Exploring how the nervous system produces movement sequences
Date Posted:
March 9, 2020
Date Recorded:
July 9, 2019
Speaker(s):
Andrew Seeds - Institute for Neurobiology, University of Puerto Rico, Medical Sciences Campus
All Captioned Videos CBMM Special Seminars
TROY LITTLETON: Good morning, everyone. Welcome to today's special neuroscience seminar. I'm delighted to introduce Dr. Andrew Seeds for today's seminar. He's been pioneering approaches to understand how sequential motor patterns are selected and executed to perform behavior. And we're in for a treat today, because Andy's work is really providing exciting roadmap for using a combination really cutting-edge techniques, including connectomics, computational modeling, and neurogenetics to begin to explore systems level questions in the fruit fly Drosophila model.
Andy earned his PhD from Duke University working with John York on the biology of inositol phosphate kinases and their role in generating various signaling lipids. He then moved to Janelia Farm to do his post-doctoral training with Julie Simpson. At Janelia, Andy identified and characterized the neural circuitry that underlies sequential grooming behavior in Drosophila. The stereotypical and sequenced behavioral pattern of how flies clean themselves provided a really beautiful sensory motor sequence that could be genetically dissected in a great system.
Both at Janelia Farm and his own lab at the Institute of Neurology at the University of Puerto Rico, Andy has used behavioral genetics and computational modeling combined with circuit tracing and optogenetics to define how neural circuits prioritize and execute specific motor sequences in a sensory-based fashion. His work has revealed some exciting concepts for sensory-based sequence generation, including a hierarchical suppression model that is used to prioritize higher value behaviors when parallel and competing sensory circuits are activated. His recent work has also revealed a persistent trace in sensory circuits that can outlast the initial sensory stimulus and be used to trigger lower priority sequences for later aspects of motor programs. So we're looking forward to learning more about Andy's research program today, so please welcome-- join me in welcoming Andy to MIT. Welcome.
[APPLAUSE]
ANDREW SEEDS: First of all, thank you for hosting me, Troy, and thanks for the introduction. It's good to be here. I used to be at Harvard for a couple of years a long time ago, so I'm familiar with Boston, and it's good to be back here.
So before I begin, I'll just briefly introduce you to some of the people who did the work that I'm going to be talking about. And I'll introduce you to all of these people throughout the talk, but I particularly want to introduce Stefanie Hampel, who is an adjunct professor, and we co-run the lab together. And then also Tony Haddock here, who I won't be discussing much of his work, but he contributes to the lab in many different ways, and he's been working on some compu-- or some automated behavioral tracking that we do in the lab that I won't get a chance to talk about.
So our lab studies how nervous systems form complex behaviors by assembling sequences of different movements. And to illustrate what I mean by a complex behavior, I'll start with a human example. And so this example starts with this team in white moving the ball up the field to score a goal. So here comes a lob pass over these defenders to this player here, who's going to score the goal.
Now, she has a repertoire of different movements that she could perform in order to score the goal. She has the ability to run, she can jump, she can use her foot to shoot the ball into the goal, she could even use her head to hit the ball into the goal. So given this repertoire of different movements that she could perform, she has to be able to select the most appropriate movements in order to score the goal. So what she ends up doing is running, jumping, and then using your head to hit the ball into the goal.
So complex behavior doesn't just involve selecting the most appropriate movement from a repertoire of possible movements, it's also important to be able to assemble those movements into an appropriate sequence. So if you didn't notice, she actually did do a little movement sequence. She first ran in the direction of the ball, she jumped up to meet the ball, and then she used her head to hit the ball into the goal.
So we can watch it again from the perspective of the movement sequence, first, running, next, jumping, and then finishing by using her head to hit the ball into the goal. So from this example, I want you to take two main points about sequential complex behaviors. First, it's important to select the most appropriate movement from a repertoire of possible movements, and then it's important to select those movements in an appropriate sequence. So this isn't just important for scoring goals, but sequential movement makes up essentially everything that we do throughout the day, so understanding the neural circuit mechanisms that drive sequential behavior really gets at the core of understanding our own behavior.
