A cubic millimeter is, by all accounts, tiny. It’s barely noticeable–a speck or fleck or crumb. But look closely enough and you can uncover an entire world inside a particle of material. A team of neuroscientists and engineers, aided by machine learning tools, have charted a cubic millimeter volume of the human brain at nanoscale resolution, tracing every neuron, synapse, blood vessel, and supporting cell within the fragment and reconstructing a 3D model of the tissue. Though it represents just one-millionth of the total brain volume, it is the most detailed map of a piece of human brain matter ever created. It could spur a wave of scientific discovery about neurological disorders, brain structure, and the origins of our behavior.
“In one respect, our data set is miniscule,” Jeff Lichtman, co-senior study researcher and a neuroscientist and professor of molecular and cellular biology at Harvard University, tells PopSci. “But it doesn’t feel small because when you get in it, you see it’s like a gigantic forest. It’s a very tiny forest, but it’s a very, very, very complicated forest,” he adds.
All that complexity is on display in a study documenting the construction of this bit of comprehensive brain map or “connectome,” published May 9 in the journal Science. The first connectome was of a nematode brain, completed in 1986. Since then, neuroscientists have continued to plot out increasingly large and complicated brains–including those of fruit flies, maggots, a tadpole, and an earthworm. Yet human brains pose a unique mapping challenge in their intricacy and inaccessibility. The new, partial human connectome is available online for anyone to explore.
“Not only is this an impressive technological feat, this is a tool and a resource that is really aimed at sharing with the world and getting all of this scientific information out there,” Tim Mosca, a neuroscientist at Thomas Jefferson University who was uninvolved in the new work, tells PopSci. “This group has done an amazing job designing all of the new tools and the pipelines to make this available to anyone who wants to look at it, wants to think about it, wants to use this in their research.”
Serving up brain pizza
The study sample was collected over a decade ago from an anonymous patient undergoing epilepsy surgery. The surgeon removed a small piece of the temporal lobe to access and treat an underlying lesion, quickly preserved the tissue, and later shared it with scientists. Though the total volume of the fragment is about 1 cubic millimeter, it is not cube shaped. Instead, “it’s like a thick piece of pizza–but it’s not that thick,” says Lichtman. This blunt, triangular chunk, longer than it is wide, enabled the researchers to capture a bit of all six layers of the 3mm thick cerebral cortex.
The first step to mapping the brain pizza was to slice it into 5,019 individual cross sections (each 30 nanometers thin) mounted on tape using a specially designed machine that cuts with a diamond knife. From there, the researchers spent a full year carefully imaging each slice via electron microscope. Then, they digitally aligned and stitched together the slices and used multiple machine learning tools to fill out the 3-D form and label and color each component.
The segment’s neuron density is 16,000 neurons per cubic millimeter– about one-third lower than a previous density estimate of the same brain section and 10 times less dense than the corresponding section of a mouse’s brain, per the study. Glial cells, the connective glue that keeps brain tissue together, outnumber neurons in the fragment by a two to one ratio.
Neural explorers
The physical size of the brain fragment may be teeny, but the level of detail means the data captured by the mapping effort is massive. The reconstructed segment is 1.4 petabytes in digital size, or 1,400 terabytes (equivalent to the storage capacity of about 2,800 average laptops). Within that, there is lots to potentially discover: individual neural circuits, previously unobserved cellular ratios and shapes, the makeup of each cortical layer, and more.
“It’s like being an explorer that lands on a new island,” says Lichtman. “You keep looking around and you’re just going to keep finding new things.”
Already, Lichtman and his many co-researchers have made some interesting observations. Amid the ~150 million synapses they mapped, they found a rare type of particularly strong connection. In the vast majority (96.5%) of cases, axons–the outgoing transmission line of neurons– formed one connection with a target cell. Some (about 3%) made 2 connections. But less than .01% forged more than four synapses, including some axons and target cells that were connected at over 50 points.
“We’ve always had a theory that there would be super connections, if you will, amongst certain cells,” says Mosca. “But it’s something we’ve never had the resolution to prove…Now we know that it exists and we can go after the question of what it does.” Lichtman’s current hypothesis is that these extra-fortified connections are the sort of hyper-fast pathways that enable “automatic use of the brain” for well-established, learned actions.
Another new observation: many dendrites (the branching extensions of neurons that generally receive inputs) seem to mirror each other–pointing symmetrically in one of just two directional arrangements out of infinite three-dimensional possibilities. “We’d never seen anything like that [before],” says Lichtman. “Why are they doing that? We don’t know… [it is] a complete mystery.”
The scientists further found a new type of unexplained structure that they’ve named an “axon whorl,” wherein long axon cables appear to tangle around themselves. Though it wasn’t every neuron, some axons contained multiple knots, says Viren Jain, co-senior study author and a senior staff scientist at Google where he leads the company’s Connectomics research team. Again, the function and cause of these whorls is unknown. “We were not expecting to find such a structure. It’s very peculiar… like a big jumble of wiring that sort of contravenes the purpose of a wire to begin with, which is to go places and contact other things.”
These three findings are likely just the tip of the iceberg. “The data set is so large that one human being or lab group can not explore it [all], but a bunch of human beings can,” Lichtman says. Because of the open nature of the project, more than 200 papers have cited the brain reconstruction since it was first released as a pre-print, Jain points out.
In addition to being a large, fundamental advance in science, discoveries resulting from this partial connectome could eventually help us better understand and treat brain diseases. “The ability to measure neural wiring of human brains in such detail opens up exciting opportunities for advancing human health,” says Andrew Leifer, a physicist and neuroscientist at Princeton University who wasn’t involved in the project. “One could imagine comparing different brains to understand how brain wiring changes when a healthy brain suffers from a disease or falls into dysfunction,” he adds.
Pushing into future frontiers
But though there is lots to be discovered, there are also limits. The automated machine learning methods which were key to enabling such a large-scale endeavor carry a margin of error that requires human oversight to correct. Editing will be an ongoing project, and is a community science effort anyone who wants to can apply to participate in.
The sample is also only one small piece of one person’s brain. There is much that can’t yet be inferred about human brains generally or other brain regions beyond the temporal lobe based on this single fragment without more samples and maps for comparison, notes Lichtman.
And, perhaps most critically, the brain segment came from someone undergoing surgery for epilepsy–it may not represent a “normal” brain and there’s no way to know for sure unless and until we have more bits to assess, say Jain and Lichtman. “But we are planning many follow-ups to this,” Jain adds.
The team has ambitions to construct multiple partial connectomes representing additional human brain samples. They are also working on zebrafish connectome, and are planning to tackle increasingly large segments of the mouse brain. Mammalian brains share many similarities, so a complete mouse connectome could offer new insights into our own brain as well as the evolution of brains across animals, Lichtman says.
At the moment, with currently available technologies (and the ethical implications), a complete connectome of the human brian is “a bridge too far,” says Lichtman. “Literally, we’re a million times away from that,” says Jain. But through this study, the scientists have taken an early (if miniscule) step in that direction, and even the smallest peephole can be a portal into a whole universe of knowledge.
“I would love people to think about this the same way they think about the Hubble or James Webb telescope,” says Lichtman. “We’re peering into an unknown domain, and one that is much more relevant to us than distant outer space. It’s this inner space that each of us have on our shoulders that we use, but know almost nothing about.”