Phenotypically and Functionally Distinct CD8⁺ Lymphocyte Populations in Long-Term Drug-Free Tolerance and Chronic Rejection in Human Kidney Graft Recipients

Abstract

A substantial proportion of long-term kidney graft recipients, including those with a stable renal function in the absence of immunosuppressive therapy, present a skewed T cell receptor (TCR) Vβ chain usage, essentially in the CD8⁺ subset. This study analyzed in more detail phenotypical and functional alterations of CD8⁺ lymphocytes in drug-free tolerant patients (DF-Tol) compared with recipients with chronic rejection (CR). Phenotyping revealed a significant increase in central memory and a decrease in effector CD8⁺ lymphocytes in DF-Tol versus CR. The expression of CD28⁺ and CD27⁺ on these effector cells was significantly decreased in CR. These profiles were stable over time and independent of treatment. Functionally, the CD8⁺CD28⁻ lymphocytes were less sensitive to apoptosis than their CD8⁺CD28⁺ counterparts, without differences in polyclonal proliferation. The CD8⁺CD28⁻ cells did not express GITR and FoxP3 but were characterized by high levels of preformed perforin and granzyme A, pointing toward a cytotoxic rather than a suppressor function. CD8⁺CD28⁻ lymphocytes did not show antigen-specific degranulation when co-cultured with targets that bear donor HLA class I antigens, suggesting that the cytotoxicity is directed either to other determinants of the graft or to nongraft epitopes. Of interest, CD8⁺ cells from DF-Tol displayed the same profile as healthy individuals, indicating an increase in CD8⁺CD28⁻ effector lymphocytes in CR rather than a decrease in DF-Tol. CD8⁺ lymphocytes from stable kidney recipients under conventional maintenance immunosuppression displayed a mixed profile, independent of treatment and time of sampling. Taken collectively, these data show a strong cytotoxicity-associated CD8⁺CD28⁻ signature in CR and suggest a suppression of pathologic cytotoxicity in DF-Tol. Further prospective studies should assess whether serial CD8⁺ phenotyping may help to identify patients who are at risk for CR when immunosuppression is tapered.

Drug-free tolerance, defined as long-term maintenance of graft integrity and function without immunosuppression, is a rare event in human kidney transplantation because interruption of immunosuppressive treatment usually leads to acute or chronic graft rejection. Nevertheless, this phenomenon is of unique interest to study the physiologic basis of graft tolerance in humans. On the one hand, long-term drug-free tolerant patients (DF-Tol) represent a unique model to study the extent to which mechanisms of tolerance defined experimentally, such as active suppression by regulatory lymphocytes, ignorance of alloantigens, chimerism, homeostatic regulation or clonal deletion, are relevant to this human situation (^1–4). Most studies in rodents analyzed the induction rather than the maintenance phase of tolerance, and discrepancies with the human situation may exist, as exemplified by the role of alloreactive CD8⁺ central memory cells in rejection and tolerance induction (^5,6). On the other hand, the characterization of peculiar immunologic profiles in DF-Tol may be clinically important to identify biologic signatures that are associated with graft tolerance. Considering the major medical and economic burden of chronic immunosuppression and that operational tolerance may be more common than expected but could be masked in long-term immunosuppressed patients, the identification of specific biologic signatures of tolerance could open new perspectives for rational rather than empiric minimizing of immunosuppressive drugs in well-selected patients (^6–10).

As a proof of concept of the relevance of the DF-Tol model to study human tolerance, we described recently a number of specific immunologic features in these patients. First, DF-Tol were characterized by a maintenance of CD25^highCD4⁺ lymphocytes that express regulation-associated molecules such as FoxP3, CTLA4, GITR, CCR4, and CD103, in comparison with patients with chronic rejection of their allograft (CR) (Louis et al., unpublished observations). Second, peripheral blood T cells of a substantial proportion of DF-Tol and CR displayed a skewed T cell receptor (TCR) Vβ chain usage, which was observed mainly in the CD8⁺ subset (¹¹). The T cells with skewed Vβ profiles from DF-Tol were characterized by a decrease in cytokine transcripts (IL-2, IL-13, and IFN-γ) compared with CR, suggesting a state of hyporesponsiveness or anergy (¹¹).

