Highlights of how single-cell analyses are illuminating differentiation and disease in the gastric corpus

Abstract

Single-cell RNA-sequencing (scRNA-seq) has emerged as a powerful technique to identify novel cell markers, developmental trajectories, and transcriptional changes during cell differentiation and disease onset and progression. In this review, we highlight recent scRNA-seq studies of the gastric corpus in both human and murine systems that have provided insight into gastric organogenesis, identified novel markers for the various gastric lineages during development and in adults, and revealed transcriptional changes during regeneration and tumorigenesis. Overall, by elucidating transcriptional states and fluctuations at the cellular level in healthy and disease contexts, scRNA-seq may lead to better, more personalized clinical treatments for disease progression.

INTRODUCTION

The stomach is a unique organ, lined with a single layer of epithelium that is constantly self-renewed and functions primarily in digestion and killing of ingested pathogens through the release of acid, digestion enzymes, and hormones. In terms of the genetic and cellular regulation of gastric structure and function, the two most studied species are mice and humans. The stomachs of mice and humans differ in that mice have a nonglandular, proximal forestomach. However, the arrangement of the glandular stomach into two anatomical zones, proximal corpus and distal antrum, is common to both species. Each of the glandular regions is lined by a single layer of epithelium that invaginates into repeating gastric units (Fig. 1A). The corpus primarily functions in pathogen killing and early food digestion, containing epithelial cells that secrete acid and digestive enzymes. Comparatively, the antral glands release hormones that regulate acid secretion and secrete mucus that helps protect the stomach lining from ingested agents and the stomach’s own acid and digestive enzymes. Major cell lineages within the corpus glands include the mucous pit/foveolar cells, isthmal stem cells, acid-secreting parietal cells, mucous neck cells, digestive enzyme-secreting chief (zymogenic) cells, and hormone-producing enteroendocrine cells. The glands of the antrum comprise mainly pit cells, endocrine cells (including gastrin-, serotinin-, and somatostatin-producing cells), and gland base mucous cells (2–6).

Figure 1.

scRNA-seq applications in the stomach. A: schematic detailing the compartments of the human stomach, the organization of the gastric units that line the gastric epithelium, and the dissociation to the single-cell level for single-cell RNA sequencing (scRNA-seq). The gastric epithelium comprises many cell types including pit cells (light purple), isthmal stem cells (burgundy), parietal cells (yellow), mucous neck cells (green), and chief cells (blue). B: t-SNE plot of scRNA-seq data from murine gastric epithelium from both corpus and antrum (with permission from Busslinger et al., 1). Cells are color coded by cell type as assigned based on gene expression profiles. Cluster identities are listed to the right. In situ hybridization can be used to validate scRNA-seq candidates and confirm expression in tissue. Demonstrative RNAscope for Esrrg in gastric tissue is shown (unpublished data). qPCR can also be used to validate scRNA-seq data. C: demonstrative pseudotime trajectory plot of isthmal stem cells/progenitors, immature, and mature parietal cells in injured murine gastric tissue (unpublished data). Trajectory analysis is a secondary analysis that uses scRNA data to infer cell differentiation and developmental trajectories along pseudotime as a progenitor differentiates into a mature cell type. Commonly used trajectory analysis methods are listed. The schematic on the right illustrates the major epithelial lineages that arise from a gastric progenitor at homeostasis. Note, that in adult gastric tissue, chief cells are self-maintained. scRNA-seq, single-cell RNA sequencing.

Although the physiological function and histological structure of the stomach have been well-characterized, the behavior of the stem cell and the molecular events underlying differentiation in the gastric epithelium are arguably the least understood of all organs that proliferate throughout life. However, we are beginning to fill this gap in knowledge as single-cell RNA sequencing (scRNA-seq) techniques have emerged. scRNA-seq permits transcriptional analysis of individual cells and subsequent analyses including unbiased clustering of cells based on similarities and differences in gene expression profiles. These analyses are being used to predict relationships between cell populations during development, in mature healthy tissues, or in disease contexts (7–12). In past several years, scRNA-seq studies of the stomach have uncovered many new markers for specific epithelial lineages (Table 1). In addition, beyond simply specifying genes enriched in each cell type, secondary bioinformatic analyses of scRNA-seq data that provide estimates of dynamic cell differentiation behavior (so-called unbiased clustering and pseudotime trajectory analysis) have begun to provide insights into critical pathways in gastric tissue development and differentiation as well as unravel how normal cells go awry in diseases like metaplasia and cancer (Fig. 1, B and C). This review will focus on recent findings made in gastric epithelium development, maintenance, regeneration from injury, and during cancer progression, with an emphasis on insights made from scRNA-seq analyses.

