The present study consecutively enrolled 38 patients with HG-GEP NENs who underwent diagnostic or therapeutic procedures (biopsy, puncture, surgery and/or consultation) at Beijing Luhe Hospital, Capital Medical University (Beijing, China) between January 2004 and May 2022. Inclusion criteria were as follows: i) Histopathologically confirmed HG-GEP NENs according to the WHO Classification Of Tumours Of The Digestive System (5th Edition, 2019) (3); ii) availability of complete clinical data with a follow-up duration of >1 year; and iii) availability of FFPE tumour tissue samples obtained after 2012 for molecular analysis. Patients not meeting all these criteria were excluded. The FFPE tissues obtained during routine clinical practice throughout this period were utilized for analysis. The cohort consisted of 9 patients with NET G3 (7 male patients and 2 female patients; median age, 66.0±12.4 years). In addition, the cohort also consisted of 29 patients with NEC who were further stratified into: i) LCNEC, 11 cases (5 male patients and 6 female patients; median age, 65.0±12.1 years); and ii) SCNEC, 18 cases (10 male patients and 8 female patients; median age, 65.0±9.6 years). The overall population included 22 male patients and 16 female patients (age range, 43-86 years; median age, 65.0±10.6 years). All patients were independently reviewed and classified by three senior pathologists according to the diagnostic criteria of the WHO classification of tumours of the digestive system (5th edition, 2019) (3). Any diagnostic discrepancies were resolved through multitiered consensus discussions (Fig. 1B). Pathological grading was performed based on comprehensive histopathological evaluation. Clinical follow-up data regarding recurrence, metastasis and survival status were obtained through telephone interviews, with the follow-up period extending from the date of initial tumour diagnosis until patient death. Complete follow-up data were available for 35 patients, while 3 cases were lost to follow-up. A total of 23 deaths were documented. Ethics approval was provided by the Institutional Review Board of Beijing Luhe Hospital, Capital Medical University (approval no. 2024-LHKY-107-01).

Soldevilla et al (32) conducted transcriptome analyses via next-generation sequencing on a cohort comprising 84 tumour tissue samples from NENs of pulmonary and GEP origin. The resulting gene dataset (GSE211485) was made publicly available through the GEO database (https://www.ncbi.nlm.nih.gov/geo/). In that study, 15 patients diagnosed with GEP NET-G1 and GEP NET-G2 were assigned to the LG-GEP NENs group, whereas 2 patients diagnosed with GEP-NEC were assigned to the HG-GEP NENs group. The present study analysed this dataset and DEG expression was assessed using R Studio (version 2023.12.1+402; Posit Software, PBC, Boston, MA, USA) and DESeq2 (version 1.40.2; Bioconductor Project). In this DESeq2 model where the HG-GEP NENs group served as the baseline, a negative log2FC indicates lower expression in the LG-GEP NENs group relative to the HG-GEP NENs group, signifying upregulation in the high-grade tumours. Batch effects were corrected using variance stabilising transformation. DEGs were identified based on dual screening criteria to ensure result reliability: i) Absolute log2 fold change (FC) >1 with a false discovery rate of <0.05; and ii) FC>2 or <0.5 with P<0.05.

GO and KEGG enrichment analyses of the identified DEGs were performed using the R package ClusterProfiler. GO enrichment analysis (http://geneontology.org/) was conducted across three domains: Biological processes, cellular components and molecular functions. KEGG enrichment analysis (https://www.genome.jp/kegg/) was performed to evaluate five domains: Metabolic pathways, signalling pathways, disease-related pathways, drug metabolism pathways and cellular processes.

The identified DEGs were subjected to PPI network analysis. Network visualisation was conducted using the Search Tool for the Retrieval of Interacting Genes/Proteins database (version 12.0; https://string-db.org). Hub genes were identified using the CytoHubba plugin in Cytoscape (v3.10.1; https://cytoscape.org).

