Human GC cell lines (AGS, HGC27, MKN45 and MKN28) and an immortalized gastric epithelial cell line (GES-1) were obtained from the Cell Bank of Type Culture Collection of The Chinese Academy of Sciences, and were subsequently authenticated by short tandem repeat profiling and routinely tested for mycoplasma contamination, with only mycoplasma-negative cultures used for experiments. All cell lines were maintained in RPMI-1640 medium (HyClone™; Cytiva) supplemented with 10% FBS (Gibco; Thermo Fisher Scientific, Inc.) and 1% 100X penicillin-streptomycin (Gibco; Thermo Fisher Scientific, Inc.) at 37˚C in a humidified incubator containing 5% CO₂. AGS cells were pretreated with LY294002 (10 µM, dissolved in RPMI-1640 medium vehicle; cat. no. S1105; Selleck Chemicals) at 37˚C for 1 h and then maintained in LY294002-containing medium for subsequent assays. LY294002 was present throughout the wound-healing and Transwell migration/invasion assays (24 h), and for pathway validation, cells were cultured under the same treatment conditions at 37˚C and harvested for western blotting of PI3K/AKT signaling proteins.

For gene knockdown experiments, cells were transfected with a GV493 short hairpin RNA (sh) plasmid targeting UPK1B (Table SI) or a negative control scrambled plasmid (sh-NC), both obtained from GeneChem, Inc. Cells were also transfected with small interfering RNAs (siRNAs) targeting CDX2 or PIK3IP1 (Table SI), with a universal non-targeting scrambled siRNA (si-NC) as the negative control (Table SI), obtained from GeneChem, Inc. For UPK1B and CDX2 overexpression, the p-TSB-CMV-UPK1B and p-TSB-CMV-CDX2 expression vectors [Shanghai Genomeditech Co., Ltd.] and the corresponding empty p-TSB-CMV vector (negative control) were used. Transfection was carried out using Lipofectamine™ 2000 (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. For plasmid transfection (overexpression vectors), 2.5 µg of plasmid DNA was used per well (in 6-well plates), whereas for siRNA transfection, the final concentration of siRNA was 50 nM. The transfection mixture was incubated with the cells at 37˚C for 6 h, after which the medium was replaced with fresh complete medium. Functional assays and western blotting were performed at 48 h post-transfection. Lentiviral particles were generated using a second-generation packaging system. 293T cells (Cell Bank of Type Culture Collection of The Chinese Academy of Sciences) were co-transfected with 20 µg of the GV493-based shRNA plasmids (Shanghai GeneChem Co., Ltd.) and 10 µg of the Lenti-Easy Packaging Mix (containing the packaging plasmid pHelper 1.0 and the envelope plasmid pHelper 2.0; Shanghai GeneChem Co., Ltd.) using Lipofectamine™ 2000 and a plasmid ratio of 2:1. Transfection was carried out at 37˚C for 6 h. Viral supernatants were collected at 48 h post-transfection. Target cells were infected at a multiplicity of infection of 10 for 24 h. Stable cell lines were selected using puromycin starting 48 h post-infection, the concentration used was 2 µg/ml for selection and 0.5 µg/ml for maintenance. Subsequent experiments were performed 5 days after transduction.

Cells (GES-1, AGS, HGC27, MKN45 and MKN28) were lysed using RIPA buffer (Sigma-Aldrich; Merck KGaA) supplemented with a cOmplete™ Protease Inhibitor Cocktail (Roche Diagnostics), and protein concentrations were measured using a BCA assay kit (Thermo Fisher Scientific, Inc.). Equal amounts of protein (30 µg) were separated on 5-20% SDS-PAGE gels, followed by transfer onto PVDF membranes (MilliporeSigma). After transfer, membranes were blocked in 5% BSA (Sigma-Aldrich; Merck KGaA) dissolved in PBS for 2 h at room temperature to reduce non-specific antibody binding. The membranes were then incubated overnight at 4˚C with primary antibodies against UPK1B, GAPDH, phosphorylated (p-)PI3K, PI3K, p-AKT, AKT and CDX2 (Table SII). Primary antibodies were generally diluted at 1:1,000, except for GAPDH (1:5,000). HRP-conjugated secondary antibodies were used at a dilution of 1:5,000. After washing in TBST (containing 0.1% Tween-20), membranes were incubated for 2 h at room temperature with HRP-conjugated secondary antibodies (Table SII). Protein bands were visualized using Tanon™ Femto-sig ECL Western Blotting Substrate (Tanon Science and Technology Co., Ltd.).

