Skip to main content

Transforming the understanding
and treatment of mental illnesses.

Celebrating 75 Years! Learn More >>

Workshop: Gene-based Therapeutics for Rare Genetic Neurodevelopmental Psychiatric Disorders


Geetha Senthil: Good morning. Everyone, I welcome you all to the NIMH virtual workshop on gene based therapeutics for genetic neurodevelopmental disorders. Today we have a panel of experts from academia, government and industry to share some success stories and their perspectives on challenges and opportunities in developing gene based therapeutics for individuals with rare genetic neurodevelopmental disorders. Specifically we will review the current state of science in gene based therapeutics including gene targeted approaches, vector design and delivery options, characteristics of suitable gene molecular targets and mechanisms amenable to gene based therapies. To set us off with overview talks, followed by a set of talks on key studies of gene targeted approaches that are currently in the clinic or in the development to highlight the scientific rationale for treatment options and also the milestones for reaching key stages of therapeutics development, as well as we will talk about the challenges of these individuals or experts faced while developing the gene based therapeutics and lessons learned along the way.

So I want to thank the members of workshop planning committee for their immense efforts in developing the program of the workshop. My special thanks to the chairs of the workshop for the relentless efforts in developing the agenda for this entire workshop.  It's a two day workshop.  Please note day two will be on the 29th. We will release a summary of that discussion following the meeting in a month or so.  The chairs of the workshop I want to thank are Drs. Mustafa Sahin from Boston Children's Hospital, Guangping Gao from the University of Massachusetts, Sitra Tauscher-Wisniewski from Novartis Gene Therapies, Beverly Davidson from Children's Hospital of Philadelphia, Cynthia Tifft from National Human Genome Research Institute and Terrance Flotte from University of Massachusetts. I also want to thank my NIH colleagues from the program. There several of them listed. They have played a key role in developing the goals of this workshop. I thank them all, especially Meg Grabb and Ann Wagner and David Panchision for the countless hours they spent with me discussing various strategies for this workshop.

We have a five hour meeting today. So the first part of the meeting will be a set of four 20 minute talks, these are overview talks, followed by a 10 minute break. Please stay on Zoom.  Don't go anywhere. Commute yourself and join back on time. During the second part of the meeting, we'll have six talks on case studies focusing on rare genetic neurodevelopmental disorders to highlight some success stories and also the challenges for specific disorders. These include Rets syndrome, Angelman syndrome, MeCP2 duplication, and spinal muscular atrophy and giant axonal neuropathy. We will have a 40 minute time at the end of these talks for taking questions from the audience. So audience, please use the Q&A feature on Zoom to type up your questions. And please add the name of the speaker to whom you're directing the question. And the audience have an option to vote up a set of questions to prioritize for the speakers to address them. We will adjourn at 4:00 p.m. and thank you all for joining the meeting. We'll now welcome Dr. Josh Gordon, director of NIMH to say a few words, opening remarks, for this workshop.

Josh Gordon: Thank you, Geetha. I want to thank you, I want to thank all the NIMH staff and the members of the organizing committee for putting together the program today. I'm really excited that we're hosting this workshop in an era where gene targeted therapeutics have begun to be applied in the nervous system and have already achieved FDA approval even in a couple of cases, and we need to think about approaches that will be appropriate to treat the disorders that NIMH is responsible for. We need to think about it from the perspective of the long term. In terms of developing platforms that will enable a precision medicine approaches in the central nervous system in much the way that other Institutes are pioneering them for other organ systems. And we also, I think, have the right and responsibility to think about it in the shorter term which of our disorders are most amenable to this approach, to pioneering this approach, in as safe and ethically responsible a way as possible and in a way that's demanded by the tremendous morbidity and mortality that is conferred by some of the more severe neurodevelopmental disorders and other disorders that we care about. So I want to thank you for getting together as a group. I apologize, I'm going to have to leave to go back to another meeting.  I'll be popping in and out the rest of the day. I hope you have productive discussions and I look forward to hearing more about the conclusions that arise from the day. Thanks.

Geetha Senthil:  Thank you, Josh. I will turn this now over to the chairs, Mustafa and Beverly and Sitra and others to continue our first part of the meeting.

Mustafa Sahin: Thank you very much. And good morning everybody. I want to thank Josh Gordon, Geetha Senthil and the whole NIMH team for sponsoring and organizing this workshop and the co chairs as well as all the speakers for their contributions in preparation of this meeting. We all think this is a very timely and impactful meeting. We have a packed agenda today, an ambitious agenda, so to be able to stay on time as co chairs, we've decided to keep the introductions to a bare minimum, but I would urge you to look up the bios of the distinguished speakers that will be presenting today. Our first speaker and co chair of this workshop is Dr. Guangping Gao from the University of Massachusetts Medical School and he's going to be presenting about the gene targeted approaches and tools for CNS disorders.

Guangping Gao: Can you see my screen?

Mustafa Sahin: Yes, we can.

Guangping Gao: Okay. Thank you. My name is Guangping Gao and first just like a mask update, I want to thank our NIH colleagues and leaders for their tireless effort in organizing this timely and important workshop. I also want to thank our speakers and chairs as well as the panelists for their participation and important contributions to the workshop and thanks to all the attendees to the workshop. 

So my task today is give an overview on approaches and tools of genetic based therapy for rare disease. This is my disclosure. And human diseases can be categorized into two categories.  One is acquired disease and one is inherited genetic diseases. The causes for genetic diseases includes autosomal recessive mutations, autosomal dominant mutations, X linked mutations and haploinsufficiency. So the consequences of those genetic mutations include loss of function, gain of function or gain of toxicity, haploinsufficiency. So gene therapy actually is a highly attractive approach for rare genetic diseases, particularly neurodevelopmental and psychiatric disorders. This is because most of them caused by genetic defect and most of them are monogenic disorders. And the gene therapy actually offers the most direct therapeutic potentials. 

