New concepts on the immune modulation mediated by mesenchymal stem cells

Mesenchymal stem cells (MSCs) are the nonhematopoietic multipotent progenitor cells found in various adult tissues. They are characterized by their ease of isolation and their rapid growth in vitro while maintaining their differentiation potential, allowing for extensive expansion in culture that yields large quantities suitable for therapeutic use. This article reviews the immunomodulatory activities associated with MSCs. Numerous studies have demonstrated that MSCs are potently immunosuppressive in vitro and in vivo. However, this article presents a new paradigm in MSC biology, in which MSCs, at least in vitro, can undergo polarization into either a pro-inflammatory or an immunosuppressive phenotype.


Introduction
Mesenchymal stem cells (MSCs) were originally described by Friedenstein and colleagues [1,2] in the nonhematopoietic component of bone marrow where they participate in regulating hematopoietic stem cell maturation and emigration into the circulation. Since that time, they have been isolated from the connective tissue of almost all organs, including adipose, periosteum, synovial fl uid, muscle, hair follicles, root of deciduous teeth, articular cartilage, placenta, dermis, umbilical cord, Wharton's jelly, lung, liver and spleen [3][4][5]. It has been posited that MSCs in these organs, like other stem cells, function as a source of cells for replacement and regeneration during normal cellular turnover, repair of injured tissue, or in response to biological aging.
MSCs were identifi ed based on their ability to undergo diff erentiation into mesenchymal lineage cell types, including bone, cartilage, adipose tissue, muscle and tendon [4]. Th e diff erentiation capacity of MSCs was initially thought to be limited to their tissue of origin; however, studies have demonstrated that MSCs have the capacity to diff erentiate into cells of mesodermal, endodermal and ectodermal origins, at least in vitro [4,[6][7][8]. Th e therapeutic application of MSCs was suggested from early observations in preclinical animal models of disease, in which transplanted MSCs homed to sites of infl ammation within damaged tissues where some of the transplanted cells underwent diff erentiation to replace injured cells. However, it quickly became evident in a variety of disease models that the levels of improvement mediated by MSCs do not always correlate with the levels of cellular engraftment and diff erentiation observed. As such, diff erentiation may not be a primary mechanism by which MSCs mediate tissue repair. Rather, it has been widely reported that MSCs secrete bioactive levels of soluble factors (growth factors and cytokines) capable of paracrine regulation of diverse disease-associated processes, including activation of tissue-resident stem/progenitor cells, apoptosis, stimulation of vasculogenesis and inhibition of infl ammation [9][10][11][12][13][14][15]. A rapidly growing body of literature indicates that MSCs possess immunosuppressive properties [16][17][18][19][20][21][22][23]. Th erefore, the reparative function of MSCs observed in so many injury models may be, at least in part, attributed to the production of paracrine factors that direct inhibition of immune responses and function. It is also apparent that MSCs secrete pro-infl ammatory cytokines that may enhance innate immunity. Increasing evidence suggests that activation of Toll-like receptors (TLRs), one of the early immune sensors, modulates this distinct MSC activity [24][25][26][27][28][29]. Th e underlying factors produced by MSCs and their immunomodulatory mechanisms are reviewed here. Additionally, a newly described mechanism of polarization in which MSCs can be induced to be either pro-or anti-infl ammatory through diff erential TLR activation is presented.
with bone marrow-derived mesenchymal stem cells (BMSCs) [16,[30][31][32]. Th ese early observations instigated numerous studies exploring the immunomodulatory eff ect of MSCs derived from a variety of sources and species. BMSCs express low levels of human leukocyte antigen (HLA) major histocompatibility complex (MHC) class I, do not express co-stimulatory molecules (B7-1 and -2, CD40, or CD40L), and can be induced to express MHC class II and Fas ligand. Th ese features are often used to explain their 'immune privileged' status in allogeneic hosts. Furthermore, BMSCs inhibit dendritic cell maturation, B and T cell proliferation and diff erentiation, attenuate natural killer cell activity, as well as support the production of suppressive T regulatory cells (Tregs) [16,[33][34][35][36]. While the mechanisms underlying the immune modulating property of BMSCs are not fully understood, they are dependent on the secretion of soluble factors as well as direct BMSC-to-immune cell contact [33]. To date, at least 14 factors produced by BMSCs have been associated with their anti-infl am matory properties. Some soluble factors are produced by BMSCs constitutively while others are induced by BMSC interaction with infl ammatory cells and factors commonly present in sites of tissue injury.
