Abstract

Background

   Serine/threonine kinase 3 (STK3) is recognized as a key regulator in
   Hippo pathway and a tumor-suppressing gene in various cancer types.
   However, its non-canonical role has been gradually revealed in cancer
   development.

Methods

   Our objective is to elucidate the upregulation pattern and molecular
   mechanisms of STK3 in advancing gastric cancer (GC) progression. The
   regulation of YAP1 on STK3 was assessed through a combination of bulk
   and single-cell RNA-sequencing, Western blot, ChIP-qPCR, gene knockout
   mouse models, and functional rescue assays. The oncogenic roles of STK3
   were confirmed through subcutaneous xenograft formation models and
   functional assays including spheroid formation and organoid growth. The
   phosphorylated target of STK3 was revealed by co-immunoprecipitation
   and in vitro kinase assays. STK3-targeted drugs were screened out by
   molecular docking and cellular thermal shift assay (CETSA).

Results

   Reduction of YAP1 significantly impaired STK3 expression at both mRNA
   and protein levels, and deletion of STK3 partially attenuated the
   oncogenic activity of YAP1. Notably, MNNG-induced tumors in
   Yap1^−/−Taz^−/− mice exhibited decreased STK3 expression. Knockdown of
   STK3 led to reduced expression of stemness markers and xenograft
   growth, while sensitizing GC organoids and xenografts to 5-fluorouracil
   treatment. Mechanistically, the direct interaction between STK3 and
   GSK-3β promoted GSK-3β phosphorylation and β-catenin nuclear
   accumulation, and thus the activation of Wnt signaling. Furthermore,
   aminopterin demonstrates as a promising STK3-targeted small molecule
   with remarkable effectiveness in inhibiting GC cell malignance and
   xenograft growth.

Conclusions

   STK3 was identified as a transcriptional target of YAP1, leading to
   enhanced DNA repair ability and stemness acquisition during GC
   progression by activating Wnt/β-catenin activity through GSK-3β
   degradation. Moreover, STK3-targeted therapy offered a novel approach
   to concur acquired chemo-resistance in GC patients.

Graphical Abstract

   [78]graphic file with name 12943_2025_2391_Figa_HTML.jpg

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12943-025-02391-x.

   Keywords: STK3, YAP1, Wnt signaling, Gastric cancer

Introduction

   Gastric cancer (GC) remains an increasingly severe world health issue
   worldwide and is more common in Asian countries. According to the
   Global Cancer Statistics 2022, GC was certified as the 5th most
   frequently occurring carcinogenesis and the 4th leading cause of
   cancer-related death all over the world [[79]1]. Comprehension and
   treatment strategies for GC patients have undergone significant
   development over recent decades. Various clinical approaches have been
   employed for GC patients, showcasing notable clinical benefits while
   maintaining acceptable side effects [[80]2, [81]3]. However, the
   emergence of drug resistance due to prolonged treatment remains a
   formidable challenge in GC therapy [[82]4]. Therefore, there is still
   an urgent need to unravel the underlying molecular mechanisms to
   address the acquired resistance to conventional therapies effectively.

   The Hippo pathway is commonly recognized as a significant participant
   in multiple cancer progression. Briefly, the serine/threonine-protein
   kinases 4/3 (STK4/3, also known as mammalian Ste20-like kinases
   MST1/2), along with salvador homolog 1 (SAV1), initiate the signaling
   transduction by phosphorylating the large tumor suppressor 1/2 kinases
   (LATS1/2) and subsequently the downstream transcriptional coactivator
   Yes-associated protein 1 (YAP1). Phosphorylated YAP1 will be
   sequestered in the cytoplasm and interact with 14-3-3 complex, leading
   to its degradation through a ubiquitination-dependent process [[83]5].
   In pathological conditions, the deactivation of the Hippo pathway will
   result in the de-phosphorylation and nuclear translocation of YAP1.
   Within the nucleus, YAP1 can engage with transcription factors to
   stimulate cell proliferation and suppress apoptosis, potentially
   causing organ enlargement and fostering tumorigenesis [[84]6]. Mainly
   induced by Helicobacter pylori (H. pylori) infection, the
   hyperactivation of Hippo-YAP1 signaling is commonly observed in GC
   patients and demonstrates close correlation with aggressive
   characteristics of cancer cells [[85]7–[86]9]. Meanwhile, Our previous
   work demonstrated that the YAP1-mediated CTGF activation contributed to
   the resistance of chemotherapy in GC patients by enhancing
   microenvironment stiffness [[87]10]. Therefore, there is noteworthy
   clinical potential in elucidating the transcriptional targets of YAP1
   that could accelerate GC progression and therapy resistance, providing
   valuable insights for clinical implementation.

   We commenced this study with an initial endeavor to identify novel
   downstream oncogenes of the YAP1 signaling by investigating the RNA
   sequencing (RNA-seq) results of siRNA-mediated YAP1 knockdown in GC
   cell lines. Surprisingly, STK3 emerged as the sole downregulated member
   among all serine/threonine kinases following YAP1 deletion. This
   unexpected discovery prompted further experiments to elucidate the
   potential bidirectional regulation between YAP1 and STK3. Upon deeper
   analysis of STK3 expression patterns in GC patients, we unveiled a
   frequent amplification of STK3 and significant correlation with poorer
   clinical outcomes, suggesting that this widely recognized tumor
   suppressor gene may exhibit oncogenic properties in GC progression.

   Building upon these findings, our objective was to delve into the
   oncogenic roles of STK3 and the underlying molecular mechanisms.
   Through bioinformatic analysis and both in vitro and in vivo
   validations, we illustrated the pro-tumorigenic functions of STK3 in
   regulating DNA damage response and stemness acquisition.
   Mechanistically, STK3 was found to directly interact with GSK-3β,
   leading to its phosphorylation, which would abolish the nuclear
   translocation of β-catenin and subsequent activation of the Wnt
   signaling pathway. In addition, we performed virtual screening and in
   vivo validation of novel STK3 inhibitors, aiming to enrich our
   armamentarium against chemoresistance in GC treatment.

Materials and methods

Cell lines, organoids, primary samples, and clinical cohorts

   Human GC cell lines MKN1, MKN7, MKN28, and NCI-N87 were purchased from
   American Type Culture Collection (ATCC) and routinely maintained in a
   humidified incubator at 37ºC in 5% CO[2]. The culture medium was
   composed of RPMI-1640 medium (Gibco), 10% fetal bovine serum (FBS,
   Gibco), and 1% penicillin-streptomycin (PS, Gibco). The conduction of
   GC organoid models was in accordance with previously instituted
   protocol [[88]11] and as follows: GC tumor tissues, sizing from 0.5 to
   1 cm^3, were collected, rinsed, minced, and incubated at 37◦C for 1 h.
   The suspension was quenched using a cold culture medium, followed by
   filtration through a 70-micrometer strainer and then centrifugation at
   400 g for 5 min. Cell pellets were resuspended with Matrigel (BD
   Biosciences) and seeded in the well to establish a 3D culture model.
   The Hong Kong GC tissue microarray cohort comprising 278 cases was
   collected between 2002 and 2014 at Prince of Wales Hospital.

