Our knowledge of the traits possessed by extraintestinal isolates of Escherichia coli, necessary for growth and survival in urine, is limited. To identify such determinants, transposon (TnphoA'1,4) mutant libraries of a clinical isolate (CP9) were generated and screened for derivatives exhibiting decreased growth in urine in vitro, and for mutants with active lacZ fusions that were induced in urine relative to laboratory medium. Using this approach we identified two genes, guaA (CPA24) and argC (CPI-1), which were previously unrecognized as being important for growth in human urine. Unexpectedly, not only does CPA24 (guaA) not grow in human urine in vitro, but it is sensitive to its effects, undergoing a 2-3 log loss of viability over 6 h. By contrast, CPA24 neither grows nor is killed in M9 minimal medium and artificial urine. Therefore, we postulate that lack of guanine or its derivatives in urine, and the inability of CPA24 to synthesize these compounds de novo, prevents CPA24 from synthesizing other guanine (or derivatives)-dependent products that are critical for growth and survival in urine. Although it seems logical that decreased growth in urine in vitro should correlate with diminished urovirulence, this concept was tested by challenging mice with CPA24 in vivo in a mouse model of urinary tract infection (UTI). Indeed, CPA24 was found to be significantly less virulent compared with its wild-type parent CP9. CPI-1 (argC) was identified because of the significant induction of its argC::lacZ fusion in urine. Subsequent testing in urine demonstrated that its growth was significantly diminished in all urine samples tested (four females, three males). Polyamine synthesis is dependent upon, in part, the arginine biosynthetic pathway. Therefore, we tested whether the induction of argC in urine and/or the decreased growth of CPI-1 was a result of low levels of polyamines or arginine in urine. The results suggest that low levels of arginine, but not polyamines, in human urine are responsible. When tested in vivo in the mouse model of UTI, CPI-1 was also found to be significantly less virulent than CP9. In summary, we have established that guaA and argC are the first genes, which we are aware of, that have been shown to contribute to the growth of E. coli in urine in vitro and both have diminished urovirulence in vivo. These results support the concept that urine can be used in vitro as a screening tool to identify urovirulence traits.