CD8⁺ immunoregulatory cells in the graft-versus-host reaction: CD8 T cells activate dendritic cells to secrete interleukin-12/interleukin-18 and induce T helper 1 autoantibody

Abstract

Initiation of cell-mediated immunity or autoimmunity requires secretion of interleukin (IL)-12 from dendritic cells (DC), which drives the generation of T helper 1 (Th1) effector cells in synergy with IL-18. Induction of IL-12 can be triggered by microbial stimuli but also requires signals from activated T cells. We investigated interactions between alloreactive CD4 and CD8 T cells in mixed lymphocyte reactions (MLR) in vitro and in the graft-versus-host reaction (GVHR) in vivo. In a parent-into-F1 model of GVHR, donor CD8 cells were found to suppress the hyper-immunoglobulin E (IgE) syndrome, anti-DNA immunoglobulin G1 (IgG1) autoantibodies and donor CD4-cell expansion, but were essential for Th1-dependent immunoglobulin G2a (IgG2a) autoantibody production and release of serum IL-12 p40. In vitro, addition of alloreactive CD8 cells to CD4 cells and mature DC enhanced Th1 development. CD4 and CD8 T cells induced IL-18 from DC and primed for IL-12 p70 secretion via interferon-γ (IFN-γ) or tumour necrosis factor-α (TNF-α). However CD8 T cells, but not CD4 cells, released IFN-γ/TNF-α after primary stimulation. The data suggest that rapid release of inflammatory cytokines from central memory-type CD8 cells early in immunity is critical for induction of Th1 cells via DC activation and IL-12 production. This pathway could provide a means for amplification of cell-mediated autoimmunity in the absence of microbial stimuli.

Introduction

Immune responses frequently involve simultaneous activation of both CD4 and CD8 T cells, but the interactions between these subsets that control the nature of the resulting immune response have not been completely elucidated. In responses to foreign antigen, cross-presentation of antigen on major histocompatibility complex (MHC) class I to CD8 T cells¹ results in immune deviation of the CD4 T-cell response away from T helper 2 (Th2) and towards a T helper 1 (Th1) phenotype.^2–4 In response to self-antigens, CD8 cells are known to promote the Th1-dependent autoimmune disease, herpes stromal keratitis,⁵ and to be required for development of islet-infiltrating CD4 T cells in type 1 diabetes.⁶ In contrast, interaction between large numbers of CD4 T cells responding to the same antigen favours Th2 immunity.⁷ Therefore, it is possible that CD4 and CD8 T cells collaborate to break down immune tolerance and induce cell-mediated, rather than humoral, immunity. Dendritic cells (DC) induce Th1 development when their Toll-like receptors are triggered by infectious agents, in the presence or absence of CD8 cells.⁸ It is therefore probable that CD8 T-cell interactions with DC antigen-presenting cells (APCs) are crucial for their immunoregulatory effects.

We have addressed the regulation of CD4 T-cell responses by CD8 cells using alloreactivity in vitro and in vivo. Graft-versus-host disease (GVHD) is a frequent and potentially fatal complication of stem cell transplantation procedures and results from donor T-cell reactivity against host tissue alloantigens. Studies of GVHD in murine parent-into-F1 models showed that two types of GVHD could be induced: acute and chronic disease. Acute disease involves depletion of host haemopoietic cells, hypogammaglobulinaemia and immunosuppression.⁹ Chronic GVHD involves production of a variety of autoantibodies (e.g. anti-DNA, erythrocyte and thymocyte antibodies) which result in immune-complex glomerulonephritis.^10,11 This syndrome resembles the autoimmune disease, systemic lupus erythematosus (SLE), in humans. Acute GVHD is caused by cytotoxic CD8 T-cell responses,¹² while chronic GVHD is associated with Th2-type CD4 T cells,¹³ which stimulate host B lymphocytes to secrete autoantibodies and induces hyper-immunoglobulin E (IgE) syndrome.¹⁴ Chronic GVHD can be induced by the injection of DBA/2 lymphoid cells into (B6 × DBA/2)F1 recipients, either because DBA/2 T cells preferentially generate Th2 responses or because of a defective CD8 T-cell response.^15,16 However, the interactions between CD4 and CD8 T-cell populations that might regulate the outcome of a GVHR have not been fully investigated. The presence of CD8 T cells in bone marrow donor-cell populations, as well as causing host cytotoxicity, contributes to stem cell engraftment and the graft-versus-leukemia effect.^17,18 Furthermore, perforin-mediated cytotoxicity by CD8 cells against B cells suppresses autoantibody production in GVHD.¹⁹

Here we show that donor CD8 cells regulate engraftment and differentiation of donor CD4 T cells during the GVHR and thus the subsequent isotypes of autoantibody secreted. Interferon-γ (IFN-γ) is not essential for CD8-mediated immune deviation in vivo, as has been proposed in other models.^2,5,20,21 Both CD8 and CD4 cells were able to enhance DC secretion of IL-12 and IL-18 in vitro. However, only CD8 cells were able to secrete inflammatory cytokines early after primary stimulation and were important for IL-12 secretion in vivo. We also show that host CD8 T cells suppress Th2-type, chronic GVHR and can therefore regulate immune deviation via an alloantigen non-specific pathway.

