Skip to main content

Transforming the understanding
and treatment of mental illnesses.

Celebrating 75 Years! Learn More >>

The NIMH Director’s Innovation Speaker Series: Advancing Therapies for Central Nervous System Disorders


>> ALEXANDER DENKER: Good afternoon.  My name is Alexander Denker.  On behalf of NIMH I'm very pleased to welcome Dr. Beverly Davidson as this month's Directors Innovation Series Speaker.

Dr. Davidson is the director of the Center for Cellular and Molecular Therapeutics and Chief Scientific Strategy officer at the Children's Hospital at Philadelphia and holds the Arthur Meigs Chair in Pediatrics.  She's also the professor in the Department of Pathology and Laboratory Medicine at the Perelman School of Medicine at the University of Pennsylvania and co-founder of Spark Therapeutics, Inc., a biotechnology company at CHOP created to bring new gene therapies to market.  The research is focused on understanding the molecular basis of neurodegenerative disorders and development of novel therapies to treat inherited brain disorders.  She is associate director of the Center for Gene Therapy for cystic fibrosis and other genetic diseases.  Her talk is titled "Advancing Therapies for CNS Disorders."  Welcome, Dr. Davidson.

>> BEVERLY DAVIDSON: Thank you very much, Alexander for that kind introduction.  I'm excited to be here today.  And to be -- to share you some of our work in gene therapy for the brain and with application to neurodegenerative diseases in particular.  And I hope today that some of what I share with you, you will find applicable as we begin to envision using gene therapies for mental health disorders.  These are my disclosures.  As mentioned briefly by Alexander.  So I want to focus the first part of my presentation on repeat expansion disorders that encompass a range of inherited diseases that are caused by repeat expansion in various regions of the genome from the five-prime UTR to the three-prime UTR and these include very, very large expansions such as seen in Fragile X and myotonic dystrophy, and the smaller expansions we see in the coding axons for Huntington's disease and spinal cerebellar ataxias.  For the CAG polyglutamine repeat diseases such a Huntington's and spinal cerebellar ataxias is a toxic gain of function mechanism.  So the idea is we would like to develop methodologies to reduce the presence of that toxic protein and the messenger RNA that encodes it.

There's a number of different approaches that are out there.  I'm not going to read through all of these, but I think it's important to know that various groups are targeting the mutant protein for enhancing degradation.  Examples include shepherding it to the proteosome.  There are a number of approaches up until recently that; were in the clinic that involved antisense ASOs.  As some may be aware, antisense oligonucleotide programs spearheaded by Roche as well as therapeutics have since put on hold and no longer moving forward.  There are clinical trials using RNA interference in a non-allele specific way, which will presumably provide reduction in the transcripts and hopeful this will be therapeutically relevant to HD patients.  But what I want to focus on today is not going after the protein or going after the RNA, but thinking about the DNA that actually encodes this toxic mutant allele.  And for that I want to focus on some CRISPR cas approaches, and this was largely spearheaded by a senior scientist in the lab, now assistant professor at the University of Pennsylvania, Alex Mas Monteys.  When Alex and I thought about CRISPR Cas gene editing, here we are either going to repress the gene expression or we're going to cause deletion in the actual gene encoding mutant Huntington.  And unlike RNA interference and ASOs which target the transcripts, we thought it very important to think about where we should focus the deletions of the mutant gene.  There were two recent papers that came out of Joe Bait's lab and David House's lab, together they collaborated to identify aberrant splicing within the first axon of Huntington.  This is the region of Huntington that contains the polyglutamine repeat region and, of course, the expanded polyglutamine repeat region.  What they found is that this alternative splicing within exon 1 in and of itself can give rise to a fragment where there is expression of this mutant exon 1 protein that may be pathogenic.  In addition to David and Julie's studies was the finding from Laura Random's lab where she identified that these repeat expansion regions similar to what occurs in other microsatellite diseases, these repeat expansions can actually give rise to aberrant transcription and translation.  This is known as RAN translation because it is ATG independent.  And these toxic products, whether RNA or the encoded proteins could also be contributing to the pathogenic sequelae within HD.

