 For the introduction and thanks for having me here. Is this is this on and working? Yeah, okay. Yeah, I'm really having fun I'm I'm here all week for the course and I'm learning a lot myself Okay, so I'm going to talk about homeostasis in neuronal circuits and What I want to do is first kind of Introduce what do I mean when I talk about homeostasis and then I want to give you just like a brief overview Over some examples of homeostatic processes that occur in different neuronal systems just to show you how widespread a phenomenon it is and that it's not just occurring in the particular model system that I'm studying Although I will then spend about the bulk of the talk on that system Which is the stomatogastric nervous system of crustaceans or lobsters and crabs, which is why I'm wearing this t-shirt and I will end with some conclusions and I think some Some kind of a look forward as to where all of this is going all of this homeostasis stuff so I First heard the term homeostasis. I was thinking earlier today. When did I first hear about homeostasis? And I realized it was kind of late in my Graduate career so that means that none of you need to be embarrassed So I'm just curious about a show of hands who has never heard the term homeostasis There are some there. Okay, so hopefully I can Educate you about a really important physiological process So what is homeostasis? So the source of all knowledge? Wikipedia says that homeostasis is the property of a system Regulates its internal environment and tends to maintain a stable constant condition of properties such as temperature or pH So here we're already getting into biological examples. And so temperature is a prime example We are we have these processes in our bodies that that are geared toward maintaining a fixed temperature and so if if the outside temperature falls lower than what we want to maintain we start shivering and and Producing releasing heat to to raise our body temperature or if it's too hot we start sweating And so that already tells you that What we're talking about are always negative feedback loops were if the the property that we're trying to maintain stable Rises too high we want to have a negative feedback that brings it back down and vice versa. So a very fundamental process so It's kind of obvious that Yeah, sure you want to maintain things in your body stable But I'm going to argue that it's going to be even more important than with other physiological Processes if you look at some neural systems and in particular at the kind of neural system that I'm going to talk about later on Which are which is a central pattern generating system? So these are rhythmically active neural circuits and they are underlying any kind of periodic behavior such as walking or chewing or Breathing and if you think about these examples it already becomes clear why it's so important that you maintain these circuits stable You definitely don't want to miss more than a few breaths or a few heartbeats So that stability is is really vital for in many cases survival of the animal And it's not trivial that that these things can be stable So Walter Cannon who was one of the first who systematically studied homeostasis said that somehow The unstable stuff of which we are composed has learned the trick of maintaining stability So there's kind of this amazement in here. How can this happen? And I want to emphasize that it's really not trivial because you think if you think about all the underlying Molecular components that play into these systems, especially neural systems where you have ion channels and synaptic molecules receptors and release molecules all of these molecules are Inserted into the membrane and stay there for a relatively limited amount of time some of them only hours or days and then they get turned over and so you have a Machinery where all the components are in constant flux, but the output still needs to be maintained stable so this is not at all a trivial task and Furthermore, it's also important because the thinking is that if we understand how this stability Rises and is maintained Maybe we can also learn something about what goes wrong when it's when it fails to be maintained So in diseases, so this is another Big player in the field historic player who says Kind of in a pompous way I think only by understanding the wisdom of the body by which he means homeostasis Shall we attain that mastery of disease and pain? Which will enable us to relieve the burden of mankind. It's kind of a big a big Task here. I don't think what I'm going to show you today is going to relieve the burden of mankind But in general that makes the point that understanding these processes can be pretty important Okay, so I already said that a Basic feature of a homeostatic Mechanism which here is shown kind of from an engineer's point of view is that you have a a feedback loop where you have I Think I pulled this again from the web any system is not exactly a very informative label here But you have a system and let's think of a particular example. Let's think of Let's think of this as say the furnace that's heating your house and You you have a setting that you give that furnace you say I want the temperature to be such and such and it generates an output and You then need to in some way sense that output to monitor am I doing a good job with my furnace and The information provided by that sensor through some feedback system needs to then possibly adjust the control Parameter for that furnace So if you realize if your sensor tells your feedback system that the temperature is really higher than what you require Then you may have to reduce that that setting and so as I pointed out before it's important that this feedback be negative and Those of you who are familiar with engineering know that if you have a feedback system where you actually have positive feedback You tend to get runaway rather than stability right your your temperature is too high and you'll make it even higher and your house Will burn down I guess So two things that I'm going to emphasize a little bit more later in the talk is In neural systems, what exactly is that output that we're trying to maintain? And you'll see that it can be very different things in different systems and in many cases in the homeostasis literature people are kind of a little bit Vague about what they really think that output isn't I think that's in in large part because we don't really know What exactly is it that the neural system cares about so we'll talk a little bit about that and I'll also talk a little bit about different kinds of sensors that could potentially be in play in a neural system to Implement such a feedback loop. I won't get very much into the feedback systems themselves because In some cases, we know kind of the molecular pathways that could be involved there But but that's still much more a black box than some of the other parts Okay, so that was kind of a general homeostasis overview So let's now look at some specific examples and again I will just I'm just going to show you these to give you a flavor of what are different types of homeostatic Processes without going too much into into the molecular detail or the physiological detail but just to give you a sense of what are things that can be maintained and What what parts of the neural system are kind of in play in all of this So the first system that I'd like to just spend two or three slides on is the drosophila neuromuscular junction So I know that there are some very mathy people in the room who may not have much of a biology background So so just very briefly the name says it all right. It's a it's the junction from motor neuron on to the muscle in drosophila in the fly and so The purpose of this junction is and it's a special type of synapse and the purpose of this junction is that a spike in the In the motor neuron is supposed to lead to a contraction in the muscle And so that makes it kind of a unique synapse in that you know in your central synapses as we were talking about earlier today You you have like lots and lots of synaptic inputs Each of which typically makes only a few millivolts of a change in the post synaptic membrane potential And a lot of them have to come together to lead to a post synaptic spike