linical Relevance of Helicobacter pylori cagA and vacA Gene Polymorphisms
ANIELA BASSO,* CARLO–FEDERICO ZAMBON,‡ DARREN P. LETLEY,§ ALESSIA STRANGES,* ALBERTO MARCHET,�
OANNE L. RHEAD,§ STEFANIA SCHIAVON,* GRAZIELLA GUARISO,¶ MARCO CEROTI,# DONATO NITTI,�
ASSIMO RUGGE,� MARIO PLEBANI,* and JOHN C. ATHERTON§
epartments of *Laboratory Medicine, ‡Medical and Surgical Sciences, �Oncological and Surgical Sciences, and ¶Pediatrics, University of Padova, Padova, Italy;Wolfson Digestive Diseases Centre, and Institute of Infection, Immunity and Inflammation, University of Nottingham, Queen’s Medical Centre, Nottingham, England;
nd #Molecular and Nutritional Epidemiology Unit, CSPO-Scientific Institute of Tuscany, Firenze, Italy
cs-alCcstcsmssoAmpfev(dvcaawsaatccasa
cm
CA
LIM
E
ackground & Aims: The Helicobacter pylori geneagA and s1 or m1 forms of vacA are more common inisease-associated strains. Recently, forms of cagA en-oding multiple type C EPIYA segments (which in-rease phosphorylation-dependent CagA activity) andnew type i1 “intermediate region” polymorphism in
acA (which confers toxicity) have been described. Wessessed the association of new and established cagAnd vacA polymorphisms with disease. Methods: Wetudied 203 H pylori–infected subjects (53 gastric can-er [GC], 52 peptic ulcer [PU], and 98 gastritis). vacAignal, mid and intermediate region polymorphisms,agA presence, and EPIYA-C segment number werenalyzed by polymerase chain reaction. Results:agA-positive strains were significantly associatedith GC and PU (P < .001 and P < .05). GC risk was
urther associated with the number of cagA EPIYA-Cegments (odds ratio [OR] � 7.37, 95% confidencenterval [CI] � 1.98 –27.48 for 1 EPIYA-C segment;
R � 32.5, 95% CI � 8.41–125.58 for 2 or morePIYA-C segments). Increasing number of EPIYA-Cegments also increased the risk of intestinal meta-lasia. Type s1 and i1 vacA alleles were also associatedith GC and type i1 vacA with PU (OR � 2.58, 95% CI1.19 –5.61), including a significant association with
uodenal ulcer. In multivariate analysis, the associa-ions of cagA EPIYA-C segment number with GC andntestinal metaplasia as well as vacA i1 type associa-ion with PU remained. Conclusions: We confirmedhe associations of cagA and vacA polymorphismsith disease but now define their most important
eatures. For cancer risk, among Western strains, theost important factor is the number of cagA EPIYA-C
egment. For PU risk, it is the intermediate regionype of vacA.
he main virulence determinant of Helicobacter pyloriis a pathogenicity island called cag. cag pathogenicity
sland–positive strains have been shown in many studieso be closely associated with peptic ulcer (PU) and gastricancer (GC).1,2 Genes in the cag pathogenicity islandncode a type IV secretion system through which an
ffector protein, CagA, is translocated into the host cell
ytoplasm.3 Once inside the host cell, CagA can disruptignaling pathways by phosphorylation-dependent andindependent mechanisms, leading to abnormal prolifer-tion, motility, and cytoskeletal change in gastric epithe-ial cells.4 –10 CagA is phosphorylated by Src kinases in the-terminal region on tyrosine phosphorylation motifs
ontaining the Glu-Pro-Ile-Tyr-Ala (EPIYA) amino acidequence. CagA is characterized by structural diversity inhe EPIYA-repeat segment, which results in 2 major spe-ies: Western and East Asian CagA. Four types of EPIYAegments are described: A, B, C, and D. Western strains
ost commonly have an ABC pattern and East Asiantrains an ABD pattern. A, B, and D segments are occa-ionally absent or repeated. In contrast, C segments areften duplicated and Western strains with ABCC andBCCC patterns are relatively common. As EPIYA-C seg-ent numbers increase, the magnitude of CagA phos-
horylation increases and the CagA-induced cellular ef-ects are more profound. This has been suggested tonhance cancer risk.8,11–16 A second virulence factor is theacuolating cytotoxin, VacA.17 The gene encoding thisvacA), detectable in all strains, is polymorphic, withifferent types encoding VacA with different levels ofacuolating cytotoxin activity. The major variation oc-urs in the vacA signal region (which can be type s1 or s2)nd the mid region (m1 or m2).18,19 Type s1m1 strainsre highly toxigenic and s2m2 strains are nontoxigenic,hile s1m2 strains are often intermediate. Few s2m1
trains have been described,20 suggesting the existence ofstrong negative selection toward this allele. The inter-
ction of active VacA toxins with specific cellular recep-ors evokes a cascade of cellular events: induction of largeytoplasmic vacuoles, mitochondrial damage, cyto-hrome c release, inhibition of T-lymphocyte activation,nd interference with antigen presentation.3,17,21-23 vacA1/m1 and s1/m2 strains have been associated with PU,nd s1/m1 strains have been associated with GC.18,19,24
Abbreviations used in this paper: DU, duodenal ulcer; GC, gastricancer; i, intermediate; IM, intestinal metaplasia; m, mid; PCR, poly-erase chain reaction; PU, peptic ulcer; s, signal.
