Here, There, and Everywhere: Myeloid-Derived Suppressor Cells in Immunology

Myeloid-derived suppressor cells (MDSCs) were first identified in patients with advanced cancer where they are profoundly immune-suppressive cells that inhibit antitumor immunity. With the notoriety of MDSCs as potent inhibitors of cancer immunotherapy and the knowledge that chronic inflammation is the dominant driving force for the development and function of MDSCs, investigators in other fields in which inflammation is also prevalent have asked whether MDSCs are induced. The answer has been a resounding “yes,” and MDSCs are now recognized as regulatory cells in many pathological settings as well as in normal physiological conditions. In contrast to cancer where MDSCs are exclusively detrimental because they are potent inhibitors of natural antitumor immunity and obstacles to cancer immunotherapies, MDSCs can be either detrimental or beneficial in noncancer settings. In addition, their potent immune-suppressive activity is being exploited in settings where cell-mediated immunity is damaging (Table I) (1–85).

Historically, MDSCs have predominantly been characterized in mouse and human cancer, but MDSCs in other settings appear to share similar key characteristics. As it has become apparent that MDSCs are involved in many immunological processes, MDSC development and function are also being explored in noncancer immune conditions. We briefly summarize the salient features of MDSCs and then focus on MDSCs in settings other than cancer. More detailed descriptions of the development and function of cancer-induced MDSCs can be found in several excellent recent reviews (1–5).

MDSCs are a diverse population of myeloid cells that span the differentiation pathway of the common myeloid progenitor cell. MDSC accumulation is initiated when NOTCH and IRF8 are downregulated and “emergency myelopoiesis” occurs. The expansion of MDSCs occurs at the expense of other cells in the common myeloid progenitor lineage.

There are two dominant phenotypes of MDSCs: monocytic (M-MDSCs) and polymorphonuclear or granulocytic (PMN-MDSCs). In mice, MDSCs are phenotypically identified as Gr1⁺CD11b⁺ cells, with M-MDSCs being CD11b⁺Ly6C⁺Ly6G⁻ and PMN-MDSCs being CD11b⁺Ly6G⁺Ly6C^−/low. Neutrophils also have the latter phenotype in infectious disease settings. In humans, MDSCs are CD33⁺CD11b⁺HLA-DR^low/−, with M-MDSCs being CD14⁺CD15⁻ and PMN-MDSCs being CD15⁺CD14⁻CD66b⁺. Cells with a third phenotype termed early stage MDSCs (eMDSCs) are present in humans; these are CD33⁺CD11b⁺ and lack other myeloid markers. The absence of HLA-DR distinguishes human MDSCs from monocytes. MDSCs are primarily identified by the immune regulatory molecules they produce. M-MDSCs generate NO synthase 2 (NOS2), NO, TGF-β, and IL-1β. PMN-MDSCs contain NOX2 and reactive oxygen species (ROS, H₂O₂, O₂⁻). Both types of MDSCs can produce arginase 1 (Arg1), peroxynitrite (ONNO⁻), PGE₂, and IL-10. These phenotypic markers in combination with assays demonstrating functional suppression of T cells (e.g., inhibition of T cell proliferation, T cell activation including IFN-γ or IL-2 production, T cell CD3ζ downregulation, and for CD8 T cells, T cell cytotoxic activity) are the defining but not exclusive parameters of MDSCs (86) (Fig. 1).

In the presence of inflammation in general, as well as certain noninflammatory molecules, MDSCs expand and gain increased suppressive potency. Proinflammatory cytokines, including IL-6, IL-1β, TNF-α, VEGF, GM-CSF, and G-CSF, the bioactive lipid PGE₂, and the DAMP/alarmins HMGB1 and S100A8/A9, all upregulate MDSCs, as do the noninflammatory molecules adenosine (87), lactoferrin (48), and leptin (33). In combination with adrenergic signaling these molecules activate MDSCs through the STAT3 and NF-κB pathways. In the tumor microenvironment (TME), inducers are produced by tumor cells or are generated by the hypoxic TME resulting in activation of an endoplasmic reticulum stress response (reviewed in Ref. 2) (Fig. 1).

MDSCs exist in the circulation of healthy individuals at low levels. When they enter the TME or other proinflammatory locales, they undergo metabolic changes. Within the TME, MDSCs become dependent on the uptake of lipids and fatty acid oxidation (FAO). Polyunsaturated fatty acids increase the differentiation of mouse MDSCs and upregulate MDSC suppressive activity by increasing ROS production via JAK-STAT3 signaling (88). Pharmacologic inhibition of fatty acid uptake and FAO reduces mouse and human MDSC suppressive potency and reprograms MDSCs to contain more mitochondria, synthesize enzymes essential for FAO, increase the oxidation rate (89, 90), and increase the expression of fatty acid transport proteins (FATPs) (91). G-CSF and GM-CSF induce mouse and human MDSCs by upregulating the expression of lipid transport receptors such as FATP2 and fatty acid metabolic pathways via STAT3 and STAT5 signaling. Increased FATP2 expression in PMN-MDSCs facilitates the uptake of arachidonic acid, which is converted intracellularly to PGE₂ (92), a dominant driver of MDSC suppressive potency (93, 94). This suppressive activity is enhanced by mast cells (95). Metabolic-driven accumulation of methylglyoxal also increases the suppressive activity of MDSCs (96) as described in MDSC suppressive mechanisms.

