Autoimmune diabetes is caused by a destruction of pancreatic β-cells by autoreactive immune response, leading to insulin insufficiency/deficiency and hyperglycemia and fatal complications. This disease afflicts up to 10 million people worldwide. There is no cure for autoimmune diabetes. Insulin injection is the only supportive medication, which always accompanies fatality. Apart from replacement therapy using insulin and/or β-cells, immune interventions hold the key to stopping this illness. Myeloid-derived suppressor cells have emerged as a new regulator in harnessing immune response. In this review, we first up-dated the advances on etiology, development and immune interventions of autoimmune diabetes. Next, we highlighted the origin, development, tolerogenic mechanisms of myeloid-derived suppressor cells with an emphasis of the signaling pathways in their development and action. Finally, we summarized and discussed the recent progress in exploring the potential and mechanism of myeloid-derived suppressor cells in autoimmune diabetes. A novel vista on MDSC-based immune intervention with AID development was also discussed.
Keywords: Autoimmune diabetes; MDSC; Immune cells; Immune intervention; MDSC development; Soluble mediators; Cell contact; Mechanism and immune tolerance
AID: Autoimmune Diabetes; MDSC: Myeloid- Derived Suppressor Cells; NOD: Non-Obese Diabetic; B: Biobreeding LETL: Long Evans Tokushima Lean; APC: Antigen-Presenting Cells; Teff Cells: Effector T Cells; Treg Cells: Regulatory T Cells; MHC II: Histocompatibility Complex Class II; CLTA4: Cytotoxic T-Lymphocyte Antigen 4; IL: Interleukin; IFN: Interferon; TGF: Tumor Growth Factor; Inos: Inducible Nitric Oxide Synthase; NO: Nitric Oxide; TNF: Tumor Necrosis Factor; ROS: Reactive Oxygen Species; CCR2: CC Chemokine Receptor; CCL2: CC Chemokine Ligand; MMP: Matrix Metalloproteinase; TLR: Toll-Like Receptors; IMC: Immature Myeloid Cells; PIR: Paired Immunoglobulin-Like Receptors; ICAM-1: Intercellular Adhesion Molecule 1
Autoimmune Diabetes (AID)
Cause, pathogenesis and current therapy of AID: In 2005, the US National Institutes of Health estimated that 23.5 million people, ~ 8% of Americans, suffer from autoimmune diseases with direct health care costs totaling 100 billion dollars annually . Among over 100 autoimmune diseases whose causes are identified, autoimmune diabetes, known as type 1 diabetes, is estimated to afflict as many as 10 million people worldwide .
As with other autoimmune diseases, AID is initiated and developed by an interaction of environment, genes and immune system . Environmental factors such as infectious agents, commensal microbiota, sex hormones and diets contribute to the establishment and (re)shaping of the immune system [3-5]. A number of genes such as major histocompatibility complex class II (MHC II), cytotoxic T-lymphocyte antigen 4 (CLTA4), insulin and many others are implicated in the immune response that regulates AID outcome . Over-reactive immune system also occurs in AID patients . As a result of their interplay, AID stems from a loss of insulin-producing pancreatic β-cells caused by infiltrating immune cells, resulting in hypoinsulinemia, hyperglycemia and fatal complications.
So far, there has been no cure for AID. Daily injection of insulin is the only medication. However, this treatment cannot match the naturally precise timing and dosing of insulin secretion of the pancreas in response to hyperglycemia, leading to severe complications, namely, kidney failure, retinopathy, cardiovascular disease, and chronic ulcers . A variety of strategies has been developed, aimed at re-establishing physiological insulin production in diabetic patients . Despite some progress, devising a means capable of restoring self-tolerance or specifically correcting autoimmunity is a crucial step toward reversing AID. In this respect, regulatory Treg and Myeloid-Derived Suppressor (MDSC) have received particular attention [8-10].