With this in mind, what we decided was to take a look at the mutant allele and scan upstream of the mutant allele and upstream of the axon 1 and downstream of five prime UTR very close to the promoter region and look for single nucleotide polymorphisms that were linked to mutant allele that would give rise to a PAM motif that would allow directed editing.  Fortunately we found a number of SNPs that seemed to be linked to mutant allele that would give us a handle for allele-specific editing.  So if we combine the upstream PAM with the downstream PAM, the idea would be that would promote exon 1 deletion in the mutant allele only and leaving the normal allele intact.  We first tested in common cell lines that just happened to contain many SNPs that were surveyed among HD patient samples.  And so we tested a number of different guide pairs, upstream guides and downstream intronic guides as you can see indicated by intro 3 and 4 guides or upstream guides steps 1 through 6.  You can see very effective silencing of Huntington expression levels on the left and denoted by the red arrows.  And then the simultaneous decrease in the protein that you see on the western blot over on the right and quantified in the lower bar graph below.  This was encouraging, suggested that the SNPs were useful and could be used for active editing.  And next we moved into Huntington cells.  So these are cells derived from Huntington patients that contain the SNP.  I'm showing a small sample of the data presented earlier in the paper that I have cited below.  So I just want to show you our ability to use this method to selectively edit the mutant allele.  So here is a situation in this cell line where we have a polyglutamine repeat expansion in the mutant allele with a linked single nucleotide polymorphism.  When we apply the guide RNAs directed to SNP1 with intronic 3 you see roughly a 50% decrease in the levels of transcript.  And if we look by western blot and you look at the mutant protein by the red arrow and compare that to the normal protein with the blue arrowhead, you can see that in the control treated cells without guides, the relative levels of the mutant to the wild type protein, and then when we add the SNP targeting guides in you can see selective decreases  in the levels of the mutant protein.  And this very much correlates with what we saw in the RNA, again, with about a 50% reduction.  What about in vivo?  Fortunately for us, William Yang made an HD mouse model that contains a full-length human gene with an expanded polyglutamine repeat, and it turns out this particular allele also has many of the SNPs that we were interested in assessing.  And so we took our candidate guides and Cas 9 and packaged into recombinant RNA viruses and injected into one side of the mouse brain and then compared the level of editing on the treated side of the brain compared to the untreated side.  What I'm showing in this first panel here on the lower left is editing as evidence by the -- in the splicing assay you can see that the right striatum injected with the virus, you can see a very obvious editing here, whereas we're not seeing that in the left striatum, and tail DNA which the virus does not travel you can see as a negative control.  The consequence of this was a 50% reduction in the levels of human Huntington mRNA from this single infusion into the right striatum, so this is very encouraging.  So this data and additional data that I didn't present here show, I think very effectively, that you can take cells that expressing both the normal and mutant protein, selectively edit the mutant protein to give you a cell that now only contains normal Huntington.  And we know from others' work that 50% of normal Huntington is well-tolerated in the brain and will not cause any deleterious sequence of events down the road.  So this would be an ideal situation.  But where are we going to -- where are we going to go with this?  So, there are several things that I want to spend the rest of seminar on showing you, how we're kind of thinking about advancing this into the clinic.  One is to improve AAV targeting.  I'll talk about that in a minute.  The other is phasing of this SNPs, the highly prevalent single nucleotide polymorphisms on the mutant versus the normal alleles.  And I'm not going to have time to show you that data today.  And the other is improving the safety of AAV delivered editing machinery.  Why do we think that's important?  Well, cas9 is a bacterial protein.  It's not a human protein.  We only need to edit for a short period of time.  We don't need it on forever.  And so it would be ideal if we had a pulsive editing that then waned and then essentially the AAV machinery would be silent in cells when we didn't need it anymore.  So how are we going to achieve that?  I'm going to kind of step back and tell you how we got to being able to do regulated editing.

