The following is a transcript of the presentation video, edited for clarity.
LAUREN: Good morning everybody. Hi. My name is Lauren Calandruccio. I’m an Associate Professor at Case Western Reserve University in Cleveland Ohio, and the chair for the 2019 Hearing Research Symposium. I am honored to introduce our second invited guest for this symposium, Dr. Jeffrey Holt. Dr. Holt earned his PhD from the Department of Physiology at the University of Rochester. He completed his postdoctoral training in the Neurobiology Department at Harvard Medical School. Dr. Holt began his academic career at the University of Virginia, but in 2011 moved his lab to Boston Children’s Hospital, Harvard Medical School, where he is now a Professor of Otolaryngology and Neurology at Harvard Medical School and Boston Children’s Hospital. He is also the Director of Research for the Department of Otolaryngology at Boston Children’s Hospital, and on the Health Sciences and Technology Faculty at Harvard/MIT. Dr. Holt is an ad hoc reviewer for 17 different prestigious journals in our field. He is a section editor for Hearing Research. He serves as a grant reviewer for NIH, NSF, and several other review panels in the U.S. and across the world. He has mentored 8 PhD students, and over a dozen postdoctoral fellows. Dr. Holt’s NIH funded research program aims to understand how sensory cells and neurons in the normal inner ear function. His lab works in identifying the molecular components required for sensory function, and identifying why mutations in those molecules lead to dysfunction. The long term goal of this work is to use this information to design and translate therapeutic strategies aimed at restoring sensory transduction and inner ear function. His work in this area led to receiving the 2019 Balochi Prize for Hearing Research. Dr. Holt has also received awards for teaching and mentoring from Harvard Medical School. You will not question that recognition once you hear him speak. He is a master lecturer. I was inspired about the future of hearing science and science education the first time I heard him give a presentation. It is a true privilege to have Dr. Holt with us here this morning. Please welcome, join me in welcoming Dr. Holt. (Applause)
JEFFREY HOLT: Thanks Lauren for that nice introduction. I’m happy to be here with you all today. One thing that didn’t make it on that list is last week I was in Paris, and I just wanted to point this one out. Work from our lab was recognized with the scientific grand prize from the Foundation Pour L’audition. Not to pat myself on the back for this one, but I wanted to call your attention to this foundation. It’s a French organization, but they’re, they’ve been based in France, and they’re trying to promote auditory research and assistance to folks who have auditory dysfunction in France, but now they wanna extend globally and have more of a worldwide reach. So if you’re interested to learn about the foundation, you can see their website right there.
So I’d like to begin more from an evolutionary perspective. I probably don’t have to emphasize the significance of auditory function for this crowd. But thinking about hearing and how hearing evolved can give us some insight. For example, the auditory system is clearly important for evolution if you imagine in the dark of night having auditory sensitivity may allow you to detect the sound of an approaching (a roaring lion drowns out the words). Or likewise the sound of a prey species (an animal noise that sounds like a cow). Thus, would clearly be an advantage for survival. In addition, many species have coevolved vocalizations that are useful for communication and identification of individuals with, of the same species which could be useful during mating (a whistle is heard). These sorts of vocalizations can help identify a mate, attract one, or perhaps avoid one. In addition, this kind of frequency selectivity can be useful for uniquely identifying one individual, such as a mother’s ability to identify the unique sound of their offspring (a baby’s sounds are heard). So, as humans we’ve taken advantage of this precise ability to encode both frequency and sounds over time, which has allowed the evolution of spoken language and the ability to detect over 6,000 spoken languages here on earth by humans. In addition, this has allowed for us the ability to detect and appreciate music (the beginning of Beethoven’s 5th symphony is heard).
So I’d like to tell you about an evolutionarily conserved molecule that we discovered in my lab that is the key to auditory sensitivity and allows the conversion of sound information into electrical signals. It’s known as Transmembrane Channel-LIKE One, and it’s right here at the tips of these sensory hair cells. The hair cells themselves as you know, are the key to convert mechanical stimuli of sound and head movement into electrical signals. Within the human auditory organ of course there are about 16,000 of these in the cochlea, and about 30,000 sensory hair cells throughout the vestibular organs. As a model, we use the mouse inner ear. The mouse is an excellent model system because the genetics are quite similar to that of the human. It’s also a good anatomical model; the structures are quite similar to those of the human ear, and we can then access them, these tissues, remove them, and then image them. This is a scanning electron micrograph of a mouse cochlea, where you can see a single row of inner hair cells and 3 rows of outer hair cells. Cell bodies here are shown in blue. And if we zoom in on a single hair bundle, you can see the staircase array of this outer hair cell bundle.
Experimentally, if we excise these, we can place them in a recording chamber, and then I can stimulate this hair bundle. The back end of this stimulus pipette is placed on a piezoelectric bimorph, and we’re able to then impose deflections of the hair bundle. The cell body is mostly out of the image, but if we place a recording pipette just here, we can wiggle the bundle back and forth, I plot the stimulus position, and then record the response. These are electrical currents flowing into the cell when the hair bundle swings in the positive direction toward the tallest stereocilia. When it swings back in the negative direction, the currents drop back to about zero.
So the model for how this occurs suggests there’s a, a series of mechanically gated ion channels here at the tips of the stereocilia. And when the hair bundle swings back and forth, that mechanical stimulus pulls these channels open and closed. I show a single pair here on the right. So when the, this sensory transduction apparatus, is excited with a bundle deflection to the right, these ion channels will open, allowing cations to flow into the cell. Primarily calcium and potassium.
