Top

iBiology

Bringing the World's Best Biology to You

Optical Studies of Individual Synapses

Part 1: Optical Studies of Individual Synapses

00:00:02.06 Good day.
00:00:03.08 My name is Karel Svoboda.
00:00:04.21 I'm a Group Leader at the
00:00:05.23 Janelia Farm Research Campus
00:00:07.14 of the Howard Hughes Medical Institute.
00:00:09.21 Today I'd like to tell you about synapses,
00:00:11.29 the tiny junctions between neurons
00:00:14.09 that underlie your perception of the world,
00:00:16.27 and also the places where memories
00:00:19.07 are stored in the brain.
00:00:21.29 I'd like to focus on studies of individual synapses,
00:00:25.25 optical studies,
00:00:27.06 where microscopy techniques have had a huge impact.
00:00:29.25 So in the first part of the lecture,
00:00:31.26 I'd like to focus on these imaging techniques,
00:00:35.22 after an introduction to synaptic transmission and plasticity,
00:00:38.14 and then give you a few major results
00:00:41.07 that were arrived at
00:00:42.26 with these new optical techniques.
00:00:44.20 And in the second part of the seminar,
00:00:46.19 I will focus on spatial interactions
00:00:49.23 between synapses,
00:00:51.12 focusing on calcium-dependent signal transduction mechanisms.
00:00:57.26 So the neocortex is the largest part of the mammalian brain,
00:01:02.11 and it's really a complex network
00:01:04.05 made out of many neurons
00:01:06.02 connected into complex networks
00:01:08.13 by synapses.
00:01:10.24 Synapses are these tiny structures
00:01:13.06 in the neurophil.
00:01:14.24 They consist of presynaptic terminals,
00:01:17.05 swellings in the axons,
00:01:19.01 which are opposed to dendritic spines,
00:01:21.04 which are these hair-like structures
00:01:23.07 that come off the dendrite.
00:01:25.04 And there's a one-to-one correspondence in the cortex
00:01:27.16 between these presynaptic terminals
00:01:29.13 and these dendritic spines.
00:01:31.24 So let's focus on this cartoon of a synapse
00:01:34.03 here over on the right.
00:01:35.24 When an action potential zips through the axon,
00:01:38.23 it invades these synaptic terminals
00:01:41.09 and causes glutamate release,
00:01:43.09 which then binds to postsynaptic receptors
00:01:46.07 on the postsynaptic dendritic spine.
00:01:50.25 And so there are many questions
00:01:53.18 that remain to be answered
00:01:55.26 about synaptic transmission and plasticity.
00:01:57.27 For example, how spatially localized
00:01:59.26 is synaptic transmission and plasticity?
00:02:03.29 For example, is glutamate released
00:02:06.06 from individual synaptic terminals
00:02:09.09 sensed by receptors on only one dendritic spine,
00:02:12.00 or does it spill over to activate neighboring dendritic spines?
00:02:15.18 And similarly for synaptic plasticity,
00:02:18.18 long-term changes in the efficacy of synaptic function.
00:02:22.01 Can individual synapses change,
00:02:24.06 or do synapses always change in groups?
00:02:26.18 And answers to these questions are important
00:02:29.03 because they challenge our understanding
00:02:31.25 of the information processing capacity of the brain.
00:02:35.29 The more independent synaptic transmission is, of course,
00:02:38.27 the larger the information processing capacity of the brain is,
00:02:43.03 because there will be more independent information transmission
00:02:45.27 channels in the brain.
00:02:47.07 Similarly, in terms of synaptic plasticity,
00:02:49.12 the more independent synaptic plasticity is,
00:02:52.27 the larger the memory storage capacity of the brain is,
00:02:55.23 because it can store more bits.
00:02:59.22 And also, if synaptic transmission and plasticity
00:03:02.28 are independent at the level of individual synapses,
00:03:05.29 that in turn challenges our understanding
00:03:08.08 of the spatial regulation
00:03:11.10 of signal transduction in synapses.
00:03:13.28 Activation of synapses
00:03:15.21 activates complex biochemical pathways
00:03:18.06 which feed back onto synaptic function,
00:03:21.01 which then changes the structure and function
00:03:23.28 of synapses, for example,
00:03:25.19 through the insertion of AMPA receptor,
00:03:27.24 glutamate receptors,
00:03:29.06 into the postsynaptic density.
00:03:32.13 And synapses are very tightly packed along the dendrites:
00:03:35.16 about one per micrometer of dendritic length.
00:03:38.26 And so the question then is,
00:03:40.11 how can these intracellular biochemical pathways
00:03:43.11 be regulated at the micrometer level
00:03:45.23 in these neuronal microcompartments.
00:03:49.17 I'd like to remind you that neurons
00:03:51.05 have an output side, the axon here in blue,
00:03:54.27 and an input side, the dendrite here in red.
00:03:58.26 And when an action potential is generated
00:04:00.27 by presynaptic neurons
00:04:03.08 -- in green is the action potential --
00:04:05.27 it propagates throughout the axon arbor
00:04:08.25 and then, where axons and dendrites overlap in the neurophil
00:04:12.19 sometimes synapse occur
00:04:14.23 and when the axons potential reaches this synapse,
00:04:18.18 it causes synaptic transmission.
00:04:22.05 There are a large number of synapses
00:04:23.28 in the dendritic arbor
00:04:25.20 and the axonal arbor of individual neurons.
00:04:28.29 Neurons have about 10,000 outputs
00:04:32.08 and 10,000 inputs,
00:04:34.01 and individual synapses are really quite weak.
00:04:37.16 It takes about 10-20,
00:04:39.10 perhaps 30 synapses
00:04:41.25 firing almost simultaneously
00:04:43.26 to drive the postsynaptic neuron to fire.
00:04:46.27 And so individual synapses
00:04:48.23 have a very weak impact
00:04:50.15 onto the postsynaptic dendrite,
00:04:53.13 onto the postsynaptic neuron.
00:04:55.07 So how do these synapses look?
00:04:57.15 Here's an electron micrograph
00:04:59.16 -- electron microscopy image --
00:05:02.18 of a thin section of cortical tissue,
00:05:06.01 a cross-section through a synapse.
00:05:08.09 And notice the tiny length scale here,
00:05:10.06 the entire image is only about
00:05:13.01 2 micrometers (um) of so in diameter.
00:05:15.15 And you see the presynaptic terminal,
00:05:17.13 chock full of vesicles
00:05:19.23 filled with the excitatory neurotransmitter glutamate,
00:05:22.22 opposed to a dendritic spine,
00:05:25.11 the postsynaptic specialization.
00:05:27.28 Here is the spine head of the dendritic spine,
00:05:31.07 which is connected through a thin neck, not shown here,
00:05:34.05 to the parent dendrite.
00:05:36.17 Notice, also,
00:05:38.00 between the presynaptic specialization
00:05:40.19 and the postsynaptic specialization
00:05:42.20 a light area, which is the synaptic cleft
00:05:46.13 into which the neurotransmitter spills
00:05:49.01 upon glutamate release.
00:05:51.18 Notice also the electron-dense,
00:05:54.14 or dark structure on the postsynaptic side,
00:05:57.07 which is the postsynaptic density,
00:05:58.24 which contains the receptors for neurotransmitter,
00:06:02.05 signaling molecules,
00:06:04.09 scaffolding protein,
00:06:06.01 and also cell adhesion molecules
00:06:07.23 that keep pre- and postsynaptic structures together.
00:06:13.01 Now, you can,
00:06:15.05 from these electron microscopy images,
00:06:17.01 from individual images of individual 50 nanometer (nm) sections,
00:06:20.08 you can build up a picture of how...
00:06:23.01 a 3D picture of how these synapses look,
00:06:25.16 and here is an example.
00:06:28.00 In green is the axon,
00:06:29.24 in yellow one 5 micron length dendritic segment,
00:06:34.05 with one dendritic spine, here.
00:06:37.21 And you can see from this electron microscopy
00:06:40.10 3D reconstruction
00:06:42.09 that along the dendrite synapses are packed rather tightly.
00:06:46.07 There's about one synapse,
00:06:47.29 one to two synapses
00:06:49.17 per micrometer of dendrite
00:06:52.10 in these pyramidal cells.
00:06:54.17 And also in the neurophil,
00:06:55.29 if you now reconstruct all of the axons
00:06:58.04 in the neurophil
00:06:59.22 you can see that it's really quite dense:
00:07:01.25 there are about 1,000 axonal elements
00:07:04.27 in this 6x6x6, roughly, micrometer cube
00:07:09.04 of cortical tissue,
00:07:10.27 and that sort of highlights the complexity
00:07:13.09 of the tissue in which synapses are embedded.
00:07:17.10 And synapses really have to studied
00:07:19.18 in intact tissues
00:07:22.22 because synapses
00:07:27.02 really don't have the same properties
00:07:28.28 in cultured preparations.
00:07:33.04 So let's look at this process of synaptic transmission
00:07:35.29 in a little bit more detail.
00:07:37.29 So when an action potential
00:07:41.09 propagates and invades a synaptic terminal
00:07:43.27 it opens voltage-gated calcium channels,
00:07:45.26 which then produce a glutamate release
00:07:48.13 in a stochastic manner.
00:07:49.23 Often there are no vesicles released,
00:07:52.29 sometimes one, sometimes two
00:07:55.04 vesicles of glutamate released,
00:07:57.02 which then leads to stochastic amounts of glutamate
00:07:59.26 released into the synaptic cleft,
00:08:01.24 which then activates, principally,
00:08:03.27 two types of ionotropic receptors,
00:08:06.03 that is, glutamate receptors that admit, directly,
00:08:09.26 ions into the postsynaptic cell.
00:08:12.20 There are the AMPA receptors and the NMDA receptors
00:08:16.18 and these receptors have different properties.
00:08:19.11 The AMPA receptors
00:08:21.05 mainly conduct sodium
00:08:23.07 at normal physiological conditions.
00:08:25.15 NMDA receptors, in addition,
00:08:27.22 conduct calcium,
00:08:29.10 and calcium is very important in synaptic biology
00:08:32.19 because it couples electrical excitation
00:08:35.28 to intracellular signal transduction pathways
00:08:38.15 that drive synaptic plasticity.
00:08:44.07 So calcium is required
00:08:46.16 to produce synaptic plasticity
00:08:50.10 and the calcium that produces synaptic plasticity
00:08:53.15 comes into synapses through NMDA receptors,
00:08:56.25 a special of glutamate receptor.
00:09:02.17 So the classic way of studying synapses
00:09:05.09 is using electrophysiological methods,
00:09:07.23 mainly in brain slices.
00:09:09.17 A brain slice is
00:09:11.27 just a slice of tissue
00:09:14.11 with intact neural networks inside it,
00:09:18.05 and one example is the hippocampal brain slice.
00:09:20.26 So here we have a hippocampal brain slice,
00:09:23.03 the CA3 region of the hippocampus,
00:09:25.14 and then the so-called Schaffer collateral axons
00:09:28.03 that go into CA1.
00:09:30.14 And in these kind of experiments,
00:09:32.02 one has a stimulating electrode
00:09:33.20 that is placed into the Schaffer collateral axons
00:09:36.22 and one often records from a CA1 pyramidal cell
00:09:40.20 using whole-cell methods.
00:09:43.01 Electrical stimulus then evokes
00:09:45.08 synchronous action potentials
00:09:47.18 in these Schaffer collateral axons
00:09:50.00 and in response one records postsynaptic currents
00:09:53.24 produced by groups of synapses
00:09:56.11 in individual CA1 pyramidal cells.
00:09:59.14 So down here is a current trace
00:10:02.04 recorded in such an experiment.
00:10:04.07 The sharp spike down there is due to the stimulus,
00:10:08.13 the synchronous activation of the action potentials,
00:10:12.02 and the slow downward deflection
00:10:14.09 is the excitatory synaptic current
00:10:17.08 that tends to depolarize the postsynaptic neuron.
00:10:20.19 So this has been an incredibly important technique
00:10:23.19 for the study of dynamics of synaptic transmission,
00:10:26.11 but it has important limitations
00:10:28.28 that I would like to highlight.
00:10:30.06 Some of these limitations
00:10:32.06 can be overcome with imaging techniques,
00:10:33.26 that's why I bring these limitations up.
00:10:36.13 First of all,
00:10:40.20 these kinds of measurements can provide only limited mechanistic insight,
00:10:44.21 because the time course
00:10:47.12 and the amplitude of these currents
00:10:49.26 is shaped both by presynaptic factors,
00:10:52.15 for example, the amount of glutamate released,
