Establishment and characterization of adult human gastric epithelial progenitor‐like cell lines

Abstract Objective Human gastric epithelial stem/progenitor cells are important for stomach homeostasis; however, the in vitro culture system of these cells remains immature. Although three‐dimensional (3D) organoid culture has fundamentally changed the in vitro study of gastrointestinal tract, its use is limited by inaccessible luminal compartment, and difficulties of imaging and manipulation. To overcome these limitations of 3D organoid culture system, we established adult human gastric epithelial progenitor‐like (hGEPL) cell lines using a novel robust monolayer cell culture system. Materials and Methods We established an in vitro gel‐based monolayer culture system for normal human adult gastric epithelium, and compared it with traditional two‐dimensional (2D) and 3D organoid culture systems using transcriptomics, immunofluorescence and cell viability experiments. At the same time, we used single‐cell transcriptomics to compare the differences of the hGEPL cells in conditioned medium (Cond.) and in chemically defined medium (Chem.), the two most common media for organoid culture, in maintaining the stemness and proliferative activity of hGEPL cells. Finally, we explored the role of key niche factors in inducing hGEPL cell differentiation. Results The hGEPL cells were similar to the in vivo gastric epithelial stem/progenitor cells, which could stably proliferate in culture for a long time. Based on the established culture system, we explored signalling pathways that were important for the homeostasis of hGEPL cells. We found that after blocking the WNT signalling pathway or activating the BMP signalling pathway, hGEPL cells could differentiate into mucous surface cells. Conclusion Our culture system of hGEPL cells from adults is robust and easy to operate, and has the transformative potential of personalized and precision medicine, laying a solid foundation for studying the self‐renewal and differentiation potentials of gastric epithelial stem/progenitor cells as well as modelling of related gastric diseases.


