Rapid turnover of the CD8⁺CD28^- T-cell subset of effector cells in the circulation of patients with head and neck cancer

Patients and controls

Thirty-two patients (aged 36 to 79 years; mean 61 years) with histologically proven squamous cell carcinoma of the head and neck (SCCHN) participated in the study after signing the IRB-approved informed consent. All patients asked to participate in the study were seen at the outpatient clinic of the Department of Otolaryngology. The patients were divided into two groups based on clinical evaluation: (1) 18 patients with active disease (AD), aged 36 to 76 years, mean 61 years; and (2) 14 patients with no evidence of disease (NED), aged 44 to 79 years, mean 61 years. The patients were not treated with chemotherapy or radiotherapy at the time of blood draws. Also, 30 healthy adult volunteers (aged 36 to 79 years; mean 59 years) participated in this study as normal controls (NC).

Lymphocyte separation

Venous blood was obtained from patients and controls. Peripheral blood mononuclear cells (PBMC) were separated by Ficoll-Hypaque gradient centrifugation, washed and counted in the presence of the trypan blue dye to determine the cell number and viability. The cells were immediately evaluated for Annexin binding or cryopreserved for other studies.

Antibodies

The fluorochrome-labeled monoclonal antibodies used for staining of PBMC were FITC- or PE-conjugated anti-CD28 (CD28.2), ECD-conjugated anti-CD3 (UCHT1), anti-CD45RO (UCHL1), and PC5-conjugated anti-CD4 (13B8.2). All were purchased from Immunotec, France. In addition, PC5-conjugated anti-CD8 (SCCI21Thy2D3) purchased from Beckman Coulter (Miami, Florida); PE-conjugated anti-CD4 (Leu-3a) purchased from Becton Dickinson (San Jose, California); and PE-conjugated anti-CD95 (DX2) purchased from Pharmingen (San Diego, California). As negative controls, the respective Ig isotypes were purchased from Immunotec.

Staining of PBMC

In preparation for flow cytometry, staining of PBMC was performed as follows: cells (1×10⁶) were placed in wells of a 96-well round bottom plate. An aliquot (1 μl) per well of an antibody was added, and the plate was incubated at 4°C for 30 min. After incubation, the cells were washed twice in FACS buffer and transferred to a flow tube. All antibodies were pretitered on normal PBMC to determine their optimal dilutions. Ig isotype controls were included in all experiments.

Flow cytometry

Four-color flow cytometry was performed on a Coulter Epics XL cytometer with a single 488-nm argon ion laser. At least 1×10⁵ events were acquired for each sample. The amplification and compensation were set according to the standard procedure, using negative controls and tested cells stained in a single color or combination of colors (FL1, FITC-CD28; FL2, PE-CD4; FL3, ECD-CD3; and FL4, PC5-CD8). For evaluations of CD8⁺ T-cell subsets the gate was set on CD8^bright T lymphocytes.

Annexin V binding assay

Annexin V binding to T lymphocytes was evaluated by multicolor flow cytometry. After surface staining with anti-CD28, anti-CD45RO, and anti-CD4 or anti-CD8 Abs, PBMC were washed once with FACS buffer and resuspended in Annexin-binding buffer (PharMingen), followed by incubation with FITC-conjugated Annexin V (PharMingen) for 15 min at room temperature. The events were acquired by flow cytometer within 60 min of staining. The percentages of apoptotic cells were calculated by scoring Annexin V–binding cells after backgating on CD4⁺ cells or CD8⁺ cells. All gated mononuclear cell populations were visualized on forward angle scatter/side angle scatter (FSC/SSC) dot plots, and the gate was set to eliminate cellular debris, which binds Annexin V and still may express CD3. Next, a cutoff was set using unstained control cells. To set the second (higher) cutoff, Jurkat cells pretreated with CH11 Ab to induce apoptosis were utilized. This allowed for the elimination of "dim" Annexin-staining cells, which were not apoptotic based on simultaneous staining of the positive control cells with Propidium Iodide (PI). This strategy of four-color staining combined with stringent gating provided the means for eliminating a majority of debris and for clear-cut discrimination between live and apoptotic cells among subpopulations of T lymphocytes in the gate. The debris, which was excluded using this gating, generally contained from 23±8% to 22±13% of Annexin⁺ events in patients and controls alike.

Intracytoplasmic staining

Effector cells were stimulated with PCI-13 or 1 ng/mL of phorbol 12-myristidate 13-acetate (PMA; Stigma, St. Louis, Missouri) and 1 μM of ionomycin (Sigma) for 24 hr at 37°C, in an atmosphere of 5% CO₂ in air. At the end of the incubation period, cells were suspended in FACS buffer and stained with labeled antibodies selected to surface markers as described above. Next, cells were fixed for 10 min with 1% (w/vol) paraformaldehyde in PBS and washed with the permeabilization buffer (0.1% [w/vol] saponin in PBS). The cells were stained with FITC-conjugated anti-IFN-γ Ab (Pharmingen) or anti-granzyme B Ab (Alexis, San Diego, California) for 30 min at 4°C. As a negative control, cells were stained with mouse IgG1 or IgG2a isotype control (Becton Dickinson). Cells were washed once with permeabilization buffer and once with FACS buffer, and then analyzed by flow cytometry.

