Skip to main content
NCBI home page
Springer logoLink to Springer
. 2025 Nov 25;26(1):44. doi: 10.1007/s10238-025-01963-x

Impact of single-cell cell cycle regulation of intercellular communication on the prognosis of hepatocellular carcinoma in the tumor microenvironment

Cong Hu 1,✉,#, Rui Deng 2,#, Shuxiong Nong 2,✉, Xinglang Mou 3,✉
PMCID: PMC12644129  PMID: 41286375

Abstract

Despite the emergence of new therapies such as, immunotherapy, the treatment and diagnosis of Hepatocellular Carcinoma (HCC) still face many challenges, and the therapeutic outcomes for patients remain less than satisfactory. The study identified differentially expressed genes in HCC through differential analysis and then used univariate cox regression analysis to identify genes associated with prognosis. The intersection of these two sets of genes was used to obtain differentially expressed prognostic genes in HCC, which were then subjected to enrichment analysis. We analyzed two single-cell RNA sequencing (scRNA-seq) datasets from HCC patients, comprising 24,637 cells. Non-negative Matrix Factorization (NMF) clustering was used to identify cell cycle regulation in HCC tumor microenvironment (TME) cells, including three cell subpopulations: proliferating cells (PC), dendritic cells (DC), and macrophages (MAC). We employed the CellChat package to analyze cell–cell communication, the Monocle package for pseudotime trajectory analysis, and the SCENIC software package to study gene regulatory networks. Survival analysis was also performed using cell cycle-related features. A total of 26 clusters, including 15 major cell types, were identified in the HCC samples. Complex cell–cell communication networks were observed among these cell types. Enrichment analysis revealed that these cells were mainly enriched in pathways related to the cell cycle. The expression of cell cycle-related genes was elevated in tumor samples, and changes in cell cycle-related genes in specific subtypes were associated with different overall survival rates. The study focused on single-cell level data analysis of the cell cycle. The bubble plot results showed that the cell cycle scores were significantly upregulated in the PC, DC, and MAC subpopulations. Further subtyping revealed that these subtypes exhibited distinct biological states, cell–cell communication, and metabolic pathways. This study demonstrates that cell cycle regulation and cell–cell communication within the HCC tumor microenvironment impact tumor progression and patient prognosis. Cell cycle dysregulation in TME cells correlates with poor prognosis and immunotherapy efficacy, suggesting cell cycle targeting as a therapeutic strategy.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10238-025-01963-x.

Keywords: Cell cycle, Prognosis, Hepatocellular carcinoma, Tumor microenvironment

Introduction

Hepatocellular carcinoma (HCC), a common malignant tumor of the digestive tract that is oneof the leading causes of cancer-related deaths, is rapidly increasing [1–3]. The major therapeutic options for HCC include early surgical resection, targeted therapy, and immunotherapy [4–6]. These standardized treatment regimens have greatly improved HCC patients’ survival rates. In contrast, HCC is characterized by a high rate of metastasis [6], recurrence [5], and drug resistance [5], all of which have a major effect on patient survival. Therefore, finding new biomarkers and learning more about their molecular mechanisms might lead to new understandings of HCC diagnosis and treatment and eventually enhance HCC therapy and prognosis.

The G1 phase (pre-DNA synthesis), S phase (DNA synthesis), G2 phase (pre-mitotic phase), and M phase (mitotic phase), is referred to as the cell cycle [7–9]. The normal progression of the cell cycle relies on a series of complex regulatory mechanisms, among which cyclin-dependent kinases (CDK) and their regulatory subunits, cyclins, are the core regulatory factors. Normal cell proliferation is strictly regulated by the cell cycle, whereas tumor cells often exhibit dysregulation of cell cycle mechanisms [8, 9]. Tumor cells have inactivated checkpoint regulation, giving cells the ability to skip typical cell cycle checkpoints, quicken the cell cycle, and thus lead to rapid proliferation of tumor cells.

Fibroblasts, malignant mesenchymal tumor cells, a complex vascular network, and immune cells are among the many invading cell types that make up the extremely complex tumor microenvironment (TME) of HCC [10–14]. The TME appears to play an important role in tumor formation, metastasis, and treatment resistance. In addition, single-cell transcriptomics has shown complicated cell–cell communication across several TME cell subtypes in cancer. Cancer-associated fibroblasts (CAFs), T and B cells, and tumor-associated macrophages (TAMs) are TME cells [15–17]. Different tumor cell types and TME have been shown to interact on multiple levels to support the progression of cancer. It is unclear what the likely mechanisms of these TME cells in HCC are. Understanding the precise mechanisms and relationships between cell clusters and the TME can therefore provide a deeper understanding of HCC and direct the development of diagnostic and treatment plans.

The impact of the cell cycle on primary HCC TME cells (such as CAFs, T cells,TAMs, and B cells) and the functional effects of the cell cycle on various cell types within the TME were studied using single-cell RNA sequencing (scRNA-seq). Non-negative Matrix Factorization (NMF) clustering was utilized to assess the various cell cycle-regulated patterns of HCC TME cells. The purpose of the study was to find distinct biological states and cell–cell communication patterns in HCC cells with cell cycle characteristics.

