Multi-omics analysis identifies CKLF as a promoter for HCC progression by regulating the AKT, ERK pathways, and infiltrating immune cells Authors: Lei Peng, Jingjing Li, Ji Yuan, Shuangshuang Ma, Lei Jin, Guizhou Zou, Lei Tang Published: 2026 ABSTRACT Highlights•CKLF is significantly upregulated in hepatocellular carcinoma (HCC) and correlates with advanced tumor stage, aggressive molecular subtypes, and poor patient survival.•Single-cell and spatial transcriptomics reveal dual CKLF expression in both malignant hepatocytes and cytotoxic immune infiltrates (CD8⁺ T cells and M1 macrophages), implicating its role in tumor–immune crosstalk.•CKLF knockdown suppresses HCC cell proliferation, migration, and clonogenicity in vitro, highlighting its functional importance in tumor aggressiveness.•Oncogenic effects of CKLF are mediated through activation of the AKT/ERK signaling axis, as constitutively active Ras rescues phenotypic deficits caused by CKLF silencing.•Integrative multi-omic analyses position CKLF as a pivotal molecular node linking genomic alterations, stromal remodeling, and immunophenotypic shifts, underscoring its potential as a biomarker and therapeutic target in HCC.Keywords: CKLF, AKT/ERK pathway, Tumor stroma, Omic analysis, Hepatocellular carcinoma Introduction Hepatocellular carcinoma (HCC) remains a leading cause of cancer-related mortality worldwide, with dismal outcomes for patients diagnosed at advanced stages despite recent advances in kinase inhibitors and immunotherapy [[1], [2], [3]]. The primary challenge lies in the profound molecular heterogeneity of HCC and the lack of precise, actionable targets that can effectively disrupt tumor progression or remodel the immunosuppressive microenvironment [4]. Therefore, identifying novel drivers that bridge intrinsic oncogenic signaling with extrinsic immune regulation is critical for developing individualized therapeutic strategies. Among potential candidates, Chemokine-like factor (CKLF) has emerged as a molecule of significant interest. Structurally distinct yet functionally similar to conventional chemokines, CKLF binds to G protein–coupled receptors (e.g., CCR1 and CCR4) to activate downstream cascades governing inflammation, tissue repair, and cell migration [5,6]. While aberrant CKLF expression has been implicated in the proliferation and metastasis of various malignancies [[7], [8], [9], [10]], its role in HCC remains paradoxical and poorly defined. Unlike classical chemokines that primarily recruit immune cells, CKLF appears to exhibit a "dual role," potentially acting directly on tumor cells to drive growth while simultaneously modulating the tumor microenvironment (TME) [[7], [8], [9], [10]]. However, a comprehensive understanding of CKLF’s genomic landscape, its regulation across specific HCC molecular subtypes, and its precise mechanistic contribution to HCC progression is currently lacking. Despite these insights, critical knowledge gaps regarding CKLF in HCC persist. Specifically, the genomic landscape of CKLF remains uncharacterized. Furthermore, while its prognostic value is suggested, the precise correlation between CKLF expression levels, distinct molecular subtypes, and specific immune phenotypes within the HCC microenvironment has not been systematically defined. Most importantly, the mechanistic underpinnings of CKLF-driven tumorigenesis are unclear: it remains to be determined whether CKLF promotes HCC progression by directly activating oncogenic signaling axes (PI3K/AKT or MAPK/ERK) or indirectly by reprogramming the functional state of infiltrating lymphocytes and macrophages to foster immune evasion. Resolving these ambiguities is a prerequisite for validating CKLF as both a robust biomarker and a viable therapeutic target. To address these challenges, we established a multi-omics platform integrating bulk transcriptomics, single-cell RNA sequencing (scRNA-seq), spatial transcriptomics, and genome-wide screening to dissect the role of CKLF in HCC comprehensively. In this study, we not only validate CKLF as a top-ranked prognostic driver through rigorous machine learning approaches but also elucidate its dual mechanism: promoting tumor cell proliferation via key signaling pathways and reshaping the immune landscape towards an exhausted phenotype. Our findings provide a biological rationale for targeting CKLF, offering new insights into patient stratification and potential combination therapies for HCC. Materials and methods Single-cell preprocessing and integration We performed computational analysis in R programming environment using Seurat package and special codes. Quality control of cells was kept for expression from 200 to 2500 genes but removing any cell that had >10% of the expressed genes on mitochondria; then we removed duplicates and poor-quality clusters.Data were normalized with the SCTransform method and integrated across different HCC scRNA-seq data sets based on the Harmony algorithm,where we used UMAP for visualizing dimensionality reduction. We clustered using the Louvain algorithm with resolution value [11,12]. Spatial data preprocessing and normalization Raw count matrices were processed using the Seurat packages in R. Quality control (QC) was strictly applied to filter out low-quality spatial spots: spots containing fewer than 500 detected unique molecular identifiers (UMIs) or genes, and spots with a mitochondrial gene percentage exceeding 20%, were excluded from downstream analysis. To ensure robust comparability of gene expression levels across different patient samples and to mitigate technical variations (e.g., sequencing depth), data normalization was performed using the SCTransform method . This approach stabilizes variance and removes the confounding effect of sequencing depth while regressing out the percentage of mitochondrial genes. To resolve the cellular composition within each spatial