Genomic Changes during Chronic Helicobacter pylori Infection

The gastric pathogen Helicobacter pylori shows tremendous geneticvariability within human populations, both in gene content andat the sequence level. We investigated how this variabilityarises by comparing the genome content of 21 closely relatedpairs of isolates taken from the same patient at different timepoints. The comparisons were performed by hybridization withwhole-genome DNA microarrays. All loci where microarrays indicateda genomic change were sequenced to confirm the events. The numberof genomic changes was compared to the number of homologousreplacement events without loss or gain of genes that we hadpreviously determined by multilocus sequence analysis and mathematicalmodeling based on the sequence data. Our analysis showed thatthe great majority of genetic changes were due to homologousrecombination, with 1/650 events leading to a net gain or lossof genes. These results suggest that adaptation of H. pylorito the host individual may principally occur through sequencechanges rather than loss or gain of genes.

INTRODUCTION

Top

Introduction

Helicobacter pylori is thought to be the second-most-prevalenthuman bacterial pathogen, infecting the stomachs of more thanone-half of the human population (20). H. pylori induces chronicgastritis in all infected persons and causes complications suchas peptic ulcers, gastric cancer, and gastric lymphoma of themucosa-associated lymphoid tissue in subgroups of patients.H. pylori shows exceptional genetic variability and intraspeciesdiversity. Allelic diversity is so high that almost every unrelatedisolate of H. pylori has a unique sequence when a fragment ofseveral hundred base pairs is sequenced from housekeeping orvirulence genes (7, 12, 19). This allelic diversity is the resultof the combination of a high (mutator-type) mutation rate (4),a high frequency of recombination between strains during mixedcolonization (8, 13, 19), and the ability of H. pylori to integrateunusually small pieces of exogenous DNA into its chromosome(6, 19).

Much of the current knowledge about genetic diversity of H.pylori has been gained from population genetic analysis of sequencedata obtained from H. pylori strains isolated from unrelatedpersons of diverse geographical origins (1, 17). Since H. pyloriinfection is chronic and can persist for decades, an alternativeapproach is the analysis of genetic relationships between H.pylori strains that have been isolated sequentially from thesame patient. We have used mathematical modeling on sequencedata from 24 pairs of sequential H. pylori isolates from Colombiaand the United States to estimate basic parameters of recombinationand mutation of H. pylori in vivo. The data showed that theestimated mean size of imported fragments was only 417 bp, theshortest import length that has been reported for a bacterium.Half of the genome is replaced by homologous sequence in a timeframe estimated to be from 40 to 2,000 years (6). The wide credibilityregion of this estimate is principally due to a high degreeof uncertainty about the time since the last common ancestorof each pair of strains, which were taken from middle-aged patientsat intervals from 3 months to 4 years, but even the higher valuesin this range represents extremely rapid import of DNA comparedwith other bacteria (6).

H. pylori strains not only show allelic diversity but also differin their gene contents. The two completely sequenced genomesof H. pylori strains 26695 and J99 share only 94% of their genes(2), whereas approximately 7% of the genes are unique for eachstrain, respectively (22). A comparison of 15 strains with DNAmicroarrays showed that in this set of strains, 1,281 of 1,643genes present on the array (combined from 26695 and J99 genomes)were shared by all strains, whereas 22% of genes were only presentin some of the strains (15). We have recently further refinedthe estimate of the number of genes present in all H. pyloristrains to 1,111 by comparative genome hybridizations performedwith a globally representative collection of 54 H. pylori strains(9). Furthermore, Israel et al. found numerous genomic differencesbetween H. pylori J99, a strain whose genome sequence was publishedin 1999, and multiple isolates cultured from the same patientat the same time as J99 and 6 years later (10). Similar resultswere found in another study, where two clones of a patient weretaken at the same time and compared to each other (3).