On the basis of the data of this latter study, we performed here a systematic analysis of CD8 phenotypes in DF-Tol versus CR and compared this with healthy individuals (HI). Our data show that a population of CD8⁺CD28⁻ lymphocytes was dramatically increased in CR. This subpopulation was characterized further with regard to apoptosis and proliferation, regulatory markers such as GITR and FoxP3, and general as well as donor-specific cytotoxicity. Finally, we evaluated the presence of this CD8⁺ population in kidney graft recipients with stable graft function under standard maintenance immunosuppressive therapy (Sta).

Materials and Methods

Patients

A total of 38 individuals were included in the study. The protocol was approved by the University Hospital Ethical Committee and the Committee for the Protection of Patients from Biologic Risks. All patients signed a written informed consent before inclusion.

The study included six DF-Tol. Operational tolerance was clinically defined as stable graft function and absence of clinical or biologic signs of chronic rejection (blood creatinemia <150 mmol/L, proteinuria <1.5 g/24 h) for at least 2 yr (median 8 yr; range 2 to 12 yr) after complete interruption of all immunosuppressive therapy. Immunosuppressive treatment was stopped because of lymphoma in two patients and noncompliance in the four others (¹¹). These normally functioning kidneys were not biopsied for ethical reasons. The CR group included 14 kidney graft recipients with a degradation of the renal function (blood creatinemia >150 mmol/L) and histologically proven chronic rejection lesions. As control groups, we included six HI and 12 kidney graft recipients with stable renal function under Sta. Demographic and clinical data are shown in Table 1.

Phenotypic Characterization by Flow Cytometry

Peripheral blood mononuclear cells (PBMC) were isolated using a Ficoll gradient (Eurobio, Les Ulis, France) and incubated for 30 min with the following fluochrome-labeled mAb for phenotypic characterization: Anti–CD45RA-FITC, anti–CD57-FITC, anti–KIR-NKAT2-FITC, anti–CD94-FITC, anti–CD28-PE or anti–CD28-APC, anti–CD27-PE, anti–CD69-PE, anti–GITR-PE, anti–CCR7-PC7, anti–CD3-Cy-Chrome or CD3-PC7, anti–CD8-PE-Cy5.5 or anti–CD8-APC, anti–CD95-APC, and anti–NKG2D-APC (all mAb from BD Biosciences Pharmingen, San Diego, CA). For the assessment of intracellular proteins, PBMC were permeabilized with saponin 1% for 15 min, and intracellular perforin and granzyme A were stained with PE-labeled mAb (BD Biosciences Pharmingen). The labeled PBMC were washed, fixed in PBS/formaldehyde 1%, and analyzed by four-color flow cytometry (FACSCalibur, Becton Dickinson, San Diego, CA) using Cellquest Pro software (Becton Dickinson). T lymphocytes were identified using a forward and side scatter gate for lymphocytes in combination with a gate on CD3⁺ cells. Nonspecific staining and autofluorescence were determined by isotype-matched control mAb. Results are expressed as mean percentage of positive cells or as mean fluorescence intensity.

Detection of Apoptosis

For in vitro induction of apoptosis, PBMC were cultured for 18 h at 37°C in serum-free RPMI 1640 medium (Sigma, St. Louis, MO). As negative controls, PBMC were cultured in RPMI with 10% heat-inactivated FCS. After 18 h, PBMC were labeled with phenotypic markers as described and with annexin V-APC (BD Biosciences Pharmingen). The percentage of annexin V–positive cells was measured by flow cytometry after exclusion of dead cells by propidium iodide labeling.

CFSE Proliferation Assay

PBMC were stained with 1 μM of carboxyfluorescein diacetate succinimidyl ester (CFSE) for 3 min, washed extensively, and adjusted to 5 × 10⁵ cells/ml in RPMI 1640 with 10% FCS. CFSE-labeled PBMC were stimulated with 1 μg/ml plate-bound anti-CD3 antibody (Orthoclone OKT3; Janssen-Cilag, Germany) with or without 100 UI/ml IL-2 (Proleukin; Chiron Corp., Emeryville, CA). After 72 h, cells were stained for 15 min with anti–CD8-PE-Cy5.5 and anti–CD28-APC antibodies. Proliferation was analyzed by measuring the CFSE signal on gated CD8⁺CD28⁺ and CD8⁺CD28⁻ cells by flow cytometry.

