Mesenchymal stem cells control alloreactive CD8⁺CD28⁻ T cells

Abstract

CD28/B7 co-stimulation blockade with belatacept prevents alloreactivity in kidney transplant patients. However, cells lacking CD28 are not susceptible to belatacept treatment. As CD8⁺CD28⁻ T-cells have cytotoxic and pathogenic properties, we investigated whether mesenchymal stem cells (MSC) are effective in controlling these cells. In mixed lymphocyte reactions (MLR), MSC and belatacept inhibited peripheral blood mononuclear cell (PBMC) proliferation in a dose-dependent manner. MSC at MSC/effector cell ratios of 1:160 and 1:2·5 reduced proliferation by 38·8 and 92·2%, respectively. Belatacept concentrations of 0·1 μg/ml and 10 μg/ml suppressed proliferation by 20·7 and 80·6%, respectively. Both treatments in combination did not inhibit each other's function. Allostimulated CD8⁺CD28⁻ T cells were able to proliferate and expressed the cytolytic and cytotoxic effector molecules granzyme B, interferon (IFN)-γ and tumour necrosis factor (TNF)-α. While belatacept did not affect the proliferation of CD8⁺CD28⁻ T cells, MSC reduced the percentage of CD28⁻ T cells in the proliferating CD8⁺ T cell fraction by 45·9% (P = 0·009). CD8⁺CD28⁻ T cells as effector cells in MLR in the presence of CD4⁺ T cell help gained CD28 expression, an effect independent of MSC. In contrast, allostimulated CD28⁺ T cells did not lose CD28 expression in MLR–MSC co-culture, suggesting that MSC control pre-existing CD28⁻ T cells and not newly induced CD28⁻ T cells. In conclusion, alloreactive CD8⁺CD28⁻ T cells that remain unaffected by belatacept treatment are inhibited by MSC. This study indicates the potential of an MSC–belatacept combination therapy to control alloreactivity.

Introduction

CD28/B7 co-stimulation blockade to prevent T cell activation and proliferation has been of interest for many therapeutic areas [1]. Belatacept, the latest immunosuppressive drug approved for therapy of kidney transplant recipients, utilizes this blocking mechanism. It is a fusion protein consisting of the extracellular domain of cytotoxic T lymphocyte antigen-4 (CTLA-4) and the Fc region of a human immunoglobulin (Ig)G1 immunoglobulin. By binding to CD80 (B7.1) and CD86 (B7.2) with a higher affinity than CD28, belatacept blocks the co-stimulatory signal [2]. However, as a consequence, belatacept treatment is not effective in impairing T cells that lack CD28 expression. While at birth all T cells express CD28, the CD8⁺ T cell compartment of an adolescent individual contains CD28⁻ cells at a frequency of up to 20–30% [3,4]. Persistent antigenic stimulation during ageing or, in an accelerated manner, through infection with cytomegalovirus (CMV) causes down-regulation of CD28 expression on CD8⁺ T cells [5,6]. The presence of these CD8⁺CD28⁻ T cells is associated with oncological diseases and autoimmune diseases such as rheumatoid arthritis, multiple sclerosis and diabetes [7–10]. In addition, their highly antigen-experienced nature and cytotoxic phenotype may pose a risk for graft rejection after organ transplantation. The insusceptibility of alloreactive CD8⁺CD28⁻ T cells to belatacept discloses a gap in the immunosuppressive activity of this drug. Therefore, CD28/B7-blocking agents may need to be combined with a therapy that targets CD28⁻ T cells.

A potential therapeutic approach could be the administration of mesenchymal stem cells (MSC). MSC possess immunomodulatory properties and their function has been established in vitro and in animal models [11,12]. First MSC trials in humans for multiple disease areas such as autoimmune diseases, graft-versus-host disease (GVHD) and allograft rejection produced encouraging results [13–16]. Activated MSC inhibit cells of the innate and adaptive immune system and of central interest in MSC research is their suppression of T cell-mediated immunity, as MSC inhibit the proliferation of CD4⁺ and CD8⁺ T cells [17]. MSC mediate their immunosuppressive effect in an CD28-independent manner through direct contact with their target cells and through various soluble factors such as human hepatocyte growth factor (HGF), indoleamine 2,3-dioxygenase (IDO), interleukin (IL)-10, prostaglandins and transforming growth factor (TGF)-β [18].

