Targeting HK3 in tumor-associated macrophages enhances antitumor immunity through augmenting antigen cross-presentation in cervical cancer

Abstract

Background

Tumor-associated macrophages (TAMs) are among the most prevalent cells within the tumor microenvironment (TME) of cervical cancer (CC). Although TAMs frequently exhibit an immunosuppressive phenotype, their plasticity enables them as an intriguing reprogrammable target for immunotherapy of CC.

Methods

Consensus clustering was employed to delineate immune infiltration patterns in a cohort of 119 patients with CC. Single-cell RNA sequencing, complemented by flow cytometry analysis, was used to characterize hexokinase 3 (HK3)-expressing cell populations. In vivo tumor models were established to assess the functional impact of HK3-expressing cells on the TME, with interventions including Hk3 knockout and CD8⁺ T-cell depletion. A comprehensive approach involving bulk RNA sequencing, immunoprecipitation assays, confocal microscopy imaging, and in vitro co-culture systems was implemented to elucidate the mechanisms underlying HK3 inhibition-mediated enhancement of antitumor immunity. Furthermore, the therapeutic efficacy of HK3 inhibition, both as a monotherapy and in combination with immunotherapeutic strategies, was systematically evaluated in preclinical tumor models.

Results

We elucidated a cross-regulation between TAMs and CD8⁺ T cells, with HK3 serving as a central regulatory node. Upon HK3 expression was upregulated by CD8⁺ T cells through the IFN-γ-STAT1 signaling axis, TAMs exhibited impaired cross-presentation capacity, which in turn attenuated CD8⁺ T cell-mediated antitumor immunity. Mechanistically, HK3 physically interacted with mechanistic target of rapamycin (mTOR), promoting nuclear translocation of transcription factor EB (TFEB) and resulting in excessive lysosomal activation and antigen degradation. Moreover, targeting HK3 in combination with immune checkpoint blockade yielded a synergistic effect in enhancing antitumor immunity.

Conclusions

Targeting HK3 in TAMs represents a promising therapeutic strategy capable of enhancing antitumor immunity and synergizing with immune checkpoint blockade by restoring efficient antigen cross-presentation.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• Here, we identify hexokinase 3 (HK3) expression as a master role in regulating the interplay between CD8⁺ T cells and TAMs. We observed that HK3 in TAMs accelerates antigen destruction and suppresses antigen presentation, thereby dampening CD8⁺ T cell-mediated antitumor immunity. We provide a compelling rationale for targeting HK3 to overcome immunosuppression in the TME and improve the outcomes of cancer immunotherapy.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our findings suggest that targeting HK3 in TAMs enhances CD8⁺ T cell-mediated antitumor immunity and synergizes with immune checkpoint blockade by restoring efficient antigen cross-presentation.

Background

As the fourth most prevalent malignant disease among women worldwide, cervical cancer (CC) poses a substantial and significant public health challenge.¹ Despite the widespread availability of human papillomavirus (HPV) vaccination for prevention and extensive screening programs of high-risk HPV, the incidence of CC remains high, particularly in low- and middle-income countries. Moreover, although early-stage CC is often curable, advanced or recurrent cases still have a poor prognosis.² These challenges underscore the urgent need for innovative strategies to improve patient outcomes, particularly in preventing disease progression or recurrence.

Potentiating antitumor immunity by immunotherapy has been increasingly recognized as an important therapeutic modality for treating cancers including CC. Since successful antitumor response largely relies on tumoricidal immune cells, especially cytotoxic CD8⁺ T lymphocytes (CTLs), current immunotherapies mainly focus on activating/reinvigorating and/or replenishing CTLs. However, the therapeutic efficacy of these strategies remains limited, particularly in CC, and is therefore demanding improvement.³ Given that the activation and effector differentiation of CTLs require antigen cross-presentation by antigen-presenting cells (APCs),^{4 5} a process where APCs internalize extracellular antigens and present them to CTLs in the context of major histocompatibility complex (MHC) class I molecules, enhancing cross-presentation therefore holds promise for improving current immunotherapies.

In antitumor immunity, cross-presentation is typically performed by conventional dendritic cells (cDCs) in the tumor-draining lymph node (TDLN); however, recent evidence suggests that cross-presentation or cross-dressing in the tumor microenvironment (TME) is required for CTLs primed in TDLN to be restimulated followed by fully differentiating into cytotoxic effectors before exerting their tumoricidal activity.⁶ Intratumor immune cells that are capable of cross-presenting and cross-dressing include cDC1 and inflammatory monocytes/macrophages, among which the latter (tumor-associated macrophages, TAMs) is of special interest given its abundance in the TME.⁷ However, TAMs frequently exhibit functional impairment in antigen presentation and T cell activation. Therefore, reprogramming TAMs to restore their functionality has emerged as a promising strategy in boosting antitumor immunity for cancer immunotherapy. However, the mechanism by which cross-presentation is regulated in TAMs remains poorly understood, hindering the application of this therapeutic strategy.

In this study, we unveil a novel and pivotal role of hexokinase 3 (HK3), an isoenzyme of hexokinases that are responsible for the first and critical step in the glycolytic pathway by catalyzing the rate-limiting phosphorylation of glucose to form glucose-6-phosphate,⁸ in regulating the crosstalk between TAMs and CD8⁺ T cells. Beyond its well-known enzymatic activity, HK3 exhibits a previously unrecognized non-metabolic function through suppressing cross-presentation in TAMs. Importantly, targeting HK3 in TAMs restored anti-CC immunity and resulted in tumor regression and yielded a synergistic effect with immune checkpoint blockade in vivo. Targeting HK3 to enhance the cross-presentation capability of TAMs emerges as a novel and promising approach to improving the efficacy of immunotherapy.

Methods

Cell lines

Human CC cell line HeLa, human myeloid cell line THP-1 and mouse CC cell line TC-1 were obtained from the American Type Culture Collection. Cells were maintained at 37°C in a 5% humidified incubator using either Dulbecco’s Modified Eagle’s Medium (DMEM, 11965092, Gibco) or Roswell Park Memorial Institute (RPMI) -1640 medium containing 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 µg/mL streptomycin.

Mice and in vivo experiments

Hk3^−/− mice (C57BL/6JGpt-Hk3^em35Cd8063/Gpt) were bought from the Nanjing Biomedical Research Institute of Nanjing University (Nanjing, China). Female C57BL/6 mice aged 6 weeks were obtained from Bestcell (Wuhan, China). Mice were housed in the specific pathogen-free Animal Laboratory at the Laboratory Animal Center of Huazhong University of Science and Technology, with approval from the university’s Laboratory Animal Ethics Committee (ID: (2024) IACUC Number: 4176).

Mice were randomly assigned to each group (n=5 per group; total 90 mice sacrificed, including those for bone marrow-derived macrophages (BMDMs) generation). The randomization sequence was generated using a computer-based system. The sample size (n=5 per group) was chosen based on previous studies investigating similar interventions in mouse models. No animals were excluded, and all data were included in the analysis. Investigators performing outcome assessments and data analysis were blinded to group allocation, while personnel handling animal allocation and experimental procedures were unblinded. The tumor volume was calculated as (length×width²/2).

For the BMDMs and tumor cells co-implant mouse model, TC-1 cells were mixed with an equal number of wild-type (WT) BMDMs from WT C57BL/6 mice or HK3 knockout (Hk3-KO) BMDMs from Hk3^−/− mice. A total of 1×10⁶ cells in 100 µL were injected subcutaneously into C57BL/6 mice. Tumors were monitored every 2–3 days post-establishment.

For in vivo CD8⁺ T-cell depletion, in vivo anti-mouse CD8α blocking antibody (clone 2.43, Selleck) or phosphate buffered saline (PBS) was administered intraperitoneally (i.p. 100 µg/mouse) at days 1, 4, and 9 after tumor implantation. The effectiveness of the depletion was assessed by flow cytometry of splenocytes at day 14.