This is very different in that you this this neuromuscular junction is very reliable in that every spike Reliably leads to a contraction of the muscle and then therefore a movement so this is studied a lot in the lab of great Davis at the University of California, San Diego and He has found a couple of different homeostatic mechanisms and in this system and here's of one example Again not getting into the molecular detail just the the overall phenomenon. So what he finds is that? the presynaptic terminal from the Motor neuron you know releases transmitter on to the muscles He's not showing the receptors here and that causes a depolarization in the muscle that eventually leads to a Muscle excitation and contraction and that's how it's supposed to function if he now Manipulates the system and basically impairs the neurotransmitter receptor sensitivity So now these receptors that for some I think did actively it's a little odd that he's not even showing the receptors here But he manipulates them and makes them less sensitive to transmitter And so that means the same amount of transmitter released from the presynaptic terminal will now Will now activate fewer receptors and will lead to a smaller excitation and if he does that the system will react and I have to say homeostatic processes like this can be anywhere from You know half hour several hours two days or longer, but typically much slower than than other neural processes What happens is that the system reacts and says okay? My normal release doesn't cause enough muscle contraction enough excitation I'm going to release more and so there's a homeostatic process right you have overcome the fact that your Small amount of transmitter can no longer activate the muscle enough and you're just releasing more you increase the release And that Gets you back to your set point muscle excitation. So that's your first example of a homeostatic process At the same neuromuscular junction and again from the same lab If they do it another manipulation, they also see a homeostatic reaction So what they do here is they over express a potassium channel in the muscle So what's the consequence of that? We talked about channels and how they influence the membrane potential earlier so one first effect is that if you have more potassium channels your Baseline membrane potential will be more hyper polarized so that takes you further away from threshold for contraction for excitation and contraction and also by simply by having more channels you have more holes in the membrane and So that makes your muscle fiber more leaky and that means that the same the same Synaptic current that comes at the neuromuscular junction will will lead to a smaller Change in membrane potential in the muscle. So basically this over expression here Weakens the ability of your of your motor neuron to excite the muscle and what does it do in reaction? again, it reacts in a homeostatic way it again increases the amount of neurotransmitter that's being released and That not only overcomes the fact that it has to start from a lower potential But also the fact that it has to lead to a total of more synaptic current flowing into the muscle and as an end result It reacts in a way that makes sure that it again is able to reach the threshold for muscle contraction So this thing is really trying to do whatever it can to reliably reach that threshold whenever there's a presynaptic spike Oh, I went the wrong way and the final process at the same synapse. They are really studying this in great detail they can They can manipulate the morphology of this and so here you have your muscle and you have your motor neuron making Synaptic contacts onto that muscle and you can cause the system to be to get hyper Innovated that is to cause to form more synaptic connections than normal onto the muscle fiber And so if each of them did what what it normally does namely release a certain amount of transmitter That hits the post-synaptic receptors and leads to a certain post-synaptic current and therefore a depolarization so if each of these now Super abundant Synaptic terminals did the same thing then you would get over excitation of the muscle And so what what happens instead is that in reaction to this again over over a slow timescale The system adjusts itself and what it now does it again adjusts how much transmitter does it release? It releases less transmitter at each of these synapses But because there's more of them together You will add up now several of these and you'll you'll end up with the same overall post-synaptic effect So again homeostatic and again It seems like the system can do a lot by manipulating the amount of neurotransmitter that's being released But if you go in the other direction something interesting happens. So if you now Cause hypo innovation that means fewer terminals than normal The system Makes sure that those few terminals cause a bigger current in the muscle so that you eventually again reach threshold But they don't do it by releasing more More neurotransmitter instead they now do it by putting more receptors into the post-synaptic membrane so you start seeing that at the at the same synapse in the same system Multiple homeostatic processes and they can use different mechanisms to achieve the same aim But this thing is really insistent. It really wants to make sure that this This transmission is very reliable Okay, moving on to another system. This is work by Gina Turigiano at Brandeis Who is a very big name in the field of homeostasis and in particular synaptic homeostasis and what she does is she does a Lot of her work in these cultured cortical networks. So we heard a little bit about cultured neurons before So you basically take a chunk of cortex There's there's lots of subtleties here including that she has to do this in juvenile animals or it won't really work as well as as It won't work as well in adults you you grind that up and you You put all the individual neurons in a dish and what they will do is they will reform a network in the dish And you now have basically have like a cortex in the dish where you have much more access than in Regular tissue because it's all two dimensional You can access every neuron and you can easily exchange the bath that they are bathed in and Manipulate, you know content of the bath. So if she just cultures I'm saying just this is really not that easy, but it's now an established technique if she cultures these networks and just measures what what are the what is the activity of a Single neuron in this dish. It'll look something like this So this is the membrane potential recorded from one of these neurons you see that occasionally it makes an action potential It also receives these excitatory post synaptic potentials that are sub threshold and you did all of this happens at a certain frequency So that's kind of your Normal activity this is very far from what happens in a real cortex, but it's normal for these cultured cortical networks So now she does one of two manipulations And both of them she does for quite a long time namely for two days So one manipulation is that she blocks electrical activity either by blocking Voltage-dependent sodium channels that you need to make an action potential or by by blocking excitatory Synaptic transmission. So while she does that for these two days, there's no activity So that's just showing that your your manipulation works And the other manipulation is she reduces the amount of synaptic inhibition that also happens in these Networks so they have excitatory and inhibitory neurons and if you do that then obviously during those two days that you apply This drug you will get increased activity here in the form of a higher firing rate So that's why we're going to call this firing rate homeostasis But what happens if you now take off your drug after having applied it for two days? So in the case where you have suppressed activity if you now wash off the drug It turns out that the network is now much more excitable than it was before So now without the drug it makes a lot more action potentials than it did initially It has somehow made itself more excitable and conversely if you have Reduced if you have reduced inhibition and so if you're if your network has been overactive for two days And