CD28-mediated costimulation impacts on the differentiation of DC vaccination-induced T cell responses

Abstract

Costimulatory signals such as the ones elicited by CD28/B7 receptor ligation are essential for efficient T cell activation but their role in anti-tumour immune responses remains controversial. In the present study we compared the efficacy of DC vaccination-induced melanoma specific T cell responses to control the development of subcutaneous tumours and pulmonary metastases in CD28-deficient mice. Lack of CD28-mediated costimulatory signals accelerated tumour development in both model systems and also the load of pulmonary metastases was strongly increased by the end of the observation period. To scrutinize whether lack of CD28 signalling influences priming, homing or effector function of Trp-2_180−188/K^b-reactive T cells we investigated the characteristics of circulating and tumour infiltrating T cells. No difference in the frequency of Trp-2_180−188/K^b-reactive CD8⁺ T cells could be demonstrated among the cellular infiltrate of subcutaneous tumours after DC vaccination between both genotypes. However, the number of IFN-γ-producing Trp-2-reactive cells was substantially lower in CD28-deficient mice and also their cytotoxicity was reduced. This suggests that CD28-mediated costimulatory signals are essential for differentiation of functional tumour-specific CD8⁺ T-effector cells despite having no impact on the homing of primed CD8⁺ T cells.

Introduction

Naïve T cells require two signals for proper activation and acquisition of effector function. One signal derives from the TCR recognizing its cognate peptide in the context of self-MHC and the second one is triggered by costimulatory molecules. In this regard CD28 is appreciated as being the most important molecule capable of costimulating naïve T cells [1,2]. To investigate its role in more detail, CD28-deficient mice were generated in 1993 [3]. Although Ig class-switch and germinal centre formation are impaired in these mutants, a significant residual T cell function was shown to persist in CD28-deficient mice. This means that these mice are capable of mounting significant immune responses against transplanted tissues and some viral infections [3–6. Furthermore, despite some reports suggesting that CD28 is necessary for mounting a Th2 response [7,8], CD28-deficient mice on a Balb/c genetic background unexpectedly exhibit normal susceptibility to L. major infections accompanied by unaltered IL-4 production [9].

The relevance of costimulatory signals for anti-tumour T cell responses can be deduced from the observation that expression of CD28 ligands in tumour cells results in their rejection (reviewed in [10,11]). Recent results further demonstrated significantly enhanced activity of tumour-specific T cells after local administration of stimulating anti-CD28 antibodies [12]. Furthermore, the growth of tumours could be accelerated by inhibiting costimulation using anti-CD28 antibodies [13]. Analysis of tumour patients revealed that lymphocytes derived from the blood or the tumour-infiltrate have a reduced functional capacity which could be overcome by CD28 engagement [14,15]. However, the function of the CD28/B7 costimulatory system in therapeutically induced T cell responses to tumours has not yet been addressed. Therefore, we used the well-established Trp-2_180−188 melanoma associated antigen in conjunction with a DC-based vaccination protocol [16,17] to compare the resulting therapeutical effects on experimental metastases of B16 melanoma. Our data suggest that the lack of CD28 signalling in this model primarily impacts on the effector function of tumour-specific CD8⁺ T cells resulting in an accelerated formation of subcutaneous tumours and pulmonary metastasis.

Materials and methods

Mice

C57BL/6 mice were purchased from Charles River, Germany. CD28 ko. mice were purchased from Jackson Laboratory, USA, via Charles River, Germany. All mice were housed under conventional conditions in the Institute for Immunology and Virology of the University of Würzburg, Germany, according to the animal care guidelines.