For in vivo programmed cell death protein-1 (PD-1) blockade experiments, in vivo anti-mouse PD-1 blocking antibody (clone RMP1-14, Selleck) or PBS was given i.p. (100 µg/mouse) at days 1, 4, and 9.

For metformin-pretreated BMDMs co-implanted with the TC-1 model, WT BMDMs were pretreated with metformin (HY-B0627, MCE) at 200 µM for 24 hours.

For metformin combined with PD-1 blockade experiments, in vivo anti-mouse PD-1 blocking antibody or PBS was given i.p. (100 µg/mouse) at days 5, 8, and 11. Metformin was given i.p. (50 mg/kg) at days 5, 7, 9, 11 and 13.

For generation of single-cell suspensions, harvested tumors were chopped into fragments and then digested with 0.1% Collagenase IV (abs47048003, absin), and 0.01% deoxyribonuclease (abs47047435, absin) at 37°C for 1 hour, then filtered through a 40 µm cell strainer.

Generation of BMDMs

Bone marrow cells were extracted by dissecting femurs and tibias of mice aged 4–6 weeks. The cells were then cultured for 7 days in DMEM containing 10% FBS, 100 U/mL penicillin, 100 µg/mL streptomycin, plus 20 ng/mL M-CSF (HY-P7085, MCE) to promote macrophage differentiation.

Plasmids, lentiviral construction, and cell transfection

HK3 shRNA lentivirus and control lentivirus were constructed using the hU6-MCS-CBh-gcGFP-IRES-puromycin vector by GeneChem (Shanghai, China). The sequences are detailed in online supplemental file 1. HK3 overexpression lentivirus and vector lentivirus were synthesized using the pcSLenti-EF1-EGFP-P2A-Puro-CMV-MCS-3xFLAG-WPRE vector by ObiO Technology (Shanghai, China). The stably expressing cells were selected using puromycin (HY-B1743, MCE).

Flow cytometry

Cells were stained with Fixable Viability Stain (564406 or 564997, BD). CD16/CD32 monoclonal antibody (101320, BioLegend) was used to block the Fc gamma receptor for 15 min before immunostaining. The cells were subsequently incubated with suitable dilutions of different antibody combinations as outlined in online supplemental table S3. For intracellular staining, cells were fixed and permeabilized with a Fixation/Permeabilization Kit (554714, BD). Of note, for detection of interferon (IFN)-γ and tumor necrosis factor (TNF)-α, cells were first activated with Cell Activation Cocktail (with Brefeldin A) (423304, BioLegend) at 37°C for 5 hours. Cells were collected using a Beckman CytoFlex S Flow Cytometer, and the data were analyzed with FlowJo.

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)

RNA was extracted with an isolation kit (RC101-01, Vazyme) as per the manufacturer’s instructions, then reverse transcribed into complementary DNA with HiScript III RT SuperMix for qPCR (+gDNA wiper) (R323-01, Vazyme), and subsequently analyzed with the SYBR Green PCR Master Mix (Q711-02, Vazyme). The expression of target genes was normalized to β-actin and calculated using the comparative CT method (2^−ΔΔCT). The primers used are detailed in online supplemental table S2.

Transwell assay

To assess the migratory capability of macrophages, phorbol 12-myristate 13-acetate (PMA)-treated THP-1 cells (5×10⁴) or BMDMs (2×10⁵) were added to the upper chamber (3,422, Corning). After 24 hours, non-migrating cells were discarded, and the cells at the chamber’s bottom were fixed using paraformaldehyde, stained with crystal violet, and inspected through an inverted microscope.

Phagocytosis

Prior to the assay, target cells HeLa or TC-1 were labeled with CellTrace Violet (CTV, C34571, Invitrogen) for 20 min at 37°C. Next, CTV-labeled tumor cells were added to PMA-treated THP-1 or BMDMs at a 5:1 ratio and cultured at 37°C for 6 hours. Phagocytosis of macrophages was analyzed using flow cytometry.

Co-culture assay

Mouse naïve CD8⁺ T cells and CD4⁺ T cells from spleen were extracted with an isolation kit (480044 or 480005, BioLegend). BMDMs were pulsed with ovalbumin (OVA) (100 µg/mL, vac-stova, InvivoGen) overnight in advance. With or without fixation with 0.05% paraformaldehyde, BMDMs were co-cultured with isolated T cells for the indicated time at a ratio of 5:1. The culture medium is RPMI 1640 containing 10% FBS, 1% penicillin/streptomycin, 50 µM 2-mercaptoethanol, 50 IU/mL rhIL2 and anti-CD28 antibodies (5 µg/mL, clone 37.51, BE0015-1, Bio X Cell).

T-cell cytotoxicity assay

Naïve CD8⁺ T cells from OTI mice co-cultured with BMDMs for 3 days were incubated with TC-1-OVA cells for 6 hours at a ratio of 5:1. The cytotoxicity capacity was measured by Annexin V-EGFP/PI Apoptosis Detection Kit (KTA0005, Abbkine)

IFN-γ blockade by antibody

Mouse CD8⁺ T cells from spleen were stimulated with PMA and ionomycin (00–4970, eBioscience) for 6 hours and then co-cultured with BMDMs for 24 hours. During co-culture, anti-IFN-γ neutralizing antibody (100 µg/mL, clone R4-6A2, A2105, Selleck) was added.

Immunohistochemistry and immunofluorescence

A section from the paraffin block was deparaffinized and rehydrated, followed by boiling for 15 min in an antigen retrieval buffer (G1203, Servicebio). Once cooled, sections were treated with 5% bovine serum albumin (BSA) for 1 hour. Overnight incubation of the sections with the designated antibodies was done at 4°C. For immunohistochemistry, the HRP-DAB polymer kit was used for visualization (G1212, Servicebio). For immunofluorescence, the ABflo 488-conjugated goat anti-rabbit IgG (H+L) or ABflo 594-conjugated goat anti-mouse IgG (H+L) (ABclonal) was used as secondary antibodies. The sections were stained using DAPI (HY-D0814, MCE) for 10 min to label nuclei.

RNA-sequencing

Transcriptome sequencing was conducted using Illumina HiSeq by Sangon (Sangon Biotech, China). RNA-sequencing (RNA-seq) reads were aligned with HISAT2 (V.2.2.1) with the default parameters to the Mus musculus reference genome Mus_musculus.GRCm38. Transcripts per million were used for normalization. R (V.4.2.3) package DESeq2 (V.1.38.0) was employed for differentially-expressed genes (DEGs). DEGs were defined as genes with a p value<0.05 and log2 (fold change) >1.5.

Immune stratification through consensus clustering

To classify samples into immune high-infiltration and low-infiltration groups, immune cell abundance was first quantified using single-sample Gene Set Enrichment Analysis (ssGSEA) based on a curated gene set of 28 immune cell types.^{9 10} The resulting ssGSEA scores were then applied to perform consensus clustering using the R package ConsensusClusterPlus, with the following parameters: 1,000 iterations, pItem=0.8, pFeature=0.8, and Pearson correlation as the distance metric.¹¹ The consensus matrix for k=2 demonstrated a clear separation between the two clusters. Stromal, Immune, and Estimate scores were calculated using the ESTIMATE algorithm (R package “estimate”). Specifically, the Stromal score (reflecting stromal cell abundance), the Immune score (indicating immune cell infiltration), and the combined Estimate score (sum of Stromal and Immune scores), which collectively provide a comprehensive evaluation of the TME composition. In the heatmap, the color legend represents z-score normalized ssGSEA enrichment scores across samples. A z-score of 0 indicates the mean enrichment level, while positive (red) and negative (steel blue) values denote higher or lower immune cell abundance compared with the mean, respectively.