Materials and methods

Mixed lymphocyte reactions (MLRs)

Mature DCs were induced using contact sensitization in vivo. Twenty-five microlitres of ¹ dinitrochlorobenzene (Sigma, Poole, UK) in acetone : olive oil (4 : 1) was applied to the ears of 4–6-week-old C57BL/6 mice (Harlan Olac Ltd, Bicester, UK). Eighteen hours later, draining auricular lymph nodes (LN) were excised and single-cell suspensions labelled with anti-CD11c–biotin (0·5 µg/10⁶ cells) followed by anti-biotin microbeads (Miltenyi Biotec, Bisley, UK). CD11c⁺ cells were purified by magnetic antibody cell sorting (MACS™; Miltenyi Biotec), according to the manufacturer's instructions, and were > 80% CD11c⁺ MHC class II^hi CD14⁻ DCs expressing > 45% CD86. For experiments involving the MACS™ secretion assay, DCs were purified using density-gradient centrifugation (600 g, 20 min, 22°) of the LN cells over 14·5% metrizamide in phosphate-buffered saline (PBS) containing 1% fetal bovine serum (FBS); cells at the interface were collected and were > 70% DC. CD4 and CD8 T cells (>98% purity) were isolated from LN and spleens (SP) of untreated BALB/c mice using MACS™. A total of 2 × 10⁴ B6 DCs were cultured, with or without 2 × 10⁵ purified BALB/c CD4 and/or CD8 T cells, in Dulbecco's modified Eagle's minimal essential medium (DMEM) supplemented with 10% FBS, l-glutamine (2 mm), non-essential amino acids (1-mm each), gentamicin (50 µg/ml) and 2-mercaptoethanol (50 µm) (300-µl cultures in U-bottom wells). Supernatants were collected after 48 hr. In some experiments, cells were washed after 48 hr and added to monolayers of CD40L-transfected NIH3T3 fibroblasts (kindly provided by Dr J. Gäken, King's College, London, UK) or control NIH3T3 monolayers. Supernatants were harvested after a further 24 hr. In some experiments, anti-tumour necrosis factor receptor I (anti-TNFRI) (55R-170) or anti-IFN-γ (R4-6A2) were added (both azide-free; BD Pharmingen, Oxford, UK).

GVHR induction

Female 4–8-week-old BALB/c, C57BL/6 and [BALB/c × C57BL/6]F1 (CB6F1) mice were purchased from Harlan Olac Ltd. IFN-γ^−/− mice²² (C57BL/6 background) were purchased from Jackson Laboratories (Bar Harbour, ME) and bred in isolators in our facility. The GVHR was induced by intraperitoneal (i.p.) injection of 4 × 10⁷ SP and LN cells from BALB/c or C57BL/6 donors into untreated CB6F1 recipients. When CD8-depleted cells were used, the numbers of injected cells were adjusted to give equal numbers of CD4 cells in control and depleted populations. Mice were bled serially from the tail vein at weekly intervals. Animal care was performed according to UK Home Office regulations. To determine engraftment of donor T cells, 5 × 10⁵ pooled cervical, axillary and inguinal LN or SP cells were stained with 0·1 µg of anti-H-2K^bD^b or anti-H-K^d fluorescein isothiocyanate (FITC)-labelled monoclonal antibody (mAb) (VHBio Ltd, Newcastle, UK) and anti-CD4– or CD8–CyChrome (Pharmingen, Oxford, UK) followed by flow cytometric analysis (FACSCalibur™; Becton-Dickinson, Oxford, UK). To determine cytokine profiles, mice were killed and LN cells cultured at 5 × 10⁶/ml in DMEM in 24-well plates. Supernatants were harvested after 48 hr.

Donor CD8-cell depletion

Pooled SP and LN cells (cervical, inguinal and axillary) were depleted of CD8⁺ cells using MACS™ (Miltenyi Biotec). Cells at 10⁸/ml were labelled with anti-CD8 microbeads (1 : 20 dilution) and passed through a BS negative-selection column with a 23-gauge needle attached, in a varioMACS™ device. Control cells were not labelled with anti-CD8 microbeads. Depletion was assessed by staining with anti-CD8–CyChrome (Pharmingen) and flow cytometry. CD8-depleted cells were consistently <0·1% CD8⁺ compared with >10% CD8⁺ in control populations.

In vivo CD8-cell depletion

Mice were injected i.p. at 3-day intervals with 1 mg of rat anti-mouse CD8 53-6·7 mAb, purified from the supernatant of hybridoma cells (ECACC, Porton Down, UK). Control mice were injected with 1 mg of purified rat IgG. Numbers of CD8 cells in peripheral blood of anti-CD8-injected mice were <0·1% compared with 13·0% in control mice.

Enzyme-linked immunosorbent assays (ELISAs)

