The mechanisms underlying the immunomodulatory effects of mesenchymal stem cells (MSCs) have been investigated under extreme conditions of strong T cell activation, which induces the rapid death of activated lymphocytes. The objective of this study was to investigate these mechanisms in the absence of additional polyclonal activation. In co-cultures of peripheral mononuclear blood cells with human MSCs (hereafter referred to as hMSCs), we observed a striking decrease in the level of CD8 expression on CD8+ cells, together with decreased expression of CD28 and CD44, and impaired production of IFN-gamma and Granzyme B. This effect was specific to hMSCs, because it was not observed with several other cell lines. Downregulation of CD8 expression required CD14+ monocytes to be in direct contact with the CD8+ cells, whereas the effects of hMSCs on the CD14+ cells were essentially mediated by soluble factors. The CD14+ monocytes exhibited a tolerogenic pattern when co-cultured with hMSCs, with a clear decrease in CD80 and CD86 co-stimulatory molecules, and an increase in the inhibitory receptors ILT-3 and ILT-4. CD8+ cells that were preconditioned by MSCs had similar effects on monocytes and were able to inhibit lymphocyte proliferation. Injection of hMSCs in humanized NSG mice showed similar trends, in particular decreased levels of CD44 and CD28 in human immune cells. Our study demonstrates a new immunomodulation mechanism of action of hMSCs through the modulation of CD8+ cells towards a non-cytotoxic and/or suppressive phenotype. This mechanism of action has to be taken into account in clinical trials, where it should be beneficial in grafts and autoimmune diseases, but potentially detrimental in malignant diseases.
Mesenchymal stem cells (MSCs) are non-hematopoietic stem cells, which were first identified in the bone marrow cavity by Fridenshtein (Fridenshteı�n, 1982); their ability to differentiate into various types of mesoderm tissues was then demonstrated (Pittenger et al., 1999; Wakitani et al., 1995). MSCs can be isolated from bone marrow, skeletal muscle, adipose tissue and synovial membrane, as well as cord blood (da Silva Meirelles et al., 2006) and are defined by using a combination of phenotypic markers.
It is well established that human MSCs (hereafter referred to as hMSCs) possess suppressive capabilities on several subsets of immune cells, namely T and B cells, dendritic cells (DCs) and natural killer (NK) cells, in response to polyclonal stimulation (Uccelli et al., 2007). MSCs have a unique immunophenotype, with a low expression of major histocompatibility complex (MHC) class I and an absence of co-stimulatory molecules. Although several groups have demonstrated that hMSCs have the capacity to inhibit T cell proliferation (Benvenuto et al., 2007; Glennie et al., 2005), controversy remains regarding the effect of hMSCs on T cell activation. Le Blanc and colleagues (Le Blanc et al., 2004) and Groh and colleagues (Groh et al., 2005) who conducted their research on cultures of phytohaemagglutinin (PHA)-stimulated lymphocytes and alloreactive T cells, reported that hMSCs prevent the expression of CD25, CD38 and CD69 activation markers on T lymphocytes. However, in a similar model, Aggarwal and colleagues (Aggarwal and Pittenger, 2005) argued that hMSCs actually induce a slight increase in the proportion of CD25+ T cells. In addition, some studies have claimed that the inhibitory effect of MSCs on T cells is confined to cellular proliferation rather than to the effector function of T cells (Ramasamy et al., 2008).
CD8 is commonly used as a cytotoxic T cell marker, whereas its downregulation has been suggested as one of the mechanisms for peripheral tolerance (Rocha and von Boehmer, 1991; Schönrich et al., 1991; Zhang et al., 1995). Xiao and colleagues (Xiao et al., 2007) have argued that the downregulation of CD8 expression and the loss of specific peptide–MHC binding during the immune response following bacterial infection are subjected to detuning during normal immune responses. Different CD8+ T cell subsets have been identified based on the expression of cell surface markers, such as the CD28 co-stimulatory molecule, which is necessary for the initiation of most T cell responses. CD8+CD28− cells are defined as suppressor T cells that have been shown to down-modulate the antigen-presenting cell (APC) function by inducing immunoglobulin-like transcript 3 and 4 (ILT-3 and ILT-4) inhibitory receptors, leading to the inhibition of CD4+ T-cell proliferation by antigen presenting cells (APCs) (Chang et al., 2002).
The difficulty in achieving long-term allograft survival has been attributed to the resistance of effector CD8+ cells (Trambley et al., 1999) and/or memory T cells (Lakkis and Sayegh, 2003; Valujskikh et al., 2002) to co-stimulatory blockade. Preclinical studies reveal that hMSCs are capable of preventing graft rejection in a rat model for cardiac allograft (Zhou et al., 2006) and in rat kidney transplantation, as well as reducing the number of CD8+ cells in the infiltrates (De Martino et al., 2010). Injection with hMSCs prolonged the survival of skin transplant in a baboon model (Bartholomew et al., 2002) and extended heart allograft survival when administered in a mouse model, with 33% of recipients showing long-term tolerance (Casiraghi et al., 2008).
