Dysplastic stem cell plasticity functions as a driving force for neoplastic transformation of pre-cancerous gastric mucosa

Abstract

Background & Aims:

Dysplasia carries a high risk of cancer development, however, the cellular mechanisms for dysplasia evolution to cancer are obscure. We have previously identified two putative dysplastic stem cell (DSC) populations, CD44v6^neg/CD133⁺/CD166⁺ (DP) and CD44v6⁺/CD133⁺/CD166⁺ (TP), which may contribute to cellular heterogeneity of gastric dysplasia. Here, we investigated functional roles and cell plasticity of non-cancerous Trop2⁺/CD133⁺/CD166⁺ DSCs initially developed in the transition from pre-cancerous metaplasia to dysplasia in the stomach.

Methods:

Dysplastic organoids established from active Kras-induced mouse stomachs were utilized for transcriptome analysis, in vitro differentiation and in vivo tumorigenicity assessments of DSCs. Cell heterogeneity and genetic alterations during clonal evolution of DSCs were examined by next-generation sequencing. Tissue microarrays were used to identify DSCs in human dysplasia. We additionally evaluated the effect of CK1α regulation on the DSC activities using both mouse and human dysplastic organoids.

Results:

We identified a high similarity of molecular profiles between DP- and TP-DSCs, but more dynamic activities of DP-DSCs in differentiation and survival for maintaining dysplastic cell lineages through Wnt ligand-independent CK1α/β-catenin signaling. Xenograft studies demonstrated that the DP-DSCs clonally evolve towards multiple types of gastric adenocarcinomas and promote cancer cell heterogeneity by acquiring additional genetic mutations and recruiting the tumor microenvironment. Lastly, growth and survival of both mouse and human dysplastic organoids were controlled by targeting CK1α.

Conclusions:

These findings indicate that the DSCs are de novo gastric cancer-initiating cells responsible for neoplastic transformation and a promising target for intervention in early induction of gastric cancer.

Graphical Abstract

LAY SUMMARY

Dysplastic stem cell (DSC) activity regulated by CK1α/β-catenin signaling can drive dysplastic cell evolution to adenocarcinoma in gastric carcinogenesis by acquiring cellular and genetic heterogeneity and recruiting microenvironment.

INTRODUCTION

Gastric cancer is the fifth most common cancer worldwide and the third leading cause of cancer-related mortality^{1, 2}. Intestinal-type gastric cancer occurs more frequently than diffuse-type gastric cancer and develops within a cascade of pre-cancerous metaplasia progression to dysplasia and adenocarcinoma^{3, 4}. Gastric dysplasia is the focal neoplastic lesion, which has the highest risk of intestinal-type gastric cancer development, and has been generally recognized as the initiation event in neoplastic transformation from pre-cancer to cancer⁵. To detect the risk of cancer development in dysplasia, histopathological investigations have been conducted with several markers highly expressed in both gastric dysplasia and adenocarcinoma, such as Cox-1 and Cldn7^{6, 7}, and our group has identified Trop2 as a novel dysplasia marker specifically upregulated in the transition between incomplete intestinal metaplasia (IM) and dysplasia⁸. However, the cellular plasticity or regulatory mechanisms of dysplastic cells during the neoplastic transformation remains largely unknown.

Stem cells are considered a key source for cellular biological functions with self-renewal and differentiation capacities. In cancer development, cancer stem cells (CSCs) are a small subset of cancer cells that contribute to intratumor heterogeneity and cancer progression or drug resistance in tumor lesions⁹. In particular, CSCs may evolve from pre-cancerous stem cells by clonal expansion, followed by sequentially acquired mutations or epigenetic regulation^{10, 11}, and those pre-cancerous stem cells may lead to the carcinogenic transition of dysplasia. Therefore, identification of distinct cell populations, which function as stem/progenitor cells that likely represent the foci of pre-cancerous initiation and progression, is crucial. Several studies in the field of gastric carcinogenesis have reported the expression of putative CSC markers such as CD133, Musashi-1 and CD44v6 in gastric pre-cancerous tissues^12–14. Nevertheless, the presence of de novo stem cells which function as cancer-initiating cells in the pre-cancerous lesions still remains obscure due to the lack of surrogate markers or appropriate model systems.

Kras activation and amplification is one of the major oncogenic mechanisms frequently present in intestinal-type gastric cancer^{15, 16}. We have previously studied the roles of Kras activation during gastric carcinogenesis using a transgenic mouse model which induces active Kras expression in normal chief cells¹⁷. We also studied the cellular characteristics and dynamic behaviors of dysplastic cells using a dysplastic organoid model established from the active Kras-induced dysplastic glands¹⁸. The dysplastic organoids recapitulate common cytological and histological phenotypes of human dysplasia and highly express Trop2, a dysplastic marker^{8, 17, 18}. We identified several dysplastic subpopulations, including putative dysplastic stem cell (DSC) subpopulations such as CD44v6^neg/CD133⁺/CD166⁺ (double-positive, DP) and CD44v6⁺/CD133⁺/CD166⁺ (triple-positive, TP) cells¹⁸.

Here, we have sought to understand the biological functions of the two putative DSC subpopulations in homeostatic regulation and neoplastic changes of dysplasia using dysplastic organoids established from the Kras-induced mouse model and from human patient samples with dysplasia. We have identified Trop2⁺/CD133⁺/CD166⁺ DSCs as a key stem cell population which perpetuates heterogeneous cell lineages in dysplasia through Wnt ligand-independent CK1α/β-catenin signaling activation. We have also defined the oncogenic potential and clonal evolution ability of the DSCs during dysplasia evolution to adenocarcinoma leading to further tumor cell heterogeneity and mutational burden. We also suggest activation of CK1α as a potential therapeutic target for intervention in dysplasia progression.

MATERIALS AND METHODS

Organoid culture and drug treatment

Meta4 dysplastic organoids were previously established from corpus stomachs of Mist1-CreERT2Tg/+;LSL-K-ras(G12D)Tg/+ (Mist1-Kras) transgenic mice^{17, 18}. The Meta4 organoids have been maintained without morphological and phenotypical changes by continuous freezing, thawing and passaging over 4 years after derivation. All of the Meta4 organoids, DSC-derived spheres and tumor spheres were cultured in Matrigel (ECM, Sigma) with Mouse IntestiCult medium (StemCell Technology) supplemented with 1% of penicillin/streptomycin (Gibco) in 48-well plates and medium was replaced every 3 days. The organoids were split every 5–7 days before they formed budding structures.