Degradation by DQ-OVA

Macrophages were treated with 50 µg/mL of DQ-OVA (D-12053; Invitrogen) at 37°C for 30 min or otherwise indicated time. The fluorescence of DQ-OVA was obtained throughflow cytometry.

Immunoprecipitation

Cells were lysed for 30 min on ice in immunoprecipitation lysis solution (G2038, Servicebio) containing protease inhibitor cocktail (HY-K0010, MCE). The clear supernatants were incubated with Protein A/G Magnetic Beads (HY-K0202, MCE) and rotated at 4°C for 2 hours. The precipitates were later subjected to western blot (WB) analysis.

Western blotting

Cells were lysed using cell lysis buffer (G2002, Servicebio) containing protease inhibitor cocktail, followed by boiling in SDS-PAGE loading buffer (HY-K1100, MCE). After electrophoresis, the gel was transferred to polyvinylidene fluoride (PVDF) membranes (1620177, Bio-Rad). Following a 5% BSA block, the membranes were left overnight at 4°C with primary antibodies. Subsequently, species-matched secondary antibodies were applied for 1 hour at 37°C. The bands were visualized with an electrogenerated chemiluminescence Kit (HY-K1005, MCE).

Bioinformatics

The Cancer Genome Atlas – Cervical Squamous Cell Carcinoma and Endocervical Adenocarcinoma (TCGA-CESC) data were acquired from the UCSC Xena platform (http://xena.ucsc.edu/). The bulk transcriptional data covering normal to CIN to CC tumor tissues can be found under accession number GSE63514. Processed single-cell RNA sequencing (scRNA-seq) data of human CC can be found at Science Data Bank (https://doi.org/10.57760/sciencedb.11624). For immunotherapy datasets, IMvigor 210 data and its clinical characteristics were gathered from the IMvigor210CoreBiologies R package. The other immune checkpoint blockade (ICB) data were acquired from GSE91061, GSE165252, and PRJEB23709, all acquired from the TIGER database (http://tiger.canceromics.org). Stat1 ChIP-Seq data were downloaded from GSE43036.

Statistical analysis

GraphPad Prism (V.9, San Diego, California, USA) was employed for all statistical analyses. Bar graphs were displayed as the mean±SEM. If the data distribution was normal and variances were equal, mean differences were assessed using Student’s t-test or a one-way analysis of variance. If not, the Wilcoxon non-parametric test was used. The statistical significance was established at p<0.05.

Results

HK3 expression correlates positively with tumor immune infiltration but predicts poor clinical prognosis

Given the profound influence of immune infiltrations on the prognosis of cancers including CC,¹² we sought to explore biomarkers that may contribute to or represent different status of immune infiltrations. To this end, we first extracted gene signatures of 28 distinct immune cell types from a previous study.⁹ Using these signatures, we calculated ssGSEA scores based on RNA-seq data from our CC cohort (n=119).¹⁰ The raw bulk RNA-seq data reported has been deposited in the Genome Sequence Archive in the National Genomics Data Center.¹³ Through consensus clustering analysis,¹¹ we stratified patients with CC into two distinct clusters, namely Immune_H (high immune infiltration) and Immune_L (low immune infiltration) (figure 1A). To identify differential pathway activities between clusters characterized by varying levels of immune infiltration, we performed Gene Set Variation Analysis (GSVA) using 50 Hallmark gene sets from the Molecular Signatures Database (MsigDB).^{14 15} Comparative analysis revealed significant downregulation of 12 pathways in the Immune_H cluster versus Immune_L cluster (figure 1B). Given the well-established role of aerobic glycolysis (the “Warburg effect”) in tumor progression and immune modulation,¹⁶ we focused particularly on the glycolysis pathway, which ranked in the top three. Within the glycolysis pathway, HK3 emerged as the most differentially expressed glycolytic gene between the two clusters (figure 1C). Moreover, immunohistochemical staining of CC tumor tissues confirmed a strong correlation between HK3 and CD8A at the protein level (figure 1D). This association was further corroborated in the TCGA-CESC dataset (online supplemental figure S1A–D). Collectively, these findings establish HK3 as a potential biomarker associated with enhanced immune infiltration in CC.

Figure 1. HK3 expression correlates positively with tumor immune infiltration but predicts poor clinical prognosis. (A) Identification of immune clusters in patients with cervical cancer (CC). Heatmap depicts the stratification of 119 patients with CC into two distinct immune clusters based on consensus clustering analysis. Rows and columns represent z-score normalized enrichment scores of 28 immune cell-specific gene signatures and samples, respectively. Estimate score, Stromal score, and Immune score were calculated using the ESTIMATE algorithm. (B) Downregulated pathways in the Immune_H compared with Immune_L clusters, as identified by Gene Set Variation Analysis (GSVA) using Molecular Signatures Database Hallmark gene sets. (C) Heatmap illustrating differential expression levels of glycolysis-related genes between the two identified clusters. Rows and columns represent z-score normalized gene expression (TPM) and samples, respectively. (D) Representative immunohistochemical images showing CD8 and HK3 expression in CC tumor tissues, respectively (left). Correlation analysis of HK3 and CD8 area fraction in IHC staining results of 30 CC tumor tissues (right). Immune stratification strategy was illustrated in figure 1A. Scale bar, 50 µm; r, Spearman correlation coefficient; p, p values for Spearman correlation. (E) Kaplan-Meier curve for progression-free survival based on HK3 expression in the CC cohort. The optimal cut-point was determined using the Survminer R package. (F) HK3 expression in tumor tissues versus paired adjacent non-tumor tissues (ADJ) in the CC cohort (RNA sequencing data). (G) HK3 expression profile across normal, cervical intraepithelial neoplasia (CIN), and tumor tissues in the GSE63514 CC dataset. (H) Objective response rate to programmed death-ligand 1 blockade between HK3^high and HK3^low groups (optimal prognostic survival cut-off) in the IMvigor 210 cohort. (I) Kaplan-Meier curve for overall survival based on HK3 expression in the IMvigor 210 cohort. The optimal cut-point was determined using the Survminer R package. All values are expressed as the mean±SEM. *p<0.05, **p<0.01, ***p<0.001 and not significant (ns) by Student’s t-test, one-way analysis of variance test or χ² test. CR, complete response; FC, fold change; HK3, hexokinase 3; Immune_H, high immune infiltration; Immune_L, low immune infiltration; LACC, locally advanced cervical cancer; PD, progressive disease; PR, partial response; SD, stable disease; TPM, transcripts per million.

Surprisingly, despite the observed positive association between HK3 expression and enhanced immune infiltration, our survival analysis instead revealed that HK3 predicts worse clinical outcomes. Specifically, patients with elevated HK3 expression exhibited inferior progression-free survival (figure 1E). Additionally, comparative evaluation of tumor and adjacent non-tumor tissues showed markedly higher HK3 expression in malignant tissues (figure 1F). Longitudinal analysis using the GSE63514 dataset further revealed a consistent and gradual upregulation of HK3 expression throughout cervical carcinogenesis from normal tissues to cervical intraepithelial neoplasia (CIN), and ultimately to invasive carcinoma (figure 1G).¹⁷ To further explore the clinical implications of HK3, we evaluated its predictive value in immunotherapy response. Analysis of the IMvigor 210 cohort,¹⁸ comprising patients with metastatic urothelial cancer treated with programmed death-ligand 1 (PD-L1) inhibitor atezolizumab, demonstrated that lower HK3 expression was associated with prolonged overall survival and a more favorable clinical response (figure 1H,I). These findings collectively reveal a complex, seemingly paradoxical role of HK3 in cancer immunology: while positively correlated with immune cell infiltration, it associates with poorer prognosis and diminished response to immunotherapy. This complexity suggests that HK3 may function as a critical modulator of tumor-immune interactions, potentially influencing both the TME and therapeutic outcomes.

