The immune microenvironment in gastric adenocarcinoma

Abstract

Like most solid tumours, the microenvironment of epithelial-derived gastric adenocarcinoma (GAC) consists of a variety of stromal cell types, including fibroblasts, neuronal, endothelial and immune cells. In this article, we review the role of the immune microenvironment in the progression of chronic inflammation to GAC, primarily the immune microenvironment driven by the gram-negative bacterial species Helicobacter pylori (H. pylori). The infection-driven nature of most GACs has renewed awareness of the immune microenvironment and its effect on tumour development and progression. About 75–90% of GACs are associated with prior H. pylori infection and 5–10% with Epstein Barr Virus (EBV). Although 50% of the world’s population is infected with H. pylori, only 1–3% will progress to GAC, with progression the result of a combination of the H. pylori strain, host susceptibility and composition of the chronic inflammatory response. Other environmental risk factors include exposure to a high-salt diet and nitrates. Genetically, chromosome instability occurs in ~50% of GACs and 21% of GACs are microsatellite instability-high (MSI-H) tumours. Here, we review the timeline and pathogenesis of the events triggered by H. pylori that can create an immunosuppressive microenvironment by modulating the host's innate and adaptive immune responses, and subsequently favour GAC development.

Introduction

The focus of this Review is to explain how the chronic inflammatory microenvironment in the gastric mucosal epithelia during Helicobacter pylori (H. pylori) infection can stimulate intracellular signalling pathways that can lead to gastric adenocarcinoma (GAC). As with many solid tumours arising from epithelial cells, cells in the surrounding support structures have multiple roles in tumour growth, as initiators, promoters or enablers. These supporting cell types include stromal cells, neuronal cells, endothelial cells and a vast array of immune cells¹. H. pylori infection has been shown to be associated with a sustained immune infiltrate (chronic gastritis) that can contribute to the development of gastric neoplasia. Therefore, adenocarcinoma of the stomach is now considered a cancer that can be initiated by infectious agents^2,3. About 75–90% of GACs exhibit positive H. pylori serology^4-7. The incidence of adenocarcinoma of the gastric cardia is increasing in the US and epidemiologically correlates with increases in the prevalence of obesity, gastrooesophageal reflux disease and oesophageal adenocarcinoma^8,9. By contrast, non-cardia gastric cancers, which include GAC in the stomach body or antrum, show a stronger association with H. pylori infection, low socioeconomic status and minority ethnic groups^10-12. Epstein-Barr virus (EBV) infection is associated with about 5–10% of cases of GAC^7,13,14. Owing to these associations, some GACs should be preventable either at the level of public health interventions, bacterial eradication, vaccination or suppression of the immune response; however, the latter two approaches have not been overly successful^3,15. Most individuals infected with H. pylori are asymptomatic¹⁵ and antibiotic treatment fails to eradicate the organism in at least 20% of patients due to the existence of resistant strains and a lack of patient compliance¹⁶. Only about 20% of infected individuals develop serious complications related to a combination of environmental factors (smoking and occupational tobacco exposure), dietary factors (high intake of smoked foods, high intake of foods high in nitrates and/or a high-salt diet) and host susceptibility (particularly single nucleotide polymorphisms (SNPs), iron deficiency and blood type O^15,17-23. In particular, several cytokine variants affect an individual’s susceptibility to GAC^24-26.

During the 1860’s, Rudolf Virchow proposed that cancer is the response to a “wound that will not heal”, and since then investigators have sought to understand the role of immune cells in various cancers^27,28. The immune infiltrate accompanying an H. pylori infection has a prominent role in creating and shaping the tumour immune microenvironment (TIME) that promotes progression to GAC. Moreover, GAC has become a paradigm for inflammation-driven cancers^21,29. This Review discusses how acute and chronic inflammatory infiltrates induced by H. pylori infection transform epithelial cell populations in the pathogenesis of GAC.

Gastric adenocarcinoma histopathology

In 1975, Pelayo Correa published the astute observation that intestinal type GAC was associated with an inflammatory process in the stomach³⁰. This publication occurred ~10 years prior to the seminal report that showed that a particular bacterial species, H. pylori (previously known as Campylobacter pylori), colonizes the stomach and initiates an inflammatory response or gastritis ^31,32. Approximately 50% of the world’s population is infected with H. pylori; about 20% of those infected develop chronic gastritis and/or more serious complications such as duodenal or gastric ulcers^20,33 and approximately 1–3% of those infected subsequently develop GAC⁵ (Fig. 1). In 2014, researchers performed a comprehensive molecular evaluation of 295 human GACs and divided them into four molecular subtypes—tumours with chromosome instability (50%), tumours positive for EBV (9%), microsatellite instability-high (MSI-H) tumours (21%) and genomically stable tumours (20%) ³⁴. These four subtypes exhibit distinct molecular signatures that correlate with methylation status, EBV DNA and the presence of mismatch repair gene mutations, or show no genome perturbations (genomically stable and microsatellite stable tumours). In this study, activated signalling pathways were only inferred from the genes mutated as concurrent gene expression profiling was not performed³⁴. An association with H. pylori could not be accurately assessed owing to the absence of serology and because H. pylori DNA was only sporadically detected in the GACs. Understanding the genomic changes that occur in GAC is useful but does not provide complete insight into the stepwise sequence through which multiple pathways lead to gastric epithelial cell transformation, nor does it provide information on the TIME.

Figure 1 ∣. Timeline of complications from Helicobacter pylori infection.

A schematic of the timeline from H. pylori infection to the development of gastric ulcers versus gastric atrophy, intestinal metaplasia, dysplasia and gastric cancer. Key genomic changes were identified by profiling of human gastric intestinal metaplasia and cancer^123,260. Vertical arrows indicate direction of the change; for example, FBXW7 mutations identified in intestinal metaplasia and more numerous in the WD40 domain are represented by larger arrow. Bidirectional arrow indicates increase or decrease. Infection of the gastric mucosa initiates acute inflammatory responses most of which result in an increase in NFκB. Injured parietal cells release Shh within 2 days of the infection and Shh can be detected in the circulation, which subsequently recruits myeloid cells to the stomach²¹⁰. Why some individuals do not develop atrophy and instead develop duodenal ulcers owing to the hyperacidity is unknown, but RNA and DNA profiling suggests that hypermethylation, loss of parietal cells and a LOF in the ubiquitin ligase FBXW7 are early steps that lead to the pathway from intestinal metaplasia to GAC¹²³.

In 2020, analysis of the TIME was performed on a subset of TCGA cancers using RNA sequencing and showed that the chromosomally unstable tumours could be subdivided into T-cell-rich ‘hot’ gastroesophageal cancers (GEAs) and T-cell-poor ‘cold’ GEAs. The cold chromosomally unstable GEAs showed enrichment of cMYC and amplification of the cyclin E1 gene (CCNE1) compared with ‘hot GEAs³⁵. In a separate study, an immunohistochemical analysis of immune subsets—including tumour-infiltrating lymphocytes (TILs) and tumour-associated macrophages (TAMs)—was performed in 43 GACs classified according to the four TCGA subtypes³⁶. EBV-positive (n=6) and MSI-H (n=11) GACs showed the highest number of immune cells (‘hot’ tumours) and highest expression of checkpoint inhibitors while intestinal (a surrogate for chromosomal instability; n=14) and diffuse (a surrogate for genomically stable; n=12) cancers showed the least number of immune infiltrates (‘cold’ tumours). In the latter two categories, H. pylori infection was documented in only three of 25 patients. However, this retrospective analysis was based on immunohistochemistry (IHC), not correlation with H. pylori serology. Thus, whether immune-poor GACs (70–75% of GACs in this study) represent the initiation of GAC pathogenesis by mechanisms other than H. pylori, EBV or MSI-H, is unclear. Nevertheless, the findings underscore the need to quantify and characterize the TIME as a function of the cancer subtype to identify which cancers are most responsive to treatment.

