[00:00:05]
>> Thank you. Thanks a lot below for the introduction and thank you for inviting me it's really a pleasure to be here so what I would like to do today is 1st to say I'm understanding as the audience is pretty broad that you come from different fields please do interrupt and ask questions if anything is unclear or new to you.

[00:00:27]
It's more fun for me of course if it's more of a discussion but. So what I'd like to do today is tell you about some very recent work and ongoing work which is part of a collaboration primarily with Byron you Carnegie Mellon and also with Christian rock and such a problem the work I'm going to be presenting today was the work of a really talented graduate student medo.

[00:00:50]
And a student in my lab a means on Fox. So very broadly what we as North scientists are interested in is understanding how the nervous system takes information from the environment and uses it to guide behavior and action and a really simple example of that mapping from sensory inputs to motor output is the needy reflex arc which is shown for you here so we all know that if you go to the doctor's office and then bang you on the knee that is going to cause a jerking of your foot and that's due to a really simple pathway like an aunt and sister 2 week Ok so a really simple pathway that goes from the need to the spinal cord and then back to the muscles that control the new position the position so this is a really simple input output relationship and critically it's a very fixed relationship so every time that you give the same sensory input you get exactly the same motor output.

[00:01:48]
So fortunately most of our behavior is not reflexive like this but we have some control of how we map different inputs to different outputs. A really nice example of that was a recent study from the new some lab done by. This is a monkey performing a task the monkey task is shown here at the top so the monkey stares at a screen 2 targets come on the screen.

[00:02:12]
As well as a cue that tells the monkey the type of task it's supposed to do the monkey then sees a bunch of dots which are very either in their very both in their color composition from lots of green dots to lots of red dots and also vary in their.

[00:02:29]
Motion So they're either drifting to the left or the right and then depending on the cue that the animal receives it's going to either report the color content of the stimulus or the motion content So here's the evidence for that. On the left here this is when the monkey is doing the motion task when the motion is right word the monkey almost always says that the motion is to the right almost never says it's to the right when the motion is actually moving to the left and importantly the color content of the stimulus whether the motion is being carried by red or green dots doesn't really influence the monkeys behavior then you give a cure to the monkey just change this little symbol of the bottom of the screen and then sort of interleave trial basis the animal can switch so now that its behavior is being governed almost entirely by the color content of the stimulus with very little influence of the motion coherence Ok So the critical thing here is we have a fixed sensory input and we have a very different behavior being driven by different aspects of the stimulus based on the animal's goals.

[00:03:27]
So we'd really like to know how is it that brains as opposed to spinal cords allow us to change our behavior in this way on a moment to moment basis and we think that this is going to be intimately related to this question of how signals are struggling with this pointer.

[00:03:48]
So how are signals selectively inflexibly routed in the brain. Ok so let me tell you a little bit about the system that the study was done in and this is all this in the system that all of the work today is going to be. Performed then which is the visual systems of the visual system of the monkey This is a side view on the left of the Monkey Brain The labelled areas are the ones involved in vision so about 40 percent of the monkey cortex thank you is.

[00:04:18]
There that one try to find out if it's me or the pointer. Ok so. About 40 percent of the visual Court of the cortex is involved in processing different aspects of visual motion what's shown on the right here is an engineer's sort of wiring diagram of the system where each box here is a distinct part of the cortex the lines indicate the number of connections between areas and so you can see that processing visual information involves more than a dozen different areas and we know from decades of work that each of these boxes performs not a unique but somewhat distinct function so there are areas that are more involved in processing color information and more areas more involved in processing motion information so one way that you can get the sort of flexible behavior that you saw in the Monterey study is by selectively routing either color or motion system motion information through this system by changing on a moment to moment basis how different areas communicate with each other.

[00:05:17]
And indeed this is kind of a widespread hypothesis in the field that flexible behaviors involve sort of dynamic reorganization of the way in which different brain regions communicate with each other. It's not just a hypothesis we in fact in many systems including the visual system have some evidence that this is true that is that the interactions between areas can change on a moment by moment basis based on the task so this is one more study I'm going to tell you about which illustrates that and it sets up some of the sort of competing conceptual framework for the work that I'll be showing you so this is again a monkey doing a task stares at a dot a bunch of stuff comes on the screen the monkey sees a color in the center the color indicates to the animal which one of these 3 stimuli it should pay attention to and then at a certain point in time random point in time the brightness of the stimulus will change in the monkey has to report the change of the cued stimulus so this is an attention task where monkeys can learn that the cue is useful for doing the task well and so like you the monkeys will allocate attentional resources whatever that means to focus on this stimulus at this location in space.

