Tip:
Highlight text to annotate it
X
So, what do our neurons do?
Our neurons are namely the networks that they make,
serve as an interface between us and our environment.
They allow us to optimize our behaviors, so that we can survive and reproduce.
Now in order to use information that we learn today
to help us tomorrow, we need memory.
And indeed this is one of the most important things
that our nervous systems do, they carry out
and etch into the neural networks a history for us.
Our life history is embedded in our neural networks.
And what I'd like to discuss for you today and to unpack
is the fact that although we know memories can last a lifetime,
memories are made from unstable elements.
They're made at neural networks that make synapses that possess proteins
that only last for a couple of days.
So how do we get stability out of unstable elements.
First I want to remind you we've seen so many beautiful pictures of neurons,
Jeff's images, you have to wipe the drool off of your collar,
but neurons are unique, they are the only cells in our body
that look like this.
Most cells are rather round and boring.
Neurons, by contrast, have that same round cell body
but have spread out miles and miles of cable,
that you saw in Jeff's images, to make connections,
to touch other cells.
In fact, most other cells in the body touch only a handful of cells,
but neurons contact 10,000 other neurons.
And that's shown in this image here; you see a neuronal cell body in green
and the processes or the cable emerging, and in red, you see individual synapses.
And if you looked at just the single cell,
and that's all I'm going to tell you about today is a single cell,
that single cell talks to 10,000 other cells.
And it talks to those 10,000 other cells at the synapses that you've heard about.
Synapses are the junctions between two cells.
One cell's axon, its transmission cable, comes in very close contact
to another cell's dendrite, the receiving aspect of a neuron.
And if we were to zoom in on the synapse,
we would see that the synapse is jam-packed with protein.
So, I would like you to think about this,
because synapses is where the information is stored and proteins turn over.
Let's just think about how much protein is in a single cell.
So, for the last 20 years, we've spent a lot of time,
"we" meaning neuroscientists,
characterizing the proteins that inhabit synapses.
Most people would agree that on average there are 500 different kinds of proteins
that are really resident at the synapse.
And for each one of those proteins,
on average, and we know this number for some of the proteins,
there's about 50 copies.
So that gives us 25,000 proteins in a single synapse.
And I already told you that there's 10,000 synapses per cell,
so that means we have 250 million proteins
and that's just in the dendrites of the cell.
We also know the cell has an equally elaborate architecture
for sending information in the axon,
and let's just add another 250 million proteins there.
So, on average, we have something like 500 million proteins per cell,
but unlike memories, unlike humans, unlike most cells in the body,
the proteins only last a little while, so on average, the half-life of a protein,
the time when it's gone through half of its lifetime, is about 24 hours.
So that means that in a single cell, every 24 hours,
we need at least 250 million proteins, just to keep the neuron at status quo.
So, I want to remind you about where proteins come from.
Information is stored in the genes in the form of DNA,
and then DNA is transcribed into RNA, and that happens in the cell body
in the nucleus of cells.
Those RNA molecules, or so-called messenger RNA molecules,
serve as a template for protein-synthesis machines
to read out the code to generate a protein.
And in most cells that happens just outside the nucleus, in the cytoplasm.
Now, in neurons where does that happen?
Does it happen in the cell body, like all the other cells?
And if that's true,
how can the cell body generate those 250 million proteins it needs
and send those proteins out to those 10,000 synapses that a single cell has?
Well, in fact, neurons have come up with a very clever solution:
they've distributed the goods.
So instead of exclusively synthesizing proteins in the cell body,
what they've done is to take the protein synthesis machinery, the ribosomes,
and send them out close to the synapses.
You can see here on the left-hand side an image of a dendritic spine
and the little arrow is pointing to a cluster of polyribosomes.
These are protein-synthesis machines
that are in the active process of making a protein.
And on top of that dendritic spine, I hope you can see
the presynaptic nerve terminal
filled with the tiny little vesicles of neurotransmitter.
So the machinery is there out in the processes, and not only that,
we can visualize the process of protein synthesis.
In fact, on the image shown in the middle and on the right-hand side,
you can see images that were taken here at Caltech together with Rich Robertson,
most recently in a much bigger form with Dave Tirrel,
where we've devised chemical techniques to label newly synthesized proteins.
And in the image shown here you see a little stream or plume
where we've stimulated a very small fraction of the dendrite
and then looked to see if we could see evidence for new protein synthesis.
And in the blown-up image shown on the extreme right,
you see these little clusters of fluorescent particles
and those are labeling newly synthesized proteins.
So, how ubiquitous is this distributed process that cells have come up with?
Now, in order to understand that, what you would really like to know
is just how many different proteins could be synthesized out in the processes.
And to know this, you could say,
"What is the population of those messenger mRNAs that are there?"
And so as a starting point for this set of experiments,
what we did was to look, of course, and see what other people had done.
And we could see 3 big studies that had been done
and in those studies they had identified about 100 mRNAs that were present.
And we took those 3 studies and said,
"Okay, is there anything left to discover? Have all of these 3 studies
found the whole population?"
And we asked, "How much overlap is there between the studies?"
We were quite surprised to see there was not a single mRNA
that had been identified by all 3 studies.
