Linc-ROR orchestrates autophagy suppression and marks gastric cancer via the miR-145-5p/CARMIL1 axis.

Introduction
Gastriccancer (GC), a predominant and lethal gastrointestinal tract malignancy, is distinguished by a multifactorial etiology involving complex genetic and molecular dysregulation (Bray 2022). Despite significant advances in diagnostic and therapeutic strategies, most GC cases are still diagnosed at intermediate or advanced stages, where curative treatments (e.g., surgery, radiotherapy, and chemotherapy) offer limited benefit (Chandra 2021), (Wang 2023). These challenges underscore a critical requirement for the identification of innovative and reliable biomarkers to enable early detection, monitor treatment response, and improve prognostic stratification in GC.
Long non-coding RNAs (lncRNAs) are regulatory transcripts > 200 nucleotides that exert pivotal functions in cancer pathogenesis, progression, metastasis, and immune evasion (Dragomir et al. 2020), (Kapranov 2007). Functionally, lncRNAs modulate essential cellular processes through diverse mechanisms, including epigenetic regulation, alternative splicing, and the translation of functional micropeptides (Bridges et al. 2021), (Gil and Ulitsky 2020). The activities of lncRNAs are highly dependent on their subcellular localization. In the cytoplasm, lncRNAs commonly serve as competing endogenous RNAs (ceRNAs) by sponging microRNAs (miRNAs), thereby preventing miRNA-mediated suppression of target mRNAs. Additionally, cytoplasmic lncRNAs can directly interact with specific proteins to modulate their stability, localization, or activity (Liu et al. 2021), (Statello et al. 2021). Linc-ROR is aberrantly dysregulated in several malignancies and has been implicated as an oncogenic modulator of tumor cell growth, apoptosis, and invasion through various signaling pathways (Peng et al. 2017), (Tian et al. 2021), (Wang 2017), (Zhan 2016), (Mo 2025). Earlier investigations have illustrated that Linc-ROR is noticeably upregulated in GC tissues and is closely related to malignant progression. Silencing Linc-ROR significantly suppresses epithelial–mesenchymal transition (EMT), thereby limiting GC cell growth and invasiveness (Liu et al. 2022), (Mi 2021). Moreover, elevated Linc-ROR expression has been associated with immune evasion mediated by natural killer (NK) cells (Niu et al. 2024). Despite these outcomes, the precise function and mechanisms of Linc-ROR in GC remain mostly unclear and warrant further investigation.
Autophagy is an extremely conserved catabolic process in eukaryotic cells (Miller and Thorburn 2021), (Kawamata et al. 2008). Its role in cancer remains paradoxical: moderate autophagy supports tumor survival under stress conditions (e.g., nutrient deprivation and hypoxia), whereas excessive autophagy can trigger cell death and exert tumor-suppressive effects. Thus, context-dependent modulation of autophagy may be a prospective therapeutic approach in GC (Levy et al. 2017), (Li et al. 2020), (Debnath et al. 2023). The core genetic machinery governing autophagy has been increasingly elucidated. Emerging evidence indicates that lncRNAs play critical regulatory roles in autophagy by orchestrating gene expression across multiple biological pathways (Yang et al. 2017). For instance, upregulation of WT1-AS or suppression of the PI3K/Akt/mTOR pathway can restrain GC progression by triggering cell-cycle arrest and enhancing autophagic activity, with the combination of WT1-AS and rapamycin exhibiting the most pronounced antitumor effects across multiple assays (Zhang 2024). Moreover, Zhang et al. (Wang 2021) demonstrated that MALAT1 competitively binds to ELAVL1 to stimulate the AKT/mTOR pathway, thereby blocking autophagic flux and accelerating GC progression. However, the molecular mechanisms by which Linc-ROR may orchestrate GC progression through the modulation of autophagy remain insufficiently elucidated.
Capping protein and Myosin 1 linker 1 (CARMIL1) is a key regulator of actin dynamics and has an essential function in cytoskeletal remodeling, cell migration, and maintenance of cellular architecture. Previous studies have shown that mutations within critical CARMIL1 domains markedly impair RAC GTPase activation and attenuate migratory capacity (Lim 2025). While phosphorylation of its C-terminal region modulates actin organization and thereby influences cell-cycle progression and proliferation (Sheela 2025). In addition, CARMIL1 has been reported to promote liver cell growth by activating the ERK/mTOR pathway through the TRIM27/p53 axis (Ge 2024). Despite these observations underscoring its potential relevance in cancer biology, the expression profiles and functional roles of CARMIL1 in GC are still mostly undefined, specifically with respect to its possible involvement in autophagy regulation and tumor progression.
Exosomes, a subtype of extracellular vesicles (EVs), are nano-sized lipid bilayer-enclosed vesicles with a range of 40–150 nm in diameter. They carry a diverse repertoire of biomolecules, including nucleic acids, lipids, and proteins, and serve as critical mediators of intercellular communication, with established roles in disease diagnosis, therapy, and prognostic assessment (Kalluri and LeBleu 2020), (Dai 2020), (Zhang 2023). Notably, exosomes are enriched in non-coding RNAs (ncRNAs) (e.g., lncRNAs, miRNAs, and circRNAs), many of which have emerged as promising biomarkers (Li 2021), (Yu 2022). Among them, exosomal Linc-ROR has been shown to facilitate intercellular signaling and promote tumor progression (Sun 2021), (He 2019). However, its diagnostic potential in GC remains unexplored.
Our previous study demonstrated that exosome-associated Linc-ROR was significantly correlated with reduced disease-free and overall survival in individuals with GC. Cox regression analyses (univariate and multivariate) further identified Linc-ROR as an independent predictive factor in GC (Ding 2025). Herein, we investigated the oncogenic function of Linc-ROR in GC progression and uncovered a previously unrecognized mechanism whereby Linc-ROR promotes tumor aggressiveness by modulating autophagy, as supported by extensive molecular and cellular evidence.

