CD8⁺ T Cells Are a Biologically Relevant Source of Macrophage Inflammatory Protein-1α In Vivo¹

The recruitment of leukocytes from the peripheral blood to infected tissues is essential for the clearance of many pathogens. However, much of the tissue damage seen in chronic inflammatory diseases such as multiple sclerosis (1) and rheumatoid arthritis (2) results from infiltrating leukocytes. An understanding of leukocyte recruitment is therefore essential to the design of rational therapies aimed at modulating inflammation in ways that favorably alter the progression of these and other diseases.

Leukocyte recruitment to inflamed tissues is a multistep process that includes the initial adherence of leukocytes to the vascular endothelium, followed by diapedesis into the underlying tissue, and then migration toward target cells within the parenchyma. These events are mediated by interactions between the molecules displayed on the leukocyte cell surface and those presented on the endothelium and subvascular tissue. Members of the selectin family and their cognate receptors promote the initial tethering of leukocytes to the endothelium, whereas integrins and their receptors mediate the firm adherence of the leukocytes to the vessel wall (reviewed in 3). The molecular interactions that direct the subsequent steps of diapedesis and leukocyte migration within the extravascular tissue are less well understood, but are thought to involve chemokines and their receptors.

Chemokines comprise a large family of structurally homologous, low-m.w. secreted proteins that induce chemotaxis of specific classes of leukocytes in vitro (4) and can mediate the tissue-specific recruitment of neutrophils and monocytes in transgenic mice (5, 6). Chemokines also possess other activities, including the capacity to induce the proliferation and activation of T lymphocytes in vitro (7). Although their expression in inflammatory settings is well documented, knowledge of the function of chemokines in vivo is limited. In particular, the requirement for the production of individual chemokines in inflammatory responses and in pathogen clearance is not clear, and the biologically relevant cellular sources of chemokines have not been investigated.

Macrophage inflammatory protein-1α (MIP-1α)⁴ is a CC chemokine that has multiple activities in vitro, including chemotaxis of monocytes and CD8⁺ T lymphocytes (8, 9), activation of basophils and mast cells (10), and enhancement of the proliferation and activation of T lymphocytes in vitro (7). MIP-1α is also an important mediator of inflammation in vivo, because gene-targeted mice that cannot produce MIP-1α (MIP-1α^−/− mice) have a reduced mononuclear cell content in virus-infected tissues compared with that seen in wild-type (wt) mice (11). However, the mechanism of action of MIP-1α has not been elucidated, and it is not clear whether any of the cell types that express this chemokine in vitro (i.e., macrophages (12), T lymphocytes (13, 14), and fibroblasts (15)) are biologically relevant sources in vivo.

Murine listeriosis is a suitable model to investigate the relevant cellular sources of MIP-1α in vivo because the clearance of the causative organism, Listeria monocytogenes (LM), is mediated primarily by macrophages and T cells, the cell lineages which respond to MIP-1α in chemotaxis assays. Activated macrophages are necessary to control the early stages of a primary infection, whereas CD8⁺ T cells are important at later stages (reviewed in Refs. 16 and 17). The function of T cells in listeriosis can be studied directly in adoptive transfer experiments, in which naive recipient mice receiving a lethal dose of LM survive when infused with T lymphocytes from an immunized donor mouse (18). Here, we have used MIP-1α-deficient mice or wt mice as recipients and as donors of T lymphocytes to determine the biological relevance of different sources of MIP-1α. We show that donor T lymphocytes, but not cells in the recipient mouse, must produce MIP-1α for these T cells to migrate to the infected spleen and provide immunity against LM.

The derivation of MIP-1α gene-disrupted mice has been described previously (11). In the present studies, two different kinds of mice carrying the MIP-1α disruption were used: mice of a mixed genetic background derived from the two inbred strains, 129 Ola and C57BL/6, and mice in which the MIP-1α gene disruption had been backcrossed onto the C57BL/6 background for seven generations. Similar results were obtained with both sets of mice. Age-matched mice having a similar genetic background but having an intact MIP-1αgene were used as controls. The mice were maintained under specific pathogen-free conditions and were used when they were between 7 and 16 wk of age. All experiments were conducted according to institutional guidelines for the University of North Carolina at Chapel Hill.

