Activation-induced cytidine deaminase sustains glycolytic flux through MYST1 acetyltransferase-dependent lactylation in hepatocellular carcinoma

AID is upregulated in HCC and promotes tumor progression

The relevance of AID expression and overall survival (OS) of HCC patients were analyzed using the TCGA database (https://kmplot.com/). The results showed a significant decrease of OS along with high AID expression (Fig. 1A). To verify the oncogenic role of AID in HCC, immunohistochemistry (IHC) was used to examine the level of AID protein in HCC specimens. Remarkably, the IHC analyses revealed minimal detection of AID expression in the para-cancerous tissues, while AID expression was significantly increased in these HCC samples (Fig. 1B).

To verify the pro-tumor effect of AID, some tumor-associated molecular characteristics were evaluated in AID knockdown (AIDKO) and AID over-expression (AIDOE) HCC cells (HCC-LM3 and MHCC-97L). The clone formation units of AIDKO and AIDOE HCC cells indicated that loss of AID alleviated while abundant AID promoted HCC cell proliferation (Fig. 1C). The protein level of these tumor-associated factors, such as MMP2, CXCR5, pro-Caspase3, BCL2 and AID, showed a decline in AIDKO cells while elevation in AIDOE cells (Fig. 1D). Meanwhile, the transcript level of some tumor-associated markers such as AICDA, CDH1, CDH2, VIM, CXCR5 confirmed the same trend of AID regulating tumor-associated factors (Fig. 1E).

Additionally, the cellular ROS generation also detected, and the results obviously showed that rich AID promoted ROS accumulation in HCC cells, while loss of AID decreased HCC cells to generate ROS (Fig. 1, F and G). The apoptosis rate of AIDKO and AIDOE HCC cells were detected using flow cytometry to detect Annexin V and 7-AAD double positive cells (Fig. 1, H and I). These results indicated that AID promoted HCC cell proliferation along with increased intercellular ROS generation.

These data collectively showed that AID apparently promote HCC progression.

Loss of AID suppresses glycolytic metabolism in HCC cells

To search the mechanism of AID promoting HCC progression, we performed transcriptome sequencing to WT and AIDKO HCC cells. Through analyzing these data, we noticed the HIF-1 signal pathway (Fig. 2A). GESA analysis indicated that HIF-1 signal decreases at the largest extent along with AID deficiency (Fig. 2B). By further analyzing the involved factors in HIF-1 signal, we identified these factors SLC2A1 (GLUT1), PGK1 and LDHA in HIF-1 signal pathway are closely related with glucose uptake and glycolysis-related metabolism (Fig. 2C). These three factors were key regulators of glucose metabolism, including the glucose transporter GLUT1 and the glycolytic enzymes PGK1 and LDHA (Fig. 2D). Therefore, the relevance of the glucose transporter GLUT1 (SLC2A1) and the glycolysis-related factors PGK1 and LDHA in the glycolytic metabolism and the OS of patients with HCC were analyzed using PanCanSurvPlot database (https://smuonco.shinyapps.io/PanCanSurvPlot/). The data showed that high SLC2A1 and PGK1 expression are apparently relevant to the OS of HCC (Fig. 2, E and F). Additionally, considering the key metabolites of the glycolytic pathway, we detected the lactate (L-lactate) content in AIDKO and AIDOE HCC cells, indicating that AID deficiency downregulates the L-lactate content while AID overexpression upregulates the L-lactate content (Fig. 2, G and H). In addition, the transcript and protein levels of AICDA, SLC2A1 (GLUT1), PGK1 and LDHA showed a confirmed trend; that is, AID promoted SLC2A1 (GLUT1), PGK1 and LDHA expression and L-lactate content in HCC cells (Fig. 2, I–K).

These results indicated that loss of AID suppresses glucose uptake and glycolysis-related metabolic reprogramming in HCC.

AID interacts with MYST1 through its 95 to 198 aa domain to regulate non-histone lactylation

To investigate the mechanism of AID modulating glucose uptake and glycolytic metabolism, we first use molecular dynamics (MD) simulations to evaluate the affinity of AID to epigenetically relevant factors. The results showed that AID has a higher affinity to acetyltransferases than to methyltransferases (Fig. 3A). Acetyltransferases are involved in two PTM including acetylation, lactylation, succinylation and butylation modification, so we tested the utilization of AID-acetyl-CoA, AID-lactyl-CoA, AID-succinyl-CoA and AID-butyryl-CoA by various acetyltransferases. The results showed that acetyltransferase MYST1 (KAT8) has the highest utilization of both, and the affinity score of MYST1 to AID-lactyl-CoA is the highest (Fig. 3B).