Total RNA was isolated from cultured cells using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. Reverse transcription was carried out using the PrimeScript RT reagent kit (Takara Biotechnology Co., Ltd.) according to the manufacturer's protocol. Furthermore, qPCR was conducted using SYBR Green Master Mix (Takara Biotechnology Co., Ltd.) on a LightCycler^® 96 system (Roche Diagnostics). β-actin served as the internal reference gene. All primer sequences are listed in Table SII. The thermocycling conditions were as follows: Initial denaturation at 95˚C for 30 sec, followed by 40 cycles of denaturation at 95˚C for 5 sec and annealing/extension at 60˚C for 30 sec. A melt-curve analysis was subsequently performed to confirm the specificity of amplification. Relative mRNA expression levels were calculated using the 2^-ΔΔCq method (22).

Cell migration and invasion were assessed using Transwell chambers equipped with 8-µm pore size polycarbonate membranes (BD Biosciences). For the migration assay, uncoated chambers were used, whereas for the invasion assay, the membranes were pre-coated with Matrigel (BD Biosciences) at 37˚C for 2 h to allow matrix polymerization. GC cells were serum-starved in RPMI-1640 medium without FBS for 24 h prior to the assay. A total of 1x10⁵ cells in 0.5 ml serum-free medium were seeded into the upper chambers, while the lower chambers were filled with complete medium containing 10% FBS (Gibco; Thermo Fisher Scientific, Inc.) as a chemoattractant. After incubation for 24 h at 37˚C, non-migratory cells remaining on the upper membrane surface were gently removed using a cotton swab. Cells that had traversed to the underside of the membrane were fixed with 4% paraformaldehyde for 5 min at room temperature, stained with crystal violet for 10 min at room temperature and visualized under an inverted light microscope. Migratory or invasive cells were counted in five random fields.

GC cells were seeded into 6-well plates and cultured in complete medium until reaching 95-100% confluency. A sterile 200-µl pipette tip was used to create a linear scratch across the cell monolayer. Detached cells and debris were removed by gentle washing with serum-free medium. Cells were then maintained in serum-free medium for the remainder of the assay to minimize the influence of cell proliferation. The initial wound area was imaged under a light microscope at a magnification of x100. Subsequently, cells were incubated at 37˚C for an additional 24 h, after which images were captured again to evaluate the extent of cell migration. The relative wound closure area was quantified to assess the migratory capacity. The percentage of wound closure was calculated as follows: [(Wound area at 0 h-wound area at 24 h)/wound area at 0 h] x100.

IF staining was performed on cells cultured on glass coverslips. When cells reached 30-40% confluency, they were fixed with 4% paraformaldehyde for 15 min at room temperature and permeabilized using PBS containing 0.1% Triton X-100. After blocking with 10% goat serum at room temperature for 1 h (Solarbio Science & Technology Co., Ltd.), cells were incubated overnight at 4˚C with primary antibodies against PIK3IP1 (1:100; sc-365778; Santa Cruz Biotechnology, Inc.) and UPK1B (1:100; PA5-72711; Thermo Fisher Scientific, Inc.). Subsequently, cells were incubated with the corresponding species-specific fluorescence-conjugated secondary antibodies: Goat anti-mouse IgG H&L (AF594; 1:200; 550042; ZenBio, Inc.) and goat anti-rabbit IgG H&L (AF488; 1:200; 550036; ZenBio, Inc.) for 1 h at room temperature in the dark. Nuclei were counterstained with DAPI (D9542; Sigma-Aldrich; Merck KGaA) for 10 min at room temperature. Images were acquired using a Zeiss LSM 980 confocal microscope (Zeiss GmbH).

Cells were lysed in IP Lysis Buffer (P0013; Beyotime Institute of Biotechnology) containing the cOmplete™ Protease Inhibitor Cocktail (Roche Diagnostics) on ice for 30 min. For each IP reaction, equal amounts of total protein (1 mg) were incubated with 2 µg of specific primary antibodies targeting PIK3IP1 (671477; ZenBio, Inc.) and UPK1B (PA5-72711; Thermo Fisher Scientific, Inc.) overnight at 4˚C. The next day, 30 µl of protein A/G agarose beads (sc-2003; Santa Cruz Biotechnology, Inc.) were added for incubation for an additional 6 h at 4˚C with gentle rotation. Parallel reactions using control rabbit IgG or mouse IgG (A00002 and A00001; ZenBio, Inc.) served as negative controls. After washing three times with cold IP Lysis Buffer, the immunoprecipitated complexes were collected by centrifugation at 1,000 x g for 2 min at 4˚C. The pellets were then resuspended in SDS loading buffer and boiled for 5 min before being subjected to downstream analysis through western blotting as aforementioned.

