Myeloid-derived MIF is a central regulator of MDSC-driven T-cell dysfunction in head and neck squamous cell carcinoma

Abstract

Background

Head and neck squamous cell carcinoma (HNSCC) exhibits a profoundly immunosuppressive tumor microenvironment (TME) enriched for myeloid-derived suppressor cells (MDSCs), regulatory T cells (Tregs), and dysfunctional CD8⁺ T cells, limiting therapeutic benefit from immune checkpoint blockade. Macrophage migration inhibitory factor (MIF) is elevated in HNSCC and linked to poor outcomes, yet the cellular source and functional role of tumor-promoting MIF in shaping antitumor immunity remain unclear.

Methods

We used a myeloid-specific MIF knockout mouse (mMIF KO) in an orthotopic MOC2 HNSCC model and evaluated immune function using ex vivo co-cultures of tumor-derived MDSCs with naïve T cells under Treg-skewing, basal, or Th1-polarizing conditions.

Results

Myeloid-restricted MIF deletion significantly reduced tumor growth and increased CD8⁺ T cell infiltration, accompanied by reduced CTLA4, TIGIT, and TIM3 expression, with elevated PD1 consistent with antigen-engaged effector activation. High-dimensional immune profiling revealed expansion of cytotoxic CD8⁺ T cell states (GranzymeB^hi/Perforin^hi) and contraction of IL10–producing CD4⁺ regulatory (Tr1-like) populations, findings corroborated by decreased intratumoral CD4⁺FoxP3⁺Tregs. Within the myeloid compartment, MIF deletion selectively depleted polymorphonuclear MDSCs (PMN-MDSCs), including CSF1R^hi and PD-L1^hi subsets. Functionally, tumor-derived MDSCs lacking MIF showed impaired capacity to induce Tregs and to promote CD8⁺ T cell exhaustion in ex vivo co-cultures, effects partially overcome by TGFβ/Th1-polarizing signals.

Conclusions

These findings identify myeloid-derived MIF as a central regulator of the MDSC–Treg–CD8⁺ axis in HNSCC and support myeloid-targeted MIF inhibition as a strategy to reprogram the TME and enhance responses to immune checkpoint blockade.

Article highlights

What is already known on this topic

Macrophage migration inhibitory factor (MIF) is elevated in head and neck squamous cell carcinoma and has been linked to tumor progression and immune suppression, but the dominant cellular source of tumor‑promoting MIF and its role in coordinating myeloid–T cell immunoregulatory networks in the tumor microenvironment have remained unclear.

What this study adds

This study identifies myeloid‑derived MIF as a central upstream regulator of MDSC‑driven T‑cell dysfunction in HNSCC, demonstrating that myeloid‑restricted MIF sustains suppressive polymorphonuclear MDSC (PMN-MDSC) subsets, promotes regulatory CD4⁺ T‑cell programs, and enforces CD8⁺ T‑cell exhaustion in a context‑dependent manner.

How this study might affect research, practice or policy

These findings establish myeloid‑derived MIF as a precise immunoregulatory node that can be selectively targeted to reprogram the tumor immune microenvironment, providing a mechanistic foundation for future strategies aimed at alleviating immune dysfunction and improving immunotherapeutic responsiveness in HNSCC.

Introduction

Head and neck squamous cell carcinoma (HNSCC) remains a major global health challenge with substantial morbidity and mortality despite advances in multimodal care and the introduction of immune checkpoint blockade. Clinical responses to PD-1/PD-L1 inhibitors occur in only a subset of patients and durable benefit is uncommon, underscoring the need to define the mechanisms that sustain immune resistance in this disease. ^1-4 A distinctive hallmark of HNSCC is a profoundly immunosuppressive tumor microenvironment (TME) characterized by the accumulation of regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs), together with dysfunctional, exhausted CD8⁺ T cells. These cellular programs act in concert to blunt antitumor immunity and limit the efficacy of immunotherapies in otherwise immunogenic tumors. ^5-15

Among upstream cytokine pathways implicated in these processes, macrophage migration inhibitory factor (MIF) stands out for its breadth of action across innate and adaptive immunity. MIF is produced by multiple stromal and immune cell types and signals through CD74/CD44 and CXCR2/CXCR4/CXCR7 to engage MAPK/ERK, PI3K/AKT, NF-κB, and AMPK pathways linked to tumor growth, survival, angiogenesis, and immune evasion. ^16-22 Elevated MIF expression and circulating MIF have been associated with advanced stage and poor outcomes in HNSCC, and mechanistic studies, including prior work from our group, implicate MIF in promoting tumor progression, expanding suppressive myeloid populations, and constraining antitumor T-cell responses. ^23-29 Collectively, these studies point to MIF as a pivotal coordinator of immune dysfunction in HNSCC.

Despite this, the cellular origin of tumor-promoting MIF within the HNSCC TME and the specific mechanisms by which MIF coordinates myeloid and lymphoid suppression have remained incompletely defined. While tumor cells can express MIF, converging evidence suggests that host-derived MIF is a dominant driver of immunosuppression in HNSCC, yet it is unresolved which host compartment supplies the MIF that sustains immune dysfunction, particularly within the interconnected MDSC–Treg–CD8 axis. This uncertainty limits the rational design of targeted strategies because systemic MIF inhibition may carry on-target liabilities, whereas cell type–selective blockade could dismantle immunosuppressive circuits while sparing essential physiological functions of MIF. ²⁷

Myeloid cells are compelling candidates for the dominant tumor-promoting source of MIF. They are abundant producers of MIF in inflamed tissues, centrally shape the immune landscape, and include polymorphonuclear MDSCs (PMN-MDSCs) that potently suppress CD8⁺ effector functions and foster Treg expansion; both MDSCs and Tregs express MIF-responsive receptors, supporting the plausibility that myeloid-derived MIF licenses these regulatory programs. ^28-34 However, a causal, cell type–restricted demonstration that myeloid MIF orchestrates the MDSC–Treg–CD8 circuit in HNSCC has been missing.

Here, we address these unresolved questions using a myeloid-specific MIF knockout (LysM-Cre; Mif^fl/fl) mouse in an orthotopic MOC2 HNSCC model, coupled with high-dimensional flow cytometry (opt-SNE/FlowSOM) and functional ex vivo MDSC–T-cell co-culture assays. MOC2 is an aggressive, poorly immunogenic, HPV-negative murine HNSCC cell line that gives rise to tumors characterized by a highly immunosuppressive tumor microenvironment enriched in MDSCs and FoxP3⁺ regulatory T cells. These features make MOC2 particularly well suited for investigating mechanisms of tumor-associated immune suppression. MOC2 also recapitulates several features of human HNSCC, including commonly observed mutations (e.g., Tp53, Notch1, Fat1, Kras, and Hras), reduced MHC class I expression, and high Treg infiltration, supporting its relevance as a clinically representative HNSCC model. ³⁵ In contrast, MOC1 tumors are indolent, more immunogenic, and more responsive to immunotherapy than MOC2 tumors. ³⁶ Therefore, both MOC2 and MOC1 models were utilized in this study to represent distinct immunologic contexts of HNSCC, with MOC2 serving as the primary model because of its aggressive growth and highly immunosuppressive, myeloid-rich tumor microenvironment. ³⁷

This approach enables a cell type–restricted dissection of how myeloid-derived MIF shapes both the myeloid compartment and T-cell states in vivo. The work advances the field in three conceptually important ways: it identifies the myeloid compartment as the dominant source of tumor-promoting MIF in HNSCC; it demonstrates that myeloid-derived MIF sustains PMN-MDSC abundance and immunosuppressive subsets while maintaining regulatory CD4⁺ programs (Tr1-like cells and FoxP3⁺ Tregs); and it shows that MDSC-intrinsic MIF is required to drive CD4⁺ Treg differentiation under immunosuppressive contexts and CD8⁺ T-cell exhaustion under effector-stimulating contexts, thereby mechanistically linking myeloid MIF to the failure of antitumor immunity. ^{29 , 34} Moreover, by identifying the myeloid compartment as the predominant source of tumor-promoting MIF, this study provides a foundation for therapeutic strategies that selectively target MIF signaling in myeloid cells.

Methods

Ethics approval

All animal studies were conducted in accordance with institutional and national guidelines and were approved by The Ohio State University Institutional Animal Care and Use Committee (IACUC) under Protocol number 2018A00000054. Research involving recombinant or biohazardous materials was reviewed and approved by The Ohio State University Institutional Biosafety Committee (IBC) under Protocol number 2018R00000038-R1. This study adhered to the ARRIVE guidelines for reporting in vivo animal research.

