A total of six pathologically diagnosed HCC tissues and matched tumor-adjacent tissues were obtained from patients at the Third Affiliated Hospital of Sun Yat-sen University. The use of clinical samples was approved by the ethics committee of the Third Affiliated Hospital of Sun Yat-sen University, and written informed consents were obtained from all enrolled patients. All patients did not receive preoperative therapies.

HCC cell lines (Hep3b, Huh7) were purchased from the Cell Bank of Type Culture Collection of Chinese Academy of Sciences (Shanghai, China). All cells were maintained in a six-well plate in RPMI-1640 medium (Corning, 10-040) supplemented with 10% fetal bovine serum (FBS) (Hyclone, SH30084), 100 μg/mL streptomycin/penicillin (Hyclone, SV30010) in a humidified 37°C incubator with 5% CO₂. The PGLS siRNA (siPGLS) was designed and obtained from Gene Pharma (Guangzhou, China). The transfection assay was carried out using Lipofectamine 2000 (Invitrogen, 11668030) following the protocols. The culture media was supplemented with 6-aminonicotinamide (6-ANA) (10 μM, Target-mol, T7545) 48 h.

Total RNA was extracted from HCC cell lines or clinical samples using a TRIZOL reagent (Magen, R4801-02) according to the manufacturer’s instructions. Quantitative real-time polymerase chain reaction (PCR) was performed using SYBR Green PCR kit protocol (Bio-Rad, 1725150). PCR primer sequences are listed in Supplementary Table 1.

The same number of cells from each population to be analyzed was sorted into phosphate-buffered saline (PBS) with 2% FBS. The cells were washed with PBS and lysed by RIPA. Equal amounts of protein extracts were fractionated by 12.5% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred to a polyvinylidene fluoride membrane (IPVH00010, Merck Millipore). After blocking with 5% non-fat milk in Tris-buffered saline with Tween-20 (TBST, pH 7.6) for 1 h at room temperature, the membranes were incubated with primary antibodies including anti-PGLS (rabbit, 1:1,000, GTX120327, Genetex), anti-CK18 (rabbit, 1:1,000, 10830-1-AP, Proteintech), and anti–β-actin (rabbit, 1:1,000, 4970s, Cell Signaling Technology) overnight at 4°C and then incubated with secondary antibodies (rabbit, 1:10,000, W401B, Promega) for 1 h at room temperature, which was detected by digital imaging with a charge-coupled device camera system (Odyssey Fc). The images shown are representative of images from at least three experiments.

For apoptosis, the cells were fixed, permeabilized, and stained by Tunel Detection kit (C1086, Beyotime) according to manufacturer’s instructions. For reactive oxygen species (ROS) activity analysis, the cells were stained by 5 μM DCFDA (D6883, Sigma). Cell sorting and analysis were performed using an Attune NxT analyzer (Thermo Fisher Scientific) or InFlux Cell Sorter (BD Biosciences). Data analysis was performed using FlowJo software.

PGLS^high and PGLS^low cells were sorted and then lysed; intracellular NADP⁺/NADPH ratio was measured using the NADP⁺/NADPH Assay Kit (KA1663, Abnova) according to the manufacturer’s instructions.

HCC in vitro proliferation was measured by calcein-AM/PI kit (C2015S, Beyotime) according to the manufacturer’s instructions.

Hep3b and Huh7 cells were added into the upper chambers of Matrigel-uncoated (cell migration) or coated (cell invasion) Transwells (ET BIOFIL, Guangzhou, TCS004024). The lower chambers were added medium with 10% FBS, and the upper chambers were serum-free medium. After 24 h culture, the migrated or invaded cells (on the bottom of the filters) were fixed using 4% paraformaldehyde (Mei Lun, China, MA0192) and stained with 0.5% crystal violet for 1 h. The number of migrated or invaded HCCs was counted under a light microscope by randomly selecting five fields.