Table 1.

Markers for gastric epithelial lineages used in scRNA-seq data

Cell Type	Marker	System	References
Pit	Aqp3	Mouse	(13)
	Dpcr1	Mouse, Human	(1, 14)
	Gkn1	Mouse, Human	(1, 14)
	Gkn2	Mouse, Human	(1, 13, 14)
	Krt19	Mouse, Human	(1, 14)
	Muc5ac	Mouse, Human	(1, 13–16)
	Tff1	Mouse, Human	(1, 14, 15)
Mucous Neck	Muc6	Mouse, Human	(1, 13–16)
	Tff2	Mouse, Human	(1, 13)
Parietal	Atp4a	Mouse, Human	(1, 14–16)
	Atp4b	Mouse, Human	(1, 13–16)
	Clic6	Mouse, Human	(1, 14)
	Ckb	Mouse, Human	(1, 14)
	Gif	Human	(1, 14, 16)
Chief	Chia1	Mouse, Human	(1)
	Gif	Mouse	(1, 13, 15, 17)
	Pga4	Human	(14, 16, 18)
	Pga5	Human	(1, 14, 16, 18)
	Pgc	Mouse, Human	(1, 13–15)
Isthmal Stem Cell	Iqgap3	Mouse	(13, 14, 19)
	Fabp5	Human	(14)
	Lefty1	Mouse, Human	(20)
	Nme1	Human	(14)
	Stmn1	Mouse, Human	(1, 13, 14)
Metaplastic Cell	Aqp5	Mouse	(16, 21)
	CD44v	Mouse, Human	(18, 21)
	Dmbt1	Mouse	(15, 21)
	Gkn3	Mouse, Human	(15, 22)
	Hoxb13	Human	(20)
	Sox9	Mouse, Human	(16, 18, 21)
	Wfdc2	Mouse	(21)

scRNA-seq, single-cell RNA sequencing.

EMBRYONIC AND DEVELOPING STOMACH

The morphological changes that occur as the gastrointestinal (GI) tract and stomach develop have largely been described at the anatomical level with later molecular studies elucidating that combinatorial expression of transcription factors (TFs) Sox2, Gata4, Pdx1, and Cdx2 defined the regions within the developing stomach and intestine (23–27). More recently in 2018, a large-scale study using a combination of high-performance liquid chromatography-mass spectrometry (HPLC-MS) for global protein and RNA-seq for global transcript analysis was used to define the global network of transcriptional changes occurring in the stomach as the key TFs discussed above worked to specify the gastric epithelium. The authors identified large-scale molecular changes that occur in a developing gastric epithelium through to adulthood (E12.5 to postnatal week 8) while linking changes in transcript and protein abundance to concomitant gross anatomical and functional changes. Specifically, the authors were able to discern three distinct phases in stomach development including an initial phase associated with cell division, a second phase during which the stomach expands the most in size, and a postweaning stage where metabolic functions are upregulated as the diet changes (28). Such bulk analyses of global proteome and transcriptome changes tend to be quite sensitive and can identify even relatively rare changes in gene expression; however, bulk analysis can be used to identify changes in specific cell populations only when painstakingly validated by low-throughput in situ techniques like immunohistochemistry.

On the other hand, scRNA-seq provides the resolution to identify genome-wide changes in gene expression in individual cells. In the stomach, for example, scRNA-seq was used to characterize cell-type-specific expression patterns in embryonic murine gastric tissue as early as E9.5. Regional heterogeneity in both the developing epithelium and mesenchyme was identified at that timepoint (29). Single-cell assay for transposase accessible chromatin with sequencing (scATAC-seq), a technique to assess chromatin accessibility, of E9.5 endoderm-derived tissue revealed that open regions of chromatin correlated with expression of region-specific TFs that are observed by E13.5. Specifically, chromatin binding sites for Sox2 and Cdx2 were open in the foregut and hindgut, respectively, corresponding with their region-specific expression later in development (30). This indicates an interaction between the chromatin landscape and TF function in regulating gastrointestinal regionalization and organ patterning. A different approach that combines near-single-cell resolution with histological information is spatial transcriptomics, which involves RNA-seq on intact tissue sections. Spatial transcriptomics at E13.5 and E15.5 confirmed that stomach versus intestinal boundaries were established by restricted expression of Sox2 in the gastric epithelium and Cdx2 in the intestinal epithelium (31). The role of Cdx2 in specifying gastric versus intestinal epithelium was also previously observed in single-cell analysis of human intestinal organoids (HIOs) as they were induced to undergo endoderm specification. Specifically, when CDX2 was knocked out in HIOs, more stomach markers were expressed (SOX2 and MUC5AC); the epithelial cells in the HIOs lost their intestinal identity, clustering with stomach epithelium cells based on unbiased clustering analysis (32).