The aim of our study was to investigate whether MSC can inhibit the alloreactivity of CD8⁺CD28⁻ T cells which escape belatacept treatment and to explore whether MSC are a potential candidate for combination therapy with belatacept.

Material and methods

Origin, isolation and culture of human MSC

Perirenal adipose tissue was surgically removed from living kidney donors and collected in minimum essential medium Eagle's alpha modification (MEM-α) (Sigma-Aldrich, St Louis, MO, USA) supplemented with 2 mM L-glutamine (Lonza, Verviers, Belgium) and 1% penicillin/streptomycin solution (P/S; 100 IU/ml penicillin, 100 IU/ml streptomycin; Lonza). Samples were obtained with written informed consent as approved by the Medical Ethical Committee at Erasmus MC, University Medical Center Rotterdam (protocol no. MEC-2006-190).

MSC were isolated, cultured and characterized as described previously [19]. In brief, perirenal adipose tissue was disrupted mechanically and digested enzymatically with collagenase type IV (Life Technologies, Paisley, UK). MSC were expanded using MSC culture medium consisting of MEM-α with 2 mM L-glutamine, 1% P/S and 15% fetal bovine serum (FBS; Lonza) in a humidified atmosphere with 5% CO₂ at 37°C. Culture medium was refreshed twice weekly. At subconfluency, MSC were removed from culture flasks using 0·05% trypsin–ethylenediamine tetraacetic acid (EDTA) (Life Technologies) and reseeded at 1000 cells/cm². MSC were characterized by means of immunophenotyping and by their ability to differentiate into adipocytes and osteoblasts. MSC cultured between two to six passages were used. MSC from these passages did not differ in their ability to differentiate or to exert their immunosuppressive functions.

Isolation of peripheral blood mononuclear cells

Peripheral blood mononuclear cells (PBMC) were isolated from buffy coats of healthy blood donors (Sanquin, Rotterdam, the Netherlands) by density gradient centrifugation using Ficoll-Paque PLUS (density 1·077 g/ml; GE Healthcare, Uppsala, Sweden). Cells were frozen at −150°C until further use in RPMI-1640 medium with GlutaMAX™-I (Life Technologies) supplemented with 1% P/S, 10% human serum (Sanquin) and 10% dimethylsulphoxide (DMSO; Merck, Hohenbrunn, Germany).

Mixed lymphocyte reaction and suppression assays

Mixed lymphocyte reactions (MLR) were set up with 5 × 10⁴ effector PBMC and 5 × 10⁴ γ-irradiated (40 Gy) allogeneic PBMC in round-bottomed 96-well plates (Nunc, Roskilde, Denmark). MLR were cultured in MEM-α supplemented with 2 mM L-glutamine, 1% P/S and 10% heat-inactivated human serum for 7 days in a humidified atmosphere with 5% CO₂ at 37°C. Effector–stimulator cell combinations were chosen on the basis of a minimum of four human leucocyte antigen (HLA) mismatches.

The immunomodulatory capacities of MSC and belatacept (Bristol-Myers-Squibb, New York, NY, USA) on MLR were determined in suppression assays. For flow cytometric analysis, effector PBMC were labelled with BD Horizon violet cell proliferation dye 450 (VPD450; BD Biosciences, San Jose, CA, USA). For distinction from effector PBMC, γ-irradiated allogeneic stimulator PBMC (40 Gy) were labelled using the PKH26 Red Fluorescent Cell Linker Kit (Sigma-Aldrich). When cell proliferation was assessed by thymidine incorporation, [³H]-thymidine (0·25 μCi/well; PerkinElmer, Groningen, the Netherlands) was added on day 7, incubated for 8 h and its incorporation was measured using the Wallac 1450 MicroBeta TriLux (PerkinElmer).

