From the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/), the ESCC gene expression datasets GSE20347, GSE23400, GSE53625, GSE67269 and GSE161533) were downloaded (15–18). Furthermore, information on the gene expression and survival of 95 patients with ESCC was gathered from The Cancer Genome Atlas (TCGA; http://cancergenome.nih.gov/). R v4.2.1 software was used for analysis in the current study (19,20). All transcriptome data underwent the necessary preprocessing before statistical analysis, including averaging multiple expression values of the same gene in each dataset, followed by logarithmic transformation and normalization of the dataset using the R package ‘limma’ (21).

Furthermore, paraffin blocks of cancerous and healthy tissues adjacent to the tumor were collected from patients with ESCC who visited The Second Affiliated Hospital of Zhengzhou University (Zhengzhou, China) between January 2022 and March 2023. The inclusion criteria were as follows: All patients had ESCC, were aged 18–75 years, all surgical specimens had been taken before any neo-adjuvant anti-tumor treatment and all patients had corresponding informed consent. Exclusion criteria were the presence of other malignant tumors, psychological abnormalities or contraindications to anti-tumor treatment.

ECA109 and KYSE450 cells were obtained from the American Type Culture Collection, while KYSE150 cells were purchased from Shanghai Zhongqiao Xinzhou Biotechnology Co., Ltd. The culture medium used contained 89% RPMI 1640 medium (cat. no. ZQ0449; Zhongqiao Xinzhou Biotechnology Co., Ltd.), 10% fetal bovine serum (cat. no. 04-001-1ACS; Biological Industries), 1% 100 IU/ml penicillin and 100 g/ml streptomycin (cat. no. CSP006; Zhongqiao Xinzhou Biotechnology Co., Ltd.). Cells were cultured at 37°C and 5% CO₂ saturated humidity.

First, the ‘limma’ package was used to perform differential analysis on the transcriptome data of the datasets. Next, patients with ESCC in the GSE53625 and the TCGA dataset containing clinical data were divided into two groups using the median expression value of NLRX1: GSE53625 high: n=89, low: n=90; and TCGA high: n=47, low: n=48. Subsequently, Kaplan-Meier survival analysis was performed on them.

The GSE53625 dataset, which had the largest sample size, was used for subsequent analysis using the Wilcoxon signed-rank or χ² tests to analyze the relationship between NLRX1 and clinical reference data. Subsequently, a receiver operating characteristic (ROC) curve was plotted to evaluate the diagnostic value of NLRX1, and Cox regression analysis was used to help judge the prognostic value of NLRX1. Finally, the ‘survival’ package was used for regression analysis. A nomogram was constructed and a calibration curve was drawn to verify its reliability.

First, based on the transcriptome data of the GSE53625 dataset, the correlation between NLRX1 and the expression of other genes was analyzed, and genes highly correlated with NLRX1 were identified with |correlation coefficient|>0.5 and P<0.05 as selection criteria. Next, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG)-enrichment analyses were performed on these genes, while Gene Set Enrichment Analysis (GSEA) and Gene Set Variation Analysis (GSVA) were used to explore the pathways enriched in the low expression group of NLRX1 in GSE53625. This step utilized the ‘clusterProfiler’, ‘GSEABase’ and ‘GSVA’ packages.

Based on the data from GSE53625, patients were divided into high and low groups according to the median expression value of the NLRX1, and immune-related functional analysis was performed using the ‘GSEABase’ and ‘GSVA’ packages (22,23). Next, the ‘Corrplot’ package was used for immune checkpoint correlation analysis, while the ‘CIBERSORT’ package was used to analyze the degree of immune infiltration in samples and screen out meaningful samples for differential analysis and correlation testing. Finally, from the Genomics of Drug Sensitivity in Cancer website (https://www.cancerrxgene.org/), drug information was downloaded and differences in drug sensitivity between patients with high and low expression of NLRX1 were analyzed.

Immunohistochemical staining was performed as previously described (24). In brief, paraffin-embedded tissues were dewaxed, hydrated and subjected to antigen retrieval and blocking of endogenous peroxidase prior to incubation with anti-NLRX1 antibody (1:200 dilution; cat. no. ab107611; Abcam) at room temperature for 30 min. Next, a conjugated secondary antibody was added (1:100 dilution; cat. no. SA00001-2; Proteintech Group, Inc.) and incubated at room temperature for 30 min, followed by staining with DAB, hematoxylin and alkaline bluing solution. Next, a neutral resin was applied for sealing after dehydration before observing the slides under an optical microscope at ×200 magnification and acquiring images.

