Read our COVID-19 research and news.

Beth Stevens's work on how microglia (green cells) prune connections between neurons (purple) may help explain conditions ranging from schizophrenia to Alzheimer's disease.
Ken Richardson

This woman may know a secret to saving the brain’s synapses

In 2010, neurobiologist Beth Stevens had completed a remarkable rise from laboratory technician to star researcher. Then 40, she was in her second year as a principal investigator at Boston Children’s Hospital with a joint faculty position at Harvard Medical School. She had a sleek, newly built lab and a team of eager postdoctoral investigators. Her credentials were impeccable, with high-profile collaborators and her name on an impressive number of papers in well-respected journals. 

But like many young researchers, Stevens feared she was on the brink of scientific failure. Rather than choosing a small, manageable project, she had set her sights on tackling an ambitious, unifying hypothesis linking the brain and the immune system to explain both normal brain development and disease. Although the preliminary data she’d gathered as a postdoc at Stanford University in Palo Alto, California, were promising, their implications were still murky. “I thought, ‘What if my model is just a model, and I let all these people down?’” she says. 

Stevens, along with her mentor at Stanford, Ben Barres, had proposed that brain cells called microglia prune neuronal connections during embryonic and later development in response to a signal from a branch of the immune system known as the classical complement pathway. If a glitch in the complement system causes microglia to prune too many or too few connections, called synapses, they’d hypothesized, it could lead to both developmental and degenerative disorders.

Some people are drawn to science because of the challenge—they prefer pain, and want to suffer. Beth really enjoys what she’s doing, which makes her a joy to work with.

Cagla Eroglu, Duke University

Since then, finding after finding has shored up and extended this picture. This year alone, Stevens and her collaborators have published papers in Science and Nature linking the complement pathway and microglia to diseases such as schizophrenia, Alzheimer’s, and cognitive problems from infection with West Nile virus. A study on Huntington disease is forthcoming, Stevens says. Although some scientists say that such research is unlikely to produce therapies any time soon, clinical trials of antibodies that block the complement system in the brain could start for glaucoma and other neurodegenerative diseases by the beginning of 2017. Stevens’s decision to stick with her hypothesis says neuroimmunologist Richard Ransohoff of the biotech company Biogen in Cambridge, Massachusetts, has “worked out spectacularly.” 

A force of nature

Athletic, with a mop of blond curls, Stevens has piercing blue eyes that seem capable of knocking a glass off a table with sheer concentration. “She’s like a four-shot espresso,” says Cagla Eroglu, a neuroscientist at Duke University in Durham, North Carolina, who met Stevens at Stanford, where both completed their postdoctoral training. 

Downing a Diet Coke in her office in the Center for Life Science at Boston Children’s Hospital, Stevens gestures at a large whiteboard, where she has scribbled a list of projects and grant applications “to keep track of what’s cooking” for her and the 14 postdocs, graduate students, and technicians in her lab. At any given point, one team in her lab may be looking for molecular triggers of the complement system while a second observes microglia in vivo and another investigates why certain types of synapses get pruned more often than others. Stevens’s many-pronged strategy is a smart move that keeps her lab productive, and “can open up many new directions for the field,” says Eric Huang, a neurobiologist at the University of  California, San Francisco. 

From her office window, Stevens can almost see her hometown of Brockton, Massachusetts, where she went to public school. Her father was a principal, her mom an elementary school teacher. “Even all their friends were teachers,” she says. “I think my parents are still wondering what happened there.”

Stevens didn’t get interested in science until her senior year in high school, when her Advanced Placement biology teacher told stories about his other job in a clinical microbiology lab. At Northeastern University in Boston, she followed his example and took a job in a hospital laboratory. Her favorite case involved an episode of food poisoning that she helped tie to a sausage contaminated with the Listeria monocytogenes bacterium. Although Stevens planned to be a physician, she realized then that she was more attracted to research. “I wasn’t really interested in hanging out with the patients as much as figuring out what was wrong with them.” 

