Most strains of Helicobacter pylori from patients with peptic ulcer di
sease or intestinal-type gastric cancer carry cagA, a gene that encode
s an immunodominant protein of unknown function, whereas many of the s
trains from asymptomatically infected persons lack this gene. Recent s
tudies showed that the cagA gene lies near the right end of a approxim
ate to 37 kb DNA segment (a pathogenicity island, or PAI) that is uniq
ue to cagA(+) strains and that the cag PAI was split in half by a tran
sposable element insertion in the reference strain NCTC11638. In compl
ementary experiments reported here, we also found the same cag PAI, an
d sequenced a 39 kb cosmid clone containing the left 'cagII' half of t
his PAI. Encoded in cagll were four proteins each with homology to fou
r components of multiprotein complexes of Bordetella pertussis ('PtI')
, Agrobacterium tumefaciens ('Vir'), and conjugative plasmids ('Tra')
that help deliver pertussis toxin and T (tumour inducing) and plasmid
DNA, respectively, to target eukaryotic or prokaryotic cells, and also
homologues of eukaryotic proteins that are involved in cytoskeletal s
tructure. To the left of cagll in this cosmid were genes for homologue
s of HsIU (heat-shock protein) and Era (essential GTPase); to the righ
t of cagll were homologues of genes for a type I restriction endonucle
ase and ion transport functions. Deletion of the cag PAI had no effect
on synthesis of the vacuolating cytotoxin, but this deletion and seve
ral cag insertion mutations blocked induction of synthesis of proinfla
mmatory cytokine IL-8 in gastric epithelial cells. Comparisons among H
. pylori strains indicated that cag PAI gene content and arrangement a
re rather well conserved. We also identified two genome rearrangements
with end-points in the cag PAI, One, in reference strain NCTC11638, i
nvolved IS605, a recently described transposable element (as also foun
d by others). Another rearrangement, in 3 of 10 strains tested (includ
ing type strain NCTC11637), separated the normally adjacent cagA and p
icA genes and did not involve IS605. Our results are discussed in term
s of how cag-encoded proteins might help trigger the damaging inflamma
tory responses in the gastric epithelium and possible contributions of
DNA rearrangements to genome evolution.