Materials and methods

Data collection

In the study, the single-cell RNA sequencing (scRNA-seq) data were derived from a single GEO dataset (login number GSE166635). This dataset includes the scRNA-seq spectra from two hepatocellular carcinoma (HCC) samples. The analysis is based on two HCC samples (GSM5076749 and GSM5076750) in the single dataset GSE166635. The transcriptome data, somatic mutations, and clinical materials of normal (n = 50) and HCC (n = 374) samples were also collected from TCGA. Samples were taken from two HCC patients’ tumours and surrounding tissues. After integrating the samples, we used a batch correction technique to retrieve the whole matrix data, including gene expression and phenotypes.

Visualization of cell types and subtype clustering in HCC samples

The scRNA-seq matrix was built using the Seurat software. We started by creating a Seurat object with the following filtering thresholds: mitochondrial numbers less than 20% and cells expressing fewer than 4,000 but greater than 500 genes. Cells that didn’t fit these requirements were deemed low-quality and eliminated. The top 2,000 genes with the highest variability were selected as features. These data were used as the basis for further analysis, and we employed the FindVariableFeatures function from the Seurat package. Subsequently, we used the ScaleData and RunPCA functions to determine the number of principal components, setting the threshold for the RunPCA function to 1:20. We further reduced the dimensionality using “t-SNE” and “UMAP” techniques. Using data derived from earlier HCC research, we annotated cell types, such as macrophages (CD163 and CD68), epithelial cells (EPCAM and CDH1), cancer-associated fibroblasts (CAFs) (ACTA2 and COL1A2), T cells (CD3, CD8A, CD4), dendritic cells (ITGAX), endothelial cells (PECAM1), NK cells (FGFBP2), and B cells (CD19, CD79A, JCHAIN). GPTCelltype automated annotations were manually verified [30]. The HCC samples were further annotated and visualized using this data.

Analysis of cell–cell communication

To examine possible relationships and evaluate cell–cell communication, CellChat software was utilized. First, we used NMF-processed scRNA data to create a CellChat object. After that, ligand-receptor pairings between various cell types were found, and the CellChat database was used to investigate cell communication. Through these interactions, specific communication patterns were discovered. The probability of cell–cell Communication was also calculated to provide insight into the molecular interaction networks across several cell types, including dendritic cells, CAFs, macrophages, endothelium cells, B cells, epithelial cells, and T cells. We were specifically interested in the interaction between tumor endothelial cells and TME subtypes that were identified by cell cycle-related genes.

NMF analysis of cell cycle-related genes in HCC TME cells

We performed NMF on the 29-gene cell-cycle signature matrix using the NMF R package (version 0.24). iteration stops when relative change < 1 × 10⁻4 for 50 consecutive steps or max-iter = 2 000 is reached. For each major cell type we evaluated ranks k = 2–10. We were subsequently able to identify different cell subtypes within these TME cell populations in order to identify potential patterns or structures with fewer components.

Pseudotime trajectory and cell cycle regulator analysis

Different NMF cell types in HCC were analyzed using the Monocle program. An empirical dispersion > 1 × dispersion fit and an average expression level ≥ 0.1 were the first criteria we used to identify highly variable genes. The “pseudotime heatmap” application was then used to generate a heatmap that displayed the dynamic expression patterns of cell cycle regulators along the pseudotime trajectory in different TME cell types in HCC.

Identification of marker genes for cell cycle-related genes in HCC TME cell subtypes

We utilized the FindAllMarkers method to find marker genes for the NMF cell clusters in HCC cell types after isolating the target cell populations. After establishing the log fold change (logFC) threshold at 0.5, we selected genes with logFC > 1 according to the list’s top-ranked cell cycle-related genes. These genes then served as the basis for renaming the NMF clusters. We displayed the top genes in each NMF cluster with the greatest expression levels using the DotPlot program to illustrate the results. The distribution of certain cell cycle genes in the HCC TME was also depicted using the feature plot tool.

Functional enrichment analysis of cell cycle-related NMF subtypes

Following the identification of cell cycle-related genes in several TME cell types, we used the clusterProfiler R tool to analyze Reactome pathway databases, gene ontology, and KEGG based on these marker genes. These pathways were arranged and visualized using Cytoscape’s enrichment map feature. Gene sets having an adjusted p-value < 0.05 were deemed significant by us. The three most important routes linked to these cell cycles were given priority.

SCENIC analysis of NMF cell cycle-related subtypes

We investigated the function of transcription factors (TFs) involved in gene regulatory networks in HCC using the aertslab/SCENIC software package, which we downloaded from GitHub. In order to facilitate the identification of transcription start sites and the exploration of gene regulatory networks in HCC scRNA-seq data, two gene motif rankings—hg19-tss-center-10 kb and hg19-500 bp-upstream—from the RcisTarget database were utilized. Following the Benjamini–Hochberg correction, TFs with p < 0.05 were chosen for additional examination.