MLR with sorted CD8⁺CD28⁻ T cells, CD28⁻ T cells or CD28⁺ T cells

PBMC were stained with monoclonal antibodies (mAbs) against CD3 (AmCyan), CD4 [allophycocyanin (APC)], CD8 [fluorescein isothiocyanate (FITC)], CD28 [peridinin chlorophyll-cyanin 5·5 (PerCP-Cy5·5)] and either CD3⁺CD8⁺CD28⁻ cells and CD3⁺CD4⁺ cells or CD3⁺CD28⁻ cells and CD3⁺CD28⁺ cells were sorted on the BD FACSAria II cell sorter (BD Biosciences). Effector populations for MLR consisted either of CD3⁺CD28⁻ cells only (mean purity 97·8%, range 96·3–98·8%), CD3⁺CD28⁺ cells only (mean purity 96·2%, range 93·0–99·5%) or a combination of 10% CD3⁺CD8⁺CD28⁻ cells (mean purity 92·3%, range 88·4–94·72%) and 90% CD3⁺CD4⁺ cells to provide help (mean purity 98·2%, range 97·2–99·5%). All effector fractions were labelled with VPD450 (BD Biosciences) to track cell proliferation. To distinguish irradiated allogeneic stimulator PBMC from effector cells they were labelled with PKH26 (Sigma-Aldrich). Effector–stimulator cell combinations were chosen on the basis of a minimum of four HLA mismatches. MLR were set up in the absence or presence of MSC (1:10; MSC/effector cells) and belatacept (1 μg/ml). After a 7-day incubation period, cells were restained with mAbs against CD3 (AmCyan), CD4 (APC), CD8 (FITC), CD28 (PerCP-Cy5·5) and analysed on the BD FACSCanto II flow cytometer using the BD FACSDiva software (BD Biosciences).

Intracellular and extracellular staining of CD8⁺CD28⁻ T cells

MLR were set up in the absence of MSC. To track cell proliferation, effector PBMC were labelled with VPD450. After 7 days, cells were restimulated with phorbol 12-myristate 13-acetate (PMA; 50 ng/ml; Sigma-Aldrich) and ionomycin (1 μg/ml; Sigma-Aldrich) in the presence of GolgiPlug (BD Biosciences). Following a 4-h incubation period, cells were stained with mAbs against CD3 (AmCyan), CD4 (APC), CD8 (FITC), CD28 (PerCP-Cy5·5), tumour necrosis factor (TNF)-α [pyycoerythrin (PE)], interferon (IFN)-γ (PE; all BD Biosciences) and granzyme B (PE; Sanquin). Intracellular staining for TNF-α, IFN-γ and granzyme B was performed according to protocol B for staining of intracellular antigens for flow cytometry (eBioscience, San Diego, CA, USA) using the described buffers. For the identification of extracellular CTLA-4 expression and the expression of programmed death ligand-1 (PD-L1) in proliferating CD8⁺CD28⁻ T cells, MLR were set up as described above, but cells were not restimulated. After 7 days, cells were harvested and stained with monoclonal antibodies (mAbs) against CD3 (AmCyan), CD4 (PE), CD8 (FITC), CD28 (PerCP-Cy5·5), CTLA-4 (APC) (all BD Biosciences) and PD-L1 (PE-Cy7; eBioscience). Fluorescence minus one (FMO) controls were used to determine negative expression. Flow cytometric analysis was performed using the BD FACSCanto II flow cytometer using the BD FACSDiva software (both BD Biosciences).

Flow cytometric analysis of apoptotic cells

MLR were set up in the absence or presence of MSC (1:10; MSC/effector cells). Effector PBMC were labelled with VPD450 (BD Biosciences) and γ-irradiated, allogeneic stimulator PBMC were labelled using the PKH67 Green Fluorescent Cell Linker Kit (Sigma-Aldrich). Cells were incubated for 4 or 7 days. Apoptotic cells were identified using the annexin V PE Apoptosis Detection Kit I (BD Biosciences), according to the manufacturer's instructions, in combination with mAb labelling against CD3 (AmCyan), CD8 (APC), CD28 (PerCP-Cy5·5). Flow cytometric analysis was performed using the BD FACSCanto II flow cytometer and BD FACSDiva software (both BD Biosciences).