First, we used Co-IP experiments to verify the MD prediction results in HCC cells, AID and MYST1 did form a complex, while AID and P300 did not interact with each other (Fig. 3, C and D). The experimental data were consistent with the prediction results, fully proving the accuracy of MD prediction. Second, the co-localization of AID and MYST1 was examined using confocal microscopy. The results revealed a clear overlap between AID (red) and MYST1 (green), appearing as white or light-yellow signals, indicating the formation of an AID–MYST1 complex (Fig. 3E). Third, based on our previous MD analysis showing a stronger predicted binding affinity of MYST1 for L-lactyl-CoA, we focused on investigating the potential role of AID/MYST1 in lactylation regulation. WB analysis using a pan-lactylation antibody showed that AID overexpression and AID deficiency were associated with increased and decreased Pan-Kla signals, respectively, as confirmed by densitometric quantification (Fig. 3, F and G). Notably, the lactylated proteins mainly distributed above 25 KDa rather than at 10 to 25 KDa, which encompasses the molecular-weight range of core histones. This pattern suggested that AID may predominantly regulate non-histone protein lactylation rather than histone lactylation. To directly assess the contribution of histone lactylation, we isolated histone fractions and examined Pan-Kla levels. Histone-associated Pan-Kla levels showed no appreciable changes following AID overexpression or deficiency, suggesting that the AID-dependent changes in overall cellular Pan-Kla levels are predominantly attributable to alterations in non-histone protein lactylation (Fig. 3, H and I). Collectively, these findings support a predominant role for AID in regulating non-histone lactylation rather than histone lactylation.

In addition, we used MD to predict the amino acid residues of AID/MYST1 interaction (Fig. 3J). Based on analyzing these amino acid residues, we established constructs with truncated AID functional domains (Fig. 3K). The Pan-Kla was detected in HCC cells with these truncated AID constructs. The results indicated that AIDΔ95-198aa domain induced slightly higher histone lactylation (15–25 KDa) modification than other domains, while non-histone lactylation (more than 25 KDa) decreased (Fig. 3L left), especially AIDΔ190-198aa mediated a decrease in GLUT1 and PGK1 protein levels and inhibited HCC progression, which indicated by elevated pro-caspase three and BCL2 protein (Fig. 3L right). Collectively, these data indicated that AID’s 95-198aa domain and MYST1 form complex to modulate non-histone lactylation.

MYST1 is required for AID-associated non-histone lactylation

We then performed rescue experiments to identify the essential role of MYST1 in AID-mediated lactylation modification. MG149, a inhibitor to MYST1 was introduced, and AID overexpression was also introduced to rescue. The used concentration of MG149 were selected to be 10 μM using gradient concentration of MG149 treatment to HCC cells, and the protein level of MYST1 and CXCR5 decreased obviously in 10 μM MG149 treated HCC cells (Fig. S1). The Pan-Kla level obviously decreased in MG149 treated HCC, and AIDOE apparently counteracted the inhibition effect, along with altered GLUT1 and PGK1 protein level, and changed MMP2 and AID protein level (Fig. 4A). To determine whether MG149 specifically affects the lactylation of GLUT1 and PGK1, we performed Co-IP using an anti-Pan-Kla antibody, followed by immunoblotting for GLUT1 and PGK1. MG149 treatment resulted in reduced lactylation-associated signals for both GLUT1 and PGK1, providing further evidence that MYST1 activity contributes to the lactylation of these proteins. These findings further support the involvement of MYST1 in AID-associated non-histone lactylation (Fig. 4B).

The clone formation units in AIDOE cells are obviously more than the WT counterparts, and MG149 inhibited the proliferation trend in AIDOE cells (Fig. 4C). The scratch test also showed the same trends in HCC cells (Fig. 4, D and E). The apoptosis rate of these HCC cells were detected using flow cytometry to detect Annexin V and 7-AAD double positive cells (Fig. 4, F and G). The cell cycle green reagent was used to stain these HCC cells, and the cell cycle was also measured by flow cytometry (Fig. 4, H and I). The AID overexpression obviously promoted HCC proliferation, while MG149 apparently turned the trend. These results suggest the critical role of MYST1 in AID mediated non-histone lactylation.

AID enhances non-histone lactylation of key glycolytic enzymes in HCC

To further confirm the role of AID in HCC progression via lactylation modification, we introduced L-lactate and the inhibitor of the glycolytic pathway, 2-DG, to treat the WT, AIDKO and AIDOE HCC cells (Figs. S2 and S3). Except that the increased lactylation mofidication (manifested by elevated Pan-Kla) in AIDOE HCC cells, while decreased lactylation modification in AIDKO HCC cells compared with these WT conuterparts (Fig. 5, A and B), we observed elevated Pan-Kla level in L-lactate treated HCC cells (Fig. 5A lane 2, 5, 8 and Fig. 5B lane 2, 5, 8), and apparently decreased Pan-Kla level in 2-DG treated HCC cells (Fig. 5A lane 3, 6, 9 and Fig. 5B lane 2, 5, 8). Meanwhile, we performed Co-IP using an anti-Pan-Kla antibody, followed by immunoblotting for GLUT1 and PGK1 in AIDKO and AIDOE HCC cells. Alteration of AID expression resulted in corresponding changes in the Pan-Kla-associated signals of both GLUT1 and PGK1, providing further evidence that AID promotes non-histone lactylation of these key glycolytic enzymes in HCC cells (Fig. 5C).