The complement protein C4 (green) often overlaps with synaptic markers (red and white dots) 
in this culture of neurons (blue marks main cell bodies), a sign of how it may flag synapses 
for pruning in brain development and disease.
Heather De Rivera/McCarroll Lab/Harvard

As Stevens approached graduation in 1993, her professors told her to go to the National Institutes of Health (NIH) for more research experience. When her new husband got a chance to work in Washington, D.C., she went with him, determined to get a job at the NIH campus in nearby Bethesda, Maryland. 

Stevens trekked to NIH weekly to scan sheets of job postings—“This was before internet job listings,” she says. She began waiting tables at a nearby Chili’s so she could easily dash over to NIH to check for new jobs. Months passed; her CV languished. One day, neuroscientist Douglas Fields, who had a habit of leafing through the rejected CVs submitted to NIH, cold-called Stevens and offered the 22-year-old a job as a technician. Even though she was “totally green,” she says, he soon made her the manager of his lab at the National Institute of Child Health and Human Development in Bethesda. 

Fields was interested in how brain activity increases the expression of certain genes in neurons, including one encoding an adhesion protein called L1. This molecule helps cells called glia wrap the wirelike neuronal projections known as axons in layers of fatty insulation, or myelin. Stevens spent hours in the lab trying to model the process in a dish. Eventually, she succeeded. Fields listed her name as a co-author on the resulting paper—a rare honor for a technician (Science, 24 March 2000, p. 2267). And Stevens was left with a passion for glia, cells that neuroscientists had long viewed as “housekeepers,” passively providing neurons with nutrients and sponging up excess ions and neurotransmitters. 

Stevens decided to get a Ph.D. in neuroscience at the University of Maryland, College Park, while continuing to work at NIH. Upon finishing, she returned to full-time work as Fields’s technician and lab manager. By that point, she had caught the attention of Story Landis, then-director of the National Institute of Neurological Disorders and Stroke (NINDS) in Bethesda. The NINDS head told Stevens she was “crazy” not to pursue a career as an independent scientist, advising her to do a postdoctoral position elsewhere and offering help with contacts.

Barres, one of the world’s leading glia researchers, ultimately invited Stevens to join his lab. She couldn’t have landed in a better place to pursue ambitious, big-picture science, Ransohoff says. Although there are many ways to motivate people, he explains—“Love the hell out of them, scare the hell out of them, and work the hell out of them”—Barres “supports the hell out of them with resources, advice, and scientific direction when needed, then lets them go.” 

Stevens was “a force of nature,” Barres recalls. “She was always working late at night, on holidays, nights, weekends.” She also revved up the lab’s social life, organizing happy hours and parties for everyone’s birthdays, Eroglu says. “Some people are drawn to science because of the challenge—they prefer pain, and want to suffer. Beth really enjoys what she’s doing, which makes her a joy to work with.”

Barres and his team had long studied star-shaped glia in the brain called astrocytes, which secrete chemicals that can influence neuronal growth. Karen Christopherson, also a Barres postdoc, discovered another remarkable property of astrocytes: They appear to induce neurons to massively increase their production of a protein called C1q. Elsewhere in the body, C1q is known to trigger the complex molecular cascade of the classical complement pathway.

Weeding out the weak

Two types of immune cells, plus complement proteins, work together to prune less active synapses in brain development—and this process may abnormally reactivate during some diseases.

1  Signaling through TGF-b, astrocytes in the developing brain induce neurons to make C1q.

2  C1q initiates the complement cascade, which marks weak or superfluous synapses with C3
and other proteins.

3  Microglia ingest, or prune, complement-tagged synapses, leaving the strongest connections.



V. Altounian/Science

Among other roles, the complement system helps label pathogens and damaged cells as cellular trash throughout the body, affixing them with protein tags that serve as an “eat me” signal for immune cells called macrophages. Christopherson’s finding led Barres and Stevens to wonder whether the complement system also plays a role in a key process as the brain develops in the womb and after birth: tagging and pruning back the thicket of newly formed synapses and leaving only functional connections. If C1q were necessary for proper pruning, they hypothesized, synapses in mice without the protein should be disrupted. 

Stevens and Barres obtained mice in which the gene for C1q had been knocked out, then looked for alterations in a deep region of the brain’s visual system called the visual thalamus. Before a newborn animal has even opened its eyes, neurons in this region undergo massive pruning of synapses, leaving a neatly organized system in which most cells receive inputs from only the right or left eye.  

