Skip to main content
Download PDF

Endogenous protein S100A14 stabilizes glutaminase to render hepatocellular carcinoma resistant to sorafenib

Journal of Translational Medicine volume 23, Article number: 435 (2025) Cite this article

Abstract

Background

Many cases of advanced hepatocellular carcinoma (HCC) are resistant to the widely used drug sorafenib, which worsens prognosis. While many studies have explored how acquired resistance emerges during drug exposure, the mechanism underlying primary resistance before treatment still remain elusive.

Methods

Single-cell lineage tracing and RNA sequencing were performed to identify primary sorafenib-resistant lineages in HCC. Differential gene expression analysis was employed to identify the biomarkers of drug-resistant lineage cells. Cell viability and colony formation assays were adopted to assess the involvement of S100A14 on sorafenib resistance. Co-immunoprecipitation (CO-IP) and mass spectrometry analysis were conducted to uncover the downstream targets and regulatory mechanisms of S100A14 in primary resistance to sorafenib. In vivo mouse xenograft experiments were carried out to assess the effect of S100A14 or its interacting protein glutaminase (GLS) on primary resistance to sorafenib in HCC.

Results

Single-cell lineage tracing identified a cluster of sorafenib primary resistant cells, and S100A14, a Ca2+-binding protein, was determined to be a critical biomarker for primary resistance to sorafenib. Knockdown of S100A14 significantly increases sorafenib treatment sensitivity in HCC cells. Mechanistically, S100A14 binds to GLS and blocks its phosphorylation at residues Y308 and S314, which in turn inhibits its ubiquitination and subsequent degradation. By stabilizing GLS, S100A14 reduces oxidative stress in HCC cells, thereby antagonizing sorafenib-induced apoptosis. Inhibiting S100A14 or GLS significantly improved sorafenib efficacy against xenograft tumors in vivo.

Conclusions

Our results demonstrate that S100A14 plays a pivotal role in promoting primary resistance to sorafenib by stabilizing GLS to reduce oxidative stress, and acts as a potential therapeutic target to enhance the efficacy of sorafenib in HCC patients.

Graphical Abstract

Introduction

Hepatocellular carcinoma (HCC), which accounts for 75–85% of primary liver cancers, affects a growing number of individuals worldwide and causes an increasing number of premature deaths every year [1]. The pathophysiology of hepatocarcinogenesis is closely tied to the progression of cirrhosis and chronic liver disease [2]. Most patients with HCC are diagnosed at an advanced stage where surgical treatment is no longer feasible, and are mainly treated through chemotherapy, targeted therapy, and immunotherapy [3, 4]. The targeted therapy drug sorafenib, a tyrosine kinase inhibitor, is a well-established first-line treatment for HCC [5]. However, only about 30% of patients respond to sorafenib, and even those usually develop resistance to it within 6 months [6]. Therefore, investigating the mechanisms of sorafenib resistance is of great significance for improving the prognosis for HCC patients.

Currently, studies of sorafenib resistance primarily focus on the mechanisms underlying acquired resistance, such as alterations in signaling pathways, autophagy, drug transport, cancer stem cells, tumor microenvironment, and non-coding RNAs [7,8,9]. While these alterations explain some cases of resistance, other cases may arise when the drug selects for pre-existing lineages of tumor cells that are already resistant to the drug (referred to as primary resistance), leading to their expansion [10]. Targeting sorafenib-resistant subpopulations could therefore greatly overcome sorafenib resistance, leading to more effective treatments in HCC. However, the identification and mechanistic exploration of cell lineages with primary resistance to sorafenib are still elusive. Recently, emerging single-cell lineage tracing technology [11, 12], combined with single-cell RNA sequencing (scRNA-seq) [13, 14], has already proven effective in identifying and transcriptionally characterizing subpopulations of cancer cells differing in their ability to regenerate [15], to metastasize [16], as well as to resist to therapeutic drugs [17]. Here, we combined single-cell lineage tracing with scRNA-seq to identify a S100A14-expressing subpopulation within primary HCC tumors that exhibits resistance to sorafenib, and investigated the mechanisms by which S100A14 influences primary resistance to sorafenib.

Methods

Cell culture

Huh7 (#TCHu182) cells were purchased from National Collection of Authenticated Cell Cultures. Hep3B (#HB-8064) and HEK293T (#CRL-3216) cells were purchased from American Type Culture Collection. Cells were cultured in DMEM (#11995065, Gibco) supplemented with 10% fetal bovine serum (#A5256701, Gibco), penicillin (100 units/mL) and streptomycin (100 µg/mL) at 37 °C in an incubator with 5% CO2. Mycoplasma infection was verified by the specific PCR assay using MycoBlue Mycoplasma Detector Kit (#D101-01, Vazyme).

Chemicals

Sorafenib (#HY-10201), MG132 (#HY-13259), and Chloroquine (#HY-17589A) were purchased from MCE (Shanghai, China). Cycloheximide (#S7418) and CB-839 (#S7655) were purchased from Selleck (Shanghai, China).

Plasmids

Barcode plasmid library was constructed by GENEWIZ (Suzhou, China). In briefly, a static 14-base pair-randomer barcode was cloned into the Sal I site of the plasmid pCDH-CMV-MCS-EF1-CopGFP-T2A-Puro (#BR541, Fenghui Biotechnology). The promoters of ANGPTL3 (-994/+109), FXYD3 (-1570/+123), S100A4 (-956/+529), and S100A14 (-372/+34) relative to the + 1 transcription start site (TSS) were amplified by PCR using genomic DNA from Huh7 cells, which were cloned into the BsaA I/BamH I sites of the plasmid pCDH-CMV-T2A-EGFP-EF1A-Puro (#BR544, Fenghui Biotechnology) to replace the CMV promoters. The shRNAs of S100A14 or GLSKGA were inserted into the plasmid pLKO.1-TRC (#10878, Addgene). To construct overexpression plasmids with Flag and HA tags, the full-length coding DNA sequence (CDS) of S100A14 or GLSKGA were amplified by PCR using 1st Strand cDNA from Huh7 cells, and then they were cloned into the BamH I/Not I sites of the plasmid pCDH-CMV-T2A-EGFP-EF1A-Puro by a ClonExpress® Ultra One Step Cloning Kit (#C115, Vazyme). The Flag or HA tags were inserted into the overexpression plasmids by primer design. Different human GLSKGA point mutants were generated using a ClonExpress® Ultra One Step Cloning Kit by primer design. All constructs were verified by DNA sequencing (Genewiz, China). All shRNA sequences and primer sequences are listed in Supplementary Table 1.

Lentiviral infections

The plasmids constructed above were transfected into HEK293T cells together with pSPAX2 (#12260, Addgene) and pMD2.G (#12259, Addgene) plasmids. The cell supernatants were collected after 48 h and were filtered using a 0.45 μm sterile filter (#SLHVR33RB, Millipore), then concentrated using a 100 kDa ultrafiltration centrifugal tube (#UFC810096, Millipore) to obtain the viruses. The viruses were added to Huh7 or Hep3B cells with 10 µg/ml polybrene. Puromycin were used to select stably transfected cells.

Sample preparation, library construction, data analysis for scRNA-seq and barcode sequencing

Single-cell suspension preparation was performed as previously described [18]. Briefly, the drug-naive tumors and sorafenib-treated tumors from NOD/SCID mice were minced on a plate and enzymatically digested with collagenase IV (#C5138, Sigma) and DNase I (#MB3069, Meilunbio) for 30 min at 37℃ with agitation. After digestion, samples were sieved through a 70 μm cell strainer. After staining with Propidium Iodide (PI), samples were sorted by BD FACSAria cell sorter (BD Biosciences) to obtain CopGFP-positive and PI-negative cells. Single-cell suspensions with phosphate-buffered saline (PBS) were loaded onto microwell chip using the Singleron Matrix® Single Cell Processing System. The barcode libraries and scRNA-seq libraries were constructed according to the protocol of the GEXSCOPE® Single Cell RNA Library Kits (Singleron) [19].

To obtain the barcode sequence of a single cell, we filtered the raw reads from the barcode library by retaining only those that contained the target sequence “AATCCAGCTAGCTGTGCAGC”, which served as a prefix for the intended barcode. Further analysis was conducted on the lineage barcodes that were annotated with Unique Molecular Identifiers (UMIs) using the R programming language. For a given cell barcode, the barcode sequence that was present in at least 95% of the reads was considered to be the barcode of this single cell.

