NAME: Arul Chinnaiyan
POSITION: Professor of pathology and urology; and director of cancer bioinformatics, pathology research informatics, and the Michigan Center for Translational Pathology, University of Michigan Medical School
As one of the first researchers to investigate the role of long non-coding RNA markers in the progression of cancer, the University of Michigan's Arul Chinnaiyan is helping pave the way for future research in this area.
In particular, Chinnaiyan and colleagues have used a number of genomics technologies, including microarrays and RNA-seq, to uncover a set of previously undiscovered prostate-specific lncRNA markers — some of which may be diagnostic or prognostic, and some of which may help elucidate biological pathways in the disease.
In some ways, identifying candidate lncRNAs was the easy part. However, validating these markers can be tedious and time-consuming. In order to speed up this process, Chinnaiyan's lab has been working with the WaferGen Biosystems SmartChip real-time PCR system, running custom lncRNA qPCR panels.
The SmartChip platform consists of the SmartChip cycler, which runs microscale qPCR reactions in 5,184-well consumable chips that either the company pre-loads with target-specific primers or the customer designs using the SmartChip nanodispenser module.
WaferGen originally designed the system for highly parallel gene expression studies, but, after struggling to gain traction with the platform, has recently refocused its efforts to market it for niche applications such as biomarker validation and target enrichment for next-generation sequencing (PCR Insider, 01/10/2013).
In collaboration with Chinnaiyan's lab, WaferGen recently published a white paper describing the SmartChip's utility for validating prostate-specific lncRNAs. Under a previously inked agreement, WaferGen also holds a license to this and other lncRNA-based discoveries from the lab.
This week, Chinnaiyan, who is also a member of WaferGen's scientific advisory board, discussed these initiatives in an interview with PCR Insider. Following is an edited transcript of the conversation.
Can you provide me with a little more context of the broad work you've been doing with long non-coding RNAs?
We began our work using microarray technology to profile cancer progression. I think our initial focus was really prostate cancer progression from benign to clinically localized to metastatic disease – essentially profiling the alterations that occur in mRNA during cancer progression for the purpose of developing novel biomarkers for the disease, as well as potentially nominating therapeutic targets.
In the evolution of technologies, we and others have transitioned into using RNA-seq or transcriptome sequencing-based approaches to get more of an unbiased [view] of the transcriptome relative to microarray technology, where you sort of have to know what you're looking for.
Actually, initially, our work was in identifying gene fusions or rearrangements in about 50 percent of prostate cancers. We essentially wanted to use the technologies to call out gene fusions across different cancer types. As sort of a byproduct of that work, since we are essentially interrogating the transcriptome in a relatively unbiased fashion — although we are poly-A enriching the RNAs that we're interrogating — this allowed us to not only look at transcripts of coding regions of the genome, but also expressed regions that are from the non-coding genome.
That's how we got into lncRNAs in that the RNA-seq data essentially allowed us to look at regions of the genome that don’t encode proteins. That's exactly how we began to pick out the different lncRNAs that appeared to correlate or become elevated with prostate cancer progression.
Had there been prior evidence of these lncRNAs being linked to disease progression of any sort?
It's still early days. We had a paper published in Nature Biotechnology, sort of a broad-based assessment of lncRNAs, in 2011.
We identified maybe 1,000 novel lncRNAs in prostate cancer. But we focused in on a single lncRNA in that particular study called PCAT-1. We actually ended up calling out 121 prostate cancer-specific lncRNAs that we called PCAT-1 all the way to PCAT-121.
So by using RNA-seq technology to profile prostate cancer tissue samples, we identified P-CAT lncRNAs that were associated with prostate cancer progression primarily at the level of expression. We did do some functional characterization of PCAT-1, which appears to be linked biologically to prostate cancer development, but since then we've identified a couple of lncRNAs that may actually be involved in the disease. So I guess there are multiple aspects to these lncRNAs. Some may just be useful as biomarkers of the disease, and involved in disease progression, while a subset of these may be [indicative] of the biology of the disease.
Is your group one of the first to have explored this area, the role of lncRNAs in cancer?
Yes, in the context of cancer, and probably the first to have explored this in prostate cancer. We have been looking at PCAT-1 and a new lncRNA called SCHLAP-1, which we haven't yet published on, that is associated with prostate cancer progression.
What technology have you been using for RNA-seq?
We've been using the Illumina HiSeq platform.
Regarding the validation of these lncRNAs that you've identified, have you been using the WaferGen platform for validation right from the start?
We were using primarily [rapid amplification of cDNA ends] technology to sequence the 5' and 3' ends of the transcripts that we nominated by RNA-seq. But then we would certainly confirm expression of the lncRNAs that we identified by using conventional qRT-PCR. But this of course became quite cumbersome to do on many lncRNAs that were nominated, so that's why the WaferGen platform came in handy. You can essentially do massively parallel qRT-PCR without amplification. It allowed us to basically validate and quantitate hundreds of lncRNAs that were nominated by RNA-seq, and we would get sort of the same sensitivity and dynamic range that you would get with standard qRT-PCR.
It allows us to be cost-effective as well as do many of them in a parallel fashion. And of course the other aspect is that we don't need to do a pre-amplification step.
These were custom panels from WaferGen?
Yes, it was essentially using content that we provided, and they would make probes based on that.
And you have a SmartChip platform in house now, along with the dispenser?
Did you look at other highly parallel gene expression platforms for this validation work, for example a Fluidigm platform?
We looked at other formats, but I think this one was the best. At the scale we were doing, between 100 and several hundred, this seemed to fit this particular use the best. We did look at Fluidigm and we have access to it through our core facilities, but this was the most easily translatable from qRT-PCR assays. It's essentially just doing qRT-PCR but on a microscale. We didn't have to change platforms that markedly.
The white paper also mentions using the SmartChip platform to evaluate PCR primer design. Can you talk a little more about that?
Often times when you are designing PCR primer pairs, there could be a variety of reasons that the PCR won't work. Sometimes people have to include multiple designs, so maybe five pairs or what have you, and that allows you to still measure the transcript even if a couple of the designs are inadequate. That becomes a bit more cumbersome when you're trying to do that by traditional qRT-PCR, because then you have many wells that you have to generate. Here you can pretty easily do this on the 5,000-well WaferGen system. You can do five primer pairs per gene and make sure you really cover expression of that gene. We just try to do a single pair. Often times you might have misdesigned your primers, or for whatever reason your primer design won't work, and you won't have a readout for your gene of interest.
Can you discuss any current limitations of the SmartChip platform for this kind of work?
I think one of the improvements that we've suggested that they've acted upon was giving the users the flexibility of loading content themselves. Previously we'd have to send our designs to WaferGen and they would have to develop it and send you the chips. Now they've enabled us to use the dispensing system to create our own. We don't have to wait for them to custom design it. This has been a significant enabling aspect.
What's next for this lncRNA research, now that you've identified and validated some of these novel lncRNAs?
Initially we will characterize them in a descriptive fashion, then figure out which of the lncRNAs might be useful in a prognostic context to identify aggressive forms of cancer, especially prostate cancer; and which subset of those lncRNAs are actually involved in cancer progression and development, and if they are involved, what are the mechanisms of action — how are they functioning? Are they working and interacting with different protein complexes? Are they functioning as antisense or using multiple different mechanisms? That will be the next major hurdle, and it's certainly early days in the field.