The chemokine environment in CRC guides both MDSCs and Tregs toward the tumor (18). Tumor cells, stromal cells, and immune cells release CCL2, CCL3, CCL4, CCL5, CXCL8, and other chemokines that create gradients for cell movement (19). MDSCs strengthen this process. They produce CCL3, CCL4, and CCL5, which attract CCR5⁺ Tregs and increase their buildup in the tumor (20). MDSC-derived S100A8/A9 also helps keep MDSCs coming in and affects how other immune cells move (21). Tregs add to this loop. They do not directly recruit MDSCs, but they secrete CCL22 to support CCR4⁺ Treg buildup and release IL-10 and TGF-β, which increase CXCL8-driven recruitment of CXCR2⁺ PMN-MDSCs (22, 23).

These chemokine signals work together to form a positive-feedback loop. CCR5 ligands from MDSCs and chemokines from Tregs maintain constant recruitment (24). In vivo, higher levels of CCL4/CCL5 lead to more Treg infiltration, and CCR5 loss reduces Treg buildup and weakens immunosuppression (24). In short, chemokine-mediated recruitment is the first step in forming the MDSC–Treg axis and sets up close spatial contact between the two cell types (20).

Once they are located in the same area, MDSCs and Tregs strengthen each other through soluble factors and direct contact. This cooperation helps maintain long-term immune tolerance in CRC (25). MDSCs strongly promote Treg expansion. High levels of ARG1 and iNOS change the metabolic environment, limit effector T-cell growth, and push CD4⁺ T cells to become FoxP3⁺ Tregs (26, 27). MDSCs also produce TGF-β and IL-10, which are essential for peripheral Treg induction (14). Engagement of PD-1 by PD-L1 on MDSCs further dampens T-cell activation and promotes tolerance (28). Although M-MDSCs retain partial antigen-presenting capacity, insufficient co-stimulation results in low-reactivity or regulatory T-cell outcomes (29). Consistently, tissue studies demonstrate a strong correlation between MDSC infiltration and intratumoral FoxP3⁺ Treg abundance (30, 31).

Tregs, in turn, sustain and enhance MDSC suppressive activity. Their secretion of IL-10, TGF-β, and IL-35 activates key pathways in MDSCs and increases ARG1 and iNOS expression, strengthening their ability to suppress effector T cells (32, 33). Through CTLA-4-mediated interactions with dendritic cells, Tregs reduce co-stimulatory signaling and induce IDO, indirectly creating a tolerogenic microenvironment favorable to both cell types (34). In addition to these cytokine- and DC-dependent mechanisms, CD40–CD40L engagement provides a critical membrane-bound licensing signal: MDSCs in CRC exhibit high CD40 expression, and interaction with CD40L on activated Tregs drives NF-κB–STAT3 activation and upregulates IL-10, TGF-β, ARG1, and iNOS, thereby amplifying both Treg induction and MDSC suppressive function. Blockade of CD40 or CD40L has been shown to disrupt this reciprocal reinforcement and restore antitumor T-cell activity in multiple tumor models (25, 35).

Beyond cytokine signaling and metabolic suppression, direct contact further stabilizes this crosstalk. β₂ integrins mediate physical interactions between MDSCs and Tregs, enhancing their suppressive output and facilitating immune evasion (6). Transcriptomic analyses show that this suppressive circuit appears early in CRC development and can already be found in nearby mucosal tissues. This means the MDSC–Treg axis begins to change local immune balance at the very early stages of tumor formation (36, 37). These mechanisms work together to form a strong positive-feedback system that keeps chronic immunosuppression in the CRC microenvironment.

Tregs in the CRC TME display marked heterogeneity and phenotypic plasticity, which amplify their contribution to immunosuppression beyond a uniform suppressive role (38). Human FOXP3⁺ Tregs subdivide into naïve-like central Tregs, with the latter showing stronger suppressive activity and predominant enrichment in CRC tissues (39). Furthermore, tumor-infiltrating Tregs frequently co-express helper lineage-specific transcription factors, generating “Th-like” hybrids such as RORγt⁺ Th17-like Tregs or T-bet⁺ Th1-like Tregs (40). This adaptability enables Tregs to respond to inflammatory signals in the TME, stabilizing their suppressive phenotype while retaining high FOXP3 levels.

