Transketolase-like 1 potentiates PD-1 blockade in hepatocellular carcinoma by glycolysis to prime dendritic cell lactylation

Hengsong Cao; Tian Huang; Yue Chai; Zhengqing Lu; Shipeng Dai; Danyang Xu; Yanshu He; Na Yu; Yuliang Wang; Ziyuan Liu; Yidong Xia; Jie Zhao; Hongyu Wang; Dawei Rong; Yuchen Xie; Chongyu Wang; Pengyu Chen; Juyue Zhou; Li Liu; Zhuozheng Li; Hanyuan Liu; Xiao Li; Xuehao Wang; Weiwei Tang; Yongxiang Xia

doi:10.1038/s41392-026-02875-2

. 2026 Sep 28;11:418. doi: 10.1038/s41392-026-02875-2

Transketolase-like 1 potentiates PD-1 blockade in hepatocellular carcinoma by glycolysis to prime dendritic cell lactylation

Hengsong Cao ^1,^2,^#, Tian Huang ^3,^#, Yue Chai ^1,^#, Zhengqing Lu ^1,^#, Shipeng Dai ^1,^#, Danyang Xu ⁴, Yanshu He ¹, Na Yu ⁴, Yuliang Wang ⁵, Ziyuan Liu ¹, Yidong Xia ¹, Jie Zhao ¹, Hongyu Wang ¹, Dawei Rong ¹, Yuchen Xie ¹, Chongyu Wang ¹, Pengyu Chen ¹, Juyue Zhou ⁶, Li Liu ⁶, Zhuozheng Li ⁷, Hanyuan Liu ³, Xiao Li ², Xuehao Wang ^1,^✉, Weiwei Tang ^1,^✉, Yongxiang Xia ^1,^✉

PMCID: PMC13616917 PMID: 42802226

Abstract

Hepatocellular carcinoma (HCC) exhibits a suboptimal response to immune checkpoint blockade (ICB) therapy; to overcome this resistance, we aimed to delineate key immune resistance factors via multi-omics analysis, develop strategies to block their immunosuppressive axes, and engineer a targeted nanosystem to enhance immunotherapy efficacy against PD-1 resistance in HCC. Using transcriptomic and proteomic data from anti-PD-1-treated HCC patients, along with functional validation in murine models and mechanistic molecular and cell biology studies, we identified transketolase-like 1 (TKTL1) as a dual-nature biomarker where overexpression predicted poor baseline prognosis yet enhanced response to ICB. Mechanistically, TKTL1 diverts glucose flux into glycolysis rather than pentose phosphate pathway (PPP), recruiting USP9X to deubiquitinate and stabilize HIF-1α, which upregulates HK2 to amplify glycolytic output and lactate accumulation. This metabolic rewiring orchestrates dual immunosuppressive circuits through HIF-1α-driven CCL4 secretion recruiting PD-L1^high dendritic cells (DCs), coupled with lactate-induced TRIM28^K408 lactylation that stabilizes PD-L1 by blocking ubiquitin-mediated degradation. We engineered a hepatoma-membrane-coated MnO₂ nanosystem (CQLH) co-delivering a TKTL1 inhibitor and lactate oxidase, which disrupted the TKTL1-HIF-1α-HK2 axis, depleted lactate, and reprogrammed the tumor microenvironment, thereby enhanced anti-PD-1 therapy to suppress tumor growth, especially in TKTL1^high tumors. These findings define a critical “TKTL1-glycolysis-lactate-DC” axis driving anti-PD-1 sensitivity in HCC, position TKTL1 as both a potential biomarker for ICB response and a tractable therapeutic target, and demonstrate that the targeted CQLH nanosystem overcomes resistance and enhances anti-PD-1 efficacy, offering a precision immunotherapeutic strategy for TKTL1^high HCC.

graphic file with name 41392_2026_2875_Figa_HTML.webp

Subject terms: Cancer metabolism, Cancer therapy

Introduction

Hepatocellular carcinoma (HCC) ranks as the sixth most common malignancy and third leading cause of cancer-related death worldwide, accounting for approximately 90% of primary liver cancer cases and imposing a devastating global health burden, particularly in Asia and sub-Saharan Africa. The rising incidence of HCC is closely linked to the high prevalence of chronic hepatitis B and C virus infections, excessive alcohol consumption, and the rapidly growing epidemic of non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH) globally.¹ While tyrosine kinase inhibitors (TKIs) like sorafenib and lenvatinib remain important backbone options for first-line systemic therapy for advanced HCC, their clinical benefits are limited by modest survival improvements, frequent dose-limiting adverse events, and inevitable acquired resistance that develops within months of treatment initiation.² In recent years, immune checkpoint inhibitors (ICIs) have emerged as a transformative approach that has revolutionized cancer treatment by reactivating the host’s endogenous anti-tumor immune response. Notably, the combination of atezolizumab (anti-PD-L1) plus bevacizumab became the first regimen in over a decade to significantly surpass sorafenib in overall survival, extending median overall survival from 13.4 months to 19.2 months and marking a new era in HCC systemic therapy.³ However, despite these remarkable advances, our phase II clinical trial (NCT04297202) demonstrated that perioperative camrelizumab (anti-PD-1) combined with apatinib only achieved an objective response rate of 33.3% in resectable HCC patients, with a substantial proportion of patients experiencing disease progression within 12 months,⁴ highlighting the critical need to elucidate the underlying mechanisms of ICI resistance and develop more effective therapeutic strategies.

The development of resistance to ICIs in HCC is underpinned by a complex and multifaceted interplay of mechanisms operating at both the tumor cell and immune system levels. These include the establishment of an immunosuppressive tumor microenvironment (TME) characterized by abundant regulatory T cells, M2-polarized tumor-associated macrophages, and myeloid-derived suppressor cells that secrete anti-inflammatory cytokines such as IL-10 and TGF-β, impaired antigen presentation capabilities of dendritic cells, significant intratumoral genetic and epigenetic heterogeneity, and alterations in gut microbiota composition that modulate systemic immune function.⁵ Clinically, ICI resistance is broadly categorized into three distinct types: primary resistance (patients do not respond to initial treatment), adaptive resistance (tumors evade immune attack after an initial transient response), and acquired resistance (tumors progress after a period of sustained clinical benefit).⁶ A critically important and understudied factor contributing to primary resistance is target deficiency, where the loss or downregulation of key ligands such as PD-L1 or PD-L2 on tumor cells has been shown to underlie 60% to 85% of ICI treatment failures in various solid tumors.⁷ Although PD-L1 immunohistochemistry has been validated as a predictive biomarker for ICI response in non-small cell lung cancer,⁸ breast cancer,⁹ and esophageal carcinoma,¹⁰ its utility in HCC remains controversial and unvalidated due to high intratumoral heterogeneity, dynamic regulation of PD-L1 expression, and inconsistent results across different clinical trials. Thus, defining the molecular drivers of PD-1/PD-L1 deficiency and identifying reliable predictive biomarkers are urgent clinical priorities.

Building on integrated multi-omics data (RNA sequencing and proteomic profiling) from our prospective clinical trial (NCT04297202),⁴ we identify transketolase-like 1 (TKTL1) as a novel metabolic master regulator of anti-PD-1 response in HCC. Metabolic reprogramming is a hallmark of cancer, and accumulating evidence suggests that altered tumor metabolism plays a central role in shaping the immunosuppressive TME and mediating ICI resistance. TKTL1 is an isoform of transketolase, a key enzyme in glucose metabolism that controls the non-oxidative branch of the pentose phosphate pathway (PPP) and is frequently overexpressed in various human cancers, where it is associated with poor prognosis.^11,12 However, contrary to previous reports, we found that TKTL1 promotes HCC progression and modulates immunotherapy sensitivity primarily through the glycolytic pathway rather than the PPP. Mechanistically, multi-omics analysis revealed that TKTL1 recruits the deubiquitinase USP9X to stabilize hypoxia-inducible factor-1α (HIF-1α), which in turn drives hexokinase 2 (HK2)-mediated glycolytic flux and lactate accumulation while inducing HIF-1α/CCL4-dependent recruitment of dendritic cells (DCs) to the TME. Notably, accumulated lactate further triggers TRIM28 K408 lactylation in DCs, which stabilizes PD-L1 protein expression and establishes a local immunosuppressive state. To therapeutically target this newly identified axis, we engineered a biomimetic nanoplatform with excellent biocompatibility and tumor-targeting ability that co-delivers a novel TKTL1 inhibitor (Quercetin 3-glc-7-gent) and lactate oxidase, synergistically disrupting both TKTL1/HIF-1α signaling and lactate-driven PD-L1 stabilization for precision HCC immunotherapy.

Collectively, our study uncovers a previously unrecognized metabolic-immune crosstalk axis centered on TKTL1 that governs PD-1 blockade response in HCC. We demonstrate that TKTL1-driven glycolysis not only fuels tumor cell proliferation and survival but also orchestrates DC-mediated immunosuppression through a previously undescribed lactylation-dependent mechanism. These findings establish TKTL1 as a novel predictive biomarker for ICI efficacy and a promising therapeutic target to overcome immunotherapy resistance in HCC. Furthermore, the biomimetic nanoplatform we developed provides a rational combinatorial strategy that simultaneously targets tumor metabolic reprogramming and immune suppression, offering a new avenue for the development of precision immunotherapies for HCC and potentially other glycolysis-dependent cancers.

Results

Integrated multi-omics profiling unveils TKTL1 upregulation in anti-PD-1 responders in HCC

Leveraging NCT04297202 trial HCC anti-PD-1 patient biospecimens, we performed integrated RNA sequencing (RNA-seq) and proteomic profiling (Fig. 1a). Comparative analysis revealed distinct transcriptional/proteomic signatures between responders and non-responders via heatmaps and volcano plots (Fig. 1b, c). Transcriptomics identified upregulated AQP3, RDH5, TKTL1 and downregulated HBB, NCF1, CYBA in responders (Fig. 1b). Proteomics showed elevated TMED5, BUB3, TKTL1 and reduced ACSM2A, CD99, PIN4 in responders (Fig. 1b). Cross-omics integration highlighted TKTL1 RNA/protein concordant upregulation in responders (Fig. 1c), prompting focus. Given TKTL1’s metabolic role, pathway enrichment was performed. RNA KEGG revealed responders-enriched chemokine signaling, antigen presentation, and metabolic pathways; GO emphasized immune regulation (Fig. 1d). Proteomic KEGG specifically highlighted glycolysis/gluconeogenesis activation, corroborated by GO glucose catabolism enrichment (Fig. 1e). Multi-omics findings indicate heightened glycolytic activity coupled with immune activation signatures in anti-PD-1 sensitive HCC patients.

To investigate TKTL1 in HCC, we performed immunofluorescence staining on a tissue microarray containing 98 paired tumor and adjacent tissues. The analysis revealed elevated TKTL1 expression in tumor tissues (Fig. 1f). Based on expression levels, 70 cases were classified as TKTL1-high and 28 as TKTL1-low. Analysis of The Cancer Genome Atlas (TCGA) data demonstrated that high TKTL1 expression was associated with worse overall survival (OS) (Fig. 1g). Consistently, our cohort data also indicated that high TKTL1 expression correlated with poorer OS and disease-free survival (DFS) (Fig. 1h). To clarify the cellular expression pattern of TKTL1 in hepatocellular carcinoma (HCC), we analyzed single-cell RNA sequencing (scRNA-seq) data from dataset GSE146115. Five major cell clusters were identified using canonical gene markers: non-tumor hepatocytes, lymphocytes, malignant cluster 1, malignant cluster 2 and myeloid cells (Supplementary Fig. 1a–c). TKTL1 was exclusively highly expressed in malignant cluster 2 but barely detectable in other cell populations (Supplementary Fig. 1b–d). Immunofluorescence co-localization further validated that TKTL1 was restricted to HCC tumor cell (Supplementary Fig. 1e), consistent with the scRNA-seq results. To further validate TKTL1 expression at the protein level, we performed immunofluorescence staining on HCC patient samples from anti-PD-1 therapy responders and non-responders. The results revealed that responders exhibited more pronounced tumor necrosis and higher TKTL1 protein expression intensity (Fig. 1i). Both TKTL1 RNA and protein levels were elevated in responders compared to non-responders (Fig. 1j, k). In summary, TKTL1 overexpression predicts a poor prognosis in treatment-naive HCC but is associated with increased sensitivity to anti-PD-1 therapy.

TKTL1 drives tumor progression while potentiating anti-PD-1 immunotherapy sensitivity rather than apatinib in HCC

To extend beyond correlative analyses of TKTL1 in clinical specimens, we systematically investigated its functional impact on HCC cells. Western blotting screening confirmed that HCC-LM3 and Hep-3B cell lines exhibited relatively high and low endogenous expression of TKTL1, respectively (supplementary Fig. 2a), selected as model systems. Three TKTL1-specific short hairpin RNAs (sh-TKTL1) were designed, with western blot confirming target-selective knockdown without affecting other TKT family members (TKT and TKTL2) (supplementary Fig. 2a). Sh1 demonstrating superior silencing efficacy (supplementary Fig. 2a) was employed for subsequent experiments. First of all, overexpression of TKTL1 in HCC-LM3 and Hep-3B cells did not significantly alter their sensitivity to apatinib (supplementary Fig. 2b). EdU incorporation and colony formation revealed that TKTL1 knockdown significantly suppressed HCC cell proliferation (supplementary Fig. 2c–f) along with impaired migratory and invasive capacities (supplementary Fig. 2g–j). Conversely, TKTL1 overexpression exacerbated these oncogenic phenotypes (supplementary Fig. 2c–j).