Three major histologic cancer subtypes arise from the gastric epithelium—intestinal, diffuse and mixed³⁵. Pelayo Correa described a cascade of events (the Correa cascade or paradigm) that precedes GAC based upon epidemiological observations³⁰. The Correa cascade was ultimately incorporated into the Sydney Classification system, which describes sequential events in histological terms: chronic inflammation proceeds to intestinal-type GAC through distinguishable lesions consisting of atrophic gastritis and intestinal metaplasia³⁷ (Fig. 1). Nevertheless, H. pylori infection is not as strongly correlated with diffuse GAC as with intestinal type GAC^38,39. However, the evidence is mixed, with some studies reporting that up to 80% of diffuse GAC cases were positive for H. pylori^40,41. H. pylori infection increases the likelihood that patients with MSI-H will develop GAC^42,43. A 2020 study suggests that different cag pathogenicity island (cagPAI) genes, such as CagA and the vacuolating cytotoxin (VacA) gene (CagA⁺ and VacA⁺) affect which GAC histologic subtypes (diffuse versus intestinal) develop⁴⁴. COX-2, DAPK and CDH1 have been shown to be the most highly mutated genes during infection by CagA⁺ H. pylori⁴⁵. Thus, the predisposing factors driving specific GAC subtypes are not clearly defined.

Diffuse-type GACs

Hereditary diffuse gastric cancers (HDGC) develop in young adults and comprise 1–3% of GACs^46,47. Affected individuals frequently carry loss of function (LOF) mutations in the genes encoding tumour suppressors cadherin 1 (CDH1) or catenin α1 (CTNN1A), a gain of function (GOF) mutation in the locus encoding transforming protein RhoA (RHOA; a downstream GTPase signalling component of the actin cytoskeleton), or have a fusion gene CLDN18-ARHGAP^46-50. Disruption of CDH1 and loss of cell junction protein expression renders the gastric mucosa more susceptible to transformation, cancer invasion and metastasis²¹. HDGC is inherited in an autosomal dominant pattern, is highly penetrant and predominates in young women, especially in Asia^51,52. CDH1 mutations are also associated with lobular breast cancer in women⁵³ and with colon cancer in women and men⁵⁴. Despite a strong association of H. pylori with sporadic diffuse-type and intestinal-type GAC, the organism is rarely detected in total gastrectomy specimens prophylactically removed from asymptomatic patients with HDGC^55,56.

Sporadic diffuse GACs make up ~20–30% of GAC cases in Western countries⁵⁷. However, gastric atrophy and intestinal metaplasia are not prominent histological features of diffuse GACs⁵⁸. Thus, commitment to tumour pathogenesis is less obvious in the absence of precursor lesions than with intestinal type GACs, with hypermethylation causing LOF or GOF in tumour suppressor or proto-oncogenes controlling tight junction proteins such as CDH1 and CTNN1A^46,57. A transgenic model of diffuse GAC in which the most common oncogenic variant (Y42C) of the RhoA locus was conditionally activated together with inactivation of Cdh1 mimicked features of diffuse GAC such as lung, liver and peritoneal metastasis⁴⁸. In particular, the RhoA^Y42C mutant shows impaired GTP hydrolysis and exhibits enhanced interaction with its effector kinase, ROCK. The subsequent stimulation of actin stress fibres and cytoskeleton rearrangements activate focal adhesion kinase (FAK) that modulates downstream pro-proliferative pathways phosphoinositide-3 kinase (PI3K)–AKT, β-catenin and the Hippo signalling pathway through YAP–TAZ⁴⁸. In the absence of Helicobacter infection in this diffuse GAC mouse model, loss of the Cdh1 tumour suppressor gene uncovered the effects of oncogenic RhoA^Y42C, suggesting that FAK might be a relevant therapeutic target.

Galectins are prominent glycoproteins that contribute to the peritoneal metastasis of diffuse cancers⁵⁹. Plasma membrane glycosylation is frequently expressed aberrantly in GACs and can also contribute to heightened invasiveness^21,60,61. For example, the overexpression of Galectin-3, a β-galactosidase specific binding protein, localizes to the cytoplasm of normal gastric mucosal cells, where it is expressed in both the cytoplasm and nucleus of GAC cells⁶². When expressed in the epithelial cell compartment, Galectin-3 increases cell motility⁶³. Galactin-9 is a ligand for T-cell immunoglobulin and TIM-3 (also known as hepatitis A virus cellular receptor 2), encoded by the HAVCR2 gene and expressed on almost all immune cells as an immune checkpoint receptor²¹. TIM-3 expression is significantly upregulated on peripheral blood monocytes from patients with GAC⁶⁴. Importantly, the TIM-3/Gal-9 axis might have a critical role in T-cell exhaustion within the GAC microenvironment⁶⁵. Therefore glycan-binding proteins might be useful prognostic markers for patients with GAC^63,66.

Diffuse GAC also correlates with H. pylori-mediated methylation of CpG islands in the CDH1 gene and other promoters^40,67-69. Genomic suppression of CDH1 is often observed in the CpG island hypermethylator phenotype (CIMP) GAC subtypes, as observed with patients with EBV or Lynch syndrome in which hypermethylation or inherited mutations in the loci encoding mismatch repair proteins leads to high microsatellite instability (MSI-H) ^70,71. By contrast, APC or CTNN1B mutations correlate with low microsatellite instability (MSI-L) and microsatellite stable (MSS) intestinal-type gastric cancers of the antrum^72-74. MSI-H cancers, including some GACs, display a higher mutational burden than non-MSI-H cancers and create important neoantigens that are subsequently recognized by the immune system⁷⁵. As such, these MSI-H tumours exhibit higher densities of immune infiltrates and are more susceptible to immune checkpoint blockade than non-MSI-H ⁷⁶. MSI-H GACs typically account for only 10–20% of MSI-H tumours^34,77-79, which might explain the overall poor response rate for GACs treated with the current immune checkpoint inhibitors^76,78-81.

EBV-positive GACs

EBV-positive GACs are typically of the diffuse-type and are located in the gastric body and cardia rather than the antrum^82,83. EBV infection creates a CIMP genomic environment of hypermethylation. The hypermethylation in EBV-driven cancers is non-random, with DNA repair genes being a frequent target, suggesting some genomic mutational overlap with MSI-H GACs ⁸². Normal gastric tissue displays a promoter CpG island methylation level of 1–2% ^34,84, whereas EBV-positive GACs display a CpG island methylation frequency of 19% and MSI-H GACs show a CpG island methylation frequency of 10%³⁴. EBV-driven GACs also exhibit a greater degree of immune infiltration and increased expression of immune checkpoint proteins (for example, programmed cell death protein 1 (PD-1; also known as CD279)³⁶, CTLA-4 (also known as CD152)) compared with non-EBV cancers and thus show a better response to immune checkpoint inhibitors⁸⁵. One study suggested that 15–38% of all GACs are EBV-positive, depending on the number of mismatch repair loci evaluated^84,86. The EBV latent membrane proteins (LMP1 and LMP2A) induce methylation of the tumour suppressor PTEN promoter by increasing STAT3-mediated expression of DNA methyltransferase1 (DNMT1)^87,88. This dual-specificity phosphatase inhibits PI3K–AKT signalling and in many cancers is suppressed through epigenetic silencing⁸⁹. CDX2, KLF4 and TFF1 are representative pro-intestinal differentiation genes that are epigenetically silenced in EBV-driven cancers⁸². A review of EBV-positive GACs indicates that EBV-mediated hypermethylation does not occur without pre-existing conditions such as chronic atrophic gastritis, intestinal metaplasia or dysplasia ⁸², suggesting that coinfection or sequential infection with H. pylori might have occurred^82,90, as reported for late-stage or metastatic GAC tumours^90,91.