In vitro functional assays

   Cell transfections in this work were carried out using Lipofectamine
   2000 Transfection Reagent (ThermoFisher Scientific). The information of
   siRNAs and plasmids adopted in this study was recorded in Supplementary
   Table [89]S1-[90]2, and primary and secondary antibodies adopted in
   Western blot were recorded in Supplementary Table [91]S3. Detailed
   procedures of functional assays, such as monolayer colony formation,
   cell invasion, and spheroid formation assays, were recorded in
   Supplementary Materials and Methods.

Immunohistochemistry (IHC) staining

   The information on antibodies adopted in IHC staining was provided in
   Supplementary Table [92]S4. The procedure of IHC was as follows: after
   deparaffinization and dehydration, tissue sections in 5 μm were treated
   with ethylenediaminetetraacetic acid (EDTA) antigen retrieval solution
   (Beyotime) for antigen retrieval. The sections were then incubated with
   primary antibodies overnight at 4℃ and subsequently incubated with
   HPR-conjugated secondary antibodies. Treatment with DAB Substrate Kit
   (Dako) was performed to achieve visualization of protein expression.
   The immunoreactive score for protein expression was quantified based on
   the percentage of positively stained tumor cells and staining
   intensity, and analyses were conducted using ImageJ software and
   recorded in Supplementary Table [93]S5.

Fluorescence in situ hybridization (FISH)

   STK3 probe (Empire Genomics) was employed to detect the copy number
   changes. PreTreatment Kit 1 (KA2375, Abnova) was performed to pretreat
   formalin-fixed, paraffin-embedded (FFPE) tissue sections. Copy number
   gain (CNG) (3–5 spots per cell) and gene amplification (copy number > 5
   per cell) were defined depending on the probe signal.

Immunofluorescence (IF) assay

   Slides embedded with treated cells were initially fixed by 4%
   paraformaldehyde (PFA) and then blocked in a Phosphate buffer saline
   (PBS) buffer containing 0.1% Triton X-100 (Sigma-Aldrich) and 1% bovine
   serum albumin (BSA, Millipore). Subsequently, the slides were incubated
   overnight at 4℃ with primary antibodies, followed by a one-hour
   incubation at room temperature with corresponding secondary antibodies.
   Details of the primary and secondary antibodies are provided in
   Supplementary Table [94]S6. For nuclear staining, the sections were
   treated with 1 µg/mL DAPI (Sigma-Aldrich) for 15 min at room
   temperature. Imaging was performed by adopting a Zeiss laser scanning
   microscopy (LSM) 880 confocal microscope.

RNA extraction and RNA-seq analysis

   The total RNA of YAP1 or STK3 knockdown cells were extracted by RNeasy
   kit (Qiagen). RNA quality was assessed by Tapestation (Agilent). The
   library was prepared using the Illumina Truseq RNA Kit (Illumina). The
   sequencing was conducted by the NovaSeq 6000 platform (Illumina) with
   single end read 100 bp. Reads were quality-checked with “FastQC”
   (v0.12.0), and “Trim_galore” (v0.6.4) was adopted for sequence
   trimming. The raw sequencing reads were aligned to the Homo sapiens
   genome assembly GRCh37 (hg19) from the NCBI database using “HISAT2”
   (v2.1.0) [[95]12]. Gene expressions were quantified by “SAMtools”
   (v1.14) [[96]13] and “FeatureCount” (v1.6.4) [[97]14]. The
   differentially expressed genes (DEGs) were identified by the R package
   “DESeq2” (v1.38.1) [[98]15]. The enrichment analysis based on Gene
   Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG)
   database was performed by the R package “ClusterProfiler” (v4.6.2)
   [[99]16]. The gene set enrichment analysis (GSEA) quantifies the
   enrichment level through the running sum score beneath the enrichment
   curve, where the absolute value of enrichment score (ES) reflects the
   strength of enrichment, and the sign of ES reveals its direction of
   signaling regulation. The GSEA was conducted by R package
   “ClusterProfiler” (v4.6.2). We also uploaded the processed bulk-RNA-seq
   data generated in this work to Supplementary Table [100]S7.

Public dataset-based bioinformatic analysis

   Two public GC cohorts were adopted in this work, namely The Cancer
   Genome Atlas-stomach adenocarcinoma cohort [[101]17] (TCGA-STAD) and
   the Asian Cancer Research Group (ACRG) cohort [[102]18]. For the
   functional enrichment analysis regarding the STK3 expression level, we
   first evaluated the whole genomic expression level alteration by
   comparing the 10% samples (n = 37) with the lowest STK3 expression
   (STK3^− samples) and the 10% samples with the highest STK3 expression
   (STK3^+ samples). DEGs were identified by R package “DESeq2” with the
   criteria of| log[2](Fold Change)| > 1 and P-value < 0.05. The
   enrichment analysis were performed using R package “ClusterProfiler”.
   Pearson’s correlation analysis was conducted based on data from TCGA,
   ACRG, Cancer Cell Line Encyclopedia (CCLE,
   [103]https://sites.broadinstitute.org/ccle/), and Clinical Proteomic
   Tumor Analysis Consortium (CPTAC,
   [104]https://proteomics.cancer.gov/programs/cptac) (calculated by R
   package “stats” (v4.1.3)). The CCLE database includes the mRNA
   expression profile of 1450 cell lines belonging to multiple organs
   including stomach. The expression data was presented in log[2](TPM + 1)
   and downloaded from online database DepMap portal
   ([105]https://depmap.org/portal/data_page/) [[106]19]. The binding
   motifs of YAP1/TEAD4 on the corresponding promoter region of STK3 were
   predicted by the Eukaryotic Promoter Database
   ([107]https://epd.epfl.ch//index.php) and JASPAR 2022 database
   ([108]https://jaspar.genereg.net). The drug sensitivity analysis was
   conducted by an online server Gene Set Cancer Analysis
   ([109]https://guolab.wchscu.cn/GSCA/#/drug, GSCA). The IC50 of 265
   small molecules in 860 cell lines and its corresponding mRNA gene
   expression profile from Genomics of Drug Sensitivity in Cancer (GDSC)
   were collected and merged by GSCA. Pearson correlation analysis was
   performed to get the correlation between STK3 mRNA expression and drug
   IC50. P-value was adjusted by false discovery rate (FDR) [[110]20]. The
   original data of chromatin immunoprecipitation sequencing (ChIP-seq)
   can be accessed in GEO database under the accession number
   [111]GSE44416 [[112]21]. The analysis and visualization were achieved
   by the online tool UCSC Genome Browser ([113]https://genome.ucsc.edu/).
   Generally, the ChIP-seq data was retrieved from Cistrome Data Browser
   ([114]http://cistrome.org/db/#/) and then imported into UCSC Genome
   Browser. Then the region of observation in the track illustrations was
   narrowed into STK3 encoding area (chr8:98433047–98825712) and then
   zoomed out by 1.5 times to include the promoter region. The
   illustration was downloaded through the “View– PDF” module.

