Abstract

INTRODUCTION

Gastric adenocarcinoma is the third leading cause of global cancer mortality, and the leading infection-associated cancer, driven by chronic infection with Helicobacter pylori.(1, 2) There is also remarkable geographic variability of gastric cancer (GC) with respect to both incidence(3) and mortality.(4) Latin America has among the highest GC incidence rates in the world. There is an excess burden in the mountainous regions along the Pacific littoral as compared with coastal populations, even though the two regions have high endemic H. pylori infection; this has been termed the “Latin American altitude enigma”.(3) Altitude appears to be a surrogate for the clustering in the mountain villages of high-risk interactions between host genetic and H. pylori genetic factors(5), with modulation by dietary and environment factors.(6)

In the United States, GC represents a marked cancer disparity, with an excess burden among minorities. The GC incidence is approximately doubled among non-whites, including Hispanics.(7) Notably, immigrants from high GC incidence areas are at-risk. Those who emigrate from high-to-low incidence regions maintain the risk of their nation of origin, most likely due to ‘importing’ their host and H. pylori genetic risks.(8)

The gastric carcinogenesis pathway is a multifactorial process that progresses from pre-malignant to malignant phenotypes through several histopathology stages: normal mucosa, chronic gastritis, atrophic gastritis, intestinal metaplasia, dysplasia, and adenocarcinoma.(9) Progression is driven by host genetics and H. pylori virulence and oncogenic factors, as well as dietary and environmental influences. H. pylori accounts for much of the attributable risk with cagA being the principal H. pylori risk genotype, and the CagA protein as the dominant virulence factor for gastric adenocarcinoma.(10-14)

Human germline mutations are thought to be important drivers in up to 10-15% of incident GC cases.(15) An elevated risk of GC has been associated with gene polymorphisms including those in the inflammation pathway, such as genes encoding IL-1β, TNF-α, and IL-10.(12, 16, 17) Additionally, associations between GC and variants in PSCA and MUC1 (18-20), and at a locus that includes PRKAA1 and PTGER4 (20, 21) have been detected in genome-wide association studies of GC in Asian populations. Few studies have explicitly examined the host genetic basis for GC in Latin American populations.(22) The interaction of genetic variants in both the host and H. pylori affect disease progression and may explain some of the geographic variation in GC risk, yet the specific human loci and their relationship to disease risk remain poorly understood.(5, 23, 24)

Knowledge of human cancer risk loci may identify at-risk populations and has potential for targeting chemoprevention. For example, in colorectal cancer with the adenoma precursor lesion, there are potential targeted treatments for chemoprevention such as those based on ornithine decarboxylase (ODC1) gene variants. The ODC enzyme converts L-ornithine to the first polyamine putrescine.(25, 26) Putrescine is sequentially converted into the polyamines spermidine and spermine via spermidine synthase (SRM) and spermine synthase (SMS), respectively; both reactions require the donation of an aminopropyl group from decarboxylated S-adenosylmethionine (dcSAM), which is synthesized by the sequential conversion of i) methionine to S-adenosylmethionine (SAM) by S-adenosylmethionine synthase (METK) and ii) SAM to dcSAM by S-adenosylmethionine decarboxylase (SAMDC), encoded by AMD1.(27) Polyamines are essential for cellular growth and survival and are important in the development of the GI tract and the immune system.

In humans, an ODC1 variant, rs2302615, has been associated with adenoma risk, as well as an augmented chemoprevention response to alpha-difluoromethylornithine (DFMO). Patients with the CC genotype have a higher risk of colon adenomas, yet are more responsive to DFMO and sulindac.(28, 29) This ODC1 SNP is located in intron 1, a region known to affect ODC1 transcription (28), but no data exists on the relationship of this polymorphism and human GC. However, this gene may also be relevant to GC as myeloid-cell specific deletion of the Odc1 gene in mice results in enhanced host immune response to H. pylori and diminished bacterial load in the stomach.(30) H. pylori-induced ODC activity is associated with macrophage apoptosis.(26, 31) In addition, studies have identified variants in ODC pathway genes that associate with GC.(32) Based on these prior studies, we hypothesized that ODC1, and rs2302615 in particular, which associates with transcription, will affect the risk for GC.