The RT-qPCR also indicated the same trend in L-lactate and/or 2-DG-treated WT, AIDOE and AIDKO HCC cells (Fig. 5, D and E). L-lactate content assay showed the role of L-lactate and 2-DG in Pan-Kla levels, which confirmed the AID-mediated non-histone lactylation modification in HCC.

Additionally, the combined cytotoxicity effect of AID deficiency and L-lactate and/or 2-DG-mediated lactylation modification was evaluated by detecting the apoptosis rate and intratumor ROS generation. The apoptosis rate of HCC cells treated with both AID overexpression and L-lactate treatment induced the smallest amount of cell apoptosis, while AID deficiency and 2-DG treatment caused the highest rate of cell apoptosis compared with other treated HCC cells (Fig. 5, F and G). AID overexpression combined with L-lactate treatment mediated the least ROS generation, while loss of AID combined with 2-DG-induced glycolytic metabolism inhibition caused the greatest decrease in HCC cellular ROS generation (Fig. 5, H and I).

These results indicated that AID-mediated HCC progression occurs via non-histone lactylation modification to glycolysis.

Site-specific lactylation of GLUT1 (K6) and PGK1 (K6/K17) is required for enhanced glycolysis and HCC progression

To explore potential non-histone targets involved in AID/MYST1-associated regulation of glycolytic metabolism, we performed MD simulations to evaluate the predicted interactions between GLUT1, PGK1, LDHA and L-lactyl-CoA. The simulations suggested favorable interactions between L-lactyl-CoA and GLUT1 or PGK1, whereas no apparent interaction was observed with LDHA (Fig. 6A). Based on these predictions, GLUT1 and PGK1 were selected as candidate proteins for further investigation. To determine whether GLUT1 and PGK1 are associated with the AID/MYST1 complex, we performed Co-IP assays. Co-IP using anti-AID and anti-MYST1 antibodies demonstrated interactions of AID/MYST1 with both GLUT1 and PGK1 (Fig. 6, B and C). These findings suggest that GLUT1 and PGK1 may serve as potential non-histone targets associated with AID/MYST1-mediated regulation.

Then, we established the GLUT1 overexpression (GLUT1OE) and PGK1 overexpression (PGK1OE) constructs, and GLUT1OE or PGK1OE WT AIDKO and AIDOE HCC cells. The lactate content increased along with elevated GLUT1 or PGK1 overexpression in AIDOE HCC cells (Fig. 6, D and E). Immunoblots indicated that non-histone lactylation modification (≥25 KDa Pan-Kla) increased while histone lactylation modification (<25 KDa Pan-Kla) decreased in GLUT1OE or PGK1OE HCC cells (Fig. 6, F and G), along with increased GLUT1, AID, CXCR5 and BCL2 levels, indicating HCC progression (Fig. 6, F and G). Both AID and GLUT1 or both AID and PGK1, overexpression increased the transcript level of GLUT1 and PGK1 (Fig. 6, H and I). Both the apoptosis rate and ROS generation showed the same trend in AID, GLUT1, and PGK1-mediated HCC progression (Fig. 6, J–M).

Considering lactylation modification is a conservative PTM, and it has been found that the lactyl group transfer in lactylation modification is mostly achieved by acetyltransferases (22, 23, 24). Therefore, we used bioinformatics databases to search for the acetylation modification sites of AID/MYST1 on GLUT1 and PGK1 and compared and analyzed the differences among different species. The results showed that the modification score of MYST1 on GLUT1 K6 and PGK1 K6/K17 was higher than that of other acetyltransferases, and GLUT1 K6 and PGK1 K6/K17 are more conserved among various species, which may be the site of lactylation modification of AID/MYST1 on GLUT1 and PGK1 (Fig. 6, N and O). To identify these selected precision modification sites from database, we established mutant constructs GLUT1^K6R and PGK1^K6R, PGK1^K17R. The Pan-Kla level slightly decreased in GLUT1^K6R and PGK1^K6R/K17R HCC cells, along with decreased AID, MYST1, CXCR5 and BCL2 level (Fig. 6, P and Q), confirming the precise sites of AID-mediated GLUT1 lactylation modification are K6, and the precise sites of AID-mediated PGK1 lactylation modification are K6 and K17. Furthermore, to clarify the To further investigate the potential lactylation sites predicted by our computational analysis, we generated GLUT1 and PGK1 mutants in which the predicted lactylation-associated residues were substituted. Importantly, immunoblotting confirmed comparable expression of the corresponding WT and mutant proteins, allowing us to directly assess their lactylation-associated signals. Co-IP using an anti-Pan-Kla antibody followed by immunoblotting for GLUT1 or PGK1 in GLUT1^K6R and PGK1^K6R, PGK1^K17R HCC cells revealed that mutation of the predicted residues resulted in altered Pan-Kla-associated signals compared with their respective full-length (FP) proteins (Fig. 6, R and S). These findings provide experimental support for the contribution of the predicted residues to the lactylation of GLUT1 and PGK1.