So just a brief introduction to movement control in animals, here, I'm showing you a cartoon of the vertebrate central nervous system that you could essentially divide up into the brain and the spinal cord. Now, brains can be thought of as taking sensory inputs and integrating that with the animal's internal state, so whether or not it's hungry, if it needs to mate, something like that, plus, the animal's past experience, and these things get integrated to select the most appropriate movement. The spinal cord can be thought of as the executioner of the movement, and so humans and other animals have an impressive ability to perform different learned movements, but the kind of movements we study in the lab are those you might consider to be innate, those that are genetically hardwired into the nervous system.
And so one way that you can produce innately-determined movements is through what we call movement pattern generators. And what pattern generators are are neural networks that coordinate the firing of different motor neurons that coordinate the muscle contractions that make up different stereotyped movements. One example of kind of a famous pattern generator would be that for locomotion, so walking or running. But we also have other pattern generators that drive other stereotyped movements, for example, chewing and swallowing, and what we'll be focusing on today, which are grooming or scratching movements.
So this type of organization is not limited to vertebrates, but other animals have similar organization. Today, we'll be discussing the fruit fly. And fruit flies also have brains and their version of a spinal cord, the ventral nervous system, which also has pattern generators that drive different movements, such as locomotion and grooming.
So our lab focuses on two main questions. First, how are specific movements selected? And then second, how are these movements selected in an appropriate sequence?
These remain major outstanding questions in the field of neuroscience for a couple of reasons. First, complex behaviors are formed through the interactions of many different parts of the nervous system, and most animal models don't allow for the precise targeting and manipulation of specific circuit components that would allow us to probe how these different brain regions interact. The second problem is that complex behaviors and sequential behaviors can be difficult to elicit in the lab, and they can also be highly variable. So we study grooming behavior in fruit flies, and the reason we study grooming behaviors is because it consists of a highly stereotyped sequence of movements and it's really the predictability of these movements that allows for a great behavioral readout for probing how different manipulations will alter the sequence.
We use fruit flies, as you'll see, because we have the ability to gain access to genetic tools that allow for precise targeting and manipulation of behavioral specific neurons. So one way that we can easily elicit grooming in the lab with flies is to just make them completely dirty by coating them in this bright yellow dust. So here, you can see this fly here is coated in dust, and what that does is it causes the fly to need to remove that dust from its body using distinct movements that are targeted to specific body parts. I don't know why that cursor keeps-- so here, you can see an example of the fly performing head grooming movements.
And the front legs perform these movements, and you can see as the fly removes dust from its head, it accumulates that dust onto the legs, and that's why you see the fly rubbing its legs together, because it's trying to remove the dust from the legs, so grooming is characterized by oscillations between head grooming and leg rubbing. The fly on the right is performing posterior grooming movements using its hind legs. And here, you can see the fly grooming its abdomen, and then it transitions to grooming its wings.
So during my postdoc in Julie Simpson's lab, I found that coating the body of the fly in dust causes them to groom in an anterior to posterior order. And so here's an example of what that looks like. Here, what I've done is to coat a single fly in dust and allow it to groom its body over the course of 25 minutes, and then what I've done is to manually annotate the different grooming movements that the fly performed over that period of time.
And you can see very early on, after being coated in dust, the flies spend the majority of the time grooming their heads. So you can see those oscillations between the whole head grooming and front leg rubbing. And then at later time points, the fly transitioned to grooming its abdomen, and even later, it transitioned to grooming its wings.
So based on this and other data, we find that grooming occurs in a sequence that starts with the eyes, moves to the mouth parts, at antenna, abdomen, wings, and then finishes with the thorax. So we've been interested in understanding the neural mechanisms by which the nervous system drives this grooming sequence, and so the way that we've figured this out was to take advantage of the Drosophila genetic toolkit. And as I'll show you in the next few slides, what we did to start was to gain control over each of the different movements in the grooming sequence and then use that control to test hypotheses about how the sequence occurs.
So the way that we started was to do a neural activation screen to identify neurons that are involved in grooming. And what we did was we took a large collection of GAL4 lines that were produced by Gerry Rubin's lab at Janelia Research Campus, and these GAL4 lines express in different subsets of neurons. And so here, what GAL4 allows you to do is to both visualize and manipulate neurons that are targeted by a particular line. Here, you can see four examples of different GAL4 lines marked by GFP in green, and in magenta, you can see the synapses or the neuropill of the brain and the ventral nervous system.
