Nuclear translocation of metabolic enzyme PKM2 participates in high glucose-promoted HCC metastasis by strengthening immunosuppressive environment

Abstract

Although some cohort studies have indicated a close association between diabetes and HCC, the underlying mechanism about the contribution of diabetes to HCC progression remains largely unknown. In the study, we applied a novel HCC model in SD rat with diabetes and a series of high glucose-stimulated cell experiments to explore the effect of a high glucose environment on HCC metastasis and its relevant mechanism. Our results uncovered a novel regulatory mechanism by which nuclear translocation of metabolic enzyme PKM2 mediated high glucose-promoted HCC metastasis. Specifically, high glucose-increased PKM2 nuclear translocation downregulates chemerin expression through the redox protein TRX1, and then strengthens immunosuppressive environment to promote HCC metastasis. To the best of our knowledge, this is the first report to elucidate the great contribution of a high glucose environment to HCC metastasis from a new perspective of enhancing the immunosuppressive microenvironment. Simultaneously, this work also highlights a previously unidentified non-metabolic role of PKM2 and opens a novel avenue for cross research and intervention for individuals with HCC and comorbid diabetes.

Graphical abstract

A diagram illustrating the mechanism through which hyperglycemia contributions to HCC metastasis.

Highlights

• •Diabetes promotes HCC metastasis by strengthening the immunosuppressive microenvironment.
• •High glucose level obviously enhances PKM2 nuclear translocation via its lactylation.
• •Nuclear PKM2 binding to TRX1 significantly downregulates the expression of chemerin.

1. Introduction

The prevalence of diabetes among patients with nonviral etiology HCC (hepatocellular carcinoma) has greatly increased over the past twenty years [1]. The risk of developing HCC in individuals with diabetes is much higher than that in non-diabetes individuals, regardless of the presence of the underlying liver disease cause [[2], [3], [4]]. On the other hand, diabetic patients have a significantly higher HCC mortality rate, more than 1.5 times that of non-diabetic individuals [5], and diabetes has become an autonomous risk factor to indicate unfavorable prognosis of HCC patients [[6], [7], [8]]. Thereby, hyperglycemia, a prominent manifestation of diabetes, may serve as an important initiator to participate in or promote HCC development and progression. Oxidative stress, hypoxia-inducible factor-1α (HIF-1α), glycation, and epigenetic alterations have been documented to explain the association between hyperglycemia and tumor progression [[9], [10], [11], [12]]. However, whether hyperglycemia enhances the strength of immunosuppression microenvironment and subsequently contributes to HCC metastasis remains largely uncharacterized. It is generally known that immunosuppressive microenvironment is a typical microenvironment characteristic of HCC [13,14], and the accumulation of infiltrated immunosuppressive cells is often taken as biomarker to reflect the strength of immunosuppressive microenvironment [15,16]. Analysis of infiltrated immunosuppressive cells in HCC tissue reveals that the increase of regulatory T cells (Tregs), Macrophage M2 cells, and myeloid-derived suppressor cells (MDSCs) within the tumor microenvironment hinders anti-tumor immune responses and facilitates HCC growth and metastasis [17]. On the contrary, targeting infiltrated immunosuppressive cells such as Macrophages M2 can trigger a T cell-mediated antitumor immune response and impede HCC metastasis [18]. These findings sufficiently suggest that infiltrated immunosuppressive cells play a pivotal role in promoting HCC metastasis. Types and levels of chemokines generally determine the accumulation of immunosuppressive cells. Pro-tumor chemokines (such as CCL22 and CCL17) can effectively recruit immunosuppressive cells into tumor microenvironment to augment its immune suppressive nature [[19], [20], [21]], and the presence of antitumor chemokines (such as CXCL10 and CXCL16) facilitates the migration of effector immune cells [22,23]. Thereby, chemokines derived from tumor generally possess a strong ability to recruit immunosuppressive cells, which is beneficial to enhancing the strength of immunosuppressive microenvironment and subsequently promoting HCC metastasis.

The primary role of glucose metabolism enzymes in tumor progression is to fulfill the energy demand of accelerated tumor growth through their classical metabolic function. Recent evidence shows that metabolic enzymes also have non-classical/non-metabolic functions and they can undergo nuclear translocation to alter the modification of histone and other proteins, thereby influencing gene transcription [24,25]. Except that, nuclear translocation of metabolic enzymes also exerts significant impacts on DNA repair, cell proliferation, cell apoptosis, and the regulation of the tumor microenvironment [26]. Thus, metabolic enzymes can effectively regulate gene transcription and govern disease progression via their nuclear translocation. An elevated concentration of glucose typically enhances glucose metabolism and promotes tumor growth in HCC [[27], [28], [29]]. However, little is known about whether metabolic enzymes undergo nuclear translocation in HCC cells under high glucose conditions, especially whether nuclear accumulation of metabolic enzymes influences the expression and secretion of chemokines. Considering that some external stimuli, such as IL3 (interleukin 3), H₂O₂ (Hydrogen peroxide), and ultraviolet radiation, not only affect glucose metabolism but also cause nuclear translocation of metabolic enzymes by post-translational modifications and other pathways [30,31], we here hypothesize that high glucose environment can effectively increase nuclear accumulation of metabolic enzymes to modulate the expressions of pro-tumor or antitumor chemokines, and then strengthens the immunosuppressive microenvironment to facilitate HCC metastasis.

In the study, we used a novel HCC model in SD rat with diabetes and a series of cell experiments with high glucose intervention to explore the impact of hyperglycemia on HCC metastasis and its relevant molecular mechanism. Our data unlocked a new mechanism by which high glucose stimulation obviously promotes nuclear translocation of PKM2 (Pyruvate kinase isozyme type M2) through changing its lactylation modification, and nuclear PKM2 interacting with the redox protein TRX1 (Thioredoxin-1) downregulates chemerin expression, thereby enhancing the immunosuppressive microenvironment to boost HCC metastasis. These findings highlight novel potential intervention targets for HCC patients with diabetes.

2. Results

2.1. Hyperglycemia obviously facilitates HCC metastasis in vivo

By joint modeling approach (Fig. 1A), we creatively developed a new HCC model in SD rat with diabetes to explore whether hyperglycemia affected HCC metastasis. As depicted in a flow chart of animal experiment design (Fig. 1A), SD rat models with diabetes were first induced using intraperitoneal injection of STZ (streptozocin), and then HCC model in SD rats with diabetes were further formed by hepatic subcapsular injection of the suspended McA-RH7777 cells in Matrigel combined with short-term dexamethasone intervention. The established HCC rat models were divided into four groups including group HCC-Control, group HCC-DM (diabetes), group HCC-DM + INS (insulin) and group HCC-DM + CANA (canagliflozin). The rat models in the group HCC-DM + INS were additionally intervened with subcutaneous injection of insulin, and the rat models in the group HCC-DM + CANA with oral administration of canagliflozin (Fig. 1B). In the group HCC-Control, the level of random blood glucose was always distributed in the normal range during the entire animal experiment. However, in three groups of HCC-DM, significant increases in the levels of random blood glucose were detected on the 3rd to 4th day after STZ injection, and high levels of random blood glucose (≥16.7 mmol/L) were consistently maintained until the end of animal experiment (Fig. 1C), indicating that the establishment of diabetic rat models is successful. On the other hand, orthotopic HCC models, which grew on the liver of diabetic rats, were also successfully obtained in all groups and confirmed by pathology (Extended Data Fig. 1A and B). Compared with those in the group HCC-Control, the rats in three groups of HCC-DM were all found to have significant weight loss (Fig. 1D). After the intervention with hypoglycemic agents, the random blood glucose levels in both the group HCC-DM + INS and the group HCC-DM + CANA decreased to the normal range (Fig. 1C). By comparative analysis of tumor size and metastasis occurrence in HCC animal models among four groups, we discovered that the size of HCC tumors in group HCC-DM was remarkably larger than that of the other three groups (Fig. 1E and F), indicating that hyperglycemia promotes the growth of HCC tumor. More importantly, the incidence of lung metastasis in group HCC-DM (5/6) was much higher than that in group HCC-Control (1/6) and in two intervention groups (2/6 and 1/6). Meanwhile, compared with group HCC-DM, two intervention groups exhibited a reduced occurrence of lung metastasis and a smaller size in HCC tumor (Fig. 1G). These results suggest that in addition to promoting tumor growth, hyperglycemia also facilitates HCC metastasis. Consistently, cell tests in vitro revealed that high glucose levels significantly improved cell migration and metastasis-associated genes expressions (Extended Data Fig. 1C–F), also supporting that high glucose environment facilitates HCC metastasis. Taken together, there exists a positive correlation between hyperglycemia and HCC progression, and hyperglycemia may serve as an important initiator to promote HCC metastasis.

