Helicobacter pylori genetic diversification in the Mongolian gerbil model

H. pylori colonization of Mongolian gerbils

H. pylori strain B128 was isolated from a human with gastric ulceration (McClain et al., 2009). H. pylori strain 7.13 is an in vivo-adapted strain that was isolated from a Mongolian gerbil infected with H. pylori strain B128. Unlike the parental strain B128, the output H. pylori strain 7.13 reproducibly includes cancer in the Mongolian gerbil model (Franco et al., 2005). The H. pylori strains analyzed in this study were isolated from a previously described cohort of Mongolian gerbils that were infected with strain 7.13, using a protocol approved by the Vanderbilt University IACUC (protocol M/14/021) (Beckett et al., 2016). Briefly, male gerbils (aged 3–5 weeks) were fed one of three diets: a normal diet (Test Diet AIN-93M, Purina Mills), a high salt diet (modified to contain 8.25% salt compared to 0.25% salt in the normal AIN-93M diet) or a low iron diet (manufactured to contain 0 ppm iron compared to 39 ppm iron in the normal AIN-93M diet) (Beckett et al., 2016). After receiving these diets for a period of two weeks, gerbils received two orogastric inoculations with H. pylori strain 7.13 (Beckett et al., 2016). At 16 weeks post-infection, gastric tissue was collected and H. pylori was cultured as described previously (Beckett et al., 2016). The analyses of gastric pH, gastric histology and hematologic parameters in this cohort of gerbils have been described previously (Beckett et al., 2016). A pool of colonies isolated from each gerbil was frozen at −70 °C until the time when genome sequence analysis was undertaken.

Isolation of H. pylori chromosomal DNA

H. pylori output strains cultured from gerbils were minimally passaged on trypticase soy agar plates containing 5% sheep blood (Hemostat Laboratories, Dixon, CA, USA), and then were streaked for single colony isolation. Individual colonies were isolated and expanded by growth on separate plates. Bacteria harvested from one-day-old plates were resuspended in 1 ml of phosphate buffered saline, and genomic DNA was isolated using a Wizard Genomic purification kit (Promega, Madison, WI, USA) and eluting the DNA into water.

H. pylori genome sequencing

H. pylori DNA samples were subjected individually to enzymatic fragmentation using the NEBNext™ dsDNA Fragmentase kit (NEB) according to the manufacturer’s instructions, with average fragment length of 600 bp (range of 400–1,000 bp). Libraries of DNA were prepared from purified fragmented DNA samples using the Kapa Hyper Prep library kit with unique indexes as per the manufacturer’s protocol (Kapa Biosystems, Inc. Wilmington, MA, USA). Quantification of these library preps was performed with the Kapa library quantification kit (Kapa Biosystems, Inc. Wilmington, MA, USA), and sequenced on a MiSeq sequencer using the 600V3 kit (Illumina Inc., San Diego, USA). For the analysis, raw reads were quality trimmed and aligned to the reference sequence (H. pylori strain B8) (Farnbacher et al., 2010) using CLCbio Genomics workbench version 8.5 (Table S1). Data pertaining to read counts, fold coverage, and percent unmapped reads are shown in the Table S1. Alignment parameters were as follows: Mismatch cost = 2, Insertion cost = 3, Deletion cost = 3, Insertion open cost = 6, Insertion extend cost = 1, Deletion open cost = 6, Deletion extend cost = 1, Length fraction = 0.5, Similarity fraction = 0.8. The alignment files exported from the CLCbio workbench in BAM format were then imported into an in-house-developed application (VGAS) for further coverage and SNP analysis. SNP reports were generated using a 10% cut-off, and genome-wide comparisons of SNPs in the output strains compared to the input strain were then performed. Sequence data were deposited in NCBI (Bioproject ID: PRJNA414609).

Three single H. pylori colonies cultured from each gerbil were isolated, expanded and sequenced individually, and three colonies of the input strain were also sequenced in the same manner. All H. pylori sequence data from each animal (three single colonies per animal) were analyzed as a group, and all of the sequence reads of the input strain (3 single colonies) were analyzed as a group. We sought to identify polymorphisms that were detected in ≥75% of sequence reads of output strains from individual animals, and ≤10% of sequence reads from the input strain. This approach allowed identification of polymorphisms that were maximally different when comparing output strain populations with the input strain. Mean annualized SNP rates per site were determined by calculating a ratio of the total number of SNPs in each strain to the genome size (based on the colonization of gerbils for 16 weeks), and then multiplying the values by 3.25 to approximate the number of mutations anticipated to arise over a period of one year.

