Article Text


Analysis of iceA genotypes in South African Helicobacter pylori strains and relationship to clinically significant disease


BACKGROUND South African Helicobacter pylori isolates are characterised by the universal presence ofcagA but have differences in vacuolating cytotoxin gene (vacA) alleles which correlate with clinically significant disease. However, the candidate virulence marker gene iceA has not been investigated.

AIM To characterise the genetic organisation and heterogeneity oficeA genotypes in different South African clinical isolates.

PATIENTS AND METHODS We studied H pylori strains isolated from 86 dyspeptic patients (30 with peptic ulcer disease (PUD), 19 with distal gastric adenocarcinoma (GC), and 37 with non-erosive gastritis) for the presence oficeA1 or iceA2genes, and for differences in the genetic organisation oficeA2 by polymerase chain reaction, Southern hybridisation analysis, and sequencing.

RESULTS Genetic analysis of iceA1 demonstrated significant homology (92–95%) with the USA type strain 26695 and probably functions as a transcriptional regulator, while a novel variant (iceA2D′) oficeA2 and marked differences in predicted protein secondary structure of the iceA2protein were defined. iceA1 was detected in 68% and iceA2 in 80% of all clinical isolates. Although approximately 40% of patients had both strains, a higher prevalence (p< 0.01) of GC patients were infected withiceA1 isolates which were invariablyvacA s1/iceA1(p< 0.005 v gastritis). Isolates from PUD patients were distinguished by the structurally alterediceA2D variant (53%; p<0.03v gastritis) while theiceA2C variant distinguished isolates from patients with gastritis alone (67%; p< 0.005v PUD).

CONCLUSION In this study, an association between iceA1 and GC was noted while differences in variants oficeA2 differentiated between PUD and gastritis alone. Combination analyses oficeA genotypes andvacA alleles supported these associations.

  • adenocarcinoma
  • gastritis
  • Helicobacter pylori
  • iceA
  • peptic ulceration
  • protein prediction
  • sequencing
  • South Africa

Statistics from

Helicobacter pylori induces gastric inflammation in virtually all colonised individuals and such gastritis increases the risk of peptic ulcer disease (PUD) and non-cardia/distal gastric adenocarcinoma (GC).1-3 However, only a minority of patients carrying H pylori develop clinical sequelae, suggesting that particular bacterial products may contribute to pathogenesis.4 Recently, a novelH pylori geneiceA was identified following transcriptional upregulation on contact with gastric epithelial cells.5 iceA exists as two distinct genotypes, iceA1 andiceA2, and onlyiceA1 RNA is induced following adherence in vitro.5 H pylori iceA1 demonstrates strong homology to a restriction endonuclease nlaIIIR inNeisseria lactamica,6 and in vivo carriage of H pylori iceA1 strains has been reported to be associated with peptic ulceration and enhanced acute neutrophilic infiltration.5 7 8 However, linkage between the iceA1 genotype and ulcer disease is not universal,9 and thus may be population dependent.

In contrast with iceA1,iceA2 has no significant homology to known proteins and its structure reveals patterns of repeated protein cassettes. Recently, the genetic organisation and sequence heterogeneity of iceA2 has been studied,6 revealing five distincticeA2 subtypes. WhileiceA2 strains are more prevalent among patients with asymptomatic gastritis and non-ulcer dyspepsia,7 a statistically significant relationship between iceA2 subtypes and disease has not yet been defined.

South African H pylori isolates are characterised by the universal presence ofcagA but differences in the 3′ region ofcagA and vacAalleles correlate significantly with clinical disease.10 11 Our hypothesis was thatH pylori isolates from South African patients may also exhibit variability in theiriceA alleles and that this variability may be related both to clinical outcome and tovacA allelic status.

Materials and methods


H pylori isolates, cultured from single biopsies from 86 clinically defined patients (37 patients with gastritis alone, 30 patients with PUD, and 19 patients with GC), were examined. Fifty six of the resultant isolates hadvacA and cagAdetermined in a previous study.11 Putative clonal isolates were obtained from eight of the parent strains by serially reculturing single colonies twice. Reference strains 26695 (iceA1) isolated from a patient in the UK with gastritis12 and J99 (iceA2) isolated from a USA patient with duodenal ulcer disease13 were used asiceA1 and iceA2positive and negative controls, respectively, in this study.


