Gastric cancer (GC) is a significant health concern, ranking fifth and fourth in terms of morbidity and mortality [1]. Early diagnosis rates for gastric cancer worldwide are currently below 20%, considerably lower than the 50%-60% rates seen in Japan and South Korea. This is primarily due to the hidden onset of gastric cancer, and the substantial cost of intensive early screening, which is not currently feasible on a widespread scale [2,3,4,5,6]. Targeted surveillance of precancerous lesions, including them in the high-risk group for critical monitoring, can improve the efficiency and financial benefits of screening for early gastric cancer (EGC). Chronic stimulation of normal gastric mucosa by factors like Helicobacter pylori (HP), carcinogens, high salt, bile acids, tobacco, or alcohol can lead to pathological progression from chronic atrophic gastritis (CAG), intestinal metaplasia (IM), gastric dysplasia, and ultimately adenocarcinoma, a process known as the "Correa cascade" [7, 8]. Intestinal metaplasia represents a significant portion of precancerous lesions associated with gastric cancer. Based on Lauren's classification, approximately 80% of gastric cancer cases are categorized as the intestinal type, which originates from the mucosa undergoing intestinal metaplasia [9,10,11]. Intestinal metaplasia results from chronic inflammatory stimulation of the stomach's normal mucosal epithelium, leading to atrophy of parietal cells and the formation of goblet cells and enterocytes. While enteroid epithelium replaces the lost gastric glands, this process is marked by impaired differentiation and atypical regeneration, increasing the risk of cancer [8, 12,13,14]. Consequently, patients with intestinal metaplasia have a significantly higher risk of developing cancer compared to healthy individuals [13, 15].

Incomplete intestinal metaplasia (IIM) is associated with a 4 to 11-fold higher risk of developing gastric adenocarcinoma compared to Complete intestinal metaplasia (CIM) [16,17,18]. Since various subtypes of intestinal metaplasia exhibit different levels of cancer risk, recognizing IIM is essential. Currently, HE staining and AB-PAS staining are the primary methods used for assessing IIM. However, staining interpretation can be challenging for non-pathologists as well as non-gastrointestinal pathological experts [13]. Therefore, developing appropriate markers for IIM diagnosis, identifying "high-risk" IIM groups, and defining key screening targets can enhance endoscopy efficiency, increase EGC diagnosis rates, and significantly reduce associated healthcare costs.

OLFM4, also known as Olfactomedin-4, GW112, or hGC-1, is a glycoprotein belonging to the olfactory regulatory protein family [19]. OLFM4 expression is absent in normal gastric mucosa but is present in the small intestine and colon [20, 21]. Researches have highlighted OLFM4's crucial role in regulating intestinal stem cells [22, 23]. In fact, one of OLFM4's key biological functions is to regulate cell adhesion and cell migration by interacting with adhesion molecules, the cytoskeleton, and the extracellular matrix [24]. Studies have shown elevated OLFM4 expression in gastric cancer and colorectal cancer, particularly in the early stages of tumor formation [25]. However, previous studies have not explored the differential expression of OLFM4 in different severity of intestinal metaplasia and the mechanism by which OLFM4 promotes the progression of intestinal metaplasia. The activation of the Wnt/β-catenin signaling pathway is crucial for tumor invasion and metastasis [26,27,28]. Phenotypically, Wnt/β-catenin expression is positive in 86% of intestinal metaplasia and 95% of gastric cancer cases [29].

Expanding upon the aforementioned literature review, our research study centered on evaluating OLFM4 expression in IIM. We aimed to construct a predictive model utilizing OLFM4 as a classifier for distinguishing different types of intestinal metaplasia. Additionally, we investigated potential mechanisms and activated signaling pathways by which OLFM4 contributes to the progression of gastric mucosal intestinal metaplasia.

Methods and materialsPatient samples and cell lines

This study received approval from the Medical Ethical Committee of the Seventh Affiliated Hospital of Sun Yat-sen University and the People's Hospital of Fengqing in Yunnan Province (KY-2021–105-01, 2021/12/02). Written informed consent, following the Declaration of Helsinki guidelines, was obtained from every patient. Paraffin-embedded archived samples were collected between 2018 and 2022 from two distinct clinical centers. The samples of the first clinical center included 78 newly diagnosed intestinal metaplasia patients and 40 healthy individuals from the Seventh Affiliated Hospital (Training set). The samples of the second clinical center consisted of 63 newly diagnosed intestinal metaplasia patients and 40 healthy individuals from the People's Hospital of Fengqing (Validation set). The supplemental materials include the clinical data of the Training set and Validation set. Histological diagnoses were based on the Third edition of Pathology [30], and the classification of intestinal metaplasia was determined according to Pathology and classical literature [30,31,32,33]. GES-1 cells were purchased from ATCC and underwent short tandem repeat (STR) analysis.