A semi-quantitative scoring technique based on the percentage of positive cells and staining intensity was used to grade the immunohistochemical staining results as follows: 0, negative staining; 1, pale yellow (weak); 2, yellow (moderate); and 3, brown (strong). The positivity score was assigned as 0 for <5% positive cells; 1 for 6–25% positive cells; 2 for 26–50% positive cells; 3 for 51–75% positive cells; and 4 for >75% positive cells. The two aforementioned scores were then multiplied and 0 was considered to indicate negative expression, <5 low expression and ≥5 high expression. Two pathologists independently evaluated the immunohistochemical staining.

The plasmids (pcDNA3.1-NLRX1) and small interfering RNAs (siRNAs) used in the present study were produced by TsingKe Biological Technology and plasmid extraction was carried out in line with the instructions provided by the manufacturer of FastPure EndoFree Plasmid Mini Kit (cat. no. DC203-01; Vazyme Biotech Co., Ltd.). Approximately 4×10⁵ KYSE450 cells or 3×10⁵ EAC109 cells were seeded into 6-well plates and transfection was performed the following day when the cell density reached 70–80%. Transfection was carried out according to the manufacturer's instructions of Lipo8000 transfection reagent (cat. no. C0533; Beyotime Institute of Biotechnology) and subsequent experiments were conducted within 48 h. The siRNA sequences were as follows: siNLRX1 forward, 5′-GGACUACUACAACGAUGAUTT-3′ and reverse, 5′-AUCAUCGUUGUAGUAGUCCTT-3′; and negative control (NC) forward, 5′-UUCUCCGAACGUGUCACGUTT-3′ and reverse, 5′-ACGUGACACGUUCGGAGAATT-3′.

RIPA lysis buffer (cat. no. R0010; Solarbio Group, Inc.) was used to lyse cells, and protein was quantified using the BCA method. Total protein (20 µg of protein was added to each lane) was separated by 7.5% SDS-PAGE prepared according to the instructions of the Epizyme PAGE Gel Express Preparation Kit (cat. no. PG111; Epizyme, Inc.; Ipsen Biopharmaceuticals, Inc.). The electrophoresis and PVDF membrane transfer times were adjusted according to the molecular weight of the proteins. A fastblocking solution (cat. no. G2052; Wuhan Servicebio Technology Co., Ltd.) was used to block the membrane at room temperature for 10 min, while an antibody diluent (cat. no. BMU103-CN; Abbkine Scientific Co., Ltd.) was used to dilute the primary antibodies to an appropriate concentration prior to incubation at 4°C overnight. Next, the membrane was incubated with a secondary antibody at room temperature for 2 h. The expression of proteins was then visualized by enhanced chemiluminescence agents (cat. no. BMU102-CN; Abbkine Scientific Co., Ltd.). Anti-NLRX1 (1:1,000 dilution; cat. no. ab107611; Abcam), anti-PI3K/AKT signaling pathway panel (1:1,000; cat. no. ab283852; Abcam) and anti-GAPDH (1:10,000; cat. no. 10494-1-AP; Proteintech Group, Inc.) were used as primary antibodies. The phosphorylation sites of phosphorylated (p)-AKT were S472, S473 and S474, and anti-GAPDH was used as a loading control for normalization. HRP-conjugated Affinipure goat anti-rabbit IgG (H+L) (1:2,000; cat. no. SA00001-2; Proteintech Group, Inc.) was used as the secondary antibody.

For the cell-proliferation experiment, cells transfected on a 6-well culture plate were collected and 2,000 cells per well were added to a fresh 96-well culture plate. When the cells had adhered to the wall after 6 h of incubation, 10 µl Cell Counting Kit-8 (CCK-8) reagent (cat. no. BMU106-CN; Abbkine Scientific Co., Ltd.) was added. After 2 h, the optical density (OD) values at 450 nm were measured by a microplate reader (Thermo Fisher Scientific, Inc.) and considered as time=0 h. Next, the OD values at 450 nm were measured at 24, 48 and 72 h.

At the back of a 6-well culture plate, three horizontal lines were drawn in advance with a black pen, while three vertical lines were drawn on the 6-well culture plate with a pipette tip at 48 h post-transfection (cell confluence rate reached >90%) (25). Cells were cultured in medium containing 2% serum (26) and the migration of cells at the intersection of the horizontal and vertical lines was recorded at 0, 24, 48 and 72 h. Results were calculated as Mobility (%)=(A0-AN)/A0, where A0 represents the initial wound width and AN represents the remaining wound width at the metering point.