The mechanisms underlying the immunomodulatory effects of MSCs are still poorly understood, and most studies have used strong T cell activators that represent extreme situations and induce rapid death of activated lymphocytes. The objective of this study was to investigate the effect of hMSCs on peripheral mononuclear blood cells without additional exogenous stimulation. Our findings indicate that hMSCs shift the CD8+ cytotoxic cells towards a suppressive phenotype, an effect that depends on CD14+ monocytes, whose phenotype is strikingly regulated by hMSC.
hMSCs decrease expression of CD8 on CD8+ T cells
We evaluated whether hMSC lines that originate from adipose tissue had the potency to modify the balance between the main T cell subsets in direct co-cultures without additional stimulating molecules. Kinetics analysis in living cells showed that CD8, but not CD4 expression, was significantly decreased on day 6 of incubation (P<0.001) and was prolonged over time, with >70% reduction in the level of CD8 expression on day 11 (P<0.001) (Fig. 1A). The viability of CD4+ and CD8+ T cells at the 7th day of co-culture was similar (∼80%). The absolute number of cells was also similar in the absence or presence of hMSC. However, hMSCs slightly decreased the viability of CD8+ T cells (from 81% to 75%) and had no effect on CD4+ cells (78.9% versus 78.7%) (supplementary material Fig. S1A). Unlike the dramatic decrease in CD8 expression, the relative percentage of CD8+ cells was only slightly decreased (Fig. 1B), given that the major effect occurred within the positive CD8+ population that shifts towards the negative peak (Fig. 1E). The percentage of CD8 positive cells was evaluated by setting the cut-off at the end of the negative peak (Fig. 1E). To determine whether the decrease in CD8 intensity was dependent on the origin of the hMSCs cell lines, we investigated three adipose hMSC cell lines, AD1, AD4, and AD5, derived from the abdomen, breast and thigh, respectively. All hMSC derivatives dramatically downregulated CD8 expression, with the AD5 cell line demonstrating a >70% reduction in geometric mean (GM) of CD8 fluorescence (Fig. 1C,E). This effect was specific to CD8+ T cells, because no such effect was observed in the CD4+ cell subset (Fig. 1D). The downregulation of CD8 fluorescence intensity in hMSCs was observed with several anti-CD8 antibodies generated from different clones, demonstrating that this effect was not epitope dependent (data not shown). The same results were obtained for the CD8β chain as well (supplementary material Fig. S1B). In order to determine whether downregulation of CD8 was specific to hMSCs, peripheral blood mononuclear cells (PBMCs) were co-cultured with other cell lines, such as human neonatal fibroblasts (Fig. 1F), the HaCat cell line (Fig. 1G), or mouse embryonic fibroblasts (MEF) (data not shown). None of these cell lines, either of juvenile or adult origin, were able to downregulate CD8 fluorescence intensity after 7 days of direct co-culture. Altogether, these results demonstrated a striking effect of hMSCs on the level of CD8 expression in a culture free of exogenous polyclonal stimulation.
Taking into account that hMSC-mediated immune suppression is considered to be the result of both cell-cell contact and soluble factors (Liu et al., 2012; Uccelli et al., 2008), we investigated whether the down-modulation of CD8 surface expression required direct contact or was mediated by soluble factors. In the absence of direct interactions (transwell supports), CD8 fluorescence intensity was downregulated by 50% (P<0.001), suggesting the involvement of soluble factors (Fig. 1H). Using hMSC-conditioned medium collected after 7 days of culture, only a 30% reduction in CD8 expression was observed (Fig. 1H). The difference between the transwell and conditioned medium co-culture could be explained by the limited pool of soluble factors in the conditioned medium, whereas the hMSCs continue to grow and produce soluble factors during the entire incubation time when seeded in transwell cultures. These results demonstrate that the capacity of hMSCs to downregulate CD8 is essentially mediated by soluble factors. Previous studies have suggested that TGF-β1 and PGE2 contribute to the immunoregulatory properties of hMSCs (Aggarwal and Pittenger, 2005; Groh et al., 2005). We examined the potential role of TGF-β1 in our cell culture conditions. Neutralization of TGF-β1 did not reverse the effect of hMSCs on CD8 and CD44 expression (supplementary material Fig. S2A, S2B). Considering that PGE2 has been previously shown to mediate the immune function of hMSCs mediated by monocytes (Chen et al., 2010; Cutler et al., 2010), and because PGE2 is considered responsible for the inhibitory effect on DC differentiation and function (Spaggiari et al., 2009), we evaluated the role of PGE2 in our experimental system. Production levels of PGE2 in the supernatants of hMSCs or PBMCs that were cultured alone were very low, and a significant increase (P<0.001) was detected in co-cultures of hMSCs with PBMCs (supplementary material Fig. S2F). However, inhibition with indomethacin did not restore the fluorescence intensity of CD8 reduced by hMSCs (supplementary material Fig. S2E). These data suggest that in our experimental model, PGE2 production is not involved in the downregulation pathway of CD8 expression.
hMSC-mediated inhibition of CD8 expression occurs at the post-transcription level
In order to address whether the downregulation of CD8 expression results from reduced CD8 transcription, the CD8 mRNA expression in PBMCs cultured either alone or with hMSCs in transwell conditions was quantified using real-time RT-PCR. There was no change in the mRNA levels for CD8 (Fig. 2A), suggesting that the reduction in CD8 expression in the CD8+ cells when co-cultured with hMSCs was not due to reduced CD8 transcription. Consequently, it is most likely that the mechanisms that regulate CD8 expression levels in hMSCs are post-transcriptional. Both the surface and cytoplasmic fluorescence intensity values of CD8 were reduced by ∼60% when co-cultured with hMSCs (Fig. 2B), whereas the relative percentage of CD8+ cells, both surface and cytoplasmic, was only reduced by ∼15% (Fig. 2C). The decrease in expression of CD8, without changes in the mRNA levels of CD8, suggests that hMSCs regulate CD8 by reducing its protein expression, possibly through proteolytic cleavage (Fujimoto et al., 1984) or ubiquitylation and degradation mechanisms (D'Agostino et al., 2011).