Animals

Animals were housed and maintained in accordance with University Laboratory Animal Resources (ULAR) guidelines. C57BL/6 male and female mice aged 8–12 weeks, including WT and mMIF KO (n ≥ 10 per group), were used in this study. Sample sizes were selected based on previous studies from our laboratory using the orthotopic MOC2 HNSCC model and MIF knockout mice, ^{26 , 27} which provided estimates of effect size and variability for the primary experimental endpoints. The number of animals per group was chosen to provide adequate statistical power to detect biologically meaningful differences while adhering to ARRIVE guidelines and minimizing animal use. Animals were kept under a 12-h d/night cycle with food and water ad libitum. Mice were monitored regularly for tumor growth and overall health and were euthanized according to humane endpoint criteria approved under the institutional IACUC protocol. Humane endpoints included clinical signs of distress, such as reduced activity, hunched posture, impaired mobility, or tumor ulceration. Therefore, animals could be euthanized prior to reaching the maximum allowable tumor burden if animal welfare concerns were observed. The harvest time point used in this study was consistent with our previous studies employing the MOC2 model. At terminal sacrifice, mice were euthanized according to American Veterinary Medical Association guidelines. Euthanasia was performed with compressed CO₂ using a regulator and flow meter, delivered into an uncharged chamber at 30%–70% volume displacement per minute. CO₂ flow was maintained for ≥1 min after respiratory arrest, followed by cervical dislocation.

Generation of myeloid-specific MIF knockout mice

Myeloid-specific mMIF KO mice were generated using the Cre-LoxP recombination system. C57BL/6 mice homozygous for a floxed Mif allele (Mif ^fl/fl), produced in our laboratory, were crossed with C57BL/6 LyzMcre transgenic mice (Jackson Laboratory #026861), which expresses Cre recombinase under the control of the lysozyme M (Lyz2) promoter. After multiple generations of breeding, offspring included Mif ^fl/fl mice lacking Cre expression (WT controls) and Mif ^fl/fl mice expressing Cre recombinase (LysMcre;Mif ^fl/fl or mMIF KO), resulting in myeloid-specific deletion of MIF.

Genotyping and knock out mouse validation

Genotyping of ear tissue DNA was performed using PCR analysis to confirm the presence of the LysMcre transgene and floxed Mif alleles. PCR products were run on a 1.5% agarose gel with GeneRuler Ready-to-Use 50 bp DNA Ladder (Thermo Scientific Cat#FERSM0373). Myeloid-specific deletion of MIF was validated by flow cytometric analysis of splenic and lymph node cell populations. MIF expression was assessed in major myeloid cell populations and compared with non-myeloid populations, including T cells, endothelial cells, and fibroblasts to confirm efficient and selective deletion of MIF in myeloid cells of mMIF KO mice while preserving MIF expression in non-myeloid compartments. GFP reporter expression was also evaluated in these cell populations to further assess the specificity of Cre-mediated recombination in this model.

Cell lines

Murine metastatic MOC2 and non-metastatic MOC1 head and neck squamous cell carcinoma cells, generated on a C57BL/6 background, were purchased from Kerafast (Boston, MA, USA). The cells were cultured as monolayers in IMDM/F12 (2:1) supplemented with 5% FBS, 1% penicillin–streptomycin–glutamine, 5 µg/mL insulin, 40 ng/mL hydrocortisone, and 5 ng/mL human recombinant EGF at 37 °C and 5% CO₂. ^38-42

HNSCC orthotopic tumor model

MOC2 and MOC1 cells were cultured to approximately 75% confluency and dissociated using TrypLE Express (Life Technologies). Cells were resuspended in 1 × PBS immediately before injection. A total of 1 × 10⁴ MOC2 cells and 5 × 10⁵ MOC1 cells were injected into the right buccal mucosa of mice under isoflurane anesthesia (4%–5% induction, 1%–2% maintenance) delivered via a calibrated vaporizer with oxygen support, in accordance with Ohio State University ULAR guidelines. mMIF KO mice and their corresponding WT littermates, aged 8–12 weeks, were used for all experiments. To minimize potential confounding effects related to handling or procedure timing, the order of cell injections was randomized for each experimental cohort. Tumor measurements were performed by investigators blinded to genotype to reduce bias. Body weight was recorded once weekly, and tumor dimensions were measured twice weekly for 27 d for MOC2 and 37 d for MOC1. Tumor volume was calculated using the formula V = (A × B²)/2, where A represents the longest and B the shortest tumor diameter. At terminal collection, primary tumors, draining cervical lymph nodes, and spleens were harvested for downstream analyses.

Flow cytometry

Single cell suspensions were prepared from the spleens, lymph nodes, and tumors of mice. Tumors were digested with Type V collagenase (0.5 mg/mL) for 1 h at 37 °C, followed by mechanical dissociation to generate single cell suspensions. Spleen samples were then treated with ACK lysis buffer to remove erythrocytes prior to staining. Cells were stained with fluorochrome-conjugated antibodies against extracellular markers including CD45 (Cat#103140), CD3 (Cat#100216), CD4 (Cat#100734), CD8 (Cat#100414), CD69 (Cat#104532), PD-1 (Cat#135218), LAG-3 (Cat#125219), CTLA-4(Cat#106310), TIM-3(Cat#134014), TIGIT(Cat#156104), CD11b (Cat#101206), F4/80 (Cat#123132), CD206 (Cat#141716), Ly6C (Cat#128041), Ly6G (Cat#127608), CSF1R (Cat#135532), IA/IE, (Cat#107622), PD-L1 (Cat#124324), Arginase-1 (Cat# 165804), iNOS(Cat#696808), CD31(Cat#), CD140a(Cat# 135921), CD172a(Cat# 144010), XCR1(Cat# 148204), and CD103(Cat# 121439). For intracellular staining of lymphoid cells, antibodies targeting granzyme B (Cat#372208), IFN-γ (Cat#505822), TNF-α (Cat#506328), IL-10 (Cat#5050345), IL-2 (Cat#5050345), and perforin (Cat#154404) were used following a 6-h stimulation with PMA and ionomycin. For CD107a degranulation assay, single-cell suspensions from tumors, spleens, and draining lymph nodes were stimulated with PMA and ionomycin in the presence of fluorochrome-conjugated anti-CD107a antibody (Cat# 121615), brefeldin A, and monensin for 6 h at 37 °C. Following stimulation, cells were stained for CD45, CD3, and CD8 and processed for flow cytometric analysis. All antibodies were obtained from BioLegend (San Jose, CA, USA). Flow cytometric data were acquired using a BD FACS Celesta flow cytometer (BD Biosciences, San Jose, CA) and analyzed using FlowJo software (Tree Star, Inc., Ashland, OR). High-dimensional analysis of the flow cytometry data was performed using OMIQ software (Dotmatics, www.omiq.ai), where dimensionality reduction (opt-SNE) and FlowSOM clustering were used to identify and classify cell populations. A cluster-by-marker heatmap was generated based on median marker expression. Differential abundance of clusters between experimental groups was assessed using EdgeR analysis. To verify myeloid-specific deletion of MIF in mMIF KO mice, additional flow cytometric analyses were performed using an anti-MIF antibody (Invitrogen, Cat# 20415-1-AP) to assess MIF expression across defined immune cell populations.

MDSC adoptive transfer

Tumor-derived MDSCs were isolated from WT or MIF KO mice using the (StemCell Technologies, 18103). 1 × 10⁶ MDSCs suspended in sterile phosphate-buffered saline (PBS) were administered to tumor-bearing mMIF KO recipient mice by intravenous injection on day 7 following tumor implantation. At the experimental endpoint, tumors were measured, collected and processed for immunohistochemical analysis.

Histological analysis

Tumor tissues were formalin-fixed and paraffin-embedded as previously described. ⁴³ Sections (5 μm) were used for both immunofluorescence (IF) and immunohistochemistry (IHC). For IF, tissue sections were stained with primary antibodies against CD4 (Invitrogen, 14-0042-82) and FoxP3 (Cell Signaling Technology, 12653S), followed by Alexa Fluor 555-conjugated goat anti-rat (Invitrogen, A21434) and Alexa Fluor 488-conjugated goat anti-rabbit (Invitrogen, A11034) secondary antibodies. Nuclei were counterstained with DAPI (BioLegend, San Diego, CA). For IHC, tissue sections were stained with antibodies against Ki-67 (ThermoFisher Scientific, MA5-14520) and Cleaved Caspase-3(Cell Signaling, 9664), followed by a biotinylated goat anti-rabbit secondary antibody (Vector Laboratories, BA-1000). Staining was visualized using DAB (Fisher, SK4100) and counterstained with hematoxylin. Confocal images for IF were acquired using a Zeiss LSM 700 microscope (Carl Zeiss, Munich, Germany), and staining was quantified using ImageJ software.

Real time quantitative PCR

RNA was extracted from tumor tissues with TRIzol, and cDNA was synthesized using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). RT-qPCR was conducted using SYBR Green Master Mix (ThermoFisher) on a Bio-Rad CFX384. Mouse primers for CXCR3, CCL5, CXCL9, CXCL10, and CXCL11 were obtained using PRIMER BANK (https://pga.mgh.harvard edu/primerbank/index.html). PCR amplification was performed using the PowerUp SYBR Green Master Mix (Thermofisher Scientific, Foster City, CA, USA) for detection. Data normalization was performed to housekeeping genes GAPDH and β-actin.

Enzyme-linked immunosorbent assay (ELISA)

Tumor tissues from WT and mMIF KO mice were weighed and homogenized in 1 × RIPA buffer containing protease inhibitors. The homogenates were centrifuged, and the resulting protein lysates were collected for quantification of MIF protein levels by ELISA using MIF capture and detection antibodies (BioLegend; Cat# 532502 and Cat# 524004).

