Name: Mats Nilsson
Title: Associate professor, Department of Genetics and Pathology, Uppsala University, Sweden; chairman, Q-Linea and Olink Genomics
Background: Nilsson is associate professor and group leader at the Rudbeck Laboratory, Uppsala University. He is also a co-founder of Olink, Q-linea, and ParAllele Bioscience. His research is focused on developing single-molecule and single-cell gene analytic techniques, employing advanced molecular tools combined with microscopy and microfluidics.
The recent arrival of tools for performing digital molecular analysis has raised questions about how these new systems will fit into research markets in which traditional, fluorescence-based microarray technology is still the preferred technology.
Platforms such as Seattle-based NanoString Technologies' nCounter or South San Francisco, Calif.-based Fluidigm's BioMark have been positioned as both alternative and complementary technologies to arrays and second-generation sequencers.
Now, a team of researchers at Uppsala University has added another platform to the mix.
Using "single-molecule arrays," the approach employs a random array format and a decoding scheme for targeted multiplex digital-molecular analyses. As discussed in the Jan. 1 issue of Nucleic Acids Research, multiplex sets of padlock or selector probes are used to create circular DNA molecules upon target recognition that are subsequently amplified through rolling-circle amplification to generate amplified single molecules [Göransson J, et al. A single molecule array for digital targeted molecular analyses. 2009 Jan;37(1):e7.]
A random array is generated by immobilizing all ASMs on a microscopy glass slide. According to the authors, the ASMs are identified and counted through serial hybridizations of small sets of tag probes, according to a combinatorial decoding scheme. A similar method for flow cytometry is currently being commercialized by Olink spinout Q-Linea (see BAN 7/22/2008).
To learn about the new single-molecule array approach and for what applications it could perform most ideally, BioArray News spoke with Q-Linea and Olink Genomics chairman and co-author Mats Nilsson this week. Below is an edited transcript of that interview.
You seem to move between academic roles and working in biotech startups. What's your story?
I did my PhD at Uppsala University. My thesis subject was about padlock probes. I did the first development of that technology which was later commercialized by ParAllele Bioscience in California. I took a two-year postdoc at Leiden University in Holland and returned to Uppsala, started my own research group and from that we spun out Q-Linea and Olink Genomics, which is the daughter company of Olink. Olink is focusing on protein analysis techniques and Olink Genomics is focused on genomic analysis techniques.
I am currently leading a research group of about 10 PhD students and postdocs at Uppsala. I am interested in single-molecule analysis techniques, such as the single-molecule array and the single-molecule cytometer that has been developed at Q-Linea, as well as in situ genotyping techniques, in order to make point detection possible in cells and tissue samples in diagnostics. The third research area is in enrichment techniques for next-generation sequencing instruments. That is what we are commercializing in Olink Genomics. In regards to that method, we are currently working with our first customers in the company. It is not publicly available yet, but we have our first customers.
How is the single-molecule array technology described in this article different from the technology being developed by Q-Linea?
The basic concept is the same in that we can make the digital quantitation of molecules based on counting individual rolling circle amplification products. What makes this different from what we have done before is that, in the past, we have counted molecules in solution. Now, we have mobilized the molecules on solid supports and so it is easier to discriminate more molecules. It is in a sense actually more similar to the new sequencing platforms than the fluidic single-molecule device that was published previously.
[ pagebreak ]
You call your platform a "single-molecule array." Why have you chosen this term?
It's not an array really, but we use this for lack of a better term. It is a completely random position of molecules on a flat surface, but we put identities to the positions, and so one could say it is a bit similar to the Illumina BeadArrays. On Illumina's arrays, the identities are random, so you cannot call it ordered, but the positioning is ordered. On our arrays, it is random position, random identity. I think that Illumina's sequencing substrates are also called "single-molecule arrays" because of the way molecules are immobilized on the substrates.
Could you be more specific about how your approach differs from Illumina's?
