Bench philosophy: Pocket-sized RNA-Seq

Fishing for non-coding RNA
by Steven D. Buckingham, Labtimes 06/2015

More and more roles are found for non-coding (nc)RNAs in fine-tuning the transcriptome. Hence, the race is on to find the full range of what these RNAs are up to. Until now, ncRNAs have proved to be challenging to work with, however, all that is starting to change, thanks to a burgeoning crop of new techniques.

The poor old central dogma has taken a bit of a battering over the past few decades. Gone are the days when we could talk about genes making RNA making proteins with a straight face. Let's face it, RNA itself was simpler in those days, too. The genes stored the information, ribosomes ran the proteins off from it and all the RNA had to do was to shuffle between the two, like a faithful messenger. Ah, such simple times ...

In a way, the trouble began when the first genomes started coming through and workers started noticing just how little of the genome is dedicated to making proteins. If they're not making proteins, then what could all those non-coding stretches be there for, we wondered. It gradually emerged that RNA was not quite the simple messenger we thought it was. The thing is, RNA was doing a bit of serious moonlighting. In fact, the messenger wasn't just relaying the message. It was editing it.

Finding unknown, low abundant non-coding RNAs is challenging. A new RNA-Seq strategy facilitates the search. Photo: Nicolle Rager Fuller, National Science Foundation

The past decade, in particular, has seen a major shift in our appreciation of RNA and its role in shaping the phenotype. We now know of the existence of several classes of non-coding RNA – that is, RNA that does not serve as a template for the production of a protein but plays some other role. These non-coding RNAs (ncRNAs) are highly abundant in all living organisms and play a lot of fundamentally important roles. For instance, there is the well-known transfer RNA, that as well as ribosomal RNA plays a part during translation on ribosomes. But there are others, such as microRNAs (miRNAs) and long non-coding RNAs, that play different roles in editing, splicing or even completely erasing the coding RNAs.

Take miRNAs, for instance. These enigmatic entities are encoded by their own independent transcription units in the garbage in between "real" genes, or even lurking in the introns of protein-coding genes. They are produced in a standard pattern. First, after being transcribed into long precursors (primary miRNAs), they are passed on to a unit in the nucleus called the "microprocessor" assembly. The microprocessor, which consists mostly of the "Drosha" RNAse, chops the precursor into a short hairpin form. This hairpin is then picked up by Exportin 5, which escorts it to the next processing stage, the Dicer assembly in the cytoplasm. Here, the hairpin is cleaved and one strand is discarded. The remaining strand, under the supervision of the Argonaute protein, is matched against passing messenger RNA strands and if there is a close (but imperfect) match, the messenger RNA is degraded.

Given this influence on mRNA, miRNAs have a potentially significant effect on fine-tuning the transcriptome. Not surprisingly, miRNAs have been regarded with suspicion in a number of diseases, ranging from neurodegeneration to cancer, hence biologists are scrambling to find out just what they are doing, and how we can manipulate them to develop new therapies.

Rare and small

But working with miRNAs is not straightforward. Although they seem to get everywhere, they do their work in small numbers. In some cases, a few transcripts are all you find in a single cell. Sure, there have been stunning advances in techniques like Next Generation Sequencing (NGS) and in micro-array technologies, and several new approaches have burst onto the scene, such as bead-based capture methods.

But these have their drawbacks, too. Interpreting NGS data is, to put it mildly, challenging. And the low levels of expression of some miRNAs mean you get low read counts and poor consistency between replicates, the very things that the data pipeline excludes to avoid sequencing errors.

So, just how do people go about monitoring miRNA activity? A common way of detecting a miRNA is to use quantitative PCR (qPCR), for which there is no shortage of protocols, off-the-shelf assays and kits. qPCR is quite sensitive – you can spot up to a couple of hundred miRNAs from a single cell. But there is a down-side: it can be quite error-prone, with false positives being a particular hazard. There are, of course, ways to get around these issues but the fact remains that the trade-off between high sensitivity and accuracy will always be a challenge, when you are trying to detect rare transcripts using this technique.

Another popular way to monitor ­miRNAs in a sample is to use in situ hybridisation but the shortness of miRNAs makes optimising detection quite tricky. Similarly, you can use micro-arrays that rely on hybridisation between the target and a number of nucleic acids, and again you can take your pick from a lot of off-the-shelf miRNA micro-arrays.

Micro-arrays, of course, work on the principle of getting the target miRNA to hybridise to a probe. But there are other techniques that use a similar principle coupling it with exponential amplification to increase the sensitivity. This was an approach taken recently by Xiabo Zhang et al., at the Shaanxi Normal University in Xi'an, China (Chem. Sci., 2015,6, 6213-18). Here, you start with a probe with a double design: it is targeted rationally towards specific miRNAs but is also designed to make it amenable to isothermal exponential amplification (IEA). This probe is stuck onto a streptavidin micro-bead, where it hybridises to the target miRNA. The probe also contains a nicking site and the idea is that once the target miRNA hybridises, it starts extending but gets nicked at this nicking site, releasing the newly synthesised single-stranded DNA. This DNA, in turn, hybridises with another bead-bound probe, causing further extension. Thus, exponential amplification follows and by doing the whole thing using biotinylated dNTPs, you can label the beads with a fluorescent marker.

How to find unknown ncRNA?