To assess TKTL1 immunomodulation, we established C57BL/6 subcutaneous tumor models with anti-PD-1 (supplementary Fig. 3a). TKTL1-overexpressing tumors showed accelerated growth but enhanced anti-PD-1 response versus controls (supplementary Fig. 3b, c). Flow cytometry revealed TKTL1 promoted CD8⁺ T cell exhaustion markers (PD-1, TOX, CTLA-4) and suppressed IFN-γ-reversed by anti-PD-1 (supplementary Fig. 3d, e), linking TKTL1 to PD-1-mediated evasion. An orthotopic HCC model (Fig. 2a) confirmed accelerated TKTL1-overexpressing tumor growth, suppressed by anti-PD-1 from day 7 (Fig. 2b, c). Combination therapy enhanced necrosis, reduced Ki67⁺ cells, and improved CD8⁺ effector function versus monotherapies (Fig. 2d–i, supplementary Fig. 3f). In a chemical-induced HCC model, TKTL1 adeno-associated virus + anti-PD-1 (Fig. 2j) synergistically regressed tumors via magnetic resonance imaging (MRI) (Fig. 2k–p). Flow cytometry and immunofluorescence confirmed TKTL1 overexpression combined with anti-PD-1 attenuated CD8⁺ exhaustion and boosted cytotoxicity (Fig. 2q, supplementary Fig. 3g). Collectively, models demonstrate TKTL1 drives HCC aggressiveness but creates a PD-1-blockade-susceptible niche (Fig. 2r).

Fig. 2 — TKTL1 drives tumor progression while potentiating anti-PD-1 immunotherapy sensitivity rather than apatinib in HCC. a Schematic of the orthotopic tumor mouse model: Hepa1-6 cells transduced with OE-TKTL1 (TKTL1 overexpression) or OE-VE (vector control) lentiviruses were subcutaneously implanted, followed by tumor transplantation to the liver. Anti-PD-1 treatment was administered, with live imaging at day 21 and sample harvest at day 42. b Tumor growth curves of orthotopic models in OE-VE, OE-TKTL1, OE-VE+Anti-PD-1, and OE-TKTL1+Anti-PD-1 groups (days 21, 28, 35). c Quantification of tumor volume and tumor weight in each treatment group. Data are mean ± SD (n = 5, ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; One-way ANOVA). d TUNEL immunofluorescence staining of orthotopic tumors to detect apoptotic cells. Scale bar: 50 µm. e Ki67 immunofluorescence staining of orthotopic tumors to assess cell proliferation. Scale bar: 50 µm. f Quantification of TUNEL-positive cells in orthotopic tumors. Data are mean ± SD (n = 3, ***p < 0.001, ****p < 0.0001; One-way ANOVA). g Quantification of Ki67-positive cells in orthotopic tumors. Data are mean ± SD (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; One-way ANOVA). h Immunofluorescence staining and analysis of CD8⁺ (green) and PD-1⁺ (red) T cells in orthotopic tumors. Nuclei: DAPI (blue). Scale bar: 50 µm. Data are mean ± SD (n = 3, ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; One-way ANOVA). i Flow cytometry analysis of CD8⁺ T cell exhaustion markers (PD-1, CTLA-4, TOX) and effector cytokine (IFN-γ) in orthotopic tumor tissues. Data are mean ± SD (n = 3, ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; One-way ANOVA). j Schematic of the diethylnitrosamine (DEN)/carbon tetrachloride (CCl4)-induced orthotopic HCC model: Mice were treated with DEN+CCl4, followed by AAV-VE (vector control) or AAV-TKTL1 injection at week 20, anti-PD-1 treatment from week 24, MRI scanning, and liver tissue harvest at week 28. k Representative micro-MRI images of mouse livers pre- and post-treatment; representative photographs of excised livers. Scale bar: 1 cm. H&E staining of tumor (T) and peritumoral (P) tissues from DEN/CCl4-induced models. Scale bar: 50 µm. l TUNEL and Ki67 immunofluorescence staining of DEN/CCl4-induced tumor tissues. Scale bar: 50 µm. m Quantification of tumor size changes pre- and post-treatment in DEN/CCl4-induced models related to (k). Data are mean ± SD (n = 5, **p < 0.01, ***p < 0.001, ****p < 0.0001; One-way ANOVA). n Quantification of TUNEL-positive and Ki67-positive cells in DEN/CCl4-induced models related to (l). Data are mean ± SD (n = 3, ns: not significant, *p < 0.05, ***p < 0.001, ****p < 0.0001; One-way ANOVA). o Immunofluorescence staining of CD8⁺ and PD-1⁺CD8⁺ T cells in DEN/CCl4-induced tumors. Scale bar: 50 µm. p Quantification of CD8⁺ and PD-1⁺CD8⁺ T cell proportions in DEN/CCl4-induced tumors. Data are mean ± SD (n = 3, ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001; One-way ANOVA). q Flow cytometry analysis of CD8⁺ T cell exhaustion markers (PD-1, CTLA-4, TOX) and effector cytokine (IFN-γ) in DEN/CCl4-induced tumor tissues. Data are mean ± SD (n = 3, ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; One-way ANOVA). r Schematic illustrating the role of TKTL1 in promoting HCC progression and sensitizing tumors to anti-PD-1 immunotherapy via modulating CD8⁺ T cell function in the TME

TKTL1 reprograms glucose metabolism through glycolysis instead of the PPP to stabilize the HIF-1α/HK2/PD-L1 axis and drive anti-PD-1 sensitivity in HCC

The above experiments provide fundamental insights into TKTL1 function at the cellular and animal levels. To further explore the molecular mechanism by which elevated TKTL1 expression enhances anti-PD-1 sensitivity in HCC, we performed RNA-seq following TKTL1 knockdown in Hep-3B cells. RNA-seq revealed that HK2, HKDC1, and SLC2A1, glycolysis-related genes, were significantly decreased after TKTL1 knockdown (Fig. 3a). KEGG analysis showed downregulated differentially expressed genes (DEG) enrichment in Glycolysis/Gluconeogenesis and HIF-1α signaling (Fig. 3b, supplementary Fig. 4a–c). Non-targeted metabolomics indicated downregulated metabolite enrichment in central carbon metabolism (Fig. 3c, supplementary Fig. 4d–f). Comprehensive analysis of downregulated genes from RNA sequencing and downregulated metabolites from non-targeted metabolomics suggested that TKTL1 may regulate malignant phenotypes through glycolytic pathways (Fig. 3d, supplementary Fig. 4g). Western blot showed TKTL1 knockdown reduced GLUT1/HK2/PKM2/ENO1/LDHA, while overexpression increased them (Fig. 3e, supplementary Fig.4h). Glycolysis inhibitor 2-DG suppressed proliferation/invasion/migration in TKTL1-overexpressing lines (supplementary Fig. 4i, j). Using carbon-13-labeled glucose to trace metabolic flux in TKTL1 knockdown versus NC group HCC cell lines showed significantly reduced glycolytic flux and lactate production after knockdown, with no significant changes in the PPP or TKTL1-related metabolites (such as Erythrose 4 phosphate, Sedoheptulose-7 phosphate,3-phosphoglyceric acid) (Fig. 3f, supplementary Fig. 5a–d). Extracellular acidification rate (ECAR) measurements confirmed TKTL1 knockdown inhibited glycolysis/glycolytic capacity in HCC-LM3/Hep-3B cells, while overexpression enhanced them (Fig. 3g). Furthermore, lactic acid production was downregulated upon TKTL1 knockdown and upregulated with overexpression (Fig. 3g). These findings collectively suggest that TKTL1 primarily regulates the glycolysis pathway within glucose metabolism, affecting the malignant phenotype of HCC.

Fig. 3 — TKTL1 reprograms glucose metabolism through glycolysis instead of the PPP to stabilize the HIF-1α/HK2/PD-L1 axis and drive anti-PD-1 sensitivity in HCC. a mRNA expression of glycolysis-related genes (SLC2A1, HK2, HKDC1) in TKTL1-knockdown Hep-3B cells, determined by RNA sequencing. b KEGG pathway enrichment analysis of downregulated differentially expressed genes (DEGs) in TKTL1-knockdown Hep-3B cells, highlighting Glycolysis/Gluconeogenesis and HIF-1α signaling pathways. c Bubble plot of KEGG pathway enrichment for downregulated metabolites from non-targeted metabolomics of TKTL1-knockdown cells, focusing on central carbon metabolism. d Integrative analysis of downregulated DEGs (RNA sequencing) and metabolites (non-targeted metabolomics), visualized as a ranked plot showing enrichment in glycolytic pathways. e Western blot analysis of glycolytic enzymes (GLUT1, HK2, PKM2, ENO1, LDHA) and TKTL1 in HCC-LM3 and Hep-3B cells with TKTL1 knockdown or overexpression. β-actin serves as a loading control. f ¹³C-labeled glucose metabolic flux tracing analysis, showing the relative abundance of metabolites in glycolysis, PPP and tricarboxylic acid (TCA) cycle in TKTL1-knockdown vs. negative control (NC) cells. g Extracellular acidification rate (ECAR) measurements (left) and lactic acid production (right) in HCC-LM3 and Hep-3B cells under TKTL1 knockdown or overexpression. Data are mean ± SD (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; Student’s t-test). h, i Western blot analysis of TKTL1, HIF-1α, HK2, and PD-L1 protein expression in HCC-LM3 and Hep-3B cells with TKTL1 knockdown or overexpression. β-actin serves as a loading control. Data are mean ± SD (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; Student’s t-test). j Immunofluorescence staining of TKTL1, HIF-1α, and HK2 protein expression in HCC tissues. Nuclei: DAPI (blue). Scale bar: 150 µm

Additionally, although the expression level of HIF-1α protein significantly decreased after TKTL1 gene knockout, its mRNA level did not change in the RNA sequencing (Fig. 3h, i). We also observed a significant reduction in PD-L1 and HK2 protein levels upon TKTL1 knockdown (Fig. 3h, i). This is consistent with prior reports indicating that enhanced glycolysis promotes high PD-L1 expression through HK2-mediated phosphorylation of IκBα, facilitating tumor immune escape.¹³ While HIF-1α is a well-established regulator of glycolysis under hypoxic conditions, its precise mechanistic role in HCC remained unclear. CUT&Tag and chromatin immunoprecipitation (ChIP) assays confirmed that HIF-1α directly binds to the HK2 promoter-proximal region to drive its transcription (supplementary Fig. 6a–c). Subsequent dual-luciferase reporter assays identified a specific sequence (positions +348 to +353 relative to TSS) within the HK2 5' regulatory region as the functional binding site mediating HIF-1α-dependent transcriptional activation (supplementary Fig. 6d). Furthermore, in a mouse orthotopic HCC model overexpressing TKTL1, we tested the lactate transporter MCT4 inhibitor, MSC-4381, in combination with anti-PD-1 therapy. This combination treatment resulted in significantly enhanced anti-PD-1 efficacy specifically in the context of TKTL1 overexpression (supplementary Fig. 6e–g). Tissue microarray analysis demonstrated positive correlations among the protein expression levels of TKTL1, HIF-1α, and HK2 in clinical samples (Fig. 3j). Collectively, these findings establish a mechanistic pathway in which TKTL1 reroutes glucose metabolism towards glycolysis, stabilizes HIF-1α protein, and thereby upregulates HK2 transcription. This cascade ultimately elevates PD-L1 expression, providing a molecular rationale for the observed sensitivity of TKTL1-high HCC tumors to PD-1 blockade therapy.

TKTL1 activates glycolysis pathway by promoting HIF-1α deubiquitination through recruitment of USP9X in HCC

To clarify HIF-1α regulation in the TKTL1/HIF-1α/HK2 axis, literature indicated HIF-1α stability involves ubiquitination.¹⁴ Cycloheximide (CHX) treatment showed TKTL1 overexpression slowed HIF-1α degradation (Fig. 4a, supplementary Fig. 7a). Chloroquine (CQ) and MG132 assays revealed TKTL1 knockdown promotes HIF-1α degradation via the proteasome pathway (MG132-inhibited; CQ-ineffective) (Fig. 4b). MG132 treatment reduced ubiquitinated HIF-1α and increased HIF-1α stability in TKTL1-overexpressing cells (Fig. 4c). Co-immunoprecipitation (Co-IP) confirmed TKTL1-HIF-1α interaction in HCC/HEK-293T cells (Fig. 4d, e). Since TKTL1 lacks direct deubiquitinating activity, we hypothesized that a third protein assists TKTL1 in regulating HIF-1α stability. We then performed Co-IP to identify TKTL1-interacting proteins and found that USP9X, a member of the DUB family, interacts with TKTL1 (Fig. 4f, supplementary Fig. 7b). Interaction assays demonstrated TKTL1-USP9X and USP9X-HIF-1α binding in HCC/HEK-293T cells (Fig. 4g, h), validated by GST-pull down (Fig. 4i).