McDonald–Kreitman analysis

We performed the McDonald–Kreitman test (McDonald & Kreitman, 1991) on nucleotide sequences of the genes of interest in output strains from each gerbil, relative to the input strain. Synonymous changes were used as the neutral class to test the hypothesis that these genes maintained their sequences in a neutral fashion via mutation and random genetic drift. Consensus sequences for each output strain were derived excluding polymorphisms representing <10% of the bases at a given position, and alignments of these consensus sequences to the reference genome were performed using the MUSCLE algorithm with default parameters. As the McDonald–Kreitman test is generalizable to noncoding DNA elements (Andolfatto, 2008), we also assessed the codon neutrality of the noncoding regions upstream of two genes (fecA2 and katA), where SNPs were detected at high frequency. We also performed a separate multi-locus McDonald–Kreitman test to assess the evenness of positive selection across these regions using the Mantel-Haenszel test, as previously described (Egea, Casillas & Barbadilla, 2008). Finally, as the McDonald–Kreitman test is subject to Type I error, we used Bonferroni correction to adjust the p-values of the individual tests.

Preparation of murine neutrophils and co-culture with H. pylori

Murine neutrophils were isolated using a protocol approved by the Vanderbilt University IACUC (protocol V/15/130). Briefly, using a 21-gauge needle, 1 mL of sterile casein solution was injected into the peritoneal cavity of each mouse. An inflammatory response was allowed to develop overnight, and a second dose of casein solution was administered the following morning. Animals were euthanized 3 h after the second injection. The abdominal skin was sterilized with 70% ethanol and retracted to expose the intact peritoneal wall. The peritoneal cavity was then filled with 5 mL of sterile PBS, using a 25-gauge needle, and the abdomen massaged. The fluid was slowly removed using a 25-gauge needle, placed in a 50 mL conical flask and the procedure repeated a second time. The pooled peritoneal fluid was centrifuged for 10 min at 200× g, followed by red blood cell lysis using Ammonium-Chloride-Potassium lysing buffer (ACK; Gibco, Waltham, MA, USA). Peritoneal exudates were washed 3 times, resuspended in 1 mL media (F-12 with 5% FBS) and cell numbers were counted. Finally, the cell solution was brought to the desired concentration (5  × 10⁵ cells/mL) and 1 mL was distributed to each well of 12-well cell culture plates. Plates were incubated for 1 h prior to addition of H. pylori.

H. pylori strain 7.13 (encoding wild-type Fur) and an isogenic mutant (encoding FurR88H) were tagged with distinct antibiotic resistance markers (chloramphenicol or kanamycin resistance), as described previously (Loh et al., 2015). Overnight cultures of these strains were inoculated into separate fresh broth cultures (Brucella broth containing 5% FBS) and allowed to grow for 6 h. Murine neutrophils were then co-cultured with the H. pylori strains at an estimated multiplicity of infection (MOI) of 20:1 (based on measurement of OD₆₀₀), either individually or in competition experiments. In parallel, mock infections were carried out by addition of H. pylori to tissue culture medium (F-12 medium containing fetal bovine serum) alone. In the competition experiments, a 1:1 mixture of H. pylori strains producing WT Fur or FurR88H were co-cultured with the neutrophils. Following a 1 h co-culture, the samples were treated with saponin (0.1% final concentration) (Kwok et al., 2002). Dilutions of the saponin-treated samples were plated on Brucella agar plates containing the appropriate antibiotics, and CFUs were counted 5 days after plating.  The survival of strains co-cultured with neutrophils was compared to the survival of the same strains in medium alone to calculate percent survival. Wilcoxon matched-pairs signed rank test was used to compare the survival of WT strains with survival of FurR88H mutant strains.