So using this collection, we next expressed neural activators in different subsets of neurons and observed the different behavioral changes. So interestingly, what we found was the activation of specific subsets of neurons could drive grooming of specific body parts in the absence of dust. Here's one example of a GAL4 line that when you activate the neurons targeted by that particular line, it grooms its antenna.
And so as long as you're activating these neurons, the flies will groom their antenna. And I've done this for up to 30 minutes. So it's kind of constitutive activation of grooming.
Here's an example of a different GAL4 line. When you activate these neurons, the flies groom their wings. And so once you've identified one of these GAL4 lines, you have to determine which neurons within these broader GAL4 expression patterns are actually mediating that behavior.
And so to do this, we take advantage of an additional genetic tool called split GAL4, which allows you to target the intersection between two different expression patterns. And so using split GAL4, we can often get down to single cell resolution. So here's an example of one particular split GAL4 line that expresses in a single neuron on each side of the brain, and when you activate this neuron, it drives grooming of the antenna. And we'll be discussing this neuron in more detail later on in the talk.
So during my postdoc, I screened through 4,000 different GAL4 lines. I identified 200 different lines that could elicit specific grooming movements, and from this screen, these 200 lines covers every movement that's required to remove dust from the fly's body. And using our split GAL4 approaches, we've so far identified 15 different neural types.
And this type of screening is something we're actively doing in my own lab in Puerto Rico, and these are three of the people who are doing this work. First is Adrian Garcia, who actually did a summer research project in Troy's lab a couple of years ago, Julian Ortiz, and Sonia Soto. So we continue to screen and identify new neurons that are mediating or driving specific grooming movements.
OK, so just to summarize what I've told you so far, grooming consists of distinct movements that are targeted to specific body parts, and we know based on our GAL4 lines that each of these movements can be independently elicited by the nervous system. We also know that these movements are assembled into an anterior to posterior grooming sequence, and so what we wanted to do was to test different hypotheses about how that sequence is driven. One of the first hypotheses we tested was that of an activation chain, which has been kind of the classical model for how behavioral sequences are produced. And the way that the activation chain works is that movements that occur first in the sequence trigger movements that occur next. So eye grooming, in this case, which occurs first in the sequence, would trigger grooming of the antenna.
Based on a number of experiments that I won't go into, we were able to rule out that the activation chain as how grooming occurs. And the next hypothesis we tested was based on the idea that each of these different grooming movements has their own independent sensory inputs. So we could put dust on the fly's eyes and the fly would just groom its eyes and no other body part, or we could put dust on the fly's abdomen, and the fly would groom its abdomen and no other body part.
So in essence, when you completely coat the fly in dust, what that does is to activate all of the different movements in parallel, and now they're competing for output. And because you can only perform one movement at a time, you have to have some mechanism for selecting the first movement and suppressing the others. And so we came up with the a hypothesis that we call hierarchical suppression.
And the way that hierarchical suppression works is that movements occurring earlier in the sequence suppress later ones. So for example, eye grooming occurs first in the sequence, and so if eye grooming is active, it suppresses all of the movements that follow it in the sequence. Another example would be abdominal grooming, which comes after eye and antennal grooming, but before wing and thoracic grooming. And if abdominal grooming is active, it could-- or sorry, abdominal grooming cannot suppress grooming of the eyes and the antenna, but if it's active, would suppress grooming off the wings and thorax.
So the way that we tested this hypothesis was to take advantage of these GAL4 lines that drive grooming of specific body parts. Here, I'm showing one that drives grooming of just the abdomen. So we would artificially drive grooming of the abdomen while simultaneously coating the body of the fly in dust to induce competing grooming movements, and then we would assess how grooming progressed. And so this figure here just shows what we expect to see if our suppression hierarchy is correct.
So coming down the left side here, these are the different movements that we would artificially drive using our GAL4 lines, and then going across the top would be the body parts, which we would assess whether or not the fly has the ability to still groom that body part. So an example would be if we artificially drive grooming of the eyes while simultaneously coating the body of the fly in dust, because eye grooming would have the ability to suppress grooming of all the other body parts, we would expect to see clean eyes, but dust remaining on all of the other body parts. Another example would be abdominal grooming. So if we were to artificially drive abdominal grooming while simultaneously coating the body of the fly in dust, because it comes after eye and antennal grooming, it would fail to suppress grooming of the eyes and antenna, so those would be clean, but then it would suppress grooming of the wings and thorax.