Materials and methods

Specimen collection
We brought 20 paired GC tissues and matched neighboring non-cancerous tissues, along with corresponding serum samples from Qilu Hospital of Shandong University. All diagnoses were independently confirmed by two experienced pathologists. No patient had experienced chemotherapy or radiotherapy pre-operation. Serum samples from twenty age- and sex-matched healthy individuals were also included as controls. Informed consent was acquired from each participant. The Medical Ethics Committee of Qilu Hospital, Shandong University, gave its approval for this investigation (No. KYLL-202011–223-1).

Cells and cultures
GC cell lines (AGS and HGC-27) and the normal gastric epithelial cell line (GES-1) were obtained from the Cell Bank of the Chinese Academy of Sciences. Cell lines were validated by short tandem repeat (STR) profiling to be mycoplasma contamination-free. Cells were cultivated in RPMI-1640 medium (Gibco, USA) enriched with 10% heat-inactivated FBS (Invitrogen, USA) and 1% penicillin–streptomycin (100 U/mL penicillin and 100 μg/mL streptomycin). Cultures were maintained at 37 °C in a humidified 5% CO₂ condition, with medium changes every 2–3 days. Cells were passaged at 70%–80% confluence via 0.25% trypsin–EDTA (Gibco, USA) for subsequent trials.

Lentiviral transfection and plasmid transfection
Lentiviral vectors encoding Linc-ROR overexpression or short hairpin RNA (shRNA) for gene silencing, both carrying puromycin resistance, were purchased from GeneCopoeia (Guangzhou, China). To determine the optimal infection conditions, AGS and HGC-27 cells were first subjected to lentiviral transduction at a range of MOIs. Based on preliminary optimization experiments aimed at achieving high transduction efficiency while minimizing cytotoxicity, the MOI yielding the highest transduction efficiency with the lowest cytotoxicity was selected for subsequent experiments (MOI 20 for AGS cells and MOI 10 for HGC-27 cells). Cells were infected with the corresponding lentiviruses for 24 h and then selected with 5 μg/mL puromycin for approximately two weeks to create stable cell lines.
Transient transfection of cells was conducted with pcDNA3.1-CARMIL1 or the corresponding empty vector via Lipofectamine 3000 (Invitrogen, USA) as per the manufacturer’s guidelines. Briefly, cells were plated in plates with six wells and grown to 60%–70% confluence at the time of transfection. For each well, 2.5 μg of plasmid DNA was diluted in Opti-MEM (Gibco) and combined with P3000 reagent, followed by incubation with Lipofectamine 3000. The DNA–lipid complexes were applied to the cells, and incubation was conducted at 37 °C. Following 6–8 h, the medium was substituted with new complete medium, and cell collection was conducted 48 h post-transfection for subsequent experiments. For miRNA modulation, GC cells plated in plates with six wells were transfected with miR-145-5p mimics or inhibitors (GenePharma, China) via Lipofectamine™ 3000 (Invitrogen, USA), with scrambled oligonucleotides used as negative controls (NCs). Table S1 lists nucleotide sequences.

Quantitative reverse transcription PCR (RT-qPCR)
Total RNA isolation from tissues, serum samples, and cultured cells was conducted via the RNA Isolation and Extraction Kit (Fastagen Bio, China) as per the manufacturer’s guidelines. RNA purity and level were measured via a NanoDrop 2000 spectrophotometer, and only samples with A260/280 ratios of 1.8–2.1 were included for analysis. cDNA was generated via the PrimeScript™ RT Reagent Kit (GeneCopoeia, China), and RT-qPCR was conducted via SYBR Green Master Mix (GeneCopoeia) on a QuantStudio™ 5 Real-Time PCR System (Thermo Fisher Scientific, USA). Using GAPDH for mRNA and U6 for miRNA normalization. Relative levels were assessed via the 2−ΔΔCt technique. Table S2 lists the primer sequences.

Western blot
GC cell lysis was conducted via RIPA buffer (Beyotime, China) enriched with a protease inhibitor cocktail (1:100). Tissue samples were first homogenized using a magnetic bead grinder and further disrupted by brief ultrasonication on ice to ensure complete lysis. Protein levels were measured via a BCA Protein Assay Kit. The separation of equal protein quantities (25 μg) was conducted by 10% SDS-PAGE (80 V for stacking, 120 V for resolving) and then transferred onto 0.22 μm PVDF membranes (Millipore, MA, USA) at 300 mA for 90 min. A 1-h blockage of membranes was conducted with 5% skim milk at room temperature, and incubation was conducted overnight at 4 °C with primary antibodies. After rinsing, a 1-h incubation of membranes was conducted with HRP-conjugated goat anti-rabbit IgG secondary antibodies at room temperature. Enhanced chemiluminescence (ECL) reagents were utilized to detect protein signals, which were then photographed with an Amersham Imager 600 RGB system (GE Healthcare, USA). ImageJ (NIH, USA) was utilized to quantify band intensities, which were then normalized to the corresponding loading controls to calculate relative protein expression levels. Table S3 presents the detailed antibody information.