The BeadArrays are not digital. On their arrays one can measure molecules and genotype molecules measuring intensities of fluorescence that indicates concentration of molecules. Our array bases the estimate of copy number on the number of events that we are detecting, so it is a digital quantitation strategy. One can expect that we will have less cross activity compared to analog arrays since we are really counting molecules instead of estimating fluorescence intensities and have better quantity precision and in principle a better linear dynamic range.
In the paper, you argue that this technology is superior to existing microarray technology. Why?
I am quite familiar with the limitation of the traditional analog microarrays. If you are doing expression studies, it is difficult to make good measurements of low-abundant transcripts. There will be cross hybridizations that will obscure the result of low-abundant transcripts. That is the rationale for digital approaches like the platforms being developed by NanoString Technologies and ours.
For what purposes was this technology developed?
We didn't develop it with a specific purpose in mind. It was developed to see whether people would pick it up and apply the technique for any application requiring multiplex DNA or RNA quantification with high quantitative resolution.
How does it work?
First, we design padlock probes. These are probes that become circularized when they identify a DNA or RNA molecule, so the ends of the probe molecules end up head-to-tail and can be joined by DNA ligase. After this is performed, we have converted the target molecules into circularized probe molecules that contain sequence tags that identify the molecules. In this single-molecule array format, we have included two positions of tag molecules. That makes it easier to decode more molecules.
We then rolling-circle amplify the probe molecules and, if necessary, we do that in two sequential reactions. We can then immobilize the product on regular glass slides and this becomes a single-molecule array substrate. Then we hybridize one generic oligonucleotide that hybridizes to all products, and then we add a small library of tag oligonucleotide probes, which is also hybridized. We image the molecules, then remove the tags, and finally hybridize the array with second library of oligonucleotides. We then image, wash off, and rehybridize the array with third small oligo library, and then using the hybridization pattern, we identify the rolling-circle products, and then we just count the rolling cycle products.
In the paper, we use selector probes. They are designed to hybridize to ends of specific genomic DNA fragments. If this genomic DNA sequence is present in a sample, then it will become circularized by selector probes. So this is the targeted enrichment strategy we use for second-gen sequencing instruments. But in this way it is a way of making an array of genomic DNA fragments.
To identify the genomic fragments, we hybridize sandwich probes that contain one sequence that hybridizes to the genomic sequence and one that hybridizes to the tag sequence.
For what kinds of applications will this be suitable?
This could be used for DNA copy number estimates for detecting and diagnosing CNVs and mRNA expression. Those are the two main applications for it. We are going to apply it for single-cell mRNA expression measurements as well. That is in development.
Who will be responsible for commercializing this technology?
This is Olink property and we still need to show that it's a useful approach in an application. If we can show good and precise single-cell mRNA expression, then I think it may be time to put it to the market. Right now, it's a little premature, I think.
Is the single-molecule array technology complementary or competitive with next-generation sequencers?
It depends. I think that sequencing instruments can be used for digital quantitation. They will compete with technologies like Nanostring and Fluidigm. So, in that sense, this is competitive with sequencing.
The enrichment strategy with selector probes is something that you really need for applying next-gen sequencing in diagnostics and large-scale studies. When you have thousands and tens of thousands of samples, you cannot sequence whole genomes. You need to enrich. So this is not competitive to sequencing, it needs to be applied together with next-gen sequencing.
If you are only interested in counting events, then sequencing is overkill. But eventually sequencing may become so cheap it may be competitive with the technologies like NanoString's, Fluidigm's, and ours.
Will it compete with traditional array technology?
I think it is competitive to array technology. For questions you can answer with regular hybridization arrays, the arrays will be more cost effective, but for more difficult questions, where perhaps regular hybridization arrays do not give a full answer, I think these digital approaches will be better, but will always be a little more expensive.
Why will they be more expensive?
I think there is more processing involved in generating the data. Arrays can be made cost-effectively in large series and the processing is not very expensive. Also, data output may be simpler and easier to analyze. But, in the end, I think there is a limit in what kind of information can be expected from traditional microarray hybridizations.