The human ESCC cell lines, ECA-109 and KYSE-150, were purchased from Sun Yat-sen Memorial Hospital (Guangzhou, China). Additionally, the human normal esophageal cell line, HEEC, was obtained from the BeNa Culture Collection (BNCC359279, BNCC, Beijing, China). These cells were maintained under RIPM-1640 medium (SH30809.01, Hyclone, Logan, UT, USA) supplemented with 10% fetal bovine serum (FBS) (FSP500, ExCell Bio, Shanghai, China), 100$\times$ penicillin-streptomycin liquid (P1400, Solarbio, Beijing, China). Cultivation was carried out at 37 °C with 5% CO₂ atmosphere. All cell lines were validated by STR profiling tested for Mycoplasma contamination by MycoStripe (rep-mys-10, InvivoGen, San Diego, CA, USA).

For proteomic investigation, ECA109 cellular samples were harvested following a 48-hour incubation period post-irradiation, including three treatment conditions: non-irradiated controls (0 Gy), X-ray exposure at 2 Gy, and C-ions irradiation at 2 Gy. Specific experimental procedures were performed as previously described [11]. Following data acquisition, MS/MS results were processed through the Proteome Discoverer 2.2 platform (Waltham, MA, ̵‌USA), employing the Matrix Science’s Search and Correlation Tool (MASCOT) algorithm 2.6 (Matrix Science, London, UK) for protein identification against the UniProt (http://www.uniprot.org) reference database. Proteins were deemed significantly altered if they met both criteria: fold change $\ge$1.2 and p 0.05.

To assess variability within and between groups, Principal Component Analysis (PCA) was conducted on all reliably identified proteins using the factoextra package in R software 4.3.2 (CRAN, Vienna, Austria). The overlapping patterns of differentially expressed proteins (DEPs) were visualized via EVenn (http://www.ehbio.com/test/venn/#/). Heatmaps of overlapping DEPs were generated using the “heatmap” package of R software. Enrichment analyses for Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) were conducted using R’s “clusterProfile” package. The STRING database (https://string-db.org/) was employed to generate protein-protein interaction (PPI) networks. Then, PPI networks were analyzed in Cytoscape 3.9.1 (Cytoscape Consortium, San Diego, CA, USA) via cytoHubba, and top five hub proteins were identified based on centrality. Consensus hub proteins derived from 12 independent algorithms were determined by Venn analysis (http://www.ehbio.com/test/venn/#/). Esophageal cancer data from The Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression (GTEx) were obtained via UCSC Xena (https://xenabrowser.net). Survival analysis used R’s maxstat for optimal mRNA cut-off determination, with ggplot2 for boxplots and Kaplan-Meier curves.

Radiation treatments were conducted using two modalities: C-ions irradiation (100 MeV/u, LET 50 keV/µm) with a 5 mm spread-out Bragg peak was delivered at Heavy Ion Research Facility in Lanzhou (Institute of Modern Physics, CAS, Lanzhou, Gansu, China) using a 10 $\times$ 10 cm field at 1 Gy/min; X-rays exposure (225 kV/13.3 mA) was performed using an X-Rad 225 system (Precision, North Branford, CT, USA) at identical dose rate.

N-acetyl-L-cysteine (NAC) (C8460, Solarbio, Beijing, China) was dissolved in sterilized water and used in a final concentration of 10 mM in medium, which was administered 2 h prior to irradiation. Carbonyl cyanide m-chlorophenylhydrazone (CCCP) (C2008S-3, Beyotime, Shanghai, China) was used in a final concentration of 10 µM in medium.

GenePharma Co., Ltd. (Shanghai, China) commercially synthesized the small interfering RNA (siRNA) for PGK1 downregulation, the negative control (NC) siRNA, and the PGK1 overexpression plasmids. Following the manufacturer’s guidelines to knockdown and overexpress PGK1 in cells. Supplementary Table 1 lists the siRNA sequences used for gene silencing. The plasmid sequence utilized for PGK1 overexpression is detailed in Supplementary Table 2.