spot, we employed RCTD. Our previously generated single-cell RNA sequencing dataset from matched HCC cohorts served as the high-resolution reference profile. For the generation of heatmaps and spatial expression plots, consistent color scales were applied across all samples. RCTD was used for determining the location of macrophages and cancer cells found in the scRNAseq dataset (https://www.science.org/doi/10.1126/sciadv.abg3750). Immune features and checkpoint landscape We used bulk RNA-seq data for studying gene expression levels of critical immune checkpoint proteins, such as PD-1,CD274, CTLA4, LAG3, TIGIT), antigen processing molecules (HLA-A/B/C, HLA-DRA, B2M) and chemotactic factors (CCR1/4, CXCR3) in immune cells.In the TCGA-LIHC and ICGC databases, three kinds of software(xCell, EPIC, ssGSEA) are used for predicting infiltrating proportion of immune cells,followed by a Spearman correlation test with multiple hypothesis correction (FDR). Correlations were calculated against the expression of CKLF. Survival analyses Kaplan-Meier plots were constructed and compared via log-rank tests for significance of difference in overall survival when patients are categorized according to their respective CKLF expression level, split on the median value. Significant prognostic indicators included age ofstage of the disease, and etiology have been adjusted into multivariate Cox regression analysis. Cell lines and culture The human hepatocellular carcinoma cell lines Huh7 and LM3 as well as the immortalized hepatocytes line MIHA were purchased from the Chinese Academy of Sciences Cell Repository. The cells were cultured using Dulbecco’s Modified Eagle Medium (Gibco) supplemented with 10% FBS and antibiotic solution (penicillin/streptomycin; 1%) at standard culture conditions (37 °C,(5% CO 2 humidified atmosphere).All experiments are performed with exponentially growing cells having a cell viability of >95%. Wound-healing assay For wound healing assays, Huh7 cells were seeded into a 6 well plate and grown to confluency. A wound was introduced by dragging an unused sterile pipette tip with a volume of 200 µl,and after two PBS washes, cells were transferred into starvation conditions. In order to analyze migration without confounding cell growth,cells were pre-treated by mitomycin C(10 μg/mL, 2 h).Cellular migration into the wound was monitored every 48 h by counting cells in selected regions of interest that were visualized using phase-contrast light microscopy following fixation. CCK-8 proliferation assay Cells were seeded at a density of 3 × 10³/well into 96-well plates and the outermost wells were filled with PBS for control purposes. After indicated periods (0,24 h, 48 h, 72 h, 96 h later, 10μl CCK-8 reagent(Dojindo) was added in each hole and incubated at 37 °C for 2 h respectively. The OD values at 450 nm were measured with subtracting blank value. All cell viability curves were calculated as relative to T = 0.At least three biological replicates (with three or more technical replicates) were used for the study. Immunohistochemistry (IHC) Paraffin-embedded tissue sections fixed with formalin were deparaffinized, hydrated, and subjected to antigen retrieval in [sodium citrate/EDTA] buffer. 3% H₂O₂ was used to block endogenous peroxidase, and 5% normal serum was used to block non-specific sites. The sections were incubated with [target protein] primary antibody (dilution ratio 1:100) at 4 °C overnight, followed by incubation with HRP-labeled secondary antibody. DAB staining was performed, and hematoxylin was used for counterstaining. Two pathologists independently evaluated the staining intensity and the proportion of positive cells, and calculated the immunoreactivity score. Statistics for functional assays The results of functional tests were shown as mean ± SD based on at least three independent biological repeats; the comparison between two groups was performed using two-tailed Student’s t-test,where comparisons between several groups were performed with one-way ANOVA followed by Tukey’s test, and significance level for a given result was defined as both the absolute value of log2-fold change above 1 or equal to 1 and p-value <0.05 [24]. Result Integrated machine learning screening identifies CKLF as a core prognostic driver in HCC To systematically identify robust molecular regulators driving hepatocellular carcinoma (HCC) progression, we integrated transcriptomic data from multiple large-scale cohorts (TCGA-LIHC, ICGC-LIRI, and GEO datasets). We employed a consensus machine learning strategy combining Random Forest for feature importance ranking and Elastic Net regularized regression for robust feature selection. Our cross-cohort analysis demonstrated that the Elastic Net model provided the most stable predictive performance, significantly outperforming Lasso, Ridge, and stepwise Cox models in terms of time-dependent AUC across independent datasets (Fig. 1A). This model identified a concise 29-gene prognostic signature. Among these candidates, CKLF emerged as a top-ranked feature with a strong positive coefficient, indicating that its upregulation is consistently associated with poor overall survival (OS) and disease-free survival (DFS). In contrast, genes such as ZKSCAN1 exhibited negative coefficients, suggesting a protective role (Fig. 1B). ML methods and regularization for the discovery of biomarker predictors in HCC. (A) The importance plot shows, across 500 RF models, that genes like NUTF2,RREB1, SECISBP2, and CTYOR.(B) Elastic net regression coefficient estimates of selected SNPs along with standard errors;coefficients greater than zero associate with poor clinical outcome.