Spleen DC isolation and vaccination procedures

Low-density spleen DC were enriched by density gradient centrifugation from fresh spleens using a modified method of McLellan et al. [18]. Six spleens were teased out in Petri dishes. The fragmented spleens were transferred to a 50 ml polypropylene tube and incubated with 1 mg/ml type III collagenase (Worthington Biochemical Corporation, Lakewood, NJ, USA) and DNase I (10 µg/ml) in 15 ml Iscove's modified DMEM with 1% normal C57BL/6 mouse serum (NMS) for 30 min at 37 °C with gentle shaking. For the last 5 min 10 mM EDTA was added and spleen fragments were then passed through a metal sieve to be collected into cold 5 mM EDTA, 10 µg/ml DNase I, 1% NMS in PBS. Subsequently, the cell suspension was filtered (70 µm cell strainers, Falcon, BD Biosciences, Heidelberg, Germany), washed twice in the same medium and then resuspended in 5 ml of cold, 14·1% Nycodenz (Nycomed, Pharma Oslo, Oslo, Norway). After transfer to a 15 ml tube the suspension was overlaid with 2 ml mouse isoosmotic buffer [19] and centrifuged for 25 min at 600× g. Low density cells were washed twice followed by plastic adherence for 2 h in 6 well plates in RPMI 1% NMS. Loosely adherent and nonadherent T cells were carefully removed and discharged. Adherent DC were cultured overnight in 2 ml of DMEM supplemented with 1% NMS and 200 U/ml rmGM-CSF (Strathmann Biotech, Hamburg, Germany). The next day cells were pulsed with 25 µM Trp-2_180−188 peptide for 2-3 h at 37°C. After two washes 1 × 10⁵ DC/mouse were injected i.d. into the lateral thigh. All vaccinations were done twice with a two week interval. In selected experiments the DC vaccination effect was boosted three times by weekly application of 25 µg TRP-2 peptide emulsified in incomplete Freunds adjuvants.

Cell lines, antibodies and peptides

The murine melanoma cell line has been described previously [20]. B78-D14 was derived from the B16 melanoma by transfection with genes coding for β-1,4-N-acetylgalactosaminyltransferase and α-2,8-sialyltransferase, inducing a constitutive expression of the disialogangliosides GD2 and GD3. B78-D14 melanoma cells were maintained as monolayers in RPMI 1640 medium supplemented with 10% fetal calf serum, 2 mM l-glutamine, 400 µg/ml G418, and 50 µg/ml Hygromycin B. Cells were passaged when subconfluent. All antibodies used in this study (anti-CD4, clone RM4-5; anti-CD8, clone 53-6·7; anti-CD25, clone 7D4) were obtained from BD Biosciences.

The tyrosinase-related protein-2 (Trp-2)_180−188 (SVYD FFVWL) 9-mer peptide was synthesized with a free C terminus by Fmoc peptide chemistry on a Biosearch SAM2 peptide synthesizer (Biosearch Technologies, Novato, CA, USA). The peptide was 90% pure (as determined by HPLC), and dissolved in DMSO (Sigma-Aldrich, Taufkirchen, Germany) and stored at −20 °C.

Tumour models

Subcutaneous tumours were induced by s.c. injection of 5 × 10⁵ melanoma cells into the left flank 7 days after the second boost with peptide-pulsed DC. Tumour growth was monitored every second day by measuring its diameter. The tumour volume was calculated as previously described [21]. When the tumour volume exceeded 1 cm³ the experiment was terminated and the mice were sacrificed.

Experimental lung metastases were induced by injection of single-cell suspensions of 5 × 10⁵ B78-D14 cells into the lateral tail vein 7 days after the second boost with peptide-pulsed DC. Tumour cells were suspended in 500 µl of PBS and administered i.v. over a period of 60 s. One mouse per group was sacrificed every week starting at 2 weeks after tumour challenge and the lungs were analysed for metastases. At day 49 grossly visible metastases were present on the surface of the organ. At day 53 the remaining mice were sacrificed and their lungs were fixed in Bouin fixative and examined under a low magnification microscope for tumour foci on their surface. Furthermore, the weight of the lungs of both groups was determined. Paraffin sections from the lungs were stained with haematoxylin/eosin and examined histologically.

Immunohistochemistry

Frozen sections were fixed in cold acetone for 10 min followed by blocking of endogenous peroxidase with 0·03% H₂O₂, and blocking of collagenous elements with 10% species-specific serum. All stainings were performed with the techmate automate (DAKO, Hamburg, Germany). For single stainings serial sections were incubated for 30 min with biotinylated antibodies at predetermined dilutions (usually 20 µg/ml). Subsequently, the streptavidin-peroxidase complex (DAKO) was applied for 30 min, followed by 15 min incubation with the Chromogen AEC (DAKO). Finally, the slides were counter stained.