Fig. 1.

Hyperglycemia obviously facilitates HCC metastasis in vivo (A) A flow chart depicting the process of establishing a new SD rat HCC model with diabetes. (B) A diagram illustrating the animal grouping and experiment protocol. (C–D) The average random blood glucose level and body weight of four distinct animal groups at various time points throughout the entire experiment. The data present represented mean ± standard deviation (n = 6 per group). (E) A photograph of orthotopic HCC tumors in four animal groups. (F) The average weight and volume of HCC tumors across four distinct animal groups (n = 6 per group). (G) The occurrence of lung metastasis of orthotopic HCC tumors across four distinct animal groups. *p < 0.05, **p < 0.01. Abbreviations: STZ, streptozotocin; RBG, random blood glucose; DM, Diabetes Mellitus; HCC, Hepatocellular carcinoma; SD, Sprague-Dawley. INS, insulin; CANA, canagliflozin.

2.2. Identification of differentially expressed chemokines under high glucose conditions and their relationship with immunosuppressive microenvironment in HCC tissues

The accumulation of infiltrated immunosuppressive cells in tumor microenvironment frequently reflects the strength of immunosuppressive microenvironment, and enhanced immunosuppressive microenvironment can greatly contribute to tumor growth and metastasis [17]. Chemokines released from tumor cells effectively modulate the recruitment and function of immune cells and become an important factor determining the strength of immunosuppressive microenvironment [32,33]. So, chemokines are believed to be closely associated with infiltrated immunosuppressive cells and tumor metastasis. Since the above results in animal experiment have demonstrated that hyperglycemia obviously promotes HCC metastasis, we speculate that high glucose environment changes chemokine expression/secretion to strengthen the immunosuppressive microenvironment, thereby facilitating HCC metastasis. We respectively collected the conditioned media from HCC cells under high glucose and normal glucose conditions (HG-CM and NG-CM), and utilized a human cytokine/chemokine microarray to screen differentially expressed chemokines between HG-CM and NG-CM. We found a total of 10 differentially expressed chemokines (fold change >1.2 or < 0.8), including 7 downregulated chemokines (chemerin, CXCL16, CCL1, CXCL10, CCL26, CCL18, CX3CL1) and 3 upregulated chemokines (CCL22, CCL17, CCL28) in HG-CM (Fig. 2A). Chemokines such as chemerin, CXCL16, and CXCL10 were reported to have a strong ability to promote immunity in tumors through inhibiting the infiltration of immunosuppressive cells [34] and recruiting the accumulation of effector T cells [22,23,35], while chemokines such as CCL22, CCL17, and CCL28 enhance immune escape in HCC by recruiting Treg cells [[36], [37], [38]], suggesting a potential linkage between differentially expressed chemokines induced by high glucose environment and immunosuppressive environment formation. Subsequently, we further determined an association between high glucose environment and infiltrated immunosuppressive cells in tumor tissues from the established HCC animal models. Immunofluorescence experiments indicated that HCC tissues in group HCC-DM presented an obvious increase in the number of infiltrated immunosuppressive cells (Tregs and Macrophage M2) and a significant decrease in effector T cells compared with the control tissues in group HCC-Control (Extended data Fig. 2A). Simultaneously, intervention of hypoglycemic agents in group HCC-DM + INS and group HCC-DM + CANA partially abrogated the number changes of the above immune cells (Extended data Fig. 2A). In another similar animal experiment, we further employed flow cytometry to measure the number of infiltrated immunosuppressive cells, and testified that the proportion of infiltrated immunosuppressive cells (Tregs and Macrophage M2) were increased and the proportion of effector T cells was decreased in HCC tissues in group HCC-DM, as compared to the HCC tissues in the HCC-control, HCC-DM + INS, and HCC-DM + CANA groups (Fig. 2B). The above-mentioned results illustrate that hyperglycemia indeed exerts a promoting effect on the strength of immunosuppressive microenvironment. Using TCGA-HCC data, we continued to analyze the association between leading differentially expressed chemokines and infiltrated immunosuppressive cells, and the results demonstrated that the leading downregulated chemokine (RARRES2, the gene name for chemerin) was negatively correlated with infiltrated immunosuppressive cells, specifically Macrophage M2 and Treg cells (Fig. 2C), while the leading upregulated chemokine (CCL22) was positively correlated with Macrophage M2 and Treg cells (Extended Data Fig. 2B), manifesting that the two leading differentially expressed chemokines increase the strength of immunosuppressive microenvironment. Additionally, both the expressions of chemerin and CXCL16 in HCC cells exposed to high glucose conditions and the expression of chemerin in HCC tissues of rats in group HCC-DM were significantly downregulated (Fig. 2D and E), validating a consistency between intracellular expression of differential chemokines and their extracellular secretion. Among the identified different chemokines, chemerin was selected as the target for subsequent function analysis based on the following reasons: (1) Little is known about the effect of high level of glucose on chemerin expression, as well as the associated regulatory mechanism. (2) Chemerin is the leading downregulated chemokine in HG-CM, and both literatures and TCGA analysis support that the expression of chemerin is negatively correlated with the number of infiltrated immunosuppressive cells, and its low expression indicates an unfavorable prognosis (Fig. 2F). (3) There is currently no relevant report on chemerin participating in high glucose-promoted metastasis in HCC. In total, hyperglycemia exerts an obvious inhibitory effect on chemerin expression while concurrently strengthens immunosuppressive microenvironment.

Fig. 2.

Identification of differentially expressed chemokines under high glucose conditions and their relationship with immunosuppressive microenvironment in HCC tissues (A) The comparative analysis of differentially secreted chemokines between HG-CM and NG-CM using a human cytokine/chemokine microarray (upper panel). The semi-quantitative analysis of the differential secretion of chemokines between HG-CM and NG-CM was conducted using fold change analysis (bottom panel). (B) Flow cytometry analysis of CD4⁺CD25⁺FoxP3⁺ Treg cells, CD68⁺CD163⁺ Macrophage M2 cells, and CD8⁺ T cells in HCC tissues from four distinct groups (HCC-DM, HCC-Control, HCC-DM + INS, and HC-DM + CANA). The average prevalence of CD4⁺CD25⁺FoxP3⁺ Treg cells, CD68⁺CD163⁺ Macrophage M2 cells, and CD8⁺ T cells was calculated. Data are shown as means ± SD, n = 3∼4 per group (bottom panel). (C) The analysis of the association between the leading differentially expressed chemokine (RARRES2) and infiltrated immunosuppressive cells (Macrophage M2 and Treg cells) using TCGA-HCC data (D) Representative IHC staining images displayed the expression of chemerin in HCC tissues from four animal groups (left panel). The average density of chemerin in HCC tissues was quantified by analyzing 6 images per group using ImageJ software (right panel). (E) Western blot analysis showed the expression of chemerin and CXCL16 in HCC cells cultured in media with normal glucose, moderate glucose, and high glucose concentrations (left panel). The protein expression levels were quantified (right panel). (F) Survival analysis of RARRES2 expression using TCGA-HCC data. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Abbreviations: HG-CM, high glucose-conditioned media; NG-CM, normal glucose-conditioned media; MG, moderate glucose.

