Background
Gastric/gastro-esophageal junction (GEJ) adenocarcinomas comprise a morphologically, molecularly and clinically diverse group of cancers. TCGA studies demonstrated that gastric/GEJ carcinomas are a heterogeneous group of diseases that could be sub-categorized into four molecular subtypes, namely EBV-related, microsatellite instability (MSI), chromosomal instability (CIN) and genomically stable (GS) [
1,
2]. While the molecular heterogeneity to some extent reflects the different etiologies and pathways of tumorigenesis, such as deficiency in the mismatch repair process (leading to the MSI subtype tumors), another potential factor contributing to the molecular diversity in gastric/GEJ adenocarcinoma is the poorly explored concept of “cell-of-origin”.
Applying the cell-of-origin concept to gastric/GEJ adenocarcinoma pathogenesis is challenging, related as it is to the complex biology of stem/progenitor cells in the stomach. The gastric mucosa is a site of dynamic regenerative homeostasis, with turnover that is thought to be maintained by long-lived stem/precursor cells, with a number of studies that point to Lgr5
(+) cells as the self-renewing, multipotent stem cells [
3,
4]. The key signaling pathway behind this phenomenon is thought to be the canonical Wnt pathway, with Lgr5 functioning to potentiate Wnt/β-catenin signalling by interacting with R-spondin, among other factors [
5,
6]. This role of Lgr5 is presumably facilitated by its trans-membrane domain allowing for membranous localization, but Lgr5 is a spatially dynamic protein, with expression seen also in the cytoplasm and even the nucleus [
7‐
9]. Surface Lgr5 expression is partially regulated by endocytosis, and inhibiting clathrin-mediated endocytosis of Lgr5 diminishes intestinal epithelial cell fitness in murine models [
8]. In the correct in vitro conditions, isolated single Lgr5
(+) stem cells have the remarkable capacity to form organoids, known as “mini-guts” [
3]. Long-term ablation of Lgr5
(+) cells impairs epithelial homeostasis in the corpus [
10]. In homeostasis, the majority of the Lgr5
(+) stem cells appear to divide symmetrically [
11], and their proliferation can be accelerated by
H. pylori infection [
12]; this may represent a potential link between the epidemiological observation of increased risk of carcinogenesis associated with
H. pylori infection. Lgr5 has also been postulated to be a marker for cancer stem cells (CSCs) in gastric cancer [
10,
13], although the hypothesis is currently lacking in definitive experimental results. While concepts may be overlapping, CSCs differ from cells-of-origin in that CSCs generally refer to progenitor/stem-like cells isolated from cancer specimens, generally with ability to form tumors in xenograft models [
14]. CSCs that have been obtained and characterized from the stomach have been isolated using a number of different mesenchymal stem cell (MSC) markers such as CD90 [
15‐
17]. Interestingly, more primitive MSC/MSC-like cells are also found in normal stomach, and some studies have attributed roles in intestinal and gastric homeostasis to these MSC/MSC-like cells [
16].
When analyzed agnostic to the intrinsic heterogeneity in gastric/GEJ adenocarcinoma, Lgr5 expression does not carry a statistically significant prognostic weight, although some prognostic significance is seen in early stage (I/II) patients [
18,
19]. Studies have associated higher Lgr5 expression with more frequent nuclear β-catenin expression, a marker of active Wnt signalling, and higher Lgr5 has been associated with worse survival in a subset of patients. However, these studies predated the TCGA molecular subtyping, and cellular sub-localization of Lgr5 is now better characterized; nuclear Lgr5 has been reported in hair follicle stem cells, and their expression was observed to be limited to cycling hair follicle stem cells, after asymmetric self-renewal division [
7]. In light of these previous reports and better understanding of Lgr5 biology, we examined the prognostic value of
LGR5 mRNA and Lgr5 protein, including its nuclear expression, and we frame the discussion to focus on the potential implication in histogenesis.
Discussion
The general dogma in many studies that have examined “stemness” posit that stemness is associated with enhanced ability to invade, and thus ultimately being associated with worse survival. However, our data with Lgr5
/LGR5 suggest that this story is likely more complex—it depends on what stem signature is present. The immunohistochemical and mRNA expression profile phenotype with Lgr5
High-GSC signature combination, which may represent “retained stemness”, was associated with better survival (Fig.
3). If this is indeed a phenotype retained from its cell-of-origin, we may speculate that these tumors retained their intrinsic predisposition for glandular differentiation, explaining the association seen with better differentiated, intestinal histo-morphology, and Lgr5 expression. Lgr5
(+) cells have been postulated to be the cellular origin of invasive gastric carcinoma by others, specifically in the intestinal histotype [
27,
28]. Noting the association with the CIN molecular subtype, the Lgr5
(+) may have an intrinsic predilection for a specific tumorigenic pathway. Targeted deletion of
Smad4 and
PTEN in Lgr5
(+) gastric stem cells in mice resulted in formation of invasive adenocarcinoma, whereas the same set of genetic maneuvers failed to induce tumorigenesis in differentiated cells, including antral parietal cells, pit cells, as well as corpus Lgr5
(+) chief cells [
28]. The likely candidate precursor in this pathway would be intestinal metaplasia (IM), and both
Lgr5 mRNA and Lgr5 protein expression has been reported in IM [
27,
29,
30]. Lgr5
(+) cells are seen at the crypt base in IM, and the expression becomes more marked in carcinoma in its luminal surface, tumor center and the invasion front type [
10,
27]. Interestingly, while spasmolytic polypeptide-expressing metaplasia (SPEM), another putative precursor to gastric adenocarcinoma, was suggested to be not associated with Lgr5
(+) cells in the variegated Lgr5-EGFP-IRES-Cre (ERT2/+); in Rosa26R mice [
31], expression of constitutively active
Kras (harboring p.G12D hotspot mutation) in Lgr5
(+) chief cells was sufficient form SPEM in a non-variegated Lgr5-2A-CreERT2 mouse model [
10]. Thus, at least two different pathogenic pathways involving Lgr5-expressing cells appear to be possible, both of which may give rise to CIN molecular subtype tumors. Interestingly, while the tumors with GEJ/cardia epicenter are largely of the CIN molecular subtype [
1], these tumors expressed lower levels of
LGR5 mRNA. The abundance of Lgr5
(+) cells varies across the different anatomical regions, being most abundant in the antrum [
3,
29]. Thus, there may also be heterogeneity amongst the CIN tumors, with those cardiac/GEJ tumors potentially arising from Lgr5-negative cells, perhaps including the recently described osteopontin responsive, Lgr5-negative/CD44
(+) cells [
32]. In line with this possibility, GEJ/cardia tumors were notable for high
CD44 mRNA, although
CD44 mRNAs are likely to be expressed by a variety of non-tumoral cells, especially lymphocytes.