Fig. 4 — TKTL1 activates glycolysis pathway by promoting HIF-1α deubiquitination through recruitment of USP9X in HCC. a Cycloheximide (CHX) chase assay in OE-VE (vector control) and OE-TKTL1 (TKTL1 overexpression) cells, assessing HIF-1α protein degradation kinetics over 0, 4, 8, and 12 h. β-actin serves as a loading control. b CHX chase assay in Sh-NC (negative control) and Sh-TKTL1 (TKTL1 knockdown) cells, evaluating HIF-1α degradation with or without proteasome inhibitor MG132 treatment. β-actin serves as a loading control. c Ubiquitination assay of HIF-1α in HCC-LM3 (Sh-NC vs. Sh-TKTL1) and Hep-3B (OE-VE vs. OE-TKTL1) cells treated with MG132. Immunoprecipitation (IP) with HIF-1α antibody and antibody incubation with ubiquitin (Ub) antibody detect ubiquitinated HIF-1α. d Co-immunoprecipitation (Co-IP) assay of endogenous proteins in HCC-LM3 and Hep-3B cells, confirming the interaction between TKTL1 and HIF-1α. IgG serves as a negative control; input lysates verify protein expression. e Co-IP assay of exogenous proteins in HEK-293T cells transfected with TKTL1 and HIF-1α expression plasmids, validating the TKTL1-HIF-1α interaction. IgG serves as a negative control. f Intersection analysis of TKTL1-interacting proteins identified via IP and deubiquitinating enzyme (DUB) family members, pinpointing USP9X as a key overlapping binding partner of TKTL1. g, h Co-IP assay identifying USP9X as a TKTL1-interacting protein. Co-IP assay demonstrating the interaction between USP9X and HIF-1α. IgG serves as a negative control. i GST-pull down assay transfected with Myc-USP9X, HA-TKTL1, and Flag-HIF-1α plasmids, verifying the formation of the TKTL1-USP9X-HIF-1α ternary complex. j AI-predicted structural model of the TKTL1-USP9X-HIF-1α ternary complex, illustrating the binding interface and spatial configuration of the three proteins. k IP assay using HIF-1α domain mutants in HEK-293T cells, confirming that USP9X binds to the inhibitory domain (ID) of HIF-1α. l IP assay using TKTL1 domain mutants in HEK-293T cells, identifying that HIF-1α binds to the T-pyr domain of TKTL1 and USP9X binds to the T-pyr domain of TKTL1. m Confocal laser scanning microscopy showing the cytoplasmic colocalization of TKTL1, HIF-1α, and USP9X in HCC-LM3 and Hep-3B cells. Nuclei are stained with DAPI (blue). n Immunofluorescence staining of HCC tissue microarrays showing the colocalization of TKTL1, HIF-1α, and USP9X. Inserted merge images confirm triprotein colocalization. Correlation analysis of TKTL1 and HIF-1α expression levels in HCC tissues (Pearson’s correlation coefficient: r = 0.1107, p = 0.003). Scale bar: 150 µm

To elucidate USP9X-mediated HIF-1α deubiquitination, overexpression of a 7-branched ubiquitin chain plasmid in HEK-293T cells demonstrated USP9X specifically removes K48-linked ubiquitin chains to stabilize HIF-1α (supplementary Fig. 7c). The ubiquitination site predicted by GPS-Uber was identified as lysine 251 (K251), the key residue responsible for targeting HIF-1α for proteasomal degradation. Furthermore, querying the PhosphoSitePlus database confirmed that K251 serves as the primary lysine residue for ubiquitin binding on HIF-1α. These findings were further validated using site-directed mutation plasmids (supplementary Fig. 7d, e). AI modeling further predicted a stable TKTL1-USP9X-HIF-1α co-binding complex (Fig. 4j, supplementary Fig. 7f), guiding domain mutagenesis studies in HEK-293T cells where IP assays confirmed USP9X binds HIF-1α‘s inhibitory domain (ID), while TKTL1 binds HIF-1α‘s ID domain, USP9X interacts with TKTL1’s T-pyr domain, and HIF-1α binds TKTL1’s T-pyr domain (Fig. 4k, l, supplementary Fig. 7g). Confocal microscopy revealed cytoplasmic colocalization of all three proteins in HCC cells (Fig. 4m), with TKTL1 overexpression increasing nuclear HIF-1α (supplementary Fig. 7h). Immunofluorescence of ICI-responder and HCC tissue microarrays confirmed triprotein colocalization and showed positive TKTL1-HIF-1α expression correlation (Fig. 4n), collectively indicating TKTL1 recruits USP9X to deubiquitinate HIF-1α and activate glycolysis in HCC. Knockdown of USP9X in two HCC cell lines transfected with a TKTL1 overexpression plasmid resulted in a reduction in HIF-1α protein expression levels (supplementary Fig. 7i).

TKTL1 enhances the sensitivity of anti-PD-1 therapy by recruiting PD-L1^high DCs via the chemokine CCL4 in HCC

To investigate the impact of TKTL1 knockdown on the tumor immune microenvironment, we performed scRNA-seq and mass cytometry on HCC mouse tumor tissues (Fig. 5a). ScRNA-seq revealed notable alterations in immune cell composition, including a decrease in DC proportions, an overall increase in T cells, and a reduction in macrophages in the experimental group (Fig. 5b, supplementary Fig. 8a–d).

Fig. 5 — TKTL1 enhances the sensitivity of anti-PD-1 therapy by recruiting PD-L1^high DCs via the chemokine CCL4 in HCC. a Schematic workflow of single-cell RNA sequencing (scRNA-seq) and mass cytometry (CyTOF) analyses: scRNA-seq was performed on mouse HCC tumor tissues, and CyTOF was used to profile CD45⁺ immune cells from tumor tissues. b t-SNE visualization of scRNA-seq data from sh-NC and sh-TKTL1 groups, showing the distribution of major immune cell populations. c t-SNE plot of unsupervised clustering of CD45⁺immune cells from tumor tissues via CyTOF, identifying 39 distinct cell clusters with annotation of key subsets. d Comparative analysis of immune cell subset proportions and PD-L1 expression on PD-L1^high DC (cluster C37) between sh-NC and sh-TKTL1 groups. e Schematic workflow of CyTOF analyses used to profile CD45⁺ immune cells from peripheral blood mononuclear cells (PBMCs) of HCC patients. f t-SNE plot of CyTOF data from PBMCs shows the immune cell landscape in anti-PD-1 responders (R) versus non-responders (NR) before treatment and distinguishes non-recurrence from recurrence subgroups in HCC patients. g t-SNE visualization of CyTOF data from PBMCs of responders (R-Before vs. R-After) and non-responders (NR-Before vs. NR-After) pre- and post-anti-PD-1 treatment. h Flow cytometry quantification of DC frequencies in the TME and tumor-draining lymph nodes (TDLNs) of OE-VE (vector control) and OE-TKTL1 (TKTL1 overexpression) mouse models, with or without anti-PD-1 treatment. Data are mean ± SD (n = 3, ns: not significant, **p < 0.01, ****p < 0.0001; One-way ANOVA). i Chemokine array and validation results showing CCL4 upregulation in TKTL1-overexpressing HCC cells, confirmed by protein expression and supernatant level detection. Data are mean ± SD (n = 3, ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; One-way ANOVA). j Chromatin immunoprecipitation (ChIP) assay demonstrating HIF-1α binding to the CCL4 promoter. Input and IgG serve as controls; agarose gel electrophoresis shows the amplified CCL4 promoter fragment. k Schematic illustration of the TKTL1/HIF-1α/CCL4 axis: TKTL1 stabilizes HIF-1α, which directly binds the CCL4 promoter to drive its transcription, promoting recruitment of PD-L1^high DCs to the TME and enhancing anti-PD-1 sensitivity in HCC

Additionally, we employed mass cytometry to profile single, viable, and intact CD45⁺immune cells isolated from selected tissues. Unsupervised clustering of CD45⁺immune cells yielded 39 distinct cell clusters, each annotated based on established lineage-specific markers (Fig. 5c, supplementary Fig. 9a, b). Comparative analysis demonstrated that, relative to the sh-NC group, the sh-TKTL1 group exhibited decreased relative proportions of B cells, DCs, and neutrophils, while the proportions of CD8⁺ T cells and CD4⁺ T cells were elevated (Fig. 5d). Strikingly, we observed a significant reduction in PD-L1 expression on the surface of cluster C37, identified as PD-L1^high DC, in the knockdown group (Fig. 5d). We further performed mass cytometry on peripheral blood mononuclear cells (PBMCs) collected from 10 HCC patients at different immunotherapy timepoints (pre-treatment, post-treatment, relapse, non-relapse). Although not statistically significant, a trend toward higher pre-treatment conventional DC (cDC) frequencies was observed in responders compared to non-responders. Post-treatment responders displayed increased cDC levels relative to their pre-treatment baseline, whereas non-responders showed no notable change. Furthermore, relapsed patients exhibited reduced cDC frequencies compared to non-relapsed patients (Fig. 5e–g, supplementary Fig. 10a–f). These findings align with prior reports indicating greater DC infiltration in ICI-sensitive patients (supplementary Fig. 11a). Bioinformatic analysis of scRNA-seq data (GSE146115) confirmed that TKTL1-positive tumor cells correlated positively with both cDC1 and cDC2 subsets (supplementary Fig. 11b). Additional bioinformatics evaluation revealed that elevated TKTL1 expression was associated with increased infiltration of T cells, DC, B cells, and other immune cells—particularly across DC subpopulations (supplementary Fig. 11c–h). Integrating these multifaceted results, we selected DC as the focal point for subsequent investigations. In vivo flow cytometry further demonstrated that TKTL1 overexpression led to increased DC frequencies within the TME and tumor-draining lymph nodes (TDLNs) compared to controls, with anti-PD-1 combination therapy further enhancing DC accumulation (Fig. 5h, supplementary Fig. 12a).

The precise molecular mechanism by which TKTL1 recruits DCs to the TME remains to be fully elucidated. DC trafficking in anti-tumor immunity involves two distinct, sequentially regulated steps: first, circulating DC precursors are recruited to the tumor site to capture tumor antigens; second, antigen-loaded DCs migrate to tumor-draining lymph nodes (TDLNs) via the CCR7-CCL19/CCL21 axis to prime naive T cells. While CCR7 plays a particularly critical role in the latter step of DC migration from primary tumors to TDLNs, the initial recruitment of DCs into the TME is primarily driven by tumor-derived chemokines. To delineate the TKTL1-specific pathway mediating DC recruitment to tumors, a chemokine array was performed on HCC cells. This analysis identified CCL4 as the most significantly upregulated DC-associated chemokine in TKTL1-overexpressing cells. This upregulation was subsequently confirmed at both the intracellular protein level (in cell lysates) and the secreted protein level (in culture supernatants) (Fig. 5i, supplementary Fig. 12b–d). Mechanistically, CUT&Tag and chromatin immunoprecipitation (ChIP) assays demonstrated that HIF-1α directly binds to the CCL4 promoter to drive its transcription (Fig. 5j). This regulatory axis was further supported by a positive correlation observed between TKTL1 and CCL4 expression levels in human HCC tissue samples (supplementary Fig. 12e). The functional relevance of CCL4 was established using the inhibitor Maraviroc, which effectively abrogated the chemotactic effect of tumor-derived supernatant on DC migration in vitro (supplementary Fig. 13a). The in vivo significance of this axis was validated in a mouse orthotopic HCC model. Notably, when Maraviroc was co-administered with an anti-PD-1 antibody in the context of TKTL1 overexpression, it significantly attenuated the therapeutic efficacy of the anti-PD-1 treatment (supplementary Fig. 13b–d). This finding underscores the critical role of the TKTL1/HIF-1α/CCL4 axis in mediating the immunomodulatory effects of TKTL1 and its contribution to anti-PD-1 sensitivity.

Given that DCs are known to mediate PD-L1-dependent immunosuppression and can predict responses to PD-1 blockade,¹⁵ we further investigated the phenotype of the recruited DCs. In vivo flow cytometry revealed that TKTL1 overexpression specifically increased the frequency of PD-L1^high DCs, predominantly within the TME rather than in TDLNs (supplementary Fig. 14a, b). Clinically, immunofluorescence analysis of patient samples showed that higher infiltration of PD-L1^high DCs was associated with sensitivity to anti-PD-1 therapy (supplementary Fig. 14c). Conversely, in vivo experiments confirmed that while TKTL1 elevated PD-L1 levels on DCs, combination therapy with anti-PD-1 was able to reduce this PD-L1 expression (supplementary Fig. 14d). In summary, our data establish a novel mechanistic link wherein TKTL1 promotes the recruitment of immunosuppressive PD-L1^high DCs to the HCC microenvironment through a HIF-1α-driven transcriptional upregulation of CCL4 (Fig. 5k).

TKTL1-mediated lactate-enriched TME promotes TRIM28^K408 lactylation in DCs, thereby stabilizing PD-L1

Lactylation modification, a significant milestone in lactate research, plays important roles in tumor progression and immune regulation,¹⁴ occurring on both histones (lysine residues) and non-histone proteins where it stabilizes substrates to enhance expression.¹⁶

When HCC cells overexpressing TKTL1 were co-cultured with DCs in a TKTL1-mediated, high-lactate microenvironment, analyses via western blotting and flow cytometry revealed an upregulation of PD-L1 expression on the DCs (Fig. 6a–c). This alteration in the DC phenotype was further associated with an enhanced exhaustion state of CD8⁺T cells (supplementary Fig. 15a–d). Adding different lactate concentrations to DCs revealed peak PD-L1 expression at 12 h with 20 mmol/L lactate, while expression decreased by 24 h, indicating dynamic lactate concentration- and time-dependent PD-L1 regulation on tumor-infiltrating DCs. Treatment with the deacetylase HDAC-1 (20 mmol/L, 12 h) reduced both PD-L1 expression and pan-lactylation levels (Fig. 6d–f, supplementary Fig. 15e, f).

