Scientist Stories: Henry Colecraft, Engineering Post-Translational Regulation of Ion Channels

Hey, welcome everybody. Welcome to the Weekly Pioneers in Biomedical Research Seminar. Welcome to everybody online as well as the folks in the room. Before I turn it over for the introduction of today's speaker, just a reminder, next week, or next lecture in the series will be Control of Neural Crest Development, co-sponsored by the Center for Neurobiology Research by Dr. Marcos Simoes Costa from Harvard Med. I'll be here at that time. And without further ado, I just want to mention that in addition to the FBRI, the Center for Vascular and Heart Research is co-sponsoring today's seminar. So, in that context, I will turn it over to Rob Gordy to introduce Dr. Colecraft.

Thanks, Rob.

Thanks, Randy.

Good jacket.

That Sherri specifically told me to wear this morning on my way out the door. So, there you have it. So, it is my great pleasure to welcome Professor Henry Colecraft to the Institute. Dr. Colecraft is a professor, a John C. Dalton Professor of Physiology and Cellular Biophysics at Columbia University in New York. And Columbia University is really one of the most storied institutions in the world for cardiac electrophysiology, and a lineage that stretches back over 50 years, I think. So, he's coming from a really strong group. We hadn't met each other previously, but as we got to know each other, we realised that there were certain things in common that we had. Both of us were born in Africa, Henry in Ghana, me in Zambia. Both of us were educated to some degree in London at about the same time. Him at King's College, me at Godless University College. And we both came to the States at about the same time. He started his PhD at the University of Rochester and then went on to do postdoctoral work and then have his first faculty position at Johns Hopkins University before moving to Columbia University about 15 years ago. And really, he has done some very fine work there, including solving one of the biggest mysteries together with his collaborator Stephen Marks on the beta-adrenergic regulation of calcium-mediated calcium release in the heart and the role of phosphorylation in a protein called Rad. So, he won't be going into that work today, I don't think. But I'm sure he's going to have a wonderful story to tell us. And so, I'd like to welcome you to the microphone, Henry.

Thank you. Can you hear me in the back? Okay, okay. Thank you, Rob, for the introduction and also the invitation. This is my first visit to Roanoke. Did I say it well? It's been delightful. I've really enjoyed the interactions and heard some great science so far. I'm looking forward to some more this afternoon. Where's my advance? There it is, right in front of me. When I was young, my mom would send me to the pantry, go fetch me something. I'd go there and say, Can't find it. You'd come right over and point it to me. Okay. It doesn't seem to be advancing, so I can do that. Click the screen. Oh, click the screen. Okay. It's not advancing. Wait until you work Scott Powers. Okay. Am I frozen?

There we go.