Single-cell RNA-seq (scRNA-seq) analysis

   The scRNA-seq analysis was carried out based on a public dataset
   ([115]https://dna-discovery.stanford.edu/research/datasets/, “Gastric
   scRNA-seq” dataset) [[116]22] using the R package “Seurat” (v4.0.2)
   [[117]23]. Cells expressing fewer than 200 genes, having more than 20%
   mitochondrial genes, or exhibiting an outlier number of unique
   molecular identifiers were removed. Genes detected in fewer than 3
   cells were also excluded. Data normalization was conducted according to
   default settings [[118]22]. The gene markers for cluster labeling and
   the reference for the selection of these markers were recorded in
   Supplementary Table [119]S8. The “DimPlot”, “DotPlot” “FeaturePlot”,
   and “VlnPlot” functions integrated in “Seurat” package were adopted for
   visualization. The STK3^+ cells were identified as cells with
   upregulated STK3 expression level within the “cancer cell” cluster. The
   Gene Set Variation Analysis (GSVA) was conducted by R package “GSVA”
   (v1.46.0) [[120]24]. The relative differentiate state of cells was
   predicted by R package CytoTRACE (v0.3.3) [[121]25]. Additionally,
   single-cell pseudotime analysis was performed by Monocle2 in the R
   package “monocle” (v2.27.0) [[122]26]. After filtering and
   dimension-reducing under the recommended parameters and procedure, the
   cells were ordered and visualized with the functions
   “plot_cell_trajectory” and “plot_pseudotime_heatmap”.

Structure-based virtual screening

   The 3D structure files of STK3 were retrieved from Protein Data Bank
   (PDB, [123]https://www.rcsb.org/) with the accession number “8A66”. The
   candidate compound library was composed of 4511 small molecules with
   literature evidence proving their anti-cancer effectiveness. Molecular
   docking was performed by Autodock 4.2.6 [[124]27], and Lamarckian
   genetic algorithm was applied for the docking procedure. The detailed
   parameters were recorded in Supplementary Table [125]S9. Information of
   the compound library and their corresponding binding affinities with
   STK3 were recorded in Supplementary Table [126]S10.

In vivo studies

   The Yap1^floxed/+;Taz^floxed/+ murine line was obtained by crossing
   Yap1^floxed/+ mice (a kind gift from professor Duojia Pan, UT
   Southwestern Medical Center) and Taz^floxed/+ mice (produced by
   Biocytogen, Beijing, China). Genomic DNA extracted from tail biopsies
   was used to evaluate offspring genotype. Ubc-Cre/ERT2 mice were
   purchased from the Jackson laboratory. Yap1^−/−Taz^−/− mice were
   generated by cross-breeding Yap1^floxed/floxed; Taz^floxed/floxed with
   the tamoxifen-inducible Ubc-Cre/ERT2 mice [[127]28]. For subcutaneous
   xenograft formation assays, NCI-N87 cells (10^6 cells/mouse) were
   treated with shCtrl or shSTK3 (Vector backbone: pLKO.1; target
   sequence: 5’-CGGCGCCUAAGAGUAAACUAA-3’(shSTK3-1);
   5’-GGACUACUUUGAUAAGCAA-3’ (shSTK3-2)) and then subcutaneously injected
   into 4-week-old NOD scid gamma (NSG) mice (n = 10/group). Tumor size
   was measured every two days by a digital caliper. The mice were
   sacrificed 16 days after the injection, and the xenografts were
   harvested, weighted, and possessed for further examination. For
   evaluating the synergistic effect of STK3 depletion and 5-fluorouracil
   (5-FU), four groups of NSG mice (n = 6/group) were subcutaneously
   inoculated with NCI-N87 cells and administered with vehicle (PBS),
   shSTK3, 5-FU (20 mg/kg, twice per week), shSTK3 + 5-FU, respectively.
   For evaluating the synergistic effect of aminopterin (Amin) and 5-FU,
   four groups of NSG mice (n = 6/group) were subcutaneously inoculated
   with NCI-N87 cells and administered with vehicle (PBS), Amin
   (0.15 mg/kg, twice per week), 5-FU (10 mg/kg, twice per week),
   Amin + 5-FU, respectively. The mice were sacrificed after 28-day
   administration, and the xenografts were harvested and evaluated. All
   mice experiments were authorized by the Animal Ethics Experimentation
   Committee (AEEC) from CUHK. The records for xenograft weights are
   recorded in Supplementary Table [128]S11, and the parameters for power
   analysis are detailed in Supplementary Table [129]S12.

Statistical analysis

   Two-tailed student’s t-test was adopted to demonstrate the significance
   between assay groups and control in gene expression level comparison
   and the functional assays. Pearson’s correlation was used to conduct
   correlation analysis. Statistical analyses were performed by Graph Pad
   Prism 8.0 (GraphPad). The co-expression index of STK3 and YAP1 was
   defined as the sum of their expression levels in the same sample. The
   clinical significance of gene expression was assessed by Kaplan-Meier
   curve and conducted by R package “survival”. Data were expressed as
   mean ± standard error of the mean (SEM) of triplicate independent
   experiments. Two-tailed P-value of less than 0.05 was defined as
   statistically significant, and P-value of less than 0.001 was
   considered highly significant.

Results

   RNA-seq analysis was conducted on YAP1-deleted GC cell lines to
   identify novel YAP1 targets. And the results highlighted STK3 as the
   sole downregulated STK in both MKN28 and NCI-N87 (Fig. [130]1A). The
   knockdown of YAP1 downregulated STK3 expression in GC cell lines with
   high basal level of YAP1 at both protein and mRNA level (Fig. [131]1B,
   Supplementary Fig. [132]S1), whereas the overexpression of YAP1 led to
   an increase of STK3 expression in YAP1 lowly expressed cell lines MKN1
   and MKN7 (Fig. [133]1C). According to JASPER and EPD, a potential
   YAP1/TEAD4 binding motif was identified in the promoter region of the
   STK3 gene, 992 bps upstream of the transcription start site
   (Fig. [134]1D). A peak representing the TEAD4 binding site was
   identified in the promoter region of STK3 gene in GC cell lines MKN28
   and SNU216 (Fig. [135]1E). Further validation of the ChIP-seq
   prediction was then conducted by ChIP-qPCR assay. (Fig. [136]1F). Two
   well-known YAP1 inhibitors, verteporfin (VP) and CA3 (CIL56), were
   demonstrated to effectively reduce the expression levels of both YAP1
   and STK3 in the GC cell lines MKN28 and NCI-N87 (Fig. [137]1G-H).
   Regarding mice models, Yap1^−/−Taz^−/− transgenic mice were treated
   with Methylnitronitrosoguanidine (MNNG) oral delivery for 6 months to
   generate GC. And the expression level of Stk3 was significantly reduced
   in Yap1/Taz deficient mice (Fig. [138]1I). Rescue assays were
   subsequently conducted to ensure the engagement of STK3 in the
   YAP1-mediated oncogenic function. As revealed by the invasion and
   colony formation assays, overexpression of YAP1 led to exaggerated
   oncogenic phenotypes of GC cell line MKN28 and NCI-N87, while STK3
   depletion inhibited the malignant features. However, the upregulation
   of YAP1 was unable to activate cell proliferation and metastasis when
   STK3 was maintained at a low level (Fig. [139]1J-K, Supplementary Fig.
   [140]S2). Further exploration of the potential feedback effect of STK3
   on canonical Hippo signaling pathway was conducted in STK3-deleted GC
   cells. Western blot analysis revealed a decrease in nuclear YAP1
   expression, while phosphorylated YAP1, which is primarily present in
   the cytoplasm, was upregulated (Supplementary Fig. [141]S3). These
   results indicate a deactivation of YAP1 signaling, which was further
   proved by the reduced expression levels of YAP1 and CTGF (Supplementary
   Fig. [142]S4).