H. pylori and gastric pathogenesis

H. pylori-induced pathogenesis is largely mediated by the virulence factors VacA and cytotoxin-associated gene A (CagA)^17,92. Injection of these factors into epithelial cells via the type IV secretory system (T4SS) activates a number of pro-tumorigenic pathways that propels the gastric mucosa towards an intestinal stem-like state, for example through activation of WNT–β-catenin signalling, which in turn induces expression of pro-intestinal differentiation factors CDX1, KLF5 and LGR5^92,93. Mutations in Wnt signalling components, including gain of function (GOF) mutations, loss of function (LOF) mutations, mRNA activation and epigenetic alterations, which collectively are present in about 30% of GACs^94,95. Other studies indicate that H. pylori virulence factors initiate the transition from atrophy to intestinal metaplasia through cytokine-mediated activation of NFκb and subsequent induction of the intestinal differentiation factors CDX1 and CDX2⁹⁶. In the progression to intestinal-type GAC, H. pylori reduces CDX1 promoter methylation level at the histologic stages of gastritis and intestinal metaplasia, reaching its lowest level of methylation in dysplasia, and then rising again in gastric cancer. NFκb ultimately binds to the unmethylated CDX1 promoter to initiate gastric mucosal transdifferentiation to an intestinal phenotype. However, once dysplasia segues into adenocarcinoma, CDX1 promoter methylation rises again^96,97. H. pylori CagA subverts both phosphorylation-independent pathways (cMET, CDH1 and WNT–β-catenin) and phosphorylation-dependent kinase signalling pathways (EGFR and gp130–STAT3), in which both pathways can activate NFκB leading to changes in epithelial cell motility, the cell cycle and release of pro-inflammatory mediators⁹⁸. In addition, H. pylori lipopolysaccharide (LPS) is recognized by toll-like receptor (TLR)4, a pathogen recognition receptor (PRR), which induces pro-inflammatory cytokine expression and type 1 interferons through formation of the inflammasome and activation of transcription factors NFκB, AP1 and interferon response factors (IRFs)⁹⁹. Recognition of unmethylated CpG DNA motifs by endocytic TLR9 is another important mechanism through which H. pylori activates epithelial signal transduction cascades via the release of chemokines such as IL-8^100,101. TLR9 mediates both anti-inflammatory and pro-inflammatory responses depending on the gastric microenvironment¹⁰². During the acute inflammatory phase, TLR9 activation by H. pylori bacterial DNA suppresses the immune response, which facilitates bacterial colonization¹⁰³. In addition, LPS, peptidoglycans, flagellin (TLR5) and other non-protein molecules that enter the epithelial cell via the T4SS also activate TLR9¹⁰⁴.

The major TLRs associated with GAC are TLR2, TLR3, TLR4, TLR5, TLR7 and TLR9^105-107. Four intracellular, endocytic TLRs (TLR3, TLR7, TLR8 and TLR9)¹⁰⁸ are known to be capable of responding to H. pylori RNA or DNA, but TLR3, TLR7 and TLR8 have been infrequently studied¹⁰⁹. Activation of TLR3 requires double-stranded RNA species characteristic of specific viruses¹¹⁰. However, TLR7 and TLR8 recognize single-stranded RNA species and thus can detect H. pylori RNA in a MyD88-dependent and TRIF-dependent manner¹⁰². The cytoplasmic retinoic acid inducible gene (RIG-1; also known as DDX58), but not the melanoma differentiation associated antigen 5 (MDA5; also known as IFIH1) also recognizes bacterial RNA but its activation of type 1 interferons is independent of MyD88 and TRIF¹¹¹. Antiviral innate immune response receptor RIG-I is a non-TLR pathogen recognition RNA helicase receptor that recognizes 5-triphosphate RNA (PPP-RNA), a specific feature of viral and prokaryotic RNAs¹¹¹. The 5’ end of bacterial and viral RNA is not capped making it a genuine RIG-I ligand^112,113. Although H. pylori initiates a cacophony of signalling cascades in both epithelial cells and antigen presenting cells (APCs), the combination of H. pylori strain and host susceptibility seems to ultimately determine which pro-inflammatory or anti-inflammatory responses predominate.

TLR9 and DAMP signalling

TLR9 receptors have been best characterized on APCs such as dendritic cells (DCs), macrophages and B cells^101,114. Plasmacytoid dendritic cells (pDCs) are the major source of secreted type I interferons (IFNα and IFNβ) upon activation of endocytic TLR7 and/or TLR9 by viral or unmethylated self-derived DNA^115,116. However, in contrast to TLR9 activation in epithelial cells, H. pylori DNA in pDCs is immunosuppressive as H. pylori DNA fails to stimulate secretion of type1 interferons and IL-12 in mouse or human bone marrow-derived dendritic cells (BMDCs) and human pDCs¹¹⁷. We reported that the rise in tissue levels of IFNα produced by pDCs correlates with the appearance of spasmolytic polypeptide-expressing metaplasia (SPEM) during the chronic phase of an Helicobacter felis infection¹¹⁸. Moreover, TLR9-mediated induction of IFNα contributes to suppression of H. pylori by down-regulating T_H1-mediated gastritis¹⁰³. Although H. pylori DNA can gain entry into both gastric epithelial cells and pDCs, the overarching chronic effect of an H. pylori infection is to dampen the T_H1/T_H17 proinflammatory response by increasing the numbers of immunosuppressive cells such as T_reg cells that secrete TGFβ and IL-10¹¹⁹ (Fig. 2). The ability of the H. pylori gastric infection to inhibit the T_H1/T_H17 immune response has systemic implications, as shown by H. pylori potently mitigating inflammation in the mouse colon (colitis) and lung (asthma)^117,120-122. Thus, the phenotypic shift in the composition of immune infiltrates in response to H. pylori DNA motifs sets the initial stage for chronic gastritis. Moreover, genomic analysis of intestinal metaplasias from cancer-free patients demonstrated that CagA⁺ H. pylori DNA is present in these precursor lesions, suggesting that H. pylori makes some contribution to the progression of intestinal metaplasia to GAC, perhaps related to changes in telomere length¹²³.

Figure 2 ∣. PMC-MDSC induction during spasmolytic polypeptide-expressing metaplasia (SPEM) development.

Acutely, Slfn4⁺ myeloid-derived suppressor cells (MDSCs) regulated by Shh signalling track to the stomach, but do not develop their T-cell suppressor function until levels of tissue IFNα produced by plasmacytoid dendritic cells (pDCs) increase. SPEM development coincides with maximal expression of Slfn4 and the Slfn4⁺ MDSCs acquire T-cell suppressor function ^118,125. Slfn4⁺-MDSCs express NFκB and secrete regulatory microRNAs such as MIR130b²⁰³. DAMPs, damage-activated molecular patterns.

H. pylori DNA and heterogeneous byproducts of injury and inflammation produce various damage-activated molecular patterns (DAMPs), which are potent activators of TLR9-mediated IFNα production¹²⁴. Host cell debris can induce an increase in TLR9 tissue levels of IFNα^118,125. High mobility protein B1 (HMGB1), heat shock proteins 70 and 90 (HSP70 and HSP90), mitochondrial DNA, calreticulin, ATP, the IL-1 cytokine family and type I interferons (IFNs) are all examples of DAMPs released from host cells dying by various mechanisms including necroptosis, apoptosis, pyroptosis and ferroptosis^126,127. In particular, in response to IL-17A and tamoxifen, apoptotic parietal cells likely release bioactive molecules that activate DAMP signalling^128-130. In addition, tumour cells themselves generate DAMPs that perpetuate the immune suppressive microenvironment¹³¹.

The timeline in humans from H. pylori infection to disease complications — gastric atrophy, metaplasia and GAC — can span months to years¹³². During that time, the loss of acid-secreting parietal cells leaves the stomach relatively hypochlorhydric, which can shift the composition of the gastric microbiota^133-135. Although a meta-analysis of studies investigating the gastric microbiome in gastric cancer inferred the presence of changes in the microbiota at different stages of the Correa cascade¹³⁶ , none of the studies link the changes in the microbiome to specific inflammatory markers^137-139. In humans with chronic gastritis, Prevotella, Streptococcus, Pseudomonas, Sphingomonas, Bacillus, and Acinetobacter were also identified in the normal mucosa adjacent to the tumour; however, H. pylori was still the organism consistently identified at different stages of progression¹³⁶. This finding is consistent with mouse studies in which Acinetobacter species were shown to induce gastritis in mice and to cause a T_H1 immune response^140,141. Bacteroides and Haemophilus species have also shown elevated 16S rRNA genomic signatures in the blood of patients with GAC¹⁴². Therefore, although H. pylori seems necessary for the initial steps in the process of gastric carcinogenesis, once the mucosa becomes atrophic, it is unclear whether H. pylori, other bacteria or immune cells drive the mucosa towards metaplasia or cancer. As cell or bacterial debris is sufficient to sustain DAMP-mediated inflammation, further epithelial changes might not require live organisms and the microbial composition of the stomach at this stage might not influence what happens next. Establishing the microbiome pattern in gastric cancer has clinical consequences as it could indicate whether antibiotic-mediated eradication can prevent progression to GAC. We can only speculate what impact gastric microbiota other than H. pylori have on the immune microenvironment by correlating immunosuppressor subtypes with bacterial species in the atrophic stomach.

