This story originally ran on Dec. 16.
By Tony Fong
Name: Ulf Landegren
Position: Assistant department head, genetics and pathology, Uppsala Unversity, 2009 to present; professor of molecular medicine, Uppsala University, 1996 to present
Background: Founder of Olink Bioscience, 2004; Associate professor of medical genetics, Uppsala University, 1993 to 1996
A team of researchers in Sweden claim they have developed a new version of proximity ligation assays that are "clearly superior" to ELISAs.
PLAs were first described in 2002 for the detection of protein molecules via DNA ligation and amplification. In a study published Dec. 8 in the online version of Molecular & Cellular Proteomics, the researchers build on the technology by using "paramagnetic microparticles for robust and highly sensitive protein detection in complex biological material," the authors said.
Dubbed solid phase PLA, the new assays are of particular use for protein biomarker detection in plasma, the team added.
In their study, the researchers used the assays to detect nine proteins in plasma and serum and found "very low limits of detection and broad working dynamic ranges," they reported.
They evaluated the SP-PLA by testing its ability to detect in clinical samples the presence of GDF-15, a cardiac protein biomarker. The assay, the authors said, showed "very good" inter-assay correlation and "agreement with available clinical data as measured in a conventional ELISA."
ProteoMonitor recently spoke with Ulf Landegren, the corresponding author on the study. Landegren is a professor of molecular medicine at Uppsala University and founder of Olink Biosciences, which develops technology for the detection of proteins, DNA, and nucleic acids.
Below is an edited version of the conversation.
Describe the SP-PLA for me.
First of all, [proximity ligation assays] is not an established method for [protein identification]. The method most people are using is quite old. They're built around the sandwich immunoassay, which I think is the state of the art. It was first published in 1967, so it's really ancient.
In 2002, we first described the PLA method … and it's now available from Applied Biosystems in one version.
We use PLA either for measuring things in solution, like in blood or in plasma, or for locating proteins in microscopic specimens, and the company Olink [sells PLAs that pertain] to imaging proteins in microscopic specimens.
But the version that Applied Biosystems is now commercializing and this now new paper … deals with measurement in solution phase. We think the paper here is quite important in several ways.
Unlike other methods, we're able to require three antibodies to recognize the same protein, so only when you get three things that recognize the protein will you get a positive signal, and that means that it will become more specific. The chance of recognizing the wrong protein is much less because of this.
And that becomes a very important problem when you try to detect very low concentration components in blood. And I think many of us believe that there will be tremendous opportunities for diagnostics if you can detect the very dilute proteins present in blood, proteins that might reflect disease processes in some organ that becomes represented in plasma.
The assays become routinely more sensitive than standard assays. In the paper that you've seen, we compare ourselves generally to commercial assays [that] people have spent quite some time developing. Quite readily, we obtain assays that are maybe 100-fold more sensitive or so.
This technology depends on the quality of antibodies that are available?
Yes, that's a very important factor. In this work, we've been quite successful by using a commercial source. We're using anti-sera produced by R&D Systems.
[ pagebreak ]
We need three things that bind the same protein molecule, and the way we do that is by just taking anti-serum raised in rabbit against whole protein, so the R&D people have prepared the protein in immunized rabbits, and when they do that, the anti-serum from the rabbit will contain antibodies that bind many different portions of the protein.
When we take that anti-serum and divide it into three aliquots, we can be pretty sure there will be something, antibodies that bind different portions of the protein in each of the aliquots.
Antibody quality has been identified as one of the limiting factors in antibody-based proteomics. If that's the case, wouldn't it limit the performance of your assay?
Definitely, if you don't have good antibodies, then you can't work. But we found that with this format, it will also depend on the assay — the character of the assay — and here, we have a very high success rate just picking these anti-sera from rabbits and applying them directly in a matter maybe of a day's work or so. … We've been able to set up very sensitive assays that are more sensitive than what the state of the art allows.
So does that depend on specific protocols that you've developed to make sure the PLA works?
Well, yes. The way we make the reagents, there are some protocols for which we've published now. [It also] depends on the way you perform the assay when you mix the reagent with the sample. That is the main purpose of this paper, to describe how you do that.
This assay is more for the validation and verification part of the process, right? It's not really for discovery work?
No, it's mostly for validation. But that is also the most important problem now in the industry. The complaint is that there are so many leads, so many promising markers that people find by doing mass spectrometry or by doing … mRNA expression analysis — things that may suggest new markers.
But there's a tremendous roadblock at the level of validation.
You compared this to ELISAs in your paper, but there are also mass spec-based techniques for validating biomarkers. Have you compared your assay against those methods?
No, we haven't, but I think in general, mass spec assays that I'm aware of tend not to be quite [as] sensitive [as] ELISAs. I'm sure that will improve, but in general they have a hard time competing with ELISAs currently, but that is an area that is under rapid development.
Is there any way to quantify the false-positive rate and false-negative rate for this assay?
We don't have that data, but one thing I can say is we are now preparing a next paper … where we're taking the assay that we published in this paper and we now apply it in a multiplex fashion because it turns out that it's very easy to do this in multiplex. You can combine many assays together.
As you may have noticed, we used very little samples, we used 5 microliters serum sample per assay. … In this paper we investigated one protein, but what we're doing now in the lab is to investigate several 10's of proteins in the same 5 microliters.
The paper says you set up the assay to detect 46 proteins. Have you increased that number since you submitted the study for review?