Total serum IgE levels were determined using anti-IgE capture mAb (LOME-3, 3 µg/ml) and anti-κ–biotin detection mAb (OX20, 3 µg/ml) (both from Serotec Ltd, Oxford, UK). Purified mouse IgE (Pharmingen) was used as a standard. Recombinant cytokine standards and antibody pairs for determination of IFN-γ, IL-10 and IL-12 p40 (all used at 1 µg/ml) were obtained from Pharmingen/BD or R & D Systems (Abingdon, UK) for IL-18. For anti-DNA IgG assays, plates were coated with 5-µg/ml calf thymus DNA (Sigma) (boiled for 10 min and cooled), and standard curves were constructed using positive serum given a value in arbitrary units/ml (AU/ml). Detection antibodies were anti-mouse IgG2a- or IgG1-alkaline phosphatase conjugates (1 : 250 dilution; The Binding Site Ltd, Birmingham, UK). For all ELISAs, Immunosorp microtitre plates (Nunc-Life Technologies, Paisley, UK) were coated overnight at 4° with capture mAb/antigen in bicarbonate coating buffer (pH 9·0). Serum samples diluted in PBS containing 0·5% FBS and 0·5% Tween-20 were added to duplicate wells for 3 hr. Culture supernatants were added, undiluted, overnight at 4°. Plates were incubated for 90 min with detecting mAbs diluted as described above. Biotinylated mAbs were detected using 1-µg/ml streptavidin-alkaline phosphatase (Pharmingen) for 60 min. p-Nitro-phenyl phosphate substrate (Sigma), in 1-m diethanolamine buffer (pH 9·8), was added and the absorbance (A) at 405 nm measured using an automated reader (Molecular Devices, Crawley, UK). Between each incubation, plates were washed four times with PBS containing 0·05% Tween-20. IL-12 p70 was measured using a Quantikine™ ELISA kit (R&D Systems).

Intracellular cytokine staining

LN cells from GVHR mice were washed after 14 days of culture and restimulated with phorbol 12-myristate 13-acetate (PMA) (10 ng/ml), ionomycin (400 ng/ml) and a protein transport inhibitor (monensin; 3 µm) for 5 hr. Cells from in vitro MLRs were restimulated with immobilized anti-CD3 + anti-CD28 (1 µg/ml) (both Pharmingen) + monensin for 5 hr. Cells were fixed with 4% formaldehyde in 1·5 × PBS, washed and permeabilized in 0·1% saponin/0·5% BSA in PBS before staining for CD4, IL-4 and IFN-γ (APC–anti-CD4 RM4-5, phycoerythrin (PE)–anti-IL-4 BVD4-1D11, FITC–anti-IFN-γ XMG1.2, 0·1-µg/sample each; BD Pharmingen). Isotype-control mAbs and non-restimulated T cells were used as negative controls to set quadrant markers. CD4⁺ cells staining positive for IFN-γ or IL-4 alone were considered to be Th1 or Th2 cells, respectively. For analysis of early cytokine secretion, pure CD4 or CD8 T cells from untreated mice were labelled with anti-T-cell receptor-αβ (TCR-αβ)–APC (H57-597; Pharmingen), anti-CD44–biotin (IM7.8.1; Caltag, Towcester, UK) and anti-CD62L–PE (MEL-14; Caltag), stimulated with anti-CD3/CD28 + monensin for 16 hr and analysed for IFN-γ, as described above, or stimulated for 24 hr with monensin added for the final 6 hr for TNF-α staining (MP6-XT22; Caltag).

Magnetic enrichment of IFN-γ-secreting cells in a MLR

A total of 5 × 10⁶ LN cells from untreated BALB/c mice were stimulated with 5 × 10⁴ BALB/c or B6 DCs purified over metrizamide in 0·5 ml of complete medium. After 16 hr, IFN-γ-secreting cells were analysed using the MACS secretion assay™ (Miltenyi Biotec), according to the manufacturer's instructions. IFN-γ-secreting cells were enriched by two successive column separations using MS columns. Flow cytometric analysis was performed on cells pre- and postenrichment after labelling with anti-CD4–biotin + streptavidin–peridinin chlorophyll protein (PerCP) and CD8–FITC (all Pharmingen). Dead cells were excluded using propidium iodide staining (1 µg/ml).

Statistical analysis

Differences between experimental groups were analysed using unpaired t-tests. P-values of < 0·05 were considered significant.

Results

Donor CD8 cells regulate IgE production during the GVHR

The hyper-IgE syndrome is a characteristic feature of GVHD,¹⁴ and IgE synthesis is highly dependent on Th2 T cells and IL-4 synthesis.^23–25 When 4 × 10⁷ BALB/c LN and SP cells were injected into CB6F1 hosts, total serum IgE was elevated after ≈ 21 days (Fig. 1a); the IgE levels (12 × 10³−14 × 10³ ng/ml) remained ≈ 12-fold greater than those of controls for up to 8 weeks, at which point animals were killed. When CD8 cells were depleted from the donor inoculum, however, there was a more rapid elevation in IgE, with levels reaching ≈ 30 × 10³ ng/ml by day 14. These IgE levels remained high (34 × 10³−51 × 10³ ng/ml) for the rest of the experiment and were significantly greater than in non-depleted control GVHR (P < 0·001). Thus, the presence of donor CD8 cells suppressed IgE production by approximately threefold. Cell numbers were adjusted so that equal numbers of CD4 cells were injected into the mice of each group. None of the mice in any group showed signs of weight loss or ill-health up to the 8-week time-point.

Figure 1.

Donor CD8 cells induce production of T helper 1 (Th1)-dependent autoantibody but suppress T helper 2 (Th2)-dependent autoantibody and hyper-immunoglobulin E (IgE) syndrome in the graft-versus-host reaction (GVHR). Phosphate-buffered saline (PBS) diluent, BALB/c lymph node (LN)/spleen (SP) cells (4 × 10⁷), or CD8-depleted LN/SP cells were injected intraperitoneally into CB6F1 recipients, as indicated. Mice were bled at weekly intervals and serum antibody concentrations were determined by enzyme-linked immunosorbent assay (ELISA). Data shown represent mean values ± standard error of the mean (SEM) from groups of five mice and are representative of two independent experiments. Statistically significant differences between CD8-depleted and non-depleted groups are indicated at the *P < 0·05, **P < 0·01 and ***P < 0·001 levels.