Animal models of AID: Since the access to clinical samples of AID patients is always limited, animal models of AID are requisite for pre-clinical studies. Different animal models of AID have been used in pre-clinical research, including chemical-induced diabetic mice, Non-Obese Diabetic (NOD) mice, Biobreeding (BB) rats, Long Evans Tokushima Lean (LETL) rats, New Zealand white rabbits, Chinese hamsters, Keeshond dogs and Celebes black . These animal models have provided a rich information about the inherently complex development of this disease [11,12], albeit the fact that certain differences in AID pathogenesis between humans and animals exist .
Based on many similar genetic, immunological and pathological features with human patients, NOD mice stand out as the most commonly used model for AID study [14,15]. Using these animal models, more than 400 compounds were demonstrated effective against AID in pre-clinical settings and some of them are in clinical trials .
Development and immune intervention with AID: Similar to AID patients, NOD mice can spontaneously develop diabetes (Figure 1) . At the initiation of AID, leukocytes begin to invade the pancreatic islets . This invasion, called insulitis, gradually induces a loss of pancreatic β-cells and, eventually, gives rise to insulin insufficiency and deficiency, a hallmark of AID . Overwhelming evidence shows that T-cells play a key role in AID development, though B cells, Dendritic Cells (DC), macrophages, NK cells and other immune cells are also implicated [18,19]. During its development, Antigen-Presenting Cells (APC) infiltrate into the inflamed pancreas, capture auto-antigens and move to pancreatic lymph nodes. Upon T-cell receptor engagement by MHC and auto-antigens, effector T(Teff) cells are activated and differentiated into different T-cell subsets. Eventually, these cells go to the pancreas and release interferon (IFN)-α, tumor necrosis factor (TNF)-α and perforin leading to the destruction of β-cells. Treg cells have been recognized as an inhibitory player in AID development. In 2007, MDSC was coined to describe suppressor cells of myeloid origin  and, later on, manifested their ability to inhibit AID [9,10]. Accordingly, harnessing immune cells at the aforementioned steps may reduce autoimmunity and β-cell destruction and, in turn, lead to AID alleviation and cure (Figure 1).
Figure 1: Cause, Development and Immune intervention of AID. Genes, environment and immune dysregulation drive AID development. During disease development, antigen-presenting cells (APC) capture autoantigens, move to the pancreatic lymph nodes (PLN). After activation, Teff cells differentiate into different subsets, enter the pancreatic islets, release pro-inflammatory cytokines (IFN-γ, TNF-a and perforin) and destroy b-cells. Eight intervention steps to prevent AID are proposed. Step 1: APC activation. Step 2: Activation and differentiation of T cell subsets, respectively. Step 3: Impairment of APC by MDSC. Step 4: Inactivation of Teff cells by MDSC. Step 5: Induction of Treg by MDSC. Step 6: Treg cell inhibition of Teff activation by IL-2 deprivation and B7 reduction. Step 7: Reduction of TNF-a and IFN-γ production in Teff cells by Treg cells. Step 8: Migration of immune cells into pancreatic islets. Arrow (thin line) and inhibitory sign (thick line) indicate promotion and suppression, respectively.
Origin and development of MDSC: MDSC were found to accumulate in bone marrows, spleens, and tumors in tumor bearing hosts about three decades ago [21,22]. In the last 10 years, research and clinical interest in MDSC has increasingly grown [23-27]. MDSC represent a heterogeneous population of myeloid progenitor cells induced by inflammatory mediators in malignancies, infections, wounds, transplants, and autoimmune disorders . Their composition and percentage vary with diverse pathological conditions [28,29]. However, MDSC of different origins show great suppressive activities [30-32]. MDSC are featured in their morphological, phenotypic, and functional heterogeneity [23,24,28]. Nonetheless, further characterization of MDSC is now limited by their heterogeneous complexity and scarcity of reliable markers .
