Mass Spec Tackles Proteins

Mass spectrometry is a key method for studying proteins, playing an increasingly important role in the study of protein modifications, such as phosphorylation, and even in structural biology. These applications were very much in evidence at a gathering of more than 2,000 mass spectrometrists, the 17th triennial International Mass Spectrometry Conference, held in late August in Prague, Czech Republic. Still, much remains to be done before mass spectrometry can fully deal with biological complexity.

Although protein identification by mass spectrometry has become practically routine, identifying modifications of those proteins has not. These modifications, such as methylation, acetylation, and phosphorylation, play important roles in cell signaling and other regulatory processes. Figuring out that a peptide has been modified isn't the hard part. Figuring out the exact location of that modification is the challenge.

Most proteomics experiments start with digesting the proteins with a protease enzyme such as trypsin to form so-called tryptic peptides. Then, peptides with a particular type of modification are collected and analyzed by mass spectrometry to determine their sequences and whether any amino acids have been modified.

Advances in mass spectrometry have made possible a "historic shift" from analyzing peptides to starting with intact proteins, said Neil L. Kelleher, a chemistry professor at the University of Illinois, Urbana-Champaign.

"All the other techniques developed over the years are there because we couldn't handle the intact protein, but you really miss a lot," said Donald F. Hunt, professor of chemistry and pathology at the University of Virginia. "A tryptic peptide doesn't tell you what modifications are on the protein. It just tells you that you've got a little piece and it belongs to protein X."

The transition to analyzing intact proteins has been fueled particularly by advances in fragmentation techniques, most importantly electron-capture dissociation (ECD) and electron-transfer dissociation (ETD). In ECD, free electrons captured by a peptide or protein cause the backbone to fragment at the peptide bonds. In ETD, the electrons are delivered by an anion that transfers its electron to a multiply charged peptide or protein. This latter reaction occurs in milliseconds, making it possible to identify intact proteins on the same chromatographic timescale used to separate mixtures of tryptic peptides.

Early work that led to the development of ETD in Hunt's lab used a design that placed the ETD source and the sample ionization source on opposite sides of an ion-trap mass spectrometer. In his current research, Hunt has now placed the two ion sources on the same side of the instrument. With this new configuration, Hunt hopes to be able to couple ETD to hybrid instruments that have a higher mass range and so measure masses to much higher accuracy than is currently possible on the stand-alone ion trap. "This would allow me to distinguish a whole bunch of different posttranslationally modified states," of a given protein, Hunt said. He acknowledged that Scott A. McLuckey of Purdue University beat him to the punch, publishing similar work earlier this year (Anal. Chem. 2006, 78, 4146).

ECD and ETD are particularly powerful for analyzing protein modifications because they fragment the backbone without touching the modification, Kelleher said. ECD complements an older fragmentation method known as collision-activated dissociation (CAD), said Roman Zubarev, a professor at Uppsala University, in Sweden. ECD and CAD break different bonds on peptides, and protein identification is actually improved by using both methods, he said.

ETD "allows freedom from trypsin digests," Hunt said. The fragmentation method makes it possible to use other protease enzymes that break the protein into fewer, larger peptides that are more highly charged, he said.

Ole N. Jensen, a proteomics expert at the University of Southern Denmark, is excited about the prospects for proteomics created by ECD and ETD. "These two methods may be the clue to being able to do general posttranslational modification analysis," he said.

The quantification of protein modifications is also important for understanding the modification's biological function. Most studies don't reveal the extent of phosphorylation, said Hanno Steen, director of the Proteomics Center at Children's Hospital in Boston. His group has developed a new method for quantifying phosphorylation by using unmodified peptides to normalize the signal intensity of modified peptides.

They focused on the anaphase-promoting complex (APC), which is a complex that marks proteins for degradation during mitosis, thereby triggering both the initiation of anaphase and the exit from mitosis. They found that the phosphorylation pattern of APC components changes at different points in the cell cycle. While studying the effect of several mitotic spindle poisons on the phosphorylation pattern of APC, they found that different drugs affected the phosphorylation patterns differently, even though they arrested the cell cycle at the same point.

An even more daunting challenge in proteomics is to characterize the constellation of modifications made to histones—the proteins that make up the spools that DNA wraps around in the nucleus—and how those modifications affect gene expression (C&EN, July 17, page 13). These modifications, which make up the so-called histone code, consist mainly of methylations and acetylations that are responsible for turning gene expression on and off. They also include phosphorylations. A number of mass spectrometrists, including Hunt, Kelleher, and Jensen, have turned their attention to histones.

One of the reasons that characterizing histone modifications is so challenging, Hunt said, is that although there are only five different types of histones, each one can be modified in myriad ways. One of the histones, H3, has a highly basic N-terminal tail that is often heavily modified. More than 100,000 forms of the same H3 protein exist, and every modification is important, Hunt said.