MDSCs circulating in blood use oxidative phosphorylation for energy generation but convert to glycolysis (Warburg effect) when they enter the hypoxic TME (97). Immunometabolic and single-cell RNA sequencing studies of MDSCs demonstrate that maturation and the acquisition of increased suppressive potency requires aerobic glycolysis and the consumption of high levels of glucose. However, MDSCs themselves consume large amounts of glucose, and their limited lifespan in vivo is likely due to insufficient quantities of glucose (6).

As alluded to above, MDSCs use multiple mechanisms to suppress adaptive and innate immunity. MDSCs are best known for their ability to inhibit T cell activation and function. They produce Arg1 and IDO and sequester cysteine, thereby limiting the availability of l-arginine, cystine, and tryptophan, amino acids essential for T activation (98–100). MDSCs also deplete T cells of essential l-arginine by their accumulation of the metabolite methylglyoxal, which they transfer to T cells. Within T cells, methylglyoxal glycates l-arginase, thereby rendering it unusable by T cells (96).

MDSCs produce ROS and NOS2 to generate peroxynitrite, which nitrates TCRs, MHC molecules, and chemokines, preventing T cell recognition of tumor Ags and blocking T cell chemotaxis (101, 102). Although MDSCs produce high levels of ROS, they are not themselves affected due to their upregulation of NF erythroid-2 (Nrf2), which controls a battery of antioxidant genes (103). L-selectin is essential for naive T cells to enter lymph nodes and become activated. MDSCs perturb this trafficking by their expression of ADAM-17, which cleaves L-selectin (104, 105). MDSCs are also potent inducers of T regulatory cells (Tregs) through their production of IDO, TGF-β, and IL-10 (106). Engagement of MDSC-expressed PD-L1 and galectin-9, ligands for the T cell inhibitory receptors PD-1 and Tim-3, respectively, drives T cell exhaustion and arrest (107, 108). Activated T cells can defend themselves against PD-L1⁺ MDSCs through their expression of FasL, which mediates apoptosis of Fas⁺ MDSCs (109). As MDSCs accumulate and differentiate, they synthesize S100A9, thereby skewing myeloid cell differentiation toward MDSCs and away from dendritic cells (DCs), resulting in diminished availability of cells for Ag processing and presentation for T cell activation (110).

MDSCs also inhibit humoral immunity. Their production of Arg1, NO, and ROS decreases B cell IgM responses and induces B cell death (111), and their synthesis of TGF-β downregulates IL-7 production and downstream STAT5 signaling, which are essential for B cell development (112).

Innate immunity is also inhibited by MDSCs. MDSCs express membrane-bound TGF-β1, which inhibits NK cell expression of NKG2D and reduces NK cell production of IFN-γ, thereby anergizing NK cells (113). MDSC production of IL-10 drives the polarization of macrophages toward an M2 phenotype, which promotes tumor progression (114).

Many of the suppressive mediators are transported in exosomes and include proteins, glycoproteins, microRNAs, and mRNAs (115, 116). MDSC activity does not involve specific receptors; however, cell-to-cell proximity is required, presumably due to concentration effects of secreted/released materials. Given the diversity of MDSCs, it is likely that their phenotype and function vary depending on the physiological setting, so that not all suppressive mechanisms are relevant in all settings. Fig. 2 summarizes the suppressive mechanisms used by MDSCs.

PMN-MDSCs and M-MDSCs share some characteristics with neutrophils and inflammatory monocytes, respectively. Whether MDSCs are a distinct population of cells or are a subset of neutrophils or inflammatory monocytes remains controversial. MDSCs were originally defined by their phenotype, their accumulation under pathological (cancer) conditions, and their exceptional ability to inhibit T cell activation and function (117). Other myeloid cells including tumor-associated neutrophils and tumor-associated macrophages share some of these properties with MDSCs. However, studies have identified the lectin-type oxidized low-density lipoprotein receptor-1 (LOX1) (118) and FATP2 as specific markers for PMN-MDSCs (5). M-MDSCs, which phenotypically resemble inflammatory monocytes, can differentiate into tumor-associated macrophages (119) and fibrocytes (120), activities that are not associated with classical monocytes. Likewise, some of the mechanisms used by MDSCs to suppress T cell activation and function, such as cysteine deprivation (99) and l-selectin downregulation (104), are not functions of neutrophils or monocytes, and single-cell transcriptional analyses of PMN-MDSCs and tumor-associated neutrophils have identified transcriptomes specific for each cell type (121). These differences do not speak to the relationship between MDSCs, neutrophils, and inflammatory monocytes; however, they define MDSCs as a functionally distinct cell population. Many reports define MDSCs without using the full repertoire of MDSC characteristics, so there is some ambiguity as to whether the cells are full-fledged PMN-MDSCs or neutrophils, or M-MDSCs or other monocytes. In this review, MDSCs are defined as cells with the established phenotype and demonstrated immune-suppressive activity.

Some of the following studies distinguish between PMN-MDSCs and M-MDSCs; however, other studies have not separated the two populations. When studies identified PMN-MDSCs and M-MDSCs, then the specific identification is used. However, when reports do not separate the two subtypes, then we use the terminology “MDSCs.”

MDSCs expand in response to inflammation in all bacterial and viral diseases studied to date. The initial expansion of MDSCs presumably protects the host against excessive inflammation. However, when MDSCs persist, they inhibit the antipathogen function of T and NK cells, leading to reduced pathogen control and increased disease progression. Therefore, MDSCs can have both protective and detrimental functions.