CCK-8 assay
For the cell viability assay, 2 × 103 GC cells/well were grown on plates with 96 wells and maintained overnight under numerous treatment conditions. At 24 and 48 h post-seeding, 10 μL of CCK-8 reagent (Abbkine, China) was applied to each well, and a 2-h incubation was conducted at 37 °C. Absorbance at 450 nm was read via a BioTek Gen5 microplate reader (BioTek, USA). Each condition was experimented with at least three independent experiments with triplicate wells per condition. Relative cell viability was estimated by the normalization of the absorbance of treated wells to that of the control.

Colony formation assay
AGS and HGC-27 cells (1 × 103 cells/well) were cultured in plates with 96 wells and cultivated for 15 days, with medium replaced every four days, until visible colonies formed. A 4% paraformaldehyde solution was utilized for fixation of the colonies for 20 min, 0.1% crystal violet was utilized for staining for 30 min, and PBS was utilized for rinsing. After air drying, plates were photographed, and colonies with > 50 cells were assessed via an Olympus BX51 microscope (Olympus, Japan). Trials were conducted at least three times independently.

Wound healing assay
Log-phase AGS and HGC-27 cells were plated in plates with six wells (5 × 105 cells/well) and cultivated at 37 °C until reaching ~ 95% confluence. A uniform scratch was formed via a sterile pipette tip, and PBS was utilized to gently wash the cells to eliminate debris before adding 2 mL of serum-free medium per well. Photographs of the wound area were taken at 0 and 24 h post-scratch via an inverted microscope (Olympus, Japan), and the migration rate was quantified by measuring the reduction in wound width. Trials were conducted in triplicate.

Transwell assay
For the migration assay, GC cells were plated into the top chambers of 8-μm pore Transwell inserts (Corning, USA), while the bottom chambers were supplied with complete medium with 20% FBS as a chemoattractant. After incubating for 24 h at 37 °C, non-migrated cells on the top surface were gently eliminated with a cotton swab. Migrated cells on the bottom surface underwent fixing with 4% methanol, staining with 0.1% crystal violet (Solarbio, Beijing, China), and counting via an inverted microscope. For the invasion assay, the top chambers were pre-coated with Matrigel (Corning, USA), and all subsequent steps were identical to the migration assay. Trials were conducted at least three times independently.

Cytoskeletal/nuclear isolation
GC cell lysis was conducted on ice via a nuclear-cytoplasmic separation buffer (Beyotime, China) for 10 min, and then spinning was conducted at 800 × g for 10 min to acquire cytoplasmic (supernatant) and nuclear (pellet) fractions. Total RNA isolation from each fraction was conducted via the RNA Isolation and Extraction Kit (Fastagen Bio, China) and analyzed by RT-qPCR to estimate the subcellular localization of Linc-ROR. GAPDH and U6 served as internal controls for the cytoplasmic and nuclear fractions, respectively. Experiments were conducted in triplicate.

Dual luciferase assay
Plasmids with wild-type (WT) or mutated (MUT) Linc-ROR and CARMIL1 3′UTR sequences with predicted miR-145-5p binding sites were constructed by GeneCopoeia (MA, USA). For the dual-luciferase reporter assays, CS-MiT0036-MT06-01 was used as the WT Linc-ROR reporter. To validate the specific interaction between Linc-ROR and miR-145-5p, two single-site mutant reporters (Linc-ROR MUT1 and Linc-ROR MUT2) were generated, each harboring distinct mutations within one predicted miR-145-5p binding site. In addition, a double-mutant reporter (Linc-ROR MUT), in which both miR-145-5p binding sites were simultaneously mutated, was created. Specifically, GC cells were plated in plates with six wells and transfected at ~ 75% confluence with WT or MUT reporter plasmids together with miR-145-5p mimics or NC via Lipofectamine™ 3000 (Invitrogen, USA). Following 48 h, luciferase activities (Firefly and Renilla) were assessed via a dual-luciferase assay kit (Vazyme, Nanjing, China), and the Firefly/Renilla ratio was assessed for normalization. Trials were conducted at least three times independently.

Autophagy flux detection
GC cells were transduced with a tandem mRFP-GFP-LC3B lentiviral construct and plated on glass coverslips in plates with 24 wells. Following the suggested treatments, autophagic flux was assessed by confocal microscopy (Leica, Germany). For each condition, puncta were quantified and averaged from a minimum of five randomly selected fields per experiment using ImageJ (NIH, USA), with analyses performed from three independent biological replicates. Autophagosomes were defined as yellow puncta (mRFP⁺/GFP⁺), while autolysosomes were determined as red-only puncta (mRFP⁺/GFP⁻). Autophagic flux was considered enhanced when red-only puncta were elevated, or the ratio of yellow-to-red puncta was reduced, and impaired when red-only puncta were reduced, or the ratio of yellow-to-red puncta was elevated.