Real-Time PCR

CD8⁺ T cells from six CR were negatively isolated by magnetic bead sorting (Milteny Biotec, Bergisch Gladbach, Germany). Then, CD8⁺CD27⁺ and CD8⁺CD27⁻ subsets were separated by a positive magnetic selection (Milteny Biotec). Because CD27 and CD28 expression correlated perfectly on CD8⁺ T lymphocytes, anti-CD27 rather than anti-CD28 beads were used to avoid cell activation. The sorted CD8⁺CD27⁺ and CD8⁺CD27⁻ populations contained >90% CD8⁺CD28⁺ and CD8⁺ CD28⁻ lymphocytes, respectively, as assessed by flow cytometry. Purified cells were frozen in Trizol reagent (Invitrogen Life Technologies, Carlsbad, CA) for RNA extraction according to the manufacturer’s instructions. Total mRNA was reverse-transcribed using a cDNA synthesis kit (Boehringer Mannheim, Indianapolis, IN). Real-time quantitative PCR was performed using labeled TaqMan probes specific of FoxP3 and normalized against the hypoxanthine phosphoribosyl transferase-1 (HPRT) transcript level, as described previously (¹¹).

Cytotoxicity-Associated Degranulation Assay

CD8⁺ lymphocytes of seven CR were assessed for antigen-specific degranulation by the CD107 mobilization assay, as described previously (^12,13). Because PBMC of the kidney graft donors were not available several years after transplantation, we collected from healthy blood donors surrogate PBMC that were matched for the HLA class I molecules of the kidney graft donors. Recipient PBMC from CR were incubated for 5 h either with irradiated surrogate PBMC at a 1:1 ratio or with a pool of common viral peptides at 10 μg/ml (gift of J.-G. Guillet, Institut Pasteur, Paris, France) in RPMI medium with 10% human serum and 2 μM monensin (Sigma-Aldrich). Unstimulated recipient PBMC were used as negative control, whereas plate-bound anti-CD3 antibody or phytohemagglutinin stimulation was used as positive control. Degranulation of CD8⁺CD28⁻ lymphocytes was assessed by flow cytometric analysis with a mix of anti–CD107a-FITC and anti–CD107b-FITC antibodies (BD Biosciences Pharmingen).

Statistical Analyses

The Mann-Whitney U test (for unpaired samples) and the Wilcoxon test (for paired samples) were used to assess differences between groups. Correlations were calculated with the Spearman’s ρ rank correlation test. P < 0.05 was considered as statistically significant. Classification of Sta according to their CD8⁺ phenotype was performed by Predictive Analysis of Microarray data software (¹⁴).

Results

Increase of CD8⁺CD28⁻ Effector Lymphocytes in CR

Peripheral blood CD8⁺ T lymphocytes from DF-Tol and age-matched CR were analyzed for the surface expression of CD45RA and CCR7 to distinguish naive (CD45RA⁺CCR7⁺), effector (CD45RA⁺CCR7⁻), central memory (CD45RA⁻CCR7⁺), and effector memory (CD45RA⁻CCR7⁻) CD8⁺ lymphocytes (^15,16). As shown in Table 2, there was a significant increase in central memory (P = 0.032) and decrease in effector (P = 0.048) CD8⁺ lymphocytes in DF-Tol versus CR. Of interest, the percentage of CD45RA⁺CCR7⁻ effector CD8⁺ lymphocytes in age-matched HI was similar to DF-Tol but significantly lower than in CR (P = 0.048), indicating that these differences correspond to an increase in CD8⁺ effectors in CR rather than a decrease in DF-Tol.

Effector as well as effector memory CD8⁺ cells can be subdivided further according to the loss of surface expression of CD28 and CD27 during terminal differentiation (¹⁵). As shown in Table 2, the percentage of CD28⁺ (P = 0.040) and CD27⁺ (P = 0.018) CD8⁺ effector lymphocytes was significantly higher in DF-Tol than in CR. A similar difference in CD28 (P = 0.002) and CD27 (P = 0.002) expression was observed in the effector memory subset. This was also reflected by a significantly higher expression of both CD28 (P = 0.001) and CD27 (P = 0.006) on the global CD8⁺ population, with a high correlation between both markers (r = 0.91, P < 0.001; Figure 1). Analysis of age-matched HI revealed that the expression of CD28 and CD27 on the different CD8⁺ lymphocyte subsets was similar in DF-Tol and HI but significantly decreased in CR (Table 2). Taken together, these data indicate an increase in CD8⁺CD28⁻ effector lymphocytes in CR, whereas DF-Tol exhibited a pattern close to that of HI.

