Introduction

Notwithstanding the gradual decline in the incidence of gastric cancer (GC), it remains the fifth most common cancer and the third leading cause of cancer deaths worldwide [1]. Despite the advances in chemotherapy and the anti-vascular endothelial growth factor receptor 2 antibody ramucirumab, the majority of patients with advanced GC continue to have disease progression following treatments [2]. Recently, immune checkpoint inhibitors have demonstrated clinical efficacy with an objective response rate of 11–12% in patients with advanced GC [3, 4]. However, the survival benefit is small albeit statistically significant and the treatment efficacy is limited to a small subset of patients [3, 5]. Establishing clinically relevant and biology-informed classification systems in GC and the associated biomarkers to facilitate the development of novel molecularly targeted and rational therapies remains an urgent and highly unmet clinical need.

The Wnt-regulated self-renewal processes in epithelial tissues are frequently hijacked by cancer cells to facilitate malignant progression [6,7,8,9]. Dysregulated Wnt pathway activity plays a crucial role in sustaining the phenotypes in stem-like cancer cells or “cancer stem cells (CSC)” [6, 8, 9]. In GC, approximately 10–30% of the tumors display deregulated Wnt signaling by developing genetic lesions affecting multiple levels of the Wnt signaling pathway [10, 11]. ASPM (abnormal spindle-like microcephaly associated) is a centrosomal protein that regulates neurogenesis and brain size and is also widely expressed in a variety of tissues [12, 13]. ASPM expression is up-regulated in various malignant tumors [13,14,15,16], wherein it serves as an oncoprotein and a positive regulator of Wnt-β-catenin signaling and cancer stemness [17,18,19,20]. Interestingly, ASPM has been shown to express in the scattered stem/progenitor cells existing in the isthmus zone of gastric oxyntic mucosa, and its expression is up-regulated in GC cells [21]. The precise pathogenetic role of ASPM in GC and the underpinning mechanistic basis have yet been rigorously investigated.

FOXM1 (Forkhead box protein M1) is a member of the Forkhead transcription factors that regulate the expression of genes involved in cell cycle regulation, development, organ regeneration, and tumorigenesis [22,23,24,25,26]. Aside from serving as a transcription factor, recent studies unraveled the pleiotropic functions of FOXM1 such as its effect to facilitate the nuclear translocation of β-catenin in glioma and leukemia cells [27, 28]. Furthermore, FOXM1 has also been found to regulates CSCs in pancreatic cancer, and its expression is upregulated in GC and poor prognostic [29, 30]. Given that both ASPM and FOXM1 have been shown to regulate the Wnt/β-catenin signaling pathway and cancer stemness, we set out to determine their potential molecular and epistatic relationships in GC cells with the goal of identifying novel pathomechanisms underlying the aggressiveness of human GC, which may unlock strategies to improve the prognostic classification of patients as well as to develop novel rational therapies for advanced and aggressive GC.

Materials and methods

Bioinformatics analysis

The RNA sequencing transcript level of ASPM and FOXM1 in GC and normal gastric tissues in The Cancer Genome Atlas (TCGA) was interrogated from the GEPIA database (http://gepia.cancer-pku.cn). The survival data of 593 GC patients stratified based on the ASPM (Affymetrix ID: 219918_s_at) and the FOXM1 (Affymetrix ID: 202580_x_at) transcript level were downloaded from KM Plotter (http://kmplot.com/analysis/index.php?p=service&cancer=gastric). The survival data of 352 GC patients in TCGA were downloaded from SurvExpress (http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp).

Immunohistochemistry (IHC) and immunofluorescence (IF) analysis

The tissue microarrays (TMA) of 59 GC and the matched normal tissues were obtained from SuperBioChips Laboratories (CQ2 and CQN2; Seoul, South Korea). Formalin-fixed, paraffin-embedded GC tissues from 63 patients who received tumor resection at National Cheng Kung University Hospital (NCKUH) were acquired and used in conformity with Institutional Review Board-approved protocols (Table S1). Informed consent was obtained from all the patients. Tissue sections were deparaffinized, hydrated, immersed in citrate buffer for epitope retrieval in a microwave. The antibodies used are listed in Table S2 [20]. The staining was detected using the DAKO EnVision kit (Agilent Technologies, Santa Clara, CA, USA). All the IHC staining was evaluated by an expert pathologist (C.T. Chiang) and the staining intensities were quantified at the single-cell level with at least 300 tumor cells counted per tumor (3 tissue sections per tumor; at least 100 tumor cells counted per sections). IF staining of cells grown on culture plastics was performed using standard protocols. Confocal imaging was performed using a Nikon Digital Eclipse C1 confocal microscope system.

