Myeloid-derived suppressor cells in cancer: biology, regulatory networks and theranostic prospects

Abstract

Despite the remarkable progress in cancer treatment, drug resistance and immune escape still severely limit clinical efficacy, largely due to tumor-induced immunosuppression. The main driver of this suppressive environment is myeloid-derived suppressor cells (MDSCs). Tumor-derived cytokines and chemokines can induce the expansion, activation, and recruitment of MDSCs, enabling them to effectively protect tumor cells from being recognized and cleared by immune cells by establishing an immunosuppressive barrier in peripheral lymphoid organs and the tumor microenvironment (TME). Upon arrival in the TME, MDSCs may alter their gene expression patterns through metabolic reprogramming, undergo skewed differentiation toward tumor-associated macrophages (TAMs) and tumor-associated neutrophils (TANs), and initiate immunosuppression to promote tumor growth. Furthermore, they can shape an environment conducive to tumor development and metastasis through various nonimmune mechanisms. Currently, the overall understanding of the systematic integration of MDSC biological properties into therapeutic strategies remains underdeveloped. Therefore, in this review, we systematically summarize: (i) the classification and identification of MDSCs; (ii) their biological properties in the context of tumors and autoimmune diseases; (iii) tumor-promoting effects; (iv) genetic and signaling pathway regulatory mechanisms; (v) differentiation skewing in the TME; and (vi) cellular interactions with the TME. We also propose the clinical potential of MDSC-based predictive and prognostic biomarkers and outline strategies for targeting MDSCs for oncotherapy.

Introduction

According to data released by the International Agency for Research on Cancer (IARC), ~20 million new cancer cases and 9.7 million cancer-related deaths occurred worldwide in 2022.¹ Although significant progress has been made in cancer treatment, therapeutic resistance and immune evasion remain major obstacles and even determinants in clinical oncology, largely attributed to tumor-induced immunosuppression or immune escape induced by tumors. Myeloid-derived suppressor cells (MDSCs) represent a key component of this immunosuppressive milieu.^2,3 By establishing immunosuppressive barriers in peripheral lymphoid organs and within the tumor microenvironment (TME), MDSCs effectively protect tumor cells from immune-mediated destruction.⁴ Emerging research has revealed that MDSCs not only exert immunosuppressive functions but also promote tumor progression through various nonimmunological mechanisms, contrary to their role in some autoimmune diseases.^3,5,6

Research on MDSCs began in 1978 when Bennett et al. identified a population of cells with T-cell suppressive activity and named them “natural suppressor cells” (NSCs).⁷ With the advancement of research, these cells were formally designated MDSCs in 2007 for their derivation and function.⁸ Notably, MDSCs are not a single cell type but rather a heterogeneous population of myeloid cells derived from precursors of granulocytes, macrophages or dendritic cells (DCs) under pathological conditions.⁹ They share common characteristics, including myeloid origin, an immature differentiation state, and potent immunosuppressive activity, although their definition, classification, and phenotypic delineation remain partially controversial.¹⁰ Based on morphology and function, MDSCs are divided into two major subsets: granulocytic or polymorphonuclear MDSCs (G-MDSCs or PMN-MDSCs) and monocytic MDSCs (M-MDSCs). The former resemble neutrophils, while the latter are similar to monocytes, and both differ significantly among peripheral blood, lymphoid organs, and the TME. However, with the development of single-cell RNA sequencing (scRNA-seq) transcriptional profiles, academic controversies on MDSC classification indicate that MDSCs and mature myeloid cells exhibit a continuous distribution along the myeloid developmental lineage.¹¹ Under the influence of tumor-derived cytokines, MDSCs accumulate and become activated and are selectively recruited to the TME through various chemokine signals. Moreover, MDSCs may undergo metabolic reprogramming, which enables them to survive in the hypoxic and nutrient-deprived TME and to utilize metabolic byproducts such as lactate to reciprocally regulate stromal cells, thus continuously shaping an immunosuppressive milieu. At the genetic level, the gene expression patterns of MDSCs alter within the TME, exhibiting remarkable plasticity by dynamically adapting their transcriptional, metabolic and functional programmes in response to environmental cues.^12–14 This remarkable heterogeneity and plasticity make MDSCs one of the most challenging and decisive cell populations in tumor immunoregulation.

Despite the growing body of research on MDSCs in recent years, a systematic understanding that integrates their biological characteristics (i.e., origin, fate, and functional mechanisms) with diagnostic and therapeutic strategies remains incomplete. Therefore, this review aims to construct a comprehensive framework that bridges basic biology with clinical translation to elucidate the multifaceted roles of MDSCs in tumor development and progression, as well as their therapeutic and prognostic potential. First, the classical definition, classification, and phenotypic characteristics of MDSCs are summarized. Subsequently, the continuous developmental lineage classification model based on scRNA-seq is discussed, and early-stage myeloid-derived suppressor cells (eMDSCs) are incorporated into this model to address their classification challenges.

Furthermore, emerging molecular markers for more precise identification of MDSCs are introduced to elucidate their immunosuppressive functions. To depict the overall diagram of their regulatory networks, we focus on the origin and trafficking of MDSCs, describing how they are induced, activated, accumulated, and recruited to the TME by tumor-derived cytokines and chemokines and how they adapt to hypoxia and nutrient deprivation via metabolic reprogramming and altered gene expression to sustain their immunosuppressive function. Collectively, the multiple mechanisms by which MDSCs promote tumor progression include immune-mediated suppression of T cells, B cells, and natural killer (NK) cells, as well as nonimmune tumor promotion, such as angiogenesis, lymphangiogenesis, invasion, and metastasis. Based on these insights, the clinical potential of MDSCs as predictive and prognostic biomarkers is summarized, while current therapeutic strategies targeting MDSCs and key factors influencing therapeutic efficacy and resistance are discussed to provide a perspective on next-generation precision immunotherapeutic strategies targeting these populations. In summary, this review not only integrates current fragmented knowledge about MDSCs but also aims to bridge the gap between fundamental research and clinical intervention, providing new insights for the development of cancer immunotherapies and the improvement of patient outcomes.

Phenotypic characteristics and biological heterogeneity of MDSCs

MDSCs are a pathologically activated and highly heterogeneous cluster of immature myeloid cells that accumulate during chronic inflammation and tumor progression.¹⁰ These cells consist of precursor cells of granulocytes, macrophages, and DCs. They functionally possess immunosuppressive properties that suppress adaptive and innate immunity.⁵ Thus, in cancer, an immunosuppressive environment is created for tumor cells that promotes tumor progression.^8,15 Currently, the definition of MDSCs is still controversial, mainly because these immature myeloid cells have a high degree of phenotypic similarity to their mature counterparts, and there is a lack of a single, specific marker to differentiate them from one another.^16,17 Therefore, the precise categorization of MDSCs and their distinction from normal myeloid cells has been controversial. This uncertainty poses an obstacle to their reliability as markers for disease diagnosis and as target cells for clinical therapy and highlights the need for a more refined characterization framework (Table 1).

Table 1.

Phenotypic markers of MDSC subsets and normal myeloid lineages

Subset	Phenotypic markers (mouse)	Phenotypic markers (human)	Novel candidate markers (human)	Maturation markers	References
PMN-MDSC	CD11b+Ly6G+Ly6Clow	CD11b+CD14−CD15+CD66b+; Low-density fraction	LOX-1, FATP2, CD52/84, PTGER2, VISTA	CD10/15/16/66b	16,19,123,337
M-MDSC	CD11b+Ly6G−Ly6Chigh	CD11b+CD14+HLA−DR−/lowCD15−	S100A8/A9, CD115/124/49d, VISTA	CD64/115/124	6,9,11,190,286,338,339
eMDSC	N/A	Lin−HLA−DR−/lowCD33+CD14−CD15−	N/A	N/A	18
Monocytic lineages	CD11b+Ly6G−Ly6Chigh	CD14+HLA−DRhigh	N/A	CD14/16/64/115/124	6,9,11,190,286,339,340
Granulocytic lineages	CD11b+Ly6G+Ly6Clow	CD11b+CD14−CD15+CD66b+; High-density fraction	N/A	CD10/16/101	341–344

Classical molecular markers and phenotypic definitions of MDSCs

The phenotypic categorization of MDSCs is fundamental to their study, but it is an evolving field. There is no unique combination of markers that belong only to MDSCs. The lack of specificity arises from the shared developmental origin and overlapping transcriptional programmes between MDSCs and conventional myeloid cells,¹¹ but there are minimum requirements for their definition and characterization. The prevailing view is based on the expression of cell surface markers to identify the two main MDSC subgroups in mice versus humans.¹⁸ In mice, tumor-associated MDSCs are usually identified by staining for two types of surface antigens, CD11b and Gr-1, and are further classified into two major subpopulations based on differential expression of Ly6G and Ly6C: G-MDSCs or PMN-MDSCs, defined by the phenotype CD11b⁺Ly6G⁺Ly6C^low, and M-MDSCs, defined by the phenotype CD11b⁺Ly6G⁻Ly6C^high. In humans, PMN-MDSCs were identified as CD11b⁺CD14⁻CD15⁺ or CD11b⁺CD14⁻ CD66b⁺, and M-MDSCs were phenotyped as CD11b⁺CD14⁺HLA−DR^−/low cells.¹⁸ The low expression or even negativity of the major histocompatibility complex (MHC) class II molecule HLA-DR is a key feature that distinguishes them from classical mature monocytes.

It is important to note that the combination of these cell surface markers cannot identify all MDSC populations with complete precision, and the imprecision of this analysis has been a major obstacle to the development of targeted therapies, e.g., if an oncology therapeutic drug is effective in one clinical study against a specific target on the surface of a certain type of cell but not in another, it is difficult to determine whether the drug has failed or whether it has selected the wrong target cell population in the first place. This challenge has driven developments in the field of characterization of MDSCs, allowing researchers to look beyond phenotypic definitions alone and simultaneously identify markers associated with the immunosuppressive inhibitory function of MDSCs.

Emerging identification of subpopulations based on novel candidate markers

In recent years, with the development of transcriptomics and single-cell sequencing technology, a series of new surface molecules have been proposed to improve the identification accuracy of MDSCs. Classical markers have significantly improved the understanding of MDSCs and helped to define the suppressor cell population. However, functional markers can be used to study the identification and function of MDSCs simultaneously, which can better reflect the nature of MDSCs as a “pathological activation state” rather than simple cell lineage markers.

Classical markers are still “discriminative markers” that can distinguish MDSCs from other myeloid cells, but they do not directly assume immunosuppressive functions. For example, lectin-type oxidized LDL receptor-1 (LOX-1) is considered to be one of the most representative PMN-MDSC-specific markers. Transcriptome studies have shown that LOX-1 is significantly upregulated in cancer patient-derived PMN-MDSCs but not in normal neutrophils.^15,19 This identification method based on LOX-1 expression levels is of great value in resolving the heterogeneity of MDSCs, and it can solve the problem of distinguishing PMN-MDSCs from normal neutrophil subsets to a certain extent. In addition, scRNA-seq has further facilitated the expansion of markers, such as CD84, a signaling lymphocytic activation molecule family member, which is significantly upregulated in M-MDSCs but not expressed in monocytes, thus providing a molecular basis for distinguishing the two. Recent studies have also identified molecules such as CD52 and prostaglandin E receptor 2 (PTGER2) that are specifically expressed in tumor-derived PMN-MDSCs. These markers provide new tools for the fine typing and targeted intervention of MDSCs.¹⁶

Different from classical markers, which are only used for identification, functional markers have the dual properties of “markers” and “functional executive molecules”; that is, their expression can not only identify the MDSC population but also participate in immunosuppressive functions. Among them, S100A9 and VISTA are typical representatives. S100A9 is highly expressed in M-MDSCs and forms a heterodimer with S100A8, which is involved in the regulation of MDSC development, recruitment, and functional maintenance.²⁰ Under tumor-associated inflammatory conditions, S100A9 can translocate from the cytoplasm to the nucleus and act as a transcriptional cofactor to enhance the transcription of immunosuppressive factors such as interleukin-10 (IL-10) and transforming growth factor-beta (TGF-β), thus directly driving the establishment of immunosuppressive phenotypes.²¹ This mechanism suggests that S100A9 is not only a marker of MDSCs but also a molecular executor of their functional status. VISTA functions in both immune checkpoint and metabolic regulation. VISTA is highly expressed in myeloid cells (including MDSCs) and can inhibit T-cell proliferation and cytokine production by binding to T-cell receptors, which constitutes contact-dependent immunosuppression.²² At the same time, VISTA expression is further upregulated in the hypoxic TME, which enhances the immunosuppressive ability of MDSCs.²³ More importantly, VISTA expression levels are closely correlated with MDSC expansion and metabolism. Loss of VISTA was found to lead to decreased STAT3 activation and polyamine production in MDSCs, which impaired mitochondrial respiration and limited MDSC expansion.²⁴ In addition, VISTA is also involved in the consumption of environmental nutrients (such as L-arginine) and the production of inhibitory factors such as arginase 1 (ARG1), inducible nitric oxide synthase (iNOS), and reactive oxygen species (ROS), which exert the immunosuppressive function of MDSCs through multiple pathways.²⁵ Therefore, these functional markers have high translational value in clinical practice. For example, in oral squamous cell carcinoma models, VISTA blockade has been reported to delay tumor growth by reducing the number and suppressive function of MDSCs and enhancing the function of tumor-infiltrating T cells.²⁶ An in-depth study of these molecules is expected to overcome resistance to existing immune checkpoint inhibitors (ICIs), such as ICIs targeting programmed cell death protein 1 (PD-1) or programmed death-ligand 1 (PD-L1).

Continuous lineage of pathologically activated myeloid cells from the perspective of single-cell transcriptomics

In recent years, scRNA-seq has become an important tool to elucidate the differentiation trajectory and functional status of MDSCs due to its powerful ability to parse the entire transcriptome profile at the single-cell level. However, previous classification methods based on the identification of surface markers by traditional flow cytometry cannot fully reflect the high plasticity of MDSCs in different tissue environments. In contrast, scRNA-seq technology, by whole transcriptome clustering analysis, revealed that MDSCs are not a static monolithic entity but consist of multiple myeloid cell subsets, a continuous cell population that undergoes adaptive changes in response to specific environmental stresses.^11,17 This perspective suggests that MDSCs represent a transcriptional continuum of myeloid cells, fundamentally changing our understanding of their origin and function.

Studies using scRNA-seq-based dimensionality reduction and clustering analysis have shown that there are no clear transcriptomic boundaries between neutrophils and PMN-MDSCs or monocytes and M-MDSCs. This suggests that MDSCs are not an independent population but form a continuous distribution with mature myeloid cells and exist as a side branch of the myeloid differentiated lineage. The detailed developmental trajectory of MDSCs begins as multipotent myeloid progenitors in the bone marrow, passes through precursor cell stages such as early and late myeloblasts, and proceeds along the standard differentiation pathway of neutrophils. Upon encountering a specific pathological inflammatory signal, these cells do not deviate from the main pathway to form a new lineage but instead bifurcate or transition to the so-called “PMN-MDSC state” at the end of the trajectory. Similarly, M-MDSCs are embedded in the monocyte lineage.^6,11 At specific points in this trajectory, cells exhibit significant transcriptional upregulation of immunomodulatory or metabolism-related genes in response to inflammatory stress. For example, PMN-MDSCs generally highly express ARG1, S100A8/A9, and CXCR2. M-MDSCs usually highly express IL10, STAT3, TGFB1, and others.^27–30

The continuous model based on scRNA-seq technology has also resolved the long-standing controversy around early MDSCs (e-MDSCs). Previously, with deeper research, some scientists have also proposed that there exists a type of MDSC that expresses neither CD14 nor CD15 but has the Lin⁻HLA-DR^−/loCD33⁺CD14⁻CD15⁻ surface phenotype of MDSCs, which accounts for ~3% of the entire MDSC cell population and has been referred to as eMDSCs.^10,18 However, it remains controversial whether eMDSCs should currently be categorized separately as a class III subpopulation. The two central issues that have led to this controversy are that eMDSCs do not have immunosuppressive capacity and that their cellular phenotype bears a high degree of similarity to that of myeloid progenitor cells. The first issue is that eMDSCs do not possess immunosuppressive capabilities like MDSCs. Instead, the most important feature that defines a cell as an MDSC is its ability to suppress the immune response, especially T-cell function.¹¹ From the perspective of scRNA-seq and developmental trajectory, e-MDSCs can no longer be considered an independent cell subset. Rather, they represent an early stage in the transcriptional continuum of pathologically activated myeloid progenitor cells, serving as a precursor supply pool for circulating e-MDSCs. When they enter the appropriate activated microenvironment, such as the TME, they can be influenced by the environment to convert to the M-MDSC or PMN-MDSC phenotype. This concept is strongly supported by Jackson et al., who used scRNA-seq to study e-MDSCs identified in patients with aggressive IDH-wild-type (IDH-WT) glioblastoma and found that although these e-MDSCs lacked expression of mature myeloid markers at this stage and are in the very early stage of the differentiation pathway of bone marrow-derived myeloid cells (BMDM), transcriptome analysis reveals a clear trend of M-MDSC differentiation. Moreover, even at this early stage, e-MDSCs upregulate the expression of genes or pathways related to tumor-related metabolism and antioxidant capacity, such as through high expression of glucose transporters (GLUT1 and GLUT5) and key metabolic enzymes (HK2, GAPDH), to meet the high energy demand within the microenvironment. Furthermore, they adapt to the nutrient-deprived and hypoxic TME by upregulating heme oxygenase (HMOX1) and hypoxia-responsive genes (NUPR1, ERO1A).³¹

Thus, the findings of scRNA-seq challenge the conventional binary classification of MDSCs and instead support a continuum model of myeloid cell activation. This continuum model framework not only illuminates the developmental biology of MDSCs but also provides critical new insights into the specific mechanisms by which they develop resistance to ICIs.