So we’ve been interested in identifying some of the components of this sensory transduction apparatus. This extracellular filament here known as the tip link is composed of cadherin molecules, Cadherin 23 and Proto Cadherin 15. There’s an adaptation motor which serves to provide feedback to this structure. This is composed of myosin molecules, probably Myosin 7a. The ion channel itself however, has been a major question in the field for the past 40 years.
This work originally began 40 years ago with David Corey and Jim Hudspeth who published a series of papers that first characterized hair cell transduction. The papers were quite elegant, and proposed the following; that there, when they recorded from these hair cells, that there was a extremely rapid response which you can see right here. The time scale is in milliseconds, and you can see the rise of the response is sub millisecond, in the microsecond range. This is far too fast of a response to be mediated by a, a second messenger cascade. Instead, it suggested a direct mechanical gating. And so they proposed a rather novel idea that this is an, a mechanically gated ion channel.
Around the same time, another group led by Jim Pickles, found that these tip link structures, as you can see right here, and put together this model in which a deflection of the hair bundle to the right would stretch or tense these tip links as you can see here, which might directly pull on this mechanosensitive ion channel. Deflections in the other direction would tend to slacken the tip link, and release any tension.
Also around the same time, a study from Karen Steel and Greg Bock characterized one of the first mouse models of deafness. This was a genetic model. These mice were known to be deaf, and she characterized them electrophysiologically and found that they had no cochlear microphonic potentials. And went on to suggest that the most likely explanation for the absence of the cochlear microphonic was that the ion channels were not opening properly. So this was incredible insight based on just a little bit of data. But unfortunately, neither Steel and Bock, nor Corey and Hudspeth, put these two ideas together. They were both looking at the same phenomena, but from different perspectives. The Steel group looking at the genetics of this mouse, and the Corey and Hudspeth group looking at the physiology. Had they made the connection, it would have avoided a 40 year search for the hair cell transduction channel. Instead the Steel spent the next 40 years and came up with a pretty large list of potential hair cell transduction candidates. And this is just a partial list. I cite the, the ion channel gene, the original publication that suggested it may be involved. Some of these published in very high-profile journals such as Science, Cell, Nature, etcetera. But systematically it seems like one after another, these were dismissed from further consideration, either because the gene and protein was not present in the mammalian inner ear, wasn’t expressed in the sensory hair cells, knockouts or deletion of these genes did not cause any auditory dysfunction. So, it raised a lot of skepticism. There were so many false positives basically, that, that the field became somewhat skeptical of any new information or any new ideas about hair cell transduction.
But we were interested in this question, and decided to go back to this list, and in particular looked at the TMC1 gene. And the reason we were interested in that, well it was first published, or described in 2002, and that was a pair of papers that came out in Nature Genetics from Andrew Griffith and Tom Friedman. They found that the deafness mouse that Karen Steel originally described, encoded this gene known as TMC1. This was present in mice and in humans, and mutations in mice and humans led to deafness, suggesting that this must be important. About the same time, Karen Abraham and Karen Steel studied a second mouse, and they gave it the name Beethoven, because it had an acquired hearing loss. This was a dominant progressive hearing loss in the same gene, TMC1. So this suggested that the TMC1 gene was important for hearing, but didn’t necessarily present or identify the function of this gene.
Another group led by Walter Marcotti and Corné Kros, suggested that the TMC1 gene was also important for hearing, but they looked at hair cell function and found normal mechanotransduction currents. So when they wiggled hair bundles, the responses seemed normal. They concluded it was not necessary for hair cell transduction. But, unfortunately, what this group missed was that there was a second gene. Similar gene simply known as TMC2, which was expressed and seems to have a redundant function.
So we were interested in this. And together with Andrew Griffith’s group, we studied the TMC genes, TMCs 1 and 2. The first insight we gained was from some behavior of mice that lacked both TMCs 1 and 2. We knew that the TMC1 gene was important, but when we deleted both TMCs 1 and 2, the mice also had vestibular dysfunction. And that was evident as you can see here. This is at 4 weeks of age. This is a mouse with vestibular (plays beginning of Rock Around the Clock). They had this head bobbing behavior, head arching, they tend to have a wide ataxic gate, they have circling behavior (several words are covered by the music). So, this was an indication that TMCs 1 and 2 were involved in both auditory and vestibular function, and really got us excited that, that something might be going on of interest. When we looked in the ears of these mice, we found via a blue stain right here, that it was expressed just in the hair cells, in the auditory organ, and within the vestibular organs of the saccule and the utricle, and in the semicircular canals. So, it was present in the right cell type.
The next question is, was, what’s the function here? And so Gwynne Géléoc, who works with me, looked at this, and she studied FM143, which is a dye that will go through sensory transduction channels, and it labels functional cells with this green stain right here. In mice that lacked TMC2, we saw green stain update suggesting that these cells were functional. Mice that lacked TMC1 also had uptake of the green dye. But in the animals that lacked both TMCs 1 and 2, there was no dye uptake whatsoever, suggesting that these cells indeed were not functional. Something was wrong with their mechanotransduction.
We then went on to record from these cells directly, and we looked at families of mechanotransduction currents in a wild type mouse, a mouse that lacked TMC2, where TMC1 had fairly normal currents, but in those that lacked TMCs 1 and 2, there was no response whatsoever. This was true throughout the cochlea both from the base to the apex, and in the vestibular organs. And also true throughout development.
