A recent study from Germany reported the isolation of E. coli O157:H7/H- from patients with non-bloody diarrhoea and hemolytic uremic syndrome, questioning the role of Shiga toxin as the main trait of virulence for human disease. We isolated 6 sorbitol-fermenting E. coli O157:H- strains that do not contain Shiga toxin genes. The isolates originated from an outbreak (3 patients, 3 asymptomatic contacts) of non-bloody diarrhoea affecting two families sharing one household. Two children (age 10 months and 2 years) suffered severe diarrhoea over 30 and 10 days, respectively. Their uncle had moderate diarrhoea for 2 weeks. In contrast to the other isolates, the uncle's strain (EH109) did not harbour a chromosomal eae gene encoding gamma-intimin nor the plasmid gene E-hly; it also showed a PFGE pattern that was different from the unique pattern of the other isolates. Employing PFGE, phage typing, and P-gene typing, five of the six stx negative isolates were indistinguishable from the stx 2 positive "Bavarian outbreak strain". The only human serum tested, obtained from one asymptomatic contact, contained antibodies to the O157 lipopolysaccharide antigen. Our finding of five stx negative sorbitol-fermenting E. coli O157:H- isolates (harbouring eae and E-hly) associated with an outbreak of non-bloody diarrhoea supports the hypothesis that Stx production is not obligatory for the pathogenicity of E. coli O157 for humans.