RNA was isolated with TRIzol (15596026CN, Invitrogen, Carlsbad, CA, USA) and reverse transcribed using a cDNA reverse transcription kit (11141ES10, Yeason, Shanghai, China) per manufacturer’s protocol. qPCR was performed with a SYBR Green kit (11201ES08, Yeason, Shanghai, China) on a real-time PCR Detection System (ASA-4800, Baiyuan, Suzhou, Jiangsu, China). $\beta$-actin served as the internal reference for normalization. Gene expression was quantified via the 2^{-Δ⁢Δ⁢Ct} method. Primer sequences are provided in Supplementary Table 3.

Cellular proteins were lysed in RIPA buffer (R0010, Solarbio, Beijing, China) supplemented with protease inhibitors (P6730, Solarbio, Beijing, China) and phosphatase inhibitors (P1260, Solarbio, Beijing, China). Protein quantification was performed using a BCA assay kit (PC0020, Solarbio, Beijing, China). Samples were separated by 8%–12% SDS-PAGE and transferred to PVDF membranes (IPVH00010, Millipore, Burlington, MA, USA). After blocking with 5% skimmed milk or BSA, membranes were incubated with primary antibodies (overnight, 4 °C). Subsequently, 1.5 h incubation with HRP-conjugated secondary antibodies. Signal detection was achieved using enhanced chemiluminescence reagent (PE0010, Solarbio, Beijing, China) and visualized with a chemiluminescence detection system (GD50202, Monad, Zhejiang, Jiangsu, China). ImageJ software 1.54f (NIH, Bethesda, MD, USA) was then used to quantify the images. Antibody details are provided in Supplementary Table 4.

Cells (300–6000 cells) were plated in $\mathrm{\Phi }$ 60-mm dishes and cultured (37 °C, 5% CO₂) for 10–14 days. After formaldehyde fixation and 1.0% crystal violet (G1063, Solarbio, Beijing, China) staining, colonies ($>$50 cells) were counted by ImageJ software 1.54f. Clonogenic surviving fractions were calculated as (colonies counted/seeded cells) $\times$ 100, and normalized to non-irradiated controls. Survival curves were generated using the linear quadratic model (SF = e^(-($\alpha$D+$\beta$d^2))) in GraphPad Prism 9.0 (GraphPad Corp., San Diego, CA, USA). Relative biological effectiveness (RBE) values were determined as the ratio of X-rays to C-ions doses required for 10% survival.

2000 cells/well were seeded in a 96-well plate. 10 µL cell counting kit-8 (CCK-8) reagents (40203ES, Yeason, Shanghai, China) were introduced into each well, and the plate was subsequently maintained for 2 h. For 3 consecutive days, the absorbance at 450 nm was recorded using a microplate reader (Infinite M200 Pro, Tecan, Mannedorf, Zurich, Switzerland).

The 5-Ethynyl-2^′-deoxyuridine (EdU) detection was performed using the Cell-Light™ EdU Apollo Kit per manufacturer’s instructions (C10310-1, RIBOBIO, Guangzhou, Guangdong, China). Cells in 12-well plates were labeled with EdU for 2 h, and then stained in the dark for 30 min. Subsequently, nuclei were re-stained with Hoechst 33,342 for 30 min in the dark. Images were acquired by fluorescence microscopy (DP74, Olympus, Shinjuku, Tokyo, Japan).

The Annexin V-FITC/PI apoptosis detection kit (MA0220, MeilunBio, Dalian, Liaoning, China) was used to evaluate apoptosis. Briefly, the harvested cell were treated with annexin V-FITC and PI in 1$\times$ binding buffer (15 min, dark). Cell cycle analysis required 75% ice-cold ethanol fixation and storage at –20 °C $\ge$ 2 h. Prior to analysis, fixed cells were centrifuged and washed according to the protocol of the Cell Cycle Staining Kit (CCS012, MULTI SCIENCES, Hangzhou, Zhejiang, China). Then, cells were stained with DNA staining solution in darkness for 30 min. Intracellular ROS was measured by the fluorescent probe DCFH-DA (S0033S, Beyotime, Shanghai, China), which was diluted in serum-free medium and used to incubate cells for 30 min at 37 °C. Mitochondrial membrane potential (MMP) was evaluated using Rhodamine 123 staining (C2008S, Beyotime, Shanghai, China). Cells were incubated in a serum-free medium containing the dye solution for 30 min under dark conditions. The staining results were measured by a flow cytometer (Model 100370, Amnis/Merck Millipore, Seattle, WA, USA). Data were analyzed using IDEAS 6.0 (Amnis/Merck Millipore, Seattle, WA, USA) or FlowJo 7.6 (Tree Star Inc., Ashland, OR, USA).