Looking at gene expression, we find that the TMC2 gene was turned on early in development during the first postnatal week, and right after the gene expression began to turn on, was the onset of mechanosensitivity. Interestingly, the TMC2 gene peaked during this timeframe, and then began to decline at the end of the first postnatal week. As it was declining, we saw a rise of TMC1 expression that just preceded the onset of hearing in mice, which is at postnatal day 12.
I think it was this developmental shift where TMC2 would rise and then fall, followed by the rise of TMC1, that developmental shift is what the Marcotti and Kros group missed. They were looking at early postnatal stages when TMC2 was expressed and saw a normal function. Had they recorded it a few days later, in the second postnatal week, they may of seen that mutations in TMC1 led to a decline of transduction.
So those data really suggested that TMCs are involved in this process, but it wasn’t entirely clear exactly what they were doing. They could of been involved during development, a transport molecule, maybe a linker of some sort, some accessory structure, or the ion channel itself. And so to look at this even further, we published another paper in 2013 where we showed that there were some distinct biophysical properties that were different between hair cells that expressed just TMC2, just TMC1, or that Beethoven mutation that I mentioned. This is a single point mutation in the ion channel, in the protein sequence that leads to a difference in the biophysical properties. These are current voltage relationships which cross the X axis at different positions, and that’s a measure of the reversal potential, which we can then use to estimate the calcium permeability of these cells. What we found was that the TMC2 expressing cells had very high calcium permeability. Those that expressed just TMC1 had lower calcium permeability, and those that carried the Beethoven Point Mutation had reduced calcium permeability still. Because permeability to ions is a key property of an ion channel, this was very strong evidence suggesting that the TMCs are closely associated with the transduction channel.
Another line of evidence was to look at single channel events. So for this we could record from hair cells, and then we designed stimulus probes that were very small at their tips, about a hundred microns, where we were able to deflect just one stereocillium at a time and then record the response. And so these were the data that we saw. These individual events right here, this is one ion channel opening, allowing just a few picoamps of current to flow into the cell. So we were able to do that in a cell that expressed TMC2, and measure currents of about 20 picoamps. Cells that expressed just TMC1, had smaller responses of about 12 picoamps. So again, this is a key property of an ion channel, closely implicating these in mechanosensory function in hair cells.
In another study, Kiyoto Kurima in our lab, worked together to show that TMCs are localized to the tips of these hair cells. So he, Kiyoto made a TMC1 fusion to M cherry, and you can see little red puncta right here at the tips of the hair cells stereocilia, suggesting the proteins are localized to the right spot. Likewise, TMC2 proteins also seemed to be localized at the right spot. These green puncta that you see right here.
So together the evidence as growing that TMCs were involved. And that began to be reflective in the literature. Cartoons, transduction models, had matured some from the original, to show several other elements that were involved in transduction, some of those shown at the upper end of this tip link structure, some at the lower end. The transduction channel itself though, still had a question mark. You can see TMCs were listed there, as well as some other proteins. Here’s another version of the model published by a colleague Bechara Kachar, shows that there are maybe several proteins involved. TMCs are listed here, so the field was getting a sense that yes indeed TMCs might be involved, but exactly what they were doing was not clear. And one other version, again with a large question mark. So, these being published in textbooks, indicated that there was some acceptance of the idea, but still the function was not clear. And part of the reason for the confusion was just the name of the protein itself. TMC stands for Transmembrane Channel-like. Is it a channel or not? Right? The typology didn’t make things any more clear. To have an ion channel, one of the key things you need is a pore, right? A pore region itself. And if we look at the typology, here’s one published from Stefan Heller’s group that had 8 transmembrane domains, a little reentrant membrane spanning region here, but no clear porevdomain. Likewise, another typology published by Andrew Griffith’s group showed 6 transmembrane domains, no real pore region, this large intracellular loop here. One by Fettiplace had 6 transmembrane domains, 2 unnamed domains, I don’t know why, and one from Uli Muller’s group showing TMCs in a complex with other proteins. But all of these lacked a pore region. So, we wanted to settle this debate once and for all. And in particular we were interested to investigate this hypothesis, does TMC1 form the hair cell transduction channel or not. And if so, what are the structure function relationships? In particular where is the pore region, if it’s forming an ion channel? What are the transmembrane domains that are lining the ion channel pore, and then furthermore, what are the amino acids that are contributing to the permeability properties? What makes it calcium selective for example?
And so last year we finally feel like we’ve gained a definitive evidence showing that TMCs are indeed the hair cell transduction channel. And I’d like to spend the next few minutes just reviewing some of the highlights of this publication. This is work from 2 postdocs in my lab, Bifeng Pan and Xiao-Ping Liu. Bifeng did some heroic recording from hair cells, as well as collaboratively with a postdoc from David Corey’s lab, Nurunisa Akyuz.
So Nurunisa began with some biochemical studies where she was able to purify the TMC1 protein and then run it here. This is western blot gel, the size of a single monomer of the protein was about 88 kilodaltons. We found that it also co-assembled as a dimer, and but we never saw trimers or tetramers. So the fact that it was a dimer was interesting to us. And we were able to take the purified protein and look at cryo EM images of this. So these are low resolution images of single molecules. But it allows us to get a little bit of an insight into what this TMC molecule might look like. And you can sort of see the dimeric structure. I, I with the yellow line indicated the axis of dimeric symmetry. On either side is one subunit of TMC. And you can maybe see something that looks a little bit like a pore, maybe 2 pore regions on each side.