The raw reads obtained from scRNA-seq were processed through CeleScope v1.5.2 pipeline to generate gene expression matrixes. Cells were filtered based on the following criteria: UMI counts more than 30,000, gene counts less than 200 or more than 5,000, and mitochondrial content more than 20%. We used functions from Seurat v3.1.2 [20] for dimension-reduction and clustering. Initially, all gene expressions were normalized and scaled using the NormalizeData and ScaleData functions. Subsequently, the top 2000 variable genes were identified with the FindVariableFeatures function for Principal Component Analysis (PCA). Utilizing the top 20 principal components, cells were segregated into multiple clusters by FindClusters function. Lastly, Uniform Manifold Approximation and Projection (UMAP) algorithm was applied to visualize the cells in a two-dimensional space.

To identify differentially expressed genes (DEGs), we employed the Seurat FindMarkers function, utilizing the Wilcox likelihood-ratio test with default parameters. Genes that were expressed in more than 10% of the cells within a cluster and exhibited an average log2(Fold Change) value exceeding 0.25 were selected as DEGs. The expression of markers for each cluster was visualized using heatmaps, dot plots, and violin plots generated through the Seurat functions DoHeatmap, DotPlot, and VlnPlot.

For the survival analysis of cluster 1 in drug-naive tumors, the top 20 genes in cluster 1 were collected in a target gene set and the clinical data for HCC were obtained from TCGA database. To evaluate the relationship between the target gene set and clinical factors, ssGSEA algorithm was used to calculate a score of the target gene set for each HCC sample [21]. The surv_cutpoint function in the survminer package was utilized to analyze the best cut-off values, and then HCC samples were assigned into two groups (High group and Low group) using the best cut-off. Differences in overall survival between high and low groups were compared using Kaplan-Meier curves, with p-values calculated via log-rank test, using the Survival package in R.

Isolation of high and low expressing cell populations of marker genes via flow cytometry

Flow cytometry sorting of subpopulations of cells with high and low marker gene expression was performed as described previously [22,23,24]. Briefly, Huh7 cells infected with plasmids expressing EGFP under the promoters of the indicated genes (ANGPTL3, FXYD3, S100A4, S100A14) were harvested using trypsin digestion, centrifuged and resuspended in PBS. Samples were analyzed and sorted on a BD FACSAria (BD Biosciences). For the positive and negative populations, the top 5% cells with high EGFP expression (EGFPHigh) and the 5% cells with the weakest intensity of EGFP (EGFPLow) were sorted, respectively. The EGFPHigh and EGFPLow cells were evaluated for purity with CytoFLEX flow cytometer (Beckman Coulter).

RNA extraction and quantitative PCR analysis

Total RNA was extracted using FastPure Cell/Tissue Total RNA Isolation Kit (#RC101, Vazyme). The mRNA was reverse transcribed into cDNA by HiScript III RT SuperMix for qPCR (#R323, Vazyme), according to manufacturer’s protocol. 20 ng of total cDNA were subjected to real-time quantitative polymerase chain reaction (qPCR) using ChamQ Universal SYBR qPCR Master Mix (#Q711, Vazyme). The Real-Time PCR amplification was conducted following the manufacturer’s protocol on QuantStudio 3 Real-Time PCR System (Applied Biosystems). GAPDH mRNA was used as a loading control to normalize mRNA expression levels. The threshold cycle (CT) values were obtained at the completion of PCR. The relative transcription levels of target genes, normalized to GAPDH, were calculated by the comparative 2−ΔΔCT. Primer sequences are listed in Supplementary Table 1.

Cell viability and apoptosis assay

Stably transfected cells were seeded at specific densities in 96-well cell culture plates and added with varying concentrations of sorafenib. After 24 h of drug exposure, cell viability was assessed using the CCK-8 Cell Counting Kit (#C0037, Beyotime) according to the manufacturer’s instructions, and the absorbance was detected at 450 nm using microplate reader (Multiskan FC, Thermo Fisher Scientific) after incubation for 2 h at 37℃. Cell apoptosis was detected using the Annexin V-PE/7-AAD Apoptosis Detection Kit (#A213, Vazyme) according to the manufacturer’s instructions. The samples were analyzed using a CytoFLEX flow cytometer (Beckman Coulter) and the apoptosis ratio was evaluated by FlowJo (v10) software.

Reactive oxygen species (ROS) and Glutathione (GSH) measurement

ROS generation was analyzed in the Huh7 and Hep3B cell lines after treatment with 10 µM sorafenib for 12 h. Intracellular ROS levels were measured using a DCFH-DA probe (#S0033S, Beyotime) via flow cytometry. After sorafenib treatment, cells were incubated with 10 µM DCFH-DA for 20 min at 37℃ in the dark condition, then cells were harvested and washed with PBS. Flow cytometry analyses were performed using a CytoFLEX flow cytometer (Beckman Coulter). Data analysis was conducted by FlowJo software. GSH levels were measured using a GSH and GSSG Assay Kit (#S0053, Beyotime) according to the manufacturer’s instructions, and the absorbance was detected at 405 nm using microplate reader.

Survival clonogenic assay

Stably transfected cells were seeded in 6-well cell culture plates at consistent densities. After adherence, cells were exposed to sorafenib at 15 µM. After 36 h incubation, the cell medium was changed with fresh sorafenib-free complete medium, and then cells were cultured for 10–14 days. Next, cells were fixed with 4% paraformaldehyde for 30 min and stained with crystal violet (#C0121, Beyotime) for 20 min. After dried the plates, colonies were photographed and were quantified using ImageJ (v1.54f) software.

Co-immunoprecipitation (CO-IP) assay

For immunoprecipitation of endogenous proteins: Huh7 or Hep3B cells were lysed in IP lysis buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1% Nonidet P-40) supplemented with protease inhibitor and incubated with related antibodies overnight at 4 °C, followed by the addition of protein A/G magnetic beads (#P2108, Beyotime) for another 4 h. Then the magnetic beads were washed four times by TBST buffer (20 mM Tris, 137 mM NaCl, 0.1% Tween-20, pH 7.4), resuspended and boiled at 100 °C for 5 min in 80 µL of SDS sample buffer (#P0015, Beyotime) to obtain immunoprecipitates. Immunoprecipitates were analyzed by western blotting.

For immunoprecipitation of overexpression of exogenous HA-tagged proteins: Huh7 cell lysates overexpressing exogenous HA-tagged proteins were incubated with anti-HA magnetic beads (#P2121, Beyotime) or negative control mouse IgG magnetic beads (#P2171, Beyotime) overnight at 4 °C. The next day, the magnetic beads were washed four times by TBST buffer and eluted using 80 µL of SDS sample buffer. Interacting proteins were visualized using western blot.

Ubiquitination assay

To evaluate the effect of S100A14 on ubiquitination levels of GLSKGA, Huh7 cells were transfected with LKO-shS100A14 or LKO-NTC (non-target control) plasmids. To assess whether the GLSKGA amino acids Y308 and S314 play a role in the ubiquitination of GLSKGA influenced by S100A14, we ectopically expressed the HA-tagged GLSKGA wild type or mutant constructs in Huh7 cells and then knocked down S100A14 for 72 h. The treated cells were supplemented with 20 µM MG132 for 8 h before harvest. Following this, they were lysed using IP lysis buffer for 30 min, and co-immunoprecipitation was then performed as described above. The degree of ubiquitination was assessed by immunoblotting with an anti-ubiquitin antibody.

Immunofluorescence staining and TUNEL assay

Huh7 or Hep3B cells were stained with MitoTracker Red CMXRos (#A66443, Thermo Fisher Scientific) according to the protocol. Then, the cells were fixed with 4% paraformaldehyde for 20 min and permeabilized with 0.25% Triton X-100 for 20 min at room temperature. After blocking with 5% BSA, the cells were sequentially incubated with the indicated primary antibodies, secondary antibody conjugated with a fluorescent dye, and stained with Hoechst 33,342 (#C1025, Beyotime). The antibodies utilized are listed in Supplementary Table 2. TUNEL (TdT mediated dUTP Nick End Labeling) Apoptosis Detection Assay Kit (#40306ES20, Yeasen) was used to detect cell apoptosis according to the manufacturer’s instructions. The fluorescent signals were captured using confocal microscopy (Zeiss LSM 980, Germany).