These results suggested that the H. pylori genome has a highlyplastic gene content. To quantitate the rate of genomic changes(gene loss or acquisition) during chronic colonization and tocompare it to rates of allelic replacement events, we have usedwhole-genome DNA microarray hybridization to study gene lossand gene acquisition in the same collection of sequential H.pylori isolates previously studied to determine the rates ofmutation and recombination in vivo (6). Where the microarraydata indicated a possible genomic change, the event was verifiedby PCR and sequence analysis. Changes of gene content occurredin 5 of the 21 pairs analyzed, indicating significant plasticityof the genome content. However, quantitative analysis showsthat genetic events leading to loss or acquisition of genesare far less frequent than events leading to allelic exchangesalone.

MATERIALS AND METHODS

Top

Materials and Methods

Bacteria. Strains were isolated and single-colony purified from patientsin Narino, Colombia (10 pairs; designated the NQ strain collection),and New Orleans, LA (14 pairs; designated the LSU strain collection),as described previously (6, 21). The time interval between theisolations of the first and follow-up strains was 3 to 48 monthsfor the LSU strains and 36 months for the NQ strains. Of the24 pairs characterized in the previous study (6), 2 were excludedbecause they were completely unrelated according to their DNAsequences. Two further pairs of strains were identical at 3and 5 of the 10 sequenced fragments, respectively, making theirdegree of clonal relatedness ambiguous. These strains were analyzedby microarrays. The former pair was unrelated in gene contentaccording to microarray analysis and was excluded from furtheranalysis, while the latter pair was identical (no single difference)according to microarrays and was included, resulting in 21 pairsof related strains in total.

Whole-genome DNA microarray. The composition of the H. pylori whole-genome microarray usedin this study has been previously described (14). It is basedon PCR products derived from the two available H. pylori whole-genomesequences of strains 26695 (22) and J99 (2), and contains 1,655probes representing 96.1% of genes present in H. pylori 26695or J99. The performance of the array for H. pylori genome comparisonswas validated by experiments comparing the two strains, 26695and J99, whose genomes have been sequenced. These experimentsshowed that we detected 100% of the known strain-specific genesof J99 and 97.3% of the known strain-specific genes of 26695as defined by Salama et al. (15). Four strain-specific genesof 26695 (HP0315, HP1001, HP1516, and HP1537) were incorrectlyscored as present in J99 when a conservative cutoff of 2 forthe ratio was used but scored as absent using a cutoff of 1.5.For these genes, overall low signal intensities were obtainedunder our hybridization conditions, very likely due to shortsequence lengths and/or low G-C content of these genes, makingscoring errors more likely.

Hybridization. Fluorescent labeling of DNA and competitive hybridizations wereperformed as described by Salama et al. (15), with several modifications.Whole-genomic DNA was prepared from blood agar plate-grown H.pylori with QIAGEN Tip-100 columns. Labeling of the DNA withaminoallyl-dUTP was performed with the BioPrime DNA labelingsystem (Invitrogen) minus the deoxynucleoside triphosphates(dNTP) set. A dNTP set was made using 0.5 mM dGTP, dATP, dCTP,0.3 mM dTTP (Amersham Pharmacia), and 0.2 mM aminoallyl dUTP(Sigma). One microgram of DNA was diluted to 24 µl inwater, 20 µl of random octamer-primer solution was added,and they were heated for 5 min at 99°C. The DNA was thenplaced directly on ice, 5 µl of dNTPs and 1 µl ofhigh-fidelity Klenow enzyme were added to the reaction mixture,and the reaction was incubated at 37°C for 1 h. The reactionwas stopped with 5 µl 0.5 M Na₂EDTA, pH 8.0. Purificationwas done by a modified QIAGEN PCR purification protocol. PEand AE buffers (QIAGEN) were replaced with 5 mM KPO₄ (pH 8.0)-80%EtOH and 4 mM KPO₄ (pH 8.5), respectively. The Cy3 and Cy5 dyes(Amersham Pharmacia) were diluted using 72 µl of water,and 4.5 µl was aliquoted and dried before storage in thedark at 4°C. Labeling of the probe with the Cy dyes followedthe protocol by Salama et al. (15) with the exception that beforehybridization the Cy3 and Cy5 probes were combined and 100 µgof yeast tRNA was added. After being dried, the probe was resuspendedin 32 µl of hybridization buffer (50% formamide, 6x SSC[1x SSC is 0.15 M NaCl plus 0.015 M sodium citrate], 0.5% sodiumdodecyl sulfate, 50 mM Na phosphate [pH 8.0], and 5x Denhardt'ssolution). The probe was heated for 2 min at 99°C and appliedto the slide. Microarray scanning and data processing were preformedas previously described (11, 18).

