Genome sequencing reveals a phage in Helicobacter pylori.

UNLABELLED: Helicobacter pylori chronically infects the gastric mucosa in more than half of the human population; in a subset of this population, its presence is associated with development of severe disease, such as gastric cancer. Genomic analysis of several strains has revealed an extensive H. pylori pan-genome, likely to grow as more genomes are sampled. Here we describe the draft genome sequence (63 contigs; 26× mean coverage) of H. pylori strain B45, isolated from a patient with gastric mucosa-associated lymphoid tissue (MALT) lymphoma. The major finding was a 24.6-kb prophage integrated in the bacterial genome. The prophage shares most of its genes (22/27) with prophage region II of Helicobacter acinonychis strain Sheeba. After UV treatment of liquid cultures, circular DNA carrying the prophage integrase gene could be detected, and intracellular tailed phage-like particles were observed in H. pylori cells by transmission electron microscopy, indicating that phage production can be induced from the prophage. PCR amplification and sequencing of the integrase gene from 341 H. pylori strains from different geographic regions revealed a high prevalence of the prophage (21.4%). Phylogenetic reconstruction showed four distinct clusters in the integrase gene, three of which tended to be specific for geographic regions. Our study implies that phages may play important roles in the ecology and evolution of H. pylori. IMPORTANCE: Helicobacter pylori chronically infects the gastric mucosa in more than half of the human population, and while most of the infected individuals do not develop disease, H. pylori infection doubles the risk of developing gastric cancer. An abundance and diversity of viruses (phages) infect microbial populations in most environments and are important mediators of microbial diversity. Our finding of a 24.6-kb prophage integrated inside an H. pylori genome and the observation of circular integrase gene-containing DNA and phage-like particles inside cells upon UV treatment demonstrate that we have discovered a viable H. pylori phage. The additional finding of integrase genes in a large proportion of screened isolates of diverse geographic origins indicates that the prevalence of prophages may have been underestimated in H. pylori. Since phages are important drivers of microbial evolution, the discovery should be important for understanding and predicting genetic diversity in H. pylori.

Introduction

All environments inhabited by bacteria and archaea also house a diversity of viruses; in fact, viruses are effectively keeping microbial populations below the level dictated by nutrient availability in many ecosystems, i.e., top-down control. For instance, it has been estimated that 20 to 40% of bacterioplankton cells in the ocean are lysed by bacteriophages on a daily basis (1). Escape from viral infection is thus a major driver of microbial evolution, which is apparent in, for example, high evolutionary rates of cell surface proteins that serve as viral receptors (2) or the rapid acquisition of spacers complementary to viral DNA within the clustered regularly interspaced short palindromic repeat (CRISPR) microbial immune system (3, 4). This is countered by adaptations of the viral populations (e.g., cell attachment proteins), resulting in an evolutionary arms race between viruses and their hosts (5). Furthermore, viruses contribute significantly to microbial evolution by inserting novel genetic elements into host genomes; it has been suggested that most of the strain-specific genes are derived from viruses (6).

Helicobacter pylori chronically infects the gastric mucosa in more than 50% of the human population and has coevolved with its human host (7–11). Most of the infected individuals do not develop disease; however, H. pylori infection doubles the risk of developing gastric cancer (12), which is responsible for 10% of all cancer-related deaths in the world. It is therefore important to identify H. pylori genes and genotypes associated with disease development within the relatively unexplored high genetic diversity of H. pylori (8, 10, 13). H. pylori possesses an arsenal of tools which allow it to colonize the hostile gastric environment, including a strong urease to resist the acidic gastric pH and flagella to move in the thick mucus and evade host immunity. H. pylori pathogenicity has been associated with the production of numerous virulence factors, in particular those encoded by the cag pathogenicity island (cag PAI), which have been associated with peptic ulceration, atrophic gastritis, and gastric adenocarcinoma (14, 15). These factors can target tumor suppression functions (reduction of p53 on delivery of CagA), in a mode similar to that of DNA tumor viruses (16). Currently, eradication of H. pylori is performed with antibiotics that have severe effects on the intestinal microbiota, including persistence of elevated levels of antibiotic-resistant strains years after treatment (17); thus, novel therapies targeting H. pylori specifically are needed.

Very little is known about H. pylori phages; however, shortly after the discovery of H. pylori, Marshall et al. (18) and Goodwin et al. (19) described intracellular phage-like particles observed in human gastric mucosa. Two other studies from the early 1990s showed spontaneous production of small amounts of phage particles by the H. pylori strain Schreck (20). Furthermore, Vale et al. (21) described H. pylori temperate phage induction of phage-like particles using UV. Reports of prophages in other Helicobacter species are also rare. Until now, there have been only two reports of a prophage: one in the genome of Helicobacter acinonychis strain Sheeba (22) and one in Helicobacter felis CS1 (ATCC 49179) (23). Although the reports concerning H. pylori phages are rare, genetic characteristics of H. pylori isolates suggest that the organism is challenged by viruses; a multitude of restriction-modification (R-M) system genes are found that display high evolutionary rates (24–26). In addition, R-M system genes are highly diverse among strains, representing more than half of the strain-specific genes present in H. pylori sequenced genomes (27). The repertoire of R-M system genes in a bacterial cell influences the pattern of DNA methylation, which in H. pylori has been shown to influence gene expression (28); however, R-M system genes are generally believed to function as a protection against invading DNA (2). H. pylori cell surface proteins are rapidly evolving (29–31). This has been inferred as an adaptation to differences in mucosal surface structures among human hosts or as a selection to counteract the immune response, but it could also reflect escape mechanisms in response to bacteriophage infection.

In this study, we sequenced the genome of an isolate from a gastric mucosa-associated lymphoid tissue (MALT) lymphoma patient with the aim to identify genes that could be associated with disease development. An unexpected finding was the presence of a prophage integrated in the bacterial genome, from which we were able to induce production of phage particles.