Polymerase chain reactions (PCRs) were performed as described previously5 11 but a final [MgCl2] concentration of 3.5 mM was used for theiceA1 PCRs. For amplification of theiceA1 allele, forward primerIceA1F (5′-CGTTGG GTAAGCGTTACAGAATTT) and reverse primer IceA1R (5′-TCATTGTATATCCTAT CATTACAAG) yielded a fragment of 558 bp (fig 1). ForiceA2, primersIceA2F (5′-GTTG TCGTTGTTTTAATGAA) andIceA2R (5′-GTCTTAAACCCCACGATTAAA) yielded a fragment of 120 bp. PCR was performed in a Sprint PCR (Hybaid, South Africa) under the following conditions: four minutes of preincubation at 94°C followed by 30 cycles of one minute at 94°C, one minute at 62°C (iceA1) or 53°C (iceA2), and one minute at 72°C. Final extension was performed for 10 minutes at 72°C.

Figure 1

Primers, sites, and amplicon sizes for iceA1 (HP1209, GenBank Accession No AE000511) and iceA2 (JHP1132, GenBank Accession NoAE001541) (see text for details).

Additional genotype specific PCR assays that used primers foriceA1 (IceA1F5(5′-GTG TTTTTAACCAAAGTATC) andIceA1R4(5′-CTATAGCCASTYTCTTTGCA)) and primers flanking theiceA2 internal cassette were performed to confirm the presence of the genes and to identify the size and subtype of the iceA2 allele.7 PrimersIceA2F6 (5′-GTTG GGTATATCACAATTTAT) andIceA2R5 (5′-TTRCCCTATTTTCTAGTAGGT) yielded a fragment of 229, 334, or 439 bp according to the existence of repeated sequences of 105 nucleotides. PCR was performed under the following conditions: four minutes of preincubation at 94°C followed by 30 cycles of one minute at 94°C, one minute at 53°C (iceA1 andiceA2), and one minute at 72°C. Final extension was performed for 10 minutes at 72°C.iceA amplimers were examined by electrophoresis on 1% agarose gels according to standard procedures.11


The complete iceA1 gene, from the upstream cysE (HP1210) to the downstreamhypIM (HP1208), was amplified using F1 (5′-GGG TGCGATTTGCGTGGGCGATG) and R10 (5′-GATCATGGCCTACAACCGCATGGA), as described previously,14 whileiceA2 was amplified using primer setsIceA2F6/R5. PCR products were gel extracted (QIAEX II gel extraction kit; Qiagen, Cape Town, South Africa) and sequenced on an ABI PRISM 377 automated sequencer (ABI, Foster City, California, USA) using the ABI PRISM BigDye terminator cycle sequencing reagent kit with AmpliTaq DNA polymerase FS (PE Biosystems, Johannesburg, South Africa), as described previously.15PCR and direct sequencing were performed at least twice to determine DNA sequences for each strain.


Southern hybridisation analysis was performed essentially as described previously.16 Chromosomal DNA digested withBglII was subjected to electrophoresis on a 0.8% Tris acetate/EDTA agarose gel, transferred to a Hybond-N+ membrane (AP Biotech, Piscataway, New Jersey, USA) by capillary transfer, and UV cross linked. Prehybridisation was performed using 6× sodium chloride/sodium citrate (SSC), 5× Denhardts, 0.5% sodium dodecyl sulphate (SDS), and 100 μg/ml of sheared salmon sperm DNA at 60°C for three hours. Buffer was replaced with fresh buffer containing denatured probe and hybridisation proceeded at 60°C for 18 hours. Membranes were washed for five minutes with 2× SSC, 0.5% SDS at room temperature, 15 minutes with 2× SSC, 0.1% SDS at room temperature, two hours with 0.2% SSC, 0.5% SDS at 60°C, and 30 minutes with 0.2% SSC, 0.5% SDS at 60°C. The membranes were then subjected to autoradiography for detection.