LM (strain EGD) was prepared as described previously (19). Briefly, bacteria were grown in trypticase soy broth, and the approximate concentration of logarithmically growing cultures was determined by spectrophotometric analysis. Bacteria were resuspended in nonbacteriostatic saline, and equal volumes were injected into the lateral tail vein of each mouse. A more precise estimate of the number of organisms injected was determined by plating various dilutions of the bacterial suspension on trypticase soy agar and quantifying colonies the following day.

Splenic tissue from LM-infected mice was homogenized in RNA stat-60 (Tel Test, Friendswood, TX) or Tri-reagent (Molecular Research Center, Cincinnati, OH) using a Polytron tissue homogenizer (Brinkmann, Westbury, NY). Total RNA was extracted according to the manufacturer’s instructions, and 5 mg of total RNA was electrophoresed through a 0.8% agarose gel containing formaldehyde. The RNA was then transferred to a Hybond-N⁺ nylon membrane (Amersham, Arlington Heights, IL) by Northern blotting, and the membrane was probed with [³²P]deoxyCTP-labeled MIP-1α cDNA prepared using a random-primed DNA labeling kit (Boehringer Mannheim, Mannheim, Germany). Similar experiments were performed in which blots were probed with ³²P-labeled TNF-α, IFN-γ, and IL-1β DNA.

To obtain donor T cells, C57BL/6J mice or MIP-1α^−/− mice were injected with 5 × 10² to 1 × 10³ CFU of LM. Mice were sacrificed at day 7 postinfection (p.i.), and their spleens were removed, pooled, and minced with a razor blade in RPMI 1640 medium containing 10% (v/v) FBS, 2 mM l-glutamine, and 5 × 10⁻⁵ M 2-ME. RBCs were lysed in ACK lysis buffer (0.15 M NH₄Cl, 1 mM KHCO₃, and 0.1 mM Na₂ EDTA). The remaining splenocytes were washed twice in PBS and resuspended in nonbacteriostatic saline. These cells were enriched for T lymphocytes by passing the preparation through a nylon wool column. Flow cytometric analysis revealed that 90% of the prepared cells stained for the T lymphocyte surface marker CD3, 3% for the B lymphocyte marker CD19, and 7% for the monocyte lineage marker F4/80.

CD8⁺ T cells were prepared from bulk splenocytes using anti-CD8 Abs coupled to magnetic beads (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s instructions. Flow cytometric analyses of these preparations showed that >98% of the transferred cells stained for the CD8 marker.

Recipient mice received either 1 × 10⁷ nylon wool-purified T lymphocytes or 2 × 10⁶ CD8⁺-enriched T cells, followed 30 min later by an injection of 5 × 10⁴ to 1 × 10⁵ CFU of LM. In survival assays, the mice were followed for 10 days. Moribund animals were euthanized to minimize unnecessary suffering. These animals were included in the group that did not survive.

To measure CFU in the infected recipients, mice were sacrificed at 40 h p.i.; their spleens were minced using fine scissors or a razor blade in PBS containing 0.1% Triton X-100. Serial dilutions of this splenic homogenate were spread onto trypticase soy agar plates, and bacterial colonies were counted on the following day.

A polyclonal anti-MIP-1α antiserum raised in immunized rabbits has been described previously (20). The antiserum was injected i.p. (0.5 ml) at 1 h before an i.v. injection of 5 × 10⁴ CFU LM. A second injection of the antiserum was given at 48 h p.i.

B6.PL-Thy1^a/Cy (Thy 1.1) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Donor wt and MIP-1α^−/− mice (Thy 1.2) were infected with 1 × 10³ LM and sacrificed at day 7. T cells were prepared from bulk splenocytes with nylon wool columns and infused via the lateral tail vein into Thy 1.1 recipient mice that had been infected 45 min previously with 1 × 10⁵ LM. Mice were sacrificed at 48 h posttransfer, their spleens were removed, and single-cell suspensions of splenocytes were generated. These cells were stained with phycoerythrin-conjugated anti-CD8 (PharMingen, San Diego, CA) and FITC-conjugated Thy 1.2 mAbs (PharMingen) for 30 min at 4°C and subsequently analyzed on a FACScan (Becton Dickinson, Mountain View, CA) using Cicero Software (Cytomation, Fort Collins, CO). A total of 10,000 events were analyzed.