Ex vivo MDSC-T cell co-culture assay

MOC2 cells (1 × 10⁴) were orthotopically injected into the right buccal mucosa of WT and mMIF KO mice. Tumors were harvested three weeks post-injection, when palpable. Tumor-infiltrating MDSCs were isolated using an EasySep™ mouse MDSC isolation kit (StemCell Technologies, 18103). Naive CD3⁺ T cells were isolated from the spleens and lymph nodes of naive WT mice using an EasySep™ mouse T cell isolation kit (StemCell Technologies, 19851). MDSCs isolated from tumors of WT or mMIF KO mice were co-cultured with naive CD3⁺ T cells (2 × 10⁵ cells/well) in 48-well plates pre-coated with anti-CD3 (3 ug/mL; Biolegend, 100340 ). Co-cultures were established under three activation conditions: a Treg-skewing condition containing anti-CD28 (3 µg/mL; Biolegend, 102116), recombinant IL-2 (5 ng/mL; ThermoFisher, 212-12-5UG), and recombinant TGF-β (5 ng/mL; ThermoFisher, 100-21-2UG); a basal activation condition containing anti-CD28 alone (3 µg/mL); and a Th1-skewing condition supplemented with anti-CD28, recombinant IL-2 (5 ng/mL), recombinant IL-12 (10 ng/mL; ThermoFisher, 210-12-2UG), and anti–IL-4 (10 µg/mL; BioXCell, BE0045). For anti-MIF blockade experiments, an anti-MIF antibody (5 ug/mL; R&D Systems, AB-289-PB) was added to the basal and Treg-skewing co-culture conditions. Co-cultures were maintained at 37 °C with 5% CO₂ for 5 d. T cells were subsequently harvested and stained with fluorochrome-conjugated antibodies to assess canonical Treg markers (CD4, CD25, FoxP3, and CD8) and exhaustion markers such as PD-1, LAG3, CTLA-4, TIGIT, and TIM3. Samples were analyzed on a FACS Celesta flow cytometer (BD Biosciences, San Jose, CA).

Treg suppression assay

CD3⁺ responder T cells were isolated from WT mouse spleens and lymph nodes using an EasySep™ Mouse T Cell Isolation Kit (StemCell Technologies, 19851) and labeled with CFSE. CFSE-labeled responder T cells were stimulated with plate-bound anti-CD3 (1 ug/ml) and soluble anti-CD28 (1ug/ml) antibodies and co-cultured with ex vivo-generated induced regulatory T cells (iTregs) for 5 d. Following co-culture, cells were stained with fluorochrome-conjugated antibodies against CD3, CD4, and CD8, and responder T-cell proliferation was assessed by CFSE dilution using flow cytometry.

Ex vivo CD8⁺ T cell cytotoxicity assay

MOC2 cells (2500 cells/well) were seeded into 96-well plates and allowed to adhere overnight. Tumor cells were stained with Hoechst 33342 (Invitrogen, R37605) for 30 min in the dark prior to co-culture with tumor-derived CD8⁺ T cells isolated from WT or myeloid-specific MIF knockout (mMIF KO) mice using an EasySep™ Mouse CD8⁺ T Cell Isolation Kit (STEMCELL Technologies, Cat# 19853). CD8⁺ T cells were added at effector-to-target (E:T) ratios of 1:1 and 5:1. Annexin V apoptosis detection reagent (Invitrogen, V13245) was added at the initiation of co-culture to monitor tumor cell apoptosis. Tumor cell apoptosis was monitored by live-cell imaging using a Cytation 5 Cell Imaging Multi-Mode Reader (Agilent BioTek) every 2 h for 48 h. Cytotoxic activity was assessed by quantifying Annexin V⁺ apoptotic tumor cells using Gen5 Image + 3.17 software.

Statistical analyses

All data analyses were performed using GraphPad Prism v9.0 (GraphPad Software, San Diego, CA). For comparisons of tumor weight, FlowJo-derived flow cytometry data, and immunofluorescence quantification, unpaired two-tailed Student's t-tests were applied to normally distributed datasets, with statistical significance defined as p < 0.05. High-dimensional flow cytometry data analyzed in OMIQ underwent dimensionality reduction and clustering using opt-SNE, followed by differential cluster analysis using EdgeR. p-values were adjusted for multiple comparisons using the false discovery rate (FDR), and clusters with FDR-adjusted p-values < 0.05 were considered significant. Tumor volume measurements were evaluated using multiple t-tests, with p < 0.05 considered statistically significant.

Results

Myeloid-specific MIF deletion reduces cancer progression in an orthotopic HNSCC model

We generated a conditional myeloid-specific MIF knockout LysM-Cre; Mif^fl/fl (mMIF KO) mice. Mice carrying LoxP-flanked Mif alleles and a GFP reporter were crossed with LysM-Cre transgenic mice to achieve targeted deletion of Mif within the myeloid compartment. The resulting mMIF KO mice were homozygous for floxed Mif and positive for LysM-Cre, whereas control mice carried the LoxP alleles but lacked Cre expression (Figure 1A). PCR genotyping followed by agarose gel electrophoresis revealed the expected banding patterns, including a 404 bp product for the floxed Mif allele, a 350 bp band for the Lyz2 wild-type locus, or a 700 bp LysM-Cre band in mMIF KO mice (Figure 1B and Supplemental Figure 1). Flow cytometric analysis confirmed myeloid-restricted loss of MIF in major myeloid populations of mMIF KO mice compared with WT controls, while MIF expression remained preserved in T cells and non-myeloid stromal populations, including CD45⁻CD31⁺ endothelial cells and CD45⁻CD140a⁺ fibroblasts. MIF expression was reduced by approximately 91% in PMN-MDSCs, 83% in monocytes, and 85% in F4/80⁺ macrophages relative to WT controls in our model (Figure 1C–E and Supplemental Figure 2), consistent with the reported targeting efficiency of the LysM-Cre model. ⁴⁴ Consistent with these findings, total MIF levels within the tumor microenvironment were significantly reduced, as determined by ELISA (Figure 1F).To assess the functional contribution of myeloid-specific MIF loss to HNSCC tumor progression, MOC2 (1 × 10⁴) or MOC1 (5 × 10⁵) cells were orthotopically injected into the right buccal mucosa of mMIF KO and WT mice. In the MOC2 model, mMIF KO mice exhibited reduced tumor growth beginning on day 15 (Fig 1G,I), resulting in significantly smaller tumors and tumor weights (1.37-fold reduction, p = 0.0356) compared to WT controls at the experimental endpoint. Similarly, in the MOC1 model, mMIF KO mice displayed delayed tumor growth, and developed significantly smaller tumors with a 4.55 fold reduction in tumor weights compared to WT controls by day 37 (p = 0.0239) (Fig 1H-I). Collectively, these findings demonstrate that myeloid-derived MIF promotes HNSCC tumor progression in vivo.

Figure 1.

Myeloid-specific MIF deletion reduces cancer progression in an orthotopic HNSCC model (a) Breeding strategy for generation of myeloid-specific MIF KO (LysMcre;Mif ^fl/fl hereafter mMIF KO) mouse model by sequential crossing floxed Mif (Mif ^fl/fl; WT) mice with LysMcre mice. (b) PCR genotyping and agarose gel electrophoresis of WT and mMIF KO mice using LysM WT, LysMcre and LoxP primers. WT (red) mice lack the LysMcre transgene (−/−) and carry floxed Mif alleles (Mif ^fl/fl ), mMIF KO (blue) mice are positive for the LysMcre transgene (+/) and homozygous for floxed Mif (Mif ^fl/f ). (c and d) Gating strategy and representative flow cytometry plots demonstrating MIF expression in different myeloid cell populations including PMN-MDSCs (CD11b⁺ Ly6G^hi Ly6C^lo), macrophages (CD11b⁺ F4/80^{+ −}) and monocytes (CD11b⁺ Ly6G⁻ Ly6C^hi) isolated from the spleens of MOC2 tumor-bearing WT and mMIF KO mice. Corresponding bar graphs show the quantification of MIF expression in mMIF KO mice compared with WT controls. (e) Representative flow cytometry histogram showing MIF expression in bone marrow-derived macrophages (BMDMs), peritoneal macrophages, T cells, endothelial cells (CD45⁻ CD31⁺)and fibroblasts (CD45⁻ CD140a⁺) isolated from naïve WT and mMIF KO mice. (f) MIF protein levels in tumor tissues of WT and mMIF KO mice measured by ELISA. (g) Representative images and weights of MOC2, and (h) MOC1 tumors collected at terminal endpoint. (i) Tumor growth curves of MOC2 and MOC1 HNSCC tumors in WT and mMIF KO mice over 27 d and 37 d respectively. Data are presented as mean ± SEM. Statistical significance was assessed using a two-tailed unpaired Student's t-test or a multiple t-test. *p < 0.05; **p < 0.01; ***p < 0.001.