TCR clonotype mapping by DGGE

The DGGE analysis used for clonotype mapping of the murine TCR BV regions 1-16 has been described [22]. Briefly, RNA was extracted using the ‘absolutely RNA RT-PCR’ Kit (Stratagene, La Jolla, CA, USA), and synthesis of cDNA was performed using 1-3 µg of total RNA, oligo(dT), and SuperScript II reverse transcriptase (Invitrogen). cDNA was amplified by primers specific for BV families 1-16 and a constant region primer that contains a 50-bp GC-rich sequence at the 5′-end. DGGE analyses were performed in 6% polyacrylamide gels containing a gradient of urea and formamide ranging from 20% to 80% separating the amplicons due to their melting properties, which are based on their nucleotide sequence. Electrophoresis was performed at 160 V for 4·5 h in 1× TAE buffer at a constant temperature of 54 °C.

CTL culture, microcytotoxicity assay and IFN-γ ELISPOT assay

Splenocytes from experimental groups were isolated and kept in culture for 4-6 days with Trp-2_180−188 pulsed syngeneic LPS-blasts in complete medium supplemented with 200-500 U/ml of recombinant human IL-2 (Chiron, Marburg, Germany) The syngeneic LPS-blasts were generated by culturing 10⁶/ml splenocytes for 3 days in the presence of 25 µg/ml LPS (Salmonella typhimurium, Sigma) and 7 µg/ml Dextran-sulphate in a T75-flask, as described [23]. After 4-6 days of culture, lymphocytes were used for ELISPOT and microcytotoxicity assay. The ELISPOT assay has been described earlier [24] and was modified to detect Trp-2_180−188-specific CD8 T cells. First, 96-well filtration plates (Millipore, Schwalbach, Germany) were coated with rat anti-mouse IFN-γ Ab (clone R4-6A2; BD Biosciences). Additionally, 25 µM Trp-2_180−188 peptide was given directly in each well and as negative control no peptide was used. After 24 h, the plates were washed, followed by incubation with biotinylated anti-mouse IFN-γ Ab (clone XMG 1·2; BD Biosciences). Spots were developed using freshly prepared substrate buffer (0·3 mg/ml amino-9-ethyl-carbazole and 0·015% H₂O₂ in 0·1 M sodium acetate (pH 5). ELISPOT were counted using the ImmunoSpot® Series 2·0 Analyser (CTL Analyser, LLC, Cleveland,US).

For the microcytotoxicity assay, 2 × 10²/well target tumour cells (B78-D14) were plated in Terasaki (Nunc, Germany) plates for 24 h. After 24 h, the in vitro restimulated lymphocytes were added at 3 different E: T ratios (100: 1, 30: 1 or 10: 1) in a final volume of 10 µl. After over night incubation, the plates were washed, fixated with −20 °C Methanol and stained with Giemsa-Solution (Merck, Germany). All stained target cells were counted and the percent of lysis was calculated. As negative control only target cells, without adding effector cells, were used.

Statistical analysis

Statistical analysis was carried out using the nonparametric Mann and Whitney two-tailed-test or the χ² test for the Kaplan-Meier Plot.

