Mass spectrometry for biologists

Mass spectrometry–based proteomics is still rapidly expanding, not just in terms of the methods and instruments but also the biological questions.

Carol Robinson first used a mass spectrometer as a teenager in the early 1970s. She remembers high voltages, dangers and excitement. “You could quite regularly get shocks, you could hear crackling and sparking and pumps failing,” she recalls. Now the parameters she would spend hours optimizing every week are managed efficiently by computers. “It's a very different experience,” says Robinson, now at the University of Oxford, “but still the same excitement.”

The ways researchers use mass spectrometers have changed along with the machines themselves. Mass spectrometry (MS)-based proteomics studies are probing everything proteins do: the signaling pathways in which they participate, the other proteins and nucleic acids with which they interact, even the substrates on which they act. “The type of questions has been vastly expanded,” says Ruedi Aebersold at ETH Zurich and a veteran of the field. “If biologists have interesting research questions and become familiar with these techniques, they can really do unique types of experiments.”

Biologists have noticed. “It used to be that the core mass spectrometry laboratories would have this technology and learn biology,” says Iain Mylchreest, who heads the life sciences MS division at Thermo Fisher Scientific. “Now we are seeing the biologists buying mass spectrometers to solve their own problems.” The mindset has shifted dramatically just over the past two or so years, he says. Researchers have become much more comfortable using MS to not just identify which proteins are in a sample, but to quantify how protein abundances, interactions and modifications change across disease states, cell cycle, environmental conditions and more. Multiple factors are responsible for this shift toward quantitative proteomics, he says: instrument vendors have made machines faster, more sensitive and easier to use; better protein purification and separation techniques mean samples presented to mass spectrometers are more reproducible; and publications validating such techniques are accumulating (Box 1).

Anne-Claude Gavin uses mass spectrometry to study global protein-protein interactions.

There are several approaches to quantitative proteomics. The popularity of targeted techniques to follow specific proteins is growing as assays for protein detection become more widely available. Targeted approaches are often the only way to detect very rare proteins; they are faster and require much less sample than 'shotgun' analysis. But shotgun instruments now allow the identification of many more proteins in less time than ever, meaning that more samples can be run. Compared to even five years ago, “key experiments could be redone in one-tenth the time and at five times the protein depth,” says Steven Gygi of Harvard Medical School. The next buzzword in proteomics will be “comprehensive,” he predicts. “Everything that we do in quantitative proteomics, we are redoing with an eye to being global or comprehensive.”

Find your partner

The use of MS to find protein-protein interactions has become routine over the last few years. Techniques have become so robust that experiments can reliably determine whether different interactions for particular proteins of interest occur in different species or disease states, says James Wohlschlegel of the University of California, Los Angeles. In the past, he says, “if you didn't see something, a negative result was meaningless.”

Steven Gygi believes improvements in mass spectromety will make quantitiative proteomics comprehensive.

Currently most MS-based work on protein-protein interactions is bent on identifying and characterizing them across the proteome (Box 2). Anne-Claude Gavin of EMBL Heidelberg says moving to quantitative studies will be the next step in her work probing protein-protein interactions comprehensively in Mycoplasma pneumoniae, a bacterium with a peculiarly small genome consisting of just 689 genes. In November 2009, Gavin, in collaboration with other groups, published a series of papers analyzing the organizations of the proteome, transcriptome and metabolic networks of this organism¹. To identify protein-protein interactions, they made hundreds of mutant bacteria, each of which made a different protein fused to an affinity tag. Next they captured the tagged protein along with its interaction partners and analyzed the proteins by MS. This work identified 411 distinct proteins with confidence, representing 212 tagged proteins and their untagged binding partners. Though diligence is required to avoid artifacts, the power of the technique is in its ability to find what has not been predicted. For techniques like western blotting, researchers must ask whether a particular protein interacts with another particular protein. “With mass spectrometry you just ask 'with what else does my protein interact?',” she explains.