Western blot analysis

Western blotting was performed as previously described [25]. Briefly, the cells were lysed in cell lysis buffer (#P0013, Beyotime) supplemented with protease inhibitor. The supernatant was loaded in SDS-PAGE gels to separate proteins. The proteins were transferred to 0.45 μm or 0.2 μm PVDF membranes (#IPFL00010 or #ISEQ00010, Millipore). After blocking with 5% non-fat powdered milk (#A600669, Sangon Biotech) in TBST, the membranes were incubated with indicated primary antibodies at 4℃ overnight. The next day, the membranes were incubated with indicated HRP conjugated secondary antibodies for 1 h at room temperature. After washed with TBST three times, the protein bands were then visualized by chemiluminescence (#BL520B, Biosharp) and imaged using imaging system (Tanon 5200 Multi, Tanon). ImageJ software was utilized for the analysis of protein band intensities, which were then normalized to the loading control GAPDH. The antibodies are listed in Supplementary Table 2.

Mass spectrometry

To determine the interacting proteins with S100A14, we performed endogenous immunoprecipitation of S100A14. The magnetic beads were eluted with 80 µL of acid elution solution (0.1 M Glycine-HCl, pH 2.8) and neutralized with 8 µL of neutralizing solution (1 M Tris-HCl, pH 7.9) to obtain the protein eluate samples. The samples were used for western blotting or liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. To identify differential post-translational modification sites of GLSKGA in S100A14 knockdown or non-target control (NTC) treated Huh7 cells, we performed endogenous immunoprecipitation of GLSKGA. Immunoprecipitates were eluted with 80 µL of SDS sample buffer, resolved on an 8% SDS-PAGE gel, and stained using Fast Silver Stain Kit (#P0017S, Beyotime). Gel slices containing the GLSKGA protein near approximately 66 kDa were used for LC-MS/MS analysis. The in-solution digestion and in-gel digestion of the samples were performed as previously described, respectively [26, 27]. LC-MS/MS analysis and data analysis were performed as previously described [28]. Data were analyzed using peaks online (Bioinformatics Solutions Inc) software, and the database was downloaded from Swiss-Prot using the human database (20,334 proteins). Acetylation (K), Acetylation (Protein N-terminal), Acetylation (N-terminal), Phosphorylation (STY), Succinic, and Oxidation (M) were set to variable modification. The retrieved protein peak areas were used for subsequent statistical analysis.

Molecular docking

The 3D structures of S100A14 (PDB: 2M0R, chain A) and GLSKGA (PDB: 5UQE, chain A, residues 137–656) were retrieved from the RCSB Protein Database (https://www.rcsb.org/). All ligands and water molecules present in their crystal structures were removed. Subsequently, a docking study was conducted using the ZDOCK server (https://zdock.umassmed.edu/) to investigate the binding mode between S100A14 and GLSKGA [29]. The top-ranked predicted binding mode was visualized, analyzed, and mapped using the PyMOL program (http://www.pymol.org).

In vivo animal experiments

Four-week-old male NOD/SCID mice and four-week-old male nude mice were purchased from SLAC Laboratory Animal Company (SLAC, Shanghai, China). The mice were all randomly grouped for the experiment. For single-cell lineage tracing, four-week-old male NOD/SCID mice were injected subcutaneously with 1 × 105 Huh7 cells transfected with barcodes in 0.1 mL of PBS containing 70% Matrigel (#356234, Corning), with three mice per vehicle group and six mice per sorafenib group. When tumors reached approximately 150 mm3, the drug-naive tumors of two mice were harvested, and the remaining mice received treatment with sorafenib (25 mg/kg/day, orally) or vehicle. After 21 days, the sorafenib-treated tumors were harvested. For the knockdown of S100A14 on the effect of sorafenib treatment in vivo, four-week-old male nude mice were injected subcutaneously with 5 × 106 Huh7-shS100A14 or Huh7-NTC cells in 0.1 mL of PBS containing 50% Matrigel. When tumors reached approximately 80–100 mm3, the mice were treated with sorafenib (15 mg/kg/day, orally) or vehicle for 14 days. For inhibiting GLS on the impact of sorafenib treatment in vivo, four-week-old male nude mice were injected subcutaneously with 5 × 106 Huh7 cells in 0.1 mL of PBS containing 50% Matrigel. When tumors reached approximately 100 mm3, the mice were randomized to four groups (n = 6) and received the following treatment daily by gavage for 14 days: (1) vehicle: 12.5% Cremophor EL, 12.5% ethanol, 75% water; (2) 15 mg/kg/day body weight of sorafenib; (3) 200 mg/kg body weight of CB-839, twice daily; and (4) sorafenib combined with CB-839. The tumor volume was measured every other day. Tumor volumes were calculated with the following formula: volume = length × width2 × 0.5. At the end of the experiments, the mice were sacrificed, and the tumors were harvested and photographed.

Immunohistochemistry and H&E staining

The collected tumors were fixed with formalin, embedded in paraffin, and were cut into serial 4 µm sections. For immunohistochemistry, the sections were deparaffinized, subjected to antigen retrieval, blocked, and incubated at 4℃ overnight with the indicated primary antibodies. After incubation with the secondary antibody, the proteins were visualized with 3,3’-diaminobenzidine tetrahydrochloride (DAB) (#AR1027, Boster Biological Technology). The tissue microarray from 80 HCC patients at Fudan University, Zhongshan Hospital was stained with anti-S100A14 antibodies, and the expression levels of S100A14 were assessed by two pathologists using a semi-quantitative method. The IHC score of S100A14 in the tissue microarray was determined by multiplying the staining intensity score by the positive score of cytoplasmic staining. The staining intensity score was assigned as follows: 1 = weak, 2 = moderate, and 3 = strong. The positive score of cytoplasmic staining was defined as follows: 1 = 0–25%, 2 = 26–50%, 3 = 51–75%, and 4 = above 75%. The samples with IHC score ≥ 6 were defined as high expression, and others were considered as low expression. The antibodies used are listed in Supplementary Table 2. Clinical and pathologic information of the samples and the IHC score of S100A14 in the tissue microarray were showed in Supplementary Table 3. Hematoxylin & eosin staining (H&E) was used to detect the morphology of tumor cells in tissue sections, which was performed using the H&E kit (#G1120, Solarbio) according the manufacturer’s instructions.

Statistical analysis

All data were expressed as the mean ± standard deviations (SD) and were based on at least three independent experiments, unless otherwise indicated. All statistics were performed using GraphPad Prism (v 9.0). Statistical probabilities were determined using either one-way ANOVA or two-way ANOVA where appropriate. Overall survival (OS) and recurrence-free survival (RFS) were assessed using the Kaplan–Meier method and evaluated with the log-rank test. A p value of 0.05 was considered as the threshold for statistical significance.

Results

Barcode-based tracing of cell lineages in HCC xenografts and identification of sorafenib-resistant lineages

To identify subpopulations of hepatocellular carcinoma (HCC) cells that would exhibit primary resistance to sorafenib exposure, we labeled individual cells in drug-naive HCC xenografts on mice with randomly unique 14-base pair (bp) DNA sequences known as “barcodes” [16], then exposed the tumors to sorafenib and compared the characteristics of tumor cells before and after drug exposure (Fig. 1A). The HCC cell line Huh7 was transduced with recombinant lentivirus encoding 14 bp barcodes and the fluorescent reporter CopGFP at a sufficiently low multiplicity of infection (MOI) to ensure that most transductants would receive a single barcode. Barcoded tumor cells were sorted based on CopGFP fluorescence (Fig. S1A) and expanded under puromycin selection to ensure > 95% CopGFP positivity (Fig. S1B). At the same time, another parallel preparation of cell samples was validated by PCR using barcode library primers, showing good specificity of the barcode library (Fig. S1C). The cells were injected subcutaneously into NOD/SCID mice and allowed to grow to a volume of approximately 150 mm3, after which they were treated with sorafenib for three weeks, resulting in a significant slowdown in tumor growth (p < 0.0001, Fig. S1D). Two animals were sacrificed before sorafenib treatment and four animals were sacrificed at the end of treatment, their tumors (drug-naive tumors (DN) and drug-treated tumors (DT), respectively) were excised and dissociated, viable CopGFP-positive tumor cells were selected using a flow cytometric sorter (Fig. S1E and F), and the DNA barcodes as well as transcriptome in those cells were sequenced. We excluded cells for which we detected more than 30,000 unique molecular identifiers (UMIs) or expression of fewer than 200 genes or more than 5000 genes, and also excluded cells for which more than 20% of expressed genes were mitochondrial in origin (Fig. S2A). The final analysis included 81,400 cells containing a median of 13,504 UMIs and expressing a median of 3,611 genes and 9,579 unique barcodes.