The number 46 is quite recent. … So I'm not sure we've taken that any further, but this is work that is steadily in progress. We add new antigens all the time.
I think the ones we've selected now were selected on the basis of being relevant for, for instance, inflammatory conditions and [are] potentially of interest in malignancy and also cardiovascular disease.
In the paper, we have specifically identified one protein, GDF-15, which is a good marker for cardiovascular disease. That happens to be a relatively abundant protein, so it's not so demanding to detect that.
[ pagebreak ]
In other work that we are now currently preparing for publication, we use this method to develop assays against particularly promising markers that may be diagnostic, so we have what we think is a very promising marker for prostate cancer, and we have one that may be of importance in Type 1 diabetes, and also in Alzheimer's … we show that we can achieve sensitivity that is far greater than what has been [previously] possible.
And that is work that was recently submitted using the same protocol.
So I think where this work is going is we are now in progress using this to validate promising markers, and we're also setting up multiplex markers to be able to screen in biobank samples large numbers of markers in parallel.
The work that you've done has looked almost exclusively at serum and blood. Have you looked at using the assay in other fluids?
In the case of the Alzheimer's marker, that work was done in cerebrospinal fluid and in lysates from mouse brain. We've never tried urine [but] I would be quite optimistic that that can be done, but we haven't had any occasion to do that yet.
Are there any specific bottlenecks in urine that don't exist in serum that may affect the performance of your assay, and how would you go about solving that?
I'm relatively naïve about this because I have no experience, but I think we would attack it with the protocol that we have now and see how that goes.
One thing about urine is that, as we all know, it varies widely in concentration. In some cases you have very dilute urine, and in other cases, it's very concentrated, and it does contain things that might potentially interfere with the reaction, but unlike the other forms of PLA where there is no washing involved, we trap molecules on the support and then wash.
And that means that components in urine that might interfere with [the performance] of the assay can be removed before you go on.
Your paper mentions increasing the plexing capacity of this assay. Do you have any data you can share?
Not much. What I can say is that the sensitivity seems to be the same, so if you do the assay in single plex, or if you do the 36 that we've mostly done, the level of detection seems to be similar.
In the context of multiplexing we've done some cross-reactivity checks. So if we have the full 36-plex assay and we add the antigens one by one, we can observe some minor cross-reactivity. It's much less minor that it would be in an ELISA because there cross-activity becomes an important problem the more you multiplex.
But here we only allow the right combinations of antibodies to give rise to a signal. In an ELISA any combination that would bind an antigen would light up as a signal, but here we only permit the right combinations, and therefore the cross-activity problem is much less severe.
So we would expect to be able to multiplex to a much higher degree.
Is that how you would address the cross-reactivity issue, by finding the right combination of antibodies?
Not by finding the right combination of antibodies, but by designing the DNA part of the assays so that only the right pairs can ligate or only the right pair can amplify.
In addition to increasing the plexing capability, what else are you doing to build out these assays?
I think a very important aim is still to push sensitivity even further. The demand for sensitivity is very great. I should also say that one weakness of the assay, perhaps, is that our variability is a little greater than ELISAs.
While our sensitivity is much better, the precision of our measurements is not quite as good as an ELISA, and that's something that we're targeting now.
[ pagebreak ]
How are you addressing that? What sorts of things are you doing to decrease that variability?
This is usual lab work. We're trying to identify the level at which most variability occurs. And one of them, I'm sure, is the real-time PCR. We use PCR as the read-out, and PCR does have its own inherent variability. Even if you have a pure DNA molecule, you will not get the perfect CVs.
We are also looking at a number of other read-outs.
You've done some work with clinical samples. Are you increasing your work with those, and if so, what kind of data can you share?
Nothing much that's ready for publication, but as I mentioned this pool of multiplex assays is designed mainly [to focus] on inflammatory markers that might be relevant both in malignancy and inflammatory diseases, and also some of them in cardiovascular diseases.
So cardiovascular diseases and colorectal cancer and neurological diseases are our main applications.
How does this assay compare to the in situ PLA [Duolink PLA, commercialized by Olink in 2007] that you developed a few years ago?
In a sense, it's the same. We call both forms PLA, one is called in situ PLA. There are many variants, but for the solution phase, generally, we attach oligonucleotides to the antibody and in solution phase, those oligonucleotides are joined by ligation and that is what you amplify for detection.
For in situ, again, oligonucleotides are attached to the antibodies, but here, they are not joined by ligation but rather they are used to template [the] joining of two other oligonucleotides to form a DNA circle.
Can you use these assays to study protein-protein interactions, or are they used solely for biomarker detection work?
No, that is one of the features and … in prostate cancer, the marker that we are after are a complex composed of many proteins and the assay is designed to ask for the co-locations of all these markers, which is also what we do for the in situ [PLA]. We're very interested in interactions.
I think that's another class of biomarkers that have not been accessible to analysis before, so we open up opportunities to look for things that people [have not been] able to measure.
Would a diagnostic form of this assay be a PCR-based test?
That could be … I think PCR is very useful for amplifying the signal, but it's not necessarily the only way to record the signal … so we're looking into a number of other methods.
I think the general trend in biology is that many problems can be solved by [new genomic sequencing technologies]. It's getting so cheap and that gives you a digital count of detection, so that's certainly one read-out that we're interested in — recording how many times we saw a protein by the DNA tag that forms.