Single-Cell Transcriptome Analyses Reveal the Cell Diversity and Developmental Features of Human Gastric and Metaplastic Mucosa

The stomach is an important digestive organ with a variety of biological functions. However, due to the complexity of its cellular and glandular composition, the precise cellular biology has yet to be elucidated. In this study, we conducted single-cell RNA sequence analysis of the human stomach and constructed a 137,610-cell dataset, the largest cell atlas reported to date. By integrating this single-cell analysis with spatial cellular distribution analysis, we were able to clarify novel aspects of the developmental and tissue homeostatic ecosystems in the human stomach. We identified LEFTY1+ as a potential stem cell marker in both gastric and intestinal metaplastic glands. We also revealed skewed distribution patterns 1 for PDGFRA+BMP4+WNT5A+ fibroblasts that play pivotal roles in, or even precede, the phenotypic changes from gastric to metaplastic mucosa. Our extensive dataset will function as a fundamental resource in investigations of the stomach, including studies on development, aging, and carcinogenesis.


INTRODUCTION
stomach and colonic mucosa indicates that the regeneration of atrophic gastric mucosa generates 126 metaplastic mucosa that resembles, both histologically and functionally, genuine colorectal mucosa. Foveolar cells, surface mucous cells for which the biological characteristics are known to differ 150 between gastric and metaplastic mucosa (Kim and Shivdasani, 2016), were clustered into two distinct 151 populations: GKN1+F cells and ADH1+GKN1-F cells (Figures 1F and 1G; Figure S1A). Encoded by GKN1, 152 gastrokine-1 is a stomach-specific protein with various functions, including modulating cell cycle 153 progression, cellular proliferation, and antibiotic, anti-inflammation, and antiapoptotic actions 154 (Alarcón-Millán et al., 2019). Encoded by ADH1C, alcohol dehydrogenase 1C is often discussed in the 155 context of ethanol metabolism (Edenberg and McClintick, 2018); however, the relationships between its 156 expression and H. pylori infection and IM have been investigated previously, and it may be relevant to the LEFTY1, the product of which is a secreted protein and transforming growth factor-beta (TGF-246 β) superfamily member, has been extensively studied in the developmental stage and is known to play a 247 role in determining left-right asymmetry (Kosaki et al., 1999;Meno et al., 1998). LEFTY1 inhibits SMAD 248 signaling by binding to Cripto-1 and blocks Nodal in the development of mice (Tabibzadeh and 249 Hemmati-Brivanlou, 2006). Additionally, scRNA-seq analysis of Barrett's esophagus has shown that 250 LEFTY1 is a potential marker of Barrett's esophagus precursors in human (Owen et al., 2018). Zabala et 251 al. (2020) showed that LEFTY1 and bone morphogenetic protein (BMP) 7 maintained long-term 252 proliferation and differentiation of human mammary gland cells, respectively, through a mechanism 253 whereby LEFTY1 binds to BMPR2 and prevents BMP7/BMPR2-mediated SMAD activation. In the 254 present study, some ADH1+GKN1-F cell routes specifically found in the metaplastic lineage also 255 expressed LEFTY1 (Figures 2G and 2H). To determine whether LEFTY1 is an actual marker of stem 256 cells and investigate their possible roles in the development of stomach mucosa, we divided both 257 PG/Neck2 and ADH1+GKN1-F cells into two clusters each, i.e., LEFTY1+ and LEFTY1−, respectively. 258 First, LEFTY1+ cells showed significantly higher stemness scores in both PG/Neck2 and ADH1+GKN1-259 F cell populations (p<2.2e-16, Figure 2I). Pseudotime plotting showed that LEFTY1 was highly frequencies and were spatially located mainly around the so-called isthmus regions ( Figure 3A), 272 consistent with the consideration of the histological isthmus region as a stem cell zone ( Figure 1E; Han 273 et al., 2019;Kim and Shivdasani, 2016). Contrastingly, in the intestinal metaplastic mucosa, LEFTY1+ 274 cells were observed at much higher frequencies and were spatially located at the base of crypts ( Figure   275 3A), consistent with the knowledge that intestinal stem cells reside at crypt bases (Spit et al., 2018). 276 These spatial data support the hypothesis that LEFTY1 is a novel marker of gastric stem cells. LEFTY1+ 277 cell frequencies were highest in the metaplastic glands, followed by the PGs and fundic glands, 278 respectively ( Figure 3A). LEFTY1 staining showed two different patterns: a moderate cytoplasmic 279 staining pattern and an intense dot signal pattern. However, the functional differences, if any, of LEFTY1 280 in relation to these staining patterns are not clear, as reported previously in an esophageal study (Owen 281 et al., 2018). 282 We performed RNA in situ hybridization (RNA-ISH) of LEFTY1 and LGR5 ( Figure 3A), finding 283 that a portion of LGR5+ cells coexpressed LEFTY1 in the pyloric and metaplastic mucosa. Additionally, 284 scRNA-seq analysis and immunofluorescent staining showed that subsets of the CD44+ and/or 285 EPHB2+ possible stem cells coexpressed LEFTY1 (Figures 3B-E). In our spatial analysis of human 286 gastric tissues, the colocalization of LEFTY1 with other stem cell markers and the low frequency of 287 LEFTY1+ cells among other stem-marker-positive cells strongly suggest that LEFTY1 is an actual 288 candidate stem cell marker. LEFTY1+ cells can be considered common stem cells in both gastric and 289 metaplastic glands based on our trajectory and pseudotime analyses; however, we found that EPHB2 290 was expressed in LEFTY1+ cells in the metaplastic mucosa but not in the normal gastric mucosa 291 ( Figure 3E). We hypothesize that, during IM, a phenotypic change occurs in LEFTY1+ stem cells in the 292 normal gastric gland and they acquire the distinctive properties of intestinal stem cells by obtaining the 293 EPHB2+ phenotype. To investigate the global gene regulatory network in gastric epithelial cells, we analyzed the activity of In the present study, IHC revealed high and universal HOXB13 expression in metaplastic 336 mucosa, including in both complete and incomplete subtypes; however, in normal mucosa, protein 337 expression was negligible ( Figure 4C). These findings suggest that the HOXB13 regulon is 338 indispensable in the development of IM and that the phenotypic switch of LEFTY1+ possible stem cells 339 between gastric and metaplastic glands may require additional HOXB13 activation. Unlike in colonic 340 mucosa, the expression of HOXB13 was not observed in duodenal or iliac mucosa, suggesting that 341 metaplastic mucosa has similar characteristics to those of colonic mucosa ( Figure 4C).