Fig. 1
figure 1

Barcode-based tracing of cell lineages in HCC xenografts and identification of sorafenib-resistant lineages. (A) Schematic of the experimental procedure for barcode lineage tracing and single-cell RNA sequencing (scRNA-seq). (B) Venn diagram of the numbers of barcodes identified in replicates of DN and DT. DN, drug-naive tumors (n = 2 mice); DT, drug-treated tumors (n = 4 mice). (C) Venn diagram of the numbers of unique barcodes identified in the two types of tumors. (D) Differences in the frequencies of the 2,539 barcodes from panel C between sorafenib-treated and drug-naive tumors. (E and I) Cell clusters from drug-naive and drug-treated tumors based on uniform manifold approximation and projection (UMAP) of scRNA-seq data. (F and J) UMAP visualization of tumor cells expressing the 1,241 barcodes (termed barcodehigh in panel D) (blue dots), and cells without those barcodes (gray dots). (G and K) Proportions of cells that expressed a barcodehigh in each cluster from panels E and I. (H and L) The trend of proportion of cells with a barcodehigh ranging from 0 to 1,241 in each cluster of DN and DT

Full size image

Of the 9,579 barcodes, 2,539 barcodes were shared between tumors before and after sorafenib treatment (Fig. 1B and C) and were present in 48,926 cell clones, corresponding to 60.1% of the total number of clones. Of the 2,539 shared barcodes, 1,241 barcodes (termed barcodehigh) were enriched after sorafenib treatment (Fig. 1D), distributing in 28,582 clones, which accounts for 58.4% of all clones encoding shared barcodes (Supplementary Table 4). These clones with higher barcode abundance were defined as the drug-resistant clones.

All tumor cell clones fell into a small number of clusters that were consistent across replicate samples of xenografts before and after drug treatment (Fig. S2B and C). Among them, the resistant clones in drug-naive tumors fell primarily within cluster 1 (Fig. 1E, F, G and H, Supplementary Table 4), while the resistant clones in drug-treated tumors fell primarily within clusters 1, 2 and 4 and less so within cluster 3 (Fig. 1I, J, K and L, Supplementary Table 4). We hypothesized that cluster 1 in drug-naive HCC xenografts is a primary sorafenib-resistant subpopulation of tumor cells.

Transcriptomic characteristics of sorafenib-resistant cells in drug-naive HCC xenografts

We developed a gene set comprising the top 20 genes highly expressed in cluster 1 of drug-naive tumors, and stratified HCC patients in The Cancer Genome Atlas (TCGA) based on high or low expression of this gene set. High expression of the gene set in cluster 1 of drug-naive tumors was associated with significantly worse overall survival (p = 0.022, Fig. 2A). These results support the notion that the cluster 1 in drug-naive HCC tumors exhibits resistance to sorafenib.

Fig. 2
figure 2

Transcriptomic characteristics of sorafenib-resistant cells in drug-naive HCC xenografts. (A) Comparison of overall survival of patients with advanced HCC in TCGA, grouping patients into high or low expression groups based on gene set scores of the top 20 candidate marker genes in cluster 1 of drug-naive tumor. (B) Heatmap of the relative expression levels of candidate genes. The top 20 genes selected in cluster 1 of drug-naive tumors are displayed. (C) Dot plot of the expression of top 20 genes from cluster 1 across each cluster in drug-naive tumors. Dots are colored according to level of expression, from low (blue) to high (red). Dot size corresponds to the percentage of cells expressing each gene. Selected markers, ANGPTL3, FXYD3, S100A4, and S100A14, are labeled. (D and E) Expression in drug-naive tumors of four genes previously associated with poor prognosis in HCC (D) based on UMAP, where relative expression is indicated on a color scale, and (E) based on violin plots, where width of the data distribution reflects the kernel density of expression values

Full size image

To identify genes potentially associated with sorafenib resistance, we compared the relative expression of the 20 candidate genes across different clusters in tumors (Fig. 2B). Dot plots showed the expression of these genes among different clusters, and demonstrated their significantly high expression and high proportion in the cluster 1 subpopulation in drug-naive tumors (Fig. 2C). Among them several genes have been reported to be highly expressed and associated with poor prognosis in HCC: ANGPTL3, encoding angiopoietin-like protein 3 (ANGPTL3) [30]; FXYD3, encoding FXYD domain‑containing ion transport regulator 3 (FXYD3) [31]; S100A4, encoding S100 calcium binding protein A4 (S100A4) [32]; and S100A14, encoding S100 calcium binding protein A14 (S100A14) [33]. All these genes were highly expressed in a large proportion of cells in cluster 1, but not other clusters, in drug-naive tumors (Fig. 2D and E). Therefore, we hypothesized that these genes are biomarkers for primary resistance to sorafenib in HCC.

Knockdown of S100A14 sensitizes cultures of HCC cells to sorafenib

To validate the four candidate marker genes of primary resistance to sorafenib, we engineered recombinant lentiviruses in which the promoter of each of the four genes drove the expression of enhanced green fluorescent protein (EGFP) [23, 24], then we infected cultures of Huh7 cells with the lentiviruses (Fig. 3A). We separated the infected cells based on whether they expressed low or high levels of EGFP, which meant that they expressed low or high levels of each candidate marker gene (Fig. 3B) [22]. Finally, we exposed the cultures of “low”- and “high”-expressing cells to sorafenib and measured their viability. Expression of the two candidate genes was associated with resistance to sorafenib, with the gene encoding S100A14 showing by far the strongest effect (p < 0.001, Fig. 3C). Therefore, we focused subsequent experiments on S100A14 as a potential mediator of primary resistance to sorafenib and a potential prognostic indicator.

Fig. 3
figure 3

Knockdown of S100A14 sensitizes cultures of HCC cells to sorafenib. (A) Schematic of procedure to generate Huh7 cells expressing EGFP under the control of distinct promoters from four candidate genes associated with poor prognosis. (B) Representative results of cell sorting. (C) Viability of the sorted cells in panel B after sorafenib treatment, based on the CCK-8 assay (n = 3). (D and E) Comparison of recurrence-free and overall survival of 80 HCC patients at Zhongshan Hospital, Fudan University. HR, hazard ratio. (F, G, H, I, J, K and L) Huh7 or Hep3B cells were transfected with one of two short hairpin RNAs against S100A14 (shS100A14) or a non-target control RNA (NTC), treated with sorafenib, then their (F) viability was assessed using the CCK-8 assay, (G and I) apoptosis was assessed using flow cytometry, (H and J) production of ROS was assessed through staining with the DCFH-DA probe followed by flow cytometry, and (K and L) ability to form colonies was assessed. n = 3. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant

Full size image

Among patients in TCGA with HCC, levels of S100A14 mRNA were significantly higher in early or advanced tumors than in healthy liver tissue (Fig. S3A), and higher levels of the S100A14 mRNA were associated with significantly worse overall survival (hazard ratio (HR) = 1.58, p = 0.042, Fig. S3B). Immunohistochemical staining of HCC biopsies from 80 patients who were treated with sorafenib at Zhongshan Hospital, Fudan University (Fig. S3C and D, Supplementary Table 3) corroborated the association between higher S100A14 expression and worse survival (Fig. 3D and E).

These analyses suggest that HCC cells expressing higher levels of S100A14 can act as a “reservoir” of primary resistance to sorafenib and can expand during drug treatment, leading to the emergence of resistant tumors. Consistent with this finding, knocking down S100A14 in the two HCC cell lines Huh7 and Hep3B using either of two short hairpin RNAs (Fig. S4A and B) sensitized the cells to sorafenib (p < 0.001, Fig. 3F). This sensitization involved increases in tumor cell apoptosis and levels of ROS (Fig. 3G, H, I and J), consistent with the ability of sorafenib to induce apoptosis through oxidative stress [34]. This sensitization was sufficient to prevent formation of tumor cell colonies (Fig. 3K and L), suggesting that our culture experiments also reflect the situation in vivo. These results showed that S100A14 is a pivotal mediator of primary resistance to sorafenib and a key prognostic indicator in advanced hepatocellular carcinoma.

