In science, it is one thing to see, and another to be able to quantify. By dint of her spatial acuity and her use of unorthodox software tools, Erica Savig has accomplished the former. But the latter is where the real test lies. She is currently tackling it with Mathieu Tamby, a postdoctoral cell biologist who came to the IME last year.
In August, Savig and Tamby presented some of their research at an informal lab review. They showed still photographs and reconstructed videos of smooth vascular muscle cells under two conditions: some had been seeded in native collagen, which approximates a healthy extracellular matrix; others had been seeded in denatured collagen, which models the ECM’s deterioration in pulmonary arterial hypertension. Collagen is the most abundant protein in mammals and a major component of the extracellular matrix.
The pictures told a more or less straightforward story. As Savig had put it earlier, “You see already the cells are behaving differently. But how do you quantify this?”
Tamby cued a slide depicting some of the standard ways cell biologists characterize the behaviors he had just projected onto the screen. “These are classical ways of representing these parameters with numbers,” he said. “So we look at cell velocity on native and denatured collagen, and this was looking at the average velocity of one cell.”
Tamby’s graphs looked basically the same to an untrained eye—or to a trained one, for that matter.
“So this classical way of viewing numerical data isn’t quite capturing differences in behaviors that we’re seeing,” Savig interjected.
“But in science, we like numbers,” Tamby responded. “We like to obtain significant differences.”
For something as simple as average cell velocity, the methodology for making statistically significant comparisons is relatively straightforward. But what seemed to differentiate the two cases in question was occurring at a finer level of detail. Cells on native collagen sometimes appeared to form different patterns of alignment than cells on denatured collagen. Or they bent themselves into what looked like different sorts of shapes, which Savig would describe during brainstorming sessions as “scouts” and “oafs” and “middlemen”—sometimes sending Tamby, whose mother tongue is French, scrambling for the Web-based translator bookmarked on his browser.
Savig turned to the Rhino 3-D modeling software to translate some of these concepts into geometrical data. “Really, what we try to do as architects is take a lot of information and almost reduce it to a series of spatial relationships,” she says. “And in essence, that’s what biologists are trying to do as well. They’re trying to look at stuff and come up with some kind of relationship or rule set that explains what’s happening. And these digital techniques we use in architecture help us to reduce the information, filter through it, and find these relationships.”
The goal was to translate subtle differences in those spatial relationships into unique signatures betraying whether the underlying collagen mimicked a healthy or diseased extracellular matrix. If that was possible, it could spur a huge advance in pulmonary arterial hypertension diagnosis and treatment.
Pulmonary arterial hypertension (PAH) involves the progressive derangement of blood vessels in the lung, which overwhelms and exhausts the heart until it fails. There is evidence suggesting that PAH actually encompasses multiple diseases. At present, however, it is so difficult to distinguish between them that doctors effectively choose a drug treatment through guesswork. The diagnostic tools they use are primitive. “Right now, we have these catheters that give you pressure tracings, and X-rays that can show you two-dimensional and sometimes reconstructed three-dimensional images of what their lungs look like,” says Darren Taichman, the associate director of Penn’s Pulmonary Vascular Disease Program. “But what that doesn’t do for me is tell me very much about who should be treated which way.”
The FDA has approved six new drugs for PAH in a relatively short span, but there is a huge variation in how patients respond to them. “These drugs are toxic, like any other drug, and they’re unbelievably expensive,” Taichman says. “And when you’ve got a disease that kills people relatively quickly—and more recent data which suggest that the sooner we get people on therapy, the better they’ll do in the long run—it would really be nice if you could choose the right therapy sooner.”
What Taichman and Jones are searching for is a way to zero in on what’s happening on the cellular level. They think that whatever drives the cell changes which lead to blood-vessel derangement might be reflected in the general bloodstream. If so, it might be possible to add Savig’s smooth vascular muscle cells to a patient’s blood-tissue sample rather than collagen, and refine the diagnostic method to derive personalized signatures for PAH patients—which in turn could help doctors determine which treatment would likely be most effective.
Traditional clinical diagnostics revolves around detecting particular molecular substances in various body fluids or tissues. But the problem in a complex nonlinear system is that “there is no single molecule that really controls everything,” Peter Davies points out. “In fact, there are many biomarkers, many principal players, and the principal player at one time may be—a day later or an hour later or a minute later—no longer the principal player.” Jones has spent much of his career investigating the role of particular molecules in PAH—for instance, an ECM protein called tenascin-C that is critical in the formation of blood vessels in the lung. The LabStudio collaboration opened the possibility for a different strategy.
“There is a whole other realm of potential diagnostic tests that revolve around what you might call functional assays, or functional readouts,” says pathologist Mark Tykocinski. “So instead of trying to catalogue all those molecular drivers, you look at the other end: you look at the cell that’s affected by all those molecular drivers. In fact, you don’t even have to know all the molecular drivers—you just look at the functional readout.” Some functional readouts are easy to read, but the intricate choreography of cell movement is likely to overwhelm even a microscope-aided eye. “So having new tools to quantitatively analyze something like cell motility, that’s where the link to architecture weighs in.”
Looking at the smooth muscle cells, it had seemed to Savig (and Jones) that the movements of filopodia—slender projections of cytoplasma that extend from the leading edge of a cell wall—were important. So she resolved to trace the edges of cells in Rhino, creating a digital data set that she could manipulate and hopefully deduce rules from. But a big obstacle stood in the way: there was no system of mapping or mathematical coordinates that easily lent itself to measuring those wild, horn-shaped protrusions.
When Jones and Jenny Sabin initiated their partnership, they hoped that injecting an outsider’s mentality into the lab “would lead to both extraction and abstraction of new biological information.” That was exactly what Erica Savig was now on the cusp of doing. It would be hard to overstate the peculiarity of the tack she chose. She turned to a resource that no scientist in the English—or French—speaking world would ever have thought to consider, even after exhausting a thousand other ideas. It was a paper written in 2006 by a student of Sabin’s named Jackie Wong GAr’07. Presumably, even Wong himself never imagined that his work might have any relevance to cell biology. Titled “Dance and Space,” it was an analysis of the spatial properties of figure skating—or, to be more specific, the compulsory ice dancing event.
“The project tracked the movement of ice dancers,” Savig explains. “But not just how they moved together along a plane: also the movement of their hands and legs”—which are, in a sense, a type of bodily protrusion not that different from a cell’s filopodia.
She adapted some of Wong’s ideas to develop her own system for characterizing the irregular motion of cell shapes. More importantly, she and Tamby have been able to do so with sufficient mathematical rigor to convert some—though still not all—of the behavioral differences between their two experimental conditions into hard numerical data. Yet that leads to a further challenge. A tiny bit of tissue can yield vast quantities of data. So much, in fact, that it becomes hard to figure out how to handle it. But that too was why the architects had come into the lab.
An Architect Walks Into the Lab By Trey Popp