Name: Peter Nilsson
Title: Platform Director, Affinity Proteomics, SciLifeLab, Stockholm
Background: 2011-present, professor, proteomics, KTH Royal Institute of Technology, Stockholm, Sweden; head of protein microarray module, Human Protein Atlas project; 2002-present, group leader, protein microarray technologies, KTH; 2000-2008, head of microarray center, KTH; 1999-2008, DNA microarray technology group leader, KTH
Education: 1999 — PhD, biotechnology, KTH; 1993 —MSc, biotechnology, KTH
Peter Nilsson has many protein-profiling tools at his disposal. As platform director of affinity proteomics at Sweden's SciLifeLab, he relies on Arrayjet's Marathon arrayer to produce antigen and reverse-phase arrays, and has four Luminex instruments — two FlexMap 3Ds, an LX 200, and a MagPix — to profile antibodies using suspension bead arrays. More recently, the lab has been collaborating with Roche NimbleGen and a Danish firm called Schafer-N to access the firms' high-density peptide arrays, which Nilsson believes could replace conventional spotted protein arrays in many instances (BAN 9/18/2012).
Nilsson's lab also has exclusive access to antibodies and antigens from the Human Protein Atlas, a project funded by the Knut and Alice Wallenberg Foundation to support systematic exploration of the human proteome using antibody-based proteomics; and has validated 35,000 antibodies from the project using its in-house printed arrays since 2005.
All of this takes place at the lab's new location within SciLifeLab's Stockholm site. Nilsson moved his lab there from KTH, the Royal Institute of Technology, two years ago, when SciLifeLab was established. KTH is one of four Swedish universities and institutions taking part in SciLifeLab, a translational medicine and molecaulr bioscience collaborative. The others are Uppsala University, Karolinska Universitet, and Stockholm University. The Swedish government recently earmarked $100 million to fund SciLifeLab's's activities, including its DNA and protein array platforms, for the next four years (BAN 9/11/2012).
While Nilsson claims that his lab is more focused on technology development than biology, he acknowledges that its two main efforts, biomarker discovery and autoimmunity profiling, have led to more specific projects related to cancer, cardiovascular, autoimmune, and neurodegenerative diseases. To get a better idea of how a state-of-the-art protein array facility functions, BioArray News recently interviewed Nilsson in his lab in Stockholm. Below is an edited transcript of that interview.
I saw that you have the Arrayjet Marathon microarrayer in your lab. What are you using it for these days?
For us it is a work horse, it supports all the protein array fabrication that we do. We generate antigen microarrays, mainly. The antigens are obtained from the Human Protein Atlas project. They are the antigens that are used to generate all of the antibodies in the project. It is a really high-throughput effort. We generate antigens, recombinant antigens, we immunize rabbits, we get polyclonal serum back, and we purify that on the [rabbit's] own antigens. The next step is the first in a long row of validations is to use the antigen arrays to see if they are binding to what they should bind. So that is the first specificity analysis. We have a format where we have 384 different antigens in 24 different subarrays [printed using the Arrayjet Marathon], where the arrays are all identical, so that means that we can analyze 24 antibodies per slide and each antibody is validated on 384 antigens. We have validated 35,000 antibodies on these arrays since 2005. So it is really a routine assay for us. We didn't use the Arrayjet for all of them though. We bought that system two years ago.
What were you using before?
The Nanodropper from GeSim. We had been using that since 2005 and we looked at Arrayjet at that time, but they were too young then, they weren't really out there. So that is the main assay we are running. We make a new batch very second week, with 384 new antigens, because they correspond to the next set of antibodies to be purified and validated, since we have a flow of somewhere around 160 new antibodies per week coming in from the project. We also use these antigen arrays as a unique resource to screen for autoimmunity targets, as a source of human proteins, basically. We take samples from plasma, serum, cerebrospinal fluid, for example, and we put these samples on the antigen arrays and detect if any of the samples have antibodies directed at these proteins. That is the initial screening to find new autoimmune targets, basically where there are human antibodies recognizing human proteins. We started that project two years back and are expanding, getting more and more. For example, we have taken hundreds of samples of multiple sclerosis and profiled them on 30 batches, meaning 11,000 antigens. And then we identified interesting targets that we are now validating in another platform, using Luminex. Another application on the planar arrays is that we are using reverse-phase arrays, meaning that we print serum or plasma samples or CSF on them. There are quite many using reverse-phase arrays for cell and tissue lysates, but there are not many researchers at all who are using them for plasma or serum. We have done that for years, and that is a way for us to get thousands of samples on one slide, but then you can only detect one analyte, or two.
This is the same assay that was developed by Lance Liotta and Emanuel Petricoin at George Mason University.
Yes, but they have only worked on cell and tissue lysates. This is serum. It is the same thing, but a different type of sample. We published in 2005 the first reverse-phase serum array application, and in that case we did screening for IgA deficiency in thousands of samples. We need to have rather highly abundant proteins in that set up, because if you put in 200 picoliters of serum sample, and you start to count molecules, very soon you get down to very low levels. Those are the two formats of planar arrays that we use: the antibody arrays for antibody specificity and immunity profiling, and the reverse-phase arrays for screening of, for example, IgA deficiency, but also as a validation tool where we have done plasma profiling as part of biomarker discovery. And then we have the suspension bead arrays.
