Assistant professor of urology
Brigham and Women's Hospital, Harvard Medical School
At A Glance
Name: Brian Liu
Position: Assistant professor of urology, Brigham and Women's Hospital, Harvard Medical School, since 1999.
Background: Assistant professor, Department of Urology, Mount Sinai School of Medicine, 1990-1999.
Cancer biology research fellow, Jonsson Comprehensive Cancer Center, University of California Los Angeles School of Medicine, 1985-1988.
Postdoc, John Fahey's lab, Department of Microbiology and Immunology, UCLA School of Medicine, 1985-1988.
PhD in microbiology/molecular biology, UCLA, 1980.
At this week's PEPTALK conference, hosted by Cambridge Healthtech Institute in San Diego, Brian Liu spoke about his development of a "reverse-capture" microarray technology that can be used as a new way for discovering disease biomarkers. ProteoMonitor caught up with Liu before his talk to find out more about his new technology, and about his background.
How did you get into using proteomics to try to find diagnostic biomarkers for prostate cancer?
I got into prostate cancer because I've always worked in urology and urologic oncology. Bladder cancer was actually my first [area of] study, and then I moved into prostate cancer when I moved up to Boston. Proteomics has always been an interest of mine because I use it to look at disease biomarkers, and there's a great need, for any type of cancers, to find better biomarkers for personalized medicine.
What kind of proteomics do you do to discover biomarkers?
We did both SELDI-based proteomics, and also 2D gel-based proteomics for serum biomarkers. The difficulty of using SELDI and 2D gels is that there is a large dynamic range of proteins in serum, and we felt that unless we focus specifically on how to identify specific targets, you're basically trying to search for a needle in a haystack.
We decided to develop a technology that's based on autoantibody profiling, because one of the major components of serum happens to be antibodies. For patients, especially those with cancer and inflammatory diseases, the immune system will recognize the antigens that are presented within the cancer cell. So the antibodies act as the amplification, or enlargement of signal, so that we can identify the appropriate antigen, despite the fact that the antigen in the serum may be at a very low level.
The way the technology works — and we do have a paper in press — is we're trying to capture the antigens on a chip using an antibody-based array. Then we take a patient's serum and isolate the antibody from the patient. Then we basically tag the antibody from the patient with different fluorescent dyes. So you can have a cancer patient that's labeled in red or green, and control labeled in the other color. So if there's antibody from the patient that recognizes the specific antigen that's now captured on the array, you will have a differential fluorescence between the cancer and the control.
How this is different from the other antigen arrays that people print is that for most approaches where people are doing autoantibody reactions, people have been spotting antigens that are derived from recombinant proteins or synthetic peptides. The limitation of that type of a study is that those recombinant proteins or peptides don't always contain the appropriate post-translational modifications, like the sugars and the phosphorylated form, and so forth. So antibodies that may be recognizing the sugar portion of a glycoprotein may not be recognized and captured appropriately.
Our approach to immobilizing the antigen is based on a native configuration of the antigen.
Do you make the arrays yourself?
The arrays are from Clontech. What we use is a Clontech 500 antibody array, which is 500 monoclonal antibodies that have been spotted onto a glass surface. These monoclonal antibodies are very specific. They've been quality controlled to recognize a single band on a Western blot. They're really high affinity, highly specific monoclonal antibodies to a variety of different antigens.
We use that as a basis to capture or immobilize the corresponding antibody on the slide. It's a sandwich assay, essentially. So on the bottom is the antibody. You add to it an extract that contains all the native forms of the different antigens. Those antigens will now bind to the monoclonal antibody that has been spotted onto the slide. Now you basically use the monoclonal antibody on the bottom to capture or immobilize all the different 500 antigens.
Then you take the patient's antibody, which you've labeled green or red, depending on if they're control or patients, and you add it to that mix. And you look at the difference in fluorescence.
How does this technique compare to using SELDI or other proteomic technologies?