Impact of CagA on gastric epithelial cells

H. pylori CagA accumulates in GAC cell lines that express CD44, a cell surface marker associated with cancer stem cells, one isoform of which suppresses cell autophagy and enhances their resistance to reactive oxygen species (ROS)¹⁴³. In cells not expressing this isoform of CD44, CagA is degraded by autophagy induced by the accumulation of ROS¹⁴³. Thus, within the tumour environment, cancer cells become resistant to ROS-induced apoptosis through CD44^144,145. The CD44 gene encodes several protein variants owing to alternative splicing and post-translational modifications¹⁴⁶. CD44 variable isoforms have been implicated as key players in malignant transformation where their expression is highly restricted and specific, unlike the canonical CD44 full-length isoform¹⁴⁷. Upon delivery into host cells by the T4SS, CagA translocates into the host cell cytoplasm where it can stimulate cell signalling through interaction with several host proteins^92,148,149, including the tyrosine kinase c-MET receptor^150-152. CagA exerts pro-proliferative effects within host cells, by inducing hyperproliferation and disrupting apical junctional complexes and cellular polarity^153-155. Suzuki et al.¹⁵⁶ demonstrated that H. pylori infection resulted in CagA CM motifs interacting with c-MET, leading to sustained PI3K–AKT signalling, which in turn resulted in β-catenin activation and cellular proliferation. We have shown that CD44 variant 6 (CD44v6) acts as a co-receptor for the function of c-MET in response to H. pylori infection and bacterial-induced epithelial proliferation¹⁵⁷. CD44v9 has a central role in malignant transformation and chemoresistance in GAC¹⁵⁸. Within GACs, CD44v9 has been reported to interact with the cystine–glutamate transporter xCT, thereby potentiating defence against ROS and promoting subsequent tumour growth¹⁵⁸.

Using human organoid cultures, H. pylori-induced expression of the immune checkpoint molecule programmed cell death 1 ligand (PD-L1; also known as CD274) was shown to be mediated by the Shh signalling pathway¹⁵⁹. In this study, CD44v9⁺ SPEM cells within the gastric epithelium of human tissue-derived or iPSC-derived organoids specifically expressed PD-L1¹⁵⁹. Collectively, this study demonstrates that H. pylori-induced PD-L1 expression is an acute gastric epithelial response to infection in human cells, as previously observed in mice. By contrast, in a mouse model of PD-L1 expression, Wang and co-workers reported that the effectiveness of conventional platinum-fluoropyrimidine chemotherapy increases when PD-L1-expressing myeloid-derived suppressor cells (MDSCs) are depleted prior to treatment¹⁶⁰. Their findings suggest that the site of PD-L1 expression, specifically when overexpressed on immune cells rather than epithelial cells, might improve the response to checkpoint inhibitors and we suggest that the level of PD-L1 expression on myeloid cells has potential as a predictive or prognostic biomarker. Therapeutic trials in GAC that include immune checkpoint inhibitors such PD-L1 in combination with standard chemotherapy are ongoing, but have not significantly improved the overall response rate¹⁶¹. These results highlight the need for GACs to be better stratified according to both histologic phenotype and genomic subtypes as well as the immune cell composition^36,162,163.

Chronic gastritis

Chronic gastritis occurs when the inflammatory process fails to resolve ². If the inflammation persists, parietal cells undergo apoptosis, atrophy ensues (chronic atrophic gastritis) and metaplastic cells emerge in their place ^164,165.

Parietal cell death

Elevated levels of IL-1β are a consistent feature of both the acute and chronic immune microenvironment that can mediate transition of the mucosa from gastritis to atrophy^26,166-168. As IL-1β suppresses Shh gene expression in parietal cells by inhibiting acid secretion ¹⁶⁷, this finding suggests that parietal cells are key cellular targets that can trigger an imbalance in the immune response, tipping the mucosa towards metaplasia. However, Burclaff et al. showed that the targeted destruction of parietal cells via diphtheria toxin-mediated apoptosis in the absence of inflammation was not sufficient to induce gastric SPEM, consistent with the idea that specific cytokines have a central regulatory role in initiating the transition from atrophy to metaplasia¹⁶⁹. This finding indicates that the induction of SPEM might require a mechanism of parietal cell destruction that induces inflammation. In contrast to the 6-month timeline for Helicobacter-induced gastritis to occur in mice¹⁷⁰, chemical induction of SPEM with the protonophore DMP777 or its analogue L-635 is much more rapid, promoting the transdifferentiation of immature chief cells to become metaplastic once parietal cells have atrophied¹⁷¹. The L-635 analogue induces a more aggressive inflammatory response than DMP777 and results in full-blown SPEM within 3 days (versus 14 days with DMP777)¹⁷². DMP777 has neutrophil elastase inhibitory activity, a chemical property that L-635 lacks¹⁷³, which explains why the L-635 analogue has a more intense inflammatory response than DMP777 that is similar to a Helicobacter infection¹⁷⁴. Studies indicate that tamoxifen also acts as a parietal cell protonophore that causes a backwash of secreted acid into parietal cells, triggering cell death without prominent inflammation, similar to DMP777^175,176. L-635, DMP777 and tamoxifen can all induce SPEM, but only L-635 induces notable gastritis and causes the most extensive SPEM changes of the three drugs ¹⁷⁴. Bulk RNA-Seq profiling of macrophage markers in the corpuses of L-635-treated mice showed M2-like polarization with increased expression of Fizz1, Klf4 and IL-33¹⁷⁷. The presence of inflammation has been shown to be required for intestinal metaplasia or SPEM to progress to dysplasia in both humans and in mouse models¹⁷¹. These studies demonstrate that inflammation is essential for the epithelium to progress to SPEM, and that the composition of the immune environment (whether caused by Helicobacter infection or by the chemical destruction of parietal cells) is one that favours infiltration of M2-like TAMs and the inhibition of effector T cells. As not all mechanisms of parietal cell death induce SPEM, it is important to understand how diphtheria toxin-mediated death differs from chemical or inflammatory-mediated mechanisms¹²⁹.

Antral inflammation and gastrin

Compared with Helicobacter-infected mice where histologic changes start in the corpus of the stomach ¹⁷⁴, H. pylori infection of Mongolian gerbils or chemical carcinogen treatment with the alkylating agent N-methyl-nitrosourea (MNU) more closely mimics the Correa cascade where inflammation in the gastric antrum predominates and leads to GAC^178,179. The predominance of inflammation in the antrum is noteworthy since non-cardia cancer in humans typically develops in response to gastritis originating in the gastric antrum that then migrates proximally to the corpus^30,132. Another possibility is that gastric atrophy originating in the antrum eliminates gastrin-expressing (G) cells and gastrin peptide, a growth factor for parietal cells, which in turn contributes to inflammation-driven atrophy in the corpus ^180-182. Although most mouse models of SPEM have focused on events in the gastric corpus¹⁷⁴, the transition of atrophy to metaplasia in the antrum follows a different pathway linked to the presence or absence of the G cell as studied in the gp130 and Tff1 mouse models ^183-186. During non-atrophic gastritis, G-cell density and serum gastrin levels are initially elevated in response to a decrease in somatostatin-expressing antral (D) cells initially reported for human subjects ^187,188. Owing to the acid-mediated feedback regulation of gastrin ¹⁸⁹, corpus-dominant atrophy with preservation of the antrum correlates with hypergastrinaemia^190,191. However, in humans, corpus atrophy from autoimmune gastritis (AIG) is more strongly associated with gastric carcinoids (G-NETs) than with GAC (Box 1) ¹⁸⁹. By contrast, antral-predominant H. pylori-induced atrophic gastritis leads to decreased gastrin levels ^192-194. Therefore, modelling the inflammatory microenvironment that inhibits gastrin is critical to our understanding of distal GAC development and might be useful for distinguishing between the induction of G-NETs and GAC.