Donor CD8 cells regulate the anti-DNA IgG autoantibody response in the GVHR

Serum antibodies to single-stranded DNA were measured using isotype-specific ELISA to distinguish Th1- or Th2-associated autoantibody production. The IgG1 (Th2-associated) response to DNA followed a pattern similar to that of total IgE (Fig. 1b). In contrast, levels of the Th1-dependent isotype, IgG2a, were barely detectable in mice injected with CD8-depleted BALB/c parental cells, but were enhanced four- to sevenfold when CD8 cells were present in the donor inoculum (P < 0·01; Fig. 1c). Unlike IgE, anti-DNA IgG2a levels peaked at 3–4 weeks and then returned to baseline levels.

Donor CD8 depletion enhances CD4-cell engraftment in GVHR

To determine engraftment of donor CD4 T cells in recipient mice, SP or LN cells were stained with anti-CD4 and anti-H-2K^bD^b (Fig. 2a). Preliminary experiments showed that donor CD4 cells, determined by their lack of H-2K^bD^b expression, were most abundant 14 days after injection, but absent in control mice (results not shown). Donor cells were barely detectable in peripheral blood (results not shown), but could be found in LNs and were most abundant in the SP. Donor cells constituted only 3% of the total CD4 population in mice injected with non-depleted cells, but reached around 8% in those receiving CD8-depleted populations, despite injection of equal numbers of CD4 cells. Donor CD8 cells were also found when non-depleted inocula were used, but at lower frequencies than CD4 cells. Thus, the presence of CD8 cells restricted the engraftment of donor CD4 cells during the initial phase of the GVHR.

Figure 2.

Donor CD8-cell depletion enhances engraftment of donor CD4 cells in the graft-versus-host reaction (GVHR). The GVHR was induced with control or CD8-depleted BALB/c donor cells, as described in the legend to Fig. 1. After 14 days, recipient spleen cells were stained with anti-H-2K^bD^b–fluorescein isothiocyanate (FITC) and anti-CD4–antigen-presenting cells (APC) and analysed by flow cytometry. Donor CD4 cells were identified by their lack of H-2K^bD^b expression (a; upper left quadrants). The numbers of donor cells, expressed as a proportion of the CD4⁺ population, were calculated (b) and expressed as mean values ± standard error of the mean (SEM) from groups of five mice. Significant enhancement in percentage donor cells in CD8-depleted, compared with non-depleted donor cells, is indicated (*P < 0·05). Similar results were obtained in lymph node (LN) cells and in several independent experiments.

Donor CD8 depletion suppresses Th1-cell development, but enhances Th2-cell development, during the GVHR

The enhanced IgE/IgG1 and suppressed IgG2a levels seen after donor CD8-cell depletion clearly indicated a switch in T helper phenotype from Th1 towards Th2. To confirm this we measured spontaneous cytokine secretion from LN cells of GVHR mice 7 weeks after inoculation with cells (Fig. 3a). LN cells of mice receiving non-depleted BALB/c donor cells secreted the Th1 cytokine, IFN-γ, but not the Th2-associated cytokine, IL-10. In contrast, cells from mice receiving CD8-depleted donor cells secreted both IFN-γ and IL-10. IL-4 secretion was not detectable in any samples (results not shown). To confirm that cytokines were modulated specifically in CD4 T cells, short-term cell lines were generated from the same populations. These were analysed for intracellular IFN-γ and IL-4 in CD4-gated populations (Fig. 3b). There was no significant difference in Th1 numbers detected in non-depleted GVHR mice compared with CD8-depleted GVHR. However, CD8 depletion significantly enhanced the frequency of IL-4-producing cells.

Figure 3.

Donor CD8-cell depletion alters cytokine profiles in recipient mice. Lymph node (LN) cells from mice undergoing control or donor CD8-depleted graft-versus-host reaction (GVHR), as described in the legend to Fig. 1, were obtained after 7 weeks and cultured at 5 × 10⁶ cells/ml. (a) Supernatants were collected after 48 hr and assayed for interferon-γ (IFN-γ) and interleukin (IL)-10 content. IL-4 could not be detected. (b) Cells were also cultured for 14 days, restimulated with phorbol 12-myristate 13-acetate (PMA) + ionomycin and analysed for intracellular cytokines in CD4-gated populations. The percentage of CD4 cells staining positively for IFN-γ (Th1 cells) and IL-4 (Th2 cells) are shown. Results represent the mean values ± standard error of the mean (SEM) from groups of five animals. Statistically significant differences between CD8-depleted and non-depleted groups are indicated (*P < 0·05).

