At A Glance
Name: David Speicher
Position: Professor, Wistar Institute, Philadelphia, since 1986, and systems biology division chair.
Backgroud: Research Scientist in pathology, Yale University Medical School, 1980-86.
Post-doc in membrane biochemistry, Yale University Medical School, 1977-80.
PhD in biochemistry, Penn State University, 1977.
BS in biochemistry, Penn State University, 1972.
How did you get involved in proteomics?
I had been using protein chemistry to study diseases for many years. In fact, we were trying to do some proteome studies about 14 to 15 years ago. One of my colleagues here at Wistar had a number of cell lines in melanoma and pre-melanocytic types of cell lines and it was a natural to use 2D gels to look for differences at different stages of tumor progression. We observed changes and we could use protein sequencing techniques that were available then to identify a few of those proteins. But the critical factor was that 80 percent of those proteins were not in the database. So the majority of things we discovered were unknowns. That was really a stone wall that you hit if you were trying to do proteomics by those types of approaches in those days. So we decided that it’s a nice idea, but the technology wasn’t ready for it at that point in time. So we dropped that project and moved onto other things, and when proteomics really developed in the 1990s, it was a natural for us to go back and look at that and similar types of problems.
Tell me about how you developed the IEF pre-fractionation device that Invitrogen is now marketing as the Zoom IEF fractionator (see PM 11-28-03).
That relates to what we were just talking about in the sense that as proteomics was emerging as a new field, we said, “this is quite interesting and we want to pursue this,” [but] we already knew the limitations of 2D gels, and we recognized that we needed new technologies. We knew that 2D gels could maybe do 1,000 or 2,000 proteins with some degree of accuracy and reproducibility, but that was a tiny proportion of what was there. So we were looking at different ways of fraction-ating proteomes into a small number of well-resolved fractions. We tried taking a conventional IPG gel, fractionated some samples, cut out the individual pH ranges, and tried to elute the proteins to see if maybe we could come up with some sort of preparative approach for IEF. We realized that a few laboratories using the old IEF technology had gone to physically very long gels. So one of the things we were contemplating is trying to adapt that to IPG gels. The idea of cutting up an IPG strip and running it on narrow pH ranges was a way of circumventing physically large gels and yet trying to achieve a similar sort of result. When we did that, we saw that we were not getting ideal focusing in those IPG gels at high protein loads, and it was clear that wasn’t going to work directly. So we looked at the various existing solution IEF technologies, and decided that none of those were optimal, and then went about developing this method.
How does your method differ from existing IEF technologies?
One of the two major IEF technologies that were out when we began working on this problem was the Rotofor from Bio-Rad, which was a large volume device with 60 mL of sample, and it used soluble ampholines. There were partitions, but the partitions had large holes in them [and] they were really meant to prevent global convection currents and to try to preserve resolution after proteins were focused. But even if you achieved very high resolution, you had problems with collecting the sample. So we didn’ t think that approach was very promising. The other major tool that was out there was the [Amersham] Isoprime unit which was at that time sold by Hoefer. The Isoprime used Immobiline membranes to divide a series of compartments, but this device also had large volumes with 30 mL per fraction minimum, I believe. This device used Immobiline polyacrylamide gels to form pH barriers between the different chambers. This instrument was marketed at that time primarily to take partially purified proteins and obtain highly purified proteins under non-denaturing conditions. Hence the large volumes and use of peristaltic pumps to have continuous flow across the Immobiline gel membranes were important. We decided that we needed something that retained the high resolution that one achieves with Immobiline partitions, but in a much smaller, simpler device. The device we developed is more streamlined and is specifically suited for use with limited sample amounts and direct interfacing with multiple downstream protein profiling methods.
Tell me about the technologies you’re developing to look for biomarkers in serum.