Collectively, these results showed that AID mediates the shift of histone lactylation to non-histone lactylation of GLUT1 K6 and PGK1 K6 and K17, driving HCC progression.

Targeting the AID-glycolysis axis induces cytotoxicity in HCC cells

To evaluate the cytotoxity of AID mediated laclytation modification of glycolytic metabolism to HCC progression, we introduced AID inhibitor 4-Deoxyuridine and glycolytic metabolism inhibitor 2-DG to treat HCC cells (Figs. S2 and S4). The combined treated using 4-Deoxy and 2-DG induced the most decrease of protein level of CXCR5, BCL2 and AID in HCC-LM3 and MHCC-97L cells, and decrease of transcript level of AID, MYST1, GLUT1, PGK1 and CXCR5 (Fig. 7, A–D). The scratch test also induced the cytotoxity of combined inhibition of AID and glycolytic metabolism (Fig. 7, E and F). The CFU of combined treatment cells showed the smallest number in clone formation experiment (Fig. 7G). The apoptosis rate of HCC cells treated with 4-Deoxyuridine and 2-DG induced the highest rate of cell apoptosis compared with other treated HCC cells (Fig. 7, H and I). These data support the therapeutic potential of co-targeting AID and glycolysis in HCC.

Attenuation of AID activity combined with glycolytic inhibition restrains tumor growth and remodels TME

To evaluate the role of AID and glycolytic metabolism in remodeling suppressive TME of HCC, we elaborated the AID antagonist 4-Deoxyuridine and glycolysis inhibitor 2-DG to treat the H22 tumor-bearing mice (Fig. 8A). First, the tumor volume and weight as well as mice weight were monitored. The tumor volume in the combined treatment group treated by 4-Deoxyuridine and 2-DG showed the smallest (Fig. 8, B and C), along with the same trend of tumor weight (Fig. 8D). The mice weight showed no significant changes, indicating no obvious drug toxicity (Fig. 8E). Second, the tumors were obtained, and the ratio of immune cells in TME were measured (Fig. S5). The tumor CD4⁺ T cells (CD45⁺CD3⁺CD4⁺) increased 15%, CD8⁺ T cells (CD45⁺CD3⁺CD8⁺) showed an increase of about 8%, and B cells (CD45⁺B220⁺) increased by about 30% (Fig. 8F). Third, the expression of AID, MTST1, GLUT1, PGK1, and CXCR5 was detected by immunofluorescence (IF). The fluorescence intensity of these proteins were apparently reduced in the combined treatment group treated by 4-Deoxyuridine and 2-DG (Fig. 8G). These data indicate that simultaneously suppressing AID and glycolysis shows obvious changes in the TME.

Collectively, these data indicated that attenuation of AID activity combined with inhibition of glycolytic pathway slows tumor growth by increasing lymphocytes infiltrating to tumor.

Expanding the function of AID to lactylation in cancer metabolism

Traditionally studied for its role in antibody diversification and DNA mutagenesis, AID is now emerging as a multifaceted regulator in cancer biology (16, 25). Our work positions AID as a critical metabolic effector in HCC, primarily through its involvement in lactylation: a rapidly evolving field of PTM. Lactylation, initially described on histones, has recently been recognized as a widespread mechanism regulating non-histone proteins, linking cellular metabolic status directly to functional protein modulation (26, 27, 28). We demonstrate that AID expression is pivotal for sustaining high glycolytic rates in HCC cells, and its loss profoundly suppresses glycolysis via impairment of HIF-1α signaling (Fig. 1). This establishes AID not merely as a passive component but as an active driver of the Warburg effect, a hallmark of cancer metabolism where cells preferentially utilize glycolysis even under aerobic conditions.

The discovery that AID regulates lactylation adds a new dimension to its functionality. While recent studies have identified other enzymes like AARS1 and TIP60 as capable of catalyzing lactylation (29, 30, 31), our identification of AID, particularly in complex with MYST1, reveals a previously unrecognized mechanism for lactylation in HCC (Figs. 2, 3, 4 and 5). This suggests that lactylation catalysis might be more widespread and involve diverse enzymatic machinery across different cancer types. The metabolic dependency of HCC cells on AID-mediated lactylation underscores the importance of this pathway in maintaining tumorigenic metabolism. By linking AID to lactylation, our study suggests that AID could be a selected checkpoint in reprogramming cancer cell metabolism.

