Bench philosophy: Making Sense out of Nonsense
by Steven Buckingham, Labtimes 05/2013
You can find a worm in the best of fruits. But sometimes the best fruits are found in a worm. It was C. elegans that gave us RNA interference. And now it has given us something else – a new way of knocking genes down.
It has become something of a truism to point out that the power of genomics is limited by how fast we can do the physiology. We are crying out for fast and reliable ways to figure out what each gene in this massive bag-of-parts is doing and how they are related to an organism’s structure and behaviour. Now, a new method has come out, which lets you knock down a single gene cell-autonomously. According to the label on the tin, it is more reliable and more cell-specific than RNAi, and doesn’t dilute out over successive cell divisions. In other words, you decide exactly which cells it gets to knock down; and it lasts the generations. Well, generations of worms, at least.
The technique was announced in Genetics ( Vol. 194, 363-373, doi: 10.1534/genetics.113.149724) by Daniel Chase and colleagues at the University of Massachusetts. It is so new, it doesn’t even have a name yet. So, we at Lab Times have christened it “NMD-mediated knockdown” or NMDK for short.
The new approach takes advantage of a poorly understood mechanism in the worm that deals with dodgy mRNA. The Nonsense-Mediated Decay (NMD) mechanism identifies malformed RNA by recognising certain features in its sequence. No-one quite knows the exact details of how the NMD mechanism works, but we do know that it is fired up by mRNA with a premature termination codon (PTC), and we also know that the further the PTC is from the real stop codon, the more efficiently NMD is induced. The trick behind NMDK is to replace the gene you are interested in with one that is engineered, to get NMD to wipe out its messenger RNA (mRNA). You inhibit NMD across the whole animal and then turn it back on in whichever cells you want the gene to be knocked down. Because NMD is completely cell-autonomous, you can be sure your knockdown is restricted to specific cells.
The technique was developed by Chase to solve an age-old problem that dogs genetic studies on the nervous system: the fact that different neurotransmitters tend to use the same set of signalling proteins. On top of that, neurons are highly interconnected and influence each other all the time. “We realised early on that we couldn’t study neurotransmission using conventional genetic knockout tools that removed gene function globally,” says Chase. “We needed a way to take out genes in individual cells.”
Chase knew that Chris Link at the University of Colorado at Boulder was tagging Aβ (the amyloid protein associated with Alzheimer’s disease) with an NMD degradation sequence and inhibiting native NMD using a temperature-sensitive allele. Link was controlling degradation by cooling the animals, which allowed NMD action to kick in, degrading the Aβ transgene.
“We realised, we could put the strategy into reverse”, Chase told Lab Times, “by starting with an NMD-defective strain and then rescuing NMD function cell-specifically.” Of course, this strategy wouldn’t be useful if the tagged gene was over-expressed – the endogenous gene would have to be replaced and, therefore, it had to be expressed at normal levels.
How is the technique done in practice? First you have to get rid of the animal’s own copy of the gene, in which you are interested. You do that by introducing a null mutation into the genome. If you are not a “wormer” you will probably not realise just how easy that is in C. elegans. A couple months’ work, if all goes smoothly. Come to that, there’s a good chance you may not even have to do that – you may just be able to order your null strain from the “million mutation project” (http://genome.sfu.ca/mmp/).
That gets the real gene out of the way. Now we have to substitute it with our engineered version. Our version of the gene will have the real gene’s promoter but we will tack a 3’ UTR sequence on to the end that labels the mRNA for destruction. But what should the sequence be? It seems the trick is to keep the stop codon of the gene you are replacing, then add some other sequence with another stop codon of its own. Remember, the presence of an extra stop codon, especially at a fair distance from the proper one, is what marks the mRNA for destruction. Chase’s group used mCherry as their “target” gene, complete with its own codon, then tacked on let-858 – a highly-conserved gene that encodes, according to WormBase, a protein that “is likely to be involved in protein synthesis and/or RNA binding”. It is expressed widely and over all developmental stages, so we probably aren’t doing too much damage by adding another copy. Besides, let-858 is known to be good at firing up the NMD mechanism.
Next, we need a strain of C. elegans, in which the native NMD has been disabled. In C. elegans at least seven proteins (SMG-1 to SMG-7) are required for NMD, so a null mutation in any one of these should do the trick. We also need to be able to turn NMD back on when we want it. To do this, we will make a construct comprising the functional version of the NMD gene that is null in our host strain, along with a promoter specific to the target cell or tissue. Inject the two constructs into the germ line of the NMD-null worm, et voilà: cell-specific knockdown that lasts for generations.