Cell culture and gene expression manipulations

AGS and NCI-N87 cells (American Type Culture Collection) were maintained as suggested by the provider. The sustained knockdown (KD) of ASPM expression including all its putative transcript variants was achieved by lentivirus-mediated RNA interference (RNAi) using validated small hairpin RNA (shRNA) oligonucleotides in the lentivector pLKO.1-puro (TRCN0000118905 and SHC002V [non-target control]; Sigma-Aldrich, St. Louis, MO, USA). The specific KD of ASPM variant 1 (ASPMv1) expression, which encodes ASPMiI (NCBI RefSeq: NP_060606.3) was carried out by lentivirus-mediated transduction of small hairpin RNA (shRNA) using 2 independent RNAi target sequences (shRNA #4 and shRNA #3) specific for the exon 18 of the ASPM gene (unique to ASPMv1) and have been described previously [20]. The sustained KD of FOXM1 expression was achieved by lentivirus-mediated RNA interference (TRCN0000015545 and TRCN0000015544, MISSION shRNA lentiviruses; Sigma-Aldrich). The lentivirus-mediated overexpression (OE) vectors of dishevelled 3 (DVL3) and FOXM1 were purchased from Origene Technologies Inc. (Rockville, MD, USA). Lentivirus was produced as previously described [20].

IB and co-immunoprecipitation (IP)

IB and co-IP analyses were performed according to standard protocols as described previously [20]. Proteins were revealed after SDS/PAGE and immunoblotting with the indicated antibodies (Table S2).

qRT-PCR

qRT-PCR analysis was performed on the amplified RNA using the LightCycler FastStart DNA MASTERPLUS SYBR Green I Kit and the LightCycler System (Roche Diagnostics GmbH, Mannheim, Germany) according to the manufacturer’s instructions. Oligonucleotide primers were designed using Primer Bank (http://pga.mgh.harvard.edu/primerbank/index.html).

Chromatin IP (ChIP)

ChIP experiments were carried out using primers specific for the putative binding sites (PBS) of FOXM1 on the promoter regions of ASPM as described previously [31]. Approximately 2–5 μL of ChIP-enriched chromatins was subjected to a standard qRT-PCR and the enrichment of specific genomic regions was assessed relative to either control IgG or the control cells.

Luciferase reporter assay

For luciferase reporter assay, cells were transduced with Cignal Lenti TCF/LEF Reporter (Qiagen, Venlo, The Netherlands) according to the manufacturer’s protocol. Following stimulation of the cells with recombinant human WNT3A (250 ng/mL for 16 h; R&D Systems) or vehicle, the reporter activity was measured using the ONE-GloTM Luciferase Assay System (Promega, Madison, WI, USA).

Flow cytometry and tumorsphere assays

To analyze CSC populations by flow cytometry, cells were dissociated, antibody-labeled (1–2 μg per 106 cells × 1 h) and resuspended in HBSS/2% FBS as previously described [32, 33]. The antibodies used are listed in Table S2. The ALDEFLUOR assay (StemCell Technologies, Vancouver, BC, Canada) was performed according to the manufacturer’s recommendation. Flow cytometry was done used a FACSCanto™ II flow cytometer (BD Biosciences, Franklin Lakes, NJ, US) with the electronic gating set according to cells stained with the corresponding isotype-matched control IgG. The tumorsphere assay and the limiting dilution assay (LDA) were performed as previously described [19, 34].

Animal models

GAC cells were lentivirally transduced a green fluorescence protein (GFP) and firefly luciferase (FF-Luc) fusion vector (UBC-EGFP-T2A-Luc; System Biosciences, CA, USA), and the GFP-positive cells were sorted. The cells were then infected with the lentivirus expressing the ASPM-specific shRNA or a non-target control as described above, after which the cells (1 × 106 cells) in 100 μl (1:1 Matrigel:cells) were inoculated into the flank of 8-week-old male NOD.CB17-Prkdcscid/NcrCrlBltw (NOD/SCID) mice (BioLASCO Taiwan, Taipei City, Taiwan) and tumor mass and distribution were assessed by bioluminescence imaging (BLI; the IVIS Imaging System, Caliper Life Sciences, Waltham, MA). For the peritoneal carcinomatosis model, FF-Luc-transduced GAC cells (1 × 106 cells) were injected directly into the peritoneal cavity of NOD/SCID mice. Protocols for animal care and experimentation were approved by the Institutional Animal Care and Use Committee of National Health Research Institutes (Zhunan, Miaoli County, Taiwan), and were adhered to the NIH Guide for the Care and Use of Laboratory Animals.