A prominent manifestation of Treg plasticity in CRC is the generation of ex-Th17 Tregs, derived from Th17 precursors that acquire a FOXP3⁺ suppressive profile (41, 42). These cells, including IL-17A⁺FOXP3⁺ intermediates or IL-17A⁻FOXP3⁺ derivatives, preserve some proinflammatory traits but primarily mediate immunosuppression, facilitating tumor progression (43). Such transdifferentiation occurs in TGF-β-rich microenvironments, where Th17 cells—originally polarized by IL-6 and IL-23—downregulate RORγt and induce FOXP3 expression (44).

Th17 cells exhibit a dual, often protumorigenic influence in CRC pathogenesis and frequently intersect with the MDSC–Treg axis (45). Within the CRC tumor microenvironment, Th17-derived IL-17A promotes the recruitment and activation of immunosuppressive populations, particularly PMN-MDSCs via CXCL1/CXCL2 or IL-8–CXCR2 signaling, thereby reinforcing immune evasion (46). During CRC progression, Th17/Treg balance is progressively skewed toward Treg dominance through plasticity-driven reprogramming, establishing a self-sustaining suppressive circuit (47). Notably, γδT17 cells—one of the major sources of IL-17A in human CRC—further potentiate this axis by driving PMN-MDSC mobilization and expansion (46).

Collectively, these dynamics highlight that the MDSC–Treg axis integrates Th17-associated plasticity and inflammatory pathways, adding multifaceted mechanisms of immune escape in CRC, as schematically illustrated in Figure 1.

Tumor- and stromal-derived factors (CSF-1, IL-6, TGF-β) drive M-MDSC differentiation into M2-like TAMs (CD206⁺, ARG1⁺, IL-10⁺). Bidirectional chemokine loops (CCL2/3/4/5 from tumor/MDSCs to CCR5⁺ Tregs; CCL22, IL-10/TGF-β from Tregs enhancing CXCL8 for CXCR2⁺ PMN-MDSCs; S100A8/A9 self-recruitment) establish spatial colocalization. MDSCs promote FoxP3⁺ Treg expansion via ARG1/iNOS metabolic depletion, TGF-β/IL-10, and PD-L1/PD-1 signaling, while inhibiting DC maturation (↓MHC II). Tregs enhance MDSC suppression through IL-10/TGF-β/IL-35 (↑ARG1/iNOS), CTLA-4–DC–IDO axis, CD40–CD40L–NF-κB/STAT3, and β₂ integrin contacts. Th17 plasticity generates ex-Th17 Tregs (TGF-β-driven RORγt↓/FoxP3↑ from IL-6/IL-23-polarized precursors), with IL-17A recruiting PMN-MDSCs (CXCL1/2/8) and promoting angiogenesis. MDSCs provide rapid, non-specific suppression (ARG1/iNOS/ROS), complemented by Treg antigen-specific, plastic tolerance, forming a multilayered network in CRC.

Cytokines and coinhibitory receptors form the main signaling loop that links MDSCs and Tregs. IL-10, TGF-β, PD-L1, and CTLA-4 are the key factors that keep immune tolerance in place (48).

As described in detail in Section 2.2, IL-10 and TGF-β serve as central bidirectional mediators of the MDSC–Treg axis. IL-10 enhances MDSC suppressive function via the JAK–STAT3 pathway, upregulating ARG1 and iNOS while impairing dendritic cell antigen presentation (26, 27, 49), and simultaneously promotes peripheral Treg induction (50, 51).

TGF-β keeps FoxP3 expression active through SMAD2/3 and reduces the cytotoxic activity of CD8⁺ T cells and NK cells (52, 53). Blocking TGF-β can improve antitumor immunity, but systemic inhibition may cause strong inflammatory side effects (54).

Contact-dependent signals also strengthen the suppressive axis. PD-L1 on MDSCs limits effector T-cell activation through PD-1 and helps Treg expansion grow (55). CTLA-4 on Tregs competes with CD80/CD86, lowers co-stimulation, induces IDO in antigen-presenting cells, and reduces effector T-cell responses (56, 57). β₂ integrins on both MDSCs and Tregs help them stay in close contact and remain inside the tumor, which forms a “closed” immunosuppressive space (58).

