Comparison of Helicobacter pylori Virulence Gene Expression In Vitro and in the Rhesus Macaque

Bacterial strain and culture.

H. pylori J166 contains a functional Cag PAI and the s1m1 allele of the vacA cytotoxin (39). Recent studies by us (39) and others (15) showed that H. pylori J166 preferentially colonizes rhesus monkeys. Plate-grown bacteria were cultivated on brucella agar (Difco Laboratories, Detroit, MI) containing 5% bovine calf serum (GibcoBRL, Gaithersburg, MD) supplemented with TVPA (trimethoprim, 5 mg/liter; vancomycin, 10 mg/liter; polymixin B, 2.5 IU/liter; amphotericin B, 4 mg/liter; all from Sigma, St. Louis, MO) and incubated at 37°C in an atmosphere that contained 5% CO₂. For liquid cultures, bacteria were grown in brucella broth containing 5% bovine calf serum with TVPA and incubated at 37°C with 5% CO₂ and gentle rotation at 60 rpm. Duplicate cultures were inoculated to an optical density at 600 nm (OD₆₀₀) of 0.05 with an overnight starter culture. OD₆₀₀ was determined serially for each culture from 6 to 23.5 h after inoculation.

Animals.

Five specific-pathogen-free rhesus macaques (two males and three females) between the ages of 3 to 4 years were located at the California National Primate Research Center (CNPRC), which is accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care. All animals were confirmed to be free of H. pylori and “Candidatus Helicobacter heilmannii” (30) according to protocols described previously (37). Monkeys were housed indoors in separate cages throughout the course of the experiment. All procedures were approved by the CNPRC Research Advisory Committee and by the University of California, Davis, Chancellor's Animal Use and Care Administrative Advisory Committee.

Animal inoculations.

A 50-ml liquid culture of H. pylori J166 was cultivated as above for 19 h to an OD₆₀₀ of approximately 0.2. The culture was centrifuged and resuspended in fresh brucella broth to an OD₆₀₀ of 1.0. Each monkey was inoculated orogastrically with 7.5 × 10⁸ CFU/1.5 ml. The inoculum was examined by Gram stain, urease, and oxidase tests to ensure a pure culture of H. pylori.

Biopsies and quantitative cultures.

Each monkey was biopsied at 1 week and 1, 2, 3, 4, and 6 months postinoculation (p.i.) using a pediatric gastroscope (Pentax FG-16X) with a 1.8-mm biopsy forceps. The procedure was performed under ketamine anesthesia (10 mg/kg of body weight intramuscularly) after an overnight fast. Eight antral biopsy specimens were collected from the stomach during each endoscopy. Six antral biopsy specimens were individually placed immediately in 200 μl Trizol (GibcoBRL) and put on ice. Two antral biopsy specimens were placed in 250 μl brucella broth for quantitative culture. Biopsy specimens were homogenized with a sterile glass rod before plating and RNA extractions. The 2 antral biopsy specimens in brucella broth were serially diluted, plated onto brucella agar supplemented with 5% bovine calf serum and TVPA, and incubated at 37°C with 5% CO₂ for 5 to 6 days. H. pylori colonies were identified in the conventional manner by colony morphology, microscopy, and biochemistry. CFU were counted and CFU/g of tissue were determined for each monkey. Individual isolates were passed and frozen at −80°C in brucella broth containing 20% (vol/vol) glycerol.

RNA extraction.

At each OD₆₀₀ determination, 2-ml aliquots were removed from the liquid cultures and centrifuged in a table top Microfuge for 30 s. The supernatant was removed, and 1 ml of Trizol was immediately added. Samples were vortexed, and RNA was extracted according to the manufacturer's directions. RNA was treated with DNase I (Roche Applied Science, Mannheim, Germany), purified using an RNeasy clean up kit (QIAGEN, Inc., Valencia, CA), and suspended in molecular biology grade water (BioWhittaker, Rockland, ME) at a concentration of 20 ng/μl. Samples were stored at −80°C prior to analysis. Total RNA was extracted from the six antral biopsy specimens using the Trizol protocol according to the manufacturer's instructions. RNA was treated with DNase I and purified using an RNeasy clean up kit. Purified RNA was suspended in 200 μl molecular biology grade water (BioWhittaker), diluted 1:5, and stored at −80°C prior to analysis.