The development of MDSC in different circumstances is not well understood . In physiological conditions, Immature Myeloid Cells (IMC) differentiate from myeloid progenitors and, gradually, mature into dendritic cells, macrophages, and granulocytes/neutrophils upon migrating to the periphery (Figure 2A). In pathological conditions, abundant growth factors associated with diseases stimulate IMC expansion and subsequently, disturb their normal differentiation in bone marrow .
Moreover, inflammatory mediators of pathologies can aberrantly drive IMC to activate and polarize into MDSC with different phenotypes  (Figure 2B). As a consequence, MDSC emigrate from bone marrow and accumulate in peripheral tissues. The question as to whether MDSC in the periphery, spleens versus tumor sites, hold the same characteristics is not resolved. Based on lineage markers, MDSC can be classified into Gr1+CD11b+CD115+Ly6C+ monocytic (M)-MDSC and Gr1+CD11b+Ly6G+ granulocytic (G)-MDSC in mice [20,36,37]. A consensus in the markers for human MDSC is not apparent. Depending on cancer types, human MDSC are characterized as CD11b+CD14+CD33+ or Lin-HLA-DR-CD33+ myeloid cells [38,39].
Figure 2: Multiple Steps in the Development of Myeloid Cells and MDSC. (A) In physiological conditions, hematopoietic stem cells (HSC) undergo a series of expansion, differentiation, and maturation in bone marrow. Mature myeloid cells migrate to the periphery via blood vessels and replenish peripheral pool of myeloid cells. In pathological conditions, mediators of pathologies deter and divert normal HSC development to pathological development, distinguished by an increase of IMC expansion and activation. These immature myeloid cells, i.e., MDSC, migrate to the peripheral lymphoid tissues and sites of inflammation. MDSC can be categorized into two subsets, monocytic (M)-MDSC and granulocytic (G)-MDSC, by their markers (CD11b, Ly6C and Ly6G) and suppressive activities (20, 33, 36). (B) The inflammatory mediators of pathologies can regulate three developmental stages of MDSC from expansion to activation and polarization. In terms of polarization, these mediators dictate MDSC subsets to skew into M2 M-MDSC and G2 G-MDSC. Polarized MDSC subsets can be distinguished by a distinct set of signature genes in relation to their functions. M2/G2 cells produce arginase, anti-inflammatory cytokines and chemokines, eventually converging to the establishment of immune tolerance (and pro-tumoral activities). In marked contrast, M1 and G1 cells produce iNOS, NO, inflammatory cytokines and chemokines, leading to their immunogenic effects (and tumoricidal activities). Whether MDSC polarization is an irreversible process or a reversible hyperactivation state remains elusive (34).
Mirroring the nomenclature of type 1 classic activation-like (M1) and type 2 alternative activation-like (M2) macrophages, polarized MDSC can be defined as M1 and M2 cells based on their corresponding phenotypes and functions (Figure 2B). The molecular basis of MDSC development at the stages of expansion, activation, and functional polarization is largely unknown. One signal model was originally proposed to explain the requirement of one of tumor-associated factors for MDSC development.
More recently, this model was evolved into the “two signal model” stating that two distinct tumor-associated mediators are required at the stages of MDSC expansion and activation . Since MDSC development from expansion to activation and functional polarization is a multiple-step process, “multiple signal model” in which multiple factors/signals are necessary for this process should be considered. However, type and mechanism of the pathology-associated factors in pathogenesis of MDSC from hematopoietic progenitors remain mostly unclear.
Functional polarization of MDSC is less studied probably due to the complexity and heterogeneity of MDSC subsets. Compelling evidence support the concept that tumor-associated MDSC predominantly exhibit M2-like phenotypes and immunosuppressive and pro-tumoral activities [28,31,32,34,36,40-42]. However, co-existence of M1 and M2 phenotypes in MDSC was observed in few cases .