Statistical analysis

The statistical programming language R (cran.r-project.org) and SPSS 10.0 software (SPSS, Chicago, IL, USA) were used to conduct the statistical analysis of our data. Survival curves were generated using the Kaplan–Meier method. The curves were plotted and compared using the log-rank test using the Prism 6.01 software (GraphPad Software). A cut-off value that best discriminates between groups with respect to outcome was determined using the Concordance Index, where an index of 1.0 is perfect discrimination [35]. The data from the LDA were analyzed and plotted using the ELDA software (http://bioinf.wehi.edu.au/software/elda/index.html).

Results

The transcript level of ASPM is upregulated in GC and correlates with poor patient prognosis

Previously, Oncomine analysis based on published GC transcriptome data sets has revealed a significant increase in the transcript level of ASPM in GC tissues compared with non-tumor tissues [21]. We confirmed in RNA sequencing data that the transcript level of ASPM significantly increased in GC tissues compared with that of normal tissues (Fig. 1a). Clinical correlative analyses conducted in several large patient cohort (N = 945 in total) revealed that the tumors with a high transcript level of ASPM are associated with a significantly worse overall survival than those with a low level of ASPM (Fig. 1b). Moreover, the ASPM expression strongly correlates with the relapse-free survival of GC patients (Fig. 1c), implicating that ASPM is associated with the aggressive behavior of GC.

Fig. 1
figure 1

ASPM expression relates to poor prognosis in GC and contributes to tumor progression. a Box plots of the expression level (transcripts per million, TPM) of ASPM in normal gastric tissue (N = 211) and GC tissues (N = 408) in the TCGA RNA sequencing data set. ***P < 0.001, t test. Kaplan–Meier survival curve comparing overall survival (b) or relapse-free survival (c) of patients with GA in the KM Plotter data sets (N = 593) or the TCGA data set (N = 352) with their tumors having a high or low transcript level of ASPM. d Representative bioluminescence images (BLI) of the tumors formed by control-KD or ASPM-KD AGS cells in NOD/SCID mice at the indicated time after cell inoculation. Right, tumor bulk quantified as BLI normalized photon counts as a function of time (mean ± SEM, N = 6 mice in each group). *P < 0.01 compared with control KD, t test. e Representative BLI of the tumors formed by control-KD or ASPM-KD NCI-N87 cells in NOD/SCID mice. Bottom, tumor bulk quantified as BLI normalized photon counts (mean ± SEM, N = 6 mice in each group). **P < 0.01; ***P < 0.001 compared with control KD, t test. f Representative BLI of the peritoneally injected NCI-N87 cells with control KD or ASPM KD. Right, the magnitude of tumor cell dissemination in the peritoneal cavity quantified as BLI normalized photon counts as a function of time (mean ± SEM, N = 6 mice in each group). *P < 0.05; **P < 0.01 compared with control KD, t test

Full size image

ASPM contributes to GC progression and aggressiveness

The upregulated expression of ASPM in GC tissues and its prognostic significance in GC patients raised the possibility of its potential oncogenic role in GC. We thus stably expressed an FF-Luc reporter in primary GC-derived AGS cells with ASPM KD or control KD and injected them subcutaneously into the flank of immunodeficient NOD/SCID mice. Remarkably, KD of ASPM expression almost abolished the tumorigenic potential of AGS cells (Fig. 1d). Likewise, advanced GC NCI-N87 cells deficient in ASPM expression failed to produce xenografted tumors in NOD/SCID mice (Fig. 1e).

Since advanced GC is frequently associated with peritoneal carcinomatosis, we injected NCI-N87-FF-Luc cells with ASPM KD or control KD into the peritoneal cavity. KD of ASPM expression abrogated more than 95% of the BLI signals emitted from the abdominal cavity compared with the control KD group (Fig. 1f). Thus, the mice bearing ASPM-KD GC cells developed a significantly less amount of peritoneal fluid than those mice with control-KD (data not shown). These data lend support to the important role of ASPM in GC progression and aggressiveness.