The CRC microenvironment shows major metabolic changes that MDSCs and Tregs use to keep long-term suppression. Arginine and tryptophan metabolism, adenosine production, and lactate buildup are central parts of this process.

High ARG1 in MDSCs reduces extracellular arginine. This weakens TCR signaling and causes effector T-cell dysfunction (59). Tryptophan breakdown through the IDO–kynurenine pathway activates AhR, increases FoxP3 transcription, and raises peripheral Treg numbers (60–62). CTLA-4 on Tregs also increases IDO expression in antigen-presenting cells, which keeps this metabolic loop active (63).

In many tumors, MDSCs and Tregs use CD39 and CD73 to make immunosuppressive adenosine from extracellular ATP (64). CD39 turns ATP/ADP into AMP, and CD73 converts AMP into adenosine (65). The accumulated adenosine then activates A2A receptors on effector T cells and NK cells. This triggers cAMP–PKA signaling and reduces activation, proliferation, and cytotoxicity in these cells (66, 67). At the same time, adenosine supports the survival and stability of suppressive immune cells, including MDSCs and Tregs, and shapes the TME into a suppressive niche (65, 68). Hypoxia in the TME increases CD39/CD73 expression through HIF-1α and raises adenosine levels even more, which strengthens immunosuppression (69). Furthermore, HIF-1α signaling can induce expression of ENTPD2 (also known as CD39L1), an ectoenzyme that converts extracellular ATP to 5’-AMP, thereby preserving MDSC suppressive phenotypes by impairing their differentiation into mature immunostimulatory dendritic cells—a mechanism that contributes to sustained myeloid immunosuppression in hypoxic solid tumors (70).

Inflammation, metabolic stress, and tumor-derived signals come together at a set of transcriptional and epigenetic regulators. STAT3, NF-κB, and HIF-1α are the main factors that integrate these signals and help maintain stable suppressive phenotypes in MDSCs and Tregs.

STAT3, activated by IL-6, IL-10, GM-CSF and other upstream signals, is essential for the suppressive differentiation of MDSCs. It induces ARG1, iNOS, and PD-L1, promotes metabolic rewiring, and biases myeloid cells toward suppressive lineages (71, 72). STAT3 also modulates DNA methylation and histone acetylation to stabilize suppressive transcriptional programs in both MDSCs and Tregs (73–75). Pharmacologic STAT3 inhibition reduces the accumulation of both populations and restores CD8⁺ T-cell activity (76, 77). Moreover, the CCL2–STAT3 axis directly enhances PMN-MDSC suppressive function (78), further highlighting STAT3 as a central driver of the MDSC–Treg circuit (79).

NF-κB activation downstream of TLR4 and TNFR2 induces IL-1β, IL-6, CCL2, and CXCL8, promoting continuous recruitment and activation of MDSCs and Tregs (80, 81). Together with STAT3, NF-κB increases S100A8/A9 and PD-L1 expression, intensifying immune evasion (82–84). NF-κB also enhances chromatin accessibility at FoxP3 and CTLA-4, reinforcing Treg stability (85).

Hypoxia-induced HIF-1α serves as a crucial link between tumor metabolism and immune suppression in CRC (86). HIF-1α upregulates glycolytic enzymes and immunosuppressive genes such as ARG1, CD73, and PD-L1 (87), and directly binds the FoxP3 promoter to promote Treg differentiation. By regulating HDACs and KDM3A, HIF-1α further embeds these suppressive programs through epigenetic mechanisms (88, 89).

An integrated schematic overview illustrating the core cytokine, metabolic, contact-dependent, and transcriptional mechanisms sustaining the MDSC–Treg immunosuppressive axis in CRC is presented in Figure 2.

