Interleukin-4-induced loss of CD8 expression and cytolytic function in effector CD8 T cells persists long term in vivo

Authors

  • Stuart Olver,

    1. The Cooperative Research Centre for Vaccine Technology and the Queensland Institute of Medical Research, Herston, Qld, Australia
    Search for more papers by this author
  • Simon H. Apte,

    1. The Cooperative Research Centre for Vaccine Technology and the Queensland Institute of Medical Research, Herston, Qld, Australia
    Search for more papers by this author
  • Adriana Baz,

    1. The Cooperative Research Centre for Vaccine Technology and the Queensland Institute of Medical Research, Herston, Qld, Australia
    Current affiliation:
    1. CSL Limited, Parkville, Vic., Australia
    Search for more papers by this author
  • Anne Kelso,

    Corresponding authorCurrent affiliation:
    1. WHO Collaborating Centre for Reference and Research on Influenza, North Melbourne, Vic., Australia
    • The Cooperative Research Centre for Vaccine Technology and the Queensland Institute of Medical Research, Herston, Qld, Australia
    Search for more papers by this author
  • Norbert Kienzle

    1. The Cooperative Research Centre for Vaccine Technology and the Queensland Institute of Medical Research, Herston, Qld, Australia
    Current affiliation:
    1. Ingham Institute for Applied Medical Research, Liverpool, NSW, Australia
    Search for more papers by this author

Correspondence: Dr Anne Kelso, WHO Collaborating Centre for Reference and Research on Influenza, 10 Wreckyn St, North Melbourne, Vic. 3051, Australia. Email: anne.kelso@influenzacentre.org

Senior author: Dr Norbert Kienzle, email: Norbert.Kienzle@sswahs.nsw.gov.au

Summary

Activation of naive CD8+ T cells in the presence of interleukin-4 modulates their CD8 co-receptor expression and functional differentiation, resulting in the generation of CD8low cells that produce type 2 cytokines and display poor cytolytic and anti-tumour activity. Although this CD8low phenotype becomes stable after about a week and can persist with further stimulation in vitro, it is not known whether it can be maintained long term in vivo. Here we report that CD8low cells derived from oval-bumin257–264-specific T-cell receptor-transgenic CD8+ T cells activated in the presence of interleukin-4 could be detected in the spleen for at least 4 months after adoptive transfer into normal mice. A significant proportion of the long-term surviving cells retained their CD8low phenotype in vivo and after clonal re-activation in vitro. Although long-term surviving CD8low cells lacked detectable cytolytic activity or perforin expression, they showed some anti-tumour function in vivo. The persistence of functional cells with a CD8low phenotype in vivo raises the possibility that such cells can contribute to effector or regulatory responses to tumours or pathogens.

Introduction

Exposure of CD8+ T cells to interleukin-4 (IL-4) during primary polyclonal or antigen-specific activation in vitro leads to the generation of effector populations that express the type 2 cytokines IL-4, IL-5 and IL-10 and show markedly reduced expression of CD8α mRNA and CD8αβ surface protein.[1-4] Compared with conventional CD8high cytolytic T lymphocytes (CTL) activated in the absence of IL-4, these CD8low cells are poorly cytolytic and express reduced levels of interferon-γ (IFN-γ), perforin and granzymes.[1, 2, 5] CD8low cells with reduced cytolytic potential have also been identified in vivo among ovalbumin (OVA)257–264-specific T-cell receptor (TCR) transgenic CD8+ T cells from OT-I mice adoptively transferred into RAG-2−/− γc−/− mice and activated with tumour cells co-expressing OVA and IL-4;[4, 6] the ability of these cells to control a secondary tumour challenge in vivo was also impaired.[6]

We have previously shown that the IL-4-dependent development of CD8low cells occurs by a process of progressive differentiation and commitment: generation of these cells required exposure to IL-4 for the first few days of primary activation but they retained their low CD8 expression and cytolytic activity for many weeks in vitro, even when exogenous IL-4 was removed and endogenous IL-4 was neutralized.[4, 5] These results suggest that the IL-4-induced down-regulation of CD8α expression is heritable[5] and mechanistically distinct from the transient CD8 down-regulation observed following TCR engagement.[7, 8]

It is not known, however, whether IL-4-induced CD8low cells can also persist in vivo and, if so, whether they retain or re-acquire any functional capacity, such as cytolytic or anti-tumour activity. Here we have addressed these questions by examining the phenotypic and functional properties of activated CD8low and CD8high cells at periods up to 4 months after adoptive transfer into normal mice.