So, for this, I want to introduce you, for those who are not familiar, with spinal muscular atrophy.  And the fact that the severity of spinal muscular atrophy has to do with the dosing in the number of SMN2 pseudogenes.  Humans have an SMN1 gene and SMN2 gene, and SMN patients have mutations in SMN1.  And if you have multiple pseudogenes of SMN2 it can functionally complement for the SMN1 because a small portion of the pseudogene is correctly spliced.  And, in fact, there are drugs that are approved and some in clinical testing that can improve the level of splicing to include exon 7 and shows efficacy in SMNA patients.  In fact, these two drugs are orally bioavailable.  Patients take them orally.  Some of the dosing is more frequent than others.  You know, maybe a couple times a week.  So, the idea here is that if these drugs can promote exon 7 inclusion and allow for the expression of full-length SMN2, what if we take advantage of this and instead of driving SMN2 expression we drive our gene of interest.  In this case, Cas 9.  So in our first examples of developing what we're calling a splicing switch for regulated control of gene therapy, we took this SMN2 mini gene, E6, E7, E8, fused it upstream in our gene of interest, and the first slides I'm just going to show you a lot of reporter data.  And the idea here is when the exon is not include, it's out of frame, there's no expression.  In the presence of drug, it's in frame, and there's protein synthesis.  And, in fact, we can alter the levels of inclusion from not the wild type levels of 10% in the baseline, but to the level of 1%.  So we asked, okay, what does this do for regulating gene expression?

So, we cloned a reporter downstream of our splicing switch, and in the absence of drug, you can see this baseline level, there is some baseline level of reporter expression.  But in the presence of drug, this drug here is LMI070, and you can see about a tenfold induction in the levels of gene expression.  This also -- not only this works with Branaplam but also the other drug I showed in another slide.  Well, this was fine tenfold increase in expression, but our concern was that this baseline level of expression was too high.  We wanted something that was a lot more off, if you will.  And also this required a fairly high dose of drug that probably would not be translatable to human therapies.

So we went back to the drawing board and exposed human cells to a low dose drug and then did an RNA seq analysis to identify novel exons that we may be able to incorporate into the splicing exon system.  This work is not only done by Alex, but also Paul Ranum who did all of the computational work for these studies.  So one of the candidates listed here shows you the difference in the absence of drug in red, and the presence of drug in blue.  Again, this is low-dose drug that would be more translatable.  And you can see the inclusion of a novel exon by this yellow bar in the presence of drug.  And we can also confirm this by looking at the amount of inclusion as assessed by a splicing assay.  And so here is our top five candidates that we found.  And you can see in the absence and presence of drug that there's another PCR product, suggesting the inclusion of the exon.  How responsive are these new non-SMN2 mini gene candidates?  So it's essentially the same structure here.  However, we did one more trick, we moved the beginning codon into the included upstream exon.  Absence of drug, no exon inclusion and no protein expression.  And in the presence of drug, there's exon inclusion and protein expression.  And I hope you can appreciate now these are really very much off.  Each of these candidates.  And our level of induction now is 200, not 10.  And we think this gives us a finer control because now we can play with doses of drug and we can play with different promoters and all kinds of things I'll show you in a minute.  Is this really off?  Here is another example that is a little easier to see visually than Luciferase, and I think you can appreciate in the lower right panels in the absence of drug, you have no GFP expression in the presence of drug.  Again, this is a single pulse of drug.  You see robust GFP expression.

What about looking at different promoters to control dose as well as different doses of the drug?  Here is just a list of three that we're showing here, RSV, PGK, and mCMV.  And you can see at different doses and with different promoters, you have different levels of protein expression, which could be useful if the levels of expression you need for the experiments that you're doing are very modest, alternatively you can need very, very high-fold induction for a short period of time and need a more robust promoter.

And this is the splicing assay that shows you, with the different promoters and at different dose es, you have different levels of splicing in of that novel pseudoexon.