Schematic illustration depicting the bidirectional and interconnected pathways driving durable immunosuppression. Bidirectional cytokine loops (IL-10 and TGF-β) and contact-dependent interactions (β₂ integrins, PD-L1/PD-1, CTLA-4) mutually reinforce MDSC and Treg suppressive phenotypes. Metabolic reprogramming includes ARG1/iNOS-mediated arginine depletion, IDO–kynurenine–AhR signaling promoting FoxP3 expression, and CD39/CD73-dependent adenosine generation via the ATP–adenosine–A2AR pathway, all leading to effector T-cell/NK-cell inhibition. These processes converge on central transcriptional hubs STAT3, NF-κB, and HIF-1α (activated by upstream signals including hypoxia, IL-10, TGF-β, and inflammatory mediators), which transcriptionally upregulate immunosuppressive molecules (PD-L1, ARG1, iNOS, CD39/CD73) and stabilize suppressive programs through epigenetic modifications. Solid green arrows indicate activation/induction; dashed red lines indicate inhibition/suppression.

Hypoxia, a hallmark of the CRC tumor microenvironment, stabilizes HIF-1α, which — as detailed in Sections 3.2 and 3.3 — upregulates key immunosuppressive molecules (PD-L1, ARG1, iNOS, CD73) in MDSCs and promotes FoxP3 expression and peripheral Treg differentiation (90–93). This generates a self-reinforcing immunosuppressive loop that further stabilizes the MDSC–Treg axis under low-oxygen conditions.

In terms of metabolism, the Warburg effect causes high lactate buildup and extracellular acidification. This creates a biochemical environment that strongly supports immunosuppression (94, 95). Lactate helps MDSC expansion and ARG1/iNOS induction, and it also serves as a metabolic source that keeps Treg function stable (64, 94, 96). At the same time, glucose loss caused by high lactate use weakens CD8⁺ T-cell energy metabolism and leads to functional exhaustion (97, 98). This metabolic competition reduces antitumor immunity and gives MDSCs and Tregs a selective advantage, which leads to a stable suppressive microenvironment.

CRC is closely linked to gut ecology, so the TME is strongly shaped by microbial signals (99, 100). Microbial components and metabolites change immune signaling and metabolism. They also increase the number and suppressive strength of MDSC and Treg cells (101). The main groups—oncogenic bacteria, beneficial bacteria, and key metabolites—act through direct receptor signals, metabolite-driven changes, and immune-cell recruitment. In this way, they form a “microbiota–metabolism–immunity” loop.

Exosome-mediated communication is a central mechanism by which the MDSC–Treg axis propagates and amplifies across the CRC microenvironment (125). Tumor cells, MDSCs, and Tregs all release exosomes carrying regulatory miRNAs, proteins, and soluble mediators, enabling long-range dissemination of immunosuppressive signals within local tissues and the peripheral circulation (126).

In CRC, tumor cells exposed to hypoxia or high lactate release exosomes enriched in miR-21, miR-146a, TGF-β1, and Galectin-9, which drive the differentiation of bone marrow precursors toward MDSCs and enhance their suppressive capacity, while simultaneously promoting Treg expansion or stabilization (127–129). These tumor-derived vesicles create a “TEX–MDSC–Treg” loop that increases the suppressive environment. Several immunoregulatory miRNAs, such as miR-21, miR-146a, and miR-494, also make MDSC and Treg suppressive functions stronger by targeting pathways like PTEN/STAT3 and NF-κB (130, 131).

Microbiota and hypoxia-driven immunosuppression converge on shared effectors (STAT3/NF-κB activation, IL-10/TGF-β, PMN-MDSC/CCR5⁺ Treg recruitment), yet differ in initiation and dominance. Oncogenic bacteria and metabolites (e.g. Fn via TLR4–NF-κB/Fap2–TIGIT, 4-HPA→JAK2/STAT3–CXCL3) trigger early inflammatory recruitment, while hypoxia/HIF-1α provides sustained metabolic amplification (CD73/A2AR, ARG1/iNOS, FoxP3) in established tumors. This suggests early-stage microbiota modulation (probiotics/antibiotics) and advanced-stage hypoxia targeting (HIF inhibitors/anti-angiogenesis) may offer complementary therapeutic windows in CRC.

An integrated schematic illustrating the microbiota–metabolism–hypoxia–MDSC–Treg immunosuppressive cascade is shown in Figure 3.

Key tumor-microenvironmental factors shaping the MDSC–Treg axis are summarized in Table 1.