Are MDSCs an independent cell lineage or functionally induced state?

With the application of new technologies, an increasing number of new surface markers of MDSCs have been discovered, and scRNA-seq has reshaped our understanding of myeloid cell development. However, thus far, no unique surface markers have been identified that can clearly distinguish PMN-MDSCs from neutrophils or M-MDSCs from monocytes. Given the uncertainty of the surface phenotype of MDSCs, a question has been raised about the cellular nature of MDSCs: Do MDSCs represent a truly stable cell population, and are they immature precursors of developmentally arrested myeloid immune cells or mature myeloid immune cells that have undergone reprogramming in a pathological setting? Or are they, like regulatory T cells (Tregs), a cell population evolved by the immune system for the negative regulation of immune responses? Some researchers even wonder if “MDSC” is the wrong name. The discussion of these questions will help to understand the biological characteristics of MDSCs and contribute to the rational design of effective therapeutic strategies for MDSCs.

Comparative analysis of the developmental arrest model and pathological activation model

The traditional view is that MDSCs originate from immature myeloid cells (IMCs). Under pathological conditions such as tumors or infections, a variety of inflammatory factors, such as granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin-6 (IL-6), and vascular endothelial growth factor (VEGF), reshape the differentiation pathway of myeloid cells through paracrine or systemic effects, blocking the normal differentiation of IMCs into mature granulocytes, macrophages, or DCs. This forces them to remain in an immature state and transition into cells with an immunosuppressive phenotype.^15,18,32

With the discovery of emerging evidence (such as new surface markers), another gradually dominant view is that MDSCs are not developmentally arrested cells but a state formed by the functional reprogramming of mature myeloid cells in the pathological microenvironment. This view emphasizes the functional reprogramming of cells by the microenvironment. Due to the long-term hypoxic, acidic, and nutrient-deficient environment in the TME, as well as TGF-β secreted by the tumor, neutrophils can be transformed into a PMN-MDSC-like state. A 2021 study demonstrated through transcriptome analysis that when neutrophils are recruited to the TME, their phenotype switches to the tumor-promoting, immunosuppressive “N2” state after being influenced by mediators secreted by tumor cells and the harsh conditions within the TME.³³ Moreover, neutrophils can manifest different phenotypes depending on the specific mediators to which they are exposed. Studies using mouse tumor models have shown that the immunosuppressive cytokine TGF-β can promote the polarization of neutrophils toward a protumor “N2” phenotype,³⁴ whereas the type I interferon IFN-β induces an antitumor “N1” phenotype.³⁵

A similar transformation was also observed in monocytes/macrophages; upon infiltration into the TME, these cells transitioned to an M-MDSC-like state. The 2022 study found that cytokines such as interleukin-4 (IL-4) and interleukin-13 (IL-13), as well as metabolic abnormalities, induce a shift of macrophages to a pro-tumor “M2” phenotype in the TME.³⁶ This phenotype is also characterized by high expression of markers such as ARG1, leading to immunosuppression and promoting tumor progression and metastasis. Similarly, blocking IL-4/IL-13 signaling shifts macrophages to a more antitumorigenic “M1” phenotype. Investigation of MDSC surface markers revealed that PMN-MDSCs express the mature neutrophil markers CD10 and CD16.¹⁸ Further support is that with the broadening of research perspectives, scientists have also found the presence of MDSCs in nonneoplastic diseases. In some autoimmune disease settings, MDSCs do not always maintain their immunosuppressive function but instead switch to a proinflammatory phenotype, in which the expression of inhibitory molecules is greatly reduced.³⁷ This further supports a model of environmentally driven reprogramming.

Differences in MDSC functions under different disease backgrounds and naming controversies

Another key to the recognition of myeloid cells as MDSCs is whether they have immunosuppressive capacity. MDSCs can maintain a stable immunosuppressive state under tumor conditions. As the core hub of immunosuppression, MDSCs exert their immunosuppressive function by consuming a variety of amino acids, generating reactive oxygen and nitrogen species such as NO and ROS, and upregulating the expression of immune checkpoint molecules.¹⁵ However, in autoimmune and inflammatory diseases, MDSCs show significant functional instability, which is a manifestation of their high plasticity. Under autoimmune conditions, MDSCs can shift from their classical immunosuppressive function to a proinflammatory or pathogenic role, mainly mediated by the reprogramming of intracellular signaling pathways and the exchange of proinflammatory cytokines between cells.³⁷

Taking psoriasis and asthma as examples, psoriasis is an immune-mediated disease that involves local DCs. Upon activation, DCs secrete large amounts of interleukin-23 (IL-23), leading to continuous stimulation of T helper (Th) 17 cells and their abnormal proliferation and activation.³⁸ Activated Th17 cells release a large number of cytokines, such as interleukin-17 (IL-17), interleukin-22, and tumor necrosis factor-alpha (TNF-α). These cytokines lead to the expansion and activation of MDSCs.^39,40 However, in psoriasis, MDSCs lose their critical immunosuppressive capacity, and their ARG1 expression and NO production are significantly decreased, thereby weakening their suppression of T cells.^37,41 Conversely, they secrete a large number of inflammatory factors, among which interleukin-1β (IL-1β), IL-6, and IL-23 further promote Th17 expansion. This exacerbates the pro-inflammatory phenotype shift of MDSCs, creating a pro-inflammatory positive feedback loop.^6,42

In models of allergic airway inflammation, such as bronchial asthma, MDSCs retain their immunosuppressive function, and this immunosuppression actually has a suppressive effect on the disease. In mouse models of asthma, MDSCs accumulate extensively in lung tissue and airways, where they inhibit DCs, thereby preventing the reactivation of Th2 cells and ultimately reducing airway inflammation.^43,44

In addition to these two classical examples, a growing body of research suggests that MDSCs play an important and indispensable role in the progression of other autoimmune diseases. For example, in rheumatoid arthritis (RA), the pathogenesis involves the destruction of cartilage and bone tissue due to chronic synovial inflammation. Both M-MDSCs and PMN-MDSCs are widely recruited to the synovial fluid and microenvironment of RA. Similar to psoriasis, MDSCs in the RA microenvironment also lose their immunosuppressive function.^45,46 In systemic lupus erythematosus (SLE), the abundance of M-MDSCs in peripheral blood is positively correlated with disease severity. Contrary to previous examples, ARG1 expression in MDSCs from SLE patients is not significantly affected.⁴⁷ Although they can suppress excessive T-cell immune responses, their consumption of amino acids leads to local amino acid depletion in the microenvironment. This nutrient-deprived environment leads to the persistent differentiation of naive T cells into Th17 cells, exacerbating the organ damage caused by SLE⁴⁸ (Table 2).

Table 2.

Roles of MDSCs in immune-mediated diseases

Disease	Dominant disease immune cell	Locally recruited MDSC subset	Effect of MDSCs on disease	Mechanism of MDSC action	References
Allergic asthma	Th2	M-MDSC	Immune suppression	Relies on IL-10 and Arg-1 to block DC activation of Th2 cells, reducing airway hyperresponsiveness and allergenic inflammation	345
Psoriasis	Th17	PMN-MDSC	Inflammation promotion	Loss of suppressive function, secretes IL-1β, IL-6, and IL-23 driving Th17 polarization, inducing abnormal stratum corneum hyperplasia	38
RA	Th17	Mixed	Inflammation promotion	Amplifies local Th17 responses, M-MDSCs transition into osteoclasts under receptor activator of nuclear factor-kappa B ligand stimulation, accelerating bone erosion	6,346
Systemic sclerosis	Th17	Mixed	Inflammation promotion	Secretes TGF-β to strongly activate fibroblasts, driving excessive collagen deposition and progressive tissue fibrosis	46
Systemic lupus erythematosus	Th17	Mixed	Inflammation promotion	Metabolic pressure activates stress pathways, indirectly promoting naive T-cell differentiation into Th17 cells	347
Inflammatory bowel disease	Mixed Th1 and Th17	M-MDSC	Bidirectional	Early stage may suppress excessive T-cell responses; late stage differentiates into pro-inflammatory antigen-presenting cells, exacerbating intestinal mucosal damage	348,349

Based on the above factors, many scholars have begun to question whether this nomenclature is completely accurate. The nomenclature of these cells is an evolving and controversial area due to the complex biological roles they play during immune regulation. In 1978, they were initially named “natural suppressor cells” based on their ability to suppress T cells.⁷ Subsequently, as researchers discovered that cancer progression was associated with poor immune function and the presence of large numbers of these heterogeneous cells in the body, they were referred to as IMCs and myeloid suppressor cells.⁹ Later, the term myeloid-derived suppressor cells became widely accepted after Gabrilovich et al. proposed it in 2007.⁸ However, some researchers argue that MDSCs do not always maintain their immunosuppressive function and that their ultimate phenotype is dictated by the microenvironment. For example, in some autoimmune diseases mentioned above, MDSCs can promote inflammation and Th17 activation, while the expression of their immunosuppressive molecules is reduced. It has therefore been argued that because the term “MDSC” inherently implies inhibitory activity, it should not be used to refer to cells lacking this function; instead, the term “MDSC-like cells” (MDSC-LCs) has been suggested.¹⁸ In a study on prostate cancer, researchers also referred to PMN-MDSCs as immunosuppressive neutrophils.⁴⁹ However, these terms have not gained wide acceptance, and the current mainstream view continues to use the term “MDSCs” for the following main reasons. First, these newly proposed terms also fail to cover all aspects of this heterogeneous cell population, such as its immunosuppressive and proinflammatory effects. Second, due to historical inertia, MDSCs have become a central concept in oncology, and under pathological conditions associated with cancer, MDSCs almost always exert their immunosuppressive function, thereby contributing to tumor progression and treatment resistance. Third, because MDSCs are currently a heterogeneous cell population without specific surface markers, it is difficult to define them using the traditional methods of naming immune cells.

Controversy over whether to consider MDSCs as a homeostatic negative immune regulatory population

In nonneoplastic diseases, MDSCs do not always promote disease progression and sometimes even suppress autoimmune diseases. This raises another question: do MDSCs, like Tregs, function as a negative regulatory cell population in the immune system? However, this idea has not been widely accepted. Taken together, MDSCs have the following characteristics: they may originate from both immature and mature myeloid cells, and their phenotypes highly overlap with those of normal myeloid cells. Their transcriptional and metabolic stability is dependent on continuous environmental signals, and their phenotypes and functions can be reversed once the relevant stimuli are removed, which further indicates that the suppressive phenotype of MDSCs is highly dependent on the environment rather than being a fixed lineage-intrinsic property of the cells.^37,50 scRNA-seq has shown that MDSCs are distributed in a continuous spectrum with the myeloid cell developmental lineage but do not constitute an independent developmental lineage. There is no clear transcriptomic boundary between neutrophils and PMN-MDSCs, and monocytes and M-MDSCs also show a continuous distribution.^17,51,52 These characteristics make them highly similar to tumor-associated macrophages (TAMs), rather than to immune cell subsets with well-defined developmental trajectories and stable phenotypes, such as Tregs. Therefore, the mainstream view is that MDSCs represent a highly plastic and context-dependent immunosuppressive state of myeloid cells rather than an independent cell lineage that evolved to suppress immunity.

Classification logic emphasizing functional state rather than lineage characteristics

In summary, the best approach to the MDSC controversy is to view MDSCs as a pathologically activated state in the continuous differentiation lineage of the myeloid cell population. Under chronic tumor-associated inflammatory stress, myeloid cells can be induced into the “MDSC state”, thereby acquiring immunosuppressive functions and contributing to the establishment of a tumor-promoting immunosuppressive microenvironment. Rather than viewing them as a defined spectrum of myeloid cells or a fixed developmental stage.

Shifting the focus of the MDSC controversy from clarifying its lineage to understanding its functional status has more realistic implications for cancer therapy. Immature MDSCs can differentiate and mature into nonimmunosuppressed mature cells. For pathological activation, our treatment can focus on blocking specific activated pathways (such as TGF-β and IL-4/IL-13). Due to its immunosuppressive properties, we can treat it by neutralizing its inhibitory effector molecules (such as ARG1).⁵ These therapies are characterized by targeting the function and status of MDSCs rather than their ambiguous cellular characteristics. This insight provides new avenues for the development of more precise immunotherapies targeting MDSCs.

Hematopoietic ontogenesis of MDSCs in the TME: development, differentiation, recruitment, and activation

Under normal physiological conditions, hematopoietic stem cells differentiate into common myeloid progenitors (CMPs) in the bone marrow and further differentiate into various mature immune cells, which then enter the circulation and lymphoid tissues.^53,54 However, under tumor conditions, tumor-derived cytokines reprogram the normal developmental process of myeloid cells, diverting their differentiation trajectory toward MDSCs with immunosuppressive functions instead of producing cells with normal immune effector functions⁵⁵ (Fig. 1).

Fig. 1.

MDSC expansion and activation. Under normal physiological conditions, hematopoietic stem cells sequentially differentiate into common myeloid progenitors (CMPs) and granulocyte–monocyte progenitors (GMPs). In response to different factors, these progenitors activate lineage-specific signaling pathways and mature into monocytes, dendritic cells, macrophages, and neutrophils. In cancer, tumor-associated factors induce pathological myeloid expansion and cause immature myeloid cells (IMCs) to undergo differentiation arrest and pathological activation. These factors also promote the accumulation of eMDSCs and bias their differentiation toward M-MDSC and PMN-MDSC lineages. Rb downregulation promotes the transition from M-MDSCs toward PMN-MDSCs. Following chemokine-mediated recruitment to the TME, these cells are further reprogrammed toward tumor-associated macrophage-like and neutrophil-like states, with a bias toward the tumor-promoting M2-like and N2-like phenotypes

Cytokine networks driving MDSC expansion and activation

The generation of pathological MDSCs is a process induced by tumor-derived cytokines. At present, the two-step model of expansion and activation of MDSCs is widely recognized. The first step is the massive expansion of hematopoietic progenitor cells into IMCs. Continuous inflammatory signals induce the body into a state of emergency myelopoiesis.⁵⁶ When the generation of myeloid cells exceeds the hematopoietic capacity of the bone marrow, the body initiates extramedullary hematopoiesis in the spleen to compensate and maintain the myeloid cell supply, which marks the beginning of MDSC generation.⁵⁷ The second step is the pathological activation of these IMCs. When these immature myeloid cells migrate to the tumor or inflammatory site, they further acquire immunosuppressive functions under the influence of local microenvironmental signals. This is manifested by the upregulation of L-arginine metabolism, the production of ROS, and the expression of immune checkpoint molecules, thus completing the transformation from “immature cells” to “functional MDSCs“.⁵

In this context, granulocyte-macrophage colony-stimulating factor (GM-CSF), granulocyte colony-stimulating factor (G-CSF), macrophage colony-stimulating factor, stem cell factor, VEGF, and IL-6 are involved. These factors not only promote the massive expansion of myeloid cells but also block their terminal differentiation into mature granulocytes, macrophages and DCs.⁵⁸ This allows a large number of functionally immature but highly plastic IMCs to accumulate throughout the body. When these cells are recruited to the tumor or other inflammatory sites, the activation signals, composed of interferon-γ (IFN-γ), PGE2, IL-4, IL-13, and TGF-β, cause the immature myeloid cells to shift to the immunosuppressive MDSC state.⁵⁵

The paradigm of synergistic signaling networks in MDSC expansion and suppressive programming

Tumor-derived cytokines act on the inflammatory signals of myeloid cells and eventually converge on the regulatory network of MDSCs with the STAT3/nuclear factor κB (NF-κB), C/EBPβ, and ARG1/iNOS effector axes as the core. Through cross-talk, they jointly construct a transcriptional network characterized by positive feedback, self-amplification, and epigenetic stability to maintain abnormal MDSC expansion, differentiation arrest, and continuous differentiation into an immunosuppressive state. At the same time, this signaling network structure also forms the molecular basis of treatment resistance (Fig. 2).

Fig. 2.