The expression of ASPMiI is prognostically important in GC

Whilst the preceding data support the importance of ASPM in GC, they did not distinguish among different protein isoforms of ASPM, which may play different roles in oncogenesis [20, 36]. To this end, we carried out TMA analysis of 59 GC tissues and the paired normal gastric epithelium for the staining of ASPMiI and ASPMiII using isoform-specific antibodies recently developed by our group. We found that ASPMiI is expressed mainly in the cytoplasm of a small subset (≥ 2 + ; ~ 10.1%) of normal gastric epithelial cells, especially those localized within the isthmus zone of the gastric oxyntic glands as previously described [21] (Fig. 2a, inset). By contrast, ASPMiII is expressed in both the cytoplasm and the nuclei of normal epithelial cells throughout the glands (Fig. 2a). Importantly, GC cells significantly upregulate the expression of ASPMiI with approximately one-third (~ 35.5%) of the tumor cells exhibiting ≥ 2 + staining (Fig. 2b, left panels), whereas the expression of ASPMiII only marginally increased in GC cells (Fig. 2b, right panels). Notably, the expression of both ASPM isoforms exhibit a considerable cell-to-cell heterogeneity within the same tumors (P < 0.001 by ANOVA). Of note, the above-mentioned expression pattern of ASPM isoforms is not restricted to the intestinal-type of GC and also applies to the diffuse-type of GC (N = 8; Fig. S1). Using IF analysis, we confirmed the differential subcellular distributions of ASPMiI and ASPMiII in established GC cell lines AGS and NCI-N87 cells. Specifically, ASPMiI is predominantly expressed in the cytoplasm of GC cells, whereas ASPMiII is mainly nuclearly localized (Fig. 2c and data not shown).

Fig. 2
figure 2

The expression patterns of ASPM isoforms in GC tissues and the prognostic role of ASPM isoform I (ASPMiI). a Representative IHC staining of ASPMiI and ASPMiII in human GC tissues and the adjacent normal gastric epithelium (× 20 magnification). Inset, ASPMiI-positive-staining cells in the isthmus zone of normal gastric oxyntic glands. b Bar chart showing the staining intensities (1 + to 3 +) of ASPMiI and ASPMiII in the GC and normal epithelial tissues in a (mean ± SEM, N = 59). ***P < 0.001; NS, not significant, one-way ANOVA. c Representative immunofluorescence (IF) images showing the subcellular localization of ASPMiI or ASPMiII (green) in AGS cells. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; blue). Bottom, quantification of the subcellular localization of ASPMiI or ASPMiII (mean ± SEM, N = 4). d Kaplan–Meier survival curve comparing overall survival of the patients in the NCKUH cohort stratified according to the ASPMiI Positive Score (PS). The cut-off value of the ASPMiI PS (10%) was determined using the Concordance Index

Full size image

Given that ASPMiI expression is specifically upregulated in GC tissues, we investigated its prognostic role by performing its IHC staining in the whole-tumor tissue sections from 63 patients with resected GC (Table S1). In light of the considerable cell-to-cell expressional heterogeneity of ASPMiI, we scored its staining intensity at the single-cell resolution and calculated the percentage of the tumor cells exhibiting a ≥ 2 + staining intensity of ASPMiI for each tumor, which we denoted the “ASPMiI Positive Score (PS)”. We found that the tumors exhibiting an ASPMiI PS of 10% or higher are associated with a significantly worse survival than those with a low ASPMiI PS (Fig. 2d). Furthermore, multivariable analysis confirmed that ASPMiI PS is a significant predictor of survival (Cox regression P = 0.043) independent of standard clinico-pathologic variables (Table S3), underscoring the important role of ASPMiI in GC progression.

ASPMiI regulates the protein stabilities of DVL3 and β-catenin in GC cells

Recent studies have shown that ASPM, especially ASPMiI, augments canonical Wnt signaling in malignant cells [17, 18, 37]. We thus posited that ASPMiI also regulates Wnt activity in GC cells. Indeed, co-IF analysis on human GC tissues revealed that ASPMiI significantly colocalizes with active β-catenin in the cytoplasm of GC cells (Fig. 3a, b). To study the functional role of ASPMiI in Wnt signaling, we specifically down-regulated the expression of ASPMiI, but not ASPMiII, using lentivirus-mediated RNAi by transducing GC cells with two previously validated shRNAs (shRNA#3 and #4) targeting ASPMv1 (encoding ASPMiI) (Fig. 3c; Fig. S2a, b). Remarkably, the specific KD of ASPMv1 expression using ASPMv1-specific shRNA#4 and #3 both abolished the WNT3A-stimulated Wnt pathway activation in GC cells (Fig. 3d; Fig. S2c). Given that the ability of ASPM to augment Wnt activity has been attributed to its ability to stabilize the upstream Wnt regulator DVL proteins [17, 18], we examined their protein abundance levels in GC cells. We uncovered that down-regulating ASPMiI expression significantly reduced the level of DVL3 but not DVL1 or DVL2 (Fig. 3e; Fig. S2d). Indeed, a co-IP assay confirmed the strong interaction of ASPMiI with DVL3 (Fig. 3f). Consistently, KD of ASPMv1 expression enhanced the poly-ubiquitination of DVL3 and enhanced the recruitment of its specific E3 ligase Cullin 3 (Fig. 3g) [38]. Echoing these findings, the concurrent OE of DVL3 could restore the Wnt pathway activity that had been reduced by KD of ASPMv1 expression (Fig. 3h), suggesting that ASPMiI regulates Wnt–β-catenin signaling by stabilizing the DVL3 protein in GC cells.