Recently, Rhead et al24 described a new polymorphicegion in vacA that they termed the intermediate (i)egion, located between the s and m regions. The i regionncodes part of the p33 VacA subunit and 2 types, i1vacuolating) and i2 (nonvacuolating), were described.hey showed that this region was directly involved in
oxicity by exchange mutagenesis experiments. They theneveloped a polymerase chain reaction (PCR)-based typ-
ng system for this region and showed in an Iranianopulation that the i1 genotype was a better marker ofancer-associated strains than s1 or m1.
The main aim of the present work was to study thessociations of both new and established H pylori viru-ence markers (vacA s, m, and i region polymorphisms,agA status, and the number of CagA EPIYA-C segments)ith GC and PU. Our data set also allowed us to study
he interrelationship between these virulence factors;heir associations with bacterial density, gastric inflam-
ation, and gastric intestinal metaplasia; and, in the casef the CagA EPIYA-C segment, with the region of thetomach preferentially colonized by H pylori.
Patients and MethodsPatientsIn this retrospective study, we selected 203 H
ylori–infected subjects. A total of 150 patients with be-ign diseases were selected from a series of 1145 consec-tive outpatients admitted to the Endoscopic Unit of theniversity Hospital of Padova for investigation of dys-epsia. The selection criteria were as follows: (1) presencef H pylori infection (see Supplementary Table 1 forefinition; see supplementary material online at www.astrojournal.org), (2) no previous treatment for H pylorinfection, and (3) no recent (within 4 months) treatmentith proton pump inhibitors, antibiotics, bismuth-con-
aining compounds, or nonsteroidal anti-inflammatoryrugs. Among these, 52 (30 male and 22 female; meange, 50 years; age range, 17–73 years) had PU (46 duode-al and 6 gastric) and 98 (44 male and 54 female; meange, 45 years; age range, 8 –79 years) had gastritis (antralredominant, 47 cases; diffuse, 51 cases). During endos-opy, one antral biopsy specimen and one body biopsypecimen were taken for H pylori culture and 4 antral and
body biopsy specimens for histologic evaluation.Fifty-three H pylori–infected (see Supplementary Table 1
or definition) patients with GC (28 male and 25 female;ean age, 67 years; age range, 28 – 88 years) were selected
rom a consecutive series of 73 patients who had under-one radical resection for GC, always histologically con-rmed. Tumors were located in the upper third of thetomach in 2 cases, in the middle third in 12 cases, andn the lower third in 31 cases; 2 patients had multifocalnd 6 patients had cardia GC. From each resected stom-ch, one tumor tissue sample and one sample from the
djacent (at least 3 cm from the tumor margin) normal c
ucosa were obtained and snap frozen in liquid nitrogenmmediately after excision and stored at �80°C untilrocessing. Lauren’s classification was as follows: intesti-al, 32 cases; diffuse, 21 cases.Written informed consent was obtained from each
ubject or his or her guardian.
HistologyGC was diagnosed and classified on surgical spec-
mens after H&E staining. From the series of 150 patientsith benign diseases, antral and body biopsy samplesere used to classify type and grade of inflammation
H&E), H pylori infection, and grade (Giemsa and/orarthin–Starry) according to the updated Sydney System
riteria.25 The presence or absence of intestinal metapla-ia (IM) was evaluated in 147 patients with benign dis-ases.25 Among patients with GC, because of confusionith neoplastic tissue, IM could only be said to be defin-
tively present or absent in 38 patients. When analyzinghe association between IM and H pylori virulence deter-
inants, patients with GC were not included anywayecause of potential confounding due to the presence ofarcinoma.
Diagnosis of H pylori InfectionH pylori–positive status was defined as positive
ulture and histology for patients with benign diseases orositive ureA PCR for patients with GC26 (details inupplementary Table 1).
vacA PolymorphismsDNA was obtained from frozen gastric mucosal
amples of patients with GC and from clinical H pylorisolates of patients with benign diseases. s and m regionolymorphisms of vacA were assayed as previously de-cribed.26 vacA i region polymorphisms were studied us-ng the primers VacF1, C1R, and C2R24 (PCR conditionsre detailed in Supplementary Table 2; see supplementaryaterial online at www.gastrojournal.org).
cagA Gene Identification and Predicted CagAEPIYA Segment AnalysisThe presence or absence of cagA was studied as
reviously described.26 The 3= region of cagA encodinghe EPIYA segments was analyzed following Yamaokat al27 and Argent et al28 (conditions detailed in Supple-entary Table 3; see supplementary material online atww.gastrojournal.org). We used both approaches be-
ause our DNA samples were not derived from single Hylori colonies. Multiple infections could be detected bysing the primers described by Yamaoka et al,27 while an
mprovement in specificity for EPIYA-repeat segmentlassification could be obtained by using the primersescribed by Argent et al.28 The results obtained follow-
July 2008 H pylori cagA, vacA, AND GASTRIC DISEASE 93
In a subset of 28 samples (14 GC, 6 PU, 9 gastritis), 3f which showed multiple amplicons after PCR withrimer pair TF-TR,27 the 3= cagA region was sequencedsing the primer pairs TF/WR27 and WF/TR27 or WR/pSRE2 (5= GGCTAAAGCAATGGGCGATTT 3=) orR/HpSRE3 (5= CGCAAGCAAAAAGCGACCTT 3=). Se-
uencing reactions were analyzed on an ABI PRISM 310enetic Analyzer (Applied Biosystems, Foster City, CA),
nd data were processed with the Sequencing Analysis 3.3oftware (Applied Biosystems).