There we go. Okay. Thank you so much. Yeah. All right. First, I have to disclose that some of the work I'll be talking about, and it wouldn't be, well, cardiac electrophysiology. Some of the technology has been spun out into this startup, Stablix. As time permits, I'll try to get through 3 stories today that are all interrelated and that come out of research interests in the lab over the last 6 years or so. The first story starts from a realization that there are certain rare serious devastating diseases exemplified here by cardiac arrhythmias that cause sudden cardiac death, cystic fibrosis, epilepsy, and so forth. And these are all manifestly very different diseases affecting different organs. And at first glance, it doesn't look like they have much in common with them and couldn't have possible common solutions. But what we realized upon further reflection was that significant amounts of these diseases are all underlain by a shared mechanism in that there are mutations in ion channels that cause a decrease in their functional expression. So either less is made or less gets to the cell surface. And so from that perspective, then the simple big idea we had was that if you could figure out a way to reverse this process, i.e., to stabilize those proteins, you could have a technology that could be transformational for many different diseases. So to understand how we came to solve that problem. Let's look at the typical life cycle of the ion channels that we study. They're born in the ER, trafficked to the Golgi where they are matured, get to the cell surface where they are functional. And from there, they are either endocytosed and recycled or taken to the lysosome to be degraded. So, what we realized was that this protein ubiquitin is a master regulator of various steps of this pathway. Both the trafficking aspects as well as the stability. So, I can collapse this cartoon down to this very simplified form and just say that the surface density of the ion channels we're interested in is regulated by 3 processes: the rate of delivery, rate of removal, and the rate of degradation. And ubiquitin simply biases the system towards retention and degradation. So, for us, the light bulb moment came from the realization that many ion channel mutations really behave like normal channels with a lot of ubiquitin on them. And so this led us to ask whether we could simply strip ubiquitin out of mutant channels and cause them to be rescued and in some cases to restore function. Of course, the problem with that is that ubiquitin is called ubiquitin because it's ubiquitous. Every single protein inside the body is utilizing ubiquitin for its homeostasis. And so the true question is, how do you target a protein of choice and strip ubiquitin from it without disturbing the whole global proteostasis. So, we solved that with this molecule that we engineered, which we call an engineered deubiquitinase, or an NDUB for short. So, these are bivalent molecules that have a targeting moiety, a nanobody, and we link it to a stripped-down version of the cell's own ubiquitin urease that's called deubiquitinases. And so with this design principle then, these N-dubs are highly specific, so minimal off-target effects. They're modular, so we can simply swap out the nanobody and attack different targets. And they are very potent, or should be, because they are— it's a catalytic activity. So this is the concept of it. Can we get it to actually work? So Scott Canner is the wonderful graduate MD-PhD student who developed this technique in the lab. And so what we first did was to try to come up with a biochemical assay to determine whether these endops would work as advertised. So in this experiment, he would transfect into HEK293 cells a potassium channel KCNQ1 that's labeled YFP so that we can follow it by fluorescence. And then what we know is that when you overexpress an E3 ligase, NET4L, it ubiquitinates the channel, causes it to be endocytosed, degraded. That's something that's known. And then what we would do in this proof of concept experiment is introduce an N-DUB that recognizes the YFP and hopefully strip off the ubiquitin and reverse that process. So that's what these blots show on the right-hand side. So in the first lane here, we have a channel, a cell transfected with Q1, and you we can in the immunoblot see the monomer, the dimer, the trimer, and the tetrameric species. When we coexpress Net4L, that density goes down because the channel is endocytosed and degraded. We can then strip this blot and probe with ubiquitin. And what you can see is that under normal conditions, you see some ubiquitin, but that's highly enhanced with coexpression of the E3 ligase. And so this arm of the cartoon is sort of working as expected. So now we can introduce an N-DUB and see what happens. And you can see that you really restore the protein expression while rubbing away the ubiquitin. So biochemically, it sort of works. Next thing we're interested in was, well, does the channel really get to the cell surface and is it functional? So we developed an optical assay that allows us to simultaneously look at the channels that are expressed at the cell surface as well as the total protein expression. So, the— we introduced a Bangrotaxin binding site that allows us to measure surface channel using a fluorescent toxin that doesn't cross the membrane. And then we subject the cells to flow cytometry. So, YFP gives you the total protein expression, and then the red fluorescence gives you the cell surface. So, under control conditions, you get a robust expression of the channels of the cell surface in YFP-expressing cells. You add the NET foil and that substantially reduced both the red as well as the YFP signal. Once we add the NDEP, we completely recover both the surface trafficking as well as the total protein expression. In order to assess— so the channels do get back to the cell surface. To assess whether they are functional, we use the patch clamp assay. So these cells will beautifully make the channel and insert it into the membrane., and then we can use the whole cell patch clamp to query current through the— in the whole cell mode. And under control conditions, you record robust outward currents. When you add NET4L, because the channel is cytosited, the currents go away. We add the N-DUB and we can completely recover that. So the deubiquitinated channels are functional. So there were two key control experiments we were keen to to do before we advance to looking at a disease model. And to orient you, this is what I've already shown you before. We are looking at the surface density of the channels, and the gray bar is when under the control conditions. When we add the Net-4L, you endocytose the channel and the curve shifts to the left. You add the N-DUB on top of that and you recover that expression. That's the blue line. So the first control was whether this action actually required the deubiquinase activity of the N-DUB. So we can make a catalytically dead version by mutating the catalytic cysteine to a serine. And when you do that, the blue line no longer shifts, right? So it doesn't work anymore. So you require the enzymatic activity. The second question was whether you required it to be targeted to the protein in order to function. And we can do that simply by leaving out the YFP to which the nanobody normally binds. And when you do that as well, there's no shift. And so it's very selective for the protein that you target the end up to. Okay. So with that proof of concept, then we were ready to assess whether we could apply this to a rare disease. The one we first looked at was Long QT syndrome. Affects 1 in 2,000 live births. Accounts for about 400,000 cases of sudden cardiac death each year in the U.S. And is due— the most common type of Long QT is due to mutations in KCNQ1, a potassium channel. And that channel is normally required for the repolarization phase of the human cardiac action potential. And so, when you have a loss of function in it, you get a prolongation of the action potential that shows up as a prolongation of the QT interval in the, um, in the electrocardiogram. And the, the bad thing about that is that it, um, people who suffer from this have a propensity towards lethal cardiac arrhythmias, um, to side-depoint. All right, so turns out that many of those mutations that cause LQT1 are trafficking mutations. The channels get stuck inside the cell. And so, we asked, well, could we use NDELT to sort of reverse that? Right. So, what we did was introduced a number of patient mutations in the C-terminus of KCNQ1 from patients with the long QT. And we looked at the cell surface staining of those channels using a flow cytometry assay. And to orient you, The wild type is this dotted line. So, what you can see is that many of the mutations under control conditions would be just the naked nanobody have a trafficking defect. There's substantially less than the dotted line. When we co-express and end up with it, we're very gratified to find that in fact we could rescue many of those right back to the surface. And you can actually see that pictorially here. The wild type channel, the KCNQ1, you can stain it really well at the cell surface. In this mutant, that staining is gone, but you can recover it with co-expression of this N-DUB. Do the channels function? All right, so here's an example from one of these mutations, R591H. Here's a wild-type current. As you might imagine, because the channel doesn't get to the surface, the current is essentially very small, almost gone. There's a small molecule agonist, ML277. When you add on top of that, you're actually able to recover some of the current back. Indicating that some of the channels do make it to the cell surface. Here's what happens when you have just the N-DUB expressed with the R591H. You recover about 50% of the current. And this was actually quite interesting because if you looked at the trafficking, the trafficking was essentially recovered back to wild type. So, what this revealed was that in addition to the trafficking phenotype, there was also a gating aspect to the mutation as well. However, now we can essentially titrate the current even beyond the wild-type movement with a small molecule agonist. And so based on how much you give, you can essentially get wild-type behavior with the endo.