Identifying the non-histone lactylation function of AID in precision regulation of glycolytic enzymes

Our study identified and validated GLUT1 K6 and PGK1 K6/K17 as direct substrates (Fig. 6). Lactylation of GLUT1, the primary glucose transporter, enhances glucose uptake to fuel glycolysis. Concurrently, lactylation of PGK1, a key glycolytic enzyme, would accelerate the overall glycolytic flux. This dual lactylation strategy ensures a synchronized amplification of glucose entry and its downstream utilization, providing a continuous supply of metabolic intermediates for biomass generation and energy production, which are essential for rapid tumor growth and proliferation.

Critically, our data reveal that this axis establishes a metabolic feedforward loop to sustain the glycolytic machinery. The accelerated glycolysis induces massive lactate accumulation, which in turn elevates the levels of the substrate Lac-CoA. This enriched Lac-CoA pool is subsequently utilized by the AID/MYST1 complex to further mediate the non-histone lactylation of GLUT1 and PGK1. This site-specific, non-histone lactylation moves beyond the conventional understanding of lactylation as primarily a histone mark influencing gene expression (32). It demonstrates a direct and rapid mechanism to post-translationally hyperactivate the core metabolic machinery itself (33). Our findings align with the emerging concept that lactylation serves as a real-time metabolic sensor, allowing cancer cells to dynamically fine-tune enzyme activities in response to fluctuating nutrient and energy demands. The fact that AID-mediated lactylation promotes HCC progression highlights the pathological hijacking of this precise regulatory mechanism for tumor therapy.

The AID/MYST1 complex signifies a novel lactylation writer complex with therapeutic potential

We discovered that AID does not function in isolation but forms a stable enzymatic complex with MYST1 (also known as KAT8), a histone acetyltransferase, to execute its lactylation function (Figs. 2 and 3). This partnership is intriguing. While MYST1’s acetyltransferase activity is well-documented, its role as a lactylation transferase in complex with AID is novel. It is plausible that the AID-MYST1 complex exhibits unique substrate specificity or enhanced catalytic efficiency towards glycolytic enzymes compared to either protein alone. The formation of this complex suggests a potential crosstalk or even a switch between different acyl modifications (e.g., between acetylation and lactylation) on specific lysine residues, adding a layer of complexity to the regulatory network controlling protein function in cancer.

Synergistic co-targeting AID/MYST1-glycolysis contributes to TME reprogramming

Building on these mechanistic insights, we further explored the therapeutic implications of targeting the AID/MYST1–glycolysis axis. We demonstrate that simultaneous inhibition of AID and glycolytic metabolism induces synergistic cytotoxicity in HCC models (Figs. 7 and 8). This synergistic effect is particularly notable given that monotherapies targeting tumor metabolism frequently encounter adaptive resistance driven by metabolic plasticity and tumor heterogeneity (34, 35, 36). By co-targeting AID and glycolysis, this combinatorial strategy effectively disrupts the feedforward regulatory circuit sustaining glycolytic dependency, resulting in pronounced metabolic collapse and tumor cell death (Figs. 7 and 8).

Beyond direct tumor cell cytotoxicity, our data further indicate that attenuation of AID activity in combination with glycolytic inhibition is associated with remodeling of the tumor microenvironment (Figs. 7, 8 and 9). Specifically, combined treatment significantly slowed HCC tumor growth and coincided with increased infiltration of tumor-infiltrating lymphocytes (Figs. 8 and 9). These observations suggest that disruption of the AID–lactylation–glycolysis feedback loop may alleviate lactate-driven immunosuppressive conditions within the tumor milieu, thereby creating a more permissive environment for antitumor immune responses (Figs. 7, 8 and 9).

Collectively, these findings support a model in which co-targeting AID-associated metabolic regulation and glycolytic pathways not only compromises tumor metabolic fitness but also indirectly influences the immunological landscape of HCC. While further studies are required to define immune cell-specific mechanisms, our results provide a mechanistic rationale for integrating metabolic intervention with immunotherapeutic strategies in HCC, an avenue that warrants further investigation. Nevertheless, several limitations should be acknowledged. The in vivo findings of the present study were derived primarily from the H22 syngeneic mouse model, which, despite preserving an intact immune system, does not fully recapitulate the natural initiation, progression, and tumor microenvironment of hepatocellular carcinoma. Therefore, future studies employing more physiologically relevant models, including genetically engineered mouse models (GEMMs) and chemically induced HCC models such as DEN-induced hepatocarcinogenesis, will be important to further validate the robustness, biological significance, and translational relevance of our findings.