Empty-site PCR and sequencing. All loci where the microarray analysis indicated that loss orgain of genes had occurred were further characterized by PCRand sequence analysis. The loci were amplified by PCR with primerstargeting genes that flank the deletion-acquisition event. PCRproducts were purified and sequenced as previously described(19).

Southern hybridization. Loss of insertion (IS) element copies was verified by Southernblot hybridization under high-stringency conditions (16). GenomicDNAs of the LSU1062 strains and the pair NQ315-NQ1712 were digestedby EcoRI and electrophoretically separated on a 1% agarose gel.A 600-bp fragment of HP0414, which is part of the IS606 element,labeled with digoxigenin-dUTP was used as the probe.

Accession numbers. All new sequences generated in this study have been submittedto the GenBank/EMBL/DDBJ databases (accession numbers AM086402to AM086418).

RESULTS

Top

Results

Comparisons of the genomes of sequentially isolated H. pylori strains by whole-genome DNA microarray hybridization. We analyzed the genomes of 21 pairs of H. pylori strains isolatedfrom predominantly middle-aged, adult patients in Narino, Colombia(NQ strains), and New Orleans (LSU strains). Each pair of isolateswas taken from the same patients at two different time points(these time points are referred to below as early and later).These pairs were compared by competitive hybridization witha whole-genome DNA microarray based on the two published genomesequences (14). The strategy used to screen for genomic changesin sequential isolates of H. pylori was a variant of the classicalapproach to comparative genome hybridization, where a test strainis hybridized against the strain upon whose genome the microarrayis based (15). To permit more reliable identification of genomicchanges, pairs of sequential isolates were directly hybridizedagainst each other. Thus, genomic DNA of the paired strainswas labeled with Cy3 or Cy5 and competitively hybridized tothe microarray. The arrays were scanned for the two colors separately,and a signal ratio was calculated.

A signal ratio of 1 indicated the presence of the gene in bothstrains. Signal ratios of <0.5 or >2 were interpretedas evidence of potential loss or gain of the respective gene.When both channels yielded signal intensities of <350, thegene was excluded from the analysis, the most likely reasonbeing the absence of the gene in both strains. This cutoff formissing genes was determined empirically, with data from comparativegenome hybridizations of either strain LSU2003-1 or strain LSU1010-1,with 26695 and J99 on the same microarray (9) being used forcalibration.

26695 and J99 have 1,590 and 1,495 predicted genes, respectively,of which 96% were represented on the array. An average of 1,453spots were detected per tested genome pair. Assuming that theyhave a genome size similar to those of J99 and 26695, this indicatesthat the great majority of genes contained in the NQ and LSUstrains were assayed by our array system.

In three of seven of NQ strain pairs and 9 out of 14 of LSUstrain pairs, microarray hybridizations did not yield evidenceof single-genome content difference. These pairs were not investigatedfurther. In the remaining nine pairs, microarray hybridizationsindicated potential gene content differences at one or moreloci.

