Individual expression

First, cross two strains of laboratory mice, and take heart, brain and liver tissue samples from 31 progeny. Then use microarrays to determine how the unique genetic background of each mouse influences messenger RNA levels from a checklist of hundreds of genes in each tissue.

Result: clear evidence that we have a long way to go before we really understand the molecular mechanisms that make us individuals.

At the 2007 Lorne Genome conference, molecular geneticist Professor Peter Little, formerly from the University of NSW and now at the National University of Singapore, will present the surprising results of his correlation-based analysis of gene expression in C57Black/DBA recombinant mice.

Little's team went looking for coordinated changes in expression patterns among 755 genes that are responsive to genetic variation. Coordinated change would imply shared control mechanisms: for example, by the same transcription complexes, through shared regulatory elements in the genes' promoters.

"We imagined that all the mice derived from the C57-DBA cross were genetically different, so we should see variation in their mRNA levels," Little says. "Given that genes are controlled by molecular machines that may contain tens to hundreds of components, we would expect that any genetic variation in this control machinery would affect not just one gene, but all genes under the control of that complex. So we looked for correlated variation across the 31 individual mice, and we found it."

Little and his colleagues identified coordinated gene sets by constructing graphs showing every possible pair-wise combination of all 755 genes. Extracting significant correlations was statistically challenging, but Little describes the resulting colour-coded graphs as "strikingly beautiful".

They confirmed the correlations were not an artifact of chance recombination in any individual mouse by showing they were co-regulated in all 31 mice. But they were amazed to find that while the sets were co-regulated, the direction of co-regulation differed, not just between the individual mice, but between the three tissue types in each mouse.

"We simply asked the question: how do groups of genes behave? We weren't asking if the genes are expressed at the same level in each mouse, but whether they all head in the same direction coherently, irrespective of the amount of mRNA expression."

What the team found was that they behave paradoxically. "The same group of genes might be up-regulated in the brain of one mouse, for example, but down-regulated in the kidney or liver. Even though each mouse has the same genetic background, the influence of these genetic variations on coordinated groups of genes is essentially unpredictable." Little says the end result shows that this class of genetic variation is unpredictable across the tissues of the same mouse. The second conclusion is that the same group of genes might be up-regulated in the brain of one mouse, and down-regulated in the brain of another.

"So this class of genetic variation is also unpredictable between individuals - although statistically, we may be able to make guesses about the direction of change."

---PB--- Up-regulation, down-regulation

Normally, when a particular gene is mutated, its mutant protein affects all tissues in which the gene is expressed. But that's not the way this new class of mutations works, Little says. Expression of the subsidiary genes can be changed or unchanged, in all tissues, or only in some tissues. This, he says, has important biological implications.

"In evolutionary terms, this type of genetic variation appears to be very tissue-specific. That's interesting, because when you change the regulation of a group of genes in one tissue, but not in another, it results in a diversity of tissues.

"There is some very elegant work showing that some of the master regulators that control foetal morphogenesis are genetically variant between species - if you want to change the basic body plan of a species, you change the master regulator HOX genes.

"The very nature of genetic control is that it's a complex mixture of basal, tissue-specific and transcriptional regulation, that's exquisitely sensitive to variation. Nearly always, the end result is very complex variation between tissues, and between individuals."

Little describes the implications for human research as "a bit daunting", as it is not possible to measure mRNA levels in human brain tissue to determine how genetic variation might influence a phenotypic effect in a living human brain. "It's not going to be an easy thing to cope with," he says. "If we want to study mRNA expression levels in the individual, it's going to be complicated - although, in phenotypic terms, very interesting.

"Shared control accounts for 40 to 90 per cent of the mRNA variability, so it's a very strong influence. The startling thing is that there is almost no data on the effect of changes in mRNA expression on levels of protein in the cell."

Little points to recent research which seeks to relate mRNA levels in yeast to protein levels, but which, according to a recent paper in Nature Genetics, found only a 0.2 correlation.

"It's not as if you change mRNA levels and nothing happens. You can get negative, positive or no change in protein synthesis. But the changing mRNA level does not allow you to precisely predict the effect on protein synthesis."

Little's team believes co-regulated sets of genes arose in part because natural selection favoured arrangements with important, shared functions. In some of their groups of genes, the activity of proteins involved in processes like glycosylation and Wnt signalling was highly correlated.

"This is a result of evolutionary selection for variation for a specific function," he says. "We're probably looking at a control mechanism involving multiple factors that influence all of the processing steps between transcription of the gene, and translation of its mature mRNA into protein."

These include the machinery for excising introns and splicing exons together, and the intervention of microRNAs that bind with mRNAs to form double-stranded RNA complexes that prevent ribosomes attaching, or microRNAs that attach to mRNAs and mark them for degradation.

"We're starting to see the full genetic architecture of the cell revealed for the first time, and it's saying that variation in protein synthesis is extraordinarily complex - the genetics is bad enough, but predicting the outcome of this type of variation on protein synthesis is going to be harder still.

"We're clearly looking at numerous genetic influences that all have small effects on the genetic machinery underlying complex phenotypes."

However, there must be homeostatic regulation of these mechanisms to maintain a near-optimal balance in the abundance of the proteins specific to each cell type, he says. He believes this balance is an emergent property of many simultaneous processes competing for the cell's limited resources.

"If you were to make 1000 times more of a specific gene transcript, you wouldn't get 1000 times the quantity of the mRNA or the protein," he says. "The cell would simply be unable to process the RNA or protein in such quantities from its limited resources."

Little says if all of the non-gene RNA transcripts from the mammalian genome were deleted, genes would probably be catastrophically dysregulated, simply because a huge excess of previously sequestered control factors would have a strong, kinetically determined, influence on gene expression.

"Cells must possess networks that damp down the potential extremes of variation - putting it simply, if a cell or a molecule is doing one thing, it can't do something else.