MIP-1α^−/− and wt mice were infected with 10³ CFU of LM. The mice were sacrificed at 8 days p.i., and CD8⁺ lymphocytes were isolated from bulk splenocytes using anti-CD8-coupled magnetic beads. The cells were stimulated twice with irradiated (2500 cGy) C57BL/6J splenocytes that had been infected with LM as described previously (19). Flow cytometric analysis revealed that 98–99% of these cells stained for the marker CD8. No CD4-staining cells were detected. These T cells were tested for lytic activity on IC21 cells in a standard ⁵¹Cr release assay. All samples were run in duplicate. The experiment was repeated five times with qualitatively similar results. Percent lysis was measured as follows: 100 × ([sample (cpm) − spontaneous (cpm)]/[total (cpm) − spontaneous (cpm)]).

Total counts were measured from target cells lysed in 5% Triton X-100. Spontaneous activity was measured from supernatants of wells with target cells but no effector cells. The spontaneous release for all target cells was <30% of total activity.

The Mann-Whitney rank sum test was used to calculate the statistical significance of differences between groups of recipient mice in the number of recovered LM CFU. Fischer’s exact test was used to calculate p values for survival experiments.

To determine whether MIP-1α is expressed in mice infected with LM, Northern blot analysis was performed on RNA extracted from their spleens at various times p.i. A low level of MIP-1α expression was seen at days 1, 2, and 4 p.i., with much higher levels seen at day 7 p.i. (Fig. 1).

IL-1β, IFN-γ, and TNF-α are three cytokines that have been shown previously to be important in clearing LM (21, 22, 23). Northern blot analysis of each of these cytokines revealed that their expression levels in the wt and MIP-1α^−/−-infected mice are indistinguishable at days 1, 2, and 4 (data not shown).

The high expression of MIP-1α seen at day 7 p.i. coincided with the previously described onset of LM-specific CTL activity in infected mice (24), suggesting that MIP-1α may be important for the T cell-mediated clearance of this pathogen. However, this possibility cannot be easily tested in this type of experiment because macrophages as well as T cells are able to clear LM-infected cells during primary infections. In addition, both of these cell types produce MIP-1α in vitro. Therefore, to study the role of MIP-1α in the T cell-mediated clearance of LM, we performed a series of adoptive transfer experiments in which the survival of naive mice receiving a lethal challenge of bacteria was dependent upon the protective activity of T cells transferred from an immunized donor mouse.

We first tested whether T lymphocytes from immunized wt mice were protective when transferred to mice that cannot express MIP-1α. T cells prepared from the spleens of immunized wt mice were separately injected into naive wt and naive MIP-1α^−/− recipient mice immediately before their challenge with a lethal inoculum of LM (5 × 10⁴ CFU). Infected mice that did not receive T cells succumbed to the infection, whereas both the wt and the MIP-1α^−/− recipients were protected by the transferred wt T cells (Fig. 2 A). These results show that primed wt T cells can protect recipient mice, and that the production of MIP-1α by the recipient mouse is not essential for this protection.

To determine whether the absence of MIP-1α in the recipient mouse had an effect on the function of the transferred wt T cells that was not detected in the survival assay, additional experiments were conducted in which the recipient mice were sacrificed at 40 h posttransfer and their spleens were analyzed for LM CFU. No increase in CFU was seen in the spleens of the MIP-1α^−/− recipient mice compared with the wt recipients. Rather, the MIP-1α^−/− mice had fewer bacteria (p < 0.01) than wt recipients (Fig. 2 B).

T lymphocytes have been shown previously to produce MIP-1α in vitro (13, 14, 25). To determine whether MIP-1α production by the transferred T cells is required for their ability to mediate the clearance of LM, we compared the protective abilities of T cells harvested from wt and MIP-1α^−/− mice. As described previously, T cells from the wt donor mice protected the infected wt recipients. However, as shown in Fig. 3 A, T cells from the MIP-1α^−/− donor mice had a dramatically reduced capacity to protect the infected recipient mice (p < 0.001).