Survival analysis of cell cycle-related features in mRNA sequencing datasets

The ggplot software was utilized to visualize the variations in cell cycle levels between the tumor and normal groups. To investigate cell cycle abundance, more analysis was done. Next, we used the FindAllMarkers tool to generate cell cycle-related gene signatures for each HCC NMF cell cluster based on scRNA-seq data. The primary cell types in the HCC TME were also identified. To mitigate technical noise when projecting scRNA-derived signatures onto bulk RNA-seq data, only marker genes with log₂FC > 1 and adjusted p < 0.01 in the scRNA-seq data were retained. Genes with mean expression < 0.5 transcripts per cell or detected in < 20% of target cells were excluded to remove drop-out–prone transcripts. The final signatures were further filtered to remove genes whose expression variance across TCGA samples was dominated by low-count outliers (Cook’s distance > 4). kcdf = “Gaussian” was used to model continuous RNA-seq counts, avoiding binary (present/absent) assumptions that amplify drop-out effects. min.sz = 10, max.sz = 200 ensured each signature contained enough informative genes while excluding overly large or small sets. All bulk datasets (TCGA-LIHC, TICG, ICGC) were log₂(TPM + 1) transformed and ComBat-corrected for sequencing platform and library preparation differences before GSVA. Spearman correlation between GSVA scores and the median expression of the same signature in matched single-cell pseudo-bulk replicates was > 0.85, confirming that technical loss did not distort the signal. Log-rank tests and Cox proportional hazards regression were employed to investigate the association between cell cycle-related NMF features and illness prognosis, including overall survival. We created Kaplan–Meier survival curves by calculating the cutoff values for distinct NMF cell properties across many available datasets using the Survminer R program.

Immunotherapy response prediction

The GEO and UCSC Xena databases provided us with survival statistics and mRNA information for HCC samples. After initial screening, these records were uploaded to the Tumor Immune Dysfunction and Exclusion (TIDE) website to predict future immunotherapy responses in HCC patients. We further examined the relationships between immunotherapy response and cell cycle-related NMF characteristics by analyzing the output data. Based on their individual datasets, we also assessed immunological checkpoints sourced from public sources.

qPCR

After being seeded in DMEM culture media supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (P/S), HepG2 cells were cultivated at 37 °C with 5% CO₂. Using the Trizol reagent, total RNA was isolated from HepG2 cells. The precise procedures were as follows: Following three PBS washes, cells were digested with 0.25% trypsin, and the cell suspension was collected and centrifuged. The Trizol reagent’s instructions were followed in order to extract the RNA. Reverse transcription was used to create cDNA using an mRNA reverse transcription kit (such the PrimeScriptTM RT reagent Kit with gDNA Eraser). TaKaRa’s TB Green® Premix Ex Taq™ II is one example of a SYBR Green quantitative PCR kit that was used for qPCR. At least three independent experiments were conducted to ensure the reproducibility and reliability of the results.

Results

Differential analysis of HCC samples

Between and nearby normal tissues, 9426 differentially expressed genes (DEGs) were found; 2580 them were pregulated, while 6846 were downregulated in TCGA (Fig. 1A). Univariate Cox regression analysis identified 5582 prognosis-related genes, and the intersection of these two sets yielded 1323 differentially expressed prognosis-related genes in HCC (Fig. 1B). According to enrichment analysis, these genes were mostly found to be enriched in pathways relevant to the cell cycle (Fig. 1C), The enriched terms suggest that the analyzed gene set is heavily involved in cell division and proliferation, chromosome and spindle organization, DNA and ATP-related catalytic activities (Fig. 1D).

Fig. 1.

Fig. 1

Open in a new tab

Differential genes between hepatocellular carcinoma (HCC) and normal tissues

Single-cell visualization of HCC samples

A substantial positive connection between nCount_RNA and nFeature_RNA (correlation coefficient of 0.91) is seen in the correlation between different variables in single-cell RNA sequencing data. while the correlations between nCount_RNA and percent.mt and between nFeature_RNA and percent.mt are relatively weak, with coefficients of −0.02 and −0.13, respectively (Supplementary Fig. 1 A). Violin plots depicting the distribution of features such as nFeature_RNA, nCount_RNA, percent.mt, percent.HB, and percent.Ribo across different cell populations (Cluster 1 and Cluster 2), providing a visual reference for quality control in single-cell RNA sequencing data (Supplementary Fig. 1B and 1 C). 24,637 high-quality cells formed 26 clusters representing 15 TME lineages (Fig. 2A, B). Strong correlations were observed between PCs, DCs, and Mac (Fig. 2C). Focusing on these three subpopulations, we performed enrichment analysis based on marker genes for each subpopulation, which further validated the accuracy of cell annotation, with PC, MAC and DC showing the highest cell-cycle scores (Fig. 2D, E). T cells demonstrated increased viral reactivity, T and B cells shown robust immune activation traits, and macrophages were linked to immunoglobulin-mediated immune responses (Fig. 3).

Fig. 2.

Fig. 2

Open in a new tab

(A) Visualization of 24,832 cells using Uniform Manifold Approximation and Projection (UMAP) reveals the integration of datasets to eliminate batch effects. Clusters are divided into (left) (A) 25 subtypes and (right) (B) 15 major cell types. (C) Cell–cell communication between all cell cycle-related subtypes and endothelial cells. (D) Heatmap showing the distribution of cell cycle-related genes in major cell types. (E) Cell cycle-related scoring for all cells

Fig. 3.