But these techniques all suffer from a major drawback: they assume you know the sequence of what you are looking for. Admittedly, if all you want to do is to monitor a known miRNA, you don't need to look any further. But if it is hard enough to detect a known miRNA, what about the task of finding new ones? This problem is particularly acute with miRNAs, because we still don't know all the details of how they are processed, leaving doubt as to what their final sequence will look like. That means, we can't yet confidently predict whether a stretch of genome drives the production of ncRNAs.

To get around this, Florent Hubé and Claire Francastel of the University Paris Diderot and the Centre Nationale de la Recherche Scientifique came up with a novel way of letting the genome do the hunting (Non-Coding RNA, 1, 127-38). "Pocket-sized RNA-Seq", as they call it, uses a genomic region of interest as a bait to find mature ­ncRNAs, which is then amplified up with PCR.

Here is how it works. The basic idea takes advantage of the fact that you can use RNA complementary to your genomic region of interest (the bit of genome you suspect makes some ncRNA) as a bait to catch the ncRNA. The RNA bait is immobilised using biotin/streptavidin beads, and once any target ncRNA has been hooked, you strip it off, ligate it to a couple of adaptors and amplify the lot with PCR.

Hubé and Francastel supply all the details you need to put this cunning plan into action. So how is it done? First off, we need to make the RNA bait. To do that, take a 60-90 long nucleotide sequence covering the bit of genome you think might be encoding an ncRNA. Alternatively, you can clone the bait in a vector under a T7 promoter, amplify it up with PCR or even synthesise it directly. Then transcribe off the RNA in vitro – Hubé and Francastel used the MessageMuter shRNA kit that deploys T7 polymerase. The authors advise that a shorter bait is better than a longer one, as a longer bait may contain structured domains that you'll have a job to denature. Instead, use a set of shorter, overlapping baits.

Great – we have our bait. But before we put it in the freezer ready for tomorrow, we need to dephosphorylate it – Calf Intestinal Alkaline Phosphatase will do. Why do we need to dephosphorylate? You'll see, he says with a glint in his eye. So far, we have the bait, next we need our RNA sample. We will isolate it from our target tissue in the usual way but Hubé and Francastel recommend adding a fractionation step, to enrich it for miRNAs. That cuts down the risk of getting non-specific binding of RNAs to our bait.

The next part is the interesting bit – the whole approach rests on doing a three-phase hybridisation. The first hybridisation step is to anneal the RNA bait at the 3' end to a biotinylated oligonucleotide –this will be the bridge between the bait and the beads. The second hybridisation step is to join the bait + biotinylated oligo to the streptavidin beads (the authors say you can probably cut down on background by annealing the oligo to the beads first, then annealing to the bait). So, what we have ended up with is a string of molecules stuck more or less end-to-end: at one end we have the magnetic streptavidin bead, joined to that we have the biotinylated oligo and at the end is the bait.

We have our rod, we have our bait. Let's go a-fishing . . .The third and final hybridisation step is to anneal the RNA sample to our bead/oligo/bait assembly. Incubate the assembly for one hour at room temperature with the purified RNAs and, hopefully, an ncRNA will bind to the bait. Hubé and Francastel recommend purifying the RNA sample down to small RNAs (<200 nt). Wash a couple of times in RNA binding buffers and we're ready for the next step – recovering the RNA. Recovering the captured RNA is a matter of polyadenylation followed by ligation to an adapter, ready for transcription and PCR. We'll start with the polyadenylation. Purify the RNAs by phenol-chloroform extraction, first, then get out the poly(A)polymerase. The authors recommend a good hour of polyadenylation, so that you end up with some 50-200 A.

The last step is to amplify up our captured RNA using RACE-PCR. And this is where that bit about dephosphorylation comes in. The oligos used for pull-down were biotinylated at their 5' end, which means that (because the bait was dephosphorylated) the only place you can stick on an adaptor is the 5' end of the small RNAs you trapped with the bait. Clever, eh?

Catch me if you can

As proof-of-principle, Hubé and Francastel had a look to see if they could catch a known miRNA using their technique. They picked on hsa-mir-21. According to mirbase (, this miRNA has its own gene within the intronic region of a protein-coding (a vacuolar membrane protein) gene called TMEM49. The handy thing from the point of view of providing a proof of principle is that 1) hsa-mir-21 is under-represented (i.e. rare) and 2) it is only present in cancer cells. Hubé and Francastel isolated RNA from human breast-cancer cells (MCF-7) using three different baits and pulled out 75 clones. Of these, 47 showed a perfect match with the known miRNA in mirbase. When they did the same with myocytes, they got nothing. Sure, the whole thing could have been done using stem-loop quantitative PCR (and indeed they did this too, as a positive control and to provide a comparison) but to do that you need to know the sequence of what you are looking for.

But are they just re-inventing the wheel here? What, exactly, are the benefits of this new technique compared, say, to high-throughput approaches? First, it seems to be sensitive enough to pull out targets that are just too rare to get past the thresholding that has to be done with deep sequencing, in order to eliminate sequencing errors. Second, with this new technique you don't need to know the sequence of what you are looking for, although you do, of course, need to know where in the genome the ­miRNA is encoded. Finally, it doesn't need very much material to work on: Hubé and Francastel believe they can detect ­miRNAs in 0.1 ng of fractionated RNA, which means you can use this technique for rare samples.

Last Changed: 22.11.2015

Information 4

Information 5

Information 6

Information 7

Information 8