Results

Accelerated melanoma growth in CD28-deficient mice

Pilot experiments had revealed that the subcutaneous challenge of both wild type C57BL/6 and CD28-deficient mice with B78-D14 melanoma cells leads to robust tumour growth accompanied by almost no detectable immune response against the tumour cells (data not shown). However, two consecutive injections of Trp-2_180−188-pulsed DC [16] prior to the application of the melanoma cells resulted in a measurable response against the tumour cells. Therefore, this model was used in the following experiments to study the role of CD28 mediated costimulation on anti-melanoma immune responses. Since the peptide used in this protocol has a high affinity for MHC class I K^b molecules, it preferentially leads to the priming of CD8⁺ CTL [23]. To analyse the relevance of CD28-mediated costimulation for the T cell response against melanoma in this setting, we immunized mice twice with Trp-2_180−188-pulsed DC followed by the induction of subcutaneous tumours by s.c. injection of 5 × 10⁵ B78-D14 cells. Notably, the tumours became apparent earlier in CD28-deficient mice with the biggest difference at day 9 after challenge. At this time point 9/11 mutant mice had a visible tumour whereas this was the case for only 2/11 wild type mice (P = 0·005) (Fig. 1a,b). Interestingly, this significant difference persisted only for a discrete time window (becoming nonsignificant by day 11) and eventually all mice developed tumours within one month (Fig. 1c). In this respect it is noteworthy that we have recently demonstrated that DC-based vaccination alone was not sufficient to completely prevent tumour take or subsequent tumour growth of B16 melanoma in C57/Bl6 mice [16]. Hence it is not surprising that differences in the immunological control of tumour growth are eventually overcome by an evasion of the immune surveillance by the tumour. Histological examination of the tumours revealed that the infiltration of immune cells was similar in wild type and CD28-deficient mice, both for CD4⁺(Fig. 2a,d) and CD8⁺ T cells (Figs 2b,e). Counting of cells in 5 different fields (high power magnification) per tumour from at least 5 tumours from each genotype at the tumour/stroma border revealed 8-12 CD4+ cells and 6-10 CD8+ cells per field with no significant difference between the groups. Furthermore, the expression of the high affinity IL-2 receptor was also comparable among tumour-infiltrating lymphocytes (TIL) (Figs 2c,f). Our observation that the magnitude of the inflammatory infiltrate in tumours of wild type and CD28-deficient mice was similar supports the hypothesis that the vaccination-induced Trp-2_180−188-reactive T cells home with the same efficacy to the tumour, irrespective of their expression of CD28. In situ tetramer staining of Trp-2_180−188/K^b-reactive T cells further confirmed the presence of vaccination induced T cells in wild type and CD28-deficient mice (see below and data not shown). Despite the fact that we could detect Trp-2_180−188-reactive T cells among TIL in mice from both genotypes, this observation does not conclusively confirm that the tumour reactive T cells indeed preferentially home to the tumour. Therefore we analysed the clonality of the respective TIL using the RT-PCR/DGGE-based clonotype mapping [25]. One would expect a specific T cell infiltrate to be characterized by a limited TCR usage and the presence of clonally expanded cells, whereas a nonspecific inflammatory infiltrate should consist of polyclonal T cells. As depicted in Fig. 3 subcutaneous tumours in animals of either genotype harboured an oligoclonal infiltrate. The pattern of the amplified bands was similar for both genotypes, demonstrating the usage of the same TCRβ variable families (Fig. 3). Sequencing of discrete bands revealed that the sequences were different in individual mice, ruling out artifacts resulting from PCR amplification (data not shown).

Fig. 1.

Tumour growth after s.c. challenge with B78-D14 melanoma cells. (a,b) 5 × 10⁵ B78-D14 cells were injected s.c. into (a) C57BL/6 wild type and (b) CD28-deficient mice which were vaccinated twice with 1 × 10⁵ Trp-2_180−188-pulsed DC in a biweekly interval. Individual tumour volumes were calculated as described in Materials and Methods and is shown separately for individual animals. (c) Kaplan-Meier Plot demonstrating the effect of the CD28 molecule on the tumour appearance in C57BL/6 wild type (- - -) and CD28-deficient (——) mice (n = 11 for each group) after s.c. challenge with B78-D14 melanoma cells. The curves differ significantly (χ² test, P = 0·0075). The experiment was terminated on day 42 after tumour inoculation.

Fig. 2.

Immunhistological characterization of TIL. Subcutaneous tumours were induced in wild type (a–c) and CD28-deficient mice (d–f) by s. c. injection of 5 × 10⁵ B78-D14 cells. On day 14 after tumour induction, sections of subcutaneous tumours were subjected to anti-CD4 (a,d), anti-CD8 (b,e) and anti-CD25 (c,f) antibodies.

Fig. 3.