(C) Left panel: The lasso path of coefficients, indicating which features enter the model as we move down the regularisation path;right panel: optimalization of the model with respect to lambda, depending on classification accuracy. Using this signature, we constructed a risk score model that effectively stratified patients into high- and low-risk groups. This stratification was statistically significant across all six independent validation cohorts (log-rank p < 0.05 for all; Fig. 1C), confirming the external validity of our screening approach. Crucially, CKLF stood out not only for its statistical weight in the prognostic model but also for its consistent ranking across different algorithms. While other genes like USP2 and JPH1 were also identified, the prominence of CKLF prioritized it as the most promising candidate for further functional investigation. Consequently, subsequent analyses focused specifically on dissecting the molecular mechanisms and therapeutic potential of CKLF in HCC. Robustness validation of CKLF as a top-ranked predictor via ensemble machine learning To ensure the identified prognostic markers were not artifacts of a single algorithm, we employed an ensemble machine learning approach to validate feature importance and generalizability. First, Random Forest analysis was performed across iterative runs to assess gene stability. While several genes showed high importance scores, CKLF consistently emerged as a top-tier predictor, maintaining high ranking across diverse data subsets (Fig. 2A). Permutation tests further confirmed that the predictive power of CKLF was statistically significant and superior to many other candidates, including ITPA and ZKSCAN1, which showed weaker or less consistent associations (Fig. 2B). Validation performance and prognostic significance of an Elastic Net-derived gene signature for hepatocellular carcinoma. (A) The AUC values across different time periods indicate that the Elastic_net_0.3 performed better than other models on several datasets,showing improved performance. (B) Heatmap of regression coefficient values for the 29-gene set in Elastic_net_0 0.3, with red showing positive predictive factors and green indicating negative predictive factors; associated bar graph shows the weight ranking for key genes). C) Kaplan-Meier curves for different groups stratified based on low or high risk scores from the Elastic_net_0.3 model across all 6 cohorts,where there were significant differences between groups for the corresponding OS and RFS curves (p-value of log rank test <0.05). Subsequently, we utilized regularized regression paths to evaluate the stability of gene coefficients under varying penalty parameters (λ). As the regularization parameter decreased, CKLF was among the first genes to enter the model and maintained a large, stable coefficient, indicating its indispensable role in predicting patient outcomes (Fig. 2C). This cross-method consistency—where CKLF was highlighted by both non-linear (Random Forest) and linear (Regularized Regression) models—strongly supports its biological relevance over other potential candidates. These results provide a rigorous, data-driven foundation for prioritizing CKLF as the primary target for subsequent mechanistic and therapeutic investigations in HCC. Elevated CKLF expression patterns in hepatocellular carcinoma correlate with aggressive tumor characteristics and poor differentiation status To explore whether there was a difference of CKLF expression level in HCC between various patients, we performed Wilcoxon rank-sum test for comparing HCC vs normal liver tissue. We found that CKLF showed an up-regulation in HCC tumors versus normal liver tissue.Moreover, this result could be reproduced in the other five data sets including TCGA, GSE14520, GSE39791, GSE54236, GSE113996, and GSE144269 with statistical significance (Fig. 3A), which indicates that the high expression levels of CKLF can be observed among different cohorts or experiments. Therefore, it is reasonable to assume that CKLF may act as an important biomarker for HCC development.Furthermore, we performed a pattern analysis of the molecular classes and found that for some tumor classes there was an enrichment in those expressing CKLF (CKLF:within each of the three molecularly defined groups from the LIHC data set (Cluster 1–3), high CKLF expression levels were more prevalent in cluster 3 (P = 0.02),suggesting a possible association with more aggressive tumor phenotypes. Expression profiles and prognostic value of CKLF in HCC. (A) Box plots show the significantly higher CKLF expression in tumor tissues compared to paired nontumorous liver tissues from 6 cohorts(TCGA,(GSE14520, GSE39791, GSE54236, GSE113996, GSE144269) and differential expression was calculated by the Wilcoxon rank-sum test.(B) Violin plots represent the distributions of CKLF expression within each molecular classification group (Cluster 1–3) and tumor differentiation stage (G1– G4),which was strongly upregulated in Cluster 3 and poorly differentiated G4 tumors implying that it is associated with an increased malignant potential of the tumors. The same trend was observed in pathway analysis at the molecule level (Fig. 3B, left). Considering the four-grade system (G1- G4),that they saw an increase in CKLF expression correlated with tumour grade and that this was most pronounced at the highest grade (G4,CKLF was also significantly correlated with the PFS value of patients (P = 0.001), suggesting a potential role for CKLF during tumorigenesis (Fig. 3B, right).In summary, all the above evidences put CKLF as one of candidate markers that is constantly overexpressed in HCC.suggesting that it is closely related to the clinicopathological features, histological types of tumors. Therefore, this gene may be not only an indicator but also can be considered as a therapeutic target in future study. CKLF drives oncogenic signaling and tumor aggressiveness in hepatocellular carcinoma and predicts poor survival To elucidate the functional role of CKLF in HCC, we integrated differential gene expression profiling, pathway activation analysis, and