Statistical analysis

Statistical analyses were performed by means of paired t-tests using GraphPad Prism 5 software (GraphPad Software, San Diego, CA, USA). A P-value lower than 0·05 was considered statistically significant. Two-tailed P-values are stated.

Results

CD8⁺CD28⁻ T cells proliferate upon allostimulation and have a proinflammatory and cytotoxic phenotype

The proliferative capacity of the CD8⁺CD28⁻ T cells and their ability to express cytotoxic effector molecules was investigated in 7-day MLR by means of VPD450 dilution and flow cytometric analysis. Allogeneically stimulated CD8⁺CD28⁻ T cells proliferated as strongly as allostimulated CD8⁺CD28⁺ T cells (Fig. 1a). Both cell types expressed granzyme B, IFN-γ and TNF-α (Fig. 1b,c). Granzyme B was expressed by equal percentages of CD8⁺CD28⁻ T cells and CD8⁺CD28⁺ T cells (85 and 90%, respectively). In contrast, more CD8⁺CD28⁻ T cells than CD8⁺CD28⁺ T cells expressed the proinflammatory cytokines IFN-γ and TNF-α (83 versus 57% and 83 versus 43%, respectively). The proliferating fractions of CD8⁺CD28⁻ T cells and CD8⁺CD28⁺ T cells expressed more granzyme B and IFN-γ than the respective non-proliferating fractions; expression of granzyme B and IFN-γ in proliferating CD8⁺CD28⁻ T cells was increased by 26% (P = 0·039) and 19% (P = 0·041), respectively. Proliferating CD8⁺CD28⁺ T cells expressed 84% (P = 0·003) more granzyme B and 54% more IFN-γ (P = 0·022) than non-proliferating CD8⁺CD28⁺ T cells. TNF-α expression did not differ between the proliferating and non-proliferating fractions.

Fig. 1.

Characterization of CD8⁺CD28⁻ T cells. (a) Effector peripheral blood mononuclear cells (PBMC) were labelled with the proliferation marker violet proliferation dye 450 (VPD450) and stimulated with γ-irradiated allogeneic PBMC for 7 days. Representative examples of proliferating CD8⁺CD28⁻ T cells (black histogram) and CD8⁺CD28⁺ T cells (grey histogram) are shown. Dashed histograms depict unstimulated CD8⁺CD28⁻ T cells (black) and CD8⁺CD28⁺ T cells (grey). (b). Expression of granzyme B (black histogram), interferon (IFN)-γ (dark grey) and tumour necrosis factor (TNF)-α (light grey) by allostimulated, proliferating CD8⁺CD28⁻ T cells in 7-day mixed lymphocyte reaction (MLR) following a restimulation with phorbol myristate acetate (PMA)/ionomycin for 4 h in the presence of GolgiPlug. Expression of cytotoxic T lymphocyte antigen-4(CTLA)-4 (black histogram; c) and programmed death ligand 1 (PD-L1) (black histogram; d) by allostimulated, proliferating CD8⁺CD28⁻ T cells in 7-day MLR. Fluorescence minus one (FMO) control is depicted as dotted histogram. Representative examples are shown. Data of multiple experiments are depicted in (c). (c) Expression of granzyme B, IFN-γ, TNF-α, PD-L1 and CTLA-4 by proliferating CD8⁺CD28⁻ T cells (grey bars), proliferating CD8⁺CD28⁺ T cells (white bars), non-proliferating CD8⁺CD28⁻ T cells (chequered grey bars) and non-proliferating CD8⁺CD28⁺ T cells (chequered white bars). Unless indicated otherwise, statistically significant changes between the corresponding proliferating and non-proliferating fractions are displayed; n = 3, mean ± standard error of the mean; paired t-test; *P < 0·05.