RESULTS

The H. pylori B45 genome.

Whole-genome shotgun sequencing carried out using the 454 Titanium platform (454 Life Sciences, Branford, CT) generated 42 Mb of data that assembled into 63 contigs with a total size of 1,602,587 bp and 26× average coverage (GenBank accession numbers AFAO01000001 to AFAO01000063). Mean contig length was 25,437 bp (minimum, 581 bp; maximum, 214,447 bp), and average G+C content was 39%. Some contigs had G+C contents that strongly deviated from the genome average (minimum, 32.99%; maximum, 48.19%), which suggests integration of external DNA in the bacterial chromosome (32).

The AMIGene software (33) predicted 1,602 coding sequences (CDSs) on the B45 contigs. We used BLASTp and the Markov clustering (MCL) algorithm (34, 35) to define orthologous genes within H. pylori genomes, including that of B45 and seven reference genomes (26695, J99, HPAG1, P12, G27, Shi 470, and B38) (13, 32, 36–39). The MCL processes built 2,293 ortholog clusters. The 1,602 CDSs identified on B45 contigs were distributed over 1,182 clusters ubiquitous in all reference genomes (corresponding to 1,195 B45 CDSs), 223 clusters were absent in at least one reference genome (nonubiquitous; corresponding to 232 B45 CDSs), and, finally, 174 clusters (corresponding to 175 B45 CDSs) had no counterpart in any of the 7 reference genomes.

The 175 CDSs found in clusters unique to B45 were analyzed further by BLASTp searches against the seven H. pylori reference genomes, two H. pylori draft genomes available at that time (H. pylori strains 98-10 and B128 [40]), and H. acinonychis strain Sheeba (22) and against all other complete bacterial genomes in GenBank. Thirty-three of the CDSs lacked significant BLASTp hits to any of the above genomes. Since these CDSs were unusually short (mean length, 33 bp), we considered them artifacts. Fifteen CDSs had significant BLASTp hits with Hp_B128 CDSs (40), six had hits with Hp_98-10 (40), 27 had hits to H. acinonychis strain Sheeba (22), two had hits to other bacteria, and 92 had hits to one or more of the seven H. pylori reference genomes initially considered.

The B45 genome contains a complete cag pathogenicity island (cag PAI) comprised of 27 CDSs divided over three contigs (C26, C50, and C53), as was expected considering its ability to promote interleukin-8 (IL-8) production and hummingbird phenotype on AGS cells (41) (see Table S1 in the supplemental material) (GenBank accession numbers AFAO01000023, AFAO01000045, and AFAO01000048, respectively).

The B45 genome has a prophage.

Interestingly, among the 27 CDSs in the orthologous clusters unique to B45 with significant BLASTp hits to H. acinonychis strain Sheeba, 22 matched genes in the H. acinonychis strain Sheeba prophage II (22). These CDSs were originally found on two contigs: contig C47 (GenBank accession number AFAO01000043) and contig C21 (GenBank accession number AFAO01000020). The two contigs could be joined by PCR, using primers targeting the ends of C21 and C47 and Sanger sequencing of the resulting sequence fragment; hence, the B45 prophage constituted a continuous genomic region. A 1,665-bp region separated C21 from C47 (which corresponded exactly to a small contig C73 [AFAO01000062]). The reconstructed area (called C21B450028-C73B450001-C47B450001) contained a large CDS (5,958 bp), homologous to Hac_1615. The final prophage sequence constituted 24,645 bp with a G+C content of 37% and contained 27 CDSs encoded on the same strand (GenBank accession number JF734911).

Nine of the 32 genes carried by the 28.4-kb H. acinonychis strain Sheeba prophage II lack orthologs in the B45 prophage, while five B45 genes lack orthologs in the Ha_Sheeba prophage (Fig. 1) (see also Table S2 in the supplemental material). Six of the most upstream B45 prophage genes were also found in a 5.5-kb region of the B38 genome, which probably corresponds to a remnant prophage (39). The average amino acid identity between B45 and Ha_Sheeba prophage genes was 61.0%, and that between B45 and B38 genes was 70.5%. The reconstructed B45 prophage was flanked on each side by partial gene sequences corresponding to jhp0786, which is a type I R-M enzyme subunit. The B45 phage has therefore probably been inserted into an R-M system.

FIG 1

Comparison of the B45 prophage area with Helicobacter acinonychis strain Sheeba prophage II and Helicobacter pylori B38 prophage. Each arrow represents a CDS for which the size is proportional to the length. The dotted lines represent homologs between B45 prophage and H. acinonychis strain Sheeba prophage II (22) or H. pylori B38 prophage (39), and the gray scale represents amino acid identity to B45 prophage proteins.

The B45 prophage can be induced by UV irradiation.

To assess whether the B45 prophage is functional, several assays were used to test for phage particle production. No significant lysis plaques of B45 were observed on agar plates after mitomycin treatment (0.1 to 1 µg/ml; see Materials and Methods) or UV inductions (1 to 10 times the D value of UV irradiation; see Materials and Methods). Lysis plaques were observed using the JP1 strain, which served as a positive control (21). Liquid medium cultures did not reveal a significant decrease in optical density (OD) values (600 nm) at low concentrations (0.1 µg/ml) of mitomycin. High concentrations (1 µg/ml) could not be used since they were highly toxic to the cells (the negative-control strain 26695 displayed markedly lower OD values at this concentration). Finally, we tried to induce the B45 prophage in liquid medium by UV irradiation (1 D value after 24 h of incubation, followed by 24 additional hours of incubation; see Materials and Methods), which was followed by a precipitation step, as described in Materials and Methods. Transmission electron microscopy (TEM) using negative staining revealed numerous phage-like particles with an eggshell structure (size, 100 ± 30 nm) (Fig. 2) apparently lacking a tail. Negative staining does not usually reveal phage tails because of the limited density of the tail (42). Fixation embedding and ultrathin sectioning of the bacteria allowed us to have access to the interior of the bacterial cells. A few cells contained 1 or 2 phage particles in their cytoplasm. These were recognizable by a high-density mature head, with rounded appearance, on which a tail was attached. The phage head was ~62.5 nm (±7.3 nm) in diameter, and the tail was ~92.4 nm (±2.97 nm) long and 5 to 6 nm in diameter. The total length of the phage was ~150 nm. Overall, these structures were compatible with a Siphoviridae phage (Fig. 2).