Clonotype mapping of TIL. Subcutaneous tumours were induced as indicated in Fig. 2. Biopsies of tumours were taken 21 days after tumour cell inoculation and analysed by TCR clonotype mapping. One representative experiment out of three is depicted. The figure shows the oligoclonal expansion of TCRβ variable families 1-16 on day 21 of (a) one wild type and (b) one CD28-deficient mouse. Each distinct band represents a clonotypic transcript.

Impaired efficacy of tumour specific vaccination in CD28-deficient mice

The first series of experiments suggested that the efficacy of vaccination-induced T cell responses is initially impaired in CD28-deficient mice with respect to tumour control. To confirm this notion in a second tumour model, we took advantage of the experimental pulmonary metastases model where single-cell suspensions of 5 × 10⁵ B78-D14 melanoma cells were injected into the lateral tail vene. While this has the disadvantage of not allowing a constant monitoring of the tumour growth, it is characterized by a higher fidelity to detect differences in the efficacy of ongoing immune responses. Wild type and CD28-deficient mice were vaccinated with peptide-pulsed DC as described above. Seven days after the second vaccination 5 × 10⁵ B78-D14 melanoma cells were injected into the lateral tail vein and again 7·5 weeks later the experiment was terminated. Macroscopic examination of the lungs already showed a clearly detectable difference between the lungs from wild type and CD28-deficient mice (Fig. 4a), althougth the mutant mice fell into two categories. 5 out of 9 mutant mice were heavily affected and the lungs consisted mainly of metastatic tissue with little remaining lung parenchym. In the remaining 4 mice as well as in most of the control mice the lung architecture was intact with only a few metastatic foci visible at the surface. Quantification of the foci revealed that their number was significantly increased in CD28-deficient mice (Fig. 4b). Furthermore, the weight of the metastatic lungs of CD28-deficient mice was also higher as compared to lungs from wild type controls (Fig. 4c).

Fig. 4.

Lung metastases in vaccinated wild type and CD28-deficient mice. (a) Representative picture of lungs from CD28-deficient mice (upper row) and control mice (lower row) 7·5 weeks after i.v. challenge with B78-D14 melanoma cells. Foci are clearly visible on the surface of the lungs in both experimental groups, but the lungs from CD28 ko. mice are much more affected then from control mice. (b) Average of the numbers of foci in the lungs of control (▪) and CD28-deficient mice (▴) 7·5 weeks after induction of pulmonary metastases by i.v. challenge with B78-D14 melanoma cells. The difference is statistical significant using the Mann and Whitney two-tailed-test (P < 0·05). (c) Average of the lung weights of control (▪) and CD28-deficient mice (▴) 7·5 weeks after induction of pulmonary metastases by i.v. challenge with B78-D14 melanoma cells.

Immunofluorescence staining for GD2, an artificial antigen expressed in the tumour cell line, revealed that in one CD28-deficient mouse analysed at this time point the tumours already appeared within 2 weeks after inoculation which was not the case in the wild type control (Fig. 5a). In contrast, 4-5 weeks after challenge tumours were prominent both in wild type and CD28-deficient mice (Figs 5b,c). Infiltrates of Trp-2-reactive T cells were present in all lung metastases analysed, with no obvious differences between wild type and CD28-deficient mice (Fig. 5d–f).

Fig. 5.

In situ detection of Trp-2_180−188-specific T cells in pulmonary metastases. Pulmonary metastases were induced by i.v. injection of 5 × 10⁵ B78-D14 melanoma cells in control (c,f) and CD28-deficient mice (a,b,d,e). On day 14 (a,d), 28 (b,e) and 35 (c,f) after tumour induction, sections of pulmonary metastases were stained with anti-GD2 (a–c) for visualization of metastases, or double stained with tetrameric Trp-2_180−188/H2-K^b complexes (red)/anti-CD8 (green) (d–f). Examples for CD28 ko. mice at 2 weeks (a,d) and 4 weeks (b,e) as well as control mice at 5 weeks (c,f) are shown.