| INTRODUCTION
Located between the oesophagus and the duodenum, human stomach is an important organ for storing, mixing and digesting food. 1 Gastric mucosa is composed of thousands of gastric units with a wide variety of well-organized cell types on the surface and in the gastric glands.
The mucous surface cells secrete a large amount of alkaline mucous to protect the gastric mucosa from damage. Gastric glands contain a variety of secretory cells, such as parietal cells, neck mucous cells, chief cells and endocrine cells. 2 Gastric stem cells (GSCs) are adult stem cells existing in the gastric tissue, which can not only maintain self-renewal, but also differentiate into various gastric cell types under appropriate conditions. The stomach epithelium is capable of rapid renewal: the pit region has a rapid turnover rate of roughly 3-5 days, and the renewal rate will be faster if it is stimulated by certain external stimuli. 3 For the gastrointestinal tract, which is vulnerable to external stimuli, the gastrointestinal epithelium may undergo multiple rounds of damage and repair, often in response to excessive inflammations. Inflammatory reaction is a double-edged sword for stem cells. On the one hand, the gastrointestinal tract is the main source of reactive oxygen species (ROS).
When the epithelial layer is damaged, ROS will activate immune cells, cause a series of immune reactions and damage GSCs, thus leading to a series of gastrointestinal diseases such as gastroduodenal ulcer and inflammatory bowel disease. 4 On the other hand, studies have shown that the pro-inflammatory environment induced by hyperbaric oxygen can enhance the differentiation of human mesenchymal stem cells to an osteogenic phenotype, which helps to enhance the differentiation ability of stem cells. 5 Therefore, GSCs are important for maintaining normal physiological functions of the stomach, and the development of methods for in vitro culture of GSCs is critical for a better understanding of gastric homeostasis and stem cell biology.
Three-dimensional (3D) organoid culture, which forms an expanding, self-organizing epithelial structure with various cell types, is the most used in vitro culture system of gastric epithelium. [6][7][8] 3D organoid cultures often have multiple cell types and are heterogeneous, which is a significant feature. 2D culture systems have the characteristics of easier genetic manipulation and imaging. Besides the traditional 2D culture system (direct culture of cells on a dish without coating), many novel 2D culture systems have been established, which have higher scalability and homogeneity. There are two common methods for novel 2D cell culture. One is to culture cells on the surface of the extracellular matrix coating, which generates several different culture systems in the small intestine and colon of mice to support the rapid expansion of their epithelial cells. [9][10][11][12] The other is based on the transwell system, a combination of epithelial monolayer culture on the transwell insert and fibroblast or immune cell culture on the transwell plate. 13 However, these culture systems have some limitations, such as relatively poor long-term maintenance during culture and difficulty of manipulation.
In this study, we attempted to expand human adult gastric epithelial cells on a large scale in vitro based on 2D monolayer culture system. First, we screened and optimized the culture system to make primary human gastric epithelial cells proliferate steadily for a long time in culture. Second, we systematically evaluated the cultured cells based on the morphological and transcriptomic analyses, and found that their gene expression patterns were very similar to those of in vivo gastric stem/progenitor cells. Third, using this system, we further explored the signalling pathways important for the self-renewal and differentiation of these cultured hGEPL cells. In summary, we established a monolayer culture system for human adult gastric epithelial progenitor-like cells, which was robust, easy to operate and optimal for long-term culture. Our study lays a foundation for the understanding of critical biological features of human gastric epithelial stem/progenitor cells. two times. We isolated gastric glands based on the previously published protocol with some modifications. 7 Briefly, blood vessels, adipose tissue, and mucus were carefully removed with forceps and scissors. And the tissue was washed with wash buffer again till the supernatant was clear. We cut the tissue into pieces, collected the pieces into a 50 ml centrifuge tube, and immersed them in 20 ml cold chelating buffer (10 mM ethylene diamine tetraacetic acid (EDTA; AM9261, Ambion) and 0.5 mM dithiothreitol (DTT, 18064014, Invitrogen) in wash buffer). The centrifuge tube then was shocked in an ice bath for 1 h to disintegrate the tissue. After these steps, we placed the tissue pieces in a petri dish and squeezed them to isolate the glands using a glass slide. We resuspended the glands in advanced DMEM/F12 (12630-010, Gibco) with 10% fetal bovine serum (FBS; SE200-ES, VISTECH) and collected them into a tube. We centrifuged the tube for 5 min at 300g and washed the glands for two times. The glands were digested with TrypLE (12604-021, Gibco) at 37 C for 5 min for subsequent culture.
The culture medium 5 (Chem.5) was the same as Chem.4, except that 500 ng/ml R-spondin1 (#120-38, PeproTech) and 100 ng/ml Wnt3A (5036-WN, R&D Systems) were added. The conditioned medium (Cond.) was advanced DMEM/F12 medium with 50% Wnt3A, R-spondin1 and Noggin conditioned medium, 1Â B27 supplement, 1Â N-2 supplement, 100 units/ml penicillin and 100 μg/ml streptomycin, 1Â GlutaMax, 1Â HEPES, 1 mM N-acetylcysteine, 50 ng/mL EGF, 100 μg/ml primocin, 200 ng/ml FGF10, 1 nM gastrin, 2 μM A83-01 and 10 mM nicotinamide. And the Wnt3A, R-spondin1 and Noggin conditioned medium was generated from L-WRN cells (CRL-3276, ATCC) according to the manufacturer's protocol. The chemically defined medium we used for subsequent culture of gastric epithelial cells is Chem.5. Then, we seeded 100,000 cells on Matrigel-coated plates and for each well, cultured in 500 μl culture medium. The medium was changed on the second day after seeded and every 3 days thereafter. And 10 μM Y-27632 (S1049, Selleckchem) was added for the first 3 days. Some points that should be noted: for primary gland cells, Cond. with 10 μM Y-27632 was used for the initial 5 days, followed by Chem.; and for other non-primary cells, both Chem. and Cond. can be used according to study requirements. The above mentioned culture system was named gel monolayer (GM). Another 2D culture was the same as the above method, except that the plates were not coated by Matrigel.

| 2D monolayer culture on Matrigel
Passaging was performed when the cell coverage was greater than 80%. The culture medium was discarded and the monolayer cells were rinsed with DPBS. About 200 μl TrypLE was then added to plates per well, and the plates were incubated at 37 C for 10 min to completely digest cells. After these, we used 500 μl advanced DMEM/F12 to terminate the digestion. Cells were spun down at 500g for 5 min and resuspended for subculture. The steps of freezing and thawing were consistent with the conventional method. The frozen medium was 10% dimethylsulfoxide (DMSO; D2650, Sigma-Aldrich) in FBS.