00:10:55.06 and also postsynaptic factors,
00:10:56.26 for example, the number of receptors activated
00:10:59.22 on dendritic spines.
00:11:02.23 Whereas with imaging, of course,
00:11:04.15 one can, depending on what fluorophore
00:11:07.08 you put into the cell
00:11:09.07 and whether or not you label the presynaptic terminal
00:11:11.21 or the postsynaptic dendritic spine,
00:11:13.29 you can look at these different
00:11:16.08 pre- and postsynaptic mechanisms separately.
00:11:18.20 In addition, these kinds of techniques,
00:11:21.00 whole-cell recording techniques,
00:11:22.14 don't have single synapse resolution and sensitivity,
00:11:26.06 and therefore one always measures
00:11:29.11 these kinds of synaptic currents
00:11:31.12 at the level of populations.
00:11:32.28 They often represent
00:11:34.10 a couple of dozen synapses
00:11:35.27 acting in concert,
00:11:37.13 and that washes out all of the interesting
00:11:39.19 single-synapse fluctuations
00:11:41.21 that are mechanistically very informative.
00:11:44.28 And one can, for example,
00:11:46.01 not tell from these kinds of measurements
00:11:47.27 whether or not different synapses,
00:11:49.14 different individual synapses are different,
00:11:51.19 or whether or not all synapses
00:11:54.14 really have similar properties,
00:11:56.01 similar receptor compositions,
00:11:58.11 and so on and so forth.
00:12:01.05 Let me tell you a little bit more about
00:12:03.23 the postsynaptic receptors that are activated
00:12:05.24 during synaptic transmission.
00:12:07.23 So, when glutamate is released,
00:12:10.08 as I mentioned already,
00:12:11.26 it activates both...
00:12:13.18 it binds to both AMPA and NMDA receptors,
00:12:16.14 and the contribution of these receptors
00:12:18.22 to synaptic transmission
00:12:20.16 can be dissociated
00:12:22.21 using electrophysiological and pharmacological techniques.
00:12:28.17 So, notice on the left here,
00:12:31.02 we block...
00:12:32.27 there are experiments where we block NMDA receptors
00:12:34.25 with a specific drug that inhibits NMDA receptors,
00:12:38.12 called APV.
00:12:39.27 Down there are traces recorded
00:12:42.04 at resting conditions,
00:12:43.18 the normal state of the cell,
00:12:46.01 roughly -80 millivolt (mV),
00:12:48.02 and you see that APV blocking NMDA receptors
00:12:50.20 has almost no effect,
00:12:52.12 telling us that NMDA receptors
00:12:53.26 really don't contribute to synaptic transmission
00:12:56.05 under resting conditions.
00:12:58.29 But the situation is quite different
00:13:00.13 when one depolarizes the postsynaptic cell, here,
00:13:03.16 to about +20 mV,
00:13:05.11 and you can see that especially the late phase of the current
00:13:08.03 is now dramatically reduced
00:13:09.24 by the presence of APV.
00:13:12.01 This tells us that NMDA receptors
00:13:14.03 don't only bind glutamate,
00:13:15.25 but they also need to be depolarized to open,
00:13:18.26 to admit charge into the postsynaptic cell,
00:13:22.14 and very importantly, to admit calcium into the postsynaptic cell.
00:13:26.11 With AMPA receptors, the situation is almost the converse.
00:13:29.27 Almost all of the charge is carried by AMPA receptors
00:13:33.21 at resting conditions,
00:13:36.05 but at depolarized conditions
00:13:38.07 blockade of AMPA receptors
00:13:39.29 has relatively little effect on the charge transfer
00:13:42.10 into the postsynaptic cell.
00:13:45.01 So the take-home message is that
00:13:46.24 to open NMDA receptors,
00:13:49.09 the NMDA receptor has to both bind glutamate
00:13:51.26 and the postsynaptic membrane
00:13:54.08 has to be depolarized close to zero
00:13:56.11 to open the NMDA receptor
00:13:57.28 and to admit calcium.
00:14:00.15 So that gets me to synaptic plasticity,
00:14:03.28 in particular long-term potentiation,
00:14:06.15 which again has been studied in great depth
00:14:08.23 in brain slices.
00:14:10.15 Long-term potentiation is a long-term change
00:14:14.08 in synaptic efficacy
00:14:16.28 that is triggered by strong activation
00:14:19.03 of NMDA receptors.
00:14:20.25 Here's a typical LTP experiment:
00:14:23.02 again we record from a CA1 pyramidal cell,
00:14:25.14 and now there's a stimulus electrode both in...
00:14:29.02 there are two stimulating electrodes
00:14:31.05 on either side of the recorded cell,
00:14:33.20 and these stimulating electrodes activate
00:14:36.17 a non-overlapping set of fibers
00:14:39.03 that course in quasi-random ways
00:14:42.09 through the tissue.
00:14:44.00 So if you blow up the experimental configuration
00:14:46.14 with a situation like this,
00:14:48.17 here's the stimulating electrode 1,
00:14:52.03 which in this experiment produces a weak,
00:14:54.15 sub-threshold stimulus
00:14:57.08 that is sub-threshold to the induction of
00:15:00.19 long-term potentiation.
00:15:02.15 And a second stimulus electrode
00:15:03.28 that produces a strong stimulus
00:15:05.24 that is super-threshold for
00:15:08.09 induction of long-term potentiation, or LTP.
00:15:11.17 So if we now provide a stimulus,
00:15:14.04 we measure the baseline synaptic transmission,
00:15:16.21 baseline response,
00:15:18.12 and produce this sub-threshold stimulus
00:15:21.18 induction protocol that is not sufficient to induce LTP,
00:15:24.20 you can see it does not cause changes
00:15:27.18 in the amplitude of synaptic transmission.
00:15:31.01 Whereas if a super-threshold stimulus
00:15:33.18 is provided through the other electrode,
00:15:36.02 that does now cause a long-lasting change
00:15:38.24 in synaptic transmission.
00:15:40.26 Notice that when we produce
00:15:43.10 the strong stimulus to stimulus electrode 2
00:15:46.12 and produce LTP in the fibers
00:15:49.16 stimulated by stimulating electrode 2,
00:15:52.07 this does not cause a change in synaptic transmission
00:15:56.24 evoked by stimulating electrode 2.
00:15:59.23 And this is taken as evidence
00:16:02.28 for synapse specificity
00:16:05.14 for the induction of plasticity.
00:16:07.12 If we induce plasticity
00:16:08.24 in one set of fibers
00:16:10.14 with stimulus electrode 2,
00:16:12.09 we don't see plasticity of synaptic transmission
00:16:16.18 of a different set of fibers
00:16:18.25 stimulated by stimulus electrode 1.
00:16:22.04 But notice that in these kinds of experiments,
00:16:24.25 we know very little about the spatial relationships
00:16:27.20 of the synapses stimulated
00:16:30.12 by the two stimulating electrodes,
00:16:32.19 and in general in these kinds of experiments,
00:16:36.11 synapses are activated in a sparse manner,
00:16:41.15 and the activated synapses are dispersed,
00:16:43.23 throughout this very large dendritic arbor,
00:16:45.18 which can be about a millimeter,
00:16:48.09 and so this doesn't really answer the question
00:16:51.01 whether or not, if one induces plasticity at one synapse,
00:16:53.23 whether or not something happens at a neighboring synapse
00:16:56.03 just a micrometer or so away.
00:16:59.03 Notice also that there are, in fact,
00:17:01.09 interactions between synapses,
00:17:03.17 because if we now pair a strong stimulus
00:17:06.10 to stimulus electrode 2
00:17:07.25 and a weak stimulus to stimulus electrode 1,
00:17:10.04 now the weak stimulus
00:17:12.09 does produce long-term changes in synaptic efficacy.
00:17:15.04 So there are, in fact, interactions
00:17:19.16 between different synapses,
00:17:21.16 and the way this kind of experiment
00:17:23.16 is usually interpreted
00:17:25.16 is that the strong stimulus depolarizes the cell
00:17:28.12 and then allows the opening of NMDA receptors
00:17:32.05 also at the synapses
00:17:34.17 that are activated by stimulus 1,
00:17:37.12 causing calcium influx
00:17:39.01 and robust long-term potentiation.
00:17:41.11 But again, the mechanisms are not terribly clear
00:17:43.26 and imaging experiments can get at that
00:17:47.15 with great incisiveness.
00:17:52.13 This slide is just to illustrate
00:17:54.10 that these long-term changes in synaptic transmission
00:17:57.28 are likely to be quite important,
00:18:00.16 and likely underlie aspects of learning and memory.
00:18:03.05 There are many kinds of experiments
00:18:04.21 that speak to this point.
00:18:06.12 Here's just one example
00:18:08.05 where the same kind of measurement of synaptic transmission
00:18:10.24 was done in a living animal,
00:18:13.25 in the hippocampus of a rat,
00:18:16.06 that was trained to avoid one of two chambers.
00:18:21.10 And after the training you can see that the synaptic transmission
00:18:24.00 in some of these recorded pathways
00:18:26.23 was potentiated over hours,
00:18:32.16 while other pathways were weakened,
00:18:34.16 suggesting that long-term potentiation
00:18:36.27 is associated with learning of novel situations,
00:18:42.15 and that changes in synaptic function
00:18:45.05 might actually underlie
00:18:47.17 learning and memory in cortical networks.
00:18:52.15 So let me then tell you
00:18:56.17 what kind of questions I think
00:18:59.09 we'd like to address.
00:19:01.10 We'd like to ask some fundamental questions
00:19:03.26 about synaptic function and plasticity.
00:19:06.10 One of the, I think,
00:19:07.23 most fundamental questions is,
00:19:09.18 how long do synapses live?
00:19:12.10 How stable are synapses in the adult brain?
00:19:15.17 And are new synapses made
00:19:18.25 in response after a novel experience?
00:19:21.06 This is of course fundamental to understanding
00:19:23.18 of adaptive changes in neural networks.
00:19:27.15 Are changes in neural networks
00:19:28.29 associated with rewiring of the neural networks,
00:19:31.04 or are these changes really limited to
00:19:33.23 functional changes at existing synapses?
00:19:37.14 Then another set of questions
00:19:38.28 that I'd like to address is,
00:19:40.14 how local are synaptic transmission
00:19:42.03 and synaptic plasticity?
00:19:44.05 As I've already suggested,
00:19:47.18 these kinds of questions are really best answered
00:19:50.23 using imaging methods,
00:19:52.15 and answers to these questions
00:19:55.16 will then challenge our understanding
00:19:57.03 of calcium-dependent signaling
00:19:59.04 and the spatial localization
00:20:00.28 of calcium signaling.
00:20:03.03 And again, we and others
00:20:04.19 have developed methods to track calcium ions
00:20:07.22 in postsynaptic dendrites,
00:20:09.18 and to track calcium-dependent signal transduction pathways
00:20:13.07 that allow us to ask how local calcium-dependent synaptic signaling
00:20:17.09 is at synapses.
00:20:19.12 So these seem like fundamental questions,
00:20:22.00 and why haven't they been address previously?
00:20:24.26 Well, there are some experimental challenges.
00:20:27.00 First of all, synapses are tiny,
00:20:29.06 they're only about half a micrometer or so,
00:20:32.11 and here I've outlined a couple of synapses.
00:20:36.03 A presynaptic terminal
00:20:37.27 and a dendritic spine here,
00:20:39.20 with the thin spine neck connecting it to the parent dendrite.
00:20:43.04 These are only about half a micrometer
00:20:45.07 and, as a consequence,
00:20:47.20 individual synapses function
00:20:49.12 with only a small number of signal transduction molecules.
00:20:52.08 There are many kinds of species
00:20:54.28 of a protein in synapses,
00:20:56.25 but any one molecule is present only
00:20:59.03 in a handful of copies,
00:21:00.21 maybe 10-50 receptors or so
00:21:03.28 per synapse, for example.
00:21:06.01 So one has to really deal with the noise and fluctuations
00:21:08.15 inherent to dealing with
00:21:11.17 small numbers of molecules,
00:21:13.07 the so-called "shot noise" in synaptic transmission.
00:21:18.25 In addition, synaptic biology
00:21:21.05 plays itself out over a wide range of time-scales.
00:21:24.24 We have to,
00:21:26.09 when we're in the savannah and we see a lion,
00:21:28.15 we have to very rapidly decide
00:21:30.06 to run away from the lion.
00:21:32.00 This kind of process,
00:21:33.15 recognition of the lion,
00:21:34.27 and deciding, and running away from the lion,
00:21:37.20 takes up many steps of synaptic transmission.