Additional adoptive transfer experiments were performed to determine whether infected mice receiving MIP-1α^−/− have an increased bacterial load compared with mice receiving wt cells. As shown in Fig. 3 B, mice receiving MIP-1α^−/− donor T cells had significantly more CFU (p = 0.03) than mice receiving wt donor T cells. Thus, as measured by both survival and bacterial load, T cells from MIP-1α^−/− mice do not clear LM in vivo as efficiently as T cells from wt mice.

Adoptive T cell transfer of protection against listeriosis is mediated predominantly by CD8⁺ T lymphocytes (26, 27). Therefore, we performed additional experiments in which CD8⁺ T cells were used as the donor cells rather than unfractionated T cells. As shown in Fig. 3 C, the results of these experiments were qualitatively similar to those obtained when unfractionated T cells were used; in total, 9 of 12 mice receiving wt CD8⁺ T cells survived, whereas only 2 of 12 mice receiving primed MIP-1α^−/− CD8⁺ T cells survived (p = 0.012).

Next, we tested whether the reduced effectiveness of MIP-1α^−/− T cells could be overcome if a larger number of cells were transferred. As described previously, when 10⁷ T cells were transferred, the MIP-1α^−/− cells were less effective than wt cells: only 1 of 14 infected recipient mice receiving MIP-1α^−/− CD8⁺ T cells survived, whereas 5 of 6 mice receiving wt cells survived. When 10⁸ MIP-1α^−/− T cells were transferred, 9 of 14 recipient mice survived, indicating that this 10-fold increase in these donor cells can partially compensate for their reduced effectiveness.

The inability of the MIP-1α^−/− T cells to efficiently clear bacteria from the infected recipient mice suggested that a T cell-specific production of MIP-1α by the donor cells was required after their transfer to the infected recipient mice. If this is so, an antiserum raised against MIP-1α should be able to abrogate the ability of the primed wt T cells to protect naive mice from the bacterial challenge. Therefore, we injected a rabbit anti-MIP-1α polyclonal antiserum and primed wt donor T cells into infected wt recipients. All recipient mice (eight of eight) receiving both the anti-MIP-1α antiserum and the wt T cells died or were moribund by 60 h (Fig. 4). In contrast, only one death was seen among six recipients receiving both a control normal rabbit serum and wt T lymphocytes (p = 0.003).

We subsequently investigated whether the inability of the MIP-1α^−/− donor T cells to efficiently clear the bacteria was associated with a reduced number of donor cells in the spleens of the recipient mice compared with that seen with the wt donor cells. We used Thy 1.1⁺ wt mice as recipients to allow detection of the donor wt and MIP-1α^−/− T cells (both Thy 1.2⁺). Flow cytometric analyses of the donor cells performed at 48 h posttransfer revealed that mice receiving the MIP-1α^−/− donor T cells had 3- to 5-fold fewer CD8⁺, Thy 1.2⁺ cells in their spleens than mice receiving wt donor T cells (Fig. 5). Interestingly, the Thy1.2⁺ cells detected in the spleens of mice receiving the wt T cells were almost entirely CD8⁺, whereas in mice receiving MIP-1α^−/− T cells, the majority of Thy1.2⁺ donor cells did not stain for CD8.

MIP-1α can induce the activation and proliferation of T cells in vitro (7). To investigate whether T cells from MIP-1α^−/− mice are able to lyse LM-infected target cells in vitro, we derived T cell lines from infected wt mice and infected MIP-1α^−/− mice and compared these lines for their ability to lyse LM-infected macrophages. Both wt and MIP-1α^−/− T cells were able to lyse ⁵¹Cr-labeled LM-infected macrophage target cells in an MHC-restricted, Ag-specific manner, although the MIP-1α^−/− cells were 25–50% less efficient than the wt cells (Fig. 6). This difference, although small, was reproducibly seen in five separate experiments using independently derived cell lines.