A part of the challenge in analyzing Lgr5 and its clinical significance is related to its complex biology, including its heterogeneous sub-cellular localization. The physiologic role of nuclear Lgr5 has only thus far been described in the hair follicle stem cells, where nuclear Lgr5 expression is limited to the cycling stem cell sister after asymmetric self-renewal divisions [
7]. While the role of nuclear Lgr5 in the gastric epithelium and in cancer cells remains largely unexplored, it is worth noting that several studies examining Lgr5
(+) cells based their identification of Lgr5
(+) cells based on
LGR5 mRNA expression and/or
LGR5-driven reporters (ex.
Lgr5-lacZ), where the sub-cellular localization of reporters may differ from that of the endogenous Lgr5 [
3,
10,
29]. Inferring from studies on hair follicle stem cells [
7], nuclear Lgr5 may be highlighting the cancer cells with a yet a different pattern of “stemness”. While nuclear Lgr5 was nearly mutually exclusive with non-nuclear, strong Lgr5 expression, nuclear Lgr5 also showed similar associations with the intestinal histotype and CIN molecular subtype, suggesting nuclear Lgr5 may also represent a pattern of “retained” phenotype, similar to non-nuclear Lgr5.
Beside the Lgr5
(+)/
LGR5-expressing cells, more primitive MSC/MSC-like cells appear to be playing a crucial role in intestinal homeostasis, and MSC-like cells have been isolated from human gastric cancer tissues, marked by high expression of
THY1 [
16,
17]. Some studies have even implicated bone marrow-derived MSCs in gastric epithelium homeostasis, which have been shown to repopulate the gastric epithelium [
33,
34]. Interestingly, bone marrow-derived MSCs, at least in mouse models, appear to be able to give rise to both the carcinoma cells and the cancer-associated stromal cells, contributing to disease progression [
33,
34]. In the context of these theories of gastric cancer carcinogenesis, the clustering pattern of different stem/progenitor cell markers seen in the TCGA data is interesting (Fig.
2a), which suggest that there may be at least two different signatures of stem/progenitor gene expression, i.e. one more intestinal (cluster #1,
LGR5-high) and one more MSC-like (cluster #2,
THY1-high). However, this is likely to be a simplified version of the story, considering the complex hierarchy between the different stem/progenitor cells in the GI tract [
35,
36]. Future studies may thus benefit from examining co-expression patterns of multiple markers, including Bmi1, which appear to mark slow-cycling stem/progenitor cells that can revert to express Lgr5 [
35,
37,
38], while noting the different sub-cellular Lgr5 expression patterns. One peculiar finding was the co-clustering of
RSPO1 and
RSPO3 in cluster #2. The mRNA levels of the two genes correlated highly with one another, while correlating little or negatively with the cluster #1 genes (Fig.
2a). R-spondins are ligands to Lgr5, and the “Rspo-Lgr5 axis” plays an important role in stimualting the growth of intestinal crypts in vitro via the Wnt pathway [
39,
40]. While the “Rspo-Lgr5 axis” had been previously postulated to be merely an amplification process for Wnt signaling, more recent studies point to non-overlapping, non-interchangeable roles between Wnt and R-spondin in GSCs [
41]. As well, R-spondins, at least R-spondin 3, can be produced by myofibroblasts in the stomach [
6]. As fibroblasts/myofibroblasts are also marked by CD90 (Thy1) expression, these observations may be merely indicating enrichment of stromal cells. Alternatively, Lgr5, perhaps in its nuclear form, may be carrying out functions independent of R-spondins in gastric/GEJ carcinoma, and further studies would be required to better understand the Rspo-Lgr5 axis in gastric/GEJ carcinoma.
We encountered several challenges in this study, one of which was related to our relatively small cohort size, which likely explains the lack of statistical significance with some of the analyses (ex. OS analysis in Fig.
3a/b). Another weakness of this study is its purely correlative nature, and we cannot differentiate between retained vs. acquired Lgr5 expression. Molecular profiling of Lgr5
(+)/
LGR5-expressing cell-derived tumors through the pathogenic pathway, including the precursors, could shed further light on the tumorigenic pathway. Lgr5
(+)/
LGR5-expressing cell-derived tumors may not be restricted to a specific molecular subtype, and CIN-type tumors are unlikely to be restricted to Lgr5
(+)/
LGR5-expressing cells as their sole cell-of-origin. The hierarchy of the stem/progenitor cells in the stomach await further clarification, and we await identification of other candidate cells-of-origin and to assess the impact of genetic manipulation of other candidate cells.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.