00:21:40.16 It has to occur rapidly,
00:21:42.05 therefore synaptic transmission has to be rapid,
00:21:44.11 and really occurs over millisecond time-scales.
00:21:47.03 On the other hand, we remember...
00:21:49.17 we have memories that last most of our lifetime.
00:21:51.20 So, synaptic biology really plays itself out
00:21:53.22 over milliseconds up to years.
00:21:57.13 In addition,
00:21:58.27 and as a real experimental challenge,
00:22:02.24 synapses are deeply embedded in complex tissue
00:22:05.09 and have to be studied
00:22:07.00 in an intact nervous system,
00:22:09.20 and that really produces experimental challenges.
00:22:12.18 One can't just sort of...
00:22:14.06 synapses are too small to be penetrated by electrodes,
00:22:17.14 but they're embedded in this tissue
00:22:19.06 that is hostile to traditional types of microscopy.
00:22:24.01 And so...
00:22:26.09 and optical methods are
00:22:29.29 very nicely amenable to synaptic biology
00:22:32.05 in the sense that the resolution limit of microscopy
00:22:35.09 is well matched to the size of synapses,
00:22:38.27 but scattering produced by complex nervous tissues
00:22:43.16 -- scattering is the process
00:22:45.29 by which light is bent
00:22:49.03 in quasi-random way
00:22:50.23 when it impinges into complex tissue
00:22:53.03 like the skin, or like the brain --
00:22:55.29 really makes ordinary,
00:22:57.26 sort of wide-field fluorescence microscopy methods
00:23:01.08 difficult to apply to intact tissues.
00:23:05.08 So let me introduce, then,
00:23:06.23 to you a bunch of tools.
00:23:09.03 Four different kinds of tools
00:23:10.28 all based on two-photon excitation of
00:23:14.14 fluorescence or two-photon uncaging,
00:23:18.08 that have really revolutionized, I think,
00:23:20.08 the study of individual synapses.
00:23:23.13 I'll introduce two-photon laser scanning microscopy,
00:23:25.21 which allows us to image individual synapses
00:23:27.21 deep in intact scattering tissues.
00:23:30.28 I'll introduce calcium imaging,
00:23:32.29 which allows us to image synaptic transmission,
00:23:35.22 the activation of synaptic glutamate receptors,
00:23:38.07 at the level of individual synapses,
00:23:40.11 and to also measure the calcium signals
00:23:42.21 produced by NMDA receptors
00:23:44.14 at the level of individual synapses.
00:23:46.06 I'll then introduce you to two-photon glutamate uncaging,
00:23:48.27 which allows you to
00:23:51.10 stimulate individual synapses,
00:23:52.28 which is a great experimental convenience
00:23:55.13 and allows one to study the interactions
00:23:57.08 between neighboring synapses.
00:23:59.26 And then a technique called
00:24:01.13 fluorescence lifetime imaging
00:24:03.07 that allows one to track protein-protein interactions
00:24:06.10 within individual synapses.
00:24:08.23 Essentially, it allows one to do quantitative biochemistry
00:24:11.25 at the level of tiny neuronal microcompartments.
00:24:16.23 Now, imaging has a long tradition in neurobiology,
00:24:20.09 going back to Ramon y Cajal,
00:24:22.04 and it's really been at the forefront of neurobiological research.
00:24:25.08 Ramon y Cajal perfected the Golgi Technique,
00:24:28.15 which is a method to sparsely label neural tissue
00:24:31.21 to really come up with the neuron doctrine,
00:24:34.17 to figure out that neurons are
00:24:37.03 the elemental building blocks of brain.
00:24:39.17 And that neurons have [an input side - the dendrite,
00:24:42.28 and an output side - the axon].
00:24:47.13 Another important technique in neurobiology
00:24:49.11 is fluorescence microscopy,
00:24:51.28 which is of course an important tool
00:24:53.27 in other kinds of cell biology,
00:24:56.06 and fluorescence microscopy
00:24:57.26 has high resolution
00:24:59.19 and it has also incredible selectivity;
00:25:02.17 one only images that which one labels.
00:25:04.29 So here for example
00:25:06.15 is a red marker that fills dendrites of this neuron,
00:25:09.17 on a recent cover of the journal Neuron,
00:25:13.06 and also a green fluorophore that is tagged...
00:25:17.21 that tags synaptic proteins
00:25:19.21 that you see as little green punctae
00:25:21.28 along the dendrite of this neuron.
00:25:25.11 But as I've already mentioned,
00:25:26.26 these wide-field standard fluorescence microscopy methods
00:25:30.01 cannot be applied to scattering tissues
00:25:32.19 because the scattering reduces the resolution and contrast
00:25:36.22 of these methods as one images deep in scattering tissue.
00:25:40.21 This problem was essentially overcome
00:25:42.17 by two-photon excitation microscopy,
00:25:45.01 invented about 20 years ago now,
00:25:49.14 which is a laser scanning microscopy technique,
00:25:52.08 and I don't have time to explain how it works,
00:25:55.18 suffice it to say that it brings
00:25:58.11 all of the wonderful advantages
00:26:00.21 of fluorescence microscopy
00:26:03.01 to imaging deep in scattering tissue.
00:26:06.01 So it's a simple laser scanning microscope,
00:26:08.26 but it provides high resolution and contrast
00:26:11.17 when imaging synapses
00:26:14.18 deep in scattering tissue
00:26:16.03 like a brain slice
00:26:17.22 or even in an intact mouse.
00:26:20.09 So this slide just illustrates the kinds of things
00:26:23.15 you can do when imaging in scattering tissues.
00:26:27.23 Two-photon excitation microscopy
00:26:30.01 has high resolution.
00:26:31.28 That's illustrated right here,
00:26:34.00 where we've imaged individual axonal terminals
00:26:37.18 over almost 80 days here
00:26:40.16 -- the time stamp is in days --
00:26:42.16 and we can resolve individual terminals along the axon
00:26:45.18 and these terminals are only a micrometer or so in size.
00:26:48.10 And it's non-invasive
00:26:49.26 because we can image the same structures
00:26:52.06 over a month.
00:26:54.22 It's specific because here...
00:26:57.02 there's illustrated right here
00:26:58.23 another in vivo image, again,
00:27:00.20 an image in the intact brain,
00:27:03.08 where we've introduced two kinds of fluorophores
00:27:05.25 into neurons in the brain:
00:27:07.21 a red fluorescence protein
00:27:09.11 that labels the cytoplasm of the neuron,
00:27:14.06 and a a photoactivatable green fluorescent protein
00:27:18.05 tagged to PSD-95.
00:27:20.17 And at time zero,
00:27:21.29 that green fluorescent protein was activated,
00:27:25.07 and then the decay of the green fluorescence
00:27:28.08 can be used to study protein trafficking
00:27:30.15 even in the intact brain,
00:27:32.19 just illustrating the incredible specificity
00:27:35.18 and selectivity of this technique.
00:27:39.10 And here on the right
00:27:41.17 you see another powerful application of two-photon excitation microscopy,
00:27:44.27 namely it can be combined with calcium imaging.
00:27:47.25 So here the neuron was labeled
00:27:50.04 with a red calcium-insensitive fluorophore
00:27:53.08 and a green calcium-sensitive fluorophore.
00:27:56.18 There are lots of dendrites in this field of view,
00:27:59.17 only one dendrite is labeled with the fluorophore,
00:28:03.08 and we provide a synaptic stimulus here,
00:28:05.16 coincident with the blue square,
00:28:08.17 and you can see these calcium flashes
00:28:11.13 in two dendritic spines,
00:28:12.29 and also a calcium flash corresponding to a shaft synapse,
00:28:16.00 which correspond to synaptic activation
00:28:18.19 of NMDA receptors.
00:28:20.18 And this is really calcium through NMDA receptors
00:28:22.25 that we image in these movies.
00:28:26.09 So let me give you a couple of applications
00:28:29.03 of two-photon excitation microscopy.
00:28:31.16 The first is simply
00:28:33.21 long-term imaging of neural structure
00:28:35.26 to address the question
00:28:37.20 of whether or not new dendritic spines
00:28:40.06 and new synapses are made
00:28:42.05 in the adult brain,
00:28:44.13 and whether or not new synapses
00:28:46.13 are preferentially made in response to novel experience.
00:28:49.24 So here we image
00:28:51.12 a bunch of dendrites
00:28:53.07 and a couple of axons in the somatosensory cortex
00:29:00.19 of living mice,
00:29:03.14 and we image these dendrites
00:29:05.01 every four days.
00:29:06.11 So here is the timeline of the experiment,
00:29:08.16 and what we found in many experiments
00:29:10.05 -- previous experiments --
00:29:12.08 that most spines are stable,
00:29:15.00 perhaps 70% or so persist
00:29:17.11 for days, weeks, and months.
00:29:19.12 But other spines, in particular very thin structures
00:29:22.05 like these very thin, faint hair-like protrusions,
00:29:26.22 some people also call them filopodia,
00:29:29.05 appear and disappear
00:29:30.19 over time-scales of days.
00:29:35.07 And when a new dendritic spine appears,
00:29:39.00 one of the filopodia-like protrusions,
00:29:40.15 we could predict under baseline conditions
00:29:43.02 that it's likely to disappear
00:29:45.17 a couple of days later.
00:29:47.17 It turns out the situations is quite different
00:29:49.25 when we change the experience of the animal.
00:29:52.09 For example, one can
00:29:54.18 just trim a subset of whiskers
00:29:56.17 and the animal has then to explore the environment
00:29:59.16 with a subset of whiskers
00:30:01.05 and the cortical circuit, in response, rewires.
00:30:04.12 So one can teach the animal,
00:30:06.00 train it do a difficult task
00:30:09.12 with its whiskers,
00:30:11.17 so-called perceptual learning,
00:30:13.11 in either case what we find
00:30:14.20 is that these thin protrusions
00:30:16.27 often are stabilized
00:30:18.28 with the growth of a thick, large
00:30:21.14 mushroom-type spine head.
00:30:23.19 And I'll illustrate that right here:
00:30:25.07 so this is the last image,
00:30:27.11 after an 8 day baseline period,
00:30:29.18 this spine just grew
00:30:32.18 between this and the last imaging session,
00:30:35.17 and now you see after novel experience,
00:30:37.24 this dendritic spine now
00:30:40.08 gets a large spine head.
00:30:41.29 Here's another thin protrusion that grows,
00:30:44.09 filopodia-like protrusion,
00:30:46.08 that also appears to be stabilized
00:30:47.29 with the growth of a large spine head.
00:30:51.14 Let's go through this movie clip again.
00:30:53.13 So here's one thin protrusion,
00:30:55.24 a large spine head,
00:30:57.13 another thin protrusion,
00:30:59.00 a large spine head
00:31:00.13 -- large, bright spine head.
00:31:02.25 Now, one question you might ask
00:31:04.18 is whether or not these new spines
00:31:06.12 have anything to do with synapses.
00:31:11.25 And so to answer that question,
00:31:14.10 one has to go beyond optical microscopy
00:31:16.14 -- this really also highlights
00:31:17.29 one of the limitations of optical microscopy,
00:31:20.16 it doesn't quite have the sufficient resolution
00:31:23.03 to resolve synapses --
00:31:26.00 one has to go back to the electron micrograph/microscope
00:31:29.23 and take electron microscope images
00:31:31.17 and reconstruct these synapses
00:31:33.11 and the surrounding neurophil.
00:31:35.01 And we've done that here.
00:31:36.20 So here's the 3D electron microscopy reconstruction
00:31:39.22 of the spine that you just saw grow,
00:31:43.03 and here a blow-up of this reconstruction.
00:31:46.15 Here is the remnant thin protrusion
00:31:49.13 and this large spine head,
00:31:51.02 which is a part of a bona fide synapse,
00:31:53.22 and there are presynaptic structures
00:31:55.13 and so on that are not shown here.
00:31:58.03 So, using these kinds of imaging experiments,
00:32:01.16 we came up with the following scheme
00:32:05.01 for changes in...
00:32:07.18 the following kinds of scheme
00:32:10.11 for changes in neuronal networks
00:32:12.13 over long time-scales in the adult brain.
00:32:14.21 A small population of spines
00:32:16.29 appear and disappear over time-scales of days and,
00:32:19.15 under baseline conditions,
00:32:20.29 most of these spines are subsequently disassembled.
00:32:24.12 But with novel experience,
00:32:26.07 a subpopulation of these new spines
00:32:28.28 is stabilized
00:32:31.22 with an increase in spine volume