FIG 2

Transmission electron microscopy (TEM) images obtained after induction of the B45 prophage. (A) Examples of TEM images obtained using negative staining: numerous phage-like particles with an eggshell structure (size, 100 ± 30 nm). (B) (1) TEM images obtained after fixation embedding and ultrathin sectioning of the bacteria. Two phage particles on the top of the bacterial cell are visualized in the cytoplasm. The dark zone in the lower part of the cell corresponds to the condensed genomic DNA. On each side of the cell, two other bacterial cells are partially visible (one with genomic DNA). Some flagellar remnants can also be seen in between the cells. (2 and 3) Higher magnification focused on these 2 particles showed that they were recognizable by the high-density mature head, with a rounded appearance, to which a tail was attached. The phage head was ~62.5 nm (±7.3 nm) in diameter, and the tail was ~92.4 nm (±2.97 nm) long and 5 to 6 nm in diameter. The total length of the phage was ~150 nm.

Most phages replicate by first circularizing their linear DNA (43). We were unable to purify enough circular DNA to visualize it directly on an agarose gel, which may indicate that our induction conditions were suboptimal. However, observation of supernatant from UV-treated cultures allowed us to detect phage DNA by PCR amplification of the integrase gene after two rounds of exonuclease treatment. This indicates that circular phage DNA was present, protected from exonuclease cleavage. In contrast, the cagA gene could not be amplified after exonuclease treatment, indicating that all residual genomic DNA had been removed (Fig. 3).

FIG 3

PCR detection of free circular phage DNA. To ensure complete elimination of bacterial genomic DNA, the DNA extracted from concentrated phage particles was treated twice with exonucleases (for 4 or 24 h) in order to digest any linear bacterial genomic DNA, leaving the circular DNA (i.e., phage DNA), which cannot be degraded by these enzymes. The extracted DNA was then tested for the presence of phage DNA and bacterial genomic DNA by PCR amplification of the phage integrase gene (using the primers F1, AAGYTTTTTAGMGTTTTGYG, and R1, CGCCCTGGCTTAGCATC, generating a 529-bp amplicon) and the cagA gene (750-bp amplicon) as already described (67, 70). Lane M corresponds to the 1-kb DNA ladder (Promega). Lanes 1 and 7, B45 extracted phage DNA plus exonucleases, 24 hours; lanes 2 and 8, B45 extracted phage DNA plus exonuclease buffer only, 24 hours; lanes 3 and 9, B45 extracted phage DNA plus exonucleases, 4 hours; lanes 4 and 10, B45 extracted phage DNA plus exonuclease buffer only, 4 hours; lanes 5 and 11, B45 DNA (positive control); lanes 6 and 12, H2O (negative control).

Genome organization compatible with the Siphoviridae phage family.

Because the phage morphology observed using TEM was compatible with the Siphoviridae family of phages, we compared the B45 prophage genome organization with this family of viruses. Siphoviridae genomes fall into two size classes: 121 to 134 kb and 22 to 56 kb (44, 45). Hence, the B45 prophage (as well as H. acinonychis strain Sheeba prophage II) had a size compatible with the second class. Siphoviridae family phages usually harbor several typical modules such as head proteins, head-tail joining, DNA packaging, tail and tail fiber proteins, host lysis, lysogeny module, replication module, and transcriptional regulators (44). Most of the genes (23 of 27) did not generate significant BLAST matches to proteins other than those of the H. acinonychis strain Sheeba and B38 prophage proteins, indicating that no closely related virus had been sequenced before (Table 1; see also Table S2 in the supplemental material). However, searches based on PSI-BLAST gave functional clues about a subset of the genes. Although sometimes highly putative, the results indicated that functionally related genes could be organized in several modules. These included a putative lysogeny module, harbouring at least a phage integrase/recombinase gene (C21B450018); 2 putative transcriptional regulators (C21B450020 and C47B450016); a replication module including one putative phage replication protein (C21B450021), a DNA repair protein (C21B450024 and C47B450012), a DNA primase (C21B450025), and a putative structural maintenance of chromosomes protein (SMC) (C21B450026); a putative tail fiber protein (C47B450002); and a putative crystalline beta/gamma motif-containing protein (putative lysin) (the 5,958-bp CDS which corresponds to the fusion of C21B450028-C73B450001-C47B450001) (Table 1).

