How do you know that the spots you are detecting are single RNA molecules? Couldn't they be conglomerates?This is a good question, and one that has a variety of answers. Many of the control experiments that Sanjay did are in his excellent Vargas et al. PNAS 2005 paper. One (beautiful, in my mind) experiment that Sanjay did was the following. Hein vitro synthesized a bunch of target RNA and put it in two different tubes. In these tubes, he labeled the RNA with probes, with the RNA in each tube labeled with a different dye (say, red or green). Then he combined the two tubes, so he had one tube with RNA that was either labeled with red probe or green probe, but not both. He then injected these into the cell and observed. If the RNA were forming conglomerates, then you would expect yellow blobs containing both red and green RNA. If they were single molecules, though, you would expect the spots to be either red or green but never both. The latter is what he observed. You might question whether this holds for endogenous RNA, but he expressed that RNA and compared intensities, and it was the same. This means that the endogenous RNA was also single particles. Nice! Definitely caveats to this, and technically it applies only to this RNA, but whatever, I think this is pretty solid.
There are other things you can do. One is to measure the fluorescent intensity of the spots and show that you get a unimodal distribution of intensities. Pretty weak in my mind, because if you had some spots with two RNA and some with one RNA, these peaks would overlap so much that it would probably look like a unimodal peak anyway. But what do I know.
To me, one of the strongest experiments are some new results from Eric Lubeck and Long Cai (Lubeck and Cai, Nat Meth 2012). They use super-resolution microscopy to actually read out a barcode of different colors along a single RNA molecule. Think about how cool that is for a minute! Anyway, it's very hard to imagine that conglomerates of RNA would show anything like that sort of thing. I think Sanjay has some other similar experiments that corroborate this.e (maybe good, maybe bad) in both conditions.
Some will tell you that you can detect the same transcript with two different probe sets and look for colocalization between the colors. The idea is that if you detect with both colors, that means that your efficiency is high. I don't think that actually makes sense–if you have an RNA that is inaccessible for whatever reason, this control tells you nothing, and if you have an RNA that is accessible, then a single color will probably detect it. This two color colocalization approach is good for specificity, though...
Lubeck and Cai estimated a hybridization efficiency of around 60-70%, and we have seen similar numbers. Hard to know for sure why it's not 100%, but whatever, if you get enough oligos, you'll be fine.
We have found that, for instance, double labeling oligos can lead to greatly diminished signal. We're not 100% sure why, but we're guessing it may be dye-dye quenching, either from the two ends of the same oligo or from oligos bound to neighboring targets on the target RNA. Either way, we've found they're more expensive and don't work as well, often alarmingly so, so we don't use them. Then again, the Singer-style oligos (~50mers) had multiple labels on longer oligos, and so they can definitely work, but they are a bit of a pain to synthesize.
Why indeed! We've had success with somewhat shorter oligos, like 17mers, but below that we have seen some loss of specificity/increased background. Longer oligos can work as well, and we've tried up to 30mers with success. But they use up more real estate on the target RNA and do sometimes seem to give more background, presumably due to more opportunity for non-specific interactions. I suppose you could try messing with the stringency of the hybridization and wash conditions, but we have found little need to go to longer oligos, especially since they cost more and allow fewer oligos to bind to the target RNA.
In some of our early experiments (Raj et al. PLoS Bio 2006), we targeted oligos to the PP7 RNA hairpin, which is a very strong secondary structure, and saw great signal. Same for targeting MS2 RNA hairpins. So I'm not so worried about it.
So the "transcription sites" are bright foci that you often find in the nucleus that look a bit brighter than the rest of the single mRNA spots. Here's an example:
What we think is happening here is that there is a pileup of nascent transcripts that sit around waiting for processing (or for transcription to finish) at the site of transcription itself. We and others have confirmed by combined DNA/RNA FISH that these really are at the site of transcription itself (see Levesque and Raj, Nat Meth 2013). Note that not every cell will show this site of transcription; this is because transcription is inherently pulsatile, with the gene itself switching between transcriptionally active and inactive states. In cycling mammalian cells, you will usually see between 0 and 4 spots–0, 1, or 2 if in G1, 0, 1, 2, 3, 4 in S/G2. Sometimes you can even see a "doublet" after replication; indeed, this is the case in the depicted example, where the upper arrow points to a single transcription site, whereas the bottom arrow points to a doublet (for a total of 3 transcription sites in this cell). The intensity of the transcription sites can vary from being just as bright as a single RNA molecule (in which case you use colocalization with an intron probe to discriminate; see Levesque and Raj Nat Meth 2013) or sometimes as bright as 10-50, but usually they're around 3-10 times as bright as a single RNA. If you really want to be very careful about whether something is a transcription site or not, then we recommend costaining an intron and looking for colocalization between the intron and the exon.