S100A14 stabilizes kidney-type glutaminase in cultures of HCC cells

To explain how S100A14 mitigate sorafenib-induced oxidative stress, we examined the protein’s potential binding partners in Huh7 cells using an unbiased co-immunoprecipitation assay, followed by LC-MS/MS analysis (Fig. 4A, Fig. S5A). This screening detected an interaction between S100A14 and glutaminase (Fig. 4B, Supplementary Table 5), which are known to help maintain redox homeostasis [35] and thereby influence progression of various types of cancer, including hepatocellular carcinoma [36]. We confirmed this interaction using specific antibodies against glutaminases in Huh7 and Hep3B cells, which indicated much stronger binding of S100A14 to the kidney-type glutaminase isoform (GLSKGA) than to glutaminase isoform C (GLSGAC) (Fig. 4C, D, E and F, Fig. S5B). We corroborated these western assays by performing immunofluorescence staining on either wild-type HCC cells or those expressing FLAG-tagged S100A14 and HA-tagged GLSKGA, which localized the interaction primarily to mitochondria (Fig. 4G, H, I and J).

Fig. 4
figure 4

S100A14 interacts with kidney-type glutaminase in cultures of HCC cells. (A) Schematic of the experimental workflow to identify binding partners of S100A14 in Huh7 cells. (B) A list of representative proteins that co-immunoprecipitated with S100A14. (C and D) Immunoprecipitates were obtained as described from panel A in Huh7 and Hep3B cells, then analyzed using a pan-antibody against glutaminases (GLS) or antibodies specifically against the kidney-type isoform (GLSKGA) or isoform C (GLSGAC). (E and F) Cell lysates were immunoprecipitated using a pan-antibody against GLS or an IgG control antibody, then immunoblotted for S100A14. (G, H, I and J) Huh7 and Hep3B cells were examined using confocal fluorescence microscopy to observe interactions between (G and H) endogenous S100A14 and GLSKGA or between (I and J) FLAG-tagged S100A14 and HA-tagged GLSKGA expressed from transfected plasmids. Mitochondria were detected using MitoTracker Red CMXRos. Quantitation of fluorescence intensity tracings (along the white arrows) in “Merge” images was shown at the far right. n = 3. Scale bar, 10 μm

Full size image

Knocking down GLSKGA sensitized Huh7 cultures to sorafenib (p < 0.001, Fig. 5A), similar to the sensitization that we observed above when we knocked down S100A14. Furthermore, we could reverse sensitization due to GLSKGA knockdown by overexpressing wild-type GLSKGA but not its S286A mutant (Fig. 5B and C), which is catalytically inactive [37].

Fig. 5
figure 5

S100A14 stabilizes kidney-type glutaminase in cultures of HCC cells. (A) Huh7 cells were transfected with one of two short hairpin RNAs against kidney-type glutaminase (shGLSKGA) or NTC, treated with sorafenib, then assessed for viability using the CCK-8 assay (n = 3). (B) Cells were transfected as described in panel A but also with empty expression vector (Vector) or plasmid expressing HA-tagged wild-type GLSKGA (GLSKGA WT) or a catalytically inactive mutant (GLSKGA S286A). Cell lysates were immunoblotted with antibodies against GLSKGA, HA, and GAPDH. (C) Cells were transfected as described in panel B, treated with sorafenib, and assayed for viability in the CCK-8 assay (n = 3). (D) Relative levels of GLSKGA mRNA in Huh7 cells that had been transfected with shS100A14 or NTC (n = 3). (E and F) (E) Western blotting of GLSKGA protein in cells transfected as described in panel D which were treated with cycloheximide (CHX) from 0 to 24 h and then (F) GLSKGA protein levels were quantified by ImageJ (n = 3). (G, H, I and J) Cells were transfected with shS100A14 or NTC to knock down S100A14 but overexpress HA-tagged wild-type or inactive GLSKGA, then (G) lysates were immunoblotted with antibodies against GLSKGA, HA, S100A14, and GAPDH, (H) cell viability under sorafenib treatment was measured using the CCK-8 assay, (I) production of ROS was assessed by staining cells with DCFH-DA probe followed by flow cytometry, and (J) levels of reduced glutathione (GSH) in lysates were assayed. n = 3. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant

Full size image

We hypothesized that S100A14 might render tumor cells resistant to sorafenib by altering the levels of GLSKGA. While knocking down S100A14 did not significantly affect levels of the mRNA encoding the glutaminase (Fig. 5D), it did shorten the half-life of the encoded protein (Fig. 5E). The protein levels of GLSKGA in shS100A14 cells were reduced to approximately 60% of those in NTC cells (Fig. 5F). The sensitization due to S100A14 knockdown was largely reversed by overexpressing wild-type GLSKGA but not its catalytically inactive mutant (Fig. 5G and H). Furthermore, the levels of ROS and GSH, the primary intracellular antioxidant that neutralizes ROS, in these cells were examined [38]. The results indicated that ROS levels in S100A14-knockdown cells significantly increased following sorafenib administration, while GSH levels exhibited a negative correlation with ROS levels. Notably, the overexpression of wild-type GLSKGA, but not GLSKGA S286A mutant, could partially rescue the alterations in ROS or GSH levels in S100A14-knockout cells (Fig. 5I and J). Overall, these results suggested that S100A14 promotes primary resistance to sorafenib by stabilizing glutaminase in mitochondria, which helps eliminate pro-apoptotic ROS whose formation is induced by sorafenib.

S100A14 prevents GLSKGA ubiquitination by blocking phosphorylation of residues Y308 and S314

We hypothesized that S100A14 stabilizes GLSKGA protein by interfering with its ubiquitination and subsequent proteolytic degradation [39]. Indeed, the ability of S100A14 knockdown to shorten the half-life of GLSKGA was diminished by the proteasome inhibitor MG132, but not by the lysosome inhibitor chloroquine (Fig. 6A and B). Furthermore, S100A14 knockdown upregulated GLSKGA ubiquitination in Huh7 cells (Fig. 6C).

Fig. 6
figure 6

S100A14 prevents GLSKGA ubiquitination by blocking phosphorylation of residues Y308 and S314. (A and B) Huh7 cells were transfected with shS100A14 or NTC, then treated for 24 h with 25 µg/ml CHX, 20 µM proteasome inhibitor MG132, or 50 µM lysosome inhibitor chloroquine (CQ). Cell lysates were immunoblotted with antibodies against S100A14 and GLSKGA, and the protein levels of GLSKGA were quantified by ImageJ (n = 3). (C) Cells were transfected as described in panel A, treated for 8 h with 20 µM MG132. Lysates were immunoprecipitated with the anti-GLSKGA antibody. Precipitates were immunoblotted to detect ubiquitin (Ub) and GLSKGA. (D) Cells were transfected as described in panel A, and lysates were subjected to immunoprecipitation by anti-GLSKGA or anti-IgG antibody. Precipitates were fractionated on gels and silver stained. The band corresponding to GLSKGA (red arrow) was excised. (E) Western blot verified the immunoprecipitation efficiency of GLSKGA, which was immunoblotted with indicated antibodies. (F) Peptides of GLSKGA bearing post-translational modifications that were detected by mass spectrometry in lysates of shS100A14 cells, but not in NTC cells. AA, amino acid residue. (G) Molecular docking to predict a potential binding mode between S100A14 (green) and GLSKGA (pink). Key residues at the interaction interface are depicted in ball-and-stick mode. Y308 and S314 at the interaction interface are shown in heavy coloring. Hydrogen bonds are displayed as yellow dashed lines, and their distances are labeled. (H) Cells were transfected as described in panel A but also with a plasmid expressing HA-tagged wild-type (WT) or mutant forms of GLSKGA, then treated for 8 h with 20 µM MG132. Cell lysates were subjected to anti-HA immunoprecipitation, and precipitates were immunoblotted for Ub. (I) Lysates from Huh7 cells expressing HA-tagged wild-type or mutant forms of GLSKGA were immunoprecipitated using anti-HA antibody, and precipitates were analyzed using antibody against S100A14 or HA tag. ***p < 0.001; ns, not significant

Full size image

Given that post-translational modifications often regulate proteolytic degradation of proteins [40, 41], We analyzed such modifications of GLSKGA in Huh7 cells with S100A14 knockdown through immunoprecipitation and mass spectrometry (Fig. 6D and E). The results showed that knockdown of S100A14 increased phosphorylation of the active-site residues Y308 and S314 in GLSKGA (Fig. 6F, Supplementary Table 6). Molecular docking predicts that these residues lie at the interface of the complex between S100A14 and GLSKGA (Fig. 6G). To assess the effects of phosphorylation on the stability of GLSKGA, we knocked down S100A14 in Huh7 cells, after which the cells were transfected with HA-tagged wild-type GLSKGA or phosphorylation-resistant mutants. We observed that the knockdown of S100A14 led to a significant increase in the ubiquitination of HA-tagged wild-type GLSKGA. However, the ubiquitination levels were markedly reduced in all GLSKGA mutants, particularly in the mutant with simultaneous mutations at both residues Y308 and S314 (Fig. 6H). Furthermore, each mutation individually weakened the interaction between S100A14 and GLSKGA, while the double mutation almost completely eliminated their interaction (Fig. 6I). These results indicated that S100A14 inhibits the ubiquitination of GLSKGA by suppressing the phosphorylation of its active-site residues Y308 and S314.

