Connectomics

What is special about the structure of the brain compared to other organs? What are the obstacles to mapping all the neuron connections in the human brain? Why don’t we have a clear idea about how our brains work? These and other questions are answered by Jeremy R. Knowles Professor of Molecular and Cellular Biology at Harvard University Jeff Lichtman.

[The question is] actually quite old. People have been interested how brains work at the level of its deep structure for over a hundred years. It’s been clear that the brain is made up of nerve cells that are connected together in very complicated patterns. But the approach of trying to figure out exactly how all the brain cells are connected to each other, a kind of mapping of the entire brain, has been really impossible until relatively recently. And now it seems conceivable, at least, that one can get a sense of how brains are organized at this deep level by mapping out all the connections between nerve cells. And this is an “omics” like genomics is the “omics” of all the genes that make up the instruction book to make cells and organisms. The “omics” of the connections of the brain would be the mapping of all those wires and that would be connectomics.

I think it’s been known probably since the invention of microscopes that every organ in the body has unique cell types. And these cells are put together in these motifs that underlie the function of the organs. So, for example, in a kidney they’re a bunch of cells that form these tubules called the ‘nephron’. And the nephron has a filtering function for the blood to remove toxic waste and other ingredients that come out as urine. And it has taken some time, but thanks to microscopes it’s become pretty clear what cell types makes up a nephron and how a nephron works. And, frankly, if you understand how a single nephron works, you understand the kidney, despite the fact that a kidney is made up of hundreds and thousands, or maybe millions of nephrons, because each one is like every other one. So once you understand one nephron, you understand the whole kidney. In fact, the whole kidney is just, because it is iterating the same structure over and over again, there is nothing very special about one kidney as opposed to the other kidney. For example, you can give your kidney away, one of your kidneys away, and the other kidney will do pretty well for you. And the same is true for liver. A part of your liver does the same thing as another part. There is a little motif, knows as the portal triad. Once you understand that basically, it’s just that same motif and that same function iterated over and over again and the same for a lung.

The brain however is very different, because although I could lose a lung and survive, I could lose a half of my liver, but if I take out a half of your brain or anyone’s brain you’ll notice a difference immediately. And that’s because brains are made up of a much more diverse set of cellular organizations. The front of your brain, the frontal cortex, for example, that’s completely different things than the back in the brain, the occipital cortex. The spinal cord has a completely different role in your body than the cerebral cortex. And the cerebellum has a different role and a place called the amygdala has a different role.

Every single part of the brain has its own unique function and, surprisingly, its own unique cell types. And those cells are wired together in unique ways everywhere. So you can’t simply understand the whole brain by understanding one little piece. You actually have to understand every little piece, and every little piece is different from every other little piece.

There’s no other aspect of our body that’s like that. There’s just nothing like it anywhere else. And this has been a tremendous problem for understanding brains. Not only how normal brains work, but how brains work poorly in cases of a brain disease. And I mention this, because for most organ systems most diseases can be traced back to an abnormality in the biochemistry or the structure of the cells that make up some part of the motif.

So for most kidney diseases, most lung diseases, most liver diseases you can trace it back in the field of pathology – it is the field where people look at these abnormal organs and see something wrong. The field of neuropathology is very useful for things like tumors of the brain. But the neuropathology of schizophrenia, which is one percent of the population, or autism – just tremendous number of children with this disorder, or the wide range of other psychiatric or behavioral disorders – there is no pathology. And so some people think the brains must be normal at the level of the cells. That’s not the case. The case is actually we just don’t know, because no one has ever really looked at enough detail to see what the physical structure of the brain is at those high levels. Because it just seems insurmountable amount of complexity and for many years it has been. But now thanks to automated technologies of industrializing the looking at tissue, it becomes conceivable for the first time, I think, to imagine really having of cellular substrate of every single part of the nervous system. And that’s what connectomics ultimately helps to do.

The idea that the brain is made up of nerve cells has a history that began in around 1870-1873 – about when a very, very talented Italian histologist named Camillo Golgi. He was 30 years old and he was playing around apparently in his kitchen with a bunch of chemicals, and he mixed them together in a particular way that allowed the brain tissue to be stained in an extremely inefficient way. Normally you’d think an inefficient stain would be the worst stain in the world, but that was the magic of his stain. And to this day it is not exactly clear why it works this way, but the Golgi stain that now has his name – and he got the Nobel prize for this work, sо it was important, to be sure – was a technique that causes the crystallization of a dark reaction product in a very small subset of nerve cells. But once the dark reaction product started to crystallize it filled up the entire cell. But the vast majorities of cells were unlabeled. So you could see a brain cell in a sea of unlabeled cells, and so for the first time you could see the full complexity of these individual brain cells.