So the fact that this is looking like a dimer led us to think well, what other ion channels might this be comparable to. Most ion channels are formed as tetramers, some are trimers, some are pentamers containing 5 subunits, and very few are, are dimeric. So, we looked at other dimeric proteins and found one in particular known as TMEM16A, which is a calcium activated chloride channel. And we used a bioinformatics approach, and 2 software algorithms that would take the amino acid sequence for the TMC1 protein and compare it to a known structure. Both of these predicted that the TMC1 amino acid sequence was most similar to TMEM16. We then took that prediction and ran it through a supercomputer, this is at Ohio State, and work done in collaboration with Marco Sotomayor, which took that predicted structure, allowed it to, to rest at its lowest energy state, and this is what the computer spit out. Here’s a predicted structure for TMC1. Each of these spiral helices is a transmembrane domain. So, there are 10 transmembrane domains in one subunit shown in blue, and in a second subunit shown in gray. And, so, we wondered, is this just some fantasy dreamed up by the computer, or is there any basis in reality here. And so to test that, we designed an experiment in which we introduced mutations, 17 mutations into TMC1 where we substituted a cysteine. We could put those in particular into this region right here. These are transmembrane domains; here’s all 10 of them. But domains 4, 5, 6, and 7 we thought were of interest, because by homology with the TMEM16A protein, which those 4 domains form the, the ion channel pore for that one, so we targeted domains 4, 5, 6 and 7 in TMC. Several of these mutations were introduced because these are known sites that cause deafness in humans, so we thought these must be significant. Likewise, we figured if those are important, maybe neighboring sites in domains 4 and 7 might be important. We also looked at negatively charged amino acids. These are aspartates which carry a negative charge, and we reasoned that well perhaps negative charges could be important here to stabilize positive ions, the cations like calcium and potassium when they permeate. Based on the homology model, domains 5 and 6, we selected these residues as well. And then one negative control in a region we thought might not be involved. We took those sequences and introduced them into adeno associated viral vectors, and then we were able to inject those into the inner ears of mice that lacked native TMCs. So the idea was to introduce a TMC with a mutation that we wanted to study. We could then excise the cochleas from these animals and put them in culture, and then record sensory transduction. Bifeng Pan did some heroic work recording from over 566 hair cells.
Then we could use a drug that would modify the properties of these cysteines. And, so I’ll come back to that in a minute. So we used this FM143 assay again, and I, as I mentioned, double knockouts lack any dye uptake. But we could use this then to see which cells had been infected with the virus, then expressing the TMCs, and you could see quite clearly those that were now expressing functional transduction channels. So we could also target these for our recording experiments.
And here’s a summary of the data. So, this is all 566 hair cells. What we’re plotting is the amplitude of the current on this axis. These are the individual mutations, recorded one at a time. Most all of these gave us current amplitudes that range from about 500, uh, 200 to about 500 picoamps. One or two of ‘em didn’t work too well, but all the rest of ‘em seemed to, to provide robust currents.
And so the experimental situation then was to record from the cells, and apply this cysteine modification drug which was known as MTSET. And the chemistry here is such that a cysteine within an amino acid chain has a sulfur side group that the thinking is, if that’s projecting into an aqueous solution, it could be available to react with this drug, MTSET, which also carries a sulfur side group. These two sulfurs will bind, forming a disulfide bond on a covalent manner, and then link this bulky side group, and if this bulky side group is within a aqueous region of permeation pathway, it could block the flux of ions through a pore.
So that was the, the strategy, and here’s what it looked like. We would give a repeating series of hair bundle deflections, and then puff on the drug right here and then wash it off. And so when we recorded from cells that carried the D569 mutation, boom, as soon as we puffed on the drug, we saw this decline in the amplitude of the current. At the end, when we washed out the drug, we didn’t really see a recovery. So that was consistent with the covalent modification. An irreversible change.
Here’s a second example. It was a very robust and repeatable response, and we got this pretty much every time. But just to convince ourselves that this was specific, we did a couple control experiments. This is in the absence of the drug. We still gave a puff just of saline, but saw no change in the amplitude. And here is a wild type TMC with the drug, but we saw no real change here either. So it was the specific combination of the drug together with that mutation that caused the effect.
So we screened all 17 of those mutations that were introduced one at a time, and here’s a summary of the data. Some of them had a very fast, more complete response, some were a bit slower. Here’s one of the fastest ones that we saw. Some had no effect whatsoever. Those that had no effect were probably not lining the ion channel pore, or inaccessible to the drug, but those that did have an effect, we reasoned must be this reduction in the current amplitude, must be due to their, their location within an ion channel pore.
So to analyze those data, we took the first 5 peaks, the average of those relative to the last peaks to get a current ratio. And so again, some had no change. This is a, a ratio of 1. But five of these sites we did see a change. D, or you can see them labeled right here.