Patient-derived organoids (PDO) were cultured in the laboratory of the Seventh Affiliated Hospital. Biopsy tissues were obtained from patients with intestinal metaplasia or healthy individuals. First, we obtained biopsy samples from the gastric mucosa suspected of intestinal metaplasia and digested them into generate organoids. Subsequently, based on postoperative pathology, we selected cases showing severe intestinal epithelial metaplasia to establish organoid models mimicking intestinal epithelial metaplasia. The collected gastric mucosal tissues were washed three times in penicillin–streptomycin mixed solution in PBS, then under aseptic conditions, cut into 1mm3 fragments. These fragments were subsequently incubated in DMEM/F12 medium with collagenase IV at 37 °C with agitation for 30 min. After centrifugation, the supernatant containing single-cell suspension was collected following removal of the sediment. The cell pellet obtained after centrifugation was resuspended in a mixture of DMEM/F12 and Matrigel matrix gel solution, and seeded into 96-well plates at 10μL per well. The gastric mucosal organoids were then incubated at 37 °C until the gel solidified, followed by addition of 200μL organoid culture medium containing penicillin–streptomycin mixture. The cultures were maintained at 37 °C with medium renewal every 3 days for 14 days. We adapted a protocol for organoid culture medium from Helen H.N. Yan, incorporating components such as DMEM/F12, HEPES, GlutaMax, penicillin/streptomycin, RSPO-1, B27, N-Acetylcysteine, EGF, FGF10, Noggin, A8301, Y-27632, and Gastrin into the culture medium [34].

DEGs in intestinal metaplasia tissues

We selected the GSE78523 dataset from the Gene Expression Omnibus database (GEO), which comprises samples from both normal gastric mucosa and intestinal metaplasia tissues. Differential Expression Genes (DEGs) were analyzed using the "DESeq2" package in R. The volcano plots and heatmaps were generated using the "ggplot2" package or GraphPad Prism 8.0.2 based on the results of the DEGs analysis.

Identification of a new PLGC subgroup in intestinal metaplasia tissues

We selected and analyzed the GSE134520 dataset, a single-cell RNA sequencing (scRNA-seq) dataset that includes non-atrophic gastritis (NAG), intestinal metaplasia (IM), and early gastric cancer (EGC), using the "Seurat" R package. Data normalization was performed using the "NormalizeData" function, and inconsequential sources of variation were removed with the "ScaleData" function. The "FindVariableFeatures" function was used to identify highly variable genes (HVGs), and the "RunPCA" function identified 50 significant principal component analyses (PCA). We embedded cells into the graph structure of PCAs using the "FindNeighbors" and "FindClusters" functions. The spatial correlation of expression data was presented through Uniform Manifold Approximation and Projection (UMAP) plots based on RunUMAP and Dimplot. We selected all epithelial cells using classical epithelial markers "EPCAM" and "KRT19". A total of 11 clusters were identified based on HVGs. To assess the malignancy of glandular cells, the "inferCNV" package was used to determine cellular heterogeneity by identifying chromosome copy number variation (CNV) in scRNA-seq. Gastric cancer cells served as a positive control for CNV, and precancerous lesions of gastric carcinoma (PLGC) cells were identified by "inferCNV" in comparison. The "Monocle" function was employed to display the evolution of gastric mucosa during EGC development by performing pseudotime analysis, projecting high-dimensional data into one dimension. The "Cytotrace" function was utilized to create a critical RNA-based feature for developmental potential and to establish a platform for delineating cellular hierarchies, attempting to predict differentiation states from scRNA-seq.

Reagents

N-Methyl-N”-nitro-N-nitrosoguanidine (MNNG) was obtained from Meilunbio (MB0455-2, China).

Cell transfection

Lentivirus vectors encoding shOLFM4, oeOLFM4, shMYH9, oeMYH9, control, or HA-ubiquitin plasmids were from GeneCopoeia (Guangzhou, China). HA-ubiquitin plasmids were transfected into cells for 48 h, followed by lysis for immunoblotting with anti-HA antibodies.

