Researchers at Germany’s GSF-National Research Center for Environment and Health have developed a new method for detecting protein-protein interactions that expands upon bimolecular fluorescence complementation (BiFC), a method developed by Tom Kerppola of the University of Michigan and originally published in 2002 [see CBA News 09-07-04 for an interview with Kerppola].
In BiFC, proteins are tagged with two halves of yellow fluorescent protein, which join together and fluoresce when the proteins interact. The GSF team’s method, which they call exBiFC for extended BiFC, uses two additional fluorescent proteins — cyan and red — to help visualize the expression of individual proteins in the cell even if they don’t interact.
In a paper describing the method published in the December issue of Biotechniques, the GSF researchers note that the exBiFC assay “represents a major advance, because it allows easy identification of cells expressing both fusion proteins and the normalization of the interaction signal to expression levels of both fusion proteins in these cells.”
This week, Cell-Based Assay News spoke with Horst Wolff, the lead author on the study, about the advantages of the approach over BiFC and the GSF team’s future plans.
Can you discuss your motivation for extending the BiFC assay?
Mainly we’re doing HIV research, but as in all fields that are somehow related to cell biology, you have to investigate protein-protein interactions. We frequently had problems that students who were not so experienced with biochemical or imaging methods had a real problem getting reproducible results, and we wanted to have some comparable results from different students, and assays like FRET, for example, are much too complicated for students. We tried that and never got something that we could compare between two different students.
So we were looking for another method to investigate interactions between HIV proteins and cellular proteins, because these interactions may be useful for novel therapies. It would be good to know if you could block an interaction between an HIV protein and a cellular protein, for example.
We found this classical BiFC [biomolecular fluorescence complementation] assay that was published in 2002 by Tom Kerppola, but we found that this is not reproducible enough because it’s not internally controlled. So if you use this assay, you can’t say whether your proteins are expressed and at which level they are expressed unless they really interact, and then you get a signal. But we frequently experienced that we had many false positive interactions, especially when you have strong overexpression of proteins. Then you nearly always get a signal in the classical assay. And you don’t know that you have very strong overexpression because you can’t see it.
We just put a marker on each of the proteins of interest that you always see — not only when an interaction occurs. And this really makes it very reproducible, very easy even for students to carry out.
This was actually the real motivation — to have an easy and reproducible assay. And we found out that it’s useful for many other applications, really — for screening, for example. Not only to confirm an interaction that you have found with another assay, but also to really to identify new interaction partners.
So our first step was that we could create stable cell lines with one [protein] partner, and then on these stable cell lines, in a 96-well plate, we can put different [protein] partners and see in which well we get a real good signal.
Can you walk me through how the approach works?
You have two expression constructs, and each of the expression constructs has one half of the yellow fluorescent protein. One is the N-terminal half, and the other the C-terminal half. That is the classical BiFC assay from Tom Kerppola.
In addition to these half proteins, we have in one construct a cyan fluorescent protein and in the other construct a monomeric red fluorescent protein. And in front of these two, you can now clone just the cloning sequence for your protein of interest. So you take one cDNA and put it in front of the red and one half of the YFP and you take the other protein of interest and put it in front of the cyan, and the other half of the YFP. And then you just have to get these two constructs into a eukaryotic cell. We only use mammalian cells, but we know that this will also work in yeast and a variety of other cells.
Then the cells start to express these constructs, and … after several hours or several days, you see that cyan and red fluorescence comes up. So the cells show a pattern that should be more or less characteristic of the protein of interest in red, and for the other protein of interest in cyan. And if you wait a little longer, you see nothing when there is no interaction — nothing more, it just stays cyan and red. But when these two proteins of interest interact, then the YFP half proteins come into close proximity, and they form a really yellow fluorescent protein.
And when you have that, you can just do imaging — fluorescence microscopic imaging, taking images of the cyan, of the red, and of the yellow channel. Or you harvest cells and do a flow cytometric analysis and analyze these in three different channels. You could also use a plate reader, for example. We haven’t done that, but in principle, it’s possible.
