It's not all junk -- and it evolves!
We've known for a long time that most human DNA doesn't code for protein, that much of that noncoding DNA is junk (former genes that no longer do anything, multiple copies of selfish "jumping DNA", etc.), but that some noncoding DNA performs useful functions. Click "junk DNA" at right for past posts on this topic. This week's paper, Adaptive Evolution of an sRNA That Controls Myxococcus Development (published in Science by Yuen-Tsu N. Yu, Xi Yuan, and Gregory J. Velicer), is an example of how such functions can evolve.
Myxococcus xanthus is a "social bacterium", whose behavior somewhat resembles that of the "social amoeba", Dictyostelium. When starved, the individual bacterial cells get together in a mound and form spores. Previously, Velicer's group found a mutant that doesn't do this. Then a second mutation arose in that line that restored the original behavior. Now they report the molecular basis for this restored spore-forming ability. The product of the key gene turns out to be small RNA molecule. Its normal function is apparently to block aggregation and spore formation, except when starved. The new mutation essentially knocks out this function, restoring the ability to make spores, but without the normal link to starvation.