In conclusion, this study uncovers a previously unrecognized function of AID in orchestrating glycolytic metabolism in hepatocellular carcinoma via a MYST1-dependent non-histone lactylation mechanism. By forming a functional complex with MYST1, AID promotes site-specific lactylation of GLUT1 (K6) and PGK1 (K6), thereby enhancing their enzymatic activities and sustaining glycolytic flux. Crucially, our findings demonstrate that this mechanism establishes a self-reinforcing metabolic feedforward loop. Hyperactivated glycolysis results in the accumulation of Lac-CoA, which subsequently fuels further AID/MYST1-dependent lactylation of glycolytic enzymes. This positive feedback loop locks hepatocellular carcinoma cells into a persistently glycolytic state and drives the upregulation of oncogenic factors such as CXCR5, BCL2, and c-MYC, thereby promoting tumor progression and remodeling the tumor microenvironment. Importantly, our results reveal an AID-associated metabolic vulnerability. Rather than advocating for direct systemic inhibition of AID, this work provides a mechanistic rationale for targeting tumor-specific, self-sustaining metabolic dependencies. While primarily based on preclinical models, our study establishes a conceptual framework linking non-histone lactylation to the regulation of glycolysis. Given the emerging roles of lactylation, this regulatory principle may extend to other malignancies characterized by metabolic reprogramming, thereby offering a foundation for future investigations into metabolic–epigenetic crosstalk in cancer.

Cells and constructs

The lentiCRISPRv2 targeting human AICDA gene were constructed to generate AIDKO plasmids. The sequences of gRNAs for AICDA were listed in Table S1.

Cells lines HCC-LM3, HMCC-97L and HEK293T were cultured in DMEM medium (Life Technologies) supplemented with 10% fetal bovine serum (FBS; Hyclone) and 1% penicillin/streptomycin (Life Technologies). The AIDKO cell lines (AIDKO HCC-LM3 cells and AIDKO HMCC-97L cells) were established by positive selecting with 1 mg/ml puromycin (Sigma). All cells lines were incubated at 37 °C in a humidified 5% CO₂ atmosphere. Transfection of the plasmids and constructs into the cells was conducted using Lipofectamine 2000 reagent (Life Technologies) according to the manufacturer’s protocol. All cell lines were tested negative for mycoplasma contamination.

The modified AID protein was achieved by modifying AID functional domain at transcriptional level. Full-length human AID cDNA was amplified and cloned into the pWPI-GFP vector to generate pWPI-AID-GFP. To generate AID deletion mutants, PCR fragments lacking the indicated amino acid regions (Δ1–55, Δ1–94, Δ95–198 and Δ1–26, Δ27–56, Δ57–94, Δ95–189, Δ190–198) were amplified using deletion-specific primers. The PCR products were assembled into the linearized pWPI-GFP backbone by overlap-extension PCR and subsequently cloned into pWPI-GFP. All plasmids were verified by Sanger sequencing prior to use. Primers were listed in Table S2. The amplified fragments were cloned into the pWPI-GFP vector. The constructed vector sequences were commercially confirmed (Sunny).

Immunoblots

Extract proteins from cells/tissues using RIPA lysis buffer supplemented with protease/phosphatase inhibitors. Quantify protein concentration via BCA assay. Mix samples with loading buffer, boil at 100 °C for 5 min to denature proteins. Load 20 to 50 μg protein alongside prestained molecular weight markers to SDS-PAGE gels. Perform electrophoresis and transfer protein to PVDF membrane in transfer buffer. Incubate membrane with 5% non-fat milk at RT to reduce background. Incubate c with primary antibody at 4 °C overnight. Wash the PVDF membrane for 3 times with TBST (10 min each). Incubate HRP-conjugated antibody for 1 h at RT. Wash PVDF membrane for 3 times with TBST (10 min each). Incubate membrane with ECL substrate for 2 to 5 min. Capture chemiluminescent signals using FUSION FX7 multifunctional imaging device.

The primary antibody pan-Kla (PTM_1401RM, WB, 1:1000), anti-AID (YT_5566, WB, 1:1000), anti-MYST1 (CY_8281, WB, 1:1000), anti-GLUT1 (CY_5360, WB, 1:1000), anti-LDHA (CY_5348, WB, 1:1000), anti-PGK1 (CY5536, WB, 1:1000), anti-MMP2 (AB_3284, WB, 1:1000), anti-CXCR5 (CY_8749, WB, 1:1000), anti-c-MYC (CY_5150, WB, 1:1000), anti-BCL2 (CY_5032, WB, 1:1000), Caspase 3 (CY_5048, WB, 1:1000), anti-β-actin (AB_0011, WB, 1:2000), and horseradish peroxidase (HRP)-conjugated secondary antibodies were purchased from North American ImmunoWay Biotechnology Company and Abways Technology Co., Ltd. The ECL Substrate Kit (P0018FS) was obtained from Beyotime Biotechnology Co., Ltd.