Yes. Do you have any evidence about the density in the membrane of these trafficked channels?

Yeah. So, that's really from here. So, there's the surface. This really is measuring the surface density with this flow cytometry approach, right? And so, that allows us to be able to see the channels really getting back to the cell surface.

But I guess, yeah, I understand. Are they distributed homogeneously or do they localize differentially?

Well, in HEK cells, it's— yeah, they go all over the place, right? And I'll tell you next about real heart cells. But yeah. But we were quite excited about this because it seemed like we were able to rescue these trafficking-deficient mutations using this approach. But of course, as Michael is talking about here, these are HEK cells. These are not actual heart cells. And so it was very important for us to see whether this would actually work in real heart cells. And it's very difficult to do this in mice. Right? In fact, you can't, because mice have very short action potentials, and the channel that we are speaking about doesn't really contribute much to the repolarization phase in the mice. And so we took heart cells from guinea pigs, which we can culture for about 2 or 3 days, and we can express in them using an adenovirus either a wild type or a mutant Q1. They already have endogenous Q1, but the one that you add in would combine with the endogenous one as well. Okay. So, in the control conditions with the wild-type Q1 expressed and fused to YFP so we can see where it goes, if you do a RAMP protocol, you get this profile. And this is the important part, the outward current due to IKS, the potassium current we're interested in. And this is the normal action potential duration. If we now introduce in there, instead of the wild type, we introduce a mutant, LQT mutant KCNQ1, see that you get less of the current because it combines with the endogenous Q1 and has a dominant negative effect. And then what you see is this really classical signature of long QT, a prolongation of the action potential duration. And this really mimics what you see in patients, which is an autosomal dominant disease where they have one wild type and one heterozygous allele. Okay, so on top of that, then if we add the N-DUB, very nicely we recovered the outward current and normalized the action potential duration. And so it looked like this approach could be utilized to rescue these trafficking deficient mutations. So one of the things we're really excited about the approach was that it could be potentially utilized for anything, for all different types of ion channels and other proteins as well. So, we tested out the generality of that by looking at a second disease in a different system. And we looked at cystic fibrosis, which is the most prevalent genetic disorder in Caucasians, affecting about 85,000 people worldwide. So, the symptoms of that are that the people that suffer from this have thick mucus buildup in their airways. And that leads to trapping of bacteria and infections and ultimately damage of the lungs and other intestinal organs. Have a low life expectancy and a very poor quality of life. Okay. So, the gold standard of care for a while was antibiotics to treat the infections, as well as airway clearance approaches to get rid of the mucus. And with all those interventions, the life Expectancy peaked around 35 years or so, 35 to 40 years. So, at the molecular level, the mutation is due to a loss of function in a chloride ion channel that sits on the airway epithelial cells at the air-liquid interface. And it normally allows chloride to come out and drags water with it to hydrate the surface epithelial cells. So, when there's a loss of function in it, then you don't have that. And so, you get that buildup of mucus in the airways. So, many of the mutations actually either decrease the stability or the trafficking of the channel. So, very much like the KCNQ1 case. So, a new breakthrough, really significant breakthroughs was made in the last few years by Vertex Pharmaceuticals, use high-throughput screening to identify small molecules that they call correctors that could help channels trapped inside the cell due to traffic to the cell surface. And once they got there, they had a second class of drugs and potentiators that could cause them to open a bit better. And these drugs in combination actually made a real difference in the lives of CFT, cystic fibrosis patients. However, there are still certain limitations there. They have an incomplete efficacy and it's really hard to improve on them because the mechanism of action actually is not known. Just discovered through high-throughput screening. And they only work for some mutations. They don't work for many others. And so we wondered whether we could utilize NBAPs to sort of improve upon this situation. So that involved us, first of all, raising nanobodies against the CFTR channel itself. So we purified the NBD1 portion of the channel. And using a yeast nanobody display library that we have, we could use the NBD1 as bait and isolate nanobodies. And this just shows the evolution of binders after 2 rounds of selection. So once we have the nanobody, we use a FRET assay to confirm that it binds to the full-length channel. Here's just one of those nanobodies that we identified, and this FRET assay indicates that it binds them compared to a negative control. All right. So, with that then, we armed this nanobody with a DUB, made it into an N-DUB, and asked whether we could recover CFTR function. And so, here are two mutations. F508del is the most common mutation, and that's one for which the FDA has approved this triple combination therapy, Trikafta. And N1303K is another mutation for which Trikafta doesn't work and is not approved for use. So pharmacoresistant. So the black traces the current voltage relationship for a wild type person, and the reds are the mutants. So the mutants essentially wipe out the current. Okay. So when you put Trikafta on top of that, you see that in fact you're able to rescue some of the current, short of wild type, but certainly much more than the mutant. And this explains why this drug, the patients feel better when they take this drug. N1303K actually does marginally worse in this assay, probably explaining why it's not as effective in the patients as well. This is what happens when we add the N-dub on top of that, right? We essentially completely recover the current in the F508del, as well as the pharmacoresistance. And this is in combination with the tricapta. So, for the cystic