In 1992, Shatz moved on to the University of California, Berkeley, with her then husband, a fellow neuroscientist. The marriage was in trouble — undermined by the pressures of academic life and a failed course of fertility treatments — and it ended soon afterward.
Despite the turmoil, Shatz continued to plumb the mysteries of LGN development. And in 1994 she launched the series of experiments that would overturn long-held assumptions about the brain’s splendid isolation from the immune system.
Shatz wanted to know what genes were active in the LGN when the retina sent its prenatal signaling waves. So she and her postdocs used tetrodotoxin to halt the retinal signals in fetal cats. Then they compared the expression of thousands of genes in two samples by measuring levels of messenger RNA. (The DNA in genes code for mRNA, which, in turn, codes for proteins.) This test determined that only a few genes switched off when signaling was blocked — and one of them was the gene known to code for MHCI.
Shatz and her team were nonplussed. They knew that MHCI wasn’t supposed to be expressed in a healthy brain — only in an injured or diseased one, when the blood-brain barrier had broken down. Yet because none of them was a specialist in neuroimmunology, they didn’t quite grasp the outrageousness of their finding.
“We were blissfully ignorant,” Shatz says, “of the fact that we had run into a dogma.”
Glue for the Brain
There are several reasons why nobody before Shatz had found genes coding for MHCI in normal neurons. In part, her advantages were technical: a more comprehensive screening test than brain researchers usually employed, and a more sensitive type of follow-up testing. But perhaps most important, other scientists hadn’t been looking in the right place at the right time.
Outside the brain, MHCI proteins and the genes coding for them are hard to miss. MHCI molecules perch on the surface of most of the body’s cells, displaying peptides — snippets of proteins — that reveal what’s going on inside. They present this evidence to killer T cells, immune-system agents that hunt for byproducts of viral infections, cancerous mutations or anything else that doesn’t register as “self.”
Each killer T cell carries thousands of receptors programmed to recognize a particular foreign peptide. If it encounters such a snippet, the T cell uses chemical weapons called cytotoxins to destroy the host cell — and the invading pathogen, as well.
Inside the brain, MHCI is far more elusive. It appears at lower levels, and only in certain areas at any given moment. In fact, Shatz observed no actual MHCI proteins when she first examined those fetal cat LGNs; what she found were biochemical signs indicating that MHCI genes were active. A less intrepid scientist might have chalked up the initial “hit” as a false positive. But Shatz pressed further.
First, she and her postdocs tested for MHCI gene expression at different stages of LGN development. They found that levels peaked as the right-eye-left-eye layers finished segregating, and fell sharply afterward. Then the team searched elsewhere. They detected little or no MHCI expression in some areas of the brain, but they found it in several other places, including the visual cortex while the ocular dominance columns were forming, and in the hippocampus — an area of the brain associated with learning and memory — at all ages.
Next, the team used tetrodotoxin to block nerve impulses in a kitten’s eye when it was 6 weeks old, during the critical period of development. When nerve impulses fell, MHCI expression decreased as well.
To see what would happen when nerve cell impulses were heightened, Shatz and postdocs injected rats with a drug that causes seizures. MHCI expression in the hippocampus and cerebral cortex soared. When Shatz and her team tested for actual MHCI proteins (as distinct from switched-on genes) in slices of rat brain from those same regions, the results were positive.
Pondering these findings, Shatz developed a tentative theory about what MHCI was doing beyond the blood-brain barrier. Under normal circumstances, it clearly wasn’t mediating immune function. Instead, it seemed to be associated with remodeling the brain.
MHCI levels rose in areas undergoing changes and fell when remodeling was complete or when a shot of tetrodotoxin interrupted it. Shatz hypothesized that MHCI molecules might function as a kind of “synaptic glue,” stabilizing contacts between neurons once appropriate connections had been established.
