Highly Variable Mutation Rates in Commensal and Pathogenic Escherichia coli Escherichia coli in humans is a commensal inhabitant of the gastrointestinal tract as well as one of the most frequently isolated bacterial pathogens (1). In studying food- borne E. coli pathogens, as well as the ECOR collection of natural isolates (2), LeClerc et al. found that mutant bacteria— defective in methyl-directed mismatch re- pair (MMR)—were present in pathogenic strains at an unexpectedly high frequency (over 1%), thus raising the possibility of a link between mutator phenotype and pathogenicity (3). In an independent study, we have sur- veyed mutation rates among different E. coli populations isolated from distinct environ- ments. We studied 504 natural isolates, which represented the genetic diversity of the species as a whole, encompassing both human commensal [n = 216 (4)] and pathogenic [n = 288 (5)] strains. Our data (Fig. 1) show a frequency of strains bearing defects in MMR genes similar to that re- ported by LeClerc et al., but such defects were found in all the phylogenetic groups and independently of the commensal or pathogenic nature of the strains. The MMR defective strains were distributed 3/288 in pathogens versus 1/216 in nonpathogens, a difference that is not statistically significant ( 2 = 0.05, P  0.5). Likewise, taking the data in the study by LeClerc et al.(3) all together, MMR defective strains were dis- tributed 9/268 in pathogens versus 0/81 in nonpathogens, which is also not statistically significant ( 2 = 1.6, P = 0.2), the Yates correction for small numbers has been ap- plied to the chi 2 ’s). The two data sets do not differ significantly ( 2 = 0.96, P = 0.2 to 0.5). In short, the numbers are too small at this point to support any hypothesis. Besides MMR defect, however, there are other pathways leading to a mutator phe- notype. Therefore, to detect a wide range of mutator effects, we undertook the screen of all mutational events leading to gene inac- tivation (6), unlike LeClerc et al.(3), who could detect only a few point mutations in the essential rpoB gene that confer resis- tance to rifampicin (7). Furthermore, we could also detect clones with small increas- es in mutation rate because each papillating colony could be an independent assay for mutation rate. With this assay, we found that as much as 14% of bacteria had an enhanced mutation rate, the majority being mild mutators. These mutators were also present in all phylogenetic groups, includ- ing pathogenic and commensal strains of E. coli. When the mutation rate to rifampicin resistance of these mutators was monitored, a high level of polymorphism was observed, ranging continuously from less than 10 -8 to more than 10 -6 (Fig. 1). In general, the highest values correspond to MMR defi- ciencies, whereas other mutators are likely to be due to different mechanisms. A high incidence of mutators was ob- served not only among emerging pathogens, but also among classical pathogenic and commensal strains, such as those isolated from feces of healthy Dogons for whom there is no record of antibiotic treatment (8). This suggests that all bacterial popula- tions have recently experienced adaptive evolution (9, 10). Even a modest increase in mutation rate has been shown to be advantageous during the adaptive evolution of bacteria (10, 11). Therefore, a higher percentage of such mild mutators observed in some pathogenic isolates might be a con- sequence of stronger selection in that spe- cific environment (12). A direct link be- tween increased genetic variability and pathogenesis, however (for example, transi- tion between commensalism and parasit- ism), remains to be demonstrated. Ivan Matic Miroslav Radman Laboratoire de Mutagene `se, Institut Jacques Monod, 2 place Jussieu, 75251 Paris Cedex 05, France Franc ¸ois Taddei* Laboratoire de Mutagene `se, Institut Jacques Monod, and Ecole du Ge ´nie Rural des Eaux et des Fore ˆts, 19, avenue du Maine, 75732 Paris, France E-mail: taddei@ijm.jussieu.fr Bertrand Picard Laboratoire de Microbiologie, Ho ˆpital Morvan, 29609 Brest Cedex, France Catherine Doit Edouard Bingen Laboratoire de Microbiologie (ER321), Ho ˆpital R. Debre ´, 75019 Paris, France Erick Denamur Jacques Elion Institut National de la Sante ´ et de la Recherche Me ´dicale (INSERM) U458, Ho ˆpital R. Debre ´ * To whom correspondence should be addressed. REFERENCES AND NOTES ___________________________ 1. S. Falkow, in Escherichia coli and Salmonella typhi- murium: Cellular and Molecular Biology, F. C. Neid- hardt, Ed. (ASM, Washington DC, 1996), vol. 2, pp. 2723; R. D. Berg, Trends Microbiol. 4, 430 (1996). 2. H. Ochman and R. K. Selander, J. Bacteriol. 157, 690 (1984). 3. J. E. LeClerc, B. Li, W. L. Payne, T. A. Cebula, Sci- ence 274, 1208 (1996). 4. Two hundred and sixteen commensal E. coli strains were isolated between 1984 and 1985 from fecal samples collected from healthy and nonhospitalized adults. They were 69 Dogons from Mali, 84 inhabit- ants of Olib and Silba islands in Croatia, and 63 military conscripts in Paris, France. 5. A list of the origins of the studied pathogenic E. coli strains is available from the authors. 6. For example, the inactivation of the gene coding for the LacI repressor of the lac operon [J. H. Miller, A Short Course in Bacterial Genetics: A Laboratory Manual for Escherichia coli and Related Bacteria (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1992)]. 7. D. Jin and C. Gross, J. Mol. Biol. 202, 45 (1988). 8. Among 216 commensal E. coli strains we found 27 Fig. 1. Frequency of mutation to rifampicin re- sistance among mutator strains. Five hundred and four natural E. coli isolates were screened for forward mutagenesis in the lacI gene. A total of 69 strains (14%) has been found to form dark blue papillae on minimum medium containing limited glucose, X-gal, and P-gal (the latter can only be used as a carbon source by lacI - mu- tants). We monitored the frequency of rifampi- cin-resistant mutants in three independent cul- tures of these strains (median value is present- ed), as well as the frequency of 52 non-papillat- ing strains (data not shown). Non-papillating strains had an average mutation rate to rifampicin resistance of about 1  10 -8 (a value commonly found for wild-type laboratory strains like E. coli K-12) and a small variance (data not shown). Papillating strains had an average mutation rate of 2.6  10 -7 , ranging from less than 10 -8 to more than 10 -6 , thus validating our initial screen. Strains that do not have increased mutagenesis to rifampicin resistance [which reveals base substitutions (7 )] are likely to involve other classes of mutators such as those generating frameshifts, deletions, or insertions. This high polymorphism of mutation rates was observed in all groups [commensal strains isolated from France (), Mali (E), or Croatia (‚) and in strains involved in diverse pathologies, urinary tract infections (}), bacteremia (x), pus (Œ), neonatal meningitis (■), and haemolytic-uremic syndrome or haemorrhagic diarrhea (F)]. Defects in mismatch repair genes were identified by complementation with plasmid-carrying wild- type MMR genes. TECHNICAL COMMENTS www.sciencemag.org  SCIENCE  VOL. 277  19 SEPTEMBER 1997 1833