Myeloid-derived MIF shapes the tumor myeloid landscape by restricting PMN-MDSC accumulation and suppressive subsets

Given the established roles of MIF in myeloid cell recruitment, differentiation, and survival, we next investigated whether myeloid-specific MIF deletion altered the composition of the tumor-infiltrating myeloid compartment. We hypothesized that MIF produced by myeloid cells contributes to the accumulation and maintenance of immunoregulatory myeloid populations within the tumor microenvironment. Therefore, we assessed the abundance of MDSC subsets, macrophage populations, and conventional dendritic cell subsets to determine how loss of myeloid-derived MIF influences the overall myeloid landscape in HNSCC (Figure 2A). Flow cytometric profiling revealed a significant reduction in tumor-infiltrating polymorphonuclear MDSCs (PMN-MDSCs; CD11b⁺Ly6G^hiLy6C^lo) in mMIF KO mice compared to WT controls (p = 0.0425), while monocytic MDSCs (m-MDSCs; CD11b⁺Ly6G⁻Ly6C^hi) remained unchanged. Consistent with these findings, PMN-MDSC frequencies were also significantly reduced in the tumor-draining lymph nodes and spleen, while m-MDSCs remained unchanged (Figure 2B). To further resolve the tumor myeloid compartment at higher resolution, we performed high-dimensional analysis of CD45⁺ tumor-infiltrating myeloid cells using opt-SNE and FlowSOM, which identified 25 distinct clusters spanning polymorphonuclear, monocytic, dendritic, and macrophage lineages (Figure 2C–E). Differential abundance analysis revealed that four PMN-MDSC clusters were selectively depleted in mMIF KO tumors: Cluster 2 (canonical PMN-MDSCs; CD11b⁺Ly6G^hiLy6C^lo), Cluster 1 (CSF1R^hi PMN-MDSCs), Cluster 3 (PD-L1^med PMN-MDSCs), and Cluster 9 (PD-L1^hi PMN-MDSCs) (Figure 2F). The loss of the CSF1R^hi subset suggests that myeloid-derived MIF supports survival or differentiation pathways mediated by CSF1–CSF1R signaling, which are essential for PMN-MDSC maintenance. Meanwhile, the reduction in PD-L1^hi and PD-L1^med clusters indicates that MIF is also required for the formation or stability of PMN-MDSC states enriched for checkpoint ligand expression, subsets known to potently inhibit effector T-cell responses.

Figure 2.

Myeloid-derived MIF shapes the tumor myeloid landscape by restricting PMN-MDSC accumulation and suppressive subsets. (a) Gating strategy of MDSCs in tumors, lymph node and spleen from MOC2 tumor-bearing WT and mMIF KO mice. (b) Representative flow cytometry plots of PMN-MDSCs (CD11b⁺Ly6G^hiLy6C^lo) and m-MDSCs (CD11b⁺Ly6G⁻Ly6C^hi) within the CD11b⁺ population. The quantification of MDSC population is shown as percentage of CD45⁺ cells in the tumors, lymph node and spleen of MOC2 tumor-bearing WT and mMIF KO mice, analyzed using FlowJo software. Data is represented as mean ± SEM. Statistical significance was determined using a two-tailed unpaired Student's t-test. *p < 0.05. (c–f) High dimensional flow cytometric analysis of extracellular myeloid markers in tumors from MOC2 tumor-bearing WT and mMIF KO mice using OMIQ Software. A total of 25 concatenated samples were used for analysis. (c) opt-NSE plots showing individual marker expression pattern for multiparametric visualization. (d) CD45⁺ cells were clustered in opt-NSE space using FlowSOM, identifying 17 distinct immune cell clusters based on their marker expression profiles. (e) Heatmap showing median marker expression across all clusters. (f) Quantification of significant polymorphonuclear myeloid clusters, including Cluster 2 (CD11b⁺Ly6G^hiLy6C^lo), Cluster 1 (CD11b⁺Ly6G^hiLy6C^loCSF1-R^hi), Cluster 3 (CD11b⁺Ly6G^hiLy6C^loPD-L1^med), and Cluster 9 (CD11b⁺Ly6G^hiLy6C^loPD-L1^hi) shown as percentage of CD45⁺ cells in WT and mMIF KO mice. Data are represented as mean ± SEM. Statistical significance between groups was calculated using edgeR differential abundance analysis in OMIQ. *FDR < 0.05.

Along with these PMN-MDSC-specific effects, we further expanded our characterization of the myeloid compartment by examining dendritic cell subsets and macrophage polarization (Supplemental Figure 3A). Myeloid-specific MIF deletion was associated with a significant increase in cDC1 frequencies along with a notable decrease in cDC2 frequencies. (Supplemental Figure 3). In contrast, no significant differences were observed in M1 (IA/IE^hi iNOS⁺) or M2 (CD206⁺Arginase-1⁺) macrophage frequencies between WT and mMIF KO tumors (Supplemental Figure 3B and C). These findings highlight the selective nature of myeloid-derived MIF in shaping the tumor myeloid landscape. Rather than broadly affecting all myeloid lineages, myeloid-derived MIF preferentially sustains the abundance and suppressive programming of PMN-MDSCs, including their CSF1R-associated survival signatures and PD-L1–associated inhibitory phenotype identified by high-dimensional analysis. Collectively, these analyses demonstrate that myeloid-derived MIF is required to maintain key PMN-MDSC populations within the HNSCC tumor microenvironment, including both canonical and highly suppressive functional subsets. The selective collapse of these PMN-MDSC states in mMIF KO tumors provides a mechanistic basis for the enhanced CD4⁺ and CD8⁺ T-cell immunity observed following myeloid-specific MIF deletion.

Loss of myeloid MIF increases intratumoral CD8⁺ T cell infiltration and alleviates T cell exhaustion

Having established that myeloid-specific MIF deletion reshapes the tumor myeloid compartment, including reductions in PMN-MDSCs and alterations in dendritic cell subsets, we next sought to determine the downstream consequences of these changes on antitumor immunity. Given the critical role of myeloid cells in regulating T-cell priming, recruitment, and effector function, we hypothesized that loss of myeloid-derived MIF would promote a more anti-tumor immune microenvironment characterized by enhanced CD8⁺ T-cell responses. Accordingly, we quantified CD8⁺ T-cell infiltration and exhaustion marker expression in tumors, tumor-draining lymph nodes, and spleens (Figure 3A and B). In the MOC2 model, tumor-infiltrating CD8⁺ T cells were increased 2.35-fold compared with WT controls (p = 0.0012), whereas the MOC1 model exhibited only modest, non-significant increases. No significant differences in CD8⁺ T-cell frequencies were observed in the spleen or cervical lymph nodes in either model (Figure 3C). To investigate potential mechanisms underlying enhanced T-cell infiltration, we next evaluated chemokine expression and observed significantly increased tumor expression of CXCL9 and CCL5 in mMIF KO mice (Figure 3D). Furthermore, CD8⁺ T cells from mMIF KO tumors exhibited reduced expression of multiple exhaustion markers. CTLA-4 expression was significantly reduced by 2.05-fold (p = 0.0285) and 1.82-fold (p = 0.0355) relative to WT levels in MOC2 and MOC1 tumors, respectively. Similarly, TIGIT expression was reduced by 1.92-fold (p = 0.0221) and 2.27-fold (p = 0.0462) compared to WT levels in MOC2 and MOC1 tumors, respectively. TIM-3 expression was significantly reduced by 1.92-fold relative to WT levels in MOC2 tumors (p = 0.0471) and showed a similar trend in MOC1 (Figure 3G–L). In contrast, PD-1 expression on tumor-infiltrating CD8⁺ T cells was increased 1.92-fold in mMIF KO MOC2 tumors compared with WT controls (p = 0.0192), consistent with an antigen-experienced, effector-competent phenotype. Conversely, in the MOC1 model, PD-1 expression was reduced by 1.39-fold of WT levels in mMIF KO tumors (p = 0.0256) (Figure 3E and F). Together, these data indicate that myeloid-derived MIF limits cytotoxic CD8⁺ T-cell immunity by constraining infiltration and promoting exhaustion, while its loss generates an immunogenic TME enriched in functional CD8⁺ T cells that likely accounts for the reduced tumor growth in mMIF KO mice.

Figure 3.

Loss of myeloid MIF increases intratumoral CD8⁺ T cell infiltration and alleviates T cell exhaustion. (a and b) Gating strategy and representative flow cytometry plots of CD3⁺CD8⁺ T cells in tumors from MOC2 tumor-bearing WT and mMIF KO mice. (c) Quantification of CD3⁺CD8⁺ T cells as percentage of CD45⁺cells in the tumor, tumor-draining lymph node and spleen of MOC2 and MOC1 tumor- bearing WT and mMIF KO mice. (d) Gene expression profile of CXCR3, CXCL9, CXCL10, CCL5, CXCL11 in tumors obtained from WT and mMIF KO MOC2 tumor bearing mice, normalized to Gapdh and B-actin. (e–l) Representative flow cytometry plots and corresponding quantification of exhaustion marker expression on CD8⁺ T cells, including (e and f) PD-1, (g and h) CTLA-4, (i and j) TIGIT, and (k and l) TIM3, shown as the percentage of CD8⁺ T cells in the tumor, tumor-draining lymph node and spleen of WT and mMIF KO mice. Data are presented as mean ± SEM. Statistical significance between groups was determined using a two-tailed unpaired Student's t-test. ^# p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.