Probes for iceA1 andiceA2 were generated by PCR using primer pairs IceA1F5/IceA1R4andIceA2F6/IceA2R5, respectively, as described previously,7 with 26695 and J99 chromosomal DNA as respective templates. Products were purified using QiaQuick PCR purification columns (Qiagen, Valencia, California, USA) and labelled with [α-32P]dCTP using the PRIME-IT RmT labelling kit as recommended by the manufacturer (Stratagene, La Jolla, California, USA). Unincorporated nucleotides were removed using ProbeQuant G-50 microcolumns according to the manufacturer's instructions (AP Biotech). Probe was denatured by boiling for five minutes and then placed on ice prior to hybridisation.


DNA sequences were analysed using the National Center for Biotechnology Information (NCBI) server (USA) and consensus secondary structure prediction for proteins was obtained using the META Predict Protein (which includes the SignalP, JPRED, TopPred, and DAS Servers (Columbia University, USA)).17-20 SignalP predicts the presence and location of signal peptide cleavage sites, JPRED predicts secondary structures, TopPred is useful for prediction of the location and orientation of transmembrane sequences through the use of hydrophobicity patterns and by applying the “positive inside” rule, while DAS predicts the location of transmembrane using hydrophobicity patterns. Internet based searches were preformed at NCBI, the Institute for Genome Research (TIGR, Maryland, USA), and Astra-Zeneca (Boston, USA).


Data were examined using the χ2 test or Fisher's exact test as appropriate. Probability levels of <0.05 were considered statistically significant.



In a preliminary study, in order to establish the composition oficeA gene products in South AfricanH pylori isolates,iceA1 (primers F1/R10: 952 bp) andiceA2 (primers F6/R5: 229-334 bp) were sequenced.

Five isolates with the iceA1 genotype (two from patients with gastritis alone, one from a patient with peptic ulceration, and two from patients with gastric cancer) were sequenced and compared with the type strains 60190 (GenBank Accession No U43917) and 26695. Alignment of these sequences revealed the presence of a putative conserved initiation codon at position ATG919 in all five South African strains as well as conserved upstream transcription initiation sites, as previously reported6(fig 2). Predicted proteins ranged from the expected 128 amino acids to 136 amino acids in four strains. No full length NLAIII homologues were predicted. These results suggested that theiceA1 gene does not encode a functional restriction endonuclease but probably acts as a transcriptional regulator.14

Figure 2

Alignment of nucleotide sequences of 60190, 26695, and five South African strains showing the 3′ end of cysE, the complete iceA1 gene (ORF1 and ORF2), intervening sequences (1–3), and the 5′ end of hypIM putative start codon. ATG617 (italicised and boxed) is the start codon for nlaIIIR and is only conserved in 3/5 strains. None of the strains however is predicted to produce the NLAIII homologue of 228 amino acids. ATG919 (boxed/bold), the start codon for iceA1, is conserved in all five South African isolates, as are the upstream transcription boxes (at –35 and –10). Predicted iceA1 proteins range in size from 128 to 136 amino acids, except for strain 54 (frameshift results in a protein of 53 amino acids). All isolates shared substantial protein homology (93–96%) with HP1209. Sequence gaps are indicated by dashes (-).


Seven isolates with the iceA2 genotype (two from patients with gastritis alone, two from patients with PUD, and three from patients with GC) were sequenced and compared with the predicted gene (JHP1132) present at the same locus in strain J99 (fig3).

Figure 3

Amino acid alignment of iceA2 genotypes from seven South African isolates and homology with JHP1132. The different peptide domains are boxed and the amino acid number indicated. Asterisks (*) represent gaps in sequence homology.

Prior to our analysis, five different variants (A–E) based on size (24, 59, 94, or 129 amino acids) and/or sequence oficeA2 had been described.6 TheiceA2 gene (JHP1132) from strain J99 which conforms to the structure of repeats (14-13-162B-6-10) demonstrated by Figueiredo and colleagues6 generates a predicted protein of 59 amino acids and is an example of theiceA2B subtype.6 None of the South African isolates sequenced in this series had either a 2A structure (14-13-16-6-10) or a 2B structure. Two isolates (Hp27 and 215) however had a 14-13-162C-6-10 motif and were therefore classified as iceA2C subtypes. Four isolates had the pattern 14-13-162B-6-13-162C-6-10 and are examples of the ice A2D subtype. These subtypes would be predicted to generate proteins of 94 amino acids but one isolate (Ca102) had an inframe three nucleotide deletion within cassette 162B which resulted in a putative protein truncation of 93 amino acids. Sequence analysis of a fifth isolate (1715) demonstrated a novel variant of theiceA2D subtype that was typified by the sequence 14-13-162C-6-13-162C-6-10 (sequence identity (71%) and shared residues (80%) withiceA2D).