A comprehensive analysis was conducted using data from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/), utilizing the TCGA-COAD and TCGA-READ datasets. Kaplan-Meier survival analysis was performed using the R package survival (version 3.5-7; https://cran.r-project.org/package=survival), with patients stratified into high- and low-expression groups based on the median expression level. The diagnostic performance of these genes was examined using receiver operating characteristic (ROC) curve analysis with the R package pROC (version 1.18.4; https://cran.r-project.org/package=pROC). Based on these prognostic and diagnostic analyses, CHEK1 was identified as the key target gene with significant diagnostic and prognostic value.

Figures derived from TCGA-COAD and TCGA-READ were generated using the GEPIA2 online analysis platform (http://gepia2.cancer-pku.cn/).

FFPE samples of patients diagnosed with HG-GEP NENs after 2012 were selected for CHEK1 molecular detection. Four samples from patients diagnosed prior to 2012 (three SCNEC and one NET G3) were excluded from subsequent molecular analyses due to insufficient RNA quality or quantity, resulting in 34 cases being included in the RT-qPCR experiments. RNA was extracted from FFPE tissue samples using the Nucleic Acid Extraction Reagent (Model: FFPE RNA; cat. no. 8.02.0019; Xiamen Moyd Biotechnology Co., Ltd.,) according to the manufacturer's protocol. Sections 5-10 µm in thickness were placed in 1.5 ml DNase-/RNase-free centrifuge tubes. To remove paraffin, 1 ml xylene was added to each tube, followed by vortex mixing for 10 sec and incubation at 56˚C for 3 min. After a second vortex mix, the samples were centrifuged at 13,000 x g for 2 min at 24˚C, and the supernatant was discarded. This step was repeated if necessary to ensure complete deparaffinization. Dehydration was then performed by adding 1 ml absolute ethanol, vortexing and centrifuging at 13,000 x g for 2 min at 24˚C, after which the supernatant was discarded. The pellets were air-dried at 56˚C until no ethanol residue remained. For RNA purification, tissue lysis and digestion were initiated by adding 160 µl Buffer RTL and 20 µl Proteinase K Solution to the pellets. The samples were vortexed, briefly centrifuged at 13,000 x g for 10-15 sec at 24˚C and incubated at 56˚C for 30 min with agitation at 500 rpm, followed by a second incubation at 80˚C for 30 min with agitation at 500 rpm. After cooling to room temperature (~24˚C), on-site prepared DNase I working mixture (30 µl per sample) was added to digest genomic DNA, and the samples were incubated at 37˚C for 15 min. After centrifugation at 13,000 x g for 3 min at 24˚C, the entire supernatant was transferred to a new tube. A total of 340 µl Buffer RPB and 750 µl absolute ethanol was added, and the mixture was vortexed. The solution was transferred to an RNA spin column and centrifuged at 13,000 x g for 30 sec at 24˚C. The flow-through was discarded. The remaining solution was loaded and centrifuged again under the same conditions. The column was then washed sequentially with 600 µl Wash Buffer A and 600 µl Wash Buffer B, with centrifugation at 13,000 x g for 30 sec at 24˚C after each wash. After a final centrifugation at 13,000 x g for 3 min at 24˚C to dry the membrane, the RNA was eluted with 80-100 µl Buffer RTE mixture by incubation at 56˚C for 2 min and centrifugation at 13,000 x g for 1 min at 24˚C. The purified RNA was dissolved in RNase-free water, and its concentration and purity (A260/A280) were determined using a SMA4000 spectrophotometer (Merinton Instrument, Ltd.). cDNA was synthesized from the purified RNA using the TransScript cDNA Synthesis Kit (TransGen Biotech Co., Ltd.) according to the manufacturer's instructions. The reaction was performed under the following thermocycling conditions: 25˚C for 10 min, 42˚C for 15 min and 85˚C for 5 sec, followed by hold at 4˚C. For qPCR, each reaction contained 2 µl cDNA template, 10 µl SYBR Green qPCR Master Mix (Thermo Fisher Scientific Inc.), 0.4 µM of each forward and reverse primer and RNase-free water up to a total volume of 20 µl. The thermocycling conditions were as follows: Initial denaturation at 95˚C for 3 min, followed by 40 cycles of 95˚C for 15 sec, 60˚C for 30 sec and 72˚C for 30 sec. Amplification specificity was verified through melting curve analysis. A no-template control was included in each run to monitor potential contamination of the reaction. Target gene expression was normalised to the reference gene (GAPDH) threshold value, and the relative gene expression level was calculated using the 2^-ΔΔCq method (33).