So much for the theory. But does it work? Well, so far, we only have Chase’s paper to go on. They used the smg-5 null mutant worm as a background – these worms have their native NMD disable. They first set out to establish, as proof of concept, that an exogenously-introduced gene can indeed be completely knocked down in a circumscribed set of cells.
They injected two marker constructs: mCherry and Green Fluorescent Protein (GFP). The latter served as a positive control to show that the target tissue did indeed contain the transgenes. Only the mCherry construct contained the destruction tag but both were under the control of the pan-neuronal promoter, rab3. Sure enough, whereas mCherry was very clearly expressed in smg-5 worms, it was not visible in wild-type worms. This confirmed that adding the let-858 sequence (plus stop codon) did, indeed, result in loss of the protein in a way that depends on NMD.
Next, they showed that co-injecting with a construct that rescues NMD in neurons only cholinergic cells (using the unc17 promoter) results in loss of mCherry staining in those cells and, crucially, the presence of mCherry in non-cholinergic cells.
OK, so it works for an exogenous gene – you can put a new gene in and turn it off. But what about a real gene? Does the idea of nulling out a real gene and replacing it with our fifth column construct actually work? To answer this, Chase looked at goa-1, which encodes a G-protein. They chose goa-1 because a single null mutation completely eliminates any goa-1 mRNA production. Because so much goa-1 mRNA is usually transcribed it should be easy to spot differences in mRNA production and its effects. What is more, the phenotypic effects of goa-1 knockdown are easy to spot: it affects crawling and egg-laying. In short, not only did the NMD strategy completely wipe out goa-1 mRNA but the team also confirmed the expected effects on egg-laying and locomotion.
So what does NMDK give us that we didn’t have already? First, we have cell-specific knockdown. But didn’t we have this with double-stranded RNA interference (dsRNAi)? dsRNAi works by introducing double-stranded RNA molecules encoding the target gene. The cellular machinery realises there is something fishy about double-stranded mRNA and cuts it up. You can try to make this cell-specific by putting a construct for dsRNA under the control of a cell-specific promoter (“hairpin” RNA), but a lot of the transcription products are actively transported from cell to cell, resulting in uncontrolled off-target effects.
On top of that, the construct randomly gets left out when cells divide, with the result in the adult worm of a mosaic pattern of expression with some cells expressing the construct interspersed with those that don’t. Chase’s group found they could overcome this mosaic effect with NMDK and end up with lines with robust knockdown phenotypes.
How does it compare to RNAi for efficiency? The label on the tin says NMDK is more efficient than “traditional” RNAi. Chase gave his technique a serious challenge, in the form of unc4 – a gene which has proven pretty resistant to dsRNAi. When unc4 is taken out of action (at least, in null mutants), it messes up the command input to motor neurons that control crawling. With careful selection of their strains, they got mutant phenotypes in 100% of their animals, compared to 62% using RNAi.
So, if you want accuracy and reliability, NMDK may indeed be the way forward. But it involves a bit more work than RNAi. If you want global knockdown of a gene, you can do it really quickly using the traditional method: get hold of a bacterial strain (C. elegans’ food of choice) expressing dsRNA encoding your gene and serve it up for lunch (the worms’, not yours). The worms eat it and, conveniently, it finds its way through their system and knocks down the gene.
But is traditional RNAi really that easy? Well, it can be but as we have seen, it doesn’t work 100 % of the time or for 100 % of the genes. The different sensitivities of various genes to this method of knockdown have seriously crippled the impact of fast, genome-wide screens using this approach. What about other RNAi techniques? Again, you lose cell specificity because the transcription products are actively taken up by neighbouring cells. Besides, if you are going to inject constructs, why not go down the NMDK route?
What might be not-so-good about NMDK? As Chase readily admits, it makes a few assumptions, which might not hold up for all genes. A key element is that NMD is completely cell-autonomous. So far as we know, it is – but there is the possibility that NMD is itself under some sort of developmental control, so the technique may not work equally at all stages of development.
But there is a bigger worry. What exactly are we doing to the worms when we take down their native NMD machinery? Surely it will mean mRNAs with premature stop codons will accumulate, and what sort of weird and wonderful proteins might result?
Chase points out that up to 8 % of the worm’s genes naturally carry premature stop codons (a sort of side-effect of alternative splicing), which would rule these genes out for NMDK. But with up to 90 % of genes still amenable, that qualifies NMDK as a Swiss-army knife.
Last Changed: 17.09.2013