clinical outcome data. Our analysis revealed a significant correlation between elevated CKLF expression and the activation of key oncogenic signaling cascades. Specifically, tumors with high CKLF levels exhibited markedly enhanced phosphorylation of ERK (Wilcoxon test, p = 0.003) and AKT (Wilcoxon test, p = 0.008) compared to low-expression counterparts (Fig. 4A and 4B), suggesting that CKLF promotes cellular proliferation and survival through these critical pathways. Furthermore, Gene Set Enrichment Analysis (GSEA) confirmed that high CKLF expression is strongly associated with core biological processes driving tumor progression, including [insert specific pathways here, e.g., cell cycle regulation, EMT, or metabolic reprogramming] (Fig. 4C). Clinically, Kaplan-Meier survival analysis demonstrated that patients in the LIHC cohort with high CKLF expression suffered significantly worse overall survival compared to those with low expression (p = 0.007; Fig. 4D). These findings collectively indicate that CKLF not only fuels oncogenic signaling and tumor aggressiveness but also serves as a robust prognostic biomarker for poor survival in HCC. The role of CKLF on HCC progression. (A) Phospho-ERK staining is stronger in the high CKLF group than that of the low CKLF group, p = 0.003; (B) phospho-AKT staining is also stronger in the high CKLF group than that of the low CKLF group, p = 0.008suggesting possible AKT pathway engagement.(C) A significant enrichment of pathways is observed for CKLFhigh tumors; the top enriched pathways are shown in this plot from the GSEA. (D) The KM curves show a statistically lower OS rate in patients with higher levels of CKLF expression (CKLF-high: N = 239) than those with lower levels of CKLF expression (CKLF-low: N = 126),with statistical significance (log-rank p = 0.007).(E) Global GSEA analysis of the TCGA-LIHC dataset reveals a series of signaling pathways such as OXPHOS,DNA repair, glycolysis, immunity, which are up-regulated in CKLF-higher tumors,suggesting a possible role of SLC7A9 in metabolism and immunological response to tumour cells. CKLF mutation promotes HCC progression To comprehensively characterize the genomic landscape of CKLF in HCC, we analyzed multi-omics data from the TCGA-LIHC cohort. Copy number variation (CNV) analysis revealed a significant accumulation of GISTIC scores at the 12q24.31 locus, where CKLF resides, showing frequent amplification across 370 liver tumor samples (Fig. 5A). This suggests that copy number gain is a primary mechanism contributing to elevated CKLF levels in tumors. Genomic landscape and copy number-driven expression of CKLF in hepatocellular carcinoma. (A) Gistic score profile across the genome from TCGA-LIHC cohort (n = 370), showing a significant peak, where CKLF is located, indicating recurrent amplification. (B) Variant analysis of CKLF: left panel shows variant classification distribution; middle panel displays variant types (SNP, INDEL); right panel presents SNV class distribution, with C>T transitions dominating. Bottom panels show variant burden per sample and top 10 mutated genes, with CKLF among the most frequently altered. (C) Boxplot comparing CKLF mRNA expression across copy number states (shallow deletion, diploid, gain), revealing significantly higher expression in copy number-gained samples (p = 0.021). (D) Scatter plot showing positive correlation between normalized CKLF expression and gistic copy number score . (E) Left panel: CKLF expression increases with genomic instability (FGA), particularly in copy-number-gained genomes; right panel: expression correlates with tumor grade (G1–G4), suggesting association with disease progression. In parallel, we examined somatic mutations and found that while CKLF harbors recurrent alterations, it is not universally mutated as previously misinterpreted; rather, it displays a specific mutational profile in a subset of patients. The observed variants were predominantly missense mutations characterized by C>T or T>C base transitions (Fig. 5B, upper panels). In silico structural analysis suggests that these missense mutations may potentially affect protein conformational stability or secretion efficiency, which could theoretically modulate chemokine activity. However, statistical correlation analysis revealed no significant association between the presence of these somatic mutations and global transcriptomic dysregulation or specific structural variant (SV) patterns. This implies that while point mutations exist, they are likely secondary events rather than the primary drivers of the oncogenic phenotype. Consistent with this hypothesis, we observed that CKLF mRNA expression was significantly upregulated in tumors with copy number gains compared to those with diploid or deleted status (p = 0.021; Fig. 5C). Furthermore, a strong positive correlation was identified between CKLF expression levels and its copy number gain scores (Fig. 5D), indicating a gene dosage effect where genomic amplification directly drives transcriptional upregulation. Notably, tumors exhibiting high genomic instability (GIN), particularly those with broad regional amplifications, showed the highest CKLF expression levels (Fig. 5E). Collectively, these findings demonstrate that while CKLF acquires occasional missense mutations, its role as an oncogene in HCC is principally driven by genomic amplification. This CNV-mediated overexpression leads to a surge in CKLF signaling that correlates with chromosomal instability and aggressive tumor progression, highlighting gene dosage, rather than mutational lesions, as the key regulatory switch for CKLF in HCC. Cytotoxic T cells, as well as pro-inflammatory macrophages in HCC tumour microenvironment selectively express the chemokine like factor CKLF We next explored the immunological context of CKLF in HCC by performing scRNA-seq analysis and determining where it is