And so when we did the experiment, we found that our observed suppression hierarchy exactly matched that which we had hypothesized. So we came up with a model that could explain how this hierarchical suppression hypothesis works, or how hierarchical suppression works, and this is a three layer model that consists of a sensory layer, a hierarchical layer, and a winner take all layer. So the sensory layer detects dust on each body part.
The hierarchical layer establishes a gradient of activity levels across the different body parts where movements occurring earlier in the sequence have the highest level of activity and movements occurring later in the sequence have the lowest levels of activity. And this hierarchical layer might be a little confusing to you, like what kind of neural mechanism can drive a hierarchical layer that could establish such a gradient of activity levels? And we modeled two different scenarios.
One is that you could just make each body part differ in its sensitivity to dust. So in this case, the eyes would have the highest level of sensitivity to the same amount of dust, and the thorax would have the lowest level of sensitivity. The other way that you could imagine this occurring is through unilateral inhibitory connections between the different units.
So this hierarchical layer then feeds into a winner take all layer that selects the movement with the highest level of activity and suppresses the other ones. So in this case-- in this case, eye grooming would have the highest level of activity. It would win in competition and it's winner take all layer and it would be selected first.
The way that you move through the sequence is the act of grooming the eyes itself reduces the sensory drive on eye grooming, and then now eye grooming has an activity level that's lower of that of the next movement, which would be antennal grooming. Now, antennal grooming wins and competition in this winner take all layer and it's selected next. So many iterations of this would lead to a sequence.
So we've identified a model that can explain how the grooming sequence occurs, and when we looked through the literature to see if similar models have been used to explain movement sequences in other animals, we came across one from the field of psychology that is called competitive queueing. And this competitive queueing model's been used to explain how different human movement sequences-- human movement sequences occur, such as the sequential typing of letters on a keyboard. And both our model and the competitive queueing model have the common feature of the parallel activation of different units that compete for output through suppression.
And this is reminiscent of a scenario that all animals face no matter how simple, in that we all need to perform multiple things at any given time, but we can only do one thing at a time, so we have to be able to prioritize what we're going to do first. So we could take an example of a simple example of behavioral choice, where you have two behaviors, let's say one behavior is escape and one behavior is feeding, and given a particular set of circumstances, both of those behaviors might need to be performed at the same time. Perhaps the animal's hungry, but there's a predator in the area. So these two behaviors need to have some kind of hierarchical association for the animal to select the most important one to survive.
So in this case, escape behavior might be the most important behavior. It would suppress feeding, and then once the animal's safely away from the predator, that escape behavior would be deleted from the queue of behaviors, and now, feeding could occur. And so in kind of the scheme of behavioral choice and sequential behavior, I would place grooming more in this realm, where making the fly completely dirty causes the fly to need to prioritize a particular body part.
And so the sequence is an emergent property of these different hierarchical associations. And so it's been suggested that the kind of neural mechanisms that drive behavioral choice might have given rise to the neural mechanisms that mediate more complex behaviors, such as the sequential typing of letters on a keyboard, or playing soccer, or things like that. And so our hope is that by studying this simple grooming behavior, we might be able to understand more about more complex sequences,
OK, so this model-- it's important to remember that this model can describe how the grooming sequence occurs at an algorithmic level, but there are many different ways by which the nervous system could actually implement this model. So what I'm going to talk about now is the work that we've been doing to understand the neural circuits that drive grooming behavior in an effort to determine how the sequence is produced. And so when we first started getting into the grooming neural circuitry, there were a couple pieces of information already available to us.
So first of all, we knew where the pattern generators that drive this specific grooming movements were located in the fly's nervous system. And I'll remind you that pattern generators are neural networks that coordinate the firing of motor neurons that coordinate the stereotype movements that make up specific grooming movements. And so based on the work of G Laurent and Ari Berkowitz and Malcolm Burrows, we knew that the pattern generators that drive grooming movements are found-- particularly, the head grooming movements are found in this region of the ventral nervous system called the prothoracic neuromeres. We also knew that the pattern generators that drive hind leg grooming movements, so abdomen, wing, and thorax, are found in this region of the ventral nervous system called the metathoracic neuromeres. So nobody's actually identified the neural networks that drive these pattern generators, but we do know where they're localized.