We utilized GraphPad Prism 9.0 for both statistical computations and visualizations. For direct comparisons between just two groups, we relied on an unpaired t-test. However, for situations requiring analysis across several groups, we turned to one-way or two-way analysis of variance, complemented by post-hoc Bonferroni test. The log-rank test was employed for survival analysis. The statistical comparisons were all two-tailed, with p-values below 0.05 indicating statistical significance. Experimental data, derived from at least three independent replicates, are presented as means $\pm$ standard deviation.

First, we evaluated the consistency of the sample. The PCA results showed high aggregation among repeated samples and significant differences among different samples, indicating good quantitative repeatability of the samples (Fig. 1A). A total of 521 DEPs were detected in the 2 Gy C-ions group compared to the control group, 412 DEPs between the 2 Gy C-ions and 2 Gy X-rays groups, and 352 DEPs between the 2 Gy X-rays group and the control group. The Venn diagram shows 33 overlapping DEPs in the three groups (Fig. 1B). The protein expression heatmap shows the expression levels of overlapping DEPs in different groups and samples (Fig. 1C). We conducted GO and KEGG enrichment analyses on these 33 DEPs to determine their functional characteristics. The findings from the GO enrichment analysis indicated that merely four terms exhibited significant enrichment within the molecular function (MF) category, predominantly related to disulfide oxidoreductase activity, oxidoreductase activity, and protein-disulfide reductase activity. The biological process and cellular components were not significantly enriched (Fig. 1E). KEGG enrichment analysis suggested that 33 DEPS were predominantly associated with carbon metabolism and the pentose phosphate pathway (Fig. 1E). To screen for hub proteins, a PPI network of the DEPs was constructed, with 11 nodes and 10 edges (Fig. 1F). The top five hub proteins were identified using twelve algorithms of the cytoHubba plug-in (Supplementary Table 5). Through the Venn diagram overlap of the hub proteins derived from the twelve algorithms, we identified three shared hub proteins, namely, PGK1, thioredoxin (TXN), and glucose-6-phosphate dehydrogenase (G6PD) (Fig. 1D).

Fig. 1.

Screening of the key proteins. (A) PCA of all identified proteins. (B) The Venn diagram shows the overlapped DEPs. (C) A heatmap shows the expression levels of overlapped DEPs in each group of samples, with the proteins highlighted in red are the candidate hub proteins subsequently screened. (D) Venn diagram shows the shared proteins of the top 5 hub proteins of cytoHubba’s twelve algorithms. (E) The bar chart visualizes the GO and KEGG enrichment outcomes of overlapping DEPs. (F) The PPI network map generated by the STRING online website shows the relationship between overlapping DEPs interactions. PCA, principal component analysis; DEPs, differentially expressed proteins; GO, gene ontology; KEGG, kyoto encyclopedia of genes and genomes; PGK1, phosphoglycerate kinase 1; TXN, thioredoxin; G6PD, glucose-6-phosphate dehydrogenase.