IFN-γ secretion by CD8 cells is not required for regulation of donor CD4 responses during the GVHR

Regulation of immune deviation in other models has been shown to depend on the production of IFN-γ.^5,20,21 To determine whether this was the case for the GVHR, we transferred purified CD8 cells from IFN-γ-deficient mice (IFN-γ^−/−, C57BL/6 background) or from C57BL/6 controls (IFN-γ^+/+). These were co-injected into CB6F1 mice along with 4 × 10⁷ CD8-depleted LN/SP cells from C57BL/6 donors. The number of CD8 cells injected was 8 × 10⁶, in order to give a donor CD4 : CD8 ratio of 1 : 1. Peak antibody titres detected in these mice are shown in Fig. 4. Total IgE levels were elevated in mice given CD8-depleted cells alone, but much less so than previous experiments where BALB/c parental cells were used, reaching a level of ≈ 5 µg/ml. In mice receiving additional IFN-γ^+/+ CD8 cells, IgE levels were significantly suppressed, as expected (Fig. 4a; P < 0·05). Adoptive transfer of IFN-γ^−/− CD8 cells also reduced serum IgE (P < 0·05) to levels that were not significantly different from the IFN-γ^+/+ group. Serum anti-DNA IgG2a (Fig. 4b) showed the opposite pattern, with higher levels detectable in recipients of both IFN-γ^+/+ or IFN-γ^−/− CD8 cells. Anti-DNA IgG1 were not detectable in recipients of C57BL/6 cells (not shown). Only one animal receiving CD8 cells had reduced numbers of splenic B lymphocytes at 7 weeks, indicative of acute GVHD.

Figure 4.

Immune regulation by donor CD8 cells is not dependent on their secretion of interferon-γ (IFN-γ). Groups of five CB6F1 mice were injected with phosphate-buffered saline (PBS) alone, or CD8-depleted C57BL/6 lymph node (LN)/spleen (SP) cells (4 × 10⁷) [graft-versus-host reaction (GVHR)] with additional purified CD8 cells (8 × 10⁶) from C57BL/6 (IFN-γ^+/+ CD8) or C57BL6 IFN-γ-deficient mice (IFN-γ^−/− CD8). Peak levels (day 14) of (a) immunoglobulin E (IgE) and (b) anti-DNA immunoglobulin G2a (IgG2a) are shown. Anti-DNA immunoglobulin G1 (IgG1) was not detectable. Mean values ± standard error of the mean (SEM) are shown (*P < 0·05).

Alloreactive CD8 cells induce secretion of IL-12 and IL-18 in vitro and in vivo

The mechanisms of CD8-cell immunoregulation and their interaction with DCs were investigated using MLR in vitro. IL-12 has been implicated in CD8-cell immunoregulation and synergizes with IL-18 to induce Th1 immunity.²⁶ Alloreactive unprimed CD8 T cells were able to induce secretion of IL-18 after 48 hr of culture with DC, as were CD4 cells (Fig. 5a). Bioactive IL-12 p70 was not detected unless DCs were restimulated with CD40L-expressing fibroblasts after 48 hr. Culture of DCs with either CD4 or CD8 T cells enhanced their ability to produce IL-12 p70, and this activity was blocked by neutralizing antibodies to TNFR1 or IFN-γ. These antibodies had no effect on IL-18 secretion.

Figure 5.

Secretion of interleukin (IL)-12 and IL-18 in vitro and in vivo. (a) Approximately 2 × 10⁴ dendritic cells (DC) from B6 mice were cultured with 2 × 10⁵ purified CD4 and/or CD8 T cells from BALB/c mice. Neutralizing anti-tumour necrosis factor receptor I (anti-TNFRI) (25 µg/ml), anti-interferon-γ (anti-IFN-γ) (10 µg/ml) or isotype-control immunoglobulin G (IgG) (25 µg/ml) were added, as indicated. Levels of IL-18 and IL-12 p70 in supernatants were measured as described in the Materials and Methods. Data shown represent mean values ± standard error of the mean (SEM) from three independent experiments. (b) Lymph node (LN) cells from mice undergoing the graft-versus-host reaction (GVHR) were cultured as shown in Fig. 3 and secretion of IL-18 and IL-12 p70 measured by enzyme-linked immunosorbent assay (ELISA). Mean values ± SEM from groups of five mice are shown. (c) Groups of five CB6F1 mice were injected with phosphate-buffered saline (PBS) alone, or CD8-depleted C57BL/6 LN/spleen (SP) cells (4 × 10⁷) (GVHR) with additional purified CD8 cells (8 × 10⁶) from C57BL/6 (IFN-γ^+/+ CD8) or C57BL6 IFN-γ-deficient mice (IFN-γ^−/− CD8). Serum levels of IL-12 p40 were measured on day 14. Mean values ± SEM are shown (*P < 0·05).

Production of IL-12 and IL-18 was also investigated ex vivo. Spontaneous secretion of IL-12 p70 was only detectable from splenocytes of GVHR mice receiving both CD4 and CD8 donor cells (Fig. 5b), and IL-18 secretion was moderately enhanced by the presence of donor CD8 cells. Furthermore, injection of donor CD8 cells also induced release of IL-12 p40 into the serum (Fig. 5c); this appeared to be only partly dependent on IFN-γ. Serum IL-12 p70 was not detectable. Taken together, the data suggested that CD8 cells contribute to DC activation during cognate interactions and enhance production of Th1-promoting cytokines IL-12 and IL-18.

Central memory-type CD8 cells produce cytokines after primary stimulation and are biased towards the type 1 phenotype in the MLR

Naïve T cells are known to require priming and restimulation in order to secrete effector cytokines such as IFN-γ.²⁷ We examined the early production of IFN-γ and TNF-α in unprimed T-cell populations using intracellular staining after overnight stimulation. The results (Fig. 6a, 6b) show that CD8, but not CD4, T cells from untreated mice produce IFN-γ and TNF-α directly after ligation of their TCR with anti-CD3. These cytokines were produced exclusively by central memory phenotype (CD44^hi CD62L⁺) CD8 cells. This suggests that during a primary immune response, central memory CD8 T cells in draining LNs are a major source of inflammatory cytokines required for early maturation of DCs. TNF-α was not detectable until 24 hr after stimulation, while IFN-γ was detected after 16 hr. The cytokine profile of CD8 cells was then compared with that of CD4 cells after mixed MLRs containing both subsets (Fig. 6c). This showed that in the same cytokine environment, CD8 cells developed exclusively into type 1 (IFN-γ⁺) effectors, while CD4 cells developed a mixed Th1/Th2 phenotype. This divergent differentiation is probably critical for the regulatory potential of CD8 cells on CD4-dependent responses.