A total of 5 × 10⁴ Hep3b and Huh7 cells were plated into six-well plates. Colonies were fixed with 4% paraformaldehyde and stained with crystal violet (Beyotime, Shanghai, C0121-100ML for 30 min at room temperature. The visible colonies were counted manually.

A total of 5 × 10⁴ Hep3b and Huh7 cells were seeded into six-well plates and grown to 80% cell abundance. Then, a single layer wound was created using a pipette tip, and we took images (Olympus, BX51). Imaging was repeated at the same location and further analyzed by ImageJ software. All assays were conducted three times.

Raw genomic data have been deposited in the Gene Expression Omnibus database with accession number GSE149614. The scRNA-seq data are available from the corresponding author upon reasonable request; 28,687 non-tumor liver cells and 34,414 primary tumor cells from 10 HCC patients were included. Normalization, dimensionality reduction, and clustering were performed with the Seurat 3.2.3 R package (Butler et al., 2018) on RStudio. Cells were filtered to have > 500 and < 5,000 detected genes and < 5% of total UMIs mapping to the mitochondrial genome. Data set normalization was performed by dividing the UMI counts per genes by the total UMI counts in the corresponding cells and log-transforming, and following the results, scaling and centering. Cells underwent dimensionality reduction with the uniform manifold approximation and projection method (UMAP). HCC-like clusters were selected by HCC markers including GPC3, CD24, and MDK (Tsuchiya et al., 2015; Lu et al., 2018; Yu et al., 2018). Feature plots were generated by the Seurat function feature plot. Pseudotime trajectory was analyzed by monocle2 on basis of the Seurat clustering (Subramanian et al., 2005; Qiu et al., 2017). Signature genes of each cluster were obtained using the Seurat function FindMarkers with “wilcox” test. Venn plots were generated by Venn Diagram R packages. Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis and plots were performed using cluster Profiler and ggplot2 R package. Gene lists were preranked by the fold change values of the differential expression analysis using Seurat. Gene sets were obtained from Gene Ontology database as indicated. Heatmap was generated by the pheatmap R package.

Data are expressed as means ± standard deviation (SD). All experiments were analyzed by Student t-test, and differences were considered statistically significant if p < 0.05. Differences were considered statistically significant if p < 0.05, ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001.

From the Gene Expression Omnibus database, we downloaded scRNA-seq data of non-tumor and HCC patient liver cells. In total, 28 clusters were shown after UMAP dimensionality reduction (Figures 1A,B). We found that clusters 8, 13, 16, and 17 were specifically presented in patient samples, labeled by HCC markers (GPC3, CD24, and MDK) (Figures 1C,D). Then we extracted these four clusters for pseudotime trajectory analysis, which showed that cluster 0 was the most primitive (Figures 1E,F). Venn plot presented the overlapped marker genes among the new four clusters (Figure 1G). The specifically high expression genes, in cluster 0, were used for KEGG enrichment analysis. It showed that several top pathways were associated with carbon metabolism (Figure 1H). Next, we found the expression of PGLS in carbon metabolism was the highest (Figure 1I). Consistent with this, the UMAP plot and violin plot showed specific high expression level of PGLS in new cluster 0 (Figures 1J,K). Similar rising level could be observed in human HCC samples from the results of IHC staining and The Cancer Genome Atlas (TCGA) (Figures 1L–N). Compared with the survival probability of the high PGLS expression group, the low PGLS expression group showed a longer survival period (Figure 1O). PGLS transcripts were profoundly higher in human HCC samples by quantitative PCR. Similar rising levels could be observed in the HCC markers GPC3, CD24, and MDK (Figure 1P).

