Ask the Expert: How to Best Utilize Mass Spectrometry for Diverse Applications
Stephen Barnes, Ph.D., Professor of Pharmacology & Toxicology and Director of the Targeted Metabolomics and Proteomics Laboratory at the University of Alabama at Birmingham (UAB), talks about the changes taking place in the field of mass spectrometry (MS) as it migrates from the research lab to a clinical environment.
Stephen Barnes, Ph.D., Professor of Pharmacology & Toxicology and Director of the Targeted Metabolomics and Proteomics Laboratory at the University of Alabama at Birmingham (UAB), talks about the changes taking place in the field of mass spectrometry (MS) as it migrates from the research lab to a clinical environment, for analysis of small molecules as well as large molecules like proteins and lipids. He also discusses some of the challenges facing MS users, particularly with data analysis and storage, when working with large amounts of MS data.
Q: Can you tell us about the types of MS instruments you have and what you use them for?
A: We cover a very wide range of applications: quantitative pathway analysis; quantification of phosphosites; oxidation and other post-translational modifications of proteins; lipidomics and individual lipids (e.g., prostanoids and isoprostanes); and other small molecules (polyphenols, particularly isoflavonoids). We don’t do much in the way of discovery proteomics. We tend to focus on measuring particular compounds that people already know about before they come to us or studying a particular pathway that has been found to be undergoing a lot of changes. What we then do is set up quantitative assays for all the proteins in that pathway. We are also working in an interesting frontier that I have named “metabolo-peptidomics” or “peptidometabolomics.” Proteins are not just proteins but are sources of peptides, which have different properties than the parent protein, and we have been studying lots of small peptides involved in interesting biology. In our lab we have three hybrid triple quadrupole/linear ion trap MS systems from AB Sciex, including the Triple- TOF (time-of-fight) 5600 and an older matrixassisted laser desorption/ionization (MALDI)- TOF instrument. We have a specific set of uses for each instrument, and we find that very useful.
Q: Have you seen a shift in the use of MS in recent years?
A: It’s coming back to a more precise form of MS. I have done MS since the 1960s, and the interfaces that we take for granted today were not there then. But what we could do was measure the mass of ions very accurately. In the late 1980s with the advent of the modern MS like MALDI, and the electrospray process, the instruments had very high resolution and people were able to apply MS to a ton of things they had never been able to do before. We certainly rode that wave and acquired our first triple quadrupole instrument in 1992. But by many accounts it was a lousy mass spectrometer since it didn’t have good mass accuracy, which is what a mass spec should be. So now the field has moved back to high-accuracy MS, and this is absolutely necessary in proteomics if we have to turn the corner to get to actual clinical usage. Triple quads have been the mainstay of quantitative analysis for the past 18-plus years, and they may still have their place in low-complexity scenarios. But for complex biological matrices, they are no longer enough. The proteome is denser than people seem to understand, and if a mass spectrometer is to be used for clinically meaningful analysis, the mass analyzer for the compound’s fragment ions has to have high mass accuracy and high mass resolution—a quadrupole analyzer or an ion trap can’t provide that. With newer instruments, like the AB Sciex 5600, instead of collecting one fragment at a time, the instrument collects all the ions with a mass accuracy of about 2 to 4 ppm, and then the odds that you are measuring the right peptide are considerably enhanced. Personally, I think this is a game changer and I suspect that for quantitative proteomics, the day of the triple quadrupole is over.
Q: How should people go about choosing the right MS instrument for their needs?
A: I did an experiment several years ago with a colleague, Dr. Jim Mobley, looking at some protein samples on an ion trap LTQ and on a Q-trap. Both experiments were done on the same day, so there was no day effect. We then put the data into the same informatics format and analyzed it with a single search program. My colleague found the same number of proteins that we found, but only 25 percent of the proteins were in common. We have done other experiments since and found that this represents the bias of the instrument. The Q-trap is biased toward measuring peptides with an m/z below 1000, and the ion trap does best for peptides with an m/z above 1000. We got a center grant to perform platform analysis, and I worked with a group of statisticians to improve experimental design to make sure we got rid of such systematic bias. What we found is that each instrument sees a different picture, which is really hard for clinical purposes because you can get different outcomes in different places. The fact that MS is so variable is not quite appreciated.
Q: So where should people turn to for advice when making their buying decisions?
A: I always do a demonstration with the companies that I am interested in. I give a company one half of the sample to run and analyze and then have them analyze the other half for us in real time. We don’t just get the data from them, but we want to see how easy it is to get the data. I then ask the companies for the names of their users and I go and talk to these users, which is very important. I have had quite a few people call me, and I have given them forthright advice.
Q: Quality control is probably very crucial in mass spectrometry experiments, right?
A: Quality control (QC) is really important in every lab. There was a period of MS development that I refer to as the “cowboy period,” when, God forbid, you repeated anything. Today when we run samples, every fourth sample is a QC sample. Anyone who is a real analyst, particularly in the small-molecule field, is already doing this. People run QC samples and standards; they keep records; they have procedures for determining whether or not to accept an assay. All this has to be translated to the proteomics field if MS has to be used as a more definitive instrument. We spent a lot of money on our laboratory information management system (LIMS) and now everything is monitored by LIMS. When an assay goes wrong, we can track back and find what could have gone wrong. People who use LIMS regularly have found it to make the reproducibility much more effective and sustainable.
Q: Is there anything you do on a more routine basis to run and maintain your instruments?
A: We have service contracts on instruments and we have people who are very well trained. We have made it more efficient and less expensive by doing some of the maintenance ourselves. The companies sometimes advise us remotely and they come in only when we have a persistent problem. We have just about halved our total service cost doing things ourselves.
Q: What are some of the biggest challenges you face today?
A: With MS, we are able to see only part of the proteome, so everyone is trying to get to high resolution, mass accuracy and speed. The outcome is not an instrument problem but an informatics issue. We are generating about 2 terabytes of data a month, on one instrument. At that level we are unable to move the data around on the computer network, and we are now in consultation with the IT group at UAB to try and transfer the data out into the cloud. I am also trying to get some funding to rewrite some of the classic programs that are used in MS into a cloud format so that the data analysis in the future will be done in the cloud. This requires some serious rewriting, and then the data in the cloud can be made more secure than in your building. The other problem is, we don’t want to move the data around more than once. We have been doing cost analysis on what it would take to store the data on computers, and that model simply won’t work. We are in the same position as those people doing deep sequencing, who are probably dealing with data an order of magnitude bigger than ours. We are now talking about “deep” proteomics and “deep” lipidomics, and they all come with these huge data sets and there is going to be a huge problem with data management.