Oncogenic gut bacteria (Fusobacterium nucleatum via TLR4–NF-κB and Fap2–TIGIT; Peptostreptococcus anaerobius via integrin α2/β1–PI3K–NF-κB) and microbial metabolites (e.g., 4-hydroxyphenylacetic acid [4-HPA] activating JAK2/STAT3–CXCL3) drive the recruitment and expansion of PMN-MDSCs and regulatory T cells (Tregs). Tumor hypoxia and lactate accumulation stabilize HIF-1α, which upregulates CD73/A2AR adenosine signaling and further reinforces the suppressive phenotypes of both MDSCs (ARG1, iNOS, PD-L1) and Tregs (FoxP3, CTLA-4). Bidirectional reinforcement between MDSCs and Tregs is mediated by cytokines (TGF-β, IL-10, IL-35), CD40–CD40L contact, and shared activation of STAT3/NF-κB pathways. Exosomes released from tumor cells, MDSCs, and Tregs carrying miR-21, miR-146a, and TGF-β amplify the axis distally. The net result is profound suppression of effector CD8⁺ T cells and NK cells, leading to an immune-excluded/desert phenotype and resistance to immune checkpoint inhibitors.

Chemokines and angiogenic factors are key signals that guide MDSC and Treg movement into tumors. Blocking these signals can lower the number of immunosuppressive cells from the start. A major example is the phase III STELLAR-303 trial (136), It showed that the multi-target TKI Zanzalintinib plus Atezolizumab (a PD-L1 inhibitor) increased overall survival in patients with refractory MSS CRC (median OS 10.9 vs 9.4 months; HR 0.80, 95% CI 0.69–0.93; P = 0.0045) (136). Zanzalintinib blocks TAM kinases (TYRO3, AXL, MER), MET, and VEGFR at the same time, and this can lower MDSC and Treg levels (137). This treatment offers the first phase III evidence of better immunotherapy outcomes in MSS CRC. It also shows that blood-vessel-related side effects, like high blood pressure, should be watched.

Other approaches use anti-angiogenic agents together with PD-1/PD-L1 inhibitors. In the single-arm phase II CAPability-01 study (n=34), Sintilimab (anti-PD-1) plus Chidamide (HDAC inhibitor) and Bevacizumab (anti-VEGF) reached an ORR of 44% (95% CI 25.5%–63.9%) in previously treated MSS/pMMR metastatic CRC (138). The results look good, but the small sample and lack of a control group mean the findings still need larger studies. These combinations likely function by reducing tumor angiogenesis and associated hypoxia, thereby attenuating HIF-1α-driven immunosuppressive programs (see Sections 3.3 and 4.1), enhancing effector T-cell infiltration, and decreasing MDSC/Treg accumulation (139).

Importantly, angiogenesis-centered strategies represent only one facet of source-directed intervention. Increasing attention has been directed toward therapies that directly modulate myeloid cell differentiation and survival, particularly those targeting the CSF-1/CSF-1 receptor (CSF-1R) axis.

Other than PD-1 and CTLA-4, new checkpoints such as TIM-3, TIGIT, and LAG-3 also help maintain the MDSC–Treg axis. These molecules interact with ligands (e.g., Galectin-9, MHC-II) to cooperatively inhibit T cell activation and effector function, stabilizing the immunosuppressive network (150–153). In CRC TME, tumor-infiltrating CD4⁺ T cells upregulate PD-1, CTLA-4, TIM-3, and LAG-3, with high co-expression in the highly suppressive CD4⁺FoxP3⁺Helios⁺ Treg subset. FoxP3^high Helios⁺ Tregs express higher levels of CTLA-4, TIM-3, and LAG-3 compared with FoxP3^low T cells, conferring stronger suppressive potential (154). This heterogeneity highlights the central role of Treg subsets in IC-mediated resistance, suggesting that depleting high IC-expressing Tregs should be prioritized to reverse TME immunosuppression.

Multiple early studies have evaluated the safety and preliminary activity of the combination of anti-TIM-3 monoclonal antibodies (such as Cobolimab/TSR-022, Sabatolimab/MBG453) with PD-1 inhibitors (155–157). AMBER et al. Phase I/Ib studies have shown that this combination regimen has controllable safety and observable anti-tumor activity in various solid tumors, but there is still limited publicly available evidence of benefits for MSS-CRC (155, 156).