Signaling pathways regulating pathological expansion and activation of MDSCs. Tumor-derived cytokines, growth factors, and inflammatory mediators regulate the pathological expansion and activation of MDSCs through multiple interconnected signaling pathways. a Expansion-associated factors activate the PI3K–AKT–mTOR, JAK–STAT3/5, and Ras–Raf–Erk signaling pathways through their corresponding receptors. These signaling cascades promote pathological myeloid expansion through transcriptional activation of downstream targets. STAT3 and C/EBPβ signaling further enhance MDSC accumulation and survival, whereas STAT3/5 suppress IRF8 expression, thereby impairing normal myeloid differentiation and promoting MDSC expansion. b Activation-associated factors induce the pathological activation of MDSCs. IL-4/13 activates the JAK-STAT6 signaling pathway, whereas IFN-γ and IL-1β activate JAK1-STAT1 signaling. IL-1β and TNF-α activate NF-κB signaling through their corresponding receptors, while S100A8/A9 activates the MyD88-MAPK-NF-κB signaling axis through TLR signaling. In parallel, PGE2 signaling through EP2/4 activates both the Ras–Raf–Erk and PI3K–AKT–mTOR pathways. Collectively, these signaling cascades induce the expression of downstream immunosuppressive mediators, thereby establishing the immunosuppressive phenotype and pathological activity of MDSCs within the TME

STAT3/NF-κB axis: core hub for inflammatory signal integration and expansion

The Janus kinase/signal transducer and activator of transcription (JAK/STAT) signaling pathway regulates diverse biological functions of MDSCs. The JAK/STAT3 pathway is the core pathway driving the expansion of MDSCs. Tumor-derived IL-6, G-CSF, and VEGF can bind to their corresponding receptors on MDSCs, activate JAK, and induce STAT3 phosphorylation at the Tyr705 site, resulting in its dimerization and translocation into the nucleus.⁵⁹ At the chromatin level, pSTAT3 preferentially binds to open chromatin regions (e.g., enhancers) and recruits coactivators (e.g., CBP/p300) to induce active marks such as H3K27ac.⁶⁰ This eventually drives the expression of genes related to MDSC survival, proliferation, and immaturity, such as S100A8/A9, c-Myc, and Cyclin D1, promoting the accumulation of MDSCs in the periphery and spleen.⁵⁵

The S100A8/A9 proteins also promote MDSC migration through the NF-κB signaling pathway, demonstrating typical cross-talk between pathways. In their study, Francesco et al. found that pancreatic ductal adenocarcinoma (PDAC) directly drives ARG1 and inducible nitric oxide synthase (NOS2) expression through high levels of tumor-derived cytokines that activate STAT3. They cooperate to metabolize L-arginine and produce reactive nitrogen species (RNS), which hinder T-cell infiltration to the tumor site.⁶¹ STAT3-dependent upregulation of S100A8 and S100A9 expression in bone marrow progenitor cells from wild-type mice and S100A9 transgenic mice prevented their differentiation, leading to the expansion of MDSCs in the spleen.^15,20 In contrast, in S100A9-deficient mice, MDSCs did not expand in the peripheral blood or spleen upon either tumor cell stimulation or complete Freund’s adjuvant (CFA) stimulation. In metastatic tumor-bearing mice, blockade of S100A8/A9 binding to MDSCs using anti-carboxylated glycan antibodies reduced MDSC levels in the blood and secondary lymphoid organs.^21,62

Distinct from STAT3-dominated amplification, NF-κB is more focused on the integration and functional activation of inflammatory signals. Studies have shown that the inhibitory phenotype of M-MDSCs depends on the upregulation of the NF-κB pathway. However, NF-κB and STAT3 are not simply two parallel signaling pathways; rather, they form an “initiation-amplification-maintenance” signaling network. The activation of the NF-κB pathway in MDSCs is dependent on TNF-α provided by the TME, and IFN-γ and PGE2 can maintain or amplify this signaling pathway through synergistic action.⁵⁵ In this process, the IKK complex mediates the degradation of IκBα, allowing the p65/p50 or c-Rel complexes to enter the nucleus, after which p65 (RelA) undergoes further phosphorylation at Ser536, Ser276, and other sites. This phosphorylation enhances the binding ability of p65 to p300/CBP from the STAT3 pathway, resulting in increased transcriptional activity of p65.⁶³ Finally, this promotes MDSCs to produce immunosuppressive effector molecules such as iNOS and ROS. In addition, PGE2 can simultaneously activate the Ras/Erk pathway and increase the level of TGF-β, thereby enhancing the inhibitory effect of MDSCs on NK cells.⁵⁵

C/EBPβ pathway: molecular switch for MDSCs to acquire immunosuppressive functions

Although STAT3 and NF-κB provide the initial signals for MDSC expansion and activation, the acquisition of MDSCs’ complete immunosuppressive function is highly dependent on the transcription factor CCAAT/enhancer-binding protein β (C/EBPβ). Due to the lack of Cebpb expression, the expression and activity of immunosuppressive molecules, such as ARG1 and iNOS, are at a low level in early myeloid cells; therefore, they do not possess immunosuppressive capacity at this stage.¹⁴ Cebpb produces three distinct isoforms through alternative translation initiation from the same mRNA: LAP* (liver-enriched activator protein*), LAP, and LIP (liver-enriched inhibitory protein), which has a dominant-negative regulatory function. The immunosuppressive effect of MDSCs is strongly dependent on the ratio of LAP to LIP.^64,65 Tumor-related signaling molecules, mainly IL-6, act on myeloid cells to activate STAT3 signaling and shift the LAP/LIP ratio toward transcriptional activation through the mTOR pathway. This process is regulated by the lncRNA network.^66,67 lnc-CHOP can bind to both the ER stress protein CHOP and the inhibitory LIP isoform. This interaction abrogates the inhibitory effect of LIP, thereby enhancing LAP-mediated transcriptional activity and increasing the binding of active LAP isoforms to the promoter regions of target genes. This results in increased ARG1 production by MDSCs.^68,69 lnc-C/EBPβ, on the other hand, selectively binds to the inhibitory LIP isoform and the epigenetic regulator WDR5, thereby inhibiting the modification of H3K4me3 histone marks at downstream promoters such as IL4I1, which in turn inhibits the further differentiation of MDSCs.⁷⁰

Multi-axis synergistic regulation of suppressive effector molecule expression

The immunosuppressive function of MDSCs is ultimately achieved through the coordinated expression of effector molecules. Among them, ARG1 and iNOS/NOS2 can deplete local L-arginine and produce nitric oxide in the TME, thereby inhibiting the function of T cells and NK cells. The transcriptional activation of these molecule-producing genes is dependent on the spatiotemporal co-occupancy of STAT3, NF-κB and C/EBPβ in their promoter and enhancer regions.^71–74 In response to tumor-derived cytokines such as GM-CSF and IL-6, c-Rel translocates to the nucleus and binds directly to the promoter and enhancer regions of the ARG1 and CEBPB genes. Without the initial binding of c-Rel, the assembly of enhancers to the corresponding chromatin sites is blocked, which prevents the differentiation of myeloid cells into the MDSC state.⁷⁵ Following the initial c-Rel binding, pSTAT3, p65, and C/EBPβ (LAP) are subsequently recruited to the same promoter.⁷⁶ When these molecules are present at the promoter and form a complex consisting of c-Rel, p65, pSTAT3, and C/EBPβ, the subsequent transcription process is activated. This is followed by the recruitment of histone acetyltransferases such as p300, which acetylate C/EBPβ, thereby facilitating RNA polymerase II-mediated transcription of ARG1 and NOS2.^77–79

Taken together, NF-κB activates chromatin opening, and STAT3 provides an enhanced transcriptional signal, creating a positive feedback inflammatory circuit that drives MDSC expansion and activation.^58,80 On the other hand, C/EBPβ stabilizes the transcription complex and maintains the continuous expression of immunosuppressive genes, which eventually produce immunosuppressive effector molecules.⁸¹ This process supports the notion that MDSCs have high plasticity, explaining how inflammatory cytokines induce lineage specificity in myeloid cells and why MDSCs exhibit different functions in different environments.

These pathways are not only highly expressed in MDSCs. Tumor cells and stromal cells—including cancer-associated fibroblasts (CAFs), TAMs, and tumor-associated neutrophils (TANs)—also upregulate the expression of NF-κB and other pathways and drive the secretion of proinflammatory cytokines (such as IL-6) under the action of inflammatory signals.^82,83 These factors, in turn, activate STAT3 signaling in myeloid cells, thereby maintaining emergency myelopoiesis and MDSC expansion.

The signaling pathways within MDSCs form a coordinated transcriptional regulatory network and interact with cells in the TME, which is the basis for the continuous expansion of MDSCs and the establishment of a protumor inflammatory environment.

Chemokine-receptor axes regulating MDSC tissue infiltration

For MDSCs to further exert mechanisms such as immunosuppression to promote tumor progression, they must not only leak out of the blood and lymphoid tissues into the tumor tissue but also further migrate to various specialized regions in the TME, interact with other stromal cells, and ultimately interact with cancer cells. These interactions regulate local cell survival, proliferation, and differentiation as well as their execution of protumorigenic effector functions. These interactions do not occur randomly but are coordinated by migration-directed signals provided by chemokines and their receptors.

The infiltration of newly formed MDSCs into tumor sites is driven by tumor-associated C-C motif chemokine ligands (CCLs), C-X-C motif chemokine ligands (CXCLs), and their corresponding receptors (CCRs and CXCRs).⁸⁴ Different subpopulations of MDSCs differ in the major chemokine signals to which they respond. This variability is a key factor in determining which MDSC subset dominates a given tumor. M-MDSC recruitment mainly depends on the CCL2-CCR2 axis,⁸⁵ while PMN-MDSC recruitment mainly depends on the CXCL8/CXCL5-CXCR1/2 axis.^29,30,86 This differential recruitment of MDSCs may have profound downstream effects. The specific combination of chemokines secreted by different tumors, as well as the cellular composition of the TME, also determines the nature of the immunosuppressive landscape in the TME. In turn, differences in MDSC composition in the TME can lead to tumor progression, metastasis, and response to therapy in different directions. For example, Bonapace et al., in their 2014 study on a mouse breast cancer model, found that high CCL2 expression by tumor cells significantly recruited M-MDSCs, and these cells inhibited CD8⁺ T-cell function by secreting IL-10 and TGF-β, promoted TAM polarization to the M2 type, and accelerated metastasis formation. The TME of tumors that mainly secrete CXCL8 will be dominated by PMN-MDSCs.⁸⁷ Highfill et al. found that high expression of CXCL8 (IL-8) in melanoma drives PMN-MDSC infiltration, and these cells directly disrupt the T-cell receptor signaling pathway by releasing large amounts of ROS to destroy effector T-cell killing.⁸⁸

In addition to the above major chemokine axes, other chemokines are also able to recruit MDSCs and can complement the major chemokine axes. For example, the CCL5–CCR5 axis recruits not only MDSC subsets but also Tregs, thereby coordinating the accumulation of multiple suppressor cell types.^89,90 The CXCL12–CXCR4 axis is involved in the directional migration of MDSCs in a variety of cancers, including breast and ovarian cancer.^91–93

Moreover, the recruitment process of MDSCs is not just a bilateral interaction between circulating myeloid cells and tumor cells. MDSCs also interact with other cells and their metabolites in the TME to construct a complex MDSC chemotactic network. CAFs are the core node for MDSC recruitment signal amplification and network construction in the TME. Due to the need for rapid proliferation, tumors discharge a large amount of lactate into the TME under the “Warburg effect,” leading to acidification and hypoxia of the TME. High concentrations of extracellular lactate activate the hypoxia-inducible factor-1α (HIF-1α) and NF-κB pathways in CAF-like cells through monocarboxylate transporters, especially monocarboxylate transporter 1 (MCT1) and monocarboxylate transporter 4 (MCT4), or by binding to specific receptors on the cell membrane, resulting in the secretion of large amounts of chemokines that recruit MDSCs, such as IL-6, CXCL1, and CCL5. This intensifies MDSC infiltration into the tumor site.^94,95 Furthermore, lactate can act on the G protein-coupled receptor GPR81 (also known as HCAR1) expressed on the surface of MDSCs and initiate the downstream signaling cascade through the related Gi proteins to activate the PI3K/Akt/mTOR pathway.^96,97 This upregulates the expression of antiapoptotic proteins such as Mcl-1, promotes the growth and survival of MDSCs,⁹⁸ and leads to the chemotactic migration of MDSCs to the tumor site along the lactate concentration gradient.^99,100

This redundancy in recruitment poses a major challenge to MDSC-targeted therapy in clinical practice. When one recruitment pathway is blocked by drugs, tumors can often compensate by upregulating alternative pathways, thereby maintaining the recruitment of immunosuppressive cells and significantly reducing the effectiveness of drug therapy. In clinical studies of PDAC, it has been shown that when the CCL2-CCR2 axis is blocked by drugs, it effectively reduces the infiltration of CCR2⁺ TAMs/M-MDSCs while leading to a compensatory influx of CXCR2⁺ neutrophils/PMN-MDSCs. This alteration of the major myeloid cell population maintains the immunosuppression of the TME and limits the therapeutic efficacy.^61,101,102 Therefore, combinations of multiple chemokine axes blocking agents, or the direct targeting of downstream signaling pathway nodes such as STAT3, may be more effective because multiple upstream cytokine and chemokine signals converge on the JAK/STAT3 pathway, enabling a single agent to block multiple chemokine recruitment signals (Fig. 3).

Fig. 3.

Recruitment of different MDSC subpopulations by chemokines in the TME. A variety of cells within the TME recruit MDSCs by secreting chemokines. These chemokines recruit different MDSC subpopulations from the circulation according to the different receptors mainly expressed on the surface of different MDSCs. The recruitment of M-MDSCs is mainly mediated by the CCL2-CCR2 axis, and PMN-MDSC recruitment is mainly mediated by the CXCL-CXCR1/2 axis, where the ligands CXCL1/2/5/8 bind to CXCR2 and CXCL8 can bind to CXCR1. In addition, the CCL5–CCR5 and CXCL12–CXCR4 axes can recruit both MDSC subpopulations

The spleen as an ecological niche for extramedullary hematopoiesis

Although MDSCs accumulate at multiple sites, the spleen has an important role in tumor progression as one of the sites of its production and accumulation, as well as a key site for its immunosuppressive function.^103,104 As the largest peripheral immune organ in the human body, the spleen, with its white pulp and marginal zone, provides a sophisticated platform for dynamic immune cell interactions. The white pulp is mainly composed of lymphoid tissue, in which the T-cell-rich zone and B-cell follicles form the core of the adaptive immune response, whereas the marginal zone is the key gateway for the blood to enter the white pulp and the red pulp, which is rich in B cells, macrophages, and specialized DCs.¹⁰⁵ The vascular structure here is specialized, and blood flow is slowed down. Antigens and pathogens in the blood are first captured and processed in this region by macrophages and DCs and presented to lymphocytes (mainly B cells) to initiate an immune response.¹⁰⁶

On the basis of its unique immune microenvironmental structure, its physiological properties as an immune cell aggregation center are pathologically exploited during cancer.^107,108 Under physiologic conditions, the spleen serves as the primary reservoir for granulocytes. In the spleen, these marginalized neutrophils temporarily attach to the surface of splenic vascular endothelial cells and are exchanged between the spleen and the circulation, aiming for efficient immune surveillance.¹⁰⁹ However, under chronic inflammatory stress caused by tumors, the physiological properties of the spleen are functionally reversed. The normal differentiation process of myeloid cells is impeded by the stimulation of persistent inflammatory factors, leading to a systemic accumulation of immature myeloid cells.¹¹⁰ When the bone marrow’s hematopoietic capacity is overwhelmed, extramedullary myelopoiesis occurs, meaning hematopoietic activity takes place outside of the bone marrow, primarily in the spleen and to a lesser extent in the liver.¹¹¹ This process greatly expands the myeloid cell pool of the spleen, making it a production site for new myeloid cells. Subsequently, these cells, whose normal differentiation is impaired owing to inflammatory factors, become the main source of PMN-MDSCs in the spleen.¹¹⁰ In mouse models of chronic inflammatory diseases and cancer, the spleen enlarges owing to the accumulation of MDSCs.¹¹² This explains why even in patients with clinically low-grade gastric or pancreatic cancer, PMN-MDSC levels in the spleen are much higher than in peripheral blood.^4,113,114

The structures that make the spleen an efficient site of immune initiation also make it an efficient site of immune suppression.¹¹⁵ The sophisticated structure of the splenic marginal zone and white pulp, the high-density immune cell aggregation area and the slow blood flow microenvironment originally evolved for efficient antigen presentation and lymphocyte activation, providing a favorable environment for the immunosuppressive activity of MDSCs.¹¹⁰ The abnormal enrichment of MDSCs in the spleen, especially in the marginal zone, creates a local population of inhibitory cells that is far in excess of peripheral blood concentrations.^6,110 This high-density aggregation amplifies the paracrine effects of MDSCs through key molecules such as ARG1, iNOS, ROS, and immunosuppressive cytokines (e.g., IL-10, TGF-β).^15,116,117 Neighboring T cells, NK cells, and DCs are thus exposed to persistently high concentrations of inhibitory mediators, and their activation, proliferation, and effector functions are significantly diminished.¹⁵

The specific slow-flow vascular structure of the marginal zone prolongs the contact time between passing immune cells (especially newly recruited effector cells) and the resident MDSCs under pathological conditions.¹¹⁵ This prolonged lag period gives MDSCs a fuller opportunity to exert deep inhibition or even induce apoptosis or incapacitation of effector immune cells through direct cell-to-cell contact (e.g., checkpoint molecule interactions such as PD-1) or proximity secretion of inhibitory factors.

In addition to paracrine inhibitory factors and checkpoint molecules, MDSCs (especially monocyte-derived M-MDSCs) utilize MHC molecules expressed on their surface for more specific immunosuppression. In the spleen, a site of antigen enrichment and high T-cell aggregation, such antigen-specific interactions of MDSCs via MHC molecules can precisely target and silence T-cell clones that recognize tumor antigens, further dismantling the core of the adaptive immune response.^{2,116,118,119}

Metabolic reprogramming of MDSCs in the TME

In addition to the cytokine network and intracellular signaling pathways, the TME is the third dimension that regulates the function of MDSCs. Hypoxia, nutrient competition, and the accumulation of metabolic byproducts caused by tumor cells generate a hostile TME. This environment not only creates obstacles for immune cells to exert effective immunity but also actively remodels cells entering the TME, such as MDSCs, inducing these cells to develop a stronger tumorigenic phenotype. This process is not only the result of functional adaptation but also programmed changes driven by multilevel signaling networks to drive tumor progression and increase the difficulty of targeted therapy (Fig. 4).

Fig. 4.