So the next thing we looked at was calcium permeability. And this is a lot of data to look at. We’ve done this both with the native mutation, the cysteine substitution, but that is well after application of the drug. And so to just summarize this data, we found that the mutation itself rendered, or, or reduced calcium selectivity at 7 of these sites. The application of the drug MTSET also reduced calcium selectivity at 8 sites, and in total there were 11 sites that were affected by this manipulation. Either the mutation itself or application of the drug. So again, selectivity is a key property of an ion channel. And the fact we can manipulate that in real time with application of the drug suggested that these sites must be forming the, the permeation pathway. However, some objected and wondered, some of the skeptics in the field, and as I illustrated earlier, there, there were some, wondered well maybe these sites are on the outside, some accessory protein, and that if you do this manipulation you’re somehow changing the confirmation of the channel that gets transmitted to the pore, and affects permeability. So to address that, we wondered, well, if they’re here in the pore, we might be able to block access of this drug to those sites. And so we wondered about using known pore blockers. So there’s a couple a drugs, the aminoglycoside, dihydrostreptomycin, and amiloride, both of which will actually sit right in the ion channel pore. This has been demonstrated previously that these will block currents flowing through the transduction channel by sitting right here. So, if so, we thought well, what if we apply this first, and then the drug, maybe it would block access to these sites, which wouldn’t happen if the sites were on the outside, right? So here’s the data. We wash on dihydrostreptomycin, and then apply MTSET, and the currents are blocked because the dihydrostreptomycin. But when we wash them both out, we see the currents recover to their initial levels. With the MTSET drug alone, remember the currents declined, and we could not recover those. So I think this is evidence that the dihydrostreptomycin does indeed protect these sites. And that was true for several of the sites that we examined. The data are summarized here. So, this is the MTSET drug alone right there, and this is if we applied them together we could get this level of protection as you can see the difference between those bars. So it suggests that yes indeed by applying the drug first, we can protect those sites.
What about simply channel closure? If we push the hair bundle in the negative direction, channels should close. Maybe that would also block access of this drug to these sites. And so here we did that. Let me draw your attention to this last pair. What we do is we deflect the bundle, then we hold the bundle in the open position, apply the drug, and we see a decline in the amplitude, so that’s consistent with what I showed you previously. But if we do the opposite experiment, where we, we measure the current, then hold the bundle in the closed position while we apply the drug, wash it out, but then deflect the bundle again, we see normal amplitude currents. So the difference here is the amount of protection by closing the channels.
That was also true for a second site where we got the same sort of protection, suggesting that these sites M412C and D569, are within the channel core protected when it’s closed. So you can imagine that in this situation the channels close, when it’s open, these drugs can get in, but when it’s closed they cannot.
So key observations here are that we think we’ve got 13 sites that are affected by mutation for the MTS drug itself, 5 affecting current amplitude and 11 affecting selectivity. 3 of the sites that we studied in detail, these were the sites associated with human deafness, the G411, M412, and B569. They had reaction rates that were voltage spent, and I didn’t show you those data, but that was true. They also had reduced single channel currents, and they could be protected by pore blockers or channel closure.
So, all of these data support that TMCs include domains that line the ion channel pore. And so coming back to this, these are the reactive sites in transmembrane domains 4, 5, 6, and 7. And if we now take this data and go back to the structural model that was spit out by the supercomputer, we can look at those right here, and we find that it does indeed support this homology model for the structure. So here are the 4 transmembrane domains, 4, 5, 6, and 7. And I show the sites that were mutated in this various colors here. If I put this in motion, you can see it a bit more clearly. So we think on the outside of the cell, ions are flowing right through the ion channel pore, to the inside.
So together we think the data provides strong support for the hypothesis that TMC1 is indeed a pore forming subunit of the hair cell transduction channel in mice, and in humans. And just yesterday, this is hot off the press, another group published some, some breaking news. This was yesterday came out at noon, a paper showing that if you purify TMCs, either TMC1 or 2, you can introduce the purified protein into something known as a liposome, where you can look at the activity of single channels, and sure enough when they looked at purified TMCs just by themselves, they got these single channel events that you can see right here, very similar to what we recorded from native hair cells. And if you apply a pressure stimulus, a mechanical stimulus to these liposomes, you can get a response that looks very similar to a hair cell transduction current. So it shows that these are forming ion channels, and that they’re mechanically sensitive.
One last bit of data is a simulation now from again Marco Sotomayor, where he shows a single TMC1 protein in a membrane, together with the ions themselves. And if I play this, this little simulation here, you can see permeation events with individual ions flowing through the ion channel pore.
So I think the data now are quite strong, and with your permission, I’m gonna remove the question mark and replace it with TMC1.
So we’ve been interested in this from a basic biology perspective, but also because the TMC1 gene carries over 40 different mutations in humans that lead to deafness. 35 of those are shown right here in black, and they cause recessive hearing loss, and 4 or 5 of them shown here in red, cause dominant progressive hearing loss. And so we’ve been thinking about this from a therapeutic perspective. Is there a chance that we can take what we’ve learned from the basic biology of hair cells and TMC1 function, to perhaps restore function? And we began with an animal model, again going back to our mice, but we thought, well let’s use the same approach for replacing a cochlear implant. Maybe we could introduce the TMC1 correct sequence through the round window membrane, and have it diffuse through the perilymphatic spaces, all the way up to the apical end. And so Tina mentioned this briefly. We have been screening a number of viral vectors. We found one in particular known as Anc80, developed at Mass Eye and Ear, that is useful for transducing large numbers of inner hair cells and outer hair cells, which you can see right here. And we’ve also got another vector we like even better now, it’s known as AAV9-Php.B. This is a variant of an adeno associated vector that transduces 100% of inner hair cells, and 90 to 95% of outer hair cells. Here’s a high magnification view showing green fluorescent protein throughout the inner hair cell region, and the outer hair cells. So we’ve been thinking we could use these synthetic vectors to introduce the TMC1 sequence. So we packaged wild type sequence, this is the correct DNA sequence into the synthetic vectors, and injected them into the ears of TMC mutant mice, which are deaf. Now we’re looking at not green fluorescent protein but myosin7a, and you can see robust hair cell survival throughout the turns of the cochlea. Typically with the myosin7a stain, you can see hair cells in all these regions labeled here in green, but a TMC1 mutant mouse tends to have hair cell loss. They’re significant deaf after a couple months. But in the TMC mutant mice where we introduced the correct sequence, we find we can restore or, or promote hair cell survival in these cases. So we’ve quantified that as the different regions and we see, see robust survival, mostly at the low frequency end, but some at the high frequency end as well.