In addition to providing insight into the patterning of the gut tube and the regionalization of the stomach, scRNA-seq has also revealed how progenitor cells emerge and differentiate during development. Unsupervised clustering analysis of scRNA-seq data in murine gastric tissue at E13.5 revealed three distinct epithelial clusters representing the forestomach (exhibiting enriched expression of Krt5, Krt15, Shh, and Trp63), the corpus (Cym, Fkbp11, and Cxcl15), and the antrum (Pdx1, Nkx6-2, and Nkx6-3). Within the corpus population, cells expressing transcripts characteristic of terminally differentiated gastric epithelial lineages were already detected by this developmental timepoint: Atp4b (parietal cells), Pgc (chief cells), and Tff2 (neck cells). In both corpus and antrum populations, Muc5ac+ (pit cells) cells were present. Distinct mesenchymal clusters represented fibroblasts, myofibroblasts, and mesothelial cells. In addition, populations of enteric neural crest cells, endothelial cells, and hematopoietic lineage cells (lymphomyeloid cells and erythrocytes) were also observed (33). These findings confirm that lineage specification within the glands commences as early as E13.5. Furthermore, the authors used scRNA-seq, in situ hybridization, and organoid studies to further explore signaling pathways (including Wnt, BMP, and Notch) known to influence cytodifferentiation in the gastric unit (23, 34). E13.5 gastric epithelial cells expressed Wnt5a, Wnt5b, Bmp1, Bmp7, and Bmp1ra. Inhibition of WNT or BMP signaling in embryos led to decreased expression of chief cell and parietal cell markers, respectively, demonstrating the necessity of each pathway for distinct lineage specification. Interestingly, Notch inhibition studies in embryos revealed that it functioned differently in the two compartments. In the corpus, Notch signaling blocks the differentiation of progenitor cells to pit and endocrine lineages and is necessary to maintain the corpus identity. In the antrum, Notch signaling inhibits neck and chief cell differentiation (33). Overall, these studies confirmed that pathways previously known to regulate gastric epithelial cell specification are conserved even in organogenesis.

SINGLE-CELL RESOLUTION IN THE ADULT STOMACH

Similar to how scRNA-seq has been used to examine patterning and specification of the developing gastric epithelium, it is proving equally valuable in exploring the differentiation of the various gastric epithelial lineages in adult tissue. As an organ with rapid cell turnover, the stomach epithelium harbors stem cells responsible for maintaining the tissue at homeostasis. Although previous studies based on electron microscopy and radiolabeling focused on how, at homeostatic conditions in the murine corpus, stem cells reside in the gland isthmus, more recent work has demonstrated that the base of the glands also have their own, distinct turnover and cell division patterns (5, 35, 36). The current, more nuanced understanding of the stomach at homeostasis is that the isthmus stem cells replenish many of the epithelial lineages during homeostasis (pit, neck, and parietal cells), whereas the chief cells in the base largely maintain themselves at homeostasis via self-duplication. Chief cells can also be activated upon injury to function as reserve stem cells (37–39). scRNA-seq has been providing insight into these patterns including elucidating bona fide markers specific to isthmal stem cells. Although multiple markers for the isthmal stem cell population had been proposed based on various lineage-tracing experiments, scRNA-seq is revealing that at the transcriptional level, most of these prior, putative stem cell markers may not actually be specific to the isthmal population (13). For example, two genes have emerged as highly enriched in proliferative isthmal undifferentiated cells: Stmn1 and Iqgap3 (19, 37). Furthermore, such scRNA-seq results have been combined with secondary analysis such as trajectory inference analysis, which enables prediction of differentiation states in pseudotime among cell types based on gene expression profiles or transcript splicing (40) (Fig. 1C). scRNA-seq and trajectory analysis of Stmn1-eGFP+ cells (from Stmn1-CreERT2-eGFP mice) demonstrated that the Stmn1^high cluster differentiates into either Stmn1^low/Muc5ac^high (pre-pit cell) and Stmn1^low/Muc6^high (pre-mucous neck cell) clusters, indicating how pit and neck progenitors might arise from the stem cell, respectively (37). Although Iqgap3 and Stmn1 mark proliferative isthmal stem cells, scRNA-seq also has been used to identify a new proposed quiescent isthmal stem cell marker, Lefty1. Histological studies in healthy and injured tissue confirmed Lefty1 expression specifically in the isthmal stem cell niche of the gastric corpus. However, lineage tracing with Lefty1-CreERT2:Rosa26-tdTomato mice showed that the isthmal Lefty1+ population has limited potential as following parietal cell injury, the trace predominantly marked pit cells (20).