The primer sequences used for amplification of the target gene CHEK1 and the reference gene GAPDH were as follows: CHEK1 forward, 5'-GTGTCAGAGTCTCCCAGTGGAT-3' and reverse, 5'-GTTCTGGCTGAGAACTGGAGTAC-3'; and GAPDH forward, 5'-GGTGGTCTCCTCTGACTTCAACA-3' and reverse, 5'-GTTGCTGTAGCCAAATTCGTTGT-3'.

Transcriptomic (TPM) and corresponding clinical data were obtained from TCGA-COAD and TCGA-READ datasets, which were used as representative cohorts of digestive system tumours. Samples annotated as ‘solid tissue normal’ were included as adjacent non-tumourous tissues. As the numbers of tumour and normal samples were not identical, the comparison of CHEK1 expression between tumour and adjacent normal tissues was performed using an unpaired two-tailed Student's t-test. Statistical analyses were performed using SPSS (version 29.0; IBM Corp.) and R Studio (version 2023.12.1+402; Posit Software). Categorical variables are presented as number (n) and percentage (%), while continuous variables are expressed as the mean ± standard deviation (SD) or median with interquartile range (IQR), as appropriate. For categorical analysis, gene expression levels were dichotomized into ‘high’ and ‘low’ groups based on the median expression value of the entire cohort. Descriptive statistics, including χ² and Fisher's exact tests were employed for data summarization and clinicopathological associations. Comparisons of continuous variables between two groups were performed using an unpaired two-tailed Student's t-test. For CHEK1 quantitative analysis in the present cohort, which did not meet the assumptions of parametric tests, non-parametric tests were applied. Specifically, the Mann-Whitney U test was used for two-group comparisons, and the Kruskal-Wallis test followed by Dunn's post hoc test was used for comparisons across multiple groups. Survival assessment was conducted using Kaplan-Meier analysis and log-rank tests, whereas diagnostic performance was evaluated by ROC curve analysis. A post hoc power analysis was performed based on the quantitative RT-qPCR data of CHEK1 expression using the R pwr package (effect size d=1.5), which demonstrated 65% statistical power. P<0.05 was considered to indicate a statistically significant difference.

A total of 38 patients with HG-GEP NENs were included in the present study, comprising 9 patients (23.7%) with NET G3 and 29 patients (76.3%) with NEC. The NEC group was further subdivided into 11 patients (37.9% of NEC) with LCNEC and 18 patients (62.1% of NEC) with SCNEC. Complete follow-up data were available for 35 patients (92.1%), whereas data for 2 patients (5.3%) with NEC and 1 patient (2.6%) with NET G3 were unavailable. The median survival time for the entire cohort was 10.0 months (IQR, 5.0-25.0 months). When stratified by histological subtype, the median survival time was 25.0 months (IQR, 15.5-42.0 months) for the NET G3 group and 10.0 months (IQR, 5.0-21.5 months) for the NEC group. Analysis of clinicopathological data was conducted as described in our previous study (34). Immunohistochemical markers CLU and TP53 significantly impacted survival outcomes, with CLU serving as an independent prognostic factor. The immunohistochemical profile provided diagnostic value for NET G3 subtyping, while conventional demographic factors showed no significant associations with survival.

Specimens from 17 patients were analysed from the GSE211485 dataset, including 15 patients with LG-GEP NENs and 2 patients with HG-GEP NENs. The baseline clinical characteristics of the patients are presented in Table I. Using DESeq2 with the HG-GEP NENs group designated as the baseline, a total of 18,591 genes were assessed. Comparative analysis identified 6 of these genes that were significantly upregulated in the HG-GEP NENs group compared with those in the LG-GEP NENs group (Fig. 2A; Table II). The complete list of these upregulated DEGs is provided in Table II.