So are these function was the next question. I showed you previously that mice that lack TMCs have no mechanosensory function, but after introducing the correct sequence, we can restore the mechanosensory responses in both outer hair cells and inner hair cells, and we’ve done this as, as late as postnatal 830. They have properties that are similar to those of wild type, and we’ve quantified the amplitudes right here. So this shows TMC1 gene therapy can promote survival, we can recover mechanosensory function in single cells. What about the level of the whole cochlea? And so we record auditory brainstem responses, just like you would in a human patient. We can place scalp electrodes on the back of a mouse’s head. A wild type mouse has thresholds of around 35 decibels or so. This is at, at 8 kilohertz. A TMC mutant mouse has no responses. The flat lines are indicating these animals are completely deaf. But after introduction of our TMC1 gene therapy into the mutant animals, we find we can recover responses with thresholds that are very similar to those of wild type. So we’re restoring function at the single cell level, at the whole cochlea level, we’ve done that for a number of mice. This was one of our best performers. Some of ‘em don’t do quite as well, but I think it mostly depends on the skill of the injector. We’re injecting just one microliter with a, a injection needle that tapers down to about a micron at its tip. And getting that through the round window membrane of a mouse is pretty tricky, but this is something that may be doable in human patients. We’ve also gotten recovery of distortion product otoacoustic emission, so this is a measure of outer hair cell function, and that seems to recover as well.
Next we wondered, well if we’re recovering function in the periphery, of course we’d wanna confirm that we get recovery of function in auditory cortex to enable perception of sound. And so in collaboration with Dan Polley at Mass Eye and Ear, we recorded from the auditory cortex on the contralateral side, using a multi electrode array. So this is the injected ear. We also play sounds into this ear, and then record activity on the contralateral cortex. In a normal mouse what we’re looking at is a rasterpod. In each spot is an action potential in auditory cortex. And so with louder sounds you’d begin to see them synchronize to the sound stimulus. For a, a TMC mutant animal, we see lots of spontaneous activity, but nothing that’s synchronized to the sound stimulus, suggesting the neurons are viable but they’re not encoding sound information. However, in the injected animals, we find that indeed we can recover these responses, albeit to, to louder sounds. We looked at a range of over 300 different auditory cortex neurons. And so this is a busy slide to look at, but the, the gray curves are tuning responses of auditory cortical neurons from a wild type mouse. Those in green are a TMC mutant mouse that was injected. And what you’ll notice is that the responses tended to cluster around lower frequencies, around 8 kilohertz in a mouse, with thresholds, some as, as low as about 35dB. The shape of these curves, indicating the bandwidth of the tuning of the neurons was very similar between the wild type and the injected animals, which is indicated right here. So they’ve preserved or, or recovered these properties.
So of course the next thing then is what about behavior? And to test the auditory function at least behaviorally in a mouse, what we’ll do is play a sudden loud sound, and like you might be startled, so is the mouse. So we can then measure the amplitude of a jump when the mouse is startled like this. We find that the mutant animals are completely deaf, they don’t jump at all. Wild type animals begin to jump at around 70 to 80 decibels, shown here in black. And our injected animals also recover their startle responses.
So with these data hopefully I’ve shown you that we can recover function at the level of the single cell, at the level of the cochlea, level of auditory cortex, and here behaviorally.
What about the vestibular organs? So we’re injecting one microliter of our gene therapy vector through the round window membrane. But the perilymphatic solutions are continuous with those of the vestibular organs. And so when we look there for expression of green fluorescent protein, we find robust expression in the utricle, the semicircular canals, the saccule, and here’s the, the posterior canal. So it seems to be able to, to target sensory hair cells throughout the inner ear spaces. And we wondered if this could be then used as a therapeutic to recover vestibular function as well. And so what we’ve done in this case is to look at vestibular ocular reflexes. So for a mouse we can rotate the mouse back and forth, and like humans, the mouse will have a compensatory eye movement, where the eyes rotate in the opposite direction. So the eyes are tracked and plotted right here. In the TMC mutant animals that lack vestibular function, this is simply a flat line. They have no VOR responses. But after injection of our synthetic vectors, we find that we can recover their VORs, and they have gains in phases that are similar to wild type. So that’s for a rotational head movement, which is going to assay the semicircular canal function. We can also do this with a linear translation side to side to assay recovery of function in the otolith organs, the utricle and the saccule, and we get recovery of these responses here as well.