To perform Wnt pathway studies, Wnt and R-spondin (Rspo) conditioned media were prepared as described in Table S1. Wnt, Rspo and Noggin conditioned medium were provided from Vanderbilt Digestive Diseases Research Center. Meta4 organoids were split, embedded in Matrigel and overlayed with Wnt/Rspo conditioned media after Matrigel polymerization at 37°C. The organoids were cultured for 3–7 days. Three Wnt pathway regulators, Salinomycin, Pyrvinium pamoate and LGK-974 (all from MedChemExpress) were dissolved in DMSO at a stock concentration of 200 μM, 20 μM and 20 μM, respectively. Meta4 or human organoids were split and cultured in the Mouse or Human IntestiCult media for 1–2 days until they formed three-dimensional (3D) spherical structures, then the media was switched to Wnt and Rspo free media (W^neg/R^neg) containing either DMSO vehicle or a final concentration of 1 μM of Salinomycin, 100 nM of Pyrvinium, 100 nM of LGK-974 and cultured for 3 days. The JuLI^™ stage, a Real-Time Cell History Recorder (NanoEntek), and an EVOS M7000 inverted microscope were used to obtain phase contrast images of organoids. All experiments were repeated at least three times.

Tumor formation in nude mice

Total live dysplastic cells, CD133^neg/CD166^neg (Non-DSCs), CD44v6^neg/CD133⁺/CD166⁺ (DP-DSCs) and CD44v6⁺/CD133⁺/CD166⁺ (TP-DSCs) were isolated from five to seven 48-well plates of Meta4 organoids by FACS 5 to 7 days after splitting. Tumor spheres generated from either cystic adenocarcinoma (CysAC) or tubular adenocarcinoma (TubAC) were dissociated from twenty wells of 48-well plates 3 days after splitting. After the cell sorting or dissociation, 15,000 or 30,000 cells per injection were mixed with 100 μL of Matrigel (Corning) with 0.1% Y-27632 on ice.

Female Nu/J nude mice at 6 weeks of age (Jackson Laboratory) were used to perform the tumorigenicity assay. The care, maintenance, and treatment of mice used in this study followed protocols approved by the Institutional Animal Care and Use Committee of Vanderbilt University and each experimental group contained three to five mice. The 100 μL of Matrigel mixture containing cells were subcutaneously injected into both flanks of the mice. Tumor size and weight were measured twice a week using a caliper and tumor volume was calculated by length (L) × width (W)². At 3 to 13 weeks after the injection, all mice were sacrificed and tumor masses were resected. The resected tumors were used for hematoxylin and eosin (H&E) staining, immunostaining, establishment of tumor spheres or single cell RNA-sequencing (scRNA-seq).

Human tissue specimens and organoid generation

Tissue microarray construction from gastric cancer patient tissues was approved by the Institutional Review Board (IRB) of Seoul National University Hospital (SNUH), Seoul, Korea (IRB No. H-1209-037-424). Two tissue microarray slides with total 90 tissue cores (76 dysplastic, 8 metaplastic, 4 normal corpus and 2 normal antrum) were used for immunofluorescence staining and subsequent immunohistochemistry.

To establish human metaplastic or dysplastic organoids, fresh tissue specimens were obtained from patients who underwent curative gastrectomy at SNUH (IRB No. H-1806-166-954). About 5 cm of tissue strips were resected from intestinal-type gastric cancer to adjacent non-cancerous lesion in direction to the corpus and utilized for organoid generation. All information of each patient was anonymized and de-identified prior to experiments. The clinicopathological information such as Lauren classification, WHO classification, tumor location, TNM stages and MSI status is provided in Table S2.

The tissue strips were washed in ice-cold PBS with 100 μg/mL of Primocin and cut into pieces using a razor blade. Each tissue piece was divided into two pieces for pathological examination by H&E staining and establishment of organoids. Stomach mucosa was separated from serosa along the muscle layer using cell scrapers, minced using a tissue chopper and incubated in pre-warmed digestion buffer at 37°C with shaking at 220 rpm for 30 min. After the digestion, pre-warmed quenching buffer was added and the dissociated glands were centrifuged at 300×g for 5 min. Pellets were washed in 5 mL of pre-warmed quenching buffer, strained through a 100 μm cell strainer and centrifuged. The glands-containing pellets were mixed with ice-cold Matrigel and plated in a 48-well plate and incubated 37°C for 30 min to polymerize the Matrigel, then Human IntestiCult medium (StemCell Technology) supplemented with 1% of penicillin/streptomycin, 0.2% of MycoZap (Lonza) and 0.1% Y-27632 was added.

RESULTS

DP- and TP-DSCs display molecular similarity as cancer stem cells (CSCs), but distinct functional cell states

For a molecular characterization of the dysplastic stem cells (DSCs), we performed RNA-sequencing using FACS-isolated DP- (CD44v6^neg/CD133⁺/CD166⁺) and TP- (CD44v6⁺/CD133⁺/CD166⁺) DSCs from Trop2⁺ Meta4 dysplastic organoids¹⁸. When comparing highly expressed genes between the two DSC subpopulations, more than 85% of the genes overlapped each other with similar gene ontology profiles (Figure 1A and S1A). Many genes related to stem cell- and cancer-related pathways, such as Ctnnb1 (Wnt), Ccnd1 (Wnt/Cell cycle) and Ywhaz (EGF/PI3K/VEGF), were highly expressed in both DSC subpopulations (Figure 1B and S1B). Differentially-expressed gene analysis revealed 1,318 upregulated genes related to transcription regulation, homophilic cell adhesion and PI3K-Akt or Wnt pathways in DP-DSCs and only 53 upregulated genes of unknown function in TP-DSCs (Figure 1C, 1D, S1C and S1F). Interestingly, the Wnt pathway-related signature was observed in both profiles, and Fzd9, Wnt6 and Wnt10a were expressed only in DP-DSCs (Figure S1D and S1E). These results demonstrated similar CSC-like molecular signatures between the two DSC subpopulations, but the profiles of DP-DSCs displayed more active and functional stem cell-like characteristics than TP-DSCs.