you then take off the drug then the network by itself It's not really that clear here. You know, I see basically the same number of spikes here But I'm trusting Gina if she works out the the statistics this is going this network is going to end up being less active than than under normal conditions so This is just an observation right this looks like firing rate homeostasis in that if you decrease firing rate It'll react by increasing its excitability if you increase firing rate. It'll go the other way So how could this how could this come about? So let's think about how you could achieve this and this gets back to one in question that upi asked and that is how What could you do to make a network more excitable and at that point we were talking about balance of excitation and inhibition? so that's one thing you could do and And that's basically what's over here So so here's a typical curve that describes how a neurons firing rate depends on the amount of synaptic input it receives so little excitatory input it will not fire at all for Big input it'll fire, you know at its maximum rate that it can go at and then there's this trend this transition in between and The assumption is that the neuron wants to be in this in this range in this what's called the dynamic range Where a difference in synaptic drive can actually be encoded in a difference in spike rate? So you don't want to be over here where you'll always be active or you don't want to be here where you're not active because Then you can't convey any information about the input you receive so you want to be in this target firing range and the one the idea is that if If something happens and you fall out of this target firing range You could get yourself back by adjusting the synaptic drive you receive, you know You could basically say oh, I'm over here. I seem to be too active. I need to reduce The excitation I receive so that's one way you can adjust the excitability The excitation inhibition balance so you work with your synapses to make sure that you're in the right range The other way is that you can work with your neuronal properties. So here again is your typical input output curve and you can So another way you could react is to say I'm going to Change how I respond to the same amount of synaptic input So I'm going to change my cellular properties so that you know if I'm too active. I'm I'm overly active here I'm just going to shift my curve to the right so that For that same synaptic input that made me overly active. I'm again coming back into the dynamic range So two two kind of different ways to do this either adjust your cellular properties or your synaptic properties so which of these happens in the in the cortical systems or and Before we go to the answer as to what is happening This is a way of representing the same thing in the less mathy and more biological way So again, you have your cellular properties your sodium and potassium channels that make your Excitation and that that depolarized and hyperpolarize you and you have your excitatory and inhibitory synapses if you Get deprived of Activity so if your activity in your network gets suppressed you have two ways of reacting to that homeostatically one is You can you can leave your You can make your you can make your Inhibitory synapses now, that's not right I'm getting confused here Sensory so you deprive you deprive this thing Okay, I got it wrong the blue guys are excitatory and the red ones are inhibitory Oh my god, okay, so so you you have this normal network and you deprive it of activity How can it bring itself back one way is that it strengthens its excitatory synapses and it weakens its inhibitory synapses So that's your what was previously on the right You kind of bring yourself back into that firing range or it can change its cellular properties It now inserts more it doesn't mess with the synapses, but it inserts more depolarizing channels and fewer hyperpolarizing channels and makes the cell more excitable and both of these would be ways to homeostatically react to this decrease in excitation So what does this particular system do well? Gina found that it works mainly with the synapses. I'm simplifying her story a Lot here. There's lots of subtle stuff going on But the main effect that she finds is what's called synaptic scaling. So what does that mean? So here is a way of looking at how strong as strong that Synapses are in the system. So what she uses as a measure is what's called these minis miniature excitatory post-synaptic Currents here and so what those are is they are the synaptic current that's caused by the release of a single vesicle from the presynaptic terminal and Those are a good measure for how strong a synapse is right if a single vesicle makes a big current It's a strong synapse if a single has a lot of receptors if if it makes a small current It doesn't have so many receptors and it's a weak synapse So that's just a measure of how strong is your typical synapse in this network in control You get these minis here crunched together in time and this is a typical size of them if you if your network was in Reduced activity conditions for several days then afterwards you find that those minis are bigger So it has made its excitatory synapses stronger And if you have Enhanced the activity for two days then as a result your excitatory synapses become weaker and that's then shown over here in a cumulative histogram, so you have the amplitude of these excitatory post synaptic currents Epscs post synaptic currents and under three conditions. I wish I had a color version of this figure So here's your control cumulative histogram if you have reduced your activity you end up with Histogram like this if you measure this from a lot of neurons in the network So that means your histogram has shifted to two larger amplitudes. These are now These are now stronger synapses and vice versa You get weaker synapses if you have enhanced the activity in the network So that's exactly the same negative kind of feedback that we talked about all along But so now the cool thing that Gina noticed and that's why it's called synaptic scaling is that You can actually take this histogram here and multiply it This is the enhanced activity histogram. You can multiply it by a certain factor You multiply each of these individually measured synapse strengths by a certain factor And you end up basically on top of the control curve or you can take this Histogram that you got when you reduce the activity and you can divide by a fixed factor divide every Amplitude here by a fixed factor and you end up on the original Histogram so what that means is that basically what happened is? When you reduced activity, it's it's as though all the synapses were just increased by the same multiplicative factor or If you have reduced if you have increased activity all the synapses were weakened by the same multiplicative factor Again, there's a lot a lot of subtleties in there, right? We're looking here at Statistics across the whole network. How do we really know that? that you know That that each individual synapse got scaled by the same factor and that's still under active investigation But that's kind of what it looks like at the statistical level so that's pretty cool and a lot of people notice that and that is Something that a lot of people work on and then maybe some of you want to be working on because as uppy told us I Think it was uppy. He told us so many things If you have your normal heavy and learning going on right we learned about heavy and learning if cells fire together They wire together so if one cell keeps exciting another cell It's going to make that synapse even stronger and uppy told us that that can lead to instabilities that you basically and Tend up attend to strengthen and strengthen and strengthen your synapses and you get your network over excited and Going into an epileptic seizure. And so that's what's shown here that these are the effects of unconstrained Long-term potentiations so you start out with a bunch of synapses and this this cell a bunch of times Causes this post synaptic neuron to fire so you will get get long-term Potentiation and you will strengthen this synapse that's why it's fat now and Now the strong synapse Increases the firing