How do miRNAs affect protein production?

A recent paper from the Bartel and Weissman groups (Guo et al. Mammalian microRNAs predominantly act to decrease target mRNA levels, Nature 466 835-40, PMID: 20703300) provides an interesting snapshot of the journey of a field from consensus to controversy to (one day?) consensus again.

At issue is the question of how microRNAs — small RNAs that control gene expression — have their effect.  Clearly they bind specifically to messenger RNAs that carry a short target sequence; clearly, the overall result is that a reduced amount of protein is expressed from the targeted mRNA.  But is the translation of the mRNA blocked, or is the mRNA itself destabilized, or both?  The answer could affect everything from the way you measure an miRNA effect to the way you think about choosing targets for therapeutic applications.  And the consensus in the field seems to be swinging fairly hard — or has swung, depending on who you talk to — from one extreme (the main effect is on translation) almost all the way to the other (most, but not all, of the effect is due to mRNA destabilization).

The earliest reports of miRNA mechanism presented evidence for inhibition of translation, and this explanation dominated the field for years.  About 10 years later, controversy reared its ugly head, as evidence for mRNA destabilization started coming out.  Until relatively recently, it was possible to take the view that different miRNA targets simply behaved differently, and the destabilized mRNAs were the exception to the rule. A review in the 2010 Annual Reviews in Biochemistry devotes over twice as much space to evidence for translation inhibition as it does to evidence for mRNA destabilization.  But in the last few years the balance of evidence has been shifting, as studies surveying large numbers of miRNA targets have provided increasing evidence for mRNA destabilization as a, if not the, major factor in the reduction of protein levels caused by miRNA.  And each successive paper states this initially controversial conclusion more boldly.

Two papers in 2008 — one a collaboration between the Bartel and Gygi groups, the other from the Selbach and Rajewsky groups — used mass spectroscopy and microarrays to systematically measure how protein expression and mRNA levels changed in response to miRNA overexpression.  The main point of these papers was that each individual miRNA regulated (though to a modest degree) hundreds of targets, but both studies also found that miRNAs reduced mRNA levels.  The Bartel/Gygi paper went so far as to state in the abstract that for many targets the reduction in the mRNA level was the major effect responsible for the reduction of protein expression.

Last year, a three-way collaboration between three groups at Stanford, led by Pat Brown, set out to measure just how much effect miRNAs have on protein translation, and how much on mRNA levels.  This study measured mRNA abundance (using DNA microarrays) and translation rate for more than 8,000 genes.  To determine translation rate, they looked at two different measures of translation: (1) what fraction of the mRNA carries at least one ribosome? and (2) what is the average ribosome density on the mRNA?  This approach uses the fact that mRNAs with different numbers of ribosomes attached to them run at different densities on a sucrose gradient.  Having separated polysomes containing different numbers of ribosomes, you can then use a DNA microarray to identify and quantitate the mRNAs in each fraction.

Using data from hundreds of mRNA targets of the miRNA miR-124, this study found that although changes in translation were detectable, they were typically small.  The reductions in mRNA abundance explained, on average, about 75% of the reduction in protein level — and the reductions in protein levels were, in general, fairly modest.  The authors suggested a possible reason for the fact that previous papers had come to such different conclusions: many studies have used reporter mRNAs with artificial 3′-UTRs, carrying multiple binding sites for the miRNA being studied and lacking the binding sites for regulatory proteins and RNAs that are typically present on a mammalian mRNA.

Unless you simply disbelieve the results in these three papers, you have to start acknowledging the possibility that mRNA destabilization could be the major effect for most miRNA targets.  But there are still a few loopholes.  The largest loophole is that the paper from Brown and colleagues, though it studied hundreds of mRNAs, used only one miRNA; and the miRNA was introduced into cells that don’t normally express it by transfection.  Do these results correctly represent the behavior of endogenous miRNAs?

Enter Guo et al., who use a recently-developed technique (ribosome profiling) to study the effects of knocking out an endogenous miRNA as well as transfecting in two other miRNAs.  This approach makes use of the ability of so-called “deep sequencing” to give you a quantitative measure of the abundance of an RNA: you count how many times you read a particular sequence, and that’s a measure of how often the sequence is represented in your sample.  In ribosome profiling, your goal is to determine how many ribosomes are found on a particular mRNA.  So you arrest translation using cycloheximide, leaving the ribosomes stuck on the mRNA, then digest all the exposed bits of RNA using RNase I.  Now the only intact RNA is a collection of segments that were protected by the arrested ribosomes.  You sequence those, and you get a picture of how many ribosomes were on the mRNA and (by comparing the sequence you found with the sequence of the mRNA) where they were located at the time you stopped translation.

Guo et al. did this, and, in parallel, used mRNA-seq to quantify the level of each mRNA.  They combined this new data with the SILAC-based measurements of protein levels from the earlier Gygi and Bartel paper.  Putting all of this information together, they have information on mRNA levels and ribosome occupancy for thousands of mRNAs, and for about half of them they have a direct comparison with mass spec measurements of protein levels.  So, they can now ask how much of the change in the level of a protein is explained by the change in the level of the corresponding mRNA, and whether the change in translation rate (measured by ribosome occupancy) matches that interpretation.  Even for proteins that were not measurable by SILAC, they can look at mRNA and translation levels and ask how much translation is affected.

This is turning into a long post, so I’ll jump to the bottom line: they conclude that for the average miRNA target, for all three miRNAs they looked at, 84% of the change in protein level is attributable to changes in mRNA level.  They looked for exceptions, targets where the main effect was due to translational changes, and didn’t find any.  They also looked for evidence that miRNAs might cause ribosomes to drop off the mRNA late in the process of translation — which might make it harder to spot a difference in translation rate using the ribosome profiling technique — and found no evidence of lower ribosome density near the end of the mRNA compared to the beginning.  The small effect miRNAs have on translation therefore seems to be due to reduction in initiation of translation, not premature termination of elongation.

No chain of evidence is perfect, of course: Guo et al. point out several caveats, such as the possibility that miRNAs might slow the process of elongation on mRNAs, distorting the ribosome profiling results.  But the explanations that would be needed to make these data consistent with the idea that miRNAs mostly have their effect through inhibiting translation are fairly hard to believe.  They would be very interesting if true, though.  It’s not outside the bounds of possibility that miRNAs will provide yet another shock to the central assumptions of molecular biology.

So — do most miRNAs have their effects through mRNA destabilization?  Might long-lived mRNAs be more affected by destabilization than short-lived ones, and does this explain why some mRNAs are better targets for miRNA destabilization than others? How do the miRNAs destabilize their mRNA targets, anyway?  Will the Guo et al. paper be viewed as the last word in the field, or will someone else come along with yet stronger evidence on the other side?  Bets, anyone?

Guo H, Ingolia NT, Weissman JS, & Bartel DP (2010). Mammalian microRNAs predominantly act to decrease target mRNA levels. Nature, 466 (7308), 835-40 PMID: 20703300

Hendrickson DG, Hogan DJ, McCullough HL, Myers JW, Herschlag D, Ferrell JE, & Brown PO (2009). Concordant regulation of translation and mRNA abundance for hundreds of targets of a human microRNA. PLoS biology, 7 (11) PMID: 19901979