hMSCs downregulate the expression of activation markers on CD8+ T cells
T cell antigen receptor (TCR)-CD3-dependent responses are regulated by the constitutive or inducible expression of co-stimulatory receptors, such as CD28, and the lack of CD28 expression is generally associated with regulatory function in CD8+ T lymphocytes (Chang et al., 2002; Cortesini et al., 2001). We tested whether hMSCs modulate the expression of CD28 molecules on CD8+ cells. As shown in Fig. 3A, 3B, a substantial decrease in CD28 fluorescence intensity was achieved in the presence of hMSCs. We also analyzed the expression of several markers involved in cell adhesion and activation. The cell adhesion molecule CD31 plays a key role in leukocyte trafficking across the endothelium (O'Brien et al., 2003), whereas CD44 is a marker associated with cell adhesion and migration of lymphocytes (Johnson and Ruffell, 2009). In addition, the Fas (CD95) marker is increased during activation-induced cell death (Cohen et al., 1992; Schwartz and Osborne, 1993). As shown in Fig. 3C, a significant decrease in CD44 and CD95 fluorescence intensity (P<0.005 and P<0.02, respectively) was observed in the presence of hMSCs, whereas CD31 expression did not change. To investigate whether these phenotypic changes were concomitant with functional alterations, we analyzed the production level of Gzm B and IFN-γ. hMSCs induced a significant reduction (P<0.005) in the production of Gzm B (Fig. 3D). A significant decrease in the level of IFN-γ production was observed in CD8+ cells when co-cultured with hMSCs, without external activation (P<0.01) (Fig. 3E). Because the percentage of IFN-γ-positive cells was quite low without activation (<3%), a positive control of these experiments was performed using external polyclonal stimulation. In this case, as expected, the percentage of IFN-γ positive cells was high (24.9±2%) and was greatly reduced in the presence of MSC (10.6±1.1) (supplementary material Fig. S3). IFN-γ production levels were also evaluated in tissue culture supernatants (Fig. 3F). Interestingly, even without polyclonal stimulation, substantially high concentrations of IFN-γ (260±164 pg/ml) were produced, and were reduced in the presence of hMSC (71±23 pg/ml).
In order to address CD8+ T cell functionality further on, the CD8+ T cell population was purified using the FACSAriaTM after 7 days of culture of PBMCs, either in the presence (CD8MSC) or absence of hMSCs (CD8CONT). Because CD8 suppressor cells are known to influence the expression of co-stimulatory and inhibitory molecules on APCs (Suciu-Foca et al., 2005), we co-cultured CD8MSC or CD8CONT cells with isolated CD14+ monocytes. CD8MSC but not CD8CONT cells were able to downregulate the CD80 co-stimulatory molecule (Fig. 3G), as well as to upregulate expression of the ILT3 inhibition receptor on monocytes (Fig. 3H), whereas CD86 was slightly decreased and ILT4 was unchanged (data not shown). The effect of these purified cell subsets on lymphocyte proliferation was also tested; CD8MSC or CD8CONT cells were added to PBMCs stained with carboxyfluorescein succinimidyl ester (CFSE). CD8MSC cells significantly decreased the percentage of proliferating cells compared with CD8CONT cells (Fig. 3I). Generally, such a suppression of proliferation is attributed to CD4+ FoxP3+ Treg cells, which were not induced by hMSCs under our experimental conditions (data not shown).
The inhibitory effect of hMSCs on CD8+ is mediated by CD14+ monocytes
In order to investigate whether hMSCs exert their downregulatory effect on CD8+ T cells in a direct manner, isolated CD8+ T cells were cultured in the presence of hMSCs. The viability of isolated CD8+ T cells was similar, whether cultured alone or in the presence of hMSCs (77% versus 78%, respectively). Compared with isolated CD8+ cells that were cultured alone, not only was the level of CD8 intensity not downregulated by hMSCs, but rather the CD8 receptor intensity was upregulated by ∼30% (Fig. 4A). These results indicate that another cell population is mediating the decrease in CD8 expression. Because CD14+ monocytes play a major role in modulating immune responses, we investigated whether the down-modulation of CD8 expression was attributed to their presence. A strong downregulation in CD8 intensity levels was observed in the presence of monocytes, with a 70% decrease in expression (Fig. 4A), which was similar to the effect of the whole PBMCs. Furthermore, when monocytes were depleted from PBMCs, the capacity of hMSCs to suppress CD8 expression was completely abolished (data not shown).
We then explored whether the contact is crucial between these three key players, CD8+, CD14+ and hMSCs. When the isolated CD14+ monocytes were cultivated separately from CD8+ T cells, the hMSCs did not exert their inhibitory effect on CD8+ cells (Fig. 4B, third column), whereas direct contact between CD8+ and CD14+ subsets led to a 70% reduction in CD8 expression (Fig. 4B, fourth column). The results shown in Fig. 4B demonstrate that CD14+ monocytes require direct contact with CD8+ T cells, rather than with hMSCs, for the modulation of CD8 expression.
hMSCs sensitize CD14+ monocytes, which in turn mediate an inhibitory effect of hMSCs on CD8+ cells
We then investigated whether hMSCs are able to sensitize CD14+ cells (Fig. 5A). To this end, we divided the experiment into two steps. The first step was co-culture of isolated CD14+ with hMSCs, in an attempt to produce sensitized CD14+ cells (sCD14+), whereas the second step was co-culture between the ‘sCD14+’ cells and isolated CD8+ cells.
When co-cultured with sCD14+ monocytes (Fig. 5A, II), the levels of CD8 expression were inhibited by 20%, compared with non-sensitized monocytes (Fig. 5A, I) (P<0.005). This effect was further enhanced when CD14+ monocytes were cultured with hMSCs continuously, where a 40% inhibition in CD8 intensity was observed compared with the non-sensitized monocytes (P<0.001) (Fig. 5A, III). These results show that CD14+ monocytes that have been pre-sensitized by hMSCs can down-modulate CD8 expression, suggesting that sCD14+ monocytes acquire a suppressive phenotype in the presence of hMSCs. Moreover, we tested some representative markers that were significantly altered (P<0.005 for CD44 and P<0.02 for CD95) on CD8+ cells by hMSCs (i.e. CD44 and CD28) (Fig. 3C). The same inhibitory pattern was observed for both activation markers (Fig. 5B). These results further support the suppressor phenotype of sCD14+ cells.