To validate the results of microarray hybridizations and tofurther analyze the events underlying the genomic changes, "empty-sitePCRs" were designed for all changes indicated by the microarrayhybridizations and sequenced. Primers were designed based onthe sequences of genes flanking the predicted genomic events.The validation experiments confirmed the predicted changes insix cases (Table 1; Fig. 1 and 2A). An additional two eventsindicated by the microarray experiments could be explained byhomologous recombination events, which changed the degree ofsequence homology to the DNA spotted on the microarray but involvedno loss or gain of sequence (Fig. 2B). Analysis by PCR and sequenceanalysis did not confirm five of the changes that were predictedfrom the microarray results, because the genes were found eitherto be absent from both strains or not to show any sequence differences.Only those events validated by sequencing were included in thesubsequent calculations.

View this table:
[in this window]
[in a new window]
TABLE 1. Comparison of the numbers of genetic changes identified by multilocus sequence analysis (6) with the numbers of genomic changes identified by microarray hybridization and sequence analysis^a

View larger version (21K):
[in this window]
[in a new window]
FIG. 1. Genomic differences between paired sequential isolates of H. pylori. The line marked "consensus" represents genes present in both strains. Genes and/or sequences only present in one strain are indicated above and below the consensus lines. Mosaic sequences are indicated in red, the coding sequence is shadowed in dark gray, and the noncoding sequence is shown in light gray. (A) Loss of the complete cag PAI in strains NQ315/1712. The cag PAI was deleted at a 31-bp repeat. The sequences flanking the repeats were identical in both strains. (B) Partial loss of the cag pathogenicity island in the LSU1062 pair. The earlier strain contained an IS606 element at the 3' end of the cag PAI. The later strain lost half of the 5' end of gene HP0527, genes HP0528 to -0548, and most of the IS606 element. See the text for details. (C) Uptake or loss of the restriction-modification system HpyAIV in strains LSU1016. The RM system occurs in place of a putative open reading frame of unknown origin (green; the orf gene is indicated by an arrow). The genome alteration was mediated by a recombination event. (D) Incorporation and/or loss of a putative iron binding protein (ceuE) gene in NQ315/1712. The change in copy number of the ceuE gene was mediated by recombination. (E) Partial loss of pseudogenes (marked by asterisks) HP0903/0904 in strains LSU1040. The earlier strain contained a partially deleted gene (HP0903) and a complete gene (HP0904), whereas the deletion of gene HP0903 was longer in the later strain and included part of gene HP0904. The sequence showed a recombination event of at least 600 bp in gene HP0904. The sequence of the later strain was 1,000 bp shorter than for the earlier strain.

View larger version (18K):
[in this window]
[in a new window]
FIG. 2. Sequence mosaics created by recombination events during chronic colonization with H. pylori, leading to gain or loss of sequence (A) or to allelic replacement only (B). Numbers on the left indicate the strain pair and the genes affected by the event. The gene numbers refer to H. pylori strain 26695. Sequenced regions are indicated by rectangles. Empty rectangles represent sequence identity between the two sequential strains. Vertical lines indicate the positions of polymorphic nucleotides. Below the rectangles, the lengths of sequence mosaics and of the flanking regions with identical sequence are indicated. (A) Characteristics of mosaic sequences adjacent to recombination-mediated genomic changes. Two events showed a flanking mosaic only on one side of the gene deletion-acquisition event. The column marked "net sequence difference" indicates the difference between the lengths of sequences between the regions depicted. See the legend to Fig. 1 for details of these gene deletion-acquisition events. (B) Characteristics of allelic replacement events without gain or loss of sequence.

Changes involving the cag PAI. Two sequentially isolated strain pairs showed differences inthe cag pathogenicity island (cag PAI) (5). NQ1712 lacked theentire cag PAI and one of the two imperfect copies of the 31-bprepeat sequence flanking the island in the earlier isolate NQ315(Fig. 1A). The sequences upstream and downstream of the remainingcopy of the repeat were identical to the sequence of NQ315,indicating that a deletion had occurred by intrachromosomalrecombination between the two 31-bp repeat sequences and subsequentexcision of the PAI, rather than by allelic replacement afterimport of an empty-site allele from an unrelated strain, asdescribed by Kersulyte et al. (13).

