While all anyone can talk about is Terry Schiavo, who is, after all, just a single person without much cognitive function left, a letter in this week's Nature (confusingly, short articles in this most prestigous of scientific journals are called letters) challenges much of what we know of genetics.
As I briefly summarized yesterday, plant genetecists at Purdue have observed unusual reversions of a mutant gene to the wild-type sequence in the plant Arabidopsis. They study a gene called HOTHEAD, that causes unusual fusion of flower petals when mutated. They observed that a substantial fraction of offspring of parents with normal flowers and were puzzled. Sequencing showed that in each case the gene was identical to the "normal" sequence. They did a pretty good job of ruling out contamination of there stocks or pollenation by wild-type plants by doing experiments like looking at embryos or cross-breeding different strains.
Then they tried to exclude the possibility that another DNA sequence in the genome was acting as a "template" to restore the correct information. This kind of process would be unusual and interesting, but probably not revolutionary. In this they did a less good job, in my opinion.
They noted the sequence alwasy went back to the parental, and cite, but don't show, more data showing sequences can be recovered several generations later. They note that other related genes have other sequence differences that aren't found in the reverted gene, but don't do a search of the Arabidopsis genome to exclude a stretch of matching DNA outside a known gene, despite the fact that the Arabidopsis genome is sequenced.
They hand-wave a bit about RNA storage of genetic data, but there are several problems with this.
First, they don't attempt to document wild-type RNA sequences. This would be pretty straightforward using RT-PCR (first RNA is reverse transcribed into single-stranded DNA and then amplified by PCR). Interpretation of data MIGHT be hard because only a few percentage of RNAs would carry it and PCR enzymes are error prone, but I think you have to try.
Second, this would imply a complex genetic apparatus of which we have remained blissfully unaware despite all our knowledge of cell biology, having sequenced the genome of multiple organisms, etc. Just think of all the genes that would be needed to store copies of RNA (which would have to be passed on in pollen), activate some sort of DNA repair enzymes that use RNA as a template only under certain conditions of stress, make the repairs, etc. How many generations is RNA retained and when is it replaced. Recall that RNA is far less stable than DNA and therefore more prone to damage.
Additionally, I'd be a lot more confident in their data with a few more controls. One good one they didn't do was to make plants with genes that were slightly altered via silent polymorphisms. Would those then be altered back in HOTHEAD deficient plants? What if they were introduced into a strain with slightly different sequences. Would those be reverted too. This would be a lot of work (I know n0thing of the details of plant work, I purposely ignored those sections druing grad school), but before you publish a paper challenging the basis of genetics you have got to have your I's dotted and T's crossed
Overally, I'm a skeptic. I suspect they have a gene conversion going on and just haven't recognized the template yet.