Figure 6.

Central memory phenotype CD8 (a) but not CD4 (b) cells produce interferon-γ (IFN-γ) and tumour necrosis factor-α (TNF-α) after primary stimulation. CD4 or CD8 cells from unprimed mice were stained for T-cell receptor-αβ (TCR-αβ), CD44 and CD62L, and stimulated with anti-CD3/CD28 for 16 hr (IFN-γ) or 24 hr (TNF-α), followed by intracellular cytokine analysis. Naïve (CD44^lo, middle panels) and memory (CD44^hi) gated T cells are shown; CD62L is the lymph node (LN) homing receptor. Unstimulated controls were <0·08% positive for either cytokine (not shown). Data are representative of four similar experiments. (c) CD4 and CD8 cells exhibit divergent cytokine phenotypes in the same cytokine environment. BALB/c CD4⁺ and CD8⁺ T cells were stimulated with B6 dendritic cells (DCs) for 7 days, stained for CD4 and CD8, then restimulated with anti-CD3/CD28 + monensin for 5 hr and analysed for intracellular IL-4 and IFN-γ. Similar data were obtained in three independent experiments.

CD8 cells are the first to produce IFN-γ in a primary MLR

In order to determine early cytokine production in primary MLRs, we used a novel technique capable of detecting low-frequency IFN-γ-secreting cells by flow cytometry.²⁸ MLRs were analysed after overnight stimulation by magnetic enrichment of cytokine-positive cells (Fig. 7). This showed that the majority of IFN-γ-secreting cells after primary stimulation were CD8⁺ (462 cells versus 150 CD4⁺ cells/10⁷ LN cells); the intensity of IFN-γ staining was also higher in CD8 cells. Responder populations stimulated with control syngeneic DCs contained few cells after enrichment.

Figure 7.

CD8 cells are the first to secrete interferon-γ (IFN-γ) in a mixed lymphocyte reaction (MLR). Lymph node (LN) cells from unprimed BALB/c mice were stimulated with purified dendritic cells (DC) from control BALB/c or allogeneic B6 DC, as indicated, for 16 hr. Cells were then analysed using the magnetic antibody cell sorting (MACS) secretion assay™ to enrich IFN-γ-secreting cells, as described in the Materials and Methods. Flow cytometry was performed pre-enrichment (upper panels) and postenrichment (lower panels); CD4- and CD8-gated populations are shown. Figures in the upper right quadrants indicate the percentage of IFN-γ⁺ cells within each population (pre-enrichment) or the total number of IFN-γ-secretors detected per 10⁷ LN cells (postenrichment). Data are representative of two independent experiments.

CD8 cells regulate Th1/Th2 development in vitro

To further investigate interactions between alloreactive CD4 cells, CD8 cells and DCs, an in vitro MLR model was used in which irradiated CD8 cells were used to prevent them overgrowing the cultures. CD4 cells developed a mixed Th1/Th2 phenotype in the absence of CD8 cells (Fig. 8a); addition of CD8 cells enhanced Th1 numbers and suppressed Th2 development. In the presence of a neutralizing antibody to TNF receptor, Th1 development in the absence of CD8 cells was unaffected, but the enhancing effect of CD8 cells was partially blocked. In contrast, anti-IFN-γ blocked Th1 development in the absence of CD8 cells, but addition of CD8 cells resulted in dramatically increased Th1 development and fewer Th2 cells. The effect of CD8 cell addition could, however, be largely reversed by a combination of both anti-TNFR1 and anti-IFN-γ (Fig. 8b).

Figure 8.

CD8 cells promote T helper 1 (Th1) over T helper 2 (Th2) differentiation in a mixed lymphocyte reaction (MLR). (a) Approximately 10⁶ BALB/c CD4 T cells were cultured with 4 × 10⁴ B6 dendritic cells (DC) in 1-ml cultures, with (right-hand panels) or without (left-hand panels) 10⁶ irradiated (2000 rads) BALB/c CD8 T cells. Neutralizing monoclonal antibodies (mAbs) to tumour necrosis factor receptor I (TNFRI) (middle panels; 25 µg/ml), interferon-γ (IFN-γ) (lower panels; 10 µg/ml) or control antibody (upper panels; 25 µg/ml) were added. After 2 days, medium was replaced to remove antibodies, and after 7 days cells were restimulated with anti-CD3/CD28 + monensin and gated CD4 populations analysed for interleukin (IL)-4 (y-axes) and IFN-γ (x-axes). CD4 cells cultured without allogeneic DC did not survive in culture. Similar data were obtained in three independent experiments. (b) Cultures were performed, as described above for Fig. 8(a), but combinations of anti-TNFR1 and anti-IFN-γ were also used. Results are expressed as percentage Th1 (IFN-γ⁺) and Th2 (IL-4⁺) in CD4-gated populations and represent mean values ± standard error of the mean (SEM) from three independent experiments.