fibrosis, probably the gold standard experiment is to do these— some experiments in these airway epithelial cells grown at the air-liquid interface. So, they form this beautifully stratified layer of cells with ciliated cells at the air-liquid interface, mucus-producing goblet cells in blue. So, for cells isolated from a wild-type person, If you stain for CFTR, you'll see it highly, nicely localized at the apical surface where it's needed for the chloride transport and then subsequent water as well. If you do that for the F508 patient, that signal is almost gone, essentially not there. This is what you get with the Vertex corrector. You get some improvement, but certainly nowhere close to the wild type. But once we combine that with the end-up, we essentially get back to wild type in terms of the ability to traffic this to the cell surface. And so that's the end-up portion of the talk. So I'm going to switch gears here and now talk about a different ion channel, but this would be the channel that I would refer to as my first love when it comes to voltage-gated ion channels. That's a voltage-gated calcium channel. And these channels play a very central and indispensable role in the biology of all excitable cells by virtue of the fact that they are bilingual. They speak the language of voltage signals, but their permeant ion, calcium, is also what signals to biological responses downstream. And so when these channels open in response to action potentials, they allow calcium to come in to drive essential biological responses. And this same signaling paradigm is utilized for synaptic transmission in the brain, enables cardiac-induced calcium release for muscle excitation contraction in the heart, and is responsible for glucose-stimulated insulin release in the— I mean, in the pancreas. So, since the inception of my lab, we've been interested in these voltage-gated calcium channels from different aspects. One of the aspects we've been most interested in is that these are very important drug targets, either actual or potential. So, L-type channel blockers, for example, are regularly utilized for the treatment of hypertension as well as cardiac arrhythmias. And there are small— there are dihydropyridines and benzothiazepines are drugs that target them. For the neuronal-type calcium channels, they are targeted by toxins from snails and funnel spiders and so forth. So, in terms of targeting these channels for therapy, there are certain limitations. First of all, because the channels are prevalent in different tissues, off-target effects, off-tissue effects is a problem and can limit the therapeutic window. And also, a lack of specificity when the channels are very close to each other. For example, the CaV1.2 versus CaV1.3. There are efforts to try to develop CaV1.3 selective blockers, but it's been very difficult because it looks so close to the CaV1.2. And so, for a long time, we've been interested in seeing whether we can develop genetically encoded calcium channel blockers to really complement what's available in terms of these smaller molecule blockers. But this slide I sometimes refer to as my Steve Jobs slide, right? Because Steve Jobs was a master of selling us things we didn't even know we needed. And this is the case here. So the reason why we think genetically encoded blockers would be really advantageous is you can express them precisely where you want to express them so that you don't have the off-tissue effects that is there. And you can even think about stuff at a more microscopic level. In a single neuron, for example, we can have the channel present on the dendrites, the cell body, and the presynaptic terminals. Same type of channel, but very different functional outputs. Would it be possible to selectively inhibit this channel and not that? We think we can do that with a genetically encoded blocker, but not a small molecule blocker. So those were some of the motivations that have led to this work. So at the time I started my lab at Hopkins, a really beautiful paper came out of the lab of Susumu seen of them in nature. What they'd done was to use the calcium channel beta subunit as bait in a yeast-hybrid assay to find out things that bound to it from a particular cell type. And what they fished out was a protein called GEM, which turned out to be a member of the Ras-like G protein family. And there are now 4 members of this family: Rad, Rem, Rem2, and GEM, which was also referred to as Cur. And so RGK just is the acronym for that. So, when they co-expressed the GEM with the channel, they found that the current was wiped out, and we've reproduced that in our own experiment here. Cell expressing robust currents, if you co-express REM and come back 24 hours later, the current is wiped out. And so, in a sense, this is what we're looking for, a genetically encoded calcium channel blocker. The problem is that these would be very dirty to use as drugs because So they do more things than just block the calcium channel, regulate the cytoskeleton, regulate various kinases and so forth. And so we were very interested in really understanding how they work and trying to use those mechanisms to design our own calcium channel blocker without the limitations that these guys had. So we did work out the mechanism. Talented student, Tingting Yang, who is now a professor at Columbia herself, And what we found was that these things were able to inhibit the channels in 3 different ways. So, here's a channel with a beta bound. In the first way, it essentially promoted endocytosis of the channel. So, you removed it from the cell surface. In the second and third ways, even if you prevented that, the channels were inhibited. The channels at the cell surface were inhibited. And I won't go through all that. The important part here is that all three mechanisms relied on the RGK binding the beta subunit. And so it looked to us like the first thing we needed to do to identify, to sort of mimic these mechanisms was to find our own unique beta binders. And we did that by using identifying antibodies that would do that. And Travis Morgenson is the bold graduate student who did this in the lab. So he purified the beta-1 subunit after bacterial expression. We sent it to a farm in Massachusetts. They injected the llama, and a few months later, they sent him back blood, and he went through the process of isolating lymphocytes, making a phage library of all the nanobodies that are present, and using phage display to identify binders to the antigen that he used, right? And so what this ELISA shows is