So I want to come back to the GFP study.  We had such nice GFP data in vitro, and we wanted to move in vivo, so we packaged this system which we're calling Exxon, shh a single transgene into AAV and we injected this into mice.  This is an IV injection.  And then we analyzed tissues -- and I'm just going to show you a smattering of the tissues today.  So, we did the AAV injection and waited a few weeks and then gave a single oral dose of the LMI070 at these doses here.  You can see 5mg/kg and 50mg/kg.  And I think you can appreciate in the vehicle treated animals there's no GFP expression, and the GFP expression with the low and high dose is commensurate with the dose of the drug.  Just to convince you of that, here is a western blot where you can see the levels of EGFP in a dose-responsive way, and to convince you this is truly off, we overexposed this western blot tremendously.  And I think you can appreciate there is no GFP expression in these animals in the absence of the drug.  Well, what about the tissue we're actually interested in here, which happens to be the brain?  We repackaged the same exon e GFP cassette into AAV/PHV, which is high grain tropism.  We injected these retro orbitally into mice, and I think you can appreciate two panel series in the hippocampus and cortex and you can see both at the 5 and 50-milligram doses we have a robust turning-on of our e GFP expression that was really exciting.  And when we look at a splicing assay, you can see that in the vehicle treated animals there's essentially no splicing in of that ATG containing exon with the low dose and high dose.  We have varying levels of splicing.  And below that is the fold induction as assessed by quantitative PCR, which is relevant here.  So, again, we get 180-fold induction with 5 milligrams per kilogram, and over 1,000 fold induction based on transcript levels at the high dose.  And this is, again, a single oral dose given to these animals, and then the animals were euthanized a few days -- about a day later.

So why did we develop the system?  It wasn't to, you know, treat people with eGFP.  Instead it was to control the levels of Cas9 that was expressed and to be able to have a pulse of that.  So I'm just going to show you one of the -- a little bit of that data here.  We essentially replaced the eGFP with the cas9.  And what we wanted to know is whether or not we got induction of cas9 to a level that was sufficient for editing similar to a con constitutively on system.  First we assessed by western blot either the constitutive levels of expression or the switch varying doses of drug.  This is all in vitro.  You can see in the absence of drug, this is truly off.  There is no cas9 expression.  And as we increase the dose of drug, we can get up to the levels of constitutive expression.  And when we use this to control cas9 in the presence of our allele specific guide RNAs, our hope is we get the level of silencing that we saw with the constitutive system.  And so here are the transcript levels when we assess the ability of the exon driving cas9 to silence Huntington in an allele-specific way.  And I think you can appreciate here that -- so on the left is a control with control guides and no drug, and then the next gray bar are the guides, but no drug, but that's with a constitutive system, those left two bars.  The orange bars are controls.  Control guides.  And the green bars are the Huntington-specific guides, plus -- minus and plus the LMI070.  And I think you can appreciate that the level of reduction at about 50% is consistent whether we use the exon system or a constitutive system.  So this is very exciting and something we want to move forward for safety reasons.