Metabolic reprogramming of MDSCs driven by tumor-derived lipids, lactate, and glucose within the TME. MDSCs take up tumor-released lipids through CD36 and FATP2 and channel them into mitochondrial fatty acid oxidation (FAO), which provides acetyl-CoA for the tricarboxylic acid (TCA) cycle. Meanwhile, lactate generated by tumor cells through the Warburg effect within the TME is also taken up by MDSCs and inhibits HIF-1α degradation, thereby enhancing glycolysis and the expression of immunosuppressive molecules. Lactate also promotes MDSC survival and migration through GPR81–PI3K–AKT–mTOR signaling. Glucose and lactate supply carbon to pyruvate, acetyl-CoA and citrate, supporting glycolysis, lactate metabolism, the TCA cycle and lipid biosynthesis in MDSCs. Arachidonic acid released from membrane phospholipids (PLs) is subsequently metabolized through COX-1/2 to generate PGE2

Evolution of metabolic patterns: dynamic transition from glycolysis to fatty acid oxidation

During emergency hematopoiesis and the early expansion stage in the bone marrow, MDSCs mainly rely on efficient glycolysis to obtain ATP quickly to support their rapid proliferation.^14,120 However, scRNA-seq analysis shows that the metabolic programme of MDSCs is significantly reprogrammed from glucose-dependent to lipid-dominated after recruitment to the TME.^121,122 The most significant feature of this process is the enhanced fatty acid oxidation (FAO) capacity of MDSCs, which efficiently take up exogenous lipids and oxidized lipoproteins released by tumor cells by upregulating the fatty acid uptake receptor CD36 and the fatty acid transporter FATP2 (encoded by SLC27A2).¹²³ After entering the cell, fatty acids enter the mitochondria through the carnitine shuttle system. scRNA-seq data confirm the high expression of carnitine palmitoyltransferase 1A (CPT1A), a key rate-limiting enzyme, in MDSCs, which ensures continuous FAO. FAO provides not only a stable source of ATP for MDSC expansion and differentiation but also lipid precursors for the synthesis of immunosuppressive molecules. For example, prostaglandin E2 (PGE2) is abundantly synthesized by MDSCs, driven by high expression of cyclooxygenase-2 (COX-2) through the arachidonic acid metabolic pathway.¹²⁴ PGE2 constructs an immunosuppressive network by maintaining the suppressive phenotype of MDSCs via autocrine signaling and by inhibiting the functions of NK cells and DCs via paracrine signaling.¹²⁵ Notably, lipid accumulation also enhances the immunosuppressive function of PMN-MDSCs by inducing LOX-1 expression and the ER stress pathway.^19,126 This phenomenon suggests that lipid metabolism is not only an energy adaptation mechanism but is also directly involved in the establishment of the immunosuppressive phenotype.

Lactic acid as a key metabolic signal shaping MDSC function

Despite the importance of FAO in the TME, glycolysis remains an important component of the metabolic network of MDSCs, which is significantly enhanced under hypoxic and high-lactate conditions. Tumor cells produce large amounts of lactate through the “Warburg effect” and actively export it via MCT4 into the TME. This results in the accumulation of lactate in the TME to concentrations of 10–30 mM or higher and the acidification of the TME to a pH of 6.0–6.5.^127,128 Lactate is not merely a metabolic waste product but also a key signaling molecule that drives the functional remodeling of other cells within the TME toward a tumor-supporting phenotype, promoting tumor neovascularization and metastasis while inhibiting anticancer immunity.^96,129,130

Lactate acts on MDSCs through multiple mechanisms, such as the competitive inhibition of prolyl hydroxylases (PHDs). This prevents the hydroxylation and degradation of HIF-1α, resulting in its stable accumulation and translocation into the nucleus.¹³¹ HIF-1α then binds to hypoxia-responsive elements (HREs) to induce the expression of key immunosuppressive molecules such as ARG1, iNOS, and PD-L1, thereby directly inhibiting T-cell function.¹³² Activation of the HIF-1α pathway can further upregulate the expression of GLUT1 and lactate dehydrogenase A (LDHA), enhance the glycolytic capacity of MDSCs, and promote their own lactate production.^133,134 In models such as glioblastoma, this mechanism not only promotes ARG1 expression but also synergizes with the ROS generation pathway to amplify the inhibitory effect. There is significant cross-regulation between these mechanisms. For example, HIF-1α not only promotes glycolysis but also indirectly enhances the expression of genes related to lipid metabolism.¹³⁴ In addition, signaling molecules produced by lipid metabolism, such as PGE2, can counterregulate the stability of HIF-1α and the activity of inflammatory pathways,¹³⁵ thereby forming a positive feedback immunosuppressive network.

Differentiation potential and bias of MDSCs in the TME

The multiple immune and nonimmune tumor-promoting mechanisms of MDSCs do not act in isolation but are interconnected with the TME to form a synergistic network. Since MDSCs are cells in different states along the continuous differentiation trajectory of myeloid cells and possess strong plasticity, they interact dynamically with the TME. When tumor cells recruit MDSCs to the TME, MDSCs contribute to the formation of an immunosuppressive and structurally abnormal TME, which in turn remodels the MDSCs. This induces MDSCs to differentiate into various states.

Influence of different tissue microenvironments on MDSC differentiation trends

In peripheral lymphoid organs, MDSCs are mainly composed of PMN-MDSCs, which exhibit a relatively mild immunosuppressive mechanism. The suppressive function of these MDSCs is primarily dependent on the production of ROS.¹³⁶ In the TME, M-MDSCs account for a higher proportion and significantly upregulate the intracellular expression of ARG1, iNOS, and PD-L1.^137,138 Kumar et al. pointed out in their review that within the tumor, M-MDSCs are more prone to differentiate into TAMs.⁴ Thus, STAT3 inhibition leads to a reduction in MDSCs in the spleen of tumor-bearing mice but not within the tumors themselves. The differentiation trajectory of MDSCs varies across different types of tumors and stages of progression. For example, in a scRNA-seq study of immune cells and tumor cells from 33 patients with different grades of glioma, it was found that a new population of e-MDSCs appeared during tumor progression, and their intracellular pathways related to metabolism, oxidative stress, and hypoxia were significantly upregulated.³¹

Evolutionary patterns of gene expression profiles during cell differentiation

Harsh conditions within the TME reshape MDSCs to shift their signaling pathways and gene expression toward a pattern favoring tumor progression. The hypoxic environment generated by the TME drives high expression of HIF-1α in MDSCs, which directly induces the expression of PD-L1.¹¹⁸ In addition, hypoxia triggers the metabolic reprogramming of MDSCs to highly express a variety of immunosuppression-related genes, such as the L-arginine metabolism-related genes ARG1 and iNOS, as well as those encoding the proinflammatory calcium-binding proteins S100A8/A9 and the protumor cytokines IL-10 and TGF-β.⁵⁵ Moreover, hypoxia shifts the metabolic mechanism of MDSCs from glycolysis to FAO to provide more energy. In turn, FAO provides metabolic fuel for maintaining the high activity of ARG1 and iNOS.¹²² The above idea is supported by Corzo et al., who showed that tumor hypoxia, via HIF-1α, rapidly upregulates ARG1 and iNOS in splenic MDSCs and drives their differentiation into protumor macrophages.¹³⁹

The MDSC shift to a highly immunosuppressive TAM or TAN phenotype is driven by tumor-associated molecular signals, such as damage-associated molecular patterns¹⁴⁰ and lipid ligands—such as phosphatidylserine,¹⁴¹ oxidized low-density lipoprotein (oxLDL)¹²⁶ and apolipoprotein E^49,142,143—as well as a variety of cytokines.¹⁴⁴ These signals can upregulate the STAT3, NF-κB, and C/EBPβ signaling pathways. STAT3 is the core regulatory pathway driving the expansion and activation of MDSCs, while NF-κB and C/EBPβ can promote the expression of immunosuppressive genes and form an interactive signaling pathway network. Taking the transformation of M-MDSCs to TAMs as an example, when these molecular signals bind to TREM2 on the surface of M-MDSCs, they activate the downstream PI3K/AKT/mTOR pathway and promote the nuclear translocation of the NF-κB transcription factor.¹⁴⁵ At the same time, the TREM2-dependent NF-κB signaling axis selectively promotes the transcription and secretion of a large number of cytokines with immunomodulatory functions, such as IL-10 and TGF-β.¹⁴⁶ These cells also stably coexpress osteopontin,¹⁴⁷ fibronectin 1¹⁴⁸ and complement C3.¹⁴⁹ In addition, HIF-1α in the TME acts as a synergistic driver in this continuum. HIF-1α not only directly binds to HREs to drive the transcriptional reprogramming of M-MDSCs into TAMs but also upregulates the expression of glycolytic enzymes and ARG1 to complete the metabolic and functional remodeling of cells during differentiation.¹⁵⁰

Furthermore, Notch signaling has been reported in some studies to be involved in MDSC expansion in tumors,¹⁵¹ and c-Rel is also involved in the regulation of MDSC characteristic gene expression.⁷⁵ The S100A8/A9 protein can act as an autophagy factor that promotes MDSC expansion.¹²² Researchers have found through scRNA-seq pathway enrichment analysis that the immunosuppressive, metabolic reprogramming, and migration/chemotaxis pathways of MDSCs are highly active. In the GSE210963 study, researchers found that after treatment with the BTK inhibitor ibrutinib, the expression of genes such as GBP1, IL1B, and CXCL8 in MDSCs was significantly downregulated, while RGS2 and ABHD5 were upregulated; concurrently, the TREM1, NO, and IL-6 signaling pathways were inhibited. In addition, scRNA-seq data further reveal significant regional heterogeneity in the gene expression profiles of MDSCs, such as the enhanced expression of HIF-1α-related immunosuppressive genes in the hypoxic tumor core compared with the tumor margin. Furthermore, epigenetic mechanisms further stabilize these transcriptional programs through processes such as chromatin remodeling and histone modification, thereby reinforcing the persistent immunosuppressive phenotype of MDSCs in the TME.

Taken together, these results suggest that MDSC signaling pathways and gene expression patterns shift to a proinflammatory and migratory state under tumor conditions, encompassing multiple functions such as immunosuppression, metabolism, and differentiation. These coordinated changes form the molecular basis for their immunosuppressive function in the TME.

MDSC chemotaxis and activation networks driven by multicellular interactions in the TME

The regulatory mechanisms of the TME do not play a single role but act on MDSCs through a variety of pathways to remodel the gene expression patterns and signaling pathways of MDSCs, forming a complementary signaling network within the cells. Therefore, a positive feedback loop is formed between the TME and MDSCs, which is interrelated, mutually reinforcing, and complementary, leading to the gradual formation of an immunosuppressive malignant TME. This is one of the driving forces that allows tumors to evade the surveillance of the body’s immune system and continue to develop.

The regulatory network of immunosuppression constructed by MDSCs is first initiated by multiple chemokines and cytokines secreted by the tumor, which enable myeloid cells to expand, become activated, and then be recruited to the TME. Subsequently, MDSCs and other MDSC-affected stromal cells, such as CAFs, secrete more chemokines, which further drive MDSCs from the peripheral circulation and bone marrow into the TME. The highly plastic MDSCs are then reshaped under acidic TME conditions, hypoxia, and nutrient deprivation. FAO and glycolysis are upregulated in MDSCs, and intermediates of these metabolic processes are used as raw materials for MDSC immunosuppressive molecules such as PGE2, which in turn enhances the stability of HIF-1α. PGE2, in turn, can enhance the stability of MDSCs under hypoxia. FAO also enhances mitochondrial ROS production and activates ER stress, leading to the upregulation of LOX-1, which in turn enhances lipid uptake. Tumor-derived lactate in the acidic TME enters MDSCs through MCT1, competitively inhibits PHDs, blocks the degradation of HIF-1α, and allows it to continuously accumulate. When HIF-1α enters the nucleus, forms a transcriptional complex, and binds to HRE sequences, it can simultaneously upregulate GLUT1 to enhance glucose uptake and LDHA to further accelerate lactate production. Therefore, lactate accumulation, FAO, and glycolysis are interconnected between MDSCs and the environment, forming a self-sustaining and self-amplifying positive feedback regulatory network.

These recruited and activated MDSCs effectively inhibit the activity of antitumor T cells and protect tumor cells from elimination by the immune system. By proliferating, tumor cells, in turn, further exacerbate the Warburg effect by producing more lactate. This newly produced lactate, in turn, acts within the TME to further recruit and activate more MDSCs. The interplay between tumor metabolism and MDSC biology establishes a robust, self-perpetuating positive feedback loop that creates a stable, progressively worsening state of immunosuppression.

MDSC-mediated immunosuppressive networks and consequent tumor immune escape

MDSCs suppress the immune response using a diverse array of mechanisms that often act synergistically to suppress the body’s antitumor immunity at multiple levels, creating an extensive and complex immunosuppressive network. MDSCs suppress both adaptive and innate immunity through multiple mechanisms, including metabolic competition for essential nutrients, direct cell–cell interactions, and the secretion of soluble mediators, thereby inhibiting the function of T cells, B cells, and NK cells.^8,152 MDSC-mediated immunosuppression is initiated at early stages of immune cell development and persists throughout immune cell trafficking and their subsequent infiltration into the TME. Collectively, these activities constrain antitumor immunity by limiting immune cell activation, inducing immune apoptosis, restricting their recruitment to tumor sites, and attenuating their cytotoxic functions (Fig. 5).

Fig. 5.

MDSC-mediated suppression of T-cell antitumor immunity. In T cells, MDSCs deplete essential amino acids, release reactive oxygen and nitrogen species, and engage in contact-dependent interactions through cell-surface molecules, thereby impairing T-cell survival and migration and inducing T-cell anergy and apoptosis while favoring regulatory T-cell (Treg) differentiation

Amino acid metabolic depletion: depriving the survival foundation of immune cells

MDSCs can deplete L-arginine from the microenvironment through the high expression of ARG1 and inducible nitric oxide synthase (iNOS or NOS2).⁵⁵ Under pathological conditions associated with cancer, tumor-derived cytokines can upregulate the amino acid transporter CAT-2B on the surface of MDSCs, promoting the rapid influx of extracellular L-arginine.¹⁵³ This L-arginine is subsequently hydrolyzed by ARG1 into urea and L-ornithine or oxidized by iNOS to produce nitric oxide (NO) and L-citrulline.^154,155 The normal development and activation of T lymphocytes are highly dependent on L-arginine. MDSCs’ consumption of L-arginine in the environment deprives T cells of essential nutrients, making it difficult for T cells to stably maintain the mRNA of intracellular regulators of the cell cycle, particularly the mRNA for the two proteins cyclin-dependent kinase 4 (CDK4) and cyclin D3. Furthermore, L-arginine deficiency simultaneously inhibits the phosphorylation of the retinoblastoma (Rb) protein and suppresses the E2F1 transcription factor within T cells, causing T cells to arrest in the G0–G1 phase of the cell cycle and rendering them unable to respond to antigenic stimulation or proliferate.^155–157 Furthermore, L-arginine deficiency severely impairs the transcription of the T-cell receptor (TCR) complex and compromises the synthesis, expression, and membrane stability of the TCR ζ chain (CD3ζ),¹⁵⁷ a key intracellular signaling subunit. The absence of CD3ζ disrupts the proximal signaling cascade required for T-cell activation, leading to functional impairment and a sharp decline in interleukin-2 (IL-2) production.¹⁵⁶

In addition to consuming L-arginine from the environment, MDSCs also deplete another essential amino acid: L-tryptophan. High expression of the rate-limiting enzyme indoleamine 2,3-dioxygenase 1 (IDO1) in MDSCs leads to the degradation of L-tryptophan via the kynurenine pathway.^158–161 High expression of IDO1 in MDSCs suppresses immune responses at multiple levels. First, L-tryptophan depletion directly triggers the GCN2 kinase stress response pathway in adjacent T cells, leading to translation arrest and increased T-cell sensitivity to apoptosis.¹⁶² Concurrently, kynurenine—a metabolite produced by MDSCs during L-tryptophan degradation—acts as an endogenous ligand for the aryl hydrocarbon receptor (AHR), and when released into the extracellular environment, it activates the AHR signaling pathway in CD4⁺ naive T cells and antigen-presenting cells (APCs), causing them to differentiate into Foxp3⁺ Tregs and IDO1-expressing tolerant DCs,^163,164 respectively, thereby establishing a positive feedback loop.

Although, from a mechanistic perspective, the catabolic pathways of L-arginine and L-tryptophan are two completely independent amino acid consumption pathways in MDSCs, further research has revealed profound interactions between these two mechanisms. These interactions enhance the immunosuppressive function of MDSCs, thereby maintaining long-term immune evasion of tumor cells. When ARG1 degrades L-arginine, it produces L-ornithine. L-Ornithine is subsequently decarboxylated by ornithine decarboxylase (ODC) to form polyamines, which are further converted into spermidine and spermine within MDSCs^154,165; spermidine is then released into the TME. When spermidine in the TME is taken up by DCs, it specifically activates Src kinase. Activated Src kinase phosphorylates IDO1, causing phosphorylation of IDO1 at its immune receptor tyrosine-inhibitory motifs (ITIMs).¹⁶⁶ As a multifunctional protein, IDO1 possesses both enzymatic L-tryptophan degradation activity and nonenzymatic intracellular signaling capabilities. Spermidine- and Src-mediated phosphorylation of IDO1 triggers an intracellular signaling cascade within DCs, activating the nonclassical NF-κB pathway. This signaling axis induces high expression of transforming growth factor-beta 1,^166–168 which in turn maintains high levels of IDO1 expression in MDSCs and sustains their immunosuppressive phenotype. These findings suggest that amino acid metabolism in MDSCs constitutes a highly integrated positive feedback signaling network.