BOX 1: Autoimmune gastritis: a tale of two phenotypes.

Autoimmune gastritis (AIG) occurs only in the gastric corpus and fundus and is caused by generation of anti-parietal cell or anti-intrinsic factor antibodies, and subsequent destruction of the oxyntic glands. Destruction of the parietal (oxyntic) cells results in a rise in the pH of the stomach, which stimulates production and secretion of the hormone gastrin from G cells located in the gastric antrum. Since parietal cells produce intrinsic factor and acid, which are essential for the absorption of vitamin B12 and iron, autoimmune gastritis is frequently associated with anaemia. Prior to understanding the aetiology of AIG, anaemia associated with parietal cell atrophy was termed ‘pernicious anaemia’³³⁸.

Relevant to the TIME, AIG is associated with GAC and type 1 gastric neuroendocrine tumours (G-NETs). G-NET incidence (4–12%) is slightly more frequent than the association of AIG with GAC (0.9–9%) ³³⁸. Moreover, the overall risk for developing GAC from AIG is lower than the association of H. pylori with GAC ³³⁹, possibly because of the lower prevalence of AIG when H. pylori organisms are present^338,340. Thus, the epidemiology suggests that firstly, the inflammatory microenvironment generated by AIG is sufficient to generate both G-NETs and GAC and secondly, that the immune microenvironment of AIG differs significantly from H. pylori-induced GAC. An explanation for the differences in tumour type might be related to elevated serum gastrin, which in humans arguably targets the enterochromaffin-like (ECL) cell to a greater extent than non-endocrine epithelial cells ^341,342. The autoreactive T-cell responses in AIG are T_H1 and T_H17-mediated, which produce IL-17A, IL-21, TNFα and IL-11 (an IL-6 family member) that contribute to the development of gastric atrophy ^128,343. Thus, the cytokine profile does not seem to be a distinguishing feature of AIG-mediated GAC pathogenesis. However, aside from the role of elevated gastrin levels, TLR5 polymorphisms such as rs5744174 might explain the differential impact of whether AIG results in GAC development³⁴⁴. In summary, H. pylori does not seem to be a key risk factor for AIG, but instead exclusive corpus inflammation, sparing of the antrum and elevated gastrin levels favour the formation of G-NETs to a greater extent than GAC^345,346.

Loss of Tff1 and constitutive activation of the IL-6–IL-11 co-receptor GP130 both inhibit gastrin expression and induce tumours in the gastric antrum via STAT3^184-186,195. Tff1 becomes epigenetically silenced in MNU-dependent antral tumours, which in turn silences the expression of gastrin¹⁹⁶. In addition, gastrin-null mice develop dysplastic antral tumours and show elevated levels of IL-1β, IFNγ, IL-11, IL-6 and STAT3^181,182,197. Gastrin administration overrides the effect of MNU-induced antral cancers by inducing TFF1 expression ^196,198. Loss of Apc gene function or MNU treatment also induces tumourigenesis in the antrum but not the corpus by increasing the activity of the Cxcl12–Cxcr4 signalling axis¹⁹⁹. Β-catenin overexpression induces the chemokine CCL28, which contributes to immune suppression in the antrum by recruiting regulatory T cells (T_reg cells)^200-202. We conditionally overexpressed IFNγ and IL-1β in the gastric antrum of mice and found that IFNγ did not suppress gastrin but that IL-1β did suppress gastrin, suggesting that these two proinflammatory cytokines exert differential effects in the antrum¹⁶⁸. Conditional overexpression of IL-1β in the antrum led to suppression of parietal cell acid secretion through to the suppression of gastrin gene expression, while IFNγ overexpression in the antrum induced gastric acidity by blocking local non-canonical HH signalling via GLI2 and increased G cell numbers and gastrin peptide secretion^168,203. Therefore, as already observed in humans^189,192,193, antral inflammation in mice also modulates gastrin peptide levels depending on the type of inflammatory cytokine ¹⁶⁸.

Chronic gastritis to metaplasia

The transition of atrophy to intestinal metaplasia in the inflamed stomach is the most visible indication that chronic inflammation has induced a phenotypic change in the epithelium’s response to injury. Various reports indicate that the signalling pathways activated by H. pylori virulence factors induce CDX1 and CDX2, which are caudal-homeobox transcription factors essential to regulating intestinal identity and intestinal metaplasia⁹⁶. Genomic profiling of intestinal metaplasia samples detected more cases of active H. pylori infection than did simple histologic examination¹²³. The mere presence of H. pylori DNA, however, does not determine whether the bacteria or APC activation is ultimately responsible for the epithelial change. Whether H. pylori virulence factors or its byproducts alone are sufficient to induce intestinal metaplasia is unclear. However, the stochastic progression from chronic gastritis to intestinal metaplasia favours involvement of both mechanisms.

Genomic profiling of intestinal metaplasia

Genomic profiling of human gastric intestinal metaplasia revealed a low rate of clonal mutations compared with GAC; however, the tumour suppressor gene FBXW7 (which encodes an E3 ubiquitin ligase required for cMYC and cyclin E1 (encoded by CCNE1) protein degradation) showed missense point mutations in 4.7% of intestinal metaplasia samples (compared with 9.2–18.5% of GAC samples¹²³) (Fig. 1). FBXW7 dominant negative mutations are increased in GAC compared to normal gastric epithelium, suggesting aberrant ubiquitin-mediated protein degradation¹²³. Normally, wild-type FBXW7 promotes CDX2 ubiquitination and degradation when it docks at the WD domain of FBXW7²⁰⁴. Fbxw7^−/+ mice treated with the alkylating agent N-methyl-nitrosourea (MNU) develop antral intestinal metaplasia and dysplasia within 8 weeks, which correlates with an increase in cMYC²⁰⁵. Thus, FBXW7 can suppress expression of CDX2 through protein degradation. Interestingly, cMYC and cyclin E1 are the same gene targets elevated in the immunologically cold chromosome instability GAC subtype³⁵. Dominant-negative gastric intestinal metaplasia mutations in the WD40 domain of FBXW7 correlates with an increase in the immunohistochemical expression of cMYC. TP53 and ARID1A exhibited clonal mutations in GAC but not in intestinal metaplasia, whereas hypermethylation was more prevalent in intestinal metaplasia than in GAC¹²³. Surprisingly, genomic analysis of GACs revealed that the CDX2 gene exhibits LOF mutations in GAC but that CDX2 amplification is present in oesophageal and colon adenocarcinomas²⁰⁶. Taken together, CDX2 protein expression can increase or decrease during the transition from intestinal metaplasia to GAC through increased protein turnover or altered epigenetic status (for example, DNA methylation or demethylation) of this locus²⁰⁷.

Gastric metaplasia in rodent models

Rodent models have provided detailed insight into GAC pathogenesis by mimicking the molecular steps occurring as the mucosa transitions from chronic gastritis through the intermediate lesions to cancer ^170,208,209. Mapping the timeline of H. pylori infection from inoculation to SPEM revealed that infection in the mouse gastric corpus triggers the release of the sonic hedgehog (SHH) ligand from parietal cells into the circulation within 2 days, where it is sensed by circulating monocytes²¹⁰. Parietal cell release of SHH is therefore one mechanism that contributes to the homing of myeloid cells to the infected stomach. During the acute T_H1 response, IFNγ-driven inflammation increases²¹¹. After 2 months, the inflammatory milieu, specifically IL-1β, suppresses parietal cell expression of SHH and subsequently reduces gastric acidity¹⁶⁷. H. pylori induction of SHH expression within parietal cells is mediated by NFκB signalling^208,210. Furthermore, SHH directly stimulates the expression of H⁺,K⁺-ATPase through PKC-mediated signalling^212,213. Thus, even before parietal cell atrophy occurs, inflammation-driven signalling suppresses parietal cell acid secretion¹⁶⁷.