2.3. The level of glycolysis and nuclear accumulation of metabolic enzyme PKM2 in HCC cells under high glucose conditions

Several literatures have demonstrated that glycolysis metabolic enzymes not only possess the classic functions of enzyme metabolism to meet the energy needs of rapid proliferation of tumor cells, but also have non-classical/non-metabolic functions to modulate various complex cellular activities and diseases progression [25,26,39]. So, glucose metabolic enzymes are likely to be involved in the regulation of chemerin expression through non-metabolic functions. We first analyzed the state of glycolysis and the expressions of the metabolic enzymes in HCC cells under high glucose conditions, and found that both glucose consumption (Fig. 3A) and lactate production (Fig. 3C) in HCC cells were significantly elevated as the concentrations of glucose increased. Simultaneously, the expression level of metabolic enzyme PFKP (phosphofructokinase) was distinctly upregulated, but the expression levels of other metabolic enzymes (HK2, PKM2, and LDHA) were almost unaffected in HCC cells exposed to high glucose conditions (Fig. 3B). Hereby, high glucose concentrations enhanced glycolysis level in HCC cells mainly through upregulating PFKP expression, partially explaining why hyperglycemia facilitates the growth of tumor in SD rat HCC models with diabetes. Afterwards, we analyzed the pathological significance of metabolic enzymes in TCGA-HCC data and examined their nuclear translocation at cellular level under high glucose conditions. TCGA analysis showed that high expressions of glycolysis metabolic enzymes (HK2, PFKP, PKM, and LDHA) all indicated worse prognosis of HCC patients (Extended Data Fig. 3A). Considering that the expression levels of HK2, PKM2, and LDHA remained unchanged in HCC cells under high glucose conditions, it is easy to speculate that these three metabolic enzymes may exert non-metabolic functions by nuclear translocation to regulate downstream gene expression such as RARRES2. We comparatively analyzed nuclear accumulation of three metabolic enzymes and discovered that only PKM2 exhibited an obvious increase trend in expression in nuclear protein (Fig. 3D), particularly following a 48-h period of intervention (Extended data Fig. 3B), and a significant nuclear accumulation in HCC cells under high glucose conditions (Fig. 3E and F), indicating that high glucose may enhance nuclear translocation of PKM2 to participate in the regulation of chemerin expression. We applied tumor tissues from the established HCC animal models to further validate PKM2 distribution in the nucleus, also found that hyperglycemia indeed significantly enhances nuclear accumulation of PKM2 in HCC tissues (Fig. 3G), in accordance with the above findings in vitro.

Fig. 3.

The level of glycolysis and nuclear accumulation of metabolic enzyme PKM2 in HCC cells under high glucose conditions (A) The glucose consumption of HCC cells cultured in media with varying glucose concentrations (NG, MG, HG) over 24 h. (B) The protein expression levels of four glycolytic enzymes (HK2, PFKP, PKM2, and LDHA) in HCC cells cultured in media with different glucose concentrations (NG, MG, HG) (left panel). The quantification of protein expression levels (right panel). (C) The extracellular lactate concentration of HCC cells cultured in media containing different glucose concentrations (NG, MG, HG) over 24 h. (D) The expression of three glycolytic enzymes (HK2, PKM2, and LDHA) in nuclear protein, which was extracted from HCC cells cultured in media with varying glucose concentrations (NG, MG, HG) for a duration of 48 h (left panel). The quantification of protein expression levels (right panel). (E–F) The representative immunofluorescence images of nuclear accumulation of PKM2 in HCC cells cultured in media with varying glucose concentrations (NG, MG, HG). The percentage of nuclear PKM2 intensity was determined by analyzing 50 images per group. (G) The representative IHC staining images displayed the PKM2 distribution in the nucleus in HCC tissues from four animal groups (HCC-DM, HCC-Control, HCC-DM + INS, and HC-DM + CANA) (left panel). The average density of nuclear PKM2 in HCC tissues was quantified by analyzing 6 images per group (right panel). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Abbreviations: HK2, hexokinase 2; PFKP, the platelet isoform of phosphofructokinase; PKM2, Pyruvate kinase isozymes M2; LDHA, lactate dehydrogenase A; Nuc, nuclear; WCL, whole cell lysate.

2.4. Nuclear PKM2 binding to TRX1 decreased the level of reduced NF-κB to downregulate chemerin expression

We further explored whether and how nuclear accumulation of PKM2 influenced the expression of chemerin. Taking use of CoIP combining with LC-MS/MS, we first identified the potential candidate proteins that bind to PKM2 in the nucleus of HCC cells under high glucose conditions (Fig. 4A and Supplementary Table 1). A total of 36 candidate proteins, including TRX1, S100A7, and X-ray repair cross-complementing protein 5 (XRCC5), were obtained. Considering that TRX1 binding to NF-κB facilitates the conversion from oxidized NF-κB to reduced NF-κB in the nucleus, and the reduced NF-κB obviously enhances its DNA binding capacity to promote downstream genes expression [40], we assumed TRX1 as the target binding protein to PKM2 in the nucleus (Fig. 4B) and investigated their binding effects on the levels of reduced NF-κB and chemerin. Regardless of taking PKM2 as a bait protein to capture TRX1 or taking TRX1 as a bait protein to capture PKM2, CoIP-western blot analysis all validated that an endogenous binding exists between PKM2 and TRX1 in the nucleus of HCC cells under high glucose conditions (Fig. 4C). Some studies have documented that five cysteine residues of TRX1 (cys32, cys35, cys62, cys69 and cys73) have a very important impact on its biological function. Specifically, cys32 and cys35 are the active cysteine sites responsible for TRX1's redox function, while the other three cysteine sites (cys62, cys69 and cys73) are inactive and primarily serve as a structural role [[41], [42], [43]]. Therefore, the change of TRX1's structural cysteine site or its combination with other proteins seems to hinder the combination of TRX1 with transcription factor NF-κB. Applying Schrödinger software, macromolecular docking state of PKM2 and TRX1 was analyzed and their binding site region was estimated. The results revealed that two proteins exhibited a robust interaction, and the binding sites of TRX1 (asp58, asp60, asp61, val71, and lys72) mainly distributed at structure important cysteine site area (cys62, cys69 and cys73) (Fig. 4D and E). Based on the relevant literatures indications [[41], [42], [43], [44]] and software prediction results, we constructed two mutant plasmids of TRX1 (Flag-TRX1-C32/35S and Flag-TRX1-C62/69/73S), which contained Flag tags and amino acid site mutations at interaction area, to testify the interaction site area between PKM2 and TRX1, as well as changes in reduced NF-κB and chemerin expression in HCC cells under high glucose conditions. CoIP assays revealed that the mutation of C62/69/73S in TRX1 significantly impeded the interaction between TRX1 and PKM2, but the mutation of C32/35S in TRX1 had little effect on their interaction (Fig. 4F), suggesting that Cys62, Cys69 and Cys73 sites of TRX1 were critical sites for altering the interaction between PKM2 and TRX1. Except that, the mutation of C62/69/73S in TRX1 also significantly improved the interaction between NF-κB and TRX1 in the nucleus (Fig. 4F), meaning that the diminished interaction between TRX1 and PKM2 leads to an increased interaction between TRX1 and NF-κB. Given that the expression and distribution of TRX1 in HCC cells remained unaltered under high glucose conditions (Extended data Fig. 4A), we ascertained that the nuclear accumulation of PKM2 obviously diminishes the level of reduced NF-κB through its binding to TRX1.

