McKusick-Nathans Institute of Genetic Medicine at John Hopkins University
Who: Akhilesh Pandey
Position: Associate professor, McKusick-Nathans Institute of Genetic Medicine at John Hopkins University, departments of biological chemistry, oncology, and pathology, 2006-present
Background: Founder and chief scientific officer, Institute of Bioinformatics, International Tech Park in Bangalore, India, 2002-present; assistant professor, McKusick-Nathans Institute of Genetic Medicine at John Hopkins, departments of biological chemistry, oncology, and pathology, 2002-2006; postdoc fellow, Whitehead Institute for Biomedical Research, 1996-1999; postdoc fellow, University of Michigan, 1995-1996; PhD in pathology, University of Michigan, 1995
Pandey is the senior author of a study published this month that evaluated the use of electron transfer dissociation for a global phosphoproteome analysis. ETD is a relatively new peptide fragmentation technique that proponents say offers a more comprehensive coverage of peptide sequences and post-translational modifications than collision-induced dissociation.
Pandey and his colleagues say that by using ETD, they were able to identify 1,435 phosphorylation sites from proteins encoded by 500 genes from human embryonic kidney 293T cells. In their paper, the authors say that 1,141 of the phosphorylation sites are novel.
The research appears online in the Feb. 7 early edition of The Proceedings of the National Academy of Sciences.
Below is an edited version of a conversation ProteoMonitor had with Pandey this week.
If I’m a proteomic researcher, why should I care about your study and its findings?
The usual technology that has been used for peptide fragmentation — and that is basically what enables identification of the amino acid sequence and localizing post-translational modifications like phosphorylation — has been collision-induced dissociation. It’s been used for a long time, and it has known drawbacks. The most pertinent one is that serine and threonine phosphorylation sites are very labile. So what you see in a mass spectrometer, in a tandem MS/MS type experiment, [is] a signature of what you get after the phosphate has been lost on the peptide.
This new method, [which] has been developed not by our group but by Don Hunt’s group [at the University of Virginia], is electron transfer dissociation. … [It] allows the peptide to retain these post-translational modifications. The major one is phosphorylation, but O-glycosylation is another PTM which basically remains intact during this fragmentation process.
What ends up happening is you catch it in situ on the peptide. It is a more reliable and surer way of localizing phosphorylation sites. Basically this is similar to electron capture dissociation, which also has the advantage that the peptides are fragmented more uniformly.
Collision-induced dissociation, or CID, is the more random process. People who work on large numbers of peptides, they know many peptides will just fragment where they want to. You cannot really predict them, whereas [our method] gives you a more complete coverage. So if you wanted to have more fragment ions at some level, peptide identification is a question of match. You are matching the fragment ions that you get from MS/MS experiments against a database. So the more points you have on fragmentation, the more confident you are going to be.
Were you able to identify more novel phosphopeptides, or get a more thorough understanding of phosphopeptides that are known already?
What ends up happening in an average peptide — let’s say if you have a 10- or 12-amino acid long peptide — you have continuous sequence information maybe for four to six amino acids, and not the rest of the peptide. You have a lot of other information, but you may not have complete evidence of each amino acid in there.
[Our method] takes you closer to that goal. It’s still not complete, but it’s more uniform compared to CID, and this is also what we have addressed in the context of phosphopeptides. It gives you more uniform fragmentation and more fragmentation.
What we are also arguing is that this needs another larger scale experiment where you can combine this data not just for phosphorylation but basically for peptide analysis itself where you could surely combine the two methods. The mass spectrometers on which ETD is enabled these days, the ion trap mass spectrometers, you could take it through your average LC/MS experiment and do it on the same peptide as it eludes a CID experiment and then turn around and do an ETD experiment.
These are two orthogonal ways of getting data, and if both of them [identify] the same peptide, now you have increased your confidence.
Are there certain types of phosphopeptides that you can’t identify through CID that you can with ETD?
In fact, many of the phosphopeptides you cannot identify by CID because of that limitation. With ETD, you can identify many more of those; that advantage still holds.
If you’re able to do what you say you can with ETD, what implications does it have for proteomics research?
I think what it will allow us to do is two specific things: One, it will allow us to dig much deeper into the phosphoproteome. That is for sure. What we are seeing is that even though a method gets popular and they get better over time, the inherent weaknesses of the method, you cannot solve.
And CID has this inherent weakness that it is just not going to give you all the sites that you would otherwise like to know. This especially has relevance in the context of systems biology where what you’re trying to achieve is [to] find as many things, in this case phosphorylation sites, on as many proteins in as few experiments [as possible].
