We need to find biomarkers for prognostic, diagnostic and personalised treatment development. Notably to fight cancers that affect tissues. Since biopsies are invasive, it’s better to look for biomarkers in body fluids. Indeed, a simple blood sample becomes a kind of ‘liquid biopsy’ to reveal tissues affections. For 13 years, increasing interest has been shown for miRNA as biomarkers and it will last for sure. The 2 main reasons are that they are major regulators of cell processes and they are released from tissues into the blood. They are major biomarker candidates in serum and plasma. Thus, these circulating miRNA (cmiRNA) are the best hope for modern medicine. Still, a lot of research has to be done to determine the specific signature for each pathology, and also depending on the patient background. Obviously, cmiRNA profiling is a key step and requires sensitive and reproducible method. Sequencing, qRT-PCR, several kind of microarrays… Let’s explore together what the best approach could be.
Next Generation Sequencing to discover unknown miRNA
The most trendy method to find miRNA biomarkers is the sequencing of the small RNA transcriptome (small-RNA seq). It deserves genuine interest since it can cover the whole miRnome (miRNA transcriptome) bringing qualitative and quantitative data. It has also unique advantage over the other methods: it can reveal unknown miRNA. So to discover miRNA that are not yet in database, it is undoubtedly the most dedicated choice.
The deeper the sequencing, the better the quality of results will be and thus the analysis. So, it matters a lot to have maximum reads for miRNA sequences. I don’t mean to mention an extensive and complete list of the challenges and the solutions, nor the pros and cons of this method. Nevertheless, I think that the challenge of removing the dimer of adaptor must be withdrawn because it concerns the quality of the reads.
When preparing a small-RNA seq library, adaptors in 5′ and 3′ ends must be added to each small RNA of the transcriptome. Unfortunately, adaptors form dimers that take quite a lot of reads (Figure 1, dark grey). Now, there is a solution called CleanTag. It is a kit to ensure ligation of the adaptors reducing greatly the dimers formation.
Previously, gel purification has been used but it is a handling option that is time consuming (Figure 2). And since the dimer size is quite close to the di-tagged small RNA, gel purification is not so easy and some contamination remains. Now thanks to CleanTag, beads-based purification is far enough and compatible with automation. It saves time and improves the efficiency to remove adaptor dimers.
In conclusion, NGS is a top solution for profiling, as far as adaptor dimers are well removed and CleanTag has come to help with this. Still, it is expensive and requires several months mainly due to the difficulty of data analysis. That is why I would recommend it only if it is required to discover completely new miRNA.
The next generation microarray for top profiling
Microarray-based profiling is certainly the most affordable method to cover the whole miRnome. Since microarrays are regularly updated, they also benefit of discoveries from NGS. And actually the Human miRNA meant already 28645 entries (release 21) in June 2014, that is quite generous to screen and find a reliable signature. The microarrays analysis is less difficult than NGS analysis. Would this make it the top method?
There are numerous microarrays and it is said that they all have their advantages and inconveniences. However, despite clear efforts, they all still have low sensitivity, which is a general problem to detect cmiRNA. Furthermore, some use pre-amplification steps that can lead to bias. The specificity, the accuracy and the reproducibility are also quite limited for all the microarrays. So finally, the situation is simple : microarrays are cheaper, but offer quite poor performances to face the cmiRNA profiling challenge. Except, the new generation of microarray, called 3D-gene (Figure 3, extract from Ono et al J. Clin. Med. 2015)!
More sensitive, more accurate, more reproducible, more specific than any other microarrays and even better than the sequencing itself, the 3D-gene is a microarray-based method that renews the genre. To know more about the principle, I invite you to read my previous post. It is the closest method in performance to RT-qPCR.
RT-qPCR for the validation of the signature
RT-qPCR is the most sensitive and reproducible method. Unfortunately, to cover the whole miRnome it is very laborious and expensive (at least 7x 384-well plates for 1 sample). Thus, it is commonly used to validate a signature that is found from a previous whole miRnome profiling. The major point for RT-qPCR is the choice of the reference ‘genes’. For miRNA analysis, these references could be cmiRNA and other small RNA. Reliability, and also standardisation, rely on their quality. So how to choose the references? As for messenger RNA analysis, there are no real universal references. They depend notably on the tissues, the sampling methods (and storages) and the treatment side effects. We can say references are specific to the project. Let’s come back to the basis: good references are those that do not vary significantly through the different samples and the controls of the study. So why not use the whole miRnome profiling data to define them? As far as the profiling is accurate, sensitive and reproducible, it can be used not only to find the biomarker signature but the ‘references’ required for the RT-qPCR.
The input in question
I have noticed different and contradictory information about the input required. For example, Exiqon miRCURY LNA microRNA arrays are often mentioned in articles with input around 40-30ng of total RNA. Though the application note of Exiqon recommends 500ng of total RNA and their manual mentions 250-1000ng per labelling reaction. So please find below a recap table (Figure 4) with input of total RNA required according the respective manuals and technical notes.
The quantity mentioned for µParaflo is required for the service platform, which is why it is so high (but I can’t find any more informative documentation). The PCR arrays miScript can work with only 10pg per well, thus about 30 ng for the 7x 384-well plate are required to cover the miRnome. So we can keep in mind that a minimum of 100ng of total RNA is required for a whole miRnome profiling (without pre-amplification and without RT-qPCR). Thanks to our experience with 3D-gene, we estimate that 150µl of plasma is enough for a complete profiling. Please keep in mind that the quality of the output, the cost, and the difficulty of the analysis are not the same for all these methods.
NanoString is not mentioned in the table since it covers a maximum of 800 points and not the complete miRnome (the input is also 100ng by the way). I think it would be fairer to compare it to TaqMan array MicroRNA cards (740 points) or any Pathway PCR arrays than to microarrays, NGS and even full miRnome PCR arrays. It is a very different approach that would make sense with projects for which researchers can reduce the set of miRNA to a couple of pathways. And for those projects, the number of targets is the key question for deciding on the method, since direct RT-qPCR may be possible. It’s another topic…
My recommendation for full and fair profiling
You can find several articles (Mestdagh et al Nature Methods 2014 notably) that review the methods for miRNA profiling. As far as I know, none take any clear position, as it is not possible to take a decision among so many options. Thus, my own input here is to share with you what I have noted and to reach a simple and useful conclusion to this complicated story!
For whole miRnome profiling, 3D-gene microarray technology can reveal variations and constants through samples. And so it brings biomarker signatures and a set of references. And no other profiling method can do it with so much sensitivity, reproducibility and accuracy, except NGS. But they are far more expensive and complicated. Then RT-qPCR validation with the identified references can normalise and confirm the signature and could also help with standardisation. That seems a perfect method possible from 150µl of plasma per sample.
You might think my conclusion that the top choice is 3D-gene microarray profiling combined with qRT-PCR validation, is because tebu-bio is a European platform for both of these technologies. In fact, we are indeed a service provider of this technique combo, purposely because it is the top choice, and it’s our mission to bring researchers the best.