expressed across different immune cell types.Using the UMAP based visualization technique we were able to identify distinct clusters of immune cells that include several subpopulations such as memory and effector CD4+ T cells,CD8+ cytolytic T-cells, NK cells, Mφ/macrophage (derived from monocytes), DCs, B cells, plasma cells (secreting antibodies)and natural lymphocyte subsets (Fig. 6A).We found that there was strong CKLF staining of certain leukocyte subsets particularly including CD8+ cytotoxic T cells,classically activated macrophage cells (M1 type), gamma delta T cells, etc.). The same is true of this particular expression profile where the CKLF transcriptional signature was clustered in those specific UMAP groups (Fig. 6B).The statistical analyses showed that the CKLF transcript was expressed at a much higher level in effector CD8+ T cells, M1 macrophages, and conventional DC subtype 1 (dDC1) than in any of the remaining immune subsets (p < 0.001) (Fig. 6C,upper panel) and quantitative distributional comparison (boxplots; also reproducible. Immune cell-selective CKLF expression patterns identified by single-cell transcriptomic analysis of HCC. (A) Uniform manifold approximation projection (UMAP).with expression of the CKLF transcripts visualized by a rainbow gradient scale (note that high intensities are observed in cytolytic CD8+ T cells,Macrophages (M1 polarised), γδ T-cells and dendritic cells.(B) Heatmap plots show the distribution of CKLF across the space defined by the UMAP axes, with colour intensity reflecting expression level andindicating enrichment in certain cell types of the immune system. C). Upper panel: statistical data shows mean expression levels for CKLF genes in major immune cells,showing statistically significant overexpression in CD8+ effector T cells, M1 macrophage, and conventional dendritic cell type 1 (p < 0.001); lower panel: boxplots of the LIHC_GSE140228 dataset confirm stable expression within cytotoxic and immune cells,compared to undetectable levels in B-cells, plasma cells secreting antibodies and regulatory/suppressor T-cells. In the LIHC_GSE140228 dataset, we found that the level of CKLF was higher in cytotoxic CD8+ effector T cells, M1 macrophageand γδ T cells. On the other hand, CKLF was not expressed at significant levels in B cells, plasma cells or Treg cells (Fig. 6C, bottom panel).All these observations suggest that CKLF is differentially expressed not only by malignant cells, but may also have a potential important function in the multiple immune cell subpopulations participating to antitumoral response,particularly in cytotoxic lymphocyte, and inflammatory macrophage populations.The unique immune cell expression signature implies that CKLF might be involved in the regulation of immunity during HCC development,that could be affecting either immunity promotion or suppression in the tumor microenvironment. Overall, these results indicate that CKLF is involved both in tumorigenesis and immunology,is becoming a desirable target for treatment in liver cancer immunotherapy. Spatially resolved transcriptomics identifies region-specific CKLF expression in hepatocellular carcinoma To investigate the spatial expression pattern of CKLF in HCC tissue, we carried out multiplex spatial transcriptomic analysis based on six representative tumor samples and created spatial distribution heatmaps to show heterogeneous CKLF expressions,showing a heterogenous distribution, where regions of high CKLF levels (i.e., in red/yellow tones) are often observed in close proximity to cancer cell aggregates and/or at the edge of the invasion front (Fig. 7A-F).CKLF expression appears to be spatially graded across some samples such that increased levels of CKLF are associated with regions containing proliferating cells and/or infiltrating immune cells,whereas down-regulated in the necrosis or stroma. Interestingly, some tumours show heterogeneous or focal patterns of expression clusters,which suggests the possibility for an association with specialized niches (e.g., hypoxia, immune cell aggregates).This consistent detection of high levels of CKLF expression in specific areas of tissues in different specimens suggests that it may be playing a role in localized functions inside the tumor. Spatially variable and -dependent expression of CKLF in tumour samples by spatial transcriptomics.(A-F). Multi-dimensional spatial gene expression patterns from six representative tumour samples,where the CKLF expression is colored (from blue for low expression to red for high expression).CKLF may also be involved in the progression of tumors or immune response, because these genes were reported (https://www.science.org/doi/10.1126/sciadv.abg3750). CKLF shows a clear difference in expression between the functional regions such as proliferation regionangiogenesis, and immunity). The different localization patterns provide compelling evidence for a non-uniform production of CKLF with specific adaptation to local conditions in HCC tissues.It may therefore be considered as an ideal spatial biomarker that can indicate high-grade tumour region or immunologically active area. All these findings demonstrate that the spatial heterogeneity should not be ignored while discussing the role of CKLF in HCC. CKLF modulates the immune microenvironment in HCC through chemokine-like signaling and adaptive immune activation In order to explore more about the role of CKLF as an immunomodulator for HCC, we combined several omics data set together and analyzed the correlations between CKLF and infiltration level ofsignaling networks, and protein interaction. We found that CKLF expression is positively correlated with the abundance of cytotoxic/inflammatory T cells andlike CD8+ T-lymphocytes, M1 polarised macrophage and DCs in different patient cohorts.On the other hand, we found a negative correlation for immunosuppressive regulatory T-cells suggesting that it may also play some role in maintaining the immunologically active tumour microenvironment (Fig. 8A). Immune regulation functions, pathways and genes related to CKLF in HCC. (A) Correlation heatmap of CKLF expression with the infiltration level of different immune cells among different groups,suggesting positive associations with activated immune cells, e.g., cytotoxic T-cells (CD8+), classically activated macrophage (M1 type), etc. and negative correlations to immunosuppressant cells, such as regulatory T-cell.