Fig. 4.

Nuclear accumulation of PKM2 decreased the level of nuclear reduced NF-κB to influence the transcription of chemerin via TRX1 (A) A representative picture of Coomassie blue staining of proteins co-immunoprecipitated with nuclear PKM2 in HCC cells, which cultured in a high-glucose medium for a duration of 48 h. (B) Representative MS plots and the corresponding peptide sequences derived from thioredoxin. (C) The CoIP assay showed the endogenous binding between nuclear PKM2 and TRX1 in HCC cells that were cultured in a high-glucose medium for a duration of 48 h. (D) The molecular docking of human PKM2 protein (blue, PDB entry 3GR4) and human TRX1 protein (green, PDB entry 5DQY) was performed by Schrodinger software. (E) The non-covalent interactions (hydrogen bonds and salt bridges) between PKM2 and TRX1 of the docked model. (F) CoIP assays displayed the interactions between PKM2 and various TRX1 recombinant plasmids (Flag-TRX1-NC, Flag-TRX1-WT, Flag-TRX1-C32/35S, and Flag-TRX1-C62/69/73S), and the interactions between NF-κB and various TRX1 recombinant plasmids in the nucleus of HCC cells, which cultured in a high-glucose medium and transfected with various recombinant plasmids. (G) The assessment of the level of nuclear reduced p50 and p65 subunit of NF-κB in HCC cells using F5M staining combined with immunoprecipitation (upper panel). HCC cells were cultured in a high-glucose medium and transfected with various recombinant plasmids (Flag-TRX1-WT, Flag-TRX1-NC, and Flag-TRX1-C62/69/73S). The quantification of the reductive degree of the p50 and p65 subunit of NF-κB (bottom panel). (H) The protein expression levels of chemerin in HCC cells were analyzed using Western blot. HCC cells were cultured in a high-glucose medium and transfected with different recombinant plasmids (Flag-TRX1-WT, Flag-TRX1-NC and Flag-TRX1-C62/69/73S) (upper panel). The quantification of chemerin expression (bottom panel). (I) The dual luciferase reporter assay demonstrated NF-κB specifically binds to the promoter of gene RARRES2. (J) The schematic diagram showed a predicted NF-κB targeting sequence within the promoter region of the RARRES2 gene, and the predicted sequence was also mutated. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Abbreviations: IP, immunoprecipitation; TRX1, thioredoxin 1; MM/PB (GB) SA, Molecular Mechanics/Poisson Boltzmann (Generalized Born) Surface Area; WT, wild type; NC, negative control; Mut, mutation.

The fluorescent dye F5M, which specifically targets reduced cysteine residues, is often utilized to evaluate the reduced state of NF-κB, and fluorescence intensity of F5M can directly reflect the level of the reduced NF-κB [45]. Applying this method, we continued to measure the level of the reduced NF-κB in the nucleus of HCC cells under high glucose conditions. We used immunoprecipitation to obtain NF-κB from the stained nuclear proteins with F5M dye, and discovered that high glucose obviously decreased the levels of reduced p50 and p65 NF-κB subunit (Extended Data Fig. 4B) and the expression of chemerin in HCC cells (Fig. 2E), but the C62/69/73S mutation of TRX1 significantly increased the levels of reduced p50 and p65 NF-κB subunit (Fig. 4G) and the expression of chemerin (Fig. 4H). All these results supported that the interaction between PKM2 and TRX1 in the nucleus decreased the level of reduced NF-κB, and then downregulated chemerin expression, in agreement with our speculation. Additionally, the JASPAR database was utilized to obtain the binding sequence of NF-κB on the promoter region of gene RARRES2, and dual luciferase reporter assays were employed to determine the regulation of NF-κB on chemerin expression. The results revealed that the wild-type RARRES2-luciferase reporter, not the mutant RARRES2-luciferase reporter, exhibited a significant responsiveness to NF-κB (Fig. 4I and J), indicating that NF-κB specifically interacts with the promoter region of gene RARRES2 and influences its expression.

2.5. Lactylation level of PKM2 at K505 determined its nuclear translocation in HCC cells exposed to high glucose

Nuclear translocation of metabolic enzymes is frequently regulated by their post translational modifications such as phosphorylation, acetylation, and glycosylation [26,46]. Lactylation, a new post-translational modification type, has recently been reported to contribute to HCC cells proliferation and metastasis [47]. Based on a significant increase in intracellular and extracellular lactate content under high glucose conditions (Fig. 5A, Fig. 3C), we supposed that high glucose might change lactylation level of PKM2 and then enhance its nuclear translocation. We captured PKM2 using immunoprecipitation assay to detect its lactylation level in HCC cells under high glucose conditions. As compared to normal glucose level, elevated glucose level remarkably improved lactylation level of PKM2, (Fig. 5B). Subsequently, we analyzed the potential lactylation sites of PKM2 in HCC cells under high glucose conditions by IP-MS/MS, and identified K247, K256 and K505 as the lactylation sites of PKM2 (Fig. 5C). To further evaluate the effects of site mutation mentioned above on the lactylation change of PKM2, we respectively constructed Flag-PKM2 wild type recombinant plasmid and site mutation recombinant plasmids (Flag-PKM2-K505R, Flag-PKM2-K247/256R). The results showed that K505R mutation significantly diminished the lactylation level of PKM2, but K247/256R mutation had little effect on it (Fig. 5D), strongly supporting that the K505 was the principal site of PKM2 lactylation. Analysis of the tertiary structure of PKM2 displayed that K505 site was distributed at the dimer-dimer interface (Fig. 5E), and a cross-linking experiment revealed that mutation of PKM2 K505 site impeded the conversion of PKM2 tetramer into dimer (Fig. 5G), indicating that K505 lactylation may impede the formation of PKM2 tetramer and promote its nuclear translocation. On the other hand, we also validated the effects of K505 site mutation on the nuclear translocation of PKM2, the interaction between TRX1 and PKM2 in the nucleus, and the expression of chemerin. The lactylation inhibition of K505 site not only effectively attenuated the nuclear translocation of PKM2 (Fig. 5F, Extended data Fig. 5B), but also decreased the binding of nuclear PKM2 to TRX1 (Extended data Fig. 5A), and increased the level of the reduced NF-κB (Fig. 5H) and the expression of chemerin (Fig. 5I) in HCC cells exposed to high glucose. Accordingly, lactylation level of PKM2 at K505 site significantly controlled its nuclear translocation in HCC cells under high glucose conditions.

Fig. 5.