Interestingly, amino acid metabolism in MDSCs is negatively regulated by NO produced by MDSCs; NO generated by iNOS can directly bind to the heme cofactor of the IDO1 enzyme, thereby spatially impairing its catalytic function. This phenomenon indicates the existence of a competitive mechanism within MDSCs, in which the balance between ARG1 and iNOS determines whether the inhibitory mechanism in the TME shifts toward IDO1-mediated tolerogenic immunosuppression or severe nitrosative stress (via NO and its derivatives).^122,169

In addition to L-arginine and L-tryptophan, MDSCs also compete with T cells for sulfur-containing amino acids in the environment, thereby further impairing T-cell function. Mature T cells exhibit low activity in cysteine-related metabolic pathways, have limited capacity to synthesize cysteine from methionine, and primarily rely on the supply of exogenous cystine. Under physiological conditions, T cells rely heavily on mature APCs (such as macrophages and DCs) to supply cysteine to sustain the metabolic demands of cellular activity. In the TME, MDSCs highly express the XCT transporter and extensively uptake extracellular cystine.¹⁷⁰ However, unlike mature APCs, MDSCs do not export reduced cysteine to T cells. Cystine deprivation not only limits T-cell activation but also directly inhibits the synthesis of glutathione (a major intracellular antioxidant). This results in a significant decline in T cells’ antioxidant capacity, making it difficult for them to survive in the highly oxidized TME and rendering them highly susceptible to reactive oxygen species (ROS) and reactive nitrogen species (RNS) secreted by MDSCs. MDSC-mediated cystine depletion acts synergistically with their immunosuppressive function, further enhancing the immunosuppressive capacity of MDSCs within the TME.

Oxidative and nitrosative stress: inducing immune cell damage and inactivation

In addition to depleting the amino acid pool, MDSCs actively create an oxidative and nitrosative microenvironment by producing ROS and RNS, thereby causing direct damage to nearby immune cells. While upregulating iNOS, MDSCs also significantly upregulate the expression of the multimeric NADPH oxidase (NOX2) complex. PMN-MDSCs primarily rely on the NOX2 complex to generate large amounts of ROS, particularly superoxide anions and hydrogen peroxide.¹⁷¹ High concentrations of extracellular ROS cause direct oxidative damage to lymphocytes in the TME, which can lead to apoptosis in severe cases. ROS also oxidize key signaling molecules within T cells, such as the tyrosine kinases Lck and ZAP-70. This inhibits their phosphorylation, disrupting downstream signaling and ultimately rendering T cells dysfunctional or unresponsive.^172,173 Concurrently, M-MDSCs generate large amounts of NO via iNOS.¹⁷⁴ When NO diffuses across the T-cell membrane, it directly inhibits the synthesis of IL-2 receptors on T cells.¹⁷⁵ Furthermore, NO can promote T-apoptosis by enhancing CD95 (Fas)-mediated apoptotic signaling.^176–178 When ROS from PMN-MDSCs and NO from M-MDSCs come into contact, they form peroxynitrite (ONOO⁻). Peroxynitrite drives profound nitration of key tyrosine residues on target proteins. In T cells, peroxynitrite specifically nitrates the TCR and CD8 coreceptor, preventing T cells from effectively recognizing peptide–MHC complexes presented by APCs and thereby suppressing antigen-specific immunity.^61,179 Furthermore, MDSC-derived peroxynitrite modifies chemokines in the extracellular matrix. Taking the chemokine CCL2 as an example, nitrosylated CCL2 exhibits a significantly reduced ability to recruit effector T cells, making it difficult for them to be recruited to the tumor core.¹⁸⁰ At the same time, the ability to recruit MDSCs is selectively preserved, which exacerbates immune suppression within the TME.

Contact-dependent suppression induces cellular inactivation by MDSCs

In addition to nutrient deprivation and oxidative stress mediated by short-range diffusion, MDSCs also interact directly with immune cells. By binding cell surface ligands to receptors on lymphocytes, they cause lymphocytes to lose their immunological function, disrupt their physiological homing patterns, and even transmit signals that induce lymphocyte apoptosis.

When MDSCs are exposed to hypoxia within the TME or stimulated by tumor-derived IFN-γ, PD-L1 expression is markedly upregulated via activation of the JAK–STAT1 signaling pathway; concurrently, hypoxia stabilizes HIF-1α, which cooperates with JAK–STAT1 signaling to further amplify PD-L1 expression on the cell surface.^181,182 Notably, although IFN-γ is canonically an antitumor cytokine, its action on MDSCs paradoxically induces high PD-L1 expression, thereby establishing a negative feedback loop that reinforces an immunosuppressive microenvironment.¹⁸³ When PD-L1 on MDSCs binds to the receptor PD-1 on the surface of T cells, it triggers the recruitment of the phosphatases SHP-1 and SHP-2 within the T cells. These phosphatases dephosphorylate proximal TCR signaling molecules, inhibiting T-cell proliferation and leading to deep functional exhaustion and inactivation of lymphocytes. This is also one of the primary mechanisms of tumor immune evasion.¹⁸⁴

Similarly, Fas ligand (FasL/CD95L), which is highly expressed on MDSCs, directly binds to the Fas (CD95) receptor expressed on the surface of CD8⁺ T cells when MDSCs are recruited to the TME or reside in peripheral lymphoid organs, thereby inducing T-cell apoptosis.^185,186 In a 2011 study, it was found that MDSCs expressing FasL in the TME significantly induced apoptosis of activated CD8+ T cells, which weakened the antitumor immune response.¹⁸⁷ In 2017, researchers confirmed through a melanoma model that PMN-MDSCs lead to rapid depletion of CD8+ T cells through the FasL-Fas pathway, which is closely related to the formation of immune tolerance during cancer treatment.¹⁸⁵

As antigen-presenting cells (APCs), M-MDSCs can express MHC class II molecules and present antigens to CD4+ T cells, but this interaction is fundamentally flawed. T-cell activation requires two signals to act simultaneously: signal 1, the binding of the TCR to the peptide–MHC complex, and signal 2, the costimulation from the binding of the T-cell’s CD28 receptor to ligands such as CD80/CD86 on APCs (e.g., DCs, macrophages).¹⁸⁸ However, M-MDSCs usually express these ligand molecules at very low levels or even lack the expression of these ligand molecules owing to their immaturity.¹⁸⁹ Thus, T cells enter an inactivated state when stimulated by costimulatory-deficient signals from MDSCs.¹⁵

MDSCs express the surface metalloproteinase a disintegrin and metalloproteinase 17 (ADAM17), which cleaves the extracellular domain of L-selectin (CD62L) expressed on the surface of circulating naive T cells in peripheral blood.¹⁹⁰ L-selectin plays a critical role in the initial adhesion and rolling of naive T cells through high-endothelial venules, a process that allows them to leave the bloodstream and enter lymph nodes where antigen sensitization occurs.¹⁹¹ By cleaving L-selectin, MDSCs physically prevent naive T cells from reaching lymphoid tissues, thereby hindering their contact with resident APCs. This mechanism significantly weakens the initial activation process of the adaptive immune system, thereby suppressing the initiation of the initial antitumour immune response.¹⁹²

Immunosuppressive mechanisms targeting B-cell-mediated humoral immunity

Humoral immune responses are also suppressed by MDSCs, while MDSCs inhibit T cells, these inhibitory mechanisms also correspondingly suppress B-cell function. First, at the metabolic level, similar to the suppression of T cells, MDSCs overexpress ARG1, which consumes L-arginine, thereby inhibiting B-cell proliferation. MDSCs produce ROS and NO, which interfere with the activation of BCR downstream signaling pathways (such as SYK, BTK, and PI3K), thereby blocking B-cell proliferation and differentiation and suppressing B-cell antibody secretion.^193,194 Second, MDSCs secrete immunosuppressive factors such as IL-10 and TGF-β, which similarly diminish B-cell antibody secretion capacity and induce the generation of immunoregulatory B cells (Bregs) with immunosuppressive functions. Bregs can suppress Th1 responses; more importantly, Bregs themselves also release IL-10, which further weakens B-cell immune efficacy, leading to the production of more Bregs.¹⁹⁵ Furthermore, MDSCs exert their inhibitory effects through direct cell‒cell contact. B cells also express receptors such as PD-1 and Fas on their surfaces. When PD-L1 on the surface of MDSCs binds to PD-1 on B cells, it blocks B-cell activation¹⁹⁶; simultaneously, FasL on the surface of MDSCs binds to Fas on B cells, directly inducing apoptosis.¹⁹⁷

Suppressive mechanisms targeting the antitumor function of NK cells

As the primary effector cells of the innate immune system, NK cells are capable of recognizing and killing tumor cells. The inhibitory effects of MDSCs are mediated through a combination of mechanisms, including downregulation of activation receptors, suppression of cytotoxic functions, and evasion of NK cell recognition pathways. This disrupts the body’s innate immune barrier and drives tumor immune evasion.

First, MDSCs secrete NO and TGF-β to downregulate the expression of key activation receptors on the surface of NK cells, particularly NKG2D and NKp30.^198,199 NKG2D is essential for recognizing ligands such as MHC class I chain-related protein A and MHC class I chain-related protein B, which are highly expressed on the surface of tumor cells in cancers such as PDAC and non-small cell lung cancer (NSCLC).^200,201 This results in the innate immune system’s inability to effectively recognize malignant tumors. Furthermore, MDSCs suppress NK cells through multiple combined mechanisms. For example, the binding of PD-L1 on their cell surface to PD-1 induces phosphorylation of ITIM and ITSM and recruits SHP-2 phosphatase, thereby inhibiting the PI3K–AKT and RAS–ERK signaling pathways.¹⁸⁴ Both synergistically inhibit mTOR, leading to reduced translation of cytotoxic molecules such as granzyme B and perforin.²⁰² Second, TGF-β secreted by MDSCs binds to the TGF-β receptor complex and represses the expression of key transcription factors that regulate NK cell effector function, such as T-bet and Eomes.²⁰³ The suppression of these master regulators impairs NK cells through two parallel pathways. On a functional level, it leads to a decreased response to activation signals such as interleukin-12 (IL-12) and interleukin-15 (IL-15), impaired degranulation, and reduced overall cytokine secretion.²⁰⁴ Concurrently, at the transcriptional level, the repression of T-bet and Eomes directly downregulates the expression of core effector genes, including PRF1 (encoding perforin), GZMB (encoding granzyme B, the key protein for caspase-dependent and caspase-independent target apoptosis), and IFNG (encoding IFN-γ).^205,206 Together, these dual mechanisms ultimately lead to a greatly reduced NK cell killing effect. Concurrently, IL-10 secreted by MDSCs binds to the interleukin-10 receptor on NK cells, thereby activating the JAK1/STAT3 signaling pathway and inducing the expression of negative regulatory molecules such as SOCS3, which downregulates the expression of cytotoxicity-related genes (e.g., IFN-γ).²⁰⁷ As a key cytokine for NK cell cytotoxic activity, IFN-γ inhibits angiogenesis, induces tumor cell apoptosis or growth arrest, and promotes T-cell recognition and elimination of tumor cells. It also induces macrophages, DCs, and NK cells to produce cytokines such as TNF-α, which synergistically enhance tumor-killing capacity. Furthermore, IL-10 reduces cellular metabolic activity by inhibiting the mTOR and glycolytic pathways, thereby suppressing the synthesis of cytotoxic molecules in NK cells at the metabolic level.²⁰⁸ These three signaling pathways exhibit significant functional synergy: on the one hand, they jointly inhibit mTOR-driven protein synthesis and energy metabolism; on the other hand, they block NK cell activation, granule formation, and degranulation at multiple levels, including signal transduction, transcriptional regulation, and cytoskeletal dynamics. This results in downregulated expression of NK cell cytotoxic molecules, impaired formation of immune synapses, and significantly reduced degranulation capacity, thereby inhibiting their ability to kill tumor cells.

NK cells kill target cells that downregulate the expression of MHC class I molecules. While low or even negative expression of MHC class I molecules is a key feature of MDSCs, this low expression applies to classical antigen-presenting MHC class I molecules (HLA-A/B/C). MDSCs overexpress nonclassical immunosuppressive MHC class I molecules (HLA-E and HLA-G). These nonclassical molecules bind to inhibitory receptors on NK cells (e.g., the CD94/NKG2A heterodimer), thus helping MDSCs escape recognition and killing by NK cells.⁶ This is a complementary strategy to the low expression of classical MHC-I, which results in MDSCs being neither recognized by T cells nor killed by NK cells, thus forming a synergistic mechanism of immune escape and creating a local NK cell-tolerant zone in the TME.

In summary, the immunosuppressive effects of MDSCs on NK cells do not rely on a single, isolated mechanism but rather on the synergistic action of multiple mechanisms, forming a complex network of inhibitory signals that plays a key role in the antitumor immune response (Fig. 6).

Fig. 6.

MDSC-mediated suppression of NK-cell and B-cell antitumor immunity. In NK cells, immunosuppressive cytokines released by MDSCs act synergistically to suppress NK-cell cytotoxicity. In parallel, TGF-β signaling suppresses PI3K–AKT activity, thereby impairing immunological synapse formation and microtubule-organizing center (MTOC) polarization. MDSCs also evade NK-cell recognition by downregulating activating receptors and classical HLA-I molecules while upregulating nonclassical inhibitory molecules. In B cells, contact-dependent interactions mediated by MDSC surface molecules impair B-cell proliferation and differentiation and promote apoptosis, while acting synergistically with immunosuppressive cytokine signaling to induce regulatory B-cell (Breg) polarization

Nonimmune mechanisms in MDSCs driven tumor progression and metastasis

In addition to immunosuppression, MDSCs also promote tumor progression through nonimmunological mechanisms, including the promotion of tumor neovascularization, the promotion of cancer cell invasion and metastasis, and the establishment of a premetastatic microenvironment in distal organs.^3,55 The nonimmunological protumorigenic functions of MDSCs and their immunosuppressive effects are clearly divided and synergistic, creating a positive feedback loop between maintaining the TME in a malignant state and promoting tumor growth and metastasis. The interconnectedness of these two tumor-promoting mechanisms of MDSCs requires that the clinical treatment of tumors not only focus on individual tumor cell components but also use a systemic and holistic view, adopting a therapeutic approach that targets the entire tumor system (Fig. 7).

Fig. 7.

MDSCs promote tumor progression through multiple nonimmunosuppressive mechanisms. a Promotion of angiogenesis. MDSCs promote neovascularization, degrade the perivascular extracellular matrix (ECM), destabilize existing vessels, recruit endothelial progenitor cells (EPCs), and protect newly formed vessels through immunosuppression. b Promotion of epithelial-mesenchymal transition (EMT). MDSCs induce invasive mesenchymal-like phenotypes that enhance tumor cell invasion and intravasation. c MDSC-mediated protection of circulating tumor cells (CTCs). MDSC–tumor cell clusters protect CTCs from immune clearance and buffer shear stress during hematogenous metastasis. d Promotion of lymphatic remodeling and dissemination. MDSCs induce inflammatory lymphatic endothelial cell (LEC) transformation and integrin-dependent retention, upregulate LEC adhesion molecules, and induce tumor-derived SAA1, thereby disrupting lymphatic endothelial junctions and promoting adhesion and dissemination. e Pre-metastatic niche priming. MDSCs suppress local immune surveillance, promote angiogenesis, and degrade the ECM in distant organs. f Reciprocal interactions with the TME. Interactions among MDSCs, tumor cells, and cancer-associated fibroblasts (CAFs) create an immunosuppressive environment, promote CAF differentiation, and enhance MDSC recruitment, expansion, and activation through chemokines, cytokines, and lactate. Hypoxia-induced metabolic reprogramming further enhances MDSC immunosuppression

Regulation of tumor angiogenesis and vascular structural abnormalities

Tumor growth requires a blood supply through angiogenesis, in which MDSCs are key factors influencing the formation of tumor neovascularization.^209,210 MDSCs are able to promote angiogenesis and shift blood vessels to a phenotype that is favorable for tumor development.