The challenges for moving forward are regulation, which I feel that we're getting there.  Again, fine control of gene expression with a brain penetrable orally bioavailable drug.  For AAVs, it's the capsid.  So we have several efforts in the lab where we're doing ongoing engineering for improved potency.  And this can have several positive impacts.  One, it can reduce the needs for extensive manufacturing, so reduced costs.  The other is it gets us to the right cells and not just a whole bunch of cells that we really don't want.  And it also, together with the regulated system, can provide for the right level of expression in the right cells.  So how are we going about engineering the right capsids?  Our strategy depends on the application.  In some cases we want AAVs that go everywhere because the entire brain is degenerating.  And so for this we have created AAV libraries based on a number of different serotypes.  Each serotype library is around 10 to the 7th or so distinct capsid variance.  We infuse these into the CSF.  And these are in non-human primates.  We infuse these into the CSF and we go through a couple rounds of evolution, collect those tissues, and then subject the capsids that we recover to next-generation sequencing.  I'm just going to show you a few examples because this could probably take up 15 figures, but just to kind of highlight for you how we might be able to use this to identify capsids that are more appropriate for what we intend.  So this is just an example for one of those AAV serotypes that went through a validation.  And each one of these rows is a different capsid variant.  And what you can see by the density of the purple is the fold enrichment relative to that input.  And this is the input after the last round of maybe about 150 different capsids.  And you can see some variants seem very good for the limbic system, some for the hindbrain, some for sensory systems, etc.  And for each of these we can go in and validate them as we move along.  So, how do the best of the best peptide modified AAVs, which is what we call these, compare to the state-of-the-art?  For all practical purposes in humans and in other large animals, it's known as AAV9.  So here is an example of a peptide modified AAV, which you can't see the data for yet, and a wild type AAV9.  After injection into the CSF.  I'm showing you a region of the hippocampus.  It's a strong promoter driving the transgene M ruby.  This is four weeks after delivery into the CSF of a non-human primate.  I think you can see a smattering of red cells here, but what is really impressive is you compare the peptide modified AAV within the hippocampus with the robust transduction within this region.  And so we're very excited about this sort of broad spectrum peptide modified AAV that may provide enhanced transduction and thereby allow us to do gene therapy with much lower doses.  What about engineering vectors for network level expression?  And this can become important when you want to control a network that may impart changes in behavior or may impart -- or maybe impact specifically in neurodegenerative diseases or other mental health disorders.  And so, again, it's the same starting approach with the AAV peptide libraries.  And here we do a parenchymal injection into the network node and then do our essential evolution here again through two rounds with next-generation sequencing.  And I'm just showing you the mid-level validation experiment from about 150 capsid variants that came out of that second-round library.  And these are three different libraries, AAV, AAB and AAVC on the left, and you can see the relative fold enrichment for these different kinds of libraries.  And this is also represented in the regions of the network which are little a through little f, where we are hoping to get coverage within this network so that we can do one infusion and then impact the -- and then get the virus to all of those regions of interest.  And so each of these highlighted with the arrow here are now moving forward in final validation studies, and it will be very interesting to see these properties again from a single stereotactic infusion and then follow where they go.  So the challenges moving forward is regulation.  We are getting there.  Fine control of gene expression, again, with this oral drug, and for AAVs I think we're making a lot of progress.  We're in the state of merging our ongoing evolution studies with single cell methodologies to get at the cell level and not just regional levels.  And then we pioneered some work in developing better ways to evaluate tropism, at least in small animal models, which helps give us some indication of not only are we getting where we want to go, but are we not going where we don't want to go.  And I think you'll be surprised by some of the experiments I'm going to show you.  Now, most of this is not in the brain.  There's a little bit in the brain, but it's relevant nonetheless.

So the conventional way that we look at where A AV goes is to have them express a reporter, put them in an animal, and then look and see where it is.  And so conventionally if we're looking for something that transduces the liver, for example, you take an AAV vector and express e GFP, and you can use it to express cree to flip on sequences, but what we did here is took this Cre infused eGFP and we did what Jon Lang who did this work is calling a revised screen.  You have an A14 mouse which has a floxed stop upstream TD tomato and you infuse that same virus and now you'll be able to tag every cell that got the virus, whether it's a stable transduction event or a transient transduction event.  How does this work in practicality?  Here is what we would see a dose-dependent increase in GFP expression when we infuse IV Cre eGFP and just look for GFP.  And this is just normal black sick mice.  If we use the AI14 mice, I think you can appreciate that every liver cell is red.  And, in fact, if you quantify this, the percent transduced from your standard transduction in the gray bars you can see at best, you know, you get about 20% of the liver, but, in fact, what we're seeing is almost 100% of hepatocytes are transduced and we would have completely missed it using standard screening methods.  We kind of asked, well, is it possible that we're getting more AAV genomes into the cells?  And that turned out not to be the case.  So how relevant is this as we look all the way across tissues, particularly for IV injections?  And you know that for some brain gene therapies, there are IV injections being done with brain tropic viruses and we think most of the viruses are going there, and what is getting peripherally, you know, is largely shed, or we understand it to some degree.  So here we are again with our AV8 expressing e GFP views to Cre in standard mouse model.  You can see nothing in the brain, a little bit in the heart, nothing in the lung, same amount that we expect in the liver.  Nothing in any of the other tissues down below, kidney, spleen, skeletal muscle or testis.  Yet when we use the Ai14 mice, I think you can appreciate we're getting vessels in the brain, we're getting quite a few cardiomyocytes and certainly hitting cells in the lung.  Again, our liver is 100%.  100% of the hepatocytes took up the virus.  We see glomeruli in the kidney.  Spleen positive which may help explain some of the immune responses we see.  Skeletal muscle is positive, and support cells in the testis are positive.