00:32:33.24 and presumably synaptic strength.
00:32:36.15 And we think that this change,
00:32:39.06 this conversion from small thin spines
00:32:41.22 to large mushroom spines
00:32:43.28 is associated with what people call
00:32:47.01 long-term potentiations.
00:32:48.17 And I'll show you in a minute,
00:32:50.16 in the next part of the talk,
00:32:53.24 that spine size is very nicely correlated
00:32:56.00 with synaptic strength,
00:32:57.20 and that changes in synaptic strength
00:33:00.01 are really tightly coupled with changes in
00:33:03.18 spine size.
00:33:06.01 Now, we can also use these imaging techniques
00:33:09.11 to study synaptic transmission
00:33:13.13 on a millisecond time-scale,
00:33:14.26 and here we make use of calcium imaging
00:33:16.20 in the following kind of experiment.
00:33:18.13 We stimulate a bunch of axons
00:33:20.07 that are not labeled here
00:33:21.25 and look at calcium signals
00:33:23.10 in individual dendritic spines,
00:33:25.01 under conditions where all of the calcium signals...
00:33:29.03 all of the calcium coming in through dendritic spines
00:33:31.03 is through NMDA receptors.
00:33:32.28 And so the calcium signals
00:33:34.20 now report the activation of NMDA receptors
00:33:37.23 at the level of individual dendritic spines.
00:33:40.11 So they report the activation of NMDA receptors
00:33:42.23 in individual dendritic spines,
00:33:44.14 and NMDA receptors are marvelous for this purpose
00:33:48.03 for two reasons.
00:33:49.18 First, they are, of course, interesting in and of itself,
00:33:51.25 and they also have the highest affinity for glutamate,
00:33:54.26 about a micromolar,
00:33:56.25 of all glutamate receptors.
00:33:58.15 So if a glutamate is around
00:34:01.04 and activates NMDA receptors,
00:34:03.28 we will see it through calcium coming in
00:34:06.18 through the NMDA receptors.
00:34:08.24 In addition, the NMDA receptors
00:34:10.13 have a very large conductance for calcium
00:34:13.14 and we can, with calcium imaging techniques,
00:34:16.20 pick up the opening of individual NMDA receptors,
00:34:19.27 so we have in some sense the ultimate sensitivity
00:34:22.16 in these kinds of experiments.
00:34:24.29 So if we stimulate an axon
00:34:26.22 and look for activated synapses,
00:34:28.15 we find... we see images like these,
00:34:31.21 where you see a stochastic coupling
00:34:34.02 between presynaptic action potentials
00:34:36.10 and postsynaptic flashes of calcium,
00:34:39.10 and that is because not every stimulus
00:34:41.07 produces a glutamate release.
00:34:44.02 Moreover, you see that on...
00:34:49.05 subsequent green flashes of different amplitudes,
00:34:51.24 and this reflects the fact that only a small number
00:34:54.10 of NMDA receptors are activated
00:34:56.28 on each trial.
00:34:58.12 In the example, probably somewhere between
00:34:59.19 two and five NMDA receptors
00:35:02.21 per flash.
00:35:04.18 Notice also that the calcium signals
00:35:06.06 are limited to individual dendritic spines.
00:35:09.13 And that already tells us that the glutamate released
00:35:13.07 by a presynaptic terminal
00:35:14.29 only activates NMDA receptors
00:35:16.28 on a single dendritic spine,
00:35:18.12 it doesn't just spread out willy-nilly
00:35:20.23 to activate NMDA receptors on other dendritic spines.
00:35:24.05 That I think tells us, quite convincingly,
00:35:26.11 that synaptic transmission is point-to-point.
00:35:31.19 In addition, the calcium signals
00:35:34.20 are limited also to individual dendritic spines,
00:35:38.13 and that tells us that the calcium signals
00:35:40.17 downstream of NMDA receptors
00:35:42.21 are synapse-specific.
00:35:44.16 Now, there are some problems with these kinds of experiments,
00:35:48.09 especially if we want to look at spatial relationships
00:35:51.03 between synapses.
00:35:53.09 With these kinds of experiments
00:35:54.27 it's often difficult to find the activated synapse
00:35:57.27 and it's even more difficult to activate synapses
00:36:00.24 with well-defined spatial relationships,
00:36:02.21 and that's because these axons
00:36:05.02 course in an unpredictable way
00:36:07.13 through the tissue.
00:36:08.29 So it was another big step in the study of synapses
00:36:11.10 when it became possible
00:36:13.10 to uncage glutamate efficiently
00:36:15.15 with a two-photon laser,
00:36:17.03 and then the presynaptic terminal
00:36:18.23 could be taken out of the equation completely
00:36:20.29 and we can now mimic synaptic transmission
00:36:23.23 by uncaging glutamate wherever we'd like.
00:36:26.27 And we can mimic synaptic transmission
00:36:29.05 very nicely, as you can see,
00:36:30.23 by comparing synaptically-invoked postsynaptic currents
00:36:34.22 and glutamate uncaging-invoked
00:36:37.04 synaptic currents,
00:36:38.17 which can be made to look essentially identical.
00:36:41.07 So, we can really now replace
00:36:45.13 synaptic transmission with glutamate uncaging
00:36:47.29 and now decide which synapses
00:36:50.17 along a dendrite
00:36:52.06 we might want to activate.
00:36:53.21 And under these conditions too,
00:36:55.03 with glutamate uncaging,
00:36:57.08 the calcium influx into dendritic spines
00:37:00.15 is spine-specific.
00:37:04.12 So you can see that right here.
00:37:07.07 The blue square again corresponds
00:37:09.07 to the uncaging stimulus
00:37:11.07 and now the NMDA receptor-mediated calcium influx
00:37:14.24 is restricted to individual synapses.
00:37:19.00 Alright, I'd like to introduce
00:37:20.20 one last technique in this segment,
00:37:23.18 and that is really concerned with all of the
00:37:26.14 interesting biochemistry downstream of calcium.
00:37:30.05 There are many kinds of calcium-dependent pathways
00:37:33.29 in dendritic spines
00:37:35.11 and some of them are shown right here.
00:37:37.16 Most of them are modulators of synaptic transmission
00:37:40.26 and synaptic plasticity,
00:37:42.17 but there are a couple of pathways which are canonical,
00:37:45.13 in the sense that they're absolutely required
00:37:48.01 both for synaptic plasticity
00:37:50.22 and for structural changes in synapses.
00:37:53.29 And that's the calcium CAMKII pathway
00:37:56.22 and the Ras/MAP kinase pathway.
00:37:58.25 And how do we now track the function
00:38:01.23 and the activation of these different pathways
00:38:04.03 downstream of calcium?
00:38:05.22 And one way to do that is by using FRET
00:38:09.14 to image protein-protein interactions.
00:38:11.28 So FRET can be used to measure protein-protein interactions
00:38:14.26 and one way to do that is to measure, for example,
00:38:17.24 for Ras, the association between Ras and Raf,
00:38:20.24 which is absolutely required for MAP kinase activation,
00:38:24.11 for the function of the Ras/MAP kinase pathway.
00:38:26.25 So how does FRET work?
00:38:29.13 Well, we bind the molecules...
00:38:33.23 we tag the molecules of interest,
00:38:35.19 for example, A and B, which might be Ras and Raf,
00:38:38.02 with fluorescent proteins:
00:38:40.15 green fluorescent protein and red fluorescent protein.
00:38:43.11 And when these molecules are dissociated
00:38:45.24 far apart from each other,
00:38:47.27 no FRET occurs.
00:38:49.11 And if you now excite the donor,
00:38:51.11 the relatively blue of the two fluorophores,
00:38:54.15 so green fluorescent protein, with blue light,
00:38:56.26 we only get green fluorescence.
00:38:58.24 If A and B bind to each other, FRET occurs.
00:39:01.21 That means if we excite the green fluorescent protein,
00:39:04.12 the green fluorescence is reduced
00:39:06.12 and red fluorescence enhanced,
00:39:10.08 through fluorescence resonance energy transfer.
00:39:14.02 This has been incredibly powerful
00:39:16.10 to study protein-protein interactions
00:39:18.09 in intact cells,
00:39:20.06 but it's not very useful for studying protein-protein interactions
00:39:25.02 in tiny microcompartments,
00:39:26.20 in particular, deep in scattering tissue,
00:39:29.19 because it's prone to artifacts.
00:39:32.12 And so, for example,
00:39:34.29 FRET depends on the relative concentration of the two...
00:39:39.04 of the donor and acceptor.
00:39:41.08 And so for these kinds of reasons,
00:39:43.13 we use something like fluorescence lifetime imaging
00:39:46.18 to read out FRET,
00:39:49.00 and fluorescence lifetime imaging only looks
00:39:51.08 at the donor fluorophore
00:39:53.07 and it measures the fluorescence lifetime
00:39:55.16 of the donor fluorophore,
00:39:57.06 which is long when the donor is not bound to acceptor
00:40:01.19 and becomes shorter
00:40:03.10 when the donor gets close to the acceptor.
00:40:05.22 So the changes in fluorescence lifetime
00:40:08.11 can be used as a measure of FRET,
00:40:11.08 and in a highly quantitative manner.
00:40:13.23 So this is then,
00:40:15.28 in a fluorescence lifetime imaging experiment,
00:40:18.12 each pixel really represents
00:40:20.24 one of these fluorescence lifetime histograms,
00:40:25.04 where the brightness of the pixel
00:40:27.22 is essentially the area under the curve,
00:40:31.16 and the hue of the pixel
00:40:35.05 the fluorescence lifetime.
00:40:36.29 And the shorter the fluorescence lifetime,
00:40:39.24 the larger fraction of the donor binds to acceptor.
00:40:44.08 Using these kinds of...
00:40:46.04 and since we only look at donor fluorescence,
00:40:48.19 we can combine this kind of...
00:40:51.06 this technique is insensitive to the relative concentrations
00:40:54.09 of donor and acceptor,
00:40:56.05 we can deconvolve
00:40:58.10 the relative concentrations of donor
00:41:01.16 that is free of acceptor
00:41:03.11 and donor that is bound to acceptor,
00:41:05.18 by measuring the relative fractions
00:41:08.07 of the subpopulations of these two components
00:41:10.26 from the fluorescence lifetime curve,
00:41:13.04 and we can combine this tool
00:41:15.18 with two-photon excitation microscopy,
00:41:18.03 and therefore apply it
00:41:20.05 in scattering tissues.
00:41:21.29 And this is just one example to show you
00:41:23.24 how this works.
00:41:25.06 This is a Ras sensor, and I'll have more to say about that
00:41:27.05 in the next segment,
00:41:29.20 where we elevate intracellular calcium
00:41:31.17 here with the white flash,
00:41:33.18 and you can see that the pixels get hotter,
00:41:35.26 and that means more interaction
00:41:39.11 between Ras and Raf,
00:41:41.21 or a shorter fluorescence lifetime,
00:41:43.15 higher Ras activity.
00:41:46.13 Alright, so let me summarize the imaging tools
00:41:49.17 that I've introduced.
00:41:51.18 I've shown you that we can use two-photon excitation microscopy
00:41:54.08 to image single synapses
00:41:56.20 deep in scattering tissues
00:41:58.20 from milliseconds to months.
00:42:01.23 We can use calcium imaging to measure synaptic transmission
00:42:05.01 and calcium signals
00:42:08.21 at the level of individual synapses.
00:42:11.05 We can use two-photon uncaging
00:42:13.10 to activate any synapse,
00:42:15.22 any postsynaptic dendritic spine we'd like.
00:42:19.24 And we can use fluorescence lifetime imaging
00:42:22.18 to measure FRET
00:42:24.15 to read out protein-protein interactions
00:42:26.16 at the level of individual synapses.
00:42:30.06 I've also given you some results,
00:42:33.06 some important results
00:42:35.11 that were arrived at with these techniques.
00:42:37.05 So, most synapses in the adult brain
00:42:39.26 persist for months;
00:42:41.27 other spines appear and disappear over days.
00:42:45.20 New spines are stabilized, selectively,
00:42:48.20 after novel experience,
00:42:50.29 and new spines often have synapses
00:42:53.23 and then contribute to synaptic transmission
00:42:55.28 and changes in neural networks.
00:42:59.24 I've also shown you,
00:43:01.20 again, using these new imaging approaches,
00:43:03.24 that synaptic transmission is point-to-point,
00:43:06.15 and that NMDA receptor-mediated calcium signals,
00:43:09.24 the calcium signals that drive long-term changes
00:43:12.22 in synaptic transmission like LTP,
00:43:15.07 are limited to individual dendritic spines.