Lactate content assay

Cells were homogenized in ice-cold assay buffer (1 ml per 0.1 g tissue or 5 × 10⁶ cells), centrifuged at 12,000×g, 4 °C for 5 min; collect supernatant. Diluted the lactate standard reagent accoring to the instruction. The cellular L-lactate content was measured after incubating with the detecting reagents in the kit (Lactic Acid (LA) Content Assay Kit, Cat: BC2235; Solarbio). Then measured the absorbance at 450 nm using a microplate reader. Plot standard curve and calculated sample lactate.

RNA extracting and sequencing

Total RNA of cell pellets was extracted with TRIzol (Invitrogen, #15596026) according to the manufacturer’s instructions. cDNA was then synthesized with PrimeScript RT reagent Kit (TaKaRa, #RR037A), according to the manufacturer’s protocol. Quantitative PCR was performed with real-time PCR using Mx3000P (Agilent Technologies). Primers were listed in Table S3. The relative mRNA level of genes were calculated according to the formula 2^−ΔΔCt using β-actin as an internal control.

For RNA-seq, cDNA library construction, and paired-end sequencing with NovaSeq 6000 platform (Illumina) were completed by Shanghai Personal Biotechnology Cp. Ltd (China). Gene expression values were computed from fragments per kilo bases per million fragments (FPKM) values produced by addition of a pseudocount of one and log2 transformation of the results. These FPKM values were used for drawing the heatmap with the pheatmap R package. Paired differential gene expression analyses were performed with DEseq R package by addition of |fold change| > 2 and p-value <0.05. Volcano plots of these differential genes were drawing with ggplots2 R package. KEGG enrichment analyses of differential genes were performed using clusterprofiler version 3.16.1. GSEA analysis was performed to functionally annotate the relevant genes and assess the enriched signaling pathways using GSEA_Linux_4.1.0. Volcano plots were generated with the ‘ggplot2’ package in R (threshold p < 0.05, equivalent change>2). The RNA sequence data were stored in the Sequence Read Archive (PRJNA1209595).

Molecular dynamic simulation

The structure of AID was established according to its amino acid sequence obtained from the UniProt database (https://www.uniprot.org/). The structures of other histone modification associated proteins were all retrieved and downloaded in PDB format from the RCSB PDB database (https://www.rcsb.org/). All proteins were required to remove water molecules and complete residues before interacting with AID. The Zdock molecular docking technology were taken between AID and other histone modification associated proteins, MYST1, L-lactyl-CoA/Acetyl-CoA, and the highest Zdock score of AID and the interaction protein were chosen and compared. The simulation between GLUT1/PGK1 and L-lactyl-CoA/Acetyl-CoA was performed and the interaction were exported. The Zdock score of the 24 histone modification associated proteins were listed in Table S4.

Flow cytometry

Single-cell suspension was prepared; cells washed twice with cold PBS and adjusted concentration to 1 to 5 × 10⁶ cells/ml. Incubated cells with fluorochrome-conjugated antibody for 30 min at 4 °C in the dark. Wash the cells twice by centrifuging (300–400×g, 5 min), discard the supernatant, and resuspend in cold buffer. The cells were analyzed immediately on a CytoFLEX Flow Cytometer (Beckman Coulter).

For cell cycle detection, cells were fixed by anhydrous ethanol overnight, then were stained using Cell Cycle Reagents (Elabscience, #E-CK-A352) to stain the cells. The cells were then washed twice with 1 × PBS at 4 C and analyzed by flow cytometry. For apoptosis rate detection, Cells were collected and washed twice with 1 × PBS at 4 °C, incubated with anti-Annexin V and anti-7-AAD (Elabscience, #E-CK-A211) for 15 min at room temperature. The cells were resuspended in flow cytometry buffer and analyzed by flow cytometry. For ROS detection, the single-cell suspension was incubated with DCFH-DA (1:1000) for 30 min at 37 °C. Cells were washed twice with 1 × PBS at 4 °C, then were resuspended in flow cytometry buffer and analyzed by flow cytometry.

For detecting intratumoral immune cells, the fresh tumor tissue was cut into 1 to 2 mm³ pieces. Incubated in collagenase IV(11.5 mg/ml) and DNase I (50 μg/ml) in RPMI at 37 °C for 60 min with agitation. Filtered through 70 μm strainers. Centrifuged (300–400×g, 10 min) and lysed RBCs using ACK buffer (5 min, RT). Resuspended cells (1 × 10⁶ cells/100 μl) in PBS/2% FBS. Added fluorochrome-conjugated antibodies [e.g., anti-CD45 (BD_103140), CD3 (BD_100203), CD4 (BD_116005), CD8 (BD_100210), CD19(BD_115507)] and incubated 30 min at 4 °C in the dark. Wash twice with 1×PBS with 2% FBS. Finally, the immune cells were detected and anlyzed using a flow cytometry.

