Tony Zador is a professor of biology at the Cold Spring Harbor Laboratory who studies auditory processing, attention and decision-making in rodents. He spoke recently at the laboratory's 79th annual symposium on quantitative biology, which focused this year on the topic of cognition. Zador talked about his recent work trying to demonstrate how brain circuits might be mapped by using techniques for sequencing genes. I met Zador at the conference and an edited transcript follows—or you can watch the whole interview here.
Scientific American: First let me ask, what is a connectome?
Tony Zador: The human brain has 100 billion neurons, a mouse brain has maybe 100 million. What we’d really like to understand is how we go from a bunch of neurons to thought, feelings, behavior. We think that the key is to understand how the different neurons are connected to one another. So traditionally there have been a lot of techniques for studying connectivity but at a fairly crude level. We can, for instance, tell that a bunch of neurons here tend to be connected to a bunch of neurons there.
There are also techniques for looking at how single neurons are connected but only for individual links between those neurons. What we would love to be able to do is to tell how every single neuron in the brain is connected to every single other neuron in the brain. So if you wanted to navigate through the United States, one of the most useful things you could have is a roadmap. It wouldn’t tell you everything about the United States, but it would be very hard to get around without a complete roadmap of the country. We need something like that for the brain.
SA: What about the sequencing part of this? You’re proposing that sequencing the connectome might be based on some of the techniques developed here to do gene sequencing.
Zador: Traditionally the way people study connectivity is as a branch of microscopy. Typically what people do is they use one method or another to label a neuron and then they observe that neuron at some level of resolution. But the challenge that’s at the core of all the microscopy techniques is that neurons can extend long distances. That might be millimeters in a mouse brain or, in fact, in a giraffe brain, there are neurons that go all the way from the brain to its foot, which can be over 15 feet. Brain cells are connected with one another at structures called synapses, which are below the resolution of light microscopy. That means that if you really want to understand how one neuron is connected to another, you need to resolve the synapse, which requires electron microscopy. You have to take incredibly thin sections of brain and then image them.
People are doing this impressively well but, at least until very recently, the successes were in C. elegans, a roundworm with 302 neurons and 7,000 synapses. With over 50 person-years of work, they were able to reconstruct the entire wiring diagram of this tiny creature. Since then they’ve scaled it up and it’s working pretty well but it’s still extremely challenging. So a few years ago, because I’m at Cold Spring Harbor and steeped in all this sequencing technology, it occurred to me that gene sequencing technology actually has the capacity to figure out how billions of synapses are connected. If you have a mouse brain that has 100 million synapses and each neuron makes, say 1,000 synapses, that’s 100 billion synapses.
A sequencing run these days costs about $1,000 for one billion reads of nucleotides, the chemical components of DNA. The way we propose to do it would be to have one read equal one synapse. That price is coming down. Fifteen years ago, the cost of sequencing the first human genome was approximately $1 billion and now you can get your genome sequenced for $1,000. And within a couple of years it’ll be well under that. So sequencing costs have plummeted incredibly fast. In fact sequencing is getting better and faster at a rate that’s even faster than the rate at which computers improve their performance. My iPhone has more computing power than a computer did 20 years ago. The enhanced performance follows something called Moore’s Law and sequencing technology has been improving at a substantially faster rate than Moore's Law since 2008. There’s every expectation that sequencing will get faster. That means there’s a huge potential benefit to converting the problem of connectivity into a problem of sequencing.
SA: A circuit is not a gene so how would you sequence a circuit?
Zador: The idea is that we would endow every neuron with a unique, random sequence of DNA. We call it a barcode. This sounds fanciful at first but actually the immune system has solved this problem. Your B-cells and T-cells generate novel antibodies through a process called somatic recombination. They scramble pieces of their chromosome to produce novel antibodies.