Statistical AnalysisClustalW program (1.82) was used to align nucle-
tide and amino acidic sequences and to generate aimilarity tree (cladogram) of the amino acid sequencesstimating genetic distances between the sequencedtrains. The statistical analysis of data was made usinghe univariate and multivariate multinomial logistic
odels, the �2 test, and the nonparametric Mann–Whit-ey test and the Kruskal–Wallis tests in SAS for Windowsersion 9.1.3 (SAS Institute Inc, Cary, NC).
ResultsvacA TypingOf our 203 samples, 4 and 53 had untypeable vacA
and m regions, respectively. We performed successfulacA i region typing at a later date on DNA from 171vailable cases. Typing of 10 cases (1 GC, 4 PU, and 5astritis) showed probable mixed strain infection in thatoth s1 and s2 or both i1 and i2 or both m1 and m2 vacAlleles, or a combination, were found. These cases wereot included in subsequent association analyses.
The Polymorphic Structure of vacATo study the structure of vacA in our samples, we
xcluded strains where vacA was not fully typeable orhere we suspected multiple strain infections. This left
trains from 123 patients. By conventional vacA s and megion typing, 42% strains were s1/m1, 16% s1/m2, and2% s2/m2. In agreement with the original description ofacA i region,24 all s1/m1 strains were type i1 and virtu-lly all s2/m2 strains (51/52) were i2; we found a single2/i1/m2 isolate. Also in agreement, s1/m2 strains coulde either type i1 (74%; 14/19) or i2 (26%; 5/19).
Association of vacA Allelic Types WithDiseaseNext we assessed the association of polymorphisms
n different vacA regions with disease. The results of univar-ate multinomial logistic regression analysis are shown inable 1. Types s1, i1, and m1 vacA strains were closelyssociated with GC. i1, not s1 or m1, strains were signif-cantly associated with PU. For the vacA i region, wexamined the association with duodenal ulcer (DU) andastric ulcer strains separately; 69% of patients (27/39)
ith DU had vacA i1 strains, significantly more than the s
4% (31/71) with gastritis (P � .01); for gastric ulcer, 50%3/6) had i1 strains, no different from the proportionith gastritis, although the small number of gastric ul-
ers means our study was underpowered to show anssociation. Finally, we examined the association of vacAype with pattern of gastritis. Only i region type showed
significant association; 27% of patients (8/30) withntral-predominant gastritis had an i1 strain, signifi-antly fewer than the 56% of patients (23/41) with pan-astritis who had i1 strains (P � .02).
Possession of cagA and Its Association WithDisease and With vacA GenotypesAlthough the 10 isolates with multiple infections
vacA genotypes) were all typed as cagA positive, they werell excluded from the analysis because they could alsoave contained cagA-negative strains. This left strains
rom 193 patients for analysis, 139 (72%) of which wereagA positive. The results of univariate multinomial lo-istic regression analysis for cagA are reported in Table 1.agA was significantly associated with both GC and PU.o association was found between cagA status and dis-
ribution of gastritis in the gastritis-only patients.cagA-positive status was very closely associated with the
acA s1 genotype; all 52 s1/i1/m1 strains, all 14 s1/i1/m2
able 1. Association Between H pylori vacA s, i, and mAlleles or cagA Gene and Disease Diagnosis
OTE. PU and GC risks were estimated with respect to the referenceroup made up of patients with gastritis. The number of patientselonging to each group is reported. Odds ratios (OR) and 95%onfidence intervals (CI) were estimated by the conditional univariateogistic model. Only strains without multiple s, i, or m vacA signalsere included in the analysis.P � .001.P � .05.
trains, and all 5 s1/i2/m2 strains were cagA positive; 50
oc
Clbblorwnatpbn
0twm(piaquEtswEws
EEsgbscif
nweplciEaao(t.(Elnff
T
P
P
P
P
P
P
CLIN
ICA
L–A
LIMEN
TARY
TRA
CT
94 BASSO ET AL GASTROENTEROLOGY Vol. 135, No. 1
f 51 s2/i2/m2 strains and the only s2/i1/m2 strain wereagA negative.
Analysis of the Number of cagAEPIYA-C SegmentsThe PCR analysis of the number of predicted type
CagA EPIYA segments (EPIYA-C) was performed at aater date on H pylori DNA samples from 59 patients withenign diseases and 53 with GC. For the patients withenign diseases, data from both antrum and corpus iso-
ates were available for 22 patients; data from antrumnly or corpus only were available for 28 and 9 patients,espectively. For the patients with GC, cagA genotypingas performed on DNA samples from both neoplastic andonneoplastic tissue. PCR findings were negative from 21nd 7 neoplastic and nonneoplastic tissue samples, respec-ively. In total, cagA EPIYA results were available for 47atients with GC; 31 patients had successful typing fromoth neoplastic and nonneoplastic tissue, one from onlyeoplastic tissue, and 15 from only nonneoplastic tissue.The number of predicted EPIYA-C segments ranged fromto 3. Among patients with benign disease, 44 (75%) had
he same number of EPIYA-C segments in all analyseshereas 15 (25%) had differing numbers of EPIYA-C seg-ents either within the same sample or in different samples
indicating probable mixed strain infection); for GC, 28atients (60%) had the same number of EPIYA-C segments
n all samples and 19 (40%) had differing numbers. Theccuracy of PCR typing was confirmed by nucleotide se-uence analysis on a subset of 32 PCR amplification prod-cts (GenBank accession numbers from EF450136 toF450167; Supplementary Figure 1; see supplementary ma-
erial online at www.gastrojournal.org). Next we used theequence data to assess relatedness of cagA alleles in patientsho appeared to have strains with differing numbers ofPIYA-C segments. We studied 2 patients who had strainsith both 1 and 2 EPIYA-C segments and 1 patient who had
able 2. Number of cagA EPIYA-C Segments Found in the 8Findings (4 Patients With Benign Diseases) or Dis(4 Patients With GC)
Patient no.