rate of the post synaptic neuron and you get a lot of instances where it just so happens that some other neuron Has a couple of pairings where it fired a spike right before the post synaptic neuron fired and and and now your heavy and learning Start strengthening that synapse as well. And then eventually it's all run away and everybody gets really strong and you get over excited So now what Gina and others? Recognized is that synaptic scaling like we've just seen before can actually Counteract this run away or prevent this unconstrained Potentiation, but in a really cool way in the sense that it will preserve Potentially I mean this is at this point still hypothetical and has to be really proven experimentally But it could potentially preserve the memory that you have tried to encode with your long-term Potentiation and your synaptic learning in a network. So how what do I mean by that? So the idea now is that? Going back to this figure So let's say you have a couple of presynaptic terminals onto the same post synaptic neuron They have certain pools of vesicles and receptors And now that some learning event happened and this particular synapse got strengthened through long-term Potentiation and The thinking always is long-term Potentiation is kind of the synaptic correlate of Memory formation right you want to strengthen certain synapses and overall that will encode a memory in your network So you have your strengthening here and you want that strengthening because you want to form this memory But but it has that danger of of overexciting the whole network So the idea is now that if scaling comes in and with a constant multiplicative factor now reduces The strength of all synapses by the same factor So see here you have You have two versus six Receptors so this synapse is three times as strong as this now you're scaling it down by the same factor So that you have one and three You you maintain this relative strength of the synapses, but you're back to your same overall Excitation level so you have basically made sure that you don't get runaway Excitation but at the same time you have preserved the memory trace the relative strength of these synapses So the really cool idea now is again, I want to emphasize This has not been proven, but I think it's very tempting to hope that it is that way the thinking is that you can have heavy and learning in a circuit and encode memories and form memories and Still prevent run away Runaway excitation by synaptic scaling so that you can have the best of two worlds Basically, you can learn something but still maintain your network in a in a in the proper excite excitation range So synaptic scaling prevents unconstrained Potentation this is basically this figure showing the same as the one that I just talked you through So in this in the same cortical networks now a final another form of Homostatic regulation before we move on to the somatogastric system. Did I start at 530? Okay, I need to really speed up. Okay, so Let's let's just skip this let's just say in the same in the same cortical circuit There's also yet another form of homeostatic regulation So what I've shown you now are two systems and these are these are the two I would say best characterized Apart from the somatogastric system, of course In terms of homeostasis, but in both systems you have seen that there are several different processes in play And so I think the the basic take-home message for those of you who are maybe thinking about what could I be working on? That's really interesting. I think that we're at the point now where we have identified Identified and we're starting to understand Individually homeostatic regulatory processes, but now we need a next generation of studies to basically understand How how do they interact when they happen in the same system? And that's not just me saying that we need to do that, but also two of my heroes So here's Gina again, and she said in a recent Paper that has actually entitled too many cooks question mark intrinsic and synaptic homeostatic mechanisms in cortical circuit Refinement she says an important challenge for the field of homeostatic plasticity is to assemble our understanding of these individual mechanisms into a coherent view of how microcircuit stability is maintained during experience dependent circuit refinement So we need to get these individual things into a coherent view And similarly my postdoc advisor Eve Marder Said in a commentary that she wrote on one of great Davis's the the neuromuscular junction guys papers She wrote if as we suspect there are multiple homeostatic mechanisms called into play in maintaining Behavioral integrity it is an important challenge for the future to understand how these are coordinated Okay, so that takes me then to the stomatogastric system, and I'm going to Try to make a case that that's the system that we need to do this in because it's so tractable It has multiple homeostatic mechanisms, and we can do modeling. We can easily do electrophysiology So what is this system? So we're talking here about a nervous system shown in yellow here This is part of the whole nervous system of the animal that governs the movements of lobster and crab Stomachs, so it's kind of wrapped around the the stomach here and It it innervates different muscles that govern stomach movements and those stomach movements are important for the animal not only to to filter food and You know process food, but also even to chew food because lobsters and crabs actually have teeth inside the stomach Which when I first heard it kind of freaked me out a little bit So you can dissect this nervous system off of the stomach and place it in a dish when schematic Where schematically kind of looks like this and the core component of somatogastric ganglion or stg I'm going to I'm sure I'm going to lapse into the abbreviation a lot Is a ganglion which means a knot that contains a small number of neurons about 30 individual neurons You can there are large cell bodies, so you can easily record from them intracegularly and this whole thing Continues to generate the rhythmic patterns that govern stomach movements when you place it in the dish That's what we call a fictive preparation. It generates a fictive Motor pattern basically without any actual muscle movements There are some in anterior ganglia involved here that will come into play in a moment and you can also Even less invasively you can record this rhythmic activity with extra cellular electrodes that you place on the motor nerves So these are the nerves that would normally Deliver the command action potentials to the muscles to lead to contractions So there are two pattern generating circuits in this 30 neuron cell group But we're going to focus on one of them and that's called a pyloric pattern generator, so it's a a central pattern generator like I described earlier it's rhythmically active and It innervates the pylorus which is the part of the stomach that has a filtering system that needs to filter out food particles for further chewing and Basically the circuit can be boiled down to just three components And this is a strength that I alluded to in some of my questions for some of the other speakers We're talking here about what's frequently found in inverted bread systems, and those are that's what's called identified neurons So there is one neuron Called the lateral pyloric neuron of which there exists exactly one copy in every animal and in every animal It generates the same Motor activity the same rhythmic activity and it sends those commands to the same muscles It makes synapses with the same partner neurons, so that's what we call an identified neuron And so what are those three components? We have a pacemaker kernel. That's actually That consists of two neuron types But for all intents of and purposes we can treat them as one and that's rhythmically active It generates these rapid bursts of action potentials the time scale here is about one second So about a one Hertz rhythm and it inhibits these two types of follower neurons the lateral pyloric and the pyloric neuron Don't worry about the names. They're mostly anatomical. I'll stick with the color code as much as I can And so these are all inhibitory synapses and what these follower neurons then do is