Shatz expected these controversial notions to meet with some resistance once she made them public. Still, she was stunned when the prestigious journal Nature rejected the paper in which she reported her team’s findings. “They said we must have done something wrong,” she recalls, “and we should go back and figure out what it was.”
In college, Shatz had been a member of the Radcliffe ski team. The experience taught her a lesson that she later applied to her scientific career: “When you’re on a really steep slope, and you can’t see the bottom, you have to just push off and trust your own abilities.” So instead of packing up her metaphorical skis and going home, she sent the MHCI paper to Neuron, whose editors could find no flaws in the methodology. It was published in September 1998.
Many of Shatz’s colleagues thought the results were crazy, but others offered her equipment and materials to pursue this daring new line of inquiry. If MHCI proteins — so crucial to immune function — also helped shape neural circuitry, the implications were profound.
To begin with, what would happen to developing brains if these molecules went awry? Shatz got a hint when she looked up ailments associated with mutations in genes coding for the human version of MHCI. The list included schizophrenia and autism. Both disorders were also associated with maternal infections during pregnancy.
In genetically vulnerable individuals, Shatz wondered, could a mother’s flu — and her immune reaction to it — disrupt MHCI’s action in the fetal brain?
To answer such questions, Shatz would first need to investigate how the brain might function without any MHCI at all. Fortunately, an immunologist at Berkeley, David Raulet, had bred a line of mice incapable of expressing functional MHCI proteins. Raulet donated some of these so-called “knockout” mice to Shatz, who found two striking anomalies.
The first anomaly confirmed her hunch that MHCI helped fine-tune circuitry: In the absence of the immune molecule, the animals’ LGNs never developed layers.
The second abnormality was, literally, a shocker: The knockout’s neurons responded differently to electrical stimulation than neurons in mice that were unimpaired. In ordinary mice, persistent low-frequency stimulation to the hippocampus weakens synaptic connections between neurons; high-frequency stimulation strengthens them. In mice lacking MHCI, both kinds of stimulation made connections stronger.
So maybe MHCI didn’t really provide glue for neural synapses. Maybe it did the opposite, telling neurons in the adult brain when to let those synaptic connections weaken — and, ultimately, be pruned away. That would allow circuits (like the LGN’s layers) to stabilize, instead of perpetually change. With the ability to put the brakes on plasticity and halt its own runaway growth, the brain could organize itself efficiently.
But faulty MHCI expression could wreak havoc. Before birth, it might lead to imprecise synaptic pruning, leaving too many connections and setting the stage for autism or schizophrenia. Later in life, overactive MHCI expression might cause excess pruning, leading to the loss of connectivity seen in neurodegenerative diseases such as Alzheimer’s, Parkinson’s, ALS and multiple sclerosis. Overactive MHCI expression might also drive the harmful synaptic pruning that follows injury to neural tissue after stroke or spinal cord damage.
Many neurological ailments involve both the elimination of synapses and the inability to form new ones, resulting in a devastating loss of plasticity, as illustrated by the futile struggle of Shatz’s grandmother to regain her language and motor skills after her stroke. “I hope maybe someday we can make a pill that will enhance plasticity. Maybe we can learn to manipulate these MHC Class I molecules or their receptors at will.”
Removing the Brakes
Shatz, then 52, had just been hired by her alma mater, Harvard Medical School, as the first woman to chair the neurobiology department. Over the next few years, she began to fill in the details of how MHCI works in the brain at the molecular level, and how it affects animal behavior.
One important task was to learn how MHCI proteins deliver their instructions in the brain. The most likely scenario, Shatz believed, was that the molecules sat on a neuron’s postsynaptic terminals, where electrochemical signals from other neurons are received. MHCI would send messages back across the synapse — orders like, “Please stop trying to connect.”
To find evidence for the phenomenon, her team began looking for MHCI receptors in the brain — literally, molecules that would connect with MHCI to help get the message through. And in 2006, they identified a likely candidate: PirB, a molecule found in some immune system cells, which also turned out to be present in neurons.