TABLE 1

Annotation of B45 prophage coding sequences[

B45 prophage CDS	Hac	HELPY	Putative annotation
C21B450017	Hac_1604	1520	Unknown function
C21B450018	Hac_1606	1521	Bacteriophage-related integrase
C21B450019	Hac_1607	1522	Unknown function
C21B450020	Absent	1523	Putative transcriptional regulator
C21B450021	Hac_1609Hac_1608	1525	Putative phage replication protein
C21B450022	Absent	Absent	Putative ABC superfamily ATP bindingcassette transporter
C21B450023	Hac_1610	1527	Unknown function
C21B450024	Hac_1611	Absent	Putative DNA repair protein(putative helicase)
C21B450025	Hac_1612	Absent	Putative DNA primase
C21B450026	Hac_1614	Absent	Putative SMC (structural maintenanceof chromosomes) proteins
C21B450027	Hac_1614	Absent	Unknown function
C21B450028-C73B450001-C47B450001	Hac_1615	Absent	Putative crystalline beta/gamma motif-containing protein (putative lysin)
C47B450002	Hac_1617	Absent	Putative tail fiber assembly protein
C47B450003	Hac_1618	Absent	Unknown function
C47B450004	Hac_1619	Absent	Putative histidine kinase
C47B450005	Hac_1620	Absent	Putative histidine kinase
C47B450006	Hac_1621	Absent	Putative sensor protein
C47B450007	Hac_1622	Absent	Unknown function
C47B450008	Hac_1623	Absent	Unknown function
C47B450009	Absent	Absent	Transposase
C47B450010	Absent	Absent	Transposase
C47B450011	Hac_1631	Absent	Unknown function
C47B450012	Hac_1632	Absent	Putative DNA repair protein
C47B450013	Hac_1633	Absent	Unknown function
C47B450014	Hac_1634	Absent	Phage tail tape measure protein
C47B450015	Hac_1635	Absent	Unknown function
C47B450016	Absent	Absent	Putative transcriptional regulator

For each predicted protein, the putative annotations are provided for Helicobacter acinonychis strain Sheeba prophage II (Hac) and Helicobacter pylori B38 (HELPY) prophage homologs. C21B450028-C73B450001-C47B450001 corresponds to the fused CDSs obtained after PCR and sequencing (see Results).

Surprisingly, C47B450009 and C47B450010 had significant homologies with the transposable element called ISHp608 of H. pylori (46). ISHp608 is a member of the IS605 transposable element family. It contains two open reading frames (orfA and orfB), each related to putative transposase genes. Interestingly, orfB is also related to the Salmonella virulence gene gipA described in the lysogenic phage Gifsy-1 that affects Salmonella enterica Typhimurium survival in Peyer’s patches (47). C47B450009 had significant homology with orfA found in H. pylori Lith5 (AAL06574.1) (expect value = 4e−87, identities = 155/155 [100%]), and C47B450010 had significant homology with orfB found in H. pylori strain PeruCan2A (AF357223_2) (expect value = 0.0, identities = 379/382 [99%]).

Prevalence of the H. pylori prophage.

In order to estimate the prevalence of this prophage type in H. pylori genomes, we screened 341 strains, isolated in different geographic regions and from patients presenting different pathologies, using degenerate PCR primers targeting the B38, B45, and H. acinonychis strain Sheeba prophage II integrase genes. Surprisingly, 21.4% of the isolates (73/341) had an integrase gene (GenBank accession numbers JF734912 to JF734984), suggesting that this is a much more frequent phenomenon than initially estimated. The prevalences of the gene were similar in different pathologies: 25.6% (30/117) in duodenal ulcer strains, 20.9% (23/110) in gastritis strains, 15.9% (10/63) in gastric MALT lymphoma strains, and 19.6% (10/51) in atrophic gastritis/adenocarcinoma strains. Significant association was not found with cag PAI status. The average prevalence among European strains was 21.4% (59/276). Prevalence was higher in Sweden (18/56, 32.1%) than in France (28/125, 20%), Germany (6/53, 11.3%), and Portugal (3/29, 10.3%). No significant conclusion can be made concerning the United Kingdom and Norway, for which only 7 and 6 strains were tested, respectively. Notably, the prevalence among African isolates was low, with only one strain out of 18 (5.6%) being positive by PCR screening. This could be explained either by a lower prevalence of the prophage in African isolates or by significant integrase sequence divergence leading to mismatching of the PCR primers. The only South American strain included in the present study was negative. Thirteen of the 46 Asian strains (28.3%) included in the study were also positive.

Phylogeography of the prophage integrase gene.

For each positive strain, amplicons were sequenced on both strands. Phylogenetic reconstruction of the phage integrase gene gave two main branches (Fig. 4). One of these branches included two well-supported clusters. Interestingly, the first one (cluster A), which included the B45 integrase, was comprised of 81.0% (17/21) French isolates and the second (cluster B) was composed of 70.6% (12/17) Swedish isolates. The H. acinonychis strain Sheeba prophage II integrase formed a sister group with these two clusters. The other main branch of the tree had two well-supported clusters. The first one was comprised exclusively of Asian strains (cluster C) and included 10 of the 13 positive Asian strains included in the study. The last cluster (cluster D), found at the extremity of the second branch, was more heterogeneous in geographic composition and was comprised of 6 French, 4 German, 3 Portuguese, and 3 Swedish strains. Overall, the phylogenetic reconstruction indicated a strong biogeographic signal within the phage integrase genetic diversity.

FIG 4

Phylogenetic analysis of phage integrase sequences. Phylogenetic analysis was performed on the partial (442 bp) integrase DNA sequences obtained from the 73 integrase positive strains, and from the Hac_1606 genome. The evolutionary history of phage integrase genes was inferred using the neighbor-joining method using the Kimura two-parameter model (73). The bootstrapped consensus tree, inferred from 1,000 replicates, is presented as a radial tree. Bootstrap values (percentages of replicate trees in which the associated taxa clustered together) are shown for a selected nodes in the tree. The tree is drawn to scale, with branch lengths corresponding to the evolutionary distances used to infer the phylogenetic tree. Two main branches were identified. One of these branches included two well-supported clusters: cluster A, comprised of 81.0% French isolates, and cluster B, comprised of 70.6% Swedish isolates. The H. acinonychis strain Sheeba prophage II integrase (Hac) formed a sister group with these two clusters. The other main branch of the tree had two well-supported clusters: cluster C, comprised exclusively of Asian strains, and cluster D, comprised of 6 French, 4 German, 3 Portuguese, and 3 Swedish strains. Country of origin is indicated at the beginning of each strain designation: Fr, France; Sw, Sweden; De, Germany; Pt, Portugal; Uk, United Kingdom; Jp, Japan; Ko, South Korea; T, Taiwan; Vn, Vietnam; Eg, Egypt. Disease status is indicated in the end: G, chronic gastritis; U, ulcer; M, MALT lymphoma; C, cancer (atrophic gastritis or gastric adenocarcinoma).