Part 2: Plasticity and Signaling of Single Synapses

00:00:03.24 Good day.
00:00:05.09 My name is Karel Svoboda.
00:00:07.03 I'm a Group Leader at Janelia Farms Research Campus
00:00:09.20 of the Howard Hughes Medical Institute.
00:00:12.17 And in the first lecture
00:00:14.27 I introduced a whole bunch of
00:00:18.24 powerful imaging approaches
00:00:20.21 to study and track the function and plasticity
00:00:23.15 of individual synapses.
00:00:28.14 And now I'd like to show you how we can apply these tools
00:00:31.21 to study individual synapses,
00:00:34.08 calcium signaling in individual synapses,
00:00:36.17 and really focus on the signaling pathways
00:00:39.02 downstream of calcium
00:00:42.17 that determine the synapse specificity
00:00:45.13 and the spatial interactions
00:00:48.05 between synapses.
00:00:50.14 I've shown you movies like these
00:00:53.09 where synapses are activated
00:00:55.27 in a calcium imaging experiment,
00:00:59.14 and you can see that NMDA receptor-mediated calcium influx
00:01:03.13 is limited to individual dendritic spines,
00:01:07.09 and so that...
00:01:09.22 and these kinds of images teach us that
00:01:11.15 synaptic transmission is point-to-point
00:01:14.03 and that the calcium signals downstream of NMDA receptor activation
00:01:18.22 are synapse-specific.
00:01:22.07 This suggests, of course,
00:01:23.26 that the induction of synaptic plasticity
00:01:26.11 might also be synapse-specific,
00:01:28.24 that is, if we activate this synapse
00:01:32.21 strongly to induce synaptic plasticity
00:01:38.09 we see synaptic plasticity only at this synapse,
00:01:40.20 but not at neighboring synapses.
00:01:42.13 Two-photon glutamate uncaging
00:01:45.00 allows us to address and show this directly...
00:01:48.14 to address this point directly.
00:01:50.10 Here is the experiment.
00:01:53.04 We use two-photon glutamate uncaging
00:01:55.10 to induce long-term potentiation, or LTP,
00:01:59.05 of synaptic transmission.
00:02:02.24 We first measure a baseline amplitude
00:02:05.22 of synaptic current
00:02:07.20 and response to glutamate uncaging.
00:02:10.02 Then we induce plasticity
00:02:12.02 and measure the amplitude of the response
00:02:14.06 to uncaging glutamate thereafter.
00:02:17.12 And if we induce glutamate uncaging...
00:02:20.10 if we induce plasticity at one synapse,
00:02:22.20 we probe the current before and after at that synapse,
00:02:26.07 but also at neighboring synapses
00:02:28.21 with well-defined spatial relationships
00:02:31.27 to the potentiated synapse.
00:02:33.28 We call these nearby synapses.
00:02:36.01 They are sort of in the range of 1-4 micrometers apart,
00:02:39.03 so in very small neighborhoods of the dendritic branch.
00:02:43.25 And what you can see here is that if you
00:02:45.04 induce long-term potentiation in one synapse,
00:02:49.01 you get, in fact,
00:02:51.09 changes in the postsynaptic sensitivity to glutamate,
00:02:53.24 but no changes in the postsynaptic sensitivity to glutamate
00:02:57.14 in nearby synapses.
00:02:59.29 Here are the population data,
00:03:02.15 so the potentiated synapses change by about
00:03:05.19 a factor of two,
00:03:07.25 so the spine becomes twice as sensitive
00:03:10.22 to glutamate uncaging.
00:03:12.19 This is likely due to the insertion of AMPA receptors
00:03:16.00 into the postsynaptic dendritic spine,
00:03:18.21 but there's no change in the sensitivity to glutamate uncaging
00:03:25.00 in nearby spines.
00:03:27.11 So that tells us that long-term potentiation
00:03:30.04 is input specific at the level of
00:03:32.28 individual dendritic spines.
00:03:35.11 Notice something else interesting in this experiment,
00:03:38.09 is that the potentiated dendritic spine becomes larger.
00:03:44.04 You can see that because
00:03:46.17 there's a GFP-expressing neuron,
00:03:48.20 the spine in this GFP-expressing neuron
00:03:50.14 becomes brighter,
00:03:52.09 and the brightness of this dendritic spine
00:03:54.04 is proportional to the spine head volume.
00:03:56.22 And if we now look at this spine enlargement,
00:03:59.25 which is associated with LTP,
00:04:02.11 we find that the spine enlargement too
00:04:05.12 is limited to the dendritic spine
00:04:07.14 that receives the LTP stimulus,
00:04:09.12 the potentiated spine,
00:04:11.23 and is completely absent from nearby spines.
00:04:15.18 You can go one step further
00:04:17.20 and correlate the spine enlargement
00:04:20.03 with the change in the postsynaptic current
00:04:22.18 at the level of individual dendritic spines.
00:04:25.10 You can see that there's an excellent correlation:
00:04:28.14 the larger the amplitude of the LTP
00:04:31.17 -- change of the postsynaptic current --
00:04:33.22 the larger the change in the spine volume.
00:04:37.27 So let me summarize these points.
00:04:41.03 So, I've shown you that synaptic transmission
00:04:43.18 is point-to-point,
00:04:45.09 and that the calcium signals
00:04:47.08 coming into dendritic spines
00:04:49.28 in response to NMDA receptor activation
00:04:52.01 are synapse-specific
00:04:53.17 - they're limited to individual dendritic spines.
00:04:56.01 We also understand the mechanism
00:04:59.06 of how calcium is limited to individual dendritic spines,
00:05:01.22 which might seem surprising
00:05:03.23 because calcium is a highly diffusible,
00:05:05.22 a rapidly diffusible small ion,
00:05:08.12 but what happens is that calcium comes in
00:05:11.01 through NMDA receptors
00:05:13.07 and it's just rapidly extruded
00:05:15.15 through the spine head membrane
00:05:17.00 by a highly specialized
00:05:19.17 calcium extrusion mechanism.
00:05:21.07 Calcium simply doesn't have time to dribble out
00:05:23.16 through that narrow, constrictive spine neck
00:05:26.01 into the parent dendrite
00:05:27.29 and out
00:05:31.28 to invade neighboring spines.
00:05:34.08 And I've also shown you that,
00:05:36.10 when we induce long-term potentiation
00:05:37.29 in individual dendritic spines,
00:05:40.06 the effect of the inducing stimulus
00:05:42.13 is limited to the spine
00:05:45.07 that receives the LTP stimulus,
00:05:47.11 so LTP and spine enlargement
00:05:49.07 are also synapse-specific.
00:05:50.26 And we can really conclude that
00:05:52.22 synapses can function as independent units
00:05:55.05 of memory storage.
00:05:58.12 Now, it's not calcium...
00:06:00.23 calcium is absolutely necessary
00:06:02.17 for the induction of long-term plasticity,
00:06:05.16 and NMDA receptors
00:06:07.21 are absolutely necessary,
00:06:09.18 and in fact it's the calcium
00:06:10.01 coming in through NMDA receptors
00:06:12.00 that are critical for the induction of LTP,
00:06:15.18 but it's not calcium
00:06:17.25 that changes synaptic function,
00:06:19.29 it's a biochemical pathway,
00:06:22.00 signal transduction pathways downstream of calcium.
00:06:24.00 In particular, the CaMKII
00:06:26.07 and the Ras/MAP kinase pathways
00:06:27.28 I've already described.
00:06:29.24 We can now look at the function of these pathways
00:06:33.16 in LTP...
00:06:36.11 LTP experiments using protein...
00:06:39.17 by measuring protein-protein interactions
00:06:42.07 by FLIM microscopy.
00:06:44.27 Here is a CamKII sensor
00:06:46.24 that was developed by Ryohei Yasuda's lab.
00:06:49.23 CaMKII, when it's inactive,
00:06:51.22 is folded on top of each other,
00:06:53.27 and what Ryohei Yasuda's lab did
00:06:57.00 is they tagged CaMKII on one terminus with GFP
00:06:59.29 and on the other terminus with a dark YFP.
00:07:03.21 And so there's FRET
00:07:06.20 when CaMKII is inactive
00:07:09.06 and you get a short fluorescence lifetime
00:07:12.04 when CaMKII is inactive,
00:07:15.12 and no FRET when CaMKII becomes active.
00:07:21.03 This sensor can now be expressed in brain slices,
00:07:26.02 in a sparse subset of neurons in brain slices.
00:07:28.24 Again, in these kinds of images
00:07:30.27 there are lots of dendrites in the field of view,
00:07:33.00 but only one neuron is labeled,
00:07:35.12 expresses the CaMKII sensor.
00:07:38.09 And then we can combine fluorescence lifetime imaging,
00:07:41.07 using the CaMKII sensor,
00:07:43.17 with induction of long-term potentiation
00:07:46.18 at the level of individual synapses
00:07:49.03 to ask, what happens to CaMKII activation
00:07:52.11 when LTP is induced?
00:07:55.21 And in particular, whether or not CaMKII activation
00:07:57.24 is limited to the spine
00:07:59.27 that receives the LTP stimulus.
00:08:03.25 So let me roll this movie now...
00:08:08.09 this is the dendritic spine that receives the LTP stimulus and...
00:08:13.08 we'll show that movie again.
00:08:15.13 Here under baseline condition or the LTP stimulus
00:08:17.23 you can see you get a brief transient of CaMKII activity
00:08:21.05 -- redder, hotter colors
00:08:23.05 means more CaMKII activity --
00:08:25.10 and subsequent to this burst of CaMKII activity,
00:08:28.18 induced by the LTP stimulus,
00:08:30.23 you see this typical spine enlargement
00:08:33.23 associated with long-term potentiation.
00:08:36.24 We can also see pretty clearly that
00:08:38.19 CaMKII activity
00:08:40.22 is limited to the stimulated dendritic spine,
00:08:44.05 to the spine that received the stimulus.
00:08:47.22 When we look at population data what we find is that,
00:08:51.10 indeed, although LTP and spine enlargement
00:08:55.25 are synapse-specific,
00:08:58.22 only the spine enlargement only occurs in the stimulated spine
00:09:01.19 -- the spine that receives the LTP stimulus --
00:09:04.04 CaMKII activity
00:09:07.05 is also limited to the activated dendritic spine,
00:09:11.03 and the CaMKII transient
00:09:13.02 is really quite rapid
00:09:15.18 and really precedes the spine enlargement.
00:09:19.21 So we think that the localization
00:09:23.11 of the CaMKII transient
00:09:25.13 after LTP induction...
00:09:27.08 the localization of CaMKII activation after LTP induction
00:09:29.26 likely underlies synapse specificity in LTP.
00:09:36.05 How about Ras?
00:09:38.04 The other canonical pathway, Ras...
00:09:40.16 there are many good reasons to study Ras
00:09:42.11 in the context of synaptic plasticity.
00:09:45.01 Ras, as I already mentioned,
00:09:47.08 is critical for the induction of long-term potential
00:09:49.23 and spine enlargement.
00:09:51.22 It's been shown by many groups,
00:09:53.13 I won't show you that data here.
00:09:55.18 And in neurons,
00:09:57.02 Ras/MAK kinase pathways
00:09:59.01 are downstream of NMDA receptor activation and calcium.
00:10:04.23 And Ras and its modulators,
00:10:07.13 the GEFs that activate Ras,
00:10:11.23 and the GAPs that inactivate Ras,
00:10:14.06 are highly concentrated in the spine.
00:10:16.02 I should say that Ras is a small GTPase
00:10:18.11 that sits in the membrane.
00:10:20.20 When it's inactive it's bound to GDP
00:10:23.23 and then it's activated by exchanging GDP with GTP,
00:10:29.14 catalyzed with GEFs.
00:10:31.25 And activated Ras binds to Raf,
00:10:34.16 which then activates MAP kinase
00:10:36.21 and has many downstream effects.
00:10:42.23 And notice also that the interaction
00:10:45.01 between Raf and Ras
00:10:47.21 is nicely amenable to measurement by FRET,
00:10:51.08 because what we'll do is we'll tag Raf
00:10:53.23 with one fluorescent protein,
00:10:56.01 Ras with another fluorescent protein
00:10:57.20 to construct a FRET sensor for Ras
00:11:00.07 that is optimized for FLIM.
00:11:02.14 And there are other reasons that Ras
00:11:04.13 is of particular interest,
00:11:06.09 because mutations in the Ras pathway
00:11:08.17 are implicated in various forms of