Increase of Effector CD8⁺CD28⁻ Lymphocytes in CR Is Stable over Time and Independent of Treatment

To investigate whether the increase of CD8⁺CD28⁻ lymphocytes was a stable phenomenon related directly to CR, paired PBMC that were obtained in five patients at two different time points with an interval of 7.4 (2 to 16) months were analyzed. There was no significant variation of the percentage of CD45RA⁺CCR7⁻CD8⁺ effector cells (35.5 ± 8.5 versus 29.4 ± 7.3%) or of the expression of CD28 (19.2 ± 10.7 versus 17.5 ± 10.7%) and CD27 (32.3 ± 13.8 versus 34.4 ± 19.7%), indicating that the observed CD8 profile was stable over time (Figure 2A).

Considering the heterogeneity of treatment in CR (Table 1), we next analyzed the potential effect of treatment on these profiles. Comparison of CR who were treated or not with corticosteroids, calcineurin inhibitors (CNI), or mycophenolate mofetil did not show a significant difference for the percentage of CD45RA⁺CCR7⁻CD8⁺ effector cells (37.4 ± 12.6 versus 29.2 ± 9.4%, 30.4 ± 11.3 versus 41.1 ± 6.8%, and 37.1 ± 10.9 versus 29.4 ± 11.0%, respectively), CD28 expression (15.1 ± 9.7 versus 24.0 ± 14.3%, 21.5 ± 14.0 versus 15.4 ± 13.1%, and 18.4 ± 10.9 versus 21.5 ± 15.9%, respectively), or CD27 expression (26.8 ± 13.1 versus 36.2 ± 18.4%, 32.6 ± 16.8 versus 26.7 ± 14.4%, and 27.5 ± 12.9 versus 33.1 ± 17.7%, respectively; Figure 2, B through D). Accordingly, after exclusion of corticosteroid-, CNI-, or mycophenolate mofetil–treated CR, there was still a significant decrease of CD28 expression (P = 0.001, P < 0.001, and P = 0.001, respectively) and of CD27 expression (P = 0.032, P < 0.001, and P = 0.012, respectively) in CR versus DF-Tol with a similar trend toward increase of CD8⁺ effector lymphocytes.

Alterations of Apoptosis but not Proliferation of CD8⁺CD28⁻ Lymphocytes

To characterize further the increase of CD8⁺CD28⁻ effector lymphocytes in CR, we studied the sensitivity to apoptosis and the proliferative capacities of these cells in comparison with their CD8⁺CD28⁺ counterparts. There was no significant difference in Fas expression on the CD8⁺CD28⁻ lymphocytes in DF-Tol versus CR (data not shown). However, although nearly all CD8⁺CD28⁻ cells expressed Fas (percentage of positive cells), the expression levels (mean fluorescence intensity) were significantly lower than on their paired CD8⁺CD28⁺ counterparts (63.3 ± 9.0 versus 103.3 ± 34.2; P = 0.002; Figure 3A). Accordingly, the CD8⁺CD28⁻ subset was less sensitive to apoptosis induced by serum deprivation than the paired CD8⁺CD28⁺ cells (15.5 ± 5.2 versus 31.1 ± 10.0%; P = 0.002; Figure 3B).

As to the proliferative capacity, there was no significant difference between the CD8⁺CD28⁻ and CD8⁺CD28⁺ T subsets with regard to the percentage of cells that proliferated upon anti-CD3 stimulation (46.1 ± 17.8 and 52.1 ± 23.5%, respectively; Figure 3C). Addition of IL-2 did not further increase the proliferation of either subset. Also, the degree of proliferation assessed by the number of divisions per cell was not different between both subsets (Figure 3C). Finally, no significant difference was observed in proliferative capacity between DF-Tol and CR for these two populations (data not shown).