For a long time, the answer was no, but Marshall Levesque (awesome former PhD student in the lab) came up with a nice method based on the toehold exchange work of David Zhang. You can read about it here.
The probe designer does a bunch of bioinformatic analysis that restricts areas prone to cross-targeting and background. It also uses some more information we have gathered over the years that tells us about what sequences are more prone to background. Sorry, can't really divulge more than that for various reasons...
In our initial paper (Raj et al. Nat Methods 2008), we used 48 oligos, and the implication was the more the merrier. In some sense, that is true, but we have found that actually we seem to get the best results with around 30 oligos most of the time. If I had to guess, the reason is that each oligo also adds background, so given our currently probe design rules, 30 seems to be a nice balance between having enough oligos on there while also keeping background reasonable.
This seems to be fairly target specific. We have definitely gotten really good signal with around 12 oligos, but have gotten pretty crummy signal with 12 oligos, and crummy signal even with 48 oligos. Just seems to depend on how well the individual oligos perform, which is hard to predict a priori. Again, we think 30 is a pretty good sweet spot.
Well, depends on short. You might be able to get away with just a few oligos, potentially enabling detection of a transcript that is only a few hundred bases long. Only thing is that the probe designer may not let you do that because some of the sequences are "bad" for whatever reason. Ignore the probe designer at your own peril, though–you could end up with a lot of background, although who knows.
If you're targeting mRNA, stick to the coding sequence if possible. This is under the most selection pressure and will have the least amount of sequence "contaminants" that can cause trouble. Sometimes this just won't allow enough oligos, though, in which case, then we start venturing into the UTRs. Can cause trouble, but with the latest probe design software, it's a bit less of a risk.
For lncRNA, just use the regular probe design software. You may have to watch out for some off-target effects, though (see troubleshooting). If you get bright nuclear spots, they might be real, might be not real. Do the odds/evens approach.
Introns can work just fine. You will usually get a bright spot at the site of transcription (if the gene is actively transcribing–remember that most genes, even GAPDH, show bursts of transcription). We try to stick to the most 5' part of the intron and also try not to let the oligos spread out too far (i.e., limit to around a 2-3 kb region). You might need to go further if too much of the region is untargetable for whatever reason, but that's how we start. Also, probably best to stick to a single intron, but sometimes you have to spread across multiple introns. 20 oligos will usually do the trick for introns.
Well, depends what you want to do. If you want to make a "pan" isoform probe that targets all the isoforms, we usually use the UCSC genome browser to find the isoform that is common to all the probes and design against that. If you want to target a specific isoform, well, then you have to make probes against a unique part of the isoform, which can be hard because they're often quite short. There was a nice RNA FISH paper (Waks et al. MSB 2011) that looked at alternative splicing using our method, but targeted isoforms that had really big differences.
We have some software that we haven't yet made publicly available that can design a set of specific oligos that minimize cross targeting with a desired stringency. Could be a while before it's available for general use, though. If you want to give it a try manually, I think you want at least 4 mismatches to ensure that the probe doesn't bind off target.
As a general rule, the answer is no. Even if there is a strong degree of conservation at the protein level, there are typically enough sequence differences at the nucleotide level to prevent the probes from binding–even just a couple base mismatches can make a huge difference. You might get some weak binding, and you might even get some signal, but it won't be very robust and so we strongly recommend just getting different probe set for each species-specific transcript. In the long run, you will be much better off that way.
We've gotten a lot of questions about imaging, and this is often a point of failure for many people. Hopefully this can help answer a few questions that people often have.
What imaging setup should I use?
Well, that's a hard question, since there is so much variety, and many things can work, although not all of them will work, certainly not right away. But...
Not typically. The issue is that you will not typically get enough light because the numerical aperture of the lens is typically smaller. You might be able to see overall differences in fluorescent intensity, but you will have to do all the appropriate controls to check if it works.
The short answer is no. The longer answer is maybe. I've definitely seen some people use confocal and get decent images. Personally, I've only had good luck using a spinning disk confocal, but never managed to get anything real out of a point scanning confocal. That said, if you know enough about optics to disagree with me, then you probably know enough about optics to get it to work. If you are just starting, I STRONGLY recommend starting with just a regular widefield fluorescence microscope.
I can't see any signal even though I used the 40x scope in my lab or my core facility's fancy confocal microscope!See above. Seriously.
We used cooled CCD cameras. We have found that using a non-cooled CCD typically just has too much background to see the signals. But you don't need a super crazy cooled camera, nor do you need an EMCCD. A standard CoolSNAP-class -20C CCD will do the trick. In fact, we have found that the EMCCDs don't really buy you much, and so are not really worth all the extra expense.
Thomas Gregor at Princeton does this, I think. And I'm pretty sure other people are doing light sheet microscopy as well.