Reduced numbers of Trp-2-specific IFNγ-producing T cells in CD28-deficient mice

The enhanced growth of immunological controlled tumours in CD28-deficient mice may either be explained by impaired priming, homing or differentiation of anti-tumour T cells. In the experiments described above, we have demonstrated that Trp-2_180−188-reactive T cells are detectable at the tumour site; hence T cell priming and homing is not abrogated in CD28-deficient mice. Furthermore, it seems unlikely that the effector phase itself is substantially disturbed by the absence of CD28 because primed CD8⁺ effector cells are considered to be only slightly dependent on CD28-mediated costimulation [1,2]. Therefore, the lack of CD28 mediated costimulatory signals after the initial triggering of the TCR presumably impacts on the differentiation of Trp-2 specific T cells during or shortly after priming. To address this issue, mice were first vaccinated with peptide-pulsed DC followed by administration of peptide emulsified in Incomplete Freund's adjuvant (IFA). Subsequently, the splenocytes from vaccinated mice were restimulated in vitro with Trp-2_180−188 pulsed syngeneic LPS blasts. After 6 days, the IFNγ production of individual CD8⁺ splenic T cells was analysed in response to Trp-2_180−188 by ELISPOT. In a second protocol, mice were vaccinated twice with peptide-pulsed DC, subsequently challenged with 5 × 10⁵ tumour cells and finally the splenocytes were restimulated in vitro with Trp-2_180−188 pulsed syngeneic LPS blasts. Afterwards IFN-γ production by CD8⁺ splenocytes was investigated in response to the Trp-2_180−188 epitope. In either setting, the number of IFN-γ-producing cells in CD28-deficient mice was substantially lower than in wild type mice (Fig. 6a,b and data not shown). This suggests that the generation of fully differentiated effector cells was impaired in the CD28-deficient mice.

Fig. 6.

Functional characterization of circulating Trp-2_180−188/H2-K^b-reactive T cells. (a,b) Quantification of IFNγ-producing circulating Trp-2_180−188/H2-K^b-reactive T cells was performed in splenocytes after vaccination with peptide-pulsed DC and subsequent administration of peptide emulsified in IFA. On day 17, 21 and 24 after the last peptide boost, lymphocytes were isolated from spleens and were analysed for their reactivity against the Trp-2_180−188 epitope by ELISPOT assay. The columns represent the average number of IFN-γ-producing cells in the presence of the Trp-2_180−188 peptide subtracted by the number of respective cells in the absence of peptide of wild type (a) and CD28-deficient (b) mice. The analysis was performed in triplicates for 2 mice per group and time point and the error bars represent the SD. The experiment was performed for two different cell concentrations, 2 × 10⁵/well (▪) and 1 × 10⁵/well (□). (c) The ability to kill syngeneic tumour cells was investigated for splenocytes using microcytotoxicity assays. Splenocytes from control (♦) or CD28-deficient (▪) mice vaccinated with peptide-pulsed DCs and boosted with peptide were used on day 17 after vaccination. Percent specific lysis was plotted against effector:target ratios. The analysis was performed in triplicates and SD is shown.

To further substantiate this notion by using an independent read-out we performed an in vitro cytotoxicity assay of splenocytes obtained from vaccinated mice. Importantly, splenocytes from CD28-deficient mice recovered at day 17 after peptide boost showed decreased efficacy in killing syngeneic melanoma cells as compared to splenocytes from wild type mice (Fig. 6c). Taken together, our findings suggest that in CD28-deficient mice IFNγ-producing cells are not only reduced in number but also less potent in lysing their respective target cells.

Discussion

CD28 was identified more than two decades ago and is appreciated as one of the most efficient molecules providing costimulatory signals to T cells, especially during activation of naïve T cells. However, the role of CD28 in anti-tumour immune responses remains controversial. The finding that rejection is enhanced when tumours are transfected with B7 molecules, the natural ligands for CD28, supports a substantial role for costimulation in anti-tumour immune responses [26–28]. In line with this notion, the growth of certain tumours is enhanced when CD28/B7 interaction is blocked [13,29]. Furthermore, CD28-deficient mice are able to reject virus induced lymphomas and chemically induced fibrosarcomas [30] but not A20 lymphomas [31]. This demonstrates that the relevance of CD28 in anti-tumour responses depends at least partially on the type of the tumour, and, at least under some circumstances, CD28 mediated costimulation is dispensable. However, with regard to melanomas the role of CD28 has not yet been addressed in detail. Although it was previously shown that administration of either anti-B7-1 or anti-B7-2 leads to a decreased anti-melanoma immune response [32], the anti-melanoma immune response in CD28-deficient mice has not been studied so far.