Generation of semiallogeneic CTL

The HLA-A2⁺ SCCHN cell line, PCI-13 was previously established from a moderately differentiated tumor and retrovirally transfected with the B7.1 gene (17). Tumor cells were maintained in DMEM supplemented with 10% fetal bovine serum, 1-mM l-glutamine, 100-μl/ml penicillin and 100-μg/ml streptomycin. Tumor cells were pretreated with 1000 IU/ml of IFN-γ (Roussel UCLAF, France) for 48 hr to enhance expression of MHC class I molecules prior to irradiation and were used as stimulator cells. PBMC from three HLA-A2⁺ healthy donors were coincubated with 1×10⁵ irradiated (100 Gy) PCI-13 [14], or B7-transfected PCI-13 [17] at an effector/stimulator cell ratio of 10:1, and maintained in culture in the complete medium composed of AIM-V supplemented with 10% human AB serum. This stimulation was repeated weekly. IL-2 (20 U/ml) obtained from Chiron Cetus, Emeryville, California, was added to the medium beginning with the second cycle of stimulation.

Cytotoxicity assay

A 4-h ⁵¹Cr-release method was used for analyzing cytotoxicity of effector T cells at the effector/target ratio ranging from 2.5 to 50 as previously described [36]. To remove NK activity, 5×10⁴ of K562 cells were incubated with effector cells for 30 min prior to the cytotoxicity assay. Aliquots of 1×10³ of PCI-13 target cells were then added to each tube. The assay was performed in triplicate, and the percent specific lysis was calculated by the following formula: % specific lysis = [(maximal cpm − spontaneous cpm) / (experimental cpm − spontaneous cpm)] x 100.

Statistical analysis

The statistical significance of differences in expression of apoptotic markers between groups was determined by nonparametric Mann-Whitney U test for the unpaired analysis and Wilcoxon matched pairs test for the paired ansalysis. Linear regression analysis was performed to evaluate the influence of age, using Spearman's test. P values less than 0.05 were regarded as significant.

Distribution of naïve, memory, and effector subsets of CD8± T cells in patients and NC

It has been previously suggested [2] that CD8⁺ T cells are heterogenous, comprising at least three distinct subsets: CD28⁺CD45RO^- (naïve), CD28⁺CD45RO⁺ (memory) and CD28^- (effector cells). By multicolor flow cytometry, backgating on CD8^bright PBMC, it is possible to discriminate between these subsets (Fig. 1). In NC, CD28^- T cells are few, and the naïve CD28⁺CD45RO^- T cells are a major CD8⁺ subpopulation. However, in patients with HNC, the proportions of CD28⁺ subsets are decreased, while a significant increase in CD28^- effector T cell is apparent (Fig. 1).

Age-dependent changes in naïve, memory, and effector subsets of CD8± T cells

Patients with HNC are generally older than 50 years of age. Therefore, before comparing patients to NC, it was necessary to consider their age, since immune parameters could be altered in aging individuals. Indeed, as shown in Fig. 2, significant age-related changes were observed in the naïve and effector subsets of CD8⁺ T cells both in patients with HNC and NC. Thus, the proportions of naïve CD8⁺ T cells decreased, while those of CD28^- effector T cells increased with age in both HNC and NC. In contrast, the memory T cells did not seem to significantly change with age (Fig. 2).

The frequency of CD8 T-cell subsets in HNC patients vs NC

Our data indicated that comparisons in the frequency of the naïve and effector subsets of CD8⁺ T cells between HNC patients and NC, would require that age-matched cohorts be used. We therefore matched the cohorts of HNC patients and NC in age within a 4-year window in order to compare the relative mean percentages of CD8⁺ T-cell subsets. The mean proportion of naïve CD8⁺ T cells was found to be decreased in HNC patients relative to controls (p<0.01), while that of memory (p<0.05) and effector (p<0.05) CD8⁺ T cells increased, as illustrated in Fig. 3.