Subsequently, we examined the mRNA levels of PGK1, TXN, and G6PD in normal esophageal tissues and esophageal tumor tissues. As displayed in Fig. 2D–F, the tumor cohort exhibited notably higher mRNA levels for TXN, G6PD, and PGK1 compared to the normal group (p 0.001), and their expression levels were greater in ESCC compared to EAC. We further evaluated the prognostic effect of TXN, G6PD, and PGK1 mRNA expression levels in ESCA patients. The analysis revealed no significant association between TXN levels and overall survival (OS) in ESCA patients, including both esophageal adenocarcinoma (EAC) and ESCC (p $>$ 0.05) (Fig. 2A). Elevated G6PD expression associated with unfavorable prognosis in ESCC patients (p 0.05), while showing no significant impact on EAC (p $>$ 0.05) (Fig. 2B). In both ESCC and EAC, high PGK1 expression significantly reduced OS in these patients (p 0.05) (Fig. 2C). Consequently, PGK1 was selected as the key molecule for subsequent experiments.

Fig. 2.

Kaplan-Meier survival analysis and mRNA expression of candidate genes. Kaplan-Meier survival analysis showing the correlation between TXN (A), G6PD (B), and PGK1 (C) mRNA expression and OS of ESCA, EAC, and ESCC patients in TCGA cohort. The mRNA expression levels of TXN (D), G6PD (E), and PGK1 (F) in normal esophageal tissues, as well as in ESCA tissues, EAC tissues, and ESCC tissues. EAC, esophageal adenocarcinoma; ESCA, esophageal carcinoma; ESCC, esophageal squamous cell carcinoma. **p 0.01; ***p 0.001.

Next, we found that the mRNA and protein levels of PGK1 was significantly higher in ESCC cell lines (ECA-109 and KYSE-150) compared to normal esophageal normal cell (HEEC) (p 0.01). Specifically, the expression of PGK1 mRNA and PGK1 protein (Supplementary Fig. 1A) and protein (Fig. 3A) was greater in ECA-109 cells compared to KYSE-150 cells (p 0.01). We carried out Western Blotting experiments to explore the effects of radiation type and dose on PGK1 levels. The findings indicated that after lower doses of X-rays radiation, the PGK1 expression levels of ECA-109 and KYSE-150 cells increased significantly (p 0.05), while they decreased significantly after higher doses radiation (p 0.05). For C-ions, PGK1 expression in ECA-109 cells increased significantly after lower doses radiation and decreased significantly after higher doses radiation (p 0.05); however, unlike ECA-109 cells, PGK1 expression in KYSE-150 cells did not increase significantly after lower doses of C-ions radiation. Although PGK1 expression was inhibited after higher doses of both X-rays and C-ions radiation, X-rays induced a more significant increase in PGK1 expression after lower doses radiation (Fig. 3B,C).

Fig. 3.

Analysis of PGK1 protein expression across various cell lines and the impact of varying doses of X-rays or C-ions on clonal survival and PGK1 protein expression in ESCC cells. Protein levels of PGK1 were analyzed in HEEC, ECA-109, and KYSE-150 cell lines (A). PGK1 protein levels were assessed in ESCC cells after different doses of X-rays (B) and C-ions (C). Cloning images and cell survival curves of ECA-109 (D) and KYSE-150 (E) cells after exposure to X-rays and C-ions. Western blot evaluated the efficiency of PGK1 knockdown (F) and PGK1 overexpression (G) in ESCC cells. C-ions, Carbon ions; RBE, relative biological effectiveness. *p 0.05; **p 0.01; ***p 0.001; and ns, not significant.

Subsequently, we assessed the cell proliferation ability after exposure to different LET radiation. The findings indicated a reduction in cell clone numbers as the irradiation dose increased, particularly notable in the C-ions group (Fig. 3D,E). Specifically, the RBE values for ECA-109 and KYSE-150 cells were determined to be 2.05 and 2.14, respectively. Further analysis revealed that for the X-rays group, the irradiation dose required for cell survival fraction below 50% exceeded 2 Gy, while for C-ions, this threshold was above 1 Gy. In light of this, we selected biologically equivalent doses (4 Gy X-rays and 2 Gy C-ions) for further research in this investigation to ensure biological comparability. Additionally, based on $\alpha$ values from the Linear-Quadratic (LQ) model, ECA-109 cells showed relative resistance to radiation, while KYSE-150 cells were comparatively sensitive (Supplementary Fig. 2).