Lactylation level of PKM2 at K505 determined its nuclear translocation in HCC cells. (A) The intracellular lactate concentration of HCC cells cultured in media containing different glucose concentrations (NG, MG, HG) over 24 h. (B) The lactylation levels of PKM2 in HCC cells, which cultured in media with different glucose concentrations (NG, MG) were measured by immunoprecipitation and Western blot assays (left panel). The quantification of PKM2 lactylation was determined by calculating the fold change of lactyl-PKM2/total-PKM2 (right panel). The immunoprecipitation of cellular proteins was performed using the PKM2 antibody, followed by their detection using the anti-lactyllysine antibody. (C) IP-MS/MS analysis was conducted to detect the potential lactylation sites of PKM2 in HCC cells cultured in a high-glucose medium. The secondary mass spectrogram illustrates the presence of three potential lactylation sites of PKM2. (D) The lactylation level of Flag-PKM2 was assessed in HCC cell proteins through immunoprecipitation and Western blot assay. HCC cells were transfected with different recombinant plasmids (Flag-PKM2-NC, Flag-PKM2-WT, Flag-PKM2-K505R, and Flag-PKM2-K247/256R) and cultured in a high-glucose medium. (E) The crystal structure of the human tetrameric PKM2 protein is depicted in a ribbon diagram (PDB entry 3GR4). Each distinct color corresponds to a single PKM2 monomer, while the region encompassing the PKM2 K505 site, situated between the red and green monomers, has been magnified. (F) The expression of nuclear Flag-PKM2 and whole cell Flag-PKM2 in HCC cells. HCC cells were cultivated in a high-glucose medium and subsequently transfected with distinct recombinant plasmids (Flag-PKM2-WT, Flag-PKM2-NC and Flag-PKM2-K505R plasmids). (G) The analysis of the oligomer states of Flag-PKM2 was performed through a cross-linking experiment. HCC cells were transfected with plasmids containing either Flag-PKM2-WT or Flag-PKM2-K505R, and these cells were cultured in a medium with high glucose concentration. (H) The assessment of the level of nuclear reduced p50 and p65 subunit of NF-κB in HCC cells using F5M staining combined with immunoprecipitation (left panel). HCC cells were cultured in a high-glucose medium and transfected with different recombinant plasmids. The quantification of the reductive degree of p50 and p65 subunit of NF-κB (right panel). (I) The expression levels of chemerin in HCC cells, cultured in a high-glucose medium and transfected with different recombinant plasmids (left panel). The quantification of the protein expression (right panel). **p < 0.01, ***p < 0.001, ****p < 0.0001.

2.6. In vivo validation of the role of PKM2 lactylation and chemerin in high glucose-promoted HCC metastasis

In vitro cell experiments suggested that PKM2 lactylation and chemerin expression participated in high glucose-promoted HCC metastasis. We proceeded to confirm the roles of PKM2 lactylation and chemerin in high glucose-regulated HCC metastasis using SD rat HCC models with diabetes. We respectively applied PKM2-OE-transfected McA-RH7777 cells, PKM2^K505R-OE-transfected McA-RH7777 cells, NC-transfected McA-RH7777 cells, and chemerin-OE-transfected McA-RH7777 cells to form four groups of orthotopic transplantation HCC models with diabetes (Fig. 6A). Among four animal groups, there were no significant differences in the levels of random blood glucose and body weight (Extended data Fig. 6A). As shown in Fig. 6B-C, the size of HCC tumors in group PKM2-OE was significantly larger than that of the group NC, while there was no difference in tumor volume between the group chemerin-OE and the group NC, indicating that increased PKM2 expression promoted tumor growth and the expression of chemerin exhibited no discernible impact on tumor growth. Notably, the occurrence of lung metastasis in group PKM2-OE was much higher than that in group PKM2^K505R-OE and group NC, while the occurrence of lung metastasis in group chemerin-OE was much lower than that in group NC (Fig. 6D). These results further validated that both PKM2 nuclear translocation and chemerin expression participated in high glucose-induced lung metastasis in HCC. Additionally, we employed flow cytometry to analyze the proportion of immunosuppressive cells in HCC tissues. and found that the proportion of CD4⁺CD25⁺FoxP3⁺ Treg and CD68⁺CD163⁺ Macrophage M2 in the group PKM2-OE were significantly higher than that of group PKM2^K505R-OE and the group NC, but the proportion of CD8⁺ T cells in the group PKM2-OE was lower than that of group PKM2^K505R-OE and the group NC, indicating that PKM2 lactylation-facilitated nuclear translocation obviously enhances the strength of immunosuppressive microenvironment. Meanwhile, the proportion of CD4⁺CD25⁺FoxP3⁺ Treg and CD68⁺CD163⁺ Macrophage M2 in the group chemerin-OE were obviously lower than that of group NC, but the proportion of CD8⁺ T cells was higher than that of group NC (Fig. 6E). These data reversely confirm that low expression of chemerin strengthens immunosuppressive environment. Therefore, PKM2 lactylation and chemerin expression indeed participate in high glucose-promoted HCC metastasis in vivo, in agreement with the findings in vitro. PKM2 lactylation and chemerin expression are expected to become new potential targets for the treatment of HCC patients with diabetes (see Fig. 7).

Fig. 6.

In vivo validation of the role of PKM2 lactylation and chemerin in high glucose-promoted HCC metastasis. (A) A diagram illustrating the animal grouping and experiment protocol. (B) A photograph of orthotopic HCC tumors in four animal groups. (C) The average tumor weight and volume of HCC tumors across four distinct animal groups. (n = 6 per group). (D) The occurrence of lung metastasis of orthotopic HCC tumors across four distinct animal groups. (E) Flow cytometry analysis of CD4⁺CD25⁺FoxP3⁺ Treg cells, CD68⁺CD163⁺ Macrophage M2 cells, and CD8⁺ T cells in HCC tissues from four distinct groups. The average prevalence of CD4⁺CD25⁺FoxP3⁺ Treg cells, CD68⁺CD163⁺ Macrophage M2 cells, and CD8⁺ T cells was calculated. Data are shown as means ± SD, n = 4∼6 per group (right panel). *p < 0.05, **p < 0.01, ****p < 0.0001. Abbreviations: PKM2-OE, PKM2 overexpression; NC, negative control; chemerin-OE, chemerin overexpression.

Fig. 7.

A diagram illustrating the mechanism through which hyperglycemia contributions to HCC metastasis. Under high glucose conditions, HCC cells exhibited heightened glycolysis activity, lactate production, and lactylation of PKM2, resulting in increased nuclear translocation of PKM2. The Nuclear accumulation of PKM2 interacted with nuclear TRX1, thereby impacting TRX1's ability of reducing NF-κB and subsequently diminishing the transcription of chemerin. This process ultimately reinforced an immunosuppressive environment, thereby facilitating HCC metastasis.

3. Discussion

Although some population cohort studies have suggested a close association between diabetes and HCC [1,2,48], whether and how hyperglycemia contributes to HCC progression remain largely undefined. Cancer cells in high glucose conditions often exhibit an increasing trend in glucose uptake and glycolytic activity [49,50], and increased aerobic glycolysis intensifies the strength of immune suppression within tumors [[51], [52], [53], [54]]. These discoveries imply that there exists a potential correlation among high level of glucose, immunosuppressive microenvironment, and tumor metastasis. In the study, we developed a novel HCC model in SD rats with diabetes to clarify the association between hyperglycemia and HCC metastasis as well as infiltrated immunosuppressive cells. Our results revealed that both the incidence of lung metastasis and the proportion of filtrated immunosuppressive cells in group HCC-DM were much higher than those in group HCC-Control. Simultaneously, in group HCC-DM + INS and group HCC-DM + CANA, hypoglycemic drug intervention obviously attenuated the promoting effects of hyperglycemia on lung metastasis and accumulation of infiltrated immunosuppressive cells. These evidences in vivo suggest that hyperglycemia indeed remarkably promotes HCC metastasis and is positively correlated with infiltrated immunosuppressive cells, also indicate that enhanced immunosuppressive microenvironment may participate in high glucose-promoted HCC metastasis. Considering that chemokines are the important contributing factors in determining the strength of immunosuppressive microenvironment, we screened differentially expressed chemokines between HG-CM and NG-CM using a human cytokine/chemokine microarray, and found that most of the identified differential chemokines regulated by high glucose levels were closely associated with immunosuppressive microenvironment [22,23,35]. In combination with the results in animal experiment, it is easy to speculate that the identified differential chemokines regulated by high glucose levels exert a profound impact on strengthening immunosuppressive microenvironment and promoting HCC metastasis. A leading downregulated chemokine chemerin was selected as the target from different chemokines for subsequent function exploring. Both TCGA analysis and literatures review demonstrated that the expression of chemerin was negatively correlated with the number of infiltrated immunosuppressive cells, and its low expression indicates an unfavorable prognosis. Besides, we applied chemerin-OE-transfected McA-RH7777 cells to establish orthotopic transplantation HCC models with diabetes for detecting the proportion of immunosuppressive cells in HCC tissues, and found that the proportion of Treg and Macrophage M2 cells in the group HCC-DM & chemerin-OE were significantly lower than that of group HCC-DM & NC while the proportion of CD8⁺ T cells was higher than that of group HCC-DM & NC. These findings further confirmed that low expression of chemerin obviously strengthened immunosuppressive environment, and thereby promoted metastasis. However, little is known about the effect of high level of glucose on chemerin expression and its relevant regulatory mechanism in HCC.