The M2 M-MDSC were phenotypically characterized by a number of enhanced signature markers such as Interleukin (IL)-10, arginase, Tie-2, CD36, CD206, IL-4R and CC chemokine receptor 2 (CCR2) . Inhibitor of SHP-1/2, NSC87877, could reverse M2-like phenotypes of M-MDSC to M1-like phenotypes . M1-polarized MDSC had an elevation of signature markers such as inducible Nitric Oxide Synthase (iNOS), nitric oxide (NO), TNF-α, IFN- γR, MHC I and CCR7 . G-MDSC in tumors shares the same markers, CD11b+Ly6G+, and immunosuppressive activities as splenic G-MDSC. However, they sometimes may not be regarded as the same cells [44,45]. Like M1/M2 MDSC, G2 G-MDSC is a major population found in tumor bearing mice. G2 cells up-regulate the expression of arginase, CC Chemokine Ligand 2 (CCL2), CCL5 and Matrix Metalloproteinase 9 (MMP-9). In contrast, G1 cells show elevated expression levels of TNF-α, Fas, and Intercellular Adhesion Molecule 1 (ICAM-1). [44,46].
Polarization of MDSC involves an array of signaling cascades, leading to their acquisition of phenotypes and functionalities. Several studies showed that IFN-γ could induce iNOS expression whilst IL-4 or IL-13 increased arginase expression in MDSC as well as macrophages [4-50]. Furthermore, activation from Toll-Like Receptors (TLR), IFN- γR, IL-4R, IL-13R could modify MDSC function [32,36,37,50-53]. Consistently, membrane receptors such as TLR, Interferon- γReceptor (IFN- γR), Interleukin-4 Receptor (IL-4R) and IL-10R have been reported to participate in the function and expression of inducible Nitric Oxide Synthase (iNOS), Tumor Necrosis Factor (TNF)-α, M1 hallmarks, and arginase, M2 hallmarks [42,44]. Besides, studies in tumor-bearing hosts using pharmacological intervention and/or genetic ablation revealed that paired immunoglobulin-like receptors (PIR) and Tumor Growth Factor Β Receptor (TGF-βR) modulate the polarization of M-MDSC and G-MDSC, respectively (Figures 3A and 3B) [42,44]. Therefore, ligands, receptors, and downstream mediators of the PIR-B and TGF-βR pathways are potential targets for manipulation of functional phenotypes of MDSC that can be used for treatment of autoimmunity, cancer and other diseases. More information on the molecular basis of MDSC polarization and related functional changes is required for their further clinical applications.
Figure 3: Signaling Pathways Governing the Polarization of MDSC Subsets. (A) LPS (a TLR4 ligand), IFN-γ, IL-4, and IL-13 are present in different pathological situations. LPS and IFN-γ trigger activation of ERK, NK-κB and STAT1, leading to M-MDSC skewing to M1 cells, characterized by an up-regulation of M1 hallmark genes, iNOS and TNF-b. In contrast, IL-4 and IL-13 induce activation of STAT3/5 and M-DSC skewing to M2 cells, defined by an up-regulation of M2 related genes, arginase and IL-10 (32, 36, 37, 50-53). PIR-A and PIR-B are highly expressed in MMDSC in a paired manner (72, 73). Upon ligand binding, PIR-/Fc-γR complex is activated, resulting in enhanced M1 pathway. M1 pathway is thought to antagonize M2 pathway. The PIR-A ligands can activate PIR-B, leading to inhibition of M1 and M2 pathways (42). (B) Similar to Figure 3A, signals from LPS/IFN-γ and IL-4/IL-13 can dictate G-MDSC polarization into G1 and G2 cells, respectively. Both cell types are characterized by G1 hallmarks (TNF-a, Fas, ICAM-1, and ROS) and G2 hallmarks (arginase, IL-10 and CCL2/5), respectively. TGF-b is known as a negative regulator of G-MDSC polarization (44-46). Upon TGF-b binding in most cell types, TGF-bRII/RI dimer forms and activates MAD2/3, leading to the increase of SMAD7 expression and NF-κB inhibition (74). Current data support the concept that TGF-b inhibits G1 pathway but promotes G2 pathway. It is still unclear whether and how SAMD2/3 and SMAD7 mediated TGF-b-mediated G1/G2 polarization. Arrow (thin line) and inhibitory sign (thick line) indicate promotion and suppression, respectively.