Well, technically, because our spots are usually spread out, we are already localizing the center of the particle with precision below the diffraction limit (I think this is called FIONA?). You would only need super-resolution if your spots were so crowded that they were all over the place. The place to learn about super-resolution RNA FISH is Lubeck and Cai Nat Methods 2012.
We usually use #1 cover glass, and that seems to work for us. Which is weird, because most microscopes are designed for #1.5 cover glass. Hmm. We have gotten good signals from #1.5 cover glass sometimes, but I also had some mixed results way back in the day, so I usually just tell people to stick with #1. Anyway, that's my experience for whatever it's worth. Further note: we often use the Lab-Tek chambered cover glass for our experiments. We use the Lab-Tek 1 chambers, which have #1 cover glass. There's also Lab-Tek 2 chambers, which are like version 2.0 of the Lab-Tek chambers and are superior in every way, except that they use #1.5 cover glass. So we don't use those.
The signals are typically so dim that you can't see them without the aid of a cooled CCD camera. So you really shouldn't see anything when you look through the eyepiece most of the time. That said, if you have a nice bright probe and a broad optical filter, we have sometimes been able to see the RNA by eye through the eyepiece. Which is a particularly satisfying experience...
I think Biosearch uses VectaShield or something like that, which can preserve the signals for a long time (i.e., weeks or longer). We usually just use glucose oxidase or just 2xSSC for our imaging, and find that the signals can degrade over 48 hours. So usually, we try to image within the first 24 hours after hybridization/washing, and keep the sample at 4C when we're not actively imaging it.
It can work, but we generally recommend NOT doing this, especially if you trying this out for the first time. The reason is that nail polish can introduce a lot of background, especially in the Cy3 channel. What can happen is that the nail polish will wick into the space between the coverglass and the slide and into the sample, causing background. It is possible for it to work, though, and we do use it sometimes in the lab. What we do is make sure to squeeze out as much liquid between the coverslip and the slide as possible (wicking away all excess liquid using a Kimwipe or something) and then sealing the sample down with a thin layer at the edge. This seems to work most of the time, but you really have to be careful to pinch things down as tight as possible beforehand and also avoid imaging near the edges.
If we don't need any anti-fade buffers (i.e., we are using Cy3 and Alexa 594, etc.), then we just mount with 2xSSC. If we are using a dye that bleaches a lot (I'm looking at you, Cy5), we always just use the glucose oxidase solutions described in our Nat Meth 2008 and MIE papers. However, others have used other antifades (including those with DAPI included in them) and gotten good results. I've heard good things about some of these like ProLong Gold, but these can often interact in unpredictable ways with specific dyes. For instance, we have some anecdotal evidence that VectaShield works good with the Quasar 670 dye from Biosearch, but does not help with Cy5 from GE/Amersham. Also, we've seen that VectaShield can sometimes give a strong background fluorescence when used with Quasar 570 and so you may want to avoid it if possible. Anyway, overall, we're very happy with the glucose oxidase solution. It's a bit fussier, but it works well in a variety of situations.
Biosearch recommends doing 4 hours. Certainly, that's enough to see signals in many cases. We usually just do it overnight because the signals are a bit brighter, but it will usually give you qualitatively the same answer. (Overnight hybridization also give you the entire next day for imaging.) I think a couple people I knew did some test in the past and found that the signals didn't get much better after 12 hours, so that's probably the maximum amount of time you need. Stay tuned, though–may have some updated guidance on this soon.
You can adjust it, but I wouldn't advise deviating from the 10% we recommend. We have found that if you increase the formamide concentration, you will lose signal, and so you will end up counting fewer RNA than are actually there. In the end, if you have high background, honestly, tuning the temperature or formamide concentration won't really help you out very much. Most likely your probes are just suboptimal...
Well, if you leave it in the wash buffer forever, you will get issues, but it doesn't seem to matter that much. We do 2x 30 minute washes, but you can do another hour or two without too much trouble.
Well, basically just follow the directions from Biosearch. One thing that we often do is to put the hybridization solution on the cells and then put another cover slip on top. That spreads the hybridization solution out evenly and also helps prevent evaporation. Another thing is that we do our hybridizations at 37C. This can lead to some drying, even with the cover slip, so we often will take a kimwipe, twist into a little knot, put 1mL of water on it and then put that in the dish along with the hybridization. The increased humidity can keep stuff from drying out.
So the first thing is to check out whether the probe itself works. We usually recommend starting with a cell line that you know expresses the gene. Repeat: a cell line that you KNOW expresses the gene. If you see spots, then you know your probe is okay. Otherwise, could be a sample thing, could be an imaging thing, could be that your cells don't actually express the gene you thought, etc. If you don't have such a cell line, then you can always clone the gene and express it. Kind of a pain, but it should work. We STRONGLY recommend testing in a cell line before trying in tissue. Tissue can have a host of other problems, and so it's best to know that your probe works before even getting into it.