| 3D organoid culture
For each well of 96 well-plates, 2500-5000 cells were suspended in culture medium and mixed with five times the volume of Matrigel.
The total volume was 8 μl per well. The cell suspension was added to 96-well plates and incubated at 37 C for 15 min to solidify the Matrigel. The cells were cultured in 200 μl medium per well, and other culture conditions were the same as 2D monolayer culture.

| Cell viability assay
We compared the cell viability of 3D, 2D and gel monolayer (GM) cultures. The same number of cells were seeded into 96-well plates. And cell viability was detected using the CellTiter-Glo 3D Reagent (G968B, Promega) according to the manufacturer's instructions on Days 0, 3, 6, 9 and 15. The results were analysed by Graph-Pad Prism 6.

| Immunofluorescence
We placed a glass slide into each well of four-well plates and coated the plates with Matrigel. Then, cells were cultured according to the  Wnt3A, R-spondin1, Noggin, A83-01 or EGF were removed from Chem. to form minus culture medium. As above mentioned, the cells were digested for FACS on Days 3, 6 and 10. 7-AAD À cells were sorted into strip tubes with 8 μl lysis buffer by FACS. The lysis buffer contained 1 U/μl RNase Inhibitor (2313A, Takara), 0.475% Triton X-100, 2.5 μM oligo dT primer and 2.5 mM dNTP (4019, Takara). There were three replicates per sample for subsequent bulk RNA sequencing library construction.

| Single-cell transcriptome sequencing
Single-cell transcriptome sequencing libraries were prepared according to the previously published methods. 15

| Processing of transcriptomic data
For STRT scRNA-seq data set, we utilized UMI-tools to extract barcodes and UMIs from the R2 reads. 16 The template switch oligo and polyA tail sequence were removed from the obtained reads. Besides, reads with low-quality bases were also discarded using seqtk (https:// github.com/lh3/seqtk). Next, we aligned the obtained clean reads to the human GRCh38 genome by STAR, 17 and used featureCounts 18 to calculate the uniquely mapped reads and UMI-tools to quantify the UMIs. Based on the obtained UMI expression matrix, we filtered cells with <1000 detected genes, <10,000 detected transcripts and high mitochondrial gene-expression fractions. Harmony was used to reduce the batch effect (https://github.com/immunogenomics/ harmony), 19 and Seurat (https://satijalab.org/seurat/) was used to identify highly variable genes, dimensionality reduction analysis and clustering. 20 For bulk RNA-seq data set, the trimming and quality control of the raw fastq files were fulfilled by TrimGalore (https://github.com/ FelixKrueger/TrimGalore). Then, the trimmed fastq files were aligned to the human GRCh38 genome using STAR. RSEM was used to quantify the gene/transcript abundances. 21 For scRNA-seq data set, we identified DEGs for each cell cluster with the FindAllMarkers function in Seurat. For bulk RNA-seq data set, we conducted the DEG analysis with DE-seq2. 22 We utilized clus-terProfiler to perform enrichment analysis. 23 All the statistical analyses were performed and related figures were generated in R software.