A partial loss of the cag island was observed in a second pair,LSU1062 (Fig. 1B). The earlier strain, LSU1062-1, containeda complete cag PAI that was followed by a complete IS606 element,consisting of tnpA and tnpB homologs. The deletion in LSU1062-3started in the "middle repeat region" of gene HP0527 (cagY)and comprised all remaining cag island genes (HP0528 to -0548),as well as part of the IS606 element (tnpA and the first 192nucleotides of tnpB).

Changes involving IS elements and restriction-modification systems. Within the LSU1062 pair, the later strain showed evidence ofhaving lost its only copy of the IS606 element. The 26695-J99whole-genome microarray contains several spots that hybridizewith components of IS elements (e.g., the tnpA and tnpB homologsHP0413/0414, HP1007/1008, and JHP826/827). In the LSU1062 pair,none of these spots showed hybridization with the later strain,in contrast to the earlier strain. To ascertain that no othercopies of IS elements were still present in the genome of thelater strain, a Southern blot hybridization was performed, whichconfirmed that the later strain did not contain any copy ofthese IS elements. Southern blot hybridization also showed thatan IS606 element was lost in the strain pair NQ315/1712. Thelocation of this event could not be determined.

The later strain of the LSU1016 pair contained the restriction-modificationsystem HpyAIV (HP1351/1352), of which only small fragments werepresent in the earlier strain, interrupted by a putative openreading frame of unknown function. In the later strain, thisregion contained the complete restriction-modification system(Fig. 1C).

Changes involving the plasticity region. In addition to the cag PAI, a region termed the "plasticityzone" has been described as harboring the majority of strainspecific genes (2). Microarray hybridization indicated that14 plasticity region genes (HP0990 to -0995 and JHP929 to -936)were present in the later strain of the LSU1014 pair that wereabsent from the earlier strain. The event underlying this changecould not be further characterized by empty-site PCR, becausethe region flanking the acquired genes was too different fromstrains 26695 and J99. The microarray results were thereforeverified by single-gene PCRs for three selected genes (HP0995,JHP933, and JHP934), which were positive only in the later strainand therefore confirmed the microarray results in all cases.

Changes involving housekeeping genes. Microarray hybridization indicated a genomic difference affectingthe ceuE genes (HP1561 and HP1562) in pair NQ315/1712. PCR analysesshowed that, in fact, the earlier strain NQ315 contained a singlecopy of ceuE, while NQ1712 contained two paralogous copies arrangedin tandem, which is similar to the situation in the sequencedstrains 26695 and J99, where the two gene copies have >90%nucleotide identity (Fig. 1D). Mosaic sequences were identifiedupstream and downstream of the ceuE genes, suggesting that thefragment containing one ceuE copy present in the earlier strainhad been replaced by a fragment containing two copies in thelater strain or vice versa.

The microarray data indicated absence of a cluster of pseudogenes(HP0903/0904) that contains truncated homologs of an acetatekinase gene (ackA; HP0903) and a phosphotransacetylase gene(pta; HP0904) in the later strain of pair LSU1040. Empty-sitesequence analysis showed that HP0903 was truncated in both pairedstrains (Fig. 1E). In the later strain, a recombination eventhad caused an even larger truncation of HP0903 and also thepartial deletion of gene HP0904. The total size of the deletionwas approximately 1 kb. A mosaic sequence was only present downstreamof the deletion.