Materials and methods

Mice

Specific pathogen-free B6.SJL/J-Ptprca (CD45.1) and C57BL/6 and C57BL/6-RAG-1−/− mice (Animal Resources Centre, Murdoch, WA, Australia) were used at 6–9 weeks of age. TCR transgenic OT-I (243.2) mice (Dr William Heath, Department of Microbiology and Immunology, The University of Melbourne, Parkville, Vic., Australia) were bred at the Queensland Institute of Medical Research (QIMR). All animal studies were approved by the QIMR Animal Ethics Committee.

Antibodies for fluorescence-activated cell sorting and analysis

Antibodies to CD8α (53-6.7), CD4 (GK1.5), CD62L (MEL-14) and CD45.2 (104) and isotype controls were purchased from BioLegend (San Diego, CA). Antibodies to CD44 (IM7) and an isotype control were obtained from BD Biosciences (San Jose, CA); antibodies to Vα2 (B20.1) and an isotype control were purchased from eBioscience (San Diego, CA). Exclusion of dead cells was based on forward scatter and uptake of propidium iodide (Merck, Darmstadt, Germany).

Naive CD8+ T-cell preparation and activation in vitro

Pooled single cell suspensions from spleen and brachial, axillary and inguinal lymph nodes of OT-1 mice were enriched for leucocytes on a Ficoll–Paque gradient (GE Healthcare, Uppsala, Sweden). In most experiments, CD4 CD8+ Vα2+ CD44low cells were isolated to > 95% purity using a MoFlo flow cytometer (Beckman Coulter, Galway, Ireland) running summit Software V4.3 (Dako, Glostrup, Denmark) with or without prior depletion of B cells and MHC Class II+ cells using magnetic beads (QIAGEN, Hilden, Germany). In two experiments, CD8+ cells were positively selected on MACS Microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) to obtain CD8+ Vα2+ cells of > 90% purity. Naive OT-1 CD8+ T cells were activated by culture at 2 × 104 to 5 × 104/well in six-well plates coated with purified antibodies to CD3ε (145-2C11; 10 μg/ml), CD8α (53.6; 10 μg/ml) and CD11a (I21/7.7; 5 μg/ml) (anti-receptor antibodies) in 8 ml modified Dulbecco's modified Eagle's medium containing 50 μm 2-mercaptoethanol, 216 mg/l l-glutamine, 10% heat-inactivated fetal calf serum, 120 IU/ml human recombinant (r)IL-2 (National Institutes of Health AIDS Research & Reference Reagent Program, Germantown, MD) and the type 2 polarizing stimuli mouse rIL-4 [25 ng/ml IL-4 (ProSpec-Tany TechnoGene, Rehovot, Israel) or 3·3 U/ml baculovirus-derived IL-4][2] and 1 μg/ml anti-IFN-γ antibody (XMG1.2). After 7–8 days, cultured cells were sorted on a MoFlo to obtain Vα2+ CD8low and Vα2+ CD8high cells with purities of > 99% and > 94%, respectively.

Transfer of CD8low and CD8high cells into host mice

In some experiments, activated OT-1 cells were incubated with 2 μm carboxyfluorescein succinimidyl ester (CFSE) for 7 min, washed and sorted for Vα2+ CFSE+ CD8low and Vα2+ CFSE+ CD8high cells; 0·5 × 106 to 1 × 106 cells were then injected intravenously via the tail vein into RAG-1−/− mice. Splenic leucocytes were analysed on days 1, 2 or 4 using a FACSCalibur with cellquest version 3.1f software (Becton Dickinson, San Jose, CA). In other experiments, 0·5 × 106 to 2 × 106 CD45.2+ OT-1 Vα2+ CD8low or Vα2+  CD8high cells were injected intravenously via the tail vein into B6.SJL/J-Ptprca mice (CD45.1); where indicated, control Ptprca mice received 0·2 ml saline intravenously. CD45.2+ Vα2+ cells were analysed in the spleen 53–126 days later. In one experiment, adoptively transferred host mice also received 1 × 107 cells of the OVA-expressing EL4 thymoma E.G7-OVA subcutaneously at the base of the tail 7 days before killing. In these instances, the draining inguinal and para-aortic lymph nodes and the spleen were collected.

Clonal activation of long-term donor cells

Pooled spleens from mice that had previously received CD8low or CD8high cells were depleted of B cells, MHC Class II+ cells and CD4+ cells using magnetic beads. Donor CD45.2+ Vα2+ CD44+ cells were then purified twice by flow cytometry to obtain CD8low or CD8high cells and single cells were deposited in wells containing anti-receptor antibodies and IL-2 with or without rIL-4 and anti-IFN-γ antibody (type 2 and neutral conditions, respectively) as described above, with the exception that anti-CD3ε was used at 0·5 μg/ml. Clones were analysed after 9 or 10 days.