Given that HCC cells live in high glucose environment, we focused on observing distribution and expression changes in metabolic enzymes in HCC cells under high glucose conditions. Our results showed that except for the significant upregulation of PFKP, the expression of the other three metabolic enzymes (HK2, PKM2, LDHA) remained almost unchanged in HCC cells when subjected to high glucose conditions, suggesting that enhanced glycolysis level in HCC cells under high glucose conditions is mainly attributed to PFKP upregulation. Apart from the main contribution of PFKP upregulation in promoting Warburg effect and preserving stemness in HCC [[55], [56], [57]], there have been almost no reports on its roles in intensifying immunosuppressive microenvironment. Thus, the focus of our research transitioned to examine non-metabolic function of other metabolic enzymes. We analyzed the distribution of glycolytic enzymes in HCC cells under high glucose conditions, and discovered that only PKM2 had a significant increase in expression in nuclear protein and an obvious nuclear accumulation in the nucleus, confirming that high level of glucose effectively increases nuclear translocation of PKM2. Subsequently, we elucidate whether nuclear translocation of PKM2 participated in the regulation of chemerin expression in HCC cells. Existing researches have suggested that nuclear PKM2 regulates gene expression mainly through two ways including acting as a transcriptional coactivator to augment the expressions of target genes [58,59] and changing post-translational modifications of histones or other proteins to affect the expressions of downstream genes [60,61]. Unlike these two regulatory ways, our study unveiled a novel regulatory mechanism by which nuclear PKM2 modulates chemerin expression through redox protein TRX1 in HCC cells. Based on the results of CoIP-LC/MS/MS and CoIP-WB, as well as interaction site mutation analysis, we determined that PKM2 has a strong ability to interact with TRX1 in the nucleus of HCC cells in high glucose conditions, and Cys62, Cys69 and Cys73 sites of TRX1 were three critical sites for governing its interaction with PKM2. Subsequently, our results demonstrated that the mutation of C62/69/73S significantly increased the interaction between NF-κB and TRX1, the level of the reduced NF-κB, and the expression of chemerin, meaning that the diminished interaction between TRX1 and PKM2 can obviously increase the interaction between TRX1 and NF-κB, as well as enhancing the expression of chemerin. Additionally, luciferase assay analysis also supported the specific binding of NF-κB to the promoter region of chemerin, thereby influencing its expression. Accordingly, the interaction between PKM2 and TRX1 in the nucleus decreased the level of reduced NF-κB, subsequently downregulated chemerin expression.

Metabolic enzyme PKM2 generally exists in three forms in the cells including monomer, dimer, and tetramer [62]. Tetrameric PKM2 predominantly locates at the cytoplasm to exert glycolytic function, while dimeric PKM2 often translocate into the nucleus to modulate gene transcription [60,63]. The PKM2 monomer is comprised of four structural domains, namely A, B, C, and N domains. The stability of PKM2 tetramer is modulated by the FBP binding site (Effector site) and the allosteric activator's action site (activators binding site) [64,65]. The FBP binding site is situated within C-domain (amino acid residues 390–531), and FBP binding to this site can transition PKM2 from an inactive dimeric state to an active tetrameric state [64,65]. Notably, modifications to specific amino acids within PKM2 C-domain, such as acetylation at the K433 site, have been reported to diminish FBP binding and lead to destabilization of the tetrameric form and increase of PKM2 nuclear translocation [66]. Conversely, activators binding site is situated within A-domain (amino acid residues 44–116 and 219–389), where their interaction stabilizes the active tetrameric state of PKM2. Thus, what post-translational modifications specifically induce the shift from tetrameric to dimeric PKM2 may be key mechanisms for its nuclear translocation [46,60]. Considering a significant increase in intracellular and extracellular lactate content in HCC cells under high glucose conditions, we guessed that high glucose environment might change lactylation level of PKM2 and thereby increased nuclear translocation of PKM2 in HCC cells. Our results revealed that high levels of glucose obviously increased the lactylation level of PKM2, and its lactylation site mutation (K505) effectively attenuated the nuclear translocation of PKM2, concurrently improved the levels of the reduced NF-κB and chemerin, suggesting a correlation between the level of lactylation and PKM2 nuclear translocation, as well as chemerin expression. Intriguingly, in a previous study from other team [67], the lactylation of PKM2 at the K62 site was found to decrease its nuclear translocation in LPS-induced BMDM cells, which was different from our findings that the lactylation of PKM2 at K505 site promoted its nuclear translocation in HCC cells under high glucose conditions. Different intervention conditions and different cell types are the possible explaining for the above differences. Besides, different lactylation sites of PKM2 may result in different impacts on PKM2 activity and its nuclear translocation. Lactylation of PKM2 at the K62 site increases the stability of its tetramer form, which may be attributed to the fact that the K62 site is situated in the A domain and adjacent to the allosteric activator binding site, but distant from the FBP binding site [67]. However, the lactylation of PKM2 at the K505 site in this study was involved in tetramer instability, potentially due to its location close to the FBP binding site within the C domain. The lactylation of PKM2 at the K505 site may hinder FBP binding, consequently promoting the conversion of tetramers into dimers. Nevertheless, further investigation is still required in the future to clarify the effects of PKM2 lactylation at K505 on the binding of FBP to PKM2. Additionally, it should be mentioned that although lactylation level of PKM2 can effectively regulate its nuclear translocation, the roles of other post-translational modifications are still unable to be excluded, and whether a synergistic effect of multiple types of post-translational modifications on nuclear translocation of PKM2 also deserves further research.

In summary, our study uncovered a novel regulatory mechanism by which nuclear translocation of metabolic enzyme PKM2 mediated high glucose-promoted HCC metastasis. Specifically, high glucose-increased PKM2 nuclear translocation downregulates chemerin expression through the redox protein TRX1, and then strengthens immunosuppressive environment to promote HCC metastasis. To the best of our knowledge, this is first report to elucidate the great contribution of high glucose environment to HCC metastasis from a new perspective of enhancing the immunosuppressive microenvironment, which opens a novel avenue for cross research and intervention between diabetes and HCC.

4. Materials and methods

4.1. Cells and cell culture

HepG2 cells and Hep3B cells, two human hepatoma cells, were obtained from the Cell Bank of Shanghai Institute of Biochemistry and Cell Biology (Shanghai, China). The cells were cultured in Minimum Essential Medium (MEM, Genom, GNM41500-2) supplemented with 12.5% fetal bovine serum (FBS, Biosun, 010423-UY) and 1% penicillin/streptomycin (Ncmbio, C100C5), at 37 °C in a 5% CO2 atmosphere. Referring to the relevant reports on diabetes [68,69], 5.5 mM (NG), 11 mM (MG), and 25 mM (HG) of glucose concentrations were employed to simulate normal, moderate, and high glucose conditions. Buffalo rat HCC cells, McA-RH7777, purchased from the American Type Culture Collection (Manassas), were grown in Dulbecco's Modified Eagle's Medium (DMEM, Genom, GNM12800-2) supplemented with 10% FBS and 1% penicillin/streptomycin.