Th e expression of indoleamine 2,3-dioxygenase (IDO) and inducible nitric-oxide synthase (iNOS) by BMSCs has been associated with repression of T-cell proliferation [37][38][39]. Recently, secretion of IDO by BMSCs was shown in vivo to both inhibit the proliferation of acetylcholine receptor-specifi c T cells and B cells and to normalize the distribution of Th 1, Th 2, Th 17 and Treg cells in experimental autoimmune myasthenia gravis that is characterized by alteration in the balance of these four Th subsets [40]. IDO catalyzes the conversion of tryptophan, an essential amino acid for T-cell proliferation, into kynurenine. IDO has been shown to exert its immunosuppressive eff ect through the local accumulation of tryptophan metabolites, rather than through tryptophan depletion [41]. Expression of IDO by BMSCs was thought to be IFN-γ dependent [37,[41][42][43][44]. However, Opitz and colleagues [45] recently demonstrated that IDO expression in BMSCs can also be upregulated by activation of TLR3 and TLR4 via induction of an autocrine IFN-β signaling loop involving protein kinase R and independent of IFN-γ. Interestingly, when human BMSCs were treated with IFN-γ in vitro, they expressed extremely high levels of IDO and very low levels of iNOS, whereas mouse BMSCs expressed abundant iNOS and very little IDO. Th ese data suggest there is species variation in the mechanisms of BMSC immunosuppression [46].
Prostaglandin E2 (PGE-2) is emerging as a central mediator of many of the anti-infl ammatory properties of BMSCs [5,35]. PGE-2 is synthesized from arachidonic acid by cyclooxygenase (COX) enzymes COX-1 and COX-2. COX-1 is constitutively expressed in BMSCs and COX-2 expression can be induced by infl ammatory cytokines such as IL-1β, IL-6, IFN-γ and TNF-α [47]. Inhibitors of PGE-2 synthesis have been shown to abrogate BMSC anti-proliferative eff ects on T [16,48,49] and natural killer cells [48,50] and almost completely abrogate the immunosuppressive eff ects of BMSCs in vitro [47]. PGE-2 also plays a major role in the BMSC-mediated inhibition of dendritic cell maturation [51][52][53]. Nemeth and colleagues [34] showed in an animal model of sepsis that BMSCs (activated by lipopolysaccharide (LPS) or TNF-α) release PGE-2 that acts on the EP2 and EP4 receptors of macrophages leading to the production and release of IL-10, a potent anti-infl ammatory cyto kine, and decreased production of the pro-infl ammatory cytokines TNF-α and IL-6. Th e reprogramming of macrophages into a regulatory-like profi le was also demonstrated in vitro by Maggini and colleagues [54].
Transforming growth factor (TGF)-β1, a well-characterized anti-infl ammatory cytokine, is constitutively expressed by BMSCs [31]. Neutralizing antibodies to TGF-β1 have been reported to impair the immuno modulatory function of BMSCs on T [31] and natural killer cells [50]. However, others have reported that TGF-β1 has no eff ect on the immunosuppressive properties of BMSCs [32,55]. Th ese discrepancies may be explained by the experimental conditions used in these studies. Th e importance of TGF-β1 in vivo was recently demonstrated in a mouse model of ragweed-induced asthma [56]. Inject ing BMSCs into ragweed-sensitized mice significantly improved lung pathology upon allergen challenge. Th e presence of pre-exisiting TGF-β-specifi c neutralizing antibodies eliminated the benefi cial eff ects of the BMSCs; furthermore, injection of BMSCs derived from TGF-β1 knockout mice had no benefi cial eff ect, suggesting that the BMSC-derived TGF-β1 is critical in suppressing the allergic responses. Treatment with BMSCs also increased the total number of Tregs in this model. Although Nemeth and colleagues did not directly connect BMSCderived TGF-β1 to the recruitment of Tregs, Patel and colleagues [57] demonstrated that in co-cultures of peripheral blood mononuclear cells and BMSCs, TGF-β1 produced by BMSCs resulted in increased numbers of Tregs.