DISCUSSION

We have identified an integrated prophage in the genome of H. pylori strain B45 which is inducible by UV irradiation. This prophage is morphologically similar to phages of the Siphoviridae family. No prophage sequence has been reported for H. pylori, except for a remnant prophage sequence in the B38 strain; however, our study confirms two previously published articles reporting spontaneous production of small amounts of phage particles by an H. pylori strain (20, 48). Interestingly, all three reports describe the morphology as compatible with that of Siphoviridae; as with the phage described in the present paper, these previous articles reported infection, propagation, and electron microscopy images of a Helicobacter phage with a morphology compatible with that of Siphoviridae. The phage characteristics described by Heintschel von Heinegg et al. (i.e., phage heads of around 50 to 60 nm and DNA estimated to be 22,000 bp in length) are also in line with the B45 prophage (20).

It remains to be determined whether the capacity to produce phage is rare among H. pylori strains or whether a substantial proportion (approximately 21%) may be capable of producing phage. The methods used for phage induction in the present study are most likely suboptimal, since relatively few phage particles were observed and there were no visible plaques (Fig. 2). Other strategies may be more effective, such as exposing the bacteria to physical stress, leading to an induction of coccoidal forms, or by using a subinhibitory concentration of an antibiotic that induces DNA damage instead of chemical stress (49, 50). However, the optimal conditions for phage induction are difficult to predict. Nevertheless, it would be of interest to know whether randomly selected H. pylori strains are susceptible to infection by phage or whether most are resistant.

A transposable element homolog to ISHp608 (46) was found in the B45 prophage. Interestingly, one copy of ISHp609 has been described in the H. pylori B38 prophage. The presence of such insertion sequences (IS) in B38 has been viewed as a sign of a degenerative process (39). In prophages from Gram-negative bacteria, “extra genes” called “morons” can be found near the prophage DNA ends, at strategic positions, and these can interfere with the functions required during lysogeny and for lytic infection (45). However, in prophages some morons can be flanked by such transposase genes in their vicinity, pointing to alternative methods of mobility. Therefore, the presence of ISHp608 could be also viewed as a “poison” inside the phage DNA interfering with the inducibility of the phage, which may provide one putative explanation of the small amount of phage particles observed in the present study.

Only 23 of the 27 prophage genes displayed detectable protein sequence similarity with annotated genes in databases. This is most likely due to the high evolutionary rate of bacteriophages; identification of unknown sequences is expected when analyzing a novel phage genome (45). This could explain why some of the genes typically existing in a phage genome could not be identified, including a head-like and a tail fiber protein. The origin of this Helicobacter prophage could not be determined based on the data obtained in the present study. Siphoviridae is the best-documented phage family, with more than 60 complete phage genomes sequenced. They were described among a large panel of bacteria including Gram-positive bacteria (Lactococcus, Lactobacillus, Streptococcus, and Staphylococcus) and Gram-negative bacteria (in particular, Enterobacteriaceae members such as Escherichia coli) as well as Mycobacterium (44). Due to this large diversity, limited conclusions could be drawn regarding the H. pylori phage biology based on the observation that it resembles a Siphoviridae phage.

In order to assess the prevalence of prophages in H. pylori, we performed a PCR screen for the integrase gene and found this gene in a surprisingly high proportion of screened isolates (21.4%, 73/341). Although some H. pylori positive strains may have been missed due to primer mispairing, the prevalence was higher than that previously observed by Thiberge et al. (39), where DNA hybridization based on three prophage CDSs was carried out. In fact, due to the genetic variability of the prophage CDSs, including the integrase gene, hybridization results may be skewed. Altogether, this does suggest that prophages are relatively widespread among H. pylori strains; however, integrase-positive isolates may not carry complete prophages. To address this point, an extensive genome sequencing strategy based on H. pylori prophage-positive strains is under way.

Considering reported associations between prophages and virulence in other bacteria (e.g., Streptococcus agalactiae prophages in neonatal meningitis isolates [51]), we were expecting to find implications for H. pylori virulence. However, no significant association was found between the presence of the integrase gene and specific H. pylori gastroduodenal disease. This does not exclude the possible association of complete prophage sequences, or of specific prophage genes, with virulence. However, phylogenic analysis of the detected phage integrase gene showed a pattern of biogeographic separation. This is in agreement with a model of coevolution between the virus and its bacterial host, since biogeographic separation is also observed within H. pylori. A model of geographically constrained viral dispersal also fits, i.e., H. pylori strains from different geographical regions may have been infected by distinct phage lineages after the geographic separation of the bacterial host. Another hypothesis could be that after bacterial infection with the virus, the divergence of the bacterium is also accompanied by divergence of the integrated virus. The fact that the genetic content and organization of the B45 prophage are similar to those of H. acinonychis strain Sheeba prophage II is of particular interest. In fact, it has been argued that H. acinonychis strain Sheeba is derived from H. pylori and that the prophage was acquired after the host jump from human to feline (22). Our study indicates that an equally likely scenario is that the prophage was present in the bacterial genome before the host jump.