| Calculation of the cell cycling score
We downloaded two curated cell cycle gene sets representing G1/S phase and G2/M phase, respectively. 24 Then, we used these two gene sets to calculate the cycling score with AUCell, 25 and generated a G1/S score and a G2/M score for each sample. and proliferation ability, which was more suitable for the culture of primary human adult gastric epithelial cells than the other four media.
Based on the Chem.5 medium (referred to as Chem. in the following description), we next explored the effects of different culture modes on the primary human gastric epithelial cells. We tested conventional 2D and 3D organoid culture modes, and found that cells in conventional 2D culture mode would form round colonies, but could be only maintained for about 15 days in culture and then gradually die ( Figure 1B,C). Under traditional 2D culture mode, the support matrix is missing, which is very different from the physiological environment in vivo, and cells lose the ability to proliferate after several divisions and then undergo cellular senescence. 26,27 Our results are also consistent with the previous study that primary human gastric epithelial cells can only be cultured for 2 weeks and cannot be further passaged. 28 For 3D organoid culture mode, although cell proliferation was able to be maintained for a long time in culture, the cells proliferated quite slowly ( Figure 1C). To avoid the drawbacks of conventional 2D and 3D organoid culture modes, we combined them by first coating petri dishes with thin Matrigel, and then cultured the cells on the surface of the gel, termed as GM culture mode. This GM culture mode achieved great culture results: the gastric epithelial cells from human adults first grew as round colonies, then stretched, and proliferated at a much faster rate than in conventional 2D or 3D organoid culture modes ( Figure 1C). Therefore, GM culture mode was easy to operate and the human adult epithelial cells proliferated much faster, which integrated the advantages of the other two culture modes and achieved optimal growth in our hands.
Next, we conducted transcriptomic analysis to make comparisons among 2D, 3D and GM culture modes. As shown in Figure 1D, epithelial (EPCAM), cell cycle (MKI67), stemness-related (FABP5, NME1) and gastric epithelial cell differentiation genes (MUC5AC, MUC6) were expressed at similar levels among these different culture modes. We also found that the results of qPCR were similar to those of the transcriptomics results, and the cells in GM culture mode were less differentiated ( Figure S1A). Meanwhile, immunofluorescence staining also indicated that cells in GM culture mode actively proliferated and expressed well-known marker genes of gastric epithelial cells ( Figure 1E, Figure S1B). We also compared the differentially expressed genes (DEGs) under these three culture modes and conducted a Gene Ontology analysis (GO). We found that the DEGs of GM culture mode were mainly enriched in cell proliferation, amino acid transport and metabolism ( Figure S1C,D).
Previous studies reported two molecules, CHIR99021 (CHIR or C, activator of the WNT pathway) and valproic acid (VPA or V, activator of the NOTCH pathway), which could enhance the self-renewal ability of mouse intestinal stem cells (ISCs), and the colony formation ability was 100 times higher when these two molecules were applied. 29 Tong et al. also found that the addition of CHIR and VPA (CV) supported the growth of mouse and human intestinal epithelial cells on 2D collagen-based monolayer culture system, and promoted the enrichment of Lgr5 + population. 11 Therefore, we separately added C, V or CV to Chem. medium to expand the primary human gastric epithelial cells using the GM culture mode. The addition of either CHIR or VPA increased the expression levels of WNT signalling pathway-related gene LGR5 and NOTCH signalling pathway-related gene RBPJ, respectively ( Figure 1G). However, these two molecules did not enhance the ability of colony formation in the GM culture system. Even worse, cells cultured for 10 days with these two molecules added slowed down their proliferation, and the cell morphology also became abnormally thickened ( Figure 1F,H). Thus, the effect of CHIR and VPA for the primary human adult gastric epithelial cells was quite different compared with their beneficial effects for the self-renewal of ISCs.
Next, we assessed the transcriptomes for different passages of the cells in GM culture system. Previous work has shown that longterm culture of primary intestinal epithelium as monolayer is very difficult due to the rapid loss of stem cells and their accelerated apoptosis, and an important limitation of all monolayer culture systems that need to be solved is the difficulty to passage and propagate the cells, 30 and long-term passage and culture needs to be further optimized. 10,12,31 Through our optimized culture system, the gastric progenitor cells can be continuously passaged for 12 passages and can last for 5 months. As the passage number increased, the cultured cells remained stable for the specific expression signatures of gastric epithelial cells, and the morphologies of the colonies were very similar to those of the early primary culture ( Figure 1I,J). In summary, we established a monolayer culture system that used GM culture modes with Chem. medium to expand primary human adult gastric epithelial cells in vitro.