Myeloid-derived MIF deletion reprograms the tumor-infiltrating T-cell landscape toward cytotoxic effector dominance

Given the increased CD8⁺ T-cell infiltration and reduced exhaustion observed in mMIF KO tumors, we next asked whether myeloid-derived MIF more broadly influences the phenotypic composition of tumor-infiltrating T cells. Because MIF regulates multiple immunosuppressive circuits within the TME, we hypothesized that its loss would shift T-cell states away from regulatory programs and toward cytotoxic effector phenotypes that support antitumor immunity. To resolve these changes at high resolution, we performed intracellular cytokine and effector molecule staining on tumor-infiltrating CD3⁺ T cells from WT and mMIF KO mice, followed by high-dimensional analysis using opt-SNE and FlowSOM clustering in OMIQ (Figure 4A–C). This approach identified 17 distinct T-cell clusters encompassing diverse effector and regulatory states (Figure 4B). Differential abundance testing revealed a significant enrichment in mMIF KO tumors of Cluster 1 (CD8⁺ Granzyme B^hi Perforin^hi IFN-γ^med) and Cluster 2 (CD8⁺ Granzyme B^hi) (Figure 4D). Cluster 1 and Cluster 2 were approximately 6.85-fold (FDR = 0.0043) and 9.30-fold (FDR = 1.76 × 10⁻⁶) higher in mMIF KO tumors compared with WT tumors. Both clusters represent highly activated cytotoxic CD8⁺ effector populations, consistent with the enhanced CD8⁺ immunity observed in these mice.

Figure 4.

Myeloid-derived MIF deletion reprograms the tumor-infiltrating T-cell landscape toward cytotoxic effector dominance. (a–e) Flow cytometric analysis of extracellular and intracellular lymphoid markers in tumors of MOC2 tumor-bearing WT and mMIF KO mice, processed and analyzed using the OMIQ Software. A total of 25 concatenated samples were used for high-dimensional analysis. (a) opt-NSE plots showing individual marker expression pattern for multiparametric visualization of multicolor flow cytometry data. (b) CD3⁺ cells were clustered on opt-NSE space using FlowSOM identifying 17 distinct CD3⁺ cell clusters based on their marker expression profiles. (c) Heatmap showing the median expression of analyzed markers across all clusters. (d and e) Quantification of significant clusters, including Cluster 1 (CD8⁺Granzyme-B^hiPerforin^hiIFN-γ^med), Cluster 2 (CD8⁺Granzyme B^hi), Cluster 16 (CD4⁺IL-2^hiIL-10^hiIFN-γ^hi), and Cluster 17 (CD4⁺IL-2^hiIL-10^hiPerforin^hi), shown as percentage of total CD3⁺ T cells in WT and mMIF KO mice. Data are presented as mean ± SEM. Statistical significance between groups was calculated using edgeR differential abundance analysis in OMIQ software. *FDR < 0.05; **FDR < 0.01; ****FDR < 0.001.

In contrast, Cluster 16 (CD4⁺ IL-2^hi IL-10^hi IFN-γ^hi) and Cluster 17 (CD4⁺ IL-2^hi IL-10^hi Perforin^hi) were markedly reduced in mMIF KO tumors relative to WT (Figure 4E). Cluster 16 and Cluster 17 were reduced by approximately 4.35-fold (FDR = 2.46 × 10⁻⁸) and 2.33 fold (FDR = 0.017) compared to WT levels in mMIF KO tumors. These clusters correspond to IL-10–producing CD4⁺ regulatory cells, most consistent with Type 1 regulatory (Tr1-like) T cells, which can suppress effector responses through IL-10, IFN-γ–associated regulation, and perforin-dependent cytotoxicity. ⁴⁵ Their contraction in mMIF KO tumors suggests that myeloid-derived MIF supports the maintenance and/or functional licensing of these immunoregulatory CD4⁺ subsets within the HNSCC TME. Together, these high-dimensional data demonstrate that loss of myeloid-derived MIF fundamentally reprograms the intratumoral T-cell landscape, expanding cytotoxic CD8⁺ effector clusters while diminishing IL-10–producing regulatory CD4⁺ populations. This shift toward a more immunostimulatory T-cell composition likely contributes to the enhanced antitumor immunity and reduced tumor growth observed in mMIF KO mice.

Loss of myeloid-derived MIF enhances the cytotoxic potential of CD8⁺ T cells in the HNSCC TME

Given the expansion of cytotoxic CD8⁺ clusters observed in the high-dimensional analysis, we next examined whether myeloid-derived MIF deletion enhances the functional cytotoxic capacity of CD8⁺ T cells at the single-cell level. Cytotoxic effector activity depends not only on CD8⁺ T-cell abundance but also on expression and coordinated co-expression of molecules such as granzyme B, perforin, and IFN-γ. Therefore, we quantified both individual and combinatorial effector signatures in tumor-infiltrating CD8⁺ T cells from WT and mMIF KO mice.

Flow cytometric analysis revealed significantly increased frequencies of CD8⁺Granzyme B⁺ and CD8⁺Perforin⁺ cells in mMIF KO tumors relative to WT controls (Figure 5A and B), indicating heightened cytolytic readiness. Specifically, CD8⁺Perforin⁺ T cells were increased 2.25-fold (p = 0.001286), whereas CD8⁺Granzyme B⁺ T cells were increased 1.88-fold (p = 0.037269), indicating enhanced cytolytic potential (Figure 5A and B). To further assess functional potency, we evaluated co-expression patterns of key effector molecules, a hallmark of highly competent cytotoxic T cells. CD8⁺ T cells co-expressing Granzyme B + Perforin, Granzyme B + IFN-γ, and Perforin + IFN-γ were all markedly enriched in mMIF KO tumors (Figure 5C–E). Notably, the proportion of triple-positive Granzyme B⁺/Perforin⁺/IFN-γ⁺ CD8⁺ T cells was also substantially increased in the absence of myeloid MIF (Figure 5F), consistent with a robust, multifunctional effector phenotype. Consistent with these findings, CD8⁺ T cells from mMIF KO tumors exhibited significantly increased CD107a expression, indicative of enhanced degranulation and cytolytic activity (Figure 5G). To determine whether this enhanced effector phenotype translated into improved tumor cell killing, we performed an ex vivo CD8⁺ T cell-MOC2 cytotoxicity assay. Tumor-derived CD8⁺ T cells from mMIF KO mice induced significantly higher frequencies of Annexin V⁺ apoptotic MOC2 tumor cells than WT CD8⁺ T cells at both 1:1 and 5:1 effector-to-target ratios throughout the 48-h imaging period (Figure 5H). Representative live-cell images further confirmed the increased induction of tumor cell apoptosis by mMIF KO-derived CD8⁺ T cells as compared to WT-derived CD8⁺ T cells (Figure 5I). These data complement and extend the high-dimensional clustering results by demonstrating that myeloid-derived MIF not only limits CD8⁺ T-cell infiltration but also suppresses their cytotoxic function. Enhanced effector molecule co-expression, increased CD107a expression, and greater induction of tumor cell apoptosis in the ex vivo cytotoxicity assay collectively demonstrate that loss of myeloid-derived MIF augments both the abundance and functional capacity of tumor-infiltrating CD8⁺ T cells. Together, these findings provide a mechanistic link between enhanced antitumor immunity and the reduced tumor growth observed in mMIF KO mice.

Figure 5.

Loss of myeloid-derived MIF enhances the cytotoxic potential of CD8⁺ T cells in the HNSCC TME. (a–f) Representative flow cytometry plots and quantification of cytotoxic and effector CD8⁺ T cell subsets, including (a) CD8⁺Granzyme B⁺, (b) CD8⁺Perforin⁺, (c) CD8⁺Granzyme B⁺Perforin⁺, (d) CD8⁺Granzyme B⁺IFN-γ⁺, (e) CD8⁺ Perforin⁺IFN-γ⁺, (f) CD8⁺Granzyme B⁺Perforin⁺IFN-γ⁺, and (g) CD8⁺CD107a⁺ shown as a percentage of total CD8⁺T cells in the tumor, tumor-draining lymph node, and spleen of MOC2 tumor-bearing WT and mMIF KO mice. (h) Quantification of Annexin V⁺ apoptotic MOC2 tumor cells following co-culture with tumor-derived CD8⁺ T cells isolated from WT or mMIF KO mice at effector-to-target (E:T) ratios of 1:1 and 5:1 over a 48-h imaging period. (i) Representative live-cell images of MOC2 tumor cells co-cultured with WT or mMIF KO tumor-derived CD8⁺ T cells at an E:T ratio of 1:1, showing Hoechst-stained nuclei (blue) and Annexin V⁺ apoptotic tumor cells (green). Scale bar = 1000 um. Data are presented as mean ± SEM. Statistical significance between groups was assessed using a two-tailed unpaired Student's t-test or multiple unpaired t-test, as appropriate. *p < 0.05; **p < 0.01.