The same principle can be used for finding proteins that interact with small molecules or nucleic acids, says Giulio Superti-Furga at the Center for Molecular Medicine of the Austrian Academy of Sciences. He pioneered MS-based techniques to study protein-protein interactions but has recently turned his attention to high-throughput techniques to discover other types of interactions. “The beautiful part,” he says, “is that you don't need any previous bias as to what you will be finding.” His group has developed methods in which nucleic acids or even small molecules are immobilized onto chromatography columns and latch onto their protein-binding partners passing by. Bound proteins can then be purified and identified by MS. Whereas established techniques such as chromatin immunoprecipitation can determine what DNA sequences transcription factors bind, his techniques can identify the other components of the protein complexes that bind to a particular DNA sequence. “Instead of seeing which nucleic acid is pulled down by a protein of your choice, you see which protein is pulled down by the DNA of your choice,” he explains.

A combination of a genomic screen and mass spectrometry–based proteomics identified the best candidates for a DNA-sensing immune protein. Image reprinted from ref. 3.

Robinson uses MS in a very different way—to examine the structures of protein complexes. Although most MS-based techniques break proteins into peptides and glean information from the pieces, Robinson works with intact protein complexes. The ion mobility mass spectrometers she uses from Waters Corporation separate species not just by mass and charge, but also by shape: bulky complexes drift less than streamlined ones in an ion mobility chamber. This approach allows her to generate topological models for protein complexes that lack high-resolution crystal structures. She recently showed, for example, that post-translational modifications influence the interactions of multiple subunits of the spliceosome, an important molecular machine that controls how exons are stitched together before translation².

All together now

Carol Robinson, now at the University of Oxford, has done mass spectrometry experiments since the 1970s.

MS in combination with other analysis techniques can be particularly powerful (Box 3). Last year, for example, Superti-Furga and colleagues combined genomic and proteomic screens to find a protein that allows the innate immune system to sense and respond to pathogens³. Gavin points out that the absolute number of M. pneumoniae protein interactions identified is not so astounding: her team studied about three times as many proteins in yeast several years ago. The main advance, she says, is in the data analysis and integration that showed proteome organization was more complex than previously thought, comprised of many homo- and heteromultimeric proteins assembled into physically connected assemblies. Such integration represents the new reality for MS-based proteomics, she says. “This is a very general trend that we observe in the field, where breakthroughs aren't produced by one technology alone but rather by the clever integration of divergent datasets.” To determine this, her group needed to team up with bioinformaticians and structural biologists to combine orthogonal information, including electron microscopy and tomography data.

Aebersold recently described a 'visual proteomics' approach in which quantitative MS is combined with cryoelectron tomography, a three-dimensional imaging technique that allows researchers to map the location of protein complexes⁴. This combined approach allowed Aebersold and colleagues to count and localize protein complexes in the human pathogen Leptospira interrogans. For her work, Robinson hopes such integration can occur even as samples are processed. “We'd like to marry mass spectrometry much more closely with other structural biology techniques,” she says. For example, she would like to use her MS separation techniques to affix rare complexes to a grid that could be fed into an electron microscope for refinement of the structure.

Fixate on the substrate

MS can also be used to study proteins' catalytic activity. Benjamin Cravatt at the Scripps Research Institute uses MS to discover unknown functions of enzymes. He has homed in on specific enzyme classes: proteases, hydrolases, histone deacetylases and others. First he redesigns enzyme inhibitors into probes that can capture proteins through affinity purification. “Then you can apply mass spectrometry and get a huge boost in the kind of information you can get from a standard experiment.” Cravatt recently identified one particular lipase out of over fifty related proteins as being elevated in highly aggressive cancer cells⁵. After that, he explains, the same probe can be used in a competitive inhibition approach to identify highly selective compounds that inhibit the enzyme of interest.

With such approaches, Cravatt explains, MS becomes a technique for evaluating a compound's selectivity against many other enzymes. “MS is essentially a new way to do medicinal chemistry,” he says. “By using these methods we can go into a complex proteome and define an enzyme as the dominant target of a particular inhibitor and confirm that there really weren't other targets for the compound in vivo.”