Figure 1. RNA-seq data and cellular functions of DP- and TP-DSCs.

(A) Similarity matrix based on top 2,000 genes in DP- or TP-DSCs and the percentages of overlapping genes. (B) KEGG mapping of top 434 of 500 genes commonly observed in both DSCs. (C&D) Gene set enrichment analysis (GSEA) of biological process (C) and KEGG (D) using upregulated genes in DP-DSCs. (E) Phase-contrast images of isolated non-DSCs, DP- and TP-DSCs from Meta4 organoids or Mist1-Kras mouse stomachs in Matrigel at 0, 1 or 4 weeks. (F-G) Quantitation of the number of spheres from isolated non-DSCs, DP- and TP-DSCs from Meta4 organoids (F) or Mist1-Kras mouse stomachs (G). Mean ± SD (n=3 or 4). Paired t-test. *P < .05. (H-I) H&E and AB/PAS (H) and co-immunostaining for Trop2 and CD44v9 or Tff3 and Hoechst (I) in DSC-derived spheres at 4 weeks. Arrows (H) and dotted boxes (I) indicate enlarged area. Arrows in (I) denote CD44v9^neg/Trop2⁺ dysplastic cells or Tff3⁺ goblet cells.

Isolated DP-DSCs can give rise to multiple cell lineages and maintain the cellular composition

In dysplastic glands of Mist1-Kras mouse stomachs¹⁷, the two DSC subpopulations were distinguished by two lineage-specific markers, CD44v9, a metaplasia marker¹⁹, and Trop2, a dysplasia marker⁸ (Table S3). While TP-DSCs were present at the base of glands where the CD44v9⁺/Trop2^neg or CD44v9⁺/Trop2⁺ intermediate cells are located (Figure S2A, yellow arrow), DP-DSCs were present in the region spanning the transitioning cell zone from metaplastic to CD44v9^neg/Trop2⁺ dysplastic cells (Figure S2A, red arrow). We also observed heterogeneous cell lineages between CD44v9⁺/Trop2⁺ intermediate cells and CD44v9^neg/Trop2⁺ dysplastic cells in Meta4 organoids (Figure S2B).

To assess the DSC plasticity, we performed a 3D long-term culture of isolated DP-, TP-DSCs or CD133^neg/CD166^neg non-DSCs from Meta4 organoids. Only DP- and TP-DSCs formed spheres with distinct sphere formation efficiency, as we have previously observed¹⁸, and the spheres recapitulated the Meta4 organoid phenotypes at 1 week by expressing Trop2, CD44v9 and Cttn, an invasion marker (Figure 1E, 1F and S2C). The spheres derived from both DP- and TP-DSCs generally maintained their cell lineages and a similar cellular ratio of non-DSCs, DP- and TP-DSCs after continuous passaging (Figure S2D, S2E and S2F). While the spheres displayed dysplastic histology (Figure S2G), a distinct differentiation ability with complicated structures between DP- and TP-DSCs was observed (Figure 1E). While differentiated cell lineages from TP-DSCs were restricted to CD44v9⁺/Trop2⁺ intermediate cells, DP-DSCs showed more complicated dysplastic features including multilayering of cells with disorganized nuclei and a dynamic differentiation capacity into multiple cell lineages, including various mucus-secreting cells including PAS⁺ cells, Alcian blue (AB)⁺ or Tff3⁺ goblet cells (Figure 1H and 1I). Fully differentiated dysplastic cells, CD44v9^neg/Trop2⁺ cells, were mainly observed in spheres derived from the DP-DSCs (Figure 1I, arrow). The different sphere formation and differentiation abilities were also confirmed using DP- or TP-DSCs directly isolated from Mist1-Kras mouse stomachs (Figure 1E, 1G, 1H and 1I). Therefore, these data demonstrated that DP-DSCs have a better ability to differentiate into various cell types than TP-DSCs, implying that the DP-DSCs display an active state of dysplastic stem cells (Figure S2H).

DP-DSCs lead to the evolution of dysplasia into various stages of gastric cancer

To elucidate whether DSCs are responsible for the evolution of dysplasia towards adenocarcinoma, total Meta4 cells, sorted non-DSCs, DP- or TP-DSCs from Meta4 organoids were injected subcutaneously under the flank of immunodeficient nude mice (Figure 2A). Only DP-DSCs successfully engrafted with 33% formation rate from 15,000 cells at 7 weeks after the injection and the engraftment efficiency was increased to 75% after 30,000 cell injection (Figure 2B). The engraftments retained dysplastic phenotypes, positive for Trop2, Sox9 and Cttn, as well as CD133 and CD166, confirming the presence of DP-DSCs (Figure 2C). Importantly, groups of cells positive for Cdh17, an intestinal-type gastric cancer-related marker²⁰, were observed and EpCAM, an epithelial cell marker, was strongly positive, whereas the Meta4 organoids were Cdh17-negative (Figure 2C and S3A). Moreover, Masson’s trichrome staining demonstrated recruitment to the engraftment sites of stromal cells and collagen fibers with desmoplastic reactions, which frequently occur around malignant neoplasms (Figure 2C, arrow). Many Pdgfrα⁺ or α-SMA⁺ fibroblasts and F4/80⁺/CD163^neg non-polarized macrophages or dendritic cells were recruited to the stroma surrounding the epithelial cells. Furthermore, many Ki67⁺ cells were observed in the engraftments, while only a few cleaved caspase-3⁺ apoptotic cells were present (Figure 2C). However, no PECAM⁺ endothelial cells and only rare Vegf⁺ cells, which are pro-angiogenic cells²¹, were observed around the engraftments (Figure 2C and S3B).

Figure 2. Evolution of DP-DSCs towards heterogeneous types of gastric cancer.