Fig. 6 — TKTL1-mediated lactate-enriched TME promotes TRIM28 lactylation in DCs, thereby stabilizing PD-L1. a Schematic diagram of co-culture system: DCs (DC2.4 or primary mouse-derived DCs) co-cultured with control or TKTL1-overexpressing Hepa1-6 cells (HCC cells) to mimic TKTL1-mediated high-lactate TME. b Western blot analysis of PD-L1 expression on DCs after co-culture with control or TKTL1-overexpressing Hepa1-6 cells. c Flow cytometry quantification of PD-L1 expression on DCs co-cultured with control or TKTL1-overexpressing Hepa1-6 cells. Data are mean ± SD (n = 3, *p < 0.05, **p < 0.01; Student’s t-test). d Flow cytometry assessment of PD-L1 expression on DCs treated with gradient lactate concentrations (0–30 mM) for 12 h or 24 h. Data are mean ± SD (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; Student’s t-test). e Western blot of pan-Lactylation levels in DCs treated with gradient lactate concentrations (0–30 mM) for 12 h or 24 h. f Western blot of pan-Lactylation and PD-L1 levels in DCs treated with 20 mmol/L HDAC-1 (Deacetylase) for 12 h. g Schematic diagram of DC2.4 cells co-cultured with tumor cells and then sent for lactic acid mass spectrometry. h PD-L1 immunoprecipitation and Lactylation blots in DCs. i, j Intersection analysis of Lactylation sites and PD-L1-interacting proteins identifying TRIM28-K408 as the research target. k Computational structural model predicting the spatial interaction between TRIM28 (K408 Lactylation site) and PD-L1. l Western blot of PD-L1 protein stability in DCs transfected with Flag-TRIM28 WT or Flag-TRIM28 K408 mutant. m Ubiquitination (add MG132) and SUMOylation assays of PD-L1 in HEK-293T cells transfected with TRIM28/PD-L1 tag protein. n Western blot of TRIM28 and PD-L1 expression in DCs treated with HDAC-1 or TRIM28-specific siRNA. o Multiplex immunofluorescence was performed to analyze the co-expression of TRIM28 and PD-L1 in DCs within tissues exhibiting high or low TKTL1 expression

To further explore the effect of lactylation on PD-L1, we performed DIA-based lactylation proteomics analysis on DCs treated with standard medium or 20 mmol/L lactate for 12 h (supplementary Fig. 15g–i). Results indicated that PD-L1 itself was not directly modified by lactylation (supplementary Fig. 15g–i). However, immunoprecipitation of PD-L1 revealed its interaction with lactoylated proteins (Fig. 6g, h). Subsequent protein mass spectrometry analysis of PD-L1 immunoprecipitation products, integrated with lactoylome sequencing data, identified TRIM28 as interacting with PD-L1 (Fig. 6i, supplementary Fig. 15j). Notably, lactylation at lysine residue K408 of TRIM28 was pronounced in DCs, with AI predicting their spatial relationship (Fig. 6j, k, supplementary Figs. 15k and 16a). Following site-directed mutation at the K408 residue, a reduction in the protein level of TRIM28 was observed, concomitant with a decrease in PD-L1 protein stability (Fig. 6l). IP analysis of TRIM28 in DCs confirmed that both its expression and lactylation level increased with rising lactate concentrations (supplementary Fig. 16b). These findings suggest that TKTL1 upregulation in HCC cells mediates a high-lactate TME, promoting TRIM28 lactylation in DCs and consequently increasing PD-L1. However, the mechanism by which TRIM28 elevates PD-L1 in DCs remains unknown.

Co-IP confirmed TRIM28 interacts with PD-L1 in DCs (supplementary Fig. 16b). Consistent with prior reports, TRIM28 stabilizes PD-L1 by inhibiting ubiquitination and promoting SUMOylation and facilitates K63-linked polyubiquitination of TBK1 to enhance PD-L1 transcription,¹⁵ IP assays using plasmids in HEK-293T cells revealed TRIM28 prevents PD-L1 proteasomal degradation by enhancing SUMO modification (Fig. 6m). Given TRIM28’s role in stabilizing PD-L1, we transfected DCs with TRIM28-specific siRNA. Immunofluorescence and western blot showed that both the deacetylase HDAC-1 and si-TRIM28 reduced TRIM28 and PD-L1 expression (Fig.6n). Immunofluorescence of HCC tissue microarrays confirmed colocalization of TRIM28^K408 Lactylation with PD-L1 in DCs and positive expression correlation (Fig. 6o), indicating the TKTL1-mediated high-lactate TME increases TRIM28^K408 Lactylation to stabilize PD-L1 via suppressing ubiquitin-proteasomal degradation.

Targeted co-delivery of que and lactate oxidase via biomimetic nanosystem for synergistic TKTL1 suppression and TME remodeling in HCC

To bridge the significant gap in available TKTL1-targeting therapeutics, we employed a structure-based drug discovery approach. Virtual screening against the human TKTL1 protein identified the top-ranking natural compound HY-N1968 (Quercetin-3-O-β-D-glucose-7-O-β-D-gentiobioside, hereafter referred to as Que). Molecular docking analysis demonstrated that Que forms a stable complex with TKTL1, which is mediated by its interaction with the thiamin diphosphate-binding fold (i.e., the pyr domain) of TKTL1 involving residues VAL270, ARG355, and ASN383 (Fig. 7a, b). Notably, comparative docking analysis across the transketolase family demonstrated that while Que interacts with the N-terminal region of TKT and TKTL2, its primary binding site on TKTL1 is located within the Thiamin diphosphate-binding fold—the key functional domain central to our proposed molecular mechanism (supplementary Fig. 17a). Molecular dynamics simulations confirmed the stability of the TKTL1-Que complex, which equilibrated after 150 ns with an RMSD of ~3.1 Å. Stabilized hydrogen bonds and SASA indicated tight binding. RMSF analysis showed overall rigidity except for a flexible region at His-160, and the complex maintained a stable radius of gyration (Fig. 7c). The binding affinity was determined to be Kd = 27.53 μM by microscale thermophoresis (Fig. 7d). Cellular validation was achieved by conjugating the compound with a fluorescent coumarin C6 tag; confocal microscopy confirmed its co-localization with TKTL1 in HCC cells, and western blot analysis demonstrated that Que downregulated the TKTL1-induced overexpression of HIF-1α protein and suppressed its downstream target, HK2 (Fig. 7e–g).

Fig. 7 — Targeted co-delivery of que and lactate oxidase via biomimetic nanosystem for synergistic TKTL1 suppression and TME remodeling in HCC. a, b Virtual screening and molecular docking of Que(HY-N1968) with human TKTL1 protein. a 3D binding pose showing hydrogen bonds and π-π stacking. b 2D interaction map. c Molecular dynamics simulations of TKTL1-Que complex: RMSD (~ 3.1 Å), radius of gyration (Rg), solvent-accessible surface area (SASA), and root mean square fluctuation (RMSF) confirm complex stability. d MST analysis showing Que binds TKTL1 with Kd = 27.53 μM. e Confocal microscopy of HCC cells treated with Coumarin C6-labeled Que, confirming co-localization with TKTL1. f, g Western blot of HIF-1α, and HK2 expression in HCC cells after Que treatment. Data are mean ± SD (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001; Student’s t-test). h Schematic of the biomimetic nanosystem: Hollow mesoporous MnO₂ nanoparticles co-loaded with Que and LOX, coated with HCC cell membrane (CQLH). i Standard curves for LOX and Que quantification. j Optimization of drug loading LOX and Que. k pH-dependent Que release profile from CQLH. Data: mean ± SD (n = 3, ns: not significant, **p < 0.01, ***p < 0.001, ****p < 0.0001; Two-way ANOVA). l TEM image of CQLH nanoparticles. m Elemental mapping (Mn, O, C, N, S, P) confirming nanocomposite composition. n SEM image of CQLH nanoparticles. o Western blot of AFP expression validating HCC membrane coating integrity. p Functional analysis of CQLH in HCC cells: Lactate reduction; Oxygen elevation. Data are mean ± SD (n = 3, ns: not significant, ****p < 0.0001; One-way ANOVA). q In vivo fluorescence imaging showing tumor-specific CQLH accumulation. r Schematic of the anticancer mechanism: CQLH targets TKTL1, scavenges lactate, generates O₂, and remodels the TME

To enable targeted delivery and combination therapy, we engineered a biomimetic nanosystem. This system comprises hollow mesoporous MnO₂ nanoparticles co-loaded with Que and lactate oxidase (LOX), subsequently coated with HCC-derived cell membranes (CQLH) (Fig. 7h). Comprehensive physicochemical characterization included standard curves for quantification (Fig. 7i), optimization of drug loading ratio and efficiency (Fig. 7j), and demonstration of a pH-responsive drug release profile with maximal release at the tumor-acidic pH of 5.0 (Fig. 7k). Nanoparticle morphology and elemental composition were verified by transmission electron microscopy (TEM), scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS) (Supplementary Fig. 18d) (Figs. 7l–n, S18a–d), and successful membrane coating was confirmed via detection of the HCC-specific marker AFP (Fig. 7o). Dynamic light scattering confirmed the nanosystem’s hydrodynamic diameter and zeta potential (supplementary Fig. 19a, b). Functional assessments in vitro revealed that the CQLH nanosystem effectively reduced lactate levels and increased oxygen concentration in the TME (Fig. 7p) and promoted Mn²⁺ ion release in acidic conditions (supplementary Fig. 19c). Mechanistically, the compound inhibited HIF-1α nuclear translocation, suppressed TRIM28 lactylation, and ultimately destabilized PD-L1 protein in DCs (supplementary Fig. 19d, e). In vivo, fluorescence imaging demonstrated strong liver tumor-specific targeting and accumulation of the nanosystem (Fig. 7q). Collectively, this work identifies Que as a novel, specific TKTL1-binding compound and develops a bioinspired, multi-functional nanosystem (CQLH) for its targeted co-delivery with LOX to remodel the TME and overcome HCC immunotherapy resistance, as summarized in the proposed mechanism (Fig. 7r).

CQLH nanodrug enhanced anti-PD-1 therapy via dual TKTL1/lactate inhibition and TME remodeling in HCC models

To evaluate the antitumor efficacy of the nanodrug CQLH, we first established an orthotopic allograft model (Fig. 8a). Results demonstrated that CQLH significantly inhibited tumor growth compared to Que or anti-PD-1 monotherapy, with further suppression observed in the CQLH+anti-PD-1 combination group, and extended OS (Fig. 8b–e, supplementary Fig. 20a). Ki67 staining revealed substantially reduced tumor cell proliferation in the combination group versus monotherapies (supplementary Fig. 20b). In a DEN-induced chemical model (Fig. 8f), CQLH exhibited consistent antitumor effects (Fig. 8g–i, supplementary Fig. 20c). Immunofluorescence confirmed that CQLH effectively suppressed HIF-1α expression and the glycolysis pathway marker HK2, while combination therapy reduced PD-L1 expression on DCs (Fig. 8j). HE staining indicated no significant CQLH-related toxicity in the heart, liver, spleen, lung, or kidney (supplementary Fig. 20d). We further developed patient-derived tumor xenograft (PDTX) models with successful immune reconstitution (Fig. 8k, supplementary Fig. 21a–d). CQLH exhibited enhanced therapeutic efficacy in TKTL1-high versus TKTL1-low PDTX subgroups, and mice with TKTL1-high tumors treated with the combination of CQLH and anti-PD-1 achieved longer survival (Fig. 8l, m, supplementary Fig. 22a–c). Concordantly, in vivo data confirmed a superior anti-PD-1 response in TKTL1-high tumors (Fig. 8n, supplementary Fig. 22d). Immunofluorescence in TKTL1-high models showed that CQLH downregulated tumor HIF-1α/HK2 expression and reduced TRIM28/PD-L1 levels in DCs (Fig. 8o, supplementary Fig. 22e). Repeated HE staining confirmed sustained organ safety profiles (supplementary Fig. 22F). In summary, the nanodrug CQLH demonstrated significant antitumor activity across orthotopic, chemically induced, and PDTX HCC models-with superior efficacy in TKTL1-high tumors and enhanced anti-PD-1 therapy to suppress tumor growth while downregulating TKTL1/HK2-driven glycolysis and PD-L1 expression on DCs, all without observable organ toxicity.

Discussion

Anti-PD-1/PD-L1 therapy faces limited efficacy in solid tumors despite enhancing CD8⁺ T cell function, as tumors evade rejection via hostile TME mechanisms (e.g., antigen deficiency, T cell exhaustion, IFN-γ resistance).¹⁷ PD-L1 is an imperfect biomarker, with poor response of 20% in PD-L1-low tumors and suboptimal overall response rates.¹⁸ Mechanistic insights reveal PD-L1 regulation by MAPK/PI3K pathways,¹⁹ exemplified by PRDM1/BLIMP1 modulating USP22-SPI1-PD-L1 in HCC.²⁰ This complexity underscores the need to identify more reliable biomarkers and therapeutic targets. Our study identifies TKTL1 as an oncoprotein that promotes HCC progression yet is associated with enhanced sensitivity to anti-PD-1 therapy. Integrated multi-omics analysis of patients treated with ICIs revealed consistent upregulation of TKTL1 in responders. Although high TKTL1 expression correlates with poor prognosis in untreated HCC, it appears to foster an immune-permissive TME that increases susceptibility to PD-1 blockade, positioning TKTL1 as a dual-purpose biomarker for both prognosis and prediction of immunotherapy response.