The next piece of information that we knew was that each one of these different movements has their own independent sensory inputs. And so one of the first questions that we asked was, what are the sensory neurons that elicit these specific grooming movements? And so I'll just briefly introduce you to the mechanosensory systems in the fly, because as it turns out, mechanosensory neurons are quite good at stimulating grooming. So first, you may have noticed that the body of the fly is covered in bristles, and at the base of each one of these bristles is a mechanosensory neuron that detects when that bristle is displaced. And so the entire body of the fly has mechanosensory bristles.
The second major mechanosensory type in the fruit fly are chordotonal organs, and these are neurons that are embedded in the appendages of the fly, such as the legs, the wings, and even in the antenna, and these detect when that particular appendage is displaced. So in today's talk, I'll be describing-- whoops. I'll be describing both chordotonal neurons and bristle neurons on the head. So now, we're going to be focusing mainly on head grooming circuits.
So if you remember back to earlier in the talk when I described this screen I did to identify neurons whose activation is sufficient to stimulate grooming of specific body parts, when we looked at the expression patterns of some of these lines, we found that many of them targeted sensory neurons on different body parts. And so I'm going to show you one example on the next slide of a line that expresses in mechanosensory neurons all over the body, and when you activate these neurons, it stimulates grooming of the body. And interestingly, I don't have time to show this, but it also stimulates a sequence that precedes exactly the same as the dust induced sequence.
So here's what that kind of pan mechanosensory line looks like. Can you guys see that all right? The expression pattern of that line is marked by GFP and green, and in magenta, you can see the cuticle of the fly. And so since we're focusing on the head here, I'll just point out that in the antenna here, those are chordotonal neurons that detect displacements of the antenna, and then you can see bristle mechanosensory neurons on different parts of the head. And then there are some mechanosensory neurons on the abdomen, thorax, and along the wing.
So all of those sensory neurons form axon bundles and then project into distinct sites in the central nervous system, and depending on which body part you're coming from, you're going to a particular site in the brain. So these neurons here are mechanosensory neurons coming from different parts of the head. These are coming from the front, middle, and hind legs, and then mechanosensory neurons coming from the wings project here.
And so we've gotten pretty good at targeting specific mechanosensory neurons coming from specific body parts. So here's an example of mechanosensory bristle neurons that are coming from the eyes. And when you activate these neurons, you stimulate eye grooming. And here's another example of chordotonal neurons coming from the antenna, and when you activate these, you get grooming of the antenna. And so I'll be focusing much more on these two types of mechanosensory neurons in the coming slides.
OK, so we've been able to identify sensory neurons on the different body parts whose activation is sufficient to stimulate grooming. And so the next question that we addressed was, how do these sensory neurons actually drive specific grooming movements? And so in our kind of earlier work, we focused mainly on the antennal grooming neurons, and so that's what I'll describe next.
So this story starts when we identified two clusters of chordotonal neurons on the second segment of the antenna, and if you activate these neurons, the fly will groom its antenna. And these neurons are actually-- these clusters are actually part of a larger mechanosensory structure called the Johnston's Organ that detects displacements of the antenna. And different types of stimuli that will activate these neurons include sound, wind, and gravity.
And so these two clusters project into this region of the ventral brain of the fly. And this raises a little bit of a conundrum, because the sensory neurons projects here, but I just told you that the pattern generators that drive the specific grooming movements project to the ventral nervous system. And so we knew that there must be additional circuitry that's required for conveying the sensory stimulus from the antenna to the pattern generators, and so what I'll describe to you is what we call a command circuit or inner neurons downstream of these sensory neurons that command or elicit antennal grooming.
So from our screen, we identified three different additional neuron types whose activation is sufficient to elicit antennal grooming. Here, we're focusing on this ventral region of the brain. And you can see those Johnston's Organ sensory neurons, those chordotonal neurons projecting.
The first inner neuron type that we identified we called BN1, Brain Neuron 1. And that's a single neuron on each side of the brain. We also identified a cluster of neurons that we call BN2 that stay local within the ventral brain, just like BN1.