Fig. 3
figure 3

ASPMiI regulates Wnt pathway activity and DVL3 in GC cells. a Representative IF images showing the colocalization (yellow) of ASPMiI (green) with active β-catenin (red) in human GC tissues. Nuclei were counterstained with DAPI (blue). b Heatmaps illustrating the correlation of the staining intensity of ASPMiI with that of active β-catenin in GC tissues (N = 8 tumors). c Immunoblotting (IB) analysis showing the effect of the specific KD of ASPMv1 expression using lentivirus-mediated transduction of ASPMv1 shRNA #4 in NCI-N87 cells, which specifically reduced the protein abundance level of ASPMiI but not ASPMiII. β-tubulin was included as a loading control. d Relative Wnt-specific luciferase activity in control- or ASPMv1-KD and WNT3A-treated NCI-N87 cells (mean ± SEM, N = 3 biological replicates with two technical replicates). ***P < 0.001, t test. e IB analysis of DVL1, DVL2 and DVL3 in control-KD or ASPMv1-KD NCI-N87 cells. β-tubulin was included as a loading control. Bottom, the fold of reduction in the protein abundance levels of DVL proteins (mean ± SEM, N = 3). *P < 0.05, t test. f Co-immunoprecipitation (IP) analysis showing that ASPMiI strongly associates with DVL3 in NCI-N87 cells. g IB analysis showing that KD of ASPMv1 expression increases the poly-ubiquitination of the immunoprecipitated DVL3 protein and enhanced the recruitment of the E3 ligase Cullin 3 to DVL3 in NCI-N87 cells. The cells were treated with the proteasome inhibitor MG-132 (10 μM for 12 h) before the collection of the protein lysate. h Lentivirus-mediated overexpression (OE) of DVL3 restored the Wnt-specific luciferase activity in NCI-N87 cells with KD of ASPMv1 expression (mean ± SEM, N = 3 biological replicates with two technical replicates). ***P < 0.001, t test

Full size image

ASPMiI critically contributes to gastric cancer stemness

Malignant tumor cells frequently co-opt the Wnt-regulated self-renewal processes to acquire a stem-like phenotype to facilitate malignant progression [7]. The important functional role of ASPMiI in Wnt activity in GC cells motivated us to investigate its potential role in GC stemness. We thus measured the proportion of GC cells expressing cell-surface Leucine Rich Repeat Containing G Protein-Coupled Receptor 5 (LGR5) or aldehyde dehydrogenase 1 (ALDH1), which contain the enriched population of CSCs in GAC [39, 40]. KD of ASPMv1 expression using shRNA#4 or shRNA#3 significantly shrank the populations of LGR5+ and ALDH1+ cells compared with those with control KD in different GC lines (Fig. 4a; Fig. S2e, f), and substantially lowered the expression levels of the pluripotency- and stemness-associated genes, including EZH2, KLF4, MYC, LGR5, NANOG, POU5F1, SMO, and SOX2, in GC cells (Fig. 4b). Indeed, ASPMiI-deficient GC cells exhibited a considerably lower tumorsphere-forming ability, a functional surrogate of cancer stemness, than that of control cells (Fig. 4c). Corroborating these findings, co-IF analysis showed that ASPMiI significantly colocalizes with representative stemness markers LGR5 and SOX2 in human GC tissues (Fig. 4d, e). These results suggest that CSCs in GC may rely on ASPMiI to maintain their population as well as their tumorigenic potential.