In addition to evaluating the ODC1 variant described above, we analyzed 93 candidate single nucleotide polymorphisms (SNPs) in 53 genes potentially linked to GC based on prior association studies of gastric cancer and gastritis.(33-39) These markers have largely been studied in European and East Asian populations, with few studies in high risk Latin American populations. They fall into classes related to mechanisms of inflammation, immunity, H. pylori colonization, and oxidative stress (40-44), as well as many of the dietary and environmental markers associated with disease, but some do not have a clear mechanism.(6, 45)

Our study aimed at determining patterns of genetic association in the mountainous regions of Western Honduras, where the GC incidence is among the highest in Latin America.(3) We assessed if the genotypes of the ODC1 SNP, rs2302615, were associated with GC in the high-risk Western Honduran population and in what context they were the most strongly associated. We also examined the other SNPs with known significance in GC and asked if putative association of the other SNPs was modified by the ODC1 genotype risk.

RESULTS

Of the 1500 individuals in this population-based, case-control study in Honduras, complete genetic data were available for 1366 participants. The mean ages of GC cases and controls were 63.8 and 53.6 years, respectively. Males comprised 70% (n= 468) and 50% (n= 350) of cases and controls, respectively. H. pylori and CagA seropositive cases were determined by a multiplex serology (see Methods). Overall, a high H. pylori prevalence was confirmed, detected in 90.5% and 88.2% of cases and controls, respectively. H. pylori CagA seropositivity was 87% and 73.6% in cases and controls, respectively, and strongly associated with GC in the unadjusted analysis (Odds Ratio, OR = 3.59; 95% CI 2.56, 5.11; p = 6.0 × 10⁻¹⁴; Table 1).

Table 1.

Selected characteristics of gastric cancer cases and population controls

Characteristics	Overall (N=1366)	Gastric Cancer Cases (N=664)	Healthy Control Population (N=702)	P-value
Age				< 2 X 10−16
Mean in years (SD)*	58.5 (16.1)	63.78 (13.7)	53.52 (16.5)
Sex, n (%)				1.34 X 10-14
Male	818 (60%)	468 (70%)	350 (50%)
Female	548 (40%)	196 (30%)	352 (50%)
H. pylori serostatus, n (%)				0.009
Positive	1220 (89.3%)	601 (90.5%)	619 (88.2%)
Negative	78 (5.7%)	26 (3.9)	52 (7.4%)
Missing	68 (4.98%)	37 (5.6%)	31 (4.4%)
CagA serostatus, n (%)				6.0 X 10-14
Positive	1095 (80.2%)	578 (87.0%)	517 (73.6%)
Negative	202 (14.8%)	48 (7.2%)	154 (21.9%)
Missing	69 (5.05%)	38 (5.7%)	31 (4.4%)

Open in a separate window

Notes:

^*Standard deviation (SD)

Of the 94 selected SNPs, one SNP, rs4072037, was out of Hardy–Weinberg equilibrium (p < 1.0 × 10⁻¹⁶ for both cases and controls), indicating a likely genotyping error, and was excluded from the analyses. In addition, 11 SNPs were excluded with a minor allele frequency of less than 0.05 (Table S1, Table S2). Linkage Disequilibrium (LD) was detected between 20 SNP pairs in which 4 SNPs appeared in two pairs. A total of 28 SNPs were excluded, resulting in 66 SNPs for the final analysis (Table S3, Figure S1). Using the Benjamini-Hochberg FDR multiple test correction with α = 0.1, resulted in a multiple testing threshold of p = 0.021 (raw value).