The mice lacking C1q didn’t display any obvious visual abnormalities, Stevens says. But they had too many neural connections in a key relay center of the visual pathway, the lateral geniculate nucleus (LGN), showing that C1q was necessary for synaptic refinement. The protein is virtually absent in the neurons of healthy mice with mature brains, suggesting that it plays only a fleeting role early in brain development. In a mouse model of glaucoma, however—a disease in which neurons of the retina are destroyed—Stevens showed that C1q levels were much higher than normal. The findings, reported in Cell in 2007, “were really novel, and set the stage for the whole field” to take a closer look at the role of the complement in brain development and function, Huang says. 

In later studies of the C1q knockout mice, Stanford neuroscientist David Prince, working in collaboration with Stevens and Barres, found that the animals’ hyperconnected neural wiring in the cortex makes them prone to seizures, memory loss, and other cognitive deficits. Stevens and her colleagues focused next on what was doing the pruning in the postnatal brain. A movie clip created by a fellow Barres postdoc, Axel Nimmerjahn, recorded through a sheer window in a mouse’s skull, hinted that microglia—which play the role of microphages in the brain—were responsible. The cells continuously extended and retracted slender protrusions as if actively exploring. Stevens had never seen any other cell move so purposefully. She was smitten. “I mean, I loved astrocytes, but they don’t do that,” she says. 

Up to that point, Stevens and many others had ignored microglia because they were thought to arrive too late in the brain to affect neurodevelopment. A group led by Miriam Merad at Mount Sinai School of Medicine in New York City, however, demonstrated that microglia begin to populate the brain within days of gestation (Science, 5 November 2010, p. 841). That made them “perfect candidates” to conduct early synaptic pruning, Stevens says. Microglia were also the only known brain cells with a receptor for C3, a downstream product of the complement cascade. 

Finding a smoking gun

After completing her postdoc position with Barres, Stevens accepted the job offer from Boston and headed back east, intent on putting together the pieces of the puzzle. Finding a way to test whether microglia actually were ingesting pieces of synapses in the living brain was her first challenge. It occurred to Dori Schafer, one of the first postdocs Stevens hired, to combine mice genetically engineered to make their microglia glow bright green under ultraviolet light with a system that Barres and Stevens had used to tease apart retinal projections in the LGN. The system made synapses connected to one eye appear red and those linked to the other eye blue. All Schafer had to do was look for bits of red and blue synapses inside the green microglia. One Saturday afternoon, the first results rolled in. “I still remember the first cell I saw with bits of presynaptic terminals inside of it,” Schafer says.

Scientists had long known that neuronal activity strengthens synapses whereas less active synapses are eliminated, and Stevens and others had predicted that microglia would go after a neuron’s weaker connections. To test that hypothesis, Schafer applied pharmacological agents to the eyes of developing mice to increase or decrease the firing activity of neurons in one eye, and found that the less active synaptic connections were more aggressively eaten and pruned by microglia. She also used mice that lacked the complement receptor in microglia, and discovered that this reduced the rate at which the cells devoured synapses. The mice also had more synapses than controls, similar to the C1q knockouts,
she says.

While Schafer and Stevens were writing up these findings, a competing group led by Cornelius Gross of the European Molecular Biology Laboratory in Heidelberg, Germany, turned up the heat by publishing a conceptually similar paper in Science. The study suggested that microglia have a role in synaptic pruning in the hippocampus, but pointed to a different immune-related protein called fractalkine—which, among other roles, shepherds microglial migration around the brain—as a key player. In mice lacking a receptor for this protein, maturation of neuronal connections was delayed. Gross’s work, however, didn’t point to a clear mechanism for the pruning. Schafer and Stevens hastened to publish their complement work in Neuron. Their contribution was particularly provocative,  Gross says, because the classical complement system had long been known to be involved in the ingestion of pathogens and dead or damaged cells. “It’s really like a smoking gun.”