Fig. 3

Open in a new tab

Enrichment analysis of the 15 cell types obtained from the annotations, where the line chart on the far left depicts the transition from macrophages to mast cells from top to bottom. Meanwhile, the accompanying heatmap provides insights into the gene expression patterns relative to each cell type, while the section on the far right elucidates the enriched functional pathways in each cell type, including the corresponding highly expressed genes

The landscape of cell cycle-related subtypes in TME cells in HCC

For each major cell type (PC, MAC, DC), NMF was run with ranks k = 2–10. The optimal k was determined by the minimum Cophenetic coefficient drop (≥ 0.05), maximum average silhouette width, and stability assessed over 200 random initializations. These metrics consistently indicated k = 3–7 as the most stable and biologically interpretable range; thus, we reported subtypes within this rank window. We selected cells related to the cell cycle in HCC tissues and constructed a cell cycle-related expression matrix. After quality control, dimensionality reduction, NMF analysis, we identified three new subtypes of PCs (Fig. 4A). These three cell groups were further narrowed by examining the number of genes linked to the cell cycle, including MCM6 + PCs-C1, NOD-PCs-C3, and CCNB1 + PCs-C2 cells (Fig. 4B). Different cell cycle-related PCs subtypes exhibited distinct cell–cell communication with malignant cells, indicating that the cell cycle indeed affects the function and cell–cell communication of PCs (Fig. 4C). To assess the overall effect of clusters of PCs connected to the cell cycle on tumor endothelial cells, we also found a wide range of variations alter the typical levels of expression of cell cycle genes associated with certain functions-associated indicators (Fig. 4D). Significant differences in TF and cell cycle marker expression were found by network regulatory analysis in the cell cycle-related clusters MCM6 + PCs-C1, NOD-PCs-C3, and CCNB1 + PCs-C2.UMAP plot depicting cell cycle subtypes of NMF PC cells including 5 subtypes (Fig. 5A) and 3 major cell types (Fig. 5B). Network regulation analysis showed significant differences in the expression of PC cell markers in the cell cycle-related clusters MCM6 + PCs-C1, NOD-PCs-C3, and CCNB1 + PCs-C2 (Fig. 5C). PCs exhibited close cell–cell communication with HCC cells (Fig. 5C). CCNB1 + PCs-C2 primarily communicates with other cells via MIF and SPP1 signaling pathways, while Hepatocytes engage in communication through TGFβ and GRN signaling pathways. The intensity of MIF and SPP1 signaling received by CCNB1 + PCs-C2 is higher. Hepatocytes exhibit a higher intensity of TGFβ and GRN signaling. MCM6 + PCs-C1 receives GAS and IL10 signaling with higher intensity. Non-PCs-C3 receives TWEAK and CHEMERIN signaling with higher intensity. A greater comprehension of the relationships between cells and their roles in biological processes may be obtained by examining these patterns (Fig. 5D). Remarkably, TFs were significantly upregulated in CCNB1 + PCs-C2 cells (Fig. 5E).

Fig. 4.

Fig. 4

Open in a new tab

(A) Trajectory analysis of the dynamics of cell cycle-related genes. (B) Pseudotime analysis of cell cycle-related NMF clusters. (C) Heatmap of the average expression levels of common signaling pathway genes in three NMF cell subtypes. (D) UMAP plot showing the activity of cell cycle-related NMF PCs subtypes

Fig. 5.

Fig. 5

Open in a new tab

UMAP plot depicting cell cycle subtypes of NMF PC cells. (A) 5 subtypes and (right) (B) 3 major cell types. (C) Cell–cell communication between cell cycle-related TAMs and tumor endothelial cells. (D) Heatmap showing the relative strength of enriched output and input signals in cell cycle-related TAMs and endothelial cells. (E) SCENIC analysis, heatmap of the average expression levels of common signaling pathway genes in three NMF cell subtypes

Cell cycle-related macrophages resembled classical features

Subsequent dimensionality reduction and NMF analysis identified 7 MAC subtypes (Fig. 6A). We analyzed the abundance of cell cycle-related genes and further refined these 4 cell clusters, obtaining TGFB1 + Mac − C1, CDKN1A + Mac − C2, GADD45B + Mac − C3, and Non − Mac − C4 (Fig. 6B). Different cell cycle-related Mac subtypes exhibited distinct cell–cell communication with malignant cells, indicating that the cell cycle indeed affects the function and cell–cell communication of Mac. Pseudotime analysis showed that cell cycle genes play a crucial role in the developmental trajectory of Mac, dominating in the middle and late stages of Mac differentiation (Fig. 6C). We found that fatty acid metabolism is highly active in a variety of cell types, whereas glycosaminoglycan biosynthesis which includes heparan sulfate and heparin is lowly active in a number of cell types. The result allows for a visual comparison of the activity or expression levels of various metabolic pathways in different cell types (Fig. 6D). The number of ligand-receptor connections between cell cycle-related Mac clusters and endothelial cells varied. HCC endothelial cells are the primary source of communication exchanges between these clusters (Fig. 6E). These characteristics were assessed in every macrophage using the AddModuleScore method. Our study’s findings showed that pro-inflammatory macrophages were strongly linked with GADD45B + Mac − C3 and Non − Mac − C4, but anti-inflammatory macrophages were significantly related with TGFB1 + Mac − C1, CDKN1A + Mac − C2, and GADD45B + Mac − C3 (Fig. 6F, G).

Fig. 6.