that we did find several binders. And from that, we cloned several nanobody clones. So, a nanobody, for those who are not familiar with this, so our own antibodies have these heavy and light chains that are put together with disulfide bonds. It's large and they would fall apart inside the reducing environment of a cell. But camels and alpacas and sharks and so forth, they have these heavy single-domain antibodies. And this part, this antigen recognition part, which is about 12 to 15 kilodaltons, is the nanobody and is able to recognize its antigen inside the cell. And here's a FRET assay showing that one of these nanobodies that we isolated, NBF3, in fact binds all 4 beta subunits. So, there are 4 different beta subunits that are highly conserved in terms of the overall in terms of the overall structure. All right. So, we have a nanobody binder. But next thing we asked was if you coexpress that with the channel, that wouldn't do anything. So, this— the assay that I introduced you for at the outset, we have a flow cytometry trafficking assay to measure these surface density channels as well as the total protein expression. And we do— we coexpress that with the— either in the presence or absence of the nanobody. And we found that it had absolutely no effect on the trafficking of the channel. So, the channel made it to the cell surface. And if you measure the whole cell currents, there's absolutely no difference between having the nanobody present and no nanobody present with any— the channels were constituted with any of the four beta subunits. So, this is either a good news or a bad news scenario, right? In a good news scenario, we have a case where we have an antibody that can really sit on this complex and do nothing. And now we can address different things to it. The bad news scenario is that if you— that somehow when the beta is in complex with the channel, the nanobody doesn't recognize it. In any case, what we did then was to really the opposite of what we did with the N-dubs, right? With the N-dubs, we were removing ubiquitin from the channel complex. Now we can really add the catalytic domain of an E3 ligase, a hex-domain E3 ligase to the nanobody with the idea that now you can address this construct to the channel complex and ubiquitinate the complex. And if that happens, then you should remove the channel from the cell surface. And that's precisely what we find. So, with the naked nanobody, you get a lot of channels at the cell surface. If you have the HEK domain of NET4L fused to it, the channel is essentially eliminated from the cell surface. And that's ubiquitin-dependent. A catalytically dead version has the cell surface trafficking. And then if you measure the current, the currents are essentially wiped out. So, this is a very potent inhibitor of these calcium channels for all these beta subunits. Okay. So, there are multiple flavors of voltage-gated calcium channels, all of which rely on assembly with a beta subunit in order to form. So, we asked whether this approach works for all of them. And it does. So here are 4 variants of the voltage-gated calcium channel. And essentially they're all wiped out, right? And so Travis named this, um, Cavablader. Um, and here's the cartoon he comes up with. So what he imagined you get when you cross the Terminator with a camel, because the channel's based on the beta subunit. So it's all, it's very effective, but what's it useful for? And one place where we thought it might be useful is in chronic neuropathic pain, which is caused by lesions to the somatosensory system, occurs in things like diabetes, nerve injury, cancer, chemotherapy, so forth. Affects millions of people both here in the U.S. and worldwide. And that's a real critical need because first and second-line medications are often ineffective in reducing some of the pain, and obviously there are side effects that are commonly well-known. All right. So, if you consider the circuitry for pain, it really arises from signals that start in these somatosensory neurons. And that signal needs to be transferred at the dorsal horn of the spinal cord to be taken up to higher brain centers where the signal is interpreted and people feel pain. And so our idea is that the transfer of signal here is absolutely dependent on these voltage-gated calcium channels, right? Because synaptic release relies on that. And so the idea was if we could express the calviblader in just those neurons, we could potentially prevent that signal from getting into the brain in the first place. Okay. So based on that sort of simple idea, Chikun and Travis injected made an AAV9 that expresses mCAVablader with mTD tomato as a reporter. And they injected into GCaMP expressing mice into the paw, the AAV9, and showed that it could actually get back into the mDRGs as shown here, the red signals in the dorsal root ganglia. Then we could We could isolate the cells and put them in culture, and then we could patch cells that were red, have them have Cavablader, or cells that didn't have red didn't have Cavablader. And you can see the difference there. So Cavablader was very effective in blocking the DRG neurons in situ. So with that, then we teamed up with Guang Yang, my colleague, and her postdoc, Linlin Sun, who do animal models of of pain. And they did this paradigm where 28 days before a spared nerve injury model, they inject into the subcutaneous, into the hind paw of the leg, either control AAV9 or calvebladed AAV9. And then after the spared nerve injury model, do some behavioral experiments. Okay. So, Shortly after the SNI, these animals, the control animals, experienced a long-lasting hyperalgesia. That's shown by this decrease in the paw withdrawal threshold to mechanical stimuli that's normally innocuous. And this only happens in the ipsilateral, not the contralateral arm. And what we were gratified to see was that the animals that had Calviblader experienced significantly less pain. And this lasts for weeks. One injection and pain relief for weeks, that's as we've measured. And this happened not only for mechanical, but also for thermal as well as cold stimuli as well. And so, we think that this would be a potential therapy for pain where you could put it in for, you know, have one injection and then have it last for hopefully months to years. In certain cases. All right. So, really, why are they feeling pain? And the other thing that Guang's lab is able to do is that they're able to introduce a window right above the spinal cord so that they can spy on the activity of the DRG neurons. Yeah.