Knockdown of S100A14 or pharmacological inhibition of glutaminase potentiates sorafenib against HCC xenografts

In line with the in vitro results, knockdown of S100A14 remarkably improved sorafenib efficacy in nude mice bearing Huh7 xenografts (p < 0.001, Fig. 7A and B), accompanied with a significant decrease in Ki67 expression (p < 0.001, Fig. 7C) and an increase in the proportion of apoptotic tumor cells (p < 0.001, Fig. 7D). Additionally, sorafenib treatment did not appear to induce systemic toxicity, based on the relatively stable body weight of mice during treatment (Fig. S6A). Similarly to inhibiting S100A14, we found that inhibiting glutaminase using the specific inhibitor telaglenastat (CB-839), which has entered phase I and II trials as an anti-cancer drug [42], enhanced the therapeutic efficacy of sorafenib against subcutaneous Huh7 xenograft (p < 0.001, Fig. 7E and F). As compared to the groups treated with either drug alone, the combined treatment led to a significant reduction in Ki67 expression (p < 0.001, Fig. 7G) and a higher proportion of apoptotic tumor cells (p < 0.001, Fig. 7H). Besides, sorafenib treatment and glutaminase inhibition did not appear to induce systemic toxicity, based on the relatively stable body weight of mice during treatment (Fig. S6B). All these results suggested that inhibiting S100A14 or GLS significantly enhances sorafenib efficacy against xenograft tumors in vivo.

Fig. 7
figure 7

Inhibition of S100A14 or glutaminase potentiates sorafenib against HCC xenografts. (A and B) Representative photographs and growth curves of Huh7 xenografts that were transfected with shS100A14 or NTC, and treated with sorafenib or vehicle. n = 6 mice. (C and D) Thin sections of excised tumors from panel A were subjected to (C) staining with hematoxylin-eosin (HE) or immunostaining for S100A14 or Ki67 or (D) staining in the TUNEL assay to identify apoptotic cells, with nuclear counterstaining using Hoechst dye. n = 6 mice. Scale bar in panel C, 100 μm; in panel D, 40 μm. (E and F) Representative photographs and growth curves of Huh7 xenografts treated with the glutaminase inhibitor CB-839, sorafenib or both. n = 6 mice. (G and H) Thin sections of excised tumors from panel E were subjected to (G) staining with hematoxylin-eosin (HE) or immunostaining for GLS or Ki67 or (H) staining in the TUNEL assay to identify apoptotic cells, with nuclear counterstaining using Hoechst dye. n = 6 mice. Scale bar in panel G, 100 μm; in panel H, 40 μm. ***p < 0.001

Full size image

Discussion

Our experiments suggest that S100A14 can promote primary resistance to sorafenib and that its expression can predict prognosis of patients with advanced hepatocellular carcinoma. S100A14 may bind to GLSKGA and prevent its phosphorylation in the active site, which in turn inhibits its ubiquitination and thereby stabilizes it against proteolytic degradation. This ensures that GLSKGA can neutralize reactive oxygen species and therefore counteract the ability of sorafenib to induce apoptosis through oxidative stress. Our results imply that inhibiting S100A14 or GLSKGA can slow tumor growth, as demonstrated through tumor xenografts in mice. Our approach of combining lineage tracing and RNA sequencing at the single-cell level allowed us to identify tumor subpopulations and characterize their transcriptomes simultaneously, which may make it effective for elucidating mechanisms of primary drug resistance in HCC and other cancers. Such work is important for complementing the more extensive literature on mechanisms of drug primary resistance in cancer [8].

S100A14, a member of the EF-hand calcium-binding protein family that is closely associated with cancer development, has been reported to be overexpressed and to promote cell growth and metastasis in HCC [33]. Additionally, high expression of S100A14 accelerated the progression of pancreatic cancer and contributed to gemcitabine resistance [43]. Nevertheless, whether S100A14 affects sorafenib resistance in HCC remains unclear. Our results demonstrated the prognostic value of S100A14 in predicting sorafenib resistance in HCC and clarified the involved mechanism, which maintains mitochondrial redox homeostasis by stabilizing GLS. It would be interesting to conduct further studies on the diagnostic and prognostic role of S100A14 in pan-cancer [44, 45] and clarify whether other S100 proteins contribute to drug resistance by modulating mitochondrial metabolism.

Of note, our results indicate that the phosphorylation of Y308 and S314 in glutaminase can influence its ubiquitination and degradation, thereby expanding the list of post-translational modifications recognized to regulate the stability or activity of glutaminase. Previous studies have demonstrated that succinylation at K164 in the enzyme promotes ubiquitination at K158, which in turn promotes its degradation [46]. Conversely, succinylation at K311 promotes oligomerization of glutaminase, enhancing its activity [47]. Future research should further explore how phosphorylation of glutaminase contributes to mitochondrial redox homeostasis.

This study has several limitations. First, the tumor immune microenvironment has been suggested to play a critical role in therapeutic resistance [48,49,50]. Whether the findings in the current study are recapitulated in immune-competent context is unknown. Second, the S100A14-mediated mechanisms of drug resistance were not extensively validated using other advanced tyrosine kinase inhibitors, such as lenvatinib. Third, the lack of specific inhibitors for S100A14 or GLSKGA urgently necessitates the screening of inhibitors targeting these molecules [51].

Conclusions

By utilizing single-cell lineage tracing technology, we identified S100A14 as a prognostic biomarker and a regulator of primary resistance to sorafenib in HCC. Mechanistically, S100A14 interacts with GLS to block its phosphorylation at residues Y308 and S314, and thereby inhibits its ubiquitin-mediated degradation. The stabilization of GLS protein effectively counteracts sorafenib-induced ROS production and apoptosis, rendering cancer cells insusceptible to the drug. Inhibiting either S100A14 or GLS substantially improved sorafenib efficacy in vivo. These findings suggest that targeting the S100A14-GLS axis could serve as a novel therapeutic strategy to overcome sorafenib resistance in HCC.

Data availability

The scRNA-seq data have been deposited at the Gene Expression Omnibus (GEO) data repository with the accession code GEO: GSE282234 and are publicly available as of the date of publication. The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Abbreviations

HCC:

Hepatocellular carcinoma

GLS:

Glutaminase

GLSKGA :

Kidney-type glutaminase isoform

GLSGAC :

Glutaminase isoform C

NTC:

Non-target control

TSS:

Transcription start site

PBS:

Phosphate-buffered saline

UMAP:

Uniform manifold approximation and projection

DEGs:

Differentially expressed genes

ROS:

Reactive oxygen species

GSH:

Glutathione

LC-MS/MS:

Liquid chromatography-tandem mass spectrometry

OS:

Overall survival

RFS:

Recurrence-free survival

DN:

Drug-naive tumors

DT:

Drug-treated tumors

TCGA:

The cancer genome atlas

IHC:

Immunohistochemistry

EGFP:

Enhanced green fluorescent protein

scRNA-seq:

single cell RNA sequencing

LTR:

Long terminal repeat

Puro:

Puromycin

T2A:

2 A peptide from Thosea asigna virus

WPRE:

Woodchuck hepatitis virus post-transcriptional regulation element

Ub:

Ubiquitin

CHX:

Cycloheximide

CQ:

Chloroquine

References

  1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71(3):209–49.

    Article  PubMed  Google Scholar 

  2. El-Serag HB, Rudolph KL. Hepatocellular carcinoma: epidemiology and molecular carcinogenesis. Gastroenterology. 2007;132(7):2557–76.