00:11:11.03 mental retardation.
00:11:13.14 So here is the sensor
00:11:15.25 that we developed to measure Ras activity:
00:11:18.05 we call is FRasF,
00:11:20.08 and we tag Ras with EGFP,
00:11:23.06 and the Ras binding domain of Raf
00:11:26.24 with a pair of red fluorescent proteins.
00:11:29.14 And this sensor was really optimized
00:11:31.10 for fluorescence lifetime imaging microscopy
00:11:33.29 and also tuned
00:11:36.17 through many control experiments
00:11:38.25 to be compatible with biology;
00:11:41.00 to not perturb Ras signaling itself,
00:11:43.10 and not to really perturb the function
00:11:46.05 of the neurons that express the Ras sensor.
00:11:49.12 And we also made sure
00:11:51.14 that the sensor actually reports Ras activity
00:11:54.06 with good kinetics,
00:11:56.06 in order words,
00:11:57.28 that the fluorescent signals that we measure
00:11:59.18 with fluorescence lifetime imaging microscopy,
00:12:01.24 that report changes in the Ras sensor,
00:12:06.06 really very nicely correspond
00:12:08.27 to the dynamics of endogenous Ras
00:12:11.15 in the cell.
00:12:13.10 So again we can play the same game:
00:12:15.13 express the Ras sensor in
00:12:19.04 small numbers of neurons in a brain slice
00:12:21.22 and then induce LTP
00:12:23.28 in individual dendritic spines
00:12:25.26 and see what happens to Ras activity
00:12:27.23 and spine enlargement
00:12:29.26 after induction of LTP
00:12:31.07 at the level of individual dendritic spines.
00:12:34.14 And I'll show you a movie
00:12:36.04 where each frame is now one minute apart.
00:12:38.13 We can get away with that
00:12:40.11 because it turns out that Ras activity is slower
00:12:42.14 than CaMKII activity.
00:12:44.18 This is the spine which we induce
00:12:46.13 long-term potentiation in.
00:12:48.07 Now you can see hotter colors,
00:12:49.28 corresponding to increased Ras activity
00:12:53.17 and also the spine enlargement
00:12:56.07 that is associated with long-term potentiation, as before.
00:12:59.11 And we roll this again.
00:13:01.16 Ras activity increases in the dendritic spine,
00:13:03.10 but notice that,
00:13:05.06 in contrast to CaMKII activity,
00:13:07.07 the increased Ras activity
00:13:08.29 is not limited to the spine
00:13:11.02 that receives the LTP stimulus.
00:13:12.28 Rather, Ras activity
00:13:14.19 escapes the activated dendritic spine
00:13:17.08 and spreads through the dendrite
00:13:19.24 and invades neighboring dendritic spines.
00:13:22.03 In fact, Ras activity
00:13:23.24 spreads roughly over 10 micrometers of dendrite,
00:13:27.10 which encompasses about 10-20 synapses
00:13:29.27 in one of these neurons.
00:13:32.17 This is a quantification
00:13:34.06 at the level of populations of experiments.
00:13:37.11 Up here, Ras activity in the stimulated spine in black
00:13:42.05 and in the nearest-neighbor spine in red.
00:13:46.11 And down here, again LTP as before.
00:13:50.07 LTP and spine enlargement
00:13:52.08 are limited to the dendritic spine
00:13:54.10 that received the LTP stimulus,
00:13:56.10 but Ras is not.
00:13:58.09 Ras activity...
00:14:00.09 activated Ras escapes the activated spine
00:14:03.12 and invades neighboring spines
00:14:06.03 to produce rather large changes in Ras activation
00:14:10.12 over about 10 micrometers of dendrite.
00:14:13.29 We can look at the spatial-temporal dynamics of Ras
00:14:17.02 in images like these,
00:14:19.27 where we produce long-term potentiation
00:14:22.09 in one spine
00:14:24.04 and then look at the spatial profile of Ras activity
00:14:28.05 as a function of time.
00:14:31.06 And again we find,
00:14:33.27 even at the level of populations of experiments,
00:14:37.17 that Ras spreads about
00:14:40.24 over 10 micrometers of dendrite
00:14:43.15 and that Ras activity lasts about 5 minutes.
00:14:47.03 So this is roughly the time and length scales:
00:14:49.29 10 micrometers...
00:14:52.19 active Ras spreads about 10 micrometers
00:14:55.14 before it's inactivated,
00:14:57.22 and it also remains active
00:15:00.10 for about 5 minutes.
00:15:04.28 Using modeling approaches
00:15:07.16 that I don't have time to go through with you here,
00:15:10.14 we are pretty sure of what the mechanism is
00:15:13.24 of this spread of Ras and it's really quite simple.
00:15:16.26 We think what happens is that Ras is activated
00:15:19.07 in the activated dendritic spine,
00:15:21.12 in the spine that receives the LTP stimulus,
00:15:24.05 and then active Ras molecules
00:15:26.13 escape the dendritic spine simply by diffusion,
00:15:30.03 and then are inactivated by GAPs,
00:15:33.25 by the inactivating protein,
00:15:35.26 which are distributed relatively homogenously
00:15:37.23 throughout the dendritic arbor.
00:15:40.00 So it's just passive spread
00:15:41.28 of activated Ras
00:15:44.09 which accounts for the spread of Ras activity
00:15:47.08 after LTP induction.
00:15:50.12 So we tentatively conclude
00:15:53.12 that Ras signaling might not be synapse-specific,
00:15:56.10 but might spread to neighboring synapses,
00:15:58.02 and that conclusion is tentative
00:15:59.22 because there are some potential artifacts
00:16:02.01 in these kinds of measurements.
00:16:04.16 I'd like to just through them with you briefly
00:16:08.19 and we call these uncertainty principles,
00:16:10.21 in the sense that any kind of measurement
00:16:12.27 with an intracellular sensor of signal transduction
00:16:16.11 also perturbs the system
00:16:21.13 in rather precise ways.
00:16:23.29 So for example,
00:16:25.22 the Ras sensor that we designed
00:16:27.27 is based on Ras itself.
00:16:30.16 And then by introducing the Ras sensor
00:16:32.10 in the neurons
00:16:34.08 we over-express Ras,
00:16:36.13 and over-expression of Ras could really lead to
00:16:38.05 an artifactual increase
00:16:40.07 in the spread of Ras
00:16:43.07 by two mechanisms.
00:16:45.06 One of them is that
00:16:47.13 more Ras means that maybe the GAPs,
00:16:50.02 the neurons...
00:16:51.20 sorry, the molecules that inactive Ras,
00:16:54.27 are now less able
00:16:57.13 to keep up with activated Ras
00:16:59.17 because there are too many Ras molecules.
00:17:01.15 So the expression of Ras
00:17:03.11 might overcome or saturate the GAPs,
00:17:05.25 and therefore Ras activity
00:17:08.04 might be longer
00:17:09.26 and active Ras might spread over longer distances
00:17:12.21 before being inactivated.
00:17:14.29 Similarly, Ras might normally be bound
00:17:17.28 by immobile or slowly mobile scaffolds
00:17:20.26 and, again, the over-expression
00:17:23.09 of the Ras sensor
00:17:25.07 might overcome and saturate these scaffolds,
00:17:27.02 leading to, again,
00:17:28.15 artifactually increased mobility of Ras
00:17:31.27 in these neurons.
00:17:33.16 So the way we chose to control
00:17:38.08 for these kinds of factors
00:17:40.10 was to keep tracking of the expression of the sensor
00:17:45.22 very precisely,
00:17:48.02 and measure the mobility of Ras,
00:17:51.05 and the inactivation time of Ras,
00:17:53.08 and the spatial spread of Ras activity,
00:17:55.05 all as a function of Ras expression level.
00:17:58.20 And Ras expression level
00:18:00.11 was measured using immunofluorescence techniques
00:18:02.14 that I don't have time to show you.
00:18:05.15 And here is, for example,
00:18:07.13 the measurement of Ras mobility,
00:18:09.24 which was measured
00:18:11.27 using Ras tagged with photoactivatable GFP,
00:18:14.23 and this time constant
00:18:17.10 is simply the time constant it takes,
00:18:19.07 the time it takes for Ras
00:18:21.18 to escape a dendritic spine.
00:18:23.15 And you see that that's really quite
00:18:26.12 independent of Ras expression level.
00:18:28.23 That means that we're not really saturating
00:18:31.15 some kind of Ras scaffold,
00:18:34.09 and even if you extrapolate
00:18:36.12 to zero Ras over-expression level,
00:18:38.16 you can see that you get sort of that
00:18:41.05 this retention time of Ras in dendritic spines
00:18:43.08 is sort of on the order of 4 seconds,
00:18:45.16 and really independent to first order
00:18:49.27 of Ras expression level.
00:18:51.29 We can also look at the timescale
00:18:53.26 over which Ras is activate and,
00:18:55.25 again, that tends to be a weak function
00:19:00.06 of the Ras expression level,
00:19:01.25 but even if we extrapolate
00:19:05.05 to zero Ras expression
00:19:06.29 we find that Ras is still active for several minutes,
00:19:10.04 and isn't really strongly perturbed
00:19:12.28 by expression level.
00:19:14.24 And if we now directly look at the spread of Ras activity
00:19:18.27 as a function of Ras expression level
00:19:21.04 we find that, again,
00:19:22.23 that's at most a weak function
00:19:24.23 of the concentration of the Ras sensor,
00:19:27.21 and therefore we can conclude
00:19:29.25 perhaps a little bit more confidently now
00:19:32.06 that Ras activity
00:19:35.14 is really not synapse-specific;
00:19:37.12 that Ras activated at individual dendritic spines
00:19:39.14 in fact spreads over 10 micrometers of dendrite
00:19:43.08 and invades about 20 synapses.
00:19:47.23 So now the question might be,
00:19:51.28 and it's somewhat paradoxical,
00:19:53.28 how can a small ion like calcium
00:19:58.12 be synapse-specific
00:20:00.15 and be limited to these small microcompartments
00:20:03.10 even though ions are highly diffusible,
00:20:06.00 while a membrane-bound protein like Ras,
00:20:08.23 which is much more sluggish
00:20:10.12 and more slowly diffusing,
00:20:12.08 how can that escape a dendritic spine?
00:20:14.23 Well, what matters here is our two time scales:
00:20:17.27 the retention time
00:20:20.03 - the time over which the signaling factor,
00:20:22.09 calcium or Ras,
00:20:24.19 remain trapped in a dendritic spine
00:20:27.02 before they can escape
00:20:29.00 into the dendritic shaft
00:20:31.03 and on to other spines,
00:20:32.22 and the extrusion time for calcium
00:20:35.04 or the inactivation time for Ras.
00:20:39.19 And really the resolution of this riddle
00:20:42.04 is that the extrusion time for calcium
00:20:43.26 is incredibly fast.
00:20:45.22 A calcium ion comes into the cytoplasm
00:20:48.02 and is extruded 10 milliseconds later,
00:20:50.13 while the retention time for calcium
00:20:52.11 is much longer than that,
00:20:54.02 hundreds of milliseconds,
00:20:55.21 and therefore calcium signaling...
00:20:58.15 calcium ion signals
00:21:00.26 are synapse-specific.
00:21:02.12 In contrast, for Ras,
00:21:04.06 the inactivation time is
00:21:06.26 on the order of minutes
00:21:08.27 whereas the retention time is on the order of seconds,
00:21:10.24 therefore active Ras has a chance to escape
00:21:14.19 the activated spine
00:21:16.24 that receives the LTP stimulus
00:21:18.14 before being inactivated.
00:21:20.04 For CaMKII,
00:21:21.28 the inactivation time is about 30 seconds
00:21:24.15 and the retention time about 10 minutes.
00:21:26.08 For that reason,
00:21:27.27 CaMKII is also synapse-specific.
00:21:30.29 So what I've shown you
00:21:32.29 is that Ras signaling is not synapse-specific,
00:21:35.19 that when we use LTP
00:21:38.03 at individual dendritic spines
00:21:40.05 Ras is activated.
00:21:42.05 Synapses are potentiated and enlarge,
00:21:45.00 but the Ras signals
00:21:47.17 are not trapped in the activated dendritic spine,
00:21:49.24 but actually travel to neighboring spines
00:21:51.19 over about 10 micrometers of dendrite
00:21:54.26 and involving sort of 10-20 synapses.
00:21:59.09 So neighboring synapses see a little bit attenuated
00:22:01.29 and slightly time-delayed
00:22:04.06 sort of puff of Ras activity.