Yes, please. Is there any cases where ubiquitin itself would be limiting for this to be effective? Do you know what I'm saying? Like the substrate for the— with it being like localized to different regions of the cell, like in the neurons in this case, or are there specific pools of ubiquitin that exist in different regions?

I see. So with this, we are really— okay, I think ubiquitin is not limited inside the cell.

Never, never.

Yeah, there's a whole lot of ubiquitin floating around.

So it'll never— there'll never be a case where this wouldn't ineffective because there's no substrate. No, that's right.

Okay.

That's right.

Thank you. Okay. So we're able to spy on the activity of these DRG neurons using calcium as the readout. And pre-nerve injury, you know, the neurons are actually fairly quiet. 3 days afterwards, they begin to be more active. And after 7 days, you know, they're basically coordinated in the activity. And this rationalizes why, you know, the animals feel more pain. With CAVIBLADER, you didn't completely eliminate that, but you severely blunted it. That indicates why they felt less pain. One of the things we were concerned about with CAVIBLADER, obviously, you wouldn't want it expressed. We're talking about Hasta la Vista patient, not Hasta la Vista. The question is, how could we come around that? Be able to push the boundaries of that without really doing things like killing the mouse or the patient. So, we've thought about a couple of ways of doing that. And one of the ways relies on the fact that these beta subunits of the calcium channel actually differentially express in different tissues. So, the heart expresses, say, beta-2, and neurons express, say, beta-1 and beta-3 mostly, and some beta-4 in central neurons as well. So the idea was whether we could expand the approach and really now get selectivity for different beta isoforms. And so from that beta, from the, from the nanobody library that we got from the llama, we were able to isolate one nanobody that actually seemed to be very selective for beta-1 over the other beta isoforms. So the question was whether we could convert this into a beta-1 selective blocker of calcium channels, which you could then use in vivo without, hopefully, many off-target effects. We're actually quite surprised that we could find this, right? Because these channels, like I said, these beta subunits are highly conserved. But it turned out that— all right, so the NBF3 is the one that recognizes all the beta subunits, and that recognized sequences or residues that were absolutely conserved between all 4 betas. E8 actually found a stretch of about 3 or 4 residues which were different between beta-1 and the others, and that's how it earned its specificity. And this was done with Dan Minor at UCSF. So unlike NBF3, E8, when we expressed it together with the channel, had some intrinsic activity of its own. Was able to inhibit the channel, right? And it did it so in two ways. So, it inhibited the amplitude. And then if you look at the waveform of the current, with maintained depolarization, the current declines. So, there's something electrophysiologists refer to as inactivation. And so, this nanobody, you know, is able to both inhibit the amplitude and increase inactivation of the channel. And it does that only when the channel is coexpressed with the beta-1 subunit and not another type greater subunit. All right, so we're curious as to how it was doing this, right? And just to cut it short, it did it in two ways. You could see a decrease in the surface density of the channels, so there was less trafficking. And if you do the single-channel analysis, the channels also open with a lower open probability. Now the question was, could we actually make it even better of an inhibitor? And so we added the HEK domain to it, as we've done with Cavablator. And now you can see that now this is something now we call Chisel1. Now it's as effective as Cavablator in wiping out the current. And it does so only in channels reconstituted with beta-1, not the other beta isoforms. So these are all done in HEK cells, and so we were wondering whether, you know, would actually work in actual cells. And muscle turned out to be a really good system to test out both the efficacy and selectivity of Chisel1. That's because skeletal muscle exclusively expresses beta-1A and cardiac beta-2. And so if Chisel1 is working well in vivo, then you would expect it to inhibit the channel here and have no— and if it's selective, have no effect on the channel over here. And that's precisely what we see. So, it completely eliminates the current in the skeletal muscle, has absolutely no effect on the current in the heart. And so, there's one way that we think we can potentially get improved treatment for pain without off-target effects. The second way, and this is just the last part of the talk here, is something else that came from our understanding of how the RGKs work, and that relates to the second mechanism. So, the second way that we were looking at was to be able to control the inhibition of the channel with a small molecule. All the inhibitors I've spoken to you about so far are constitutive, right? You express them, they inhibit the channel. But could we get a version of which would of it where we could regulate the inhibition in a more acute manner with a small molecule. And it turned out we could. And I'll just go through the evidence for that very quickly here. So, here's one of these RGC proteins, Rem. And if you express Rem with— if you tag it with the CFP and express it by itself in a cell, it's actually membrane-localized because it has this polybasic region in the C-terminus that targets it to the membrane. As I've shown you before, these RGKs are very potent calcium channel blockers. If you eliminate this N-terminus domain, nothing changes— still membrane targeted, still inhibits. But if you eliminate the C-terminus, two things happen: it becomes cytosolic, no longer membrane targeted, and it completely loses the ability to block the current, right? So this could mean one of two things— that the part that we cut off is a calcium channel blocker. That's sort of a direct effect. Or more indirectly, that membrane targeting of the REM is important for the ability to inhibit the current. And it turned out to be the latter. And this is the crucial experiment that indicated that. So, in this truncated REM, we attached a C1 domain from protein kinase C. So, the C1 domain is normally cytosolic, but you can recruit it to the cell surface membrane with a small molecule, a phorbol ester. So, when you fuse it to the truncated rem under control conditions, under basal conditions, it's cytosolic. You add a phorbol ester and you rapidly recruit that to the membrane. And if membrane targeting is what's important, then this predicts we should have made a small molecule membrane-targeted inhibitor. And that's what actually happens. Now when you add the 4-benzoylester, you get inhibition of the current. The second thing that was interesting to us was whether the beta subunit binding to that was important. We made a mutation in the beta that selectively eliminated this interaction, and that's shown here. The wild-type beta pulls down with REM. This mutant beta doesn't. When you do that, you lose that effect. That means there are two things, a dual requirement, membrane targeting, and beta binding. And that led us to think, well, how could that be happening, right? So, we have this hypothesis. Here's the channel. Here's the intracellular loop that the beta in black binds to. This is the RGK. Under this condition, we said the channel can open and current passes through. When we add the phorbol ester, this hooks up to the membrane, and all we imagined was that it sort of pulled on the channel and caused a conformational change that inhibited the channel. So, the prediction was that if that's the way it's working, you don't really need RAMP to do that. We should be able to really use the beta subunit itself as a handle if we attach to it the membrane targeting domain, and that's something that we could test. So, we put the C1 domain right on the beta subunit so that we can— the beta normally is cytosolic. You can now target it to the membrane using the four-bolster. And here's what we find when you have a channel reconstituted with this beta now. Now this channel is really sensitive to, to phorbol ester, and so it's really acting in the way that we thought, and you can get a really deep inhibition and fast-acting inhibition of the channel. It turned out that this didn't just work with the beta. You could use it for— you could convert other proteins that bind with other loops of the channel into inhibitors simply by causing them to be associated with the membrane, and so we refer to this approach as as CHIMP, Channel Inactivation by Membrane Tethering Associated Protein. And so now with the nanobodies, then we could ask whether we could now convert them to an inducible inhibitor with them, with this CHIMP approach. So there's the NBF3 that I spoke to you about. We armed it with the CFP-C1, with the C1 domain from protein kinase C. And indeed, now we have a molecule where we can actually induce the inhibition simply by regulating the association of the nanobody with the, with the membrane. Okay, so that's the end here. I'll summarize by saying I've spoken to you about 4 things. Spoken to you about NDEBS. It's an approach that allows you to rescue and make more ion channels or any other protein using a targeted deubricination approach. And I've shown you how you can utilize that to rescue trafficking-deficient mutants of ion channels. Caviblator is a genetically encoded calcium channel inhibitor. That inhibits all high voltage activated calcium channels. And showing you, we think it has some potential utility for things like pain. And so, we are pursuing that. The CHISL gives you more isoform selectivity. And so, hopefully overcome some of the potential side effects of Calviblader. And then CHIMP allows small molecule control of that inhibition. Okay, so I think I'd like to thank the— I spoke about the people who did this work as I went along, but certainly Travis did develop the chimp and the— what did I speak about? Yes, he developed the COVID blader and the chisels. Scott developed the endo, helped by Zunaira, and then also Ananya did some work in that regard. Have a number of wonderful collaborators. We got the yeast nanobody display library from Andy Cruz. Manu Banjoni is a great colleague, and Guangyang and Linlin for work with the pain. And I'd like to thank our funding sources, and thank you for listening. Take questions now. Roger Bannister, who ran the formative model? I don't— looking at him, I don't think he understands. I hope Roger is not on the phone.