By 6 months, the gastric mucosa exhibits SPEM characterized by the expression of HE8, CD44v9, trefoil factor 2 (TFF2; also known as spasmolytic polypeptide (SP)). SPEM was first reported as a gastric mucous cell lineage expressing TFF2 in mice infected with Helicobacter felis²¹⁴. In a 1999 review of archival samples from 22 resected GACs, Schmidt et al. confirmed that SPEM was present in most cases, typically within the corpus of the human stomach bordering the GAC²¹⁵. More recent studies suggest that SPEM precedes intestinal metaplasia development in the human stomach²¹⁶. Gastric glands in the antrum and mucous neck cells of the corpus express TFF2^172,216,217 (Fig. 2). SPEM-expressing glands have subsequently been reported to express other markers including the GAC stem cell marker CD44v9^159,218, a target of HH signalling^219,220.

Central to the debate regarding GAC cell origin is the simple question of whether GAC arises from a metaplastic cell, a non-metaplastic stem cell, or both? Since Helicobacter infection alone in mice does not lead to progression to GAC, studying the linear progression from metaplasia to GAC is difficult to test experimentally in vivo. Perhaps the best example of non-linear progression from an inflamed gastric mucosa to cancer is non-hereditary diffuse GAC, which typically does not progress through the intestinal metaplasia histologic intermediate and could serve as an example of accumulating stem cell mutations²²¹. This pattern of progression to GAC would support a stem cell that accumulates mutations under the inflammatory milieu initiated by a H. pylori infection. As discussed by Huang et al, the inflammatory microenvironment induces epigenetic changes, particularly alterations in DNA methylation that contribute to the accumulation of clonal mutations leading to GAC¹²³.

Role of Hedgehog in immune cell signalling and metaplasia

SHH is a morphogen that has a key role in homeostasis of the stomach in addition to being involved in the transition of the gastric epithelium to GAC²²². Originally identified for its role in cell patterning during embryonic development, SHH ligand is expressed in a number of adult organs, including the stomach^223,224. In 2000, Ramalho-Santos et al. reported that Shh-null mice exhibit gastric metaplasia at birth²²⁵, indicating that Hedgehog (HH) signalling has an essential role in gastric homeostasis as well as in foregut malignancies^222,226. Subsequent studies of H. pylori-infected individuals demonstrated a reciprocal relationship between SHH and CDX2 expression in which the SHH ligand was elevated in diffuse GAC²²⁷, while, loss of SHH correlates with the aberrant overexpression of CDX2^228,229. One of the main aetiologies associated with chronic H. pylori infection and gastritis is atrophy of the acid-secreting parietal cells and the resulting loss of the major epithelial reservoir of SHH ligand²³⁰.

Suppression of Shh in the inflamed gastric corpus is an early change, which in the mouse occurs within 2 months of the infection and prior to parietal cell atrophy^167,231. Although loss of acid secretion is not synonymous with parietal cell atrophy, it does reduce Shh expression¹⁶⁷. CDX2 associated with intestinal metaplasia and intestinal-type cancers inhibits SHH expression²³². Thus, loss of the SHH ligand correlates with an intestinal-type GAC phenotype (Fig. 1). By contrast, diffuse-type GACs, which typically do not progress through an intestinal metaplasia phenotype, express the SHH ligand, its receptor PTCH and its downstream target CD44v, suggesting autocrine regulation of cell growth by Hedgehog (HH) signalling in this cancer subtype^233,234. YAP has been shown to regulate GAC growth through the HH effector protein GLI1, independently of SMO activation, consistent with non-canonical HH signalling²³⁵. In addition, non-canonical HH signalling activates RHOA²³⁶, the proto-oncogene mutated in some diffuse GACs. Although acute H. pylori gastritis is associated with elevated SHH secretion^{208,210,213,237-239}, the reduction in SHH expression during chronic gastritis is associated with disruption of normal cell differentiation in the adult stomach and reduced parietal cell numbers^{167,231,240,241}. By contrast, in a mouse model of Helicobacter infection, El-Zaatari et al. showed that NFκB mediates re-expression of SHH, which correlates with development of SPEM²⁴². Taken together, HH signalling either through ligand-receptor dependent (canonical) or ligand-receptor independent mechanisms (non-canonical) has been implicated in both histologic subtypes of GAC pathogenesis²⁴³.

Canonical HH signalling begins with de-repression of the G-protein-coupled receptor (GPCR) Smoothened (SMO) once the HH ligand binds its receptor Patched (PTCH), another GPCR²²³. Of the three glioma (GLI)-associated zinc finger transcription factors activated by the pathway, GLI1 is widely accepted as an indicator of HH signalling²⁴⁴. In the uninfected stomach, GLI1 is expressed in myofibroblasts²⁴⁵; however, during H. felis infection, Gli1⁺-myeloid cells infiltrate the mucosa²⁴⁶. Moreover, H. felis-infected Gli1-deficient mice (Gli1^+/LacZ (Gli^+/−) and Gli1^LacZ/LacZ (Gli1^−/−) do not develop SPEM at 6 months despite the presence of inflammation, indicating that Gli1⁺-myeloid cells are required for chronic inflammation to progress to SPEM²⁴⁶. Gene profiling of wild-type versus Gli1^LacZ/LacZ mice revealed elevated levels of several myeloid differentiation factors (for example, Slfn2 and Slfn4) at 6 months, but not at 2 months²⁴⁶. Between 2 months and 6 months, parietal cell atrophy occurs and SPEM develops, coincident with a shift in the immune microenvironment characterized by an increase in DAMP-induced type 1 cytokines, IFNβ and IFNα¹¹⁸ (Fig. 2). Lineage tracing revealed that the SLFN4⁺ cells are a subset of granulocytic or polymorphonuclear (PMN)-MDSCs that originate in the bone marrow and acquire their T-cell suppressive phenotype in the stomach at the same time as the emergence of SPEM at 6 months¹¹⁸. The level of Slfn4 expression reflects the presence of constitutive HH signalling, since Slfn4 was absent in myeloid cells from the infected Gli1 mutant mice²⁴⁶. Using chromatin immunoprecipitation, we showed that GLI1 binds to the Slfn4 promoter²⁴⁶. Both human (SLFN12L) and mouse (Slfn4) Schlafen loci are strongly induced by type 1 interferons, especially IFNα^118,247, which explains the peak expression of Slfn4 in this subset of Cd11b⁺Gr1 MDSCs^118,125. More recently, RNA-sequencing analysis of Slfn4⁺ MDSCs sorted by flow cytometry from H. felis-infected stomachs demonstrated elevated expression of Nos2, Arg1, Tnfα and Il-1α transcripts compared to Slfn4⁺ cells in the bone marrow or spleen²⁰³. Moreover, these Slfn4⁺ MDSCs secrete MiR130b, a microRNA that suppresses tumour suppressor genes such as Trp53INP1, Cyld, Runx3 and Pten²⁰³. Lineage tracing of this subset of HH-dependent myeloid cells enabled the identification and temporal tracking of their polarization within the H. felis-infected stomach and demonstrate how DAMP and HH signalling can shape the immune microenvironment during the progression from chronic gastritis to SPEM²⁰³.

Transgenic expression of proinflammatory cytokines

Various mouse models have been used to determine the impact of specific proinflammatory cytokines on parietal cell atrophy. The evidence from human stomachs supports the observation that the pivotal transition of the epithelium towards transformation occurs with incomplete intestinal metaplasia (IIM) rather than with complete intestinal metaplasia (CIM) beginning at the incisura on the lesser curvature of the stomach with the metaplasia eventually extending bi-directionally into both the corpus and the antrum^248,249. The dysplastic, hyperproliferative cells that overtake the mucosa presumably originate from the intestinal metaplasia in response to cells within the TIME that produce cytokines, growth factors and regulatory RNAs^29,250. A major factor predisposing humans to neoplasia is their susceptibility to elevated levels of proinflammatory cytokines, especially IL-6, TNF and IL-1β^{25,26,251,252}. In 2000, polymorphisms in the IL-1β promoter of patients with GAC were associated with increased circulating levels of this cytokine²⁶. Subsequently, Tu et al. demonstrated that parietal cell overexpression of human IL-1β in mouse parietal cells using the H⁺, K⁺-Atpase-hIL-1β transgene induced corpus atrophy and dysplasia, confirming a pathogenic role for IL-1β. These researchers also reported that IL-1β is involved in the recruitment of Cd11b⁺Gr1⁺-MDSCs to the stomach ¹⁶⁶. Macrophages recruited to the site of infection secrete the pro-inflammatory cytokine IL-1β²⁵³, which leads to inhibition of Shh expression¹⁶⁷ and loss of normal epithelial cell differentiation²³⁷. IL-1β inhibits gastric acidity but IFNγ does not, suggesting that a T_H1 response characterized by IFNγ maintains gastric acid production and might explain why the acute inflammatory response to H. pylori is linked to duodenal ulcer development¹⁶⁷.