Fig. 1.

   [143]Fig. 1
   [144]Open in a new tab

   STK3 is transcriptionally regulated by YAP1-TEAD4. (A) RNA-seq analysis
   based on YAP1-deleted cell lines. (B) Western bolt revealed the
   downregulation of STK3 protein expression level after YAP1 knockdown in
   MKN28 and NCI-N87. (C) STK3 level was upregulated after YAP1
   overexpression in MKN1 and MKN7 cells. (D) A potential YAP1/TEAD4
   binding motif was identified in STK3 promoter region. (E, F) YAP1-TEAD4
   could directly bind to the promoter region of STK3 and exaggerate its
   expression, which is confirmed by public ChIP-seq dataset analysis and
   ChIP-qPCR assay. (G, H) Both VP and CA3 could downregulate the protein
   expression level of STK3. (I) Stk3 expression was significantly
   downregulated in Yap1^−/−Taz^−/− transgenic mice. (J, K) STK3 knockdown
   impaired the colony formation and invasion capacity of MKN28 and
   NCI-N87, while overexpressing YAP1 failed to rescue the detriment. (*,
   P < 0.05; **, P < 0.01; ***, P < 0.001)

   Serial sections of GC primary samples were stained by YAP1 and STK3,
   and the results demonstrated the co-expression of them in the same
   regions. As for the subcellular localization, YAP1 upregulation was
   accumulated in nuclei, while STK3 expression was predominantly
   localized in the cytoplasm (Fig. [145]2A-B). Statistically, the protein
   expression levels of YAP1 and STK3 were correlated in our in-house
   primary GC cohort (R = 0.50) and public proteomic databases CPTAC
   (R = 0.47) (Fig. [146]2C-D). Two commonly used databases, namely TCGA
   and ACRG, were adopted to investigate the correlation between STK3 and
   YAP1/TEAD4 expression at mRNA levels. The findings further
   substantiated the positive correlation between the expression levels of
   STK3 and YAP1, indicated by Pearson’s R values of 0.62 and 0.28,
   respectively. Additionally, the expression of STK3 and TEAD4 exhibited
   a positive correlation in two TCGA and ACRG cohorts, with Pearson’s R
   values of 0.55 and 0.55, respectively (Fig. [147]2E-F). A close but not
   significant correlation with poor survival rate was identified in the
   STK3-YAP1 co-overexpression group in TCGA cohort (P = 0.094)
   (Supplementary Fig. [148]S5). Meanwhile, Co-upregulation of STK3 and
   YAP1 was also identified in both all cancer cell lines (R = 0.50) and
   STAD cell lines only (R = 0.38) (Fig. [149]2G). Public GC scRNA-seq
   dataset was adopted, and the results illustrated that STK3 and YAP1
   upregulation were largely enriched in the cancer cells within GC TME
   (Fig. [150]2H-I). By further subgrouping the cancer cells, the STK3^+
   and YAP1^+ cells were classified into the same Seurat clusters (cluster
   4 and 5) (Fig. [151]2J-K).

Fig. 2.

   [152]Fig. 2
   [153]Open in a new tab

   STK3 shows co-upregulation with YAP1 in tumor tissues and cell lines
   across various GC cohorts. (A, B) Representative IHC images of
   YAP1-STK3 dual-negative and dual-positive GC primary samples. (C, D)
   Co-upregulation of YAP1 and STK3 was observed in Hong Kong cohort and
   proteome dataset STAD-CPYAC with a significant correlation. (E, F) The
   mRNA level of STK3 was significantly correlated with both YAP1 and
   TEAD4 expression in TCGA and ACRG cohort. (G) The mRNA expression of
   YAP1 and STK3 was remarkably associated in both all cancer cell lines
   and STAD cell lines. (H) UMAP reduction and cell type classification of
   a public GC scRNA-seq atlas. (I) The scRNA-seq data revealed that YAP1
   and STK3 were largely overexpressed in cancer cells. (J-K) STK3^+ and
   YAP1^+ cancer cells were clustered in the same sub-clusters

   A large proportion of GC samples (56.1%) in the TCGA cohort showed STK3
   copy number gain among all five molecular subtypes. Several cases in
   three specific subtypes, namely microsatellite instability-high (MSI),
   genomically stable (GS), and chromosomal instability (CIN),
   demonstrated STK3 amplification. Besides, only 3 cases of GS subtype
   showed STK3 deep deletion (Fig. [154]3A). In our in-house GC cohort,
   10.1% cases demonstrated STK3 copy number gain, and 6.8% of the samples
   showed STK3 amplification (Fig. [155]3B). The expression level of
   multiple genes involved in cell cycle and DNA repair progress were
   closely correlated with the mRNA level of both STK3 and YAP1
   (Fig. [156]3C). Meanwhile, the same gene set showed opposite expression
   patterns between STK3^+ and STK3^− samples (Fig. [157]3D). Volcano plot
   demonstrated that multiple cell cycling- and DNA repair-related genes
   were significantly upregulated in the STK3^+ samples, while several
   notorious oncogenes were highlighted with significant and relatively
   higher upregulation in STK3^+ samples, such as ERBB2, GRB7, NOTUM,
   KRT23, LEFTY1, and DLX3 (Fig. [158]3E). Further enrichment analysis
   revealed several cell cycling and DNA repair processes, such as
   mismatch repair, base-excision repair, and cell cycle checkpoint
   signaling, were activated in the STK3^+ samples (Fig. [159]3F-G).
   Pseudotime analysis on the scRNA-seq data indicated that STK3
   expression was chronologically correlated with cell proliferation
   regulators, whereas the activation of apoptosis mediators was enriched
   in the other end of pseudotime schedule (Fig. [160]3H-I). GSVA
   highlighted two pathways, namely DNA replication and DNA repair, were
   both hyper-activated in STK3^+ cells (Fig. [161]3J). Besides, the
   average AUC score of these two pathways were relatively higher in the
   STK3^+ Seurat clusters (Fig. [162]3K). To validate the oncogenic
   function of STK3 inferred from public databases, RNA-seq was conducted,
   revealing downregulated expression of cell cycle regulators and the
   decrease of cell cycle- and DNA repair-related pathway activities
   (Fig. [163]3L-M). Western blot assay revealed a downregulated
   phosphorylation level of ATM, ATR, and CHK1 after STK3 depletion,
   indicating a reduced activity of DNA repair in GC cell lines
   (Fig. [164]3N). In xenograft models, mice with subcutaneous injection
   of shSTK3-treated NCI-N87 cells generated significantly smaller tumors
   (Fig. [165]3O).

Fig. 3.