Fig. 6

Open in a new tab

UMAP plot depicting cell cycle subtypes of NMF macrophages. (A) 7 subtypes and (B). 4 major cell types. (C) Pseudotime analysis of cell cycle-related NMF clusters. (D) Bubble plot showing activated KEGG pathways in cell cycle-related macrophage subtypes. (E) Cell–cell communication between cell cycle-related macrophages and tumor endothelial cells. (F) UMAP plot showing M1 and M2 activity in cell cycle-related NMF TAM subtypes. (G) Assessment of M1 and M2 activity in cell cycle-related macrophages

Cell cycle-related DCs cell phenotypes underscored the antitumor immune response in HCC

Subsequent dimensionality reduction and NMF analysis identified six new DCs subtypes (Fig. 7A). We further refined these six cell groups by examining the number of genes linked to the cell cycle, obtaining CDKN1A + DCs − C1, GADD45B + DCs − C2, PCNA + DCs − C3, Non − DCs − C6, ANAPC11 + DCs − C4, and RB1 + DCs − C5 (Fig. 7B). Different cell cycle-related DCs subtypes exhibited distinct cell–cell communication with malignant cells, indicating that the cell cycle indeed affects the function and cell–cell communication of DCs. Cell cycle genes dominate in the middle and late phases of DC differentiation, according to pseudotime analysis, which demonstrated their critical significance in the developmental trajectory of DCs (Fig. 7C). Six transcription factors linked to the cell cycle were shown in the heatmap (Fig. 7D). To further evaluate the relationship between cell cycle-related co-stimulation and co-inhibition, we also conducted an investigation of DC metabolism (Fig. 7E, F).

Fig. 7.

Fig. 7

Open in a new tab

UMAP plot depicting cell cycle subtypes of NMF DC cells. (A) 6 subtypes and (right) (B) 6 major cell types. (C) Pseudotime analysis of cell cycle-related NMF clusters. (D–F) Heatmap of the average expression levels of common signaling pathway genes in 6 NMF cell subtypes, including co-stimulation and co-inhibition

Cell cycle-related TME patterns contributed the HCC prognosis and immunotherapy

We discovered that tumor samples had much higher cell cycle levels than normal tissues, albeit in varying abundances (Fig. 8A). There were notable differences in overall survival rates across certain subtypes (such PCs, macrophages, and DCs) associated with changes in genes related to the cell cycle (Fig. 8B-G, 9A-F). Tumor tissue-specific immune cell expression was observed (Fig. 10A). Using TCGA and ICGC data, we confirmed these conclusions and found comparable outcomes (Fig. 10B). Furthermore, we found transcriptional indicators in HCC patients to forecast how well they will respond to therapy (Fig. 10C). However, the beneficial effects of current immunotherapies for cancer patients may not be effectively targeted to cell cycle-related HCC patients. Tumor tissues exhibited significantly higher expression of five genes compared to normal tissues (Fig. 11A and B). To verify the expression of the target gene through qPCR, select the gene whose expression needs to be validated, ensuring that it is consistent with the sequencing results (Fig. 11C).

Fig. 8.

Fig. 8

Open in a new tab

(A) Overall survival analysis across different datasets. (B–G) Survival analysis of different cell cycle-related TME cell subtypes

Fig. 9.

Fig. 9

Open in a new tab

Survival analysis of different cell cycle-related TME cell subtypes (A-F)

Fig. 10.

Fig. 10

Open in a new tab

(A) Differential gene expression between normal and tumor tissues. (B) Analysis of ICB efficiency in different cell cycle-related TME cell subtypes. (C) ICB analysis across different datasets

Fig. 11.

Fig. 11

Open in a new tab

Differential genes between tumor and normal tissues, (A). heatmap and (B). bar chart. (C). Verify the expression of the target gene through qPCR

Cell cycle-related TME patterns enhanced the intercellular communication

Different cell cycle-related DCs subtypes exhibited distinct cell–cell communication with malignant cells, indicating that the cell cycle indeed affects the function and cell–cell communication of DCs (Fig. 12A). Network regulation analysis showed significant differences in the expression of DCs markers in the cell cycle-related clusters CDKN1A + DCs − C1, GADD45B + DCs − C2, PCNA + DCs − C3, Non − DCs − C6, ANAPC11 + DCs − C4, and RB1 + DCs − C5 (Fig. 12B). MIF exhibits the highest outgoing signal intensity in the ANAPC11 + DOC-C4 cell type. The ANAPC11 + DOC-C4 cell type receives higher incoming signal intensity for both MIF and GALECTIN, while it receives lower incoming signal intensity for other signaling molecules. This information helps to understand the roles and interrelationships of different cell types in signal transduction (Fig. 12C). The pathways were activated in cell cycle-related DCs cell subtypes (Fig. 12D). Cell cycle-related DC clusters and endothelial cells have varying numbers of ligand-receptor interactions. HCC endothelial cells are primarily responsible for the primary communication exchanges among these clusters (Fig. 12E). It’s interesting to note that differences in cell cycle-related genes within certain cell subtypes, such DCs, macrophages, and PCs, were linked to highly different overall survival rates within these subclusters (Fig. 13A-F).

Fig. 12.

Fig. 12

Open in a new tab

(A) Different cell cycle-related DCs subtypes exhibited distinct cell–cell communication with malignant cells (B)Cell–cell communication models among cell cycle-related DCs subtypes to endothelial cells. (C) Pseudotime analysis of cell cycle-related NMF clusters. (D) Bubble plot showing activated KEGG pathways in cell cycle-related DCs cell subtypes. (E) UMAP plot showing DC activity in cell cycle-related NMF DCs cell subtypes

Fig. 13.