HK3 is predominantly expressed in immunosuppressive macrophages

To delineate the cellular sources of HK3 within the TME, we reanalyzed a single-cell RNA sequencing data of CC and found that HK3 was almost exclusively expressed in monocytes and macrophages (figure 2A).¹⁹ In contrast, tumor cells typically exhibit high expression of hexokinase isoforms HK1 and HK2, but low expression of HK3, whereas macrophages show a distinct metabolic profile characterized by high HK3 expression (online supplemental figure S2A–C). Next, TC-1 cells, a widely used murine cell line generated by HPV-16 E6/E7 transformation of primary lung epithelial cells and exhibiting characteristic features of HPV-associated tumors on inoculated into syngenic mice, was used to examine the expression profile of HK3 in mouse CC. Indeed, the HK3 expression pattern was corroborated in TC-1 tumor tissues, where myeloid cells exhibited significantly higher HK3 expression compared with other cellular compartments (figure 2B). It is important to note that TAMs display substantial heterogeneity, containing both pro-inflammatory and anti-inflammatory subsets for which CD206 is generally considered as an anti-inflammatory and protumoral marker.²⁰ Therefore, we next examined the expression profile of HK3 in macrophage subsets, finding elevated HK3 expression in CD206⁺ macrophages compared with their CD206⁻ counterparts (figure 2C). Moreover, in the single-cell dataset, HK3^high macrophages exhibited upregulated expression of canonical immunosuppressive markers (CD163, MRC1, MS4A4A, TREM2), the fibrosis-related gene FN1 and the angiogenesis-associated modulator ACE compared with HK3^low macrophages (online supplemental figure S2D). This finding was further supported by colocalization of HK3 and CD163 (another immunosuppressive marker) in human CC tissues (online supplemental figure S2E).

Figure 2. HK3 is predominantly expressed in immunosuppressive macrophages. (A) UMAP visualization of identified cell types in cervical cancer tumor tissues in the single-cell dataset (left). UMAP plot displaying HK3 expression distribution (right). (B) Representative histograms and gMFI of HK3 in the indicated cell types of TC-1 tumors at day 14 after inoculation. (C) Representative histograms and gMFI of HK3 in the macrophages of TC-1 tumors at day 14 after inoculation. (D) Expression of selected genes analyzed by RT-qPCR in NC and HK3-KD THP-1 cells after PMA treatment (20 ng/mL, 48 hours). (E) Representative histograms and gMFI of CD206 in WT and Hk3-KO BMDMs with or without IL-4 stimulation (20 ng/mL, 24 hours). (F) Representative images of migrated cells of WT and HK3-KD PMA-treated THP-1 assayed by transwell assay for 12 hours (left). Quantification of migrated cells per field (right). Scale bar, 50 µm. (G) Percentages of PMA-treated THP-1 cells phagocytosing CTV-labeled HeLa cells (left) and BMDMs phagocytosing CTV-labeled TC-1 cells (right). All values are expressed as the mean±SEM. *p<0.05, **p<0.01, ***p<0.001 and not significant (ns) by Student’s t-test or one-way analysis of variance test. BMDMs, bone marrow-derived macrophages; CTV, CellTrace Violet; gMFI, geometric mean fluorescence intensity; HK3, hexokinase 3; HK3-KD, HK3 knockdown; Hk3-KO, HK3 knockout; IL, interleukin; Mφ, macrophages; PMA, phorbol 12-myristate 13-acetate; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; UMAP, Uniform Manifold Approximation and Projection; WT, wild-type.

The predominant expression of HK3 in immunosuppressive macrophages, and the inverse correlation between HK3 expression and clinical outcomes together suggest its protumoral function in the TME. To test this, we used shRNA lentivirus to knock down HK3 in THP-1 cells, a widely accepted human monocytic cell line that recapitulates key aspects of macrophage differentiation and function on stimulation. The efficiency of the knockdown in THP-1 cells treated with PMA is shown (online supplemental figure S2F). Knockdown of HK3 led to increased gene expression of immunostimulatory markers (CD80, CD86, and NOS2), and decreased expression of CD163 (figure 2D), suggesting a shift towards an immunostimulatory phenotype. Moreover, BMDMs derived from Hk3^−/− mice showed a reduced level of CD206 expression following treatment with interleukin-4 (IL-4), along with an enhanced level of iNOS expression following treatment with IFN-γ (figure 2E, online supplemental figure S2G), indicating a diminished potential for transition into an immunosuppressive phenotype compared with HK3-sufficient counterparts. Further functional assessment revealed that HK3 deficiency not only altered the phenotype of macrophages but also affected their biological behavior. Both HK3-knockdown and Hk3-KO macrophages exhibited reduced migratory capacity but enhanced phagocytic activity toward tumor cells (figure 2F and G, online supplemental figure S2H), which was in line with previous research on immunostimulatory or pro-inflammatory macrophages.^{21 22} STAT1 and STAT6 are well-recognized as key transcription factors that mediate the pro-inflammatory and anti-inflammatory phenotypes of macrophages, respectively.^{23 24} Indeed, WT macrophages showed decreased phosphorylated STAT1 and increased phosphorylated STAT6 compared with Hk3-KO macrophages (online supplemental figure S2I,J). To conclude, these findings provide compelling evidence that HK3 mediates the immunosuppressive phenotype of macrophages.

HK3 in TAMs compromises CD8⁺ T cell-mediated antitumor immunity

To investigate the effect of macrophages expressing different levels of HK3 on tumor progression, we established a co-implantation mouse model by inoculating mice with equal numbers of either WT or Hk3-KO BMDMs with tumor cells (figure 3A). Tumors co-implanted with Hk3-KO BMDMs demonstrated significantly reduced growth compared with WT BMDM controls (figure 3B and C), accompanied by decreased proliferation (Ki67 staining, figure 3D) and increased apoptosis (TUNEL assay, figure 3E). When T cell infiltration was examined, we observed enhanced CD8⁺ T-cell infiltration in tumors with Hk3-KO BMDMs and no significant change in CD4⁺ T cells (figure 3F). Furthermore, Hk3 deficiency in BMDMs promoted CD8⁺ T cell activation, as evidenced by increased production of IFN-γ, TNF-α, and granzyme B (figure 3G and H), while CD4⁺ T cell activation remained unaltered (online supplemental figure S3B,C). We then eliminated CD8⁺ T cells in this model with a CD8-depleting antibody to verify the participation of CD8⁺ T cells in tumor regression mediated by Hk3 deficiency in BMDMs (figure 3I). Efficiency of depletion was confirmed (online supplemental figure S3E). Notably, CD8 depletion reversed the tumor-suppressive effect of BMDMs induced by Hk3 deficiency, as comparing the KO anti-CD8 group to the KO PBS group (figure 3J), suggesting that the antitumor effect of Hk3 deficiency is dependent on CD8⁺ T cells. Thus, the findings from our in vivo experiments indicate that HK3 in macrophages exerts protumor effects through the inhibition of CD8⁺ T-cell infiltration and activation, thereby facilitating tumor growth and progression. Consistently, naïve CD8⁺ T cells co-cultured with Hk3-KO BMDMs in vitro expressed higher levels of IFN-γ and TNF-α compared with those with WT BMDMs (figure 3K), while no significant difference in the activation of CD4⁺ T cells was observed in vitro (online supplemental figure S3D). These results collectively demonstrate that HK3-expressing macrophage promotes tumor progression by specifically impairing CD8⁺ T cell-mediated antitumor immunity, rather than affecting CD4⁺ T-cell function. This HK3-dependent immunomodulatory mechanism suggests that HK3 could be a potential therapeutic target in TAMs for enhancing antitumor immune response.