Usually, phage DNA is integrated in the vicinity of a tRNA sequence. For Helicobacter species, prophages are integrated into protein-encoding genes. The B45 prophage is flanked on both sides by partial gene sequences corresponding to a type I R-M enzyme subunit. Indeed, R-M systems are often linked with mobile genetic elements, such as viruses. Examples of linkage between phages and restriction and R-M systems have already been described, i.e., HindIII on Phi-flue, Sau42I on phi-42, EcoO109I on P4-like prophage, and type I RM on a prophage annotated as a genomic island 5, among others (52, 53). This integration site is not conserved among Helicobacter prophages; the H. acinonychis strain Sheeba prophage II is integrated into a camphor resistance gene and carbamoylphosphate synthase small chain, while the B38 prophage is located between a hypothetical protein and cysteine-rich protein G. The integration of the prophage in the B45 genome does not seem to create any significant genome rearrangement, because the numbering of the flanking gene on each site is conserved when taking into account all reference genomes.

There is ample evidence for continued exchange of genetic material between phages, bacterial genomes, and various other genetic elements (54). This explains the sometimes fuzzy distinction between phages, plasmids, and pathogenicity islands (PAIs) and the chimeric nature of some phage genomes (55, 56). Nonetheless, several factors are typical of a PAI, like the cag PAI, which are not shared by the H. pylori prophage: (i) GC content clearly different from that of the host, (ii) hot spot for IS, (iii) integration within an essential gene (glutamate racemase), (iv) encoding of a secretion machinery specific for effectors (type IV secretion system for cag), and (v) effector molecules (CagA and peptidoglycan) translocated into the host cell by a contact-dependent secretion system (57). The large CDSs in the middle of the B45 prophage have significant similarities to Hac_1615 (Hac_1615|-|Hac prophage II orf11|mosaic CUP0956/HP1116/jhp1044-like protein) and a JHP1044-like CDS in another part of the genome. Interestingly, JHP1044-like proteins have been described in an article aimed at identifying strain-specific genes located outside the plasticity zones of H. pylori genomes (58). An exciting hypothesis would be that the plasticity zones, which are increasingly being considered true PAIs (13, 59), have a phage origin. This kind of association between phage and PAI has indeed already been described in other bacteria such as Staphylococcus aureus (60, 61).

Interestingly, the Hac1615-like protein also shows similarity to a crystalline beta/gamma motif-containing protein, in particular with a protein corresponding to a putative lysin found in Bordetella bronchiseptica phage BPP-1 (62). The role of phage lysin is to make the bacterial membrane more permeable, thus allowing the entry of the phage DNA. Lysin targets the integrity of the cell wall and is designed to attack one of the major bonds in the peptidoglycan. During the lytic cycle of phages, lysin also participates in bacterial lysis. With few exceptions, lysins do not have signal sequences, so they are not translocated across the cytoplasmic membrane. When lysins are used as antimicrobial agents, they are believed to work only with Gram-positive bacteria, where they have access to the cell wall carbohydrates and can attack the peptidoglycan, whereas the outer membrane of Gram-negative bacteria prevents this interaction. However, the presence of the B45 phage carrying a putative lysin suggests that it may be possible to identify novel lytic enzymes for targeting Gram-negative bacteria (63, 64). These could potentially be engineered to lyse cells and be used as antimicrobial agents against H. pylori, for example, and possibly delivered through liposomes (65).

The phage acquisition can play an important role in short-term adaptation processes for H. pylori. The high percentage of isolates carrying an integrase gene highlights the idea that prophages are underestimated in H. pylori. It is possible that they shape the genome at least in terms of the diversity of strains found worldwide as well as contributing to virulence evolution. The high percentage of isolates carrying the integrase gene of a prophage strongly suggests its importance in the diversification and fitness of the genome architecture in H. pylori. We previously published a paper in which we described the discovery of new sequences in the H. pylori genome (66). One of our conclusions was that these new genes could correspond to bacteriophage receptor/invasion proteins. This highly speculative conclusion is further supported by the present study. We believe that the description of the B45 prophage is an important contribution that will aid future genome analysis in H. pylori.

MATERIALS AND METHODS

H. pylori strain B45.

The B45 strain was isolated from a 41-year-old male patient enrolled in a French multicenter study for low-grade gastric MALT lymphoma (67). B45 harbors a functional cag PAI (41), has a vacA s1m2 genotype (67), and belongs to the European multilocus sequence typing (MLST) cluster (based on data kindly provided by Sebastian Suerbaum, Hannover, Germany). This strain is susceptible to amoxicillin and clarithromycin, the two antibiotics given as the first line in eradication treatment.

Other H. pylori strains included in the present study.

The other 340 H. pylori strains included in the present study came from several origins: 134 came from the collection of the French National Reference Center for Campylobacters and Helicobacters (F. Mégraud and P. Lehours, Bordeaux, France); 78 from the Faculty of Engineering, Catholic University of Portugal (F. F. Vale); 54 from the Department of Microbiology, Tumor and Cell Biology, Karolinska Institute (Lars Engstrand); and 28 from the Klinikum Rechts Der Isar II, Medical Department, Technische Universität (M. Gerhard, Munich, Germany), and 46 Asian strains were provided by Y. Yamaoka (Department of Medicine-Gastroenterology, Michael E. DeBakey Veterans Affairs Medical Center and Baylor College of Medicine, Houston, TX) and M. Oleastro (Department of Infectious Diseases, National Institute of Health, Lisbon, Portugal).

Genome sequencing.

Genomic DNA from the B45 strain was extracted from cultures originating from a single colony, by using a commercial kit (Qiagen SA, Courtaboeuf, France) including an RNase step. The quality and concentration of the extracted DNA were verified using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Wilmington, DE) and also by running the extracted DNA on an agarose gel. Five micrograms of DNA was used for 454 library preparations according to the manufacturer’s protocol (April 2009) (454 Life Sciences, Branford, CT) and sequenced in one lane of a 4-region FLX Titanium plate on a Roche 454 FLX instrument. Data were processed with the accompanying 2.3 software package.