Fig. 4
figure 4

ASPMiI regulates the stemness properties and the population of CSCs in GC cells. a Specific KD of ASPMv1 expression diminished the population of LGR5+ or ALDH1+ CSCs in NCI-N87 cells. Shown are representative flow cytometry plots showing pattern of LGR5 expression in control-KD or ASPMv1-KD NCI-N87 cells. Right, the percentage of LGR5+ or ALDH1+ NCI-N87 cells (mean ± SEM, N = 3 biological replicates with two technical replicates). Statistical analysis was performed using unpaired t test. *P < 0.05, ***P < 0.001, t test. b The transcript levels of the indicated pluripotency- and stemness-associated factors in ASPMv1-KD NCI-N87 cells compared with control-KD cells as measured by qRT-PCR (mean ± SEM, N = 3 biological replicates with two technical replicates). *P < 0.05; **P < 0.01 compared with control KD, t test. c Representative phase contrast images of tumorspheres formed by control-KD or ASPMv1-KD NCI-N87 cells in serum-free and nonadherent culture plates for 10 days. Right, limiting dilatation assay demonstrating the tumorsphere-forming efficacy of NCI-N87 and AGS cells with control KD or ASPMv1 KD (maximum likelihood estimates with 95% confidence interval, N = 4 biological replicates with two technical replicates). d, e Representative IF images showing the colocalization (yellow) of ASPMiI (green) with LGR5 (red; d) or SOX2 (red; e) in human GC tissues. Nuclei were counterstained with DAPI (blue). Right panels, heatmaps illustrating the correlation of the staining intensity of ASPMiI with that of LGR5 (d) or SOX2 (e) in GC tissues (N = 8 tumors). Nuclei were counterstained with DAPI (blue)

Full size image

FOXM1 augments Wnt activity and gastric cancer stemness by regulating ASPMiI and DVL3 expressions

Both FOXM1 and ASPM have been shown to regulate Wnt signaling [27, 28], which prompted us to investigate their potential interplay in GC cells. We first verified that FOXM1 expression is significantly upregulated and strongly correlates with the transcript level of ASPMv1 in human GC tissues (Fig. 5a, b). Concordantly, co-IF and Venn diagram analyses confirmed that the subsets of CG cells staining positively (moderate to bright staining) for FOXM1, ASPMiI or DVL3 considerably overlaps with each other (Fig. 5c, d). In analogous to ASPM and ASPMiI, the expression level of FOXM1 correlates with the poor prognosis in patients with GC (Fig. 5e).

Fig. 5
figure 5

The roles of FOXM1 in ASPM expression, Wnt activity, and cancer stemness in GC. a Box plots of the expression level (transcripts per million, TPM) of FOXM1 in normal gastric epithelial tissue (N = 211) and GC tissues (N = 408) in the TCGA RNA sequencing data set. ***P < 0.001, t test. b Scatter plot showing the correlation between the transcript levels of ASPMv1 and FOXM1 in the TCGA RNA sequencing data set. c Representative IF images showing the colocalization of ASPMiI (green) with DVL3 (red) and FOXM1 (magenta) in human GC tissues. Nuclei were counterstained with DAPI (blue). d Venn diagram illustrating the overlap of the cell subsets stained positive for ASPMiI, DVL3, and FOXM1 in GC tissues (N = 8 tissues). e Kaplan–Meier survival curve comparing overall survival of GC patients in the KM Plotter data sets (N = 592) with their tumors having a high or low transcript level of FOXM1. f IB analysis showing the effect of KD of FOXM1 expression using two different shRNAs on the protein abundance levels of FOXM1 and ASPMiI in NCI-N87 cells. β-tubulin was included as a loading control. g Chromatin immunoprecipitation (ChIP)-qRT-PCR assays showing that FOXM1 is recruited to several putative binding sites (PBS) on the ASPM promoter in NCI-N87 cells (mean ± SEM; N = 3). **P < 0.01, ***P < 0.001, t test. h KD of FOXM1 expression reduced the transcript level of ASPMv1 in NCI-N87 cells as measured by qRT-PCR (mean ± SEM, N = 3 biological replicates with two technical replicates). **P < 0.05 compared with control KD, t test. i Co-IP analysis showing that FOXM1 associates with β-catenin in NCI-N87 cells. j KD of FOXM1 expression reduced the protein abundance levels of DVL3 and β-catenin in NCI-N87 cells. k KD of FOXM1 expression reduced the population of LGR5+ CSCs in NCI-N87 cells (mean ± SEM, N = 3 biological replicates with two technical replicates). **P < 0.01, t test. l Tumorsphere assay and LDA demonstrating the tumorsphere-forming efficacy of control-KD or FOXM1-KD NCI-N87 cells (maximum likelihood estimates with 95% confidence interval, N = 4 biological replicates with two technical replicates)

Full size image

To further substantiate the preceding clinical correlative analyses, we conducted a series of molecular studies and thereby established the crucial role of FOXM1 in the expressions of ASPMiI, DVL3, and β-catenin, the Wnt pathway activity, and the stemness properties of GC cells. First, we showed that KD of FOXM1 expression diminished the expression of ASPMiI in GC cells (Fig. 5f). ChIP-qRT-PCR analysis revealed that FOXM1 strongly binds to the promoter of ASPM in GC cells (Fig. 5g). Consistently, KD of FOXM1 expression substantially reduced the transcript level of ASPMv1 (Fig. 5h), suggesting that FOXM1 regulates ASPM expression at the transcriptional level. Moreover, affirming the findings previously reported in glioma and leukemia cells [27, 28], co-IP analysis revealed that FOXM1 strongly associated with β-catenin in GC cells (Fig. 5i). Furthermore, KD of FOXM1 expression markedly lowered the protein abundance levels of DVL3 as well as β-catenin in GC cells (Fig. 5j) and reduced the population of CSCs (Fig. 5k) and inhibited their tumorsphere-forming ability (Fig. 5l).