Tumor samples and IHC analysis

This study adhered to the Declaration of Helsinki principles. This study was approved by the Institutional Review Board of the Institute for Henan Medical University. A total of 39 cases were included in the IHC scoring, with patient details summarized in Table S5. These patients are aged 30 to 60 years old without a proprietary weight standard. The patients with surgeries to resect tumors should have adequate organ function, meeting specific thresholds for hemoglobin, neutrophil count, platelet count, liver enzymes, creatinine clearance, and bilirubin levels. All included patients were retained to the end. All tumor and para-tumoral tissue were collected simultaneously. The study included patients with cholangiocarcinoma confirmed by histology or cytology. The samples underwent deparaffinization in xylene, rehydration through graded ethanol, and dH₂O washing.

Briefly, antigen retrieval was performed using sodium citrate buffer (pH 6.0). After inhibiting endogenous peroxidase and blocking nonspecific binding, sections were incubated with primary antibody at 4 °C overnight, followed by PBS washes and incubation with secondary antibody for 1 h at room temperature. Subsequently, sections were treated with StrepABComplex/HRP for 1 h. Chromogenic detection used 0.05% DAB, and counterstaining was done with hematoxylin, with normal serum as a negative control. Protein expression in normal and HCC tissues was evaluated independently by two blinded pathologists, with disagreements resolved through consensus. The IHC evaluation of protein expression intensity in normal adjacent tissues and paired HCC tissues was performed independently by two pathologists from the Department of Pathology, Xinxiang Central Hospital. The pathologists were blinded to the patient’s clinical features; in cases of disagreement, a consensus was reached by joint review. All experimental procedures involving humans were approved by the Human Ethics Committee of Henan Medical University (approval no.XYLL-20246634).

Immunofluorescence

The excised tumor tissue was fixed in 4% paraformaldehyde (PFA) for 24 h at 4 °C, dehydrated in 40% sucrose until tissue sinks (at least 24 h), then embed in OCT compound and stored at −80 °C. Cut the tissues to be 20 to 30 μm sections. The sections were permeabilized with 0.1% Triton X-100 in PBS for 10 min, blocked with 10% normal serum (e.g., goat serum) + 1% BSA in PBS for 2 h at RT to reduce nonspecific binding. Incubated sections with primary antibody (1:200 diluted in blocking buffer) overnight at 4 °C; washed three times with PBS (5 min each); fluorochrome-conjugated secondary antibody was added (1:200) and incubated for 30 min at RT in the dark. The cells were washed three times with PBS (5 min each); nuclei were counterstained with DAPI (10 min). Sections were mounted with anti-fade mounting medium (e.g., 90% glycerol) and imaged using a fluorescence microscope. The used primary fluorescent antibodies included anti-AID (YT5566, WB, 1:200), anti-MYST1 (CY8281, IHC, 1:200), anti-GLUT1 (CY5360, IHC, 1:200), anti-PGK1 (CY5536, IHC, 1:200), and anti-CXCR5 (CY8749, IHC, 1:200).

Syngeneic tumor-bearing mouse model

BALB/C mice weighing 18 to 20 g were maintained in specific pathogen-free facilities at Henan Medical University. The mice were classified into the tumor control group (n = 5), the 4-Deoxyuridine treatment group (n = 5), the 2-DG treatment group (n = 5) and the combined group (4-Deoxyuridine+2-DG; n = 5). A murine model of HCC was established by subcutaneous injection of 1 × 10⁶ H22 cells (CVCL_H613) suspended in 0.2 ml of PBS. The treatment were performed by intratumoral injection on the fourth, eighth 11^th, and 14th days after H22 cell seeding subcutaneously. The tumor growth was monitored by measuring tumor size in two orthogonal dimensions. Tumor volume was calculated using the formula: volume = (length × width²)/2. Mice were killed after 14 days. All the mice were sacrificed to collect the solid tumor, and the tumor weights were measured. All animal procedures were approved by the Institutional Animal Care and Use Committee of Henan Medical University (approval no.XYLL-20246634).

Statistical analysis

Unpaired t-tests were performed using GraphPad Prism 8 (GraphPad Software. Data are presented as mean ± SD. The expression levels and co-expression relationships of AICDA, SLC2A1 (GLUT1), and LDHA in clinical specimens, as well as the survival rates of patients with HCC, were analyzed using GENT2 (http://gent2.appex.kr/gent2/), a web-based gene expression database that integrates publicly available microarray datasets primarily from the Gene Expression Omnibus (GEO); Kaplan-Meier Plotter (https://kmplot.com/), an online survival analysis platform incorporating gene expression and clinical data from multiple public repositories, including TCGA, GEO, and EGA; and PanCanSurvPlot (https://smuonco.shinyapps.io/PanCanSurvPlot/), a survival analysis platform based on The TCGA database. A difference was considered statistically significant at ∗p < 0.05 (∗∗p < 0.01; ∗∗∗p < 0.001).