that they also generate bursts in Rebound from inhibition that they receive from the pacemaker So they get inhibited and then released and they rebound and you have this triphasic pattern 123123 that governs these stomach movements So what does all of this have to do with homeostasis? So this system has a really cool phenomenon that kind of got me and many others Actually into this system. That's why we wanted to work with it. And that's what's called recovery of rhythmic activity. So what you can do is you have your System in them in the dish. You record extra cellular Lee here the rhythmic activity and That's what it looks like when you record it extra cellular Lee on the nerve So again 123123 and now what I didn't tell you is that production of this rhythm depends on the ongoing delivery of neuro modulators Neuro modulatory chemical substances that are released from Neurons that sit in these other ganglia here and they are released down this single nerve into the somatogastric ganglion so you have this one point where you can Manipulate a bottleneck so you can either just physically cut or you can block this nerve So you're you will no longer have those modulators delivered and what happens is the rhythm Either falls completely silent or goes into this tonic activity But no more bursting rhythmic activity. But now the cool thing is that if you wait a day or two Still this is still blocked So you're still lacking those nerve modulators But the system has now somehow reconfigured itself and is again generating the rhythm usually at a slightly slower pace but you know it has overcome a massive injury and It can again generate that vital Rhythmic activity and the animal can again eat and survive So how does that happen? So it has been found over the years that there are several Processes involved in this and I'll try to go through the published ones quickly so that I can get to two new exciting ones That we found in my lab so the first thing is it turns out that these individual neurons are Sensitive to their own electrical activity and can adjust their electrical Properties if that activity is not what it should be. So what we're looking at here. These are voltage clamp recordings of Stomato gastric neurons that were synaptically isolated so they sit in the ganglion But they are no longer receiving any inputs and you can study the properties of that neuron in isolation and Jorge Golo watch who did most of this work what he does here is he's measuring Basically three different currents So this is a potassium current the a type potassium current for those of you who know conductances and so he's applying different Voltage clamp protocols to measure how big is that current when I step the voltage to different Levels and he also measures the sum of these two other potassium currents. This is the delayed rectifier current That's typically you know it from Repolarizing the action potential and then there's also a calcium dependent potassium current and he it's pharmacologically Tricky to measure them individually. So he measures the sum of those two and that's what's shown here And we're looking at before stimulation in black So you have certain amplitudes here and what he finds is that After stimuli I have to tell you what is that stimulation so he has that synaptically isolated neuron sitting in the ganglion And he with his electrode that he has stuck in the neuron actually that's one thing I've been curious about for days How many of you have ever? Electrically recorded from a neuron three four Okay, that okay that puts things into perspective. So this is probably totally cryptic for you, but okay So what he finds is he uses the same electrode that he uses to measure these currents to force these neurons for several hours to Burst at a certain rate that's different from what they would normally do in the ganglion So he imposes an an electrical activity pattern on them and he does that for a couple hours and he sees That when he then again measures these currents they have changed. So this particular current here This potassium current is bigger After he has done this perturbation this activity to perturbation the sum of these two potassium currents is Smaller and that's quantified here. So we're looking here. I Mean to fix this now as we get so these are minutes here. So we're talking one hour two hours and so on and We're looking here at the amplitude of these this current and the sum of these two currents normalized to their initial value Here is when he starts his perturbation forcing the rhythmic activity here is when he stops it and you see the a current increases and that's reversible and Some of these currents decreases and that's also reversible. So We're calling this activity dependent changes in membrane properties So the cell is adjusting how many ion channels of a certain type does it have in its membrane in response to? Manipulation of its electrical activity and why do we say that it depends on activity? That's because you can repeat the same experiment and you Know so it I'm going down the wrong track. Hold on So we're saying it's activity dependent because we have manipulated activity and it changed, right? What I'm getting at is we know that this so what's the sensor here? So it seems like the cell is responding to activity, but how does it know what its own activity is and? We know that that intracellular calcium concentration is a big player in this And that's in a sense if you think about it an ideal candidate because you have voltage dependent calcium channels And so electrical activity with action potentials leads to influx of calcium And so the concentration somehow is a proxy of how electrically active you are But then the calcium concentration is also involved in all of these intracellular pathways that that regulate pretty much everything So calcium is an ideal Candidate to connect electrical activity to molecular changes inside the cell And so the basic idea then for this kind of calcium based activity dependent homeostatic regulation is that you use Calcium concentration inside the cell as a measure for electrical activity You have some kind of target activity level that you want to achieve if you for some reason you fall below You sense that by the fact that your calcium concentration goes down and you react in a number of different ways This is a very basic the basic idea here. What could you do? You could? Decrease the number of these blue channels that are hyperpolarizing channels You could increase the number of these red depolarizing channels or you could weaken Inhibitory synapses right there's less fewer synapses and receptors here Compared to here or you could strengthen excitatory synapses and all of that would bring you back to your target activity level and vice versa Now why do I say this is a basic idea and simplified well here? I'm saying well you want to if you're not active enough reduce all of those those Hyperpolarizing currents what just now I told you that in response to the same manipulation some potassium currents go up and some go down So it's definitely not as simple as that, but we know that it depends on calcium because you can Repeat these same experiments. That's what I wanted to say earlier in the presence You put inside the cell before you do all this a calcium buffer And so now the calcium concentration can no longer change as much when you manipulate activity And you see that those changes either don't happen at all or happen to a much lesser extent So we know from that and other pieces of evidence that calcium is important in all of this And we can model this I'm going to show you Very briefly here and in the next couple slides that in all of these things We always try to have an interplay between experiments and modeling. We can basically model Do a model-based proof of concept that based on calcium Using calcium based activity sensors you can achieve homeostasis of electrical activity so again the molecular pathways are probably very complex your electrical activity Influences your calcium level and that will influence all kinds of intracellular molecular stuff Which eventually can change your membrane properties which determine your electrical activity So here's your feedback loop. Hopefully mostly negative