And lastly, during this study, we found that there were some improvements in secondary outcomes. Mice tend to have one litter per month for a breeding pair, and that, that’s the standard rate. We found that mice that lacked TMCs, they’ve got vestibular dysfunction, they’ve got auditor dysfunction, and they’re just not very good breeders. Again, this comes back to the evolutionary point I was making at the beginning. The auditory system and the vestibular system are important for these kinds of functions. But they have reduced rates of producing litters. Maybe only once every 3 or 4 months. But breeding pairs that were injected had near normal breeding activity. In addition, we found that survival of the offspring was restored as well. Not only did these animals produce fewer litters, but they don’t survive very well either. They’re, we measured these at, at one month of age, and very few, 3 out of 12 litters actually survived. Whereas, after injection they, they promote robust survival. In addition, we tracked the weight of these animals. And so typical mice will grow from about 15 grams to 25 grams, and in the mutant animals, they do have some growth, but it’s a stunted growth rate that you can see right here. So, we wondered if injection would improve their growth rate. Interestingly, when we injected the parents, we found that there was no real recovery, but if we injected both parents and offspring, we found that their growth rates were back to those of wild type. So these are secondary outcomes, but I think it does emphasize the significance and the importance of inner ear function.
Okay, we’ve talked about gene replacement for recessive mutations which could potentially address all of those shown here. We tried it for the dominant mutations, but because they are dominant in nature, that strategy did not work. So we thought about alternate strategies, and decided to focus on something known as genome editing, using CRISPR and Cas9. Maybe you’ve heard about CRISPR or Cas9. But the approach here is if an animal has a wild type gene, and a second gene that is dominant, carrying a dominant mutation, maybe we can selectively disrupt the dominant gene, leaving the wild type one intact, and perhaps allowing normal auditory function. So the strategy is to use a guide RNA. This is an RNA sequence that’s complementary to a native sequence. And the hope is that by using this guide RNA it may be specific to this mutant gene, but not the wild type. And then the Cas9 enzyme will sit down at this green location, cut the DNA, and thereby disrupt function. This strategy using conventional Cas9 enzymes did not work too well. So we ended up using an alternate Cas9, it’s known as SaCas9kkh. And interestingly, this has a site, it’s known as a proto spacer adjacent motif or PAM site, that’s selectively recognized, the dominant allele, this Beethoven allele right here, but not the wild type. And so the guide RNA would recognize both, delivering the Cas9, but because the Cas9 enzyme was not recognizing the wild type, we got selective cutting of just the Beethoven allele. And that’s shown right here, where we quantified it and found that it was 99% specific. We chose this strategy right here. 99% specific for the Beethoven allele, and not the wild type. We were able to package that into these adenoviral vectors. Another advantage is that the sequence here is small enough it can fit within one of these vectors together with the guide RNA and injected those into the ears of mice. First we did some safety studies just to see how this affected wild type hair cells, and in this case we’re looking at hair cell numbers. We saw no change, so no toxicity or no effects of disrupting the native TMC1 sequence. In Beethoven mice where the cells begin to decay after a month or two of age, we saw that we were able to preserve a number of cells, right here particularly in these middle frequency regions. And that’s quantified just here.
When we looked in the ears of these injected animals, we could look at the hair cell more morphology. Here’s a wild type series of hair bundles, outer hair cells and inner hair cells. The Beethoven hair bundles are badly disrupted. You can see the, the morphology is, is barely disrupted relative to the wild type. But after injecting these Beethoven animals, we find that the morphology is preserved, almost identical to that of wild type. And here’s higher magnification view showing the outer hair cells, disrupted Beethoven hair bundles, and after treatment with our CRISPR Cas9 strategy.
So what about function in these cases? And so, we’ve tracked auditory brainstem response thresholds in wild type mice at 4, 8, 12, and 24 weeks of age. They’re fairly stable over this time. The Beethoven animals have this progressive hearing loss, so their thresholds are elevated until they’re profoundly deaf at about 24 weeks age. But if we introduce our gene therapy Cas9 approach at early postnatal stages, we find we can preserve their auditory thresholds at near wild type levels. And we’ve now done this out to one year of age. And as you might predict, for something that’s changing the native genome, it’s, it’s stable and is a lasting effect.
So to summarize a, a few key points. TMC1, this is the take home message, we think is indeed the hair cell transduction channel. And that’s true for mice and for humans, but we think it’s also true broadly throughout the animal kingdom. We’ve gone back and analyzed the TMC1 sequence in a number of different species. It’s present probably underlying auditory function in, in birds, fish, reptiles, and in mammals. And interestingly in the mammalian branch here, we see it’s in a distinct separate branch of this tree. The sequences are, are different enough, but we think this difference may be a specialization that enables high frequency hearing.
Secondly, we think that TMC1 gene therapy replacement approach could be used to restore auditory and perhaps vestibular function for not just the, the mutations that I showed you, but for any of the 35 recessive mutations that exist in TMC1. With introduction through the round window membrane, we think we can recover function in 6 sensory organs, the cochlea, and 5 vestibular organs.
For dominant hearing loss, I showed you the CRISPR Cas9 strategy that if we can selectively target that we can preserve auditory function. And importantly, this strategy that we’ve devised for targeting Beethoven we think may be broadly useful. We’ve gone back to look at the database, this is the ClinVar database, and in that database there’s 17,000 human mutations that are classified as dominant mutations. Of those, taking this strategy with two different enzymes, there’s about 5,000 of those mutations could be targeted with a similar strategy. So we think that by using a, a viral vector delivering this CRISPR Cas9 approach, we may be able to address other forms of inherited human disease as well. Some of the genes that are on that list are involved with Alzheimer’s, some with Parkinson’s, some with epilepsy. And there’s at least 15 other genes that cause dominantly inherited deafnesses which may be targeted as well.