| Comparison between chemically defined medium and conditioned medium for hGEPL cell lines
Although the Chem. medium combined with GM culture system can expand primary human adult gastric epithelial cells efficiently, it is a chemically defined medium, which relies on very expensive purified growth factors. Thus, we tested a conditioned medium (Cond.) 32 based on a supportive cell line that secreted three growth factors (WNT, R-spondin1 and Noggin), and made comparisons between the chemically defined medium (Chem.) and conditioned medium (Cond.).
We found that the morphologies of the colonies were similar in these two media. Colonies formed about 6 days after being passaged, and the colonies could grow to full on the petri dish after around 15 days after passage (Figure 2A). In addition, the gastric progenitor cells cultured in Cond. medium had a faster growth rate than those in Chem. medium ( Figure 2B).
To analyse the cellular composition of the human gastric epithelial cells cultured in Chem. and Cond., we profiled the transcriptomes were upregulated ( Figure 3B). After the removal of R-spondin1 or WNT3a, the WNT signalling pathway was suppressed, and the expression levels of its target genes LGR5, MMP7 and CTNNB1 were downregulated ( Figure 3B).
To evaluate the effects of these niche factors and small molecules on hGEPL cells' self-renewal and differentiation, we profiled the transcriptomes of hGEPL cells after the removal of the factors, and then calculated their signature scores compared to in vivo human gastric epithelial cell types. Interestingly, we found that the signature scores of mucous surface cells dramatically increased after Noggin or R-spondin1 were removed, and the corresponding marker genes of mucous surface cells such as GKN1, GKN2 and MUC5AC were significantly upregulated ( Figure 3C). After removing WNT3a, the hGEPL cells also showed a tendency to differentiate into mucous surface cells ( Figure 3C). We also explored whether there was the differentiation of other mature gastric epithelial cell types after niche factor removal, but the relevant characteristics were not significant, indicating that the differentiation of hGEPL cells to mature cell types other than mucous surface cells may be controlled by other additional factors ( Figure S3).
These results indicated that these niche factors play important roles in the homeostasis of hGEPL cells, and the removal of them may result in premature differentiation. The block of BMP signalling pathway and activation of WNT signalling pathway play decisive roles in maintaining the self-renewal ability of the hGEPL cells and preventing them to prematurely differentiate into mucous surface cells.

| Identification of key regulatory transcription factors in the differentiation of hGEPL cells
Next, we explored the dynamic expression patterns after withdrawing Noggin or R-spondin1. For both Noggin and R-spondin1, in the first 3 days of withdrawal, the differentiation was only slightly increased, and the number of DEGs was also small compared with the control (cells cultured in Chem. medium). However, from the 6th day to 10th day after Noggin or R-spondin1 removal, the differentiation strongly accelerated, and the number of DEGs also increased sharply ( Figure 4A). After noggin removal, representative marker genes of mucous surface cells (GKN1, GKN2 and TFF1), the receptor genes of BMP pathway (BMPR2), and the target genes of BMP signalling pathway (ID1/2/3/4, SMAD3/5/6/7 and SMURF1) were upregulated.
Similarly, after removal of R-spondin1, markers of surface mucous cells (GKN1/2, TFF1/2 and MUC5AC) and transcription factors of the WNT signalling pathway (PPARD) were upregulated. Interestingly, BMP-related genes (SMAD3 and SMURF1) were also upregulated ( Figure 4A). Since transcription factors (TFs) play crucial roles for cell differentiation, we then focused on the dynamic changes of TFs during the differentiation of hGEPL cells into mucous surface cells.
After the removal of Noggin, the expression levels of ID1, ID3 and ID4 firstly increased, indicating that the BMP signalling pathway was activated, and the hGEPL cells start differentiation, and TFs such as MYC and PPARG were also upregulated ( Figure 4B). Although no TF-related terms were enriched in the first 6 days after the removal of R-spondin1, the TF enriched by R-spondin1 and Noggin removal appeared to coincide greatly at Day 10 ( Figure 4B,C). This result implied that after the removal of either R-spondin1 or Noggin, the hGEPL cells start differentiation, and they differentiate into comparable types of mature cells at the later stage.
After removing Noggin, the cellular state gradually changed, ID1 and ID3 were continuously upregulated ( Figure 4B). After the removal of R-spondin1, the change of the cellular state is initially slow but accelerated later ( Figure 4A,C). During the differentiation of hGEPL cells into mucous surface cells, the removal of Noggin and R-spondin1 shared 45 upregulated transcription factors, suggesting that these TFs may be more universal for the mucous surface cell differentiation ( Figure 4C). We also compared these identified TFs with the TFs important for mucous surface cell differentiation in vivo, and found that several TFs were shared between them, such as MAFG, HNF4A, PPARD, FOSL2, KLF3, ID1 and KLF2 ( Figure 4C,D).