PD-L1 expression was similar in proliferating CD8⁺CD28⁻ T cells and CD8⁺CD28⁺ T cells (47 versus 44%, respectively; Fig. 1c,e). CTLA-4 was expressed at very low levels by both cell types (Fig. 1d,e).

MSC and belatacept permit each other's immunosuppressive function

To study the combined effect of MSC and belatacept on effector cell proliferation, the appropriate concentrations and the effect of both immunosuppressive agents on each other's function had to be established. Therefore, MLR were set up in the presence of various concentrations of MSC and/or belatacept. Inhibition of proliferation was assessed by means of [³H]-thymidine incorporation. MSC and belatacept inhibited PBMC proliferation in a dose-dependent manner (Fig. 2). The two highest concentrations of belatacept and MSC tested (10 μg/ml and 1:2·5; MSC/effector cells) reduced proliferation of effector cells to 19·4% (P = 0·0002) and 7·8% (P < 0·0001), respectively. When applied in combination both immunosuppressants permitted each other's anti-proliferative function. At low concentrations the combination of MSC and belatacept had an additive suppressive effect. While belatacept (0·1 μg/ml) inhibited the proliferation of effector cells by 20·7% (P = 0·0086), MSC reduced proliferation by 38·8% (P = 0·0037). Belatacept–MSC co-treatment suppressed effector cell proliferation by an additional 15·1% compared to the inhibition achieved by MSC alone (P = 0·029).

Fig. 2.

Immunosuppressive effects of mesenchymal stem cells (MSC) and belatacept. Using peripheral blood mononuclear cells (PBMC) as effector cells, mixed lymphocyte reactions (MLR) were set up in the presence of various MSC concentrations (ratio MSC/effector cells) and belatacept concentrations. After 7 days proliferation was assessed by means of [³H]-thymidine incorporation; n = 5 (mean).

MSC reduce the percentage of proliferating, alloreactive CD8⁺CD28⁻ T cells

In its function as co-stimulation blocker, belatacept only constrains the interaction of CD28 expressing CD8⁺ T cells with APC. To examine whether MSC can control CD8⁺CD28⁻ T cells which are unaffected by belatacept treatment, the effect of MSC (1:10; MSC/effector cells) and 1 μg/ml belatacept on the proliferation of CD8⁺ T cells and their CD28⁻ subpopulation was assessed. Both agents were added alone or in combination to MLR for 7 days. Belatacept and MSC reduced the percentage proliferating CD8⁺ T cells by 13·6% (P = 0·0034) and 19·2% (P = 0·012), respectively (Fig. 3a); the combination of both treatments led to a reduction by 26·7%. At these concentrations a synergistic effect of MSC and belatacept was not observed. While belatacept reduced the proliferation of CD8⁺ T cells, it did not have an effect on the proliferation of the CD28⁻ cells within the proliferating CD8⁺ T cells (Fig. 3a,b). In contrast, MSC reduced the percentage of CD28⁻ cells within the proliferating CD8⁺ T cell population by 45·9% (P = 0·009). MSC and belatacept in combination inhibited the proliferation of CD8⁺CD28⁻ T cells by 44·9% (P = 0·036), indicating that belatacept did not impair the immunosuppressive function of MSC.

Fig. 3.

Mesenchymal stem cells (MSC) reduce the percentage of proliferating, alloreactive CD8⁺CD28⁻ T cells. Effector peripheral blood mononuclear cells (PBMC) were labelled with the proliferation marker violet proliferation dye 450 (VPD450) and stimulated with γ-irradiated allogeneic PBMC (PKH26 label) in the presence or absence of MSC (1:10; ratio MSC/effector cells) and/or 1 μg/ml belatacept. After 7 days, flow cytometric analyses were performed. (a) The percentages of proliferating CD8⁺ T cells (white bar) and the percentages of CD28⁻ cells within the proliferating CD8⁺ T cells (grey bar) are shown; n = 8, mean ± standard error of the mean (s.e.m.); paired t-test; *P < 0·05; **P < 0·01. (b) Representative examples of allostimulated CD8⁺CD28⁻ T cells in the absence and presence of belatacept and/or MSC are shown. Percentages of proliferating CD8⁺CD28⁻ T cells are stated. (c) The percentages of non-proliferating CD8⁺ T cells (white bar) and the percentage of CD28⁻ cells within the non-proliferating CD8⁺ cells (grey bar) are shown; n = 8, mean ± s.e.m.; paired t-test; *P < 0·05; **P < 0·01.

