Microglia’s pruning function called into question

Microglia were once thought to have one job—as the brain’s resident garbage collectors. If neurons became damaged or diseased, microglia would spring into action, engulfing dead or infected cells and pumping up the local immune response. Between clean-up operations, scientists believed, they rested in a deep sleep.

In 2005, though, researchers got their first direct look at what microglia were doing in the brain, and they promptly tore up this cellular CV. The grainy live-cell imaging footage, published in Science, showed that the supposedly “resting” microglia were actually marauding around in the neocortex of adult mice, firing out processes and furtively feeling out the surrounding parenchyma.

“This, for me, was a game changer,” says Rosa Paolicelli, associate professor of biomedical sciences at the University of Lausanne, who was about to embark on a Ph.D. at the time. “People started to think about the physiological role of microglia. What do they do in the intact, healthy brain?”

The work kick-started two decades of research that has changed how the field classifies microglia, and led to new tools to help scientists define and scrutinize the cells’ functions in detail. Many studies have focused on “critical windows”—at the beginning and end of an animal’s life, Paolicelli says. During these time frames, microglia take on many side jobs: as sculptors of the developing brain, cultivators of new neural connections and fighters of neurodegeneration, for example.

Along with this rise in profile from garbage collector to cellular polymath has come controversy. “There are some players that are trying to bring forward some ideas that are a bit too simplistic or restrictive. And I feel it’s very dangerous not to stay open-minded regarding the implications of the findings, not to stay open-minded regarding the limitations of all these models,” says Marie-Ève Tremblay, professor of medical sciences at the University of Victoria, who studies microglial function in health and disease.

What microglia do during early development has proved particularly contentious. At least three mouse studies published this year have called into question the extent to which microglia snip away synapses to shape a budding brain.

These findings continue to refine the cells’ task list and work schedule, supporting the idea that their functions depend on timing and context, and opening new avenues for research, says Beth Stevens, associate professor of neurology at Harvard Medical School. “We need to really get mechanistic and really start to think about developing the tools and the ways to be able to manipulate the right pathway at the right time to ask specific questions,” she says.

S

oon after the 2005 Science paper came out, Stevens was a postdoctoral researcher in the late Ben Barres’ lab at Stanford University, studying developing visual circuits in newborn mice. Immune molecules called complement proteins appeared to have a role in shaping these circuits by “tagging” redundant synaptic connections for elimination, the team reported in 2007 in Cell. And microglia, a 2012 follow-up showed, devoured the tagged synaptic connections, helping to sculpt the brain.

These results unmasked a whole new identity for these cells—like finding the Batcave under a microglial Wayne Manor. But recent work has drawn some of the cells’ superhero status into question. For instance, mice depleted of microglia starting 14 days after birth develop no discernible problems with visual performance, according to a July Nature Neuroscience study.

“We didn’t find any difference for any of the measurements we made across the different experimental approaches we used to examine the functional aspects of circuitry,” says the study’s lead investigator, Aaron McGee, professor of translational neurosciences at the University of Arizona. “Microglia perform immuno-surveillance; they do not sculpt brain circuits.”

But the new findings don’t directly contradict the earlier work, says Dori Schafer, associate professor of neurobiology at the University of Massachusetts Chan Medical School. For example, in 2012, Schafer, Stevens and their colleagues found evidence of microglia sculpting circuits in the lateral geniculate nucleus of the thalamus—a region McGee’s team did not explore, leaving open the possibility that sculpting might still occur there.

“We’ve never said that microglia and complement regulate pruning through the entire brain,” Schafer says, noting that the 2012 paper describes microglial pruning in only one neural circuit at one point in time. Regardless, it has been cited more than 3,500 times, according to Google Scholar, and dozens of these articles, which The Transmitter reviewed, frame synaptic sculpting as one of microglia’s key roles throughout the brain.

“I think that people use terms like synaptic pruning sometimes too generally,” Stevens says. “Or if they’re using just a general readout and they say that’s pruning without doing the key experiments.”

O

ne such key experiment would be to show microglia devouring synapses in real time—footage that still has never been recorded, says Laetitia Weinhard, a postdoctoral researcher in Dan Littman’s lab at New York University who previously spent five years as a Ph.D. student trying to capture this elusive event. Using multiphoton microscopy and electron microscopy, she took high-resolution videos and snapshots of microglia swarming around synapses in the developing mouse hippocampus.