Loss of myeloid-derived MIF reduces IL-10 linked regulatory CD4⁺ programs and intratumoral FoxP3⁺ Tregs

Given the contraction of IL‑10-rich CD4⁺ clusters in the high‑dimensional analysis, we next assessed regulatory cytokine and effector profiles within tumor‑infiltrating CD4⁺ T cells by conventional gating. CD4⁺IL-2⁺ and CD4⁺IL-10⁺ cells were significantly reduced in mMIF KO tumors in the MOC2 model, while the MOC1 model showed similar but non-significant trends (Figure 6A–D). In contrast, CD4⁺Perforin⁺ cells and CD4⁺IL-2⁺IL-10⁺Perforin⁺ Tr1-like cells were significantly decreased in both the MOC2 and MOC1 models (Figure 6E–H). Because perforin‑dependent cytotoxicity contributes to the ability of regulatory CD4⁺ subsets (including Tr1‑like cells) to restrain effector responses, ^{46 , 47} these data indicate a functional impairment of IL‑10-linked regulatory programs when myeloid‑derived MIF is absent.

Figure 6.

Loss of myeloid-derived MIF reduces IL-10 linked regulatory CD4⁺programs and intratumoral FoxP3 + Tregs. (a–h) Representative flow cytometry plots and quantification of CD4⁺ T cell subsets, including (a and b) CD4⁺IL-2⁺, (c and d) CD4⁺IL-10⁺, (e and f) CD4⁺Perforin⁺, (g and h) CD4⁺IL-2⁺IL-10⁺ Perforin cells, shown as a percentage of total CD4⁺ T cells in the tumor, tumor-draining lymph node, and spleen of MOC2 and MOC1 tumor-bearing WT and mMIF KO mice. (i) Representative immunofluorescence images of MOC2 HNSCC tumors from WT and mMIF KO mice stained with DAPI (blue), CD4 (red), and FoxP3 (green), with composite images shown. (j) Quantification of intratumoral CD4⁺FoxP3⁺ T cells based on immunofluorescence staining. Measurements were obtained from at least four field per tumor, with N = 18–20 mice per group. Data are presented as mean ± SEM. Statistical significance between groups was calculated using a two-tailed unpaired Student's t-test. *p < 0.05; ***p < 0.001; ****p < 0.0001.

To determine whether these changes extended to conventional Tregs, we quantified intratumoral FoxP3 expression by immunofluorescence. Consistent with the flow cytometric findings, mMIF KO tumors harbored significantly fewer CD4⁺FoxP3⁺ Tregs (2.38 fold reduction, p = 0.0108) compared with WT. (Figure 6I and J). Collectively, these findings demonstrate that myeloid-derived MIF sustains multiple immunoregulatory CD4⁺ T-cell populations, including Tr1-like cells and FoxP3⁺ Tregs, thereby contributing to an immunosuppressive tumor microenvironment. The coordinated reduction of these immunoregulatory populations in mMIF KO tumors provides a mechanistic link to the enhanced CD8⁺ cytotoxicity and overall antitumor immune pressure observed in the absence of myeloid MIF.

MIF-expressing MDSCs functionally contribute to tumor progression in mMIF-KO mice

Having established that myeloid-specific MIF deletion reshapes both the myeloid and lymphoid compartments and promotes antitumor immunity, we next sought to determine whether MIF expression within MDSCs directly contributes to the observed tumor phenotype. Although loss of myeloid-derived MIF reduced PMN-MDSC accumulation and enhanced cytotoxic immune responses, these studies alone could not distinguish whether the antitumor effects were driven by intrinsic loss of MIF function within MDSCs or by broader alterations in the tumor microenvironment. To address this question, we performed adoptive transfer experiments using tumor-derived WT or MIF-KO MDSCs and assessed their ability to restore tumor-promoting activity in mMIF-KO tumor-bearing mice.

To determine whether MIF expression within MDSCs directly contributes to tumor progression, tumor-derived WT or MIF-KO MDSCs were adoptively transferred into mMIF-KO tumor-bearing mice. Following adoptive transfer, tumors in both groups grew more rapidly than those in non-transferred mice, consistent with a pro-tumorigenic role for MDSCs in this model. However, no significant differences in overall tumor growth were observed between mice receiving WT versus MIF-KO MDSCs (Figure 7A).

Figure 7.

MIF-expressing MDSCs functionally contribute to tumor progression in mMIF-KO mice and myeloid-derived MIF is required for MDSC-mediated Treg induction under immunosuppressive and basal activation conditions. (a) Tumor weights of MOC2 injected mMIF-KO recipient mice adoptively transferred with WT or MIF deficient MDSCsc. (b and c) Representative immunohistochemistry images and quantification of Ki67- and cleaved caspase-3-positive staining of MOC2 tumors from mMIF KO mice following adoptive transfer of WT or MIF KO tumor-derived MDSCs, stained for Ki67 and cleaved caspase-3. Positive staining was quantified from at three representative fields per tumor, with n = 4 mice per group. (d) Schematic of the ex vivo co-culture assay in which CD3⁺ T cells isolated from WT mice were co-cultured with MDSCs isolated from tumors of MOC2-bearing WT or mMIF KO mice. T cells cultured without MDSCs served as controls. Co-cultures were established under three stimulation conditions: Treg stimulation (CD3/CD28, IL-2, TGF-β), Basal stimulation (CD3/CD28), and Th1 stimulation (CD3/CD28, IL-2, IL-12, anti-IL-4). (e) Flow cytometry plots showing CD3⁺ T cells and MDSCs before and after isolation for use in the ex vivo co-culture assay. (f) Representative flow cytometry plots of induced Tregs (CD4⁺CD25⁺FoxP3⁺) under Treg-polarizing conditions and corresponding quantification of Tregs under Treg, basal, and Th1 stimulation conditions. (g) Bar graphs showing the quantification of Tregs following treatment with anti-MIF antibody or isotype control under Treg and basal stimulation conditions. (h) Representative flow cytometry histograms of CFSE dilution in responder CD3⁺ T cells and the corresponding bar graph showing the division index of CD3⁺ T cells cultured alone or in the presence of ex vivo-induced Tregs. Data are presented as mean ± SEM. Statistical significance was determined using a two-tailed unpaired Student's t-test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Given the rapid tumor progression observed following MDSC transfer, we next assessed tumor cell proliferation and apoptosis. Tumors from mice receiving MIF-KO MDSCs exhibited significantly reduced Ki-67 expression compared with tumors receiving WT MDSCs (Figure 7B). Conversely, transfer of mMIF-KO MDSCs resulted in significantly increased cleaved caspase-3 expression (Figure 7C), indicative of enhanced tumor cell apoptosis, and reduced proliferative activity relative to WT MDSC recipients. These findings demonstrate that MIF-expressing MDSCs promote tumor cell survival and proliferation and provide functional evidence that loss of MIF within MDSCs contributes to the antitumor phenotype observed in mMIF-KO mice. Together, these data support a role for MDSC-derived MIF in fostering a tumor-permissive microenvironment.

Myeloid-derived MIF is required for MDSC-mediated Treg induction under immunosuppressive and basal activation conditions

Because myeloid-specific MIF deletion reduced intratumoral Tregs and several IL-10–producing regulatory CD4⁺ subsets in vivo, we next asked whether MDSC-intrinsic MIF directly contributes to Treg induction. To address this, we performed ex vivo MDSC–T-cell co-culture assays in which tumor-derived MDSCs from WT or mMIF KO mice were paired with naïve CD4⁺ T cells under immunosuppressive (Treg-skewing), basal, or Th1-polarizing activation conditions (Figure 7D and E). Under Treg-skewing conditions (CD3/CD28 + IL-2 + TGF-β), WT MDSCs robustly induced CD4⁺CD25⁺FoxP3⁺ Tregs, whereas mMIF KO MDSCs exhibited a significantly diminished Treg inducing capacity (2.08-fold reduction, p = 0.0064) (Figure 7F). A similar impairment was observed under basal activation, in which Treg frequencies were significantly reduced in cultures containing mMIF KO MDSCs relative to WT control (2.63-fold reduction, p = 0.0317) (Figure 7F). These results demonstrate that MDSC-derived MIF is required for efficient Treg induction in immunosuppressive or steady-state environments. To determine whether extracellular MIF mediated this effect, WT MDSCs were cultured in the presence of either neutralizing anti-MIF antibody or an isotype control. MIF neutralization did not significantly alter Treg induction under either basal or Treg-skewing conditions, suggesting that MDSC-intrinsic MIF promotes Treg differentiation through indirect mechanisms (Figure 7G). In contrast, under Th1-polarizing conditions (CD3/CD28 + IL-2 + IL-12 + anti-IL-4), both WT and mMIF KO MDSCs induced uniformly low Treg frequencies, and the disparity between genotypes was eliminated (Figure 7F). This pattern indicates that strong pro-inflammatory signals can override or bypass MIF-dependent regulatory pathways, thereby suppressing Treg induction regardless of MIF availability. Furthermore, to confirm that the induced Tregs were functionally suppressive, we performed a Treg suppression assay, in which induced Tregs significantly suppressed CFSE-labeled responder CD3⁺ T-cell proliferation, as demonstrated by a reduced division index and diminished CFSE dilution (Figure 7H). Together, these findings demonstrate that MDSC-intrinsic MIF functions as a contextual signal that promotes Treg differentiation specifically under immunosuppressive or basal activation states, but becomes dispensable under Th1-dominant conditions. This mechanism provides a direct explanation for the reduced Treg abundance observed in mMIF KO tumors and highlights myeloid-derived MIF as a critical driver of MDSC-mediated immunoregulation in the HNSCC TME.