What about using the same system for editing?  Because remember I came at this talk from the beginning, thinking about advancing DNA editing approaches for dominant gain of function disorders similar to Huntington disease.  And so we used another mouse here, the Ai14 mouse again that expressed SPS cas9 and came back with our virus that expressed guides and said, where is editing possible?  And it was really quite surprising.  Again, we see editing in the brain in the cerebellum, heart, lung, really all the tissues we assayed, we saw evidence of editing.  We even saw evidence of editing or expression when we drove the Cre in the first experiment from a liver-specific promoter.  So, you know, I think what this study highlights for us is how unusual and how much gene transfer we're really getting in these tissues that we probably didn't realize using conventional screening methods.  And interestingly enough, if we're thinking about allele specific editing, we can also use this technology to see how effectively are we actually editing both alleles?  And it turns out we're not so good for hitting both alleles.  So this is an example using AAV 9 in the heart and some other peripheral tissues, and AAVB in the brain, throughout, the cortex, striatum, hippocampus and cerebellum.  And you can see there's very few double edited cells but mostly a single edit where you have either red or green.  And I think this will be important in informing how efficient we are in editing mutant alleles in vivo.  So just to summarize -- and I hope I left plenty of time for questions, allele specific editing of disease alleles can be accomplished in cells and animals.  We've advanced approaches for regulating knock down, essentially creating a rheostat that can be used with an orally bioavailable drug that's been in people, it's in children, and this gives us a method for drug-based regulation.  We're also devising ways where the state of the cell regulates expression.  And we have been able to identify novel vectors for focused or widespread or even network level transduction.  And I think as these emerge, we'll go back to our lower animal models and assess the footprint of transduction.  And thus far, anything that we have found in the non-human primate, actually has translated down quite well.  Which is interesting, because as the evolution work happened in mice, it didn't translate up to larger animal models.

I'd like to thank the members of my lab and, of course, the funding for this work is both the NIH and the Children's Hospital of Philadelphia Research Institute as well as foundations such as the Herd Disease Found days and Hopi's Hope.  And I think here is a happy moment after we heard some good news.  And this was obviously pre-COVID when we could get together, very high density, probably over the population of the room.  So with that, I am happy to take any questions.  I can see the questions here, and I can start to go through them.  I'll start at the top.  Ryan says, I'm confused by the translational viability of small molecule regulated splicing that uses a small molecule.  Does it also modulate endogenous genes?  Seems the system likely will have off-target effects.  Ryan, that's a really great question.  So this small molecule, obviously, it will regulate -- or it will induce the splicing of a handful of genes at this low dose in which we screened.  Again, there's only a handful of genes that -- for which there is a suicide exon incorporated.  And it's pretty short lived in cells.  So the idea is that we're using a dose to control gene expression that is not going to cause side effects in humans.  And we know a lot of the human data, quite frankly, because this has been used for several years in kids.

Jim Bloss asked why the therapies were canceled.  And I think he's referring to the Roche studies for -- the allele nonspecific Roche studies for Huntington's disease and the wave studies for Huntington's disease.  The Roche studies failed to meet their end points, to my knowledge.  And the Wave studies showed no evidence -- again, this is all on the website.  So, that's all I know is what I see.  And for Wave there was no evidence of Huntington lowering or target engagement after the infusions.  Jim also goes on to ask, can you provide a little more detail about why 50% of Huntington in the brain was tolerable?  This is from knock-out experiments in mice, and from other knock-down experiments using other approaches.  So 50% reduction of the total levels of Huntington doesn't appear to be deleterious, and if you have a Huntington deficient animal for which -- that is a heterozygous animal, they're fine.  Samuel says, at what age will the genetic intervention be implemented?  Will existing neurodegeneration reverse itself as a result of the intervention?  That's a great question.  For Huntington's disease, we think we have to go early.  There is a progression of disease in which you go from modest degeneration or modest atrophy to profound, and it would probably be right around the modest atrophy stage.  I think similar to the studies that were set up by both Roche and Wave in terms of early on either just write it onset or even a little earlier.  Kurt... any comments -- Kurt Fishbeck -- hi, Kurt.  Any comments on AAV toxicity?