You get specific interactions, so you know exactly which antigens are giving an immune response to a patient. That's number one. So out of the 500 antigens, we have a candidate list of about 40 different autoantibody reactions to it. So we can identify about 40 specific antigens that patients recognize.
You cannot do that on SELDI, because there's no way of capturing these antigens. With this technology, you have the actual protein identification there, because it's an array form.
The limitation currently is that we only have 500 antigens that are spotted on this array. But conceptually, we can expand the array to 1,600 or whatever, as you print more. We haven't done that yet.
So the limitation is that if the monoclonal antibody is not spotted onto the glass slide, then you're capturing the corresponding antigen to test for the autoantibody.
Currently, this is a proof of concept, not a final product.
What proteins have you found using this method?
We found a bunch of proteins that allow us to look at signal transduction, transcription factors, some membrane type proteins. These are all known proteins, but they had not been previously identified as being associated with prostate cancer.
Some of the newer proteins we found we validated. Some of the other proteins have been reported by others. We found, for example, C-myc. C-myc has been associated with prostate cancer, and that's been reported by others.
Are you still using other proteomic techniques, besides the antibody array?
We are using the other techniques to develop what we call a global immunome profile. We are only spotting 500 antibodies onto this array, and this array is not specific for prostate cancer. But we could pull down antigens from patient serum using the patient's autoantibody to identify what other targets are present. Once we identify these new tumor antigens, then we will find out whether there are good monoclonal antibodies to them. If there are, we will be able to spot it, and customize the array.
Are you looking at other diseases aside from prostate cancer?
We did prostate cancer as a proof of principle, but we're also beginning to look at other stuff now. There has been interest from a lot of other people asking us to look at inflammatory responses, arthritis, diabetes, all the way to other types of cancers.
No fractionation is needed with this method, right?
That is correct. You've got to purify the antibody from the patient serum, so there is that one step that you have to do, but actually that's very quick, and you don't require a lot of sample volume. We only take about 10 microliters of serum to isolate the IgG for labeling, and you can do that in a spin column that you can purchase.
Are there other people using this technique, besides your research group?
No, I think we're the first ones as far as I know that tried to put it together. We developed this technique somewhere around May of last year. We have the first paper in press now that talks about prostate cancer.
How many potential biomarkers for prostate cancer have you identified using this method?
We have about 48 candidate biomarkers. Some of them were known. About 30 of them are novel.
Some of the markers that have been reported on include the myc oncogene, protein kinase C isoforms, p53, B cell2-like proteins, and a couple of transcription factors.
Do you see this as a discovery technique for biomarkers?
Yes, this is a discovery technique. We can adapt it to a clinical practice very easily because you have the monoclonal antibody to immobilize the protein, so you can develop a platform so that it becomes sort of like an ELISA platform.
How would you go about validating the candidate biomarkers you have discovered?
We actually did validate some of the biomarkers. We did it by taking autoantibodies from patients, and using those antibodies to immunoprecipitate a reference cell extract, and then doing a Western blot using the monoclonal antibody that was used on the array to test whether or not the patient's autoantibody was able to precipitate that particular protein.
We wouldn't do that for lots of samples. Once we've checked a couple of antigens, we would just use an ELISA assay.
How many patients have you used for this study?
We've done a total of ten patients and ten benign disease controls.
Do you have any plans to commercialize this technique?
We do. We're in discussions with some companies. Patents have been filed.
What limitations are there to this approach?
I think this is an approach where we take advantage of the immune system to essentially amplify the appropriate signals so we can look at low-abundance proteins and really know what are some of the factors or mechanisms that are contributing to disease pathobiology.
There really aren't a lot of limitations that I can think of. The only limitations that are real are that there are some diseases that the immune system may not recognize. Cancer I think is clearly something that you can think of as immunogenic, and so are certain other diseases, such as inflammatory disease, or diabetes, or arthritis, but you may not have an immune response, to let's say, for example, heart attack or something.
We're happy to discuss any interest for collaborations.