Benign diseases
Antrum Co
D348, PD670, PD682 Single infection with 1EPIYA-C segment
Coinfection wEPIYA-C se
D345 Coinfection with 1 and 2EPIYA-C segments
Coinfection w3 EPIYA-C
D3497
D5951
D9094
D9446
trains with 1, 2, and 3 EPIYA-C segments (EF450138 and E
F450139, EF450148 and EF450149, EF450145 andF450146 and EF450147, respectively). The amino acidequences were further analyzed using the ClustalW pro-ram to generate a cladogram, estimating genetic distancesetween the sequenced strains (Supplementary Figure 2; seeupplementary material online at www.gastrojournal.org).agA alleles from the same patient always clustered together,mplying that they were more closely related than allelesrom different patients.
Association of EPIYA-C Segment NumberWith Site Within the StomachIt has been suggested that strains with higher
umbers of CagA EPIYA-C segments are less well able toithstand low pH.27 We thus asked whether they prefer-
ntially colonized different parts of the stomach. Foratients with benign diseases with successful strain iso-
ation from antrum and corpus, 17 had completely con-ordant but 4 had discordant results (Table 2). Interest-ngly, all 4 of these had isolates with higher numbers ofPIYA-C segments in the gastric corpus. However, takingll the results from patients with benign disease together,lthough numerically a higher proportion of corpus bi-psy specimens had at least 1 isolate with �1 segment40%; 12/30) than antral biopsy specimens (20%; 10/49),his did not reach statistical significance in our study (P �07). For patients with cancer, 61% of biopsy specimens28/46) from benign tissue had at least 1 isolate with �1PIYA-C segment, no different than 59% (19/32) from ma-
ignant tissue. In 4 cases, strains isolated from cancer oroncancer tissue gave discordant results; in 2 the isolates
rom the tumor had more segments, and in 2 the isolatesrom the normal mucosa had more segments (Table 2).
Association of Predicted cagA EPIYASegment Number With GCNext we asked whether the number of cagA
tients With Discordant Antrum and Corpus Isolateant Neoplastic and Nonneoplastic Sample Findings
GC
Normal mucosa Tumor sample
and 2ts, 2, andents
Coinfection with 1, 2, and3 EPIYA-C segments
Single infection with 1EPIYA-C segment
Coinfection with 1 and 2EPIYA-C segments
Single infection with 2EPIYA-C segments
Coinfection with 1 and 2EPIYA-C segments
Single infection with 1EPIYA-C segment
Single infection with 1EPIYA-C segment
Coinfection with 1 and 2EPIYA-C segments
Pacord
rpus
ith 1gmenith 1segm
PIYA-C segments was associated with disease. For this
July 2008 H pylori cagA, vacA, AND GASTRIC DISEASE 95
nalysis, we divided patients with cagA-positive strainsnto 2 groups: those where no isolate had �1 EPIYA-Cegment and those where at least 1 isolate had �1PIYA-C segment. Table 3 reports the results of univar-
ate multinomial logistic regression analysis madeonsidering diagnosis as the dependent variable andPIYA-C segment number as the predictor variable. GCisk was estimated to be 7- and 32-fold higher, respec-ively, in patients infected by strains with no more than 1nd by strains with at least 2 EPIYA-C segments whenompared with patients with nonatrophic gastritis in-ected by cagA-negative strains. Although we had nothown an association between cagA EPIYA-C segmentumber and GC site, we performed a further analysisestricted to strains typed from antral biopsy specimensrom noncancerous tissue to exclude possible confound-ng from our strains being isolated from different partsf the stomach. In this limited data set, the same resultas found, with 71% of patients (20/28) with canceraving at least 1 isolate with at least 2 EPIYA-C segmentsompared with 22% of patients (10/46) without cancerP � .0001).
Multivariate Analysis of Virulence FactorAssociation With DiseaseIn the univariate model (Table 1), type s1, i1, and
1 vacA and the presence of cagA were associated with an
able 3. Association Between H pylori cagA EPIYA-C Segmen
Diagnosis
cagA negative
cagA posi
No more than 1 EPI
n n OR
Gastritis 39 30 1PU 12 13 1.41GC 3 17 7.37a
OTE. PU and GC risks were estimated with respect to the referencedds ratios (OR) and 95% confidence intervals (CI) were estimated bP � .01.P � .001.
able 4. Association Between i Region Type of vacA or the N
OTE. Odds ratios (OR) and 95% confidence intervals (CI) were estimand sex.P � .05.P � .01.
P � .0001.
nhanced risk of GC. Type i1 vacA was associated with annhanced risk of PU. The number of EPIYA-C segmentsTable 3) was associated with an enhanced risk of GC. Weow performed multivariate multinomial logistic analy-is, including terms for age and sex. We considered i vacAlleles and EPIYA-C segments as variables of interest inredicting PU or GC to estimate the relative role of theseredictors. vacA s region type was excluded from thenalysis because of the very close association of the s1ype with cagA-positive status. vacA m region type wasxcluded from the analysis because of the low number ofC cases successfully typed for this region. Table 4 shows
he results obtained. Type i vacA remained an indepen-ent predictor of PU. Compared with cagA-negative in-ections, both cagA-positive infections where CagA wasredicted to have 1 EPIYA-C segment and infectionshere CagA was predicted to have more than 1 EPIYA-C
egment were independent risk factors of GC, with theatter having an odds ratio 5-fold higher than the former.