00:22:06.07 In contrast,
00:22:07.24 CaMKII activity is synapse-specific,
00:22:10.07 and it's really the synapse-specificity of CaMKII activity
00:22:13.08 which likely underlies synapse-specificity in LTP.
00:22:17.26 Now, what could be the biological meaning
00:22:21.08 of this spatial spread of Ras?
00:22:24.03 What are the consequences
00:22:26.01 of the spatial spread of Ras?
00:22:28.29 Well, inspired by previous experiments
00:22:32.16 that have shown that interactions
00:22:35.18 between synapses
00:22:37.25 when weak stimuli are given on one set of synapses
00:22:40.12 and strong LTP-inducing stimuli
00:22:42.15 on another set of synapses,
00:22:47.08 we did similar kinds of experiments,
00:22:49.23 but now with micrometer spatial control.
00:22:53.24 We had to again check
00:22:56.11 synapse-specificity and LTP,
00:22:58.22 because these experiments are done under slightly different conditions
00:23:01.09 than the ones I've shown you a few slides ago.
00:23:03.20 But under these conditions, again,
00:23:05.29 LTP is induced at one synapse,
00:23:09.02 and the LTP induction
00:23:11.27 has absolutely no effect
00:23:13.26 on nearest-neighbor synapses
00:23:15.14 just 1 or 2 micrometers away.
00:23:17.16 So induction of LTP and spine enlargement,
00:23:19.20 again, are input-specific.
00:23:22.05 Now what if we tune down the stimulus
00:23:25.08 to be sub-threshold to LTP induction.
00:23:28.20 So, here for example,
00:23:30.21 we reduced the amount of NMDA receptor activation
00:23:33.07 by a factor of 5,
00:23:35.01 and that can be done in various ways,
00:23:36.29 either by reducing the number of pulses
00:23:40.08 with which we activate NMDA receptors
00:23:42.19 or the number of NMDA receptors
00:23:45.15 activated per pulse.
00:23:47.16 In either way,
00:23:49.05 we now only get a transient spine enlargement,
00:23:51.05 which decays down to baseline,
00:23:53.08 so no long-term potentiation.
00:23:54.27 I should say that, under these conditions,
00:23:56.26 NMDA receptor activation is so weak
00:23:59.23 that no Ras is also activated
00:24:02.12 in the spine that receives this kind of weak stimulus...
00:24:06.24 we call this a weak stimulus.
00:24:08.21 Now, what if we
00:24:10.15 pair a weak stimulus on one spine
00:24:12.25 with a strong stimulus in another spine,
00:24:14.01 in a neighboring synapse?
00:24:15.29 And this is the experiment:
00:24:17.22 we first give a strong stimulus at time 0,
00:24:21.16 and then some time later,
00:24:23.07 in this case 90 seconds,
00:24:24.28 we provide a weak LTP stimulus,
00:24:29.03 a sub-threshold stimulus,
00:24:31.08 this kind of stimulus,
00:24:33.06 to a neighboring spine a couple of microns away
00:24:35.09 along the same dendrite.
00:24:37.08 And you can see that after inducing LTP on one spine,
00:24:40.19 we can now, with the weak stimulus,
00:24:43.00 produce long-term potentiation
00:24:44.27 at the weak spine.
00:24:46.29 So what this means is that
00:24:48.02 induction of plasticity
00:24:50.23 at one dendritic spine
00:24:53.05 reduces the threshold for inducing plasticity
00:24:56.15 at a neighboring spine
00:24:58.11 over some dendritic neighborhood,
00:25:01.24 at least a few micrometers or so.
00:25:04.23 Now, does this have anything to do
00:25:06.26 with the spread of Ras?
00:25:08.13 You could imagine that Ras spreads
00:25:11.28 from the spine
00:25:13.20 that receives the strong stimulus
00:25:15.17 to the spine that receives the weak stimulus
00:25:17.26 and now provides the missing Ras,
00:25:20.05 the Ras that cannot be activated
00:25:22.00 by the weak stimulus,
00:25:24.03 now allowing the weak stimulus
00:25:25.24 to trigger long-term potentiation
00:25:28.04 in the neighboring spine.
00:25:30.23 To test whether or not Ras
00:25:33.24 might in fact be involved,
00:25:35.19 we first looked at the spatial and temporal scale
00:25:38.03 of the coupling between synapses
00:25:40.11 -- we call this also crosstalk --
00:25:43.23 and compare that to
00:25:45.29 the spatial and temporal scales of Ras signaling.
00:25:48.01 So we varied the distance
00:25:50.21 between the spine that received the strong stimulus
00:25:52.24 and the spine that received the weak stimulus,
00:25:55.13 and the time between the strong
00:25:57.24 and the weak stimulus.
00:25:59.26 And what we find is that the spatial scale of coupling
00:26:02.20 is about 10 micrometers.
00:26:04.26 Here at 0 is the spine
00:26:07.11 that received the strong stimulus,
00:26:09.05 and you see you can go about 5 microns
00:26:11.10 on either side of the strong stimulus
00:26:13.25 and still allow a weak stimulus
00:26:16.08 to evoke long-term potentiation,
00:26:18.25 and that corresponds very nicely
00:26:20.20 to the spatial spread of Ras activity,
00:26:22.17 illustrated here by this green line.
00:26:24.22 And similarly,
00:26:27.14 if we wait about 5 minutes...
00:26:29.21 we can wait up to about 5 minutes
00:26:31.22 between the strong and weak stimulus
00:26:33.18 to still allow the weak stimulus to produce
00:26:37.19 solid, robust potentiation
00:26:39.27 and, again,
00:26:42.03 so the temporal scale of crosstalk
00:26:44.01 between neighboring synapses
00:26:46.01 matches very nicely to the
00:26:48.15 temporal duration of Ras activity
00:26:51.14 in these CA1 dendrites.
00:26:55.07 Now these are all still correlations of course.
00:26:59.28 We need to do a perturbation experiment
00:27:03.15 to really prove that the Ras/MAP kinase pathway
00:27:07.16 has anything to do
00:27:09.18 with the coupling between synapses
00:27:12.08 with reduction of the induction threshold for LTP
00:27:19.06 in this dendritic neighborhood.
00:27:21.14 So we came up with the following experiment:
00:27:23.13 we reasoned that, well,
00:27:25.23 maybe we can induce the weak stimulus
00:27:31.23 a couple of minutes after the strong stimulus,
00:27:35.06 after the requirement of Ras/MAP kinase activity
00:27:38.07 in the dendritic spine that receives the strong stimulus,
00:27:41.25 and we could then, perhaps,
00:27:44.02 have time to intervene with LTP induction
00:27:49.00 on the weakly stimulated spine
00:27:51.28 using pharmacological methods.
00:27:54.18 And that's what we did -
00:27:59.27 so here's the experiment that we performed.
00:28:02.16 We provided this LTP stimulus
00:28:04.15 at time equals zero
00:28:06.17 and the sub-threshold weak stimulus 5 minutes later,
00:28:10.01 and in between, at 3 minutes,
00:28:12.18 we interfered with the Ras/MAP kinase pathway
00:28:19.13 using a MEK inhibitor.
00:28:22.05 What we find is that in the absence of MEK inhibitor,
00:28:25.04 even after 5 minutes,
00:28:26.27 for nearby spines we can still evoke
00:28:29.04 a good, robust long-term potentiation
00:28:31.28 with the weak stimulus,
00:28:33.16 in the neighborhood of the spine
00:28:35.22 that received the strong stimulus.
00:28:38.04 When we now put the MEK inhibitor
00:28:39.22 on at time equals 3 minutes,
00:28:41.27 it has almost no effect
00:28:43.19 on the amplitude of the LTP
00:28:45.18 for the spine that received the strong stimulus,
00:28:48.04 but the MEK inhibitor
00:28:50.08 dramatically reduces the amplitude of long-term potentiation
00:28:53.19 in the dendritic spine
00:28:55.18 that receives the weak stimulus.
00:28:58.04 This really shows that the spread of Ras activity
00:29:02.02 is involved in reducing,
00:29:04.07 is critical for reducing the threshold for LTP induction
00:29:08.11 over this 10 micrometer stretch of a dendrite.
00:29:14.14 So what I've shown you is that plasticity at one synapse
00:29:17.24 facilitates induction of plasticity
00:29:19.29 at neighboring synapses,
00:29:22.07 so neighboring synapses
00:29:24.19 are coupled by diffusible activated factors
00:29:28.08 including Ras
00:29:30.00 and likely other players.
00:29:32.10 And that coupling,
00:29:34.11 that crosstalk of synapses
00:29:36.04 really extends over about 10 micrometers
00:29:38.07 of dendritic territory,
00:29:39.29 involving 10-20 synapses,
00:29:42.11 and one might speculate
00:29:44.11 what could be the biological functions
00:29:46.28 of this coupling between nearby synapses.
00:29:49.17 And this is only speculation,
00:29:52.08 but this kind of mechanism
00:29:55.09 could be a learning rule that binds semantically related inputs
00:30:00.16 within the dendritic arbors of individual target neurons.
00:30:05.08 So, it's been shown
00:30:07.28 using cellular physiology experiments
00:30:10.21 that neurons that make synapses
00:30:13.05 nearby within the dendritic arbors of the target neurons,
00:30:17.21 when they're coactive,
00:30:19.22 have a much larger impact on the postsynaptic cell
00:30:22.09 than if their synapses are wide apart
00:30:25.00 in the dendritic arbors,
00:30:26.15 and that's because of strong non-linear, super-linear interactions
00:30:31.11 of these synapses within the dendritic arbors.
00:30:35.03 And of course it would be quite powerful
00:30:37.28 if then neurons that process
00:30:41.24 or that encode related input
00:30:44.10 would place their synapses
00:30:46.16 nearby on their target neurons.
00:30:49.22 And so the kinds of mechanisms
00:30:53.20 that crosstalk in synaptic plasticity provides
00:30:56.15 could actually be a learning rule
00:30:58.11 to place these synapses correctly.
00:31:00.02 So, imagine here you have two neurons
00:31:03.03 that have weak synapses on one dendritic branch,
00:31:06.10 interspersed with some other neurons
00:31:08.16 from another group of neurons
00:31:13.08 that really encode unrelated information.
00:31:15.27 Now imagine that neuron 1 and 2
00:31:20.04 have firing patterns
00:31:23.16 that are coactive
00:31:25.27 over this timescale of a couple of minutes,
00:31:28.09 so neuron 1 might induce synaptic plasticity
00:31:30.14 at one synapse,
00:31:32.15 and then if neuron 2 fires sometime later,
00:31:35.13 it might also strengthen those synapses,
00:31:39.20 and so it would bind the inputs of these neurons together
00:31:46.12 at that dendritic location,
00:31:48.05 and perhaps compete away the input
00:31:50.13 from another group of neurons
00:31:52.02 that doesn't show correlated activity.
00:31:54.00 And in this fashion,
00:31:55.22 neurons that have a correlated activity
00:31:59.05 in time
00:32:01.17 that might be semantically related,
00:32:04.10 for example, neurons with related place fields
00:32:09.03 would now occupy
00:32:12.02 and target the same region
00:32:14.28 of the dendritic branch.
00:32:16.24 So in other words,
00:32:18.25 that semantically related inputs
00:32:21.10 would now be
00:32:24.05 targeting small regions of dendrite
00:32:27.09 and then would be effective
00:32:29.11 in driving more complex percepts.
00:32:32.16 So for example, here, one neuron in this cartoon, of course,
00:32:35.22 that says something about the color of orange,
00:32:38.13 another one ball,
00:32:40.12 will drive that neuron then
00:32:43.06 to perceive
00:32:45.16 or to be active
00:32:48.09 with the perception of a basketball.
00:32:52.09 So, I'd like to finally
00:32:55.00 acknowledge my collaborators
00:32:57.09 in the projects that I've shown you.
00:33:00.03 Most of the data
00:33:02.05 that I had time to talk about
00:33:04.28 were collected by Ryohei Yasuda
00:33:07.04 and Chris Harvey.
00:33:08.24 Ryohei is now a professor at Duke;
00:33:11.00 Chris Harvey is a postdoc at Princeton.
00:33:13.06 I've had time to show you bits and pieces
00:33:15.26 from experiments from other postdocs in the lab,
00:33:18.18 in particular Anthony Holtmaat,
00:33:20.06 Linda Wilbrecht,
00:33:21.21 Graham Knott,
00:33:23.11 and Thomas Oertner.