Transmission electron microscope scanning
To evaluate autophagosome and autolysosome dynamics, GC cells stably overexpressing Linc-ROR or transduced with an empty vector were cultured in 75 cm2 flasks until 75% confluence. Then, cells were exposed to 5 μM Everolimus (MCE, USA) for 24 h. Cells were collected, pelleted, and fixed in 0.5 mL electron microscopy-grade fixative (Servicebio, Wuhan, China) at room temperature for 30 min in the dark. Fixed samples were stored at 4 °C until further processing for TEM imaging, which was performed following standard protocols. For quantitative analysis, five randomly selected fields per condition were assessed in each experiment. The autophagosomes were quantified on a per-cell basis and averaged for each field. Analyses were performed from three independent biological replicates.

Xenograft mice experiment
Male BALB/c nude mice (four weeks, 18–20 g) were brought from Beijing Vital River Laboratory Animal Technology Co., Ltd. and kept in specific pathogen-free (SPF) environments. The sample size was determined using variability and effect‐size estimates derived from pilot xenograft experiments (Zhang and Hartmann 2023). Depending on the 3R principles of the institutional animal ethics committee, five animals per group were deemed sufficient to achieve adequate statistical power at α = 0.05. To establish subcutaneous xenografts, a suspension of AGS cells (1 × 10⁷) stably overexpressing Linc-ROR or vector control in 150 μL of PBS was introduced with a subcutaneous injection into the flanks of the mice. Tumor growth was observed every three days, and when tumors reached approximately 2 × 2 mm in size, mice were randomized to receive Everolimus (2.5 mg/kg/day) or vehicle control (PBS, 100 μL/day). Mice were sacrificed upon reaching the largest tumor diameter of 15 mm. Tumors were excised, weighed, and their volumes measured as: volume = (length × width2)/2. The harvested tumor tissues were subsequently processed for immunohistochemical (IHC) analysis.

Bioinformatics analysis
Gene expression profiles of GC, accompanied by matching clinical data, were acquired from The Cancer Genome Atlas (TCGA, TCGA-STAD; https://portal.gdc.cancer.gov) and Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo), including datasets GSE106817, GSE59856, GSE26253, and GSE34942. Raw count data from TCGA and GEO were preprocessed and normalized using standard pipelines in R (version 4.2.0). Differentially expressed lncRNAs between GC and healthy tissues were determined via DESeq2 with thresholds of |log₂ fold change (FC)|≥ 1 and P.adj < 0.05 (Benjamini–Hochberg correction for various testing). Spearman’s rank correlation coefficient was measured to estimate associations between lncRNAs, mRNAs, and miRNAs. Kaplan–Meier survival analyses were conducted to estimate the predictive significance of selected genes, and survival curves were visualized via the R survival and survminer and UALCAN (https://ualcan.path.uab.edu/index.html). P < 0.05 was deemed significant.

Statistical methodology
Continuous variables were reported as mean ± standard deviation (SD) from at least three independent trials. Two-group comparisons were conducted via Student’s t-test or the Mann–Whitney U test. Categorical variables were shown as counts or proportions, and comparisons were conducted via the Chi-square or Fisher’s exact tests. GraphPad Prism 9.0 was utilized for statistical analyses. The diagnostic performance of Linc-ROR was estimated by the measurement of the area under the receiver operating characteristic (ROC) curve (AUC) via the “pROC”. P < 0.05 was deemed significant.

Results

Linc-ROR triggers malignant behaviors of GC cells
We previously identified serum sEV-derived Linc-ROR as an independent prognostic factor in GC (Ding 2025). Consistently, analysis of the TCGA-STAD cohort confirmed its significant upregulation in tumor tissues relative to healthy tissues(Fig. 1A) and revealed a strong association with tumor grade (Fig. 1B). Kaplan–Meier analysis further illustrated that Linc-ROR overexpression predicted inferior overall survival (Fig. 1C). To investigate its functional role, we profiled Linc-ROR expression and found marked overexpression in AGS and HGC-27 cells comparative to GES-1 cells (Fig. 1D). Stable Linc-ROR knockdown cell lines were established, with effective silencing confirmed by RT-qPCR (Figures S1A, C). Loss of Linc-ROR significantly impaired cell proliferation and clonogenicity, as evidenced by CCK-8 and colony formation assays (Figs. 1E–H). In addition, functional assays revealed that Linc-ROR silencing markedly reduced the migration and invasion capabilities of GC cells. Wound healing and transwell assessments illustrated diminished migratory capacity (Figs. 1I-J), while Matrigel-coated chambers further confirmed reduced invasiveness upon Linc-ROR depletion (Figs. 1K-L). Furthermore, subcutaneous xenograft experiments demonstrated that Linc-ROR silencing markedly suppressed GC cell growth in vivo, as evidenced by significantly reduced tumor volume and tumor weight (Figuress S2A-C). Consistently, immunohistochemical analysis revealed a pronounced decrease in Ki-67–positive cells in tumors derived from Linc-ROR -depleted cells, indicating impaired proliferative capacity (Figure S2D). Conversely, enforced Linc-ROR expression led to a substantial increase in its transcript levels (Figures S1B, D), which significantly enhanced GC cell growth, migration, and invasion compared to controls (Figures S3A–H). Collectively, these findings underscore the oncogenic function of Linc-ROR in promoting the malignant progression of GC.