Figure 3. HK3 in tumor-associated macrophages compromises CD8⁺ T cell-mediated antitumor immunity. (A) Schema of the co-injection model using FigDraw. TC-1 cells were co-implanted with either WT or Hk3-KO BMDMs at a 1:1 ratio subcutaneously into C57BL/6 mice. Tumors were resected for measurement at day 14 after inoculation. (B) Tumor growth after tumor inoculation (n=5 per group). (C) Tumor weight at day 14 after inoculation (n=5 per group). (D) Representative immunohistochemical images and quantification of Ki67 expression in tumor tissues. Scale bar, 50 µm. (E) Representative TUNEL staining of tumor tissues and quantification of TUNEL-positive nuclei (blue: DAPI, green: TUNEL; scale bar: 50 µm). (F) Percentages of tumor-infiltrating CD4⁺ or CD8⁺ T cells out of living cells at day 14. (G) Percentages of IFN-γ⁺CD8⁺ T cells, TNF-α⁺CD8⁺ T cells, GzmB⁺CD8⁺ T cells at day 14. (H) The gMFI of IFN-γ, TNF-α and GzmB in CD8⁺ T cells at day 14. (I) Experimental design for CD8⁺ T-cell depletion. Anti-CD8 neutralizing antibody (100 µg/mouse) was administered intraperitoneally at days 1, 4, and 9. (J) Tumor growth following CD8⁺ T-cell depletion (n=5 per group). (K) Percentages of IFN-γ or TNF-α-producing OT-I CD8⁺ T cells co-cultured with the OVA-loaded BMDMs. All values are expressed as the mean±SEM. *p<0.05, **p<0.01, ***p<0.001 and not significant (ns) by Student’s t-test or one-way analysis of variance test. BMDMs, bone marrow-derived macrophages; gMFI, geometric mean fluorescence intensity; GzmB, granzyme B; HK3, hexokinase 3; HK3-KO, HK3 knockout; IFN, interferon; i.p., intraperitoneally; PBS, phosphate buffered saline; TNF, tumor necrosis factor; WT, wild-type.

HK3 inhibition enhances cross-presentation by attenuating lysosomal antigen degradation

To further investigate the molecular mechanisms underlying the antitumor effects mediated by HK3-deficient macrophages, we conducted comprehensive pathway analysis. Stratification of macrophages based on HK3 expression levels revealed significant enrichment of lysosomal and antigen processing pathways in HK3^high macrophages (figure 4A). Notably, the enrichment of the “antigen processing and presentation” pathway in HK3^high macrophages is largely attributed to the upregulation of lysosomal genes such as CTSB and CTSL. Although these genes are grouped under antigen processing and presentation in pathway analysis, their primary role is antigen degradation and processing, rather than the actual presentation of antigenic peptides. Comparative analysis between WT and Hk3-deficient mouse BMDMs, as well as PMA-treated human THP-1 cells, consistently demonstrated enhanced phagosome and antigen-processing pathways in WT macrophages (figure 4B, online supplemental figure S4A). Based on these results, and the findings that HK3 in macrophages affects the activation and cytotoxic function of CD8⁺ T cells (figure 3), we therefore hypothesized that HK3 might inhibit the capacity of macrophages to cross-present exogenous antigens to CD8⁺ T cells. Since fine-tuned lysosome-mediated and proteosome-mediated antigen processing are needed for successful cross-presentation,²⁵ we then compared HK3^high and HK3^low macrophages for their expression of lysosome-related and proteosome-related genes. Interestingly, HK3^high macrophages exhibited upregulation of lysosomal markers (LAMP1, LAMP2, CTSL, CTSB) and downregulation of proteasomal components (PSME1, PSMB8, PSMB9) (figure 4C). These results were recapitulated in the WT BMDMs versus Hk3-KO BMDMs comparison (figure 4D, online supplemental figure S4B,C), suggesting that HK3-expressing macrophages preferably rely on lysosomes rather than proteasomes to process antigens.

Figure 4. HK3 inhibition enhances cross-presentation by attenuating lysosomal antigen degradation. (A) KEGG enrichment analysis of marker genes in HK3^high versus HK3^low macrophages (median cut-off) in the cervical cancer single-cell dataset. (B) Top 20 upregulated KEGG pathways in WT versus Hk3-KO BMDMs. (C) Dot plots showing the expression of selected genes from major histocompatibility complex class I antigen processing signature in HK3^high and HK3^low macrophages. Dot size indicates the percentage of cells expressing each gene, and color intensity represents the relative gene expression level. (D) RT-qPCR showing the mRNA expression of Lamp1 and Lamp2 in WT and Hk3-KO BMDMs. (E) The gMFI of DQ-OVA in BMDMs at indicated time points post-treatment. (F) The gMFI of DQ-OVA in BMDMs pretreated with or without CQ for 30 min. (G) The gMFI of H-2K^b-SIINFEKL complexes in OVA-loaded BMDMs following 24-hour treatment with either CQ or MG132. (H) Experimental design for the co-culture of naïve CD8 T cells derived from OT-I mice with OVA-loaded and PFA-fixed BMDMs using FigDraw. (I) Percentages of CD25⁺ or CD69⁺ OT-I cells following 24-hour co-culture with the OVA-loaded and PFA-fixed BMDMs. (J) Representative plots and percentages of IFN-γ-producing OT-I T cells following 72-hour co-culture with the OVA-loaded and PFA-fixed BMDMs. (K) Quantification of Annexin V⁺ TC-1-OVA cells following 6-hour co-culture with OT-I T cells that were pre-co-cultured with OVA-loaded and PFA-fixed BMDMs. All values are expressed as the mean±SEM. *p<0.05, **p<0.01, ***p<0.001 not significant (ns) by Student’s t-test or one-way analysis of variance test. BMDMs, bone marrow-derived macrophages; CQ, chloroquine; gMFI, geometric mean fluorescence intensity; HK3, hexokinase 3; HK3-KO, HK3 knockout; IFN, interferon; Mφ, macrophages; KEGG, Kyoto Encyclopedia of Genes and Genomes; mRNA, messenger RNA; OVA, ovalbumin; PFA, paraformaldehyde; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; WT, wild-type.

To test this premise, we used DQ-OVA as an exogenous model antigen to treat BMDMs and found that WT BMDMs display heightened antigen degradation than the Hk3-KO counterparts in as early as 5 min after DQ-OVA treatment (figure 4E). Of note, degraded antigen (green fluorescence) was mainly co-localized with lysosome (red fluorescence), which is consistent with the reported predominant role of lysosome in processing exogenous antigens (online supplemental figure S4D). This differential degradation was abolished by the lysosome inhibitor chloroquine (CQ) (figure 4F), confirming the critical role of lysosomal activity in HK3-mediated antigen degradation. It has been established that excessive lysosomal degradation results in antigen destruction, thereby leading to impaired antigen presentation.^{26 27} Indeed, we observed that Hk3-KO resulted in augmented cross-presentation of the short peptide derived from OVA onto MHC class I molecules, as evidenced by upregulation of the SIINFEKL-H2k^b complex on the surfaces of macrophages pulsed with OVA. Importantly, this difference disappeared if macrophages were pretreated with CQ but not proteasome inhibitor MG132 (figure 4G). Next, to exclude the confounding effect of soluble factors secreted by macrophages, OVA-loaded BMDMs were fixed with paraformaldehyde and then used to stimulate naïve CD8⁺ T cells from OT-I mice (figure 4H). Again, Hk3-KO in macrophages significantly augmented the activation and cytotoxicity of co-cultured OT-I T cells, as revealed by the upregulation of activation-related molecules CD25, CD69, and cytotoxicity-related molecules IFN-γ and granzyme B, along with enhanced apoptosis of tumor cells (figure 4I–K, online supplemental figure S4E,F). Thus, these findings collectively demonstrate that HK3 impairs antigen cross-presentation in macrophages through accelerating lysosomal antigen degradation, thereby attenuating CD8⁺ T cell-mediated antitumor immunity.

