At A Glance:
Name: Daniel Chan
Position: Professor of pathology, oncology, urology, and radiology, Johns Hopkins University School of Medicine, since 1977.
Director, Biomarker Discovery Center, Johns Hopkins University School of Medicine, since 2000.
Director, Tumor Markers Research Laboratory and Analytical Clinical Chemistry Laboratory, Johns Hopkins Singapore, Singapore, since 1999.
Chair, Specimen Collection & Specimen Handling Subcommittee, HUPO.
Background: Post-doc in clinical chemistry, Erie County Laboratories, Meyer Hospital, Buffalo, NY, 1976-77.
PhD in biochemistry, State University of New York at Buffalo, 1976.
BA in biology, University of Oregon, 1972.
How did you establish the biomarker center in Singapore?
Johns Hopkins has a facility called Johns Hopkins Singapore. That facility was funded by the Singapore government. So what I proposed to do was start a proteomics laboratory in Singapore. Because the Singapore government provides all the funding — in a university setting like Johns Hopkins, if we want to do research, the university does not provide you with funds. And when proteomics was new, no one gave you money to do it. So I thought this was an opportunity to do something new.
We started a proteomics lab there [in 1999]. We took approaches like 2D gels, and we bought Applied Biosystems’ MALDI-TOF. We were looking for biomarkers that might give us some unique perspective for the type of cancer that was more prevalent in Singapore and other Asian countries — for example liver and gastric cancer, and also breast cancer. For some reason, in the last 10 to 15 years, there seems to be a lot of young women in Singapore that developed breast cancer. Here in the United States, people tended to develop breast cancer when they’re in their 40s, 50s, and 60s, but there for some reason people started developing cancer much earlier.
I would say research progress was slow, because running 2D gels, trying to pick out cancer versus non-cancer, trying to pick out the spots, and trying to ID the proteins — things [move] slowly
So you spend part of your time in Singapore?
Yes. The way that it worked is, I have a laboratory of about 10 people, and I would go to Singapore every three months, and spend approximately a week there. And when I was not there, obviously I’d communicate with my people by email and fax. Then I hired a person from here who at the time happened to be one of my post-docs, and I would send him over to Singapore. He worked full time and he’s in charge of the research lab there for me. We hire scientists and technicians locally in Singapore, because the Singapore government would like us to teach and train the people there.
Shortly after I set up the lab, which was basically from scratch, I wanted to develop a similar program here in Baltimore going on at the same time. By then, I was looking for another technology in addition to what I was doing on the 2D gel. I happened to talk to Ciphergen, and looking at SELDI, I found it rather interesting and exciting. I began talking to Ciphergen [in 2000]. Obviously Ciphergen was also quite interested in what I was interested in dong, and we developed a collaboration agreement.
What’s the nature of your relationship with Ciphergen?
We developed a collaboration for biomarker discovery with the understanding [that] I would use Ciphergen’s technology, which is the mass spec and the ProteinChip. And I would use my expertise at Johns Hopkins to define first what clinical questions am I asking, and then I would design the approach — how we are going to discover biomarkers. We would use [our] patients and samples. I have a collection of patient samples in a serum bank that I collected, so we can quickly take some of the specimens and do some early discovery work to see if it would be worth investigating. So my approach is, I ask the question and then design studies, and I typically do discovery [on] between 100 and 200 patient samples first, just to see if we find anything useful.
We initially just used the whole serum, without fractionation. But then we decided we need to fractionate according to pH. Typically we fractionate into six different fractions, without removing the high abundant proteins like serum albumin and IgG. Those proteins would tend to move to certain fractions that are not present in the other one, so we can look at one fraction without albumin, one fraction with albumin, and so forth. And so the six fractions we then apply to four chip surfaces that they have — IMAC, the hydrophobic chip, and two kinds of ion exchange chips. So we were trying four different chips with six fractions, and four times six is 24. Then we decided we would need to run them in triplicate, because we want to make sure that any of the variation that we see in the profile is not just random. So that’s three times 24 is 72. So that’s a lot of sample, that’s true — but we find that this way allows us to see lower abundant proteins, and we can also see [samples] with and without serum albumin.
So Ciphergen gave you the equipment?