Within the adult human gastric corpus, NME1 and FABP5 have been identified as potential progenitor cell markers based on single-cell spatial transcriptomic analysis and validation with in situ hybridization. Interestingly, other progenitor/stem cell markers proposed in mice were not specifically expressed in the same population in human tissue. Pseudotime analysis of healthy murine corpus tissue suggested that NF-κB signaling maintains isthmus progenitor cells in an undifferentiated state while EGFR-ERK signaling promotes pit lineage differentiation consistent with earlier studies (13, 41). Further analysis with single-cell ATAC-seq has allowed the examination of epigenetic regulatory mechanisms that might provide insight into how lineage specification occurs from gastric stem cells. Cell-type-specific TFs have been identified including MAFG, HNF4A, and PPARD in pit cells; KLF9, ZKSCAN1, MECOM, and BPTF in mucous neck cells; and AR in chief cells and SERRB and ESRRG in parietal cells (14). Identifying isthmal stem cell markers and lineage-specific transcriptional regulators is essential for understanding how the gastric epithelium is maintained at homeostasis.scRNA-seq is also being used to explore transcriptional differences in cell types across species. For decades, it has been known that at homeostasis, gastric intrinsic factor (GIF) is expressed in chief cells in mice but in parietal cells in humans (42), but what other cell lineage expression differences are there? Busslinger et al. (1) showed that in mice, there were two distinct chief cell populations in their scRNA-seq analysis based on unbiased clustering analysis: chitinase (Chia1) positive and negative. This distinction was not found in the human data. Only in humans was there a clear separation of a prezymogenic cell population transcriptionally distinct from the mature chief population. Although most genes have similar expression patterns in the human and mouse, dimension reduction and t-SNE plot analysis revealed a few human-specific and mouse-specific markers for each of the four most abundant gastric epithelial lineages (pit, parietal, mucous neck, and chief). Fhl1, Apoa1, Gkn3, and Clps were expressed in the pit, parietal, neck, and chief cells of murine stomachs, respectively, whereas expression of HSPB1, CCKBR, FMOD, and PGA5 was human-specific in those same lineages (1). Discovering these differences is essential for understanding the limitations of mice as a model for human stomachs and for developing tools to model the human gastric epithelium more accurately.

NORMAL REGENERATION AND ABNORMAL REGENERATION LEADING TO METAPLASIA IN THE GASTRIC CORPUS

Single-cell analyses have also begun to unveil molecular features of injury response in the gastric mucosa. Although the adult stomach is maintained mostly by the stem cells residing within the isthmus at homeostasis, in the context of gastric injury, the chief cell lineage at the base of the corpus gastric glands can also be recruited to regenerate new epithelial cells (35, 43–46). After certain injuries, the chief cells undergo an evolutionarily conserved reprogramming process known as paligenosis. Paligenosis is the process mature cells use to reprogram to a more stem cell-like state to reenter the cell cycle and regenerate damaged tissue in response to injury. Paligenosis is conserved across tissues and species and involves an initial downscaling of mature cell architecture followed by increased expression of progenitor or wound-healing genes and, subsequently, a licensed entry into the cell cycle. Thus, paligenosis describes the cellular changes that occur as a mature cell becomes proliferative, rather than the end point or goal of the reprogramming compare versus “transdifferentiation” (which describes when a mature cell becomes another mature cell of different type) or “dedifferentiation” (which describes when a mature cell reverts to an earlier developmental stage) (46–48). Specifically in the stomach, paligenotic chief cells become spasmolytic polypeptide-expressing metaplasia (SPEM) cells. SPEM cells at the base of gastric glands are characterized by their ectopic expression of spasmolytic polypeptide (TFF2) and are seen within the context of pyloric metaplasia, a lesion involving the whole gastric unit characterized by changes in other cell types. Perhaps the most dramatic alteration in pyloric metaplasia is the loss of parietal cells (46, 49, 50). Persistent, chronic pyloric metaplasia and SPEM may be the precursor to a further change in cell differentiation patterns known as intestinal metaplasia (IM). Pyloric and intestinal metaplasia can eventually increase the risk for gastric adenocarcinoma especially when the chronic injury is caused by infection with Helicobacter pylori. Based on histoanatomical analysis such as microscopy and lineage tracing, the chief cells within the stomach have been shown by numerous investigators to be the principal source of SPEM cells in the acute setting (35, 44, 45, 51–53).