HK3 induces excessive lysosomal activation through promoting TFEB nuclear translocation after binding to mTOR

To explore how HK3 contributes to the acceleration of lysosomal degradation, we first examined the number and acidification of lysosomes. Compared with WT BMDMs, Hk3-KO BMDMs exhibited a significant reduction in lysosomal number (figure 5A) and an increase in lysosomal pH, as indicated by LysoSensor Green DND-189 (figure 5B). We then assessed the expression of key lysosomal genes, including Atp6v1h (encoding ATP6V1H, a subunit of the proton pump V-ATPase that acidifies lysosome), and Ctsl (encoding the cysteine proteases cathepsin L, one of the most abundant and functionally significant lysosomal proteinases). Hk3-KO led to reduced transcription of Atp6v1h and Ctsl, as well as decreased protein levels (figure 5C, online supplemental figure S4H). Given transcription factor EB’s (TFEB’s) role as a master regulator of lysosomal biogenesis and function, we next examined if TFEB is responsible for Hk3-KO-mediated changes in lysosome. Hk3-KO significantly reduced nuclear TFEB levels, whereas HK3 overexpression enhanced TFEB nuclear translocation (figure 5D, online supplemental figure S4G). Therefore, we conclude that HK3 is essential for maintaining lysosomal activation and acidification through affecting TFEB. Activation and nuclear translocation of TFEB are negatively controlled by mechanistic target of rapamycin (mTOR)-dependent serine phosphorylation.²⁸ Previous study reported that HK2 could bind to mTOR and inhibit its activity.²⁹ Since HK3 and HK2 belong to the same family known as hexokinase, we hypothesized that HK3 might similarly regulate mTOR signaling. Indeed, we detected an interaction between HK3 and mTOR by co-immunoprecipitation (figure 5E). Furthermore, Hk3-KO resulted in a nearly twofold increase in mTOR phosphorylation (figure 5F), indicating that HK3 is capable of interacting with mTOR, thereby negatively regulating its activity and potentially influencing downstream signaling pathways. Finally, to determine whether the HK3-mTOR-TFEB axis influences lysosomal function and cross-presentation, we treated Hk3-KO BMDMs with rapamycin (an mTOR inhibitor) and TFEB activator 1 (an activator of TFEB nuclear translocation) to inhibit mTOR and activate TFEB nuclear translocation, respectively. Both interventions reversed the inhibited lysosomal gene expression and enhanced cross-presentation observed in Hk3-KO BMDMs (figure 5G–I, online supplemental figures S4H and S5A,B), confirming that HK3 modulates lysosomal activation and cross-presentation through mTOR-TFEB signaling.

Figure 5. HK3 induces excessive lysosomal activation through promoting TFEB nuclear translocation after binding to mTOR. (A) Representative confocal images of BMDMs. Hoechst 33342 (blue) and LysoTracker Red (red) were used for cell nuclei and lysosome labeling, repectively (left). Scale bars, 5 µm. Representative histograms and gMFI of LysoTracker in BMDMs (right). (B) Representative histograms and gMFI of LysoSensor Green DND-189 in BMDMs. (C) RT-qPCR showing the mRNA expression of Atp6v1h and Ctsl in WT and Hk3-KO BMDMs. (D) Representative confocal images of BMDMs. DAPI (blue) and TFEB (red) were used for cell nuclei and TFEB labeling, repectively (left). Western blotting of cytoplasmic and nuclear TFEB in WT and Hk3-KO BMDMs (right). (E) Co-immunoprecipitation for the interaction between Flag-tagged HK3 and mTOR in HK3-OE HEK293T cells.(F) Representative histograms and gMFI of phospho-mTOR (Ser2448) in BMDMs. (G) Representative histograms and gMFI of LAMP1 in BMDMs following 24-hour treatment with either rapamycin (Rapa, 25 µM) or TFEB activator 1 (TA1, 2 µM). (H) The gMFI of H-2K^b-SIINFEKL complexes in OVA-loaded BMDMs following 24-hour treatment with either Rapa or TA1. (I) Percentages of IFN-γ-producing OT-I T cells following 72-hour co-culture with the OVA-loaded and then paraformaldehyde-fixed BMDMs pretreated with Rapa, TA1 and 2-deoxy-D-glucose (2-DG, 1 mM). All values are expressed as the mean±SEM. *p<0.05, **p<0.01, ***p<0.001 and not significant (ns) by Student’s t-test or one-way analysis of variance test. BMDMs, bone marrow-derived macrophages; gMFI, geometric mean fluorescence intensity; HK3, hexokinase 3; HK3-KO, HK3 knockout; HK3-OE, HK3 overexpressed; IFN, interferon; IP, immunoprecipitation; mRNA, messenger RNA; mTOR, mechanistic target of rapamycin; OVA, ovalbumin; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; TFEB, transcription factor EB; WT, wild-type.

As HK3 is a glycolytic enzyme involved in glycolysis, we therefore investigated whether its metabolic activity contributes to cross-presentation. Unexpectedly, HK3 knockdown resulted in increased hexokinase specific activity and lactate production (online supplemental figure S5C,DH), indicating a compensatory upregulation of glycolysis in response to HK3 deficiency. Nevertheless, inhibition of hexokinase activity with 2-deoxy-D-glucose failed to accelerate antigen degradation (online supplemental figure S5E) or downregulate cross-presentation in Hk3-KO BMDMs (figure 5I, online supplemental figure S5A,B), indicating that the regulation of cross-presentation in macrophages by HK3 is independent of its metabolic activity. To conclude, these results together demonstrate that HK3 accelerates antigen degradation and modulates cross-presentation by interacting with mTOR and promoting TFEB nuclear translocation. This regulatory mechanism is decoupled from HK3’s role in glycolysis, highlighting its multifaceted functions in cellular metabolism and immune regulation.

HK3 is regulated by the IFN-γ-STAT1 signaling axis

TAMs are highly susceptible to various factors in the TME, which can readily reprogram their function and phenotype.³⁰ To determine factors regulating HK3 expression, we stratified TCGA-CESC samples into two groups based on median HK3 expression levels. GSVA using 50 MsigDB Hallmark gene sets revealed a significant enrichment of the “INTERFERON_GAMMA_RESPONSE” pathway in HK3^high samples compared with HK3^low samples (figure 6A). This finding was further supported by a positive correlation between HK3 and IFNG expression (figure 6B). Therefore, we speculated that HK3 is regulated by IFN-γ. To validate this, we exposed BMDMs to IFN-γ, lipopolysaccharide (LPS) or IL-4 for 24 hours. In line with expectations, IFN-γ but not LPS and IL-4 treatment promoted the transcription of Hk3 (figure 6C). Furthermore, HK3 was enhanced by IFN-γ in a concentration-dependent manner both at transcription and protein levels (figure 6D and E), accompanied by increased number and acidification of lysosomes (online supplemental figure S6F,G), as well as accelerated DQ-OVA degradation (online supplemental figure S6H). Therefore, these results confirm that IFN-γ is a potent inducer of HK3 expression in macrophages.