HCC cells from human liver cancer tissues were sorted by the expression level of PGLS for further exploring the difference between these two groups. PGLS was also highly expressed in PGLS^high cells (Figure 2A), with a higher expression level of CK18, which was used for HCC diagnosis in clinic (Figures 2B,C). PGLS^high cells presented a higher transcription level of GPC3, CD24, and MDK (Figure 2D). It also showed a lower NADP⁺/NADPH ratio, which was associated with the reduction of ROS production and apoptosis (Figures 2E–I), we next used scRNA-seq data to investigate the role of PGLS in HCC; we found that the expression of apoptosis-related genes was negative correlated with the expression of PGLS (Figure 2J). Furthermore, this kind of apoptosis occurs because of the activation of ROS relative signaling pathways (metabolism of xenobiotics by cytochrome P450, chemical carcinogenesis–ROS) (Figure 2K).

To investigate the effects of the rate-limiting enzymes of PGLS in HCC cell lines, Huh7 and Hep3b cells were transfected with siRNA-1, siRNA-2, or siRNA-3. We found that all three siRNAs could reduce the expression level of PGLS in Huh7 cell lines as compared to control groups (Figures 3A,B). Similar results could be observed in Hep3b cell lines (Supplementary Figures 1A,B). In addition, we found that the treatment of siRNA-1, siRNA-2, and siRNA-3 increased the NADP⁺/NADPH ratio in HCC cell lines (Figure 3C and Supplementary Figure 1C). Meanwhile, ROS and apoptosis rate in the three knockdown (KD) groups were also significantly higher as compared to the control group (Figures 3D,E and Supplementary Figures 1D,E). After transfection by siRNA-1, siRNA-2, and siRNA-3, both HCC cell lines presented a slower proliferation rate at all six time points (Figures 3F,G and Supplementary Figures 1F,G). Consistent with this, Huh7 and Hep3b cells showed a significantly smaller size and fewer numbers of CFUs (colony-forming units) as compared to the control group (Figures 3H,I and Supplementary Figures 1H,I).

In the migration assay, the two cell lines presented significantly fewer clones per field in the three KD groups as compared to the control group (Figures 4A,B and Supplementary Figures 2A,B). To analyze the effects of siRNA-1, siRNA-2, and siRNA-3 on HCC cell invasion, chamber invasion assay was performed on Hep3b and Huh7 cells. Significantly fewer cell clones were shown in the three siRNA groups, in both Hep3b and Huh7 cells (Figures 4C,D and Supplementary Figures 2C,D).

To further confirm that PGLS could regulate the migration of HCC. Wound-healing assay was performed, and the area covered by treated Huh7 cells was calculated by ImageJ. Nearly the same area was occupied at 0 h among four groups, and less migration length was shown in all three KD groups at 24 h as compared to the control group at the same time point (Figures 4E,F). Similar inhibitory effects were also observed in Hep3b cells (Supplementary Figures 2E,F).

6-ANA is a PPP inhibitor (Street et al., 1997; Arbe et al., 2020; Cheng et al., 2020). When Hep3b and Huh7 were cultured with a 10 μM 6-ANA concentration, we found that the NADP⁺/NADPH ratios increased significantly (Figure 5A and Supplementary Figure 3A), and the level of ROS and apoptosis rate were also increased in both HCC cell lines as compared to the control group (Figures 5B,C and Supplementary Figures 3B,C). Then we analyzed the effect of 6-ANA on Hep3b and Huh7 cell on proliferation. Similar to previous results, 6-ANA significantly suppressed cell proliferation in 10 μM concentration (Figure 5D and Supplementary Figure 3D). Subsequently the CFU results showed 6-ANA could also effectively reduce the size and number of CFUs in the treated group (Figures 5E,F and Supplementary Figures 3E,F). To test the effects of 6-ANA on HCC cell line migration, we did both chamber assay and wound-healing assay. The results showed that 6-ANA could reduce the clone number and migration area of Hep3b and Huh7 cells (Figures 5G,H,K,L and Supplementary Figures 3G,H,K,L). The number of invasive cells was also markedly decreased after the treatment of 6-ANA (Figures 5I,J and Supplementary Figures 3I,J). Overall, our data showed that PGLS was essential for the development of HCC.