Recently, the single-cell-based approach has been applied to characterize SPEM cells as well as their cellular origins. Now secondary analysis using single-cell transcriptomics such as trajectory analysis enables predictions about the path cells take to differentiate from other cells. Such pseudotime analyses can provide more insight into, for example, the transcriptional signature and cellular origin of SPEM cells. In mice, pseudotime analysis has predicted that SPEM cells and the metaplastic mucous neck cells that arise in pyloric metaplasia as parietal cells die are derived from both homeostatic chief and neck cells via an intermediate pre-SPEM cell stage. This might be because the SPEM cells formed from chief cells acquire a transcriptome more like mucous neck cells. Indeed, pseudotime analysis of nonmalignant human gastric epithelium inferred a trajectory axis where mature chief cells transdifferentiate into mucous neck cells and then become SPEM cells, demonstrating the plasticity of mature chief cells and indicating a potential pathway for SPEM progression at the single-cell level in human tissue (18). Implicit to the above analyses is the fact that scRNA-seq analysis indicates that mucous neck cells during pyloric metaplasia not only become hyperplastic (increased in number over homeostasis) but also themselves change gene expression and phenotype (i.e., undergo metaplasia). Because such mucous neck cell changes occur as chief cells become SPEM cells, these mucous neck cells have been called “neck-SPEM” (15, 22). Thus, as these metaplastic cells may also eventually contribute to gastric adenocarcinoma, it is important to not forget this often-neglected population. However, it has been argued that it is probably best to refer only to the cells at the base of gastric units with coexpression of mucous neck and chief cells as SPEM cells and refer to the whole pattern of changes as pyloric metaplasia (49).

Identifying novel molecular markers for different cell states in injury and disease states is another valuable application of single-cell-based methods. With the use of scRNA-seq, p57 was identified as a molecular switch regulating the ability of chief cells to function as reserve stem cells following gastric injury. Its loss was associated with increased proliferation in chief cells after injury (17). Biomarkers for early gastric cancer have also been proposed based on scRNA-seq, trajectory analysis, and validation with histology of premalignant and early gastric cancer lesions including SLC11A2 and KLK7 (16). A biomarker for the precancerous gastric lesion, SPEM, has also been found based on the scRNA-seq method. Gastrokine-3 (Gkn3) is a protein expressed in healthy mouse stomach antrum but absent from, or at low levels, in the corpus. The authors also noted immunostaining using a polyclonal anti-GKN3 antibody in human SPEM, though the significance of this is unclear, as it is designated as a pseudogene in the National Center for Biotechnology Information (NCBI) database due to key point mutations between the human GKN3 and that of other animals (54). Nevertheless, in the chronic inflammatory setting, Gkn3 is found in the mouse stomach corpus, specifically in the SPEM cells and a small subset of foveolar cells. Gkn3 was reported to be more specific in distinguishing the SPEM cell cluster in scRNA-seq versus other transcripts including those most commonly used to identify SPEM histologically. In addition, unlike Gif and Tff2, which are expressed by other cell types in healthy stomachs, Gkn3 is absent from those cell types. This implicates Gkn3 as a more specific marker of SPEM cells for identifying cell types in single-cell transcriptomics datasets. Gkn3 transcript and GKN3 protein have both been confirmed to overlap with the SPEM in mouse samples in this study (22).