Yeah, so one of the things we're working in our AAV evolution systems is not only to evolve things where we want them to go, but also evolve them to not go where we don't want them.  And in this case I think you might be referring to the dorsal root ganglion cells where there seems to be some toxicity for AAV9 particularly in adults.  And I think using the methodology I used here you can deselect for any of those capsids that tend to like that area.  Jim asked again, are you aware of any other studies that may be looking into use of CRISPR research in studying SMIs such as schizophrenia?  If so could you identify the research and perhaps some results of their work?  I am not.  But I've been in conversations with several investigators in the mental health arena, and, you know, the question is, will it be a single gene or a host of genes that you would try and regulate to have an impact.  Jonathan asks, in the revised AAV screening, if so many cells are transduced and able to express Cre and then RP, why don't those same cells express GFP?  That's a great question.  The reason they don't express GFP is because this is transient.  And that is what is relevant for editing, particularly nuclear editing with CRISPR.  So it turns out that not all transduction events are stable, but yet this data shows us that there are a number of transduction events that are occurring that we had no idea were happening, and it's really relevant for editing.

And he goes on to say, is the copy number of AAV too low to see GFP expression?  Exactly.  And that is the beauty, because what that means is that we don't need those super high doses for gene therapy in the liver for stable transduction when we are going to employ an editing strategy for something.  We may be able to go in an order of magnitude or lower to have an effect.  And I think this is, you know, relevant as we move the systems forward to humans.  Ryan asks, have you tested the exon transcriptional control system using the Ai 14 model assay?  No, we have not.  And that's a good suggestion.  Thank you.  And I think there is one that was dismissed that I want to say -- it says that -- Jim says, I'm very impressed with our signers ability to keep up with this level of scientific detail.  And I congratulate the signers as well.

>> ALEXANDER DENKER: Thank you so much for an excellent talk.  And I concur about our signers' ability, and I recruit the great team of sign interpreters who can deal with all of our jargon-y things.  I want to ask, we have more time for questions if anyone else wants to write into the Q&A box, but I want to ask if you can comment a little bit about some of the business-y side of your work and what goes into making the decisions on when to move forward with the specific gene therapy or not.

>> BEVERLY DAVIDSON: Yeah.  Money.  So, as an academic, yeah, we have to make very... we have to make tough... you know, companies do too, but we have smaller purses.  And what we -- what is required is that the data really needs to be compelling.  And then we go out and we raise money to do the I&D enabling studies.  And if the I&D enabling studies support moving to the next step, then that is what we try and do, either collaborating with biotech or pharma or trying to move it forward in house.  We're very fortunate here at CHOP to have a GFP manufacturing facility that allows us to get through the I&D enabling phase and also to first in human.  However, we still have to raise the money to pay for all of the product and the clinical trial.  And part of what I -- what we have worked toward with funding from NIDMS is advance a couple of gene therapies towards the clinic, taken about ten years.

>> ALEXANDER DENKER: Thank you.  At CHOP you're obviously working with a very vulnerable population, children and also their families.  What are some of the challenges you face in terms of expressing to families and explaining these experimental therapies to them and going through the risks and why it might be, you know, worth it for their kids and families?