Phenotypic characterization of the CD14+ monocytes is presented in Fig. 5C. Both the CD80 and CD86 co-stimulatory molecules were significantly downregulated, with a >60% reduction in fluorescence intensity. The HLA-ABC and HLA-DR histocompatibility molecules, as well as CD16 expression associated with a proinflammatory status of monocytes (Ziegler-Heitbrock, 2007), were also subjected to downregulation by hMSCs (data not shown). Both the ILT3 and ILT4 inhibitory receptors were upregulated after incubation with hMSCs. Neutralization of TGF-β1 did not reverse the effect of hMSCs on ILT3 and CD80 expression in monocytes (supplementary material Fig. S2C,D). Along with the functional sensitization of CD14+ cells, the hMSCs also displayed a direct effect on the size of CD14+ monocytes compared with control cultures through kinetics (supplementary material Fig. S4A,B). The downregulation of co-stimulatory and histocompatibility molecules by hMSCs together with the upregulation of the inhibitory ILTs receptors support the fact that hMSCs modulate CD14+ monocytes towards a suppressor phenotype.
hMSCs exert similar inhibitory effect in vivo in humanized NSG mice
In order to address whether hMSCs could present similar features in vivo, we used immunodeficient NSG mice humanized with CD34+ hematopoietic stem cells according to protocols described previously (Marodon et al., 2009). Most mice had a clear reconstitution of the human immune system in the periphery (peripheral blood, spleen) (supplementary material Fig. S5). The level of humanization was evaluated by analyzing the CD45+ human cells in the peripheral blood. We then selected the mice with the best reconstitution and injected them with hMSCs intraperitoneally. After 7 days, the mice were killed, and expression of CD8, CD28 and CD44 was analyzed within the human CD45+ (huCD45+) cells in the spleen and bone marrow from both experimental groups. Analysis of the bone marrow showed that expression of human CD8 (Fig. 6A), CD44 (Fig. 6B) and CD28 (Fig. 6C) was downregulated in the human CD45+ gate in the NSG mouse injected with hMSCs, compared with the control NSG mice. Although the fluorescence intensity of CD8 decreased, the main change observed in the NSG mice was the decreased percentage of CD8+ cells. This difference with our in vitro data could be owing to the nature of the models. The immune system in NSG mice obtained by reconstitution with neonatal CD34+ cells is not completely mature (Watanabe et al., 2009), whereas the cells in vitro originated from adult blood. Furthermore, our in vitro analysis showed that APCs were essential for the effects on CD8 expression, whereas in the NSG mice, the involvement of the APCs could not be evaluated. Another possibility is that the decrease in CD8 expression preceded the drop in CD8 cell number, and in the mouse experiments, the first step was very fast. This hypothesis fits with the in vitro experiments (Fig. 1B), which show that the percentage of CD8+ cells decreased more strikingly between 6 and 12 days, whereas the fluorescence intensity starts to decrease from day 3. Similar results were observed in the spleen (data not shown). In addition, we investigated the effect of hMSCs on the mobilization of human cells in the blood of NSG mice, under non-activated or LPS-activated conditions, by calculating the blood:BM ratio of human CD45+ cells. We observed a significant reduction in the mobilization in NSG mice treated with MSC (Fig. 6D) under both conditions, suggesting that MSCs alter the migratory properties of human lymphoid cells in the NSG mice.
The aim of this study was to investigate the immunoregulatory mechanisms of hMSCs in a steady-state situation. The main results of this study are as follows: (1) hMSCs exert a striking downregulatory effect on expression of the CD8 receptor in CD8+ cells, which is associated with features of reduced cytotoxicity (i.e. a reduction in the production of IFN-γ and Gzm B), and an increased regulatory phenotype (CD28lo, CD44lo and CD95lo); (2) CD8+ cells that are preconditioned by MSCs decrease the expression of costimulatory molecules on monocytes and inhibit lymphocyte proliferation; (3) downregulation of CD8 is essentially mediated by soluble factors and occurs at the post-transcriptional level; (4) the immunomodulatory effect of hMSCs on CD8+ T cells requires the presence of CD14+ monocytes, which acquire a suppressive phenotype; and (5) hMSCs are found to present similar features in a humanized mouse model.
hMSCs down-modulate CD8 expression
In this study, we have shown for the first time that hMSCs exert a strong inhibitory effect on expression of the CD8 receptor (∼60% decrease), whereas the relative percentage of CD8+ cells was only slightly reduced (∼15% decrease). This inhibitory effect was shown to be specific for hMSCs, because other cell lines that were tested did not show the same features. Although hMSCs have substantial effects on CD4+ cells, such as a decrease in several activation markers (data not shown), we decided to focus on CD8+ cells because of the striking effect of hMSCs on CD8 detuning, which has not been previously reported. Although most studies addressing the immunoregulatory properties of hMSCs have used different means of external activation, we have used none. Once T cells are activated through their T cell receptor, they display changes in the expression of surface markers and in the balance between proliferation and cell death. By avoiding external activation, we assume that the functional activity of hMSCs on immune cells mimics the steady state condition. This difference could explain why this down-modulation of CD8 has never been reported before.