In the Transwell migration assay, cells were collected and resuspended in serum-free medium and 40,000 cells/200 µl were added to each upper chamber of a Transwell plate (8.0-µm pore size; Corning, Inc.), while 600 µl complete medium was added to each lower chamber. After 36 h, the cells were fixed at room temperature for 30 min with 4% paraformaldehyde and then stained at room temperature for 10 min with crystal violet. After staining, the cells at the top of the pore filter were gently wiped off with a wet cotton swab. The number of migrated cells was observed under a microscope. For the invasion assay, the culture medium was replaced with serum-free culture medium for starvation treatment 1 day in advance. Next, Matrigel^® (cat. no. 356234; BD Biosciences) was diluted with serum-free medium and 100 µl of this solution was added to each well. After incubation at 37°C for 1 h, any uncured diluent was aspirated and the next steps were the same as for the migration assay.

Approximately 1×10⁵ KYSE450 cells or 7.5×10⁴ ECA109 cells into were inoculated in each well of a 24-well plate. According to the instructions of the manufacturer of the apoptosis kit (cat. no. KTA0002; Abbkine Scientific Co., Ltd.), a mixture was prepared at a ratio of apoptosis inducer to complete culture medium of 1:3,000. After 48 h of transfection, 1 ml mixture was added to KYSE450 or ECA109 cells cultured in a 24-well culture plate. After 24 h of induction, Annexin V/PI staining reagent was added and the fluorescence emission of the sample on the slide was observed using a fluorescence microscope. Green fluorescence (Annexin V) represents cell membrane staining and red fluorescence [propidium iodide (PI)] represents cell nuclear staining. Cells exhibiting only green fluorescence represented early apoptotic cells, while cells exhibiting only red fluorescence represented necrotic cells and cells exhibiting both red and green fluorescence represented late apoptotic cells. Furthermore, the cells treated with the above-mentioned kit were prepared into a suspension and analyzed by flow cytometry (BeamCyte Flow Cytometer; Bidake Biotechnology Co., Ltd.), with 10,000 effective cells recorded each time to detect the apoptosis rate, and the software used was CytoSYS v1.0 (Bidake Biotechnology Co., Ltd.).

SPSS 26.0 (IBM Corp.) was used for statistical analysis. Each experiment was performed as 3 repeats. Continuous variables were presented as the mean ± standard error of the mean. To determine the significance of the difference between the two groups, an unpaired Student's t-test, Fisher's exact test or χ² test was employed as appropriate. If the variances within the groups were not homogenous, nonparametric Mann-Whitney U or Wilcoxon signed-rank tests were used. P<0.05 was considered to indicate a statistically significant difference.

Data mining revealed that, in contrast to normal adjacent tissue, NLRX1 was expressed at lower mRNA levels in ESCC tissue (Fig. 1A). To determine whether there were changes in the expression of NLRX1 at the protein level in 36 pairs of cancerous and adjacent normal tissues [24 males and 12 females; 11 patients aged <65 years and 25 patients aged ≥65 years; median age, 68 years (range, 18–75 years), immunohistochemistry was employed (Fig. 1B). The results showed that NLRX1 protein expression was lower in cancer tissues (19 cases of low expression) than in adjacent normal tissues (10 cases of low expression), with a statistically significant difference (χ²=4.677, P=0.031, Table I).

Among the KYSE450, KYSE150 and ECA109 cell lines, NLRX1 expression was higher in KYSE450 cells, while the expression level of NLRX1 in ECA109 cells was lower (P<0.05 vs. KYSE150; Fig. 1C). Therefore, in subsequent experiments, NLRX1 expression was knocked down in KYSE450 cells and overexpressed in ECA109 cells.

Patients with low NLRX1 levels had shorter survival times in the GSE53625 and TCGA datasets (Fig. 2A). An association between NLRX1 and the pathological parameter of tumor grade was found in patients with ESCC (Fig. 2B). In the ROC curve, the area under the curve of NLRX1 expression was 0.933, indicating that NLRX1 has a good diagnostic value (Fig. 2C). Univariate and multivariate Cox regression analyses showed that NLRX1 was an independent prognostic factor for OS with a hazard ratio of 0.726 [95% confidence interval (CI), 0.574–0.919] and 0.770 (95% CI, 0.596–0.995), respectively (Fig. 2D). Next, a nomogram was created based on the NLRX1 expression and clinical data contained in the GSE53625 dataset as parameters to predict the prognosis of patients with ESCC. The calibration curves of OS showed good consistency between the predicted OS and the observed OS, indicating that the predicted results of the nomogram are in good agreement with the actual results (Fig. 2E).