342
Regardless of LEFTY1 expression, PG/Neck2 cells commonly showed high scores for SOX8 343 ( Figure 4A; Figure S3A), for which an association with stomach biology has not been reported to date; 344 thus, further investigation of its function in the stomach is warranted. Meanwhile, LEFTY1+ADH+F cells 345 showed high scores for CDX1 and CDX2 ( Figure 4A; Figure S3A), confirming that LEFTY1+ADH+F 346 cells are on the routes of differentiation of cells destined for the metaplastic lineage ( Figure 2H). Fundic 347 gland-specific cells (i.e., chief and parietal cells) showed high SP5 activity ( Figure S3A). Huggins et al.

348
(2017) reported that SP5 is a WNT target and negatively regulates WNT activity in human pluripotent 349 stem cells. We confirmed the specific expression of SP5 in fundic glands using RNA-ISH; thus, it 350 appears to be important in the development and/or maintenance of these glands ( Figure 4D were reactivated in the damaged enterocytes of both humans and mice. Given that the stomach and 354 intestine share common features, the gastric mucosa might also reactivate their progenitor programs 355 upon epithelial damage. Indeed, we found that metaplastic mucosa expressed LEFTY1 and HOXB13, 356 both of which are embryonic genes (Kosaki et al., 1999;Ma et al., 2003), at higher levels than those in     To test the aforementioned hypothesis, we conducted RNA-ISH of BMP4, one of the signature 426 genes among the PDGFR+ fibroblasts, and NOG, an intrinsic BMP antagonist (Zimmerman et al., 1996).

427
In intestinal metaplastic mucosa, BMP4+ fibroblasts were more frequently discovered surrounding the 428 metaplastic epithelial cells in the surface areas than in the gastric mucosa ( Figure 5C; Figure S4C). The 429 frequent existence of BMP4+ fibroblasts was also detected in normal gastric mucosa adjacent to 430 metaplasia ( Figure 5C; Figure S4C); however, in normal gastric mucosa distant from IM, BMP4+ 431 fibroblasts were found infrequently ( Figure 5C; Figure S4C). This suggests that composition changes in 432 the population of specific fibroblasts occur earlier than the epithelial changes over the course of IM. In colonic mucosa. We found the monotonically increasing of BMP4 from normal gastric mucosa to 440 metaplastic and colonic mucosa (p= 0.001445; Figure S4D). With these findings, we showed, for the first 441 time, our hypothesis that the increase of BMP4 in the fibroblasts precede and may even induce IM.