TKTL1, a PPP rate-limiting enzyme linked to carbohydrate catabolism, has recently been implicated in glycolytic reprogramming across malignancies, with emerging studies demonstrating its direct regulation of glycolysis beyond PPP functions. In cervical cancer, TKTL1 participates in malignant progression via AKT-signal-mediated PFKFB3 activation to enhance glycolytic flux,²¹ while in head and neck squamous cell carcinoma (HNSCC), it is activated by promoter hypomethylation and drives carcinogenesis through increased aerobic glycolysis and HIF-1α stabilization,²² collectively repositioning TKTL1 as a dual-metabolic regulator orchestrating tumor progression through both PPP-dependent NADPH generation and glycolysis-centered Warburg effects. Here, we reveal TKTL1 recruits USP9X to deubiquitinate HIF-1α at K251, activating the HIF-1α/HK2/glycolysis axis and upregulating PD-L1. We note that while other E3 ligases are well-established regulators of HIF-1α stability under normoxic conditions, our findings reveal that TKTL1 stabilizes HIF-1α under normoxia by co-opting USP9X. This represents a complementary mechanism to the canonical oxygen-dependent pathway, expanding our understanding of how metabolic oncoproteins can bypass classical degradation routes to sustain a pseudo-hypoxic state even in the presence of oxygen. Furthermore, the selective activation of HK2 by HIF-1α in our system warrants mechanistic insight. This selectivity is governed by chromatin accessibility, cofactor availability, and transcription factor competition.^23,24 We propose that in TKTL1-high HCC cells, metabolic reprogramming epigenetically primes chromatin to favor HK2 activation, a hypothesis testable by future ChIP-seq studies comparing TKTL1-high and low tumors. While glycolysis promotes PD-L1 in glioblastoma via HK2-IκBα-NF-κB pathway¹³ and histone lactylation driven by lactate facilitates PD-L1 transcription in acute myeloid leukemia,²⁵ these models incompletely replicate human HCC TME subtypes including immune-inflamed, -desert, -excluded).¹⁷ Our work specifically delineates the mechanism in immune-inflamed, TKTL1-high HCC, where elevated PD-L1 likely results from adaptive immune resistance mediated by infiltrating CD8⁺ T cells, a state that can be reversed by PD-1 blockade. This subtype-specific insight highlights the importance of tailoring therapeutic strategies based on TME classification.

In the present study, we also found that TKTL1-generated lactate induces TRIM28^K408 lactylation in DCs, stabilizing PD-L1 by inhibiting ubiquitination. Concurrently, HIF-1α transactivates CCL4, recruiting PD-L1^high DCs to establish immunosuppression. This positions PD-L1^high DCs as pivotal effectors. Inflamed triple negative breast cancer responsive to anti-PD-1 features CLEC9A⁺ DCs and T cell exhaustion markers.²⁶ PD-L1 blockade on DCs relieves B7.1 sequestration, enhancing CD28-mediated T cell priming. DC gene signatures correlate with improved survival in renal/NSCLC patients receiving atezolizumab,²⁷ confirming DCs as critical PD-L1 blockade targets. Our findings add a metabolic perspective to this model, directly linking tumor glycolysis to DC function via post-translational lactylation.

To translate these insights into a viable therapeutic approach, we engineered CQLH, a biomimetic nanosystem co-delivering TKTL1 inhibitor (Que) and LOX. Its hollow MnO₂ core catalyzes lactate degradation to pyruvate/H₂O while generating O₂ to alleviate hypoxia. HCC membrane coating enables tumor-specific targeting. This strategy advances lactate-focused TME remodeling, akin to mRBC-HMnO₂-LOX systems,²⁸ by uniquely combining TKTL1 inhibition with enzymatic lactate/O₂ modulation for combined metabolic-immunotherapy. By combining metabolic inhibition with immunomodulation, CQLH provides a new strategy to disrupt the TKTL1–glycolysis–lactate–DC axis and overcome anti-PD-1 resistance.

In summary, TKTL1 drives anti-PD-1 response in HCC primarily through the glycolytic pathway rather than the PPP. Specifically, USP9X-mediated HIF-1α stabilization triggers HK2-dependent glycolysis and lactate production alongside HIF-1α/CCL4-driven recruitment of PD-L1^high DCs. Lactate then induces TRIM28^K408 lactylation in DCs, stabilizing PD-L1 to establish immunosuppression. Targeting this axis with the CQLH nanosystem suppresses TKTL1, depletes lactate, destabilizes PD-L1, and synergizes with anti-PD-1 in preclinical models. TKTL1 thus serves as both a predictive biomarker and therapeutic target, with CQLH converting TKTL1^high immunotherapy resistance into sensitivity.

This study has several limitations to consider. First, while TKTL1 shows a paradoxical association with both poor baseline prognosis and enhanced anti-PD-1 response, its clinical utility as a predictive biomarker remains exploratory and requires validation in larger, prospective cohorts. Second, the multi-omics analyses were conducted in a modest discovery cohort, limiting statistical power and underscoring the need for confirmatory studies. Third, we acknowledge that much of the mechanistic dissection relied on DC2.4 cells and in vitro co-culture systems, which do not fully capture the heterogeneity and functional specialization of dendritic cell subsets in vivo. Nevertheless, deeper subtype-resolved analyses will be necessary to fully delineate DC subset-specific responses. Fourth, although the CQLH nanosystem demonstrates promising efficacy in preclinical models, its translational potential faces substantial challenges—including scalable manufacturing, immunogenicity of the biomimetic coating, and regulatory approval pathways for multi-component nanotherapeutics. Accordingly, we have reframed CQLH as a preclinical proof-of-concept tool that validates the therapeutic potential of targeting the TKTL1-lactate axis, rather than a near-term clinical candidate. Collectively, our findings establish a compelling mechanistic rationale but represent an early-stage investigational platform requiring substantial further development before clinical application.

Materials and methods

Clinical specimens and patient cohorts

Tumor and adjacent non-tumorous tissues were collected from 98 HCC patients undergoing curative resection at the Hepatobiliary Surgery Center of the First Affiliated Hospital of Nanjing Medical University (2016-2024). All patients provided written informed consent under Institutional Review Board approval (2019-SR-332). Fresh tissue samples from the NCT04297202 trial cohort were stratified into responders and non-responders according to mRECIST criteria. Subsets of these fresh specimens were subjected to multi-omics analyses: RNA sequencing was performed on 6 responders and 4 non-responders, while proteomic profiling included 5 responders and 7 non-responders.

RNA and proteomics sequencing

We processed cancer samples with Trizol for lysis, followed by RNA sequencing. The RNA sequencing data was then evaluated by Berry Genomics (Beijing, China). For proteomics sequencing, cell pellets/tissue were suspended on ice in 200 μL lysis buffer. Peptides were labeled with TMT (Tandem Mass Tag) reagents according to the manufacturer’s instructions (Thermo Fisher Scientific, USA). TMT-labeled peptides mixture was fractionated using a Waters XBridge BEH130 column on Agilent 1290 HPLC operating at 0.3 mL/min. Buffer A consisted of 10 mM ammonium formate and buffer B consisted of 10 mM ammonium formate with 90% acetonitrile; both buffers were adjusted to pH 10 with ammonium hydroxide. The fractions were dried for nano LC-MS/MS analysis (Shanghai Bioprofile, China).

Immunohistochemistry (IHC) and immunofluorescence (IF)

The paraffin-embedded sections underwent a process of removing wax and rehydration in preparation for immunohistochemical analysis. The activity of peroxidase was effectively blocked by a 3% hydrogen peroxide solution from Beyotime in China. Following an overnight incubation at 4 °C with primary antibodies, the tissue sections were treated with a biotinylated secondary antibody. Subsequently, the sections were exposed to a combination of streptavidin-horseradish peroxidase (Beyotime, China).

In the process of conducting fluorescent immunohistochemistry, the initial step involved fixing the sample at room temperature using a 4% solution of paraformaldehyde for a duration of 20 min. Subsequently, the sample was subjected to immersion in a 0.05% solution of Triton X-100 for a period of 5 min. Following these treatments, the samples were sealed overnight in a solution containing 2% BSA PBS and were exposed to primary antibodies at a temperature of 4 °C. The treated samples were then linked with either Alexa Fluorite or HRP (Beyotime, China) at room temperature. A combined secondary antibody was applied and left to incubate for 1 h. To maintain the integrity of the nucleus, DAPI staining (Beyotime, China) was utilized. Subsequent to drying, images were captured utilizing a laser scanning confocal microscope from Zeiss, Germany. A detailed list of the antibodies employed in this investigation can be found in supplementary table 2.

Animal models

The experimental protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the First Affiliated Hospital of Nanjing Medical University (Approval ID: IACUC-2404094). All procedures complied with the institutional guidelines established by the Nanjing Medical University Animal Experiment Center. To mitigate potential biases, all experimental procedures employed stratified randomization of animals across treatment groups based on baseline tumor volume/body weight. All mice were 6-week-old C57BL/6 mice at the start of experiments, with an initial body weight range of 18–22 g. Processing order (e.g., drug administration, measurements) was systematically rotated daily. Cage position within racks was alternated every 48 h to eliminate microenvironmental gradients (light/temperature). Outcome assessments (tumor measurements, behavioral tests) were performed by blinded operators using standardized protocols. Experimental mice were included based on verified tumor establishment (subcutaneous: 100–150 mm³; orthotopic/PDX: imaging-confirmed lesions; chemically induced: histopathological HCC confirmation), baseline health (body weight loss ≤10%, no systemic distress), and successful PDX engraftment (F1). Animals were excluded for health deterioration (body weight loss >20%, spontaneous ulceration/necrosis), treatment-related complications (nanosystem toxicity, inadequate drug uptake), or technical failures (implantation errors, poor biodistribution). All exclusions required objective validation (imaging/histology/behavioral scoring), with immediate euthanasia if ethical thresholds (body weight loss >20% or severe distress) were breached. For survival curve statistics, we defined a 20% or more weight loss, severe mobility impairment, or a tumor size exceeding 1.5 cm as the humane survival endpoint.

Subcutaneous tumor model establishment

Six-week-old C57BL/6 mice were subcutaneously inoculated with Hepa1-6 cells (1 × 10⁶ cells/mouse) transduced with lentiviral vectors for TKTL1 overexpression (OE-TKTL1) or empty vector control (OE-VE), subsequently stratified into four tumor-bearing experimental groups (n = 4/group): OE-VE, OE-TKTL1, OE-VE+anti-PD-1, and OE-TKTL1+anti-PD-1 (The random grouping scheme was generated by using SPSS to produce a random number table and then matching it with the ear tag reference table), with anti-PD-1 monoclonal antibody (10 mg/kg) administered intraperitoneally three times per week commencing on day 8 post-implantation.

Orthotopic transplantation model

Lentivirus-transduced TKTL1-overexpressing (OE-TKTL1) or empty vector control (OE-VE) Hepa1-6 cells suspended in 100 μL sterile PBS were subcutaneously implanted into the right flank of 6-week-old C57BL/6 mice, with subsequent humane euthanasia upon xenografts reaching 800 mm³; excised tumors underwent meticulous dissection to remove necrotic and stromal tissues, whereafter viable specimens were aseptically minced into 2 mm³ fragments and orthotopically engrafted into hepatic parenchyma of recipient mice (n = 5/group) under isoflurane anesthesia, followed by randomization into four cohorts: OE-VE, OE-TKTL1, OE-VE + anti-PD-1, and OE-TKTL1 + anti-PD-1 (The random grouping scheme was generated by using SPSS to produce a random number table and then matching it with the ear tag reference table), with weekly tumor monitoring via in vivo bioluminescence imaging until terminal endpoint, where final tumor burden was quantified by digital caliper measurements using ellipsoid volume formula, supplemented by necropsy documentation, gravimetric analysis, and histopathological assessment.

DEN&CCL4-induced orthotopic hepatocarcinogenesis model

C57BL/6 mice underwent diethylnitrosamine (DEN, 25 mg/kg in 0.9% NaCl, 10 μL/g injection volume) priming via intraperitoneal administration, with pre-procedural identification through toe-clipping and comprehensive injection logs. At 4 weeks of age, subjects transitioned to ear-tag identification with concomitant fasting body weight documentation. Chronic liver injury was induced from week 4 through twice-weekly intraperitoneal delivery of 10% CCl4 (0.1 mL/mouse in corn oil vehicle) over 10 consecutive weeks, with monthly fasting weight monitoring. Hepatic tumorigenesis was longitudinally tracked via MRI commencing at 20 weeks of age. Upon confirmation of orthotopic hepatocellular carcinoma establishment, mice received single-stranded AAV-TKTL1 or control vectors through tail vein systemic delivery (The random grouping scheme was generated by using SPSS to produce a random number table and then matching it with the ear tag reference table). Pharmacological interventions commenced 7 days post-viral transduction under stratified randomization, with tumor progression quantified through MRI volume measurements.

The PDTX model

All experiments were conducted using NCG mice. Two qualified xenografts were thawed and propagated subcutaneously in immunodeficient hosts until reaching 800 mm³, followed by humane euthanasia. Excised tumors underwent macro-dissection to remove necrotic cores and peritumoral stromal components, with viable tissues aseptically fragmented into 2 mm³ implants for secondary engraftment. When donor-derived tumors reached 80–120 mm³, recipient mice were stratified into cohorts (n = 3 /group, n = 5/group for OS) through block randomization balancing tumor burden and body weight (The random grouping scheme was generated by using SPSS to produce a random number table and then matching it with the ear tag reference table). On cohort assignment day (D0), human PBMCs were adoptively transferred via tail vein injection, with therapeutic interventions initiated 24 h post-PBMC infusion according to protocol-defined dosing schedules. Longitudinal monitoring included biweekly tumor volumetry and body weight tracking at critical phases: pre-engraftment, pre-treatment, during treatment cycles, and pre-termination. Immune reconstitution was validated through serial flow cytometric analysis of peripheral hCD45⁺ populations at 2-day post-PBMC injection and terminal timepoints. Terminal endpoints comprised tumor excision, gravimetric analysis, macroscopic documentation, and histopathological correlation studies for efficacy evaluation.