24 h prior to lentivirus transfection, adhere cells were seeded at a density of 1 × 105 cells per well in a 24-well plate. 5 ug/ml of polybrene was added as a membrane-disrupting agent, followed by the establishment of five different virus concentration gradients (0, 1, 2, 3, 4, 5 ul/mL) and incubation at 37 °C. After 6 h, the media was replaced, and cells were further cultured until 48 h. Fluorescence expression was observed, and an appropriate virus concentration was selected for experimentation. Cell selection was performed using the purine resistance gene contained within the lentivirus. The sequence of plasmids as follows:

 

ID

Target Sequence

NC

CSHCTR001-LVRU6GP

GCTTCGCGCCGTAGTCTTA

EC

EX-NEG-Lv201

 details in Supplementary Information

OLFM4

EX-Y2060-Lv201

 details in Supplementary Information

shOLFM4-1

HSH090725-LVRU6GP-a

CCAAAGTGAGGGAATATGTCC

shOLFM4-2

HSH090725-LVRU6GP-b

CCTAACTGTCCGAATTGACAT

MYH9

EX-T1335-Lv242

 details in Supplementary Information

shMYH9

HSH102697-LVRU6GP-a

GCAAGCTGCCGATAAGTATCT

Immunohistochemistry (IHC), Immunofluorescence (IF) staining, Alcian blue-periodic acid-Schiff (AB-PAS) straining, and hematoxylin–eosin (HE) staining

We conducted IHC, IF, AB-PAS, and HE staining using standard protocols as follows:

HE Staining

Deparaffinization of histopathological tissue paraffin sections and organoid cell paraffin sections in xylene followed by rehydration in distilled water. Sequential staining with hematoxylin, differentiation in 95% ethanol, eosin staining, dehydration in xylene, and mounting.

AB-PAS (Alcian Blue Periodic Acid Schiff) staining

Deparaffinization of histopathological tissue paraffin sections and organoid cell paraffin sections in xylene followed by rehydration in distilled water. Sequential staining with Alcian Blue, differentiation in 95% ethanol, eosin staining, PAS staining, dehydration in xylene, and mounting.

IHC or IF staining

Deparaffinization of histopathological tissue paraffin sections and organoid cell paraffin sections in xylene followed by rehydration in distilled water. High-temperature antigen retrieval with sodium citrate buffer at 121° C for 4 min, permeabilization with 3% H2O2, blocking of nonspecific binding sites with goat serum, overnight incubation at 4 °C with primary antibodies diluted accordingly, incubation at room temperature for 60 min with corresponding secondary antibodies (utilizing a universal mouse/rabbit secondary antibody labeled with horseradish peroxidase for IHC or Alexa Fluor 488/Alexa Fluor 647 fluorescent secondary antibodies for IF). IHC sections were stained with DAB (3,3'-Diaminobenzidine hydrochloride, GK600705), dehydrated in xylene, and mounted. IF sections were counterstained with DAPI fluorescence, antifade reagent before mounting and observed either through fluorescence microscopy (Leica, DM6B) or confocal microscopy (ZEISS, LSM-880). Two independent observers, unaware of the patient's clinical information, evaluated the staining results at separate intervals. The IHC score was determined using Image J to calculate the proportion of stained areas. Primary antibodies used included OLFM4 (14369S, CST,1:600), CDX2 (A19030, Abclonal, 1:800), MUC2 (sc-515032, Santa Cruz Biotechnology, 1:1000), MYH9 (14844–1-AP, Proteintech, 1:400), E-cadherin (60335–1-Ig, Proteintech, 1:200), Vimentin (60330–1-Ig, Proteintech, 1:100), and Ki67 (ab16667, Abcam, 1:1000).

Western Blot

Western Blot were performed following standard methods. Electrophoresis: Prepare Tris–Glycine gel and electrophoresis buffer. Load protein samples onto the lanes and set the voltage for electrophoresis (90 V for 30 min followed by 120 V for 90 min). Transfer: Soak PVDF membrane in methanol and then place it in transfer buffer. Place the membrane and gel in the transfer clamp, remove any bubbles, and proceed with the transfer (300 mA for 90 min). Block with 5% BSA for 60 min, incubate with the primary antibody overnight at 4℃, incubate with the secondary antibody for 60 min at room temperature, and expose using a Bio-Rad instrument after soaking in ECL developing solution. The primary antibodies included OLFM4 (14369S, CST, 1:1000), MYH9 (11128–1-AP, Proteintech, 1:5000), GSK3-β (22104–1-AP, Proteintech, 1:1000), β-catenin (51067–2-AP, Proteintech, 1:5000), β-actin (66009–1-Ig, Proteintech, 1:20000), p-STAT3 (9145S, CST, 1:2000), c-Myc (10828–1-AP, Proteintech, 1:2000), N-cadherin (22018–1-AP, Proteintech, 1:2000), E-cadherin (60335–1-Ig, Proteintech, 1:2000), Vimentin (60330–1-Ig, Proteintech, 1:20000), Snai1 (13099–1-AP, Proteintech, 1:500), Ubiquitin (10201–2-AP, Proteintech, 1:1000), and GAPDH (60004–1-Ig, Proteintech, 1:50000).