Histology and H pylori vacA andcagA PolymorphismsTable 5 reports the results of the association be-
ween H pylori virulence genes and the histologic param-ters evaluated in patients without GC. Mean and me-ian values of antral and corpus H pylori colonizationrade, mononuclear cell infiltration (“inflammation”), or
sing a conditional multivariate logistic model including terms for age
t Nu
tive
YA-C
groupy the
umb
cagA-m
R
ted u
npplmccTcad�anmdSca
gttGb
rcdcaclspfasrim1C[cj
T
v
v
v
v
v
v
c
c
2
1
c
NgN
CLIN
ICA
L–A
LIMEN
TARY
TRA
CT
96 BASSO ET AL GASTROENTEROLOGY Vol. 135, No. 1
eutrophil infiltration (“activity”) are reported after theatients have been subdivided according to vacA or cagAolymorphisms. vacA s1, i1, and m1 alleles were corre-
ated with more severe corpus inflammation and withore severe antral and corpus activity. Presence of the
agA gene was associated with more severe antral andorpus activity but not with grade of inflammation per se.he number of cagA EPIYA-C sequence did not signifi-antly affect inflammation or activity. In both antrumnd corpus, there was a close association between H pyloriensity and both inflammation (�2 � 18.92, P � .001 and2 � 12.98, P � .001) and activity (�2 � 15.62, P � .001nd �2 � 18.44, P � .001). Virulence factors were largelyot associated with H pylori density, but strains with 2 orore EPIYA-C segments colonized the antrum more
ensely than strains with only one EPIYA-C segment.urprisingly, cagA-negative strains colonized antrum andorpus as densely as cagA-positive strains and numericallys densely as strains with 2 or more EPIYA-C segments.
IM was found in similar percentages in patients withastritis alone (21.9%), PU (25.5%), and GC (24.3%). Forhe subsequent analysis of data, we considered only pa-ients without GC, because the majority of patients withC harbored cagA-positive, vacA s1/i1 strains, and we
able 5. Gastric Histopathology and H pylori Virulence Genes
acA m2 no. of cases 74 74Mean (median) 1.99 (2) 1.84 (2)Mann–Whitney test (P value) NS NS
agA positive no. of cases 85 84Mean (median) 1.98 (2) 1.86 (2)
agA negative no. of cases 51 51Mean (median) 2.04 (2) 1.90 (2)Mann–Whitney test (P value) NS NSor more EPIYA-C no. of cases 17 16Mean (median) 2.06 (2) 1.75 (2)EPIYA-C no. of cases 31 31Mean (median) 1.68 (2) 1.77 (2)
agA negative no. of cases 51 51Mean (median) 2.04 (2) 1.90 (2)Kruskal–Wallis test (P value) �.01 NS
OTE. H pylori density and mononuclear (inflammation) and polymorpraded according to the updated Sydney System.25
S, not significant.
elieved this might bias the results. Table 6 reports the l
esults of univariate binary logistic regression analysisonsidering the presence or absence of IM as the depen-ent variable and H pylori virulence genes or H pyloriolonization grade as predictors. s1, i1, m1 vacA as wells the presence of cagA or higher corpus H pyloriolonization grade enhanced the risk of IM. We se-ected i type of vacA, the number of cagA EPIYA-Cegments, and corpus H pylori colonization grade asredictors and IM as the dependent variable to per-orm multivariate multinominal logistic regressionnalysis, including terms for age and sex. EPIYA-Cegment number was an independent predictor of IMisk. The magnitude of this risk increased with increas-ng numbers of EPIYA-C segments (1 EPIYA-C seg-
ent: odds ratio � 27 [95% confidence interval �.8 –394]; 2 EPIYA-C segments: odds ratio � 30 [95%I � 1.1– 863]; 3 EPIYA-C segments: odds ratio � 50
95% CI � 1.9 –1300]). Neither vacA i1 type nor corpusolonization grade was an independent risk factor (ad-usted odds ratio � 1.34 and 0.58, respectively).
DiscussionThe most widely researched and best-estab-
lear (activity) cell infiltrate in antrum and corpus were assessed and
ri
honuc
ished H pylori virulence factors are the vacuolating
cAlgeasvwtt
assrnttcv
ctdtn
iadaTcovawfnstbave
pPtwameeta
ifipnibsHsitfoobigsdornw
T
ssiimmccc12H23H23
NwOtHna
b
CLI
NIC
AL–
ALI
MEN
TARY
TRA
CT
July 2008 H pylori cagA, vacA, AND GASTRIC DISEASE 97
ytotoxin, VacA, and the cytotoxin-associated protein, CagA.2,18,19,29 CagA has received renewed interest fol-
owing the demonstration that the 3= region of the cagAene varies in size due to a variable number of repeatsncoding type C EPIYA phosphorylation sites.8 For vacA,
new polymorphic region (the i region) has been de-cribed and shown to be a major determinant of toxicity;acA type i1 strains were shown to be closely associatedith GC in an Iranian population, and i1 type was shown
o be a better marker of disease than s1 or m1 geno-ypes.24
Like that study, we confirmed that all s1/m1 vacAlleles were type i1, all s2/m2 alleles were type i2, and1/m2 alleles could be i1 or i2. The lack of other allelictructures (eg, s1/i2/m1) could be because the specificecombination events needed to make these mosaics haveever occurred. However, it is perhaps more likely thathey have occurred but the resultant allele is unfavorableo H pylori and is selected against, as is believed to be thease for s2/m1 vacA alleles, which are described but areery uncommon.20
vacA s1, i1, and m1 strains were all significantly asso-iated with GC in our Italian population, confirming thathese vacA alleles are risk factors for GC in Italy as alreadyescribed in Iran.24 The discrimination of vacA i regionyping between cancer and “gastritis only” cases was
able 6. Association Between H pylori vacA s, i, and mAlleles, cagA gene, H pylori cagA EPIYA-C SegmentNumber, and Antral or Corpus H pylori ColonizationGrade and IM
OTE. Only patients without GC entered the analysis. Only strainsithout multiple s, i, or m vacA signals were included in the analysis.dds ratios (OR) and 95% confidence intervals (CI) were estimated by
he conditional univariate logistic model.PAD, H pylori antral colonization grade; HPCD, H pylori corpus colo-ization grade.P � .001.P � .01
umerically superior to s and m region typing; if i1 type f
s taken as a marker of a cancer-associated strain and i2s a marker of a “gastritis only” strain, 65% of cases wereiscriminated correctly. We also show, for the first time,significant association between vacA i1 type and DU.