Then, we knocked down and overexpressed PGK1 in two ESCC cells, with the knockdown and overexpression efficiency were confirmed via qRT-PCR (Supplementary Fig. 1B,C) and Western blotting (Fig. 3F,G). A plate colony formation assay revealed that both PGK1 knockdown and IR inhibited the proliferation of ESCC cells (p 0.01). Compared to the use of either modality alone, the combination of PGK1 knockdown and IR had significantly stronger antiproliferative effects (p 0.01) (Fig. 4A,C). Overexpression of PGK1 resulted in an enhanced proliferation rate and attenuated the inhibitory impact of IR on cell proliferation (p 0.01) (Fig. 4B,D). These were validated by the CCK8 assay (Fig. 4E,F) and EdU assay (Fig. 4G,H) (p 0.05). Therefore, overexpressing PGK1 can promote ESCC cells growth, whereas knocking down PGK1 enhance the antiproliferative effects induced by IR.

Fig. 4.

PGK1 impacts the regulation of IR concerning the ESCC cells proliferation. Plate colony formation assays were conducted to assess clone formation following PGK1 knockdown (A,C) and overexpression (B,D) after IR exposure. Cell viability of ESCC cells with PGK1 knockdown (E) or overexpression (F) was assessed using the CCK8 assay at 24, 48, and 72 h post-IR. An EdU incorporation assay was conducted 48 h post-IR to assess the DNA replication of ECA109 and KYSE150 cells with PGK1 knockdown or overexpression (G,H). CCK8, cell counting kit 8; IR, ionizing radiation; NC, negative control. *p 0.05; **p 0.01; ***p 0.001. Scale bar, 100 µm.

The flow cytometry results disclosed that the apoptosis level in the combination treatment group was notably higher than that in the single treatment groups (p 0.05) (Fig. 5A). However, overexpression of PGK1 reduced IR-induced apoptosis (p 0.01) (Fig. 5B). These findings display that PGK1 inhibition synergistically induce cell apoptosis with IR.

Fig. 5.

PGK1 impacts the regulation of IR concerning the ESCC cells apoptosis. The apoptosis levels in ESCC cells with PGK1 knockdown (A) or overexpression (B) after 48 h exposure to IR. *p 0.05; **p 0.01; ***p 0.001.

The results showed that regardless of whether IR exposure, PGK1 deficiency elevated the proportion of cells in the G0/G1 phase and reduced those in the G2/M phase (p 0.05) (Fig. 6A,B). Overexpression of PGK1 significantly elevated the percentage of cells in the G2/M phase while reducing those in the G0/G1 phase (p 0.05) (Fig. 6C,D). In the ECA-109 cell line, both with and without IR, PGK1 knockdown markedly reduced the S-phase cell proportion (p 0.05), whereas PGK1 overexpression enhanced it. In KYSE-150 cells, PGK1 knockdown alone diminished S-phase cells, while overexpression notably increased S-phase cell proportion (p 0.01). When exposed to X-rays, both PGK1 knockdown and overexpression resulted in decreased S-phase proportions in KYSE-150 cells (p 0.01). Under C-ions, PGK1 expression status did not significantly influence S-phase distribution (Fig. 6B,D).

Fig. 6.

PGK1 impacts the regulation of IR concerning the cell cycle of ESCC cells. Frequency histograms and statistical analysis charts illustrate the impact of downregulating (A,B) or overexpressing PGK1 (C,D) on cell cycle after 48 h exposure to IR. *p 0.05; **p 0.01; ***p 0.001; and ns, not significant.