CD8⁺CD28⁻ Lymphocytes in CR Have No Characteristics of Suppressor T Cells

To explore the functional features of the CD8⁺CD28⁻ lymphocytes in CR, we first compared these cells with the recently described CD8⁺CD28⁻ suppressor lymphocytes (^17–22). As indicated previously, the CD8⁺CD28⁻ cells that were detected in CR were CD45RA⁺CCR7⁻CD27⁻, which contrasted with the described phenotype of the suppressor lymphocytes (CD45RO⁺CD62L⁺CD27⁺) (^21,22). Flow cytometric analysis of GITR, which is expressed on suppressor cells (^19,21), revealed low expression on CD8⁺ cells of CR (2.8 ± 2.4%) compared with DF-Tol (12.1 ± 6.2%; P = 0011) and HI (16.0 ± 15.5; P = 0.033; Figure 4A). Accordingly, real-time PCR showed low levels of FoxP3 transcripts in CD8⁺ lymphocytes from CR compared with DF-Tol (P = 0.006) and HI (NS; Figure 4B). Moreover, real-time PCR on purified lymphocyte subsets in CR showed that FoxP3 expression was very low in both the CD8⁺CD28⁺CD27⁺ and CD8⁺CD28⁻CD27⁻ cells (Figure 4C). Taken together, these data clearly distinguish the CD8⁺CD28⁻ population that was detected in vivo in CR from the previously described CD8⁺CD28⁻ suppressor lymphocytes that were obtained after several rounds of in vitro stimulation.

CD8⁺CD28⁻ Lymphocytes Exhibit Markers of Differentiated Cytotoxic Cells

Considering that the CD8⁺CD28⁻ lymphocytes in CR were different from suppressor cells and belonged to the effector subset, we next analyzed the presence of markers of cytotoxicity. There was a highly significant increase of the number of CD8⁺ lymphocytes that were positive for intracellular perforin (85.6 ± 16.8 versus 30.1 ± 5.3%; P < 0.001) and for granzyme A (32.7 ± 15.7 versus 6.8 ± 8.4%; P = 0.010) in CR versus DF-Tol (Figure 1). However, there was no difference for perforin or granzyme A between HI (30.1 ± 10.1 and 5.3 ± 7.2%, respectively) and DF-Tol. Perforin and granzyme A correlated strongly (r = 0.723, P = 0.009) and both markers correlated inversely with CD28 expression (r = −0.947, P < 0.001; and r = −0.745, P = 0.007, respectively). Also the surface expression of CD57 was significantly increased on the CD28⁻ subset (67.0 ± 20.4%) compared with the CD28⁺ counterpart (22.2 ± 10.6%; P = 0.031; Figure 5). Whereas a similar difference in CD57 expression was found between both subsets in DF-Tol (70.9 ± 12.7 versus 26.2 ± 16.5%; P = 0.046), there was no difference for the expression of these cytotoxicity-associated markers on the CD8⁺CD28⁻ lymphocyte subset between CR and DF-Tol, indicating that the number of these cells rather than their cytotoxicity-associated profile differentiated both situations (Figure 5). Finally, the analysis of the expression NK-cell receptors (KIR-NKAT2, CD94, and NKG2D) on CD8⁺ lymphocytes showed no difference between DF-Tol and CR or between the CD8⁺CD28⁻ and CD8⁺CD28⁺ subsets in both situations (Figure 5).

To identify the cytotoxic targets of the CD8⁺CD28⁻ lymphocytes in CR, we next performed a CD107 mobilization assay to assess donor antigen-specific degranulation. Unstimulated CD8⁺CD28⁻ cells showed low levels of degranulation (6.0 ± 2.0%), which were significantly increased upon nonspecific stimulation with anti-CD3 antibody (14.5 ± 3.9%) or phytohemagglutinin (37.9 ± 21.4%). Incubation with surrogate donor cells, which were matched for HLA class I with the original kidney graft donor, did not increase degranulation (4.9 ± 2.5%). Similarly, stimulation with a pool of common viral peptides did not significantly increase the degranulation of CD8⁺CD28⁻ T lymphocytes (8.2 ± 3.9%). Taken together, these data do not provide functional evidence that the targets of this cell population are donor-specific HLA class I molecules or common viral antigens.