Other factors associated with the immune modulating properties of BMSCs include HLA-G (reviewed in [58]), hepatocyte growth factor (HGF) [31,59], leukemia inhibitory factor (LIF) [60,61], IL1 receptor antagonist (IL1RA) [62,63], CCL-2 [64], galectin-3, galectin-1 and semaphorin-3A [65,66]. All of these soluble factors are capable of inhibiting T-cell proliferation, are highly expressed by BMSCs and thus may contribute to their immunosuppressive properties [65,66]. More recently, a new role was advanced for TNF-α-induced protein 6 (TNAIP6 or TSG-6) in BMSC immunosuppression. Lee and colleagues [67] found that intravenously infused human BMSCs embolized in the lung and expressed high levels of TSG-6. TSG-6 secretion is known to suppress infl am mation through the inhibition of the infl ammatory network of proteases primarily by increasing the inhibitory activity of inter-α-inhibitor, sequestration of hyaluro nan fragments and decreasing neutrophil infi ltra tion into sites of infl ammation [68][69][70][71]. In a model of acute infl ammation induced by myocardial infarction, the knockdown of TSG-6 expression by small interfering RNA in human BMSCs signifi cantly reduced MSCmediated improvements of infl ammatory responses. Further more, the administration of recombinant TSG-6 protein largely duplicated the therapeutic eff ects of the delivered BMSCs on infl ammatory responses and infarct size [67]. Together these results suggest that TSG-6 may play a key role in the anti-infl ammatory eff ects of BMSCs.

Pro-infl ammatory mediators
Th ough we are beginning to better understand the many complex mechanisms associated with the secretion by BMSCs of immune suppressive mediators like TSG-6, so far only a few reports have described the contrasting proinfl ammatory activity of BMSCs. Indeed, the observation of this divergent immune eff ect by the BMSCs came from studies primarily focused on the downstream consequences of TLR stimulation within these cells. TLRs are a conserved family of receptors that recognize pathogenassociated molecular patterns (PAMPs) and promote the activation of immune cells [72][73][74][75][76]. Many TLRs (TLR1 to TLR13) have been identifi ed and characterized in a variety of immune cell types and species. Agonists for TLRs include exogenous microbial components, such as LPS (TLR2 and 4), lipoproteins and peptidoglycans (TLR1, 2, 6), viral RNA (TLR3), bacterial and viral unmethylated CpG-DNA (TLR9), and endogenous molecules shed following cell injury, including heat shock proteins and extracellular matrix molecules [72][73][74][75][76][77]. Specifi c agonist engagement of TLRs leads to the expression of infl ammatory cytokines or co-stimulatory molecules by a MyD88 (a TLR adapter protein)-dependent or MyD88-independent signaling pathways and can promote chemotaxis of the stimulated cell. TLRs are diff erentially expressed on leukocyte subsets and nonimmune cells and may regulate important aspects of innate and adaptive immune responses [24,27,75,78,79].
MSCs are among the cells that express an array of TLRs, including TLR2, 3, 4, 5, 6 and 9 [24,27,28]. Furthermore, studies by our group established that the stimulation of MSCs with TLR agonists leads to the activa tion of downstream signaling pathways, including NF-kB, AKT and mitogen-activated protein kinase (MAPK). Consequently, activation of these pathways triggers the previously unreported induction and secretion of pro-infl ammatory cytokines, chemokines and related TLR gene products. Interestingly, the unique patterns of aff ected genes, cytokines and chemokines measured identifi ed the TLRs as potential players in the established MSC immunomodulatory properties as well as their ability to migrate towards injured tissues. Surprisingly, we noted that TLR4 stimulation with LPS led to the secretion of primarily pro-infl ammatory media tors, such as IL-1β, IL-6, IL-8, IL-12, type I IFNs and TNF-α [28]. Th ough unexpected, previous observations reported by Beyth and colleagues [80] recognized that LPS priming aff ected co-cultures of leukocytes with human MSCs and attenuated the expected human MSCmediated inhibition of T-lymphocyte activation as well as aff ected their capacity to secrete interferon. More recently, Romieu-Mourez and colleagues [81] showed that TLR stimulation in murine MSCs similarly resulted in the production of infl ammatory mediators, such as IL-1, IL-6, IL-8, and CCL5. Furthermore, they demonstrated that TLR and IFN activated murine MSCs injected within Matrigel matrices into mice resulted in the formation of an infl ammatory site attracting innate immune cells and resulting in a dramatic recruitment of neutrophils. Raicevic and colleagues [82], studying the eff ect of TLR activation within MSCs in an infl ammatory milieu, observed that this environment shifted the cytokine profi le to a pro-infl ammatory one rather than the expected immunosuppressive one. Th ey similarly observed an increase in IL-1β, IL-6, and IL-12 after TLR activation in this infl ammatory context.