Linc-ROR facilitates oncogenic phenotype by targeting the miR-145-5p axis in GC
To clarify the molecular mechanism of Linc-ROR function, we first determined its subcellular distribution. Cytoplasmic/nuclear fractionation assays illustrated that Linc-ROR is predominantly found in the cytoplasm (Figs. 2A-B), suggesting its potential role as a ceRNA. Given the emerging relevance of lncRNA–miRNA interactions in GC progression (Karreth and Pandolfi 2013), (Tay et al. 2014). Given this, we integrated differential miRNA expression data from three public datasets (TCGA-STAD, GSE106817, and GSE59856) with target prediction using DIANA tools, identifying miR-145-5p as a prospective Linc-ROR downstream target (Fig.2C). Correlation analysis illustrated a significant inverse association between Linc-ROR and miR-145-5p expression in GC tissues (Fig. 2D), which was validated by RT-qPCR (Fig. 2E). Consistently, miR-145-5p expression was noticeably suppressed in GC tissues compared to controls (Figs. 2F–H), and broadly suppressed across multiple solid tumors (Figure S3A), supporting its putative tumor-suppressive role. ROC curve analysis indicated the prospective diagnostic value of miR-145-5p in GC (Figs. 2I–K).
Mechanistically, miR-145-5p levels were significantly upregulated following Linc-ROR knockdown (Figs. 2L-M), whereas Linc-ROR upregulation resulted in pronounced suppression of miR-145-5p (Figs. 2N-O). Bioinformatic analysis predicted two putative binding sites for miR-145-5p within the Linc-ROR sequence at positions 1307–1330 and 2037–2059 (Fig. 2P). Dual-luciferase reporter assays using WT and MUT constructs confirmed the direct interaction: mutations at either or both binding sites significantly restored luciferase activity in GC cells transfected with miR-145-5p mimics compared to WT constructs (Figs. 2Q-R). Collectively, these findings demonstrate that Linc-ROR directly binds to and negatively controls miR-145-5p, thereby functioning as a ceRNA to promote GC progression.
We first confirmed that transfection of miR-145-5p inhibitors or mimics into GC cells effectively suppressed or elevated miR-145-5p expression, respectively (Figures S3B-C). To assess the functional relevance, miR-145-5p inhibitors were introduced into Linc-ROR-silenced AGS and HGC-27 cells. The suppression of miR-145-5p significantly restored cell proliferation (Figs. 3A-B). Similarly, wound-healing assessments illustrated that suppression of miR-145-5p reversed the diminished migration capability induced by Linc-ROR knockdown (Figs. 3C-D). Transwell assays further confirmed that miR-145-5p suppression triggered both migration and invasion abilities in Linc-ROR-deficient cells (Figs. 3E-F). Conversely, transfection of miR-145-5p mimics into Linc-ROR-overexpressing cells partially abrogated the pro-proliferative, migratory, and invasive phenotypes (Figure S5). Collectively, these outcomes illustrate that Linc-ROR triggers GC cell malignancy at least in part by suppressing miR-145-5p activity.

CARMIL1 is a functional target of miR-145-5p in GC
To identify functional targets of miR-145-5p, we conducted cross-platform prediction via seven publicly available algorithms (TargetScan, RNA22, PITA, PicTar, microT, miRanda, and miRmap), all of which consistently highlighted CARMIL1 as a putative downstream target (Fig. 4A). Analysis of TCGA-STAD illustrated that CARMIL1 was significantly higher in tumor tissues than in controls (Fig. 4B). ROC curve analysis demonstrated high diagnostic accuracy of CARMIL1 for GC, with an AUC of 0.917 (95% CI: 0.896–0.939) (Fig. 4C). Moreover, CARMIL1 expression was inversely related to miR-145-5p levels (R = –0.344, P < 0.001; Fig. 4D), but positively associated with Linc-ROR expression in-house cohort (Fig. 4E). KM analysis indicated that elevated CARMIL1 levels predicted lower overall survival in individuals with GC (Figs. 4F–H). Bioinformatic analysis predicted conserved binding sites between miR-145-5p and the 3′UTR of CARMIL1 (Fig. 4I). Dual-luciferase reporter assessments confirmed that mutation of these binding sites significantly relieved the miR-145-5p-mediated repression of CARMIL1 (Figs. 4J-K), validating a direct interaction.
Furthermore, Linc-ROR overexpression significantly increased CARMIL1 expression at the mRNA and protein levels, an effect that was inhibited by co-transfection with miR-145-5p mimics. Conversely, Linc-ROR knockdown suppressed CARMIL1 expression, while miR-145-5p suppression rescued this phenotype (Figs. 4L–Q). Collectively, these findings establish CARMIL1 as a direct downstream effector of the Linc-ROR/miR-145-5p axis, functionally linking this ceRNA network to GC progression.