To elucidate the fate of the CD28⁻ cells, we studied the non-proliferating T cell fraction. MSC increased the percentage of CD28⁻ cells within the non-proliferating CD8⁺ T cell fraction by 58% (Fig. 3c). Further, as MSC are able to induce apoptosis, we also investigated this option by means of annexin-V staining. At days 4 and 7, the percentage of annexin V⁺CD8⁺CD28⁻ T cells was similar in MLR and MLR–MSC co-culture, indicating that MSC did not render CD8⁺CD28⁻ T cells apoptotic [day 4 (mean): 35·5 versus 32·3%; day 7: 19·9 versus 23·45%].

MSC do not affect CD28 expression of CD8⁺ T cells

The reduction of alloreactive CD8⁺CD28⁻ T cells in the proliferative fraction may not solely be attributed to the anti-proliferative effect MSC exert on these cells. Therefore, we investigated whether MSC influenced CD28 expression of CD8⁺ T cells. First, the effect of MSC on a potential gain of CD28 expression was determined. When used in MLR as single effector-cell population, proliferation of CD28⁻ T cells was limited, while allostimulated CD28⁺ T cells proliferated strongly (Fig. 4a). To provide sufficient help enabling CD28⁻ T cell proliferation, the MLR–effector population consisted of 10% sorted CD8⁺CD28⁻ T cells and 90% sorted CD4⁺ T cells. After 7 days, 48·2% of the originally CD8⁺CD28⁻ T cells had gained CD28 expression in MLR (Fig. 4b). MSC did not influence this effect on CD28 expression. In the reverse experiment to investigate whether loss of CD28 expression would be mediated by MSC, sorted CD28⁺ T cells were used as effector cells in 7-day MLR. Full CD28 expression was sustained in MLR and MSC did not affect this (Fig. 4c).

Fig. 4.

Effect of mesenchymal stem cells (MSC) on CD28 expression of CD8⁺ T cells. (a) Sorted CD28⁻ T cells and sorted CD28⁺ T cells were allostimulated with γ-irradiated, allogeneic peripheral blood mononuclear cells (PBMC) for 7 days. Proliferation of both cell populations is shown by means of violet proliferation dye 450 (VPD450) dilution. (b) Effector cell population in mixed lymphocyte reactions (MLR) consisted of CD8⁺CD28⁻ T cells (10%) and CD4⁺ T cells (90%). Effector cells were stimulated with γ-irradiated, allogeneic PBMC in the absence and presence of MSC. The percentage of CD28⁺ cells within the CD8⁺ T cell population was determined in the starting effector cell population (day 0) and in MLR (•) and MLR-MSC co-culture (1:10; ratio MSC/effector cells; ▪) after 7 days by flow cytometry. n = 6, mean ± standard error of the mean (s.e.m.). (b) Sorted CD28⁺ T cells were used as effector cells in 7-day MLR. MLR were set up in the absence (•) or presence of MSC (1:10; ratio MSC/effector cells; ▪). The percentages CD28⁺ cells within the CD8⁺ T cell population was determined by flow cytometry; n = 8, mean ± s.e.m.