When Hunt first started working with histones, instrumental limitations prevented him from working with the intact proteins. The first time he tried to analyze histones, he saw nothing but an "unresolved blob." Instead, he concentrated on the first 50 residues of the N-terminal tail, which he derivatized with propionic anhydride and then cleaved with trypsin into five peptides. He compared histones from two phases of the cell cycle, isotopically labeling one set so that they could be distinguished.

He found more than 1,000 variations in the five peptides. Some of those peptides didn't change much between interphase and mitosis, but others changed a lot, he said. The spectra revealed that sites of lysine methylation that regulate gene activation and silencing occur adjacent to potential phosphorylation sites. When these adjacent sites are phosphorylated, the negatively charged phosphate group blocks specific protein-binding partners from recognizing the methylated lysines and thus regulates both gene activation and silencing.

Interpreting the histone code requires analyzing the co-occurrence of modifications, Kelleher said, which in turn requires analyzing the intact proteins, or at least large peptides. The team of Kelleher and Craig A. Mizzen at Illinois has recently completed an analysis of all the core histones from human cells.

Jensen, collaborating with Dutch scientists, is using mass spectrometry to investigate histone modifications in the malaria parasite. They have found approximately 50 modifications, mostly acetylations but also methylations and one phosphorylation. "It tells us that on average the chromatin of the malaria parasite is encoded for very high activity," Jensen said. "Lots of genes are turned on to be transcribed and translated into products."

Mass spec's increasingly important role in proteomics is not without challenges. For example, the vast amount of data being generated from proteomics experiments means that the interpretation of mass spectra must be automated. "Fifteen years ago, we could do a protein a day. We do one a second now," said Garry L. Corthals, a mass spectrometrist at the Turku Center for Biotechnology in Finland. He warned, however, that it is imperative now to consider how the mass spectra of protein modifications are interpreted, before the protein databases become populated with incorrectly annotated spectra.

Wolf D. Lehmann of the German Cancer Research Center in Heidelberg warned about the pitfalls of automated evaluation of fragmentation patterns. Phosphopeptides, for example, are identified by their characteristic loss of 98 mass units during fragmentation. Other types of fragmentation losses, however, can fall into the window used to select putative phosphopeptide precursor ions for further fragmentation, leading to false-positive identifications. Another source of mix-up, Lehmann said, can be the misidentification of charge states.

Annotating spectra is "inherently difficult," Corthals said. He described a method that he hopes will make such annotation easier and reduce the number of false-positives. In this method, phosphopeptides are turned into their own reference peptides. First, a mass spectrum of the phosphopeptides is obtained. Then, the phosphate groups are removed by an enzymatic treatment, generating a second set of peptides from the same sample that can be used to validate the phosphorylation site assignments.

The method uses intensity distributions and mass shifts to match the phosphorylated peptides and their dephosphorylated counterparts. Corthals' team has manually crunched through more than 300 examples to convince themselves that the automated method works. They plan to make their open source software freely available on the Web soon, Corthals told C&EN.

In addition to its central role in protein characterization, mass spectrometry continues to make inroads into structural biology by helping researchers figure out how proteins interact with one another in complexes.

Mass spectrometry is certainly not the first technique that comes to mind for structural biology. It may lack the high spatial resolution of techniques like X-ray crystallography and nuclear magnetic resonance spectroscopy, but researchers such as the University of Cambridge's Carol V. Robinson are showing that it has plenty to contribute to the field.

"We can take much more heterogeneous complexes than would typically be studied by structural biology, where you are restricted to very pure complexes of single species," Robinson said. "We can take mixtures of protein complexes and hope to get information about interacting subunits and even the spatial arrangement of subunits."

Robinson and her group are now able to take complexes directly from cells rather than using reconstituted complexes. In particular, her team is getting structural information about the interactions between proteins in complexes by combining mass spectrometry with homology modeling, carried out in collaboration with Robert B. Russell's group at the European Molecular Biology Laboratory in Heidelberg.

She used such an approach to construct a model of the yeast exosome, a 10-protein complex involved in RNA processing. "We knew which subunits come off the model quite early in the research, the peripheral ones, but we had no way of getting into the core of the complex, which held the key to how to build the model," Robinson said.

A lucky break came when a postdoctoral researcher in the group began adding small amounts of the organic solvent dimethyl sulfoxide to solutions containing the exosome. This disrupted the six-protein ring at the core of the exosome into three pairs of proteins. With these smaller pieces, they could tell which proteins interact with one another, and they then simply fit the pieces together like a jigsaw puzzle. By generating 20 subcomplexes and using an algorithm to explore all possible networks, they were able to construct the first 3-dimensional model for the intact exosome (EMBO Rep. 2006, 7, 605).

"We're getting more and more ambitious and going up in size with these complexes," Robinson told C&EN. "Of course, the number of possibilities that you then generate for all the subcomplexes we use to build our models becomes much more difficult and challenging because you've got many more subunits to play with." Robinson and her team have used a similar approach to predict the structural organization of the lid of the regulatory particle from the yeast proteasome, a complex that regulates degradation of proteins (PLoS Biol. 2006, 4, 1314).