(A&D) Experimental scheme of isolated DSC-injection study created with BioRender.com. (B) Bright-field (BF) images of injection sites at 7 weeks after injection. Dotted areas indicate engraftments. Graph shows engraftment rates. (C) H&E, Masson’s trichrome (MT) or co-immunostaining for Sox9, Trop2, Cttn, CD44v6, CD133, CD166, Cdh17, EpCAM, Pdgfrα, α-SMA, Ki67, Cleaved caspase-3 (CC-3), F4/80, CD163, P120 (epithelial cells) or PECAM in DP-DSC-derived engraftments at 7 weeks after injection. Dotted boxes indicate enlarged area. (E) Tumor volumes in each group for 13 weeks. Paired t-test. *P < .05, **P < .01, ***P < .001. (F) BF and H&E images of tumors from total Meta4 cells or isolated DP-DSCs at 13 weeks after injection. Arrows denote CysAC (black) or TubAC (red) lesions. Graph shows tumor formation rates. (G) H&E or MT staining of CysAC or TubAC tissues from total Meta4 cells or isolated DP-DSCs. Arrows denote desmoplastic response areas. (H) H&E of each histological feature is indicated by arrows. Dotted boxes indicate enlarged area. (I) Co-immunostaining for Trop2, Cdh17, EpCAM, α-SMA, F4/80, PECAM or Ki67 in whole tumor tissues.

Notably, engraftment formation from DP-DSCs resulted in 100% solid tumor formation 13 weeks after the injection and total Meta4 cells also formed tumors (Figure 2D and 2F). But, no engraftment or tumor was observed from non-DSCs (Figure S3C). DP-DSC-derived tumors were significantly larger than total Meta4-derived tumors and developed central vein formation, which supports tumor growth, and 40% of DP-DSC-injected mice required sacrifice even earlier than 13 weeks due to the skin ulceration caused by tumor overgrowth (Figure 2E and 2F). Both total Meta4 cells and DP-DSCs developed heterogeneous types of tumors: either cystic adenocarcinoma (CysAC) or high-grade tubular adenocarcinoma (TubAC) including stromal invasion, desmoplastic stroma and even single invasive tumor cells (Figure 2G and 2H, arrows, Table S4). Interestingly, the Trop2⁺ dysplastic cells were still present in the CysAC lesions, but decreased in the high-grade TubAC lesions (Figure 2I, arrows, and S3D). EpCAM was expressed uniformly in the cancer cells and Ki67⁺ cells were distributed throughout the lesions (Figure 2I and S3D). The Cdh17⁺ cell zones were surrounded by α-SMA⁺ fibroblasts and F4/80⁺ immune cells, suggesting that neoplastic transformation of DSCs may be facilitated by the recruited microenvironment (Figure 2I and S3D). In particular, PECAM⁺ endothelial cells were observed around the cancer cells, indicating the intratumor vascularization (Figure 2F, 2I and S3D). Therefore, these results provide direct evidence of the spontaneous malignant transformation of DP-DSCs into multiple types of gastric adenocarcinoma.

DP-DSC-derived tumors recapitulate tumor heterogeneity and genetic alterations commonly observed in human gastric cancers

To assess the molecular and cellular heterogeneity of the DP-DSC-derived tumors, we performed single-cell RNA-sequencing²² using CysAC and TubAC cells. The t-SNE plot clearly showed distinct subpopulations of epithelial cells in both tumor types including dysplastic, cancer and dysplastic/cancer intermediates with different distribution patterns between tumor types (Figure 3A). While the dysplastic subpopulations, highly expressing Cldn7, Sox9, CD9, and a DSC marker, CD133, were more enriched in the CysAC, the dysplastic/cancer intermediates and cancer cell subpopulations, expressing many known cancer-related genes such as S100a6, Tbx20, Rack1 and Cfl1, were more enriched in the TubAC (Figure 3B). Immune cell subpopulations, expressing Lyz2, Ccr1, Slfn4 and Ccl3,²³ (Figure 3C) and a fibroblast subpopulation, expressing Pdgfra, Col5a3, Mmp2 and Loxl1,²³ (Figure 3D) were also identified.

Figure 3. Cell heterogeneity of DP-DSC-derived tumors.

(A) t-SNE plot overlayed with cells from CysAC or TubAC (top) or cell clustering into 11 subpopulations (bottom). (B-D) t-SNE plot overlayed with expression of representative genes (blue) in each subpopulation, dysplastic/cancer (B), immune (C) or fibroblast (D). (E-J) Co-Immunostaining for Sox9, Trop2, CD133, Cldn7, CD9 or Pdgfrα (E, F & G) or Ly6B2, CD4, CD19, F4/80, CD3 or Vegf (H, I & J) in CysAC or TubAC. Arrows indicate enlarged area and dotted areas denote Sox9- or Cldn7-negative cancer regions in E&F.

These heterogeneous transcriptome signatures were validated in the tumors. While dysplastic cells, positive for Trop2, Sox9, CD133, Cldn7 and CD9, were widely distributed in epithelial cells of CysAC, multiple lesions in TubAC were negative for those proteins, and especially, Trop2 was lost in the TubAC tissues with many Ki67⁺ proliferating cells (Figure 3E, 3F and 3G). Pdgrfα⁺ fibroblasts, immune cells, such as neutrophils (Ly6B2), T-lymphocytes (CD4), B-lymphocytes (CD19), macrophages/dendritic cells (F4/80) and T-helper cells (CD3), as well as Vegf⁺ cells were observed in the stromal regions of both CysAC and TubAC, indicating that the tumor formation might be supported by recruited microenvironment and increased angiogenesis (Figure 3G, 3H, 3I and 3J). Therefore, these data suggest that clonal expansion of DP-DSCs can lead to progression of dysplasia to multiple types of gastric adenocarcinomas and establish tumor microenvironments.

We additionally examined the self-renewal ability of cancer cells from both CysAC and TubAC. Cells dissociated from both tumors formed spheres and grew continuously (Figure S4A). Consistent with our observation of distinct marker expression patterns, CysAC spheres showed partial loss of Trop2 and TubAC spheres were completely negative for Trop2, indicating that the established tumor spheres recapitulate the primary tumor phenotypes (Figure 4A). In addition, sphere formation efficiency and growth rate were significantly higher in TubAC cells compared to CysAC cells (Figure 4B, 4C and 4D).

Figure 4. Genetic alterations acquired during the DP-DSC evolution.