Discussion

Belatacept is the first intravenous long-term immunosuppressive therapy for kidney transplantation and is believed to challenge the position of calcineurin inhibitor (CNI) tacrolimus as the most prescribed drug for the prevention of graft rejection in solid organ transplantation [20,21]. Despite their success as immunosuppressants, next to adverse side effects such as hypertension, malignancies and diabetes, CNIs have the major drawback of causing nephrotoxicity, indicating a need for alternative agents [22]. The BENEFIT (Belatacept Evaluation of Nephroprotection and Efficacy as First-line Immunosuppression) study compared the CNI cyclosporin A with belatacept in kidney transplantation [23,24]. Three-year outcomes of this study showed that patient and graft survival rates were similar for cyclosporin A and belatacept, but belatacept-treated patients had superior renal function and fewer adverse events [25]. In contrast, administration of belatacept led to higher frequencies of acute rejections. An underlying cause for these acute rejections might be CD8⁺CD28⁻ T cells that escape inhibition by belatacept. In the present study we investigated the effect of MSC on CD8⁺CD28⁻ T cells.

We identified CD8⁺CD28⁻ T cells as potentially harmful cells that express granzyme B, TNF-α and IFN-γ and are highly proliferative upon allogeneic stimulation. Expression of these cytolytic and proinflammatory molecules by CD8⁺CD28⁻ T cells has been observed by others [26–29]. However, data about the ability of CD8⁺CD28⁻ T cells to proliferate are ambiguous. While some reports confirm our finding [30,31], other research groups describe that the proliferative response of CD8⁺CD28⁻ T cells is inhibited [32,33]. Critical for CD8⁺CD28⁻ T cell proliferation are the stimulation conditions. Plunkett et al. describe that anti-CD3 stimulation leads only to mild proliferation, while in the presence of irradiated PBMC CD8⁺CD28⁻ T cells proliferate strongly [34]. Contrary to these results, we found that CD8⁺CD28⁻ T cells stimulated with allogeneic PBMC had restrained proliferative abilities. CD8⁺CD28⁻ T cells proliferated as strongly as their counterparts in total PBMC only when CD4⁺ T cell help was provided. This indicates that certain cytokines or co-stimulatory signals other than CD28 ligands are required for the activation and proliferation of CD8⁺CD28⁻ T cells. We determined that proliferating CD8⁺CD28⁻ T cells expressed PD-L1 but lacked CTLA-4. Upon binding to the CD80/86 complex, both molecules transmit inhibitory signals [2,35–37]. Control of cell proliferation through these inhibiting pathways can therefore be jeopardized by belatacept. However, next to its inhibitory function, PD-L1 has also been described to enhance T cell activation and thereby might contribute to the proliferative capacities of CD8⁺CD28⁻ T cells [38,39].

CD8⁺CD28⁻ T cells are found predominantly within the (terminally differentiated) effector memory CD8⁺ T cell subset [40] and they can have cytotoxic [29,41–43] or immunosuppressive functions [10,44–47]. Thus, inhibition of CD8⁺CD28⁻ T cells by MSC could not only involve suppression of the cytotoxic subset, but also affect the regulatory subset. Our study shows, however, that MSC inhibited CD8⁺CD28⁻ T cells that express the cytotoxic molecules granzyme B, TNF-α and IFN-γ. In contrast, CTLA-4, which is associated with a regulatory function, was hardly detectable on the CD8⁺CD28⁻ T cells. Earlier studies by our group demonstrated that terminally differentiated CD8⁺ T cells contain a large proportion of CD28⁻ cells, and these cells showed no immunosuppressive capacity in vitro [48]. This suggests that under the inflammatory conditions as set up in the present experiments, MSC target only the effector CD8⁺CD28⁻ T cell subset. It is possible that under different conditions CD8⁺CD28⁻ T cells with regulatory properties are more prominent, and under these circumstances the use of MSC should be reconsidered.