Reduction in CD8 expression is associated with impaired cell interaction, leading to loss of cytotoxity (Xiao et al., 2007). Therefore, we suggest that the inhibitory effect exerted by hMSCs when co-cultured with unstimulated PBMCs shifts the cytotoxic CD8+ T cells towards a less cytotoxic and more suppressive phenotype. This hypothesis is supported by our data showing that purified CD8+ cells from co-culture with hMSCs were able to reduce the expression of CD80 and to increase expression of the ILT3 receptor on monocytes. Interestingly, these cells were also able to substantially reduce the percentage of proliferating lymphocytes. In addition, because the absence of CD28 is one of the CD8+ T-cell hallmarks of a suppressor state (Cortesini et al., 2001), we analyzed CD28 expression on CD8+ cells and showed its decreased expression when co-cultured with hMSCs.
A substantial decrease in expression of CD44 was also noted in the presence of hMSCs. The cell adhesion molecule CD44 is responsible for mediating adhesion to the extracellular matrix glycosaminoglycan and hyaluronan (Johnson and Ruffell, 2009). Considering the role of CD44 in the recruitment of leukocytes in pathological situations, its decreased expression on CD8+ cells might lead to inhibition of CD8+ cell migration to inflammatory sites. Our in vivo data in NSG mice showing inhibition of cell migration after treatment with hMSCs are compatible with this hypothesis, and confirm recently published data (Chiesa et al., 2011). Given the correlation of reduced CD44 expression with reduced cytotoxic activity in CD8+ CTL clones (Rodrigues et al., 1992), our results strengthen the idea that hMSCs shift CD8+ cells towards a non-cytotoxic and/or suppressive phenotype. This assumption is confirmed further by the decreased secretion levels of Gzm B and IFN-γ in CD8+ cells when co-cultured with hMSCs.
CD14+ monocytes play a central role in the effect of hMSCs on CD8+ T cells
Zhang and colleagues (Zhang et al., 2004) showed that dendritic cells were the primary target of the immunosuppressive activity of MSCs, which affected all major stages of the DC life cycle. Here, we demonstrate that CD14+ monocytes, which were sensitized by hMSCs were crucial in the inhibitory effect that hMSCs exerted on CD8+ T cells.
DCs play a central role in the initiation and regulation of immune response (Steinman et al., 2003), and their ability to either initiate an immune reaction or induce tolerance relies on the transition of co-stimulatory molecules CD80 and CD86, as well as the upregulation of MHC class II on the cell surface (Mellman and Steinman, 2001). Immature or regulatory DCs, which are deficient in co-stimulatory molecules, can induce T cell anergy, generate regulatory T cells and promote alloantigen-specific tolerance (Smits et al., 2005). Considering that hMSCs downregulate the expression of MHC antigen and CD16 (data not shown), as well as both co-stimulatory molecules CD80 and CD86 in CD14+ monocytes, we suggest that hMSCs shift the APC population towards a suppressive phenotype. High expression of surface inhibitory molecules, such as ILT-3 and ILT-4, is considered to be a biomarker for tolerogenic APCs (Chang et al., 2002), and these molecules were also increased in the monocytes that were co-cultured with hMSCs.
To date, several groups have reported that the inhibitory effect of hMSCs is mediated by soluble factors (Le Blanc et al., 2004; Rasmusson et al., 2003), whereas others claim that cell-to-cell contact is necessary (Krampera et al., 2003). Here, we report that the down-modulation of CD8 surface expression by hMSCs is the result of soluble factors that are secreted by hMSCs, which target the monocytes. However, the CD8 cells require direct contact with the CD14+ monocytes, probably through the MHC-TCR and interactions with co-stimulatory molecules. The nature of the soluble factors involved in the down-modulation of CD8 is still unclear. Neutralization of TGF-β1 and PGE2 soluble factors did not reverse the effect of hMSCs on CD8 expression, although a substantial increase of PGE2 production levels was noted under the co-culture conditions. These data highlight that the mechanisms of action of MSCs are diverse, and that downregulation of CD8 is not dependent on PGE2, whereas inhibition of lymphocyte proliferation appeared to be, at least in part, dependent upon PGE2 (Spaggiari et al., 2009).
The mechanisms of action of hMSC are complex, and the sequence of events is not easy to define. However, based on our experiments, we can propose the following scenario (illustrated in supplementary material Fig. S6). First, a soluble factor produced by hMSCs modulates the functional properties of the CD14+ monocyte towards a suppressor phenotype (downregulating MHC antigens and co-stimulatory molecules and increasing ILT3 and ILT4 inhibitory molecules); second, upon the interaction with MSC-conditioned monocytes (and in the absence of hMSCs), CD8+ cells have a decreased expression of CD8 and CD28, and shift towards a suppressor CD28neg-low phenotype; third, the CD8lo cells (purified by FACS) were able to reduce the proliferation of fresh lymphocytes as well as to reduce co-stimulatory molecules on monocytes, probably maintaining their regulatory phenotype. Although we cannot exclude a direct effect of conditioned monocytes on the proliferation of lymphocytes, our data clearly describe one potential sequential mechanism. The mandatory presence of CD14+ cells might explain why the hMSCs inhibitory effect is more pronounced at high lymphocyte concentrations. Indeed, under these conditions, a higher number of CD14+ cells could be sensitized and exert their effects on the CD8+ cells.
Moreover, hMSCs were found to induce memory over CD14+ cells, as pre-sensitized monocytes succeeded to reduce surface expression of CD8, even in the absence of hMSCs. Considering that hMSCs are capable of inducing pre-sensitized CD14+ cells, we suggest that the ability of hMSCs to prolong allograft survival is the consequence of this tolerogenic-inducing effect. Indeed, clinical studies have revealed that a single intravenous administration of MSC might be sufficient for achieving long-term tolerance. One possible explanation is that hMSCs sensitize monocytes and dendritic cells towards a suppressor phenotype. The duration of this effect needs to be investigated further.