Th e eff ects of TLR engagement on the typical MSC properties of tri-lineage diff erentiation (chondrogenic, osteogenic, adipogenic) and proliferation have also been assessed. For instance, Hwa Cho and colleagues [27] described a role for TLRs in proliferation and diff erentiation of human adipose-derived mesenchymal stem cells (ASCs) [24]. In another report, murine BMSCs were found to express TLRs that, upon activation, aff ected their proliferation and diff erentiation. In contrast to human BMSCs, they suggested that activation of TLR2 inhibits both diff erentiation and migration of murine BMSCs, while also promoting their proliferation. Liotta and colleagues [25] found that TLR activation had no eff ect on adipogenic, osteogenic, or chondrogenic diff eren tiation in human BMSCs. Furthermore, this report suggested equivalent roles for TLR3 and TLR4 engagement in human MSC-mediated immune modulation in contrast to those reports mentioned above. Additionally, Lombardo and colleagues [26] reported that TLR3 and TLR4 engagement within ASCs increased osteogenic diff erentiation but had no eff ect on adipogenic diff erentiation potential or proliferation. Again diff erent from the observations presented by Liotta and colleagues, it was also determined that TLR2, TLR3, and TLR4 ligation does not aff ect the ability of human ASCs to suppress lymphocyte activation.
Th ough somewhat confounding, this recent body of work on the downstream consequences of TLRs provides emerging evidence for a new pro-infl ammatory immune modulating role for BMSCs. Th e identifi cation of the molecular details for this new pro-infl ammatory BMSC role, and whether it is innate or just an in vitro artifact, awaits further investigation. However, this novel observation is important to consider given the accelerated use of BMSCs in anti-infl ammatory cell-based therapies. Additionally, as Raicevic and colleagues [82] suggest, targeting of TLRs in BMSCs may avoid deleterious consequences in their use as anti-infl ammatory therapies.
By contrast, TLR-activated pro-infl ammatory BMSCs could prove useful in breaking tolerance in the therapy of immune evasive diseases, such as cancer.

New MSC paradigm: polarization into a proinfl ammatory (MSC1) or an immunosuppressive (MSC2) phenotype
TLRs are vital in coordinating not only the pro-homeostatic tissue injury responses of immune cells but also that of MSCs of various origins. Th e immune modulating activities of MSCs downstream of TLR stimulation seem to be more complex than originally anticipated. In an attempt to resolve the confl icting data between the immune modulating eff ect of TLR3 and TLR4 activation in human MSCs, our group has proposed a new MSC

Stem cell property TLR2 (Pam3Cys) TLR3 (Poly(I:C)) TLR4 (LPS+) TLR5 (Flagellin) TLR9 (CpG-ODNs)
Proliferation hADSCs NC [24,26] NC [24,26] NC [24,26]  [24] hMSCs NC [28,45] NC [86] paradigm that takes its cue from the monocyte literature (Table 1). It is based on the premise that these heterogeneous cells can be induced to polarize into two diverse but homo ge neously acting phenotypes (Figure 1) [83]. It is established that stimulation of monocytes with known cytokines or agonists to their TLRs, including IFN-γ and endotoxin (LPS, TLR4-agonist), polarizes them into a classical M1 phenotype that participates in early pro-infl ammatory responses. IL-4 treatment of monocytes yields the alternative M2 phenotype that is associated with anti-infl ammatory resolution responses [83]. Our group proposed that human BMSCs, like monocytes, are polarized by downstream TLR signaling into two homogenously acting phenotypes, classifi ed as MSC1 and MSC2, following the monocyte nomenclature. Th e short-term, low-level exposure or 'priming' we described with TLR4 agonists polarized human BMSCs toward a pro-infl ammatory MSC1 phenotype while the downstream consequences of TLR3 priming of human BMSCs polarized them toward an immunosuppressive (MSC2) phenotype. Th is novel MSC polarization paradigm is based on the consistent but novel