Based on co-expression analysis, 181 genes were found to be highly correlated with NLRX1 (Table SI), and Circos plots were used to show the 6 genes that were most positively correlated with NLRX1 and the 5 genes that were most negatively correlated with NLRX1 (Fig. 3A). In the KEGG analysis, NLRX1-related genes were mainly enriched in ‘phagosome’, ‘arachidonic acid metabolism’ and ‘endocytosis’, while in the GO analysis, they were mainly enriched in ‘skin development’, ‘cornified envelope’ and ‘structural constituent of skin epidermis’ (Fig. 3B).

Next, the tumor samples of GSE53625 were divided into two groups according to the median expression value of NLRX1, and GSEA and GSVA were performed. The results showed that pathways/functional terms including ‘taste transduction’, ‘arrhythmogenic right ventricular cardiomyopathy’, ‘melanoma’, ‘focal adhesion’, ‘small cell lung cancer’, ‘antigen processing and presentation’, ‘prostate cancer’, ‘stem cell differentiation’, ‘polymerase II specific DNA binding transcription factor binding’ and ‘histone binding’ were enriched in the NLRX1 low expression group (n=90) (P<0.05; Fig. 3C and D).

To investigate whether NLRX1 is immune-related in ESCC, an immune-related functional analysis was first performed. The findings demonstrated that in 179 ESCC samples, the group with low NLRX1 expression (n=90) had lower scores for CC chemokine receptors (CCR), dendritic cells (DCs), plasma-like dendritic cells (pDCs) and T-cell co-stimulation (Fig. 4A, P<0.05; Table SII contains definitions of the terms). Afterward, 85 meaningful samples (P<0.05) were analyzed using the ‘CIBERSORT’ package for immune infiltration analysis. According to the median expression value of NLRX1, they were divided into two groups. In the differential analysis and correlation testing, the degree of natural killer (NK) cells activated, monocytes and macrophages M0 infiltration decreased in the low NLRX1 group (n=43) and was highly positively correlated with NLRX1 expression (Fig. 4B and C). Immunological checkpoint analysis showed that TNFRSF25, TNFSF9, TMIGD2, CD244, TNFRSF18, HHLA2, PDCD1, TNFRSF4, CD40LG and CD86 were positively correlated with NLRX1, while ICOSLG, TNFSF15, CD80, CD276, BTLA, CD200R1 and CTLA4 were negatively correlated with NLRX1 (Fig. 4D, P<0.05). Drug sensitivity analysis revealed that the group with low NLRX1 expression was more resistant to 5-fluorouracil and more sensitive to irinotecan (Fig. 4E).

In order to further explore the biological function of NLRX1 in ESCC, in vitro experiments were conducted. As confirmed by WB, the expression of NLRX1 was knocked down in KYSE450 cells (Fig. 5A). The results of the CCK8 assay showed that knockdown of NLRX1 significantly promoted the growth of KYSE450 cells (Fig. 5B, P<0.05). Furthermore, the scratch wound-healing and Transwell assays showed that silencing NLRX1 expression significantly promoted the migration and invasion of KYSE450 cells (Fig. 5C and D, P<0.05). The fluorescence microscopy and flow cytometry results of the apoptosis experiment also showed that knockdown of NLRX1 significantly reduced the apoptosis of KYSE450 cells (Fig. 5E, P<0.05).

In ECA109 cells, NLRX1 was overexpressed (Fig. 6A). According to our findings, upregulation of NLRX1 expression in ECA109 cells significantly inhibited cell growth, migration and invasion, and significantly increased apoptosis (Fig. 6B-E, P<0.05).

To further explore the mechanism by which NLRX1 regulates the growth and development of ESCC, we experimentally analyzed the effects of NLRX1 on the PI3K/AKT pathway. According to the WB data, silencing NLRX1 in KYSE450 significantly activated the PI3K/AKT pathway (Fig. 7A, P<0.05). By contrast, upregulation of NLRX1 in ECA109 significantly inhibited the activation of the PI3K/AKT pathway (Fig. 7B, P<0.05).