Analysis of the protein signal sequence, secondary structures, and topology for the South African iceA2variants is provided in table 1. Of note are the observations that these proteins do not have a typical prokaryote signal sequence (from the SignalP server), and the N terminus appears to be embedded within the cell membrane. The TopPred topology program demonstrated that variants 2B and 2C had no predicted transmembrane regions while 2D and the novel 2D variant (2D′) had a single putative transmembrane region with scores of 0.758 and 0.615, respectively (fig 4).19The dense alignment surface (DAS) method confirmed that variant 2B had no transmembrane region but suggested a possible transmembrane region in 2C and confirmed two putative transmembrane regions in 2D and the 2D′ variant.20 Variant 2B can be classified as a mixed class protein but 2C, 2D, and 2D′ are all all beta proteins.

Table 1

Protein characteristics of iceA variants

Figure 4

Topology profiles of four iceA2 variants. Left: protein predictions from the TopPred program; right: predictions from the dense alignment sequence (DAS) program. The type strain JHP1132 (iceA2B variant) is indicated in the top panel and South African strains are included with the iceA2 variants signified (2C–2D′). Arrows indicate putative transmembrane regions.


After validating the PCR protocol and examining the structure oficeA1 and iceA2, PCR was performed in clinical isolates and products were analysed in terms of the distribution of genotypes. Fifty four isolates were positive for either iceA1 oriceA2 using previously described primersIceA1F/1R andIceA2F/2R,5 54 isolates were positive for both iceA1 andiceA2, and one isolate did not yield any PCR product for either of the iceA genotypes. Amplicons of the respective sizes were obtained from reference strains 26695 and J99 as expected. Twenty one of the 54iceA1+/iceA2+ isolates were re-examined with differenticeA primer sets (F5/R4 (iceA1) and F6/R5 (iceA2))6 7 to determine whether this “mixed genotype” was a consistent finding. All 21 (100%) were positive for both genotypes, indicating that these primer sets are concordant. These isolates each had a singlevacA allele, a singlecagA 3′ amplicon, which suggested the presence of multiple iceA genotypes in these strains.

We then examined the presence of these putative “mixed genotypes” further by performing repeat PCR (using two primer sets) and Southern hybridisation in multiply recultured single colonies from eighticeA1+/iceA2+ parent strains. Seven of the eight isolates had a singleiceA allele by both PCR and Southern analysis (fig 5). No hybridisation was seen in one isolate. A single PCR band using primer sets IceA2F6/IceA2R5(which identifies iceA2 sizes and subtypes) and a single Southern band was seen in the three clonal strains that were iceA2+ . These data signify the presence of multiple iceA strains rather than mixed genotypes within a single strain in these cultures.

Figure 5

Southern analysis of putatively clonal isolates derived from eight parent strains (lanes 1–8) that were originally iceA1+/2+ by polymerase chain reaction. Four were iceA1+, three iceA2+, and one had no hybridisation. Positive controls (lanes 9 and 10) were 26695 (iceA1+) and J99 (iceA2+).

The finding that the 54iceA1+/2+ strains were mixed cultures suggested identification of bothiceA1+ andiceA2+ strains from a single gastric biopsy. Fifty four (63%) of the 86 patients were therefore infected with a single iceA strain. Examination of the distribution of iceAalleles in patients with single strains demonstrated that the distribution differed between the clinical groups (χ27.13, p<0.03). Specifically, significantly more patients with GC (10 (67%) of 15) were infected with iceA1strains compared with five (23%) of 22 with gastritis alone (p<0.01) (fig 6). This was not different to patients with PUD (seven (41%) of 17; p>0.1). Examination of the distribution oficeA genotypes in all 86 patients demonstrated a significant relationship (p<0.05) between infection with iceA2 strains and patients with gastritis alone (86%) and PUD (77%) compared with patients with GC (47%).