Annexin V binding to CD8^± T cells

We previously reported that CD8⁺ T cells preferentially bound Annexin V relative to CD4⁺ T cells and that proportions of such Annexin⁺CD8⁺ T cells were found to be elevated in patients with HNC [15]. We therefore examined binding of Annexin V to CD8^bright cells and their subsets in our cohorts of patients and NC. As shown in Fig. 4A, HNC patients had a significantly higher total percentage of Annexin-binding CD8^bright cells in the peripheral circulation than did NC (p<0.01), confirming our previous results. Because the frequency of CD8⁺CD28^- T cells was significantly increased in HNC patients relative to NC, we did not expect to observe much apoptosis in this subset of effector cells. Surprisingly, CD28^- T cells contained the highest proportion of Annexin⁺ cells among the three CD8⁺ T-cell subsets (CD8⁺CD28^-Annexin V⁺), while the naïve CD28⁺CD45RO^- subset contained the lowest (Fig. 4B). This counterintuitive result could be explained either by proliferation of CD8⁺CD28^- cells or by cell influx from the naïve T-cell pool, which compensated for the loss of CD8⁺CD28^-Annexin⁺ effector cells. This, in turn, implies that CD28^- effector cells had a rapid turnover rate in patients with HNC. Both memory and effector CD8⁺ T-cell subsets contained a substantially higher proportion of CD95⁺ cells, as compared with naïve CD8⁺ T cells: 80–90% vs 15–30%, with patients consistently in the upper range.

In vivo changes in CD8^± T-cell subsets after surgery

To determine whether the observed increased frequency of CD8⁺CD28^- effector cells in HNC patients relative to NC was related to the presence of tumor, CD8⁺ T-cell subset analysis was performed before and after surgery. In five cases, we were able to obtain paired peripheral blood specimens. In three of five cases, normalization of the subset frequencies was observed after surgery, with a decrease in the percent of CD28^- effector cells and an increase in the proportion of naïve CD28⁺CD45RO^- T cells (Fig. 5). The proportion of CD95⁺ cells decreased slightly in the CD28^- effector cell subset after surgery (data not shown). Although limited, these data suggest that the presence of tumor has an effect on the homeostasis of the CD8⁺ T-cell subsets, including that of CD8⁺CD28^- effector cells.

Down-regulation of CD28 on tumor-reactive effector cells

A series of ex vivo studies was next performed to investigate whether the observed in vivo loss of CD28 from the surface of CD8⁺ T cells could be related to antigenic stimulation in the tumor microenvironment. Semiallogeneic T-cell lines were generated from PBMC of three normal HLA-A2⁺ donors in mixed lymphocyte tumor cultures (MLTC), using the PCI-13 cell line (HNC-derived, HLA-A2⁺) as a stimulator. The T-cell lines were tested for PCI-13–specific cytotoxicity in 4-h ⁵¹Cr-release assays performed after NK activity was removed by preincubation of T cells with K562 targets. Expression of CD28 molecules on CD8⁺ T cells and their cytotoxicity were repeatedly measured over a culture period of 12 weeks. Stimulations with the tumor were performed weekly. Early in MLTC, most cells had the CD8⁺CD28⁺CD45RO⁺ phenotype and mediated moderate but detectable antitumor activity (Fig. 6A). In the course of culture with repeated antigenic stimulations, gradual down-regulation of CD28 was observed in CD8⁺ cells, until more than 60% were CD28^-. This gradual shift from the CD28⁺ to CD28^- phenotype was accompanied by increasing antitumor cytotoxicity (Fig. 6A) and increased Annexin V binding, indicating that CD28^- T cells were sensitive to apoptosis. When about 50–70% of the cells became CD28^- (e.g., 9 weeks in culture), the down-regulation of CD28 leveled off and cytotoxicity of the population began to decline, while Annexin V binding was observed in 25–35% of the cells. These ex vivo experiments confirm that CD28⁺CD8⁺ T cells lose CD28, retain cytolytic function upon chronic (repeated) exposure to tumor antigens, and upon differentiating to CD28^- cells become sensitive to apoptosis.

On week 6 of these cultures, when antitumor cytotoxicity was high and CD28^- cells began to reappear and constituted a substantial proportion of CD8⁺ T cells (Fig. 6A and B), we restimulated the culture with PCI-13 cells and 24 h later, performed intracytoplasmic staining for IFN-γ and granzyme B. By flow cytometry, it was possible to discriminate between CD28⁺ and CD28^- cells and to define their functions (Fig. 7). We observed that both CD8⁺CD28⁺ and CD8⁺CD28^- cells had strong granzyme-B and IFN-γ expression in response to the tumor (Fig. 7). In these cultures, the expanding CD28^- cell population retained substantial antitumor functions.

Finally, when PBMC were cultured in the presence of IFN-γ treated PCI-13, which had significantly enhanced MHC class I expression (data not shown), down-regulation of CD28 was diminished (Fig. 6B). When PCI-13 transduced with the human B7.1 gene and overexpressing B7.1 was used for weekly stimulations, the shift from CD28⁺ to CD28^- effector cells was suppressed (Fig. 6B). These data indicate that strong ligation of TCR by MHC molecules or of B7.1 by CD28 prevented down-regulation of CD28 in tumor-stimulated CD8⁺ T cells. Furthermore, no or little apoptosis was detected in these cultures.