Linc-ROR suppresses autophagy by activating the CARMIL1/ERK/mTOR pathway
Increasing evidence suggests that dysregulated lncRNAs may contribute to cancer progression by modulating autophagy across diverse physiological and pathological contexts (Ghafouri-Fard 2022), (Zhang et al. 2017). Based on this, we hypothesized that Linc-ROR promotes GC progression by suppressing autophagy. To test this, we evaluated autophagy levels in Linc-ROR-silenced cells transfected with miR-145-5p inhibitors, as well as in Linc-ROR-overexpressing cells co-transfected with miR-145-5p mimics. Linc-ROR knockdown enhanced autophagic activity, while Linc-ROR overexpression markedly inhibited autophagy in GC cells(Figs. 5A-B). Next, we determined whether CARMIL1 is required for Linc-ROR–mediated malignant phenotypes in GC. Cell viability assays showed that overexpression of Linc-ROR significantly enhanced AGS and HGC-27 cell proliferation at 48 h, whereas silencing CARMIL1 substantially attenuated the proliferative advantage conferred by Linc-ROR overexpression (Figures S6A-B). Consistently, wound-healing assays illustrated that suppression of CARMIL1 significantly impaired the enhanced cell migration induced by Linc-ROR overexpression (Figures S6C-D). Furthermore, the enhanced migratory and invasive capacities conferred by Linc-ROR overexpression were reversed upon CARMIL1 silencing (Figures S6E-F). Moreover, immunoblot analyses revealed that CARMIL1 silencing effectively reversed Linc-ROR overexpression–triggered stimulation of ERK1/2 and mTOR signaling, as evidenced by reduced phosphorylation of ERK1/2 and mTOR. This signaling attenuation was accompanied by restoration of autophagic activity, characterized by increased Beclin-1 and LC3-II levels and concomitant reduction of p62 (Figures S6G-H). Given the close association between autophagy dysregulation and EMT (Gundamaraju 2022), we next examined EMT-related markers. Notably, CARMIL1 knockdown markedly attenuated Linc-ROR–induced EMT-associated changes in both GC cell lines, as evidenced by overexpressed E-cadherin and lessened levels of N-cadherin, Vimentin, and Twist1, indicating suppression of EMT (FiguresS6I-J). Collectively, these results demonstrate that Linc-ROR enhances oncogenic roles by activating ERK/mTOR signaling, suppressing autophagy, and promoting EMT in a CARMIL1-dependent manner.
Earlier investigations have illustrated that the MEK/ERK pathway regulates autophagy through multiple downstream effectors (Huang et al. 2023). We discovered that overexpression of Linc-ROR elevated the phosphorylation of ERK and mTOR, while this activity was weakened by miR-145-5p mimics(Fig. 5C). Conversely, Linc-ROR knockdown suppressed p-ERK and p-mTOR levels, which were reversed by miR-145-5p suppression (Fig. 5D). Everolimus, a rapamycin analog with potent mTOR-inhibitory activity (Ohtsu 2013), (Huangfu 2023), significantly upregulated Beclin-1 and LC3 expression and reduced p62 levels in GC cells (Fig.5E), indicative of enhanced autophagic activity. Dual-fluorescence LC3 assays further confirmed that Linc-ROR overexpression suppressed autophagic flux, an effect partially rescued by Everolimus treatment (Figs. 5F-G). Consistently, transmission electron microscopy revealed reduced autophagosome/autolysosome formation in Linc-ROR overexpressing cells, while Everolimus restored autophagic vesicle accumulation (Figs. 5H-I).
Conversely, Linc-ROR depletion robustly enhanced autophagic flux, as evidenced by a significant increase in total LC3 puncta, including both GFP⁺/mCherry⁺ autophagosomes and mCherry-only autolysosomes (Figures S7A-D). Chloroquine-mediated blockade of lysosomal degradation further increased autophagosome accumulation, supporting enhanced autophagic flux rather than defective autophagosome clearance. Consistent with these observations, Linc-ROR knockdown reduced p62 abundance and promoted LC3-II accumulation, effects that were reversed by chloroquine treatment (Figures S7E-F). Collectively, these findings identify Linc-ROR as a negative regulator of autophagy, whose downregulation activates autophagic flux in GC cells.

Everolimus partially counteracts Linc-ROR-mediated tumor progression in vitro and in vivo
To further estimate the therapeutic potential of Everolimus in GC, we performed CCK-8 and colony formation assessments. Everolimus treatment significantly suppressed GC cell proliferation and clonogenic capacity compared to the control group, and notably attenuated the oncogenic effects driven by Linc-ROR overexpression (Figs. 6A-C). In vivo, xenograft experiments demonstrated that Everolimus markedly reduced tumor volume and weight. Importantly, Everolimus partially reversed the tumor-promoting phenotype induced by Linc-ROR (Figs. 6D-F). Consistent with its selectivity for the mTOR axis, western blot analysis revealed that everolimus markedly decreased p-mTOR levels without affecting p-ERK (Figs. 6G-H). Immunohistochemistry of subcutaneous tumors showed that Linc-ROR overexpression increased Ki-67, CARMIL1, p62 and Vimentin while decreasing LC3B and E-cadherin; these molecular alterations were partially reversed by intraperitoneal administration of everolimus (Figs. 6I-J). Together, these results indicate that everolimus mitigates Linc-ROR–mediated tumorigenesis, potentially by inhibiting mTOR signaling to restore autophagy and decelerate EMT progression.