   [166]Fig. 3
   [167]Open in a new tab

   STK3 gene is amplified in GC tissues, playing a crucial role in the
   regulation of cell cycle and DNA damage repair progression. (A) Copy
   number gain of STK3 gene was widely observed crossing all molecular
   subtypes of TCGA-STAD cohort (POLE, polymerase epsilon-mutated and
   ultramutated; MSI, microsatellite instability-high; GS, genomically
   stable; EBV Epstein-Barr virus-positive; CIN, chromosomal instability).
   (B) FISH assay demonstrated the frequent occurrence of STK3 gene copy
   number gain and amplification in primary GC samples. (C) The mRNA level
   of STK3 and YAP1 were remarkably correlated with the expression of cell
   cycle markers and DNA repair regulators in TCGA-STAD cohort. (D) The
   expression of cell cycling and DNA repair signatures are significantly
   upregulated in STK3^+ samples when comparing with STK3^− samples. (E) A
   volcano plot illustrating different expressed genes between STK3^+ and
   STK3^− samples. (F, G) Enrichment analysis and GSEA highlighted the
   hyperactivation of cell cycling- and DNA damage repair-related
   biological processes in STK3^+ samples. (H, I) Pseudotime analysis on
   the scRNA-seq data indicated that STK3 was upregulated in the early
   stage of cancer cell evaluation, along with cell proliferation gene
   markers. (J, K) GSVA indicated a higher activity of DNA replication and
   DNA repair pathways in STK3^+ cells. (L, M) RNA-seq analysis on
   STK3-deleted GC cell lines demonstrated declined expression of cell
   cycling/DNA repair-related gene expression and impaired pathway
   activities. (N) Phosphorylation of ATM, ATR, CHK1 and the expression of
   γ-H2AX were downregulated by STK3 depletion. (O) STK3 depletion
   impaired the xenograft formation capacity of NCI-N87 cells (n = 10
   mice/group; ***, P < 0.001)

   Besides the regulatory function of STK3 on cell cycling and DNA-repair,
   former analysis also indicated the potential of STK3 in maintaining the
   population of stem cell (Fig. [168]3G). CytoTRACE analysis revealed
   that STK3^+ cells exhibited a lower differentiation grade, suggesting
   higher stemness features (Fig. [169]4A), which is consistent with GSVA
   scoring on the stem cell proliferation pathway (Fig. [170]4B).
   Experimental results indicated that knocking down of STK3 significantly
   inhibited the spheroid formation capacity of MKN28 and NCI-N87
   (Fig. [171]4C). Multiple stemness signatures, such as CD44 and LETM1,
   were co-downregulated in STK3-deleted GC cell lines (Fig. [172]4D). At
   protein level, stemness markers including CD44, Nanog, KLF4, and SOX2
   were both downregulated by STK3 depletion (Fig. [173]4E). And the
   expression of the same stemness markers increased following the
   overexpression of STK3 in MKN1, a GC cell line which was proved to
   possess a low baseline level of STK3 (Fig. [174]4F). Furthermore,
   malignant features of MKN1, such as colony formation, invasion, and
   spheroid formation, were all over-activated by STK3 overexpression
   (Fig. [175]4G-I). The in vivo validation of the oncogenic function of
   STK3 was subsequently conducted using organoid and mouse models, which
   resulted in accelerated growth of organoids and subcutaneous xenografts
   in the STK3-overexpressing groups. (Fig. [176]4J-K).

Fig. 4.

   [177]Fig. 4
   [178]Open in a new tab

   STK3 facilitates the acquisition of stemness features in GC cells,
   consequently accelerating tumor formation. (A, B) STK3^+ cancer cells
   demonstrated less differentiation grade and higher activity in stem
   cell proliferation. (C) Knockdown of STK3 impaired the spheroid
   formation ability of MKN28 and NCI-N87. (D) STK3-depletion
   downregulated the mRNA levels of multiple stemness signatures. (E)
   Western blot assay indicated downregulated protein expressions of
   stemness markers in STK3-deleted cells, such as CD44, Nanog, KLF4, and
   SOX2. (F) Overexpression of STK3 increased the expression of CD44,
   Nanog, KLF4, and SOX2. (G-I) STK3 overexpression significantly enhanced
   the colony formation, invasion, and spheroid formation ability of MKN1
   cells. (J, K) STK3 overexpression accelerated the growth of
   patient-derived organoids and cell line-derived xenografts. (n = 5
   mice/group; *, P < 0.05; **, P < 0.01; ***, P < 0.001)

   The STK3-mediated biological processes, such as DNA repair and cancer
   cell stemness maintenance, are closely correlated with chemotherapy
   resistance [[179]29, [180]30]. In the GDSC drug resistance database,
   STK3 expression was highly correlated with the resistance against
   multiple chemotherapy drugs (Fig. [181]5A). Two organoids were adopted
   for investigating the role of STK3 in regulating drug sensitivity. As
   the results suggested, the organoid growth was inhibited by 5-FU
   treatment or STK3-depletion alone, and the co-administration group
   demonstrated significant decrease of organoid sizes when compared with
   5-FU treatment group (Fig. [182]5B). In the mice models, NSG mice were
   subcutaneously injected with NCI-N87 cells treated by shCtrl of shSTK3.
   After a 4-week treatment of 5-FU or PBS, cell-derived xenografts were
   harvested, weighted, and possessed for further IHC staining
   (Fig. [183]5C). Consistently with the organoid assay results, knocking
   down of STK3 significantly increased the anti-tumor efficiency of 5-FU,
   which was represented by the largely limited xenograft formation in
   mice of co-administration group (Fig. [184]5D-F). As indicated by the
   IHC staining results, the expression level of cell proliferation marker
   Ki-67 and stemness marker CD44 were downregulated in 5-FU treatment and
   shSTK3 groups. More importantly, the co-treatment of 5-FU and
   STK3-deletion showed superior effectiveness on inhibiting cancer cell
   proliferation and stemness (Fig. [185]5G).

Fig. 5.

   [186]Fig. 5
   [187]Open in a new tab

   Knockdown of STK3 remarkably improves the GC chemotherapy sensitivity.
   (A) High STK3 expression level is correlated to the resistance of
   various chemotherapy drugs in GDSC dataset. (B) STK3 knockdown enhanced
   the sensitivity of GC organoids to 5-FU treatment, as indicated by the
   smaller organoid sizes in shSTK3 + 5-FU group. (C) Schematics of the in
   vivo drug sensitivity assay. (D) Representative photos of subcutaneous
   xenografts. (E, F) Xenograft formation was significantly restrained in
   STK3-deletion group and demonstrated heightened sensitivity to 5-FU
   treatment (n = 6 mice/group). (G) Representative IHC images of
   xenograft sections stained by STK3, cancer cell proliferation marker
   Ki-67, and stemness marker CD44 (ns, not significant; *, P < 0.05; **,
   P < 0.01; ***, P < 0.001)