GO enrichment analysis demonstrated significant enrichment of DEGs in biological processes related to ‘centromere complex assembly’, ‘negative regulation of mitotic cell cycle’ and ‘negative regulation of mitotic cell cycle phase transition’; cellular components related to ‘condensed chromosome’, ‘chromosomal region’ and ‘kinetochore’; and molecular functions related to ‘protein C-terminus binding’, ‘dynein complex binding’, and ‘histone kinase activity’. KEGG enrichment analysis revealed substantial enrichment in pathways associated with ‘Cell cycle’, ‘p53 signalling pathway’ and ‘Fanconi anemia pathway’ (Fig. 2B).

Among the six significantly upregulated DEGs, five genes (CENPF, CHEK1, UBE2T, E2F7 and ITGB3BP) formed a tightly interconnected hub module in the PPI network constructed using the STRING database (Fig. 3A and B). The identification of these hub genes reveals their functional involvement in cell cycle regulation and DNA damage repair pathways. Furthermore, the convergence of their upregulated expression and central network position highlights their coordinated role in promoting genomic instability and uncontrolled proliferation, which are key mechanisms underlying HG-GEP NENs progression. Collectively, these findings support the value of further investigating these hub genes as potential therapeutic targets.

Owing to the rarity of HG-GEP NENs, corresponding tumour classifications were unavailable in TCGA. Therefore, transcriptomic and clinical data from TCGA-COAD and TCGA-READ projects were utilized as relevant proxies for digestive system tumours. Survival analysis demonstrated that high expression of CHEK1 was significantly associated with poorer overall survival (Log-rank P<0.05; Fig. 4A-E). Concurrently, ROC curve analysis indicated that CHEK1 expression possessed high diagnostic accuracy for distinguishing tumour from normal tissues (AUC >0.85; Fig. 4F). CHEK1 was selected for further analysis since it emerged as the most prognostically and diagnostically significant gene among the five hub genes in these validation analyses. Among these, CHEK1 was significantly overexpressed in digestive system tumour tissues compared with that in adjacent normal tissues based on TCGA-COAD and TCGA-READ datasets (P<0.001, unpaired Student's t-test; Fig. 4G), suggesting that CHEK1 may serve as a diagnostically and prognostically relevant target gene in digestive system malignancies.

A total of 34 patients with HG-GEP NENs recruited in the present study were included in the RT-qPCR analysis, excluding 3 patients with SCNEC and 1 patient with NET G3 diagnosed prior to 2012. Patients were classified into high- and low-CHEK1 expression groups based on the median expression level for categorical analyses. RT-qPCR analysis revealed that CHEK1 expression was lower in the NET G3 group (n=8) compared with that in the NEC group (n=26), although the difference was not statistically significant (P=0.075; Fig. 5A). Further analysis among the NET G3 (n=8), LCNEC (n=11) and SCNEC (n=15) subgroups revealed no significant differences (P>0.05 for all pairwise comparisons; Fig. 5B). Analysis between histological subtypes indicated a significant difference between the NET G3 and NEC groups (P=0.0113; Fig. 5C), with a higher proportion of patients exhibiting high CHEK1 expression in the NEC group. Further intergroup analysis demonstrated that the proportion of patients exhibiting elevated CHEK1 expression was significantly higher in the SCNEC subgroup compared with that in the NET G3 and LCNEC groups (P=0.0075; Fig. 5D).

The median survival time was 14 months in the CHEK1 high-expression group and 16 months in the low-expression group. The Kaplan-Meier survival analysis demonstrated no significant difference in overall survival between the high- and low-expression groups (P=0.8464, Fig. 6A). ROC curve analysis demonstrated that high CHEK1 expression influenced the diagnostic performance of NEC (AUC=0.8029; Fig. 6B). Further analysis, stratifying NEC cases into LCNEC and SCNEC subtypes, revealed that high CHEK1 expression contributed to the diagnosis of LCNEC and SCNEC, with a more pronounced effect observed for SCNEC (AUC=0.8708) compared with LCNEC (AUC=0.7102) (Fig. 6C and Table III).