feedback so that the thing is stable. And so here's a model of Very hypothetical calcium based activity sensors. So here you have a model neuron that Three different versions of a model known that generate different electrical activity here It's the voltage spiking here bursting and this kind of spike with a shoulder We're modeling the intracellular calcium concentration during these different activities and here are three examples of very hypothetical mathy Sensors a fast sensor here that responds to every little blip in calcium concentration basically every action potential a slower sensor This green one here that is a little bit sensitive to the action potentials But mostly cares about the burst and then an even slower one. It's called DC here. It's not really Constant, but it's it operates on an even slower timescale and just as proof of principle here is a single model neuron Conductance-based so it has a bunch of different membrane Conductances with differential equations and everything and in this model neuron. We have implemented Calcium-based sensors that then back on to these membrane conductances They basically if the fast sensor is too active It's going to reduce the number of sodium channels because they tend to be involved in fast action potential and so it was basically put together in a kind of common sense kind of way and Look proof of principle this can work So what we're looking at here is the conductances for a bunch of different membrane currents normalized to their initial value we have simulation time in minutes here and Initially, this is what the cell is doing and what it wants to be doing its target activity It was set to tonic spiking at a certain frequency and here at this point one. I applied a perturbation I jumped the Potassium reversal potential up by I think five or ten millivolts. So what that does is initially causes a huge burst of action potentials now these sensors kick in and sense that and start adjusting these Membrane conductances and you see that some of them go up some go down. There's some activity like They change all over the place the activity goes through a couple of different regimes and eventually you settle into new new a new stable situation your Potassium is still jumped. I never jumped back, right? It's the perturbation is still going on But this the system has adjusted has found a new combination of conductances that again produces the target value So it's kind of a proof of principle using a model that You can you can achieve cellular activity homeostasis on the basis of calcium based Sensors on the basis of calcium based. Okay, so we're collecting here homeostatic mechanisms in the somatogastric ganglion We know from this and many other Lines of evidence that there is a calcium based Mechanism that senses electrical activity and adjusts can adjust iron channels in these neurons Activity dependent regulation of cellular properties. Okay Moving on so another one that has Created a lot of excitement in the field over the last couple years is mainly worked by David Schultz in Missouri who is also an offspring of Eve martyrs like myself and Before I go into his data here's more maybe more easily understandable data again by Jorge Gola wash So we're looking here again at the conductances That he measured with voltage clamp in a somatogastric neuron for two different membrane currents Here is again the a type potassium current that we saw before so this is The conductance for that current how many iron channels are in the membrane? This is the H current the hyperpolarization activated current that showed up once or twice before So it's another membrane current and what Jorge and others found is that if you just go into a bunch of different PD neurons in different animals and you measure in each animal What's the conductance for this and this current they are not they both Individually they vary over wide ranges and that's something I'm going to be talking more about on Friday this variability But they don't vary independently. There is a linear relationship. So it seems like The cell is enforcing this rule that it wants a certain ratio between these two important conductances And what David is now showing here? Maybe more interesting for the more molecularly minded here He is basically showing the same phenomenon, but at the level of MRNA So he can pull a single somatogastric neuron out of the out of the ganglion and do Single cell PCR he can measure the copy number of MRNAs for different types of iron channels That's why I was harping on the cell number earlier when we were talking about that So here you can do it with single cells because they are a hundred micrometers cell body diameter Which is easily ten times the diameter of your typical vertebrate neuron If you do that these are different cell types in color coded So, you know PD already, you know LP and these are a couple of other somatogastric cell types and On the two axes of each of these plots are two different Are the MRNAs coding for two different types of iron channels? So shab and shale some of you may know our potassium channels IH mRNA that's the The mRNA coding for this hyperporalization activated channel So there are different combination different pair-wise combinations of iron channels and whenever there is a line group of dots and the Line shown here. It means that that David found a linear correlation between That particular pair in that particular color coded cell type if there is no for example, there's no red here That doesn't mean he didn't measure it. He measured those Those two, but they were not correlated in the lateral lateral I forget lateral gastric neuron So there's a lot of information in this and I don't have time to go into it now but maybe a little bit more on Friday, but bottom line is it seems like what makes a cell type here is Not so much a particular value of a particular Membrane conductance, but it's a set of rules a set of correlation rules between pairs of conductances So the LG neuron once this to be correlated this to be correlated and this to be correlated It doesn't correlate these and other cell types have other correlation rules So it seems like that's what's making a cell type. So why am I talking about this in the context of homeostasis? Oh, let's skip over this because I can talk some more about this on Friday when we talk about More the computational stuff So this is worked by my students. So why Why what does this mean in the context of homeostasis? So remember those Recovery experiments we cut the nerve the rhythm goes away and then it comes back So here's work again by Jorge Where he's now measuring these pair-wise With voltage clamp these pair-wise correlations between pairs of membrane currents in here non Decentralized preparations of preparations where he hasn't cut the nerve and he finds these are all PD neurons. So this one pacemaker neuron type he finds these three and he finds others, but here he's focusing on these three pair-wise correlations If he just keeps the prep in the dish and maintenance it in organ culture for four days without cutting the nerve Non-decentralized here those seem to persist. So that's a stable thing that cell really wants those those relationships but if he Decentralizes the preparation and then waits for the rhythm to recover which happens easily by day four The system has reconfigured itself to again generate the rhythm and it seems like in the process It has given up on some of these correlation rules So we'll talk much more about parameter spaces on Friday But our idea is basically you have this space of parameters of the cell think of the parameters For example as being the maximal conductances for the different ion currents and the cell starts out by sitting in some point And if you look across the population of cells they sit in some sausage in that space, right? They all have a fixed ratio of something and Now if you force the if you massively perturb the cell it loses its activity And it wants to adjust its properties to again produce that activity It seems like it's giving up on these rules and it's starting to wander around more in Parameter space to find again a con a configuration that works Those are concepts that we'll talk about more on Friday and This I won't