Single-cell RNA sequencing and mass cytometry analysis in mouse tumor tissues

Single-cell RNA sequencing (scRNA-seq) was performed on dissociated tumor tissues from C57BL/6 mice (sh-NC and sh-TKTL1 groups). Following tissue dissociation (Tumor Dissociation Kit, Miltenyi Biotec) and viability assessment, single-cell suspensions were processed using the MobiNova®-100. Data were analyzed with Cell Ranger and Seurat for clustering and annotation using canonical marker genes.

For mass cytometry (CyTOF), the treatment method of mouse tissue was referred to Miltenyi Biotec mouse tumor isolation Kit. Percoll removed impurities and split red. CyTOF staining steps included 194Pt staining–Fc block–surface antibody staining–overnight DNA staining–intracellular antibody staining–data collection on computer.

Cell cultures

Mice HCC cells (Hepa1-6) and human HCC cells (Hep-3B and HCC-LM3) were supplied by the Cell Bank of Type Culture Collection. Hepa1-6 cells, Hep-3B and HCC-LM3 cells were all cultured with DMEM medium (Gibco, USA). At 37 °C and 5% CO₂ in an incubator, all the cells were supplied with 10% fetal bovine serum (Gibco, USA) and 1% penicillin/streptomycin (Gibco, USA).

DC2.4 murine dendritic cells were maintained in RPMI-1640 supplemented with 10% heat-inactivated FBS, 20 ng/mL recombinant murine GM-CSF, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37 °C/5% CO₂, with media replenished every 48 h. Hepa1-6 murine hepatocellular carcinoma cells were cultured in DMEM containing 10% FBS under identical conditions. CD8⁺ T cells were isolated from C57BL/6 mouse spleens using a CD8⁺ T Cell Isolation Kit (Miltenyi Biotec, Germany) and pre-activated for 48 h in RPMI-1640 complete medium with 2.5 μg/mL anti-CD3/CD28 antibodies and 100 U/mL recombinant murine IL-2. All cell lines were authenticated before the start of the experiment and were regularly tested for infections, including mycoplasma.

Cell transfection

TKTL1 expression was downregulated in human HCC cell lines by using shRNA (Genechem, China). For a whole day, we incubated 1 × 10⁵ cells in each well of a six-well plate with 2 ml of medium. Subsequently, the medium was increased to 1 ml, and 40 µL of polybrene (Sigma-Aldrich, USA) were combined with the relevant quantity of virus. Following a 12- to 16-h incubation period, the cells were cultivated in standard media and puromycin was added for screening (Beyotime, China). Overexpression experiments involved OE-TKTL1 lentivirus in mice or HCC cells were performed. We used qRT-PCR and Western blotting assays to monitor the transfection effectiveness. Sequences used in this study were listed in supplementary table 1.

TRIM28 knockdown was achieved by transfecting species-specific constructs: mice-targeting siRNA (Transheep, China) into murine DC 2.4 cells. Cells were plated at 1 × 10⁵ density per well in 6-well plates containing 2 ml culture medium and incubated for 24 h. The medium was then exchanged for 1 ml fresh medium supplemented with viral particles and 40 µL polybrene (Sigma-Aldrich, USA). Following 12–16 h incubation, cells underwent puromycin selection in standard culture medium. Transduction efficiency was confirmed through qRT-PCR and Western blot analysis. Sequences used in this study were listed in supplementary table 1.

RNA extraction and qRT-PCR

The Cell/Tissue Total RNA Isolation Kit (Vazyme, China) was used to extract total RNA from tissues and cells. We converted RNA into cDNA using a reverse transcription kit (Vazyme, China). Supplementary table 1 includes a list of all primer sequences. The levels of mRNA expression were normalized using the internal control (β-actin).

Western blotting

Proteins were extracted from the cells using RIPA lysis buffer (Sigma-Aldrich, USA) containing protease and phosphatase inhibitor (Beyotime, China), separated by 10% SDS-PAGE (Sigma-Aldrich, USA), and transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, USA) as per the manufacturer’s operating manual. After blocking the membranes for 30 min with blocking buffer (NCM, China), the membranes were incubated with primary antibodies for a whole night at 4 °C. The ECL signals were seen using an ECL Kit (NCM, China) following a 1-h incubation period with the corresponding secondary antibodies and three TBST-buffered saline washings every 10 min. Antibodies used in this study were listed in Supplementary table 2.

Cell proliferation assay

In the clone forming experiment, HCC cells were seeded in 6-well plates at a density of 1000 cells/well. After 10 days, the cells were fixed using methanol, followed by staining with Giemsa (Sigma-Aldrich, USA), and the colonies were then imaged and counted.

Using Cell-Light 5-ethynyl-2’-deoxyuridine (EdU) DNA Cell Proliferation Kit (RiboBio, China), we performed the EdU experiment to assess cell proliferation. We plated 5 × 10⁴ HCC cells in 24-well plates, and the cells were then cultured for 24 h. The cell lines were fixed with 4% paraformaldehyde after incubating them with 50 mmol/L EdU solution for 2 h. Following the manufacturers protocol, we treated the cell lines with Apollo Dye Solution and DAPI, respectively. EdU cell lines were captured and counted under Olympus FSX100 microscope (Olympus, Japan).

Transwell assay

By adhering to the manufacturer’s guidelines, HCC cells were placed in the upper chambers with 200 μl of serum-free RPMI 1640 medium (Gibco, USA). In preparation for the invasion and migration assays, the transwell chamber (Corning, USA) was coated with a matrigel mixture (BD Biosciences, USA). The bottom chamber was filled with DMEM medium (Gibco, USA) and 10% FBS (Gibco, USA) to serve as a cell attractant. Following a 48-h incubation period, the top chambers were fixed and stained with crystal violet dye (Beyotime, China) for 15 min. Cell counting and imaging were conducted in 3 different fields for visualization purposes.

Wound healing assay

HCC cells were evenly distributed in 6-well culture plates. A standard 20-µl pipette tip was employed to create deliberate wounds on the confluent cell layer. The controlled elimination process aimed to segregate debris and floating cells at the well’s base. Subsequently, a serum-free medium was introduced, and the plate was subjected to a 37 °C incubation. The width of the scratch was measured using an inverted microscope and images were captured at 0 and 48 h. This experimental procedure was repeated thrice to assess both the initial wound width and the extent of cell migration.

Flow cytometry detection and analysis

Single-cell suspensions were prepared from spleens, lymph nodes, and tumor tissues. Initially, all cells were stained with Live/Dead dye (BioLegend) at 4 °C for 30 min. Regarding surface staining, cells were collected and washed twice with phosphate-buffered saline (PBS). Subsequently, they were stained in PBS supplemented with 2% fetal bovine serum (FBS) using antibodies at the recommended dilution on ice for 30 min in the dark. Prior to flow cytometry analysis, the cells were washed twice with PBS. For the intracellular staining of cells to quantify cytokines, cells were harvested and washed twice with PBS. They were then fixed, permeabilized, and stained using the Intracellular Fixation & Permeabilization Buffer set (eBioscience, 88 - 8824) in accordance with the manufacturer’s protocol. Before fluorescence-activated cell sorting (FACS) analysis, the stained cells were washed twice with PBS. Antibody information is detailed in Supplementary table 2. Cells were initially gated based on forward scatter (FSC) and side scatter (SSC) characteristics to exclude debris. Subsequently, FSC - area (FSC - A) and FSC - height (FSC - H) were used for gating, followed by SSC - A and SSC - H to eliminate non-singlet cells. Target cells were then gated to obtain the population of interest through specific staining. Data were acquired using the CytoFLEX S system (Beckman) and analyzed with FlowJo (v.10.5.3).

Metabolomics sequencing and detection

A concoction of methanol, acetonitrile, and water was prepared in a 2:2:1 proportion. After cooling, 1 ml of this blend was sonicated for 60 min in ice water. The sample then underwent an hour-long incubation at −20 °C. Centrifugation at 14,000 × g at a temperature of 4 °C was performed for 20 min to facilitate LC-MS analysis. The evaluation of metabolomics utilized a UPLC-ESI-Q-Orbitrap-MS system (UHPLC, Shimadzu Nexera X2 LC-30AD, Shimadzu, Japan), complemented by Q-Exactive Plus (Thermo Scientific, USA). Data were obtained via electrospray ionization (ESI) in both positive and negative modes. Quality control (QC) samples were made by combining all aliquots, followed by data normalization. Every six samples saw the injection of QC and blank samples (75% acetonitrile in water).

Metabolic flux analysis

Cells were cultured in medium containing uniformly labeled [U-¹³C] glucose for 8–12 h to achieve isotopic steady-state. Upon reaching metabolic equilibrium, pre-cooled extraction solvent was added to the culture. The mixture was vortexed, immediately snap-frozen in liquid nitrogen, and subjected to sonication. Following low-temperature incubation for ≥1 h, the samples were centrifuged at low temperature. The resulting supernatant was collected, lyophilized, and reconstituted in solvent. After thorough dissolution, the solution was centrifuged again under cold conditions, and the final supernatant was transferred to LC-MS vials for analysis. LC-MS signals were matched to corresponding metabolites by comparison with a standard compound database and/or via algorithm-based prediction.

Seahorse assay

The extracellular flux analyzer, Seahorse XF96 (Agilent Seahorse XF Technology, USA), was used to assess the glycolytic activity of the HCC cells. In Seahorse XF96 well plates, cells were seeded at 12,000 cells per well. Right before the assays, these cells were changed from a culture medium to an assay medium and incubated for 1 h at 37 °C. Under baseline conditions and with the addition of Glucose (Agilent, USA), Oligomycin (Agilent, USA), and 2-deoxyglucose (Agilent, USA), the extracellular acidification rate (ECAR) was determined. By graphing the total ECAR as a function of time (mpH/min), respectively, in the XF96 Glycolysis report generator, the total ECAR was calculated. Cells were immediately trypsinized and normalized individual well rate data to protein concentration after the completion of the experiment.

NADPH/NADP⁺ ratio in tumor cells

Human hepatocellular carcinoma cells were cultured in DMEM high-glucose medium with 10% FBS, 100 U/mL penicillin and 100 μg/mL streptomycin at 37 °C with 5% CO₂. Logarithmic phase cells (70–80% confluence) were lysed with pre-cooled RIPA buffer on ice, centrifuged, and the supernatant was collected. NADPH/NADP⁺ contents were detected using Promega kit (G9081): standard curves were prepared, 10 μL supernatant was added to 96-well plates (3 replicates), followed by detection buffer and enzyme mixture, incubated in dark for 30 min, and absorbance was measured at 560 nm. The NADPH/NADP⁺ ratio was calculated from standard curves, with 3 independent repeats

CUT&Tag assay

Freshly harvested cells were counted and aliquoted into microcentrifuge tubes. Following low-speed centrifugation, the supernatant was discarded. Cells were resuspended in Digitonin Permeabilization Buffer and gently agitated. Concanavalin A-coated magnetic beads (ConA beads), pre-equilibrated in Binding Buffer, were added dropwise to the suspension. Samples were gently inverted for 5–10 min. After magnetic separation and buffer removal, cells were resuspended in ice-cold Antibody Buffer. Tubes were incubated on a rotator at room temperature for 2 h. Primary antibody was removed, and cells were washed. A 1:100 dilution of secondary antibody in Digitonin Permeabilization Buffer was added. Tubes were gently agitated and incubated on a rotator at room temperature for 30–60 min. Following washing, the CUT&Tag Transposome complex was diluted 1:250 in Digitonin Permeabilization Buffer 2 and added dropwise to the cells. After gentle agitation, tubes were placed on a magnetic stand. The supernatant was discarded after bead capture. CUT&Tag Activation Buffer (300 µL) was added, tubes were gently agitated, and incubated at 37 °C for 1 h. DNA was extracted (1 h), amplified by PCR (1 h; including a 5 min 72 °C step prior to cycling), purified (30 min), and subjected to sequencing.

Luciferase reporter assay

The full-length human HK2 gene promoter was cloned into the pGL3-basic luciferase reporter vector (Genechem, China) to generate the wild-type reporter plasmid pGL3-HK2-WT. A series of HK2 promoter deletion fragments of different lengths were amplified by PCR and cloned individually into pGL3-basic to identify the minimal core region required for both basal and HIF-1α-induced HK2 promoter activity. Using a site-directed mutagenesis kit (Genechem, China), the HIF-1α core binding site in the HK2 promoter was mutated using pGL3-HK2-WT as the template, generating the mutant reporter plasmid pGL3-HK2-Mut. Reporter plasmids were co-transfected with the pRL-TK Renilla luciferase internal control plasmid into cells. At 48 h post-transfection, luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega, USA), and results were expressed as the ratio of firefly to Renilla luciferase activity.

