At A Glance
NAME: Thomas Szyperski
POSITION: Associate Professor of Chemistry and Biochemistry, University of Buffalo, New York, since 1998
Director, UB high-field NMR facility, since 1998
BACKGROUND: Diploma in chemistry, Technical University Munich, 1985
PhD, ETH Zurich, Switzerland, 1992, with 2002 Nobel Prize winner Kurt W thrich in biological NMR spectroscopy.
Habilitation, ETH Zurich, 1998
Recently published a paper in the Journal of the American Chemical Society entitled “GFT NMR, a New Approach To Rapidly Obtain Precise High-Dimensional NMR Spectral Information.”
How did you become interested in protein structure analysis?
I started studying biochemistry in T bingen [in Germany], but I felt more attracted by theoretical chemistry and physical chemistry, so [I} moved to the Technical University in Munich, where I studied chemistry. When I was looking for a suitable area to do research for a PhD thesis, I was attracted by Kurt W thrich’s work at the ETH in Zurich. He had a lot of good press, because they just solved the first protein structures [by NMR]. NMR, I found, was offering a nice combination between theory and experiments, and has a very beautiful, closed theory. Bio-NMR offered a nice blend of life science and theory. My PhD thesis was about methods development in bio-NMR and the structure determination of a small protein called hirudin [from leeches].
I stayed [at ETH] for my habilitation [credential qualifying for a professorship in Germany], which I got in 1989. That work was focusing on structure determination of proteins and on metabolic profiling using NMR. That’s still one of my interests, elucidation of metabolism in support of biotechnology research.
Then I was looking for a professorship, and the State University of New York advertised a job. They had recently bought pretty impressive NMR instrumentation that’s worth a couple of million dollars, so I couldn’t resist when they came up with a very good offer.
How are you equipped at Buffalo in terms of NMR?
We have a pretty large facility here, with six NMR spectrometers: 750 MHz, 600 MHz, two at 500 MHz, 400 MHz, and 300 MHz. We are also affiliated with the New York Structural Biology Center in New York City, which has recently been established on the campus of the City University of New York. This facility will be housing a 900 MHz spectrometer, which will be delivered sometime during the next year, and three 800 MHz and 760 MHz spectrometers. NIH has funded four 900 MHz spectrometers in total.
Your group is part of the Northeast Structural Genomics Consortium, one of the nine centers funded under the NIGMS Protein Structure Initiative. How has that been progressing?
I think the consortium is pretty much on track: We have solved 70-80 protein structures, and the methods development is coming along fine. We have an X-ray division that is led by Wayne Hendrickson at Columbia University and an NMR division that is led by Gaetano Montelione at Rutgers, and both have in the last three years delivered the same number of structures, so NMR has shown to be remarkably competitive. Up to now, our consortium has been the only one with an established strong NMR component. There is another one that was funded in 2001 and is headed by John Markley at the University of Wisconsin, Madison, which will likely also become a consortium with a strong NMR effort. But the other consortia primarily or exclusively focus on x-ray crystallography. The specialty up to now of our consortium is to explore the role of NMR, and the synergy or complementarity between X-ray and NMR.
What has been your contribution to the consortium?
We solved a couple of structures, and we developed a new technique in order to determine protein structures very rapidly [G-matrix Fourier transform NMR].
How does this method speed up data collection?
If we determine protein structures, we have to collect data for many multi-dimensional NMR experiments. The problem with these experiments is that we have to sample the indirect dimensions — setting a minimum measurement time for each measurement, which is rather long and is getting longer with the increased number of dimensions. A three- or four-dimensional experiment is taking at least many hours or many days, respectively. In order to avoid these very long minimal measurement times but to get the same information, we have developed GFT NMR. In the GFT NMR experiment, you go through all dimensions simultaneously. Because you do it in parallel, you are much faster, but then you need an additional transformation to recover the information, and this is what the G-matrix does.
How much faster is it?
It very much depends on the sensitivity of your spectrometer. For a medium-sized protein and a high-end NMR machine, it can easily be orders of magnitude. But if the protein is very large and your machine is not very powerful, the gain may be not significant. It gives the possibility to take full advantage of highly sensitive NMR instrumentation.
Has this method already impacted on your structural genomics consortium in any way?
Of course there is a lag-time until we will get structures that have been solved using this new methodology. We are now implementing new experiments, and after that, we will go into the development of new software in order to efficiently analyze these new spectra. To estimate the impact of GFT NMR on the fraction of NMR structures that are eventually deposited in the PDB is a hard job. It might be significant in the long run.
Are you planning to commercialize your method?
We filed a patent application for GFT NMR last June. The technology transfer office at UB is advertising the license, and I think they have been contacted by companies to negotiate exclusive or non-exclusive licenses.
Where do you see a commercial role for high-throughput NMR? Up until now, there seem to be very few NMR-based companies— GeneFormatics comes to mind.
There is also Affinium Pharmaceuticals in Toronto. I think the new methods will make NMR much more competitive - in the long run, NMR will play a significant role for high-throughput efforts, also in industry.
Where do you see NMR-based structural genomics going?
NMR-based structural genomics will definitely grow. There are many aspects that make NMR unique in terms of investigating the protein structures. The constant comparison with X-ray is partly inappropriate. It is competing in the sense that you get a protein fold, and you get a structure, but it’s complementary in many respects, for example dynamics. We can also look at protein structures that probably can’t be solved with X-ray. On the other hand, they can go to molecular weights we can’t approach.
I could imagine that in five years we can at least triple our throughput [by NMR]: Maybe we will get a five-fold higher throughput because of developments of methods for rapid data collection. There will be new software for automated structure determination and automated analysis of the structures. Once we have the structures, we have to try to get to their function, evaluate if it’s a new fold.
A particularly important point may be the marriage of fast NMR data collection and ab initio protein fold prediction. That’s becoming more reliable with the ever-increasing speed of our supercomputers and new methods like efficient determination of structure-based potentials. The idea of this marriage is, you have a computational method to predict a structure from the sequence, but you have multiple possibilities that are equally good. It needs only relatively sparse data in order to sort out the right one or to refine the predicted one. In this respect, NMR is much better than X-ray because with X-ray, you collect a diffraction pattern and you get everything about the molecule at once. In NMR, we can look at the backbone only, we can look at the aromatic sites only if we want, we can collect orientational constraints for helices only, for example. We can pretty much fine-tune our efforts. So when it comes to a marriage between ab initio fold prediction and experimental approaches, NMR will be a strong player — you can sort out incorrect solutions very rapidly, and you can possibly very efficiently refine certain proposals that you expect to be among the correct ones.
The metabolomics aspect of NMR is also important for proteomics. NMR is becoming a more and more powerful technique to analyze reactions of metabolism online. If you go into recent literature, I would say in the last three or four years, the determination of metabolic fluxes by NMR has undergone a revolution. It’s unique for investigating metabolic regulation using carbon-13 labeling techniques.
Despite the fact that we have learned about feedback regulation since the ‘60s, the regulation of metabolic networks is largely unexplored. We know the enzyme protein structures, we know the metabolic pathways very well, but we don’t know how cells are actually regulating metabolic pathways. The networks are so complex and holistic that experimental approaches to describe the regulation and to assess changes in regulation are tremendously important for metabolic engineering and for metabolomics. We have actually founded a spinoff company at ETH called Metabolic Concepts that provides consulting for metabolic engineering and metabolomics.