Physiological relevance of the inhibitory effect of hMSCs on CD8+ T cells
In vivo, hMSCs have been shown to be capable of preventing graft rejection in several transplantation models (Bartholomew et al., 2002; Casiraghi et al., 2008). The ability of hMSCs to protect allografts from rejection could be mediated by targeting the effector function of alloreactive T cells. Because the CD8 receptor plays a crucial role during activation of CD8 T cells (Couedel et al., 1999; Gao and Jakobsen, 2000; Holler and Kranz, 2003), and as the CD8+CD28− cells could be protective for the graft (Coley et al., 2009; Colovai et al., 2003; Sindhi et al., 2005), we propose that the beneficial effects of hMSCs observed in the therapeutic context of transplantation occur through the inhibition of CD8 expression and costimulatory molecules. Conversely, this mechanism could explain why MSCs could promote cancer growth (Muehlberg et al., 2009; Prantl et al., 2010). The functional consequence of CD8 inhibition by hMSCs is associated with substantially reduced levels of IFN-γ production, as it has been reported that IFN-γ cytokine secretion enhances the rejection of skin grafts (Mattarollo et al., 2010). Our data obtained from the humanized mouse model, support that this mechanism of action does occur also in vivo. A recent paper by Perico and colleagues (Perico et al., 2011) strengthens our results, because these authors report that infusion of autologous hMSCs in recipients of kidney from related living donors induces a profound reduction of CD8+ cell activity in the transplanted patients.
The appearance of regulatory CD8+ CD28− T suppressor cells is associated with a reduced need for maintenance of immunosuppression in pediatric liver-intestine transplant recipients (Sindhi et al., 2005), as well as in adult-to-adult living donor liver transplantation (Lin et al., 2009). Studies in heart transplant patients indicate that CD8+ CD28− T suppressor cells upregulate the inhibitory receptors ILT-3 and ILT-4 on APCs (Chang et al., 2002). Therefore, it is possible that, in our system, there is a cross-talk between the CD8+ cells and the monocytes. Not only do the monocytes influence the CD8+ cell phenotype, but the newly generated CD8+CD28lo cells might also increase the immunosuppressive properties of the monocytes, as assessed by the upregulation of ILT-3 and ILT-4 inhibitory molecules. Our experiments using purified CD8+ cells after co-culture with hMSCs confirm this hypothesis, because ILT3 was significantly increased in monocytes co-cultured in presence of CD8MSC.
In conclusion, our study demonstrates a new immunomodulation mechanism of action of hMSCs through the modulation of CD8+ cells towards a non-cytotoxic and/or suppressive phenotype. This mechanism of action must be taken into account in clinical trials, where it should be beneficial in grafts and autoimmune diseases, but potentially detrimental in malignant diseases.
Materials and Methods
Generation of human adipose-derived MSCs
The hMSCs were isolated from healthy adult females undergoing a routine lipoaspiration procedure. Informed consent was obtained in accordance with the Declaration of Helsinki. Adipose tissue cells were washed with phosphate-buffered saline (PBS) (Gibco, USA), digested by using 0.1% collagenase I (Worthington, USA) for 1 hour at 37°C, and then centrifuged for 10 minutes at 2000 rpm. The cell pellet was then re-suspended and passed through a 100 µm filter for removal of debris. The resultant cellular fraction was plated in hMSC complete medium: Dulbecco's modified Eagle's medium: Ham's F12 expansion media (DMEM/F12) (Beit Haemek, Israel), containing 10% fetal calf serum (FCS) (Hyclone, USA), 5 ng/ml basic fibroblast growth factor (bFGF) (Biological Industries, Israel), 2 mM L-glutamine (Gibco, USA); 1% streptomycin and penicillin solution (Biological Industries, Israel). The cells were incubated in a humid atmosphere containing 5% CO2 at 37°C and allowed to adhere for 72 hours, after which the non-adherent cells were removed by washing with PBS. The culture medium was refreshed twice per week thereafter.
When the cells had reached 70–80% confluence, the adherent cells were trypsinized with Trypsin-EDTA solution B at 37°C for 5 minutes (Biological Industries, Israel), harvested and re-plated in 75 cm2 flasks. After the third passage, a homogenous cell population was obtained, and characterization of the adherent cells was performed by flow cytometry analysis. The adipose-derived adherent cells were phenotyped as hMSCs, using mouse anti-human antibodies against CD31, CD34, CD45, CD29, CD73, CD44, CD105, HLA-AB and HLA-DR (eBioscience). The typical phenotype of hMSCs is: CD31−, CD34−, CD45−, CD29+, CD73+, CD44+, CD105+, HLA-ABC+ and HLA-DR−. Human isotype antibodies served as respective controls (eBioscience). hMSCs were maintained in culture for no more than six passages. MSCs derived from different tissue sources, such as abdominal (AD1), breast (AD4) and thigh adipose tissue (AD5), were studied. Human fibroblasts from foreskin and the HaCat cell line were used as controls for hMSCs.
Separation of PBMCs and blood-cell subtypes from whole blood
Blood donations from healthy donors were purchased from the Magen David Adom National blood bank service (MADA, Tel Hashomer Hospital, Tel Aviv, Israel). Approximately 50 ml of peripheral blood received per donation was diluted with PBS 1∶3 and layered on Ficoll-Histopaque density gradient solution (Histopaque 1077; Sigma-Aldrich, Munich, Germany). The mononuclear cells obtained were enumerated using Trypan blue coloration before fluorescence staining. The PBMCs were cryo-preserved in freezing media containing 90% fetal bovine serum (FBS) and 10% dimethyl sulfoxide (DMSO), with a cell number adjusted to 107 cells/ml in cryotube vials.