ChIP-qPCR assay

We performed the ChIP assay using the EZ-Magna ChIP Chromatin Immunoprecipitation Kit (Abcam, UK), according to the manufacturer’s instructions. In brief, we fixed cells in 1% formaldehyde for 10 min at ambient temperature. The fixed cells were harvested, lysed, and sonicated for 10 cycles of 10 s ON/20 s OFF and 50% AMPL using Sonics VCX130 (Sonics & Materials, USA). Antibodies against HIF-1α and rabbit IgG (Thermo Fisher, USA) were used for immunoprecipitation. PCR amplification of the precipitated DNA was performed.

Immunoprecipitation (IP) assay

Cells were introduced with the specified plasmids and allowed to incubate for a period of 24 h. Following this, the cells underwent lysis in an IP lysis buffer, which had protease and phosphatase inhibitor tablets, for half an hour. Post centrifugation at 12,000 rpm at a temperature of 4 °C, the supernatant, which contained the protein, was subjected to overnight incubation with protein G agarose beads (Beyotime, China) and specific antibodies at 4 °C. Subsequently, the beads were thoroughly rinsed three times with buffers having 300 mM and 150 mM NaCl and then were treated with a 2×SDS loading buffer at 4 °C before executing the immunoblotting process. The information of antibodies could be found in Supplementary table 2.

Immunoprecipitation-mass spectrometry (IP-MS)

Target protein gel bands were excised, destained (50% ACN/50 mM NH₄HCO₃), and dehydrated (100% ACN). Reduction used 10 mM TCEP (37 °C, 30 min); alkylation used 25 mM iodoacetamide (RT, dark, 30 min). After washing (50 mM NH₄HCO₃) and dehydration, in-gel digestion used 2 μg trypsin (50 mM NH₄HCO₃, 37 °C, overnight). Peptides were extracted (50% ACN/0.1% FA), SpeedVac-dried, and resuspended in 0.1% FA. LC separation (EASY nLC-1200, 400 nL/min) used: 2–7% B (1 min), 7–35% B (35 min), 35–55% B (9 min), 55–100% B (7 min), 100% B (8 min). Mobile phase A: 0.1% FA; B: 0.1% FA/80% ACN. MS (Q Exactive HF-X) used nano-ESI (2.0 kV). Full MS: Orbitrap, 60,000 resolution. MS/MS: Top 20 precursors, NCE 27, Orbitrap 30,000 resolution, 20 s dynamic exclusion. Data searched via Proteome Discoverer 2.4 against SwissProt Human (2023_09) with Trypsin/P (≤ 2 miscleavages).

GST pull-down assay

Purified recombinant protein (Abcam, UK) was incubated with glutathione agarose gel at 4 °C for 1 h, followed by washing with binding buffer. Target protein was added and incubated for another hour under the same conditions. The protein complex was washed and eluted using elution buffer. Eluted proteins were analyzed by SDS-PAGE and Western blotting. GST (glutathione S-transferase) protein was used as a negative control.

Mass cytometry

Peripheral blood mononuclear cells (PBMCs) from 18 samples (from 9 HCC patients) underwent CyTOF analysis. CyTOF staining steps comprised 194Pt staining → Fc block → surface antibody staining → overnight DNA staining (191/193Ir) → intracellular antibody staining → collecting data on the computer. The data analysis steps comprised A. FlowJo pre-treatment: circle and select single, live and complete CD45⁺ immune cells; B. Bio-information analysis: X-shift algorithm performs cell subpopulation clustering, manual annotation, TSNE dimensionality reduction visual display, as well as statistical analysis.

Chemokine array

Prepared samples were uniformly applied to the array membrane, ensuring complete coverage of all spot wells, and incubated at room temperature for 1–2 h to facilitate specific binding of chemokines to the immobilized capture antibodies. Unbound components were subsequently removed by thorough washing with wash buffer. If required, a blocking agent was applied to minimize non-specific binding and reduce background noise. The membrane was then incubated with a biotinylated or enzyme-conjugated detection antibody specific for the captured chemokines, forming an antibody-chemokine-detection antibody sandwich complex. Finally, bound complexes were detected using an appropriate method.

Lactylation detection

Lactylated proteins were immunoprecipitated from sample lysates using a lactylation-specific antibody to form antigen-antibody complexes. Biotin-conjugated secondary antibodies were then applied, and lactylated targets were specifically enriched via streptavidin-biotin affinity purification. Enriched proteins were resolved by SDS-PAGE according to molecular weight and subsequently transferred to PVDF membranes. Membranes were probed with a primary antibody specific for lactylation modifications, followed by an appropriate HRP- or fluorophore-conjugated secondary antibody. Immunocomplexes were detected using chemiluminescent or fluorescent imaging systems, enabling qualitative and semi-quantitative analysis of protein lactylation.

Prediction of potential allosteric sites of TKTL1 and virtual screening

Using AlphaFold, we estimated potential allosteric sites within the TKTL1 protein structure. To identify promising compounds, we conducted virtual screening using two commercial chemical libraries, each containing over 67,000 compounds. The screening process employed grid-based ligand docking facilitated by the GLIDE software (Schrödinger Maestro 11.4), meticulously targeting the expected locations. Subsequently, we procured the top 1 compounds with the highest scores from MCE for experimental investigation.

Microscale thermophoresis (MST)

With the use of conventional capillaries and Monolith NT.115 (NanoTemper Technologies GmbH, Germany), MST was used to ascertain the binding affinity of TKTL1 with Quercetin 3-glc-7-gent. Lysates from HEK-293T cells expressing GFP-TKTL1 were used as a source of fluorescently labeled TKTL1. HEK-293T cells were transfected with GFP-fused TKTL1 and lysed 24 h after transfection. To assess the binding of Quercetin 3-glc-7-gent to TKTL1, cell lysates were diluted 1.5-fold with MST buffer (10 mM Na-phosphate buffer, pH 7.4, 1 mM MgCl₂, 3 mM KCl, 150 mM NaCl, 0.05% Tween-20) to provide optimal fluorescence levels. Quercetin 3-glc-7-gent titration series (0–1 mM) were incubated with diluted cell lysates. Measurements were performed in a high-quality coated capillary (NanoTemper Technologies GmbH, Germany) using an LED light source at 470 nm with 100% IR laser power and a temperature of 25 °C. The fluorescence signal was normalized, and the Hill equation was fitted using MO Affinity Analysis v2.1.3 software (NanoTemper Technologies GmbH, Germany).

Synthesis of the CQLH nanosystem

Synthesis of Hollow Mesoporous MnO₂ Nanoparticles (H-MnO₂ NPs): Hollow mesoporous MnO₂ nanoparticles were synthesized via a modified hydrothermal method. Briefly, 0.5 g KMnO₄ and 0.3 g CTAB were dissolved in 40 mL deionized water under vigorous stirring. The solution was transferred to a Teflon-lined autoclave and heated at 160 °C for 6 h. The resulting brown precipitate was collected by centrifugation (12,000 rpm, 15 min), washed three times with ethanol/water (1:1 v/v), and dried at 60 °C overnight.

Co-Loading of Quercetin 3-glc-7-gent and Lactate Oxidase (LOX): H-MnO₂ NPs (10 mg) were dispersed in 5 mL PBS (pH 7.4) containing Quercetin 3-glc-7-gent (1 mg/mL) and LOX (20 μg/mg NPs). The mixture was gently agitated at 4 °C for 24 h. The co-loaded nanoparticles (denoted as QL@MnO₂) were collected by centrifugation (10,000 rpm, 10 min) and washed twice with PBS to remove unbound cargo.

HCC Cell Membrane Coating: Cell membranes were isolated from Hepa1-6 HCC cells using hypotonic lysis and differential centrifugation. Membrane protein concentration was quantified by BCA assay. QL@MnO₂ nanoparticles were mixed with HCC membrane fragments (1:1 w/w protein ratio) and co-extruded 10× through a 200-nm polycarbonate membrane using an Avanti mini-extruder to form the biomimetic CQLH nanosystem.

Characterization of the CQLH nanosystem

TEM: Morphology of H-MnO₂ NPs and CQLH was analyzed by transmission electron microscopy (JEOL JEM-2100, 200 kV). Samples were drop-cast onto copper grids and air-dried.

SEM-EDS: elemental mapping (Mn, O, C, N) was performed using SEM coupled with EDS

Size and Zeta Potential: Hydrodynamic diameter and surface charge of nanoparticles (H-MnO₂, QL@MnO₂, CQLH) were measured by dynamic light scattering (DLS, Malvern Zetasizer Nano ZS) in PBS.

Loading Efficiency: Quercetin 3-glc-7-gent loading was quantified via UV-Vis spectrophotometry (λ = 370 nm) after dissolving nanoparticles in DMSO. LOX loading was determined by BCA assay of the unbound supernatant.

pH-Responsive Release: Quercetin 3-glc-7-gent release from CQLH was evaluated in PBS at pH 7.4, 6.5, and 5.0 using dialysis (MWCO 14 kDa). Samples were analyzed by HPLC at predetermined intervals.

Membrane Coating Validation: Protein profiles of HCC cell membranes and CQLH were compared by 10% SDS-PAGE.

AFP Immunoblotting: α-Fetoprotein (AFP) expression on CQLH was confirmed by Western blot using anti-AFP antibody (1:1000 dilution).

Mn²⁺ Release: Mn²⁺ ions generated by CQLH in acetate buffer (pH 5.0) were quantified via o-phenanthroline assay and ICP-MS.

Catalytic Activity: LOX enzymatic activity was measured spectrophotometrically (λ = 570 nm) using an Amplex™ Red Lactate Assay Kit.

Statistical analysis

All experiments were carried out based on independent biological replicates. Statistical analysis was performed using GraphPad Prism 10.0 software, with p < 0.05 indicating statistically significant differences. The specific methods are as follows: For comparison between two groups, an unpaired t-test was used to analyze the differences between biological replicate samples of the two groups; for comparison among multiple groups, one-way ANOVA or two-way ANOVA was used to analyze the differences among biological replicate samples of three or more groups; for survival analysis, the Kaplan–Meier method was used to draw survival curves, and the log-rank test was used to compare the survival differences among biological replicate samples of different groups; for correlation analysis, the Pearson correlation coefficient was used to evaluate the linear association between two variables in biological replicate samples. The detailed results of all statistical analyses (including, test type, p-value, etc.) are indicated in the legends of the corresponding figures.

Schematic generation tools

The illustrations in this article were created using Figure Draw and BioRender software, for which authorization licenses have been obtained.

Patient and public involvement

Patients and the public were not involved in this work.

Ethics statement

All human-derived samples were approved by the Ethics Review Committee of the First Affiliated Hospital of Nanjing Medical University (2019-SR-332). All in vivo animal experiments were approved by the Committee on the Ethics of Animal Experiments of Nanjing Medical University (IACUC-2404094).

Supplementary information

Supplementary^{(9.9MB, docx)}

Western blot.docx^{(5.3MB, docx)}

Acknowledgements

This work was supported by the grants from National Natural Science Foundation of China (No. 82373265, No. 31930020), Jiangsu Provincial Frontier Technology Research and Development Program (No. BF2024053), the Major Basic Research Fund of Jiangsu Province Hospital (No. TS202404), National Health Commission Project (WKZX2023CX020009), Funding of Jiangsu Province Hospital (JSPH-MA-2023-9), the Nanjing Municipal Science and Technology Bureau (Grant No. 202511068).

Author contributions

There are 5 first authors in this manuscript and they have equally contributed to this project. Dr. H.S.C., T.H., Y.C., Z.Q.L. and S.P.D. were responsible for designing and performing the experiments. All other authors participated in the conduct of the experiment. Furthermore, we have 3 corresponding authors in this manuscript. Dr. Y.X.Y., W.W.T. and X.H.W. have contributed to study design and critical revision of the manuscript. All authors read and approved the final manuscript. We would like to thank the Core Facility of the First Affiliated Hospital of Nanjing Medical University for its help in the experiment. All authors have read and approved the final manuscript.

Data availability

Animal sequencing data are available in GSA (accession: CRA043895). Raw human genomic and identifiable clinical data are restricted by ethics committee approval (2019-SR-332) for privacy protection. Non-commercial research access may be requested from the corresponding author with institutional ethics review.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Hengsong Cao, Tian Huang, Yue Chai, Zhengqing Lu, Shipeng Dai.

Contributor Information

Xuehao Wang, Email: wangxh@njmu.edu.cn.

Weiwei Tang, Email: 1243773473twww@sina.com.

Yongxiang Xia, Email: yx_xia@njmu.edu.cn.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41392-026-02875-2.