his is important because although both DU and GC areaused by H pylori, they are pathogenically and epidemi-logically different.30 Our study shows that, like cagA,acA i1 strains increase the risk of both major H pylori–ssociated diseases, not just GC. However, the associationith DU was weaker than that with cancer and other data
rom our study both support this and suggest an expla-ation. Among the noncancer patients, i1 strains wereignificantly more closely associated with pangastritishan with antral-predominant gastritis. GC develops on aackground of pangastritis and DU on a background ofntral-predominant gastritis; thus, we suggest that i1acA predisposes more strongly to pangastritis, and thisxplains its closer association with GC than with DU.
As in many other studies, we showed that cagA-ositive strains were associated with both GC and withU2,3,18,19,29 and a virtually complete concordance be-ween cagA-positive status and the vacA s1 genotypeithin strains.18,19,29,31 This means that human diseasessociation studies such as ours cannot assess which isore important in disease pathogenesis; both perform
qually as markers of disease-causing strains. Moreover, itxplains associations between cagA-positive status andhe vacA i1 and m1 types; these are entirely due to thessociation of these types with the s1 vacA type.
Our analysis of gastric histology gives further insightsnto cagA- and vacA-associated pathogenesis. Firstly, weound a good correlation between H pylori density andnflammation, in support of previous studies.31,32 cagA-ositive, vacA s1, i1, and m1 strains were associated witheutrophil infiltration (“activity”) but not lymphocyte
nfiltration (“inflammation”) in the antrum and withoth “activity” and “inflammation” in the corpus. This islightly different than the effects recently described in the
pylori–infected Mongolian gerbil model using wild-typetrains and cagA null isogenic mutants, where cagA pos-tivity increased bacterial density and inflammation inhe corpus but had no effect in the antrum.33 This rein-orces the need to study humans for the understandingf human pathology and the dangers of trying to extrap-late from animal models. We also showed associationsetween vacA i region and m region type and corpus
nflammation and activity, which were as strong or stron-er than the associations for cagA/vacA s region; thisuggests that these other vacA polymorphisms have ad-itional or independent importance in inducing pathol-gy and reinforces the importance of studying humansather than animal models; in the Mongolian gerbil, vacAull mutants are nearly as pathogenic as vacA-positiveild-type strains.34 Finally, we showed that the same
actors as those associated with GC in this and other
sw
wwsesrtshwiVrscvdawboecvctAicmbpE
tcicnptbstsmpssET(l
rcsftsd
rifisapimifidcstsptEstr
picaactwa
aG1
CLIN
ICA
L–A
LIMEN
TARY
TRA
CT
98 BASSO ET AL GASTROENTEROLOGY Vol. 135, No. 1
tudies (cagA and s1, i1 and m1 vacA) were associatedith precancerous IM.We confirmed previous studies showing that strains
ith cagA with higher numbers of EPIYA-C segmentsere much more closely associated with GC than
trains with cagA with one C-type segment only.8,14 How-ver, this is the first clinical study performed in a largeeries of European patients specifically showing that theelative risk of GC development is directly correlated withhe number of EPIYA-C segments; in patients infected bytrains with no more than one EPIYA-C, it was 7 timesigher than cagA-negative H pylori–infected subjects,hile the risk was more than 30-fold higher in subjects
nfected by strains with 2 or more EPIYA-C segments.ariation in EPIYA-C segment number is particularly
elevant in Western types of CagA present in our Italiantrains. Previous clinical association studies have beenonducted mainly on East-Asian patients, where rele-ance is minimal because the ABD form of CagA pre-ominates.27,35,36 We could not show, however, a directssociation between numbers of EPIYA-C segments andorsened gastric inflammation or activity. This may haveeen a type II error (we studied fewer strains than forther virulence factors and so could have missed a realffect), but if not it may imply that these strains increaseancer risk through a process other than increasing se-erity of inflammation. In agreement with this, a directorrelation was found in patients without GC betweenhe number of EPIYA-C segments and the risk of IM.lthough the risk of IM was increased in patients with s1,
1, and m1 vacA and in patients with increased H pyloriolonization density in the corpus, in a multivariateodel, only EPIYA-C segment number was confirmed to
e an independent risk factor for IM, the risk beingrogressively higher when infecting strains had 1, 2, or 3PIYA-C segments.When we analyzed the cagA 3= region for numbers of
ype EPIYA-C segment, we found frequent evidence ofoinfections; in 25% of patients with benign diseases andn 40% of patients with GC, amplicons of the 3= region ofagA with molecular weights corresponding to differentumbers of repeated sequences were obtained from sam-les of the same patient. These were confirmed to reflecthe number of repeated sequences in the 3= region of cagAy sequence analysis performed in a subset of 28 DNAamples. From the derived amino acid sequence, in allhese samples the EPIYA-A and EPIYA-B segments wereingly present, while a variable number of EPIYA-C seg-