Cytolytic activity of long-term donor cells

After magnetic bead depletion as described above, donor CD45.2+ Vα2+ CD44+ cells were purified by flow cytometry to obtain CD8low and CD8high cells for use as effectors in a fluorolysis assay.[9] Briefly, effector cells were incubated for 48 hr with uncoated or SIINFEKL-coated EL4 target cells stably transfected with a plasmid expressing the enhanced green fluorescent protein (EGFP) gene. Target cell lysis was measured by flow cytometry to count the number of propidium iodide-negative EGFP+ cells, standardized to a reference number of fluorochrome-linked beads. Per cent lysis was calculated as follows: [1 – (targets in wells with effectors/targets in wells without effectors)] × 100.

mRNA preparation and real-time PCR analysis

RNA was extracted from 1·5 × 103 purified CD8low or CD8high cells by Nonidet P-40 (Sigma-Aldrich, St Louis, MO) hypotonic lysis and cDNA was prepared as described elsewhere.[10] Duplicate or triplicate cDNA samples were prepared from each cell type and each cDNA sample was quantified in duplicate by real-time PCR using primers and probes as previously described by reference to titrations of cDNA standards of known copy number.[4, 6, 11] Samples were amplified in a Corbett Rotor-Gene 3000 (QIAGEN, Doncaster, Vic., Australia) with initial heating to 95° for 2 min, followed by 45 cycles of 95° for 5 seconds, and 60° for 30 seconds.

Anti-tumour activity

In some experiments, 1 × 107 E.G7-OVA cells or E.G7-OVA-luc+ cells (a subclone stably expressing the luciferase gene) were injected subcutaneously at the base of the tail into Ptprca mice that had previously been adoptively transferred with CD8low cells, CD8high cells or saline; tumours were excised and weighed 7 days later. In other experiments, a classical Winn assay was performed by injecting C57BL/6 mice (= 5) with 4 × 106 E.G7-OVA tumour cells subcutaneously with saline or 6 × 105 purified primary CD8low or CD8high cells.[12] These CD8 cells were derived from primary OT-I CD8+ cells activated in type 2 conditions for 7 days and then FACS-sorted for high and low CD8 expression. Tumour growth was monitored over 32 days and mice were culled when tumour size exceeded 1 cm3 in accordance with QIMR animal ethics guidelines.

Statistical analyses

Data were evaluated by unpaired two-tailed t-test and Mann–Whitney U-test or Log-rank (Mantel–Cox) test (prism 4.02 software package, GraphPad Software, San Diego, CA). P values are expressed as *0·01–0·05, **0·001–0·01, ***< 0·001.

Results

CD8low cells proliferate and maintain low levels of CD8 expression in vivo

Naive TCR-transgenic OT-I Vα2+ CD8+ T cells specific for the OVA257–264 epitope SIINFEKL were activated in vitro with antibodies to CD3ε, CD8 and CD11a (anti-receptor antibodies) and IL-2 in type 2 polarizing conditions. After 1 week, the cells displayed variable levels of surface CD8 that ranged from normal to undetectable, as previously observed in CD8 T cells from wild-type or OT-I mice activated in the presence of IL-4.[2, 4, 5] To determine whether their altered CD8 expression was stable in vivo under conditions in which the cells could proliferate, they were incubated with CFSE and the Vα2+ CFSEhigh cells were separated into CD8low and CD8high cells (Fig. 1a) and adoptively transferred into RAG-1−/− mice. Donor cells were identified in the host spleen 1 or 4 days later by gating on Vα2+ cells (Fig. 1b). Most of the CD8low cells retained their low CD8 expression over 4 days in vivo despite having undergone multiple rounds of cell division, as indicated by the loss of CFSE. The frequency of donor CD8low cells in spleen expanded about 240-fold, from 0·08% at day 1 to 19·5% at day 4. Most of the surviving donor CD8high cells had also proliferated by day 4 and maintained relatively high CD8 levels; the emergence of some cells expressing very low levels of CD8 in this population may be a result of their prior exposure to IL-4 in vitro as previously observed.[5] Similar results were obtained in two other experiments; a fourth independent experiment found low CD8 expression 8 days after adoptive transfer of CD8low cells (data not shown). Parallel experiments in which day 0 CFSE-labelled CD8low and CD8high cells were re-cultured with IL-2 but without anti-receptor antibodies showed that both populations proliferated and maintained their respective CD8 expression profiles over 4 days (data not shown). Collectively these data indicate that OT-I CD8low cells maintained their phenotype during proliferation in vivo and in vitro for at least 4 days in the absence of high-affinity TCR stimulation.

Figure 1.