That discovery in 1873 prompted another remarkable scientist, perhaps the greatest neuroanatomist ever, a man named Ramón y Cajal, a Spaniard, to start looking at the brain with the Golgi technique. And he was genius of a very unusual type, he was a genius who could see – then he looked at things he could see more than most people could see. In fact, most people denied that he could possibly have seen what he saw, but he definitely did see this, because of it had stood the test of time. And what he discovered is that nerve cells have two kinds of processes coming out of the cell body – these little cell bodies that are sort of football-shaped – some of them are local branches that are called dendrites, where the cell receives information, these are like antennas for the cell to collect information. And each cell also has one process that goes potentially very far distances, called an axon, which is the output of the cell. And he realized that the axons of nerve cells are touching the dendrites of other nerve cells, talking to those dendrites, sending information into the dendrites that then get to the cell body and then get sent out to the cell’s axon. So there’s a directional circuit where the axons of cells are talking to the dendrites of other cells, those cells are collecting the information and then sending the information on through their own axon to other dendrites of other cells. And in one fell swoop he sort of figured out how information flowed through the nervous system, and he was right. It’s amazing, because he was doing this from fixed material stained in a very sparse way.

And that kind of worldview has kind of dominated the way we have thought about brains since the time of his discoveries, and he shared the Nobel prize with Golgi actually for his discoveries with that technique. The strength of that approach was the profound insight that the brain is made up of nerve cells that have very peculiar shapes, there’s a wide range of them and they have very particular connection patterns. And the weakness of that approach was – unfortunately a very small subset of nerve cells were stained.

So you couldn’t say how many different axons converged on the dendrite of a cell or how many different target cells a particular axon innervated. And its only with modern techniques that allow us to see all the cells and all the connections that one can kind of fill in this sparse Golgi-like stain now with the complete rendering of what’s going on in the brain.
The big problem with a field like this is that there is still a huge chasm between what we know about illnesses of behavior, whether they’re learning disorders or psychiatric illnesses, and what we know about the structure of the brain. So we don’t even in normal brains have a good map of how brain cells are arranged relative to each other, so it’s not surprising, I guess, that we don’t yet have any really good ideas at the level of fine circuitry about what is different in the brains for many behavioral disorders and psychiatric diseases.

Brains are being mapped extremely well now with tools like functional magnetic resonance imaging or PET scanning. These are remarkable tools and they can non-invasively image entire brains at a resolution of about one cubic millimeter – is what each little spot of date is, so the brain is rendered at that resolution.

Within one cubic millimeter of brain tissue we could generate 2000 terabytes of data, 2 petabytes of data per cubic millimeter.

This is a big data problem, because a human brain is about a million cubic millimeters. So that would be about 2 million thousand terabytes, or two million petabytes, which is comparable to the digital content of the world! It’s like big number, so one of the real drawbacks of connectomics is it’s a big data problem in an enormous big data, even big data underrepresents how big this dataset is. So that is one of our problems.

The other is not just if have the data, but how do you acquire that much data? We are working with a microscope company, they are building us the fastest electron microscope ever built, we will take receipt soon. And it images at about two billion pixels per second, or three terabytes per hour. At that rate one can do a cubic millimeter in less than a monthб as opposed to fifty years. It’s a big improvement, but still to do human brain with the machine like that would be fifty million months. So over a million months, I mean, t’s still very difficult to find tools that go fast. Now I should say that genomics, when it began, was very, very slow. People were by hand pipetting little things and if you calculated the speed at which you would have taken to do a human genome people would have been estimating centuries. And now a human genome can be done in a day or two. And, I think, once you know how to do, it you can find ways to parallelize and use ever faster machine. So I don’t think this is a fundamental limit forever, but we are just at the very beginnings of this field. So this is still a tremendous problem and until computers were around, and digitized data, and automated machines that were run by computers, it was not even possible to contemplate doing this kind of project as it is now possible to at least contemplate, not sure do, but at least contemplate doing.

I am an optimist to be sure, and we are pushing as hard as we can to make this field a reality. Between the time we started and now we’ve had a speedup of several thousand fold, and when our new microscope comes – that will be another speed up of over fifty fold. And these speedups mean that we are moving in the right direction. We’re very hopeful that for small animals we will have full wiring diagrams at some time soon…
So human brain is a harder problem, but there may be reasons why after a while it’ll get boring, just as no one would do a full reconstruction of a kidney, because once you understand what the motifs are there’s no need to do more. It’s possibly that the human brain, at some point, we would understand it well enough that we’d say: “We don’t need to do anymore, we’ve got the picture.” That’s my hope.