DISCUSSION

Top

Discussion

A whole-genome microarray was used to compare the genome contentof paired sequential isolates of H. pylori. We detected onlysix genomic changes in 21 pairs, despite the fact that we sampledthe great majority of the genome. The strains are undoubtedlyclosely related to each other, because they were isolated fromthe same patient at a short time interval. The strains comefrom middle-aged individuals and thus are likely to representwell-established infections, so we may not have captured changesthat happened during the initial period of adaptation to thehost. Despite the high degree of conservation of gene content,our previous analysis of sequence variations in 10 gene fragmentsshowed that a large number of recombination events had takenplace in the same strains (6). In that study, we estimated thatacross the genome an average of 114 recombination events hadoccurred per pair. However, this estimate excluded the pairof strains with the highest number of differences (5 of 10 genefragments), which we have now shown to be clonally related andthus should be included in the analysis. Inclusion of this strainpair raises the estimated average total number of homologousrecombination events to 194 per pair or $~$ 4,000 for all 21 strainpairs. Thus, genetic changes associated with gain or loss ofsequence were 650 times rarer than simple homologous replacementevents.

In three of six cases, the later strain had fewer genes thanthe earlier strain; in the other three cases, it had more. However,we cannot be certain whether each of these events were a deletionor an acquisition, because the event may have occurred beforethe isolation of the first strain, with both variants cocolonizingthe stomach, when the first sample was taken. Indeed, sampleswere taken at intermediate time points for a subgroup of pairs(Table 2), which in one case (LSU1014) showed both strains coexistingduring the observation period.

View this table:
[in this window]
[in a new window]
TABLE 2. Genotypes of intermediate sequential isolates compared to those of first and last strain^a

With one exception, all genomic changes occurred among the 7of 21 pairs in which recombination had been detected withinthe 10 sequenced gene fragments (Table 1). Indeed, four of thesix events required homologous recombination with DNA from anotherH. pylori genome, the two exceptions being deletions of thecag PAI. This emphasizes the importance of transient or persistentcocolonization by two or more unrelated H. pylori strains forgenerating genomic changes in H. pylori.

The observed deletions could have been produced by intrachromosomalevents or by import of empty-site alleles from unrelated strains.In the latter case, the import should have resulted in mosaicsequences flanking the deletion. Surprisingly, in most cases,mosaic sequences were only identified on one side of the deletion(Fig. 1 and 2). The mechanism responsible for this phenomenonis not known. No mosaic sequences were found in the pair thatdeleted the cag PAI, indicating that recombination between theimperfect copies of the 31-bp repeat were sufficient to deletethe island, presumably by a RecA-dependent mechanism, similarto an event described by Björkholm et al. (3).

It is widely assumed that H. pylori uses genetic variation toadapt to individual hosts and specific niches within a host.While some variation can be generated in the absence of mixedcolonization (e.g., by point mutations, slipped-strand mispairing,or intrachromosomal recombination), import of DNA fragmentsfrom an exogenous H. pylori strain appears to be the principalsource of allelic variation, as well as genomic changes in H.pylori. Most of these events involve no loss or gain of sequenceand are likely to have phenotypic and fitness effects that aresubtle and difficult to detect.

The effectiveness of recombination as an adaptive strategy dependscritically on the prevalence of H. pylori in a population. H.pylori is likely to have evolved in a situation when infectionwas quasi-universal, permitting the panmictic exchange of genesand alleles. With declining prevalence of H. pylori, coinfectionsare likely to become much rarer, making genetic exchange ineffectiveas a means of genetic adaptation, which in turn may furtheraccelerate the disappearance of H. pylori from certain populations.

ACKNOWLEDGMENTS

We thank Matthias Frosch, Mark Achtman, and Thomas F. Meyerfor encouragement and support.

This work was supported by grants PTJ-BIO 031U213B from theBundesministerium für Bildung und Forschung competencecenter PathoGenoMik and SFB479/A5 from the Deutsche Forschungsgemeinschaftto S.S. and by DFG grant Jo 344/2-1 to C.J.

FOOTNOTES

* Corresponding author. Mailing address: Institut für Medizinische Mikrobiologie und Krankenhaushygiene, Medizinische Hochschule Hannover, Carl-Neuberg-Str. 1, 30625 Hannover, Germany. Phone: 49 511 532 6769. Fax: 49 511 532 4355. E-mail: suerbaum.sebastian{at}mh-hannover.de.

REFERENCES

Top