| DISCUSSION
Although the identity of stem cells in the small intestine has been thoroughly studied, 36 culture. 7 Interestingly, we found that removing only a single factor in the WNT pathway, WNT3a or R-spondin1, could induce differentiation of hGEPL cells, and R-spondin1 played a more dominant role than WNT3a ( Figure 3C). For the BMP signalling pathway, after Noggin removal, the expression of BMP target gene ID1 increased as expected ( Figure 3B, Figure 4A,B), indicating the activation of BMP signalling pathway and the shift of cell identity to mucous surface cells, which was consistent with the recently published study. 35 For the NOTCH pathway, after its activation with VPA, the proliferation of hGEPL cells became slow ( Figure 1G,H), which was contrary to the intestinal epithelial cells, reflecting the physiological differences between stomach and intestine. 29 For EGF pathway, we found that once EGF was removed ( Figure 3A), most hGEPL cells could not survive, which also indicated the importance of EGF for the self-renewal of hGEPL cells.
It is very difficult to expand the culture of adult primary cells and enrich for stem/progenitor cell types in vitro. How to better maintain the self-renewal and differentiation potential of hGEPL cells is the focus of our future research direction. The stem cell niche is the local tissue microenvironment that regulates the self-renewal and differentiation of stem cells, and is crucial for the maintenance of normal physiological functions of stem cells. 43 The function of stem cells is controlled by multiple niche signalling pathways. How to better simulate the real niche signalling pathways in vitro has always been a great challenge. Sato et al. 36 induced organoids into undifferentiated state to maintain their potential of amplification. Barker  In addition to the niche and exosomes of stem cells, the selection of novel biomaterials is also an important way to enhance the bioactivity and function of hGEPL cell lines, thus more accurately representing the characteristics of gastric tissue. The culture system we developed currently uses Matrigel to maintain the normal growth of hGEPL cells. But Matrigel is derived from mouse sarcoma cells, which exhibits batch-to-batch variability and it is difficult to control its physical and chemical properties. And Matrigel's murine origin hinders the use of organoids in human clinical transplantation. 52,53 Novel synthetic hydrogels have become potentially a better choice, which can mimic salient elements of the natural extracellular matrix, and can support cell adhesion. 54 Synthetic hydrogel supports the growth and differentiation of human and mouse intestinal stem cells into organoids. 55 Studies have shown that, 56 when Phosphorene coating is applied to hydrogel, the osteogenic ability of human dental pulp stem cells is enhanced. There are also studies combining borophene with hydrogel-based substrates, which is considered to be a highly sensitive method that can be used as a biodetector for clinical translational medicine. 57 There are three key characteristics of our hGEPL cells. First, we overcome the large gap between the 2D monolayer culture system and in vivo gastric epithelial cells. hGEPL cells have similar gene expression patterns with the in vivo gastric stem/progenitor cells, recapitulating key properties of in vivo gastric epithelium. Second, in contrast to 3D organoids, our GM culture system can be easily introduced into lumen contents without labour-intensive microinjection and are suitable for genetic manipulation and high-throughput imaging analysis. Finally, hGEPL cells have similar gene expression patterns of gastric characteristic genes with those of 3D organoids, and can be easily expanded by suspension and passage like classical cell lines, which is a great improvement over previous works. 10,12,31,58 We have established a robust monolayer cell culture system for adult gastric epithelium. The hGEPL cell lines have biological features comparable to gastric epithelial stem/progenitor cells in vivo and could proliferate steadily over a long period of time. Activation of the WNT signalling pathway and repression of the BMP signalling pathway are critical for the maintenance of the self-renewal ability of hGEPL cells. Our novel culture system lays a solid foundation for the study of the physiological function and molecular mechanism of gastric epithelium, and provides a new selection model for the in vitro drug screening for gastric diseases. Meanwhile, our system has certain limitations.
Our culture system is relatively expensive, which limits the application of large-scale experiments. Besides, cells cultured in our system are more homogeneous than 3D organoid system, so it is difficult to study the interactions between different epithelial cell types. However, our