INSIGHTS INTO THE MOLECULAR ORIGIN, PERPETUATION, AND CHARACTERIZATION OF GASTRIC CANCER

As single-cell transcriptomics has been used in understanding SPEM better in injured stomachs, they have also been used to begin to answer questions about the origin and progression of gastric cancer (GC), which is the fourth leading cause of cancer death worldwide (55). Adenocarcinoma, as one of the most common types of GC, comprises ∼90% of them (56). Two major questions about gastric adenocarcinoma are 1) the contribution of intra- and/or intertumor heterogeneity of gastric cancers to cancer outcome and 2) the mechanism of carcinogenesis and possible cell types contributing to it. One of the major issues caused by the highly heterogeneous nature of tumors is their resistance to treatment. By single-cell RNA-sequencing, neoadjuvant chemotherapy (NACT) was shown to decrease the T-cell population, possibly disable T-cell function, and increase protumor pathways such as EMT and angiogenesis (57). Thus, differences in the tumor microenvironment could change how we approach tumor biology and provide a pathway for individualized medicine. The approach highlights the benefits of using single-cell transcriptomics over bulk RNA transcriptomics in the sense of tailoring personalized treatment for patients possibly due to the highly heterogeneous nature of gastric tumors. In addition, different histology subtypes of gastric cancers based on Lauren’s classification including intestinal-type and diffuse-type, and molecular subtypes such as Epstein-Barr virus (EBV), microsatellite instability (MSI), genomically stable (GS), and chromosomal instability (CIN) make characterizing and comparing the heterogeneity of GC among samples by bulk RNA-seq limited (58). Indeed, scRNA-sequencing of human GC samples at various stages reveals that during tumorigenesis, there tends to be more heterogeneity of fibroblasts, immune, and endothelial cells within the tumor microenvironment (59, 60). With scRNA-sequencing of patients’ stomach samples across clinical stages and histological subtypes, differences in immune cell composition were found among different gastric tumor types (61–63). Plasma cells, as an example, are found enriched in diffuse-type tumors when compared with intestinal-type gastric tumors (62). Furthermore, integration of scRNA-seq with other single-cell technologies like scATAC-seq and scDNA methyl-seq could provide a more robust profiling of intratumoral heterogeneity. scTrio-seq3, which allows for RNA-seq, DNA-methyl-seq, copy number variation, and chromatin accessibility analysis at the single-cell level, confirmed that compared with healthy human gastric tissue, gastric cancer tissue upregulated normal colon gene signatures and downregulated normal stomach ones. Within the tumors of each individual patient, various differentiation states and DNA methylation heterogeneity were observed, indicating that individual variation in epigenetic factors may also influence how the transcriptional alterations that arise in gastric cancer cells occur (64).

In addition, cellular composition or tumor microenvironment differences such as distinct cancer-associated fibroblasts (CAFs) with INHBA–FAP-high expression have been discovered between different stages of cancer, which correlates with poor clinical prognosis and could be used as a predictor (62). Furthermore, scRNA seq has revealed that more differentiated gastric carcinomas tend to have more tumor-infiltrating CD8+ T cells compared with poorly differentiated gastric cancer (65). Also, studies have identified a subpopulation of cancer cells that undergoes an epithelial-myofibroblast transition (EMyoT) that is thought to differ from the canonical epithelial-mesenchymal transition (EMT). EMyoT cancer was found to correlate with a worse prognosis than EMT (61). Even beyond the stomach in gastric cancer peritoneal metastasis (GCPM), scRNA-seq and subsequent trajectory analysis of ascites from patients with GCPM have revealed that following treatment, highly plastic gastric cancer cells go through EMT and eventually express a more proliferative signature. Overall, this suggests that as highly plastic cells become proliferative, gastric cancer cells may undergo an adapted paligenosis program to escape treatment and promote metastasis (66).

Subsequently, patient-derived organoids (PDOs) as a method for disease modeling, have been shown with single-cell RNA sequencing to preserve some degree of heterogeneity from the primary tumors; however, the relative proportion of cell types might differ from that seen in tissue (62). Further investigation of how well scRNA-seq accurately captures cell populations and how the relative cell proportions compare to intact gastric tissue is necessary. Nevertheless, PDOs may have some important utility for understanding metaplasia and tumorigenesis. For example, scRNA-seq of samples from patients with GC revealed that distinct fibroblast subpopulations define each stage of gastric cancer progression from normal to dysplastic tissue. Specifically, CEACAM5 was enriched in dysplastic tissue while PDGFRA was enriched specifically in dysplastic fibroblasts. Coculturing of PDOs from metaplastic tissue directly with cancer-derived fibroblasts or with conditioned media from cancer-derived fibroblasts promoted progression to dysplasia and expression of CEACAM5, indicating that these fibroblasts may secrete factors that promote disease progression (67). Moreover, as previously discussed, based on chronology and histoanatomical evidence, gastric cancer seems to arise via a sequence of tissue and cell differentiation changes with chronic gastritis causing pyloric metaplasia with SPEM and then intestinal metaplasia followed by dysplasia and cancer (called the Correa cascade). It is impossible to see these transitions over months and years in living human stomachs (68). It is also unclear whether so-called diffuse gastric cancer arises via a similar cascade. By pooling samples from patients with different types of gastric cancers along with the paired normal and metaplastic tissues, the authors of one publication compared the development of intestinal gastric cancer and diffuse gastric cancer computationally by single-cell trajectory analysis. Intestinal-type gastric cancer follows a Correa-like cascade along the pseudotime trajectory, starting as nonmalignant cells at the earliest pseudotime and progressing to intestinal metaplasia in the middle before finally becoming intestinal gastric cancer. Diffuse gastric cancer, however, might follow a de novo carcinogenesis mechanism based on the trajectory (61).

