A rule for a loss-of-function approach by gene silencing is: "Protein follows message". This means that a complete depletion of the gene product follows Dicer/RISC-mediated mRNA cleavage in a time-dependent manner, and this critically depends on the turnover of a protein. If RNA interference of a specific gene target leads to undetectable levels of the mRNA, than the expression of the target is certainly affected. However, it may be important to show complete depletion of the target by antibodies against the target.
With a KO approach a specific gene is deleted in germ cells and not produced in the first place. Does this render a KO approach better and gene silencing less significant, respectively? Certainly not, but a loss-of-function by gene silencing has to be controlled. A complete loss of mRNA inevitably leads to the complete loss of the protein, it is only a matter of time. Hence, it is recommendable to check on protein level.
Hi David, I agree with Maria. When you check the effect of the gene/protein depletion at the level of an organism it represents a better model than looking at a cell culture.
@Dmitri...we used siRNA and shRNA in vivo on tissue. Verified a complete loss of protein at the histochemical and Western level. I am not asking cell culture versus animal model but the way of generating loss-of-function.
@Maria...why do you say this? I am looking for the rationale for this conclusion which I think is really just an opinion, but as we are scientists, we need to justify this on the facts.
The question is imprecise. If there is no effect what so ever on RNA and protein level, then there is no knockdown happening.
If the mRNA level is untouched but the protein level is clearly down regulated, that can mean several things:
a) you have some miRNA like action here, with translational arrest for example or some genes are regulated on the post-transcriptional level only
b) your results are not specfic but an artifact of whatever indirect or off target results
To discriminate between both cases you can test several shRNA's/siRNAs against the same target, they should give similar results in phenotype (but one could lower the mRNA nevertheless) and if you know of transcription factors regulating your target gene these should be unchanged.
Lets assume you have case "a", then it becomes an interesting question. The answer to this is not so simple. A conditional knockdown usually is a removed gene or a mutated one. A removed gene can mean a missing promoter which would mean transcription factor binding does not occur anymore. But promoters compete for transcription factor binding so you could get side effects on other promoters binding the same TFs in this case. If only the CDS is removed you can have still impacts on adjacent genes (naturally occurring antisense transcripts can be disturbed for example, some UTRs are transcribed over downstream genes for several kB and regulate mRNA stability etc) or the chromatin structure of the region can be impacted depending on the genes regulation. If the gene is mutated, the mutant can still be transcribed/translated and thus cause other side effects.
So a conditional knockout can cause all kinds of indirect effects too. If your gene is regularly transcribed but a specific protein knockdown seen - that is not an indirect or off target effect - that would in my opinion be far superior to what a conditional knockout could achieve or cause. Because you don't interfere with any of the above but the effector protein is still remove.
I think the problem is the detection limit. You say that you knocked down the gene to "undetectable levels" (I think only the protein level is interesting here). But, depending on the antibody, this might be still 10% of the physiological expression, you just don't see it. If the activity of your protein can be increased a lot by some kind of "activation", remaining activity can restore the phenotype to a level, that it can not be excluded that you will not see an effect even if it would be there with complete knock out...
Thanks for these answers, however I am not unfamiliar with both approaches. I can assure you that a conditional KO, which does not carry the caveat of how development proceded up to the point of analysis in the absence of a gene function, then depends upon the efficacy of the cre used. I really do not see these answers as addressing the point. If you can show that the siRNA wiped out gene expression (mRNA and protein), then what is the difference?
A rule for a loss-of-function approach by gene silencing is: "Protein follows message". This means that a complete depletion of the gene product follows Dicer/RISC-mediated mRNA cleavage in a time-dependent manner, and this critically depends on the turnover of a protein. If RNA interference of a specific gene target leads to undetectable levels of the mRNA, than the expression of the target is certainly affected. However, it may be important to show complete depletion of the target by antibodies against the target.
With a KO approach a specific gene is deleted in germ cells and not produced in the first place. Does this render a KO approach better and gene silencing less significant, respectively? Certainly not, but a loss-of-function by gene silencing has to be controlled. A complete loss of mRNA inevitably leads to the complete loss of the protein, it is only a matter of time. Hence, it is recommendable to check on protein level.
According to me a book would be not enough to do considerations on this topic.
1) When u KD a gene a part of a given tissue or cell population u have to consider that all the surrounding environment remains wt. With a KD made, let's say, by in utero shRNA electroporation u consider mainly (not only) the cell-autonomous functions of a gene. The knock-down cells, indeed, will move and interact in a wt background;
2) On the contrary, a KO (i do not consider conditional KOs) is supposed to have lost the gene of interest in each single cell of the body. So u cannot distinguish between cell-autonomous and extrinsic events. This may give, sometimes, phenotypes that are really different between KO and KD cells.