    Article  CAS  PubMed  Google Scholar 

  3. Qi F, Li J, Qi Z, Zhang J, Zhou B, Yang B et al. Comprehensive metabolic profiling and Genome-wide analysis reveal therapeutic modalities for hepatocellular carcinoma. Research (Wash D C). 2023;6:0036.

  4. Shi J, Liu J, Tu X, Li B, Tong Z, Wang T, et al. Single-cell immune signature for detecting early-stage HCC and early a Ssessing anti-PD-1 immunotherapy efficacy. J Immunother Cancer. 2022;10(1):e003133.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, et al. Sorafenib in advanced hepatocellular carcinoma. N Engl J Med. 2008;359(4):378–90.

    Article  CAS  PubMed  Google Scholar 

  6. Tang W, Chen Z, Zhang W, Cheng Y, Zhang B, Wu F, et al. The mechanisms of Sorafenib resistance in hepatocellular carcinoma: theoretical basis and therapeutic aspects. Signal Transduct Target Ther. 2020;5(1):1–15.

    Google Scholar 

  7. Zhu Y-J, Zheng B, Wang H-Y, Chen L. New knowledge of the mechanisms of Sorafenib resistance in liver cancer. Acta Pharmacol Sin. 2017;38(5):614–22.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Ladd AD, Duarte S, Sahin I, Zarrinpar A. Mechanisms of drug resistance in HCC. Hepatology. 2024;79(4):926–40.

    PubMed  Google Scholar 

  9. Mei K, Chen Z, Wang Q, Ali A, Huang Y, Yi L. A prognostic aging-related LncRNA risk model correlates with the immune microenvironment in HCC. Cancer Insight. 2024;3(2):26–41.

    Google Scholar 

  10. Dagogo-Jack I, Shaw AT. Tumour heterogeneity and resistance to cancer therapies. Nat Rev Clin Oncol. 2018;15(2):81–94.

    Article  CAS  PubMed  Google Scholar 

  11. Kester L, van Oudenaarden A. Single-cell transcriptomics Meets lineage tracing. Cell Stem Cell. 2018;23(2):166–79.

    Article  CAS  PubMed  Google Scholar 

  12. Wagner DE, Klein AM. Lineage tracing Meets single-cell omics: opportunities and challenges. Nat Rev Genet. 2020;21(7):410–27.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Yang L, He Y-T, Dong S, Wei X-W, Chen Z-H, Zhang B, et al. Single-cell transcriptome analysis revealed a suppressive tumor immune microenvironment in EGFR mutant lung adenocarcinoma. J Immunother Cancer. 2022;10(2):e003534.

    Article  PubMed  PubMed Central  Google Scholar 

  14. Franken A, Van Mol P, Vanmassenhove S, Donders E, Schepers R, Van Brussel T, et al. Single-cell transcriptomics identifies pathogenic T-helper 17.1 cells and pro-inflammatory monocytes in immune checkpoint inhibitor-related pneumonitis. J Immunother Cancer. 2022;10(9):e005323.

    Article  PubMed  PubMed Central  Google Scholar 

  15. Rodriguez-Fraticelli AE, Weinreb C, Wang S-W, Migueles RP, Jankovic M, Usart M, et al. Single-cell lineage tracing unveils a role for TCF15 in haematopoiesis. Nature. 2020;583(7817):585–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Quinn JJ, Jones MG, Okimoto RA, Nanjo S, Chan MM, Yosef N et al. Single-cell lineages reveal the rates, routes, and drivers of metastasis in cancer xenografts. Science. 2021;371(6532):eabc1944.

  17. Eyler C, Matsunaga H, Hovestadt V, Vantine S, van Galen P, Bernstein B. Single-cell lineage analysis reveals genetic and epigenetic interplay in glioblastoma drug resistance. Genome Biol. 2020;21(1):174.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Wu J-J, Zhu S, Tang Y-F, Gu F, Valencak TG, Liu J-X et al. Age- and Microbiota-Dependent cell stemness plasticity revealed by Cat Tle cell landscape. Research (Wash D C). 2023;6:0025.

  19. Dura B, Choi J, Zhang K, Damsky W, Thakral D, Bosenberg M, et al. scFTD-seq: freeze-thaw Lysis based, portable approach toward highly distributed single-cell 3’ mRNA profiling. Nucleic Acids Res. 2019;47(3):e16.

    Article  PubMed  Google Scholar 

  20. Satija R, Farrell JA, Gennert D, Schier AF, Regev A. Spatial reconstruction of single-cell gene expression data. Nat Biotechnol. 2015;33(5):495–502.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Luo Z, Zhu J, Fang Z, Xu R, Wan R, He Y, et al. Exercise-augmented THSD7B exhibited a positive prognostic implication and tumor-suppressed functionality in pan-cancer. Front Immunol. 2024;15:1440226.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Leung HW, Leung CON, Lau EY, Chung KPS, Mok EH, Lei MML, et al. EPHB2 activates β-catenin to enhance cancer stem cell properties and drive Sorafenib resistance in hepatocellular carcinoma. Cancer Res. 2021;81(12):3229–40.

    Article  CAS  PubMed  Google Scholar 

  23. Wang Q, Liang N, Yang T, Li Y, Li J, Huang Q, et al. DNMT1-mediated methylation of BEX1 regulates stemness and tumorigenicity in liver cancer. J Hepatol. 2021;75(5):1142–53.

    Article  CAS  PubMed  Google Scholar 

  24. Shan J, Shen J, Liu L, Xia F, Xu C, Duan G, et al. Nanog regulates self-renewal of cancer stem cells through the insulin‐like growth factor pathway in human hepatocellular carcinoma. Hepatology. 2012;56(3):1004–14.

    Article  CAS  PubMed  Google Scholar 

  25. Luo Z, Mei J, Wang X, Wang R, He Z, Geffen Y, et al. Voluntary exercise sensitizes cancer immunotherapy via the collagen in hibition-orchestrated inflammatory tumor immune microenvironment. Cell Rep. 2024;43(9):114697.

    Article  CAS  PubMed  Google Scholar 

  26. Shevchenko A, Tomas H, Havlis J, Olsen JV, Mann M. In-gel digestion for mass spectrometric characterization of proteins a Nd proteomes. Nat Protoc. 2006;1(6):2856–60.

    Article  CAS  PubMed  Google Scholar 

  27. Ng CKY, Dazert E, Boldanova T, Coto-Llerena M, Nuciforo S, Ercan C, et al. Integrative proteogenomic characterization of hepatocellular carcinoma across etiologies and stages. Nat Commun. 2022;13(1):2436.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Wang Z, Zhang L, Yuan W, Zhang Y, Lu H. SAPT, a fast and efficient approach for simultaneous profiling of prot Ein N- and C-Terminome. Anal Chem. 2021;93(30):10553–60.

    Article  CAS  PubMed  Google Scholar 

  29. Pierce BG, Wiehe K, Hwang H, Kim B-H, Vreven T, Weng Z. ZDOCK server: interactive Docking prediction of protein-protein Comple Xes and symmetric multimers. Bioinformatics. 2014;30(12):1771–3.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Amal S, Zidan HE, Rashad NM, Wadea FM. Angiopoietin-like protein 3 and 4 expression 4 and their serum levels in hepatocellular carcinoma. Cytokine. 2017;96:75–86.

    Article  Google Scholar 

  31. Wang LJ, Li QJ, Le Y, Ouyang HY, He MK, Yu ZS, et al. Prognostic significance of sodium-potassium ATPase regulator, FXYD3, in human hepatocellular carcinoma. Oncol Lett. 2018;15(3):3024–30.

    PubMed  Google Scholar 

  32. Zhang J, Zhang D, Jiao X, Dong Q. S100A4 regulates migration and invasion in hepatocellular carcinoma HepG2 cells via NF-kappaB-dependent MMP-9 signal. Eur Rev Med Pharmacol Sci. 2013;17(17):2372–82.

    CAS  PubMed  Google Scholar 

  33. Zhao F-T, Jia Z-S, Yang Q, Song L, Jiang X-J. S100A14 promotes the growth and metastasis of hepatocellular carcinoma. Asian Pac J Cancer Prev. 2013;14(6):3831–6.

