HD168443 might be created. Could it havethe same history as a multiple star system?These are thought to start with the collapse ofa dense molecular cloud core to form a flat-tened protostellar disk, followed by frag-mentation into multiple protostars. Ourunderstanding of this process is still hazy, butit seems an unlikely way to give birth to a tight stellar system in which the mass of the primary star is 50 to 100 times moremassive than its companions. This is becausesubsequent growth by gas accretion from thedisk and an infalling envelope of gas wouldadd mass preferentially to the smaller proto-stars, leading to more equal stellar masses4.Furthermore, fragmentation calculations donot favour the formation of low-mass proto-stars in stable, planet-like orbits around asingle much more massive protostar. Whenmultiple protostars form by fragmentation,they begin life in unstable configurations,which are then subject to gravitational scat-tering during close mutual encounters.We are therefore left with trying toexplain the formation of the HD168443 sys-tem by the process of planetary formationwhich unfolds well after the central protostarhas gained the bulk of its mass. There are two competing mechanisms for the for-mation of gas giants in a protoplanetary disk:core accretion and disk instability. The cur-rently favoured theory, core accretion, wasdeveloped to explain the formation ofJupiter and Saturn through the formation ofa rock and ice core, followed by the accretionof gas. Core accretion could account for theformation of planets with masses up to about5 MJ(refs 5–7), but whether it could produceobjects as massive as the companions ofHD168443 within the lifetime8of a typicalprotoplanetary disk (a few million years)remains to be seen. Disk instability9, on theother hand, is quite capable of rapidly form-ing massive protoplanets, as large as 17 MJ,with change to spare.Theorists have their work cut out toexplain the formation of HD168443’s un-expected companions. The planet hunters,meanwhile, are discovering bizarre solar systems at an alarming rate.■Alan P. Boss is in the Department of TerrestrialMagnetism, Carnegie Institution of Washington,5241 Broad Branch Road NW, Washington DC20015-1305, USA. e-mail: [email protected]. Mayor, M. & Queloz, D. Nature 378, 355–359 (1995). 2. Marcy, G. W. et al. Astrophys. J. (submitted). Seehttp://exoplanets.org3. Marcy, G. W., Butler, R. P., Vogt, S. S., Fischer, D. & Liu, M. C.Astrophys. J. 520, 239–247 (1999).4. Bate, M. R. Mon. Not. R. Astron. Soc. 314, 33–53 (2000).5. Bodenheimer, P., Hubickyj, O. & Lissauer, J. J. Icarus 143, 2–14(2000).6. Kley, W. Mon. Not. R. Astron. Soc. 313, L47–L51 (2000).7. Nelson, R. P., Papaloizou, J. C. B., Masset, F. & Kley, W. Mon.Not. R. Astron. Soc. 318, 18–36 (2000). 8. Briceño, C. et al. Science 291, 93–96 (2001).9. Boss, A. P. Astrophys. J. 536, L101–L104 (2000).news and viewsNATURE|VOL 409|25 JANUARY 2001|www.nature.com 463Figure 1 Pattern of discoveries for gas giantplanets and brown dwarf stars found orbitingnearby stars similar to our Sun. In most cases,only the minimum mass of the companion isknown, although this value is likely to be closeto the true mass. The star HD168443 has nowbeen found to have two massive companions(HD168443b and HD168443c) that blur thedistinction between giant planets and browndwarf stars. In particular, the mass ofHD168443c falls into the observed gap (fromabout 15 MJ to 30 MJ) between gas giants andbrown dwarfs. The radial-velocity techniquepreferentially detects massive, close-in planets,as indicated by the detection bias along thedashed line.HD168443cHD168443bJupiterSaturn1003010310.30.1EccentricCircularSeparation (Earth–Sun distance)Mass (Jupiter masses)0.01 0.1 1 10 100Humans and Escherichia coli normallylive happily together: E. coli is a benefi-cial bacterium commonly found in thehuman gastrointestinal system. But it alsoexists in harmful forms. One of the mostharmful of these, called O157:H7, was firstlinked to human disease in 1983 (ref. 1),when it was shown to have been the cause of two outbreaks of an unusual and severegastrointestinal ailment in the United Statesthe previous year. The number of docu-mented human illnesses and deaths causedby O157:H7 strains has since increasedsteadily worldwide, and these strains arenow considered to be both emerging patho-gens and major threats to public health2.Studies of O157:H7 receive a boost from apaper on page 529 of this issue3, in whichPerna and colleagues describe and analysethe genome sequence of one strain of E. coliO157:H7.Outbreaks of O157:H7 infections inhumans have been traced primarily toinfected cattle4, which are the source of contamination of ground beef, milk and —indirectly, through fertilizer — many fruitand vegetable products. Fortunately, propercooking or pasteurization can preventhuman infection by contaminated food. But infections can also be transmitted bysewage-contaminated water and person-to-person contact.Genetically, the pathogenicity (ability to cause damage) and virulence (degree ofpathogenicity) of O157:H7 strains dependon several factors. For example, these strainspossess genes encoding the so-called shigatoxin, as well as small, circular DNA mol-ecules that encode ‘virulence factors’. AndO157:H7 has at least one pathogenicityisland — a section of chromosomal DNAcontaining many genes that contribute topathogenicity5. Evolutionary studies haveshown that all O157:H7 strains are closelyrelated, and that they share a common ancestry with pathogenic O55:H7 strains6MicrobiologyGastrogenomicsJonathan A. EisenThe genome of an Escherichia coli strain that is emerging as a severe threatto human health has been sequenced. Comparing it with that of a harmlessstrain suggests why some forms of this bacterium cause disease.Ancestorof species1 and 2Species 1Species 2Insertion in 1Insertion in 2DeletionabcFigure 1 Model of the generation of genomeislands specific to just one of a pair of relatedstrains or species. a, The ancestral circulargenome. b, Modern genomes — the results ofdifferential insertion and deletion of genomicsegments in each lineage. c, Genomes 1 and 2linearized and aligned. ‘Islands’ within eachgenome are highlighted on alignment (blue for species 1 and purple for species 2). Note that islands in one genome are generated bothby insertion of genetic segments into thatgenome and deletion of segments from the other genome.© 2001 Macmillan Magazines