Interestingly, scRNA-seq has also proven valuable for understanding how H. pylori (HP), a carcinogenic bacterium that colonizes nearly 50% of the world’s population, changes the epithelium at a single-cell level (69). For decades, the histopathological changes that occur during HP infection including progression from atrophic gastritis (parietal cell loss) to intestinal metaplasia, and to gastric cancer has been well detailed (68). In murine stomachs, scRNA-seq comparing Kras-mutation-driven intestinal metaplasia (a chronic inflammation injury model in which Kras is constitutively active in chief cells) with or without HP infection informed about how the interaction of genetic factors and microbiome contribute to tumorigenesis. HP-infected Kras mutants exhibited pit cell hyperplasia and increased expression of intestinal mucin Muc4 compared with non-HP-infected mutants. Further scRNA-seq analysis of human samples corroborated the observation as MUC4 was highest in pit cells from patients with cancer or metaplasia and MUC4 was shown to be positively associated with MKI67 expression (proliferation) (70). Although additional investigation into whether Muc4 can act as a driver for metaplasia progression is needed, this provides some hints into how genetic factors may play a role in HP-associated gastric cancer. scRNA-seq of HP-infected two-dimensional (2-D) and three-dimensional (3-D) PDOs suggests that HP preferentially binds large, mature pit cells, an interaction that is driven by chemotaxis of HP to host-produced urea (71). Pit cells produce the highest concentration of urea compared with other lineages but illicit the lowest immune response (71, 72). Accordingly, the authors suggested that this increases the likelihood that HP would escape an immune attack (71). Not surprisingly within human tissue, scRNA-seq has shown that HP infection alters immune cell composition in the stomach as seen in mouse. However, contrary to mouse studies where type 2 innate lymphoid cells (ILC2s) predominate, they were absent in human tissue. Instead, IL3Cs were the foremost population, demonstrating additional differences in the murine versus human system (73).

Finally, single-cell analysis has also been opening new lines of investigation across tissue silos. For example, single-cell profiling has shown that Barrett’s esophagus, which involves intestinal metaplasia of the esophagus, is molecularly similar to intestinal metaplasia in the stomach (74). This is consistent with recent histoanatomical findings (75). The progression of the precursor lesions to adenocarcinoma is also similar, specifically though of the intestinal or chromosomal instability subtypes of cancer, suggesting that both the esophageal adenocarcinoma and the bulk of gastric adenocarcinoma arising from metaplasia may be essentially the same entity (76). Such cross-organ comparisons in characterizing carcinogenesis will be limited by bulk RNA-seq but are more accessible only with the aid of scRNA-seq and the secondary analysis associated with it (21, 77).

Limitations of scRNA-Seq in Gastric Tissue

Although scRNA-seq has its benefits, there are still technical limitations to overcome. Although filtering parameters can be used to remove poor-quality cells before downstream analysis, artifacts from the initial single-cell dissociation can persist. Specifically for the stomach, filtering algorithms that use a fraction of mitochondrial genes as a threshold to determine signals from viable cells are problematic because parietal cells harbor enormous numbers of mitochondria, resulting in investigators discarding viable parietal cell populations. Also, for epithelial cells in general, dissociation and removal from the surrounding tissues can cause stress and alter gene expression leading to observations that might not exist in tissue (78, 79). This is particularly an issue during the study of paligenosis of chief cells to SPEM, as the isolation of chief cells for scRNA-seq causes a fraction of those cells to show early paligenosis changes not representative of normal cells in their undisturbed niche (unpublished observations). Thus, candidate biomarkers, novel populations, or pathways identified using scRNA-seq should be corroborated and validated using techniques like in situ hybridization (RNAscope) and general histology (immunohistochemistry and immunofluorescence). To mitigate dissociation-induced artifacts, several alternate techniques are becoming more accessible but have their own drawbacks. Single-nuclei RNA sequencing is emerging as an alternate transcriptional analysis technique where tissue can be snap frozen and RNAs barcoded before the isolation. Unfortunately, fewer mRNA transcripts per cell are captured (80). Spatial transcriptomics allows for RNA sequencing of cells in tissue. Currently, though, commonly used spatial platforms are limited in that the approaches that can assay the whole transcriptome cannot resolve to single-cell level, and those that have closer to single-cell resolution have a much more limited cohort of transcripts that can be assayed.