In H⁺,K⁺-Atpase-Ifnγ transgenic mice, TNF, IL-10, IL-6, IL-1β levels increased at 7 weeks remained elevated at 15 months, while IL-11 levels increased only at 15 months coincident with SPEM development²⁵⁴. The stomachs of H⁺,K⁺-Atpase-Ifnγ mice showed upregulation of phosphorylated STAT3 compared with the stomachs of control mice. Both IL-6 and IL-11 are important inducers of STAT3 phosphorylation and IL-11 has been showed to induce parietal cell apoptosis²⁵⁵. Transgenic expression of IFNγ from parietal cells or gastric organoids treated with IFNγ stimulates apoptosis, contributing to the progression of gastritis to atrophy and SPEM^254,256. IFNγ^−/− mice do not develop severe gastritis following H. pylori infection²⁵⁷, whereas infected IL-4^−/− mice develop more severe gastritis compared to wild-type mice. Furthermore, a model of autoimmune gastritis in which gastritis is induced by CD4⁺ cells that are autoreactive against the H⁺,K⁺-ATPase antigen induces T-cell mediated destruction of parietal cells^128,129,258 . In this autoimmune gastritis model, IL-17A is an important mediator of parietal cell death¹²⁸. Taken together, either cytokine-induced apoptosis of parietal cells or T-cell-mediated autoimmune destruction of parietal cells is sufficient to induce inflammation and initiate gastric atrophy and SPEM ¹²⁹.

Immune microenvironment and immune escape

A unique feature of H. pylori infection is that the bacteria itself orchestrate immune suppression through the early recruitment of T_reg cells and myeloid cells to the stomach, in part as a result of the composition of its DNA, LPS and other bioactive molecules^15,101. T_reg cells, DCs and pro-inflammatory M1 macrophages, like neutrophils, are early inhabitants of the H. pylori-infected stomach^17,211. Single-cell RNAsequencing of human GACs has been limited by the low number of samples, but has consistently shown active TNFα–NFκB and IL-6–JAK–STAT3 cytokine signalling^259-261.

H. pylori and mechanisms of immune escape

The acute immune response to H. pylori consists of a mixture of neutrophils, pro-inflammatory M1 macrophages and CD4⁺ T lymphocytes that produce cytokines, which induce IFNγ, IL-17, IL-21 and IL-22^211,262,263. IL-21 is required for the expansion and maintenance of the T_H1 and T_H17 responses ²⁶⁴. Ultimately, the balance between T_H1, T_H17 and immune suppressor cells such as T_reg cells and tumour associated macrophages (TAMs) determine the switch from acute to chronic inflammation²⁶¹, which is mediated in part by an increase in indoleamine 2,3-dioygenase-1 (IDO1) from DCs²⁶⁵. H. pylori eradication only moderately reduces GAC risk^266-268 because inflammation persists in the gastric mucosa²⁶⁹. A 20-year follow-up study of pre-neoplastic lesions showed that IIM, when extended into the corpus, increases the risk of GAC compared with CIM, gastritis and antrum-restricted intestinal metaplasia¹³². However, the immune cell population was not analysed in that study. We performed a limited analysis of the immune microenvironment in tissue from a 13-year follow-up study of patients with GAC and found that increased SLFN5 expression in immune cells is associated with progression to GAC in patients with IIM²⁷⁰. Single-cell RNA sequencing of GAC as well as precursor lesions such as atrophic gastritis and metaplasias has revealed extensive heterogeneity within immune subpopulations^259,261.

In individuals infected with H. pylori, the acute (neutrophilic) gastritis can progress to a chronic (monocytic) infiltrate²⁷¹. H. pylori can initiate the acute immune response in a number of ways: through the activation of pathogen-activated molecular patterns (PAMPs) by bacterial antigens such as LPS and urease; through the injection of CagA into host cells with the subsequent induction of intracellular protein phosphorylation leading to changes in cell mobility; through the detection of soluble H. pylori antigens by resident APCs; and through the release of soluble factors from injured epithelial cells (for example, Shh from parietal cells)^{100,118,210,272-274}. The transition from acute to chronic gastritis is orchestrated by a cascade of cytokines (including hepatocyte growth factor (HGF), IL-4 and macrophage colony-stimulating factor (M-CSF), which promotes polarization to resolve the inflammation and contribute to tissue remodelling⁹⁹. The acute inflammatory response segues to a chronic infiltrate comprised of T_reg cells (FOXP3⁺CD4⁺CD25^hi), anti-inflammatory macrophages (M2 macrophages) and mononuclear cells^99,275. Differences in the ratio of inflammatory cytokines initiated by H. pylori-specific molecules (for example, bacterial DNA and LPS) skew the cytokine profile to one that favours a muted immune response¹¹⁹. IDO1, a metabolite of tryptophan, is produced by activated DCs, and has a key role in creating an environment of immune tolerance and escape^276,277. In addition to DCs, IFNγ induces IDO1 production by fibroblasts, macrophages, endothelial and stromal cells²⁷⁶. IDO1 induces differentiation of CD4⁺ T lymphocytes into T_reg cells and promotes TAM polarization²⁷⁶. Thus accumulation of IFNγ also polarizes the cell population that will terminate the T_H1 pro-inflammatory response.

GAC and mechanisms of immune escape

Once GAC is established, the tumours themselves sustain the immune suppressive environment, which enables them to escape immune surveillance^123,206. Specifically, PD-L1, PD-1 and CTLA-4 are a subset of immune checkpoint receptors expressed on GAC and immune cells in the neoplastic microenvironment ^278,279. These immune checkpoint receptors are targets for immunotherapy now approved for use in GAC²⁸⁰. Tumours can evade immune surveillance by expressing PD-L1 that interacts with the T-cell protein PD-1, subsequently inhibiting CD8⁺ cytotoxic T-lymphocyte proliferation, survival and effector function^281-283. Inhibition of the PD-1–PD-L1 axis has produced effective response rates in various malignancies, including melanoma, renal and non-small cell lung cancer^284-287. Despite the fact that PD-L1 is expressed in ~40% of GACs, the overall 5-year response rate to standard chemotherapy with or without checkpoint inhibitors is low at ~30%^80,288-290. A plausible explanation for this low response rate is the overlap of immune suppressive mechanisms. Specifically, multiple immune suppressor populations, such as TAMs, tumour-associated neutrophils (TANs), MDSCs, regulatory B cells (B_reg cells) and T_reg cells, accumulate in GAC²⁹¹ and facilitate tumour evasion by preventing cytotoxic T cell from attacking the tumour. Moreover, chemotherapy can also induce PD-L1 expression on GAC cells, reducing the effectiveness of immunotherapy when combined with standard chemotherapy regimens such as gemcitabine plus 5-fluorouracil¹⁶⁰.

Helicobacter-mediated mouse models typically do not progress past the metaplastic stage (SPEM) 4–6 months after the infection is initiated depending on the genetic background ¹¹⁸, which has permitted extensive examination of how various suppressive immune populations interact with the epithelium during this process¹²⁵. In humans, best estimates suggest a 6–10 year period for the initial immune response to H. pylori to polarize the inflammatory environment and alter epithelial cell fate¹³². Reports in humans indicate that the suppressive microenvironment exists before cancer develops^{80,123,260,261}, a concept that has been confirmed in mouse models ^118,125,170. Cytokines that are commonly detected prior to and after GAC develop include IL-1β, TNF, IL-6, TGFβ, type 1 interferons and WNTs. IL-6 and TNF secreted by TAMs induce PD-L1 expression on GACs through NFκB and STAT3 signalling²⁹². TGFβ has a critical role in shaping the GAC immune microenvironment by inducing B_reg cells, (CD19⁺CD24^hiCD38^hi), a subset of B lymphocytes that express IL-10, suppress T_H1/T_H17 proinflammatory mediators and promote the conversion of CD4⁺CD25⁻ effector T cells to CD4⁺CD25⁺FoxP3⁺ T_reg cells^293,294. TGFβ1, a potent inducer of Foxp3 expression²⁹⁵, promotes T_reg cell development²⁹⁶. Researchers have also shown that extracellular vesicles from tumours laden with HMGB1, a potent DAMP–TLR9 activator, induce neutrophils to suppress T-cell immunity by secreting CCL17²⁹⁷. Moreover, MDSCs and TAMs express regulatory RNAs (for example, microRNAs), which can be packaged and released from extracellular vesicles and modulate pre-cancerous gastric epithelial cells by targeting multiple signalling pathways^203,298 (Fig. 3).