Quantitative Reverse Transcription-PCR (qRT-PCR)

Total RNA was harvested, and cDNA was generated by a reverse transcription reagent kit (AG11706, Accurate Biology, China). Then, the cDNA template was used for amplification with specific primers. qRT-PCR was conducted using SYBR-green PCR Master Mix and 45 cycles of 95℃ for 10 s, 60℃ for 20 s, and 72℃ for 20 s. These sequences of primers are defined as follows:

 

Forward

Reverse

CDX2

TTCACTACAGTCGCTACATCACCA

CTGCGGTTCTGAAACCAGATT

MUC1

TTCACCACCACCATGACACC

GGGGCTGTGGTAGCTGTAAG

MUC2

GGGGAGTGCTGTAAGAAGTGTGA

GTTGGAGACGGACGAGATGAG

OLFM4

GAGAAATCGTGGCTCTGAAGAC

CAGACGGTTTGCTGATGTTC

GSK3β

CATCCTTGGACTAAGGTCTTCCG

CATTTGTGGGGGTTGAAGCAG

β-actin

TCAAGATCATTGCTCCTCCTGAG

ACATCTGCTGGAAGGTGGACA

Cell proliferation assay, colony-formation assay, EdU assay, wound healing assay, and transwell assayCell proliferation

We assessed cell proliferation using the Cell Counting Kit-8 (CCK8) assay kit (Biosharp, China) and the Microplate Reader (BioTeK, USA). We seeded 2,000 cells into 96-well plates and cultured them for 1–5 days. Each day, we mixed the CCK8 reagent with the cell culture medium at a 1:9 ratio and incubated the cells for 90 min. We measured absorbance at 450 nm using a spectrophotometer in each culture dish.

Cell colony formation

For colony formation, we inoculated 800 cells in six-well plates and cultured them for 14 days. The number of cell colonies was determined by microscopy after staining with crystal violet dye.

Cell viability (EdU Assay)

We measured cell viability using the EdU assay. We plated 6,000 cells into 96-well plates, treated them with EdU reagent (10 μM, Beyotime, China), and observed them with fluorescence microscopy (Leica, DMI8).

Wound healing assay

Cells were plated and grown to confluence in six-well plates. We created scratches with a pipette tip and examined the cell migration process under a microscope at 0 and 24 h.

Cell migration and invasion

We evaluated cell migration and invasion using 24-well transwells (8.0 μm, Corning, USA), precoated with Matrigel in invasion assay but without Matrigel in migration assay. In the lower chamber, we added 500 μL RMPI-1640 with 10% FBS (Nanjing Ozfan). We seeded 5 × 104 treated cells suspended in 500 μL RMPI-1640 without FBS in the upper chamber and cultured them at 37 °C for 36 h. We counted the number of GES-1 cells in the lower chamber using a cell counting plate.

Co-immunoprecipitation (IP) and mass spectrometry

Protein extraction and purification were performed using primary antibodies for IP and Protein A/G Magnetic Beads (B23202, Selleck). Mass spectrometry was performed by Baiqu Tech. co. LTD (Hangzhou, China) and results were provided in Table 5. The primary antibodies included OLFM4 (14369S, CST), MYH9 (11128–1-AP, Proteintech), and GSK3-β (22104–1-AP, Proteintech).

Cycloheximide (CHX) chase assay

Three groups of GES-1 cells were incubated with 2 μM MG132 (HY-13259, MCE, USA) for a duration of 12 h prior to protein extraction, while another groups of cells remained untreated. After the treatment of 20 μg/mL CHX (C7698, Sigma-Aldrich) for different times, cells were harvested and prepared for Western Blot.

PLGC animal model

The Institutional Animal Care and Use Committee (IACUC) (TopBiotech Co., LTD., Shenzhen) approved the experimental methods and animal use and care protocols. We obtained twenty male Sprague Dawley (SD) rats five-week-old for each group from Gempharmatech company (Jiangsu, China).

We prepared an MNNG solution with a concentration of 170 µg/ml by dissolving MNNG in drinking water containing 5% alcohol. The rats received the MNNG solution by gavage every two days, with a regimen of one day of a normal diet and one day of fasting. This procedure continued for 24 weeks.

Statistical analysis

Statistical analyses were performed by using SPSS 22.0 (SPSS Inc., Chicago). Histogram Graphing was performed with GraphPad Prism 8.0.2 (GraphPad Software). Each in vitro experiment was repeated three times or more and experimental data were depicted as mean ± standard deviation (SD). Quantitative variables were analyzed using a Student t-test for Gaussian distribution and non-parametric tests for non-Gaussian distribution. Differences were considered statistically significant at p < 0.05 (*p < 0.05, **p < 0.01, ***p < 0.001).