(A) Phase-contrast images or immunostaining for Trop2 in Meta4 organoids, CysAC- or TubAC-derived spheres in Matrigel. (B-D) Phase-contrast images (B), quantitation of the number (C) or diameters (D) of spheres derived from 1,000 or 5,000 cells dissociated from the tumor spheres at day 12. Mean ± SD (n=3). Each dot in D indicates a sphere diameter cumulated from three independent experiments. Unpaired t-test. **P < .01, ***P < .001, ****P < .0001. (E) Oncoplot of coding mutations in CysAC or TubAC. Top row indicates top 100 genes abundantly mutated in human gastric cancer cases (n=165) and the following five rows indicate gene mutations in 5 independent tumors, CysAC (n=2) or TubAC (n=3). (F) Venn-diagram presenting unique or overlapping genes between CysAC and TubAC. Genes affected by at least one mutation were classified into CysAC, TubAC or both. (G) Schematic domain structures of Rnf43, Cdkn2a and Apc with somatic mutations in Meta4 organoids, CysAC or TubAC created with BioRender.com. SP, signaling peptide; TM, transmembrane domain, ARM; armadillo repeat, BD; binding domain.

We next performed whole-exome sequencing using dissociated cells from CysAC (n=2) or TubAC (n=3) spheres to identify additional somatic mutations acquired during the DSC evolution to cancer cells. In comparison with the mutation patterns in cells from Meta4 organoids, more than 10,000 single-nucleotide variants (SNV) and 2,500 indels were identified in both tumor types (Figure S4B) and about 15–20% of the mutations occurred in coding regions (Figure S4C and S4D). The mutation profiles were compared to the genomic mutation data available from The Cancer Genome Atlas (TCGA)¹⁵. 67 genes out of the top 100 genes abundantly mutated in human gastric cancers were also mutated in CysAC and/or TubAC, demonstrating similar genetic mutation signatures between the mouse DP-DSC-derived tumors and human gastric cancers (Figure 4E). Both tumor types repeatedly showed mutations in genes which have critical roles in human gastric cancer development and progression, including Arid1a (2/2 of CysAC and 2/3 of TubAC), Cdh1 (all 5 tumors), Fat4 (all 5 tumors) and Cdkn2a (all 5 tumors). Also, unique gene mutations, including Wnt pathway-related genes, such as Rnf43 (1/2 of CysAC) and Apc (2/3 of TubAC), were also identified in either CysAC or TubAC cells (Figure 4F). In particular, many mutations occurred in functional domains of the key genes such as Rnf43, Cdkn2a and Apc, suggesting genetic alternations in key genes involved in signal transduction or interaction with other molecules during the dysplasia evolution (Figure 4G). Therefore, these results indicate that the mutations acquired during the DP-DSC evolution may facilitate development of multiple types of gastric adenocarcinoma.

We further characterized the progression of the two tumor types by reimplantation of dissociated CysAC or TubAC cells from the established tumor spheres under the flanks of immunodeficient nude mice. The injected tumor cells rapidly formed solid tumors within 3 weeks. As we observed that TubAC cells showed high proliferation activity and contained more cancerous cells, tumor sizes from the TubAC cells were about 5 times bigger than those from CysAC cells (Figure 5A and 5B), but their original histology did not change (Figure 5D and 5E and Table S4) and increased PECAM⁺ and Vegf⁺ cells were observed only in TubAC cell-derived tumors (Figure 5E). However, the CysAC cell-derived tumors continuously grew and eventually progressed to TubAC-type, negative for Trop2 but positive for Cdh17, with vascular invasion at 7 weeks (Figure 5A, 5C, 5D and 5E and Table S4). Thus, these results suggest that clonal evolution of DSCs can lead to the sequential progression of gastric adenocarcinomas (Figure 5F).

Figure 5. Examination of tumorigenic abilities of DP-DSC-derived tumor cells.

(A) BF images of tumors developed from CysAC or TubAC cells at 3 and 7 weeks after injection. (B-C) Tumor volumes in each group at 3 weeks (B) and 7 weeks (C). Two-way ANOVA. ****P < .0001. (D) Quantitation of the proportion of cystic or tubular histology in the CysAC (n=10)- or TubAC (n=20)-derived tumors. Mean ± SD. (E) H&E or co-immunostaining for PECAM and Trop2 or Vegf and Cdh17 in CysAC- or TubAC-derived tumors. Nuclei were counterstained with Hoechst and dotted boxes indicate enlarged area. (F) Graphical diagram of clonal evolution of DP-DSCs towards CysAC or TubAC created with BioRender.com.

Regulation of CK1α activity suppresses growth and survival of DSCs

Since we identified many Wnt pathway-related signatures in the transcriptome profiles of DSCs, we next examined whether the DSC plasticity is maintained by Wnt pathway. We cultured Meta4 organoids in three different Wnt and R-spondin (Rspo) conditioned media, Wnt/Rspo-supplemented (W⁺/R⁺), Wnt-depleted (W^neg/R⁺) and Wnt/Rspo-depleted (W^neg/R^neg). All spheres grown in the three conditions displayed similar morphologies with Trop2 and Ki67 expression (Figure S5A and S5D) and underwent differentiation into the Tff3⁺ goblet cells (Figure S5E). β-catenin activation was confirmed at the protein level in both cytoplasmic and nuclear fractions of the organoids (Figure S5B). Importantly, more than 75% of Trop2⁺ cells were DP-DSCs in all three conditions (Figure S5C) and the proportion of non-DSCs and DSCs in W^neg/R^neg condition was maintained with a similar level over time (Figure S5F). These data suggest that DSCs can self-renew and differentiate independently from external Wnt/Rspo stimulation.

