Genome-wide association study of gastric cancer- and duodenal ulcer-derived Helicobacter pylori strains reveals discriminatory genetic variations and novel oncoprotein candidates

Abstract

Genome-wide association studies (GWASs) can reveal genetic variations associated with a phenotype in the absence of any hypothesis of candidate genes. The problem of false-positive sites linked with the responsible site might be bypassed in bacteria with a high homologous recombination rate, such as Helicobacter pylori, which causes gastric cancer. We conducted a small-sample GWAS (125 gastric cancer cases and 115 controls) followed by prediction of gastric cancer and control (duodenal ulcer) H. pylori strains. We identified 11 single nucleotide polymorphisms (eight amino acid changes) and three DNA motifs that, combined, allowed effective disease discrimination. They were often informative of the underlying molecular mechanisms, such as electric charge alteration at the ligand-binding pocket, alteration in subunit interaction, and mode-switching of DNA methylation. We also identified three novel virulence factors/oncoprotein candidates. These results provide both defined targets for further informatic and experimental analyses to gain insights into gastric cancer pathogenesis and a basis for identifying a set of biomarkers for distinguishing these H. pylori-related diseases.

Keywords: GWAS; Helicobacter pylori; duodenal ulcer; gastric cancer; population genomics; recombination.

Fig. 1.

Core-genome phylogeny and metadata of the 240 strains from an hspEAsia population. Left: mid-point rooted core-genome phylogeny. Heatmap: column 1, host disease status (DU or GC). Column 2: country of isolation. Columns 3–14 correspond to the genes carrying a nucleotide associated with GC. GC, gastric cancer; DU, duodenal ulcer.

Fig. 2.

Q-Q plot to assess the GWAS results. Each dot in (a) indicates a SNP, whereas that in (b) indicates a unitig. Y-axis: observed $-{\log }_{10}\left(P\right)$ of each SNP or unitig, where P is its P-value. X-axis: expected $-{\log }_{10}\left(P\right)$ under the null hypothesis of no association. Non-synonymous and intergenic SNPs with P<10^-4 and associated with GC are presented as red and green, respectively. GC, gastric cancer; SNP, single nucleotide polymorphism.

Fig. 3.

Manhattan plot summarizing the GWAS results. The nonsynonyous SNPs, integenic SNPs, and DNA motifs associated with gastric cancer are colored in red, light green, and orange, respectively. The black dots corresponds to all the other SNPs used in the SNP-GWAS.

Fig. 4.

Predicted structures of proteins with discriminatory non-synonymous SNPs. (a) TlpC (HPF57_0094). (i) Model on the homolog of strain 26695 (PDB 5wbf) [38]. K55 in HPF57_0094 corresponds to E217 in HP0082, which is split into two genes in the Japanese reference strain F57. (ii)–(iii) Surface electric charges. E55 mutant protein was generated from the model by mutagenesis in silico (PyMOL). (iv) N51 has direct interaction with lactose. (v) S51 from mutagenesis (PyMOL). S51 is farther from lactose and would accommodate larger ligands. (b) HsdM (HPF57_0490). (i) Reaction steps of a Type I modification enzyme [41]. (ii) Model on 5ybb in PDB, two methyltransferase molecules, each 2M+1S, of Caldanaerobacter subterraneus . D115 corresponds to L95. (iii)–(v) Model on 2y7h in PDB, a model of EcoKI methyltransferase based on EMD-1534 [74]. D115 corresponds to R72. SNP, Single nucleotide polymorphism.

Fig. 5.

Predicted structures of three new virulence factor/oncoprotein candidates. (a) Isp (inactive Ser protease). (i) F57_1355 modelled on E. coli DegP. (ii) Active site of F57_1355 modelled on 3lgi.1 in PDB (E. coli DegS). The three amino acids (HDS triad) responsible for activity are all replaced. (iii)–(v) Surface electric charge distribution in E. coli DegS without PDZ [75] (3lgi.1 in PDB), HPF57_1355 modelled on it, and the E178G mutant generated in silico. (b) TriH, Triple halves. HPF57_0151 (HP0130). (i) Map. ‘Disordered’ is from UniProt. Nuclear localization signal is by cNLS Mapper. ‘Diversifying selection’ is from a previous study [49]. (ii)–(iv) Similarity to three half domains. (iii) Modelled on NapA (strain YS39, 4evd in PDB) and aligned with iron-soaked NapA (YS39, 3ta8 in PDB). Fe-interacting residues as well as the GWAS residues are in sticks. The 2c6r in PDB is Dps2 in Deinococcus radiodurans . Note the difference in NapA coordinates in the literature [48, 76]. (iv) HPF57_0151 modelled on PDB 4pjw (human).