go into in detail Jorge then asked well What I'm doing when I cut this nerve is really two things one thing is I'm taking away the modulators and that Causes the activity to go away. So which of these two is the system reacting to does it react to the lack of neuromodulators and And and start regulating or does it react to the lack of activity? What he basically showed by teasing these things apart is that in this case This homeostatic process actually reacts Reacts to the lack of neuromodulators. So it's that's why he he started calling it an activity Independent process it it doesn't matter what the electrical activity is what triggers this regulation is The taking away of the modulators. So we're adding a second thing here Okay, is it okay if I go like five minutes over time? Good. Okay, so we now have a second process here It seems to be modulator rather than activity dependent that also affects the ion channels or correlations between them There's a little side thing here with a question mark that I can get into for the aficionados Which I know that there's one here, but only one probably Okay, so moving on to really new exciting processes that I'm really excited about and so are my students We have now in the last year added two more homeostatic processes that also happen in the same system And I have to be very brief the second one is very brief anyway, but I have to cut this one short So what we were saying is basically so far in the somatogastric gangly and all the regulation Processes we know act on the and the cellular properties on the ion channels What about the synapses and nobody had really looked at those mainly for technical reasons because these processes are slow They these homeostatic processes they happen over hours, and so you would basically have to measure synapse strengths over many hours and Do that in a stable way and that's challenging But my students Santiago managed to pull it off. So he's now routinely doing these six-hour experiments With two electrodes stuck into the same neuron. So those of you who have done electrophysiology. It's not easy It's it's hard, but he can do it. So he sticks two electrodes into the pacemaker neuron, and he does one of three protocols Either he does voltage clamp and he just keeps that pacemaker at minus 60 millivolts throughout the six hours That's going to be in blue or he Has it at minus 60 for two hours then jumps to minus 35 and jumps back or Jumps to minus 85 and jumps back. So a depolarizing or a hyperpolarizing perturbation and during those six hours every 400 second every I Think every form no, I think every four minutes he does a little he interrupts this voltage Does a little protocol where he measures this Po the synaptic Currents received through this one particular synapse. So here's another nice feature of the circuit. We have Few identified neurons with known synaptic connections And it just so happens that this pacemaker kernel receives only one chemical synapse And so you can really you know study the same synapse in every animal In in isolation and you can really tease that circuit apart. So he measures that that that synaptic current and from that he can get a Slope which is the synaptic conductance basically the strength of the synapse and he can get the synaptic reversal potential And I can't go through all of the details of this But basically what he finds is if he does this perturbation I Have to say so he's voltage clamping one of three pacemaker neurons that he's he's keeping that cell body at a fixed voltage but because of the extended dendritic trees of those neurons the the other pacemaker neurons There are two PDs and one AB. So the other Neurons are still going so what during all of this one cell body is voltage clamped one post-synaptic cell body But the rhythm is still going on which is actually a neat feature here And so he can observe all kinds of rhythmic features. He sees that the cycle period when you when he depolarizes speeds up that's not surprising When he hyperpolarizes it slows down and he sees Also not surprisingly when he when he starts depolarizing the post-synaptic cell You you suddenly get more chloride flowing in Right you get more chloride because you're moving further away from the chloride reversal potential So more negative ions flow in into the cell So your current your synaptic current jumps up and that leads to an accumulation of chloride in the cell which changes the Reversal potential so he can observe all kinds of features of this activity, but most importantly we were amazed to see Actually the synaptic conductance does change when he does that so the synapse reacts to this depolarizing or hyperpolarizing perturbation by changing its strength and he has worked all of this out Again, I'll get into the modeling part later and he has basically found That the direction in which the synapse changes He basically asked is it homeostatic for anything? You know if that synapse changes as it does does that help maintain the cycle period? Does that help maintain the burst duration of the pacemaker? What what is it in the right direction for any homeostasis and he found that Through a combination of experimental and modeling work. It's wrong It's exactly in the wrong direction for all of this stuff. So why is the synapse doing it? It seems like the only thing That it's homeostatic for it seems to try to maintain the amount of chloride flowing in through the synaptic channels and the post-synaptic chloride concentration so and we have Modeling now and experiments and we're in the process of writing this up So it seems like this is a local homeostatic mechanism It doesn't care about the network activity. It cares about maintaining the amount of Chloride flowing in and the amount of chloride present inside the post-synaptic neuron So that gets added to our list here a chloride dependent by now. I'm so convinced I would strike this question mark Chloride dependent regulation of synaptic properties in the very final one This is worked by my student Amber Hudson. She got Interested in in this particular type of extracellular matrix Molecules so you know that the neurons Are sitting next to each other and there's glia glia cells But then the space around that is filled with what's called the extracellular matrix Which is a mesh of lots of different molecules that have More long thought to just hold the cells together, but are now Increasingly appreciated is having important signaling roles also and one of them is what's called Condroitin sulfate proteoglycans. So those are basically little protein chains that have sugar side chains proteoglycans and There is there is literature from Birds and mammals that that says that on the one hand These molecules get released into the extracellular space to build up that matrix in an activity dependent way So how much gets released depends on how active the neurons are and on the other hand They can influence in a lot of different ways neuronal excitability. So Amber said well That's a cool candidate for a homeostatic regulation Process and what she did indeed find and this is the last data slide So here's your recovery that you're now familiar with right you you have your network ticking along at a certain frequency and if you cut or Block your nerve that rhythm goes away basically and then over the course of hours here So this would be five days The the rhythm comes back typically to a lower frequency So that's your typical recovery experiment. She now discovered that if you pre-treat Before you do the cut if you pre-treat the preparation with an enzyme that eats up the chondroitin sulfate Molecules it chops off those sugar side chains. So it makes that extracellular matrix part that part of the matrix breaks it apart Nothing happens acutely like she's treating it with the enzyme here. The rhythm just keeps going But if you now do your de-affentation your decentralization you cut that nerve the rhythm Doesn't come back as much as it normally would so somehow without interfering with the ongoing rhythm You have interfered with the ability of the system to reconfigure itself and again generate that rhythm and that's quantified here So your frequency Recovers to a certain level in control conditions it recovers to a significantly