Usually, if you see the nice tight discrete spots that are in all the pictures, then your signal is most likely specific. You might, however, see some background. Here are some notes:
The answer is most definitely yes! You can do this in a number of different ways. The easiest way, although it doesn't always work, is to just add the primary antibody in with the hybridization overnight and then add the secondary during the second wash. Doesn't always work, though, because sometimes antibodies are not compatible with the hybridization conditions. In that case, you can either do the RNA FISH first and then do the IF, or do the IF first and then the RNA FISH. I would probably do the former. Just be sure to do the IF relatively quickly, and also try to keep the salt high, which will help keep the oligos on the target RNA during the IF procedure. Also, try not to use serum as a blocking agent. We have found that those serums can have a lot of RNases in them, and will often kill off all your RNA FISH signal.
Answer is a qualified yes. The problem is that the signal is often very dim, so it's hard to get any real dynamic range. For highly abundant RNA (e.g., viral), you can really see it, though. If you're going to try it, be sure to go in with an unstained control as well as a non-specifically stained control (e.g., stain for GFP even though your cells don't have GFP or something like that). The latter control will tell you (sort of) how much signal you get from background.
That's tricky. The question is negative control for what? If you want to show that your signals are specific, we think the strongest experiment is the odds/evens dual labeling that we talked about earlier in the "How do you know you're detect the right RNA?" section. You can of course stain with a GFP probe or a scramble probe or whatever, but that's not really a negative control. Those are just a different set of probes that will stain different background things in the cell. For instance, you might find some bright non-specific nuclear blobs with your GFP "negative control" that don't show up when you use your actual probe. What does that mean? Just means the GFP probes are different, that's all–doesn't tell you anything about what your probes are binding to...
Can I check whether my RNA and a certain protein of interest colocalize using combined IF and RNA FISH?
Well, this is tricky. In some cases, maybe yes. But often times, the protein of interest will just give a relatively uniform staining in the nucleus, which makes it very difficult to say whether the RNA and the protein "colocalize".
We often use Cyclin A2 mRNA, which only shows up in S, G2 and M phase (and a little bit early in G1). We've validated that this actually happens by doing a Click-iT EdU staining at the same time, which stains cells that are in S phase through pulse labeling of newly incorporated DNA nucleotides (see Levesque and Raj, Nat Meth 2013). Here's a pic of Cyclin A2 in foreskin fibroblasts where you can clearly see some cells as positive and some are negative. The positive ones were Click-iT EdU positive.
Umm... working on it.
The issue with miRNA is that they are very short and so you can really only get one oligo on there. That means you lose all the specificity and signal you gain from the tiling approach we use on longer RNA. That said, others here at Penn have used LNA and signal amplification to do this sort of detection. Haven't tried it myself, but the results look really nice.
This is all courtesy of Lenny Teytelman, who spent a long time working through the details. Main thing: you really need to watch out that you have completely removed the cell wall with zymolyase, otherwise you can sometimes get "variability" that's actually just variable permeability. Here's a more detailed protocol. Again, huge thanks to Lenny for figuring this stuff out.
these pics. There are a couple of considerations when doing RNA FISH in tissue, though, especially if you're doing it for the first time:
Also, check out Shalev Itzkovitz's excellent Nature Protocols paper for some other considerations.
this example of LGR5 in a mouse adenoma.
These are usually the easiest. We typically use chambered coverglass from Lab-tek. These are very handy because you can grow your cells directly in the chamber and then do all your fixation, hybridization, washing and even imaging right in the same well. Note that they also have a version 2 of this product, but we have honestly had less success with that product than with the original Lab-tek chambered coverglass. We usually get the two well chambers because you can slip a 18x18mm coverslip on top of you hybridization solution in order to spread out the solution and reduce evaporation, but the other sizes also work just fine.
We recommend using our RNAquant software: http://rajlab.seas.upenn.edu/StarSearch/launch.html
We wrote this software largely with the same algorithms that we use in house for quantifying spots. Hopefully it's pretty self-explanatory.
Well, turns out that computers are pretty bad at that sort of thing. It's often very hard to get the computer to reliably assess where one cell ends and the next begins. Although it may be possible in some situations, it often requires a ton of work to make it robust, and the resulting code is seldom useful in other contexts. So the most efficient method we've found overall is to just circle cells by hand. But if you know of a better way, let us know!
Part of the beauty of having absolute molecule counts is that you don't have to normalize. You get absolute numbers! But you may also want to stain against GAPDH or something at the same time just to make sure your RNA FISH worked (and to make sure the cells aren't dead or something like that). Then again, there are some fundamental questions here about what the right number to report is. Stay tuned...