DNA fingerprinting.

Repetitive extragenic palindromic PCR (Rep-PCR) was used to type strains recovered from each monkey at 1 week and at 1 and 2 months p.i. using methods previously described (22). Chromosomal DNA was prepared from plate-grown cells using the cetyltrimethylammonium bromide method (6). Degenerate oligonucleotide primers (50 pmol each) REP1R-Dt (5′-IIINCGNCGNCATCNGGC-3′) and REP2-Dt (5′-NCGNCTTATCNGGCCTAC-3′) were added to a 25-μl PCR that contained 100 ng of template DNA, 6 mM MgCl₂, 0.6 mM concentrations of each deoxynucleoside triphosphates, and 2 units AmpliTaq DNA polymerase (Applied Biosystems, Foster City, CA). Amplification conditions were initial denaturation at 94°C for 2 min, 30 PCR cycles (94°C for 30 s, 45°C for 1 min, and 72°C for 3 min), and a single extension of 72°C for 5 min. PCR products were separated on 1.5% agarose gels, and fragments were visualized by ethidium bromide staining.

Primer design.

Gene-specific oligonucleotide primer pairs were utilized from a previous experiment (10), and new ones were designed for additional genes (Table 1) using Oligo 6.0 software (Molecular Biology Insights, Cascade, CO) and the known genome sequences of H. pylori 26695 (44) and J99 (3). All primer pairs had a calculated melting temperature of 68 to 70°C, amplified products between 100 to 300 bp, and ended with a double dA 3′ terminus homologous to the template (36). Every primer pair was first used to amplify DNA from H. pylori J166, and the predicted amplicon size was verified by agarose gel electrophoresis prior to RT-PCR.

qRT-PCR.

qRT-PCR was performed with gene-specific primer pairs (Table 1) using methods previously described (10). Briefly, RT and PCR were performed in a single 20-μl reaction mixture using the thermostable recombinant Tth (rTth) DNA polymerase (Applied Biosystems), which in the presence of Mn(OAc)₂ has reverse transcriptase activity and DNA polymerase activity. The RNA template was either 100 ng RNA from in vitro-grown cells or a 1:5 dilution of in vivo RNA. To eliminate PCR carryover contamination, each reaction mixture also included 0.4 U uracil-DNA-glycosylase (New England Biolabs, Beverly, MA). A two-step amplification was performed for 45 cycles at 95°C for 20 s followed by 59.5°C for 1 min. Accumulation of PCR product was detected during each cycle by excitation of SYBR green at 490 nM. Relative fluorescence was characterized by a cycle threshold (C_t) value, which was defined by the crossover point of the kinetic curve with an arbitrary fluorescence level set at 150 relative fluorescence units. The absence of contaminating DNA was examined by performing the RT-PCR with 2.4 mM MgCl₂, in which rTth has DNA polymerase but no RT activity. If the observed C_t with RNA template was not at least 2 cycles less than that of the no-template control (water), the primer pair was eliminated from the analysis.

Primer efficiency.

Since even primer pairs with 100% homology to their template can amplify with markedly different efficiencies, we first tested primer pairs with H. pylori J166 DNA. Assuming there is one DNA copy per cell for each amplified gene, a C_t was determined and used to calculate the primer efficiency. In our laboratory, efficient amplification of 10⁵ bacterial cell equivalents of DNA (0.186 ng) yielded a C_t equal to 18.5, which was arbitrarily defined as 97% efficiency. A primer pair with 97% efficiency would give a standard yield equal to 1.94^18.5 copies after 18.5 cycles of PCR. Therefore, cycle efficiency as a function of the C_t observed for each primer pair can be defined as:

If the observed C_t for a primer pair is 18.5, then the cycle efficiency is 97%; higher C_t values yield lower cycle efficiencies. Only primer pairs with efficiencies of 90% or better were used. Corrected C_t, which takes into account primer efficiency, can be determined from the following equation, where cycle yield = 2 × cycle efficiency:

Calculation of absolute gene expression in vitro.