In addition to the use of mass spectrometry as an independent tool, it can also be used to help interpret data from other techniques. Robinson believes mass spectrometry and electron microscopy will make a powerful team for predicting structures of protein complexes. Electron microscopy measurements reveal the overall shape of the complexes. Mass spectrometry can help locate the subunits within the electron density. In addition, Robinson said, charge separation with mass spectrometry can be used to purify samples for electron microscopy.

She and her coworkers are putting electron microscopy grids directly into a mass spectrometer and using a technique known as soft landing to deposit samples on the grid. Implementation has proven difficult because the grid attenuates the ion beam, but she reported that her team is finally succeeding in simultaneously obtaining mass spectra and landing proteins on the grid.

In addition to structure, conformational changes in proteins and protein complexes are of great interest to researchers seeking to understand protein-protein interactions. These changes can be seen with ion-mobility spectrometry, a cousin of mass spectrometry that is sensitive to the overall shape of the molecules and complexes. Perdita E. Barran of the University of Edinburgh, in Scotland, described work to determine whether other divalent metals cause the same conformational changes as calcium causes in the calcium-binding protein calmodulin. Without calcium ions, calmodulin is globular, but calcium binding changes its shape to a dumbbell.

Barran's team used ion mobility spectrometry to study complexes of calmodulin with an enzyme called neuronal nitric oxide synthase, which prefers to bind calcium-loaded calmodulin. By looking for complexes of the enzyme with dumbbell-shaped calmodulin, they figured out whether different divalent metals cause the structural changes in calmodulin necessary for the interaction.

Calmodulin prefers calcium over most of the metals tested, but lead showed interesting behavior. Lead can't bind metal-free calmodulin, but it efficiently displaces calcium from calcium-loaded calmodulin. This efficient binding of lead occurs even when the lead concentration is only one-tenth of the calcium's. Barran called this observation "slightly worrying" and suggested that it may partially explain lead's toxicity.

Mass spectrometry can also help characterize complexes that are too fragile for other methods. For example, Jennifer L. Beck, a senior lecturer at the University of Wollongong, in Australia, described work with the Escherichia coli replisome, the cellular machine that carries out DNA replication. This complex has not been isolated intact in sufficient quantities for biophysical characterization, she said.

She and her coworkers are using electrospray mass spectrometry to study oligomeric forms of the portion of this machine that unwinds the DNA, called the helicase. So far, they have observed heptameric forms of the helicase, which is usually made of six identical proteins.

Although mass spectrometry is making tremendous strides in biological applications, mass spectrometrists have wish lists for further improvements.

Key advances that would enhance the identification of posttranslational modifications, whether in peptides or intact proteins, include increasing the dynamic range and increasing the duty cycle of the mass spectrometer. The dynamic range affects the ability to measure both low-concentration and high-concentration species in the same sample. The wider the dynamic range the better, because without good separation methods, the most abundant peptides or proteins can swamp the more interesting ones, which tend to be present at low concentrations.

The duty cycle refers to how quickly the instrument can make measurements. "If you want to inject crude samples, you are limited by duty cycle," Jensen said. "You simply cannot cope with that many peptides."

The way around these problems is with better separation methods. Kelleher would like to see methods that allow the characterization of 1,000 intact proteins in a single sample. Then, "top-down" proteomic approaches with intact proteins would be competitive with bottom-up proteomic approaches with protein digests.

Hunt called the dynamic range the Achilles' heel of mass spectrometry. "There's no mass spectrometer that has a dynamic range higher than 5,000-10,000," he said. "If I have two peptides coming into the mass spectrometer at the same time and they go through electrospray ionization, the one that is 5,000 times more concentrated will absorb all the charge. You'll miss the little guy." Unfortunately, Hunt said, the dynamic range problem is amplified by the fact that the dynamic range of proteins in the cell is more like a million to one.

Yet Hunt is optimistic that such dynamic range limitations can be overcome. "The biology is driving innovation," he told C&EN. "Three years ago, if you looked at serum, the only thing you'd see was serum albumin. Now, people are going deeper and deeper because they can find a way to get rid of the top 10 proteins."

Often, it is necessary to make trade-offs between accuracy, sensitivity, and dynamic range. "Mass spec can do many of the things that I would like to do, but they cannot be done by one instrument," Jensen told C&EN. "You want high sensitivity, high dynamic range, high duty cycle, high precision, and high accuracy all in one instrument. We don't have that yet. Right now, we need to use different instruments for different purposes."

In addition, both Kelleher and Robinson would like to see improvements in electrospray ionization. Robinson seeks greater efficiency in electrospray. "A lot of what we spray is wasted," she told C&EN. "We know we don't have huge efficiency in that step."

Kelleher wants to find a way to "supercharge" samples in electrospray by driving all species to their highest charge state. Most species, he said, have 60 to 70 distinct charge states. Having such a wide range of charge states in a single mass spectrum significantly reduces the signal for each species.

With these advances and others still to be dreamed up, mass spectrometry will become an even more indispensable tool for understanding proteins.