So patients with cystic fibrosis are treated with for wrinkles and potenciators through their whole life, their whole life basically. So how do you envision to implement your nano-based, you know, ubiquitinase in patients? I mean, they, you know, in vitro everything looks very good, right? And very promising. You have the current channel being depolarized and it goes back. But the reality, you need to have the the chloride channel right there and working all the time. So how do you envision that application?

Yeah, it's a really good question. Thank you for it. So yeah, the part— what I've spoken to you about right now, if you were going to implement it, would be a gene therapy, right? Yeah. You'd have to be AAV9. So when Scott and I were doing this work, we also realized that for certain applications, it would be really great if you had a small molecule that can essentially do what we are doing here. And so that's what actually led to the startup. So what the startup is doing is very much like the— how PROTACs are huge in the pharmaceutical industry. So PROTACs are these bivalent molecules which one part binds to a target and then the other part recruits an endogenous E3 ligase to it, and that degrades the target protein. So we figured that we could do that in reverse, right? So instead of recruiting E3 ligase, we could recruit an endogenous deubiquitinase. And so that's what the company is doing. And that's the approach that's going to be utilized for diseases where you need a systemic expression of something. That's right.

You're going to be thinking about the cavablader and the pain experiment. So I'm going to make a parallel. In the auditory system, there's a condition called tinnitus, where you're kind of a ringing in your ears, right? What often happens is the peripheral hair cells will be damaged and degenerate. And what you've done is kind of vacated activity on the synaptic space that's normally going into these higher neurons in the cortex. And they start to generate this intrinsic activity that you perceive as sound. So I'm thinking in the nociception model with those selective calcium channel blockers on the nociceptive inputs, the downstream synaptic targets that project centrally are now not getting normal input. And whether— kind of like Carla's question— for the long-term treatment, whether there'll be a perception of pain that will start centrally even though you're not getting the peripheral activation of the nociceptors.

I don't know if you've studied animals long enough Yeah, it's a really interesting question. Um, I think the thing I would rest my laurels on there is that there are people who suffer from mutations in sodium channels, Nav1.7 loss-of-function mutations, who absolutely feel no pain, right? And those people, those neurons won't fire action potentials. So in a sense, it's precisely that what you're talking about, right? That, but beyond just going on and breaking their bones and things like that, they don't seem to have other neurological deficits, right? That would suggest that if you don't, if they don't fire, things downstream seem, still seem to be fine, at least from a gross level in terms of— That's a good point.