Since the maintenance of dysplastic organoid growth and cell lineages by the DSCs was independent from exogenous Wnt stimulation, we next evaluated whether inhibition of downstream signaling of Wnt pathway can control the DSC activity. We treated Meta4 organoids with two FDA-approved drugs, Pyrvinium, a CK1α activator, and Salinomycin, a LRP5/6 receptor inhibitor, in W^neg/R^neg media for 3 days (Figure 6A). The Meta4 organoids treated with either Salinomycin or Pyrvinium displayed a significant reduction in size with only 13.5% and 5.0% of survival rates, respectively (Figure 6B, 6C and 6D). While surviving organoids displayed clear internal lumens (Figure S5G), Pyrvinium-treated organoid structures were more disrupted and contained many pyknotic cells (Figure 6B and 6E). Live/dead cell staining and FACS analysis confirmed the significant increase of organoid death after Pyrvinium treatment, whereas many live cells still remained in the Salinomycin-treated organoids (Figure 6F and 6G). Importantly, the proportion of DSCs in the remaining live cells was also significantly decreased after Pyrvinium treatment indicating that CK1α regulation targets DSCs (Figure 6H). These results were confirmed by decreased expression in Pcna and Ki67, and increased expression in cleaved caspase-3 (Figure 6I). In addition, Trop2 expression was decreased in Pyrvinium-treated organoids compared with Salinomycin-treated organoids and Cttn was not detected in either condition, suggesting loss of the invasive phenotype (Figure 6I). We would note that blocking Wnt ligand secretion using LGK-974, a porcupine inhibitor, did not affect organoid growth or survival (Figure S5H), but a direct targeting of the downstream mediator, CK1α, using the Pyrvinium with the minimum effective dose at 100 nM was sufficient to suppress the organoid growth and survival (Figure S5I).

Figure 6. Treatment of Salinomycin and Pyrvinium in Meta4 organoids.

(A) Schematic illustration of target molecules of Salinomycin and Pyrvinium in Wnt pathway created with BioRender.com. (B) Phase-contrast images of Meta4 organoids treated with DMSO vehicle, 1 μM of Salinomycin or 100 nM of Pyrvinium for 3 days. (C&D) Quantitation of diameters (C) or the number of surviving organoids (D) before and after the treatment. Mean ± SD (n=4). Each dot in C indicates an organoid diameter cumulated from four independent experiments. Paired t-test. **P < .01, ****P < .0001. (E, F and I) H&E (E), live/dead (Calcein AM/EthD-1) cell staining (F) or co-immunostaining for Trop2, Cttn, Ki67, Pcna, Cleaved caspase-3 (CC-3) or P120 (I) after the treatment. Dotted boxes depict enlarged area. (G-H) Quantitation of the number of live and dead cells (G) or non-DSCs and DSCs among live cells (H) after the drug treatment. Mean ± SD (n=3). Paired t-test. *P < .05. (J) Phase-contrast images of DP-DSCs isolated from Meta4 organoids 7 days after drug treatment and 7 days post drug withdrawal. (K) Quantitation of the number of surviving organoids after drug treatment and withdrawal. Mean ± SD (n=3).

We further examined whether Fzd9, an upregulated Wnt receptor in DP-DSCs, is associated with the β-catenin activation. Although Fzd9 was detected only in dysplastic organoids and expressed in cell membranes with or without Wnt/Rspo addition (Figure S5J, S5K and S5L), Fzd9 knockdown in Meta4 cells did not show any changes in Tcf7 expression, a transcription factor regulated by β-catenin activation (Figure S5M and S5N), indicating that Fzd9 might not be a key receptor controlling the DP-DSC activity. However, a direct knockdown of Ctnnb1 in Meta4 cells significantly reduced Tcf7 expression (Figure S5O and S5P). We additionally treated isolated DP-DSCs from Meta4 organoids with either Salinomycin or Pyrvinium to confirm whether the drugs just arrested proliferation activity in surviving DSCs. DP-DSCs treated with DMSO formed many spheres within 7 days and further showed dynamic morphological changes 7 days post DMSO withdrawal. While the DP-DSCs treated with Salinomycin successfully reformed spheres after the drug withdrawal, 100% of Pyrvinium-treated DP-DSCs were failed to reform spheres indicating direct effects of Pyrvinium on the DSC survival (Figure 6J and 6K). Therefore, these data suggest that the DSC activities and survival are dependent on downstream mediators of the Wnt pathway, such as CK1α and β-catenin, rather than Wnt ligands or receptors.

Cellular activity of TROP2⁺/CD133⁺/CD166⁺ DSCs in human dysplasia can be controlled by Pyrvinium

Given that our findings indicate DSCs are a fundamental source for maintaining dysplastic cell lineages and the evolution of dysplasia towards cancer, we next investigated the existence of DSCs in human dysplasia. We first examined the expression of three DSC markers, CD44v6, CD133 and CD166, in human patient tissues with normal, metaplastic, dysplastic and intestinal-type gastric cancer lesions. While CD166 was present in all stages and CD44v6 was expressed from metaplasia, CD133 was the only marker, specifically expressed in dysplasia and cancer (Figure S6A). To evaluate further the presence of DSCs during metaplasia progression to dysplasia, we immunostained a set of human tissue microarrays for CD44v6 and CD133 to distinguish two DSC populations as well as CD44v9 and TROP2 to discern transitioning or dysplastic glands.

CD133⁺ cells were only observed between the CD44v9⁺/Trop2⁺ metaplasia-to-dysplasia transitioning zone and the CD44v9^neg/Trop2⁺ dysplastic cell zone, suggesting the de novo production of DSCs during metaplastic cell progression to dysplastic cells (Figure S6B). In particular, 32.9% of dysplastic cores (25/76) contained CD44v6^neg/CD133⁺ cells (DP-DSCs) predominantly in CD44v9^neg/Trop2⁺ dysplastic cells (84%, 21/25) and 38.2% (29/76) contained CD44v6⁺/CD133⁺ cells (TP-DSCs) mainly in the CD44v9⁺/Trop2⁺ transitioning zone (96.6%, 28/29) (Figure 7A, 7B, 7D and S6C). Multiplex immunostaining for KI67 in the same tissue arrays demonstrated a significantly higher proliferation rate for DP-DSC-containing glands (80%, 20/25), compared to those of TP-DSC-containing glands (31%, 9/29) (Figure 7C and 7D). These findings are consistent with the presence of DSCs in Mist1-Kras mouse stomachs (Figure S2A)¹⁸.

Figure 7. Evaluation of DSC activities in human gastric dysplasia.