IL-15 is a cytokine that promotes CD8⁺CD28⁻ T cell proliferation [30]. Interestingly, IL-15, next to IL-7, is crucial for the homeostatic maintenance of T cells in the absence of antigenic stimuli and expedites the loss of CD28 expression [49]. During normal exposure to antigen CD28 expression is transiently reduced but returns quickly to basal expression levels. Repeated antigen exposure due to the natural ageing process, viral infections or viral reactivation in immunocompromised patients causes a decline in CD28 expression, leading eventually to total loss of CD28. Surprisingly, we found that in our setting CD28⁺ T cells did not lose CD28 during allogeneic stimulation with PBMC, confirming that extended rounds of antigen exposure are required to initiate reduction of CD28. Permanent decline of CD28 expression entails telomere shortening and reduction of telomerase activity and is attributed to a defect in the CD28 promotor leading to transcriptional inactivation [50–54]. We, however, found that CD8⁺ T cells that were initially CD28⁻ gained CD28 expression during allogeneic stimulation with PBMCs. Reinduction of CD28 expression in CD4⁺CD28⁻ T cells is a known phenomenon and only possible until CD28⁻ T cells have reached terminal differentiation. Warrington et al. described that combined stimulation of T cell receptor (TCR) and IL-12 receptor restored CD28 transcription and protein expression, while single stimulation of either the TCR or the IL-12 receptor was not sufficient [55]. IL-12 is produced by phagocytic cells, B cells and other antigen-presenting cells [56] and therefore potentially contributes to the CD28 re-expression in originally CD8⁺CD28⁻ T cells in MLR. Although CD28 expression can be influenced up to a certain stage during T cell differentiation, MSC did not affect the immunophenotypical changes of CD8⁺CD28⁻ T cells, nor did they cause loss of CD28 expression in CD8⁺CD28⁺ T cells. Further, we found that MSC did not induce apoptosis in CD8⁺CD28⁻ T cells, despite their ability to express Fas ligand (FasL) or to initiate the programmed death (PD)-1/PD-ligand 1 (PD-L1) pathway [57,58]. These observations indicate that MSC solely have an anti-proliferative effect on CD8⁺CD28⁻ T cells.

Co-administration of MSC with other immunosuppressive drugs is not always encouraged; agents such as tacrolimus, mammalian target of rapamycin (mTor) inhibitor rapamycin and rabbit anti-thymocyte globulin (rATG) negatively affect the suppressive capacity of MSC in vitro [59–61]. At same time, MSC are able to reduce the efficacy of tacrolimus and rapamycin [59,60]. As MSC lack expression of the CTLA-4 ligands CD80 and CD86, it was not surprising that belatacept did not diminish MSC function [62]. Conversely, MSC did not affect the immunosuppressive capability of belatacept. In the presence of belatacept and lower MSC/effector cell ratios we even observed an additive suppressive effect.

MSC exert their immunomodulatory function not only by suppressing the proliferation of various immune cells; in a previous study we have shown that MSC also induce functional de-novo regulatory T cells (T_reg) [63]. CD28/B7 co-stimulation in T_reg is required for their differentiation [64]. T_reg-specific deficiency of CD28 and CTLA-4 leads to an impaired immunosuppression by T_reg and the development of autoimmunity and rejection in transplant models [65,66]. The effect of CTLA-4-Ig therapy on T_reg is controversial. Administration of CTLA-4-Ig to a skin transplant mouse model abolished T_reg-dependent graft acceptance and expansion of T_reg [67]. In contrast, CTLA-4-Ig therapy in rheumatoid arthritis patients reduced the frequency of peripheral T_reg but enhanced their function [68]. Therefore, alongside the alloreactive CD8⁺CD28⁻ T cells that escape belatacept therapy, the possible diminution of T_reg in patients receiving belatacept might contribute to the increased frequency of acute rejections reported for belatacept-treated kidney graft recipients [25].

In conclusion, CD8⁺CD28⁻ T cells sustain their proliferative capacity in the presence of belatacept, and secrete cytolytic and cytotoxic effector molecules. As MSC are able to control these CD8⁺CD28⁻ T cells by inhibiting their proliferation, our study suggests a potential for MSC–belatacept combination therapy to prevent alloreactivity after solid organ transplantation.

Author contributions

A. U. E. performed the experiments and participated in the writing of the manuscript. M. G. H. B. participated in the writing of the manuscript. C. C. B, N. H. R. L., M. F., W. W. and M. J. H. participated in the study design and the writing of the manuscript.

Disclosure

The authors of this manuscript have no financial or commercial conflicts of interest to disclose.