>> BEVERLY DAVIDSON: Sure.  As a non-physician, I don't see these patients, but I work closely with physician scientists and physicians that do take care of these patients.  And I think we need to be 100% upfront about the risk of these therapies.  We can't over-promise.  You know, these are experimental therapies at this stage and we hope to have a positive impact and reduce the disease burden in these individuals.  I think gene therapy has very -- has a lot of promise.  We have been able to, you know, make headway into some very severe diseases with great success.  You know, I think the SMA story is an example of that.  But there's a lot of disorders that may not be minimal to the current suite vectors that we have available and we have to be smart about what we do, and really understand fundamentally and convince ourselves that, you know, we can make a positive impact on these patients before we would move into kids.

>> ALEXANDER DENKER: Thank you.  We do have another question that came into the box.  A couple questions now.

>> BEVERLY DAVIDSON: Yeah.  Anthony... yeah, I can read it if you want.


>> BEVERLY DAVIDSON: Anthony asks if the approach has been assessed in larger animal models or plan in the near future, would similar coverage of transduction be expected?  I'll answer that.  So all of the work that we're doing in the evolution experiments is in large animal models.  It's not in small animal models.  However, we don't have a way to assess that footprint of transduction in large animal models yet because we don't have any Cre expressing large animal models.  But we could do that with Cas and then evaluate editing.  With the fail of Roche's study, due to insufficient transduction resulting in insufficient suppression?  Roche's study is tin fusion of oligo nucleotides into the CSF via the lumbar puncture.  It's given repetitively, I think monthly or every other month.  And there's evidence of Huntington lowering in the CSF from the therapy, so one would assume there would be benefit.  What we don't understand is where that Huntington is coming, where that mutant Huntington is coming from, what part of the brain or what -- even what part of the, you know, the spinal column, and whether or not that is reflecting where the ASOs are going.  So I think there's really a lot we don't understand in the instance of the antisense oligonucleotide therapies.  The mRNA lowering therapies, we worked on in a number of years and continuing to work on in the spinal attacks.  That has worked to the clinic I think with Unicure.  Jonathan asked if CRISPR mediated double strand breaks has been shown to lead to chromosomal rearrangements and does this concern me.  Yeah, it would concern me.  And we are in the process of evaluating what is going on with an active CRISPR to be quite honest.  All the work that has been done to date has in vivo has mostly been done with viral delivered CRISPR.  So this is constitutively on.  And in the case where you have active editing and viral genomes and it's on for a long period of time, I think that you will increase the likelihood for unwarranted and unwanted editing events.  And this is why I think this regulated expression is so critical.  Because you really only need it there for a short period of time and it should reduce the numbers of aberrations.  Other alternative approaches are to not use active nuclease at all but just to take advantage of the CRISPR's ability to guide machinery to the region of interest for either gene silencing or some other alteration that will inhibit mutant allele expression.  Ryan asked if there's a comment on the challenge of dosing when using viral vectors and are there good mitigation strategies beyond immunosuppression?  There are some recent papers that have come out a looking at transient diminution of anti-A AV antibodies that allow for redosing for the liver and peripheral tissues.  In the brain, we're not sure when redosing would happen.  The data from Chris Bankovitch suggests that expression is maintained for at least a decade, and this is in non-human primates.  So we don't think it will diminish.  That's why we think with an editing machinery we really need to control the levels of gene expression.  And the other thing to keep in mind in the brain is while we may induce an antibody responses, what we want to be worried about not so much neutralizing antibodies but cell mediated toxicity and need to fully understand the induction of toxicity in the brain.  It's not happened.  I think the biggest other concern in the brain particularly for gene replacement is you develop antibodies to the recombinant protein that the individual has never seen before because they're null for the expression of that protein.  And that is where the immunosuppression I think is doing a very good job in inhibiting an antibody response to neutralize any ability of that protein to do what it needs to do.

>> ALEXANDER DENKER: Thank you so, so much.  Thank you so much for taking the time to be with us today.  And NIMH is very happy that you were able to take the time.  We sorry we are not in person, but grateful that it worked even in this medium.


>> ALEXANDER DENKER: And thank you to all the attendees.

>> BEVERLY DAVIDSON: I want to thank everybody for their great questions.  And, yeah, it's fun to go and talk in person, but this is sure easy on the travel budget.

>> ALEXANDER DENKER: Thank you so much, everyone!