Figure 6. HK3 is modulated by the IFN-γ-STAT1 signaling axis. (A) Upregulated pathways in HK3^high versus HK3^low samples (median cut-off) in the TCGA-CESC dataset using GSVA with Molecular Signatures Database Hallmark gene sets. (B) Correlation between HK3 and IFNG in TCGA-CESC dataset. r, Spearman correlation coefficient; p, p values for Spearman correlation. (C) RT-qPCR showing the mRNA expression of Hk3 in BMDMs treated with IFN-γ (20 ng/mL), LPS (100 ng/mL) or IL-4 (20 ng/mL) for 24 hours. (D) Dose-dependent effects of IFN-γ on Hk3 mRNA expression in BMDMs. RT-qPCR analysis was performed after 24-hour treatment with indicated concentrations of IFN-γ (0, 20, 50, 100 ng/mL). (E) Representative histograms and gMFI of HK3 in BMDMs treated with indicated concentrations of IFN-γ for 24 hours. (F) Experimental design for IFN-γ blockade. Splenic CD8⁺T cells were activated with PMA and ionomycin for 6 hours, followed by 24-hour co-culture with BMDMs in the presence of IFN-γ neutralizing antibody. (G) Representative histograms and gMFI of HK3 in BMDMs following 24-hour co-culture with CD8⁺T cells in the presence of IFN-γ neutralizing antibody. (H) Representative histograms and gMFI of HK3 in BMDMs treated for 6 hours with IFN-γ (20 ng/mL), JAK1/2 inhibitors ruxolitinib (1 µM) or STAT1 inhibitors fludarabine (1 µM). All values are expressed as the mean±SEM. *p<0.05, **p<0.01, ***p<0.001 and not significant (ns) by Student’s t-test or one-way analysis of variance test. BMDMs, bone marrow-derived macrophages; FACs, Fluorescence-Activated Cell Sorting; gMFI, geometric mean fluorescence intensity; GSVA, Gene Set Variation Analysis; HK3, hexokinase 3; IFN, interferon; IL, interleukin; LPS, lipopolysaccharide; mRNA, messenger RNA; PMA, phorbol 12-myristate 13-acetate; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; TCGA-CESC, The Cancer Genome Atlas – Cervical Squamous Cell Carcinoma and Endocervical Adenocarcinoma.

Given that activated CD8⁺ T cells are a major source of IFN-γ in the TME, we investigated whether they could induce HK3 expression in BMDMs. As expected, CD8⁺ T cells activated by PMA and Ionomycin increased HK3 expression in BMDMs, and this effect was abolished by adding anti-IFN-γ neutralizing antibody (figure 6F and G), demonstrating the capacity of activated CD8⁺ T cells to promote HK3 expression in macrophages through IFN-γ secretion. IFN-γ signaling is known to activate the JAK-STAT1 pathway, leading to the transcription of IFN-stimulated genes.³¹ To determine whether this pathway regulates HK3, we treated BMDMs with a JAK1/2 inhibitor (ruxolitinib) or a STAT1 inhibitor (fludarabine). Both inhibitors significantly suppressed HK3 expression (figure 6H), suggesting a role of the IFN-γ-STAT1 signaling axis in regulating HK3. Furthermore, chromatin immunoprecipitation sequencing (ChIP-Seq) analysis from GSE43036 identified HK3 as a direct target gene of STAT1 in CD14⁺ monocytes treated with macrophage colony-stimulating factor (M-CSF) (online supplemental figure S6I).³² Together, our results indicate that HK3 is an IFN-γ-stimulated gene regulated by the IFN-γ-STAT1 signaling axis. Activated CD8⁺ T cells in the TME secrete IFN-γ, which promotes HK3 expression in macrophages, thereby enhancing lysosomal function and antigen degradation. This regulatory mechanism highlights the interplay between CD8⁺ T cells and macrophages.

Targeting HK3 in TAMs synergizes with immune checkpoint blockade in CC

Given that HK3 impairs the antitumor function of CD8⁺ T cells, it is reasonable to hypothesize that HK3 might influence the efficacy of ICB therapy. In line with this hypothesis, patients with high HK3 expression have a markedly worse clinical outcome compared with those with low HK3 expression, regardless of their CD8⁺ T-cell infiltration levels (online supplemental figure S7A). Moreover, analysis of multiple immunotherapy datasets revealed that HK3 expression is significantly elevated during ICB treatment compared with baseline levels (figure 7A).33,35 We then sought to test whether targeting HK3 could enhance the efficacy of ICB therapy in the TC-1/BMDMs co-implantation mouse model (figure 7B). The combinatorial therapy resulted in superior tumor suppression compared with anti-PD-1 or Hk3-KO monotherapy (figure 7C, (online supplemental figure S7B). Thus, our results suggest that macrophage-specific HK3 deletion could be combined with ICB therapy to unleash the optimal therapeutic potential.

Figure 7. Targeting HK3 in tumor-associated macrophages synergizes with immune checkpoint blockade in cervical cancer. (A) Expression of HK3 in multiple immunotherapy datasets before and during immune checkpoint blockade therapy. (B) Treatment schema of anti-PD-1 antibody using FigDraw. Anti-PD-1 antibody (100 µg/mouse) was administered intraperitoneally at days 1, 4, and 9 after inoculation. (C) Tumor growth after inoculation (n=5 per group). (D) Representative histograms and gMFI of phospho-STAT1 (Tyr701) in BMDMs treated with or without metformin (Met, 200 µM) for 6 hours. (E) The gMFI of HK3 in BMDMs treated with indicated concentrations of Met for 24 hours. (F) The gMFI of LysoTracker in BMDMs treated with or without Met (200 µM) for 24 hours. (G) Tumor growth after inoculation (n=5 per group). BMDMs were pretreated with or without Met (200 µM) for 24 hours before co-implantation with TC-1 cells. (H) Treatment schema of anti-PD-1 antibody and metformin using FigDraw. (I) Tumor growth after inoculation (n=5 per group). (J) Representative histograms and gMFI of HK3 in tumor-associated macrophages. All values are expressed as the mean±SEM. *p<0.05, **p<0.01, ***p<0.001 and not significant (ns) by Student’s t-test or one-way analysis of variance test. BMDMs, bone marrow-derived macrophages; gMFI, geometric mean fluorescence intensity; HK3, hexokinase 3; HK3-KO, HK3 knockout; i.p., intraperitoneally; PBS, phosphate buffered saline; PD-1, programmed cell death protein-1; WT, wild-type.

Although no specific inhibitors for HK3 are currently available, metformin has been reported to inhibit HK3 expression in tumor cells.³⁶ To determine whether metformin could similarly inhibit HK3 in macrophages, we treated BMDMs with metformin. As expected, metformin treatment effectively decreased phosphorylated STAT1 and thus HK3 expression in BMDMs (figure 7D and E). Consistent with the role of HK3 in lysosomal regulation, metformin treatment also led to a reduction in lysosome number in BMDMs (figure 7F). Importantly, in vitro pretreatment of BMDMs with metformin suppressed tumor growth in the TC-1/BMDMs co-implantation mouse model (figure 7G, online supplemental figure S7D). Similarly, the combination of metformin with anti-PD-1 antibodies also achieved a synergistic tumor-suppressive effect, outperforming either treatment alone (figure 7H and I, online supplemental figure S7E). Consistent with in vitro results, metformin treatment indeed reduced the HK3 expression in TAMs in vivo (figure 7J), further supporting its role as an HK3 inhibitor in the TME. Collectively, these findings indicate that targeting HK3 synergizes with anti-PD-1 antibodies in CC, and metformin could serve as an HK3 inhibitor for reprogramming macrophages to suppress tumor growth and boost ICB therapy.

Discussion

TAMs are well-documented to promote tumor progression through multiple mechanisms, including epithelial to mesenchymal transition, angiogenesis, matrix remodeling and immunomodulation.³⁷ However, accumulating evidence from numerous studies has demonstrated the feasibility and therapeutic potential of TAM reprogramming as a strategy to enhance antitumor immunity.^{38 39} Our findings significantly contribute to this emerging field by identifying HK3 as a critical regulator of TAM function. We have demonstrated that elevated HK3 expression in TAMs compromises their antigen cross-presentation capability, thereby attenuating their ability to activate CD8⁺ T cell-mediated antitumor immunity. This discovery not only reveals a novel mechanism by which TAMs maintain their immunosuppressive phenotype but also establishes HK3 as a promising therapeutic target for TAM reprogramming.