Despite close contact, the microglia she saw engaged in little more than the occasional “munching” of parts of the presynaptic compartment—a process called trogocytosis. This distinction is important, Weinhard says; classic microglial phagocytosis, by contrast, would involve the engulfment of the entire axonal structure.

Other common imaging techniques have also proved too imprecise to catch microglia in the act, Weinhard adds. She is particularly critical of colocalization analysis experiments, in which light microscopy images that show overlap between proteins fragments and microglia are used to suggest the cells have consumed the molecules. But the light microscopy in these studies has poor axial resolution, Weinhard says, and so the “consumed” dots of protein could simply be above or below microglia. “Because the depth resolution is bad, it just looks like it’s inside, but it’s not,” she says.

Weinhard, despite not finding any synapses devoured by microglia, did find some that appeared to be formed by microglia. “Very often when [microglia] would get close to a dendrite, they would have this sprouting of filopodia,” she says.

Independent evidence supports this idea, Weinhard says. Microglial contact with dendrites caused new filopodia to sprout in the developing mouse somatosensory cortex, according to a multiphoton imaging study. And conversely, mice that lack microglia didn’t show the increase in dendritic spines in the motor cortex that typically accompany learning a new motor task, such as treadmill running, according to another study. “This increase in spine formation was reduced or impaired without microglia,” Weinhard says.

O

ne issue McGee’s study faces is that the compound the team used to deplete microglia, PLX5622, actually kills them, leaving a swath of cellular corpses throughout the brain. “This has huge outcomes on other microglia that remain, huge outcomes on astrocytes, on oligodendrocyte precursor cells, on so many other cell types”—in a way that could affect the team’s conclusions, Tremblay says.

A mouse model genetically engineered to lack microglia from birth avoids these limitations. Csf1r(delta FIRE) mice show no changes in hippocampal synapse number or synaptic spine density, according to a study published in April in EMBO Reports. And many other aspects of the animals’ early neurodevelopment remain unchanged, according to a preprint posted on bioRxiv in September.

“There’s no phenotype. It’s amazing,” says preprint investigator Clare Pridans, an immunologist at the University of Edinburgh who co-developed the Csf1r(delta FIRE) mice. Based on what she has seen, she says she does not think microglia sculpt the brain.

In response to The Transmitter’s request for comment on this preprint, Stevens and Schafer provided a joint statement, in which they offered several alternate hypotheses. “Microglia have been shown to play roles in synaptogenesis and pruning. The net effect without microglia could be what the authors observed,” they wrote.

What’s more, microglial pruning in the mouse barrel cortex was apparent only after Schafer’s team removed the animals’ whiskers, according to a 2019 study. “It is possible that one needs to push the system to see effects,” she and Stevens wrote in the statement. They also added that Csf1r(delta FIRE) have a different genetic background from the mice tested in earlier studies, which might suppress the magnitude of any pruning.

Other cell types, possibly directed by the same complement mechanism that Stevens identified in 2007, could be taking up microglia’s pruning function in these animals, Stevens and Schafer wrote in their statement.

This is supported by the April EMBO Reports study, which noted an increased uptake of synaptic material by Csf1r(delta FIRE) mice’s hippocampal astrocytes. But these mice show no transcriptional upregulation of any other brain cell types, according to a second study.

“At this point, it is not clear to what extent astrocytes may be taking over this role in the absence of microglia,” says David Munro, a postdoctoral researcher in Josef Priller’s lab at the University of Edinburgh and an investigator on both studies.

The Csf1r(delta FIRE) mice do show unusual brain calcifications, particularly in the thalamus, the study finds. And when crossed with mice that develop Alzheimer’s-like neuropathology, the animals show early and lethal vascular pathology—a change that microglial transplants prevent. That finding chimes well with genetic evidence that implicates microglial genes in neurodegenerative disease.

Overall, microglial research is, by most metrics, in good health, but the current controversies leave its foundations vulnerable to critique. “The tension on the string of logic that connects findings that yield to an eventual interpretation of a model should be really taut,” McGee says. “And when there’s slack in that line, I think it’s inherent on the field to try and sort that out.”