   The KEGG pathway enrichment analysis on the TCGA cohort revealed that
   Wnt signaling pathway was the most upregulated pathway in STK3^+
   samples (Fig. [188]6A). Besides, the expression of STK3 and multiple
   Wnt signaling regulators and effectors, such as GSK3B, MYC, CCND1, and
   SNAI1, were significantly correlated (Fig. [189]6B). To add up with,
   the STK3^+ cancer cells in scRNA-seq atlas also demonstrated relatively
   higher activity of Wnt signaling (Fig. [190]6C, Supplementary Fig.
   [191]S6). The activity of Wnt signaling and cell proliferation were
   investigated by IHC staining, which indicated that the inhibition of
   STK3 expression largely decreased the expression of active β-catenin
   and Ki-67 (Fig. [192]6D-E). Immunofluorescence was then conducted on
   two GC cell lines to demonstrate the impact of STK3 on Wnt signaling
   activity. As revealed by the results, the active β-catenin was largely
   accumulated in the nucleus of cancer cells in regular conditions,
   executing its oncogenic function. After STK3 deletion, the nuclear
   active β-catenin was dismissed and then distributed to the cytoplasm,
   indicating the dysfunction of Wnt signaling (Fig. [193]6F).
   Furthermore, overexpression of β-catenin partially rescued the
   inhibitory effect of STK3 depletion on the cancer cell stemness
   acquisition and GC organoid growth (Supplementary Fig. [194]S7).
   Western blot results revealed that the phosphorylated GSK-3β, an
   upstream regulator of the phosphorylation of β-catenin, was
   significantly downregulated after knocking down of STK3, along with the
   active β-catenin, c-Myc, and Cyclin D1. Meanwhile, the expression level
   of GSK-3β and β-catenin remain undisturbed (Fig. [195]6G). In MKN1
   cells, overexpressing STK3 would remarkably activate the
   phosphorylation of GSK-3β, and subsequently the expression of Wnt
   signaling markers c-Myc and Cyclin D1 (Fig. [196]6H). To demonstrate
   the direct interaction between STK3 and GSK-3β, AlphaFold 3 was adopted
   and the predicted binding pattern showed close affinity between STK3
   and the N-terminal of GSK-3β, which includes the phosphorylation site
   that inactivates GSK-3β (Ser 9) (Fig. [197]6I). Further
   co-immunoprecipitation (Co-IP) assay demonstrated that STK3 could
   directly bind to GSK-3β and induce the ubiquitination of it
   (Fig. [198]6J-K). The in vitro kinase assay results demonstrated that
   the proposed binding and phosphorylation-dependent regulation could be
   reconstituted in a purified system without cellular components. This
   was evidenced by the increased ATP consumption in the STK3-GSK-3β
   co-existence system compared with the STK3-only or GSK-3β-only systems
   (Supplementary Fig. [199]S8). Functional assays demonstrated that STK3
   overexpression exhibited no regulatory effect on the malignance of MKN1
   cells when co-transfected with S9A mutated GSK-3β. However,
   overexpression of STK3 partially rescued the proliferation and invasion
   ability in cells with overexpression of unmutated GSK-3β.
   (Supplementary Fig. [200]S9).

Fig. 6.

   [201]Fig. 6
   [202]Open in a new tab

   STK3 mediates Wnt signaling transduction by directly interacting with
   GSK-3β. (A) Wnt signaling pathway was highlighted by enrichment
   analysis. (B) STK3 mRNA expression was positively correlated with
   multiple Wnt signaling signatures. (C) STK3 positive cells presented
   higher Wnt signaling activity. (D, E) Representative IHC images of
   xenograft sections stained by STK3, active β-catenin, and Ki-67. (F)
   Knockdown of STK3 inhibited the nuclear accumulation of active
   β-catenin. (G) STK3 depletion inhibited the phosphorylation of GSK-3β,
   leading to the upregulation of β-catenin degradation and downregulation
   of Wnt signaling signatures. (H) Overexpression of STK3 upregulated the
   phosphorylation of GSK-3β, and the expression of Wnt signaling markers.
   (I) The predicted binding pattern of STK3 and GSK-3β. (J) Co-IP assay
   proving the direct interaction between STK3 and GSK-3β. (K)
   introduction of STK3 accelerated the ubiquitination process of GSK-3β
   (***, P < 0.001)

   High throughput virtual screening was performed on an anti-tumor
   chemical library to identify novel inhibitors of STK3. Ranked by
   binding affinity and referring to the root mean square deviation (RMSD)
   of binding conformation, the top 10 candidates were screened out for
   further evaluation (Fig. [203]7A-B). Finally, aminopterin was selected
   out for its advantages in molecular weight, internal stability, and the
   composition of intermolecular interactions (Fig. [204]7C-D). The
   cellular thermal shift assay (CETSA) revealed that STK3 protein would
   be largely degraded when the sample was heated to 59 ℃, while
   aminopterin treatment remarkably increased the stability of STK3
   protein, indicating the direct binding between STK3 and aminopterin
   (Fig. [205]7E). The in vitro kinase assay results demonstrated a
   significant inhibitory function of aminopterin on the STK3 kinase
   activity (Supplementary Fig. [206]S10) Functional assays proved that
   aminopterin treatment could inhibit the malignant features of GC cell
   lines, such as colony and spheroid formation, in a dose-dependent
   manner (Fig. [207]7F-G). As indicated by the dose-dependent expression
   variation of corresponding markers, aminopterin treatment succeeded in
   inhibiting STK3 expression and the oncogenic functions activated by it,
   including Wnt signaling transduction, cell cycling, and stemness
   maintenance (Fig. [208]7H). Furthermore, the co-administration of
   aminopterin and 5-FU regained a better anti-tumor effect than 5-FU
   monotherapy in our organoid models (Fig. [209]7I) and mice models
   (Supplementary Fig. [210]S11). The IHC staining results demonstrated
   significate downregulation of active-β-catenin and Ki-67 while the
   expression level of cleaved-PARP was upregulated following both
   aminopterin treatment alone and in combination with 5-FU (Supplementary
   Fig. [211]S12). As for the safety profile, administration of
   aminopterin demonstrated limited impact on mice body weight
   (Supplementary Fig. [212]S13). Hematoxylin and eosin (H&E) staining
   results also demonstrated no severe histopathological changes in the
   major organs of control and treatment groups (Supplementary Fig.
   [213]S14).

Fig. 7.

   [214]Fig. 7
   [215]Open in a new tab

   Aminopterin is screened out by in silico modeling as a highly promising
   inhibitor for STK3. (A-B) Amin was identified as the most effective
   STK3 inhibitor among 4511 candidates with anti-tumor properties. (C-D)
   Predicted binding pattern and intermolecular interactions of Amin-STK3
   complex (E) CETSA was conducted to further prove the binding between
   Amin and STK3. (F, G) Colony and spheroid formation ability of GC cell
   lines were restrained by Amin treatment in a dose-dependent manner. (H)
   Amin treatment downregulated the expression level of Wnt signaling
   signatures, cell cycling regulators, and stemness markers. (I) Amin
   administration impaired the growth of GC organoids and enhanced their
   sensitivity to 5-FU treatment (*, P < 0.05)

