Previously, on the story of MS2 in yeast: Last year, Roy Parker published a short article, in which he claimed that using the MS2 system in yeast causes the accumulation of 3′ RNA fragments, probably due to inhibition of mRNA degradation by the 5′ to 3′ exoribonuclease Xrn1. He argued that these findings put in question all the work on mRNA localization in yeast using the MS2 system. About a year later, we wrote a response to that article. We argued that, yes, such fragments exist, but 1. most of it stems from over-expression of the labeled mRNA. Parker agreed with that. 2. That these fragments accumulate in P-bodies, and are distinguishable from single mRNAs and we can discard cells which show these structures. 3. We argued that this might not be the case for every mRNA and should be tested on a case by case basis. 4. We and Parker agreed that the best way to determine if such fragments exist is by performing single-molecule FISH (smFISH) with double labeling – a set of probes for the length of the mRNA and a set of probes for the MS2 stem-loops. Now, a new paper from Karsten Weis’ lab shows more evidence, by doing smFISH, for the existence of these fragments.
Weis and his students read our papers and were concerned. So they took our advice and performed smFISH on unlabeled vs. MS2 or PP7 labeled mRNAs. The labeled three different mRNAs: PGK1 (which also appears in Parker’s paper), FBA1 and GFA1. These seem to be “cytoplasmic” mRNAs – no distinct localization, so the only aberrant localization is assembly into large granules. Furthermore, they compared the localization in cells cultured in the presence or absence of glucose. Glucose starvation is a known inducer of P-bodies formation in yeast.
So, they mainly looked at two parameters: 1. co-localization with P-bodies (labeled with the decapping protein DCP2 fused to GFP); and 2. the abundance and brightness of transcription sites (TS; they call it nuclear foci, or nuclear co-localization and not transcription sites, for some reason). They also performed Northern blot analysis to reproduce the results from the Parker paper.
Their conclusions: both systems cause accumulation of 3’end fragments which seem to accumulate in P-bodies. Under glucose starvation, this mis-localization is exacerbated. There are differences to the worse between using the ORF probe set and the MS2 probe set, indicating that there are many 3′ fragments. In contrast to Parker, and more closely to our results, they show that fragments occur even in the absence of the MS2 or PP7 coat protein (CP). Their Northern analysis (which looks so much better than Parker’s) shows that there are more fragments with CP expression, but in both cases there is similar P-body co-localization. So, this is a little inconsistent. Or maybe it’s the math. They show % of P-bodies that co-localize with mRNA granules, which is close to 100% only with MS2/PP7 probes. But they don’t show % co-localization of mRNA with P-bodies. This is not the same. Look at our Response to Parker, figure 2E & F.
Their TS analysis shows that under glucose starvation conditions, there are more PGK1-MS2 TSs compared to endogenous unlabeled mRNA. When labeled with PP7, this is true only if using the ORF probe. no increase is detected when using the PP7 probes. So that’s odd. For FBA1-PP7 we see an increase under both conditions; again only with the ORF probes. GFA1-PP7 is interesting, because we see a slight decrease in number of TSs, but only with the PP7 probe. In all of these cases the TS of MS2 or PP7-labled mRNAs is brighter (i.e. contains more transcripts) compared to the endogenous unlabeled mRNA.
At the end, they say that one need to be careful when interpreting data of MS2-like systems, which is more or less what we said in our response. They also point out that the U1A system might not be free of problems too. Parker showed that labeling PGK1 mRNA with the U1A does not produce fragments. However, Parker published before that PGK1-U1A accumulate in P-bodies under glucose starvation conditions, whereas here the unlabeled PGK1 does not.
This is a nice addition to the evidence for problems with MS2. But i’m still not happy with it. The biggest problem I have is that they did not do co-FISH! They have done separate FISH for each probe set. Therefore, we cannot assess, quantitatively or even qualitatively the extent of co-localization of the two sets (ORF with MS2/PP7) and determine if there are indeed 3’end fragments (i.e. only MS2/PP7 spots). Also, we cannot determine tri- localization with P-bodies.
What happens at the TS? Is it accumulation of full-length transcripts at the TS (full co-loc of both sets)? Why with PP7 labeling we see more TSs only with the ORF set? Is transcription stuck at the PP7 stem-loops?
They chose two very abundant mRNAs, PGK1 and FBA1, for most of their analysis. I suspect that these are the more problematic mRNAs, similar to what I showed with over-expression. Note that the less abundant mRNA, GFA1, has less P-body colocalization and no TS problem. They do not have any quantification of transcript number/cell for each probe set. How do we know if there are more MS2/PP7 spots compared to ORF spots? Also, just based on qualitative analysis of the images, I think there is less mRNA when labeled compared to unlabeled. Is it because of the MS2/PP7 loops that affect transcription/decay? or is it the insertion of a large piece of 1.2kb size at the 3’UTR, unrelated to whether it is stem loops or not? By the way, the Northern of the PGK1-PP7 strains here, and the PGK1-MS2 from Parker’s paper also shows less PGK1 expression when labeled. This is worse under glucose starvation, but it is hard to say from the images if there really is a decrease like that, because the large granules are so bright the singles are less visible, it seems. They do not discuss that at all.
Regarding their TS results, this could be a feedback mechanism from the decay machinery.
Why do these fragments accumulate in P-bodies? I have a theory: If they are indeed produced by Xrn1 that gets “stuck” and can’t process any further, then it could be that these fragments remain bound by Xrn1, along with other decay factors of the decaysome. We know that Xrn1D208A, a mutant form that binds mRNA but cannot degrade it, accumulates in P-bodies. So my guess is that this Xrn1*MS2 stem-loop complex accumulates and, because these are very abundant mRNAs, gets assembled into P-bodies. Maybe, the aggregation of these fragments into P-bodies protects the cell from unwanted RNA fragments, and maybe the assembly in P-bodies helps to finally degrade these fragments, even if more slowly than they accumulate.
So what’s next? On my part, I think I will do some co-FISH by myself to some of the mRNAs that we looked at in our Response, as well as maybe some others that were previously published by the Gerst lab. I am also very interested in the role of Xrn1 in this story. I might look into that if I have the time.
But more importantly, expect a new paper from the Singer lab soon. I’ve attended their recent lab meetings. Evelina & Maria did an amazing work and will soon publish the proper response to all of this…
Heinrich, S., Sidler, C., Azzalin, C., & Weis, K. (2016). Stem loop RNA labeling can affect nuclear and cytoplasmic mRNA processing RNA DOI: 10.1261/rna.057786.116