Sta Recipients Display a Mixed CD8⁺ Lymphocyte Profile

Finally, we analyzed whether the described CD8⁺ lymphocyte profile that was associated with the chronic rejection process could be observed also in immunosuppressed kidney graft recipients without clinical or biologic signs of rejection. The CD8⁺ lymphocytes of five patients showed a similar CD28, CD27, and granzyme A expression as DF-Tol, whereas two had an intermediate profile and five had similar expression levels as CR (Figure 6, A through D). As to the percentage of effector CD45RA⁺CR7⁻CD8⁺ cells, seven Sta recipients had similar numbers as CR, whereas five had low numbers such as in DF-Tol. Similar to our observations in CR, these profiles were not dependent on the treatment with CNI (Figure 6E) or other immunosuppressive drugs (data not shown), and a kinetic analysis in five patients at a median of an 18-mo interval (16 to 24 mo) showed that these profiles were stable over time (Figure 6F). To analyze the Sta profiles in more detail, we set up a model using Predictive Analysis of Microarray data software based on the CD8⁺ lymphocyte phenotypes in DF-Tol versus CR. Cross-validation of this model using the DF-Tol and CR profiles classified correctly 13 of the 14 CR samples and five of the six DF-Tol samples (Figure 6G). Applying this model to the Sta patient cohort, the CD8⁺ lymphocyte phenotype of seven patients resembled CR, whereas the others had a profile more closely related to DF-Tol (Figure 6H).

Discussion

Considering the major burden of immunosuppression in organ transplantation and the discrepancies between tolerance in rodents and humans, spontaneous drug-free tolerance in humans is a rare but unique model to study clinically relevant immune phenomena related to graft integrity and survival. In this context, we analyzed alterations of the CD8⁺ lymphocytes in kidney graft recipients who tolerated their graft for several years without any immunosuppressive or corticosteroid treatment. The main finding is the reduced numbers of circulating CD8⁺CD28⁻ effector lymphocytes compared with patients with chronic graft rejection. However, the DF-Tol had a similar number of CD8⁺CD28⁻ effector lymphocytes as HI, indicating an abnormal increase of this cell population in CR rather than a primary alteration in DF-Tol. Because the expression of CD28 on CD8⁺ cells decreases with age (²³), it is important to notice that the study groups were age matched. A bias caused by immunosuppressive therapy is also unlikely because we found no effect of the different treatment regimens on the CD8⁺ lymphocyte profiles in CR. Finally, the increase in CD8⁺CD28⁻ effector cells was stable over time in CR.

Several mechanisms can lead to an increase of CD8⁺CD28⁻ lymphocytes. During aging, effector CD8+ lymphocytes can lose the surface expression of CD28 by replicative senescence as a result of extensive homeostatic proliferation (^16,23). A distinct mechanism is observed after TCR-mediated activation, which is associated mostly with a decreased susceptibility to undergo apoptosis (²⁴). In this respect, CD8⁺CD28⁻-cell populations can appear as a consequence of extensive rounds of antigen-induced division such as in chronic infections or malignancies (^25–27) and, in contrast with senescent cells, are characterized by increased effector functions rather than functional anergy (^13,28,29). Investigation of the sensitivity to apoptosis and the proliferative capacity of the CD8⁺CD28⁻ lymphocytes in CR indicated that the loss of CD28 was associated with a decreased susceptibility to apoptosis, which related both to the Fas-mediated pathway and to the sensitivity to growth factor withdrawal (^30–32). However, in contrast with replicative senescence, which is characterized by an irreversible nondividing state (^30,33,34), the CD8⁺CD28⁻ lymphocytes in CR proliferated at a similar level as their CD28⁺ counterparts upon CD3 stimulation. Because CD28 is required for IL-2 production and subsequent sustained proliferation, it is likely that both in our assay and in vivo IL-2 could be provided in a paracrine manner by CD28⁺ lymphocytes rather than in an autocrine manner by the CD8⁺CD28⁻-lymphocytes themselves (³⁵). The impaired sensitivity to apoptosis and the maintained proliferative potential of the CD8⁺CD28⁻ lymphocytes may influence the balance between clonal expansion and contraction and thereby contribute to the increase of this cell population in CR.

A subset of CD8⁺CD28⁻ cells have been described recently as suppressor lymphocytes, which, in contrast to our study, were obtained by multiple rounds of stimulation in vitro (^17,18). These cells were described to express GITR and FoxP3 and to suppress CD4 responses by tolerization of antigen-presenting cells through an upregulation of the inhibitory molecules ILT3 and ILT4 (^19,20). Whereas animal models support an in vivo function for these suppressor lymphocytes (^36,37), their natural presence, exact phenotype, and suppressive function in humans still are a matter of debate (²²). The suppressor CD8⁺CD28⁻ lymphocytes seem to be central memory cells (CD62L⁺CD45RO⁺) that co-express CD27 (^21,22), in contrast with the cell population in CR, which are CD27⁻ effector cells (CCR7⁻CD45RA⁺). Moreover, the regulatory-associated markers GITR and FoxP3, which were described on the suppressor cells (^19,21), were not expressed on CD8⁺ lymphocytes in CR, thereby clearly indicating that the CD8⁺CD28⁻ cells that we described here are different from suppressor lymphocytes.