4.2. Establishment of a new SD rat HCC model with diabetes

Referring to and modifying the methods reported previously [70], we developed a new SD rat HCC model with diabetes using a joint modeling approach. Eight-week-old male SD rats (Beijing Weitong Lihua Experimental Animal Technology Co., Ltd.) were used for animal model development. According to the relevant report [71], the rats with diabetes were firstly induced by a single intraperitoneal injection of streptozotocin (60 mg/kg, STZ, Sigma-Aldrich, V900890). Three days later, random blood glucose levels of rats were measured. Diabetic rats are defined as random blood glucose level exceeding 16.7 mmol/L, moreover their blood glucose levels (≥16.7 mmol/L) can be maintained until the end of animal experiment. Subsequently, SD rat HCC models with diabetes were further established by injecting McA-RH7777 cells (5 × 10⁶) suspended in Matrigel (Corning, 356230) into hepatic subcapsular of diabetic rats. From two days before transplantation to three days after transplantation, diabetic rats received dexamethasone intramuscular injection (2.5 mg/day). Simultaneously, they also received penicillin (50,000 U/day) from the day of surgery until the second day post-surgery. On the 12th day after orthotopic transplantation, SD rat HCC model with diabetes were successfully established and their blood samples, tumors, and lung tissues were collected for subsequent analysis. During the development of the HCC animal model with diabetes, some surgical reasons such as wound rupture, and infection resulted in a limited number of rat death within our estimate, but there were no deaths caused by the model itself.

Animal models in Fig. 1B were divided into four groups including group HCC-DM, group HCC-Control, group HCC-DM + INS and group HCC-DM + CANA. In the two intervention groups, the injection dosage of insulin (Novo Nordisk) and oral administration dosage of canagliflozin (Selleck, S2760) were 10∼20 U/kg and 30 mg/kg, respectively. Animal models in Fig. 6A were divided into four groups including group HCC-DM & PKM2-OE, group HCC-DM & PKM2^K505R-OE, group HCC-DM & NC and group HCC-DM & chemerin-OE. Group HCC-DM & NC served as a shared control group for both group HCC-DM & PKM2-OE and group HCC-DM & chemerin-OE, owing to the utilization of an identical vector (GV492 vector). All experimental procedures and animal care strictly followed the guidelines outlined in the "Care and Use of Laboratory Animals" publication by the US National Academy of Science (Washington, WA, USA), and the experimental design for the animal model was approval by the Animal Care Ethical Committee of Fudan University (Shanghai, China), the ethics approval number: 202306005Z.

4.3. Analysis of differentially expressed chemokines between HG-CM and NG-CM using human chemokine array

The conditioned media from HCC cells cultured in high glucose and normal glucose conditions (HG-CM and NG-CM) for 48 h were collected respectively, and their differentially expressed chemokines were analyzed using a Proteome Profiler™ Human Chemokine Array Kit (R&D Systems, Cat. No. ARY017) in accordance with the instructions and the published studies [[72], [73], [74]]. Briefly, the conditioned media was added with reconstituted detection antibody cocktail and incubated at room temperature for 1 h. Subsequently, the sample/antibody mixtures were administered to each array and incubated at 4 °C overnight. After washing, the array was further incubated with diluted Streptavidin-HRP for 30 min, and then reacted with Chemi Reagent Mix. Signals of the array were captured using the Bio-Rad Chemi Doc XRS imaging system (Hercules), and their signal intensity was quantified using ImageJ software. The relative expression of chemokine was normalized to the average intensity of the positive control spots on the same array. The differentially expressed chemokines between HG-CM and NG-CM were defined using a fold change cut-off of >1.2 or <0.8.

4.4. Analysis of the correlation between chemokines and infiltrated immunosuppressive cells using TCGA-HCC data

The RNA-sequencing expression profiles and clinical data for HCC were obtained from the TCGA dataset (https://portal.gdc.com). The association between the expression of chemokine genes and immune scores was assessed using spearman's correlation analysis. The ggstatsplot package in R software was utilized to visualize the relationships.

4.5. Recombinant plasmid construction and transient transfection

Recombinant expression plasmids (using GV657 vector) for human TXN, human TXN (C32/35S), human TXN (C62/69/73S), human PKM, human PKM (K505R), and human PKM (K247/256R) were designed and constructed by Shanghai GeneChem, Co.Ltd., China. For transient transfection, HCC cells were cultured in 6-well plates and reached a confluency of 70–90%. A mixture (2.5 μg recombinant plasmid, 5 μL P3000 reagent and 5 μL Lipofectamine 3000 reagent, Invitrogen, L3000015) was prepared, and the resulting mixture was added into the culture medium for 16 h transfection. Afterwards, the medium was replaced with fresh MEM medium, and the cells were further incubated for approximately 48–72 h. The efficiency of recombinant plasmid transfection was evaluated using fluorescence microscopy, PCR, or Western blot.

4.6. Lentivirus infection

The tool vector plasmids (utilizing GV492) containing target genes (rat PKM, rat RARRES2) or target sequences (rat PKM^K505R) were designed and constructed by Shanghai GeneChem, Co.Ltd., China. To produce lentivirus, HEK293T cells were seeded in 10-cm dishes and co-transfected with the tool vector plasmids and packaging plasmids Helper1.0 and Helper2.0 using Lipofectamine 3000 (Invitrogen, L3000015). The lentivirus-containing supernatant was collected between 48 and 72 h after transfection and subjected to centrifugation for the removal of cell debris. For lentivirus infection, McA-RH7777 cells were seeded in 6-well plates at a density of 5x10^4 cells per well. The lentivirus (MOI = 20) was added to the medium along with a lentivirus infection enhancement solution (4% HitransG P, GeneChem, REVG005) and incubated with cells for 16 h. Then, the infection solution was substituted with regular complete medium and the cells were further incubated for 48–72 h. These cells were subsequently selected using puromycin (2.5 μg/mL, GeneChem, REVG1001). The efficiency of infection was assessed through qRT-PCR or Western blot.

4.7. Glucose consumption assay and lactate production assay

Glucose uptake and lactate production were assessed using the Glucose Assay Kit (Njjcbio, A154-1-1) and Lactic Acid Assay Kit (Njjcbio, A019-2-1) according to the instructions. Briefly, HCC cells were cultured for 24 h in media containing various concentrations of glucose (5.5 mM, 11 mM, and 25 mM). Subsequently, the cells and their culture media were collected separately. The cells were subsequently lysed in lysis buffer, and protein concentration was determined using the BCA Protein Assay Kit (Beyotime, P0010). Glucose consumption was calculated by comparing the initial and final concentrations of glucose in the medium and expressed as μmol/gprot. For lactate production assay, extracellular lactate concentration was calculated based on the final concentration in the medium, and intracellular lactate concentration based on the final concentration in the cells. The lactate concentration was represented as a relative value compared to the NG group.

4.8. Molecular docking analysis

The crystal structures of two proteins, PKM2 (PDB entry 3GR4) and TRX1 (PDB entry 5DQY), were retrieved from the Protein Data Bank (PDB) database (https://www.rcsb.org/). The chosen structures were then subjected to a series of treatments, and the OPLS_4 force fields were utilized to minimize the protein energy. For molecular docking, the Schrödinger software's protein docking module (Piper) was employed and numerous conformations were obtained. Piper clustered the initial 1000 rotating conformations by evaluating the RMSD between each atom, subsequently identifying representative conformations within each class based on the highest number of neighbors. The generated conformations were then sorted by the quantity of clusters present in each class. The conformation with the greatest number of clusters was ranked highest, and subsequently served as the initial conformation for further analysis.