(B) Schematic of potential protein interactions between CKLF with its putative ligands such as CXCR3, CCR1 and CCR5 which are expressed on Th1,Th2, and NKs; the color intensity indicates cell-cell interactions strength and function (inhibitory or stimulatory).(C) Scatter plots of significant correlations found between the levels of CKLF and immune-related genes: positive correlation with CD4+ helper T (Th) cell marker 1 and marker 2,and inverse relationships with vascular endothelium markers and granulocyte infiltration, indicating participation of these pathways to certain immune response and the control of tumour micro-environment. CKLF may interact with the chemokine receptors CXCR3, CCR1, and CCR5 (which are mainly expressed by Th1,Th2, and natural killer cells. This result supports its putative function of being a chemotactic agent which may promote the movement and activation of these immune cells (Fig. 8B). We also performed pathway enrichment analyses to show that CKLF expression is highly correlated with Th1 and Th2 differentiation pathways,its dealings with The adaptive arm of the immune system is inversely correlated with CKLF; conversely, CKLF is positively correlated with angiogenesis markers as well as neutrophil-related genes,suggesting some anti-angiogenic activity and mixed effect in immune response modulation at early stages of infection (Fig. 8C).Together, these findings highlight CKLF as an important link between tumour-intrinsic and extrinsic (immune) regulation of HCC,as it can act simultaneously as an early predictor of the disease and target of immunotherapy. Elevated CKLF expression occurs in hepatocellular carcinoma at both transcriptional and translational levels The immunoblot showed that the expression level of CKLF was significantly higher in HCC tissues than that in adjacent normal tissues,with a mean fold change of approximately 1.5-fold (p < 0.0001) (Fig. 9A). Similarly, with real time PCR analysis, we found that expression levels for CKLF were significantly upregulated in HCC compared to matched non-tumoral tissues (p < 0.0001),verifying increased gene expression activity (Fig. 9B, E).To further assess the extent to which it is expressed in different cell line models for HCC, we tested MIHA (normal liver cell line),Huh7, and LM3 (with a highly metastatic capacity). Both protein (Fig. 9C), and mRNA levels (Fig. 9D) increased progressively along the cell line series,from MIHA via Huh7 up to LM3 and was most strongly expressed by the aggressive LM3 cell line (p < 0.0001 for all comparisons).Taken together, these results demonstrate that CKLF is consistently highly expressed on the mRNA and protein level in human liver cancer tissue specimens as well as in vitro cells of hepatocarcinogenesismost commonly seen in progressive/metastatic disease. CKLF is highly expressed at the mRNA and protein level in human HCC tissue and cell lines. (A) The expression of CKLF was elevated in HCC tissues compared with that in paired NAs as shown by western blot analysis, and there was an increase of more than one hundred percent (p < 0.0001).(B) qPCR analyses reveal significantly higher levels of CKLF transcripts in the HCC tissues than in the NC samples (* p < 0.0001). (C) Western blo, (E) IHC and (D) qPCR analysis of CKLF expression. These results together indicate that CKLF is a consistently upregulated factor in HCC, which is positively associated with the development of disease. Reducing the expression of CKLF inhibits the proliferation of hepatocellular carcinoma cells As shown in our study, CKLF downregulation significantly alters several important aspects of cancer cells’ behavior. To determine whether CKLF knockdown altered cell growth rates, we performed Cell Counting Kit-8 (CCK-8) tests and found thatwe found that both si-1 and si-2 led to significant reduction of the number of cells surviving for four days compared with NC,demonstrating the important role of CKLF for cell proliferation (Fig. 10A).Western blot was used to detect protein levels in cells after transfection of CKLF-specific siRNAs (si-1 & si-2). The results showed that both siRNA inhibited the expression of CKLF; however, the latter had a stronger inhibition effect on CKLF.while si-3 failed to significantly reduce CKLF protein expression (Fig. 10B). Consistently, qPCR quantifications revealed significant reductions of CKLF mRNA expression in cells transfected by si-1 and si-2 (p < 0.0001),reinforcing the target specificity of such interventions (Fig. 10C).Flow cytometric analysis for cell cycle distribution showed that knockdown of CKLF led to accumulation of cells at G0/G1 stage and reduction in the fraction of cells entering S phase (Fig. 10D,left column), indicating that CKLF promotes progression through the cell cycle. In addition, we tested for cell motility using a wound-healing assay and found. Silencing CKLF leads to reduced proliferation, cellcycle progression, migration and invasion in HCC cells. (A) CCK-8 assay showed that the proliferation was inhibited by silencing CKLF for 96 h with two different small interfering RNA (siRNA: si-1 and si-2),compared with controls. (B) Western blotting and densitometry confirmed efficient knockdown of proteins by si-1 and si-2,whereas si-3 had no effect; GAPDH was used as a normalizer.