Copies per cell were calculated based on three factors: (i) a 10-fold change in starting template concentration corresponds to a change in C_t of 3.3 cycles, (ii) the assumption that 100 ng of RNA equals 10⁶ H. pylori cells, (iii) the empirically derived observation that a C_t of 18.5 corresponds to 1 × 10⁵ copies of starting DNA template (assuming 1 copy on the chromosome). For example, an observed C_t of 18.5 for a primer pair with 97% efficiency is calculated as 0.1 mRNA copies/cell. Since the starting amount of RNA template (100 ng) represents 10⁶ bacterial cells, a C_t of 18.5 for RNA indicates 10-fold less starting template per cell than the same C_t observed with DNA, which is performed with 10⁵ cell equivalents. Since a 10-fold change in template concentration corresponds (with efficient amplification) to 3.3 cycles (2^3.3 = 10), we calculate absolute gene expression as:

We have previously shown that calculation of mRNA copies/cell using C_t corrected for primer efficiency yields values that are essentially identical to those obtained by the more conventional method using standard curves (10).

Calculation of relative gene expression in vivo.

Absolute gene expression could not be calculated for in vivo samples because the number of bacterial cells in each sample was unknown. Therefore, to account for differences in bacterial load among animals, all C_t values were normalized to the C_t of H. pylori 16S rRNA for each monkey sample. Data were therefore expressed as:

Primer efficiency and control samples.

All 46 primer pairs amplified H. pylori J166 DNA with at least 90% efficiency. Single bands were detected for each primer pair, and the predicted amplicon length was verified on agarose gels (data not shown). No signal was detected when qRT-PCR was performed with control tissue from uninfected primates nor when Mn(OAc)₂ was replaced with 2.4 mM MgCl₂, in which rTth has DNA polymerase activity but no RT activity (negative control).

In vitro gene expression. (i) Transcript copies/cell at mid-log growth.

Calculation of mRNA copies/cell at mid-log-phase growth (15 h; OD₆₀₀, 1.0) showed that transcript levels of virulence genes varied by 4 orders of magnitude (Fig. 1A), ranging from 0.004 to almost 20 copies/cell. Expression of ureA and ureB was the highest, with nearly identical amounts of transcript. This was expected, since urease is one of the most abundant proteins found in H. pylori (26) and the genes are transcribed on a single operon. Expression levels of ureG and ureI were approximately 10- and 100-fold lower than that of ureAB, respectively. These accessory urease genes are thought to be transcribed on a second operon that consists of ureIEFGH. However, the steady-state level of message from individual genes is complex because both the amount and size of mRNAs from this operon are thought to be affected by a pH-dependent posttranscriptional regulatory mechanism (1). Also among the most highly expressed genes were napA, vacA, and babA (omp28), the ABO blood group adhesin (5). NapA was originally described as a promoter of neutrophil adhesion to endothelial cells (19). It was subsequently shown to be a multifunctional protein related to bacterioferritins and to the Escherichia coli Dps, a nonspecific DNA-binding protein that is induced by environmental stress and probably protects DNA from oxidative damage (9, 13). It seems likely that NapA and VacA, which was shown recently to inhibit H. pylori-specific T-lymphocyte activation (21, 41), are both critical for avoiding innate and adaptive host immunity and maintaining chronic infection. Transcript levels for genes on the Cag PAI also varied by more than 4 orders of magnitude (Fig. 1B), ranging from 0.001 (cag15) to 22 (cag25) copies/cell. There was no apparent relationship between expression level and whether a gene on the PAI is required for CagA tyrosine phosphorylation or induction of IL-8.

(ii) Growth-phase-dependent gene expression.