RNA-sequencing data from The Cancer Genome Atlas Stomach Adenocarcinoma (TCGA-STAD) cohort were retrieved from TCGA portal (https://portal.gdc.cancer.gov) and normalized to transcripts per million (TPM) values. Gene set enrichment analysis was conducted using the ‘clusterProfiler’ R package version 4.4.4(23) from R Studio version 2023.12.1 (Posit Software, PBC). Survival analysis was conducted using the Kaplan-Meier (KM) Plotter database (https://kmplot.com) [including all datasets: GSE14210(24), GSE15459(25), GSE22377(26), GSE29272(27), GSE51105(28), GSE62254(29)] with Auto select best cutoff percentiles. Protein-protein interaction (PPI) information for the target gene(s) was retrieved from two curated repositories: The Human Integrated Protein-Protein Interaction rEference (HIPPIE; https://cbdm-01.zdv.uni-mainz.de/~mschaefer/hippie/) and the Biological General Repository for Interaction Datasets (BioGRID; https://thebiogrid.org/). Gene symbols (and corresponding UniProt accessions when needed) were queried using each database's web interface, restricting results to Homo sapiens and experimentally supported physical PPIs. Furthermore, the ChIP-X Enrichment Analysis (ChEA; https://maayanlab.cloud/Harmonizome/) and ChEA3 (https://maayanlab.cloud/chea3/) platforms were used to predict transcription factors for specific genes or gene sets.

All experiments were performed in triplicate independent replicates. Statistical analysis was conducted using GraphPad Prism (version 8.0; Dotmatics). Normally distributed data are presented as the mean ± SD. Comparisons between two groups were made using an unpaired two-tailed Student's t-test. For comparisons among ≥3 groups, one-way ANOVA followed by Tukey's post hoc test was performed. Kaplan-Meier survival curves were compared using the log-rank test, and Cox proportional hazards regression models were used for univariate and multivariate analyses of prognostic factors. P<0.05 was considered to indicate a statistically significant difference.

Lymph node and distant metastases are two major manifestations of GC dissemination (30). To identify genes associated with these traits, differential expression analysis between tumors with and without lymph node involvement (N0 vs. N1-3) and with and without distant metastasis (M0 vs. M1) was performed based on TCGA-STAD RNA-seq dataset. Genes with log₂ (fold change) >0.585 and adjusted P<0.05 were considered significant. The intersection of these two gene sets yielded 26 upregulated genes involved in both lymphatic and distant metastases of GC cells (Table SIII; Fig. 1A).

Prognostic data were further obtained for the TCGA cohort, and Least Absolute Shrinkage and Selection Operator Cox regression analyses were conducted (Fig. 1B), which identified a 12-gene prognostic signature. The corresponding risk score formula was defined as follows: Risk score=(0.0710) x UPK1B + (0.0911) x TFPI2 + (0.0372) x IL6 + (0.1959) x ST8SIA6-(0.0546) x LGALS9C + (0.0244) x PSAPL1 + (0.3766) x SYT16-(0.1152) x DCHS2 + (0.1222) x IRS4 + (0.0024) x VMO1 + (0.0082) x UPK2 + (0.1254) x chorionic gonadotropin subunit β-5 (CGB5). Survival analysis using TCGA-STAD data demonstrated that patients with higher risk scores exhibited a significantly poorer prognosis (Fig. 1C). Cox regression analysis was also performed on the 26 metastasis-related genes, identifying eight genes as potential prognostic factors after univariate analysis (Table SIV) and two genes, UPK1B and CGB5, as independent prognostic factors after multivariate analysis (Table SIV). Fig. 1D and E show that high expression of UPK1B and CGB5 is significantly associated with diminished overall survival in patients with GC in TCGA-STAD dataset. Collectively, these analyses defined a metastasis-related 12-gene prognostic signature, and highlighted UPK1B and CGB5 as independent prognostic markers in GC, providing a basis for risk stratification of patients with advanced disease.