Figure 6

Distribution of iceA genotypes in all isolates. **p<0.01 versus gastritis alone; †p<0.04, ††p<0.003 versus gastric cancer.


The size and subtype of the iceA2alleles were next investigated using primer setIceA2F6/IceA2R5. As previously reported,6 7 most isolates could be divided into two types according to the presence of 105 nucleotide repeat sequences which resulted in PCR products of either ∼229 bp (iceA2B oriceA2C) or ∼334 bp (iceA2D). A number of isolates had either both 229 and 334 bp products or a combination of 229 bp and/or 334 bp and ∼449 bp (iceA2E) products. Subsequent analysis of three clonally derived isolates from the latter group demonstrated only single PCR and Southern bands (fig 5), which suggested the existence of multiple iceA2 strains, rather than multipleiceA2 gene copies in these isolates.

Sixty five (75%) of 87 patients were infected with a singleiceA2 strain. The distribution of theiceA2 subtypes was significantly different between the different clinical groups (χ2 9.4, p=0.009). Specifically, the iceA2B oriceA2C subtype (229 bp) was predominant in gastritis cases (26 of 35 (67%); p<0.02 vPUD) (fig 7) while the iceA2D subtype (334 bp) was predominant in PUD cases (14 of 21 (67%)). No differences in distribution were noted among the 22 patients with multipleiceA2 subtypes (data not shown).

Figure 7

Distribution of iceA2 subtypes in patients with gastritis alone or peptic ulcer disease.


All strains were cagA + and gave a single 3′ amplicon. By using the method of van Doorn and colleagues,7 eight different combinations based on analysis of the vacA s region (s1 and s2), and the iceA type (iceA1, iceA2B/C,iceA2D, andiceA2E) were examined in all patients (fig8). Significant associations between genotypes and clinical outcome were noted (χ2 57.8, p<10−6). Significantly more PUD isolates (69%; p<0.005v gastritis alone) and GC isolates (81%; p<0.0005 GC v gastritis) werevacA s1/iceA1. In addition, significantly more PUD isolates werevacA s1/iceA2D(53%; p<0.006 v gastritis alone). In contrast, isolates from patients with gastritis alone were identified as either vacAs2/iceA1 (21%; p<1×10−5 v PUD; p<0.0001v GC) or vacAs2/iceA2 (22%; p<1×10−5 v PUD; p<3×10−5 v GC). In addition, a similar analysis was undertaken in patients with a single iceAstrain, with concordant results (table 2). Significant associations between genotypes and clinical outcome were noted (χ230.4, p<0.0008). Significantly more PUD isolates (47%; p<0.01v gastritis) and GC isolates (67%; p<0.0004 v gastritis) werevacA s1/iceA1. In contrast, significantly more (p<0.009) isolates from patients with gastritis alone were vacAs2/iceA2B/C.

Figure 8

Combination analysis of vacA/iceA alleles and disease outcome in all clinical isolates. **p<0.01, ***p<0.005 versus gastritis alone; †p<0.05 versus peptic ulceration or gastric adenocarcinoma.

Table 2

Distribution of vacA/iceA alleles in single genotype isolates (n (%))


These findings support the hypothesis that strain differences iniceA genotypes may partially explain the differences in disease outcome associated with H pylori infection in South African populations.

Almost 40% of South African patients were distinguished by a high prevalence of mixed iceA strains while examination of cultures demonstrated that ∼50% were comprised of strains that contained different iceA types (either iceA1/2 or >1iceA2 subtypes). The reason for the high incidence of infection with mixed iceAstrains in African patients may simply reflect multiple colonisation but mixed strains do appear to be commonly found and range from 4% in the USA9 to 15% in the Netherlands,7 17% in Japan,9 and 22% in Colombia.9

In the present study, analysis of iceAgenotypes in South African patients demonstrated that althoughiceA1 was more prevalent in PUD patients than in gastritis alone patients, this was not statistically significant. Previous studies in the USA5 and the Netherlands7 but not in Southeast Asia9 have demonstrated a strong association between this allele and PUD. It is possible that the high prevalence (∼50%) of mixediceA strains in PUD patients may obscure any potential relationship between the allele and the disease. There was however a strong relationship between the combination ofvacA signal sequence subtype andiceA alleles, and PUD. Significantly more PUD isolates were vacAs1/iceA1 compared with 40% of isolates from patients with gastritis alone.