Recent progress has revolutionized our understanding of antigen presentation in tumor immunity. According to the traditional concept, antigen presentation and/or cross-presentation occur in the second lymphoid tissues, predominantly in TDLN for solid tumors, where dendritic cells migrate from the tumor bed present tumor peptide antigens in the context of MHC II and MHC I molecules, respectively, to CD4⁺ and CD8⁺ T cells, resulting in the activation and differentiation of these cells.⁴⁰ Activated T cells then migrate from TDLN into the tumor bed under the guidance of chemo-attractive cues to perform their effector function. However, recent investigations have demonstrated that after entering into the tumor, CD8⁺ T cells primed in the TDLN require further restimulation by intratumoral myeloid cells, and in some cases need license provided by CD4⁺ T cells engaged on the same myeloid cells to acquire cytotoxic capacity.^{41 42} Indeed, such multicellular networks were reported to be associated with favorable patient response to immunotherapies with ICB and adoptive transfer of tumor-infiltrating lymphocytes, highlighting the therapeutic potential of harnessing these myeloid cells.⁴³ Initially, cDC1 was considered the principal intratumoral myeloid cells with cross-presentation/cross-dressing capacity, while recent studies suggest that inflammatory monocytes and CXCL9⁺CXCL10⁺ TAMs were armed with this capacity as well.⁷ It is important to note that although cross-presenting TAMs were documented, most TAMs still lack this capacity and are often functionally impaired.⁴⁴ Therefore, our finding that immunosuppressive TAMs can be reprogrammed to possess the capability of cross-presenting and activating antigen-specific CD8⁺ T cells adds a new weapon with sufficient supply to the existing arsenal of antitumor cells, therefore may shed new light on future immunotherapies.

Efficient antigen presentation relies on optimal lysosomal activity, and dysregulated lysosomal activity can have detrimental effects on the antigen process and presentation. Specifically, hypoactivity of lysosomes suppresses the generation of antigenic peptides, while hyperactivity results in the destruction of putative epitopes before they can be effectively presented to T cells.^{26 45} Studies have confirmed the antigen-destroying properties of immunosuppressive TAMs,⁴⁶ yet the precise underlying mechanisms for this phenomenon are not thoroughly comprehended. Our study provides significant mechanistic insights into this phenomenon by identifying HK3 as a key regulator of lysosomal function in TAMs. We demonstrate that HK3 modulates lysosomal activity through a novel molecular pathway involving its interaction with mTOR. This HK3-mTOR axis facilitates the nuclear translocation of TFEB, a master regulator of lysosomal biogenesis and function. The resulting lysosomal overactivation is characterized by two distinct features: increased lysosomal quantity and enhanced acidification, both of which contribute to excessive antigen degradation. Genetic ablation of HK3 in TAMs significantly attenuates their lysosomal activity, creating a more favorable environment for antigens to escape destruction and subsequently be cross-presented to CD8⁺ T cells. These findings not only elucidate a previously unrecognized mechanism of immune evasion in the TME but also establish HK3 as a critical molecular switch controlling the balance between antigen preservation and degradation in TAMs. By targeting the HK3-mediated pathway, we may be able to modulate the antigen-presenting capacity of TAMs, potentially converting them from immunosuppressive actors to effective antigen-presenting cells that can stimulate robust antitumor T-cell immunity.

Glycolysis is a highly conserved metabolic process that plays a fundamental role in modulating cellular metabolism across diverse cells types.⁴⁷ This pathway not only provides ATP to meet cellular energy demands but also generates essential metabolic intermediates for biosynthetic processes.⁴⁸ At the core of glycolysis regulation lies the hexokinase (HK) enzyme family, which catalyzes the initial and rate-limiting step of glucose metabolism through the ATP-dependent phosphorylation of glucose to glucose-6-phosphate.⁸ Mammalian cells contain four distinct isoforms of HKs: HK1, HK2, HK3, and HK4 (which is often referred to as glucokinase). While HK1 and HK2 are ubiquitously expressed across various cell types, the less well-studied HK3 exhibits a more restricted expression profile, predominantly in cells of myeloid origin.⁴⁹ Our findings align with and extend this understanding by demonstrating that HK3 expression is specifically localized to TAMs within CC tissues. This specific expression pattern of HK3 in TAMs underscores its unique and crucial role beyond its well-established function in regulating cellular metabolism. Moreover, HK3^high TAMs exhibit immunosuppressive phenotypes, which are marked by elevated expression of CD206 and increased phosphorylation of STAT6. These phenotypic changes suggest that HK3^high TAMs contribute to the establishment of an immunosuppressive TME, potentially through multiple mechanisms including the suppression of antitumor immune responses.

Intriguingly, our analysis of the local CC cohort, complemented by the TCGA-CESC dataset, has unveiled a striking positive correlation between HK3 expression and immune infiltration. These results can be explained by two different mechanisms, either HK3-expressing TAMs promote the infiltration of CD8⁺ T cells, or activated CD8⁺ T cells induce HK3 expression in TAMs. Several lines of evidence provided by us support the latter mechanism. First, exposure to IFN-γ, which could be secreted by activated CD8⁺ T cells, markedly increased HK3 expression in BMDMs. Second, activated CD8⁺ T cells were capable of inducing HK3 expression and this effect could be reversed by adding anti-IFN-γ neutralizing antibody. Third, blockade of the JAK-STAT1 pathway, which mediates the signal transduction on IFN-γ exposure, with either JAK1/2 inhibitors or STAT1 inhibitors could counteract IFN-γ-induced HK3 expression in BMDMs. These findings together imply the role of the CD8⁺ T-IFN-γ-STAT1 signaling axis in enhancing HK3 transcription. Finally, considering the immunosuppressive phenotype exhibited by HK3-expressing TAMs, we reason that the upregulation of HK3 is likely not a primary driving force of immune infiltration, but rather a secondary response to the recruitment and activation of immune cells within the TME. It is worth noting that although the IFN-γ-STAT1 signaling axis has been initially considered as the major driver in activating macrophages including TAMs, its role in inducing TAMs to express immunoinhibitory molecules including PD-L1 and BTN3A1 has also been reported.⁵⁰ The dual role of IFN-γ in the polarization of macrophages warrants further exploration to determine which end of the spectrum it induces macrophages to tilt toward. Nevertheless, the regulatory relationship between HK3 expression in TAMs and immune cell infiltration reported by us can offer fresh understanding of the process through which T cells subvert TAMs to facilitate tumor progression.

In summary, our findings revealed closely intertwined negative feedback regulatory crosstalk between TAMs and CD8⁺ T cells, in which HK3 serves as a critical regulatory molecule. TAMs with low HK3 expression have the capability to efficiently stimulate and activate CD8⁺ T cells through enhanced cross-presentation. However, once activated, CD8⁺ T cells secrete IFN-γ, which in turn modulates TAMs to limit the exacerbation of inflammation. This modulation is achieved through the upregulation of HK3 in TAMs, which subsequently accelerates antigen destruction and suppresses antigen presentation, thereby dampening CD8⁺ T cell-mediated antitumor immunity. Consistent with this proposed model, our experimental data demonstrate that inhibition of HK3 in macrophages significantly suppresses tumor growth and synergistically enhances the therapeutic efficacy of ICB therapy. These findings not only highlight the pivotal role of HK3 in regulating the TAM-CD8⁺ T-cell axis but also provide a compelling rationale for targeting HK3 as a promising strategy to overcome immunosuppression in the TME and improve the outcomes of cancer immunotherapy.

Our present study does have some limitations. First, although we used BMDMs derived from Hk3^−/− mice and co-implanted them with TC-1 tumor cells to establish a mouse tumor model, which effectively elucidated the role of HK3 in TAMs within the TME, the use of a macrophage-specific Hk3 conditional KO mouse model could further enhance the precision and reliability of these findings. Second, given the unavailability of a specific HK3 inhibitor, we employ metformin to suppress HK3 expression in macrophages. Nevertheless, this approach does not rule out other potential antitumor mechanisms of metformin. Further research is needed to develop agents specifically targeting HK3 in macrophages.

Data availability statement

Data are available in a public, open access repository.