Exosomal Linc-ROR serves as a robust biomarker for GC diagnosis
Exosomal ncRNAs, highly abundant in exosomal cargo, are gaining recognition as diagnostic and prognostic biomarkers (Tang 2021). To assess the clinical relevance of exosomal Linc-ROR in GC, we first isolated exosomes from the conditioned medium of GES-1, AGS, and HGC-27 cells. TEM illustrated the characteristic cup-shaped morphology of isolated vesicles (Fig.7A), while nanoparticle tracking analysis (NTA) illustrated a particle size distribution centered around 130 nm (Fig. 7B). Western blotting validated the enrichment of canonical exosomal markers (ALIX, TSG101, CD63) and the absence of the non-exosomal marker Calnexin (Fig. 7C-D). Subsequent RT-qPCR analysis demonstrated that Linc-ROR was significantly enriched in GC cell-derived exosomes relative to those from GES-1 cells (Fig. 7E). Moreover, RNase digestion assays revealed that exosomal Linc-ROR was resistant to RNase A treatment unless combined with membrane-disrupting reagents (Triton X-100 or proteinase K), indicating its vesicular encapsulation (Figs. 7F-G).
Clinically, Linc-ROR was significantly overexpressed in GC tissues compared to neighboring healthy tissues (Fig. 7H). Consistently, serum-derived exosomal Linc-ROR levels were noticeably higher in GC patients than in healthy controls (Fig. 7I), and a positive relationship was detected between Linc-ROR levels in serum exosomes and matched tumor tissues (Spearman R = 0.565, P < 0.001; Fig. 7J). ROC analysis demonstrated that tissue Linc-ROR yielded an AUC of 0.693 (Fig. 7K), while serum exosomal Linc-ROR achieved superior diagnostic performance with an AUC of 0.861 (Fig. 7L). Notably, exosomal Linc-ROR outperformed carcinoembryonic antigen (CEA), and their combination further improved diagnostic accuracy (AUC = 0.876; Fig. 7M). Collectively, these findings highlight that circulating exosomal Linc-ROR is significantly upregulated in GC and may act as a prospective non-invasive biomarker for GC diagnosis.
Together, Linc-ROR promotes GC aggressiveness by sequestering miR-145-5p, thereby derepressing CARMIL1 and activating the ERK–mTOR axis, which suppresses autophagic flux and enhances malignant phenotypes. Pharmacological inhibition of mTOR by Everolimus restores autophagy and mitigates Linc-ROR–driven tumor progression. Clinically, circulating exosomal Linc-ROR is significantly elevated in GC patients and demonstrates high diagnostic performance, highlighting its prospective as a non-invasive marker and therapeutic target (Fig. 8).