3) The timing of the loss of a given genes sometimes is crucial to the final phenotype. Let's do an example. I work on brain development. Some genes, like Lhx2 an Fgf8 seem to have an important role in the formation of the midline of the cerebral cortex. Their KD in mice before E0.5 strongly impairs midline formation while their manipulation after this stage has no or poor effect on midline cell specification. So it's conceivable that phenotype shown by a complete KO , where a given gene is absent since the very beginning of embryonic development, will be, potentially, very different from that induced by the KD of the same gene at different stages.
4) Finally, if we are not talking about development, but the experiments are performed on the adults, we have to consider that in vertebrates, and expecially in mammals, there are lots of redundant systems that are activated when a given gene is mutated or totally lost. These systems are easely activated and regualted, let's adapted to the needs of a given KO organism, while it will be difficult for a system abruptely unbalanced by a KD to set finely the compensatory mechanisms, expecially the extrinsic ones.
In conclusion I think that the two methods KD and KO, depending on the function, importance and stage of aexpression of a given gene, can give really different results and should be , when possible, both performed and, then, compared.
Just a little correction to my answer....a little typo...in my intentions the stage of development in which KD were performed (point 3) was E10.5...but the number 1 magically disappeared. Have a nice day
There are a bunch of good opinions and advices here...Just something else to consider when using KOs: let's say you KO gene A and as a result you get an increased expression of the product of gene B. Conclusion: gene A product is a repressor of gene B product. Right? Maybe not...The increase in gene B expression can perfectly be a attempt of the system to compensate for the loss of gene A. All that to say that results with KOs need to be more carefully analyzed and therefore, should not be always be considered the best model to go. Unfortunatelly, the question was not clear enough. I would perform both (if possible), taking special attention with low turnover proteins if you choose to go only with KD.
David, love your work, you've contributed to answering some very sticky Qs and re-livened some Qs that should still have been being asked. Thanks, and keep up the good work! The only sticky thing with this particular wicket is the difference between Ab/si/primer specificity and gene function. If one knocks out a particular gene (specifically the entire gene, which, in reality, is almost never done), one abrogates its function. Antisense RNA targets the specific RNA it is designed to target. Not many alternate transcripts (as far as I'm aware) or miRs emanating from the same locus. The biological conclusion that one draws from this is up to the conclusions they wish to make, and there's evidence of people discovering covert miRs this way. Yes, si/shRNAs are subject to off-target effects, but I "feel" that complete immediate specific knockdown gives a far greater intuition into the process governed by a particular protein at a particular instance. And all the rest is noise!
To RObert...thanks. My sense of the state of the art today is that siRNA approaches are not considered to be as 'valid' as a KO. However, generating KOs for each gene of interest remains complex (despite the recent advances in this technology). I just sense a certain bias in the field.
True, there is a bias for KO but it has its background in a solid facts.
In case of one gene I have done both KO (complete and conditional in mice) and RNAi (in tumor cells). The KO was done in such a way that the protein was still produced and only an active site of the enzyme was missing so the disruption to the chromatin structure and the metabolism was kept to the minimum. Interference, when cells were in normoxia was showing almost no protein, however, in hypoxia the expression and the activity of the protein was so enhanced that the effect of the stable knockdown was significantly decreased. With knocked-out gene it would have never happened as a lack of functional protein due to the KO means no increased activity (even if the nonfunctional protein is still produced) no matter the conditions. So the KO, if done well can give a very nice clean results, much better than in case of knockdown, but RNAi is still the faster method (and the level of the knockdown even in hypoxia was enough to get us a many interesting results).
Thanks to everyone for your answers. It seems there is no clear consensus. I guess, in referring to my initial question, the editor has the trump card, regardless of whether he/she is correct.
Dear Professor Samson, the question you asked created a very valuable discussion, I've learnt a lot. You politely described a good example of editorial/referee policy should not be a standard. My honest opinion will be less polite...
They ask you if a 1-3 year experiment would be better than the one you want to publish. I agree to complain "the results of RT-PCR should be verified by Western-blot", since having the material and antibody will take you two days to do it. My only comment is that with your 695 Impact Points on ResearchGate (regard:), there are many competitors in your field, have the interest to delay your editorial process.
Agree with your sentiment, but try to remain optimistic. With time, even editors come around. As one publishes more, we take the hard knocks more in stride.