In contrast to a suppressor function and in agreement with the association between loss of CD28 and marked cytotoxicity of CD8⁺ effector cells (^13,28,29), the CD8⁺CD28⁻ effector population in CR was characterized by high levels of perforin, granzyme A, and CD57. Of interest, however, we found no differences for these markers in CD8⁺CD28⁻ effector lymphocytes between DF-Tol and CR, indicating a quantitative rather than a qualitative difference between both situations. An important question raised by this observation is the target of these cells and the potential functional consequences for the graft. On the basis of reports in animal models (^38–41), it would be tempting to speculate that these cells are induced by donor antigens by either a direct or an indirect pathway of allorecognition. Using donor HLA class I–matched cells as surrogate targets, however, our attempt to provide functional evidence that the described CD8⁺CD28⁻ lymphocyte subset is indeed committed to donor determinants was unsuccessful. Whereas our study is hampered by the lack of original donor cells, future prospective and/or experimental research is needed to address this in more detail and to evaluate the eventual contribution of other donor determinants. An alternative hypothesis would be that pre-existing cytotoxic CD8⁺ lymphocytes directed against viruses and pathogens cross-react with the graft, as suggested by the fact that heterologous immune memory has been described as a potential to transplantation tolerance (^5,6,42,43). Also here, however, we were unable to demonstrate antigen-specific degranulation of CD8⁺CD28⁻ lymphocytes against a pool of common viral peptides. A third and most interesting possibility that should be explored further is reactivity against self-determinants as suggested by recent experimental evidence (^44,45).

Independent of the precise primary target of the CD8⁺CD28⁻ effector cells in CR, the normal expression level of the activating cytotoxic receptor NKG2D, which is not counterbalanced by an increase of MHC class I–binding inhibitory receptors such as CD94/NKG2A and KIR-NKAT2, is compatible with functional cytotoxicity of these cells (^46–49). In this context, it is interesting to note that cell-mediated alloimmunity has been demonstrated to contribute to chronic allograft nephropathy in renal transplant recipients (⁵⁰). Moreover, Li et al. (⁵¹) indicated recently that operational tolerance in liver transplantation was associated with a decrease of NK cells, a distinct cytotoxic lymphocyte population that was not investigated in this study.

A final important issue is raised by the fact that some of the patients who had stable graft function and were analyzed in our study display a similar CD8⁺ lymphocyte phenotype as CR, whereas others had an intermediate profile. This indicates that the described CD8 profile is strongly but not exclusively associated with active chronic rejection. Because an increase of granzymes and perforin was also reported to precede allograft rejection in rodents as well as humans (^52,53), it would be interesting to analyze whether such a CR-like signature in seemingly stable patients may help to identify a poorer prognosis and eventually subsequent rejection. All stable patients who were analyzed in our study, however, maintained a normal graft function for now almost 3 yr of follow-up.

These data also emphasize the complexity of such studies in human subjects. First, the large interindividual variability precludes conclusive statements in cross-sectional studies and requires validation by serial, prospective analyses in which each individual with changes in clinical status can serve as his own control over time. In this context, a gradual increase of a cytotoxic CD8⁺CD28⁻ population over time may be more relevant than an absolute value as such at the individual level. Second, because most of the patients with kidney transplantations remain stable for years under conventional immunosuppression, only large studies over a longer period will be able to relate the described phenotype to poor clinical outcome. An alternative would be a prospective tapering of the immunosuppression in patients with a profile close to DF-Tol, but this is still medically and ethically unacceptable without a clear, coherent, and reliable signature based on a broader panel of immunologic parameters (¹¹). Finally, studies on spontaneous tolerance in kidney transplantation are severely hampered by the extreme rarity of these patients. Our studies have analyzed the largest number of these patients in the current literature, making it difficult to validate these findings in an independent cohort. Performing similar analyses in other types of solid organ transplantation, where tolerance is more frequently observed, such as for liver, therefore may be an interesting alternative (⁵¹).

Figure 1: Analysis of the expression of CD28 on the cell surface and of intracellular perforin in peripheral blood CD8⁺ lymphocytes in drug-free tolerant kidney graft recipients (DF-Tol) and in patients with chronic rejection of the graft (CR). Representative histograms show the increased expression of CD28 and the decrease of intracellular perforin in DF-Tol versus CR.

