A Science News Focus piece by Don Monroe ("Genomic Clues to DNA Treasure Sometimes Lead Nowhere") presents the concerns that "not all conserved sequences are important and, worse, that not all important sequences are conserved." While I think that formulation is a bit misleading, it does point to some very interesting and timely questions in genomics. Eddie Rubin and colleagues have shown that "deletion of ultraconserved sequences yields viable mice" (PLOS Biology 2007). While this is not the same as showing that the sequences are not important, it does point to an important specific question ("What are these noncoding ultraconserved sequences in vertebrate genomes doing?") and an important general question ("Why is the correlation between gene importance and gene evolutionary rate so weak?" Wang and Zhang 2009). The article got me thinking about those questions.
However, the conservation of nonessential sequences is not new, and there are several well-established means by which the loss of sequences important enough to be maintained by purifying selection can fail to produce a phenotype. First, the specific sequences tested can be redundant. Second, the process under selection can be important without being essential. Examples of widely conserved processes that are not essential in all species include telomerase and nonsense-mediated decay. Third, the selection can be imposed by something (such as a rare pathogen) that does not arise in the experimental system. However unlikely these cases may seem, I know of no means other than purifying selection by which a sequence can be maintained unchanged for millions of years.