Analysis of catalase enzymatic activity

Overnight broth cultures of H. pylori were inoculated into fresh broth cultures and grown to an OD₆₀₀ of ∼0.5–0.6. Samples were normalized to an OD₆₀₀ of 0.1, and catalase enzymatic assays were then performed with the Amplex red catalase kit (Life Technologies). To measure catalase activity, the culture samples were serially diluted 2-fold, and the catalase activity of the diluted cultures was compared to that of purified catalase standards provided in the Amplex red catalase kit (Life Technologies, Carlsbad, CA, USA). Enzymatic measurements were performed in accordance with the manufacturer’s instructions.

Identification of single nucleotide polymorphisms (SNPs)

To gain a better understanding of how H. pylori adapts to different gastric environments, we investigated the genetic diversification of H. pylori that occurs during colonization of Mongolian gerbils. We analyzed H. pylori strains isolated 16 weeks post-infection from a previously described cohort of gerbils (Beckett et al., 2016). To maximize the number of H. pylori genetic adaptions detected, we analyzed H. pylori strains cultured from five gerbils that were fed different diets and that exhibited substantial variation in gastric pathology (Beckett et al., 2016). One animal (Gerbil #1) was maintained on a normal diet and had severe disease (defined as high inflammation scores, gastric cancer and ulcer, increased gastric pH and/or anemia) (Table 1). Two gerbils (Gerbil #2 and #3) were maintained on a high salt diet; Gerbil #2 had severe disease and Gerbil #3 had less severe disease (defined as relatively low inflammation scores, lack of gastric cancer and ulcers, normal pH and not anemic) (Table 1). Gerbils #4 and #5 were maintained on a low iron diet; Gerbil #5 had severe disease and Gerbil #4 had less severe disease (Table 1). Three individual H. pylori colonies isolated from each gerbil (total of 15 single colony isolates, designated as output strains), as well as three individual colonies of the input strain, were analyzed by whole genome sequencing. We then identified differences in the genomes of the output strains compared to the genome of the input strain, using the stringent criteria described in the Methods, which were designed to identify genetic changes that occurred in response to strong selective pressure.

Collectively, the output strains contained 25 unique SNPs that were either not detected or detected at very low levels in the input strain (Table 2). Twenty-one were in coding regions and four were in non-coding regions (Table 2). All 4 of the SNPs in non-coding regions (Table 2) were localized less than 60 nucleotides upstream of a translational start site (katA, fecA2, frpB3, and alpA) (Table 2 and Fig. 1). Two of these were downstream of transcriptional start sites, within 5′ untranslated regions of the mRNA (katA and alpA) (Fig. 1). Among the 21 mutations in coding regions, 17 were non-synonymous and 4 were synonymous (Table 2). Four of the non-synonymous SNPs were in cagY, which encodes a component of the type IV secretion system that translocates the CagA effector protein into host cells. Two of the non-synonymous changes in cagY were instances in which a sense codon was mutated to a stop codon. The SNPs in cagY were all identified in output strains from the same animal (Gerbil #5), and were localized within regions that are repeated multiple times within the gene. Mapping the precise sites of such mutations within repeat regions is challenging using the sequencing technology used in this study. We did not conduct additional studies to verify the precise sites of the cagY mutations listed in Table 2.

The identification of predominantly non-synonymous SNPs in the output strains supports the hypothesis that these mutations were positively selected. Formal testing of this hypothesis is difficult due to the small number of strains analyzed, but we nevertheless performed a McDonald Kreitman analysis to compare the genome sequences of three single colony isolates of the input strain compared to the output strains. This analysis provided evidence that the FurR88H SNP and two mutations in non-coding regions were positively selected (Table 2).

In a previous study, we analyzed the genome sequences of H. pylori strains cultured from two gerbils using 454 sequencing methods (Loh et al., 2015). Five SNPs were detected in 100% of sequence reads of isolates from the animal on a high salt diet, but were not detected (or detected in low abundance) in the input strain or isolates from the animal on a regular diet (Loh et al., 2015). Two of these SNPs were identified in multiple output strains in the current study. Specifically, a mutation upstream of fecA2 was identified in all 5 output strains in the current study, and a FurR88H mutation was identified in 4 of the 5 output strains (Table 2). The only output strain that did not have the FurR88H mutation was isolated from a gerbil consuming a normal diet.