Subsets of mononuclear cells, including CD4+ T cells, CD8+ T cells and CD14+ monocytes, were enriched using the RosetteSep negative-selection method, according to manufacturer's instructions (StemCell Technologies, BC, Canada). The purity of the cell populations was 90±5% for CD4+ T cells, 88±5% for CD8+ T cells and 79±6% for CD14+ monocytes (± s.e.m.), according to the flow cytometry data.
For depletion of CD14+ monocytes, PBMCs were labeled with mouse anti-human CD14-PE antibody (ImmunoTools, Friesoythe, Germany), followed by incubation with mouse anti-PE microbeads. The magnetically labeled CD14+ cells were bound on an MS column using the MiniMACS separator (Miltenyi Biotech, Bergisch Gladbach, Germany), and non-CD14+ cells were recovered.
Direct and transwell co-cultures
The hMSCs (105/well) were seeded in a complete growth medium in six-well culture plates and were allowed to attach overnight. Allogeneic PBMCs were added on the following day, and cells were co-cultured for 3 to 11 days. PBMCs from several donors were tested at different ratios, and the inhibitory effect was found to be optimal for the hMSC∶PBMC ratio of 1∶100. Unless otherwise specified, the experiments were performed with the time frame of 7 days.
For the investigation of CD8+ T cell functionality, CD8+ cells were purified using FACSAriaTM cell sorter (BD Biosciences, Immunocytometry Systems, Mountain View, CA, USA) from PBMCs cultured with (CD8MSC) or without hMSCs (CD8CONT). The purity of CD8+ T cells was >95%. The CD8MSC or CD8CONT cells were then cultured for 72 hours at 5×105 per well, in fresh medium in the presence of PBMCs to assess their effect on the proliferation of lymphocytes, or in the presence of purified CD14+ cells to assess their effect on co-stimulation markers.
For co-culture experiments using transwell supports, preventing direct cell interaction, the hMSCs were seeded in six-well culture plates and were allowed to adhere overnight. Allogeneic PBMCs were added to the upper transwell chamber (30 mm diameter) with a 0.4 µm pore membrane (transwell chamber, Costar).
In several experiments, the hMSCs were replaced by conditioned medium from hMSC cultures. One day before passaging, the hMSC culture supernatant was harvested, centrifuged and filtered through a 0.2-µm Millipore filter. All experiments were performed in duplicates.
Inhibition assay and ELISA
In some experiments, anti-TGF-β1 antibodies (R&D Systems) or indomethacin (Sigma-Aldrich) were added to the co-cultures at concentrations of 2 µg/ml and 5–50 µM, respectively. Controls were treated with an equivalent volume of the relevant carrier solution. IFN-γ and PGE2 levels were measured in tissue culture supernatants by ELISA following the manufacturer's instructions (R&D Systems, MN, USA).
NOD SCID RAG2−/−γc−/− (NSG) immunodeficient mice strain were purchased from the Jackson Laboratories (Bar Harbor, ME, USA). All animal experiments were approved and properly conducted according to the local Institutional guidelines IACUC (IL-035-04-2008).
For humanization, 24–48-hour-old NSG mice were irradiated at 100 cGy, and 105 purified human CD34+ cells were directly injected into the liver using a 1 ml insulin syringe (Terumo, Tokyo, Japan), as described previously (Marodon et al., 2009). For LPS-activated conditions, mice were intraperitoneally injected with LPS (Escherichia coli, Sigma-Aldrich, St Louis, MO) dissolved in saline solution at a dosage of 50 µg/kg body weight, or given an equal volume of PBS, serving as control PBS group.
CD34+ hematopoietic stem-cell purification
Human umbilical cord blood (hUCB) was obtained from healthy women undergoing full-term caesarian deliveries at the Rambam Health Care Campus (Haifa, Israel). Informed consent of the mother was obtained in accordance with the Declaration of Helsinki. Approximately 30 ml of hUCB received per donation were diluted with PBS 1∶3 and layered on Lymphoprep solution (Axis-Shield PoC AS, Oslo, Norway) for density gradient separation.
The mononuclear cells were then labeled with mouse anti-human CD34-PE antibody (ImmunoTools, Friesoythe, Germany), followed by incubation with mouse anti-PE microbeads. The magnetically labeled CD34+ cells were purified on an MS column by using the MiniMACS separator (Miltenyi Biotech, Bergisch Gladbach, Germany).
Transfer of hMSCs in humanized NSG mice
1×106 hMSCs (AD5) at passage three were re-suspended in 1 ml PBS and injected intraperitoneally into each humanized mouse at the age of 7–9 months. Seven days after the transfer of hMSCs, the mice were killed and samples of peripheral blood, bone marrow and spleen were harvested from the recipient mice. Levels of CD8, CD28 and CD44 expression were determined in the human CD45+ cell gate by flow cytometry. The blood mobilization rate was determined according to the blood:bone marrow ratio of human CD45+ cell number in the NSG mice, treated or not with MSC. Non-activated and LPS-activated conditions were tested.
Flow cytometry analysis
Cell phenotype was evaluated using conjugated antibodies. To determine the phenotype of the cell surface antigen, single-cell suspensions were incubated with the following antibodies: anti-CD3-PE-Cy5, anti-CD4-PE-Cy5, anti-CD8α-FITC, anti-CD8α-APC, anti-CD8β-PE, anti-CD25-PE, anti-CD14-PE and anti-CD19-FITC antibodies (ImmunoTools, Friesoythe, Germany); anti-CD105-PE and anti-CD95-PE antibodies (eBioSciences, San Diego, CA); anti-CD8β-FITC antibodies (Immunotech, Marseille, France); anti-CD16 PE, CD80-Alexa 647 and anti-CD86-APC antibodies (BioLegend, San Diego, CA); and anti-HLA-ABC-PE antibodies and anti-HLA-DR-PE mAb (Dako, Glostrup, Denmark). Isotype-matched antibodies were used as control. The cells were incubated with the respective fluorescent antibodies at a concentration of 1 µg per 106 cells per 100 µl for 30 minutes at 4°C in the dark. The cells were then washed twice with ice-cold PBS (pH 7.2), containing 1% FCS and 1% bovine serum albumin (BSA), and re-suspended in 300 µL PBS. In order to distinguish between surface and intra-cytoplasmic expression of CD8, surface CD8 was stained (using anti-CD8α-APC antibodies) before the inner CD8 (using anti-CD8α-FITC antibodies), saturating CD8 surface expression.