Fig. 13

Open in a new tab

(A-F) Survival analysis of different cell cycle-related TME cell subtypes

Discussion

Cell cycle is an important factor in the progression of HCC. Cell proliferation depends on the control of the cell cycle, and changes in cell cycle regulation are a feature of HCC development [18–21]. One of the main characteristics of HCC development is cell cycle dysregulation. It has been confirmed that blocking cell proliferation by targeting cyclin-dependent kinases (CDKs) is an effective anticancer treatment [19]. Furthermore, the progression of the cell cycle in HCC cells is controlled by Cdk/cyclin complexes, whose activity is tightly regulated by Cdk inhibitors [20–22]. For instance, p27 is highly expressed when hepatocytes are quiescent but decreases as the cell cycle progresses, whereas its homolog p21, in contrast to the inhibitory effects seen in several investigations, is quickly generated in the early G1 phase of hepatocytes. The molecular processes and prognostic aspects of HCC advancement are revealed by single-cell study of the cell cycle, which offers a fresh viewpoint on the intricacy of HCC. This analytical method not only aids in discovering new therapeutic targets but also provides a scientific basis for developing more effective treatment strategies.

To improve medication research and treatment approaches, it is essential to comprehend the precise processes behind the cell cycle in malignancies. Few research have examined the relationship between the cell cycle and the pathophysiology of HCC at the single-cell level, despite the fact that several studies have examined this relationship [23, 24]. We looked closely at cell cycle-related genes in the primary cell types present in the HCC TME for the first time. Additionally, we discovered unique patterns of cell–cell contact between tumor cells and cell cycle-related TME subtypes, with an emphasis on endothelial cells. Our knowledge of how the cell cycle in various TME cell components influences the clinical outcomes of HCC patients is improved by this fresh viewpoint. Invasion, immune evasion, tumor development, metastasis, and responsiveness to therapy are all significantly influenced by tumor endothelial cells [15]. More and more research is highlighting the importance of communication between endothelial cells and other cells within the TME. Prognostic results and patients’ response to therapy are impacted by this variation in cell communication [25]. In our investigation, we found that TME cells exhibited a variety of cell cycle regulation patterns and engaged in significant interactions with tumor endothelium cells. Additionally, the analysis of cell–cell communication proved that ligand-receptor pairings and pathways associated to angiogenesis were activated, elucidating the function of cell cycle-related cell subtypes in angiogenesis. We focused on proliferating cells (PC), dendritic cells (DC) and macrophages (MAC) for two pragmatic and biological reasons. Among the 26 clusters, PC, DC and MAC were the only populations whose aggregate cell-cycle scores showed both (i) the highest absolute expression and (ii) the largest variance across patients (σ2 > 0.35). In contrast, CAFs and most T-cell subsets exhibited uniformly low or intermediate scores (σ2 < 0.12), making downstream NMF sub-clustering unreliable with the current sample size. CAFs are largely quiescent in HCC; their cell-cycle genes are expressed only transiently during activation, yielding < 5% cycling cells—below the threshold for robust subtype discovery. T cells displayed heterogeneous but predominantly post-mitotic signatures (naïve/exhausted states). Thus, excluding CAFs and T cells preserved statistical clarity and focused the study on the three TME compartments where cell-cycle regulation demonstrably drives functional heterogeneity and clinical relevance.

The crucial function of immune cell control and reprogramming in malignancies, especially in TAMs and DCs, has been the subject of increased investigation in recent years [26]. We found a significant feature in our study: tumor endothelial cells and cell cycle-related macrophages and DCs demonstrated considerable cell–cell communication. According to our research, certain subtypes of DCs and macrophages associated with the cell cycle showed a marked activation of metabolic pathways. These metabolic changes lead to the reprogramming of macrophages and DCs [27–29]. Immune cells including dendritic cells (DCs) and macrophages undergo immunore programming in large part due to genes associated with the cell cycle. Within the tumor microenvironment, the metabolic reprogramming of DCs and macrophages is intimately linked to cell cycle regulation, and these mechanisms work together to impact immune evasion and tumor growth. In macrophages, high expression of cell cycle genes is associated with the characteristics of proliferative macrophages, indicating that these cells may be in an active phase of the cell cycle, thereby playing a role in tumor progression. Metabolic reprogramming in macrophages and DCs exerts its effects by influencing immune functions. The Warburg effect in macrophages and DCs, which involves increased glycolysis and decreased mitochondrial respiration, is related to the activation and maturation of immune cells. DCs also undergo metabolic reprogramming during activation, which is associated with metabolic shifts driven by the PI3K/Akt pathway, similar to the Warburg effect observed in tumor cells. This metabolic transition is characterized by a shift from mitochondrial oxidative phosphorylation (OXPHOS) to aerobic glycolysis, marked by an increase in glycolytic metabolism triggered by TLR signaling, which is a major pathway for DC activation. Cell cycle-related genes affect the metabolic reprogramming and immune functions of macrophages and DCs, thereby influencing the progression of liver cancer and immune evasion. These findings provide potential targets for the development of new immunotherapy strategies.

Our data suggest that targeting cell-cycle genes in the HCC TME could improve prognosis, but translating this concept to the clinic faces well-documented hurdles. First-generation pan-CDK inhibitors produced dose-limiting toxicities such as secretory diarrhoea and myelosuppression, partly because they could not discriminate between cancer and normal cycling cells. Second-generation CDK4/6 inhibitors have a narrower spectrum and are now standard-of-care in hormone-receptor-positive breast cancer, yet 15–40% of patients develop primary or acquired resistance through cyclin E–CDK2 up-regulation, PI3K/AKT/mTOR hyper-activation, or RB loss. In HCC specifically, phase II trials of CDK4/6 monotherapy have shown modest activity and similar haematological toxicities. patient selection using the cell-cycle sub-signatures identified here to enrich for RB-proficient tumours, combination regimens (CDK4/6 + checkpoint inhibitors or PI3K inhibitors) to bypass resistance, and next-generation modalities such as PROTAC-mediated CDK degradation or CDK2/4/6 multi-specific inhibitors that are currently in early-phase testing. Rigorous pre-clinical validation in patient-derived organoids and liver-specific toxicity models will be essential before these strategies can enter HCC trials.