outcomes observed for MSC1 when compared with MSC2 for several parameters, including dissimilar patterns of secretion of cytokines and chemokines and diff erences in diff erentiation capabilities, extracellular matrix deposition, TGF-β signaling pathways, and Jagged, IDO and PGE-2 expression [29]. Th e most compelling outcome was opposite eff ects of each cell type on T-lymphocyte activation [29]. Recently, our group has also observed that the infusion of MSC1 and MSC2 cells into murine models of infl ammatory lung injury had disparate outcomes. Th e MSC1 cells aggravated the infl ammatory injury, while the MSC2 cells were anti-imfl ammatory (unpublished observations).
We suggest that an immunosuppressive phenotype has been detected for most current BMSC preparations because of the manner in which the cells are isolated from the host and the way they are expanded in ex vivo culture. Furthermore, we propose that the default BMSC phenotype must be immunosuppressive to avoid pro found and deleterious consequences of a pro-infl ammatory MSC1 phenotype on the hematopoietic stem cells that BMSCs maintain and support. While circulating or quiescent stem/progenitor cells are both equipped to respond to environmental cues, they must not actively engage immune cells or mediate repair while circulating throughout the body or maintaining hematopoietic stem cells in the bone marrow niche. In a manner analogous to the immature state maintained for monocytes, dendritic cells, and other immune cells, BMSCs are immune suppressive until a pro-infl ammatory role is required to promote tissue repair. More than likely, TLR4-priming is not the optimal way to induce the MSC1 phenotype. It is expected that a combination of other factors, such as IFNs, or direct contact with other pro-infl ammatory cells and their microenvironments along the lines of that reported by Romieu-Mourez and colleagues [81] and Raicevic and colleagues [82] will more readily induce the MSC1 phenotype. It also remains to be seen whether these newly described phenotypes are found in nature. Nonetheless, we believe that a greater understanding of the complexities driving the immune modulating properties of pro-infl ammatory MSC1 and immunosuppressive MSC2 phenotypes will allow for more successful and eff ective targeted MSC-based therapies.

Conclusions
Current evidence supports the utilization of MSCs for the treatment of the infl ammatory component of numerous diseases and autoimmune disorders. It is evident that the mechanisms by which MSCs enhance tissue repair are more complex than originally believed. MSCs secrete a wide variety of pro-and anti-infl ammatory factors that have the potential to aff ect multiple processes, such as apoptosis, angiogenesis and infl ammation. Our group has defi ned a mechanism by which MSCs can undergo molecular signaling to be either antiinfl ammatory, as widely described, or pro-infl ammatory based on the diff erential engagement of TLRs. As the application of MSC-based therapy moves forward to the  [29]. The proposed agonists for the toll-like receptor (TLR) priming schemes are listed in the yellow boxes above the specifi c priming scheme. We speculate that MSC1 may contribute to early tissue injury responses while MSC2 may contribute to later tissue resolution responses based on similar contributions by polarized monocytes in wound healing [83][84][85]. Poly(I:C), polyinosinic:polycytidylic acid. Figure adapted from [85]. clinic, it is essential that we expand our understanding of the molecular mechanisms governing both the pro-and anti-infl ammatory properties of MSCs so that we can enhance their therapeutic effi cacy.
Competing interests AMB has a pending provisional patent application fi led for the polarization of human MSCs into MSC1 and MSC2 and their clinical use in cell-based therapies.

Authors' contributions
All authors prepared the outline and wrote the initial draft of the manuscript, edited and revised both the outline and draft manuscript, providing additional references and insights.