MDSCs play both detrimental and beneficial roles in bacterial infection, with MDSC-derived IL-10 being a key factor. Metabolic profiling demonstrated that limited amounts of environmental glucose in invasive Staphylococcus aureus disease inhibit the differentiation of MDSCs into more mature nonsuppressive cells (6). Nonetheless, MDSCs generated in S. aureus infection produce IL-10, which supports biofilm formation (7). In contrast, M-MDSC–derived IL-10 promotes clearance of Klebsiella pneumoniae in mice and reduces inflammation, thereby improving host survival (10, 11). MDSCs also accumulate in patients infected with Mycobacterium tuberculosis as well as in individuals recently exposed to M. tuberculosis (8). The MDSCs typically inhibit immunity to M. tuberculosis. However, if their leukocyte Ig-like receptor B is blocked by Ab, then human MDSCs are converted to M1-type macrophages, which mediate the intracellular killing of M. tuberculosis (9).

In general, MDSCs play a detrimental role in viral infections and the generation and maintenance of antiviral immunity. M-MDSCs accumulate in blood and the liver of hepatitis B patients where they inhibit antiviral T cell responses (12). M-MDSC–derived IL-10 promotes the development of induced Tregs, which further inhibit antiviral immunity (13). These findings have been replicated in chronic hepatitis C patients (14, 15). In HIV-infected individuals, M-MDSCs accumulate in response to IL-6 and drive Treg levels via their production of IL-10 (16). The persistence of immune dysfunction in AIDS patients on combined antiretroviral therapy is thought to be due to the persistence of M-MDSCs (17).

M-MDSCs accumulate in the blood of SARS-CoV-2–infected patients with both mild and severe disease (122, 123). Higher levels of M-MDSCs and PMN-MDSCs at the time of hospital admission are predictive of severity (19), and recovery is associated with reductions in MDSCs (122). PMN-MDSCs similarly expand in patients with severe acute respiratory distress syndrome accompanying SARS-CoV-2 infection (20), and they correlate with COVID-19–associated lymphopenia. The continued accumulation of MDSCs parallels fatal disease, although it is unclear whether MDSCs contribute to fatality through their inhibition of antiviral immunity (124). COVID-19–induced MDSCs suppress T cell activation in vitro and, as for other pathogens, the anti-inflammatory and immune-suppressive functions of MDSCs work antagonistically in that they limit the damaging inflammation, but at the same time suppress antiviral immunity. Transcriptional analysis suggests that PMN-MDSCs of COVID-19 patients with severe disease have increased Arg1 expression relative to asymtomatic patients. People who die of COVID-19 have large infiltrates of Arg1⁺ PMN-MDSCs in their lungs, suggesting that ROS production by PMN-MDSCs may contribute to lung pathology (125). In patients with asymptomatic or mild disease, MDSCs remain elevated for at least 3 mo postrecovery (126). Given the excessive levels and persistence of MDSCs, it is questionable whether new T cells are activated in immunized patients who succumb to SARS-CoV-2 infection, or whether resolution of the disease is due to adaptive immunity induced predisease by vaccination.

Myeloid cells are key players in antiparasitic immunity with different roles in protozoan and metazoan infections. Although not extensively studied, it has been reported that MDSCs are produced in tandem with enhanced myelopoiesis during several parasitic infections (127). MDSCs are also considered to be part of “trained” immunity during fungal infections including Aspergillus (128), Candida (128–130), and Cryptococcus (131). The effects of MDSCs on immunity to complex pathogens can be detrimental in some infections where T cell–mediated immunity is protective. For example, arginase inhibitors, specifically the p38 inhibitor SB202190 or the receptor tyrosine kinase vandetanib, suppress PMN-MDSCs in Cryptococcus neoformans fungal infection and enhance protective T cell responses (131).

MDSCs, particularly M-MDSCs, play a key role in kinetoplastid infections. In mouse models they resolve inflammation during Trypanosoma cruzi acute infection (21, 22) and prevent T. cruzi infiltration of the heart causing Chagas-associated myocarditis (132). A cellular therapy with mesenchymal stem/stromal cells overexpressing the MDSC inducer G-CSF increased MDSC migration to the heart (133) where they suppressed cardiac pathology. MDSCs suppress T cell function by both Arg1 (22) and NO (134). In mouse T. cruzi infection, suppression prolongs parasite survival in heart tissue. In Chagas disease, MDSCs alter ROS and NO-mediated mechanisms of T. cruzi parasite killing (22). In contrast, MDSCs in leishmaniasis expand in response to parasite Ag (23) and suppress T cell proliferation (23, 24), possibly contributing to progressive visceral leishmaniasis (25). Surprisingly, the NO induced in MDSCs is lethal to intracellular amastigote stages and protects against propagation of infection (24). In contrast, glycyrrhizic acid inhibits Cox-2–mediated MDSC suppression and restores T cell proliferation leading to control of parasite numbers in BALB/c mice (23).

There is little information on MDSCs in apicomplexan infection. This is surprising given the burden of malaria, the recent surge in understanding Cryptosporidium with the emergence of a tractable model of study, and the historically strong focus on Toxoplasma. An initial report described the generation of PMN-MDSCs in controlled human malaria challenge with Plasmodium falciparum and their suppression of CD3/CD28-mediated T cell proliferation (26). A recent paper indicates the presence of potential MDSC populations in PBMCs during naturally acquired infections with severe P. falciparum and uncomplicated Plasmodium vivax malaria, albeit suppressive capacity was not tested (135). Splenic MDSC populations, in particular NO-expressing PMN-MDSCs, expand in C57BL/6 mice infected with Plasmodium berghei ANKA (27), a mouse model of cerebral malaria. Expansion was dependent on STAT3 signaling emanating from the IL-6 receptor and correlated with promoting Th17 responses. The significance of this finding is unclear because other studies suggest that neurologic manifestations of P. berghei ANKA in mice are independent of IL-17 (136).