Study limitations

The study explicitly noting the absence of functional validation. The study infers that cell cycle states “impact” prognosis but lacks functional experiments (e.g., knockdown of CCNB1 to test communication changes). Only 5 genes were tested in HepG2 cells, future work should therefore incorporate primary HCC cells freshly isolated from surgical specimens and patient-derived organoids (PDOs) to confirm the cell-cycle signature observed in HepG2 cells. Crucially, it lacks functional validation of CCNB1/GADD45B subtypes’ roles in communication and spatial context. Immunotherapy prediction rely solely on computational TIDE analysis without in vitro/in vivo validation. The inferred metabolic shifts in MACs/DCs rest solely on pathway-enrichment data. Future studies incorporating real-time metabolic measurements will be required to validate these observations.

Conclusions

This study investigates the role of cell cycle regulation in intercellular communication within the tumor microenvironment (TME) of hepatocellular carcinoma (HCC) using single-cell RNA sequencing (scRNA-seq). NMF clustering revealed cell cycle-dependent subtypes (e.g., CCNB1 + PCs-C2, GADD45B + Mac-C3) in proliferating cells (PCs), dendritic cells (DCs), and macrophages (MACs), linking them to distinct biological states and communication patterns. Demonstrated that cell cycle dysregulation in TME cells correlates with poor prognosis and immunotherapy efficacy, suggesting cell cycle targeting as a therapeutic strategy. Combined scRNA-seq (CellChat, Monocle, SCENIC) with bulk RNA analysis (TCGA, ICGC) to map cell–cell communication networks and regulatory pathways.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 41522 KB) (40.5MB, docx)

Acknowledgements

The authors declare that no financial support.

Abbreviations

HCC

Hepatocellular carcinoma

NMF

Non-negative matrix factorization

TME

Tumor microenvironment

PC

Proliferating cells

DC

Dendritic cells

MAC

Macrophages

CDK

Cyclin-dependent kinases

Author contributions

CH, SXN, RD: Conceptualization, Validation, Visualization, Writing & editing and participated in the design of the study and CH performed the PCR and statistical analysis. CH, XLM Supervision, editing. All authors reviewed the manuscript.

Funding

No Funding.

Data availability

All data pertinent to this study, whether generated or analyzed, are comprehensively presented in this manuscript.

Declarations

Competing interests

The authors declare no competing interests.

Consent for publication

All the authors agree to publish this paper.

Ethics approval and consent to participate

All procedures conformed to the Helsinki Declaration for the research on humans.

Footnotes

Publisher’s Note

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

Cong Hu and Rui Deng have contributed equally to this article.

Contributor Information

Cong Hu, Email: conghu@whu.edu.cn.

Shuxiong Nong, Email: shuxiongnong@sr.gxmu.edu.cn.

Xinglang Mou, Email: djxzyymxl@163.com.