And then interestingly, we identified a descending neuron that has its dendrites in the ventral brain region, but then sends axons down to the prothoracic neuromeres where those pattern generators are located. And so activation of any of these neurons is sufficient to stimulate grooming. If you computationally align these neurons in the same brain, you can see that they're all very close together and intimately associated, and so this led us hypothesize that maybe these neurons are part of a functionally connected circuit that's responsible for conveying the sensory stimulus on the antenna to elicit grooming.
And so we tested functional connectivity between different neuron pairs using the following experimental approach. So here, I'm showing an example of the Johnston's Organ's sensory neurons, the JOs, in testing whether or not those neurons can activate the BN2 neurons. And so what we do is we express the light-gated neural activator CsChrimson that probably everybody here is familiar with from the work of Ed Boyden. And so we, in this example, were expressing CsChrimson in the JO neurons, and then in the putative downstream BN2s, we expressed GCaMP6m which would detect increased calcium in the active neuron.
And so we explant the entire nervous system and put it in a dish and used two-photon imaging to image BN2. And so the experiment looks like this. When you see the red hashes, that's when we deliver a red light stimulus to activate the JO neurons, and then we observed fluorescence changes in BN2. And you can see with the red light stimulus, you get a nice fluorescence response in BN2.
And so that's just plotted there. You can see the increase in fluorescence with the light stimulus. And so if we do this with the different inner neuron pairs, this has allowed us to form a functional connectivity wiring diagram. And several different interesting features became apparent from this functional connectivity diagram.
First, we found evidence of reciprocal connectivity between BN1 and BN2. And this was very interesting to us, because in work that I don't have time to describe today, we found that activation of this circuit with a brief stimulus induces grooming that can persist for tens of seconds after the stimulus. And so we've been interested in understanding how you can get this kind of persistence within the circuit and this kind of putative reciprocal excitatory motif is one way you could get that persistence. We also found evidence with our imaging studies for a putative inhibitory neuron that's downstream of BN2 that exerts feed-forward inhibition on DN2. And finally, we find that the circuit consists of parallel descending commands, and by that, I mean there are different descending neurons who, if you activate any one of them, you can stimulate grooming of the antenna.
OK, so this wiring diagram shows us the functional connectivity between these different neurons, but it doesn't really reveal to us the actual synaptic connectivity. So what we've been doing in the past year and a half is to do EM reconstructions of this circuit in collaboration with [? Davi Bach, ?] who was formerly at Jenalia Research Campus, but has since moved to Vermont, and also, more recently with Greg Jefferis at Cambridge University. And so I'll just play a brief video that introduces you to this amazing data set. And this movie was produced by Steven Calle, who is a contractor in our lab who does EM reconstructions for us.
And so what [? Davi ?] did was he took the entire fly brain and then serial sectioned it and then imaged using EM the brain and then stitched all of those images together to produce a 3D volume. And so this is what the actual data looks like. You can see that volume stitched together.
And so we can zoom in on particular parts of the brain and identify individual neurons, and then we can follow those neurons through the data set. And so we can trace these neurons, and we can also identify their synaptic connections. And so here's an example of one neuron that's been reconstructed in green.
And so because this is the entire brain, you have the ability to map entire neural circuits within the brain. So this is that neuron completely reconstructed. And because you have the inputs and outputs, you can now identify connected partners, and so you can reconstruct these circuits. So Steven Calle and Stefanie Hampel and postdoc in the lab Katharina Eichler have been reconstructing the grooming neural circuitry from this volume, and basically, what they do is they go through, they identify a particular neuron, and then they identify its synapses and its synaptic partners. So we just start with particular neurons, and then we trace downstream and upstream.
And so the way that we started getting into these circuits was by taking the approach of starting with the sensory neurons and reconstructing in. And so here are those Johnston's Organs sensory neurons that I described to you. And I told you about how those neurons project into the ventral brain. And so we've been able to identify those neurons in this EM volume and reconstruct them. And you can kind of see the comparison between the light level and the EM level.
We've done the same thing with the eye bristle mechanosensory neurons. So these are mechanosensory neurons on the eye that detect stimuli to the eye. And you can see that those mechanosensory neurons form this bundle. They go to a distinct sight in the ventral brain. And we've been able to similarly identify those eye bristle mechanosensory neurons and reconstruct some of them.
So the next thing that we've done is to then reconstruct the postsynaptic partners of those neurons. And so here's just an example of what the internal grooming circuit looks like. Here are those sensory neurons, those JO sensory neurons that are projecting into the ventral brain.