Schistosoma japonicum worm Ag and egg Ag also drive differentiation of MDSCs via STAT3 (137) with ROS-dependent suppression on antischistosome T cell responses (28). Similarly, PD-L1 expression by PMN-MDSCs has been suggested to suppress antischistosome Tfh1 cells (28), potentially impacting antischistosome humoral immunity. MDSCs also expand in Echinococcus granulosus infection of BALB/c mice (138). These MDSCs use NO to suppress Th2 cell development that drives antihelminth mechanisms of clearance (29). E. granulosus–driven MDSCs have elevated transcription of VEGF and induce angiogenesis in cultured HUVECs (139). Furthermore, multiple microRNAs involved in immunoregulatory pathways (139) as well as long coding RNAs (140) are enriched in MDSCs from E. granulosus–infected mice.

Mast cells are key mediators in parasite expulsion, but they effect MDSC function, which can feed back into antiparasite immunity. In the TME, mast cells augment accumulation of M-MDSCs (141, 142), in part through histamine released during mast cell degranulation (143). However, mast cells are critical mediators of parasite expulsion, and PMN-MDSCs enhance antiparasitic immunity and have been correlated with elevated levels of the Th2 cytokines IL-4, IL-5, IL-13 (144), as well as IL-17 and the alarmin IL-33. Depletion of MDSCs by gemcitabine enhanced infection with the rat hookworm Nippostrongylus (144) and the nematode Trichinella spiralis (145). Although off-target effects of gemcitabine were assessed, there was no separation of neutrophils from PMN-MDSCs, making it difficult to separate effects of PMN-MDSCs from the protective effects of neutrophil extracellular traps produced from mature neutrophils in the Nippostrongylus model (146). This point is pertinent because of the role of mast cell tryptase in the induction of neutrophil extracellular traps (147). Nonetheless, the enhancement of IgE-mediated mast cell function, a key mechanism in mast cell activation during helminth infection, suggests how MDSCs augment immunity to intestinal helminths. This activity may also impact other infections such as malaria, where the role of mast cell degranulation promotes parasite survival in the P. berghei ANKA mouse model in some (148) but not all (149) genetic backgrounds.

Sepsis results when innate immunity overresponds to infection, and the release of proinflammatory cytokines and chemokines causes cardiovascular dysfunction and organ destruction. Hyperinflammation, cytokine release syndrome, tissue hypoxia, hypercoagulation, and endothelial cell destruction are other potentially lethal events (150). Medical advances have improved survival of patients with acute sepsis; however, sepsis survivors can have long-term complications including chronic inflammation, immune suppression, tissue wasting, and lymphopenia. A study of 74 patients with severe sepsis/septic shock detected increasing levels of MDSCs throughout the 28-d study period (151), and high levels of MDSCs during early sepsis are predictive of a poor prognosis (32). Chronic immune suppression in post–severe sepsis/septic shock patients has a poor clinical prognosis due to infection and is associated with abnormal myelopoiesis and the accumulation of MDSCs (152, 153).

Studies of postsurgical sepsis patients underscore the detrimental effects of MDSCs, particularly M-MDSCs, in suppressing immunity to infection. In 267 survivors of surgical sepsis, MDSC levels were elevated for at least 6 wk postinfection (30), suppressing Ag-driven T cell proliferation and cytokine production and leading to increased nosocomial infections. Cardiac surgery patients with elevated levels of M-MDSCs had higher incidences of postsurgery sepsis-induced immune suppression and subsequent mortality due to infection (154). Elevated levels of M-MDSCs were similarly found to predispose to sepsis-induced immune suppression and mortality due to infection in 301 septic shock patients, 50 of whom exhibited post–septic shock immune suppression (31). A single-cell RNA sequencing study of MDSCs from two 21-d postsepsis patients and two healthy controls identified common transcripts as well as transcripts unique to post sepsis MDSCs, suggesting that postsepsis MDSCs may have a distinct transcriptome (155).

Studies in mice have provided mechanistic insight into MDSCs in sepsis. A study using cecal ligation and puncture followed by polymicrobial infection demonstrated that late-phase MDSCs accumulate in the vasculature of nonlymphoid tissues, including the lungs (156). In the same model, COX2 (157) and S100A9 (158) are key drivers of MDSCs in postsepsis immune suppression, and in a mouse LPS-induced sepsis system Nrf2 is essential for MDSC survival (159). In a neonatal sepsis model (bacterial infection of 3- to 4-d-old mice) MDSCs contain elevated transcripts for NOS2, Arg1, and IL-27p28 and express TLR2, TLR4, and TLR5, which recognize multiple pathogen-associated molecular patterns of Escherichia coli (160).

The chronic presence of multiple proinflammatory mediators such as IL-6, TNF-α, PGE₂, and IL-1β in obesity is reminiscent of the TME. Because these molecules are also inducers of MDSCs, it is not surprising that MDSCs are elevated in obese patients (161, 162). Mice that become obese through consumption of a high-fat diet (HFD) have elevated levels of MDSCs in their circulation and adipose tissue, and tumor growth is accelerated (33, 37, 163). Although MDSC-like cells are elevated in morbidly obese patients and the levels decrease following bariatric surgery, these cells have low levels of ROS and do not suppress T cell proliferation (164). Therefore, not all MDSCs in obese individuals have the same characteristics.

Obesity is a risk factor for cancer development and progression (165–167). The discovery of MDSCs in obese individuals led to the concept that immune suppression by MDSCs contributes to cancer risk (36), and studies in tumor-bearing obese mice, including breast (39), ovarian (35), and renal (37) cancers, support this hypothesis.