Genome assembly.

The sequence generated included at least 26.4-fold coverage of Roche 454 Life Sciences FLX fragment data. The Roche 454 Life Sciences sequence was assembled using the Roche 454 Life Sciences Newbler Metrics assembler (software release 2.0.00.20). Only contigs longer than 500 bp were considered for data analysis.

Gene predictions and orthology analysis.

We used the free access AMIGene (Annotation of Microbial Genes) pipeline available on the Genoscope website (http://www.genoscope.cns.fr/agc/tools/amiga/Form/form.php) (33), which is an application designed for automatically identifying the most likely coding sequences (CDSs) in a large contig or complete bacterial genome sequence. The contigs were numbered according to the Newbler Metrics assembler, and the CDSs identified among each contig were numbered according to AMIGene prediction: for example, C1B450001 corresponded to the first CDS identified in contig 1.

For defining H. pylori orthologous genes, the proteins of seven reference H. pylori genomes and of Hp_B45 were cross-compared using BLASTp (68) and clustered into orthologous groups using the Markov clustering (MCL) algorithm (34, 35). The complete reference H. pylori genomes available at the time of the analysis (Hp_26695, Hp_AG1, Hp_B38, Hp_G27, Hp_J99, Hp_P12, and Hp_Shi470) were downloaded from NCBI.

Before MCL, the BLAST output file was filtered such that (i) alignments with lower than min_identity sequence identity were removed and (ii) alignments with lower than min_length_ratio (alignment length)/(protein length of the longest of the two proteins) were removed. To find optimal values for min_identity and min_length_ratio, as well as for the granularity parameter I of the MCL algorithm, simulations were run using only the reference H. pylori proteomes. All combinations of a series of different values for the three parameters were run (min_identity: 50, 60, … 90; min_length_ratio: 0.3, 0.4, … 0.9; I: 1, 2, … 5), and the combination rendering the largest number of clusters with exactly one protein per genome was selected. The rationale for this was that too-stringent criteria would likely split orthologous groups and produce groups lacking representatives in some genomes, while too-sloppy criteria would likely expand groups to also include paralogs and produce groups with multiple representatives in some genomes. The optimal conditions were found to be min_identity = 70, min_length_ratio = 0.5, and I = 1 (I had only minor effect compared to the other parameters). This parameter setting was then used for running the clustering procedure on all H. pylori proteomes, including Hp_B45.

H. pylori strains and growth conditions for phage induction.

The strain 26695, whose genome does not have any integrated prophages (36), was selected as the negative control. As positive control, the selected strain was JP1, which produces lysis plaques after UV induction (21). H. pylori strains were grown on conventional in-house selective agar medium (Wilkins-Chalgren agar supplemented with 10% human blood and a mixture of antibiotics: vancomycin, trimethoprim, amphotericin B [Fungizone], and cefsulodin) or liquid medium (brucella broth supplemented with 10% Polyvitex [bioMérieux, Marcy l’Etoile, France] and 10% filtered fetal bovine serum [Gibco Invitrogen, Cergy Pontoise, France]) under microaerophilic conditions at 37°C.

For prophage induction, both chemical (mitomycin C from Streptomyces caespitosus; Sigma-Aldrich, St. Louis, MO) and physical (UV irradiation at 254 nm) agents were used. For mitomycin C induction, strains were resuspended in brucella broth and mitomycin C was added to final concentrations of 1 µg/mL, 0.5 µg/mL, and 0.1 µg/mL. Twenty µL spots of this bacterial suspension were applied on H. pylori selective medium, dried, incubated under standard conditions, and observed for lytic plaque formation. Mitomycin C induction was also performed in liquid medium at the same concentrations, and the absorbance at 600 nm was checked at regular intervals. Controls without mitomycin C were used.

The induction with 254-nm UV irradiation (Bioblock Scientific VL 6C/6W UV lamp, 254 nm; power, 12 W; or Bulbworks BW.G8T5 UV lamp, 8 W, 253.7-nm UVC) was performed by exposing inoculated (20-liter spots) H. pylori selective medium plates to different periods of UV irradiation at a distance of 13 cm from the UV source. For each UV lamp used, the D value (decimal reducing time, which is the time necessary to reduce 90% of the bacterial population) was determined for Escherichia coli K-12. The D value was used for exposure of the H. pylori cultures to 0, 1, 5, and 10 times the D value of UV irradiation. The plates were incubated and checked for lytic plaques. The induction of H. pylori in liquid medium was performed by exposing 500 ml of brucella broth culture to 1 D value after 24 h of incubation, followed by 24 additional hours of incubation. Then, after phage precipitation, the broth was used for transmission electron microscopy (TEM) analysis after phage precipitation and phage DNA extraction (see below).

Phage particle concentration.

Phage particles from H. pylori strain B45 were concentrated using a protocol adapted from that of Henn et al. (69). Briefly, a 48-h culture of H. pylori grown in liquid medium was centrifuged to pellet cells (4,000 rpm, 10 minutes), which were discarded. The supernatant was gently decanted to a new tube. Four milliliters of phage precipitant (33% polyethylene glycol [PEG], 3 M NaCl) was added to the supernatant, followed by incubation overnight at 4°C. To pellet phage particles, a centrifugation step (10,000 rpm, 10 min at 4°C) was applied. The supernatant was gently discarded, and residual liquid was drained using a paper towel. The phage pellet was resuspended in 500 liters of phage buffer (150 mM NaCl, 40 mM Tris-HCl [pH 7.4], and 10 mM MgSO4) and transferred to a 1.5-ml Eppendorf tube for further processing.

Extraction of nucleic acid from concentrated phage particles.