References

1.Sung, H. et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin.71, 209–249 (2021). [DOI] [PubMed] [Google Scholar]
2.Kudo, M. et al. Lenvatinib versus sorafenib in first-line treatment of patients with unresectable hepatocellular carcinoma: a randomised phase 3 non-inferiority trial. Lancet391, 1163–1173 (2018). [DOI] [PubMed] [Google Scholar]
3.Finn, R. S. et al. Atezolizumab plus bevacizumab in unresectable hepatocellular carcinoma. N. Engl. J. Med.382, 1894–1905 (2020). [DOI] [PubMed] [Google Scholar]
4.Xia, Y. et al. Efficacy and safety of camrelizumab plus apatinib during the perioperative period in resectable hepatocellular carcinoma: a single-arm, open label, phase II clinical trial. J. Immunother Cancer10, e004656 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Shi, X. et al. Mechanism insights and therapeutic intervention of tumor metastasis: latest developments and perspectives. Signal Transduct. Target Ther.9, 192 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Sharma, P. et al. Primary, adaptive, and acquired resistance to cancer immunotherapy. Cell168, 707–723 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Kim, T. K. et al. Adaptive immune resistance at the tumour site: mechanisms and therapeutic opportunities. Nat. Rev. Drug Discov.21, 529–540 (2022). [DOI] [PubMed] [Google Scholar]
8.Reck, M. et al. Updated analysis of KEYNOTE-024: pembrolizumab versus platinum-based chemotherapy for advanced non-small-cell lung cancer with PD-L1 tumor proportion score of 50% or greater. J. Clin. Oncol.37, 537–546 (2019). [DOI] [PubMed] [Google Scholar]
9.Muller, K. et al. What’s new in breast pathology 2022: WHO 5th edition and biomarker updates. J. Pathol. Transl. Med.56, 170–171 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Leone, A. G. et al. Efficacy and activity of PD-1 blockade in patients with advanced esophageal squamous cell carcinoma: a systematic review and meta-analysis with focus on the value of PD-L1 combined positive score. ESMO Open7, 100380 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Pinson, A. et al. Human TKTL1 implies greater neurogenesis in frontal neocortex of modern humans than Neanderthals. Science377, eabl6422 (2022). [DOI] [PubMed] [Google Scholar]
12.Hao, S. et al. The role of transketolase in human cancer progression and therapy. Biomed. Pharmacother.154, 113607 (2022). [DOI] [PubMed] [Google Scholar]
13.Guo, D. et al. Aerobic glycolysis promotes tumor immune evasion by hexokinase2-mediated phosphorylation of IκBα. Cell Metab.34, 1312–1324.e1316 (2022). [DOI] [PubMed] [Google Scholar]
14.Chen, J. et al. Lactate and lactylation in cancer. Signal Transduct. Target Ther.10, 38 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Ma, X. et al. TRIM28 promotes the escape of gastric cancer cells from immune surveillance by increasing PD-L1 abundance. Signal Transduct. Target Ther.8, 246 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Gaffney, D. O. et al. Non-enzymatic lysine lactylation of glycolytic enzymes. Cell Chem. Biol.27, 206–213.e206 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Mellman, I. et al. The cancer-immunity cycle: indication, genotype, and immunotype. Immunity56, 2188–2205 (2023). [DOI] [PubMed] [Google Scholar]
18.Doroshow, D. B. et al. PD-L1 as a biomarker of response to immune-checkpoint inhibitors. Nat. Rev. Clin. Oncol.18, 345–362 (2021). [DOI] [PubMed] [Google Scholar]
19.Yamaguchi, H. et al. Mechanisms regulating PD-L1 expression in cancers and associated opportunities for novel small-molecule therapeutics. Nat. Rev. Clin. Oncol.19, 287–305 (2022). [DOI] [PubMed] [Google Scholar]
20.Li, Q. et al. PRDM1/BLIMP1 induces cancer immune evasion by modulating the USP22-SPI1-PD-L1 axis in hepatocellular carcinoma cells. Nat. Commun.13, 7677 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Zhu, Y. et al. TKTL1 participated in malignant progression of cervical cancer cells via regulating AKT signal mediated PFKFB3 and thus regulating glycolysis. Cancer Cell Int.21, 678 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Sun, W. et al. TKTL1 is activated by promoter hypomethylation and contributes to head and neck squamous cell carcinoma carcinogenesis through increased aerobic glycolysis and HIF1alpha stabilization. Clin. Cancer Res.16, 857–866 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Chen, Y. et al. Mechanisms governing target search and binding dynamics of hypoxia-inducible factors. Elife11, e75064 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Bouthelier, A. et al. Differential contribution of N- and C-terminal regions of HIF1α and HIF2α to their target gene selectivity. Int. J. Mol. Sci.21, 9401 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Huang, Z. W. et al. STAT5 promotes PD-L1 expression by facilitating histone lactylation to drive immunosuppression in acute myeloid leukemia. Signal Transduct. Target Ther.8, 391 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Hammerl, D. et al. Spatial immunophenotypes predict response to anti-PD1 treatment and capture distinct paths of T cell evasion in triple negative breast cancer. Nat. Commun.12, 5668 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Mayoux, M. et al. Dendritic cells dictate responses to PD-L1 blockade cancer immunotherapy. Sci. Transl. Med.12, eaav7431 (2020). [DOI] [PubMed] [Google Scholar]
28.Gao, F. et al. Intra/extracellular lactic acid exhaustion for synergistic metabolic therapy and immunotherapy of tumors. Adv. Mater.31, e1904639 (2019). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary^{(9.9MB, docx)}

Western blot.docx^{(5.3MB, docx)}

Data Availability Statement

[CR1] 1.Sung, H. et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin.71, 209–249 (2021). [DOI] [PubMed] [Google Scholar]

[CR2] 2.Kudo, M. et al. Lenvatinib versus sorafenib in first-line treatment of patients with unresectable hepatocellular carcinoma: a randomised phase 3 non-inferiority trial. Lancet391, 1163–1173 (2018). [DOI] [PubMed] [Google Scholar]

[CR3] 3.Finn, R. S. et al. Atezolizumab plus bevacizumab in unresectable hepatocellular carcinoma. N. Engl. J. Med.382, 1894–1905 (2020). [DOI] [PubMed] [Google Scholar]

[CR4] 4.Xia, Y. et al. Efficacy and safety of camrelizumab plus apatinib during the perioperative period in resectable hepatocellular carcinoma: a single-arm, open label, phase II clinical trial. J. Immunother Cancer10, e004656 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR5] 5.Shi, X. et al. Mechanism insights and therapeutic intervention of tumor metastasis: latest developments and perspectives. Signal Transduct. Target Ther.9, 192 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR6] 6.Sharma, P. et al. Primary, adaptive, and acquired resistance to cancer immunotherapy. Cell168, 707–723 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Kim, T. K. et al. Adaptive immune resistance at the tumour site: mechanisms and therapeutic opportunities. Nat. Rev. Drug Discov.21, 529–540 (2022). [DOI] [PubMed] [Google Scholar]

[CR8] 8.Reck, M. et al. Updated analysis of KEYNOTE-024: pembrolizumab versus platinum-based chemotherapy for advanced non-small-cell lung cancer with PD-L1 tumor proportion score of 50% or greater. J. Clin. Oncol.37, 537–546 (2019). [DOI] [PubMed] [Google Scholar]

[CR9] 9.Muller, K. et al. What’s new in breast pathology 2022: WHO 5th edition and biomarker updates. J. Pathol. Transl. Med.56, 170–171 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Leone, A. G. et al. Efficacy and activity of PD-1 blockade in patients with advanced esophageal squamous cell carcinoma: a systematic review and meta-analysis with focus on the value of PD-L1 combined positive score. ESMO Open7, 100380 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Pinson, A. et al. Human TKTL1 implies greater neurogenesis in frontal neocortex of modern humans than Neanderthals. Science377, eabl6422 (2022). [DOI] [PubMed] [Google Scholar]

[CR12] 12.Hao, S. et al. The role of transketolase in human cancer progression and therapy. Biomed. Pharmacother.154, 113607 (2022). [DOI] [PubMed] [Google Scholar]

[CR13] 13.Guo, D. et al. Aerobic glycolysis promotes tumor immune evasion by hexokinase2-mediated phosphorylation of IκBα. Cell Metab.34, 1312–1324.e1316 (2022). [DOI] [PubMed] [Google Scholar]

[CR14] 14.Chen, J. et al. Lactate and lactylation in cancer. Signal Transduct. Target Ther.10, 38 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR15] 15.Ma, X. et al. TRIM28 promotes the escape of gastric cancer cells from immune surveillance by increasing PD-L1 abundance. Signal Transduct. Target Ther.8, 246 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Gaffney, D. O. et al. Non-enzymatic lysine lactylation of glycolytic enzymes. Cell Chem. Biol.27, 206–213.e206 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR17] 17.Mellman, I. et al. The cancer-immunity cycle: indication, genotype, and immunotype. Immunity56, 2188–2205 (2023). [DOI] [PubMed] [Google Scholar]

[CR18] 18.Doroshow, D. B. et al. PD-L1 as a biomarker of response to immune-checkpoint inhibitors. Nat. Rev. Clin. Oncol.18, 345–362 (2021). [DOI] [PubMed] [Google Scholar]

[CR19] 19.Yamaguchi, H. et al. Mechanisms regulating PD-L1 expression in cancers and associated opportunities for novel small-molecule therapeutics. Nat. Rev. Clin. Oncol.19, 287–305 (2022). [DOI] [PubMed] [Google Scholar]

[CR20] 20.Li, Q. et al. PRDM1/BLIMP1 induces cancer immune evasion by modulating the USP22-SPI1-PD-L1 axis in hepatocellular carcinoma cells. Nat. Commun.13, 7677 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Zhu, Y. et al. TKTL1 participated in malignant progression of cervical cancer cells via regulating AKT signal mediated PFKFB3 and thus regulating glycolysis. Cancer Cell Int.21, 678 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Sun, W. et al. TKTL1 is activated by promoter hypomethylation and contributes to head and neck squamous cell carcinoma carcinogenesis through increased aerobic glycolysis and HIF1alpha stabilization. Clin. Cancer Res.16, 857–866 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Chen, Y. et al. Mechanisms governing target search and binding dynamics of hypoxia-inducible factors. Elife11, e75064 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Bouthelier, A. et al. Differential contribution of N- and C-terminal regions of HIF1α and HIF2α to their target gene selectivity. Int. J. Mol. Sci.21, 9401 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Huang, Z. W. et al. STAT5 promotes PD-L1 expression by facilitating histone lactylation to drive immunosuppression in acute myeloid leukemia. Signal Transduct. Target Ther.8, 391 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR26] 26.Hammerl, D. et al. Spatial immunophenotypes predict response to anti-PD1 treatment and capture distinct paths of T cell evasion in triple negative breast cancer. Nat. Commun.12, 5668 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Mayoux, M. et al. Dendritic cells dictate responses to PD-L1 blockade cancer immunotherapy. Sci. Transl. Med.12, eaav7431 (2020). [DOI] [PubMed] [Google Scholar]

[CR28] 28.Gao, F. et al. Intra/extracellular lactic acid exhaustion for synergistic metabolic therapy and immunotherapy of tumors. Adv. Mater.31, e1904639 (2019). [DOI] [PubMed] [Google Scholar]

PERMALINK

Transketolase-like 1 potentiates PD-1 blockade in hepatocellular carcinoma by glycolysis to prime dendritic cell lactylation

Hengsong Cao

Tian Huang

Yue Chai

Zhengqing Lu

Shipeng Dai

Danyang Xu

Yanshu He

Na Yu

Yuliang Wang

Ziyuan Liu

Yidong Xia

Jie Zhao

Hongyu Wang

Dawei Rong

Yuchen Xie

Chongyu Wang

Pengyu Chen

Juyue Zhou

Li Liu

Zhuozheng Li

Hanyuan Liu

Xiao Li

Xuehao Wang

Weiwei Tang

Yongxiang Xia

Abstract

Introduction

Results

Integrated multi-omics profiling unveils TKTL1 upregulation in anti-PD-1 responders in HCC

Fig. 1.

TKTL1 drives tumor progression while potentiating anti-PD-1 immunotherapy sensitivity rather than apatinib in HCC

Fig. 2.

TKTL1 reprograms glucose metabolism through glycolysis instead of the PPP to stabilize the HIF-1α/HK2/PD-L1 axis and drive anti-PD-1 sensitivity in HCC

Fig. 3.

TKTL1 activates glycolysis pathway by promoting HIF-1α deubiquitination through recruitment of USP9X in HCC

Fig. 4.

TKTL1 enhances the sensitivity of anti-PD-1 therapy by recruiting PD-L1high DCs via the chemokine CCL4 in HCC

Fig. 5.

TKTL1-mediated lactate-enriched TME promotes TRIM28K408 lactylation in DCs, thereby stabilizing PD-L1

Fig. 6.

Targeted co-delivery of que and lactate oxidase via biomimetic nanosystem for synergistic TKTL1 suppression and TME remodeling in HCC

Fig. 7.

CQLH nanodrug enhanced anti-PD-1 therapy via dual TKTL1/lactate inhibition and TME remodeling in HCC models

Fig. 8.

Discussion

Materials and methods

Clinical specimens and patient cohorts

RNA and proteomics sequencing

Immunohistochemistry (IHC) and immunofluorescence (IF)

Animal models

Subcutaneous tumor model establishment

Orthotopic transplantation model

DEN&CCL4-induced orthotopic hepatocarcinogenesis model

The PDTX model

Single-cell RNA sequencing and mass cytometry analysis in mouse tumor tissues

Cell cultures

Cell transfection

RNA extraction and qRT-PCR

Western blotting

Cell proliferation assay

Transwell assay

Wound healing assay

Flow cytometry detection and analysis

Metabolomics sequencing and detection

Metabolic flux analysis

Seahorse assay

NADPH/NADP⁺ ratio in tumor cells

CUT&Tag assay

Luciferase reporter assay

ChIP-qPCR assay

Immunoprecipitation (IP) assay

Immunoprecipitation-mass spectrometry (IP-MS)

GST pull-down assay

Mass cytometry

Chemokine array

Lactylation detection

Prediction of potential allosteric sites of TKTL1 and virtual screening

Microscale thermophoresis (MST)

TKTL1 enhances the sensitivity of anti-PD-1 therapy by recruiting PD-L1^high DCs via the chemokine CCL4 in HCC

TKTL1-mediated lactate-enriched TME promotes TRIM28^K408 lactylation in DCs, thereby stabilizing PD-L1