Multiple mechanisms of MDSC in immune regulation: MDSC are one of the pivotal regulators of innate and adaptive immunity. They act as a “hub” to link and cross-talk with other immune cells in favor of immune tolerance in order to maintain disease progression and persistence. The details regarding the coordinated regulation of MDSC and other immune cells were summarized in Figure 4A [38,54,55]. MDSC exert immune suppression by cross communication with T cells, NK cells, DC , macrophages, and other immune cells via cell contact (MHC/peptide/TCR, CD28) and soluble mediators (Reactive Oxygen Species (ROS), NO, IL-10, TGF-β) [54,56,57]. MDSC can impair DC functions by decreasing maturation, antigen uptake and migration and skewing DC cytokine profile from inflammatory phenotype to anti-inflammatory one . Additionally, MDSC interact with macrophages. MDSC diminish inflammation by downregulating macrophage production of IL-12, IL-6 and MHC II. This down-regulation appears to require IL-10 and cell contact . MDSC also suppress development and function of NK cells and this suppression can be enhanced by inflammation [59,60]. As far as T cells are concerned, MDSC can induce Teff cell inactivation and apoptosis [61-65] and expand Treg cells [9,32,36,66-69]. T cell suppression and Treg expansion by MDSC are cell contact-, NO- and/or arginasedependent [36,61,70-74]. M2-like M-MDSC possess higher abilities to suppress Teff cell activation and proliferation than M1-like counterparts in the co-culture of T cells with M-MDSC and in vivo . Moreover, M-MDSC with M2 functional phenotype possess higher potency in Treg expansion than those with M1 phenotype in vitro and in vivo . M2 M-MDSC-induced Treg increase seemed to be IL-10, IL-4 and IL-13- mediated arginase-dependent . GMDSC could inhibit CD8 T cell activity in tumor-bearing hosts . However, the ability of G-MDSC to induce Treg expansion is not corroborated. Overall, MDSC with M2 functional phenotype induce higher immune tolerance than those with M1 phenotype.
Potential and mode of action of MDSC in suppressing AID
MDSC have emerged as one of key immune regulators, raising a hypothesis that MDSC can treat AID and other autoimmune diseases. This hypothesis was first assessed in mouse models of AID as evidenced by two seminal studies [9,10]. One study from our group, for the first time, demonstrated that MDSC isolated from tumor-bearing mice mediated Treg induction or Teff suppression dependently on a MHC II-dependent antigen presentation . The mechanism of action of MDSCs is via secretion of anti-inflammatory cytokines (TGF-β and IL-10), induction of CD4+CD25+Foxp3+ Treg or suppression of Teff proliferation that are beneficial for creating host immune tolerance . To understand the role of MDSCs in murine diabetes models, we showed that adoptive transfer of MDSCs reduced diabetes by 75% compared with control group in RIP-HA/Rag2-/- mice . Moreover, the protective role of MDSCs in NOD/SCID mice was investigated . NOD/SCID mice were injected with diabetogenic T cells from diabetic NOD mice in the presence of MDSCs. Consistently, protective efficacy of MDSCs is dose-dependent and single dose treatment of MDSCs showed significant long-term protection, i.e. 60% remained diabetes free over 14-week observation . The overall data prove the concept that MDSCs can suppress AID via regulation of T cell-mediated tolerance. It is worth mentioning that the MDSC were characterized as M2 MDSC. Later on, the other study confirmed the function of MDSC in AID development. They first showed that temporary B-cell depletion by anti-hCD20 antibody increased CD11b+Gr1+ splenocytes by 6% in h-CD20/NOD transgenic mice .