Figure 3 ∣. Differentiation and cell signatures of common immune cells recruited to the stomach in response to H. pylori infection.

Within the bone marrow, hematopoietic stem cells (HSCs) differentiate into common myeloid progenitor cells (CMPs) that then give rise to granulocyte-macrophage progenitors (GMPs). GMPs differentiate into myeloblasts, promyeloblasts, myelocytes and granulocytes in response to granulocyte colony-stimulating factor (G-CSF). By contrast, macrophage colony-stimulating factor (M-CSF) induce GMP differentiation towards monocytes, macrophages or dendritic cells via monocyte/macrophage and dendritic cell progenitors (MDPs)³³⁰. In response to specific environmental cues, hemangiocytes, TIE2-expressing monocytes (TEMs), monocytes, dendritic cells, MDSCs, neutrophils, eosinophils and immature mast cells all originate from the same HSC and circulate within the blood. Once recruited to the stomach in response to H. pylori infection or tissue injury, these immune cells express specific cell signatures that orchestrate neovascularization, create an immune suppressive environment and lead to development of metaplasia ^{177,203,210,217,323,331-337}. APC, antigen- presenting cell; IM, intestinal metaplasia; iNOS, inducible nitric oxide synthase; LPS, lipopolysaccharide; M- MDSC, monocytic myeloid- derived suppressor cell; PMN, polymorphonuclear; SPEM, spasmolytic polypeptide- expressing metaplasia.

During the initial stages of an H. pylori infection, disruption of the epithelial apical-junction complex by the bacteria¹⁵³ likely facilitates migration of DCs and their luminal and subepithelial interactions with CagA and other bacterial antigens^299-301. H. pylori outer membrane proteins such as Omp18 and HpaA, intact bacteria and bacterial membrane preparations have all been shown to induce maturation of DCs in cell culture models^302-304. Once activated, different subsets of DCs initiate either T_H1, T_H2 or T_reg cell responses³⁰⁵. Mature DCs secrete IL-12 and IL-10 when stimulated by H. pylori^306,307. Classic or conventional DCs (cDCs) are matured by TLR2-6 and TLR8 and induce T_H1 effector cells, while DAMPs activate TLR7 and TLR9 which polarize pDCs^115,308. In response to H. pylori infection, tolerogenic pDC populations that express CD11c, DC-SIGN and CDH1 expand, secrete type 1 interferons and prime T_reg cells, which support immune evasion^119,309. As with other immune cell types, single-cell profiling has revealed multiple T_reg cell subsets, not all of which are immunosuppressive ^310,311. In gastric tumours, T_reg cell suppression requires the presence of the TCR-inducible co-stimulatory molecule (ICOS)³¹². Recruitment of immunosuppressive ICOS⁺ T_reg cells requires activation of pDCs, which as previously stated is activated by either H. pylori DNA, unmethylated host DNA or specific molecules released from damaged epithelial cells¹⁰¹. Therefore, the subtype of DCs activated likely contributes to the persistence of suppressive T_reg cells, which shifts the immune microenvironment from a pro-inflammatory environment to immune quiescence. In contrast to T_reg cells, B_reg cells are polarized by TGFβ, with the likely source of this cytokine from cancer-associated fibroblasts (CAFs), M2-like TAMs and the cancer itself³¹³. In addition to monocytic-MDSCs and PMN-MDSCs³¹⁴, a third category of MDSCs in GAC known as immature MDSCs (iMCs) has been reported³¹⁵.The lectin-type oxidized LDL receptor 1 (LOX-1) is a unique marker for human polymorphonuclear (hPMN)-MDSCs³¹⁶. hPMN-MDSCs upregulate genes associated with ER stress; for example, transcription factor genes encoded by spliced XBP1 and CHOP³¹⁶. LOX⁺CD15⁺ hPMN-MDSCs are activated by endoplasmic reticulum stress and accumulate in the peripheral blood of patients with hepatocellular carcinoma³¹⁷. In patients with gastric and pancreatic cancer, PMN-MDSCs accumulate in the spleen and can suppress T-cell proliferation, unlike PMN-MDSCs found in the peripheral blood, which do not display T-cell suppressor activity³¹⁸. The conversion of mature neutrophils to PMN-MDSCs in the spleen can occur via two signalling pathways that induce immune suppressive activity in these cells. One of the pathways expands immature myeloid cells by inhibiting their terminal differentiation while the other pathway reprogrammes immature myeloid cells into suppressor cells ^319,320.

As immunosuppressive cells, TAMs persist within the gastric immune microenvironment following H. pylori-mediated chronic inflammation because DCs and other immune cells express IDO1^321,322. Single-cell RNA sequencing of the gastric TIME revealed that the macrophage population is quite heterogeneous and does not necessarily conform to categorizing them simply as M1 or M2²⁶¹. Mouse models have demonstrated that M2-polarized macrophages emerge with the appearance of SPEM³²³ and are associated with IL-33-mediated induction of IL-13 signalling^177,324, demonstrating that M2-like polarization in mice occurs during the transition of the mucosa to SPEM. Taken together, M2-polarized macrophages likely create a suppressive immune environment before GAC develops, in part because of arginase1 (ARG1) catalyses the hydrolysis of L-arginine, an essential substrate for T-cell proliferation^325-327.

Conclusions

Bacterial and viral infections collectively have a key role in the initiation of GACs and early detection, eradication and prevention of these infections is therefore important. Most people carrying these infectious agents will not develop symptoms, and only a small but clinically significant percentage of those infected will develop complications. By delving deeper into the suppressive nature of the immune response and understanding host risk factors, we should eventually be able to predict which individuals are at risk of these complications and develop biomarkers to detect the epithelial changes early and/or prevent progression to GAC. EBV-positive and MSI-H GACs represent a small subset of GACs that generate a robust inflammatory response from neo-antigens. The generation of neo-antigens and the type of immune response that results makes EBV-positive and MSI-H GACs more susceptible to chemotherapies in combination with immune checkpoint inhibitors^58,76,85. In comparison, most GACs are immunologically ‘cold’ and include cancers initiated by LOF mutations in CDH1 and other tight junction protein components (diffuse-type GACs) and intestinal-type GACs initiated by H. pylori infection^57,58. The fact that H. pylori-induced GACs are so unresponsive is somewhat surprising, but possibly results from the organism initiating immune evasive strategies^119,328. A central question is by what mechanism does the organism dampen the immune response, which benefits its persistence yet allows GAC to emerge. In short, both bone-marrow-derived and tissue resident cells become activated and over time differentiate into immune suppressive phenotypes under the instruction of the microenvironment before GAC develops. The type and extent of the metaplasia is a crucial inflection point. Although H. pylori DNA persists in intestinal metaplasia, what role it has, if any, is unclear¹²³. One possible explanation is that H. pylori DNA contains immunoregulatory sequences that contribute to immune suppression by inducing T_reg cells¹¹⁹. Another possibility is that the debris created by injured and dying cells (e.g., parietal cells) accumulates to a threshold level that eventually activates pDC DAMP signalling^125,217. Owing to the multiple immune cell types involved, their phenotypic plasticity and metabolic signatures will need to be temporally studied using single-cell RNA-sequencing and other unbiased analyses of the multiple immune cell populations recruited to the stomach prior to GAC. Understanding the transitions between atrophy, SPEM, CIM and IIM in the human stomach will provide insight into the conditions under which the heterogeneous immune populations develop and become dominant, since once GAC has emerged, the dampened immune environment has already existed for several years²⁶¹.