MDSCs secrete a variety of proangiogenic factors to promote tumor neovascularization, and VEGF plays a major role in these cytokines. The hypoxic environment in the TME leads to the upregulation of HIF-1α expression in MDSCs, which drives them to secrete VEGF.²⁰⁹ When VEGF binds to vascular endothelial growth factor receptor 2 on vascular endothelial cells, it induces vascular endothelial cell proliferation, migration, and formation of new tubular structures through the PI3K/AKT and MAPK/ERK pathways. It can also enhance the permeability of the blood vessels themselves, thereby facilitating hematogenous metastasis.^211,212 In addition to VEGF, MDSCs can release other signaling molecules to enhance VEGF signaling. Bv8/prokineticin 2 (PROK2) released by MDSCs acts synergistically with VEGF on vascular endothelial cells by activating their prokineticin receptors (PKR1/2).^213,214 Other key mediators released by MDSCs that assist in promoting angiogenesis include angiopoietin-2 (ANG-2) and placental growth factor (PlGF). ANG-2 can destabilize existing blood vessels and facilitate vascular remodeling in the presence of proangiogenic signals,^209,215 and PlGF may recruit bone marrow-derived endothelial progenitor cells or act via vascular endothelial growth factor receptor 1 signaling to assist neovascularization.^213,216

In addition to VEGF, immunosuppressive cytokines secreted by MDSCs, such as TGF-β and IL-10, also participate in tumor neovascularization. TGF-β, beyond its immunosuppressive roles, acts as an inducer of endothelial-to-mesenchymal transition through Smad signaling, thereby modulating endothelial cell proliferation, migration, and phenotypic plasticity.^217,218 Concurrently, IL-10 helps establish an immunosuppressive microenvironment that mitigates immune detection and clearance of nascent vessels, thereby indirectly facilitating angiogenesis.¹³⁵

MDSCs can produce basic fibroblast growth factor (bFGF/FGF2) under inflammatory or tumor microenvironmental stimuli.^219,220 bFGF binding to endothelial cell FGFRs may activate downstream signaling including Ras–MAPK (and possibly PI3K/AKT), driving endothelial proliferation and migration.^221,222 Moreover, FGF signaling is known to influence vascular smooth muscle cells and can enhance CAF proliferation or activation, which potentially contributes to stabilization or remodeling of perivascular structures.²²³

In addition to secreting cytokines that directly promote angiogenesis, MDSCs release other key cytokines that act synergistically with angiogenic factors to amplify angiogenic signals and collectively promote the formation of structurally and functionally aberrant tumor vascular systems. MDSCs assist in the promotion of angiogenesis through the release of matrix metalloproteinase-9 (MMP-9), which is essential for the development of the tumor vasculature.²²⁴ MDSCs assist in promoting angiogenesis by releasing MMP-9, which degrades the extracellular matrix (ECM) and vascular basement membrane components, thereby removing physical barriers to vascular endothelial cell migration and by releasing growth factors (including VEGF) that are sequestered in the matrix during the degradation of the ECM.²²⁵

Angiogenic factors produced by MDSCs, such as VEGF, can also act as chemoattractants, recruiting more MDSCs from the circulatory system and bone marrow to the tumor site. These newly recruited MDSCs further enhance the production of proangiogenic signaling molecules.²²⁶

In addition, tumor neovascularization differs significantly from normal blood vessels in that neovascularization in tumors usually exhibits a disorganized morphology and the size of the vascular lumen varies, and this structure leads to hypoperfused regions in the tumor vascular system that are unable to deliver oxygen efficiently.²²⁷ This exacerbates hypoxia in the TME, which further enhances HIF-1α expression in MDSCs, thereby releasing more VEGF into the TME, thus creating a positive feedback loop that continues to exacerbate the malignant nature of the TME and tumor. Moreover, the enhancement of HIF-1α expression also enhances the expression of immunosuppressive molecules such as PD-L1, ARG1, and iNOS in MDSCs, which strengthens their immunosuppressive ability.²²⁶

Driving lymphatic vessel remodeling and tumor lymphatic metastasis

In addition to immunosuppression and the promotion of tumor neoangiogenesis, MDSCs can enhance tumor cell metastasis by promoting tumor-associated lymphangiogenesis. Metastasis of tumor cells to regional lymph nodes is an early sign of tumor spread in the patient’s body.²²⁸

When tumor metastasis occurs, lymphatic endothelial cells (LECs) express chemokines and various adhesion molecules, which recruit and capture specific metastatic tumor cells and assist tumor metastasis through the lymphatic system. At the same time, tumor cells secrete vascular endothelial growth factor-C (VEGF-C), which binds to vascular endothelial growth factor receptor 3 on LECs, thereby inducing the proliferation of LECs as well as their inflammatory transformation.²²⁹

Inflamed LECs secrete chemokines that recruit MDSCs, such as CXCL1, CXCL2, CXCL5, and CXCL8, and upregulate the expression of adhesion molecules such as intercellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion molecule 1 (VCAM-1).^180,230 These chemokines recruit MDSCs from the circulatory system to the corresponding inflamed lymphatic vessels and lymph nodes. Subsequently, CD11a and CD11b on MDSCs bind to the adhesion molecule ICAM-1, and CD49d binds to VCAM-1, thereby prolonging the residence time of MDSCs in the lymph and promoting their accumulation.¹⁸⁰

When MDSCs are recruited to inflamed lymphatic sites, they not only inhibit immune cells in the lymphatic vessels to exert their immunosuppressive effects but also assist tumor cells in remodeling the structure of lymphatic vessels, thereby increasing the permeability of lymphatic vessels, allowing tumor cells to break through the physical barrier of lymphatic channels, and facilitating the metastasis of tumors.²³¹ MDSCs induce tumor cells to secrete serum amyloid A1 (SAA1), which triggers the phosphorylation of tyrosine residues of vascular endothelial cadherin (VE-cadherin) in nearby LECs in a paracrine manner.^232,233 This disrupts the integrity of the adherens junctions between LECs, causing the lymphatic vessels in the vicinity of the tumor to become “loose and porous,” thus significantly increasing the permeability of the lymphatic vessel walls. This allows larger molecular weight substances to enter the lymphatic system from the tissue fluid, making it easier for tumor cells to invade the lymphatic system.²³⁴

In addition to contributing to tumor neovascularization, inflammatory cytokines such as IL-1β secreted by MDSCs induce LECs to express E-selectin, which binds to the glycosylated ligands of tumor cells, slowing down the movement of the tumor cells in the lymphatic vessels and enhancing the ability of the tumor cells to adhere to and stay in the lymphatic vessels, thus contributing to the establishment of tumor micrometastases.^235,236

Mechanisms of tumor cell stemness maintenance and enhancement

Apart from being able to construct the environment required for tumors, MDSCs also modulate the intrinsic biology of tumor cells to support tumor progression. Cancer stem cells (CSCs) are a subpopulation of tumor cells that possess self-renewal and differentiation abilities, which are key factors leading to tumor metastasis and recurrence. MDSCs can exacerbate the malignancy of tumors by paracrine secretion of signaling molecules that enhance the stemness of cancer cells. These signaling molecules mainly include IL-6, IL-1β, and TGF-β. IL-6 can activate JAK/STAT3 in the tumor cell pathway, thereby inducing the expression of CSC core stemness transcription factors (e.g., SOX2, NANOG, OCT4), which enable tumor cells to maintain their self-renewal capacity and remain undifferentiated. A typical example is in a breast cancer model, where MDSCs can activate the JAK/STAT3 signaling pathway via IL-6, inducing tumor cells to acquire a CSC phenotype and enhancing the formation of tumorspheres.²³⁷ NO released by MDSCs also activates the Notch pathway, which contributes to and maintains the sustained activation of the IL-6-mediated JAK/STAT3 pathway.²³⁷ In contrast, IL-1β upregulates aldehyde dehydrogenase 1 (ALDH1) expression via the NF-κB pathway, thereby enhancing drug tolerance and metabolic adaptation in CSCs.²³⁸ In addition, the NF-κB pathway synergizes with STAT3 to jointly promote the expression of NANOG.

TGF-β secreted by MDSCs binds to transforming growth factor-beta receptor (TGF-βR) on tumor cells to activate the SMAD2/3-SMAD4 complex and synergistically enhances the activation of the Notch pathway, which can synergistically interact with the JAK/STAT3 signaling pathway to enhance and maintain STAT3 activation, which further maintains and enhances the stemness state of tumor cells. There is significant synergy between these three pathways, and the cascade effect of these pathways enhances the stemness state of tumor cells and increases the difficulty of drug therapy.

In addition to affecting tumor cell signaling pathways by secreting signaling molecules, MDSCs also shape the local oxidative environment to provide a favorable TME for the survival and stability of the CSC subpopulation. NO and ROS produced by MDSCs not only inhibit T-cell activity but also trigger protective antioxidant pathways in cancer cells, such as the Nrf2-antioxidant response element (ARE) pathway, which helps CSCs survive better in hypoxic TME.^171,239,240 By inducing and shaping the oxidative TME through cytokine signaling, MDSCs create an environment that consistently produces CSCs and favors their survival. These CSCs themselves are highly resistant to conventional chemotherapy and radiotherapy, while MDSCs inhibit immune cells capable of destroying CSCs, which exacerbates the difficulty of treating tumors clinically. Thus, the ability of MDSCs to induce CSC formation and protect their survival makes them a high priority target for therapeutic intervention.

Induction of epithelial–mesenchymal transition (EMT) and extracellular matrix remodeling

The prerequisite for tumor cells to undergo metastasis is to acquire the ability to migrate and invade and to break through the ECM limitation.²⁴¹ MDSCs help tumors to better survive and metastasize by inducing EMT and remodeling the ECM.

EMT is a major paracrine process of migration and a process by which tumor cells change from an epithelial phenotype to a fully or partially mesenchymal phenotype, in which cells lose polarity, weaken intercellular adhesion, and acquire an invasive mesenchymal phenotype. EMT facilitates the invasion and migration of tumor cells across the ECM and the vascular system.²⁴²

MDSCs can promote EMT by secreting a variety of signaling factors, including TGF-β, IL-6, and TNF-α.²⁴³ TGF-β binds to the TGF-β receptor on the membrane of tumor cells and activates the phosphorylation of the Smad2/3 protein. It then binds to Smad4 and enters the nucleus, which upregulates the expression of EMT-related transcription factors such as Snail, ZEB1, and Slug.²⁴⁴ This downregulates the expression of E-cadherin, leading to the disruption of the adhesion structure between tumor cells, and upregulates the expression of N-cadherin and vimentin, which enhances the migration ability and mesenchymal phenotypic characteristics of tumor cells.²⁴¹

IL-6 secreted by MDSCs activates intracellular JAK signaling by binding to the receptor complex interleukin-6 receptor alpha-glycoprotein 130 on tumor cells. This induces STAT3 phosphorylation; phosphorylated STAT3 then enters the nucleus, activating the expression of EMT-associated target genes, including Twist1, ZEB1 and Snail.^{237,245–247} This process also causes downregulation of E-cadherin expression and upregulation of mesenchymal-related markers, resulting in tumor cells with greater plasticity and invasiveness.²⁴⁸ In addition, MDSCs activate the NF-κB pathway in tumor cells through the binding of released TNF-α to TNFR.^249,250 Although activation of the NF-κB pathway does not directly induce EMT, activation of this pathway enhances the persistence of the IL-6 and TGF-β pathways, thereby maintaining EMT.⁸⁰ Some studies have found that in colorectal cancer cells, the NF-κB pathway can be inhibited by inhibiting TNF-α, which in turn inhibits EMT.²⁵¹ In triple-negative breast cancer, researchers have found that TNF-α also enhances TGF-β-induced phosphorylation of Smad2 and upregulates the expression of Snail, which in turn enhances the process of EMT.^252,253 In addition, TNF-α can also build a pro-migratory, immunosuppressive microenvironment by promoting the formation of an inflammatory environment that allows EMT to continue to operate in a state of chronic activation.²⁵⁴

The ECM is a complex structure composed of multiple proteins, including structural proteins (collagen), proteoglycans and glycoproteins (hyaluronic acid) and cell adhesion proteins (fibronectin), which together form the scaffolding of the TME.²⁵⁵ MDSCs can secrete enzymes that hydrolyze these proteins, the most critical of which are matrix metalloproteinase-2 (MMP-2) and MMP-9. These enzymes can hydrolyze ECM, promoting EMT and facilitating tumor progression and metastasis.^225,256,257

Immune shielding and physical protection afforded to circulating tumor cells (CTCs)

The process of hematogenous metastasis is dangerous for CTCs, which must survive hemodynamic shear forces and escape destruction by immune cells, especially NK cells.²⁵⁸ MDSCs provide protection for CTCs during hematogenous metastasis by forming heterotypic cell clusters with CTCs, which are mediated by adhesion molecules, such as VCAM-1, on the MDSCs and corresponding ligands on the CTCs.²⁵⁹ MDSCs form cell clusters with CTCs that are more stable than single, isolated CTCs during hematogenous transit because MDSCs physically buffer CTCs from destructive intravascular shear forces.^260,261 MDSCs also provide an immunoprotective barrier to CTCs by forming a local immunosuppressive microenvironment. Thus, CTCs are protected from circulating NK cells and cytotoxic T lymphocytes. As a result, such cell clusters formed by the combination of MDSCs and CTCs are able to more efficiently arrest in the microvasculature of distal organs and enter the metastatic microenvironment. Thus, they are an efficient vector for tumor dissemination.

Construction of premetastatic niches in distant organs

During metastasis, tumors not only need to leave the primary site but also need to create a suitable environment for tumor colonization in distal organs. MDSCs arrive at the future metastatic site of the tumor before the arrival of CTCs, forming a premetastatic niche. The tumor mobilizes MDSCs from the bone marrow to the circulatory system by releasing factors such as G-CSF and VEGF, which then reach the distal organ site.²⁵⁶ At this site, MDSCs exert their immunosuppressive capacity by depleting local L-arginine through the upregulation of ARG1 and iNOS, while inhibitory cytokines such as IL-10 and TGF-β are secreted to suppress immunosurveillance within the premetastatic microenvironment to ensure that all CTCs arriving here are protected from immune cell attack. Second, MDSCs induce tumor-associated angiogenesis and remodeling in the region to facilitate the arrival of CTCs. MDSCs also release MMPs to degrade the local ECM and secrete VEGF to increase vascular permeability, which makes it easier for CTCs to extravasate from the vasculature into the organ. The VEGF secreted by MDSCs also causes more MDSCs to be mobilized to the premetastatic niche, thus effectively enhancing the formation of the premetastatic microenvironment.^225,259,262

Synergistic mechanism between MDSCs and CAFs

In the TME, the function of MDSCs is not only regulated by intracellular signals but also involves a complex interaction network with other tumor-associated cells. CAFs are the key cells by which MDSCs construct an immunosuppressive network in the TME. These interactions are driven by tumor metabolites, such as lactate, which jointly shape the highly immunosuppressive malignant microenvironment. Tumor cells produce large amounts of lactate through the Warburg effect, which not only acts directly on MDSCs but also induces the differentiation of a variety of precursor cells—such as fibroblasts, mesenchymal stem cells (MSCs), resting stellate cells, endothelial cells, epithelial cells, and adipocytes—into CAF phenotypes.

Mechanistically, lactate activates and maintains high expression of two key signaling axes, HIF-1α and NF-κB, in CAFs. The synergy of these two pathways enables CAFs to maintain the release of a large number of cytokines, growth factors, and chemokines, even in the absence of actual hypoxia or exogenous inflammatory cytokine stimulation, thereby recruiting MDSCs to the tumor site. HIF-1α induces the expression of chemokines such as CXCL12 and CCL2 to achieve the targeted recruitment of M-MDSCs, while NF-κB acts as an inflammatory signal amplifier to drive the continuous secretion of IL-6, CXCL1, CXCL5, and CXCL8. IL-6 promotes the expansion of MDSCs and the maintenance of immunosuppressive function by activating STAT3 signaling, thereby enhancing the recruitment of PMN-MDSCs and the intensity of local inflammation.

A classic example is in the esophageal squamous cell carcinoma model, in which bone marrow-derived MSCs cocultured with tumor cells differentiate into fibroblast activation protein-expressing CAF-like cells, which subsequently secrete a large number of MDSC-recruiting chemokines. This suggests that adjusting the therapeutic strategy to target lactate levels during treatment can indirectly interfere with MDSCs and impair their tumor-promoting effects by blocking CAF formation. It is worth noting that tumor cells themselves, driven by lactate in the environment, also activate the HIF-1α and NF-κB pathways and secrete similar chemokines (such as CXCL1/2/5), which complement CAFs and make MDSC recruitment to the tumor site more stable.

In the process of cancer treatment, even if a certain treatment can inhibit the secretion of chemokines by CAFs, tumor cells may compensate and supplement the gap in chemokines caused by the drug, thereby maintaining the continuous recruitment of MDSCs. At the same time, the redundancy of chemokine and chemokine receptor pairing further reduces the efficacy of drugs targeting chemokines or their receptors for tumor treatment. This indirect mechanism of MDSC recruitment by affecting the secretion of other cytokines complements the direct recruitment effect of lactate on MDSCs and together constitutes the complete mechanism of lactate-driven MDSC recruitment in the TME. The key role of lactate in constructing an immunosuppressive microenvironment rich in MDSCs by orchestrating multicellular interactions within the TME is highlighted. At the same time, multiple mechanisms of MDSC recruitment complement each other, highlighting the highly integrated and adaptive ecosystem created by tumors.

Dynamic feedback and self-expansion of MDSCs within the TME

The multiple immune and nonimmune protumor mechanisms of MDSCs do not operate in isolation but are interconnected with the TME and form synergistic networks with each other. When tumor cells recruit MDSCs to the TME, the MDSCs shape an immunosuppressed and structurally abnormal TME, which in turn reshapes the MDSCs. Hypoxic, acidic, and nutrient-deficient conditions within the TME upregulate signaling pathways in MDSCs that are favorable for tumor progression.

Hypoxia is one of the potent regulators of MDSCs in the environment, and the hypoxic environment created by the TME drives high HIF-1α expression in MDSCs, which directly induces PD-L1 expression, resulting in a substantial increase in the ability of MDSCs to induce T-cell depletion.¹¹⁸ In addition, hypoxia can induce the metabolic reprogramming of MDSCs, which enhances not only glycolysis but also lipid uptake and FAO activity. This promotes their dependence on FAO for energy supply, thereby maintaining their immunosuppressive function and reflecting a high degree of metabolic plasticity. In turn, FAO provides metabolic fuel for maintaining high activity of ARG1 and iNOS. This metabolic shift further enhances the immunosuppressive capacity of MDSCs.¹³⁹

This also explains the differences in MDSC function in infiltrating tumors and in peripheral lymphoid organs (e.g., the spleen), where splenic MDSC immunosuppressive function is dependent on the production of ROS, whereas MDSCs in the TME significantly upregulate the cellular expression of ARG1, iNOS, and PD-L1.¹⁵

It was noted that in peripheral lymphoid organs, MDSCs are mainly composed of PMN-MDSCs with milder immunosuppressive mechanisms, and the differentiation of M-MDSCs to macrophages and DCs is inhibited in these tissues.