ents were found. We also sequenced 7 gel-eluted am-licons: 2 (1 and 2 EPIYA-C) were from one patient’sample (PD5951K), 2 (1 and 2 EPIYA-C) were from aecond patient’s sample (PD2968K), and 3 (1, 2, and 3PIYA-C) were from a third patient’s sample (PD3497K).he sequences from the same patient were very similar
Supplementary Figure 2; see supplementary material on-
ine at www.gastrojournal.org), suggesting that they de-
ived from a common strain rather than from unrelatedoinfecting strains. This suggests that the presence in theame stomach of H pylori cagA-positive strains with dif-erent EPIYA-C segments is probably not due to coinfec-ions, but rather represents rapid evolution of a singletrain with survival of parent and daughter substrains, asescribed by Aras et al.37
One hypothesis to explain the persistence of closelyelated substrains with different CagA EPIYA-C segmentsn the same stomach is that such substrains might betted to colonize different anatomic areas within thetomach. Our data give some support to this idea. Firstly,mong the patients with benign diseases, all 4 with Hylori substrains with different cagA EPIYA-C segmentssolated from antrum and corpus had the substrain with
ore EPIYA-C segments (and so likely more active CagA)n the corpus. Also, overall, twice as many strains isolatedrom the corpus had �1 EPIYA-C segments as strainssolated from the antrum in these patients with benignisease, although this just failed to reach formal statisti-al significance (P � .06). Admittedly this pattern was noteen in patients with cancer, but these patients are likelyo have atrophic change throughout their stomach ando might be expected to lose regional gastric selectiveressures based on differential acidity. Also, these pa-ients, who tended to have strains with more CagAPIYA-C segments, would likely have higher pH in theirtomachs due to the atrophy. This fits well with the ideahat strains with more EPIYA-C segments may be lessesistant to acid.27
In conclusion, we have shown the importance of Hylori vacA polymorphisms in disease, particularly themportance of the newly described i1 type in PU. We haveonfirmed the pathogenicity of cagA-positive strains butlso shown that among these strains, those with morective forms of cagA are associated with a further in-reased risk of GC. Our analyses of distribution of gas-ritis and histopathologic changes and of co-colonizationith different H pylori substrains support these findingsnd give insights into possible mechanisms.
Supplementary Data
Note: To access the supplementary materialccompanying this article, visit the online version ofastroenterology at www.gastrojournal.org, and at doi:0.1053/j.gastro.2008.03.041.
References
1. Peek RM Jr, Blaser MJ. Helicobacter pylori and gastrointestinaltract adenocarcinomas. Nat Rev Cancer 2002;2:28–37.
2. Palli D, Masala G, Del Giudice G, et al. CagA� Helicobacter pyloriinfection and gastric cancer risk in the EPIC-EURGAST study. Int JCancer 2007;120:859–867.
July 2008 H pylori cagA, vacA, AND GASTRIC DISEASE 99
changes by Helicobacter pylori. Proc Natl Acad Sci U S A1999;96:14559–14564.
5. Stein M, Rappuoli R, Covacci A. Tyrosine phosphorylation of theHelicobacter pylori CagA antigen after cag-driven host cell trans-location. Proc Natl Acad Sci U S A 2000;97:1263–1268.
6. Mimuro H, Suzuki T, Tanaka J, et al. Grb2 is a key mediator ofHelicobacter pylori CagA protein activities. Mol Cell 2002;10:745–755.
7. Hirata Y, Maeda S, Mitsuno Y, et al. Helicobacter pylori CagAprotein activates serum response element-driven transcriptionindependently of tyrosine phosphorylation. Gastroenterology 2002;123:1962–1971.
8. Hatakeyama M. Oncogenic mechanisms of the Helicobacterpylori CagA protein. Nat Rev Cancer 2004;4:688–694.
9. Bagnoli F, Buti L, Tompkins L, et al. Helicobacter pylori CagAinduces a transition from polarized to invasive phenotypes inMDCK cells. Proc Natl Acad Sci U S A 2005;102:16339–16344.
0. Kurashima Y, Murata-Kamiya N, Kikuchi K, et al. Deregulation ofbeta-catenin signal by Helicobacter pylori CagA requires theCagA-multimerization sequence. Int J Cancer 2008;122:823–831.
1. Higashi H, Tsutsumi R, Fujita A, et al. Biological activity of theHelicobacter pylori virulence factor CagA is determined by varia-tion in the tyrosine phosphorylation sites. Proc Natl Acad Sci USA2002;99:14428–14433.
2. Zhang Y, Argent RH, Letley DP, et al. Tyrosine phosphorylation ofCagA from Chinese Helicobacter pylori isolates in AGS gastricepithelial cells. J Clin Microbiol 2005;43:786–790.
3. Azuma T, Yamazaki S, Yamakawa A, et al. Association betweendiversity in the Src homology 2 domain—containing tyrosinephosphatase binding site of Helicobacter pylori CagA protein andgastric atrophy and cancer. J Infect Dis 2004;189:820–827.
4. Yamaoka Y, Kodama T, Kashima K, et al. Variants of the 3=region of the cagA gene in Helicobacter pylori isolates frompatients with different H. pylori-associated diseases. J Clin Mi-crobiol 1998;36:2258–2263.