442
NOG expression was neither obvious in our scRNA-seq analysis nor was it observed in any cells in the 443 mucosal layers of the stomach ( Figure 5C; Figure S4A); however, neuron cell clusters, including 444 ganglion cells, in submucosal layers expressed NOG in the stomach ( Figure 5D). Drokhlyansky et al.

445
(2020) showed that NOG is expressed in neuron cells in the colonic submucosa, but we are the first to 446 report that neuron cells are the intrinsic source of NOG in the stomach. A previous study found that

451
In our dataset, KLF+ fibroblasts ( Figure 5A) characteristically expressed SFRP1, SFRP2, 452 PI16, and CD34 ( Figure 5B; Figure S4A), but no such fibroblasts existed in the intestine in the public 453 fibroblast atlas, suggesting that the KLF+ fibroblasts in our dataset are unique to the gastric mucosa.

454
RNA-ISH of SFRP1 showed KLF+ fibroblasts existed in the submucosa ( Figure S5B). CD34 is a  , 2006). We also found that expression of BARX1, which encodes a stomach fibroblast-specific 468 transcription factor, was highest in FibSmo cells ( Figure 5B; Figure S4A). Kim et al. (2005)  downstream genes ( Figure 6A). In contrast, BARX1+ fibroblasts were not observed in genuine colonic the frequencies of BARX1+ fibroblasts were comparable to those detected in the gastric mucosa ( Figure   477 6A). We also found clear coexpression of BARX1 and HHIP in parts of the stomach fibroblasts, although 478 BARX1 expression was detected more broadly than HHIP expression in fibroblasts ( Figure 6A). Based on 479 these findings, we confirm the existence of a unique subset of stomach-specific fibroblasts, i.e., with BMP4 expression ( Figure 6B). HHIP expression was also observed broadly in both gastric and metaplastic mucosa; however, in contrast to WNT5A, HHIP+ fibroblasts were not limited to the surface 528 areas but found at broader depths of the stomach mucosa ( Figure 6C). Intriguingly, the close lining of 529 WNT5A+ fibroblasts just behind the epithelial cells was frequently observed throughout the stomach 530 mucosa ( Figure 6B). Conversely, HHIP+ fibroblasts were distributed in stromal spaces at distances from 531 the epithelial cell layers (Figures 6A and 6B). Therefore, WNT5A seems to act locally, i.e., only on the 532 neighboring epithelial cells, whereas HHIP acts more broadly by spreading to distant epithelial cells.

533
Notably, although HHIP+/WNT5A+ double-positive fibroblasts were frequently observed in colonic 534 mucosa, such coexpression of HHIP and WNT5A was not found among stomach fibroblasts, regardless 535 of gastric or metaplastic conditions ( Figure 6C). In conclusion, although gastric and colonic fibroblasts 536 share some characteristics, they also harbor their own specific features and probably have distinct 537 functions.

538
We performed gene regulatory network analysis on the fibroblasts ( Figure S5D), finding that

621
In our scRNA-seq dataset, surface epithelial cells, such as metaplastic enterocytes and ADH1+GKN1-F 622 cells, showed the highest levels of EFNB2 expression ( Figures S8D and S8E). Using IHC, we showed 623 that EPHB2 was only positive in the crypt base of IM, whereas EFNB2 was positive in other regions of 624 IM and the superficial region of gastric mucosa ( Figure S8F). This confirms that the function of Eph-625 ephrin repulsion in intestinal metaplastic mucosa is similar to that in colorectal crypts; however, this 626 interaction was not observed in stomach mucosa. This implies that different combinations of Eph-ephrin interactions or unknown regulation mechanisms might play some roles in stomach glands.

628
Although our CCC analysis did not identify clear enrichment of the major signaling pathways of 629 BMP or WNT between epithelial cells and fibroblasts, the spatial distribution of WNT5A+ fibroblasts 630 ( Figures 6B and 6C) suggested that interesting cellular interactions possibly occur around WNT/BMP.

631
According to RNA-ISH analysis, WNT5A+ fibroblasts were mainly found in the surface area of stomach 632 mucosa ( Figures 6B and 6C); however, following closer observations, we found that WNT5A+ fibroblasts 633 were also found in line with LEFTY1+ possible stem cells at the crypt bases of IM ( Figure S8G). This    with those of the Tsinghua University dataset and our own dataset.  We extracted green or red signals from the image of RNAScope and magnified these signals 100 times 1174 using python library cv2 and PIL. Green and red signal thresholds were defined manually.