Finally, while hailed as a panacea for identifying rare or elusive cell populations, scRNA-seq still has limited sensitivity to detect lowly expressed transcripts (81, 82). Unfortunately, this means that transcripts for elements with cell fate regulatory functions like long-noncoding RNAs, (lncRNAs), transcription factors (TFs), and nuclear receptors, which are often only scantly expressed, are underrepresented in the scRNA-seq generated gene lists (83, 84). Consequently, this can be detrimental to correctly identifying cell types or unveiling biologically relevant relationships between different cell states. However, there are emerging methods to ameliorate this problem. SCENIC constructs gene regulator networks with scRNA seq data to infer TF activity and scCapture-Seq was adapted from a tool originally used to capture TFs for bulk RNA seq (83, 85). We can hope that the implementation of similar methods further improves the utility of scRNA-seq to identify TFs and pathways that govern cell differentiation and lineage maintenance of the gastric epithelium where only a handful of TFs have been implicated.

CONCLUSIONS

Transcriptional analysis at the single-cell level is becoming a powerful tool, especially as researchers move to create atlases for every organ during development and in adult healthy and diseased tissues. Consortiums like The Human Cell Atlas (HCA), a collaborative project to compile single-cell data across multiple organs, are readily emerging. Unfortunately, not all organs have been represented in these datasets and platforms as the stomach is currently not included (86). Only recently have large cell atlases that feature adult human gastric tissue across various healthy and/or disease states have been published (1, 20). Nevertheless, scRNA-sequencing of gastric tissue by individual laboratories has already proven valuable, with payoffs ranging from detecting critical pathways during development to identifying transcriptionally distinct cell populations in adult tissue and during the progression of gastric diseases (87, 88). Various gene lists generated from these studies can also provide additional markers to complement and extend and enhance purely histological studies. Specifically, pseudotime trajectory analysis has begun to elucidate the cells of origin of metaplasia in the stomach in ways not possible by looking at fixed human tissue. The combination of PDO and scRNA-seq techniques in particular is proving powerful in dissecting temporal molecular and phenotypic progressions as cells exhibit lineage plasticity in the setting of metaplasia and provide targetable candidates to prevent or reverse the progression of gastric cancer.

Nevertheless, these emerging technologies that allow interrogation of molecular and transcriptional states with single-cell resolution have greatly supplemented our understanding of the development and maintenance of healthy gastric epithelium, and the technology seems to be evolving even as we are writing these words. As scRNA-seq platforms and associated downstream analyses like those for cell-cell communication, trajectory inference, and pathway analysis evolve, we can continue to improve our ability to answer questions about how the gland recognizes cell census, what the origin of metaplastic cells in the corpus is, and how lineage allocation decisions are made. This review only highlighted a select number of studies that used scRNA-seq to examine the transcriptional changes that occur in the gastric epithelium during development, at homeostasis, during regeneration, or the progression of disease. The use of scRNA-seq beyond just the epithelium by including the gastric stromal and vasculature compartments would also be exciting, bettering our understanding of nonepithelium-intrinsic factors that aid the development and maintenance of the stomach, and some of these studies are being done (67). Additional analysis with ever-improving technology will continue to provide insight into important pathways in the progression of gastric diseases like gastric cancer, possibly identifying druggable targets and impacting clinical treatments.

GRANTS

J.C.M. and Y-Z. H. were supported by National Institute of Health Grants R01DK094989, R01DK105129, and R01CA239645. M.A-T. was supported by National Science Foundation- Graduate Research Fellowship DGE-2139839 and DGE-1745038.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.A-T. and Y-Z.H. prepared figures; M.A-T., Y-Z.H., and J.C.M. drafted manuscript; M.A-T, Y-Z.H., and J.C.M. edited and revised manuscript; M.A-T., Y-Z.H., and J.C.M. approved final version of manuscript.