Viability of T cell subsets
Dying lymphocytes are characterized by morphological changes, and have a lower forward scatter (FSC) and a higher side scatter (SSC) compared with the living cells, which in conjunction with Annexin-PI staining, allow the definition of living and dead cell gates (Moulian et al., 2001). Accordingly, cell viability was determined in CD4+ and CD8+ cell subsets, cultured either alone or with hMSCs.
PBMCs (105/ml in PBS) were incubated with 10 µM CFSE (Invitrogen) at 37°C for 15 minutes. Cells were then re-pelleted and re-suspended with fresh pre-warmed medium for another 30 minutes at 37°C. CFSE-labeled PBMCs were washed twice in PBS and stimulated with beads coated with anti-CD2, anti-CD3 and anti-CD28 antibodies, at a 1∶2 bead-to-cell ratio (Milteny Biotec) to induce a significant proliferation. After 24 hours of culture, CFSE-labeled mononuclear cells were incubated with allogeneic highly purified CD8+ T cells (1∶5 ratio) from conditioning cultures, in the presence (CD8MSC) or in the absence of hMSCs (CD8CONT). After 3 days, the cells were washed and resuspended in PBS, and the CFSE level was measured in the CFSE-labeled cells.
For intra-cytoplasmic cytokine expression, a fixation and permeabilization kit was used according to manufacturer's instructions (eBioscience, San Diego, CA). The cells were then stained with anti-IFN-γ-PE mAb (eBioscience, San Diego, CA) or anti-Granzyme B-FITC antibodies (BioLegend, San Diego, CA). Intracellular FoxP3 was stained using the same kit that included the anti-FoxP3 antibody (eBioscience, San Diego, CA).
Data for three- and four-color analysis were collected on a FACSCaliburTM (BD Biosciences, Immunocytometry Systems, Mountain View, CA, USA) and analyzed using the Flow Cytometry Analysis Software ‘Flowjo’ (Tree Star, Inc.). GraphPad Prism software (San Diego, CA, USA) was used for presentation of the results and statistical analysis.
For real-time RT-PCR analysis, the PBMCs were cultured in transwell conditions to avoid any potential contamination with detaching MSCs. Isolation of RNA from PBMCs and conversion to cDNA were performed according to manufacturer's instructions, using the Aurum total RNA kit (Bio-Rad, CA, USA) and the iScriptTM cDNA synthesis kit (Bio-Rad, CA, USA). PCR amplification was performed using the SensiMix Plus SYBR (Quantance, CA, USA) according to manufacturer’s instructions, in the Stratagene Mx3005P real-time PCR platform (Stratagene, La Jolla, CA). The following primers were used: CD8 5′ left primer, 5′-CCCTGAGCAACTCCATCATGT-3′; CD8 3′ right primer, 5′-GTGGGCTTCGCTGGCA-3′; CD4 5′ left primer, 5′-GTCCCTTTTAGGCACTTGCTTCT-3′; CD4 3′ right primer, 5′-TCTTTCCCTGAGTGGCTGCT-3′; UBE202 5′ left primer, 5′-AATGGCAGCATTTGTCTT-3′; UBE202 3′ right primer, 5′-CACACAACAGAGAACAGATGGAC-3′. UBE202 was used as a housekeeping gene, as described previously (Aquea et al., 2008).
Most experiments were performed independently at least three times, and technical duplicates were systematically performed. Results are given as mean ± s.e.m. of independent experiments throughout the manuscript. The statistics were analyzed using GraphPad Prism software (San Diego, CA). Throughout the manuscript, either the paired or unpaired two-tailed Student's t-test (for paired and unpaired data, respectively) was used for comparison of the two groups. We chose to use the Student’s t-test and not a nonparametric test because of the low statistical power of the nonparametric test for small samples. However, we checked the equality of variances. In addition, because in the t-test the assumption is that the data follow a Gaussian distribution, we tested the normality of the decreased CD8 expression induced by hMSCs, by combining the results from all the experiments (n = 42) that were performed in the study by Kolmogorov-Smirnov, D'Agostino-Pearson and Shapiro-Wilk tests, and we found that the distribution was normal. Therefore, we assumed that the distribution was also normal for other markers. However, when n>3, the nonparametric test was also used and similar results were obtained. To compare three or four conditions, we used one-way variance (ANOVA) and post tests: Dunnet's multiple comparison test was used to compare the different experimental conditions with the control, and the Bonferroni test was used to compare all pairs of columns. Throughout the manuscript, the P values are as follows: ***P<0.001; **P values between 0.001 and 0.01; *P values between 0.01 and 0.05.
We thank I. Petit for helpful discussions and critical review of the manuscript, V. Morad for her contribution in the establishment and characterization of the human MSC lines, and O. Shenker and Y. Sakoury for their help in flow cytometry experiments.
This study was supported in part by the French Association Against Myopathies; MYASTAID [grant number LSHM-CT-2006-037833 to S.B.-A]; and FIGHT-MG [grant number HEALTH-2009-242-210 to S.B.-A.] from the European Community.