Discussion

   Until now, GC remains a major threat to human health. Especially in
   Asia, due to the consistent concerns in the high salt/oil diet and
   high-pressure living patterns, the occurrence of GC is shifting to
   younger generation [[216]31]. Among all kinds of biological pathways
   involved in tumorigenesis, the Hippo pathway is notorious for its
   predominantly regulatory function in cancer cell malignancy and
   cross-talks with other oncogenic pathways. Studies have demonstrated
   that the high expression level and nuclear transition of YAP1 would
   induce the enhancement of multiple aggressiveness features of cancer
   cells, such as proliferation, migration and cancer stem cell population
   maintenance [[217]32, [218]33]. Some studies also highlighted that
   Hippo pathway activity can directly mediate the ferroptosis status
   within cancer cells [[219]34]. Mechanistically, the regulation of Hippo
   pathway on carcinogenesis is mainly facilitated by the regulation on
   YAP1. In the Hippo-off status, the phosphorylation cascade
   Mst1/2-Lats1/2 would be deactivated, and unphosphorylated YAP1 was
   transported into the nucleus, binding to TEADs directly to facilitate
   its oncogenic roles. Subsequently, the co-transcription factor would
   activate the expression of multiple oncogenic genes, such as CTGF and
   CYR61, to encourage tumorigenesis [[220]35, [221]36]. The YAP1-5SA
   mutation, in which five serine residues (S61, S109, S127, S164, and
   S397) are substituted with alanine, represents a constitutively active
   form of YAP1 that functionally mimics YAP1 overexpression [[222]37,
   [223]38]. These five serine residues serve as the primary
   phosphorylation sites for Lats1/2 kinases within the Hippo pathway. By
   abolishing these critical phosphorylation sites, the 5SA mutant is
   capable of escaping Hippo cascade-mediated inhibitory phosphorylation.
   Functionally, this mutant recapitulates the phenotypic effects of YAP1
   overexpression, including promotion of cell proliferation,
   epithelial-mesenchymal transition, and other oncogenic processes, while
   remaining unresponsive to normal Hippo pathway regulation.

   In this work, we identified STK3 as a novel and indispensable YAP1
   downstream target which could largely utilize the oncogenic role of
   YAP1. Starting from a coincident observation, we found that
   YAP1-deleted GC cells and CAFs demonstrated concordant and significant
   decrease in STK3 expression. We then further demonstrated the
   co-downregulation in protein level by siRNA-mediated YAP1 knockdown,
   YAP1 inhibitor treatment, and Yap1^−/−Taz^−/− transgenic mice models,
   and all the results proved that the downregulation of YAP1 can inhibit
   STK3 expression. Until now, the hypothesis generated from the data is a
   totally reverse against former knowledge, as STK3, also named as Mst2,
   is long-believed and well-proven to be an upstream inhibitor of YAP1.
   To further validate the realness of the regulation between YAP1 and
   STK3, we conducted correlation analysis on multiple databases. The
   results strongly supported that the expression of YAP1 and STK3 are
   significantly and positively correlated, no matter in primary samples
   or cancer cell lines, in mRNA or protein levels, in bulk scale or
   single cell scale, and in in-house or public databases. Besides, a
   TEAD4 binding motif was identified in the promoter sequence of STK3,
   and this binding site was further validated by both ChIP-seq and
   ChIP-qPCR. Taken together, we conclude that STK3 is a direct downstream
   target gene of YAP1/TEAD4.

   Given its promoting function in the YAP1 phosphorylation and
   degradation process, STK3 is canonically considered as a tumor
   suppressor gene. However, recent studies have unmasked the oncogenic
   potential of STK3 in breast and prostate cancer progression [[224]39,
   [225]40]. In our previous study, STK3 was also demonstrated to be a
   promoting factor for the proliferation and migration activity of GC
   cell lines [[226]41]. Again, we challenged the classic thinking by
   checking the genome stability of STK3, which revealed a notable
   percentage of STK3 amplification/copy number gain cases while less than
   1% GC patients showed STK3 deletion. Further analysis of the TCGA-STAD
   cohort demonstrated a similar pattern of STK3, which indicates a strong
   oncogenic potential of this traditionally recognized tumor suppressor
   gene. As we dig deeper into the expression pattern of STK3, it was
   observed that while STK3 amplification was noted in a limited group of
   patients, the overexpression of STK3 is more commonly observed in GC
   patients: In the TCGA cohort, upregulated STK3 levels were identified
   in 76 GC patients (19.8%), whereas only 24 patients (6.3%) exhibited
   STK3 gene amplification. Similarly, in our in-house cohort, 19 patients
   (6.8%) demonstrated STK3 amplification, while 63 cases (22.7%) showed
   elevated STK3 levels. These findings suggest that the oncogenic
   function of STK3 is not solely a result of gene amplification, but is
   also mediated by upstream mechanisms. This underscores the involvement
   of the previously established YAP1/TEAD4-STK3 regulatory cascade in the
   activation of STK3.

   The potential negative feedback of STK3 upregulation on YAP1
   phosphorylation was also investigated. The canonical tumor suppressor
   function of STK3 was considered to be largely dependent on regulating
   the phosphorylation and then degradation of active YAP1. However, we
   observed a slight downregulation of YAP1 and upregulation of
   phosphorylated YAP1 in STK3-deleted GC cell lines. Meanwhile, the
   expression level of CTGF (a classic downstream transcriptional target
   of YAP1) was also downregulated after STK3-deleption, which reflects a
   decreased transcription activity of the co-transcription factor
   YAP1/TEADs. More precisely, the expression level of YAP1 in nuclear was
   decreased in STK3-deleted cells, while phosphorylated YAP1, which is
   primarily present in the cytoplasm, was upregulated. All these results
   suggested that the knockdown of STK3 would not protect active YAP1 from
   being phosphorylated by Lats1/2 in GC cell lines as canonically
   hypothesized. In contrast, STK3 deletion led to the decreased activity
   of the oncogenic YAP1 signaling, which might result from its
   significant inhibition on the tumor cell malignancy.

   Subsequent biological analysis based on the TCGA cohort and scRNA-seq
   atlas also emphasized the oncogenic potential of STK3 in regulating
   cell cycling and DNA repair progress. Notably, some stemness-related
   processes were also included in the enrichment analysis. The
   maintenance of cancer cell stemness is largely dependent on the
   activation of a series of stemness regulators. Within this criterion,
   several stemness markers, namely CD44, Nanog, KLF4, and SOX2, were
   selected as the predominant ones in GC progress [[227]42, [228]43]. In
   the following in vitro and in vivo validations, STK3 knock-down and
   overexpression assays harvested significant variations in marker gene
   expression levels and malignant phenotypes, which coherently meet the
   former expectations of the oncogenic roles of STK3.

   The drug resistance to chemotherapy remains a major challenge in GC
   treatment, and the detailed molecular mechanisms within this biological
   mutation are still not thoroughly elucidated due to the complexity and
   variability of the cancer cell genome [[229]44, [230]45]. The
   Hippo-YAP1 signaling has been long identified as an indispensable
   participator in the occurrence of chemotherapy resistance [[231]46]. As
   acknowledged, elevated transcriptional activity of YAP1 would
   significantly enhance the response to DNA repair of cancer cells, which
   would largely abolish the lethality of chemo drugs. Besides,
   YAP1-mediated increase of cancer stem cell population also contributed
   to the recurrence of GC after first-line treatment. Now as STK3 has
   demonstrated regulatory functions in both DNA repair and cancer stem
   cell population maintenance, we wondered if the expression level of
   STK3 can affect the drug sensitivity to chemotherapy. By both
   patient-derived organoid and cell-derived xenograft models, we observed
   a better inhibitory effect of 5-FU when co-administrated with shSTK3.
   As indicated by the IHC staining, the proliferation and cancer stem
   cell population were impaired by 5-FU treatment, while more severe
   inhibition was observed in the shSTK3 group. Collectively, we conclude
   that STK3 overexpression in GC patients could lead to the
   hyperactivation of malignant phenotypes and resistance to chemotherapy.
   Mechanistically, the activation of STK3 accelerated the cell cycling
   and DNA repair response of GC cells while contributing to the
   maintenance of cancer stem cell population as well.