Diabetes can also be a consequence of obesity and is characterized by elevated fasting glucose levels and high insulin. HFD obese mice have elevated TNF-α in the circulation and adipose tissues, as well as elevated blood glucose, and are insulin resistant. Depletion of MDSCs further increased blood glucose levels and insulin tolerance, indicating that these cells protected the mice against more extreme metabolic dysfunction, while concomitantly enhancing tumor progression (33, 34). Leptin is produced by adipose cells in response to inflammation and is an appetite suppressant in nonobese healthy individuals. It regulates the balance between metabolism and food intake and is frequently overexpressed in obese individuals who become nonresponsive to its regulatory function. Leptin levels are elevated in the blood of HFD obese mice. Depletion of MDSCs increases leptin levels and blocking the leptin receptor in HFD obese mice lowers circulating MDSC levels. Hence, leptin serves as a driver for MDSC accumulation while MDSCs downregulate leptin levels (33).

During healthy pregnancies women are tolerant to their semiallogeneic fetus, a phenomenon known as maternal–fetal tolerance. It is likely that multiple mechanisms are responsible for maternal–fetal tolerance. MDSCs appear to be one of the key mechanisms. PMN-MDSCs are present in human cord blood, accumulate in the peripheral blood and endometrium during healthy pregnancies, and are diminished in women experiencing spontaneous abortions and miscarriages (168, 169). Following parturition, blood PMN-MDSC levels decrease. Pregnancy-associated PMN-MDSCs suppress T cell activation, contain Arg1 and NOS2, produce large amounts of ROS, and polarize CD4⁺ T cells toward a Th2 phenotype (170, 171). Decreases in MDSCs during early miscarriage track with declines in estrogen and progesterone and the diverting of CD4⁺ T cells away from a Th2 phenotype and toward a Th1 phenotype, although it is unclear whether these effects are linked to MDSC levels (172).

Studies in mice demonstrated PMN-MDSC infiltration into the uterus of pregnant females. The MDSCs are activated via STAT3 and used ROS to suppress T cell activation (40, 41). Depletion of MDSCs early in gestation caused implantation failure, enhanced T cell activation, and increased T cell migration into the uterus. Induction of MDSCs by G-CSF reversed these effects. Naive T cells of pregnant mice had reduced l-selectin and were impaired in their ability to enter lymph nodes and become activated (39). Mice that contain HIF-1α–deficient myeloid cells contain fewer uterus MDSCs, decreased implantation rates, and increased abortions, indicating that hypoxia in the uterus is critical for the development of MDSCs and successful pregnancy (38).

PMN-MDSC deficiency contributes to pre-eclampsia in women. PMN-MDSCs express the HLA-G receptors ILT2 and ILT4, which are increased during pregnancy. A soluble form of HLA-G increases PMN-MDSC suppressive potency by activating STAT3 (44). Mouse studies using Qa-2, the mouse analog of human HLA-G, demonstrated that Qa-2 deficiency resulted in a pre-eclampsia–like condition and abortion that was reversed by soluble HLA-G (42). In a meta-analysis study elevated circulating levels of PMN-MDSCs were associated with and predictive of a more favorable outcome for women undergoing in vitro fertilization (43, 45).

PMN-MDSCs are also present in preterm and normal term infants but decrease rapidly after the 28-d neonatal period. Preterm delivery is a major cause of perinatal morbidity and mortality, and neonates are typically highly susceptible to infection. The presence of MDSCs may contribute to these conditions (46, 47). However, mouse studies indicate that MDSCs may also play a protective role in that their transitory presence downregulates potentially lethal inflammatory conditions that can occur in newborns (48, 49).

Acute and chronic graft-versus-host disease (GVHD) is the dominant complicating factor for allogeneic hematopoietic stem cell transplantation (alloHCT), while host-versus-graft responses are the major deterrent for successful solid organ transplantation. MDSCs naturally occur in alloHCT; however, the levels are insufficient to prevent GVHD. AlloHCT into irradiated recipients produces an inflammatory environment conducive to MDSC development. M-MDSCs and PMN-MDSCs are associated with reduced incidence of acute GVHD in humans (50, 51), and adoptive transfer of M-MDSCs protects against acute GVHD in mice (52). Likewise, increased M-MDSCs in posttransplant renal patients may protect indirectly against host-versus-graft rejection by induction of Tregs (53). Therefore, MDSCs have the potential to play an important role in reducing unwanted immune responses in transplant settings.

The immunosuppressive activity of MDSCs led to studies aimed at using in vivo induction or adoptive transfer of in vitro–generated MDSCs for combating GVHD in alloHCT and for minimizing host-versus-graft responses in solid organ transplantation (173, 174). Early mouse studies used GM-CSF in combination with G-CSF to expand MDSCs from bone marrow, which were then adoptively transferred into allogeneic mice providing proof of principle of this strategy (56). Dexamethasone in combination with GM-CSF extends survival of allogeneic cardiac transplants in mice via MDSCs that increase Tregs and use NOS2 to suppress T cell function (175). A comparison of naturally occurring MDSCs, GM-CSF–induced MDSCs, and G-CSF–induced MDSCs in a mouse cardiac transplant model identified G-CSF as most effective in extending graft survival (176). GM-CSF also expands MDSCs from human cord blood CD34⁺ cells, and inclusion of stem cell factor generates more robust MDSCs. In a xenogeneic NSG mouse system adoptive transfer of these human MDSCs decreased GVHD and upregulated Tregs (177). Thus, multiple strategies have been used to generate MDSCs in mouse and human settings.

