 4B). Although we noted some variability in these cross-absorption experiments, we have previously confirmed the specificity of Abs used for preabsorption (33) and therefore can conclude that the biologically active form of CD8⁺T_E–secreted CCL3/4/5 consists in part of hetero-oligomeric complexes. Beyond the apparently intimate coordination of CCL3/4/5 activities, this finding also emphasizes important limitations for the interpretation of any experiments that employ Ab-mediated in vivo neutralization of these chemokines.

Specific antiviral T cell immunity in the absence of systemic chemokine deficiencies

Despite their prominence among T cell–produced effector molecules, CCL3/4/5 are apparently dispensable for the control of an acute LCMV infection (102–104). Accordingly, we found that LCMV-challenged B6.CCL3^−/− and B6.CCL5^−/− mice generated a diversified virus-specific T cell response and controlled the infection with kinetics comparable to B6 wt mice (Supplemental Fig. 4C, 4D; similarly, B6.CCL1^−/− mice mounted normal T_E responses and readily controlled LCMV; Fig. 2D, data not shown). Because CCL5 may exert direct apoptotic functions (105) and in vitro degranulation and killing by CCL5^−/− CD8⁺T cells in the context of a chronic LCMV infection is reportedly impaired (104), we also examined the in vivo–killing kinetics by LCMV-specific CD8⁺T_E in the absence of CCL5. As shown in Supplemental Fig. 4E, however, in vivo target cell killing proceeded with the same rapid kinetics in wt and CCL5-deficient mice, and the lack of a role for CCL5 in this assay was further confirmed by treatment with a CCL5-neutralizing Ab.

However, a careful analysis of T_E chemokine expression profiles in B6.CCL3^−/− and B6.CCL5^−/− mice demonstrated some unanticipated quantitative differences. In comparison with B6 mice, B6.CCL3^−/− but not B6.CCL5^−/− mice generated a slightly reduced antiviral CD8⁺ but increased CD4⁺T_E response (Supplemental Fig. 4C, 4D). Furthermore, in B6.CCL3^−/− mice, CCL4 and CCL5 production by specific CD8⁺ and CD4⁺T_E was significantly diminished in comparison with B6 mice, and a somewhat lesser reduction of induced CCL3 and CCL4 expression was also observed for B6.CCL5^−/− mice (Fig. 8A, 8B, Supplemental Fig. 4F). These differences were even more pronounced when chemokine production by specific CD8⁺T_E was quantified in vivo following a 1-h peptide inoculation (Fig. 8C, data not shown). Lastly, both B6.CCL3^−/− and B6.CCL5^−/− mice also exhibited a modest but significant impairment of CCL9/10 production capacity by GP₃₃-specific CD8⁺T_E interrogated in vitro (Supplemental Fig. 4F). Altogether, functional impairments extending beyond specific chemokine gene deficiencies may be related to the cooperative regulation of chemokine expression/secretion and/or to artifacts arising in the mutant mice due to the proximity of respective gene loci on chromosome 11 (Fig. 1D); they will also need to be considered for any interpretation of relevant observations made with CCL3- or CCL5-deficient mice.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 8.

Impact of CCL3 or CCL5 deficiency on related chemokine production capacity by antiviral T_E. (A) Induced CCL3/4/5 production by GP₃₃-specific CD8⁺ and GP₆₄-specific CD4⁺T_E analyzed on day 8 after LCMV challenge of B6, B6.CCL3^−/−, and B6.CCL5^−/− mice (all plots gated on CD8⁺ or CD4⁺T cells). (B) CCL3/4/5 content of stimulated GP₃₃-specific CD8⁺ and GP₆₄-specific CD4⁺T_E in wt and chemokine-deficient mice (n = 3 per group) for one of three similar experiments; asterisks indicate significant differences between B6 and mutant mice by one-way ANOVA. (C) One hour in vivo CD8⁺T_E activation assays were performed on day 8 after LCMV infection of B6 and B6.CCL5^−/− mice by i.v. injection of GP₃₃ peptide, as detailed in Materials and Methods (saline injection was negative control). Note the reduced CCL3 induction in CCL5-deficient D^bGP₃₃⁺CD8⁺T_E (n = 3 mice per group) for one of two experiments; all plots were gated on splenic CD8⁺T cells; GP₃₃ peptide also activates K^bGP₃₄⁺CD8⁺T_E, accounting for the D^bGP₃₃ tetramer-negative population in the LR plot quadrants). *p < 0.05, **p < 0.01, ***p < 0.001.

Last, fatal lymphocytic choriomeningitis following i.c. LCMV infection of immunocompetent mice is contingent on a potent virus-specific CD8⁺T_E population that may recruit pathogenic myelomonocytic cells into the CNS through secretion of CCL3/4/5 (27, 106, 107). Prior work with CCL3- and CCR5-deficient mice, however, has demonstrated a normal lethal phenotype after i.c. LCMV challenge (102, 103), leaving the possibility that CCL5 may uniquely contribute to the fatal disease course. In this study, we used a set of chemokine- and chemokine receptor-deficient mice to assess whether the lack of any CD8T_E-produced chemokines delayed or prevented lethal choriomeningitis. Specifically, we employed CCL1^−/−, CCL3^−/−, CCR5^−/− (CCL3/4/5 receptor), CCL5^−/−, CCR1^−/− (CCL3/5/9/10 receptor), and CCR3^−/− (CCL5/9/10 receptor) mice and found that all of them succumbed to lethal disease with kinetics comparable to B6 or BALB/c control mice (Fig. 9). Even CCL3-deficient mice lacking CCR5 and thus exhibiting a reduced CCL4/5 production capacity (Fig. 8A, 8B) as well as decreased or absent responsiveness to CCL5 or CCL4, respectively, readily died after i.c. LCMV infection (Fig. 9). Although we cannot rule out that more complex compound chemokine/receptor deficiencies may alter the course of lethal disease and a potential contribution of the XCL1:XCR1 axis remains to be investigated, the fatal course of i.c. LCMV infection appears largely independent of chemokines produced by virus-specific CD8⁺T_E.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 9.

No role for antiviral T_E-produced chemokines in the development of lethal choriomeningitis. The wt, chemokine- and/or chemokine receptor-deficient mice were infected with LCMV i.c., and survival was monitored (as per IACUC guidelines, we employed a scoring matrix to measure morbidity, and terminally ill mice were euthanized and scored as deceased). Multiple independent experiments were performed with matched experimental and control mice each, and the data displays feature the cumulative total (n) of individual mice analyzed in four separate experiments. The lower-right insert displays T_E-produced chemokines and their respective receptors with specific chemokines/chemokine receptors interrogated in the present analysis highlighted in black.