Some patients with schizophrenia experience auditory hallucinations. Such illusions are among the most disturbing aspects of the disease. Patients often hear people talking to them or about them. The voice can be a running commentary on what the patient is doing at that moment, a conversation, an outright command, or even a warning. Pitt psychiatry faculty members Robert Sweet and Matthew MacDonald wanted to find out whether a protein-signaling problem in the brain might play a role in those hallucinations.
Based on previous research, the two hypothesized that such a signaling error would probably occur within the synapses in the auditory cortex of the brain. But, says Sweet, “the synapse has close to 2,000 proteins.” To look at each one would be time consuming and labor-intensive. Plus, he says, “when you just study a protein at a time, you’re not really getting a picture of the global function or dysfunction of the synapse in disease.”
Here’s where Nathan Yates, associate professor of cell biology and scientific director of Pitt’s Biomedical Mass Spectrometry Center, enters the picture.
MacDonald, PhD assistant professor of psychiatry, and Sweet, MD professor of psychiatry, obtained auditory cortex tissue of deceased patients from Pitt’s brain bank, both from people with no known history of mental illness and from people with schizophrenia. Then, with Yates’s guidance, they used a mass spectrometer, which helps scientists identify relevant biological molecules in a sample by converting them to ions and sorting those by their mass-to-charge ratios. With this approach, they were able to quantify more than 150 synaptic proteins in that auditory cortex tissue, and they observed significant differences in the way 17 of those proteins were organized into a network in the tissue from people with schizophrenia.
“We found a group of proteins that hang out together in schizophrenic patients that don’t really hang out in normal controls,” MacDonald says, adding that they are pursuing more studies in this area.
Published in the June 2015 issue of Biological Psychiatry, this experiment is but one of several that Yates has been involved in since he joined the faculty in 2011. Yates, a 49-year-old chemist, is giving biomedical scientists all over the Pitt campus new perspectives on their work. As he puts it, his specialty areas of mass spectrometry and proteomics “provide a large amount of robust quantitative data that can help scientists observe and understand complex biological systems.”
In a sense, Yates’s expertise and the center’s tools (which include powerful cloud computing platforms that Yates was instrumental in developing) enable investigators to pull back and identify important biological players that they may never have recognized otherwise.
Before Yates, a PhD, came to Pitt, he was at Merck and Co. When he was hired as a senior research chemist there in 1995, he was tasked with using his knowledge of mass spectrometry to screen large libraries of small molecules for drug discovery. The company was interested in evaluating thousands of molecules at once to identify those that bind proteins and alter given disease states (proteins involved in glucose regulation, for example, or perhaps those involved in regulating blood pressure). This required new software, which Yates created.
While doing that work, Yates began to think about how similar bioinformatics methods might be developed to analyze data on not just small molecules, but proteins, too. Then in 2002, Merck launched a major effort to discover protein biomarkers for a number of therapeutic areas, including neurology and oncology.
Such discoveries would make testing potential drugs more accurate and efficient; they might also help in the development of drugs that could halt progression of diseases before symptoms appeared.
Sometimes scientists use mass spectrometers in a targeted way, focusing the instrument on specific proteins or other biomolecules believed to be important to a disease or biological process. This approach offers “absolute molecular specificity that can’t be achieved with traditional assays,” says Yates. (Sweet, MacDonald, and Yates took a targeted approach in their schizophrenia study.)
With a targeted approach, you pretty much know what you are looking for—at least as a starting point.
But Merck and Yates wanted to take a different path. They wanted to somehow measure and identify biologically relevant proteins without having to tell the machine what they were searching for. Such “unbiased” methods existed but weren’t always reliable or scalable to large numbers of samples.
Yates set out to create a new bioinformatics method that could do this and do it quickly. He initiated a collaboration with mathematician colleague Matthew Wiener.
The men spent their lunch hours hunkered in front of their computers, eating peanut butter and jelly sandwiches as they tinkered with formulas. Months later, they had a powerful algorithm. Yates dubbed it “differential mass spectrometry.”
“Let’s say you take healthy and diseased cells, and you want to ask, ‘What’s different?’” Yates explains. “There have been lots of quantitative proteomics approaches for doing that, but what would typically come back is—instead of a short list of two or three or five proteins that you could follow up on—you could get a list that had a couple hundred proteins that were changing. That list might contain those five proteins that actually were changing, but it would also contain [dozens of ] other proteins that were just changing due to random chance, so the experiments weren’t very good at finding just the needle. They would find a lot of hay, too. You went from a big haystack with a needle in it to a not-so-big haystack with a needle in it.
“With differential mass spectrometry,” he says, “it brings back the needle (or needles if two or three proteins change) itself, or nothing at all.”
After creating the algorithm, Yates worked with software developer Andrey Bondarenko, who was employed by Rosetta Biosoftware, a subsidiary of Merck, to turn it into commercial software.