Clinical data support a role for MDSCs in limiting GVHD in transplant settings. Posttransplant treatment with cyclophosphamide improved PMN-MDSC generation, maintenance, and suppressive activity and reduced grade II to IV acute GVHD in children undergoing alloHCT (MCC protocol 19295) (54). AlloHCT patients adoptively transferred with cells from pegylated GCF-mobilized MDSCs had reduced rates of grade III–IV acute GVHD compared with patients given nonpegylated GCF-mobilized MDSCs, suggesting that pegylation increased MDSC half-life. Overall survival rates did not differ between the pegylated and nonpegylated groups; however, pegylation may reduce the number of MDSC transfers and enhance MDSC function when using growth factor–induced MDSCs (178).

Studies have also focused on location of both solid organ and alloHCT grafts to determine whether MDSCs suppress at the graft site or systemically. CCR2-mediated migration of MDSCs to the graft site is critical in a mouse allogeneic islet transplant system and resulted in site-specific inhibition of CD8⁺ T cell function and enhancement of Tregs (179). Adoptive transfer of γδ Th17 cells in a mouse alloHCT model increased MDSC suppressive potency and drove infiltration of MDSCs into the inflamed intestine, thereby reducing local GVHD (180). In a mouse allogeneic cardiac transplant model, MDSC activation required host NKT cells and their production of IL-4 (181), and umbilical cord mesenchymal stromal cells secreting HLA-G enhanced MDSC generation, suppressive activity, and the ability to reduce acute GVHD in a mouse alloHCT system (55). Thus, other cells influence MDSC activation and function, and graft location may dictate MDSC efficacy.

In patients with leukemias it is essential that GVHD be reduced but the graft-versus-leukemia (GVL) effect be maintained. This outcome requires that MDSCs suppress GVHD without impacting the GVL response, a requirement that is supported by several studies (55, 56). This observation is perplexing because MDSCs are neither Ag specific nor MHC restricted. A study using mouse models of allogeneic T cell transfer into lethally irradiated recipients followed by engrafting with tumor cells suggested that NKG2D⁺CD8⁺ T cells with a memory phenotype are essential for elimination of leukemic cells and are resistant to MDSC-mediated suppression (57), providing a potential explanation for how GVHD can be reduced by MDSC therapy without diminishing the GVL response.

Autoimmune diseases are typically associated with local chronic inflammation and the activation of autoreactive CD8⁺ T cells coupled with excessive differentiation of CD4⁺ Th1 and Th17 cells and inhibition of protective Tregs. MDSCs are often present in animal models of autoimmunity and in patients with autoimmune diseases but do not completely prevent pathology. In several autoimmune conditions MDSCs arise spontaneously prior to remission, and some mouse models indicate that MDSCs may downregulate pathogenic Th1 and Th17 cells, suggesting that MDSCs may contribute to resolution of disease. However, there are also studies suggesting that MDSCs may exacerbate autoimmunity. Therefore, whether MDSCs can be exploited as therapeutic agents is controversial.

Most studies use adoptive transfer of MDSCs into mice with an established autoimmune disease. Experimental autoimmune encephalomyelitis (EAE) is an accepted mouse model for multiple sclerosis (MS). Mice are immunized with myelin oligodendrocyte glycoprotein (MOG)_35–55 and develop Ag-specific Th1 and Th17 cells. EAE mimics human MS in that there are cycles of disease and remission. Plasmacytoid DCs suppress MDSC accumulation in EAE (182); however, glycolipid-activated invariant NKT cells favor the accumulation of MDSCs (183). Adoptive transfer of syngeneic MOG-induced MDSCs into EAE mice reduces EAE severity by downregulating myelin-reactive Th1 and Th17 cells. PD-L1 expression by the transferred MDSCs is essential for disease remission in mice and probably in SLE patients (58). B cell production of GM-CSF increases EAE severity. PMN-MDSCs are beneficial in that they downregulate B cell synthesis of GM-CSF. MS patient data support a role for PMN-MDSC B cell interactions because cerebral spinal fluid frequencies of CD138⁺ B cells negatively correlate with PMN-MDSC and cytokine levels (184).

B regulatory cells (Bregs) play an important role in some autoimmune diseases by their production of IL-10, which downregulates proinflammatory cytokines and drives Treg differentiation (185). C57BL/6 female Roquin^san/san mice spontaneously develop systemic lupus erythematosus (SLE). IL-10–secreting Bregs expand when cocultured with MDSCs, and their adoptive transfer into SLE mice decreases renal pathology by reducing anti-DNA Abs, effector B cells, germinal center B cells, and follicular Th1 and Th17 cells (64). As with EAE, expression of PD-L1 is important for MDSC-mediated protection against SLE (186).

Similar MDSC protective effects are seen in mice with dextran sodium sulfate and/or 2,4,6-trinitrobenzene sulfonic acid–induced autoimmune colitis (66). Adoptively transferred MDSCs suppressed T cell activation in vitro and decreased intestinal inflammation, IFN-γ, IL-17, and TNF-α (60). Autoimmune uveitis is induced in mice by immunization with the retina-specific Ag interphotoreceptor retinal-binding protein in CFA followed by an injection of pertussis toxin. M-MDSCs increase in these mice during and before spontaneous resolution of disease, and adoptive transfer of these M-MDSCs accelerates remission and reduces Th1 and Th17 cells. Similar increases in M-MDSCs were associated with disease remission in patients with autoimmune uveitis (61). Adoptive transfer of mesenchymal stem/stromal cells to mice with uveitis also increased MDSC levels, resulting in decreased Th1 and Th17 cells (67).