In contrast, M-MDSCs predominate within the TME, where their conventional differentiation into mature myeloid cells, such as macrophages and DCs, is more strongly constrained. However, they can undergo rapid reprogramming toward TAMs. Consistent with this compartment-specific regulation, inhibition of STAT3 reduces MDSC accumulation in the spleen of tumor-bearing mice but has limited effects within tumors.

These facts suggest that the TME remodels MDSCs into more inhibitory and tumor-promoting pathological cells by actively shaping them, and vice versa. This interaction suggests that tumor therapeutic strategies targeting the TME also act indirectly on MDSCs, which broadens our search for therapeutic agents for MDSCs (Fig. 8).

Fig. 8.

The cytokine network secreted by MDSCs promotes cancer stemness, EMT, angiogenesis, and lymphatic remodeling. MDSC-derived inflammatory mediator signaling enhances cancer cell stemness and antioxidant capacity through the Nrf2, Notch, JAK–STAT3, and NF-κB signaling pathways. Meanwhile, JAK–STAT3, together with the SMAD pathway, induces EMT in tumor cells. With the assistance of Bv8, PlGF, and bFGF, VEGF activates the RAS–RAF–ERK and PI3K–AKT–mTOR signaling pathways in endothelial cells to promote angiogenesis. When VEGF-C binds to its receptor, it drives lymphatic vessel remodeling. These pathways show significant synergistic interactions within the same cells in the TME, mutually maintaining and amplifying one another

MDSCs as a clinical biomarker

An important application of MDSCs in tumor therapy is as a clinical biomarker.²⁵ MDSC accumulation is not only a result of cancer progression but also a key cause of tumor immune escape and treatment resistance.^6,15 Therefore, the evaluation of MDSC levels in patients can help to make prognostic judgments during the treatment process and predict tumor progression with real-time monitoring of drug treatment effects^4,263 (Table 3).

Table 3.

Prognostic and predictive value of MDSCs in various cancers

Cancer type	MDSC subset analyzed	Clinical setting	Key finding (correlation with OS, PFS, or ICI response)	Reference
Melanoma	M-MDSC, PMN-MDSC	Advanced/metastatic (ICI therapy)	High baseline M-MDSC levels predict poor response and shorter OS/PFS to anti-PD-1/CTLA-4. On-treatment decrease predicts favorable outcome.	274
NSCLC	Total MDSC, M-MDSC	Advanced/metastatic (ICI therapy)	High baseline M-MDSC levels are associated with shorter PFS and OS in patients receiving ICIs.	350
Breast cancer	Total MDSC, M-MDSC	Advanced/metastatic breast cancer	High circulating MDSC levels are associated with greater metastatic burden, disease progression, and worse OS.	268,351,352
Pancreatic cancer	PMN-MDSC	Radiotherapy	Radiation-induced PMN-MDSC infiltration is associated with reduced treatment response and poor prognosis.	353
Renal cell carcinoma (RCC)	Total MDSC	Metastatic (tyrosine kinase inhibitor (TKI) therapy)	High MDSC levels are associated with resistance to TKI therapy (e.g., sunitinib) and poor prognosis.	295

Prognostic value of MDSCs in cancer patients

A large body of clinical evidence has clearly established that elevated levels of circulating MDSCs in patients are a negative prognostic factor in a wide range of malignant tumors.²⁶⁴ High levels of MDSCs, especially the M-MDSC subgroup, are closely associated with adverse clinical features such as advanced cancer stage during treatment, tumor progression and the immunosuppressive status of the body.^265–267 According to a meta-analysis that pooled data from several solid tumor studies, this correlation was confirmed to have a direct impact on patient survival. This analysis showed that patients with higher levels of circulating MDSCs had a poorer prognosis, with a hazard ratio (HR) of 1.796 for overall survival (OS) and an HR of 2.459 for progression-free survival (PFS), confirming a close correlation between the level of MDSCs and the survival status of patients.²⁶⁸

Predictive value for ICI treatment

MDSCs, as one of the key mechanisms leading to drug resistance in tumor ICI therapy (e.g., anti-PD-1 and anti-PD-L1 antibodies), can be used to predict the therapeutic ability of ICIs.²⁶⁹ MDSCs directly disrupt the conditions under which ICIs exert their therapeutic efficacy by establishing an immunosuppressive TME.^88,270 The mechanism of action of an ICI is to act directly on T cells and activate their immune-killing ability; i.e., ICI treatment of tumors requires the presence of T-cell infiltration in the environment. Therefore, in an MDSC-rich TME, even with ICI treatment, there are still not enough T cells capable of exerting a tumor-clearing effect, leading to therapeutic failure.^271,272

In a meta-analysis of 17 studies with a total of 1035 patients, patients with low circulating MDSC levels showed significant improvement in OS (HR = 2.13) and PFS (HR = 1.87) after ICI treatment compared to patients with high MDSC levels. In the subgroup analysis, the HR for OS was 2.45, and that for PFS was 2.03.²⁷³ This suggests that the baseline MDSC level is a significant negative predictive biomarker. Therefore, by monitoring the level of MDSCs in patients, it is possible to screen out those patients who will not benefit from ICI therapy. These patients may require alternative or combination strategies to overcome this myeloid-driven resistance.

MDSCs can also be used as pharmacodynamic markers

MDSC levels in patients not only provide prognostic and predictive information but are also capable of serving as real-time pharmacodynamic biomarkers. For example, in a study of patients with metastatic melanoma treated with ICIs, it was found that M-MDSC levels in some patients, which were high prior to treatment, decreased rapidly during therapy. The long-term OS and PFS exhibited by such patients whose MDSCs decreased rapidly from high levels in their bodies were comparable to, or even better than, those of patients with tumors that had low levels of M-MDSCs to begin with.^274,275

This fact suggests that the absolute baseline value of MDSCs in a patient is less important than the trajectory of its response to therapy.²⁷⁵ An effective therapeutic strategy disintegrates the regulatory network of immunosuppression mediated by MDSCs. The speed and magnitude of this disintegration can be measured by the metrics of MDSC decline. Changes in the number of MDSCs can therefore be used as an early indicator that a drug is effective.^269,276 This has far-reaching clinical implications, and the quantification of MDSCs serves as a dynamic tool for real-time monitoring of efficacy during treatment, allowing earlier communication of patient information. A decrease in circulating M-MDSC levels in response to a drug can provide initial evidence that the antitumor drug has a therapeutic effect. In contrast, the presence of elevated or persistently high levels of M-MDSCs during treatment can serve as an early warning that the tumor is developing acquired or primary drug resistance. Thus, it suggests that the physician needs to reformulate the treatment strategy.

Cancer treatment strategies by targeting MDSCs

On the basis of the critical role that MDSCs play in promoting tumor progression, they are an important therapeutic target in tumor therapy.^25,277,278 Researchers are currently developing therapeutic strategies against MDSCs based on four directions: depletion of MDSCs, inhibition of their functions, blocking of their recruitment or promotion of their differentiation into mature cells without inhibitory functions.^4,279,280 However, these therapeutic strategies often face many problems to be overcome in the translation from laboratory mouse tumor models to mature clinical treatment protocols.² Some drugs that have good efficacy in cell models or animal experiments have limited therapeutic effects on patients in the actual clinic and even have serious toxic side effects.²⁸¹

Therefore, it is crucial to guide future drug development and clinical research by discussing the clinical trial results, combination therapy capabilities, and limitations in the clinical use of therapeutic drugs for MDSCs. By unraveling the biology of MDSCs, it is possible to target them to overcome the dilemmas that arise with current therapeutic strategies. This chapter will examine some of the therapeutic strategies targeting MDSCs and their clinical translation and discuss the reasons leading to therapeutic failures, aiming to inform the future development of drugs with more precise targeting effects and research into more effective therapeutic measures.

Therapeutic strategies and clinical translation of MDSCs

To date, a variety of drugs targeting MDSCs have been investigated. Some of these drugs have shown good promise, while others have failed to demonstrate good clinical efficacy. This highlights the complexity of treating tumors by targeting MDSCs in cancer patients. A discussion of these clinical results will help to reveal the key influencing factors that lead to the success or failure of clinical trials (Table 4).

Table 4.

Clinical translation of key MDSC-targeting agents

Drug/agent	Target/mechanism on MDSCs	Cancer type(s)	Trial phase	NCT number(s)	Key reported outcome(s)	Limitations	Reference
All-trans retinoic Acid (ATRA)	Promotes differentiation into mature DCs, macrophages, and granulocytes; reduces ROS production.	Melanoma, NSCLC, Renal Cancer	I/II	NCT03200847	In combination with pembrolizumab for advanced melanoma, ATRA was well-tolerated, reduced circulating PMN-MDSCs, and enhanced T-cell activity. Reported objective response rate (ORR) was 71% with a 5-year OS rate of 54.7%.	Early-phase, single-arm trial (n = 24); efficacy requires validation in larger, randomized studies.	354,355
Low-Dose Chemotherapy (e.g., Gemcitabine, 5-FU)	Direct depletion/selective killing of proliferating MDSCs.	Pancreatic Cancer, Breast Cancer	III (Adjuvant/Advanced)	NCT00003216	In resected pancreatic cancer (RTOG-9704), adjuvant gemcitabine improved median survival over 5-FU.	Nonspecific to MDSCs. Significant on-target, off-tumor toxicity, primarily myelosuppression, limits dosing and duration. MDSC depletion is a secondary finding, not a primary endpoint.	356
Sunitinib (TKI)	Multitarget TKI; reduces peripheral MDSCs and reverses MDSC-associated T-cell suppression in RCC patients.	mRCC	III (Approved)	NCT00083889	First-line sunitinib improved PFS and ORR, and was associated with longer OS, compared with IFN-α in mRCC; separate translational studies showed reduced peripheral MDSCs.	Assessment of OS was confounded by crossover and subsequent lines of therapy, with treatment toxicity and acquired resistance remaining major limitations.	295,357,358
PDE5 Inhibitors (e.g., Sildenafil, Tadalafil)	Inhibit immunosuppressive function by reducing ARG1 and iNOS expression.	Head & Neck Cancer (HNSCC)	I/II	NCT00894413	In a phase II trial in HNSCC, tadalafil reduced peripheral MDSCs, decreased ARG1/iNOS expression, and augmented general and tumor-specific T-cell responses.	Early-phase data; not designed to evaluate clinical efficacy endpoints like survival. Immune effects may be influenced by patient BMI.	359
MMP-9 Inhibitors (e.g., Marimastat)	Inhibit MDSC migration by blocking extracellular matrix degradation.	Breast Cancer, Pancreatic Cancer	III (Development Terminated)	NCT00003010	The ECOG E2196 trial in metastatic breast cancer found marimastat did not improve PFS versus placebo.	Broad-spectrum MMP inhibition led to debilitating musculoskeletal toxicity, a major dose-limiting factor. The drug failed to show clinical benefit, leading to termination of its development for cancer.	360
COX-2 Inhibitors (e.g., Celecoxib)	Inhibit MDSC expansion and function by blocking PGE2 production.	NSCLC	III	NCT00104767	Multiple phase III trials found that adding celecoxib to standard chemotherapy for advanced NSCLC did not improve PFS or OS.	Failed to demonstrate clinical benefit despite a strong biological rationale. COX-2 expression by IHC was not a predictive biomarker of response.	283

These clinical cases reveal a common problem that arises in the transition of MDSC therapeutics from experimentation to the clinic, i.e., targeted drugs developed on the basis of the biology of MDSCs often do not work well when they enter clinical trials.^2,4 Although the drugs may successfully act on the targets on MDSCs or block the relevant pathways, there is no guarantee that they will enhance the patient survival rate and other indicators in the clinic.

A typical example is the COX-2 inhibitor celecoxib. From a theoretical point of view, inhibition of COX-2 reduces the production of PGE₂. This in turn inhibits the differentiation of MDSCs and their secretion of iNOS and ROS, ultimately weakening the suppression of antitumour immunity by MDSCs.²⁸² However, celecoxib did not significantly improve survival in patients with advanced NSCLC in several phase III clinical trials.^283–285 This result suggests a high redundancy of tumor-promoting and tumor cell signaling pathways in the TME.²⁸⁶ Even when one immunosuppressive signaling pathway is inhibited by a drug, MDSCs can continue to maintain immunosuppressive capacity by upregulating other compensatory pathways that can play a similar role. A second example is the broad-spectrum MMP inhibitor Marimastat, which is associated with musculoskeletal toxicity despite its ability to inhibit MDSC migration.^287–290 This example illustrates the need for drugs to target MDSCs while also minimizing the impact on normal tissues.

Combination of MDSC-targeted drugs and immune checkpoint inhibitors

Based on the complexity of the tumor ecosystem, therapeutic strategies targeting MDSCs alone are seldom applied in the clinic but are instead combined with other drugs, especially immune checkpoint inhibitors (ICIs), to enhance their therapeutic effects.^4,6 This is because the immunosuppressive effects of MDSCs—such as inducing apoptosis in T cells by directly binding to PD-1 on T cells via PD-L1 or indirectly suppressing immune responses by depleting key metabolites essential for T-cell activation, such as arginine—are one of the main causes of primary and acquired resistance to ICIs.^291–293 Therefore, the effectiveness of ICI therapy can be enhanced by depleting MDSCs or inhibiting their function. This therapeutic strategy has entered clinical trials, and several clinical trials have demonstrated favorable results. This further reveals the superiority of drug combination therapy over single-targeted MDSC treatment strategies²⁹⁴ (Table 5).

Table 5.

Key clinical trials of MDSC-targeting agents in combination with ICIs

MDSC-targeting agent	ICI partner	Cancer type(s)	Trial phase	NCT number	Key reported outcome(s)	Limitations	Reference
ATRA	Pembrolizumab (anti-PD-1)	Melanoma	I/II	NCT03200847	Showed a 71% ORR and 5-year OS of 54.7% in anti-PD-1-naive metastatic melanoma. The combination was well-tolerated and associated with reduced circulating PMN-MDSCs and enhanced T-cell function.	Small (n = 24), single-arm study. Efficacy requires confirmation in a randomized Phase III trial.	354
Entinostat (HDAC Inhibitor)	Pembrolizumab (anti-PD-1)	NSCLC, Melanoma	I/II	NCT02437136	In NSCLC patients who progressed on prior anti-PD-1/L1 therapy (ENCORE 601 cohort), the combination showed a modest ORR of 9.2% and a median OS of 11.7 months. Benefit was enriched in patients with high baseline classical monocytes.	Primary endpoint of ORR was not met. Grade ≥3 AEs occurred in 58% of patients.	361
CXCR1/2 Inhibitor (SX-682)	Pembrolizumab (anti-PD-1)	Melanoma	I	NCT03161431	In metastatic melanoma patients refractory to prior anti-PD-1 therapy, the 200 mg dose cohort had an ORR of 21% and a DCR of 63%. The combination showed tolerable safety and meaningful disease control in heavily pretreated patients.	Early-phase trial; efficacy data are preliminary. Potential for additive immune-related toxicities.	362
Arginase Inhibitor (INCB001158)	Pembrolizumab (anti-PD-1)	Solid Tumors	I/II	NCT02903914	Combination was generally well-tolerated and showed pharmacodynamic evidence of arginase inhibition. However, ORRs were low and did not exceed historical rates for pembrolizumab monotherapy in most cohorts. The highest ORR was 19.2% in anti-PD-1-naive HNSCC.	Limited clinical activity despite target engagement, suggesting L-arginine depletion alone is insufficient to overcome resistance in most tumors.	363
Axitinib (TKI)	Pembrolizumab (anti-PD-1)	RCC	III (Approved)	NCT02853331	KEYNOTE-426 demonstrated superior OS and PFS versus sunitinib in first-line advanced RCC, leading to FDA approval. The combination reduces tumor angiogenesis and MDSCs, creating an inflamed TME.	Higher rate of Grade ≥3 AEs compared to sunitinib, including hypertension (22%) and hepatotoxicity (13% ALT increase).	364

Not all combination therapies have been successful; for example, the combination of an arginase inhibitor with pembrolizumab has not achieved favorable therapeutic outcomes. This suggests that combination therapy is not simply a matter of using MDSC-targeted drugs together with ICIs but that the key lies in choosing drugs that target different, nonoverlapping, and interdependent signaling pathways to complement each other’s strengths and weaknesses. For example, the success of combining axitinib with pembrolizumab lies in the action on two key pathways: axitinib’s inhibition of VEGF-driven angiogenesis and VEGF release from MDSCs and pembrolizumab’s inhibition of MDSC-induced T-cell apoptosis. This leads to broader and deeper remodeling of the entire TME, which is at the core of the success of this combination therapy.

In the treatment of the HDAC inhibitor entinostat in combination with pembrolizumab, although it showed some efficacy, it was accompanied by adverse events, which suggests that we also need to pay attention to the balance between efficacy and safety in the selection of drugs for combination therapy.

Therefore, the key to the success of drug combination therapy lies in understanding the complexity of the tumor ecosystem. The direction for the selection of future therapeutic strategies should be to choose drugs that can simultaneously block multiple key tumor signaling nodes and act on the target cells more precisely to achieve synergistic efficacy and maximize the antitumor effect.