Recipient CD8 cells suppress donor Th2-cell development during the GVHR

Finally, the potential of non-alloreactive CD8 cells to regulate GVHR was investigated. Evidence has indicated that host CD8 T cells, as well as donor CD8 cells, are capable of regulating the GVHR.^29,30 To determine whether host CD8 cells were able to regulate either the Th1 or Th2 components of the GVHR, we used in vivo CD8 depletion. CB6F1 mice were injected with anti-CD8 mAb or control rat IgG. One day later, both groups were injected with 4 × 10⁷ CD8-depleted BALB/c LN/SP cells. CD8-depleted mice had a significantly greater number of splenic donor CD4 cells on day 14 (P < 0·05, Fig. 9a). Anti-DNA IgG2a antibodies were not detectable (not shown), but IgE and anti-DNA IgG1 levels were elevated by day 14 in CD8-depleted mice (P < 0·05; Fig. 9b). Analysis of cell lines derived from LN cells indicated that CD8-depleted mice had a much greater Th2 response than non-depleted recipients (Fig. 9c), although the numbers of Th1 cells detected were similar. These data indicate that host CD8 cells are able to specifically suppress the development of Th2, but not Th1, donor cells during the GVHR and therefore possess regulatory properties similar to those of donor CD8 cells.

Figure 9.

Host CD8 cells suppress (a) donor CD4 cell engraftment, (b) T helper 2 (Th2)-dependent antibody secretion and (c) Th2 cytokine production in the graft-versus-host reaction (GVHR). CB6F1 recipients were injected with anti-CD8 to deplete CD8 cells in vivo or with control antibody. The GVHR was then induced by transfer of 4 × 10⁷ CD8-depleted BALB/c lymph node (LN)/spleen (SP) cells. Control animals received phosphate-buffered saline (PBS) alone. After 14 days, donor CD4 cell engraftment was determined in the spleen, as described in the legend to Fig. 2, serum antibodies were measured as described in the legend to Fig. 1, and Th1/Th2 profiles of recipient LN cells were determined as described in the legend to Fig. 3(b). * Indicates a P-value of < 0·05 compared with the control GVHR. Data represent mean values ± standard error of the mean (SEM) from groups of five animals and are representative of two independent experiments.

Discussion

In this study we used the GVHR model to demonstrate an important immunoregulatory pathway that links the development of CD4 T-cell-mediated and CD8 T-cell-mediated immunopathology. Initiation of Th1-type CD4 responses is generally thought to require inflammatory signals that activate APC, or the provision of a cytokine environment that favours Th1-cell differentiation. Our study suggests that CD8 T cells can provide signals that initiate Th1 immunity, and that DC activation, rather than IFN-γ secretion, is the critical mechanism in this process. Injection of CD8 T cells led to higher levels of IL-12 p40 release into the circulation, as well as Th1-associated autoantibody production. In vitro, CD8 cells are able to enhance DC secretion of IL-12 and IL-18, which synergize to promote Th1 development. This is consistent with a recent study which showed that human CD8 cells prime DCs for IL-12 secretion.³¹ Interestingly, IL-12 induction was dependent on IFN-γ and TNF-α, while IL-18 secretion from DC was not. It is therefore probable that other cytokines or chemokines assist in the activation/maturation of DC by T cells. Both CD4 and CD8 cells were able to activate DC in vitro. Therefore, the unique ability of unprimed CD8 populations to secrete inflammatory cytokines early after stimulation probably explains their regulatory potential, as DC that do not receive sufficient signals upon entering lymph nodes die or induce tolerance.⁸ Central memory type CD8 cells, which have been previously stimulated by environmental antigens but retain LN homing receptors, may fulfil this role. Our data indicate that both TNF-α and IFN-γ contribute to the regulatory potential of CD8 cells, explaining the ability of IFN-γ^−/− CD8 cells to regulate the GVHR. It should be noted that the ability of IFN-γ^−/− cells to regulate the GVHR was demonstrated using B6, rather than BALB/c donor T cells, as in other experiments, and these induce a more Th1-like GVHR with stronger cytotoxic T-lymphocyte (CTL) activity.³² This could have altered the requirement for IFN-γ. However, in vitro, BALB/c CD8 cells promoted Th1 development in the presence of anti-IFN-γ. This was presumably through secretion of TNF-α, as blockade of TNFR1 partially reversed the effect of CD8 cells.

An alternative mechanism for CD8-cell immunoregulation is release of cytokines that directly regulate CD4 cells or B-cell isotype switching. IFN-γ has recently been shown to activate the T-bet transcription factor in CD4 cells which subsequently induces IL-12Rβ2 expression.³³ Thus, early secretion of IFN-γ could directly enhance Th1 development by promoting CD4 IL-12 responsiveness. Direct regulation of B cells is unlikely given our data that CD8-derived IFN-γ is not required for the induction of IgG2a autoantibodies, as IFN-γ is the major switch factor for IgG2a.³⁴ CD8 cells are also known to be highly suppressive for antibody production during the GVHR as a result of cytotoxicity against host B cells.¹⁹ This may explain the rapid downregulation in anti-DNA IgG2a titres after 28 days (Fig. 1).