Cases were associated with ODC1 genotype considered as cancer risk (CC) or protective (CT/TT) based on the colon cancer literature using a chi-square analysis (p = 0.021).(46-48) In the unadjusted logistic regression model, the ODC1 SNP was associated with GC (OR=1.21, 95% CI 1.14, 1.80; p=0.027). In the model adjusted for age, sex, and CagA status, and FDR corrected for multiple testing, the ODC1 SNP was significantly associated with GC (OR = 1.36, 95% CI 1.05, 1.76; p=0.018) (Table 2, Table S4).

Of the remaining 65 SNPs, initial unadjusted association analyses identified seven SNPs with risk or protective associations with GC where the major allele was used as the reference. In an additive regression model, GC was associated with rs35686 in the ghrelin and obestatin prepropeptide (GHRL) gene, (p = 0.015), rs4129009 in toll-like receptor-10 (TLR10) (p = 0.014), rs2294008 in prostate stem cell antigen (PSCA) (p = 0.009), rs1927914 in toll-like receptor-4 (TLR4) (p = 0.0004), and rs530537 in caspase-1 (CASP1) (p = 0.0007) (Table 2). Lastly, two SNPs, rs1056836 and rs1800440, in cytochrome P450 family 1 subfamily B member 1 (CYP1B1) were both associated with GC (p = 0.001 and p = 0.011, respectively) (Table 2). The two CYP1B1 SNPs are not in LD (R² = 0.034). In the models adjusted for age, sex, and bacterial CagA serostatus, only the SNPs in the TLR4 and CASP1 genes remained significant: TLR4 (OR = 1.23; p = 0.021) and CASP1 (OR = 1.23; p = 0.019) (Table 2, Table S5).

In the final model with the three SNPs (ODC1, TLR4, CASP1) and adjusted for age, sex, and CagA serostatus, only the ODC1 SNP remained significant (OR = 1.35; p = 0.021), as well as age (OR 1.03, 95% CI 1.02, 1.04), sex (OR 2.65, 95% CI 1.91, 3.71), and CagA (OR 3.16, 95% CI, 1.93, 5.26). SNPs in TLR4 (OR = 1.04; p = 0.30) and CASP1 (OR = 1.02; p = 0.27) were not significant in the final model. There were no detected interactions between CASP1 and ODC1 (OR = 1.02; p = 0.68) nor TLR4 and ODC1 (OR = 1.04; p = 0.52). However, noting the effects of the TLR4 and CASP1 loci on ODC1-related risk, we performed stratified analyses on the subsets of these two loci to assess how ODC1 risk changes among strata. The effect of the ODC1 genotype in the TLR4 TT subset alone (n = 563; OR = 1.77; p =1.85 × 10⁻³) was significant, and with a larger effect size (Table S6).

We also performed a stratified analysis restricting to the 1095 high-risk individuals who were CagA seropositive. Unadjusted analysis of these individuals were significant (p < 0.021) for the previously identified SNPs: rs1927914/TLR4 (OR = 1.34; p = 0.005), rs1056836/CYP1B1 (OR = 0.77; p = 0.008), and rs2302615/ODC1 (OR = 1.35; p = 0.020). One additional SNP was identified in this high-risk cohort, rs4129009/TLR10, also passed the FDR threshold (OR = 0.717; p = 0.018). The ORs were of similar magnitude and size in the unadjusted and adjusted models for all individuals (Table 2, Table 3).

A network analysis of gastrointestinal tract tissue was performed for the three genes of interest, TLR4, CASP1, and ODC1. TLR4 and CASP1 were shown to interact (interaction = 0.19). ODC1 had no currently known interaction with TLR4 or CASP1 (Figure S2).

DISCUSSION

We found significant associations of the ODC1 rs2302615 SNP, TLR4 rs1927914, CASP1 rs530537, and H. pylori CagA seropositivity with gastric adenocarcinoma in the rural high incidence region of Western Honduras. The ODC1 rs2302615 SNP is of particular interest as it remained significant when adjusted for the other SNPs and is an actionable SNP previously seen in colorectal cancer.(28) CagA is the principal H. pylori oncogenic virulence factor for GC, and in studies with adjustment for CagA serostatus, the effect sizes of the human genetic variants are usually significantly reduced, (49, 50); however, in our study the effect sizes for H. pylori remained robust for the ODC1 rs2302615 SNP following adjustment.

