"Spreading silence" is not unique to plants - it was PhD student Su Guo's observation of the phenomenon in the nematode C. elegans in 1994 that alerted US geneticists Professor Craig Mello and Professor Andy Fire to the presence of a systemic, RNA-based gene-silencing phenomenon in nematodes, earning them the 2006 Nobel Prize for Medicine and Physiology.

When used as a rootstock, the Dcl3 mutant failed to transmit a GFP-silencing signal to a GFP-expressing scion. Yet, when used as a scion, it continued to glow green. So a Dcl3-diced RNA could not be the messenger.

"It was unable to receive the signal and convert it into GFP silencing," Waterhouse says. "That was a real eye-opener."

If the elusive signalling molecule wasn't a small, diced RNA, some other mechanism had to be at work: possibly an epigenetic effect, operating at the level of the GFP gene, or its messenger RNA, that repressed transcription.

Five days after dye was injected into the rootstock, it began moving into the aerial parts of the plant, including the leaves.

By this stage, the scion had developed five or six leaves, from tiny bumps on the stem, called leaf primordia, that were already present at the time of grafting. These unsilenced leaves expressed GFP.

But the seventh and all subsequent primordia, newly differentiated from meristem tissue, produced red leaves: the GFP-silencing signal had reached these tissues.

And while the evidence was that it had arrived via the vasculature, the phloem tubes ended in disorganised tissue below the meristem, as the cells in the elongating shoot continued to differentiate.

It was as if the genes detecting the signal were silencing GFP expression by some epigenetic mechanism blocking the transcription machinery's access to the gene's promoter.

But this process clearly did not occur in pre-formed leaves which still glowed green - they were already committed to express GFP.

So silencing must be occurring in the undifferentiated meristem cells.

The CSIRO team's working model involves three layers of meristem cells. The third, lowermost layer forms the elongating vasculature as the shoot grows, while gene-silencing occurs in the second "action" layer.

As the cells of the second layer divide and differentiate, they are pushed sideways and upwards, carrying the third layer of undifferentiated meristem cells on top of the growing shoot.

"If the signal is coming up the vasculature and flooding the region in which the stem cells are located, they are possibly perceiving the signal and silencing the GFP gene, so the tissues no longer glow green," Waterhouse says.

"Now imagine that these stem cells are dividing and producing the tissues for the next. So what we might be seeing, instead of the mobile signal spreading through the leaves, is that it is converting the stem cells and the differentiated tissues derived from them to the same state, so the leaf is red instead of fluorescent green.

"We seem to have stumbled on some epigenetic process going on in meristem cells.

"Interestingly, if you do a graft, then cut off the head and grow it in nutrient media, it grows poorly, producing apical growth that throws out leaves all over the place, but no flowers. But if we make the graft lower down, the top part of the graft throws out lateral green roots, that continue to grow, allowing the plant to grow and flower."

With this approach, Waterhouse's team made a plant that produces a pulse of the silencing signal, then allowed it to grow normally and set seed. Arabidopsis seed forms in small pod-like structures called siliques.

When they broke open the siliques, the inner surface was still silenced, i.e. red - GFP silencing had occurred. But the newly formed seed fluoresced bright green. The silencing signal had been lost, or erased, from the germline cells that formed the seed.

"The more boring possibility is that the signal is lost during meiosis, when the pollen or the ovule forms. But curiously, if you look at the forming floral parts in the silenced plants, you can see that GFP is actually 'on' in the younger flowers - and it seems to be 'on' just as strongly in the pollen cells and ovules.

"So it doesn't seem that the silencing signal is erased during meiosis."

Insurance policy

The more intriguing possibility, says Waterhouse, is that the plant maintains a separate population of undifferentiated, virginal meristem cells. (He was delighted to find that French researchers had first observed this effect half a century ago and with typical Gallic panache, dubbed it meristem detente. Needless to say, the idea was pooh-poohed at the time.)

These cells divide and give rise to floral meristem cells, which in turn form the floral organs. But the virginal meristem cells remain unchanged and insulated against malign influences within the sanctum sanctorum of the floral meristem.

If so, these cells are the plant equivalent of the immunologically privileged germline cells that give rise to sperm and ova in animals - they are a form of insurance for the next generation against viral infections, and potentially deleterious epigenetic influences.

They are now designing experiments to test the hypothesis, which they believe makes evolutionary sense.

"Tissue culture uses these meristem cells - it's a way of clearing virus infections from clonally propagated plants," Waterhouse says. "It's possible that some of these virginal cells have been maintained unchanged for millions of years."

But it is now clear, the CSIRO researchers say, that whatever spreads the gene-silencing or virus-quelling signal through plant tissues involves more than simple diced RNA molecules.