    Article  PubMed  Google Scholar 

  34. Wei L, Lee D, Law CT, Zhang M, Shen J, Chin DC, et al. Genome-wide CRISPR/Cas9 library screening identified PHGDH as a critical driver for Sorafenib resistance in HCC. Nat Commun. 2019;10(1):4681.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Gao P, Tchernyshyov I, Chang T, Lee Y, Kita K, Ochi T, et al. c-Myc suppression of miR-23a/b enhances mitochondrial glutaminase expression and glutamine metabolism. Nature. 2009;458(7239):762–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  36. Li B, Cao Y, Meng G, Qian L, Xu T, Yan C, et al. Targeting glutaminase 1 attenuates stemness properties in hepatocellular carcinoma by increasing reactive oxygen species and suppressing Wnt/beta-catenin pathway. EBioMedicine. 2019;39:239–54.

    Article  PubMed  Google Scholar 

  37. Thangavelu K, Chong QY, Low BC, Sivaraman J. Structural basis for the active site Inhibition mechanism of human kid ney-type glutaminase (KGA). Sci Rep. 2014;4:3827.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Matés JM, Campos-Sandoval JA, de Los Santos-Jiménez J, Márquez J. Glutaminases regulate glutathione and oxidative stress in cancer. Arch Toxicol. 2020;94(8):2603–23.

    Article  PubMed  Google Scholar 

  39. Song Z, Jia-Mei W, Jing Y, Da-Lin Z, Bao-Qin L, Jing-Yi J, et al. BAG3 promotes autophagy and Glutaminolysis via stabilizing glutaminase. Cell Death Dis. 2019;10:284.

    Article  Google Scholar 

  40. Hunter T. The age of crosstalk: phosphorylation, ubiquitination, and beyond. Mol Cell. 2007;28(5):730–8.

    Article  CAS  PubMed  Google Scholar 

  41. Shimizu K, Gi M, Suzuki S, North BJ, Watahiki A, Fukumoto S, et al. Interplay between protein acetylation and ubiquitination controls MCL1 protein stability. Cell Rep. 2021;37(6):109988.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Wicker CA, Hunt BG, Krishnan S, Aziz K, Parajuli S, Palackdharry S, et al. Glutaminase Inhibition with Telaglenastat (CB-839) improves treatment response in combination with ionizing radiation in head and neck squam Ous cell carcinoma models. Cancer Lett. 2021;502:180–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Zhu H, Gao W, Li X, Yu L, Luo D, Liu Y, et al. S100A14 promotes progression and gemcitabine resistance in pancreatic cancer. Pancreatology. 2021;21(3):589–98.

    Article  CAS  PubMed  Google Scholar 

  44. Zhang X, Shao S, Li L. Characterization of Class-3 semaphorin receptors, neuropilins and plexins, as therapeutic targets in a Pan-Cancer study. Cancers. 2020;12(7):1816.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Luo Z, Zhu J, Xu R, Wan R, He Y, Chen Y, et al. Exercise-downregulated CD300E acted as a negative prognostic implicati on and tumor-promoted role in pan-cancer. Front Immunol. 2024;15:1437068.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Greene KS, Lukey MJ, Wang X, Blank B, Druso JE, Lin M-CJ, et al. SIRT5 stabilizes mitochondrial glutaminase and supports breast cancer tumorigenesis. Proc Natl Acad Sci USA. 2019;116(52):26625–32.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Tong Y, Guo D, Lin S, Liang J, Yang D, Ma C, et al. SUCLA2-coupled regulation of GLS succinylation and activity counteracts oxidative stress in tumor cells. Mol Cell. 2021;81(11):2303–e168.

    Article  CAS  PubMed  Google Scholar 

  48. Zhu J, Chang R, Wei B, Fu Y, Chen X, Liu H et al. Photothermal nano-vaccine promoting antigen presentation and dendritic cells infiltration for enhanced immunotherapy of melanoma via transde rmal microneedles delivery. Research (Wash D C). 2022;2022:9816272.

  49. Wan R, Chen P, Guo S, Zhu J, Mei J, Mai CW, et al. Editorial: the immunological regulation of extracellular vesicles on chronic diseases. Front Immunol. 2024;15:1442387.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Luo Z, He Z, Qin H, Chen Y, Qi B, Lin J, et al. Exercise-induced IL-15 acted as a positive prognostic implication and tumor-suppressed role in pan-cancer. Front Pharmacol. 2022;13:1053137.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Barreto C, Jandus A. Role of natural products in combating cancer. Cancer Insight. 2022;1(1):35–52.

    Google Scholar 

Download references

Acknowledgements

Not applicable.

Funding

This work was supported by the National Natural Science Foundation of China (82173161, 82473124), the Natural Science Foundation of Shanghai (22ZR1446800), Shanghai Municipal Key Clinical Specialty (shslczdzk00101, shslczdzk06403) and Baoshan District Health Commission Key Subject Construction Project (BSZK-2023-A03).

Author information

Authors and Affiliations

  1. Department of Pharmacology, School of Pharmacy, Fudan University, Shanghai, 201203, P. R. China

    Menghui Wang, Junhui Su, Xinjue Dong, Yuqi Yang, Xinyi Tang, Ruijie Chen & QingQuan Li

  2. Department of Pancreatic Surgery, Zhongshan Hospital, Fudan University, Shanghai, P. R. China

    Yueheng Li

  3. Department of Geriatrics, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, P. R. China

    Ao Liu

  4. Medical Center on Aging of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, P. R. China

    Ao Liu

  5. Department of General Surgery, GI Surgery, Zhongshan Hospital, Fudan University, Shanghai, 200030, P. R. China

    Hongshan Wang

  6. Department of General Surgery, Shanghai Baoshan District Wusong Central Hospital (Zhongshan Hospital Wusong Branch Fudan University), Shanghai, P. R. China

    Hongshan Wang

  7. Baoshan Cancer Center, Baoshan District, Shanghai, P. R. China

    Hongshan Wang

  8. Department of Infectious Diseases, Shanghai Public Health Clinical Center, Fudan University, Shanghai, 201508, P. R. China

    Hong Xiao

Authors
  1. Menghui Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Yueheng Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Junhui Su
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Xinjue Dong
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Ao Liu
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Yuqi Yang
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Xinyi Tang
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Ruijie Chen
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. QingQuan Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  10. Hongshan Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  11. Hong Xiao
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

Menghui Wang: Methodology, Validation, Formal Analysis, Investigation, Data Curation, Writing-Original Draft, Visualization. Yueheng Li, Junhui Su, Xinjue Dong, Ao Liu, Yuqi Yang, Xinyi Tang, Ruijie Chen: Investigation, Visualization. Qingquan Li: Conceptualization, Methodology, Resources, Writing-Original Draft, Writing-Review & Editing, Supervision, Project Administration. Hongshan Wang: Conceptualization, Methodology, Resources, Writing-Original Draft, Writing-Review & Editing, Supervision, Project Administration, Funding Acquisition. Hong Xiao: Conceptualization, Methodology, Resources, Data Curation, Writing-Original Draft, Writing-Review & Editing, Supervision, Project Administration. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to QingQuan Li, Hongshan Wang or Hong Xiao.

Ethics declarations

Ethics approval and consent to participate

All daily animal care and the experimental procedures were approved by Institutional Animal Care and Use Committee of Fudan University. Animal handling was in accordance with ARRIVE Guidelines 2.0. The maximal tumor size was permitted by Institutional Animal Care and Use Committee of Fudan University, no subcutaneous tumors were exceeded 2,000 mm3 in this study. The human HCC tissue microarray was obtained from Zhongshan Hospital, Fudan University, with approval from the Ethics Committee of Zhongshan Hospital, Fudan University. Informed consent was obtained from patients before tissue collection.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s note

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

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Supplementary Figures S1-6.

Supplementary Table 1. Primer sequences.

Supplementary Table 2. Source and characteristics of antibodies in this study.

12967_2025_6333_MOESM4_ESM.xlsx

Supplementary Table 3. Pathological information of patients from tissue microarrays and the IHC scores of S100A14 expression.

12967_2025_6333_MOESM5_ESM.xlsx

Supplementary Table 4. Statistical analysis of barcode frequencies and the percentages of cells expressing those barcodes.

Supplementary Table 5. The detailed results of mass spectrometry for the interacting protein with S100A14.

Supplementary Table 6. The detailed results of mass spectrometry for the post-translational modifications of GLSKGA.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Wang, M., Li, Y., Su, J. et al. Endogenous protein S100A14 stabilizes glutaminase to render hepatocellular carcinoma resistant to sorafenib. J Transl Med 23, 435 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12967-025-06333-5

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12967-025-06333-5

Keywords

Journal of Translational Medicine

ISSN: 1479-5876

Contact us