CD8low cells proliferate and maintain low CD8 expression in vivo. Naive CD44low CD4 Vα2+ CD8+ cells from OT-I mice were activated with immobilized anti-receptor antibodies, interleukin-2 (IL-2), IL-4 and anti-interferon-γ (IFN-γ) antibody for 7 days in vitro, then labelled with CFSE. Vα2+ CFSEhigh cells were separated into CD8low and CD8high cells and adoptively transferred into RAG-1−/− mice (day 0). Control mice (‘No transfer’) received no cells. (a) The left panel shows binding of anti-CD8 antibody (filled histogram) and isotype control antibody (open histogram) and sort gates (horizontal bars) of the activated cells at day 0. The right panels show the CD8 and isotype control profiles of the two populations after sorting; figures within the panels represent CD8 expression levels expressed as CD8 units [ratio of median fluorescence intensity (MFI) of anti-CD8 antibody binding to MFI of isotype control antibody binding]. (b) After 1 or 4 days, donor CD8 T cells were identified in individual spleens by Vα2 expression (rectangles) (upper panels); figures within the panels are the percentage of spleen cells expressing Vα2+. CD8 and isotype control profiles and relative CD8 expression of the gated donor cells are shown (lower panels).

CD8low cells survive long term in vivo

To assess the stability of the CD8low phenotype in vivo over longer periods, in vitro generated OT-1 CD8low and CD8high cells (Fig. 2a) were adoptively transferred into CD45.1+ congenic host mice in which the donor T cells could be discriminated from endogenous Vα2+ cells by their expression of CD45.2. As shown in Fig. 2(b), the median fluorescence intensity (MFI) of CD8 expression by CD45.2+ Vα2+ CD4 CD44+ donor cells in individual spleens of mice that had received CD8low cells 53 days earlier was lower than in mice that had received CD8high cells and a significant subpopulation of the donor CD8low cells was CD8-negative. This experiment (experiment 4a) and others (experiments 1–3) summarized in Table 1 showed that surviving CD8low and CD8high T cells maintained distinct CD8 expression profiles for at least 53–90 days after transfer, although in all experiments there was some convergence of their CD8 expression levels compared with the original transferred populations. Another experiment showed that CD8low cells could be recovered after 126 days (see below, Fig. 3). The majority of recovered splenic CD8low cells and CD8high cells expressed high levels of the activation/memory marker CD44 (means of 94–99% and 92–99% respectively in experiments 1–3). Expression of CD62L was more variable: CD8low cells were 38–62% CD62L+ and CD8high cells were 79–85% CD62L+ (data not shown).

Table 1. Long-term stability of CD8low and CD8high cells in vivo
ExptPhenotypePrimary in vitroLong-term in vivo
No. transferred cells (× 10−6)aCD8 expressionbTumourcDay in vivoOrganCD8 expression (n)d P e No. CD8 cells (× 10−4)f
  1. a

    Number of donor Vα2+ CD8 T cells transferred per mouse (day 0).

  2. b

    Ratio of MFI of anti-CD8 to isotype control antibody binding to Vα2+ CD8 T cells before transfer (day 0).

  3. c

    Where indicated, E.G7-OVA tumour cells were injected 7 days before analysis.

  4. d

    Ratio of MFI of anti-CD8 to isotype control antibody binding on long-term surviving donor Vα2+ CD4 CD44+ CD45.2+ T cells (mean ± SD) (n, number of mice analysed).

  5. e

    P value in a two-tailed t-test of difference in CD8 expression between CD8low and CD8high cells.

  6. f

    Number of long-term surviving donor CD4 CD44+ CD45.2+ Vα2+ T cells (mean ± SD) (n, number of mice analysed).

  7. g

    Two-tailed t-test indicated significant (**P 0·001–0·01) difference in CD8 levels between CD8low cells in spleen in experiment 4a versus 4b.

  8. h

    Two-tailed t-test indicated significant (***P < 0·001) difference in CD8 levels between CD8high cells in spleen in experiment 4a versus 4b.

  9. i

    DLN, draining lymph nodes.

1CD8low21·378spleen6·6 ± 1·0 (5)***5 ± 3 (5)
CD8high21878spleen16 ± 2·9 (5) 16 ± 6 (5)
2CD8low11·662spleen3·0 ± 0·57 (6)***5 ± 2 (6)
CD8high13062spleen13 ± 1·6 (6) 10 ± 5 (6)
3CD8low10·6590spleen2·3 ± 1·1 (3)*2·8 ± 0·9 (3)
CD8high0·58·490spleen7·3 (6·8, 7·8) 7·2 ± 3·3 (2)
4aCD8low0·61·753spleen2·2 ± 0·46 (3)g**0·8 ± 0·2 (3)
CD8high0·62153spleen6·0 ± 0·61 (3)h 0·4 ± 0·1 (3)
4bCD8low0·61·7+60spleen5·8 ± 1·2 (3)***3·6 ± 1 (3)
     DLNi4·3 ± 0·93 (3)***0·7 ± 0·5 (3)
CD8high0·621+60spleen19 ± 2·0 (3) 0·5 ± 0·4 (3)
     DLN18 ± 1·0 (3) 0·05 ± 0·07 (3)
Figure 2.