Growth curves for the duplicate H. pylori cultures were nearly identical (Fig. 2). Therefore, we combined the C_t values determined independently from each culture to calculate a mean copy/cell (Fig. 2). C_t values for 16S rRNA varied by less than twofold during the growth curve. The change in gene expression during the growth cycle was expressed as a ratio of mRNA copies/cell at each time point relative to that at 6 h. From late-log-phase (18 h; OD₆₀₀, 1.4) to stationary-phase (23.5 h; OD₆₀₀, 1.8) growth, most genes (61%) showed a decrease in expression, though expression increased in some cases (15%). Genes with a change in expression of ≥2-fold are listed in Table 2. Of the 7 genes whose expression was induced, most were also found by whole-genome DNA microarray analysis to be induced by entry into stationary-phase growth (43) and by iron starvation (28). Among the induced genes were napA and katA, which likely play an important role in protection against oxidative DNA damage during starvation or other environmental stress. There was little correspondence between the genes whose expression we found to be reduced during stationary phase and the results from DNA microarray studies (28, 43). This may in part reflect the fact that repressed genes were expressed at a lower level than induced genes (Table 2) and may have been below the level of detection of the DNA microarray.

In vivo gene expression. (i) Quantitative culture and Rep-PCR.

Two antral biopsy specimens were used to determine quantitative bacterial load from each monkey at 1 week and 1, 2, 3, 4, and 6 months p.i. At 1 week p.i., all 5 monkeys were infected with 5 × 10⁴ to 4 × 10⁷ CFU/g of tissue. However, between 1 and 6 months p.i., only 1 monkey remained infected with a bacterial load that was sufficient for transcript quantitation. This was likely due to the fact that the inoculum was grown from a single colony, whereas we have previously used a mixture of 6 rhesus-passaged J166 strains (39). A mixed inoculum probably contains greater genomic diversity that more efficiently colonizes multiple hosts (23). Similar observations have recently been made with the mouse model (17). Agarose gel electrophoresis of the Rep-PCR products showed that the output strains from each monkey were identical to the input J166 and differed from representative strains of H. pylori that are enzootic in socially housed rhesus monkeys at the CNPRC (Fig. 3). These results confirm that our expression data reflect infection with the inoculated strain (J166) and not naturally acquired infection.

To provide an estimate of bacterial load, we normalized the C_t for in vivo expression to that for 16S rRNA. We therefore compared the quantitative culture results to the C_t for 16S rRNA (Fig. 4). The results showed a close correspondence between CFU/g and C_t over a broad range of bacterial load, with a Pearson product correlation coefficient of 0.80. Since quantitative culture of H. pylori from gastric biopsy specimens lacks sensitivity, with a detection limit of 10² to 10³ CFU/g of tissue, qRT-PCR can be used as a more accurate measure of infection when bacterial load is low. For genes with high expression levels, it may sometimes be possible to reliably quantitate gene expression from biopsy specimens that are negative by culture but which are positive by qRT-PCR for 16S rRNA.

(ii) Relative gene expression at 1 week p.i.

To account for differences in bacterial load, all C_t values were normalized to the C_t of 16S rRNA for each animal, which was 1.0 by definition. Since 16S rRNA is more abundant than bacterial message, relative expression ranged from 0 to 1. Expression levels for most genes represent mean data from the results from 5 animals, though in some cases, due to low bacterial load, expression could only be determined from 3 or 4 animals. Relative gene expression of the virulence and Cag PAI genes was between 1 and 5 logs lower than 16S rRNA (Fig. 5A and B). Genes that were highly expressed in vitro, such as ureA, ureB, napA, cag1, and cag25, were also the most highly expressed genes in vivo. The lowest expression among the Cag PAI genes was cag22, which is required for IL-8 induction and CagA translocation (20). This suggests that even expression levels more than 4 orders of magnitude lower than 16S rRNA are biologically relevant.

In vivo versus in vitro gene expression.