Gygi offers another technique for profiling kinase activity. Researchers incubate synthetic peptides with cell lysates⁶. Kinases within the lysate phosphorylate particular peptide sequences, and the modified peptides are analyzed by MS. Not only can as many as ninety synthetic peptides be quantified in a single MS run, the enzyme activity amplifies the signal, so the technique can work even when study material is limited, as is often the case for cells from clinical samples. Indeed, Gygi predicts that the technique could work with mere nanograms of cell lysate. Though the technique does not directly match enzymes with substrates, Gygi's team has been able to identify kinases and even associated protein complex members involved in activating particular signaling pathways.

Getting to Goldilocks

Many emerging applications are possible because of improved instrumentation and sample preparation, but a key ingredient is scientists' creativity. Indeed, when asked what improvements are on the horizon, Ron Hendrickson, head of proteomics at Merck Research Labs, mentions the ability to frame questions before advances in instrumentation. “One of the biggest challenges,” he says, “is on the application side and identifying the most critical questions that can be solved with these enhanced MS-based proteomics capabilities.”

No matter what technique is used, researchers still face a balancing act between analyzing overly complex samples and sacrificing information by making samples easier to handle. Most labs now dealing with yeast can identify as many as 3,500 proteins after separating the lysate into a dozen or so fractions, says Harvard University's Gygi. Getting past that gets exponentially harder, he notes: “There are many peptides from each protein and you tend to just find more and more peptides from proteins you have already.” The other problem, he says, is that there is really no way to know when a study is comprehensive. After all, no cell makes every protein in the genome, and cells frequently make just a few copies of some proteins. Bioinformatic strategies and careful experimental design helps. The proteins that are found most often in protein-protein interaction studies are most likely sticky proteins binding nonspecifically, explains Gavin, so she and another group at Heidelberg developed programs to eliminate these false positives. These programs evaluate whether proteins find each other reciprocally, as well as how often a protein is found.

At the same time, however, Gavin realizes her studies miss interactions. Not all genes can be successfully cloned or resultant proteins purified. And some components of protein complexes are certainly lost during purification and thus never identified by MS. The problem is not specific to studying protein-protein interactions. Targeted techniques, for example, identify proteins by looking for only prespecified peptides, making post-translational modifications on other peptides essentially invisible.

Researchers agree that ultimately, the art of a good MS study is making smart trade-offs. Shotgun analysis can identify more proteins the more times a sample is run through a mass spectrometer, for instance, but samples decay over time, and the extra proteins identified may not be worth the machine time, says Gygi. “We spend most of our time focused on this idea of balancing what's a reasonable amount of analysis time to get comprehensive data.” Similarly, analyses of purified samples may not require the fastest or the most sensitive machines, explains Gavin. “You have to find the best protocol in terms of quality but also in terms of economy.”

Table 1 Suppliers guide: companies offering mass spectrometry and proteomics tools

Box 1 Three recent picks

During interviews for this Technology Teature, researchers described several 2009 publications that represent the current state of the field. A small, nonrepresentative sample is included below.

Quantitation of human acetylation

Researchers led by Matthias Mann of the Max Planck Institute identified and classified 3,600 lysine acetylation sites on 1,750 proteins after exposure to deacetylase inhibitors. They used stable isotope labeling of amino acids in cell culture, or SILAC, a technique that can distinguish proteins from cells grown under different conditions⁷. For mass spectrometry analyses, they used Agilent Technologies 1200 nanoflow system connected to the LTQ Orbitrap XL.

Finding phosphorylation targets

Researchers led by Steven Gygi at Harvard Medical School compared the phosphorylation status of proteins in yeast cells with or without active forms of the cyclin-dependent kinase (Cdk1)⁸. The researchers identified 547 phosphorylation sites on 308 Cdk1 substrates and observed changes in the abundance of some phosphopeptides depending on Cdk1 activity, yielding clues about how kinase-signaling circuits evolve. For mass spectrometry analyses, they used a ThermoElectron LTQ Orbitrap.