Merck, and Yates, used the new method in the hunt for protein biomarkers for a number of conditions. It allowed them to, for example, align and compare complex proteomic profiles obtained from the spinal fluid of patients with Alzheimer’s disease and controls to reveal differences. Eventually, the approach was successful in identifying two robust candidate biomarkers that continue to show promise in independent studies for Alzheimer’s disease. (Discovering and verifying such markers is a more involved and lengthy process than people probably realize, notes Yates. “Even with the financial muscle of large pharmaceutical companies,” he points out, it can take decades.) Their work is outlined in the August 2015 issue of the journal PLOS ONE.
Merck knew what it was doing when it asked Yates to reimagine existing technology to find a biomarker. Yates had been an innovator and problem solver for every other lab he’d worked in, starting as a grad student.
Yates was first introduced to mass spectrometry in the late 1980s when he attended the University of Florida for graduate school. His mentor there, Rick Yost, had made a huge contribution to the field. Fifteen years earlier, when Yost was a graduate student at Michigan State University, he had coinvented the first triple quadrupole mass spectrometer.
Mass spectrometers work by first converting molecules to charged particles (ions). Yost’s triple quadrupole analyzed these ions in two steps—before and after they are broken apart—by colliding them into a gas.
“The [triple quadrupole] has revolutionized analytical science,” Yates says. It’s now the most widely used mass spectrometer in the world.
Yates adds, “It’s really great for measuring trace-level analytes present in ultracomplex mixtures such as amino acids in a baby’s blood, or drug stimulants in blood sampled from racehorses, or harmful gases in the air on the space station.”
When Yates arrived in Florida, Yost’s lab was focused on a new type of mass spectrometer, a quadrupole ion trap, which was just coming on the market. It had the promise, but not yet proven capability, of performing very rapid ion analysis—which would make it less labor-intensive and costly to run.
Yates has always enjoyed finding out how things work. In fact, he likes to joke that he’s more mechanic than chemist. During their honeymoon, he and his wife, Jan, drove across the country in an old VW van with an engine that Yates rebuilt himself and tinkered with constantly to keep running. And when their historic Colonial home needed new air conditioning, rather than pay $8,000 for a new system, Yates decided to get certified in air conditioning repair and install it himself.
With that kind of outlook, it was only natural that he would try to solve the practical problems of the quadrupole ion trap. So, with Yost’s blessing, he began trying different things to speed up the ion analysis. “He set up this very intense 24–7 use of the mass spectrometer,” Yost recalls with a hearty chuckle. “We figured afterward he’d done the equivalent of two years of mass spec in a week.”
Later, Yates briefly left graduate school to undertake a six-month co-op in research and development at Finnigan (now part of Thermo Fisher), a Silicon-valley company that manufactures high-end mass spectrometers. While there, he helped to advance the capabilities of ion traps.
Today, the mass spectrometers that he uses for proteomics research at Pitt have ion traps inside them. Says Yates, “they’ve advanced into high-performance instruments that are a mainstay in proteomics and metabolomics research.”
After Florida, Yates took a postdoc at the University of Virginia in the lab of chemistry professor Donald Hunt. This was 1993; the field of proteomics was emerging, and Yates wanted to learn all he could about it.
Hunt was studying proteins in the immune system and was interested in finding out what an ion trap could do to improve his measurements. Yates went there to build an ion trap and then apply it to the analysis of proteins.
In the ’80s, a Yale chemist named John Fenn had invented a technique called electrospray ionization. (He won the Nobel Prize in Chemistry for this in 2002.) Fenn’s method allowed large biomolecules to be analyzed by mass spectrometers—a game changer for biomedical science.
Yates coupled electrospray ionization with the ion trap he was building. He also wrote software that modified how the ion trap operated and gave the software to the 20 to 30 labs around the world doing similar work.
From there, he took the Merck job opportunity, where he ended up developing differential mass spectrometry. Once Yates had developed this new tool, he was like a kid with a new computer game. He couldn’t wait to use it. He envisioned himself performing drug-target analyses to help Merck better understand the medications it was bringing to market. He knew there was value in it. Beyond the possibility of making drugs more effective, there was the issue of side effects. If drug firms don’t know to what proteins their drugs bind, they risk having to pull them off the market should toxic effects emerge. But company officials resisted Yates’s overtures.
“I was fascinated about the possibilities of using differential mass spectrometry for drug-target analysis,” he says. “And it was disappointing that I couldn’t communicate my excitement to the company.”
Yates began to look for a new job. He was close to moving to another position in industry when a contact at Pitt called and asked him to consider applying to the School of Medicine. Yates jumped at the chance, knowing the eventual move to academia would give him the freedom to pursue his drug-target research, as well as other interests.