To identify genes that were induced or repressed in vivo, we first compared expression in vitro at mid-log phase (15 h; OD₆₀₀, 1.0) to in vivo expression at 1 week p.i. Relative gene expression for the in vitro data was calculated exactly as for the in vivo data (see Materials and Methods). Gene expression was generally lower in vivo compared to in vitro expression during mid-log phase (Fig. 6A). The mean (±standard deviation) ratios of in vivo to in vitro expression during mid-log phase were 0.34 (±0.36) and 0.24 (±0.23) for the virulence and Cag PAI genes, respectively. We next compared in vivo expression at 1 week p.i. to stationary-phase expression (23.5 h; OD₆₀₀, 1.8) in vitro (Fig. 6B). This analysis showed that 9 genes were more highly expressed in vivo than in vitro (Fig. 6B). Of these, 6 genes (cag5, virB11, cag7, cag8, cag9, and cag10) are found contiguously on the PAI (absent cag6), are oriented with open reading frames (ORFs) in the same direction, and are flanked by genes (cag4 and cag11) whose ORFs are in the opposite direction (Fig. 1B). Most of these genes are implicated by experimental analysis (32, 35, 42) or by sequence homology (11) to form portions of the type IV secretion pilus or its accessory proteins. In addition, another gene that appeared induced in vivo (comB8) is thought to encode a portion of a second H. pylori type IV secretion system that is utilized for DNA uptake and is required for competence (25). These results suggest the possibility that at least some of the genes encoding the H. pylori type IV structural pilus and its accessory proteins may form an operon that is induced during growth in vivo.

In vivo gene expression over time.

We analyzed in vivo H. pylori gene expression for one monkey at 1 week and 2, 3, 4, and 6 months p.i. To simplify the analysis, we grouped the data into early (1 week)-, mid (2 to 3 months)-, and late (4 to 6 months)-acute-phase infection. Interestingly, almost all genes that showed more than a twofold induction were induced during early- to mid-acute-phase infection (between 1 week and 2 to 3 months p.i.) (Table 3). These genes included outer membrane proteins (OMPs), superoxide dismutase, and Cag PAI genes, which may all play a role in establishing the infection and evading the initial host immune response. Early in infection, we saw apparent repression of babA (omp28), which codes for the ABO blood group binding adhesin (5), and induction of babB (omp19), a gene encoding for a related OMP with unknown function. However, we recently reported that infection of rhesus macaques with H. pylori J166 results in selection for strains that have undergone a gene conversion, whereby babA has been deleted and replaced by a duplicated copy of babB. Therefore, this apparent change in expression of H. pylori OMPs is mediated by genomic recombination and not by regulation of expression in the conventional sense. The apparent changes in cag7 (cagY) expression (Table 3) may also be mediated by genomic events, since cag7 has extensive repeats that mediate in-frame deletions and duplications (4). These findings emphasize that, particularly in H. pylori, apparent changes in expression may sometimes reflect modifications in the genome rather than bacterial sensing of environmental signals via traditional two-component regulatory systems.

Comparison of H. pylori gene expression in rhesus macaques and humans.

We recently used qRT-PCR to perform H. pylori transcription profiling in chronically infected human patients (10). In general, the results are similar to those described here, which further validates the use of the macaque as a model of human infection. The Pearson product correlation coefficient between in vivo gene expression levels in humans and macaques was 0.89 (Fig. 7), and the trend line had a slope of 1.0. This is a striking finding in view of the fact that these data were collected not only from different host species but also from different bacterial strains and at different time periods. The generally reduced expression levels in vivo compared to in vitro also mimics the results seen for humans. We cannot fully exclude the possibility that this finding is spurious. For example, if extraction efficiency for mRNA versus stable RNA is lower in vivo than it is in vitro, we might incorrectly conclude that expression is lower in vivo. If this were the case, the reduced extraction efficiency of mRNA from in vivo samples might be particularly troublesome for low-abundance message, which would lead us to conclude erroneously that reduced expression in vivo was exaggerated for genes with low transcript abundance. However, this is the opposite of what we observed (Fig. 6B). Furthermore, the fact that the result is robust over differences in host and bacterial strains suggests that reduced expression in vivo for many (but not all) genes reflects the biology of H. pylori infection and is not an artifact of the methods used here.

These results provide, for the first time, a quantitative fingerprint of H. pylori gene expression in vitro and in vivo in experimentally infected nonhuman primates. Comparison of H. pylori gene expression in vivo to expression in vitro during stationary phase suggests the possibility that some genes encoding the H. pylori type IV structural pilus and accessory proteins may form an operon that is induced during growth in vivo.