CD8low cells survive long term in vivo. In vitro activated OT-I CD45.2+ Vα2+ CD8 cells were separated into CD8low and CD8high cells and adoptively transferred into CD45.1+ congenic mice (day 0). (a) The panels show binding of anti-CD8 antibody (filled histogram) and isotype control antibody (open histogram) of the sorted cells before transfer; figures within the panels represent CD8 expression levels (see legend to Fig. 1). (b) After 53 days, donor CD45.2+ Vα2+ CD4 CD44+ cells (rectangles) were identified in individual spleens of mice (= 3) that had received CD8low (upper panels) or CD8high (lower panels) cells. CD8 profiles and relative CD8 expression of the gated donor cells are shown; the same isotype control staining of pooled spleens from mice that had received either CD8low or CD8high panels was used. The results are summarized in Table 1 (experiment 4a). (c) At day 53, mice (= 3) that had received CD8low (upper panels) or CD8high (lower panels) cells were challenged subcutaneously with ovalbumin-expressing E.G7-OVA tumour cells. One week later donor cells were analysed as in (b). The results are summarized in Table 1 (experiment 4b).

Figure 3.

Long-term surviving CD8low cells can maintain low CD8 expression levels during clonal activation in vitro. In three experiments (day 53, experiment 4a; day 90 or day 126, experiment 3, Table 1), donor CD45.2+ Vα2+ CD4 CD44+ cells were sorted twice for CD8low or CD8high cells before single cell deposition into wells containing immobilized anti-receptor antibodies, interleukin-2 (IL-2) and neutral or type 2 polarizing conditions. Clones were analysed after 9 or 10 days. (a): The panels show binding of anti-CD8 antibody (filled histogram) and isotype control antibody (open histogram) of the sorted day 53 cells before cloning; figures within the panels represent CD8 expression levels. (b) CD8 expression profiles (filled histogram) and levels (in CD8 units) are shown for two clones of day 126 CD8low cells grown in neutral conditions for 9 days; the same isotype control staining (open histogram) of a pool of clones is shown in both panels. (c) CD8 expression levels are shown for all tested clones derived from cells recovered on day 53, day 90 and day 126. Most clones were estimated by microscopic inspection to contain between 100 and 1000 cells. Each symbol represents a clone; the broken line marks the median CD8 expression level in each set. For each group of clones in each experiment, isotype control staining was carried out using pooled CD8high or CD8low clones from the same group. The following cloning efficiencies were obtained: Day 53: CD8low cells, 20% in neutral conditions, 20% in type 2 conditions; CD8high cells, 30% in neutral conditions, 28% in type 2 conditions. Day 90: CD8low cells, 31% in neutral conditions, 38% in type 2 conditions; CD8high cells, 44% in neutral conditions, 36% in type 2 conditions. Day 126: CD8low cells: 37% in neutral conditions, 45% in type 2 conditions.

The effect of tumour challenge on the phenotype of long-term surviving donor cells was tested by injection of cells of the OVA-expressing EL4 tumour line E.G7-OVA 53 days after adoptive transfer and 1 week before analysis. CD8 levels on donor CD8low cells in the spleen and lymph nodes draining the site of tumour injection were significantly lower than on the corresponding donor CD8high cells (Fig. 2c; Table 1, experiment 4b). CD8 levels of both CD8low and CD8high donor cells from tumour-challenged mice were significantly higher than the corresponding group of donor CD8 cells from non-challenged mice (Table 1, compare experiments 4a and b). Although differences were noted between experiments in the absolute numbers of donor CD8low or CD8high cells surviving after adoptive transfer, these differences were not consistently higher for one population than the other (Table 1). We conclude that many CD8low T cells survive and maintain their low CD8 expression long term in vivo, even after challenge with an antigen-bearing tumour.

The progeny of many single long-term CD8low cells maintain low CD8 expression

We next investigated whether long-term surviving CD8low cells up-regulate CD8 expression during re-stimulation in vitro. To ensure that the cells analysed after expansion were the progeny of long-term surviving CD8low cells rather than contaminating CD8high cells, these analyses were performed by sorting twice followed by single-cell cloning. In independent experiments (Table 1), donor cells were isolated from spleens 53 days (experiment 4a), 90 days (experiment 3) or 126 days (an extension of experiment 3) after adoptive transfer of CD8low and/or CD8high cells and sorted first into CD8low and CD8high populations, respectively, and then a second time to deposit single cells. After this double purification regimen, CD8 levels were similar for day 53 donor cells (CD8low cells, 1·3 CD8 units; CD8high cells, 7·6 CD8 units; Fig. 3a), day 90 donor cells (CD8low cells, 1·0 CD8 units; CD8high cells, 10·5 CD8 units) and day 126 CD8low donor cells (2·2 CD8 units). Individual cells were then cloned in neutral or type 2 conditions for 9 or 10 days before analysis of CD8 expression. Secondary culture conditions did not markedly affect cloning efficiencies (see legend to Fig. 3).