Right. Ciphergen gave us the equipment and the ProteinChips, and also some support for other expenses related to some experiments. And we provide the know-how, the specimens, the study design. Also, we developed our own bioinformatics program. We hired a bioinformatician, Zhen Zhang, who developed a program that takes a small set of samples, rapidly identif[ies] the peaks without knowing what protein it is first, and then we calculate sensitivity and specificity, based on the clinical outcome, and then identify the best set of peaks to identify the protein.
The reason we do it that way is because if I do a tryptic digest on whole serum, I cannot deal with all these pieces of information — I’d rather simplify the spectrum into a manageable number of peaks that I would look for, and once I find they have good sensitivity and specificity I select those peaks, and then I have to go through the tedious process of purification and identification of those proteins. At that point we do use trypsin to digest the protein into fragments, and then we use standard peptide mapping. To me, it’s important to know what specific proteins were captured by the ProteinChips. I’m not interested just in the fragments — I want to know exactly what the native protein that might exist in there was.
Is the software you developed the same as what Ciphergen is selling?
No. Currently they are not selling it, [although] they might be interested. Ciphergen has [its] own software, which we also have.
You run samples through with intact proteins, identify them, and then do the digest?
Right. We select those peaks that give the best separation between cancer and non-cancer for groups, and then we identify exactly what protein might be present.
Now you’re working on a validation study?
Yes. In the case of ovarian cancer, we did an initial discovery study which we published, and so that’s kind of a proof of principle to us. That was all [on] Johns Hopkins specimens. After we found we could possibly use this approach, we were able to obtain specimens from a number of sites from the US, Europe, and Australia. We then used two sites for discovery, and cross-matched the two sites — we wanted to make sure that the markers we found were the same. And then we further validated on two other sites.
At that time we had two existing immunoassays that we could quickly run, but they were not for the specific truncated molecule. For example, a protein we found to be clinically useful was a truncated form [of transthyretin] lacking 10 amino acids But there was no antibody available for that particular protein. So we have been working with a company developing a monoclonal antibody for that particular protein. Before an antibody was available, we’d take the immunoassay against the transthyretin and we tested on the sample just to see if we got any differentiation. And then we compared it to another protein, apolipoprotein A1, and we were able to see better separation between cancer and non-cancer [samples], and were able to show that for other types of cancer there seems to be more association with the non-cancer group. We are developing specific assays against those proteins right now, and once we have the antibody for each one of them specifically, we would have a better assay [in that case]. At that time we will do a specific immunoassay validation. But with this current approach we validated at five sites, and we were able to show reasonable good performance for separating ovarian cancer. The multi-center study was for approximately 500 patients.
Now you’re doing a larger validation study in collaboration with Ciphergen?
This paper [on the 500-patient study] has been submitted for publication, but in addition to that, what we would like to do is develop specific assays in addition to what I just told you. A specific assay could be two approaches: one approach would be to develop an immunoassay with the antibody and just run it like a standard ELISA immunoassay. The other one would be, we would take the antibody and immobilize it on the Ciphergen chip surface, or somehow have the antibody on some other type of solid phase. We would capture those proteins, and then we would use the mass spec to read it out. So the mass spec would be the detector in that case.
In the case of transthyretin, we actually found an additional protein with a different truncation, so we hypothesized that the process of cancer might be taking proteins and cleaving them at many different places. So you might have a set of biomarkers — not just one with this truncation of 10 amino acids. Maybe there’s one truncated with 20 amino acids, for example. So if you use an antibody to capture the protein in the serum and then use a mass spec to read it out, you can actually find out whether it’s one protein that’s binding to the antibody, or more than one.
Let’s say you have three different forms of transthyretin, and they all have slightly different molecular weights, but they might be recognized by this antibody. So you would see three peaks. And those three would be potentially all biomarkers. That’s one reason that we think that identification of proteins is an important process. Because you would then understand and begin to learn about why those proteins are there and maybe there are additional ones that could be identified.
So that’s why we haven’t ruled out either of those approaches. Because ELISA, everyone is familiar with, and the acceptance would be much greater. On the other hand, if you want to use mass spec coupled with chips, maybe half the people may think, ‘Hey, I don’t believe this approach, because I’m not familiar with it.’ But there might be some advantage to use this approach. As of today, we don’t know yet, so we’re working on both approaches.