(A) Quantitation of the number of dysplastic tissue cores (n=76) containing DP-DSCs (CD44v6^neg/CD133⁺) and/or TP-DSCs (CD44v6⁺/CD133⁺). (B&C) The percentages of dysplastic tissue cores containing DP- or TP-DSCs in CD44v9⁺ and/or TROP2⁺ cell zones (B) or in Ki67⁺ cell zones (C). (D) Representative images of H&E or multiplexed-immunostaining for TROP2, CD44v6, CD44v9, CD133 and KI67. Dotted boxes indicate enlarged area of a DP- or TP-DSC zone. Arrows indicate a KI67-positive DP-DSC-zone in CD44v9^neg/TROP2⁺ dysplastic glands or a KI67-negative TP-DSC-zone in CD44v9⁺/TROP2⁺ transitioning glands. (E) Representative picture of surgical specimens from gastric cancer patients. Box indicates the collected region. (F) H&E (top) or co-immunostaining for TROP2 and CD44v9 in patient tissues with metaplasia or dysplasia (middle) or co-immunostaining for TROP2, CD44v9 and AQP5 in organoids derived from the tissues (bottom). (G) The percentages of TROP2⁺ cells or TROP2⁺ DSCs in dysplastic organoids. Mean ± SD (n=3). (H) Phase-contrast images of isolated non-DSCs and DSCs from human dysplastic organoids in Matrigel at 0, 1 or 4 weeks. Dotted boxes indicate enlarged area. (I&K) Images of H&E (I) or AB/PAS (K) staining. Arrows denote enlarged area. (J&L) Co-immunostaining for TROP2, CD44v9 and AQP5 (J) or TFF3 and Hoechst (L) in DSC-derived spheres at 4 weeks. Dotted boxed indicate enlarged area. Arrows denote TROP2⁺ dysplastic cells (J) or TFF3⁺ goblet cells (L). (M) Phase-contrast images at day 0 or merged with images captured after live/dead (Calcein AM/EthD-1) cell staining 3 days after the treatment in metaplastic or dysplastic organoids. (N) Quantitation of the number of metaplastic (n=2) or dysplastic (n=5) organoid lines positive for Calcein AM (live) or EthD-1 (dead) after treatment with either Salinomycin or Pyrvinium. Mean ± SD. Paired t-test. *P < .05. (O) Phase-contrast images of isolate DSCs from human dysplastic organoids 7 days after treatment and 7 days post drug withdrawal. (P) Quantitation of the number of surviving organoids after the drug treatment and withdrawal. Mean ± SD (n=3).

We next established metaplastic and/or dysplastic organoids from intestinal-type gastric cancer patient tissues with non-cancerous lesions adjacent to cancer (Figure 7E). The collected tissues were identified as metaplasia or dysplasia by histological examination and predominant expression of lineage markers (Figure 7F). Dysplastic organoids displayed strong TROP2 expression with only a few cells positive for metaplasia markers, CD44v9 and AQP5²⁴, and consisted of 87% of TROP2⁺ DSCs (Figure 7G). However, metaplastic organoids showed strong expression of AQP5 and CD44v9 with no TROP2 expression (Figure 7F). TROP2⁺/CD133⁺/CD166⁺ DSCs isolated from the dysplastic organoids self-renewed and formed many dysplastic spheres with multilayering of cells and disorganized nuclei (Figure 7I), while no sphere formation was observed from non-DSCs (Figure 7H) as observed in the mouse DSCs and non-DSC cultures (Figure 1E). The human DSCs also differentiated into TROP2⁺ dysplastic cells (Figure 7J) as well as various mucin-secreting cells including PAS⁺ cells and AB⁺ or TFF3⁺ goblet cells (Figure 7K and 7L).

To assess inhibitory effects of the Wnt pathway in human dysplasia, we treated human metaplastic and dysplastic organoids with Salinomycin or Pyrvinium in W^neg/R^neg media for 3 days. Metaplastic organoids did not grow well, but they remained viable and retained spherical structures (Figure 7M). However, growth of dysplastic organoids was remarkably diminished by the two drugs, and especially Pyrvinium treatment significantly increased organoid cell death (Figure 7M and 7N). We additionally treated two mixed-type organoid lines which expressed all three markers, Trop2, CD44v9 and AQP5, with the drugs (Figure S6D). We observed similar morphological and histological changes in some organoids after the treatment, however, only one organoid line treated with Pyrvinium showed a significant increase in organoid death (Figure S6E and S6F). This result suggests that the organoids that responded to Pyrvinium might contain more dysplastic cells. Furthermore, direct targeting of isolated DSCs from dysplastic organoid lines resulted in 100% DSC death after the Pyrvinium treatment. However, Salinomycin-treated DSCs survived and reformed spheres 7 days post drug withdrawal, consistent with the data in mouse DSCs (Figure 7O and 7P). Therefore, these results suggest that targeting a downstream molecule, CK1α, can selectively control the growth and viability of human DSCs and may be a promising strategy to prevent neoplastic transformation from dysplasia to gastric cancer.

DISCUSSION

Genetic and cellular pathogenesis of dysplasia progression or evolution is a major question for a mechanistic understanding of the carcinogenic process from non-cancerous stages. Our results suggested for the first time that Trop2⁺/CD133⁺/CD166⁺ DSCs are de novo stem cells initially developed in pre-cancerous lesions and drive the carcinogenic process in gastric cancer development. Transcriptome profiling revealed highly similar molecular identities of two DSC subpopulations showing that DP- and TP-DSCs are not two different types of stem cells, but their stem cell states and functional activities were distinguishable. While the TP-DSCs were likely dormant stem cells only related to the transitioning cell lineages between metaplasia and dysplasia, DP-DSCs were more associated with manifestation of the proliferative dysplasia phenotype with tumorigenic potential.