Tumors resist strategies targeting MDSCs through multiple pathways

Another reason for the failure of clinical drug treatment strategies is the acquisition of drug resistance by tumors. This development of drug resistance is an evolutionary choice made by the tumor and the cells in the TME in response to the pressure exerted by the drug. In therapeutic strategies targeting MDSCs, tumors and MDSCs alter themselves in various ways, such as upregulation of compensatory signaling pathways, immune editing, metabolic reprogramming and epigenetic alterations. These theories provide a plausible explanation for acquired resistance in tumors in the course of drug therapy.

Compensatory signaling bypass activation with compartmentalization leads to the development of drug resistance

In addition to their ability to promote tumor progression, MDSCs are also involved in the development of drug resistance in tumors. MDSCs can inhibit the action of drugs in the absence of genetic changes in tumor cells, thereby reducing the therapeutic ability of the drug to treat the tumor. A typical example is the development of resistance to the multitargeted TKI sunitinib in RCC.

Sunitinib carries out antitumour therapy by modulating immune effects, and its core mechanism is the inhibition of STAT3, which plays a key driving role in the expansion and immunosuppressive function of MDSCs.^295,296 However, during tumor treatment, acquired drug-resistant tumor cells and associated stromal cells synergistically upregulate the expression of GM-CSF, which activates a parallel survival pathway in MDSCs that is insensitive to sunitinib, namely, the JAK2-STAT5 axis.^{277,297–299} The activation of STAT5 is sufficient to compensate for the lack of STAT3 signaling due to sunitinib, thereby promoting MDSC survival and proliferation and maintaining their immunosuppressive function.^300,301

This phenomenon also raises another issue, namely, the phenomenon of compartmentalization faced during treatment. Although sunitinib successfully depletes STAT3-dependent MDSCs in the peripheral blood,^295,296 tumor cells and associated stromal cells in the TME maintain the survival and immunosuppressive function of MDSCs in the TME through the secretion of GM-CSF, which activates a STAT5-dependent compensatory pathway.^298,300 This results in the fact that although various indicators, such as the number of circulating MDSCs in the periphery, are reduced under drug treatment, the drug fails to eliminate MDSC-induced immunosuppression at the tumor site.³⁰¹

Therefore, MDSC-targeted drugs need not only to monitor the number of MDSCs in the peripheral blood but also to assess the level of local immunosuppression at the tumor site during the development process. In this way, the therapeutic effect of a drug can be truly assessed, avoiding the result of treating the symptoms but not the root cause due to misleading indicators.

Tumor-intrinsic resistance and immunoediting

Even when a targeted drug against MDSCs acts precisely on its target and successfully lifts the immunosuppressive state, the tumor itself can evolve to resist the pressure of the body’s immune system and continue to deteriorate.³⁰² This process is called “immunoediting” and involves three stages in tumor development: elimination, equilibrium, and escape.³⁰³

The elimination phase begins when MDSC-targeting drugs successfully exert their therapeutic effects, at which time the immunosuppressive state of the body caused by MDSCs is lifted, resulting in the elimination of most tumor cells by the body’s immune system. However, a few tumor cells can still survive, and then the tumor will enter the equilibrium phase of elimination and proliferation. At this point, tumor proliferation is still under the control of the immune system.^304,305 These tumor cells may have downregulated the expression of MHC I molecules or acquired mutations in genes essential for antigen processing and presentation, such as β-2-microglobulin or the antigen processing-associated transporter associated with antigen processing 1.^306–309 Loss or downregulation of MHC-I is a common mechanism by which tumors evade immune surveillance and has been associated with both intrinsic and acquired resistance to immunotherapy in a variety of cancers, thereby making them less likely to be recognized by T cells.^310,311

Finally, malignant cells possessing mutated genes occupy the vast majority of the tumor ecological niche, and the tumor enters the escape phase. This escape is mediated by two types of tumor-intrinsic mechanisms: mutation of drug-related targets, which prevents the binding of the drug to the therapeutic target, and activation of oncogenic bypass pathways, which reduces the effect of the original targeting pathway on the cells after it is blocked by the drug, thus maintaining the proliferation of the tumor cells.^312–315 Tumor immune escape reveals that when an effective combination therapy strategy (e.g., MDSC inhibitor + ICI) is applied to a tumor, it leads to dynamic changes in the tumor ecology from two aspects: on the one hand, remodeling the TME to reestablish its immunosuppressive environment, and on the other hand, screening the tumor cells so that those that are able to resist the attack of T cells or are insensitive to the drug effect are retained.^316–319

This suggests that a successful therapeutic strategy will also need to overcome alterations at the genetic level of both the TME and the tumor. This informs the design of the next generation of antitumor therapeutic strategies: how to design therapeutic strategies that can address both the tumor ecosystem and the intrinsic plasticity of tumor cells.

Metabolic and epigenetic reprogramming

Tumor resistance is not just a transient upregulation of a signaling bypass but also evolves this shift into a heritable phenotype through epigenetic reprogramming. For example, the metabolism of MDSCs in the TME undergoes alterations, such as a shift from being glycolysis-dependent to FAO-dependent, which significantly enhances MDSC survival in the hypoxic and nutrient-poor TME. At the same time, lactate released into the TME by the tumor due to the Warburg effect also upregulates the expression of ARG1 and iNOS in MDSCs, which actively promotes the formation of an immunosuppressive environment.

These metabolic shifts are consolidated epigenetically in MDSCs, where they maintain their immunosuppressive phenotype through epigenetic modifications under the constant pressure of the TME. EZH2 (enhancer of zeste homolog 2) is a key epigenetic regulator of gene expression in MDSCs.³²⁰ Researchers found that when its activity was inhibited with GSK126, the number of MDSCs in patients increased, while the number of CD4⁺ and IFN-γ⁺ CD8⁺ T cells decreased.³²¹

Epigenetics explains why once drug resistance is established in a tumor, resistance persists even after a period of drug discontinuation. Under drug pressure, tumor cells adjust the expression of relevant pathways for better survival. However, with continued drug use, tumor cells turn to more stable solutions to gain a survival advantage, i.e., by epigenetically locking in the expression of genes in pathways that favor tumor survival.^322,323 Epigenetic inheritance also poses higher requirements for drug development; drugs need to not only block signaling pathways in therapy but also possess the ability to block epigenetic inheritance in tumor cells.^324,325

New generation of MDSC-targeted therapies

With the growing recognition of the central role of MDSCs in promoting tumor progression by creating an immunosuppressive TME, overcoming their protumorigenic effects has become the next major challenge and opportunity in the field of oncology. Therapeutic strategies against MDSCs have evolved from simple cell removal to re-engineering the TME, altering cellular metabolism, and more precisely targeted cellular therapies. This section focuses on the new generation of antitumor therapeutic strategies currently under development that are not limited to eliminating MDSCs but rather to modulating the complex immune networks of their environment (Table 6).

Table 6.

Next-generation strategies to overcome MDSC-mediated therapeutic resistance

Strategy	Specific approach/agent	Mechanism of action	Developmental status/key finding	Reference
Armored CAR-T engineering	IL-12 secretion	CAR-T cells locally produce IL-12 to counteract TME suppression and inhibit MDSCs.	Preclinical; enhanced CAR-T persistence and antitumor efficacy.	365
	Dominant-negative TGF-βRII	CAR-T cells express a truncated TGF-β receptor, rendering them resistant to suppressive TGF-β.	Preclinical; increased CAR-T proliferation and function in TGF-β-rich models.	326
Direct MDSC targeting	CD33-CAR-T	CAR-T cells are engineered to recognize and kill MDSCs expressing the surface antigen CD33.	Early Clinical (NCT02958397); effectively depleted MDSCs preclinically.	366
	TR2-CAR-T	CAR-T cells coexpress a chimeric receptor that induces apoptosis in TR2-expressing MDSCs.	Preclinical; restored cytotoxic activity of tumor-targeting CAR-T cells in breast cancer models.	327
Metabolic targeting	FAO Inhibition (e.g., Etomoxir)	Selectively inhibits FAO, a key energy source for MDSCs in the TME.	Preclinical; cripples MDSC suppressive capacity. Clinical use limited by off-target toxicity on mitochondrial complex I.	113
Bispecific engagers	CD33/CD3 BiTE (e.g., AMG 330, AMV564)	Redirects cytotoxic T cells to kill CD33-expressing monocytic MDSCs.	Preclinical/Early Clinical (NCT03144245); demonstrated MDSC depletion and anti-leukemic activity.	333
Engineered oncolytic viruses	OV suppressing GM-CSF	Oncolytic adenovirus designed to suppress GM-CSF production in the TME.	Preclinical; reduced MDSC accumulation and enhanced PD-L1 blockade efficacy in pancreatic cancer models.	336

Overcoming resistance to chimeric antigen receptor T-cell (CAR-T) therapy in solid tumors

CAR-T cell therapies have shown promising results in hematologic-related tumors, but their therapeutic efficacy in solid tumors has not yet been clarified.^326,327 The major obstacle to the role of CAR-T therapies in solid tumors is the poor TME, of which MDSCs are one of the key components.³²⁸ During CAR-T therapy, the immunosuppressive barrier formed by MDSCs inhibits the transport, proliferation, and activation of CAR-T cells.^328,329

To eliminate the obstruction of CAR-T cells by the immunosuppressive barrier constituted by MDSCs, researchers are investigating several innovative strategies. One is the development of CAR-T cells capable of directly recognizing and killing MDSCs, which take advantage of specific antigens expressed on the surface of MDSCs, such as CD33 or tumor necrosis factor-related apoptosis-inducing ligand receptor 2 (TR2).^327,330 CAR-T cells targeting TR2 can express a chimeric receptor that induces apoptosis in MDSCs upon binding to TR2.³²⁷ This approach not only removes MDSCs from the TME but also further enhances the overall antitumour immune response.³²⁷

Targeting the weak link in MDSC metabolism

As the metabolic pressure exerted by the harsh environment of the TME is capable of remodeling the immunosuppressive function of MDSCs, another therapeutic strategy is to switch from targeting the MDSC signaling pathway to selectively targeting the weak links in the internal metabolism of MDSCs.

In the hypoxic and nutrient-deprived TME, MDSCs undergo a metabolic shift from glycolysis to the use of FAO for ATP production, thus maintaining high enzymatic activity of ARG1 and iNOS to sustain their immunosuppressive functions.^5,113 T cells, on the other hand, utilize glycolysis more to provide energy to support their rapid proliferation and immune function.³³¹ This difference in cellular metabolism provides a rationale for the elimination of MDSCs: FAO inhibitors such as etomoxir can be utilized to block the main energy source of MDSCs, thus weakening their immunosuppressive capacity without affecting normal immune cells.^113,332

Novel therapeutic modalities for MDSC clearance and reprogramming

In addition to the above therapies, other therapeutic strategies for MDSCs with innovative and promising approaches to precisely target MDSCs in the complex TME are being developed:

Bispecific T-cell Engagers (BiTEs): BiTEs are antibody fragments engineered to connect two different cells. Against MDSCs, a BiTE can be engineered that binds to the CD3 complex on T cells at one end and specific surface antigens (e.g., CD33) on MDSCs at the other end, thus accurately directing T cells to kill MDSCs.^330,333 BiTE exerts a localization function that does not depend on TCR recognition and antigen presentation by MHC; it can overcome impaired T-cell recognition owing to defective antigens on the surface of MDSCs.³³⁴

Engineered oncolytic viruses (OVs): Ovs are genetically modified viruses that can selectively infect tumor cells. Therefore, an OV that reduces or remodels MDSCs can be designed to improve the TME.³³⁵ An OV can also be designed to inhibit tumor secretion of cytokines that promote the accumulation and activation of MDSCs, such as GM-CSF, thereby reducing the number of MDSCs to enhance the effects of other immunotherapies.³³⁶ Since OVs can act precisely on tumor cells, they can inhibit tumor progression while substantially reducing cytotoxicity to normal cells. OVs can be used to inhibit tumor progression while minimizing systemic toxicity by acting on tumor cells and precisely modulating the TME.

Conclusions and perspectives

MDSCs are a population of immature myeloid cells formed under conditions of persistent pathological inflammation and characterized by high plasticity and heterogeneity. Their identification, classification, and nomenclature remain an evolving field. The functions of MDSCs are markedly dependent on the surrounding environment. In autoimmune diseases, MDSCs may exhibit immunosuppressive or proinflammatory states depending on the disease type and local environment.

Under tumor conditions, large numbers of immature myeloid cells are generated in the bone marrow and spleen. These cells undergo expansion and activation under the influence of tumor-derived cytokines, become immunosuppressive MDSCs, and are selectively recruited to the TME by chemokine signals. After reaching the TME, MDSCs may alter their gene expression patterns through metabolic reprogramming and preferentially differentiate toward TAM-like and TAN-like states. MDSCs have complex biological characteristics and promote cancer progression through a dual strategy. They suppress immunity through multiple mechanisms, including amino acid depletion, oxidative and nitrosative stress, the release of soluble inhibitory mediators, and contact-dependent inhibition mediated by cell-surface molecules. Through these mechanisms, MDSCs inhibit the functions of T cells, B cells, and NK cells and thereby promote tumor progression. MDSCs are also involved in multiple nonimmune protumor processes, including the promotion of tumor-associated angiogenesis and lymphatic remodeling, induction of epithelial–mesenchymal transition, enhancement of tumor cell stemness, promotion of metastatic capacity and protection of CTCs, and establishment of premetastatic microenvironments in distant organs. These immune and nonimmune protumor mechanisms do not act independently but are interconnected and cooperate with the TME. They amplify and enhance one another, forming a complex tumor-promoting network.

The deep involvement of MDSCs in tumor progression makes them key therapeutic targets in the clinical treatment of cancer. Based on the characteristics of MDSCs, researchers currently develop MDSC-targeted cancer therapies mainly in four directions: depleting MDSCs, inhibiting their functions, blocking their recruitment, or promoting their differentiation into mature cells without suppressive functions. However, because of the complexity of the tumor ecosystem, strategies targeting MDSCs are rarely used alone in clinical practice. Instead, they are combined with other drugs, particularly ICIs, to improve therapeutic efficacy. In clinical treatment, the baseline levels of MDSCs in patients may provide prognostic information and predict therapeutic responses, whereas changes in their numbers after treatment can be used as pharmacodynamic biomarkers.

Despite the progress made in MDSC research and the development of MDSC-targeted drugs, many challenges remain. MDSCs do not always exhibit immunosuppressive properties, particularly in some autoimmune diseases, which has raised questions about the term “myeloid-derived suppressor cells.” Moreover, no single specific marker for MDSCs has yet been identified. This creates difficulties in studying their biological mechanisms and developing MDSC-targeted drugs, and some drugs targeting MDSCs may also cause toxic effects on other normal cells. Because different MDSC subsets have different functions, therapeutic strategies directed against one MDSC subset may show limited efficacy in clinical applications. In addition, because of the redundancy of chemokine and cytokine signals, even when a drug inhibits MDSC recruitment or an immunosuppressive signaling molecule, MDSCs may continue to maintain their immunosuppressive capacity by upregulating other compensatory pathways. Although combinations of multiple drugs may achieve better effects, tumors and MDSCs can acquire drug resistance through immune editing, metabolic reprogramming, and epigenetic changes, greatly limiting therapeutic efficacy. Furthermore, differences in MDSC subset composition between the TME and other tissues result in compartmentalization during MDSC-targeted treatment. A decrease in the number of MDSCs in peripheral blood does not necessarily indicate that local immunosuppression within the tumor has been relieved, further complicating the use of MDSCs as predictive and prognostic biomarkers.

To address these obstacles, more precisely targeted drugs need to be developed, and additional therapeutic strategies need to be explored. The application of scRNA-seq has, to some extent, overcome the limitations of traditional classification based on surface markers and has allowed MDSCs to be viewed as branches along the continuum of myeloid cell differentiation. The focus of MDSC research should therefore shift from simply distinguishing their subsets toward an integrated view that combines their lineage origin, phenotype, and functional state, which may have more practical significance for cancer treatment. This shift suggests that drug development should target the functions and states of MDSCs rather than their poorly defined cellular characteristics. With the development of transcriptomics, emerging functional markers such as VISTA have been identified, providing candidate targets for MDSC-targeted therapy. The clarification of key signaling mechanisms also enables the development of drugs targeting major regulatory nodes in downstream signaling pathways, which may partly overcome the redundancy of chemokine and cytokine signals. The success of combination therapy depends on a deeper understanding of the complex cooperation between MDSCs and the TME and on selecting drugs that can simultaneously block multiple key tumor-promoting signaling nodes to achieve better synergy and maximize antitumor effects. During treatment, a prolonged tug-of-war with the tumor should be avoided as much as possible because it may lead to the development of drug resistance. The effects of MDSC-targeted drugs should be evaluated not only by monitoring the number of MDSCs in peripheral blood but also by assessing the local level of immunosuppression at tumor sites to monitor treatment progress more accurately. With the application of new technologies, MDSC-targeted strategies have expanded to include redesigning the TME, altering cellular metabolism, and developing more precise targeted cell therapies. In recent years, novel therapeutic platforms, including CAR-T cells, BiTEs, and engineered OVs, have also provided new options for MDSC-targeted therapy. Future cancer treatment strategies should combine more precise MDSC identification, functional state assessment, local and systemic monitoring, and rational drug combinations to improve therapeutic outcomes in patients.

Author contributions

R.J.D. and Y.Y.L. wrote and conceived the manuscript. X.Y.B., J.Y., W.T.Y. and Q.Q.L. collected the literature. Y.X. and Y.X.Y. designed the figures. R.J.D., Y.Y.L., A.P.L. and X.J.H. supervised and revised the manuscript. All authors read and approved the article.

Competing interests

The authors declare no competing interests.