References

  • 1.Kamal MA, Mandour YM, Abd El-Aziz MK, Stein U, El Tayebi HM. Small molecule inhibitors for hepatocellular carcinoma: advances and challenges. Molecules. 2022;27(17):5537. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Wang X, Cao J, Li J. Anatomic liver resection based on portal territory with margin priority for hepatocellular carcinoma. JAMA Surg. 2024;159(6):710–1. [DOI] [PubMed] [Google Scholar]
  • 3.Vogel A, Grant RC, Meyer T, Sapisochin G, O’Kane GM, Saborowski A. Adjuvant and neoadjuvant therapies for hepatocellular carcinoma. Hepatology. 2023. 10.1097/HEP.0000000000000726. [DOI] [PubMed] [Google Scholar]
  • 4.Sun R, Moraleda JM, Wei LJ. Quantification of treatment effect of tislelizumab vssorafenib for hepatocellular carcinoma. JAMA Oncol. 2024;10(5):674. [DOI] [PubMed] [Google Scholar]
  • 5.Huang Q, Zhong X, Li J, Hu R, Yi J, Sun J, et al. Exosomal ncRNAs: multifunctional contributors to the immunosuppressive tumor microenvironment of hepatocellular carcinoma. BioMed Pharmacother. 2024;173:116409. [DOI] [PubMed] [Google Scholar]
  • 6.Tang M, Zhang S, Yang M, Feng R, Lin J, Chen X, et al. Infiltrative vessel co-optive growth pattern induced by IQGAP3 overexpression promotes microvascular invasion in hepatocellular carcinoma. Clin Cancer Res. 2024;30(10):2206–24. [DOI] [PubMed] [Google Scholar]
  • 7.Shi K, Peng X, Xu T, et al. Precise electromagnetic modulation of the cell cycle and its applications in cancer therapy. Int J Mol Sci. 2025;26(9):4445. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Glaviano A, Singh SK, Lee EHC, et al. Cell cycle dysregulation in cancer. Pharmacol Rev. 2025;77(2):100030. [DOI] [PubMed] [Google Scholar]
  • 9.Chen D, Lu S, Huang K, et al. Cell cycle duration determines oncogenic transformation capacity. Nature. 2025;641(8065):1309–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Park JG, Roh PR, Kang MW, Cho SW, Hwangbo S, Jung HD, et al. Intrahepatic immunoglobulin a complex induces polarization of cancer-associated fibroblasts to matrix phenotypes in the tumor microenvironment of hepatocellular carcinoma. Hepatology. 2024;80(5):1074–86. [DOI] [PubMed] [Google Scholar]
  • 11.Jiang S, Lu H, Pan Y, Yang A, Aikemu A, Li H, et al. Characterization of the distinct immune microenvironments between hepatocellular carcinoma and intrahepatic cholangiocarcinoma. Cancer Lett. 2024;588:216799. [DOI] [PubMed] [Google Scholar]
  • 12.Dou T, Li J, Zhang Y, Pei W, Zhang B, Wang B, et al. The cellular composition of the tumor microenvironment is an important marker for predicting therapeutic efficacy in breast cancer. Front Immunol. 2024;15:1368687. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Ansell SM, Vonderheide RH. Cellular composition of the tumor microenvironment. Am Soc Clin Oncol Educ Book. 2013; 10.1200/EdBook_AM.2013.33.e91. [DOI] [PubMed]
  • 14.Sadeghi M, Dehnavi S, Sharifat M, Amiri AM, Khodadadi A. Innate immune cells: key players of orchestra in modulating tumor microenvironment (TME). Heliyon. 2024;10:e27480. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Leone P, Malerba E, Susca N, Favoino E, Perosa F, Brunori G, et al. Endothelial cells in tumor microenvironment: insights and perspectives. Front Immunol. 2024;15:1367875. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Chen M, Zhang Q, Zheng S, Guo X, Cao L, Ren Y, et al. Cancer-associated fibroblasts promote migration and invasion of non-small cell lung cancer cells via METTL3-mediated RAC3 m(6)a modification. Int J Biol Sci. 2023;19:1616–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Kim DK, Jeong J, Lee DS, Hyeon DY, Park GW, Jeon S, et al. PD-L1-directed PlGF/VEGF blockade synergizes with chemotherapy by targeting CD141(+) cancerassociated fibroblasts in pancreatic cancer. Nat Commun. 2022;13:6292. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Panagopoulos A, Stout M, Kilic S, et al. Multigenerational cell tracking of DNA replication and heritable DNA damage. Nature. 2025;642(8068):785–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Song CH, Lin CW, Han KH. Cell cycle-based antibody selection for suppressing cancer cell growth. FASEB J. 2025;39(4):e70402. [DOI] [PubMed] [Google Scholar]
  • 20.Kumarasamy V, Wang J, Roti M, et al. Discrete vulnerability to pharmacological CDK2 inhibition is governed by heterogeneity of the cancer cell cycle. Nat Commun. 2025;16(1):1476. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Gao X, Wang Y, Li Y, et al. Huganpian, a traditional chinese medicine, inhibits liver cancer growth in vitro and in vivo by inducing autophagy and cell cycle arrest. Biomed Pharmacother. 2019;120:109469. [DOI] [PubMed] [Google Scholar]
  • 22.Xiang X, Deng L, Xiong R, et al. Tex10 is upregulated and promotes cancer stem cell properties and chemoresistance in hepatocellular carcinoma. Cell Cycle. 2018;17(11):1310–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Wang J, Su Q, Chen K, Wu Q, Ren J, Tang W, et al. Pyrimethamine upregulates BNIP3 to interfere SNARE-mediated autophagosome-lysosomal fusion in hepatocellular carcinoma. J Pharm Anal. 2024;14:211–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Luo P, An Y, He J, Xing X, Zhang Q, Liu X, et al. Icaritin with autophagy/mitophagy inhibitors synergistically enhances anticancer efficacy and apoptotic effects through PINK1/Parkin-mediated mitophagy in hepatocellular carcinoma. Cancer Lett. 2024;587:216621. [DOI] [PubMed] [Google Scholar]
  • 25.Santi A, Kay EJ, Neilson LJ, McGarry L, Lilla S, Mullin M, et al. Cancerassociated fibroblasts produce matrix-bound vesicles that influence endothelial cell function. Sci Signal. 2024;17:eade0580. [DOI] [PubMed] [Google Scholar]
  • 26.Li SL, Hou HY, Chu X, Zhu YY, Zhang YJ, Duan MD, et al. Nanomaterials involved tumor-associated macrophages’ reprogramming for antitumor therapy. ACS Nano. 2024;18(11):7769–95. [DOI] [PubMed] [Google Scholar]
  • 27.Ammarah U, Pereira-Nunes A, Delfini M, Mazzone M. From monocyte-derived macrophages to resident macrophages-how metabolism leads their way in cancer. Mol Oncol. 2024;18(7):1739–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Geiß C, Salas E, Guevara-Coto J, Régnier-Vigouroux A, Mora-Rodrıǵ uez RA. Multistability in macrophage activation pathways and metabolic implications. Cells. 2022;11(3):404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Ye L, Jiang Y, Zhang M. Crosstalk between glucose metabolism, lactate production and immune response modulation. Cytokine Growth Factor Rev. 2022;68:81–92. [DOI] [PubMed] [Google Scholar]
  • 30.Hou W, Ji Z. Assessing GPT-4 for cell type annotation in single-cell RNA-seq analysis. Nat Methods. 2024;21:1462–5. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials

Supplementary file1 (DOCX 41522 KB) (40.5MB, docx)

Data Availability Statement

All data pertinent to this study, whether generated or analyzed, are comprehensively presented in this manuscript.

Articles from Clinical and Experimental Medicine are provided here courtesy of Springer

RESOURCES