MDSC-derived MIF drives CD8⁺ T-cell exhaustion under effector-stimulating conditions

Given the enhanced CD8⁺ cytotoxicity observed in mMIF KO tumors, we next investigated whether MDSC-intrinsic MIF contributes to CD8⁺ T-cell exhaustion, a hallmark of MDSC-mediated suppression. Using the same ex vivo co-culture system, naïve CD8⁺ T cells were activated under Th1-polarizing, basal, or Treg-skewing conditions and co-cultured with MDSCs isolated from WT or mMIF KO tumors. Under Th1-polarizing conditions, CD8⁺ T cells co-cultured with WT MDSCs exhibited robust induction of exhaustion markers—including PD-1, CTLA-4, LAG-3, TIGIT, and TIM-3—whereas CD8⁺ T cells cultured with mMIF KO MDSCs showed significantly reduced expression of all markers (Figure 8A–E). Remarkably, exhaustion levels in mMIF KO co-cultures approximated those of T-cell-only controls, indicating that MIF is required for MDSC-mediated enforcement of exhaustion during effector-stimulating activation. A similar pattern was observed under basal activation, where mMIF KO MDSCs failed to induce exhaustion marker upregulation to the extent seen with WT MDSCs, further supporting a direct role for MIF in licensing MDSC suppressive function.

Figure 8.

MDSC-derived MIF drives CD8⁺ T cell exhaustion under effector-stimulating conditions. (a–e) Representative flow cytometry plots (Th1 stimulation condition shown) and quantification across all stimulation conditions depicting (a) CD8⁺PD-1⁺, (b) CD8⁺CTLA-4⁺, (c) CD8⁺LAG3⁺, (d) CD8⁺TIGIT⁺, and (e) CD8⁺TIM3⁺ cells, shown as percentage of total CD8⁺ T cells. CD3⁺ T cells were cultured either alone (T cells only) or co-cultured with MDSCs isolated from tumors of WT or mMIF KO mice under three stimulation conditions: Treg stimulation, Basal stimulation, and Th1 stimulation. Data are shown as mean ± SEM (N = 4). Statistical significance between groups was determined using a two-tailed unpaired Student's t-test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

In contrast, under Treg-skewing conditions, most exhaustion markers (CTLA-4, TIGIT, TIM-3, LAG-3) remained comparably elevated across all conditions, with only PD-1 showing a modest reduction in mMIF KO co-cultures. This suggests that the strong TGF-β–driven environment characteristic of Treg induction programs can maintain CD8⁺ exhaustion independently of MIF, again highlighting the context-specific nature of MIF-mediated immunoregulation. Together, these findings demonstrate that MDSC-derived MIF is a key driver of CD8⁺ T-cell exhaustion under effector-stimulating and basal activation conditions, but its influence is masked in strongly tolerogenic settings. This mechanism provides a direct cellular explanation for the pronounced reduction in CD8⁺ exhaustion observed in mMIF KO tumors and reinforces the central role of myeloid-derived MIF in sustaining dysfunctional CD8⁺ T-cell states within the HNSCC TME.

Human single-cell transcriptomic analysis validates association of MIF with immunosuppressive myeloid programs in HNSCC

To evaluate the translational relevance of our findings, we analyzed the published single-cell RNA-sequencing dataset GSE232240 derived from patients with recurrent/metastatic HNSCC treated with immune checkpoint blockade. ⁴⁸ Unsupervised clustering identified multiple myeloid populations, including conventional dendritic cells, inflammatory myeloid cells, activated interferon-responsive myeloid cells, and a population of immunosuppressive myeloid cells, characterized by expression of APOE, GPNMB, TREM2, and C1Q family genes (Supplementary Figure 4A). MIF expression was broadly detected throughout the myeloid compartment, demonstrating that multiple tumor-associated myeloid populations express MIF in human HNSCC (Supplementary Figure 4B). To determine whether MIF expression was associated with suppressive myeloid programs, we calculated an MDSC-related transcriptional score, and assessed its relationship with MIF expression within individual myeloid clusters. MIF expression positively correlated with MDSC scores in immunosuppressive myeloid cells (cluster 1; R = 0.22, p = 7.1 × 10⁻⁸) and activated interferon-responsive myeloid cells (cluster 4; R = 0.19, p = 8.3 × 10⁻⁵), indicating that elevated MIF expression is associated with suppressive myeloid transcriptional states (Supplementary Figure 4C). We next examined MIF expression in relation to response to immune checkpoint blockade. MIF expression was observed in both responders and non-responders and remained detectable following treatment, suggesting that MIF-associated myeloid programs persist during immune checkpoint blockade (Supplementary Figure 4D). Together, these analyses provide human validation of the association between MIF and suppressive myeloid programs and support the clinical relevance of the myeloid MIF pathway identified in our murine studies.

Discussion

HNSCC persists as one of the most immunotherapy-refractory cancers due to a deeply entrenched immunosuppressive TME dominated by MDSCs, regulatory CD4⁺ T-cell subsets, and dysfunctional CD8⁺ T cells. ^5-15 Although MIF has long been implicated in promoting immunosuppression, tumor growth, inflammation, and immune evasion across malignancies, ^16-22 the specific cellular source of tumor-promoting MIF in HNSCC, and the integrated immunoregulatory circuits it maintains, have remained poorly defined. The present work identifies myeloid-derived MIF as the central upstream driver of the MDSC–Treg–CD8 axis in HNSCC, positioning it as a keystone regulator of immune dysfunction in this disease.

Myeloid-specific deletion of MIF fundamentally reprogrammed the tumor immune landscape. Tumors in mMIF KO mice exhibited significantly reduced growth, consistent with prior reports that host-derived MIF, rather than tumor-intrinsic MIF, drives HNSCC progression. ^25-27 Importantly, this antitumor effect was observed despite substantial MIF expression by MOC2 tumor cells, ²⁶ highlighting the critical contribution of myeloid-derived MIF to tumor progression in this model. This antitumor effect was accompanied by a marked shift toward a pro-inflammatory, cytotoxic environment enriched in activated CD8⁺ T cells. These cells displayed a phenotype indicative of antigen engagement rather than terminal exhaustion, characterized by elevated PD-1 but diminished CTLA-4, TIGIT, and TIM-3, consistent with literature describing PD-1⁺ effector-competent CD8⁺ T cells in solid tumors. ^49-52 Their enhanced multifunctional cytotoxicity, demonstrated by increased co-expression of granzyme B, perforin, and IFN-γ, together with elevated CD107a expression and enhanced ex vivo cytotoxicity against MOC2 tumor cells, indicates that myeloid-derived MIF acts as a critical suppressive checkpoint constraining CD8⁺ effector activation in the HNSCC TME. Importantly, these effects were largely confined to the tumor microenvironment, as no significant differences were observed in the spleen or cervical lymph nodes, suggesting that myeloid-derived MIF primarily regulates the local tumor immune microenvironment.

The enhanced CD8⁺ T cell infiltration observed in mMIF KO tumors may be driven, at least in part, by alterations in chemokine-mediated T cell trafficking. Consistent with this possibility, mMIF KO tumors exhibited significantly increased expression of chemokines CCL5 and CXCL9, whereas CXCL10, CXCL11, and CXCR3 expression remained unchanged. Importantly, increased CCL5 and CXCL9 expression was also observed in our previous study using the MIF inhibitor CPSI-1306 in MOC2 tumors, indicating that both pharmacologic inhibition and myeloid-specific genetic deletion of MIF promote a similar chemokine signature. These findings suggest that suppression of the CCL5-CXCL9 axis may represent an important mechanism by which MIF restricts CD8⁺ T-cell infiltration and promotes immune evasion within the tumor microenvironment.

Previous studies in solid tumors have identified a cooperative role for tumor-derived CCL5 and IFN-γ-inducible CXCL9 produced by myeloid cells in promoting cytotoxic T-cell recruitment and establishing immunoreactive tumor microenvironments. High co-expression of CCL5 and CXCL9 has been associated with increased CD8⁺ T-cell infiltration, enhanced responsiveness to immune checkpoint blockade, and improved anti-tumor immunity in solid tumors. ^{53 , 54} In our model, increased CCL5 and CXCL9 expression was accompanied by elevated IFN-γ production, enhanced CD8⁺ T-cell accumulation, and increased expression of cytotoxic effector molecules, suggesting that loss of myeloid-derived MIF may promote a chemokine milieu favorable for CD8⁺ T cell recruitment and function. Collectively, these findings support a model in which myeloid-derived MIF contributes to immune exclusion, whereas its deletion promotes a more inflamed and immunologically active tumor microenvironment through mechanisms involving the CCL5-CXCL9 axis.