And then you just [calculate] a ratio of the CFP and RFP signal of the cell and you divide it by the YFP signal, and that ratio gives you a normalized YFP fluorescence. So if you have a very strong red and cyan fluorescence, then you would also expect a very strong YFP signal.
On the other hand, if you have a limited expression of one partner, so if you have a very, very low red expression, for example, then this limits the maximum of YFP that you can get, because you can never get more YFP than you have of one partner.
And by simply having these numbers, this normalization of the expression levels of the red and cyan tag partners, this is really crucial, and this turned out to be the real advantage over the classical BiFC assay because you now know what to expect and what could be an artifact because you simply have high overexpression, or you have no expression of one partner at all.
How does this compare to other protein-protein interaction methods like yeast two-hybrid?
Yeast two-hybrid has one major difference. In yeast two-hybrid you have to target both proteins to the cell nucleus. You have to put a signal in front of the construct that drives everything into the nucleus, and then transcription of a reporter gene is driven.
We don’t need that. I also don’t know of a real yeast two-hybrid assay that’s internally controlled. I think there is a two-luciferase assay that gives at least a rough estimation of transfection efficiency, but not of the levels of the two partners. So in this case, I don’t know of a real three-signal two-hybrid assay.
As far as FRET is concerned, there are several differences. FRET is very transient, so that means that you can also look at rapid changes of interaction. This is an advantage of FRET, but also a disadvantage because you don’t really have a stable signal. You have to look at ratios of images and then the fluctuations. And once an exBiFC complex is formed, it is stable. So the signal stays there and you can image it a day or an hour later. It’s more robust, I would say. But you can’t look at transient fluctuations in interactions with the exBiFC assay.
Are there any particular limitations that you would need to overcome in further developing this?
The major limitation was finding flexible peptide linkers between the protein partners in one construct — between the one half of the YFP and the whole fluorescent protein like red or cyan and the real protein of interest. So there needed to be real flexible peptides for this. And that’s not trivial because if they’re too long, they tend to fragment or get digested by proteases, and if they are too short, they’re not flexible enough.
There are still some limitations. For example, YFP is not so suitable for fluorescence at 37 degrees centigrade, so sometimes you have to do a little cold shock to the cell, so the YFP becomes stronger. And one could use different proteins for that. We’re looking [to see] if it makes sense to invest more time to change YFP to a more modern variant of yellow fluorescent protein. There are constant developments in the field of fluorescent proteins. Nearly every week someone has a new mutant.
You mentioned that this assay might be applicable in a number of areas. Are you looking into any other particular applications yourselves or with partners?
We’ve had several requests for the plasmid constructs, and there are several interesting applications. We’ve had requests from parasite researchers and from tumor researchers. But actually, and the moment we’re trying to find someone who can provide us with a small molecule library — molecules that could be tested to inhibit interactions of HIV proteins.
But right now, we’re not doing this. We’re confirming and trying to find new interaction partners. This is ongoing work and we’ve had some success with it that’s not published yet.
What are your longer-term goals in this area?
You always have more ideas than you have time or manpower, but we’ve thought of screening a real library with this. So you would have one [protein] partner fixed and just screen a cDNA library for the other [protein] partner to really find all potentially interacting proteins with one or two important viral proteins. But that would be a project for two or three PhD students, which we don’t have at the moment.
But in principle it’s a nice idea. We found it to be very robust, so we would expect to have very, very few false positives, and that’s always good to have in [large-scale] screening. You don’t want too many false positives because that really spoils your fun. If you’re trying to confirm all of the false positives, then what [remains] is only one percent of the hits that you got, and that’s boring and tiring.
We have made some advances since we published the paper. The image analysis is now mainly automatic. In the paper we used data generated by manually marking the cells, at least for the microscopy. Now we’re doing that with an automatic setup that really analyzes several hundred cells within minutes.
There are several new [directions] that the technology is going that are really suitable for doing such analysis. That would be a prerequisite for really scaling it up.