Bacterial genome GC content varies from ~13% to 75% GC. Understanding the source and maintenance of this variation is important for understanding the evolution of codon usage bias and protein-coding sequences. Both neutral processes (like mutation) and selective processes (environmental selection) are thought to shape GC content. One environmental factor, nitrogen limitation, is thought to select for AT-rich genomes. We hypothesized that if this is true, over evolutionary time scales genomes with an underlying GC->AT mutational bias would have a selective advantage over the wild-type under nitrogen-limitation. Our data suggest that under nitrogen-limitation, mutants of Escherichia coli with altered mutational biases vary in initial fitness, suggesting that under selection, these mutants would have different probablilities of becoming fixed in the population. We find that a mutant with a GC->AT bias does not have a competitive selective advantage over WT, but a mutant with an AT->GC bias has 30% higher fitness compared to the WT in nitrogen limitation, in a 24-hour competition. We find that the GC->AT biased mutant has a selective advantage over WT in rich media. This suggests that under nitrogen limitation, a GC->AT biased mutant would only rise to high frequencies under genetic drift, and we are currently testing whether this trend continues over longer timescales. We are addressing this question by evolving mutants with biased mutational spectra under various environmental conditions implicated in GC content evolution and asking whether the spectrum of mutations fixed under selection is different from that fixed under drift.