4.9. Assessment of reduced NF-κB by immunoprecipitation and F5M fluorescence

The redox of the NF-κB subunit (p50 and p65) was assessed as the method described previously [45]. HCC cells were subjected to trypsinization and the nuclear proteins were extracted using the general method, except that the used buffers contained 0.3 mM F5M. Immunoprecipitation was employed to obtain NF-κB protein from the nuclear extracts. Specifically, the extracts were incubated overnight at 4 °C with an antibody targeting NF-κB. The next day, the complex was captured using Protein A/G Mix Magnetic Beads and then denatured. Subsequently, the denatured protein was subjected to SDS-PAGE. The fluorescence emitted by the subunit of NF-κB bands on the freshly prepared gel was observed and captured using the Tanon-5200 Multi chemiluminescent imaging system (Tanon) with an excitation wavelength of 494 nm. The intensity of the fluorescence bands served as an indicator for the extent of reduced NF-κB subunit, and was semi-quantitatively analyzed using ImageJ software. Western blot was employed to measure the overall quantity of NF-κB subunit.

4.10. PKM2 cross-linking experiment

The transfected HCC cells with recombinant plasmid (Flag-PKM2-WT and Flag-PKM2-K505R) were cultured in high glucose environment. The cells were then collected and washed with PBS, subsequently incubated for 30 min in 0.5 mM disuccinimidyl suberate (DSS) at 37 °C. After the incubation, the cells were further lysed according to standard protein extraction protocols. Next, the protein samples were boiled and subjected to Western blot analysis. The resulting membranes were reacted with a primary antibody (Flag, 1:5000, Proteintech, 66008-4-Ig) and a secondary HRP-conjugated antibody, ultimately the target protein bands were visualized using an electrochemiluminescence kit (Tanon, 180–5001). The densities of protein bands were quantified using ImageJ software. The molecular weight of the tetrameric PKM2 was 240 kDa (kDa) and the molecular weight of the dimeric PKM2 was 120 kDa.

4.11. Statistical analysis

Statistical analysis was conducted using R software (version 4.2.1) and GraphPad Prism (version 9.0.0). The data were presented as mean ± standard deviation (SD). Multiple group comparisons were statistically analyzed using one-way analysis of variance (ANOVA), and two groups comparisons using Student's t-test. The correlation between variables was analyzed using Pearson's or Spearman's rank correlation coefficient. Kaplan-Meier analysis was performed to assess the relationship between gene expression and overall survival (OS) in HCC patients,. A significance level of P < 0.05 was considered statistically significant.

4.12. Other methods

For additional methods about hematoxylin and eosin (HE) staining, immunohistochemistry (IHC) and Immunofluorescence (IF) staining, Western blot, co-immunoprecipitation (CoIP) assays, CoIP-LC-MS/MS analysis, dual-luciferase reporter assay, flow cytometry analysis, quantitative reverse transcription polymerase chain reaction (qRT-PCR), transwell migration assays, and wound healing assay, please see the supplementary materials and methods.

CRediT authorship contribution statement

Jiali Qian: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Chuxin Huang: Project administration, Methodology, Investigation. Mimi Wang: Project administration, Data curation. Ying Liu: Resources. Yingying Zhao: Project administration, Methodology. Miao Li: Methodology, Investigation. Xi Zhang: Project administration. Xiangyu Gao: Methodology. Yawen Zhang: Methodology. Yi Wang: Methodology. Jinya Huang: Methodology. Jiajun Li: Project administration. Qiwen Zhou: Project administration. Rui Liu: Methodology. Xuanchun Wang: Methodology. Jiefeng Cui: Writing – review & editing, Methodology, Funding acquisition, Data curation, Conceptualization. Yehong Yang: Project administration, Methodology, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no competing interests.

Appendix A. Supplementary data

The following are the Supplementary data to this article.

Supplementary Table1

A total of 36 candidate proteins identified through CoIP in conjunction with LC-MS/MS.

Supplementary Table2

The primer sequences for qRT-PCR

Extended Data Fig.1 Validation of SD rat HCC models with diabetes and in vitro studies showed high glucose promoted the migration and invasion of HCC cells. (A) Hematoxylin and eosin (H&E) staining of HCC tissues from four animal models (HCC-DM, HCC-Control, HCC-DM + INS, and HCC-DM + CANA). (B) Immunohistochemical staining for Ki67 of HCC tissues from four animal groups (left panel). The average density of Ki67 in HCC tissues was quantified by analyzing 6 images per group using ImageJ software (right panel). (C) qRT-PCR analysis of genes correlated with tumor invasion and metastasis (CD44, OPN, MMP2, and MMP9) in HCC cells cultured in media with varying glucose concentrations (NG, MG, HG) over 24 h. (D-E) The typical pictures of scratch-wound healing assay of HCC cells cultured in media with varying glucose concentrations (NG, MG, HG) over 48 h (left panel). Quantification of the scratch-wound healing assay by calculating 6 images per group (right panel). (F) The typical pictures of transwell assay of HCC cells intervened with varying glucose conditions (normal, moderate and high) over 36 h (upper panel). Quantification of the transwell assay by calculating 5 images per group (bottom panel). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Extended Data Fig.2 (A) Representative images of immunofluorescence staining for FoxP3⁺ Tregs, CD68⁺CD163⁺ Macrophage M2 cells, and CD8⁺ T cells in HCC tissues from four animal groups (HCC-DM, HCC-Control, HCC-DM + INS, and HC-DM + CANA) and the quantification of infiltrated immunosuppressive cells (Tregs and Macrophage M2) and effector T cell (CD8⁺ T cells) in HCC tissues was conducted by analyzing 6 images per group. (B) The analysis of the association between the leading differentially expressed chemokine (CCL22) and infiltrated immunosuppressive cells (Macrophage M2 and Treg cells) using TCGA-HCC data. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Extended Data Fig.3 (A) Survival analysis of HK2, PFKP, PKM, and LDHA using TCGA-HCC data. (B) The nuclear expression of PKM2 in HCC cells under high glucose conditions at different time points (upper panel). The quantification of protein expression levels (bottom panel). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Extended Data Fig.4 (A) The whole cell and nuclear protein expression levels of TRX1 in HCC cells cultured in media with different glucose concentrations (NG, MG, HG) (left panel). The quantification of protein expression levels (right panel). (B) The assessment of the level of nuclear reduced p50 and p65 subunit of NF-κB in HCC cells using F5M staining combined with immunoprecipitation (left panel). HCC cells were cultured in media with normal or high glucose concentrations. The quantification of the reductive degree of p50 and p65 subunit of NF-κB (right panel). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Extended Data Fig.5 (A) CoIP assays displayed the interactions between TRX1 and various PKM2 recombinant plasmids (Flag-PKM2-NC, Flag-PKM2-WT, and Flag-PKM2-K505R), and the interactions between NF-κB and TRX1 in the nucleus of HCC cells, which cultured in a high-glucose medium and transfected with various recombinant plasmids. (B) The quantification of nuclear Flag-PKM2 and whole cell Flag-PKM2 expression in HCC cells was performed for Fig. 5F.

Extended Data Fig.6 (A) The average random blood glucose level and body weight of four distinct animal groups (HCC-DM & PKM2-OE, HCC-DM & PKM2^K505R-OE, HCC-DM & NC, and HCC-DM & chemerin-OE) at various time points throughout the entire experiment. The data presented represent mean ± standard deviation (n = 6 per group).

Supplementary Fig. 1.

Gating strategies for flow cytometry analysis in Fig.2B.

Supplementary Fig. 2.

Gating strategies for flow cytometry analysis in Fig.6E.

Data availability

Data will be made available on request.