Again he knocked on the door of “crackerjack” software designer Bondarenko, who was using cloud computing to develop advanced software tools for scientific analysis. The two teamed up with Amazon and created a cloud-based platform, CHORUS (chorusproject.org), that allowed mass spectrometry labs around the world to work together by sharing computational tools and data.
“Instead of paying large upfront costs to purchase software, install computers, and hire experts to analyze mass spectrometry data, any researcher with a connection to the Internet can now log on to CHORUS and begin analyzing data in minutes,” Yates says.
CHORUS, which is operated under a not-for-profit public-private partnership, went live in 2013; already 1,200 scientists and 220 labs around the world are using it. These days, Yates and Bondarenko (now heading Seattle-based InfoClinika) are working on a version of differential mass spec to make available on the site. With the software, says Yates, “It’s practical to measure and analyze millions of data points in a few minutes, a task that was not possible with older manual methods.”
Yates is all about getting datasets out of individual labs and into collaborative environments. At a recent conference in São Paulo—after a friendly soccer match among the attendees and some beer drinking—Yates befriended a Brazilian researcher doing proteomics work on the Amazon rain forest and convinced him to put his data up on CHORUS. Yates advocates for such data sharing because he believes it will help propel the field of proteomics forward at a much faster pace and eventually bring about better health therapeutics.
His approach is revving up proteomics research at Pitt and elsewhere. In addition to his work with MacDonald and Sweet, Yates has worked with Robert Sobol, formerly a member of the University of Pittsburgh Cancer Institute, now molecular and metabolic oncology program director at the University of South Alabama Mitchell Cancer Institute. The two studied protein pathways involved in DNA repair. They hope to uncover previously unknown protein partnerships.
“When you do this on the entire proteome, you’re able to identify novel things, things you would not have thought about,” Sobol says.
“You know, Mother Nature is pretty funny; it tends to want to do what it wants, not what we think it does. So using this unbiased discovery method that Nathan created will help us find partnerships we had never considered.”
Yates is now performing the drug-target research he so badly wanted to do while at Merck as he collaborates with Lans Taylor, a PhD and director of the University of Pittsburgh Drug Discovery Institute and Pitt’s Allegheny Foundation Professor of Computational and Systems Biology. (Interestingly, Merck is now intrigued by the idea and is collaborating with Pitt and Yates.) Among other projects, Yates is looking at the widely used diabetes drug metformin, a medication whose molecular mode of action is not completely understood. He wants to use differential mass spectrometry to uncover what it binds to.
“It’s the most widely prescribed treatment for type 2 diabetes,” he says. (He notes that he’s “swinging for the fences” on the metformin effort but remains optimistic.)
“Finding the molecular target of metformin could lead to the development of new and improved treatments for patients,” he says.
In Yates’s vision of the future, his efforts (and those of others who are advancing the use of mass spectrometry) will help people take charge of their own health. Imagine, he says, if people could monitor their own proteomes on a daily or weekly basis. That would offer doctors and patients new information that could reveal, say, evidence of muscle damage or perhaps even cancer before a tumor forms.
Yates’s mentor in this area is Mike Lee, who also trained as an analytical chemist under Rick Yost and was with Bristol-Myers Squibb before leaving to become a consultant. Lee specializes in bringing together people in academia and industry to work on difficult problems in health care. He likens today’s health measurements—heart rate, height, weight, and the like—to dropping a fishing line into a murky pond and hoping for a bite. Yet being able to monitor one’s proteomic or metabolic profile would be akin to “draining the pool so you can see all the fish.”
Both Yates and Lee believe sports teams will be early adopters of this sort of personal health-monitoring technology. They imagine teams would be able to keep an eye on their players’ protein levels and determine whether anyone is showing signs of physical stress before a full-fledged injury occurs.
Explains Lee: “The coach might say to the pitcher, ‘Son, you look tired,’ and the pitcher will say, ‘I’m not tired!’ Now, what if before each inning, the pitcher spit into a cup, and Nathan was monitoring that, and he saw markers for stress? He could say to the coach, ‘Uh-oh, this guy’s about to blow out an elbow.’”
Lee notes that even if individual health tracking can’t stop a condition from developing, it is likely to offer advantages in terms of choosing a treatment. For instance, two people might get the same kind of cancer but need different combinations of medications to fight the disease as it evolves.
“In addition, most drug dosages are set for an ‘average human’—40-year-old white men,” Lee says half-jokingly. But with personalized proteomics, doctors could better tailor drug regimens to a petite woman or an Asian man.
How close is a future of personal proteome monitoring? Hard to say. The human proteome is a highly complex system; tens of thousands of proteins in our bodies interact with each other in myriad ways. Unraveling those mysteries won’t be easy. And giving clinicians and laypeople information they can act on is another issue.
But, Yates says, “It’s an exciting time to be working on some of these questions.”