We’re now to the stage where we’re combining not only 2D approaches but 3D and 4D approaches. So we are now using, in the case of serum markers, major protein depletion followed by Zoom IEF, followed by 2D gels. Or there is another very new approach that we are in the process of developing for identifying serum markers of disease, called Protein Array Pixelation. Basically what we are doing here for serum markers is major protein depletion, then Zoom IEF, but as an alternative to 2D gels, we are then running the Zoom IEF fractions out on 1D gels for a short distance. This gives us a 2D protein array, where in the first dimension we have separation by isoelectric point and in the second dimension separation by size. We then cut out each of these lanes into a uniform number of slices. Each one of these slices is digested with trypsin and analyzed by LC-MS/MS. So we have the potential, as with alternative LC-MS/MS analyses of complex mixtures, to identify a large number of proteins in each one of these samples. Because each one of the pixels in the array is being analyzed by LC-MS/MS, we can identify a much larger number of proteins than by most alternative methods.
What are you using for protein depletion?
We’re participating in the HUPO Plasma Proteome Project. HUPO had a workshop in July, and I was quite impressed with data shown by Agilent on the antibody depletion system — the Multiple Affinity Removal System (see PM 8-15-03, 10-10-03, 10-17-03). Even though the capacity is low and the column is expensive, we purchased a column and tried it, and indeed it is doing a very efficient job of depleting the top six proteins. So we took a HUPO sample and depleted the top six proteins and compared them to results without depletion, and then we also used a Zoom IEF and compared the results with and without Zoom IEF. Our two downstream analysis methods were the two I mentioned: slightly overlapping 2d gels, and the Protein Array Pixelation approach. Both are giving us good data. They showed that major protein depletion was very efficient, and we seem to not be losing a lot of minor proteins with the major proteins. Zoom IEF further simplified the separation, and the narrow range gels could identify an impressive number of protein spots. In some cases, for example, it was obvious that removing the major proteins allowed us to see many bands that were underlying the depleted proteins. On the other hand, what was also clear is that now the next most abundant proteins were a problem.
We estimate that when the entire plasma proteome is analyzed on a series of 2D gels, we are going to see at least 4,000 spots. By the Protein Array Pixelation approach, in our initial attempt, we saw slightly more than 700 proteins, but feel that this can be further optimized and many more protein identifications are feasible. We think as in many LC-MS/MS methods, a few of those protein identifications are not accurate and it requires further validation. At the same time, we now have a more sensitive mass spectrometer, and we’re pursuing enhancements of our methods that will further increase the number of proteins we can identify. So how many proteins we can ultimately identify by this technique we’re not sure, but we anticipate several thousand.
Would it make sense to develop another column to take away the next six most proteins, or is that a never-ending cycle?
We’ve discussed this with Agilent and I think they’ve probably gotten similar feedback from others as well. They’re actually pursuing removing some of the additional proteins. How many they’re going to remove and how successful they’re going to be remains to be seen of course, but there are a number of publications describing the plasma proteome that point out that approximately 22 proteins account for 99 percent of the total protein mass. So if we can remove the top six successfully, there is perhaps hope that we can remove the top 22, which leaves us with the 1 percent that we really care about. It clearly is going to get us down a number of orders of magnitude over no protein depletion.
What do you think about the idea put forth by Thomas Conrads that some potential biomarkers are attached to high-abundance proteins, which act as molecular sponges (see PM 10-17-03, 11-28-03)?
I think there’s a lot of merit to it. There are two types of questions. One is, ‘let’s remove the big proteins as efficiently as possible so we can identify those small ones’; the other is, ‘let’s look specifically at the proteins that are bound and see if there is some physiological method there,’ which is clearly what Conrads is doing. Since they’re pursuing that, I don’t think there’s a lot of advantage to our pursuing it. What I see with the Agilent column is that they seem to have done a good job of coming up with conditions where they retain good binding to the antibody and are minimizing protein-protein interactions.
So I think the molecular sponge theory is an exciting one, [but] the Agilent column to our first approximation seems to be removing that interaction. If you’re trying to get as many of your biomarkers as possible into one pool for further analysis, which is what we’re doing, it’s a promising approach.