Here, we demonstrate in a prophylactic vaccination melanoma model that tumours grow faster and that the tumour burden is larger in CD28-deficient mice as compared to wild type animals. This suggests that CD28 plays a role in vaccination induced anti-tumour immune responses in this setting. An obvious explanation for the failure to mount a full anti-tumour immune response would be inappropriate proliferation of tumour specific T cells in response to the antigen. However, this seems not to be the case. Priming of Trp-2 specific T cells by DC vaccination, a method found to ensure efficient priming of CD8⁺ T cells in previous experiments [16,23] leads to an infiltration of the tumour by comparable numbers of Trp-2 specific CD8⁺ T cells irrespective of their genotype. Thus homing of Trp-2 specific CD8⁺ T cells to the tumour is not compromised in CD28-deficient mice. In contrast, functional analysis of CTL revealed important differences. In CD28-deficient mice the number of cells producing IFN-γ in response to Trp-2 was substantially lower as compared to splenocytes from wild type controls. Although it might be explained by a direct perturbation of the CD8⁺ T cell effector function in the absence of CD28 this appears unlikely since it was previously shown that primed CD8⁺ effector cells are less dependent on CD28-mediated costimulation than naïve CD4⁺ T cells [1,2]. Therefore, we suggest that the CD28-deficient CD8⁺ T cells are normally primed and that also the homing to the tumour is unaffected. In contrast, we rather assume that the differentiation into bona fide effector cells is impaired in the absence of CD28, resulting in reduced effector function in response to secondary antigen encounter. This is in line with the previous observation that CD8⁺ effector T cells expand properly after vaccination with heat-killed L. monocytogenes and even differentiate into memory cells although they do not acquire effector functions and do not protect the mice from subsequent infection [33]. One may speculate that due to the heat-treatment these bacteria are not able to evoke efficient costimulation, leading to impaired differentiation and lack of effector functions. In this respect it is noteworthy that CD28 deficiency also causes resistance to experimental autoimmune encephalitis (EAE). This is neither to be due to an impaired proliferative response of encephalitogenic T cells nor due to impaired cytokine responses [34,35] but rather derives from compromised effector functions of primed T cells [36]. In other autoimmune disease models, e.g. experimental mycocarditis or collagen-induced arthritis, costimulation by CD28 also seems to be crucial [7,37]. Hence, in response to weak antigenic stimuli such as self antigens, T cells may be incompletely activated in the absence of costimulation and therefore rendered ‘unfit’, e.g. characterized by insufficient expression of homeostatic cytokines and their respective receptors (such as IL-7 and IL-15). In contrast, optimally primed T cells accumulate sufficient levels of cytokines, antiapoptotic molecules and cytokine receptors to sustain their metabolism, viability and effector functions [38,39].

In a recent study it was demonstrated that induced immune responses to viral peptides in CD28-deficient mice differ quantitatively but not qualitatively as compared to wild type mice [40]. In our model system, we did not observe a difference in DC vaccination induced T cell responses between CD28-deficient and wild type mice. This discrepancy may be ascribed to the repeated vaccinations used in our model, as it was previously demonstrated that TCR stimulation by repetitive peptide vaccination can overcome T cell unresponsiveness in CD28-deficient mice [41]. Nevertheless, our results suggest a qualitative difference in the effector phase between CD28-deficient and wild type mice even after repeated vaccinations. It should also be noted that viral epitopes per se generally induce strong immune responses in vivo[42].

Taken together, our results suggest that CD28 plays an important role in DC vaccination-induced T cell responses, in particular by impacting on the differentiation of CD8⁺ effector T cells. In contrast, expansion and homing of CTL to the tumour does not require CD28 engagement. Our proposed model therefore may explain the seeming controversy of the role of costimulation in tumour immunology, e.g. strong antigens may be deleted in the absence of CD28, whereas weak antigens such as self-antigens crucially depend on it.