Lower level and later actually when you treat with the enzyme and this is another control if you treat with a denatured enzyme that you have made dysfunctional by heating it up That enzyme is not effective and you it behaves like control so that adds another process here We now have an extracellular player these chondroitin sulfate proteoglycans that through some membrane Proteins can talk to the intracellular milieu and they seem to be also Doing something to this. This is speculative at this point. That's part of a grant that I just submitted We think that they're probably also acting on the cellular rather than the synaptic properties just for Parsemonious reasons at this point, but that's part of what we're trying to test in the grant. So here's the summary Okay, so in I've shown that homeostatic Processes occur in lots of neural systems. I am tempted to say in all neural systems and You often have multiple homeostatic mechanisms in a given system. They can be Very there's a broad range of those processes. They can be Activity dependent or activity independent. They can act on cellular properties or synaptic properties They can be local like they care about that particular synaptic environment that one synapse or they can be global and care about the overall network activity And there is intracellular and extracellular stuff involved and in all of this I didn't get as much into the modeling as I wanted to but I will on Friday in all of this it's helpful to have kind of a crosstalk between experiments and modeling and so again my kind of my Message for you know, what what would be cool stuff to do in the future I think now we're in a position where we we can start asking well What happens when you have multiple processes in the same system? What if they have What if they have different aims like what if this guy says I need to strengthen the synapse and this one says I need to weaken it Can one of them overrule the other is maybe is are some of them sufficient? Are all of them necessary to achieve network stability I would speculate that probably some of them are better at compensating for this type of perturbation some better at compensating for that type Also, they could be acting on somewhat different time scales. So there's lots of open questions of how do you put all of these? All of these regulatory pathways together so that overall you have a stable system Okay, so that's that was my story. I wanted to thank you for your attention for ten minutes over time I Showed data briefly from my grad student Santiago and other grad student amber and I also want to acknowledge funding sources Down here. Thank you very much So that is a very big question that a lot of people are thinking about what How do you create that set point and especially like if you think about the molecular biology? how can you use molecular biology to create a set point and We really don't know at this point There are some ideas out there that I think maybe offline we can talk a little bit more because it's it's a pretty complex question, but at this point we we simply don't know now This is not going to be very helpful But in the models what we typically do is like we have it We have a differential equation for example for For the strength of that synapse I Totally rushed through that but Santiago did a little bit like a very basic model of that Chloride-dependent regulation of the synapse strength and we basically set up a very simple differential equation that changes the strength of the synapse Depending on the chloride concentration and and you can easily just stick a target value in there Right by by saying I'm going to increase it if it's bigger than this value I'm going to decrease it if it's smaller, but what does that it? What does that mean? Molecularly is really a big question. So I It's kind of giving away a secret, but I I reviewed a proposal by Gina Recently where she had some very interesting ideas in that proposal that get exactly at that question I Don't think I can really show you that proposal that She has a really cool figure there and I also don't think I remember enough to reproduce it in my head but there are ways to kind of set up molecular pathways that would ensure that you you have some kind of a set point that if you're say your Signaling molecule concentration your whatever your car modeling concentration is larger than such and such you can have you can have a Process that reduces it and then you can have another process that increases it if it's smaller than such and such And you can basically set up a molecular A molecular set point that way now whether that's really what's underlying this we don't really know Measuring it There are definitely other candidate things so I already I had to speed up at that point I showed you that they're That there are processes that are not thought to be looking at electrical activity and therefore probably not at calcium But that are dependent on the presence or absence of modulators and there's there's another process that I I didn't mention My colleague Pete Wenner who is in the physiology department at emory he studies Regulation of synapse strength in the developing chick spinal cord So he has these Frankenstein preparations with the spinal cord and the legs attached and they can record from them But that's a different story but what he what he actually showed in a in a nice pair of PNAS papers is that There is homeostatic regulation of those synapses But he could show that it's not triggered again like like here. It's not triggered by the by elect by the electrical consequences of synaptic transmission But it's triggered by the absence or presence of modulator of transmitter in that case So I think that some of some things rely on Calcium probably the things that care about electrical activity other things rely on concentrations of molecules There's also one slide in the cortical culture thing that I Glossed over is kind of using what what people think of as kind of a global activity signal And that is that BDNF, which is brain-derived neurotrophic factor is released by cortical neurons in an activity dependent way So the the more excited they are they more they release BDNF and then you have a network wide Concentration and if that concentration rises it increases the strength of inhibitory synapses and decreases the strength of excitatory synapses again working on that balance to bring you back and so there the signal is clearly not calcium, but the extracellular concentration of some signaling molecule Okay, this is going to be great for Friday. So on Friday. I'm going to talk a lot about Neuronal variability Exactly getting at that question and that will be variability both in the in terms of the properties You know, we already saw some of these plots were the conductance for a particular ion channel varies widely And the other one varies widely, but they're correlated. So that's properties of the Neurons, but what also maybe your question is getting more at the output characteristics like how variable for example is the rhythm period or the things like that and so we'll get at all of those things more on Friday Okay, and how to deal with that kind of variability when you want to model things Yeah Yeah, so We don't really understand it yet, but the thinking is that you have basically a Kind of a simple set of rules that allow you to function under normal conditions And then if something dramatic happens that I again more on that on Friday, I think of You know, there's a solution space a part of parameter space in which you function properly And if something dramatic happens that kicks you out of there Then the only way that you may be able to get in there is to give up on some of these constraints, but But that's a very abstract kind of answer Or to a right point the point the point is going to be or you already saw You get actually you get back to a new configuration that allows you to function But it's that's usually different from the old one because those perturbations still persist Yeah, that's one way of thinking about it But I have to say and that's what makes it exciting like this This stuff is not really super clear in everybody's head just yet. We were still kind of exploring what could be going on there Okay