But a big difference is they didn't have the activity from birth. This is the genetic mutation, right? So the central neurons were not used to that activity, that they could be a difference.

But that's true. That's right. But then, and then, and then these, the second part of it is that, that's, um, Tavablader. Like I said, we're always concerned about, about its use because it's so potent and so non-selective and stuff. And so that's part of the reasons why we're developing these, um, other approaches, ChIMP and so forth, so that you can titrate the amount of You don't always want it off. You can, you know, turn it on and off as you would. I'm sorry.

All right. In your AAV9 catabolator work, did you look at function of other tissues that are targeted by AAV9, like the heart or the liver? Did you look at, like, expression if, like, if these were taken up by those tissues and then ultimately functional output, like echo or, like, ALT and AST?

That's right. Yeah. So, yeah, we did. We did do some of that work. The way that we delivered the AAV9 was also very localized, right? Just a subcutaneous injection into the hind paw, trying to avoid it getting into the systemic circulation as much. So we looked at several places. We found maybe a couple of cells lighting up in the liver, but yeah, mostly not much. We suspect that if we really jack up the amount that we do, we are going to see it off-target. Get both sides, right? And you can sort of deal with that with the promoter that you use so that even if it gets somewhere else, you don't at least express the protein that you want in other tissue you don't want it expressed.

We're going to go Mark, Sora, and then Anthony. So go, Mark. Yeah, exactly. So it was really, really interesting and nice talk. I have a question. So the cell has its own intrinsic mechanism to inhibit calcium channels, calcium currents through this That's independent inactivation, right? I was wondering if this is a bit of a generic question, but I was wondering if this last technology you talked about, this could be implemented to, to probably shape that, for example.

Yeah, so it's a really great question, right? I think so the genetically encoded inhibitor aspect of what we do, I think there are so many different applications of it, right? And Because they're, like you said, they're endogenous natural systems that regulate these channels. And they're actually very important, right? So, the channels, for example, are inhibited by— the neuronal channels are inhibited by G beta gamma subunits. And that's one of the mechanisms for how you feel less pain. The calcium-dependent activation that you speak about is a feedback mechanism where calcium binds to calmodulin that's on the channel. And that sort of inhibits the channel, and that's very important for maintaining the cardiac action potential duration and, and so forth. And so you can think about these— this program that we're doing is really trying to mimic what nature is doing and actually expanding on that. One of the things we've been thinking about is, can we confer calcium-dependent regulation to channels and so forth like that? And so, yeah, there are many interesting avenues to go with that. Great question.

All right, Laura.

When you rescue pain with CAMP emulator, I'm wondering what part of the mechanism of calcium channel has been changed by CAMP emulator? Because when I look at some in vivo imaging data, I can see the amplitude is usually diminished, but the frequency is still there. So what are your thoughts?

Can you repeat that? I didn't quite get it.

So I'm wondering what part of the mechanism of some CAMP the pressure channel has been changed and affected by capillar. So when I look at the in vivo imaging data, there are two parts of some amplitude and the frequency. So what I can see, the change in the amplitude but not the frequency.

Ah, so you're talking about the genetically encoded calcium sensor data. All right. So, all right. So with the capillar, there is a decrease in the frequency of it. As well as you get less of the phasic activity. So when Guang did this experiment, so she had a paper where she published in Nature Communications where they did this imaging after the spared nerve injury. And actually the first time she showed me the data was quite remarkable because what you see, and this would be interesting to the gap junction sort of people here, because it seemed like all the channels were just all the cells would just light up together in terms of their activity. And I was asking her, you know, why are they all of a sudden behaving like that? And she said, oh, I don't know, right? Could be a fracture. Or a fracture, yes. I forgot where I was. But whatever. But the coverblader, it doesn't eliminate that signal, right? But it certainly blunts it down. And the other thing you need to remember is that for these, for calcium signaling, these high voltage activated calcium channels are not the only source of calcium. There are low voltage activated calcium channels as well, the T-type. Capablader isn't going to touch that. And so, you know, so there you will see some calcium signals, but what we see is an overall blunting of the, of of the response.

So particularly for the CFTR, how sensitive is the deubiquitinase mechanism to mutations in the channel? Because there are a lot of mutations that are considered disease-causing. And I'm wondering whether or not that will limit the efficacy to a subset that don't have mutations that might inactivate.

That's a great question. Is this going to work for all mutations or not? And the answer is no, right? Because there are some mutations for which the channel is hopelessly misfolded. And actually, for the KCNQ1 work, if you saw, there were certain mutations for which we actually introduced that because we knew that the channel was misfolded and seeing what's going to happen.

That's what the— from the KCNQ1.

Exactly. And so those channels, do not. So you can make more of them if you look at the YFP signal, it's a hub. But there appears to be another gate which tells them, you know, if the channel is hopefully completely misfolded, not to advance, which, you know, is a good thing. So it works for mutations where there is some residual activity and things that if you get—

Is it possible to devise an algorithm to predict where this would be efficacious and where it would not be based on—

That would be very useful. Yeah, that's true. Yeah. Yeah. Thanks.

All right.

Well, I just needed to say, thank you, Henry, again.