Clones derived from single CD8low or CD8high cells displayed a range of CD8 expression profiles; two examples are shown in Fig. 3(b). Among the clones generated in neutral conditions, CD8 levels were biased toward the phenotype of the founder cell and differed significantly between the 21 clones of CD8low and 26 clones of CD8high cells (3·3 versus 21·9, < 0·0001, two-tailed Mann–Whitney U-test) derived from day 53 and day 90 donor cells (Fig. 3c, upper and middle panels). Cloning in type 2 conditions did not significantly alter CD8 expression: clones of CD8low cells grown in neutral (= 21) and type 2 (= 22) conditions displayed similar median CD8 levels (3·3 versus 4·7) and clones of CD8high cells grown in neutral (= 26) and type 2 (= 12) conditions displayed similar median CD8 levels (21·9 versus 24·3). Median CD8 levels were also similar among clones derived from day 126 CD8low cells grown in neutral (= 36) and type 2 conditions (= 25) (12·7 versus 7·7; Fig. 3c, lower panels).

These data suggest that long-term surviving CD8low and CD8high cells have variable potential to alter their CD8 levels following re-activation and expansion in vitro. Although some expressed significantly higher or lower levels than their founder cells, on average the distribution of CD8 expression profiles was strongly influenced by founder cell phenotype. The observation that exogenous IL-4 did not significantly reduce CD8 expression among clones derived from CD8low or CD8high cells suggests moreover that the clones were no longer dependent on or responsive to the CD8-modulating effects of this cytokine.

Long-term CD8low cells are poorly cytolytic and express low levels of mRNA for effector molecules

To determine whether the functional properties of CD8low cells were altered by long-term survival in vivo in the absence of OVA, CD8low and CD8high cells were isolated 62 days after adoptive transfer (experiment 2, Table 1) and tested for cytolytic function and expression of mRNAs encoding cytolytic molecules and cytokines. For cytolytic assessment we used the fluorolysis assay which detects very small numbers of CTL by measuring the lysis of EGFP-expressing target cells.[9] When assayed ex vivo, long-term CD8high cells showed strong specific cytolytic activity against SIINFEKL peptide-coated target cells whereas CD8low cells displayed negligible activity (Fig. 4a). Similar results were obtained in another experiment with sorted donor CD8 T cells collected after 78 days in vivo (experiment 1, Table 1) (data not shown). The cytolytic activity of long-term CD8low and CD8high cells was similar to that of primary CD8low and CD8high cells after activation for 8 days in vitro (Fig. 4b). In three other independent experiments, primary CD8low cells showed sixfold to tenfold lower lytic activity per cell than CD8high cells; similar relative activity was observed in conventional 51Cr-release assays with primary CD8low and CD8high cells and SIINFEKL peptide-coated target cells (data not shown). We concluded that long-term surviving CD8low cells retained the poor CTL function typical of in vitro activated CD8low cells.

Figure 4.

CD8low cells have poor cytolytic function. (a) Donor CD45.2+ Vα2+ CD4 CD44+ cells were identified in spleen 62 days after adoptive transfer and sorted for CD8high and CD8low cells (experiment 2, Table 1). (b) In an independent experiment, primary OT-I CD8 T cells were activated for 8 days in vitro and then separated into CD8high and CD8low populations. Effector cells from both regimens were tested for cytolytic activity at the indicated effector to target (E : T) ratios in a fluorolysis assay using enhanced green fluorescent protein-expressing EL4 target cells coated with or without SIINFEKL peptide; the means ± SD of three replicates are shown.

When assayed ex vivo 62 days after adoptive transfer, CD8low cells expressed levels of CD8α mRNA that were undetectable by real-time PCR and at least 14-fold lower than those expressed by CD8high cells (Fig. 5). Average CD8α mRNA levels in the CD8high population were about 100-fold lower than those detected in CD8high cells isolated from 7-day primary type 2 cultures (data not shown), suggesting that co-receptor turnover was low in these long-term surviving cells in vivo. By contrast, CD8low and CD8high cells expressed similar levels of mRNA for IFN-γ, granzymes A and B and the housekeeping gene β2-microglobulin. The IL-4, IL-10 and perforin mRNAs were not detected. Similar results were obtained in another experiment with cells assayed 78 days after adoptive transfer (experiment 1, Table 1) (data not shown). The data show that the long-term surviving cells had low or undetectable expression levels of all the tested effector molecules; the minor differences observed in the expression of the cytolytic mediators by CD8low and CD8high cells appeared insufficient to account for differences in their cytoloytic activity.