IR can cause damage to cells by inducing a substantial increase in ROS levels [12]. Mitochondria serve as the primary source of ROS in mammalian cells. Excessive ROS can lead to mitochondrial dysfunction, which subsequently decreases the MMP and promotes further ROS generation, thereby establishing a detrimental feedback loop that exacerbates cellular damage [13]. We investigated ROS and MMP levels to assess if PGK1 modulates the radiosensitivity of ESCC cells by affecting mitochondrial dysfunction through ROS. The examination of the flow cytometry data indicated that both IR and PGK1 deficiency elevated the fluorescence intensity of DCF staining, with the effect being more significant in the combined group (p 0.05) (Fig. 7A,B). Similarly, both IR and PGK1 deficiency reduced MMP of ESCC cells, and the combined group had a more significant effect (p 0.05) (Fig. 7C,D). Following the use of NAC, an ROS scavenger, the damage to the MMP caused by PGK1 deficiency and IR was restored (Fig. 7E). CCCP, as an uncoupling agent of mitochondrial proton carriers, can destroy the mitochondrial membrane and causing mitochondrial dysfunction. Our findings indicate that overexpressing PGK1 mitigated apoptosis induction (p 0.05) (Fig. 7F) and inhibited proliferation (p 0.05) (Fig. 7G,H) induced by CCCP. Consequently, PGK1 deficiency and IR have synergistic effects on intracellular ROS accumulation and exacerbate mitochondrial damage.

Fig. 7.

PGK1 deficiency-initiated ROS accumulation and induced mitochondrial dysfunction. Figures display cytograms and statistical charts illustrating ROS levels in ESCC cells with PGK1 knockdown after 12 h exposure to IR (A,B). Figures display cytograms and statistical analyses of MMP levels of ESCC cells with PGK1 knockdown after 24 h exposure to IR (C,D). The changes in MMP of ESCC cells following PGK1 knockdown or exposure to IR after NAC treatment (E). Apoptotic levels (F) and cloning capacity (G,H) in ESCC cells with PGK1 overexpressed after CCCP treatment. ROS, reactive oxygen species; MMP, mitochondrial membrane potential; NAC, N-acetyl-L-cysteine; CCCP, carbonyl cyanide m-chlorophenylhydrazone. *p 0.05; **p 0.01; ***p 0.001; and ns, not significant.

To investigate whether PGK1 modulates the radiosensitivity through ROS, we used ROS scavengers. EdU (Fig. 8A,B) and CCK8 (Fig. 8D) assays showed that NAC partially rescued the impact of PGK1 deficiency and IR on ECA-109 cell proliferation. Similarly, NAC alleviated the apoptosis in ECA-109 cells caused by PGK1 knockdown and IR exposure (Fig. 8C). NAC also had a similar effect on the KYSE-150 cells (Supplementary Fig. 3).

Fig. 8.

Following NAC treatment, ECA-109 cell proliferation and apoptosis were assessed with PGK1 knockdown or IR exposure. And the impact of PGK1 on the Akt/mTOR pathway. The DNA replication activity of ECA-109 cells with PGK1 knockdown or IR was detected by EdU incorporation assay following the use of NAC (A,B). The viability of ECA-109 cells with PGK1 knockdown or exposed to IR was determined by CCK8 assay after NAC treatment (D). Apoptosis levels of ECA-109 cells were assessed after PGK1 knockdown or IR with NAC treatment (C). Protein expression levels of Akt, pAkt, mTOR, and pmTOR were analyzed in ECA-109 cells subjected to PGK1 knockdown and IR (E). Protein expression levels of Akt, pAkt, mTOR, and pmTOR were analyzed in ECA-109 cells with PGK1 overexpression and IR (F). *p 0.05; **p 0.01; ***p 0.001; and ns, not significant. Scale bar, 100 µm. Akt/mTOR, protein kinase B/mammalian target of rapamycin.

Inhibition of the Akt/mTOR pathway can effectively augment radiosensitivity [14]. Next, we explored whether PGK1 affects radiosensitivity through the Akt/mTOR pathway. Changes in pathway proteins were detected by Western blotting. Knockdown of PGK1 dramatically curbed the phosphorylation of Akt and mTOR, and intensified the inhibition of this pathway by IR (p 0.05) (Fig. 8E). PGK1 overexpression alleviated the inhibitory effects of IR on the Akt/mTOR pathway (p 0.05) (Fig. 8F). Such findings imply that PGK1 may regulate the radiosensitivity of ESCC cells via the Akt/mTOR pathway.