So even though we were not the ones to come up with the technique, what we wanted to find out was: In the context of a global phosphoproteomic experiment, what is it that one can get? For the peptides that we had identified, we went back and we looked at how many sites were reported on the peptides that we were able to capture on these experiments.
We have found many phosphorylation sites, of which 80 percent were novel. But when we took the other ones [that] are described in the literature, that’s the 20 percent, we [went] back and we saw on those peptides, how many more could we have ever found.
That means: What are all the sites that have ever been described? And what we found is that the 20 percent of known phosphorylation sites that we found was described in more than 100 different publications. And we could capture them in one. We basically were able to find 80 percent of all the phosphorylation sites ever described.
Does the CID method have any advantages over ETD?
I would say [that] in the context of looking at peptide sequences and post-translational modifications, perhaps not. For a given peptide, it is still possible that CID gives you nice fragmentation when ETD gives you a similar but different pattern.
If you could combine those two, it would be better than [either one by itself]. It would not be difficult at all [to combine the two methods]. Because using these ion trap mass spectrometers, you can easily set your experiment to do both for every peptide. So you alternate between ETD and CID, and we have done that actually.
For some peptides, you would be no better off than [using] ETD alone. For most peptides, you will be better off than with CID alone. But in a small set alone, perhaps, where even ETD data exists but isn’t of good quality, if CID complemented that data, complementarity of the method is the main thing. Many people in the field say these methods are orthogonal. Really, what they mean is, they are complementary.
What specifications for a mass spec do you need to do this?
I would say that the mass spectrometers to which this method can be applied are being expanded as we speak. What is really more limiting in a practical world today is which company supplies which mass spectrometer that has the necessary hardware that allows you to do this.
What sort of hardware do you need?
The commercial instruments that are available now with ETD capability are all ion trap mass spectrometers. But then there are occasional published reports where instrument developers are applying this to other kinds of mass spectrometers such as a quadrupole time-of-flight mass spec. So what we’re going to see is in the future, ETD capability will be an option that will be available on other types of mass spectrometers. Today, it is available on ion trap mass spectrometers.
In the context of phosphorylation, which is very labile, [the ion trap mass spectrometer] has already done what people by and large have not been able to do with Fourier transform mass spectrometers and electron capture dissociation. Those have significant hardware requirements in the sense that you must have a Fourier transform mass spectrometer, and those are very expensive and very few people are experts at using them … though more and more people are getting those types of instruments.
But if you go back and see how many of the groups that have had the Fourier transform mass spectrometer with this electron capture dissociation option available have been able to engender these kinds of data sets, now that number has been vanishingly small.
What has really happened is that the groups that have been pioneers in the electron capture dissociation field have generated slightly larger, yet still proof-of-principle, data sets saying it is feasible.
But people have not been able to do global phosphorylation studies. What we are doing is using a technology that is very similar but which is using a much cheaper mass spectrometer, the ion trap mass spectrometer. This is the garden variety [instrument] that is found in many different labs. And more people can use this to try to come up in a way to the specs of FTMS for the purpose only of phosphorylation analysis.
What we have to do is be prepared to accept the lower resolution. The resolution is the issue.
Does the lower resolution skew the results?
I think we would all be happier with higher resolution. It is hard to say whether it, by itself, skews the results. What may happen is we may have to discard results. But I don’t think we are skewing. We basically would like to put our names on data sets that are valid.
ETD coupled with even higher resolution would be the answer, but today, that platform, the middle-of-the-road platform, is not available. We keep hearing there are other instruments out there, like [Thermo Fisher’s} Orbitrap, which companies are scrambling to try to put ETD [capability to].
What sort of kinks do you need to address as you further develop this ETD method?
You might think, ‘Well, the next obstacle might be in further developing the hardware.’ We have tried to address many of the biological issues like, which enzyme should be used, what are the pitfalls.
I think one of the bigger obstacles that many of the people are not thinking about [is the fact that ] these are the early days of generating large data sets of this type. [We need] to be able to devise the algorithms and make them better to search databases.
At the end, we are left with tens of thousands of mass spectra. The way we used to do things five years ago was to interpret everything manually. But as people develop more and more faith in the technology, they will even develop hundreds of thousands of MS/MS spectra.
The search algorithms are not up to speed. So what we are now doing is we are carrying out a systematic evaluation of many of the search programs out there to see how often they agree and what the differences are. And then maybe we are going to learn from the data to try to get these programs to work better.