So I’d like to conclude just with acknowledgements. This is the group from my lab who contributed this work, as well as a number of our collaborators in, in the Boston area and, and elsewhere. And also I’d like to acknowledge the funding that supported this, including the NIH and then most recently the Foundation Pour L’audition in France. And, and there’s their website. So, thank you very much, and I’ll take any questions. (Applause) I’ve convinced you all that TMC1 is the channel. Great. Yes.
(NOTE: No microphone was used for the questioners so the volume level was very low.)
Q: Thank you so much. This talk was amazing. And I, I would love to talk about the deafness, and what I don’t understand. But I will, I’ll skip that and go to kind of a, a different perspective, which is can you talk about the way science is emerging across this (inaudible words) supercomputing conversation of ours, the semantics of searchability of the data that’s out there at big databases, for this kind of data, how, how the world is merging that collaboration.
JEFFREY: Yeah, that’s a big question. Things are, things are changing rapidly, and I think using AI, artificial intelligence, we may be able to tackle some a these problems that, that have vexed us for, for years, right. Some of the strategies I’m talkin’ about here, CRISPR Cas9, there, there’s all sorts of implications for the future, being able to change the native DNA in just subsets of cells. You know I’m not talking about at the level of changing it in reproductive cells. So, so any changes we are making here are gonna be local just in the ear, and, and we don’t want to, to start changing future generations. I, I think we’re not ready for that yet. But, so, so there’s ethical things that we need to be concerned about there a bit. But the strategies and, and some of the, the prospects for this in addressing inherited human disease I think is enormous. And so, I’m particular rly excited about that. And bein’ able to screen databases looking for some a these mutations with these strategies I think is, is not too far off. Yeah. Yes.
Q: So, you’re obviously looking at congenital or genetic types of hearing losses. What implications does this have for acquired hearing loss?
JEFFREY: Yeah, right. So the gene delivery approaches may also be applicable. So the one strategy I think Dr. Stankovic mentioned earlier, was using an HO1 transcription factor which is important during development for making hair cells. The thinking in that case is if you use one of these gene therapy vectors to deliver that sequence into a supporting cell in the inner are, you might be able to convert it into a hair cell for patients who have acquired hearing loss. That, the, you know, there’s some questions about that. I don’t think we’re, we’re fully there yet. There are other strategies as well for acquired hearing loss, but, but some of this could be applicable. Yeah.
Q: I didn’t know (inaudible words) a lot of this I didn’t understand. So just from applied, help me understand applied using (inaudible words), You’ve got a family that has a gene and the children are, this is before you were talking about the adults, and the (inaudible words) the child and adult, the child (inaudible words).
JEFFREY: yeah, that’s an important question, and that’s something we’re still exploring right now. We do know that at least in the animal model, that if you’ve got a TMC1 mutation, the hair cells will eventually die. So if that’s also happening in the adult, once the hair cells are gone, this strategy is not gonna be effective. So there’s probably a, a window of opportunity, and, and if we intervene during that window, we may be able to recover function. But beyond that, it might not work. The key thing is how wide is that window? Does it really have to be right at birth, or can it be before age 10, or, or something like that? And that’s, that’s what we’re working on now, tryin’ to figure that out. Yeah.
Q: I’d like to take, it’s kind of a follow-up (inaudible words). In this situation now. What is the timeline though? Is it involved TMC mutation for what? Because currently we know that is the connected (inaudible words) that gene is more (inaudible words) you know actually averaging at (inaudible words). So especially now is looking at TMC1, with the (inaudible words) genetic mutation, right, what is the healthy (inaudible words) for that because it’s, and so basically essentially what I’m asking is that prevalence in the (inaudible words) the TMC mutation.
JEFFREY: Sure. So I agree. Connection 26 is certainly the most common, affecting you know estimates vary, but around 35, 40% of genetic hearing loss being due to that one. For TMC1, we think it’s within the top 5 or so genes that, that are mutations cause genetic hearing loss. And so the numbers we’ve come up with that are based on limited datasets, but we think in the range of 5 to 8,000 patients in the United States carrying TMC1 mutations. So economically is that, is that viable? From a business point of view, I don’t know. I’m, I’m a scientist. I’m gonna let the businessmen figure that out. But, we hope that we, we could be able to help some a those patients. Yeah.
Q: Probably after or in that injection that (inaudible words) have an injection that typically those results you know where you have stages right after. I mean what was the timeframe for that?
JEFFREY: yeah, so we were usually injecting during the first postnatal week, and then we could record the responses at about a month of age. So we would wait some time, allow the vector to get in, the gene to be expressed, the protein to get to the right spot, and we could then do the assay. So it, it takes a little bit of time before we see the effect. But within a month, it was pretty clear. And we’re able to track it then out to a year of age, and it seems table that long. Other questions?
Q: I just have a question. So, the very first part of this, why are there two peaks?
JEFFREY: So in the very first one I, we’re wiggling the hair bundle back and forth. Whenever the bundle is deflected in a positive direction, we see an inward current. As it’s moving back in the other direction, in the negative direction, we see that current decline. Is that what you’re referring to?
Q: I’ll just ask you later.
JEFFREY: Okay, that’s fine. Yeah.
Q: So while I still have your attention, i just wanna remind everybody to please fill out your comment forms at the end of convention. And if you really enjoyed these sessions, please let ASHA know that, ‘cause it’s very helpful to us for planning future hearing research symposiums. And with that, I hope you all enjoyed this as much as I did. I thought it was absolutely fantastic. Thank you so much for coming.
JEFFREY: My pleasure. (Applause) .