ASPMiI is indispensable for FOXM1-mediated Wnt/β-catenin activation

The findings that FOXM1 can transcriptionally regulate the expression of ASPMiI as well as directly regulate β-catenin translocation and expression [27, 28] (Fig. S3) raise an interesting question as to their relative importance in the Wnt signaling transduction and activity in GC cells. Since KD of ASPMv1 expression almost completely abrogated the WNT3A-induced Wnt activation in GC cells (Fig. 3d) despite a high endogenous level of FOXM1 expression (data not shown), we surmised that ASPMiI-mediated DVL3 and β-catenin stabilization may be more important than the direct effect of FOXM1 on β-catenin. ASPM is a huge (~ 409 kDa) protein whose OE in cells is highly technically challenging as shown previously [18]. We thus sought to address the above possibility by overexpressing its downstream effector DVL3 in GC cells. Indeed, while KD of FOXM1 expression markedly diminished β-catenin expression, the concurrent OE of DVL3 was able to significantly restore the protein abundance level of β-catenin (Fig. 6a). Conversely, forced OE of FOXM1 even at a super-physiological level failed to rescue the protein levels of β-catenin or DVL3 in ASPMiI-deficient GC cells (Fig. 6b). In accordance with these molecular studies, we confirmed at functional level that the concurrent OE of FOXM1 could not restore the Wnt pathway activity (Fig. 6c) or the tumorsphere-forming capacity of GC cells deficient in ASPMiI expression (Fig. 6d). These data underscored the indispensable role of ASPMiI in the FOXM1-dependent regulation of β-catenin and Wnt signaling activity in GC cells.

Fig. 6
figure 6

FOXM1-mediated β-catenin/Wnt activation and GC stemness requires ASPMiI. a IB analysis showing that OE of DVL3 significantly restored the protein abundance level of β-catenin in FOXM1-deficient NCI-N87 cells. b OE of FOXM1 failed to rescue the protein abundance levels of β-catenin and DVL3 in ASPMiI-deficient NCI-N87 cells. c Wnt-specific luciferase activity in NCI-N87 cells with OE of FOXM1 with or without the concurrent KD of ASPMv1 expression (mean ± SEM, N = 3 biological replicates with two technical replicates). ***P < 0.001, t test. NS, not significant. d LDA demonstrating the tumorsphere-forming efficacy of NCI-N87 cells with control KD, ASPMv1 KD or FOXM1 KD with or without the concurrent OE of FOXM1 (maximum likelihood estimates with 95% confidence interval, N = 4 biological replicates with two technical replicates)

Full size image

Discussion

GC is a molecularly heterogeneous disease with marked phenotypic and genomic diversity. Recent advances in genomic analyses have led to the identification of novel molecular subtypes in GC [10, 41]. Of note, the molecular patterns identified from a developed tumor may only reflect the accumulative effect of the malignant transformation process without providing information reflecting the pathogenetic processes. The only biology-informed biomarkers currently used in the clinic are the analysis of HER2, PD-L1 and/or microsatellite instability, which guides the use of anti-HER-2 therapies and immune checkpoint inhibitors [4, 42, 43]. Unfortunately, these therapies only benefit a small subset of GC patients and additional biology-informed and clinically applicable analysis that can guide the use of next-generation rationale therapies remains an unmet clinical need.

GC develops Wnt dysregulations at multiple levels ranging from ligands, receptors, upstream regulators and β-catenin [10, 11]. Notably, these genetic lesions were discovered by genomic analysis on whole-tumor samples, which do not consider the potential intra-tumoral and cell-to-cell heterogeneity in Wnt signaling as reported in other types of solid cancers [44, 45]. We have recently discovered the novel Wnt co-regulator ASPM, especially its isoform I (ASPMiI), which displayed a high intra-tumoral heterogeneity in prostate, pancreatic and liver cancers [18,19,20]. We here extended our findings to GC by demonstrating the considerable cell-to-cell heterogeneity in ASPM expression and Wnt activity in human GC tissues. Using a series of molecular and functional studies, we uncovered the Wnt-regulatory function of ASPMiI in stem-like GC cells and the intriguing molecular crosstalk of ASPMiI with another important Wnt and stemness regulator, FOXM1. Importantly, we elucidated that FOXM1 regulates the stability of β-catenin, Wnt pathway activity and cancer stemness in GC cells in an ASPMiI-dependent manner. Our findings thus shed new lights on the molecular basis of the dysregulated Wnt signaling in GC cells and underscore the critical and indespensable role of the FOXM1/ASPMiI/β-catenin interplay in the Wnt pathway activity and the aggressiveness of GC.