If PirB partnered with MHCI in neural circuits, Shatz thought, the absence of either molecule should have similar effects on the brain. That theory was confirmed by experiments in which her postdocs removed one eye in mice bred to lack PirB. In adult knockouts as well as juveniles, the ocular dominance columns associated with the missing eye expanded much more markedly than in ordinary mice. Whenever you removed the MHCI receptor, the brakes were lifted and plasticity took off.
To see how those qualities played out behaviorally, Shatz gave her lab mice some challenging tasks. First she deprived juveniles of vision in one eye so that the corresponding brain cells failed to make connections; once the mice reached maturity, they were put in a water maze that required them to recognize a pattern of fine lines to find a floating platform. MHCI knockout mice, with no MHCI and more plastic brains, did far better than their ordinary counterparts.
In another experiment, using animals with unimpaired vision, mice unable to produce MHCI were quicker to learn how to keep from falling off a rotating rod; they also remembered the trick for longer time periods. These animals, Shatz said, “performed like Olympians.”
A downside, of course, was that the MHCI knockouts were seriously immunocompromised. Mice without MHCI were also more vulnerable to epilepsy, an unfortunate side effect of untrammeled neuroplasticity. Nor could anyone be sure that their unusual strengths in some cognitive areas weren’t balanced by unusual weaknesses elsewhere; in animals that can’t speak, such tradeoffs are hard to gauge. What was certain was that Shatz opened a vast field of exploration.
If immune molecules acted as brakes on plasticity, those brakes might be pharmacologically adjusted. For a host of neurological ills, Shatz says, that meant “a new kind of hope.”
New Avenue for Therapy
In 2007, Shatz returned to Stanford. She’d been hired to run Bio-X, an interdisciplinary program meant to foster collaboration among life scientists, medical scientists, computer scientists, engineers and physicists. The program’s home, the James H. Clark Center, is an aggressively biomorphic structure whose two curving wings could be taken for the hemispheres of a giant brain.
In her office (in the left hemisphere), Shatz’s face lights up as she describes her latest collaboration. It began when she and Rona Giffard, an anesthesiologist and a longtime stroke researcher, were whispering in the corner at a boring staff meeting.
“I said, ‘We’re working on these really cool molecules, and if you knock them out, the animals have more brain plasticity,’ ” Shatz recalls. Giffard was fascinated, and soon the pair designed a series of experiments aimed at advancing the quest for the “plasticity pill” Shatz had long dreamed of.
The team took mice lacking MHCI and had them perform athletic feats: balancing on a spinning rod, crossing a horizontal ladder. Then they gave the animals strokes, cutting off blood to the artery supplying motor and sensory areas of the brain. Days later, when these knockout mice were given the same tasks, the animals regained their expertise faster and more fully than ordinary mice with strokes. What’s more, lab tests indicated that while the areas of brain damage started off the same for both sets of mice, the damage diminished more for the knockout mice over the following days.
“This opens a new avenue for therapy,” says Shatz, whose team published their findings in Neuron in March 2012. Right now, the only widely available treatment for preventing brain damage from stroke is tissue plasminogen activator (tPA), which breaks up blood clots; it must be given within a few hours to be effective, and though it limits initial damage, it doesn’t help the brain restore lost synapses or form new ones.
Much of a stroke’s destruction, indeed, occurs after the initial injury, as even healthy synapses within the affected area are pruned away — a process apparently mediated by MHCI proteins. If a pharma company can find a way to temporarily disable those molecules (most likely by blocking their receptor, PirB), that secondary damage might be avoided, and the brain might also do a better job of repairing itself.
Shatz knows it may take many years to develop such a drug, but she’s savoring the proof-of-concept moment. Her only regret is that her grandmother can’t be here to witness it. “To think that this could be applied clinically …” she pauses, remembering her beginnings.
“It’s kind of like a big, wonderful circle.”
[This article originally appeared in print as "Brain Benders."]