The immunoregulatory effects of CD8 T cells in the GVHR parallel those reported in rodents immunized with exogenous proteins, infected with virus or transplanted with foreign grafts, although these have generally been dependent on IFN-γ secretion.^{2,5,20,21,35,36} The dependence of IL-12 secretion on the presence of CD8 T cells has been demonstrated in both DNA and protein/alum immunization procedures.^37,38 Blockade of cytotoxic T-lymphocyte antigen-4 (CTLA-4) during the GVHR has been reported to increase expansion of donor CD8 cells and results in the production of lower levels of IgE and IL-4 in the chronic model.³⁹ Together, these data suggest that a common mechanism governs such CD4 : CD8 : DC interactions, regardless of whether antigen is derived from host tissue or foreign proteins. As cross-presentation of exogenous antigen into the MHC class I pathway occurs in DCs but not in B cells,¹ and because DCs are major producers of IL-12 and T-effector-cell inducers, it is probable that the DC is crucial to this mechanism in vivo as well as in vitro.

In addition to regulation of Th1/Th2 development and subsequent isotype switching in B lymphocytes, we observed that CD8 cells suppressed the expansion of donor CD4 cells in host animals. Several mechanisms may account for this. CD8 cells induce Th1 cells, which are more susceptible to activation-induced cell death than their Th2 counterparts⁴⁰ and may therefore be partially deleted in host tissues. This might be enhanced further by CD8 cells themselves, which express high levels of CD95L and are known to induce CD4-cell apoptosis during superantigen responses.⁴¹ Alternatively, CD8-cell suppression or cytotoxic attack of host B cells may prevent them from presenting antigen effectively to primed donor CD4 cells. This would suppress later stages of donor CD4-cell expansion.

The regulatory function of CD8 cells could also be performed by host CD8 cells, as demonstrated by in vivo CD8-depletion experiments. Although depletion of lymphoid cells from host animals in itself may have allowed donor T cells to expand more rapidly, this would not explain our observation that host CD8 cells specifically suppressed Th2 cells. The effect of depletion was unlikely to depend on CD8-expressing non-T cells such as DC, as DC IL-12 secretion is unaffected by injection of anti-CD8.³⁷ Our observation agrees with previous reports that recipient CD8 cells have a suppressor function in GVHD.^29,30 Recipient mice lacking the cytotoxic molecule, CD95L, are more susceptible to the disease,⁴² confirming the presence of a host protective mechanism. In the study of Fast et al.,²⁹ depletion of host CD8 cells enhanced donor CTL function in acute GVHD, but host CD4 depletion had the opposite effect. Our data show that chronic, CD4-dependent responses can also be suppressed. How CD8 cells are able to respond to reactive T cells, despite being unresponsive to the stimulating antigen, remains unclear. One mechanism is that CD8 cells recognize epitopes derived from TCR or immunoglobulin on activated T or B cells, respectively, via classical MHC class I presentation⁴³ or via non-classical class I molecules such as Qa-1.^5,44 Such a response would exist to dampen the excessive T-cell reactivity that occurs after exposure to bacterial superantigen or overactivation of B cells after infection with a B-cell tropic virus. In the latter case it is interesting to note that GVHRs induced with cells lacking key cytotoxic molecules, CD95L and perforin, produce higher levels of B-cell function or autoantibody production, respectively.^19,45 Our data suggest that non-cytotoxic CD8-cell functions that result in immune deviation also play a part in such processes.

Our results have implications for antibody-mediated autoimmune disease because they indicate, for the first time, that CD8 T cells can play an important role in the induction of Th1-associated autoantibodies. Production of IgG2a is associated with autoimmunity in the MRL-lpr mouse^46,47 and Th1-associated autoantibody subclasses also exacerbate collagen-induced arthritis⁴⁸ and predominate in cardiomyopathy in humans.⁴⁹ Fulpius et al. directly demonstrated that a switch in the heavy chain of a rheumatoid factor from IgG3 (Th1-associated) to IgG1 (Th2-associated) abrogated its pathogenicity in vivo.⁵⁰

Because T cells of the CD8 lineage, unlike CD4, do not require IL-12,⁵¹ STAT-4⁵² or T-bet⁵³ signalling in order to differentiate into type 1 effector cells, they develop rapidly into predominantly Tc1 effectors. This is demonstrated by our observation that CD4 and CD8 cells develop type 0 and type 1 cytokine profiles, respectively, in the same cytokine environment during a MLR. Although CD8 cells are known to secrete high levels of IFN-γ and TNF-α, we show here, for the first time, that CD8 cells with a central memory phenotype secrete these cytokines before equivalent CD4 cells and that CD8 cells are the first to produce IFN-γ in a primary MLR. This early source of cytokines would be important for maturation of DCs into highly immunogenic APCs. Together, these data strongly implicate CD8 cells as initiators of inflammation and DC activation in alloresponses. Recognition via classical or non-classical MHC class I molecules would provide early secretion of inflammatory cytokines that allow Th1 activation via IL-12-secreting DC. Such a pathway could provide one route whereby viral infections that stimulate strong CD8 responses precipitate cell-mediated autoimmunity. Furthermore, because activation of Th1 cells results in IL-2 production, an important costimulator for CD8 T-cell responses, two-way collaboration between CD4 and CD8 T cells may act to exacerbate both cell-mediated autoimmune pathologies and transplant rejection.

Abbreviations

DC

dendritic cell

GVHD

graft-versus-host disease

GVHR

graft-versus-host reaction

LN

lymph node

MACS

magnetic antibody cell sorting

MLR

mixed lymphocyte reaction

SP

spleen