The molecular mechanism by which ASPM augments Wnt signaling in cancer cells has been attributed to its ability to interact with selective DVL proteins, whereby inhibits their proteasome-dependent degradation [17, 18, 20]. Notably, we have recently mapped the region of ASPM that mediates its interaction with DVL, which is located within the region encoded by exon-18 (data not shown) and is missing in another major isoform of ASPM, ASPMiII [12, 20]. Consistently, we showed that the expression of ASPMiI is specifically upregulated in GC tissues, where it colocalizes with active β-catenin in the cytoplasm of cancer cells. By contrast, ASPMiII is expressed mainly in the nuclei of both normal epithelial cells and cancer cells (Fig. 2b). Notably, the differential subcellular distribution of ASPMiI and ASPMiII is not only mechanistically important but also prognostically significant. Whist a previous analysis interrogating into whole-tumor transcriptome datasets has demonstrated a potential poor prognostic role of ASPM in GC [21], this study did not distinguish among the transcript variants of ASPM. With the advent of novel isoform-specific antibodies, our study provides an unprecedented opportunity of elucidating the prognostic role of ASPM in GC in an isoform-specific manner. We showed that the protein abundance level of ASPMiI is upregulated mainly in the cytoplasm in approximately a third of GC tissues, which correlated that of active β-catenin and with the poor prognosis in patients with GC, whereas the expression level of ASPMiII did not. We thus envisage that the IHC staining of ASPMiI may serve to identify a novel subset of GC cells in the tumors or a subset of GC tumors that are Wnt active and may respond to various Wnt-targeted molecular therapeutics currently undergoing preclinical or clinical developments [46], potentially yielding novel therapeutic opportunity for advanced GC.

Whilst the expression of ASPM is found to be upregulated in a wide variety of human malignant tumors [13,14,15,16], the mechanisms giving rise to its elevated expression remain poorly understood. Also elusive are the molecular mechanisms underlying the elevated expression of ASPM in CSCs. The strong regulatory role of FOXM1 on the transcript level of ASPM in GC cells as identified in the present study may provide a plausible explanation as FOXM1 is known to be involved in development, regeneration, tumorigenesis, and its expression is upregulated in stem cells [27, 28]. Our rigorous molecular studies indicate that FOXM1 and ASPM work in concert to promote the stability of β-catenin, resulting in the activation of Wnt signaling as well as the stemness properties of GC cells. FOXM1 and the ASPMiI–DVL3 signaling axis converge on β-catenin through protein–protein interactions, wherein ASPMiI plays a crucial and indispensable role in this novel Wnt-regulatory module in CSCs (Fig. 7). Notably, our bioinformatics analysis revealed that many malignant tumors concurrently upregulate the expressions of FOXM1 and ASPM (data not shown), raising a compelling question as to whether the interplay between FOXM1, ASPM and β-catenin may also play an important role in Wnt activation and cancer stemness in other types of malignant tumors. Further studies are warranted to confirm the generality of the role of this novel oncogenic mechanism.

Fig. 7
figure 7

Model depicting the FOXM1/ASPMiI/DVL3/β-catenin regulatory module that reinforces Wnt signals to control GC stemness and aggressiveness. CSCs in GC express a high level of FOXM1, which induces the transcription of ASPM. The ASPMiI protein specifically interacts with DVL3 in the cytosol and prevents it from the proteasome-dependent degradation, thereby stabilizing the β-catenin protein. In parallel, FOXM1 also directly mediates the nuclear translocation of β-catenin. This multi-mode regulatory module promotes β-catenin-dependent Wnt pathway activation, cancer stemness and aggressiveness in GC

Full size image

In conclusion, our findings illuminate a novel Wnt-related pathomechanism giving rise to and sustaining GC stemness and aggressiveness. We elucidate the interplay and the epistatic relationship among ASPMiI, FOXM1, DVL3, and β-catenin, which augments Wnt activity in CSCs in a cooperative manner. Our data not only highlight a promising role of ASPMiI and FOXM1 in the diagnosis of a novel Wnt-activated and stemness-high subtype of GC but also provides key molecular hubs toward which Wnt- and/or CSC-targeted rational therapies can be developed.