Getting rarer

Researchers led by Ruedi Aebersold of ETH Zurich quantified yeast proteins of widely varying abundance, ranging from 1,300,000 copies to only 41 copies per cell, using a targeted mass spectrometry approach called multiple reaction monitoring (MRM)⁹. They also measured absolute abundance of each member of a network of 45 proteins involved in carbon metabolism in a one-hour mass spectrometry run. For mass spectrometry analyses, they used hybrid triple quadrupole/ion trap mass spectrometer, equipped with a nanoelectrospray ion source (4000QTrap, ABI/MDS-Sciex).

Box 2 Multiple prisms on protein interactions

In addition to mass spectrometry (MS), two other high-throughput techniques to identify protein-protein interactions are yeast two-hybrid systems and protein microarrays. Though experimental procedures are very different, both approaches rely on signals created when a protein of interest interacts with another protein. In most MS-based techniques, in contrast, a protein of interest is purified along with associated components.

To identify protein interactions by MS-based techniques, the protein's interaction partner must be associated with it in the particular cells under the conditions studied. “The MS approach is therefore not a screen for all potential partners and clearly will not identify interactions [that] are only seen in other cell types or under alternate conditions,” explain Mike Taussig and Oda Stoevesandt of Babraham Bioscience Technologies. In principle, microarray screening can identify such additional interactions, they say, but requires researchers to have some prior knowledge about what to look for.

In his research at the Dana-Farber Cancer Institute, Marc Vidal uses cell-based genetic screens to characterize the 'interactome', the full network of protein-protein interactions in a biological system. He readily accepts criticism that not all interactions identified are biologically relevant, but says that applies equally to MS-based techniques. Such experiments require lysing cells, extensive protein purification and processing, so proteins are measured under very different conditions than those in living cells. “We are all measuring things that can happen,” Vidal says, adding that what researchers need to do next is to make sure that the interactions happen in vivo.

Instead of arguing over which technique is best, researchers should appreciate approaches as “different prisms to see the same reality,” Vidal says. If MS can identify members of a particular team, signal-based techniques can identify brief but important interactions such as how a ball is passed among pairs of players, says Vidal, using metaphors for multiprotein complexes and signal-transduction pathways. And many prisms are necessary, he says. The possible binary interactions in the yeast interactome comprise some 18 million combinations. Figuring out which interactions actually happen, and when, is going to require multiple approaches.

Box 3 Keys to successful collaborations

Not all of the advancements in mass spectrometry (MS)-based proteomics are technological. “I see mass spectrometry–savvy labs becoming more and more biological,” says Giulio Superti-Furga, director of the Center for Molecular Medicine at the Austrian Academy of Sciences. “I see the community becoming more knowledgeable about what MS can and cannot do and thus more efficient in how it structures collaborations with expert labs.” The best collaborations are those where both the biologists and the mass spectrometrists are in-house, he says.

Steven Gygi at Harvard Medical School thrives on collaborations with biologists, but it is not always easy. “The first time you do an experiment in proteomics, it can be wildly disappointing,” he says. Often biologists have focused on optimizing conditions for their particular protein and have not considered how to generate materials appropriate for unbiased mass spectrometry studies. “For every [study] that shows up in Science or Cell or Nature, there may be ten that don't work.”

Benjamin Cravatt of the Scripps Research Institute says that more researchers would like to move into MS, but few laboratories have the funds or expertise to house suitable mass spectrometers, and relying on core facilities is not feasible because cutting-edge experiments cannot be standardized. He hopes the new Center for Physiological Proteomics at Scripps could provide a solution: it is tasked to tackle fundamental biological problems that are also technically difficult.

This is also the approach taken by James Wohlschlegel at the University of California, Los Angeles. “I was a biologist before I was a mass [spectrometry] person,” he says. He has seen researchers focused on MS techniques do beautiful work, he says, “but they weren't biologists, so they didn't know how to follow up. The observation would stop there.” That presents an opportunity for biologists to learn some skills from the MS experts. “For us, MS isn't the end, it's just the tool that we use to study the process we're interested in.”