This is the BN1 neuron. That's the single neuron that stimulates grooming of the antenna. And just to give you an idea of the amount of effort that goes into reconstructing one of these neurons, BN1 took [INAUDIBLE] six weeks to reconstruct, and it was around 70,000 clicks. So it's a significant amount of work to do this. More recently, we've been working with the Google segmentation, which is an automated reconstruction of chunks of these neurons, and it's made our lives a little easier.
So that's the JO connected to BN1. Then we find the BN2 neuron in magenta here. So it has its dendrites in the ipsilateral side and it extends an axon contralaterally. And then we've begun reconstructing descending neurons as well. And so that's what the partially reconstructed circuit looks like at the moment.
So this was the functional connectivity wiring diagram that we came up with just using calcium imaging studies, and so the EM reconstructions have now allowed us to revise our wiring diagram and know exactly how the circuit is organized. It's also allowed us to identify additional neurons that we weren't expecting. So we've identified additional descending neurons that we've called DN4 and 5.
OK, so we've been able to describe a circuit that elicits grooming of the antenna. One of the next things that we wanted to know is how these other movements are stimulated. Do they have a similar kind of organization to the antennal grooming circuit?
And so from our screen, we identified GAL4 lines that could stimulate grooming of other parts of the head. So here's the antennal grooming line that I described. But we also identified neurons that stimulate grooming of just the mouth parts of the fly, and other neurons that stimulate grooming of the eyes.
And so if you look at these neurons, what we find is that the circuits, the putative circuits, look very similar to the antennal grooming circuit. Here, I'm showing the eye grooming circuit and the mouth parts grooming circuit right next to the antennal, and I've color-coded the different neurons in the circuit based on whether they're sensory neurons in blue, inner neurons in magenta, or descending neurons in green. So this has revealed to us that the circuits are likely very similarly organized.
And so what we've been doing in the lab is to address this question directly, whether or not the circuits are functionally and organizationally similar. And also, this has stimulated a new direction in our lab to focus on the development of these circuits and address whether or not these circuits have the same developmental origins. So I'm going to give you a few slides that kind of gets into this, but I'll give away the punch line first.
So this is our wiring diagram for the antennal grooming circuit, where you have the chordotonal neurons connected to BN1, and then the BN1 is connected to BN2, but what we found with the other circuits that are largely stimulated by mechanosensory bristles is that they don't have a BN1. So they're a little bit different in that the mechanosensory bristle neurons are actually connected to BN2. So we consider the BN1 on the internal grooming circuit to be somewhat of an adapter neuron that allows the chordotonal neurons to plug in to the BN2 or the antennal grooming circuit. And I'll give you a brief description of what I mean by that.
And so for the rest of the talk, I'll basically be focusing on this connection here, the sensory neuron to BN2 connection. So just to briefly introduce you into a little more depth on the mechanosensory bristles on the head, we've identified driver lines that target the eye bristle mechanosensory neurons, but we've yet to identify drivers that target bristles on other parts of the head, such as the top of the head or the bristle down here. So what we've been doing is, with the help of Adrian, is to do dye fills from different bristles on the head to figure out what their specific projection patterns are so we can identify kind of the somatotopic map of the head.
And so what Adrian has done is to remove specific bristles from particular parts, so maybe bristles on the antenna, and then he fills the mechanosensory neurons to identify their morphology. So for example, the antennal bristles. So the antenna has chordotonal neurons, but they also have bristles on the outside of the antenna which detect different stimuli and induce antennal grooming.
And these neurons have a particular morphology in that they come in from the more top part of the brain and then they project across the midline. So they have this very characteristic pattern. This is in contrast to bristles that are kind of lower on the head that come from-- that project in from the bottom part of the brain and also cross the midline.
And so in conjunction with Adrian's bristle fills and our EM reconstructions, we've been able to kind of figure out that the bristles are actually connected to BN2 type neurons. So here, I'm showing you the antennal BN2 that has a ipsilateral dendritic field and extends an axon contralaterally, and we've been able to identify these bristle mechanosensory neurons connected to that BN2. And it's the same story with BN2 that we found that stimulate other grooming movements.