You are referring to Luminex's xMAP technology?
We prefer to call them suspension bead arrays. We see them as arrays. We immobilize antibodies on the beads. We have developed a format with a rather unique quality, which is direct labeling of samples, as we have biotin on all of the proteins, and then we have selected antibodies, taking up to 384 antibodies, they have up to 500 identities on the bead colors these days. We use the format of 384, so that means we can bind antibodies to 384 beads and keep them together as one bead pool. Then we use that on 384 labeled samples. So with one assay, you can profile 384 antibodies on 384 plasma samples in one go. These numbers are really attractive because you are able to get up to the hundreds in both parameters. In other array formats we can have thousands in one dimension, but only one, two, three in the other dimension. So that is what we see as the advantage of that context.
In the biomarker discovery program, is it simply mass screening or is there a particular disease focus?
The situation we have is very unique. No one has access to these antibodies outside of the Human Protein Atlas project. And we also have a unique array set up. When you combine that, it means that there are many people who would like to interact with us. It is hard to keep up with all the requests. We have learned the hard way that we should be rather picky when it comes to collaborators. They must have access to a very good collection of samples. We saw in the early days that we started a collaboration with a few hundred samples, but when we wanted to have more, it was difficult. The criteria are that there should be a verification cohort available when we start. The disease focus is a bit scattered. There are three areas where we focus. One is, of course, cancer, within the Human Protein Atlas, and that is where we connect with plasma profiling. What can we take from tissues to prepare plasma and vice versa? The next main area is cardiovascular diseases, where we now have rather large funding from the Swedish Agency for Innovative Research together with the Karolinska Hospital, and where we run all of these 10,000 antibodies. The third focus is on neurodegenerative diseases. But we are not biologists. We develop technology and we use antibodies. On the other hand, we depend on collaborators, and we are getting more focused.
How are you making all of the data you generate available to the public?
First of all, the Human Protein Atlas is publicly available, but there is no plasma data there. One reason is that we are in a phase where we are really trying to explore the potential of what can be found if we explore these 10,000 antibodies, though we have not reached this stage yet. What is also very much the case is that the individual variation is in many cases enormous. In order for us to publish, let's say, a plasma atlas or profiles, it will have to wait until we have generated much larger sets of data, because it will not make any sense to publish them with limited numbers. But our main focus is biomarker discovery, and of course we are trying to publish as much as we can. That is based on each case and project. We have two markers that we have identified and published and filed patent applications for. One is in the context of kidney disorders, and one is in the context of prostate cancer. But we also have a lot in the pipeline. It is easy to discover biomarker candidates, [but] it is difficult to verify them, because diseases are heterogeneous [and] you need larger sample sets, you need very well characterized and stratified sample cohorts. [Our] strategy [is] doing discovery doing single-binder assays. Then going to dual-binder assays — meaning two antibodies — and then it is up to the biology to determine if they are really markers in these specific instances. So the data is available in the form of ordinary publications but not in the Atlas. We are discussing how that could be done in the future. It might be … that we define 500 plasma proteins that we really know … are detectable [and that] we have good antibodies for … Maybe combine all of our data from these antibodies to generate a profile.
Protein arrays have never become, in terms of a market, as lucrative as gene expression or genotyping arrays. Going forward is it still going to be an important research tool?
Absolutely. I am totally convinced. For example, the Invitrogen ProtoArrays are more or less the only commercial purified protein arrays. We here at SciLifeLab have access to more than 35,000 different antigens. We will make larger arrays, getting up to 10,000 or 15,000 antigens on one array, and use that in certain cases … We have spent some time together with Roche NimbleGen to work on high-density peptide arrays, which is really something that can totally change the protein array field. Before Affymetrix was called Affymetrix, they were called Affymax, and they were trying to generate peptide arrays in the early 1990s, which they never succeeded with, so they turned to oligos. Now Roche NimbleGen are able to make arrays with whole-proteome coverage of over 2 million peptides, and we have used that for epitope mapping of antibodies and also for autoimmunity profiling. It is still in an early phase and they have not been commercialized yet. But I think that is something that really has potential and the data is amazing.
We have another publication through an EU project, where a small Danish company, called Schafer-N, produced the peptide arrays. There are some German companies making peptide arrays, such as JPT and PepPerPrint, but that's totally different scale in terms of density. The arrays produced by NimbleGen and Shafer-N are in the range of millions of peptides. I think they could potentially replace spotted protein arrays. But each peptide has its own configuration, it's very flexible, each amino acid may enable different combinations, may enable different types of binding while the same thing in a protein context might be different. So I still think we will have a need for protein arrays as a verification step. We are actually planning on [using] the NimbleGen arrays … as a national services provider here at SciLifeLab. So when we are looking to the future of how to provide service in biomarker discovery using antibodies on beads, and autoimmunity profiling on arrays, then the high-density peptide arrays would be the perfect starting point.