[00:06:30]
Now in this work from the doesn't own lab while the animal was doing this task they recorded from 2 brain regions involved in visual processing f e f and v 4 and they're just 2 of those boxes I showed you on the preceding slide and what's plotted here is the coherence of the statistical relationship as a function of frequency of spikes recorded in one part of the visual system with the local field potentials recorded in another and what you can see comparing the blue to the red is when the animal is attending to the stimulus which falls in the relevant receptive fields that the animal the coherence of the relationship between these 2 signals are stronger specifically in these frequencies between about 30 and 70 Hertz Ok so again on a moment by moment basis depending on the animal whether the animal is attending to this stimulus or a different one you can see a change in the functional interaction between these 2 parts of the brain so this is a signature of the something in the brain is changing when the animals task goals are are altering.

[00:07:29]
So we have a hypothesis that brain interactions and changes in those interactions are important we have some evidence that these interactions are robust and we can measure them with electrophysiology So this is sort of the starting or launching point for what we wanted to do and in particular we were kind of troubled by one aspect of a lot of this literature which the previous study is only one example that really in a large number of studies that have done similar experiments and that is that the goal our goal is to understand how populations of neurons in one area interact with populations of neurons in another area but the measurements and the analysis usually focuses on recording one neuron and one brain area and the local field potential in another brain area so the local field potential it is a summary of the activity in this area but it's really unclear how the activity of these different neurons relates to this sensor this recorded signal of the f.p. So this using this as a way to assess inter Li interactions while it's informative to tell you that something has changed really doesn't give you access to the sorts of signals that you want which are the population spiking responses that are actually relayed between these different populations so that's what we wanted to do we wanted to switch away from individual neurons and l.f. piece to understand this entire area interaction problem at the level of spiking responses of groups of neurons located in 2 different structures.

[00:09:01]
Ok. So the thing that we're after then is what are the principles that govern how populations spiking responses in one brain area are relayed or affect another brain area. So this is the sort of the outline of what I'd like to tell you today I'm going to outline for you the approach that we took that is the electrophysiology experiments and some sort of descriptions of that data set and I'm going to talk to you a little bit about the conceptual framework that we adopted to try and address this question and then going to March you through a set of analyses in 4 distinct but interrelated analyses that get at this question of inter-league interactions and they give rise to sort of a new view of how areas might interact differently from moment to moment I'm going to discuss the implications of that and compare it to some alternative schemes namely the ones that involve changes in the oscillation frequency is like you saw in the one of the preceding slides and then if there's time I'll tell you about some published work.

[00:10:05]
Building on this framework looking at feed forward feedback interactions in the cortex. Or it so again this is our model system we're going to be using the mechanics visual system we're going to be recording specifically from the 1st 2 stages of that system areas of the one in v 2 so these are the 1st 2 stages of the monkey visual processing it's cascade these are large areas so each of these is about 10 percent of the cortex in the monkey and they're also very strongly interconnected which is captured here by this big black line.

[00:10:39]
So what we do to record from the want to be 2 is we anesthetize animals we implant them with recording devices which are shown here in the one we place what's known as a Utah or a so this is a grid of $100.00 electrodes one millimeter in length which we insert into the superficial layers of the one and then at the same time we lower a system of tetrodes through v one through the white matter into v 2 and specifically and I'll show you evidence for this in the next slide we can place these tetrodes specifically in the middle layers where the axons of the neurons that live here project so we're recording similar tenuously from an output population and one brain region and exactly from the downstream target one of the downstream targets of that population in featured.

[00:11:28]
So the yield that we get is typically on the order of about $100.00 cells in the one when I say cells I mean both individual neurons and multi-unit clusters and we're going to group those together and then be to our population sizes on the order for about 30 cells.

[00:11:44]
And then while we're. Monitoring these this activity we're going to present drifting sinusoidal gratings the stimuli actually for this talk are not particularly important these stimuli are effective at driving neurons in both of these brain regions and what we're actually going to do is show the same gratings over and over again and look at trial to trial variations in the response to those stimuli which I.

[00:12:07]
Begin to in a in a few more slides. Ok so there are many possible brain regions that you could do this type of work in but I think there are a couple of real advantages to asking questions about Enter early communication in the context of. Using the want to be true to ask that those sorts of questions so the 1st is you know we want you to have both been studied extensively also in an anatomical level so we know a great deal about the connections that go from one brain area to the other the layers that they come from the density of the connections the degree to which they diverging converge so we can use that existing literature to help us interpret the functional measurements will be making with spiking activity.

[00:12:52]
We also know the v 2 very strongly depends on the one for its activity and so the illustration of that is this very old cooling study by Shell or. So this is a little sagittal slice of monkey cortex they lower an electrode into v 2 which is labeled area 18 to confuse you they show a visual stimulus which evokes a response in the neuron they're recording and they then cool the surface of the cortex this is where the one is to transitively and activate that part of the brain when you do that when you activate the one you find that activity in the 2 basically goes away and we know similar results come from permanent lesion studies that if you damage this part of the cortex that is the one you're going to basically remove visually driven activity in the 2.