5. Argent RH, Kidd M, Owen RJ, et al. Determinants and conse-quences of different levels of CagA phosphorylation for clinicalisolates of Helicobacter pylori. Gastroenterology 2004;127:514–523.
6. Naito M, Yamazaki T, Tsutsumi R, et al. Influence of EPIYA-repeatpolymorphism on the phosphorylation-dependent biological activ-ity of Helicobacter pylori CagA. Gastroenterology 2006;130:1181–1190.
8. Atherton JC, Cao P, Peek RM Jr, et al. Mosaicism in vacuolatingcytotoxin alleles of Helicobacter pylori. Association of specificvacA types with cytotoxin production and peptic ulceration. J BiolChem 1995;270:17771–17777.
9. Atherton JC, Peek RM Jr, Tham KT, et al. Clinical and pathologicalimportance of heterogeneity in vacA, the vacuolating cytotoxingene of Helicobacter pylori. Gastroenterology 1997;112:92–99.
0. Letley DP, Lastovica A, Louw JA, et al. Allelic diversity of theHelicobacter pylori vacuolating cytotoxin gene in South Africa:rarity of the vacA s1a genotype and natural occurrence of ans2/m1 allele. J Clin Microbiol 1999;37:1203–1205.
1. Ji X, Fernandez T, Burroni D, et al. Cell specificity of Helicobacterpylori cytotoxin is determined by a short region in the polymorphicmid region. Infect Immun 2000;68:3754–3757.
2. Sundrud MS, Torres VJ, Unutmaz D, et al. Inhibition of primaryhuman T cell proliferation by Helicobacter pylori vacuolating toxin(VacA) is independent of VacA effects on IL-2 secretion. Proc Natl
Acad Sci U S A 2004;101:7727–7732.
3. Yamasaki E, Wada A, Kumatori A, et al. Helicobacter pylorivacuolating cytotoxin induces activation of the proapoptoticproteins Bax and Bak, leading to cytochrome c release and celldeath, independent of vacuolation. J Biol Chem 2006;281:11250–11259.
4. Rhead JL, Letley DP, Mohammadi M, et al. A new Helicobacterpylori vacuolating cytotoxin determinant, the intermediate region,is associated with gastric cancer. Gastroenterology 2007;133:926–936.
5. Dixon MF, Genta RM, Yardley JH, et al. Classification and gradingof gastritis. The updated Sydney System. International Workshopon the Histopathology of Gastritis, Houston 1994. Am J SurgPathol 1996;20:1161–1181.
6. Zambon CF, Basso D, Navaglia F, et al. Pro- and anti-inflammatorycytokines gene polymorphisms and Helicobacter pylori infection:interactions influence outcome. Cytokine 2005;29:141–152.
7. Yamaoka Y, El-Zimaity HM, Gutierrez O, et al. Relationship be-tween the cagA 3= repeat region of Helicobacter pylori, gastrichistology, and susceptibility to low pH. Gastroenterology 1999;117:342–349.
8. Argent RH, Zhang Y, Atherton JC. Simple method for determina-tion of the number of Helicobacter pylori CagA variable-regionEPIYA tyrosine phosphorylation motifs by PCR. Clin Microbiol2005;43:791–795.
9. Basso D, Navaglia F, Brigato L, et al. Analysis of Helicobacterpylori vacA and cagA genotypes and serum antibody profile inbenign and malignant gastroduodenal diseases. Gut 1998;43:182–186.
0. Hansson LE, Nyrén O, Hsing AW, et al. The risk of stomachcancer in patients with gastric or duodenal ulcer disease. N EnglJ Med 1996;335:242–249.
1. Zambon CF, Navaglia F, Basso D, et al. Helicobacter pyloribabA2, cagA, and s1 vacA genes work sinergistically in causingintestinal metaplasia. J Clin Pathol 2003;56:287–291.
2. Atherton JC, Tham KT, Peek RM, et al. Density of Helicobacterpylori infection in vivo as assessed by quantitative culture andhistology. J Infect Dis 1996;174:552–556.
3. Rieder G, Merchant JL, Haas R. Helicobacter pylori cag-type IVsecretion system facilitates corpus colonization to induce pre-cancerous conditions in Mongolian gerbils. Gastroenterology2005;128:1229–1242.
4. Ogura K, Maeda S, Nakao M, et al. Virulence factors of Helico-bacter pylori responsible for gastric diseases in Mongolian gerbil.J Exp Med 2000;192:1601–1610.
5. Azuma T, Yamakawa A, Yamazaki S, et al. Correlation betweenvariation of the 3= region of the cagA gene in Helicobacter pyloriand disease outcome in Japan. J Infect Dis 2002;186:1621–1630.
6. Choi KD, Kim N, Lee DH, et al. Analysis of the 3= variable regionof the cagA gene of Helicobacter pylori isolated in Koreans. DigDis Sci 2007;52:960–966.
7. Aras RA, Fischer W, Perez-Perez GI, et al. Plasticity of repetitiveDNA sequences within a bacterial (Type IV) secretion systemcomponent. J Exp Med 2003;198:1349–1360.
Received September 24, 2007. Accepted March 20, 2008.Address requests for reprints to: Daniela Basso, MD, Department of
aboratory Medicine, University Hospital of Padova, Via Giustiniani 2,5128 Padova, Italy. e-mail: [email protected]; fax: (39) 0498212785.Supported by AIRC regional grant 2005 and by Cancer Research UK
nd CORE, the digestive diseases charity.
The authors report they have no conflict to disclose.