Collagen-induced arthritis (CIA) in mice is a model for human rheumatoid arthritis and is initiated by injection of type 2 xenogeneic collagen in IFA. MDSCs develop spontaneously in CIA mice by day 35 postinitiation of CIA but do not prevent disease. However, when splenic PMN-MDSCs, M-MDSCs, or total MDSCs from CIA mice are adoptively transferred into CIA mice at day 21 after CIA induction, then arthritis symptoms are reduced. MDSC production of IL-10 is essential for the therapeutic effect and acts by increasing Tregs and downregulating Th17 cells, TNF-α, IL-6, and IFN-γ (59, 63). Adoptive transfer of MDSCs in a mouse model of psoriasis diminished imiquimod-induced skin inflammation by reducing TNF-α and IFN-γ and by increasing Tregs (62). Likewise, adoptive transfer of MDSCs reduced or prevented the onset of diabetes by 60 and 70%, respectively, in a type 1 transgenic NOD mouse diabetes model (65).

Although many mouse models suggest that MDSCs protect against autoimmune-induced inflammation, contrasting studies indicate that MDSCs may exacerbate autoimmunity by enhancing Th17 differentiation. In a MOG EAE study, the MDSCs suppressed T cells in vitro but lost potency when adoptively transferred and instead enhanced the differentiation of Th17 cells, increasing disease severity. Depletion of these MDSCs reduced disease severity (68, 69). In SLE patients, M-MDSCs and PMN-MDSCs correlate with disease severity and increase Th17 cell differentiation in vitro. Studies with humanized mice confirmed that SLE-induced MDSCs drive the differentiation of Th17 cells through an Arg1-dependent process (70). Other studies in MRL/lpr mice with SLE have confirmed that adoptive transfer of PMN-MDSCs impairs Treg development whereas M-MDSCs promote Th17 polarization via IL-1β (71). In a mouse CIA model MDSCs promoted Th17 differentiation through an IL-1β–dependent process (187). The MDSCs in this study were osteoclast progenitors that facilitate arthritis bone resorption. MDSCs in NOD mice decreased Th2 responses, and depletion of MDSCs improved Sjögren-like autoimmune symptoms (188).

Chronic stress can diminish cell-mediated immunity. The decrease in immunocompetence is commonly attributed to glucocorticoids and catecholamines via T cells that have receptors for these hormones, thus polarizing the T cells toward type 2 immunity. However, stress also induces IL-6, IL-1, and VEGF (189), raising the possibility that MDSCs may contribute to stress-induced immune effects.

Stress studies in patients are challenging because of the difficulty in evaluating the severity of stress and the multiple conditions that lead to it. In cancer, stress can accompany the period immediately following removal of primary tumor and is associated with decreased T cell function, which may contribute to metastasis. Mouse breast cancer studies indicate that removal of primary tumor increases MDSCs and enhances metastasis (72), supporting the hypothesis that metastasis increases postsurgery due to the accumulation of MDSCs. However, a clinical trial (NCT03578627) of 16 patients following breast cancer surgery indicated that circulating MDSC levels do not increase in the period immediately after surgery but are elevated in postsurgery patients experiencing additional life-induced stress factors (190).

Physical conditions such as crowding, food and/or water deprivation, and improper day/night lighting have been used to induce physical and psychological stress in animal models. In a mouse restraint model stress-induced IL-6 led to an increase in PMN-MDSCs via STAT3 signaling, an effect that was reversed by administration of β-adrenergic inhibitors (191–193).

Clinical studies suggest that MDSC differentiation is induced following trauma-induced inflammation but indicate that the MDSCs may be beneficial in the immediate aftermath of the incident. Patients with mild to severe trauma develop high intracellular ROS and elevated levels of activated PMN-MDSCs. Because MDSCs produce the anti-inflammatory cytokine IL-10, they downregulate trauma-associated inflammation by decreasing Th17 and Th1 cells (75). Immune deficiency can occur following a severe traumatic event, and patients can succumb to infection (194) possibly via MDSC-mediated T cell suppression. In a mouse model of abdominal trauma, splenic MDSCs were elevated 6–72 h following the initiation of trauma, and T cell dysfunction ensued via an Arg1-dependent mechanism (73), supporting this hypothesis. In rat femur fracture and polytrauma models, MDSC levels peaked in the spleen 2 h after initiation of trauma, returning to normal within 6–18 h. MDSCs also increased in a mouse pseudofracture trauma model where neutralization of the DAMP HMGB1 reversed the MDSC increase and restored T cell function (74).

As laboratory mice age, hematopoiesis becomes skewed toward myelopoiesis and this is associated with increased differentiation of functional MDSCs (195). Myeloid skewing also occurs in humans (196), but whether functional MDSCs are increased with age in the absence of pathology remains unclear. The wealth of mouse data are matched by knowledge of human MDSCs in cancer, but not in healthy aging, although sparse data suggest there may be parallels with the mouse. Despite the many reviews on this topic, the marked differences between humans and mice (197) are not usually made clear. Hence, the limited data solely on human MDSCs and aging are considered here.