Fur is a regulatory protein that controls gene expression in response to iron availability (Pich & Merrell, 2013). We showed previously that the FurR88H mutation confers increased resistance to high concentrations of salt or conditions of oxidative stress (Loh et al., 2015). Resistance to oxidative stress would presumably provide an important selective advantage in the context of the H. pylori-induced gastric mucosal inflammatory response, which is characterized by an infiltration of neutrophils, macrophages, and other immune cell types. To further define how the FurR88H mutation might confer a selective advantage in the gastric environment, we conducted studies in which isogenic H. pylori strains producing wild-type Fur or FurR88H (each strain harboring a different antibiotic marker) were co-cultured with neutrophils. We detected a significant survival advantage of the isogenic mutant strain containing the FurR88H mutation, compared to wild-type strain (Fig. 2). These data indicate that the FurR88H mutation confers a survival advantage to H. pylori in a neutrophil-containing environment.

H. pylori catalase (encoded by katA) confers resistance to oxidative stress (Benoit & Maier, 2016), and catalase is essential for H. pylori infection of mice (Harris et al., 2003). In a previous study, we noted that output strains cultured from gerbils had markedly higher catalase activity than the input strain (Loh et al., 2015). The input strain used in the previous study contained a frameshift mutation within katA and many of the output strains contained an intact katA ORF, which accounted for the difference in catalase activity (Loh et al., 2015). In contrast to the previous study, the input strain used for the current study had an intact katA ORF. We analyzed the catalase activity of the output strains in the current study compared to the input strain, and found that all of the output strains had increased catalase activity compared to the input strain (Fig. 3). Further experiments will be necessary to evaluate whether this change is a consequence of a mutation downstream of the katA transcriptional start site (within 5′ untranslated region).

Mutation rate

In an effort to quantify the rate of genetic change during the four months in which gerbils were colonized with H. pylori, we calculated the annualized SNP rate per site, as described in the Methods. Overall, the mean annualized SNP rate per site (the number of SNPs that would be expected to occur per site, over the course of one year) among all output strains was 1.5e⁻⁵, with a range of 7.77e⁻⁶ to 3.11e⁻⁵ among individual output strains (Table 1). Interestingly, the mutation rate detected in output strains was positively correlated with the gastric pH in the corresponding gerbils (i.e., higher numbers of SNPs were detected in strains from animals with a high gastric pH) (r = 0.93, Pearson correlation coefficient, p = 0.0204).

Deletions and insertions

We detected five unique deletions (ranging from one to four consecutive nucleotides deleted in individual genes) and three unique insertions among the output strains (Table 3). Three of the deletions were in coding regions and two were in intergenic regions. The genes containing deletions were oipA (also known as hopH, encoding an outer membrane protein), tonB1 (encoding a protein required for activity of outer membrane receptors involved in iron acquisition), and a gene encoding a hypothetical protein (Table 3). The intergenic region deletions were upstream of genes encoding an LPS 1,2-glucosyltransferase and a hypothetical protein. Among the deletions in coding regions, all were frameshift mutations. One of the insertions was in a coding region and two were in intergenic regions. The insertion in a coding region was in the gene encoding the outer membrane protein FecA3, and it was a frameshift mutation. Fifty percent of the deletions or insertions occurred within polynucleotide tracts (Table 3). All of the insertions or deletions (indels) that occurred in coding regions resulted in protein truncation (Fig. 4).

Detection of genetic changes in multiple animals

Most of the SNPs were detected in H. pylori isolates from only one of the gerbils analyzed, but 3 were detected in isolates from multiple animals (four or five gerbils) (Tables 2 and 4). The FurR88H mutation discussed earlier was detected in output strains from four of the five animals. Two SNPs (in non-coding regions upstream of fecA2 and upstream of katA) were found in output strains isolated from all five animals (Tables 2 and 4, Fig. 1).

Most of the indels were detected in H. pylori isolates from a single animal, but two deletions were detected in isolates from multiple animals. Specifically, a two-nucleotide in-frame deletion in tonB1 was detected in isolates from three animals (Tables 3 and 4). Additionally, a one base pair deletion in a gene encoding a hypothetical protein (HPB8_1200) was detected in two animals (Tables 3 and 4). The presence of SNPs or indels in strains cultured from multiple animals suggests that these mutations conferred a selective advantage.