Figure 2: Stability of the effector CD8⁺CD28⁻CD27⁻ lymphocyte population in CR. Peripheral blood CD8⁺ lymphocytes were analyzed by flow cytometry for the percentage of CD45RA⁺CCR7⁻ effector (EFF) cells, CD28⁺ cells, and CD27⁺ cells. Comparisons were performed between two different time points in five patients (A). In addition, patients with (n = 6) and without (n = 8) corticosteroid treatment (B), patients with (n = 11) and without (n = 3) calcineurin inhibitor (CNI) treatment (C), and patients with (n = 6) and without (n = 8) mycophenolate mofetil treatment (D) were compared. Data are represented as mean ± SD. None of the comparisons was statistically significant.

Figure 3: Analysis of apoptosis and proliferation of CD8⁺CD28⁻ and CD8⁺CD28⁺ peripheral blood lymphocytes. Expression of Fas (CD95), assessed as mean fluorescence intensity (MFI) by flow cytometry, was significantly lower in CD8⁺CD28⁻ lymphocytes than in the paired CD8⁺CD28⁺ counterparts (P = 0002; A). Accordingly, these cells were less sensitive to apoptosis upon serum deprivation, as assessed by annexin V staining (P = 0002; B). In contrast, there were no significant differences for the total number of cells that proliferated or for the number of divisions per cell upon anti-CD3 stimulation of CD8⁺CD28⁺ and CD8⁺CD28⁻ lymphocytes, as assessed by CFSE-based flow cytometry (C).

Figure 4: Analysis of the expression of the regulatory markers GITR and FoxP3 on CD8⁺ peripheral blood lymphocytes in CR in comparison with DF-Tol and healthy individuals (HI). Expression of GITR as assessed by flow cytometry was significantly lower on CD8⁺ lymphocytes of CR compared with DF-Tol (P = 0011) and HI (P = 0033; A). Similarly, real-time PCR indicated that the FoxP3 mRNA levels were decreased in CD8⁺ cells from CR compared with DF-Tol (P = 0006) and HI (NS; B). Moreover, real-time PCR analysis of purified lymphocyte subsets of CR showed that FoxP3 expression was very low in both the CD8⁺CD28⁺(CD27⁺) and CD8⁺CD28⁻(CD27⁻) cells (C). Data are represented as mean ± SD.

Figure 5: Flow cytometric assessment of the expression of markers associated with cytotoxicity and NK-cell receptors on the cell surface of peripheral blood CD8⁺ lymphocytes in CR and in DF-Tol. The CD8⁺CD28⁻ subset was compared with the CD8⁺CD28⁺ subset. Data are represented as mean ± SD. *P < 0.05.

Figure 6: CD8⁺ lymphocyte profiles as assessed by flow cytometry on peripheral blood lymphocytes of 12 kidney graft recipients with stable renal function under immunosuppressive therapy (Sta). The percentage of CD8⁺ lymphocytes that expressed CD28 (A), CD27 (B), and intracellular granzyme A (C) and the number of effector CD45RA⁺CCR7⁻CD8⁺ lymphocytes (D) are shown. For comparison, the mean ± SD is depicted for DF-Tol (n = 6) and CR (n = 14). Similar to the data in CR, there was no influence of treatment with CNI on the CD8⁺ T lymphocyte profiles in Sta (E), and kinetic analysis of five patients at an interval of 18 mo showed that these profiles were stable over time (F; data are represented as mean ± SD). Using the CD8⁺ lymphocyte phenotypes from CR and DF-Tol, a model was set up for classification of individual samples. Cross-validation indicated that 13 of 14 CR and five of six DF-Tol were classified correctly (G). When applied to samples of Sta patients, the model classified the CD8⁺ lymphocyte profiles of seven of 12 as CR, whereas five showed an intermediate profile (H).

Table 1: Demographic and clinical data of the patient cohortsa

Table 2: Phenotypic analysis of CD8⁺ T lymphocytes in DF-Tol, CR, and age-matched HIa

D.B. is a senior clinical investigator of the Fund for Scientific Research-Flanders (FWO-Vlaanderen).

We thank Drs. Bignon and Cury (EFS, Nantes, France) for providing the HLA class I matched target cells and Dr. J.-G. Guillet (Cochin Hospital, Paris, France) for providing the viral peptide pool.