Videos in this Talk
  • Part 1: Optical Studies of Individual Synapses
    Part 1: Optical Studies of Individual Synapses
    Audience:
    • Researcher
    • Educators of Adv. Undergrad / Grad
  • Part 2: Plasticity and Signaling of Single Synapses
    Part 2: Plasticity and Signaling of Single Synapses
    Audience:
    • Researcher
    • Educators of Adv. Undergrad / Grad
Speaker: Karel Svoboda
Total Duration: 1:16:51
Recorded: July 2009
All Talks in Neuroscience

Talk Overview

Neurons are connected to form complex networks by tiny junctions called synapses. Svoboda reviews the process of synaptic transmission and the methods traditionally used to image neurons and their synapses. He then describes powerful new methodologies that allow synaptic transmission and calcium signaling in individual synapses to be measured in intact tissue. This lecture is suitable for neuroscience graduate students.

In the second part of this lecture, Svoboda focuses on how the imaging techniques introduced in his first lecture can be used to study calcium signaling and downstream pathways such as Ras, at the level of single synapses. This lecture is also suitable for neuroscience graduate students.

Speaker Bio

Karel Svoboda

Karel Svoboda

Karel Svoboda is a group leader at The Howard Hughes Medical Institute’s Janelia Farm Research Campus. His lab is studying synaptic plasticity in response to new experiences in the developing and adult neocortex. By developing new biochemical tools and new microscopes, Svoboda’s group has been able to track changes in individual synapses in intact tissue…. Continue Reading

Playlist: Imaging in Biology

Related Resources

Svoboda, K. and Yasuda, R. (2006) Principles of two-photon excitation microscopy and its application to neuroscience. Neuron 50:823-39.

Holmaat, A. and Svoboda, K. (2009) Experience-dependent structural synaptic plasticity in the mammalian brain. Nat. Rev. Neurosci. 10:647-58.

Reader Interactions

Leave a Reply

Your email address will not be published. Required fields are marked *