The DNA from concentrated phage particles was extracted using a Qiaprep miniprep kit (Qiagen SA, Courtaboeuf, France) and recovered in 50 liters of sterilized distilled water. To ensure complete elimination of bacterial genomic DNA, the minipreparation was treated sequentially with two exonucleases and resuspended in the same volume of water to eliminate the enzymes (after heat inactivation and DNA precipitation). The enzymes used were (i) exonuclease I (E. coli) (New England Biolabs, Ipswich, MA) and (ii) lambda exonuclease (New England Biolabs) sequentially. The goal was to cut any genomic remnant DNA trace, whereas putative circular closed DNA (i.e., phage DNA) could not be degraded by any of these enzymes. The extracted DNA was then tested for the presence of phage DNA and bacterial genomic DNA by PCR amplification of the phage integrase gene (see below) and the cagA gene as already described (67, 70).

Transmission electron microscopy.

Two kinds of electron microscopy analyses were performed: negative staining or fixation embedding and sectioning. For negative staining, Formvar membrane-covered 300-mesh copper grids were floated for 1 to 2 minutes on a small drop (10 μl) of the virus suspension for particle adsorption to the membrane. The grids with adsorbed particles were floated on a drop (50 μl) of aqueous 1% uranyl acetate in bidistilled water for 1 min and then transferred to a second similar droplet for 1 additional minute. In some preparations, additional samples were stained with 2% aqueous phosphotungstic acid using a similar procedure. Excess stain was drained with filter paper and then air dried. The resulting samples were observed using a JEOL 100SX electron microscope. For fixation embedding and sectioning, bacteria in suspension were prefixed in 0.2 mol/liter cacodylate buffer (pH 6.8) containing 5% glutaraldehyde (50/50 mixture with the culture medium) for 2 h at 4°C. After centrifugation at 8,000 rpm for 3 minutes, bacteria were postfixed in 1% osmium tetroxide buffered in 0.1 mol/liter of cacodylate (pH 6.8) for 1 h at room temperature in the dark. Pellets were included in 1% agar and cut in small pieces (2 mm3), and specimens were dehydrated in an ethanol series (50%, 70%, 95%, 100%). Dehydration was completed in propylene oxide, and the specimens were embedded in epoxy resin (Epon 812; EMS, Hatfield, PA). The resin was polymerized at 60°C for 48 h. The samples were sectioned using a diamond knife on an ultramicrotome (Ultracut-E; Leica Microsystems, Nanterre, France). Thin sections (70 nm) were picked up on copper grids and then stained with uranyl acetate and lead citrate. The grids were examined with a transmission electron microscope at 120 kV (H7650; Hitachi, Tokyo, Japan).

Determination of the prevalence of the prophage integrase gene in a collection of H. pylori isolates.

To identify the prevalence of prophage sequences in H. pylori strains, a PCR screening strategy was used with degenerated primers F1, AAGYTTTTTAGMGTTTTGYG, and R1, CGCCCTGGCTTAGCATC, designed using Primer 3 (http://frodo.wi.mit.edu/primer3/) on the aligned integrase sequences of B38 (HELPY_1521), B45, and H. acinonychis strain Sheeba (Hac_1606) obtained using multiple sequence alignment with hierarchical clustering (71). These primers generated a 529-bp PCR product. A total of 341 H. pylori strains were screened (117 from duodenal ulcer, 110 from gastritis, 63 from gastric MALT lymphoma patients [including strain B45], and 51 from atrophic gastritis or gastric adenocarcinoma patients). Most of these strains came from Europe (France, 125; Sweden, 56; Germany, 53; Portugal, 29; United Kingdom, 7; Norway, 6), 46 strains came from Asia (Japan, 26; Thailand, 3; Vietnam, 7; Taiwan, 4; South Korea, 6), 18 strains came from Africa (Egypt, 7; Burkina Faso, 11), and 1 strain came from South America (Costa Rica). For each positive strain, amplicons were purified using MicroSpin S-400 HR columns (GE Healthcare, Saclay, France) and directly sequenced on both strands as previously described (67).

Phylogeny analysis of phage integrase sequences.

Phylogeny analysis was performed on the integrase sequences obtained on the 73 positive strains screened in the present study: a 442-bp sequence was available for each sequenced PCR product. We also included the internal fragment of the Hac_1606 integrase sequence (YP_665314.1|) corresponding to the amplified product.

The evolutionary history of phage integrase genes was inferred using the neighbor-joining method. Neighbor-joining phylogenetic tree topologies of nucleotide alignments were constructed using the MEGA (Molecular Evolutionary Genetics Analysis) 3.1 software (72), on the basis of distances estimated using the Kimura two-parameter model (73). This model corrects for multiple hits, taking into account transitional and transversional substitution rates. Branching significance was estimated using bootstrap confidence levels by randomly resampling the data 1,000 times with the referred evolutionary distance model.

Nucleotide sequence accession numbers.

GenBank accession numbers for the 63 contigs obtained after genome assembly are AFAO01000001 to AFAO01000063. The GenBank accession number for the final prophage sequence is JF734911. The corresponding GenBank accession numbers for prophage integrase gene sequences are JF734911 to JF734984.

B45 coding sequences with significant homologies with the cag pathogenicity island. The cluster numbers are indicated on the left side of the table. The corresponding CDSs identified in the H. pylori reference genomes are indicated. Absent, no homolog. The B38 genome is a cag PAI-negative reference genome (39).

Table S1, XLS file, 0.1 MB.

B45 coding sequences with significant homologies with Helicobacter acinonychis strain Sheeba prophage II and Helicobacter pylori B38 prophage. For each CDS, the following are provided: NCBI accession number of match, gene and synonym, annotation, protein length, alignment identity, alignment E value, and alignment length. C21B450028-C73B450001-C47B450001 corresponds to the fused CDSs obtained after PCR and sequencing (see Results).

Table S2, DOC file, 0.1 MB.