That isn’t literally the approach we’re taking. We’re not using the particular collection of enzymes used in the immune system because they are not really convenient for us. But that principle is used by many organisms and we’re hijacking similar proteins from other organisms to try to do the same things in neurons. The idea is that we’re going to put in every neuron a cassette of nucleotides that express a particular protein. The protein will scramble the nucleotides and generate a novel sequence in every neuron in the brain. Every neuron sounds like a lot but combinatorics work in our favor. A sequence of 30 random nucleotides has a potential diversity of 430 because there are four nucleotides.
That number is way more than the number of neurons in the brain. So if we can cause enough scrambling to occur by chance, the probability that two neurons will have the same barcode is infinitesimal. If we can do that then the next step is to express little pieces of RNA that encode that little random barcode. Then we will have engineered proteins that drag those RNAs to the synapse. At each synapse then there’ll be a presynaptic RNA barcode and a postsynaptic one. After that, it’s just sort of biochemistry to link the presynaptic and postsynaptic barcodes together so you'll have a single piece of DNA. Finally, all we have to do is read out those codes and we get, in principle, this huge connectivity matrix.
SA: There are lots of different types of neurons and lots of different structures in the brain. How far, in principle, do you think you could go with this?
Zador: This won’t work in humans because it requires manipulating the neurons. But there’s no reason we couldn’t do an entire mouse brain and, in fact, there’s no reason we couldn’t do many mouse brains once we get the transgenic mice working. So the sequencing costs will be not negligible but will be well within the range that would make this project worthwhile. What we envision is getting the connectivity of not just a portion of the circuit but the entire circuit. I’ve told you the barebones version, but there are bells and whistles that we can add that we think will allow us to not only have the connectivity but the precise location of those cells and their gene expression pattern. That, in turn, will tell us about their cell type so what we really would like to have, although it’s ambitious, is the complete connectivity matrix and then associated with each one of those elements the name of what cell type it is.
SA: As with gene sequencing, this could be done very quickly.
Zador: Once we have a transgenic animal, extracting the DNA takes a few days, and sequencing takes two weeks or less per-individual. The whole process could take a month.
SA: There’s a lot of emphasis in the U.S. on spending substantial sums on developing new technologies to understand the brain better. Are you envisaging this as one of the techniques that might be used?
Zador: I’ve applied for grants. I would love to do this. What I hope to have soon is a proof of principle that will convince people that this is something that’s worth pursuing. But to actually scale all this up is something beyond what one lab could possibly do. So what I would love to see is after we’ve shown a proof of principle, then lots of labs would get involved and come up with much better ideas for doing the things that we’re trying to do now in the quickest way possible. I would be thrilled if resources were devoted to this because I think having the complete connectivity of organisms would be incredibly useful and I think having this ability to generate that could transform how we do neuroscience.
SA: Just to wrap up, at the end of your recent paper on this topic in PLOS Biology, you mentioned that your technique might be good for testing what happens when brain circuits go awry in a disorder like autism. Can you talk about that a bit.
Tony Zador: One of my research interests has been autism and there’s been a lot of progress recently in identifying the genes involved in causing autism. But it turns out that there are dozens and perhaps hundreds of genes, which when perturbed can cause autism. Still, autism, although it’s heterogeneous, can be diagnosed in a meaningful way. So there must be something in common. There’s an appealing hypothesis that autism results from disruptions of circuitry. It’s possible that that happens because there are genetic lesions that we know cause autism in people.
We can recapitulate these lesions in mice and then we can ask the question: ‘What goes awry in the circuits of mice that express the same genetic lesions that people have?’ The hope is that what we could do is take 20 mouse models of autism, look at their brains, look at their connectivity and say that we notice that in 17 out of these 20 there’s a disruption in this circuit compared with non-autistic mice. That would lead us to look closely at that circuit and that circuit might connect the front and back of the brain or one subset of neurons to another subset either locally or in another brain region. The space of possible hypotheses is enormous. Testing them all at this point is not possible unless you have some method like this.
SA: Thanks very much.
Image Source: Cold Spring Harbor Laboratory