(C) qPCR results showed significant decrease in CKLF mRNA expression by transfection of si-1 & si-2 (* p < 0.0001,with si-3 showing little effect. (D) Flow cytometric cell-cycle analysis revealed a G0/G1 arrest and decreased progression into the S phase upon CKLF knockdown.(E) Scratch assay showed poor wound closure ability for the cells transfected by si-1 and si-2 (at both 24- and 48-h timepoints,with significant difference vs. control (p < 0.0001), which altogether indicates that CKLF is essential for the maintenance of hepatocellular carcinoma cell proliferation,cell proliferation, cell cycle progression, motility. Decreased CKLF expression inhibits Ras AKT and Ras ERK signaling pathways to mitigate the development of hepatocellular carcinoma To further understand the effects of CKLF knockdown, we analyzed downstream protein levels following knockdown and recovery from knockdown. As shown by immunoblot analysis, knockdown of CKLF using two independent siRNAs significantly reduced p-AKT and p-ERK levels,that are known to be important downstream targets of both PI3K/AKT and MAPK/ERK signaling but with no statistically significant difference across the total levels of these proteins (Fig. 11A).Quantification demonstrated significant decreases of p-AKT / AKT ratio and p-ERK / ERK ratio after treatment with si-2 (*p < 0.0001,which confirms that CKLF is essential for maintaining these key cell survival/proliferation signals. In order to determine if restoration of downstream signalling would ameliorate the effects of CKLF loss,we transfected the cells using an active, constitutively active Ras construct.The subsequent proliferation assay with CCK-8 also indicated that Ras overexpression could rescue the cell proliferation ability of si-2-treated cells,achieved similar effects as the control, non-treated cells (Fig. 11B), suggesting that Ras-dependent pathways are able to compensate CKLF’s action in a functional sense. This was also corroborated by wound healing assays. CKLF promotes hepatocellular carcinoma progression through activation of the AKT and ERK signaling pathways, which can be rescued by constitutively active Ras. (A) Western blot analysis showing that CKLF knockdown (si-1 and si-2) significantly reduces phosphorylation of AKT and ERK, with no change in total protein levels; quantification reveals marked decrease in p-AKT/AKT and p-ERK/ERK ratios (p < 0.0001). (B) CCK-8 assay demonstrating that overexpression of constitutively active Ras restores cell proliferation in si-2-treated cells, rescuing the growth defect induced by CKLF silencing. (C) Wound healing assay showing that Ras expression reverses the impaired migration caused by CKLF knockdown, as evidenced by increased scratch recovery at 48 h (p < 0.0001). (D) Flow cytometry analysis indicating that Ras re-expression alleviates G0/G1 phase arrest, restoring cell cycle distribution toward control levels. These findings confirm that CKLF exerts its oncogenic effects primarily via the Ras/AKT and Ras/ERK signaling axes, positioning it as an upstream regulator of key proliferative and migratory pathways in HCC. In wound healing assays, we showed that re-expression of Ras was able to rescue the migratory defect induced by CKLF knockdown,wherein we found a statistically significant increase in wound closure rate, p < 0.0001 (Fig. 11C). Flow cytometric cell cycle analysis demonstrated that rescue of Ras expression relieved the CKLF knock-down-induced cell cycle arrest at G0/G1,resulting in the normalization of the S and G2/M phases population (Fig. 11D).Taken together, our data suggest a key role of CKLF in HCC growth and migration through activation of both the Ras/AKT and Ras/ERK pathways,and that these oncogenic activities could potentially be reversed by the restoration of this pathway. Such knowledge places CKLF in an important position, acting upstream of key oncogenic processes, and further supports it being considered a therapeutic target in liver cancers. Discussion HCC is characterized by profound alterations in the function of cytotoxic immune cells, including CD8+ T cells, γδ T cells, and M1-polarized macrophages. These immune deficiencies, coupled with a background of chronic inflammation (often driven by HBV/HCV infection or metabolic stress), create a unique and highly immunosuppressive tumor microenvironment (TME) [[13], [14], [15], [16], [17], [18],[25], [26]]. In this study, we present compelling evidence that Chemokine-like factor (CKLF) serves as a critical clinical biomarker for HCC progression. By integrating multi-omics approaches with experimental validation and spatial localization, we demonstrate that CKLF is not merely a bystander but an active driver of HCC development. Our data suggest a dual mechanism of action: first, by directly stimulating pro-oncogenic pathways within malignant hepatocytes, and second, by orchestrating complex shifts in the immunocomposition of the TME. Taken together, our data indicate that CKLF is consistently upregulated across various HCC patient cohorts. This aberrant expression appears largely driven by genomic amplifications, suggesting that CKLF may function as a bona fide oncogene subject to copy-number gain. This genomic dependency aligns with the "oncogene addiction" hypothesis, wherein cancer cells become reliant on overexpressed drivers for proliferation and survival, thereby rendering them susceptible to targeted interventions. Functionally, we observed that silencing CKLF significantly reduces HCC cell proliferation, motility, and cell cycle progression. Mechanistically, these effects are mediated through the activation of the PI3K/AKT and MAPK/ERK pathways, evidenced by the downregulation of phospho-AKT and phospho-ERK following CKLF knockdown. Importantly, the partial rescue of these defects via activation of Ras, a critical upstream regulator, positions CKLF as a potential non-canonical upstream activator of well-known oncogenic signaling