Figure 5.

Long term CD8low cells have low CD8α mRNA expression. Donor CD45.2+ Vα2+ CD4 CD44+ cells were identified in spleen 62 days after adoptive transfer and sorted for CD8high and CD8low cells (experiment 2, Table 1). The indicated effector cell populations were assayed for expression of mRNAs encoding β2-microglubulin (β2M), cytokines [interferon-γ (IFN-γ), interleukin-4 (IL-4), IL-10], cytolytic molecules (perforin, granzyme A, granzyme B, granzyme C) and CD8α by real-time PCR; the means, and where applicable ± SD, of two or three replicate cDNA samples, each calculated from duplicate PCR runs, are shown. The broken line indicates the threshold of detection.

Long-term CD8low cells have protective anti-tumour function in vivo

In view of their poor CTL function ex vivo, long-term CD8low cells were tested for anti-tumour activity in vivo. In two independent experiments, congenic mice received CD8low or CD8high cells or saline; after 53 or 79 days, the mice were challenged with OVA-expressing tumour cells and 7 days later the mice were culled and tumours were excised and weighed. Long-term CD8high cells significantly inhibited tumour weight in both experiments. CD8low cells also reduced tumour weight but the effect was significant in only one of the experiments (Fig. 6a).

Figure 6.

CD8low cells have protective anti-tumour function. (a) In two independent experiments mice adoptively transferred with CD8low or CD8high cells or saline were injected with E.G7-OVA tumour cells at day 53 (experiment a) or with E.G7-OVA-luc+ tumour cells at day 79 (experiment b). Tumour weights were measured 7 days later. Experiment a corresponds to day 60 (experiment 4b, Table 1) and experiment b to day 86 long-term CD8 donor cells. Each data point represents one mouse and horizontal lines indicate the mean of each group; groups were compared by unpaired t-test. (b) CD8low or CD8high cells, purified from primary in vitro activated OT-I CD8 T cells, or saline were co-injected with E.G7-OVA tumour cells into normal mice (five mice per group). Tumour growth was monitored over 32 days and mice were culled when tumour size exceeded 1 cm3. Representative results of one of three similar experiments are shown.

In vitro activated primary CD8low or CD8high cells also exhibited anti-tumour activity when they were injected with E.G7-OVA cells into normal mice in a classical Winn assay. Both CD8low and CD8high cells prolonged survival but again CD8low cells were less effective than CD8high cells (= 0·011, Log-rank test; Fig. 6b). From this and three other independent Winn assays, we concluded that long-term surviving CD8low cells possessed some anti-tumour function which was weaker than that of CD8high cells.

Discussion

Here we show that CD8 T cells that had down-regulated CD8 during primary activation in vitro in the presence of IL-4 could survive for at least 4 months after adoptive transfer into normal mice. These long-term surviving cells on average displayed significantly lower levels of surface CD8 than those recovered after transfer of CD8high cells. The CD8low cells among the long-term survivors of CD8low donor cells expressed markedly lower levels of CD8α mRNA and antigen-specific cytolytic activity than long-term surviving CD8high cells. Moreover, many single long-term surviving CD8low cells gave rise to CD8low clones when restimulated in vitro in the absence of exogenous IL-4. We conclude that some IL-4-induced CD8low cells can retain their phenotypic and functional properties for several months in vivo and through multiple subsequent divisions in vitro.

We have previously reported that IL-4-induced CD8low cells also retained their CD8low phenotype and impaired cytolytic activity long-term in vitro, through many rounds of cell division in the absence of IL-4 and even when endogenous IL-4 was neutralized.[4, 5] As for long-term surviving CD8low cells in vivo (this study), their CD8low phenotype was associated with reduced CD8α mRNA expression. The heritability of this property suggests regulation by epigenetic mechanisms. Others have demonstrated an association of various epigenetic modifications of the CD8α locus with changes in CD8α gene expression as T cells differentiate from double-negative progenitors in the thymus and then migrate to the periphery,[13-16] as well as with maintenance of CD8 expression following activation of peripheral CD8+ T cells.[17] Work is in progress to determine whether some of the epigenetic modifications observed in naive CD8+ T cells are retained or reversed in CD8low cells in vitro or in long-term surviving CD8low cells ex vivo.