Even now, however, no one has definitively shown that microglia eat synapses in a living animal’s brain. The evidence is circumstantial, Gross notes—from before-and-after shots showing either microglia hovering near a synapse or microglia that have engulfed pieces of the synapse. It’s not clear in those images whether the microglia actively ate the synapses or merely gobbled up pieces that had already weakened or fallen off. But Schafer, now an assistant professor at the University of Massachusetts Medical School in Worcester, thinks she may have a way to catch pruning in action. She is studying the mouse’s barrel cortex, the part of the rodent brain that is wired to the animals’ whiskers. “If we manipulate one whisker, we know exactly where to look” for synaptic changes, she says, which should increase the odds of capturing images of microglia in the act of eating synapses. 

A new way to treat brain diseases

Although it's now well accepted that C1q is necessary for proper neuronal wiring early in life, evidence is mounting that the molecule can be detrimental later on. As mice and humans age, C1q levels rise in their brains up to 300-fold, Barres has found. Reducing its levels or blocking its ability to start the complement cascade limits cognitive and memory decline in aging mice compared with untreated controls, Stevens and others have further demonstrated. 

Studies of human disease also hint that the complement can trigger harmful synapse loss. In January, geneticist Steven McCarroll at Harvard Medical School and the Broad Institute in Cambridge, Massachusetts, reported evidence that the C4A gene, which encodes a complement protein downstream of C1q, may contribute to the synapse loss and brain tissue thinning that characterizes schizophrenia. After analyzing genome data from more than 64,000 people, they found that a subset of those with the mental disorder were more likely than controls to have an overactive version of C4A. When McCarroll and Stevens teamed up with Harvard’s Michael Carroll, who knocked out the mouse version of this gene, they found reduced synapse pruning during postnatal development in the altered rodents. Gross and others hailed the potential new schizophrenia mechanism, published in Nature, as a major advance.

The next month, in Science, Stevens described results suggesting that overhungry microglia are responsible for the early loss of synapses in Alzheimer’s disease (Science, 6 May, p. 712). In several mouse strains bred to produce excessive amyloid, a protein that forms plaques in the brains of people with Alzheimer’s, abnormally high levels of C1q set off a microglial feast, which destroyed functional synapses long before plaque formation and symptoms of cognitive impairment set in. That pattern of decline is consistent with observations that synapse loss is a more powerful predictor of Alzheimer’s symptoms than amyloid plaques, and “brings into light what’s happening in the early stage of the disease,” says Jonathan Kipnis, a neuroscientist at the University of Virginia School of Medicine in Charlottesville.

Most recently, in Nature, a collaboration with Stevens’s group led by neurobiologist Robyn Klein at Washington University in St. Louis in Missouri demonstrated in mice that the classical complement pathway also revs up during recovery from infection by West Nile virus, driving microglia to engulf synapses at a dangerous rate. That could help account for the chronic memory impairments that more than half of people experience after infection, says Huang, who calls the research “fascinating.” 

Such findings have piqued interest in targeting C1q clinically. Annexon Biosciences, a South San Francisco, California–based company, is leading the way. Co-founded by Barres back in 2011, after mouse data suggested blocking C1q could be beneficial for multiple neurodegenerative and autoimmune diseases, Annexon has developed several antibodies that can bind and block the action of the complement protein.

The company, in which Stevens is a shareholder, plans to launch human clinical trials of the drugs in people with Alzheimer’s, Huntington, and glaucoma by next year, but Huang and others warn that the drugs may not make a dent in more complex disorders such as schizophrenia. Another challenge will be to show that the antibodies really are preserving synapses. One tool may grow from a study published last month in Science Translational Medicine in which a research team showed that it could use a positron emission tomography scan to quantify synapse numbers, and loss, in living people. That could be “really important, Stevens says.  

Now past the crucible of starting her own lab, Stevens’s anxieties of just a few years ago have dissipated. She’s confident in her science and settling into a new role as a mentor, seeking to repeat what others once did for her. Stevens has “already transferred” the Barres-style incubator environment to Boston, and produced exceptional “scientific grandchildren,” Ransohoff says. 

Stevens recently started another happy hour—this one for other junior principal investigators in the Boston area. When people see other people launching their careers, and it looks effortless, “they don’t know what you’re really going through,” the former Chili’s waitress and lab technician says. “When you get a chance to get together and have a beer, you realize we’re all going through the same thing.”


Read more from our Neuroimmunology special package.