   Although the oncogenic potential of STK3 has been observed in several
   reports, the deep mechanism of this non-canonical function was hardly
   explored. By enrichment analysis on the genomic profile of STK3-deleted
   samples, we found the Wnt signaling pathway ranks as one of the most
   distributed pathways. As a well-recognized oncogenic factor, this
   notorious pathway is vital for the genomic activation of multiple
   oncogenes, especially cancer cell migration and stemness markers. More
   interestingly, the crosstalk between Hippo and Wnt signaling has been
   identified in multiple patterns, such as Wnt5A-ROR2-YAP1 [[232]47] and
   FZD10-Lats1/2-YAP1 axis [[233]48]. Some studies also reported the
   mutual-indispensable and direct binding of YAP1 and β-catenin in
   intrahepatic cholangiocarcinoma [[234]49]. Recent works further
   demonstrated the crosstalk could be established by dual directions, as
   YAP1 was identified as a crucial promotor for the adrenocortical cancer
   progression by mediating the transcription activity of β-catenin
   [[235]50]. Another study focused on the GC hepatic premetastatic niche
   revealed that the ephrin A1-initialized YAP1-CCL2-Wnt/β-catenin axis
   bridged the malignant communications between cancer cells and hepatic
   stellate cells during GC hepatic metastasis [[236]51]. Inspired by the
   bioinformatic findings, we firstly validated the regulation of STK3 on
   Wnt signaling activation, and the data proved that the expression of
   active (non-phosphorylated) β-catenin and nuclear accumulation of
   β-catenin were consistently decreased after knockdown of STK3. However,
   the biological function of STK3 is to mediate the phosphorylation of
   substrate proteins on the serine or threonine residues [[237]52,
   [238]53], which indicates the existence of other intermediators within
   the STK3-β-catenin signaling axis. Among all direct regulators of
   β-catenin, GSK-3β is particularly manipulated by its phosphorylation
   status: the phosphorylation at Ser9 of GSK-3β is the most common
   mechanism of GSK-3β inactivation, which would lead to the deactivation
   of β-catenin phosphorylation and ubiquitination [[239]54]. To further
   elucidate the detailed mechanisms within the STK3-Wnt signaling
   transduction, we conducted both STK3 knockout and overexpression assays
   in GC cell lines, which demonstrated a hypothesis-endorsing expression
   variations of Ser9-phosphorylated GSK-3β and non-phosphorylated
   β-catenin. Furthermore, the Co-IP and ubiquitination assay confirmed
   the direct binding of STK3 and GSK-3β, which subsequently led to the
   degradation of GSK-3β. Taken together, these results revealed a novel
   STK3/GSK-3β/β-catenin signaling axis which was responsible for the
   oncogenic functions of YAP1-derived STK3 upregulation. Meanwhile, this
   work also proposed a novel crosstalk between Hippo and Wnt signaling,
   enriching the understanding of the signaling transduction during GC
   progression.

   Given the high incidence of STK3 gene amplification in GC patients and
   its significant effects in promoting oncogenesis and chemo-therapy
   resistance, there is considerable potential in developing STK3-targeted
   treatments. Classic Mst1/2 inhibitors like XMU-MP-1 were well developed
   with good efficiency, appreciable pharmacokinetics, and tolerable side
   effects, aiming to fulfill the requirements of tissue repair and
   regenerative medicine [[240]55]. Given the promising records in
   efficiency and safety profiles of STK3-targeting strategy, coupled with
   the absence of STK3 inhibitors specially designed for cancer treatment,
   we employed high throughput virtual screening to identify candidate
   STK3 inhibitors based on a previously constructed library comprising
   4511 chemicals with anti-tumor potentials. After a comprehensive
   consideration combining the druggability evaluation of small molecules
   and the stability of binding conformation, aminopterin emerged as the
   prior candidate and demonstrated significant STK3-inhibiting and
   tumor-suppressing effects in vitro.

   Aminopterin, an analogue of folic acid, has long been recognized for
   its significant anti-tumor effects at relatively low doses. Its
   application has been widely investigated in clinical trials focusing on
   patients with endometrial carcinoma, leukemia, and psoriasis
   [[241]56–[242]58]. Notably, aminopterin exhibits excellent solubility,
   allowing for substantial dissolution in water. Furthermore, its
   relatively low molecular weight and acidic structure facilitate both
   absorption and degradation. However, the toxicity of aminopterin raises
   major concerns and limitations for clinical use, leading to its
   discontinuation in routine applications. In mouse models, doses above
   0.37 mg/kg resulted in 80% mortality, significantly lower than that
   observed with another analogue, methotrexate [[243]59]. Given these
   concerns, we examined the safety profile of aminopterin administration
   in mice. The results demonstrated that administration of aminopterin
   significantly inhibited subcutaneous xenograft growth while
   demonstrating limited impact on mice body weight. Additionally, H&E
   staining results indicated that both aminopterin alone and in
   combination with 5-FU exhibited tolerable organ toxicity in
   tumor-bearing NSG mice, as no severe histopathological changes were
   observed in the major organs of control and treatment groups, except
   for slight congestion in kidneys and lungs in the treatment groups.
   Collectively, these characteristics of aminopterin highlight
   limitations in its therapeutic window and underscore the need for
   further investigation into structure-based optimization in the drug
   discovery process to enhance its therapeutic potential.

Conclusion

   In summary, we demonstrated that STK3, against the common knowledge, is
   transcriptionally regulated by YAP1, and the high expression of STK3
   can promote GC progression by manipulating cancer cell malignance.
   Meanwhile, STK3 upregulation is vital for cancer stem cell population
   maintenance and DNA repair progress, and thus contributes to the
   acquisition of chemo-therapy resistance. Mechanistically, STK3 can
   directly bind to GSK-3β and facilitate its ubiquitination, which leads
   to the nuclear accumulation of unphosphorylated β-catenin and then the
   activation of the oncogenic Wnt signaling signatures. Given the
   structural feasibility and good safety profiles of targeting STK3, we
   also delivered a promising candidate for STK3-targeted GC therapy. In
   conclusion, our findings elucidated the unexplored oncogenic function
   of STK3 and the underlying mechanism, providing a brand-new insight
   into the crosstalk between Hippo and Wnt signaling.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [244]Supplementary Material 1^ (1.5MB, pdf)
   [245]Supplementary Material 2^ (1.1MB, xlsx)
   [246]Supplementary Material 3^ (2MB, pdf)

Acknowledgements