The initial establishment of the CD8low state depends on TCR stimulation[4] but its maintenance in vivo apparently does not require continuing TCR ligation by specific peptide–MHC complexes, consistent with other evidence that neither peptide–MHC nor TCR signalling is needed for memory CD8+ T-cell survival in vivo in many systems.[18, 19] Maintenance of the CD8low state was observed whether the cells were transferred into RAG-1−/− mice (analysed at 4 or 8 days) in which they underwent homeostatic proliferation, or into wild-type mice (analysed at 53–126 days). We did not examine whether transferred cells proliferated in wild-type mice but Perret and Ronchese[20] detected bromodeoxyuridine uptake by transferred effector CD8+ T cells as late as 83–90 days after adoptive transfer into intact congenic mice. Proliferation itself is insufficient to induce the CD8low state, however, as adoptively transferred naive CD8+ T cells did not down-regulate CD8 during homeostatic proliferation in the absence of antigen for 5 days in vivo.[4] Endogenous IL-4 may have helped to maintain the CD8low phenotype in vivo but it was notable that long-term surviving CD8low cells themselves did not express detectable IL-4 mRNA when analysed immediately ex vivo.

While long-term surviving CD8low cells displayed significantly lower levels of surface CD8 than those recovered after transfer of CD8high cells, median expression levels were reproducibly higher than those of the starting population. This partial reacquisition of CD8 in vivo could reflect the absence of the down-regulating signal provided by IL-4 and/or exposure to an up-regulating signal. A strong candidate for the latter is IFN-γ. We have reported elsewhere that IFN-γ and IL-4 exert opposing effects on CD8 expression during primary T-cell activation in vitro and in vivo such that, for example, IL-4-mediated CD8 down-regulation was enhanced in IFN-γ-deficient compared with wild-type CD8+ T cells.[4, 6] In addition, both CD8low and CD8high cells expressed somewhat higher levels of CD8 following OVA tumour challenge in vivo, raising the possibility that specific TCR stimulation led to partial recovery of CD8 expression (e.g. by inducing IFN-γ production) and/or selective recruitment or expansion of cells with higher CD8 levels. Further work will be necessary to determine whether IFN-γ can reverse the effects of IL-4 in cells that have already down-regulated CD8 mRNA and protein expression.

It was also notable that the cells that survived long-term after adoptive transfer of CD8low cells displayed some anti-tumour activity in vivo, albeit weaker than that of long-term surviving CD8high cells. Cells that retained a CD8low phenotype long-term in vivo lacked cytolytic function and expressed perforin and granzyme mRNA levels near or below the threshold of detection, at least before re-exposure to antigen on tumour challenge. It is not known whether tumour clearance was mediated by these persisting non-cytolytic CD8low cells (perhaps via IFN-γ) or by cells that reacquired some cytolytic function during the preceding weeks or following challenge. Overall, however, the data are consistent with a number of earlier studies showing that IL-4 modulates anti-tumour function in CD8 T cells, in some cases activating alternative pathways of tumour clearance.[6, 21-25]

Given that functionally similar CD8low cells can be generated in response to antigen under certain circumstances in vivo,[4, 6] the ability of transferred CD8low cells to persist for months in vivo raises the question whether this ever occurs under physiological conditions. Those cells expressing the lowest levels of CD8 identified here would be phenotypically indistinguishable from some other CD4 CD8 αβ T cells which can be found at low frequency in normal mice and humans and which have been ascribed regulatory function in a variety of conditions.[26, 27] In the future, identification of epigenetic modifications may enable distinction of IL-4-down-regulated CD8low cells from double-negative cells of other origins. Importantly, the cells we have described are not functionally inert. Although their sensitivity to peptide–MHC may be reduced in the absence of the CD8 co-receptor,[4] they retain the capacity to proliferate and synthesize (predominantly type 2) cytokines in vitro and, as shown here, they exert some anti-tumour activity in vivo. Understanding their induction, regulation and persistence may therefore contribute to mapping the full range of differentiation pathways available to naive CD8+ T cells during immune responses in vivo.

Acknowledgements

We thank Dr William Heath for the generous gift of mice, Grace Chojnowski and Paula Hall for assistance with cell sorting, Suzanne Cassidy and staff from the QIMR animal facility for animal husbandry, and Penny Groves for technical contributions. The following reagent (human rIL-2) was obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases, National Institutes of Health: human rIL-2 from Dr Maurice Gately, Hoffmann-LaRoche (Nutley, NJ). This work was supported by grants from the National Health and Medical Research Council of Australia, the Cancer Council Queensland and the Cooperative Research Centre for Vaccine Technology. The Cooperative Research Centre for Vaccine Technology was established and supported under the Australian Government's Cooperative Research Centres Programme. S.H.A. was supported by an Australian Postgraduate Award and the Basil Shaw Fellowship of the Australian Rotary Health Research Fund.

Disclosures

The authors state that there are no financial or commercial conflicts of interest.

Ancillary