A much-cited seminal paper (198) reported that the mean absolute numbers of peripheral blood cells with the PMN-MDSC phenotype CD11b⁺CD15⁺CD33⁺HLA-DR⁻ were on average slightly but significantly higher in healthy older-versus-younger Canadians (61–76 versus 19–59 y of age) but notably higher in frail seniors (67–99 y of age), associated with higher levels of proinflammatory cytokines that may facilitate MDSC production. Hence, higher levels of cells with the phenotype of MDSCs may primarily be associated with the degree of individual frailty and “inflammaging” (199). Other investigations on MDSCs in human aging have suggested that very old Brazilians (80–100 y of age) possess a higher percentage of circulating PMN-MDSCs than do 20- to 30-y-old controls, but absolute numbers per milliliter of blood were not different in this case (200). However, suppressive capacity was not tested in either of these studies, so the cells could not be confirmed as MDSCs. As far as we are aware, there is a single paper addressing functionality, showing that bone marrow MDSCs from older donors retain their suppressive functionality (76). The papers (196, 198, 200) with quite sparse data are constantly cited in support of the contention that MDSCs increase with age in healthy humans, but the actual published evidence for this is not overwhelming. It is nonetheless highly likely that the commonly slightly enhanced systemic levels of inflammatory mediators seen in older humans, especially in frailty, coupled with the likely myeloid skewing of hematopoiesis contribute to higher levels of immunosuppression by MDSCs in older adults as a further negative impact of inflammaging, also associated with poorer responses to vaccination. As we have argued before (201), there are few data on the numbers, types, and functions of MDSCs in older humans. However, the much larger datasets in the clinical context are consistent with their important role in healthy aging.

Perhaps surprisingly, there are few studies on the impact of MDSCs on either prophylactic or therapeutic vaccination, but most of the available data suggest that MDSCs reduce vaccine efficacy. Vaccination itself induces a local inflammatory reaction and MDSCs. Surprisingly, multiple doses of SARS-CoV-2 vaccines do not induce MDSCs (202), perhaps due to the unique lipid composition of the vaccines. In mice, bacillus Calmette–Guérin recruits immune-suppressive M-MDSCs that impair T cell priming in the draining lymph node (77). In rhesus macaques immunized with an mRNA influenza vaccine, M-MDSCs and PMN-MDSCs invade the injection site but not the draining lymph nodes (78), suggesting that vaccine immune responsiveness may not always be affected. This possibility is supported by a study of South African infants up to 1 y of age that showed no correlation between MDSC levels and responsiveness to common childhood vaccines; however, the levels of MDSCs were only <1% at 6 wk of age and decreased thereafter (203). Elevated levels of MDSCs that occur in diet-induced obese mice inhibit Ab production and T cell proliferation to a hepatitis B vaccine (79). MDSCs play a unique role in HIV vaccines as shown in a mouse model using EcoHIV infection. Initial infection produces HIV-1 GAG-specific CD8⁺ T cells that reduce virally infected cells. However, within 7 d CD8⁺ T cells are rapidly reduced and PD-L1–expressing HIV-infected MDSCs appear and persist (80), potentially reducing vaccine efficacy and duration of protection.

In disease-prevalent low- and middle-income countries, individuals in endemic areas of parasite infection may have infection-induced MDSCs, and vaccine design for use in such countries needs to consider MDSC immune suppressive activity regardless of vaccine target (81). Therefore, vaccine design against parasitic infection in these areas needs to consider MDSC immune-suppressive activity. Indeed, in a mouse model of Chagas disease, depletion of MDSCs enhanced the efficacy of an anti–T. cruzi vaccine (82).

Along similar lines, therapeutic vaccines for cancer (e.g., tumor peptides, recombinant viral vector vaccines, DC-based vaccines) must consider the presence of existing MDSCs, given that MDSCs impair DC function (204). In an ovarian cancer mouse model, a prime/boost vaccine strategy with an Ag-armed oncolytic virus combined with depletion of MDSCs and PD-1 checkpoint blockade was most efficacious in restoring antitumor CD8⁺ T cell activity (83). Immunization with a viral vector–based cancer vaccine targeting a mouse human papillomavirus (HPV) tumor similarly resulted in optimal activation of HPV E7-specific CD8⁺ T cells and maximal mouse survival when combined with sunitinib to deplete MDSCs (84). In a clinical trial of small-cell lung cancer patients (NCT00617409), 41.7% of patients given DC-transduced with wild-type p53 in combination with all-trans retinoic acid to deplete MDSCs developed p53-reactive CD8⁺ T cells compared with 20% of patients receiving only the vaccine (85).

The presence of MDSCs in many diseases and nonpathological physiological conditions has made them an obvious target for intervention. In diseases where they are detrimental, strategies are being developed to eliminate them. In conditions where they are beneficial, MDSCs are being exploited as therapeutic agents. For example, in pregnancy, MDSCs facilitate maternal–fetal tolerance and could be induced or adoptively transferred to enhance in vitro fertilization and successful pregnancies.

Where MDSCs exhibit a dual protective/detrimental role it is unclear whether they should be targeted. Their protective effects in reducing metabolic dysfunction in obesity are countered by their association with increased cancer risk. Likewise, MDSCs provide protection against excessive inflammation during acute sepsis, but they contribute to postsepsis immune suppression. MDSCs may also play a dual role in immunity to infections in that their anti-inflammatory activity prevents excessive inflammation while impairing cell-mediated immune mechanisms for clearing pathogens. The same situation accompanies normal aging where MDSCs could temper inflammaging but at the cost of depressing immunity to infection. Similarly, the immune-suppressive potency of MDSCs makes them a prime cell for controlling GVHD in transplant patients, and multiple mouse studies have shown that adoptive transfer of large quantities of MDSCs can produce disease remission in autoimmune conditions. However, this strategy has the potential to increase the risk of infection in the recipients.

In pathologic situations such as cancer where MDSCs are exclusively detrimental, therapeutic strategies to eliminate them are already benefiting patients in clinical trials. However, for situations in which MDSCs play a dual beneficial and detrimental role, a comprehensive understanding of the quantity, timing, location, and specific phenotype of MDSCs is needed to intervene specifically and effectively. Given the variation in pathologies and individuals, efficacious intervention in the latter conditions is likely to require customization.

The authors have no financial conflicts of interest