Discussion
LncRNAs, a burgeoning class of transcripts implicated in diverse cellular processes, are frequently dysregulated in various cancer types, exerting significant influence on oncogenic progression, diagnostics, and prognostication (Yuan 2020), (Wei 2020). This work presents numerous interesting findings: First, Linc-ROR promotes GC progression by sponging miR-145-5p and elevating CARMIL1, which stimulates the ERK/mTOR pathway and suppresses autophagy. Second, Pharmacological inhibition with Everolimus reverses Linc-ROR-induced oncogenic effects, highlighting a potential therapeutic vulnerability. Third, serum exosomal Linc-ROR is significantly elevated in GC patients, serving as a promising noninvasive biomarker.
Growing evidence indicates that lncRNAs deeply contribute to the control of gene expression, differentiation, growth, and apoptosis across diverse cellular contexts (Herman et al. 2022), (Ebrahimi 2022). Beyond their functional roles, lncRNAs often display cancer-type-specific expression patterns and are noticeable in bodily fluids, making them attractive candidates for noninvasive cancer biomarkers (Beylerli et al. 2022), (Badowski et al. 2022), (Jiang et al. 2017). Herein, we detected a marked overexpression of exosomal Linc-ROR in the serum of individuals with GC, which correlated strongly with its expression in tumor tissues, indicating its prospective as a circulating biomarker for GC. Notably, the functional impact of lncRNAs is closely tied to their subcellular localization (Ferrer and Dimitrova 2024), (Bhan et al. 2017); cytoplasmic lncRNAs are often considered ceRNAs by sponging miRNAs and modulating post-transcriptional regulation (Beermann et al. 2016). Consistent with this model, our nucleocytoplasmic fractionation assays confirmed that Linc-ROR is predominantly found in the GC cells’ cytoplasm. Mechanistically, we found that Linc-ROR triggers GC cell aggressiveness by sequestering miR-145-5p, thereby derepressing CARMIL1. Together, these findings uncover a novel Linc-ROR/miR-145-5p/CARMIL1 regulatory axis facilitating GC progression. Nonetheless, whether cytoplasmic Linc-ROR exerts additional functions through direct interaction with RNA-binding proteins or signaling complexes warrants further investigation. Notably, several previous studies have reported downregulation of Linc-ROR in cancer, which appears to be inconsistent with our findings (Yu 2020), (Soghala 2022). Such discrepancies may arise from differences in cohort size, patient composition, or intratumoral heterogeneity across studies. Beyond these factors, the divergent observations may also reflect context-dependent regulation of Linc-ROR across distinct GC molecular subtypes and disease stages. Emerging evidence indicates that ncRNA expression profiles vary substantially among GC subtypes, including intestinal versus diffuse types and microsatellite-stable tumors (Han 2017), (Zhu 2023). Importantly, our findings are in line with a growing body of literature supporting an oncogenic function of Linc-ROR in multiple malignancies, wherein Linc-ROR promotes tumor progression by modulating cell motility, survival, and stress-adaptive signaling pathways (Wen et al. 2023), (Shi et al. 2017), (Hou 2018). In this work, we provide new evidence that Linc-ROR exerts pro-tumorigenic effects in GC in a CARMIL1-dependent manner. Taken together, these results underline the context-specific nature of Linc-ROR function in GC and underscore the importance of considering tumor subtype, disease stage, and molecular source when interpreting its biological role.
Increasing evidence highlights the ability of lncRNAs to fine-tune autophagy by controlling the expression of autophagy-related genes and pathways. For instance, EIF3J-DT has been shown to trigger autophagy and promote chemoresistance in GC by sponging miR-188-3p and upregulating ATG14 (Luo 2021). In contrast, DANCR inhibits autophagy to drive gastric tumor progression (Cheng et al. 2021). Herein, we illustrated that Linc-ROR suppresses autophagic activity in GC cells in vitro and in vivo, an effect that was reversed by miR-145-5p restoration. Mechanistically, Linc-ROR was considered a ceRNA for miR-145-5p, thereby derepressing CARMIL1, which subsequently activates the ERK/mTOR signaling cascade. Given that mTOR functions as a master negative regulator of autophagy, its activation by upstream signals (e.g., MAPK/ERK or Akt) inhibits autophagosome formation (Kim and Guan 2015), (Bork 2020), (Rakesh et al. 2022). Our outcomes illustrate that the Linc-ROR/miR-145-5p/CARMIL1 axis contributes to autophagy suppression via ERK/mTOR activation.
To explore therapeutic relevance, we employed everolimus, a clinically approved mTOR inhibitor and a Rapamycin analog with improved pharmacological properties (Houghton 2010), (Yao 2016), (Morviducci 2018). Treatment with everolimus effectively reversed Linc-ROR-induced malignant phenotypes, suggesting that autophagy restoration may underlie its anticancer efficacy in GC. However, the clinical benefit of everolimus as monotherapy in solid tumors remains limited, largely due to incomplete mTORC1 inhibition and activation of compensatory feedback loops (Ohtsu 2013), (Fukamachi 2019). Moreover, adverse events and emerging resistance remain challenges that warrant further investigation.
Mounting evidence suggests that circulating lncRNAs, particularly those encapsulated within exosomes, hold promise as noninvasive biomarkers for cancer diagnosis due to their stability and cancer-type specificity (Tang 2021). Several investigations have underlined the potential of plasma-derived exosomal lncRNAs as diagnostic biomarkers for early GC. For instance, GClnc1 has been reported to effectively distinguish early GC from precancerous lesions, such as chronic atrophic gastritis and intestinal metaplasia, even in patients with negative conventional gastrointestinal tumor markers (CEA, CA72-4, and CA19-9) (Guo 2023). Similarly, LNCUEGC1 is highly expressed in plasma exosomes from early GC patients, achieving an AUC of 0.8760, reliably differentiating early GC from precancerous conditions (Lin 2018). Serum-derived exosomal lncRNAs have increasingly attracted attention as non-invasive biomarkers for GC. For example, (Cai 2025) developed a diagnostic model comprising four lncRNAs (RP11.443C10.1, CTD2339L15.3, LINC00567, and DGCR9), which demonstrated promising potential for detecting individuals with GC who are negative for conventional gastrointestinal tumor markers. In this work, a strong correlation was observed between serum exosomal and tissue-derived Linc-ROR, with the exosomal form demonstrating superior sensitivity and specificity. Notably, the diagnostic performance of serum exosomal Linc-ROR surpassed that of CEA. These outcomes underline the potential of exosomal Linc-ROR as a reliable biomarker for GC detection. The enhanced diagnostic accuracy may be attributed to the active secretion of Linc-ROR by tumor cells into the circulation via exosomes, providing a dynamic and real-time reflection of tumor burden. Unlike tissue-based assays or conventional serum markers, exosomal Linc-ROR are highly stable in serum and more likely to capture tumor heterogeneity. Moreover, compared with previously reported multi-lncRNA panels, exosomal Linc-ROR offers the practical advantage of a single, serum-detectable marker, while its involvement in autophagy and tumor progression provides mechanistic insight and suggests potential as a therapeutic target.
Nevertheless, our investigation has limitations that warrant mention. First, the relatively small clinical cohorts analyzed in this study may limit the generalizability of our outcomes. Large-scale independent clinical cohorts are required to further validate the diagnostic values of serum exosomal Linc-ROR in GC. Second, our mechanistic investigations of Linc-ROR were largely confined to cellular and animal models, necessitating additional validation of its clinical relevance in GC patients. Third, the interactions of Linc-ROR with other signaling pathways and its impact on the immune microenvironment merit more in-depth investigation.

Conclusion
In summary, our study identifies serum exosomal Linc-ROR as a promising noninvasive biomarker for GC. Mechanistically, Linc-ROR promotes tumor progression by sponging miR-145-5p, thereby upregulating CARMIL1 and stimulating the ERK/mTOR pathway, which leads to autophagy inhibition. Importantly, our data suggest that therapeutic co-targeting of Linc-ROR and mTOR with Everolimus may represent a novel and effective strategy for GC treatment, providing valuable insights into both molecular pathogenesis and clinical intervention.

Supplementary Information
Below is the link to the electronic supplementary material.