This is the first study, to our knowledge, demonstrating an association of an ODC1 SNP with GC. Notably, in a Korean cohort investigation, other variants in this pathway were associated with GC, indicating that ODC may play a role in risk even if specific genes and variants differ by population.(21) In conjunction with the colon cancer and gut microbiota literature, these results indicate a broader role for this pathway in gastrointestinal cancers.(21, 51-56) ODC has been shown to contribute to the pathogenesis of colitis and associated carcinogenesis by impairing M1 macrophage responses for antitumor immunity in a mouse model.(57) Loss of ODC in myeloid cells enhances host defense against H. pylori.(30) (58)ODC generates polyamines that regulate the host immune response and is associated with DNA damage, due to the release of hydrogen peroxide by the back conversion of spermine to spermidine by spermine oxidase (SMOX), as spermidine and SMOX are associated with beta-catenin activation in gastric epithelial cells.(59-61) Both our current study and the colon cancer study by Zell et al. report an effect for ODC1 variation but our results were for risk of cancer and theirs was for survival .(47)

In addition to the ODC1 SNP, rs2302615, we found that SNPs in the CASP1 and TLR4 genes were also associated with GC. The ODC1 SNP genotype is highly significant when including covariates age, sex, CagA serostatus, and stronger in the TLR4 TT subset. CASP1 encodes for caspase-1, which is an important inflammasome mediator that cleaves precursors of key inflammatory cytokines into active proteins, including the interleukin-1 family and the pyroptosis inducer gasdermin D. (62-64) These CASP-1-dependent processes are critical in mediating an immune response that effectively dampens H. pylori infection and regulates the host immune response, especially to cagA-positive isolates.(65-70) TLR4 encodes the Toll-like receptor-4, which elicits a strong inflammatory immune response by binding lipopolysaccharide ligands on the surface of Gram-negative bacteria and triggering the MyD88 inflammatory pathway to activate monocytes and neutrophils to clear infections.(71) Variants of TLR4 have been associated with risk for chronic H. pylori infection in multiple studies in humans and mice.(72-74) They have also been implicated in carcinogenesis and metastasis and associated with a broad range of pathology, including recurrent spontaneous miscarriage and diabetic retinopathy in type-2 diabetes.(75, 76) Potential links between TLR4 and CASP1 have been described in human and mouse models in a variety of disease phenotypes, including sterile and infectious inflammatory diseases, idiopathic pulmonary fibrosis, and traumatic spinal cord injury.(77-79) However, there is currently no research that supports a clear functional relationship between ODC1 and either TLR4 or CASP1.

Our findings may have potential GC chemoprevention implications. The administration of DFMO to patients in high-risk GC regions where pre-malignant lesions are prevalent may be warranted. This concept is supported by findings that DFMO significantly attenuates GC development in H. pylori-infected Mongolian gerbils.(59, 60) If the mechanism is similar to colon cancer (28, 29), DFMO could be used for chemoprevention in humans with a precision medicine approach and focused on those with the higher risk CC ODC1 genotype, and especially in those who also have the TLR4 TT genotype, which was observed in 40.6% of the patients in our population. The genotype CC ODC1 and diplotype ODC1 CC/ TLR4 TT have a prevalence of 29.5% and 11.5%, respectively, in the Honduras population.

In summary, our findings suggest that ODC1 genotyping and H. pylori CagA seropositivity is significantly associated with gastric adenocarcinoma in Western Honduras. In addition, its association and potential therapeutic efficacy may be enhanced in those with the TT genotype at TLR4 rs1927914. Nonetheless, replication studies in diverse populations will be necessary to confirm possible generalizability. If confirmed by independent studies, this finding will be an important innovation in guiding a precision-based risk reduction intervention for GC in this and perhaps other high-risk populations and potentially reducing the GC mortality in the high incidence regions of Latin America and Asia.

Supplementary Material

Acknowledgements

Footnotes

REFERENCES