[00:13:40]
So this is important because it means that when we're recording activity in b. 2 and we're monitoring this output population in the one we're really recording from a group of neurons this functionally important for the neurons that we're. Measuring downstream. And yet. It is getting very very little input from the phallus So in the cat I think as you know both area 18 and 17 can be driven directly from the phallus but in the monkey that's not the case there are a few axons but they don't seem sufficient able to drive activity in v 2.

[00:14:16]
Ok and then the 3rd thing that's really important is that both be one of the 2 because the early visual areas have very nice retina topic maps so that means that we can place our electrode in a way that we can sample from groups of neurons that encode information from the same part of the visual space and therefore are likely to be communicating with each other so this is actual data the blue dots here are the center of the spatial receptor fields of different neurons and the red dots are the corresponding data for v 2 and the shading here just gives you a sense of the size of this spatial receptive field to the extent over which the neurons will respond in space and you can see we can record from groups of neurons that have receptive fields that are very precisely aligned down to a fraction of a degree that actually turns out to be quite important for most of the results that I'll be showing you if you record from groups of neurons whose receptive fields are offset you won't necessarily see this phenomenology that I'll be telling you about.

[00:15:10]
Ok so that's my sales pitch for the one of the 2 let me tell you a little bit about sort of the basic aspects of these paired recordings so I mean some vocally when he started these recordings in my lab did the following thing he implanted the electrode the array and the one he lowered the tax roads into the 2 and as soon as he encountered cells in v 2 he stopped moving the electrodes recorded activity for a few hours and then moved the electrodes a few 100 microns repeated and did that over and over again over about a 24 hour period.

[00:15:43]
And what's shown here on the right are the cross correlation functions that relate the one activity to be 2 activity looking at all possible pairings of a v. one neuron and a v. 2 neuron Ok so these are been filtered in a particular way so that only the brief time scale events are left what you'll notice is that the cross correlation functions at a bunch of depths in v 2 are flat that means there is no statistical relationship between activity and be one and v 2 at those locations but at a subset of locations specifically the you can see that there is a sharp peak meaning that after a v. one cell one of the one cell occurs there is some change in the probability of seeing a spiking event in the 2.

[00:16:24]
And these sharp peaks are evident particularly in the layer for which we know for making lesions during these recordings there specifically the layers that are receiving directly the v one. So only at specific layers of the cortex the ones that anatomically we know are connected do we see a statistical relationship in the activity between these 2 brain regions.

[00:16:48]
This is the typical relationship so the probability of seeing structure like this in the cross correlation function depends a lot on the separation of the receptive fields that I told you about when the receptive fields are offset by a degree or more the chance of seeing this is basically a chance when they're well aligned the probability is still really small so I mean recorded about 300 pair 300000 pairs of cells and then end up with about $300.00 pairs of cells that were sort of directly connected but the probability of those small drops precipitously once you start looking at neurons that aren't functionally aligned.

[00:17:22]
So the sharp eeks are different from the ones that you would observe if you recorded from 2 neurons in the same brain region and that's shown to you here so this is a plot of where the peak of this cross correlation function is for pairs recorded either within 2 neurons recorded within the 12 neurons recorded within the 2 or one neuron and we want to one neuron of the 2 so when you record pairs of neurons in the same brain region what usually find is they tend to fire synchronously some of their spikes and that's consistent with both of those neurons receiving some form of common input both a and b. are driven by some 3rd neuron and so some of their spikes will occur at the same time between be one of the 2 you see a different structure where the peak of the cross correlation function is offset by 3 or 4 milliseconds and that's consistent with neurons in structure a namely the one providing input synaptic input to area b. and therefore elevating the probability that those neurons will fire Ok so this is a functional evidence for this anatomical projection that we know occurs between these layers they were recording from groups of neurons where when the priest and optic side the v one side fires there's a better chance of seeing something happening between us.

[00:18:41]
Yeah so a couple things about that so yes there are massive feedback connections so this is a general feature of the brain. As you know which is every time an x. on goes from area one to area to there is usually an x. on coming in the opposite direction and that's also true between the 2 and be one of those connections do not go specifically between the layers we're sampling here but we do later on see some signatures potentially of a feedback interaction which I may get too importantly the feed forward in feedback directions although anatomically they're equivalent in terms of number of axons functionally in terms of spiking activity they're not meaning if you lesion the one you get no activity and b. 2 like I showed you if you lesion v 2 although there are just as many x. ones going back to the one the one doesn't really care as far as we can tell that is activity one continues as normal and terms of that structure that you're seeing so.

[00:19:37]
There is sort of a trough in a secondary peak some of that is related to the filtering some of it is related to the fact that the neurons have an autocorrelation function so when you detect a peak there's a reduced probability of a spike both.