So that suggested to us maybe there was something left to learn.
(Laughter)
And at this point I looked at this
and I said to one of my favorite students, Georgi Tushev,
"Just with this information, we can probably estimate
how big that pool of messenger RNA's is."
And Georgi said, "You're right, and I'm going to get back to you on this."
And the next day he showed up with something remarkable.
A paper from 1930. This meant Georgi had to go to the library!
(Laughter)
For the students this is a place where we used to store books.
(Laughter)
Georgi brought me this paper by Zoe Emily Schnabel,
a field biologist and mathematician
who was working in the Midwest of the United States,
and was faced with the task of estimating how many fish are in a lake.
Same problem as us. She did not want to drain the lake.
But rather she reasoned that if she took a net of fish,
and marked the fish she had captured, and threw the net back,
and then took another net of fish, by doing this several times,
and each time noting how many fresh fish and how many previously tagged fish,
she could come you with an estimate of how many fish were in the lake.
So we did this, we took Schnabel's method and we looked to see how many mRNAs
would we predict are out in the processes, and we found a much bigger number
than we had even dreamed of, which was around 2,000 mRNAs.
So, maybe the universe of this local translation was much bigger
than we had thought.
So, to discover those mRNAs,
we used a technique called "deep sequencing."
And unlike other techniques to discover things,
this is a technique where you don't have to know what you're looking for.
You isolate mRNA from the dendrites and axons and then you directly sequence,
read out the nucleotide sequence,
to discover what those mRNAs are coding for,
what proteins are being coded for by the mRNAs.
So, when we did this, we were really pleased to see
we actually got a number very close to what the estimate had come up with.
We discovered there were 2,500 mRNAs out in the processes.
And now what we could see is that what we had amassed
when we looked at what proteins could be made there
is almost a toolkit for building a synapse.
So, instead of having all of the proteins being shipped out to build synapses,
this suggested that locally, right there near the synapse,
was the protein synthesis machine and all of the mRNAs we would need
to actually build the synapse.
Now we discovered these by sequencing,
but we wanted to be able to visualize them,
and so to do this we used a new technology which allows us
to visualize individual mRNA molecules. It's a kind of fluorescent barcode.
So for each kind of mRNA you would like to detect
you develop a probe that has a sequence
that's complementary to your mRNA of interest.
So the probe will bind specifically.
And the way that we can see the mRNA
is for each mRNA, we have a string of fluorescent molecules that are unique.
So it's like a fluorescent barcode.
We can design many such probes and mix them with our dendrites and axons
to see which mRNAs are there and to count which mRNAs are there.
So the probes are mixed and then we stick them down on a slide so we can image them.
We image them and count them and this is what they look like.
Now when I first saw this, it took my breath away.
What you can see here, hopefully, is little strings of fluorescent barcodes.
You also see some bright red dots which are used
to help us re-register the images which are taken at high-magnification.
And if you blow up, you see these little mRNA molecules
which are coming from the dendrites and the axons.
And when I first saw this, it's sort of a testimony
to how seeing things is so wonderful, it reminded me in a very small way
of when I was pregnant with my daughters and I got my first ultrasound.
So I knew that they were there but seeing them somehow was super-special,
and that's how it was for us when we could actually visualize
these mRNAs there and see them, knowing they had been there for so long.
Now these techniques allow us to actually tell that the mRNA is there,
but we'd actually like to visualize it back in situ.
We'd like to see these mRNAs back in the dendrites.
And to do this, we can use similar strategies,
where we use fluorescent molecules that can recognize these mRNAs
and we can visualize them back in the neurons.
So shown here is a single neuron from the hippocampus in the brain,
the cell body is highlighted in blue,
and the dendrites are the green things that stick out.
And each one of those red dots is a single mRNA molecule.
So what we can do is, for all of those 2,500 molecules,
we can look at the distribution pattern, and we can start to see
how synapses assemble this local machinery.
You can tell already that this solution is a great one,
because it puts this machinery in close proximity to the business end,
to the synapse,
and allows the machinery to, first, respond to synaptic signals,
to keep synapses stable,
if they need to be stable when proteins are turning over,
but also allows for the modification of synapses by changing
the local protein complement that's there.
Now this kind of problem, this distribution of goods,
the optimization of distribution of goods, is one that's ubiquitous in society.
And if you think about it, if you have to deliver something to a population
there's several different models that might work.
One would be to have a large production center
that would distribute the goods to the entire population.
Another might be a solution where you'd have multiple production centers
with sort of random routes going to local end points,
and something in between appears to be the solution
that the neurons and other systems have come up with,
and this is the so-called "small-world network,"
where you can see, for example,
analogies with the distribution of power in the US,
where the task is to provide electricity to all households in the US,
and you can see a similar system
where you have local hubs that subserve a series of local endpoints.
And even more perhaps gratifying is something like the worldwide web
which uses a similar model where you have a hub subserving these endpoints.
So, in conclusion, from the synapse to the Internet,
it appears that the same kind of scheme has been implemented
and local control makes sense.
This is my team in Frankfurt, who are amazing and motivating
and fabulous people and funding
and I thank you.
(Applause)