Next, they found that these myeloid cells inhibited T cell proliferation in vitro in a NO- and cell contact dependent fashion, suggesting that this subset had MDSC characteristics . Strikingly, they were able to employ one single dose of anti-Gr1 antibody (RB6- 8C5 clone) to induce a significant expansion of CD11b+Gr1+ cells in NOD mice whose diabetic incidence was reduced by ~40%. Besides, anti-TGF-β neutralizing antibody almost abolished the reduction of diabetic incidence in NOD mice, suggesting the implication of TGF-β in the function of CD11b+Gr1+ cells. The CD11b+Gr1+ cells showed perfect traits of MDSC as evidenced by in vitro Teff suppression and Treg induction assays . Taken together, MDSC suppress AID via multiple mechanisms involving Teff inactivation, Treg induction, cell contact and soluble mediators (TGF-β, IL-10, NO, etc.) (Figure 4B). Besides, MDSC polarization could affect a potency level of MDSC in AID prevention and/or therapy.
Figure 4: Mechanism of MDSC in Cross-talk with Immune Cells and AID Suppression. (A) MDSC is thought to be a central player in regulating immunity via interplay with macrophages (MΦ), dendritic cells (DC), NK cells, regulatory T cells (Treg), effector T cells (Teff), etc. MDSC regulate immune cells via cell contact and soluble mediators. The interaction of MDSC with different immune cells is indicated with thin-line arrow (activation) and thickline inhibitory sign (inhibition). (B) M1/M2 polarization of MDSC may affect disease and health. M1 and M2 MDSC show pro-inflammatory and antiinflammatory (tolerogenic) activities dependently on the expression of their respective signature genes. Therefore, manipulation of MDSC polarization can used to treat diseases. In this context, M2-polarized MDSC can be used as a novel cell-based therapy for AID. However, whether M1-polarized MDSC aggravate this disease is not ascertained. Arrow (thin line) and inhibitory sign (thick line) indicate promotion and suppression, respectively.
Several lines of evidence have proved the principle indicating a great potential of MDSC-based strategy for AID prevention in mouse models [9,10]. Clearly, immunotherapy with MDSC underscores the establishment of long-term immune tolerance before a complete destruction of remaining β-cells or β-cell replacement/regeneration in hosts, leading to the AID cure. However, such immunotherapy is a double-edged sword. On one hand, it can suppress aberrant autoimmunity. On the other hand, this therapy may increase the risk of infections and malignancy. Ideally, manipulating MDSC to establish antigen-specific immune tolerance can minimize the above risk, which was proven possible in the mouse model [9,10]. MDSCbased immunotherapy for AID from bench side to bed side needs to overcome several hurdles, i.e., reliable source of human MDSC, in vivo establishment of auto-antigen-specific immune tolerance by MDSC and re-establishment of MDSC induced immune tolerance after loss. Before fully exploiting MDSC for AID, more questions remain to be addressed, whether or not MDSC exert their action on macrophages, DC, B and NK cells in AID protection, the mechanism by which MDSC induce Treg cells, whether MDSC are effective for AID therapy, costeffective way of producing enough and safe MDSC for clinical trials, the relationship of MDSC polarization and AID prophylaxis/therapy and impact of MDSC on β-cell function.
AID is an autoimmune endocrine disorder with premature death. Mounting data have clearly pointed to a critical role of MDSC in autoimmune diabetes. Although some advances have been made in understanding MDSC development from expansion, activation to polarization stages in recent years, relatively little is known about the multi-stage process. Here, we brought up an evolving concept of the multiple signal model in regulating MDSC development. Moreover, the signaling cascades involving PIR and TGF-β receptors were discussed for the polarization of M-MDSC and G-MDSC, respectively. Control over this polarization might have an impact on the clinical potential of MDSC in AID therapy. A special emphasis was placed on recent progress in understanding the therapeutic potential and mechanism of action of MDSC in AID. A new view on MDSC-based interference with AID development was also discussed.
We thank the authors whose publications we cited for their contributions. This work was supported by Grants NSC 100-2320-B-001-005 and NSC 101-2320-B- 001- 027 (to WCY) from National Science Council, Taiwan.