So in kind of light blue here is BN2 that when you activate this neuron, it stimulates grooming of the eyes. And it has a very similar organization, the dendritic field on the ipsilateral side with the contralateral projecting axon, and we find that it has neurons coming from kind of the top of the head near the eye, and also eye bristle mechanosensory neurons. Similarly, this green BN2 stimulates grooming of the mouth parts, and it is connected with neurons that we've yet to identify, but we hypothesized that these mechanosensory neurons are bristles-- are coming from bristles on the mouth parts.
And so this is what all those BN2s and mechanosensory neurons look like in the same brain. And interestingly, none of these BN2s talk to each other. So they still form these parallel distinct pathways, but they're connected to distinct bristles.
OK, so our current model that we're working on now is that the head has this sematotopic map where bristle mechanosensory neurons coming from different head parts connect with distinct neurons that we call BN2. So just to summarize what we know so far, we know that the antennal grooming circuit and the-- that there are kind of two types of ways that mechanosensory neurons can connect with these circuits. So in the case where the mechanosensory neurons are coming from chordotonal neurons on the antenna, they connect through a BN1-like neuron that then is connected to be BN2, and if the mechanosensory neurons are bristle mechanosensory neurons, they're connected directly to BN2.
OK, so the last thing that I want to mention here is the fact that all of these BN2s seem to be very similar has led us to address the question of whether these circuits arise from common developmental origins. And so what Amelia Merced in the lab has done is to start a developmental project where she hypothesized that neurons in the grooming circuits arise from the same stem cell lineages. And so the ventral part of the brain is established by 13 different stem cell lineages that form these clusters of neurons that have very similar neurite projections.
And so if you look at these BN2 neurons, they all have similar organs-- or similar morphology in that their cell bodies are clustered together, they have the ipsilateral dendritic field with the contralateral axon projections, so what Amelia has done to start is to ask whether or not these neurons arise from the same stem cells. And so based on the work of Jim Truman and Haluk Lacin, we know that particular stem cells give rise to particular lineages in the brain, and so we hypothesized that the DNs were derived from this stem cell, and the BN2s were derived from-- were from a lineage 23b, which was derived from the NB7-4 lineage.
And so each of these different lineages has their own characteristic transcription factor kind of signature. And so what Amelia has done is to do stains to-- for example, for the BN2s, we have hypothesized that it's lineage 23b, and lineage 23b should have UNK4 and ACJ6 positive-- or be positive for UNK4 and ACJ6 transcription factors. And so she's just done some co-stains with-- in this case, these are BN2s that drive grooming of the ventral head. And so in magenta, you can see ACJ6 positive cells, and these cells probably make up-- well, we know they make up the entire lineage 23. And you can see that the ventral head BN2s co-stain with ACJ6, and we can also see that these neurons are UNK4 positive.
So just to summarize, the different circuits that drive these different grooming movements are both functionally and anatomically similar, and we know that these circuits arise from the same developmental origins. So now, what we're doing is trying to identify the developmental cues that give these circuits their kind of specificity. What makes a antennal grooming circuit an antennal grooming circuit, and what makes a ventral head grooming circuit a ventral head grooming circuit.
So just to summarize, we know that we have-- that grooming behavior is formed by parallel circuits that are responsible for eliciting each of the different grooming movements, and we also know that when we put dust-- whoops. When we put dust on the entire body of the fly, somehow, these circuits are prioritized to allow an anterior to posterior grooming sequence. And so some of the work that we're doing now is to try and understand how you get this prioritization. And so one of our main questions is to now try and identify-- now that we have the EM reconstructions of these different circuits, to look for potential connections between these neurons, and also search for neurons that are mediating hierarchical suppression and understand how those neurons impact the circuits to produce a sequence.
And I'll just finish by throwing up the people in the lab. I think I mentioned everybody in here except for Alexandra, who's a new undergrad in the lab who's also doing some of the screening for us. And this is the Institute of Neurobiology. It's located in the middle of Old Town San Juan. And our lab looks out over an old Spanish Fort that's on the other side of this building, and then there's the ocean right there. So any questions?
[APPLAUSE]
have an interactive transcript feature enabled, which appears below the video when playing. Viewers can search for keywords in the video or click on any word in the transcript to jump to that point in the video. When searching, a dark bar with white vertical lines appears below the video frame. Each white line is an occurrence of the searched term and can be clicked on to jump to that spot in the video.