cascades. Although CKLF lacks classic kinase domains, its structural homology to chemokines and its ability to bind G-protein coupled receptors (such as CCR1 and CCR4) suggest it may act as an agonist that triggers the Ras cascade via receptor-mediated signaling, a mechanism increasingly recognized in chemokine-driven malignancies. A particularly novel and intriguing finding of our work, revealed through single-cell and spatial transcriptomics, is the high expression of CKLF not only in malignant hepatocytes but also in cytolytic effector immune cells, specifically CD8+ T cells, γδ T cells, and M1-polarized macrophages. At first glance, this observation appears contradictory to the pro-tumorigenic role of CKLF described above. Typically, the presence of CKLF in cytotoxic cells might be interpreted as a robust anti-tumor immune response. However, given its concurrent role in driving tumor progression, we propose two non-mutually exclusive hypotheses to explain this paradox, drawing on established principles of immune regulation in chronic inflammation. First, the high expression of CKLF in these immune populations may represent a "hijacked" immune response. In the context of chronic antigen stimulation within the TME, persistent exposure to chemokines can drive T cells and macrophages toward a state of exhaustion or dysfunction rather than effective killing. Similar to other chemokines that sustain chronic inflammatory loops, CKLF signaling via CCR1/CCR4 on CD8+ T cells or M1 macrophages might initially promote recruitment and activation but ultimately lead to desensitization, upregulation of immune checkpoint molecules, and loss of cytotoxic potency. In this scenario, the tumor effectively co-opts CKLF to maintain an influx of immune cells that are present in number but functionally impaired, thereby creating an illusion of immune surveillance while facilitating immune evasion. Second, the elevated CKLF levels in cytotoxic cells could reflect a compensatory expression under a chronic inflammatory environment. The HCC TME is dominated by pro-tumorigenic signals (e.g., IL-10, TGF-β, M2 macrophage polarization). The upregulation of CKLF in CD8+ T cells and M1 macrophages may be a desperate, endogenous attempt by the host immune system to counteract this suppression and restore a Th1-dominant, anti-tumor phenotype. Our data showing an association between CKLF signatures and increased Th1/Th2 ratios supports this view. However, in the face of overwhelming oncogenic CKLF signaling from the tumor cells themselves, this compensatory effort is likely insufficient, resulting in a net pro-tumor outcome. This duality—where a molecule acts as both a tumor driver and a marker of immune struggle—is reminiscent of recent findings for other tumor antigens that concurrently trigger anti-tumor activity and mechanisms of acquired resistance [[19], [20], [21], [22], [23]]. This dual nature of CKLF presents a unique therapeutic opportunity. Unlike traditional targets that solely inhibit tumor growth or solely boost immunity, targeting CKLF could attack the disease on two fronts: disrupting oncogenic signaling in hepatocytes and potentially reversing the exhaustion or dysfunction of infiltrating immune cells. By blocking CKLF, we may not only halt tumor proliferation but also "release the brakes" on the hijacked immune response, allowing endogenous CD8+ T cells and M1 macrophages to regain their cytotoxic efficacy. This synergistic effect could significantly enhance the responsiveness of HCC to existing immunotherapies, such as immune checkpoint inhibitors. Furthermore, the strong association between CKLF expression levels and tumor differentiation, as well as its enrichment in invasive regions, underscores its value as a prognostic marker. The spatial heterogeneity of CKLF suggests it could be utilized to identify high-risk zones for local relapse or metastasis, guiding clinicians in post-surgical management and follow-up strategies. Although this study is comprehensive, several limitations still need to be acknowledged to ensure a balanced interpretation of the research results. Firstly, we are unable to delve deeply into its underlying regulatory mechanisms. Secondly, our clinical samples are also relatively scarce, and a large-sample cohort investigation is needed. Informed consent All procedures were approved by the ethics committee of the Second Affiliated Hospital of Anhui Medical University (approval number: SL-YW2022–152) before study enrollment, and all patients provided written informed consent for their inclusion in this study.written informed consent was obtained from each subject at the time of first medical visit, and this study complied with the Declaration of Helsinki’s ethics guidelines. Data availability Data underlying the results presented in this article are available upon reasonable request from the corresponding author. The data are not publicly available because of privacy and/or ethics reasons. Funding The study was funded by the projects on key research program of Anhui Provincial Department of Education(No.2024AH050809). CRediT authorship contribution statement Lei Peng: Data curation, Conceptualization. Jingjing Li: Data curation, Conceptualization. Ji Yuan: Data curation, Conceptualization. Shuangshuang Ma: Data curation, Conceptualization. Lei Jin: Data curation, Conceptualization. Guizhou Zou: Data curation, Conceptualization. Lei Tang: Funding acquisition, Formal analysis, Data curation, Conceptualization. Declaration of competing interest The authors confirm no conflicts of interest exist regarding this study. Acknowledgments Not applicable. References Associated Data Data Availability Statement Data underlying the results presented in this article are available upon reasonable request from the corresponding author. The data are not publicly available because of privacy and/or ethics reasons.