Consistent with the signature-based findings, in the TCGA-STAD cohort and KM Plotter database, higher UPK1B expression was associated with shorter overall survival (Fig. 1E and F), and UPK1B levels were significantly increased in tumors with distant (M1) and advanced nodal (N2) metastasis compared to those with M0 and N0 status, respectively (Fig. 1G and H). As CGB5 has been extensively studied in gastric cancer (31-34), the present analysis was focused on UPK1B to avoid redundancy and highlight novelty. These data indicated that UPK1B marked a clinically high-risk, highly metastatic subset of GC, and supported its selection for subsequent mechanistic studies.

To explore the mechanism linking UPK1B to GC metastasis, transcriptomes of UPK1B high- and low-expression tumors in the TCGA-STAD cohort were compared. Genes in the PI3K/AKT signaling pathway were significantly enriched among those upregulated in the UPK1B high-expression group (Fig. 2A), implicating UPK1B in PI3K/AKT activation.

UPK1B expression was first assessed across a number of GC cell lines, with the highest levels observed in MKN45 and the lowest in AGS (Fig. 2B). UPK1B knockdown in MKN45 cells reduced PI3K/AKT phosphorylation (Fig. 2C), and markedly decreased invasion and migration (Fig. 2D and E), whereas UPK1B overexpression in AGS cells enhanced PI3K/AKT activation (Fig. 2F), and increased the invasive and migratory capacity (Fig. 2G and H). Notably, the PI3K/AKT inhibitor LY294002 (10 µM) abrogated the pro-invasive and pro-migratory effects of UPK1B overexpression (Fig. 2F-H), demonstrating that UPK1B-driven metastatic phenotypes are PI3K/AKT-dependent. Together, these findings suggested UPK1B as a functional driver of GC cell invasion and migration through activation of the PI3K/AKT pathway, and indicated that tumors with high UPK1B expression may be particularly sensitive to PI3K/AKT-targeted therapies.

To investigate how UPK1B activates PI3K/AKT signaling, HIPPIE and BioGRID interaction databases were interrogated, and PIK3IP1, a negative regulator of PI3K, was identified as a putative UPK1B binding partner (Fig. 3A; Table SV). Consistently, IF staining showed co-localization of UPK1B and PIK3IP1 at the plasma membrane (Fig. 3B), and co-IP assays demonstrated their physical interaction in MKN45 cells (Fig. 3C).

Dual knockdown experiments demonstrated that silencing of PIK3IP1 in UPK1B-deficient MKN45 cells (Fig. 3D) restored PI3K/AKT activation (Fig. 3E), and improved invasion and migration (Fig. 3F and G). These data support a model in which UPK1B promotes PI3K/AKT signaling and pro-metastatic behavior by functionally antagonizing the inhibitory regulator PIK3IP1. Therefore, by releasing PI3K from PIK3IP1-mediated restraint, UPK1B may amplify PI3K/AKT signaling in high-risk GC, providing a mechanistic rationale for targeting this axis in patients with UPK1B upregulation.

To identify upstream regulators responsible for UPK1B upregulation during invasion and migration, the top 20 predicted transcription factors (ESR1 and SOX2 appeared twice) (Table SVI) regulating UPK1B were first retrieved using the ChEA platform within Harmonizome. The ChEA3 platform was then employed to predict the top 20 upstream transcription factors (Table SVII) for the genes co-expressed with UPK1B (R>0.4; adjusted P<0.05) based on TCGA-STAD dataset. CDX2 emerged as the top-ranking transcription factor for these co-expressed genes (Table SVII), and the intersection of both analyses (Fig. 4A) suggested that CDX2 acts as an upstream transcriptional regulator of UPK1B.

Consistent with this prediction, CDX2 knockdown in AGS cells (Fig. 4B) increased UPK1B mRNA and protein expression (Fig. 4C and D), whereas CDX2 overexpression in MKN45 cells reduced UPK1B protein levels (Fig. 4E). Clinical data from TCGA-STAD and the KM Plotter database further showed that low CDX2 expression in GC tissues was associated with a poorer prognosis compared with that in the high CDX2 group (Fig. 4F and G). These findings indicated that loss of the tumor-suppressive transcription factor CDX2 could de-repress UPK1B, thereby associating CDX2 downregulation with the activation of the UPK1B-PIK3IP1-PI3K/AKT axis and the adverse clinical outcomes observed in GC.