In CRC TME, immunophenotyping and histology indicate TIGIT is frequently expressed on tumor-infiltrating T cells, including FoxP3⁺ Tregs, often co-expressed with other inhibitory checkpoints, suggesting a role in maintaining intratumoral immunosuppression (158, 159). Clinically, anti-TIGIT antibody Tiragolumab combined with anti-PD-L1 Atezolizumab has shown early activity in other cancers such as NSCLC (160). Analyses of CRC tumor samples revealed that this combination increased Th1/Treg and Tc1/Treg ratios, indicating an association between high cytotoxic T cells, low immunosuppressive cells, and combination therapy (161). However, efficacy in MSS/pMMR CRC remains unconfirmed in large, reproducible cohorts.

LAG-3, an inhibitory checkpoint expressed on activated CD4⁺/CD8⁺ T cells, Tregs, and NK cells (162, 163), suppresses antigen-specific T cell activation and proliferation in CRC TME while enhancing Treg stability and immunosuppressive function (164). Clinically, Relatlimab (anti-LAG-3) combined with Nivolumab (PD-1 inhibitor) has shown durable benefit in MSI-H/dMMR metastatic CRC (165). In the CheckMate 142 trial (NCT02060188), this combination achieved an ORR of 50% (95% CI 36%–65%) in previously treated MSI-H/dMMR patients, with manageable safety (mainly diarrhea and fatigue) (165).

MSI-H/dMMR CRC remains sensitive to PD-1 inhibitors but resistance occurs in some patients. Single-cell transcriptomic analyses reveal enrichment of MDSCs, particularly PMN-MDSCs, in resistant tumors, suggesting compensatory MDSC upregulation contributes to PD-1 resistance (166). Clinically, low-dose Ipilimumab (1 mg/kg) combined with Nivolumab yielded an ORR of ~55% in dMMR/MSI-H metastatic CRC (167). In first-line cohorts, the combination achieved an ORR of ~69% in phase II studies (168).

For MSS/pMMR CRC, removing Tregs alone cannot reverse the strong immunosuppression. So, “dual” or “multi-step” strategies are needed. These include TKI combinations or new checkpoint co-blockade. They aim to block MDSC recruitment and also weaken Treg function. In this way, they can better overcome immune resistance.

Beyond immune checkpoint blockade, disrupting the MDSC–Treg axis holds promise for improving vaccine-based immunotherapies in MSS/pMMR CRC, where “cold” tumors often limit antigen-specific T-cell priming and effector function (169). Dendritic cell (DC) vaccines aim to overcome this by loading autologous DCs with tumor antigens to stimulate antitumor immunity. The GEMCAD 1602 Phase I/II trial evaluated an autologous DC vaccine combined with avelumab (anti-PD-L1) in pretreated MSS metastatic CRC patients, demonstrating favorable safety and modest clinical efficacy, including immune activation in a subset of cases (170). Although primary endpoints were not fully met, these results highlight the potential of DC vaccines in pMMR disease when combined with strategies to alleviate immunosuppression.

Preclinical evidence indicates that MDSCs and Tregs directly impair vaccine efficacy by suppressing DC maturation, antigen presentation, and effector T-cell responses (171, 172). Depletion or reprogramming of these cells can synergize with vaccines: for instance, CSF-1R inhibitors reduce MDSC/TAM accumulation, enhancing CD8⁺ T-cell infiltration and vaccine-induced immunity in multiple tumor models (146, 173). Low-dose cyclophosphamide selectively depletes Tregs, amplifying responses to peptide or whole-cell vaccines (e.g., GVAX) (174, 175). Emerging modalities, such as neoantigen vaccines or viral vector-based platforms, may similarly benefit from axis disruption to convert immunosuppressive niches into supportive environments for durable T-cell memory (169).

Integrating MDSC/Treg-targeted agents (e.g., CXCR2 antagonists, all-trans retinoic acid, or STAT3 inhibitors) with vaccines represents a rational approach to sensitize MSS/pMMR CRC, potentially extending beyond ICI combinations (176, 177).

A summary of the main therapeutic strategies that target the MDSC–Treg axis, along with related clinical evidence, is shown in Table 2.