The function of iceA2 is unknown. In this study, amplicon size of this gene appeared to discriminate between PUD and gastritis alone. iceA2C (229 bp) occurred in ∼62% of all isolates from South African patients with no clinically significant disease while iceA2D(334 bp) was more prevalent in isolates from patients with PUD. Combination vacA/iceA analysis demonstrated that the vacAs1/iceA2D was more prevalent in PUD patients (53%) than in patients with gastritis alone (23%). Consistent with these findings we have recently reported a significant relationship between the cassette structure oficeA2 and expression in vivo;iceA2A/B/C mucosal transcript levels were higher than iceA2D levels.21This suggests that downregulation of iceA2expression, in addition to induction oficeA, may also contribute to disease outcome. It is possible therefore that South African strains from patients with clinically significant disease may not readily expressiceA2 as the majority of these isolates encode the iceA2D variant. Alternatively, alterations in the number of repeat cassettes may result in potential changes in protein secondary structure. For example, variants 2A and 2B may exist as a single globular entity outside the cell while variants 2C and 2D are predicted to display one and two surface exposed regions, respectively. These changes may translate into differential binding and/or function of the protein.

There is a paucity of investigations of iceAalleles in isolates from patients with GC. In the one small study, gastric cancer isolates from Japan and Korea were distinguished by the prevalence of iceA1 (67%) while 75% of cancer isolates from the USA wereiceA2.9 The numbers in this study were relatively small which limited statistical analysis. In a second Japanese gastric cancer study, iceA1was associated with enhanced gastric inflammation but not adenocarcinoma per se.8 South AfricanH pylori cancer isolates, similar to Southeast Asian strains, were distinguished by the presence oficeA1 and the absence oficeA2. In contrast however the prevalence of these alleles in cancer patients was significantly different to that found in patients without clinically significant disease in this study. In addition, gastric cancer isolates, unlike those from South African patients with PUD or gastritis, were not distinguished by any specificiceA2 subtype. These data were strengthened when analysed in combination with the vacAstatus of the strains. Significantly more gastric cancer isolates were distinguished by vacAs1/iceA1 compared with patients without disease. The putative role of IceA1 in gastric cancer is not defined but experimental studies have demonstrated that long term colonisation in both mouse and monkey models of experimental H pylori infection require a wild-typeiceA1 strain.6 22 In the present study, iceA1 does not encode a restriction enzyme and probably, as has been previously suggested,14 acts as a transcriptional regulator for downstream genes. It is however possible that the higher percentage oficeA1+ strains in South African gastric cancer patients may simply reflect the observation that these patients have been potentially infected for a longer time period. Alternatively, analysis of differences iniceA1 gene transcription and therefore protein production may reveal a functional role for this protein in this condition.14

In summary, the sequences and prevalence oficeA1 and iceA2in South African H pylori isolates has been investigated, the observation that ∼40% of patients are infected with mixed strains made, and a novel iceA2Dsubtype identified. The potential association betweeniceA1 and GC,iceA2D and PUD, andiceA2C and gastritis alone has been demonstrated in our study populations. No isolate from patients with clinically significant disease exhibited either thevacA s2/iceA1 orvacA s2/iceA2allele. Overall, these data support the hypothesis that there is a difference between organisms associated with and without disease. Absolute separation cannot be made which suggests that other factors must play a role in disease pathogenesis. Analysis oficeA allelic types however is useful in South Africa and certain combinations of virulence factors may provide excellent negative markers for disease.


This study was partially funded by a David and Freda Becker Trust Award (to JAL) and by the National Institutes of Health Grant DK 02381 (to RMP). MK is a recipient of the Claude Harris Leon Foundation Fellowship. The two reference strains, 26695 and J99, were kind gifts of Dr JC Atherton, Division of Gastroenterology and Institute of Infectious Diseases, University of Nottingham, UK. Ms T Marks performed a number of the PCR reactions.


View Abstract


  • Abbreviations used in this paper:
    peptic ulcer disease
    gastric adenocarcinoma
    polymerase chain reaction
    sodium dodecyl sulphate
    dense alignment surface
    sodium chloride/sodium citrate

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.