Carcinogenic transformation of dysplasia into the multiple tumor types was triggered by clonal expansion and evolution of the single-type of non-cancerous stem cells. Although both mouse and human DSCs showed dynamic differentiation capacity into multiple cell lineages such as dysplastic cells and intestinal goblet cells (Figure 1 and 7), the DSCs did not evolve into cancerous cells in vitro. As previous studies reported that pre-malignant cell lineages were still present during cancer initiation and more cancer-cell related markers are increased during cancer progression in human gastric cancer^{23, 25}, the DP-DSCs promoted an activated niche and evolved into heterogeneous cell lineages, by recruiting all the components necessary for the tumor microenvironment, including a desmoplastic stroma, tumor vascularization and various types of immune cells from the host mice. Thus, the recruited microenvironment may be important for the expansion and evolution of DSCs towards cancer cells. In particular, genetic mutations commonly observed in human gastric cancers were also acquired during the DSC evolution. Sequential genetic mutations, such as in APC, KRAS, RNF43 and CDKN2A, have been well-defined in a multistep process of colon and pancreatic carcinogenesis^{26, 27}. Gastric cancer has a molecular complexity with diverse mutational signatures^{15, 16, 28, 29}, however, it is not yet clear whether any sequential genetic changes can be observed in gastric carcinogenic cascade. The two types of DSC-derived tumors demonstrated distinct mutational heterogeneity and showed several genetic signatures of human gastric cancer molecular subtypes, such as Cdh1 mutation (genomically stable; GS)¹⁵ (Figure 4). Therefore, the acquisition of additional mutations following Kras activation may be a sequence of genetic priming in DSCs as cancer-initiating cells during the evolution process. Wnt pathway is critical for stem cell homeostasis, and mutations in Wnt pathway-related genes can cause cancer initiation³⁰. DSCs could maintain dysplasia homeostasis without external Wnt/Rspo, whereas normal gastric organoids require external Wnt ligands for growth³¹. Inhibition of Wnt ligand secretion did not affect dysplastic organoid growth or survival (Figure S5H), indicating that the β-catenin activation status in DSCs is maintained in a Wnt ligand-independent manner^{32, 33}. Thus, Wnt ligand-independent β-catenin activation through CK1α controls the DSC stemness and mutations in the Wnt pathway downstream genes, such as Rnf43 and Apc, might drive neoplastic transformation of DSCs by dysregulation of CK1α activity.

In our previous study, inhibition of Kras signaling using a MEK inhibitor, Selumetinib, diminished aggressive behaviors of dysplastic organoids, but did not affect the dysplastic cell heterogeneity or survival, which is maintained by DSCs¹⁸. Pyrvinium, a CK1α agonist, is a widely used anti-anthelmintic drug which has anti-cancer and anti-CSC effects^{34, 35}. CK1α is not only a downstream regulator of the Wnt pathway, but also mediates crosstalk between Wnt pathway and other signaling pathways such as hedgehog, autophagy and cell cycle³⁶. Notably, CK1α activation using Pyrvinium could selectively block regeneration and viability of both mouse and human DSCs, while the inhibition of LRP5/6 receptor using Salinomycin did not show a significant effect on organoid survival. Current management guidelines for gastric dysplasia are limited to endoscopic resection or local surgical excision only for small and focal lesions^{37, 38}. Therefore, controlling the DSC activity by CK1α regulation may be an effective approach to reduce the risk of dysplasia progression. Also, identification of other signaling pathways that also regulate CK1α activity in DSCs merits further investigation.

In summary, our findings not only highlight key mechanisms of dysplasia evolution to heterogeneous types of adenocarcinomas through the DSC stemness and plasticity, but also provide important insights that could allow targeted therapeutic intervention by controlling the DSCs and a preventative treatment approach to reduce the risk of early gastric carcinogenesis.

WHAT YOU NEED TO KNOW.

Background and context:

Intestinal-type gastric cancer develops by a multistep process of pre-cancerous metaplasia and dysplasia progression and evolution. In gastric carcinogenesis, dysplasia is especially considered the strongest predictor of adenocarcinoma development.

New findings:

Trop2⁺/CD133⁺/CD166⁺ dysplastic stem cells (DSCs), non-cancerous stem cells, are a key source for maintaining dynamic dysplastic cell lineages and clonal evolution of dysplasia to multiple types of gastric cancer including high-grade adenocarcinoma.

Limitations:

Although this study identified acquired mutation signatures associated with human gastric cancer during the DSC evolution, further studies seeking definitive roles of genetic alterations in DSCs are needed to better define molecular mechanisms of DSC evolution to adenocarcinoma.

Impact:

These findings provide novel insights in a carcinogenic process that leads to gastric cancer development through pre-cancerous stem cell activity and plasticity. Our results also suggest a preventative and therapeutic treatment approach that can reduce the risk of gastric cancer development in patients with metaplasia and/or dysplasia.

Abbreviations

AB

alcian blue

ARM

armadillo repeat

BD

binding domain

BWA

burrows-wheeler alignment

CK1α

casein kinase 1 alpha

CSC

cancer stem cell

CysAC

Cystic adenocarcinoma

DMSO

dimethyl sulfoxide

DN

double negative

DP

double positive

DSC

dysplastic stem cell

EthD-1

ethidium homodimer-1

FACS

fluorescence-activated cell sorting

FBS

fetal bovine serum

FDA

food and drug administration

FFPE

formaldehyde-fixed paraffin-embedded

GEO

gene expression omnibus

GO

gene ontology

GSEA

gene set enrichment analysis

HRP

horseradish-peroxidase

H&E

hematoxylin and eosin

IM

intestinal metaplasia

IRB

institutional review board

KEGG

kyoto encyclopedia of genes and genomes

LRP5/6

lipoprotein receptor-related proteins 5 and 6

MT

masson’s trichrome

NBF

neutral buffered formalin

PAS

periodic acid–schiff

PFA

paraformaldehyde

qPCR

quantitative real-time PCR

RNA-seq

RNA sequencing

Rspo

R-spondin

scRNA-seq

single-cell RNA-sequencing

SNV

single-nucleotide variants

SP

signaling peptide

TCGA

the cancer genome atlas

TM

transmembrane domain

TP

triple positive

t-SNE

t-distributed stochastic neighbor embedding

TubAC

Tubular adenocarcinoma

3D

three-dimensional

Data availability

The transcriptome datasets reported in this paper have been deposited in the NCBI Gene Expression Omnibus (GEO) database under accession code: GSE202191 (bulk RNA-seq) and GSE202192 (scRNA-seq). The whole exome sequencing dataset is available in Sequence Read Archive (SRA) under accession code: PRJNA834800. Further information and requests regarding materials and reagents should be directed to and will be fulfilled by Dr. Eunyoung Choi (eunyoung.choi@vumc.org).