Consistent with these findings, deletion of myeloid MIF also caused a profound depletion of multiple CD4⁺ regulatory T-cell populations that maintain immune suppression in HNSCC. In addition to a significant reduction in classical FoxP3⁺ Tregs, we observed a collapse of IL-10–producing, IFN-γ–co-expressing Tr1-like CD4⁺ cells, a regulatory population increasingly recognized for its potent suppressive properties in tumor contexts. ^{45 , 55} These Tr1-like subsets suppress immunity through IL-10, IFN-γ–driven regulation, and perforin-dependent cytotoxicity, and our observed reduction in perforin within the IL-2⁺ IL-10⁺ compartment further supports functional impairment of this regulatory axis. ^{46 , 47} Because Tregs and Tr1-like cells both express high levels of MIF-responsive receptors such as CD74 and CXCR4, ^{34 , 56 , 57} their coordinated contraction in the absence of myeloid MIF underscores MIF's essential role in stabilizing multiple immunosuppressive CD4⁺ T cell programs.

Upstream of these T-cell changes was a striking depletion of polymorphonuclear MDSCs, particularly the highly suppressive CSF1R^hi and PD-L1^hi subsets implicated in immunosuppression and tumor progression. ^58-61 Given that PMN-MDSCs express MIF receptors, ²⁹ these findings suggest that MIF provides essential survival and programming cues for PMN-MDSC differentiation, maintenance, and suppressive activity. Prior reports demonstrating MIF's role in MDSC expansion and tolerogenic programming in cancer ^{32 , 33} align with our observation that myeloid MIF sustains the cellular and molecular architecture of suppressive myelopoiesis in HNSCC.

To extend these findings to human disease, we analyzed the published single-cell RNA sequencing dataset GSE232240 from patients with recurrent/metastatic HNSCC treated with immune checkpoint blockade. Consistent with our murine observations, MIF expression was broadly detected throughout the tumor-associated myeloid compartment, including immunosuppressive myeloid cells and activated interferon-responsive myeloid populations. Importantly, MIF expression positively correlated with an MDSC-associated transcriptional program within both populations, supporting a conserved relationship between MIF and suppressive myeloid biology in human HNSCC. While MIF expression was observed in both responders and non-responders to checkpoint blockade, its association with MDSC-related transcriptional programs was maintained, suggesting that MIF-associated suppressive myeloid states persist during immunotherapy. Although these analyses do not establish causality, they provide independent clinical support for our mechanistic findings in mice and reinforce the concept that MIF contributes to the maintenance of immunoregulatory myeloid programs in HNSCC. Collectively, these human data strengthen the translational relevance of our findings and support targeting the myeloid MIF axis as a strategy to disrupt MDSC-mediated immune suppression and improve antitumor immunity.

The ex vivo MDSC–T-cell co-culture assays confirm that MDSC-intrinsic MIF is required for both Treg induction and CD8 exhaustion under conditions that favor effector activation. WT MDSCs robustly induced FoxP3⁺ Tregs and upregulated exhaustion markers on CD8⁺ T cells under Treg-skewing and basal stimulation respectively, whereas mMIF KO MDSCs lacked this capacity. Under strongly tolerogenic TGF-β–rich conditions, CD8 exhaustion was maintained regardless of MIF status, highlighting that TGF-β can dominate over MIF in enforcing suppression. ⁶² These results position MIF as a context-dependent amplifier of MDSC suppressive programs: dispensable under strongly tolerogenic environments, but indispensable in settings where MDSCs must actively enforce immune suppression.

Together, these findings support a model in which myeloid-derived MIF functions as the architect of a multilayered immunosuppressive hierarchy. MIF sustains suppressive PMN-MDSC subsets, which in turn drive both FoxP3⁺ Treg expansion and IL-10⁺ IFN-γ⁺ Perforin⁺ Tr1-like regulatory programs. These regulatory CD4⁺ populations, together with MIF-licensed MDSCs, converge to enforce CD8⁺ exhaustion through checkpoint engagement, cytokine suppression, metabolic restriction, and perforin-mediated cytolysis of effector cells. Removal of MIF from the myeloid compartment collapses this architecture, restores effector function, and shifts the TME toward an antitumor state.

Although our data support a central role for the MDSC–Treg–CD8 axis in mediating the antitumor effects observed in mMIF KO mice, additional mechanisms downstream of myeloid-derived MIF may also contribute. Beyond its immunomodulatory functions, MIF has been implicated in tumor angiogenesis through the induction of pro-angiogenic mediators, including VEGF-A and IL-8, as well as through its effects on myeloid cell recruitment and activation. ⁶³ Myeloid populations, including MDSCs, have been shown to promote tumor neovascularization through the production of angiogenic factors and matrix-remodeling enzymes, thereby facilitating tumor progression. ⁶⁴ Given the reduction in PMN-MDSC populations observed in mMIF KO tumors, it is conceivable that loss of myeloid-derived MIF may also impair angiogenic remodeling within the tumor microenvironment in addition to alleviating immune suppression. Furthermore, angiogenesis and immune suppression are increasingly recognized as interconnected processes, with pro-angiogenic factors such as VEGF capable of impairing dendritic cell maturation, limiting effective T-cell priming, and fostering an immunosuppressive microenvironment. ^{64 , 65} Collectively, these observations suggest that the antitumor effects of myeloid-specific MIF deletion may extend beyond restoration of CD8⁺ T cell responses and involve concurrent disruption of protumor angiogenic pathways. Future studies are needed to define the contribution of angiogenic mechanisms to myeloid MIF-driven HNSCC progression. While this study provides compelling evidence for the centrality of myeloid-derived MIF, several limitations must be acknowledged. The LysM-Cre model, although widely used, affects macrophages, neutrophils, and subsets of dendritic cells; more lineage-refined approaches or MIF-rescue strategies would more precisely identify the myeloid subsets indispensable for mediating these effects. Furthermore, ex vivo co-culture systems cannot fully recapitulate the stromal, spatial, or metabolic interactions present within the TME. Future CD8⁺ T-cell depletion studies will also be important to establish a direct causal role for CD8⁺ T cells in mediating the reduced tumor growth observed following myeloid-specific MIF deletion. Finally, while prior work indicates that host-derived MIF is dominant over tumor-intrinsic MIF in HNSCC, ²⁷ future studies could expand this comparison using tumor-specific knockouts.

Despite these considerations, the therapeutic implications are substantial. Existing MIF inhibitors, including tautomerase inhibitors (ISO-1, 4-IPP, CPSI-1306), allosteric inhibitors (ibudilast), and monoclonal antibodies (Bax69/imalumab) have demonstrated antitumor activity in preclinical or early clinical settings, ^{22 , 66-71} but systemic inhibition is complicated by MIF's physiological roles. Our data provide a compelling rationale for myeloid-targeted MIF inhibition, such as delivery via CSF1R-targeted nanoparticles, ⁷² to dismantle the MDSC–Treg–CD8 axis at its source. In addition to reducing immunosuppressive myeloid and regulatory T-cell populations, myeloid-specific MIF deletion promoted a more T-cell-inflamed tumor microenvironment characterized by increased CD8⁺ T-cell infiltration, elevated expression of the T-cell-recruiting chemokines CCL5 and CXCL9, and enhanced cytotoxic function. Notably, although expression of multiple inhibitory receptors, including CTLA-4, TIGIT, and TIM-3, was reduced, PD-1 expression remained elevated on tumor-infiltrating CD8⁺ T cells of MOC2 tumors. These findings suggest that while myeloid MIF blockade may partially relieve immune suppression, the PD-1 axis may remain therapeutically relevant, providing a strong rationale for combining myeloid-targeted MIF inhibition with immune checkpoint blockade, including anti-PD-1 or anti-CTLA-4 therapy in future studies. By defining both the cellular origin and mechanistic impact of tumor-promoting MIF, this work establishes myeloid-derived MIF as a precise and highly targetable vulnerability in HNSCC and provides a mechanistic blueprint for therapeutic strategies aimed at reprogramming the TME and overcoming immunotherapy resistance.

Supplementary Material

Supplementary material

Supplemental figure 1.tif

Supplementary material

Supplemental figure 4.tif

Supplementary material

Supplemental figure 2.tif

Supplemental figure 3.tif

Supplemental figure 3.tif

Supplementary material

Author Checklist Oncoimmunology v2

Funding Statement

This work was funded by the National Institutes of Health grants R01DE033906, DP1DA054344 (NIH), and the American Cancer Society (ACS), grant RSG-19079-01-TBG awarded to SO.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Data availability statement

Flow cytometry data are available at Zenodo: https://zenodo.org/records/21384789. All other data that support the findings of this study are available from the corresponding author, [S.O], upon reasonable request.

Ethics statement

All animal studies were conducted in accordance with institutional and national guidelines and were approved by The Ohio State University Institutional Animal Care and Use Committee (IACUC) under Protocol number 2018A00000054. Research involving recombinant or biohazardous materials was reviewed and approved by The Ohio State University Institutional Biosafety Committee (IBC) under Protocol number 2018R00000038‑R1. This study adhered to the ARRIVE guidelines for reporting in vivo animal research.

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/2162402X.2026.2719293.