The present experiments were undertaken to identify the cellular sources of MIP-1α that mediate T lymphocyte function in vivo. Several lines of evidence show that T cells, in particular CD8⁺ T cells, are a biologically relevant source of MIP-1α. First, CD8⁺ T lymphocytes from MIP-1α^−/− mice were significantly less effective than their wt counterparts in promoting the clearance of LM from infected naive mice. Second, the ability of wt T cells to clear the bacteria was abrogated by the injection of anti-MIP-1α-specific Abs into wt recipient mice. Third, infected MIP-1α^−/− recipients and wt recipient mice were equally well protected by the wt donor T cells, demonstrating that endogenous production of MIP-1α by the recipient mice was not required for the protective function of the donor T cells.

There are several possible explanations for this demonstrated requirement for CD8⁺ cell-produced MIP-1α. First, it may be required to recruit or to maintain recruited CD8⁺ T effector cells to the infected spleen. In support of this possibility, fewer donor CD8⁺ T cells were detected in the spleens of mice receiving MIP-1α^−/− donor T cells than in mice receiving wt donor T cells. In contrast, the number of cells that did not express CD8 was not reduced, suggesting that MIP-1α may specifically recruit CD8⁺ cells in vivo. This observation is consistent with the finding that CD8⁺ cells are much more responsive to MIP-1α in chemotaxis assays than are CD4⁺ cells (8, 9).

A second possible function of the CD8⁺ T cell-produced MIP-1α may be to participate in the activation of these cells by an autocrine mechanism. We observed a small (25–50%) but reproducible decrease in the in vitro cytotoxic activity of T cells cultured from the MIP-1α^−/− mice compared with T cells cultured from wt mice. This reduced cytolytic activity could result from the absence of MIP-1α during the assay itself or from a slight reduction in the priming of T cells in the donor MIP-1α^−/− mice. Regardless, it is unlikely that this difference is sufficient to account entirely for the dramatic difference seen in vivo between the wt and MIP-1α^−/− T cells. Moreover, the ability of the anti-MIP-1α antiserum to block the protective function of wt donor T cells indicates that MIP-1α must be produced after transfer of these cells to the recipient mice.

It is possible that the T cell-produced MIP-1α induces the expression of other cytokines that in turn mediate bacterial clearance. We did not detect any differences between the infected wt and MIP-1α^−/− mice in their levels of mRNA encoding IL-1β, TNF-α, or IFN-γ, three cytokines that are important for macrophage-mediated clearance of LM. However, we cannot exclude the possibility that cytokines other than those tested may be induced by MIP-1α and are required for macrophage-mediated clearance.

Expression of MIP-1α by CD8⁺ T cells has been observed in vitro (13, 14, 25) and in vivo (28), but the biological relevance of this expression has not been demonstrated until now. Interestingly, MIP-1α expression can be induced in T cells by anti-CD3 Ab, and this induction is dependent upon CD28/B7 costimulation (25). In addition, we have recently found using Listeria-specific cell lines that the production of MIP-1α by CD8⁺ T cells is dependent upon the presence of Ag (J.S.S. et al., unpublished data). MIP-1α has also been implicated in the recruitment of CD8⁺ T cells in vivo (29). Thus, an attractive hypothesis is that MIP-1α is produced in vivo by CD8⁺ T cells upon their recognition of the Ag presented on pathogen-infected cells, and this production results in a local gradient of the chemokine with the highest concentration near the infected cells. This gradient could direct the migration of other T lymphocytes to these “marked” foci of infected tissue. These recruited cells would in turn participate in an escalating production of more chemokine, serving to amplify the inflammatory response. In this regard, it is noteworthy that the MIP-1α^−/− recipient mice receiving wt T cells had significantly fewer bacteria than the wt recipients. The absence in these mutant mice of MIP-1α expression in cells other than the Ag-stimulated donor CD8⁺ T cells may actually increase the effectiveness of the suggested CD8⁺ T cell-generated gradient of MIP-1α, resulting in a more efficient recruitment of nearby effector cells and consequently in increased clearance of the bacteria.

In summary, we have conducted experiments in which MIP-1α^−/− mice were used as donors and as recipients of adoptively transferred CD8⁺ T lymphocytes to demonstrate that CD8⁺ T cells are a biologically relevant source of MIP-1α. The general experimental design used here is likely to be useful in identifying biologically relevant sources of other secreted molecules.

We thank Jennifer Wilder, Raymond Reilly, and Katy Migliarese for excellent technical assistance and Suzanne Kirby and Robert Reddick for critical reading of the manuscript.