Now I want to give an overview on strategies for gene based therapy. The first approach called the in vivo gene therapy, in this approach there are two categories of this disease and once viral gene therapy. And I used FDA approved AAV drugs as example. The second category is non viral and I use FDA approved nucleic acid drugs. As examples, this include ASO and SIRNA. The reason we call in vivo is because those drugs we give you just like any other pharmaceutic drugs you directly give to human and treat the disease. The second category called ex vivo, where you take out human patient cells, genetically modified, expanded and infused back and those modified cells will function as living drug in patient to complete therapeutic outcome. And there are four drugs approved by FDA and EMRA. 

That's overall strategies and in terms of conventional approaches for gene based therapy I would like to summarize into the following categories. The first when you have a loss of function, then you replace these normal gene and accomplish gene therapy, that’s called gene replacement. The second category you may not have mutation on your gene, but you could get exogenous or endogenous gene through overexpression to accomplish gene addition therapy. The third is when you have a loss of function and the gene become toxic, then you can deliver a molecule to silence those genes. And finally, you can modulate gene and its expression through gene editing. 

So those are the approaches, but what are the key components for gene therapy research? I will summarize into the four: therapeutic gene, vector, route-of-administration to target tissues safely and efficiently, and animal models to study efficacy and safety. And among them, vector is most important component for the gene therapy. 

As the gene therapist, our desired features of a viral vector for in vivo gene delivery are the following. First we want high efficiency just like adenovirus, we want long term stability like lentivirus, we also do not want immunogenicity and toxicity, and we do not want genotoxicity. Actually, now all the viral vectors we have, Adeno-associated virus has it all. This virus is a single stranded adenovirus and is a protein capsid and is consisted of regulatory proteins and capsid proteins. In addition there is some accessory proteins such as AAP and MAAPs. Most importantly to date there is no direct evidence of a causative association base any disease in this virus.

How do we convert this virus into the vector? Basically you take out the viral gene and then you replace with your therapeutic gene cassette and then you provide a helper function and Rep/Cap function and put all those components into a cell that can be transient or stable transfection or to infection such as baculo-, adeno- or herpes virus. And then you generate recombinant AAV. The beauty of the system is you can simply replace the capsid, in this case we replace with AV9, and then become AV9 virus because the biology of the vector is primarily determined by the caps itself. So now AAV mediated gene therapy actually is a teamwork inside out. The Capsid is very important. How it dresses matters. It directs tissue tropism and intracellular trafficking, it dictates host immune responses, it delivers the therapeutic genes to the nucleus. 

Because Capsid is most important for AAV vector, so vector development is centered at Capsid discovery. The first strategy is looking to natural reservoir. In this case you identify sequences, capture it, and then through high throughput in vivo evaluation. The second strategy is called  directed evolution. There are three strategies, the first strategy is called capsid shuffling, and second called error prong PCR, by those means can generate the variants of the new AAV capsid. The third approach is peptide insertion, it basically conducts capsid surface panning of random peptides and you create new Capsid with those insertions. And third strategy is rational design. You basically, based on Capsid structure-function relationship or you redirect a tissue tropism receptor/ligand engineering. And the final strategy is called in silico design or machine learning. This is based on the phylogenetic analysis and computer science. And the most popular current approach is really the pathway insertion, many vectors in the past several years have been generated through this pathway. 

My AAV gene therapy career starts in 90s, when I joined Dr. James Wilson at the University of Pennsylvania searching for the next generation of AAVs. This strategy we started 2001. We designed primers, a neotube conserve reaching and the general library of the Capsids. And really I want to give you example, that's AAV9, which is the most popular AAV today for rare disease.  This was discovering a human liver samples on January 21, 2003. As showing here, that's the PCR band that contains, we amplified on human tissue contains this AAV9 and this is the topochromic clones, and this marks the beauty of the virus at the strong transcytosis and across vascular structure spread transgene widely through muscle, liver and most importantly can cross blood brain barrier lead to widespread seeing this transaction. We also recently discovered another AAV2 natural variant called a V66. As you can see, it's a very close relative to AAV2.  And this virus only has 13 amino acid difference from AAV2. However this difference makes it 13 fold more neurotropic after intrahippocampal injection. 

And so that's the Capsid. However, as I said, AAV is teamwork, inside out. The genome itself is very important. This therapeutic payload carrier it mediates therapeutic benefit, achieves long term stay in the transmuted cells, triggers innate and adaptive transgene immunities, if you do not have good design but it causes potentially other transgene-related toxicity. And actually engineering the vector could start from ITRs, which is terminal repeats, as showing here actually pioneer work by [name] and Dr. McCarty back in early 2000s, they have done this and created a potent vector itself called [Away from mic] vector. With single strand DNA we package that can go up to 5KB, if packed as single stranded DNA and once get into cells, you go through this second strand synthesize or self allele-ing, you can generate virus that is transductionally active. This is a two difference dose in mouse variants, however, those scientists, they modify the one ITR, make a mutant ITR, then when package, you package as a mirror image of two copies and folding back from double stranded. By doing so, as you can see here, you bypass the second strand synthesize, the early onset of expression and high level of efficiency. As you can see here, there's at least 10 folds difference between single strand and double stranded. The second engineering could be a transcriptional regulation. When you use ubiquitous promoter single IV injection AAV9 it transduces many different tissues.  If you add different tissue-specific promoter you can limit transaction in that tissue. You could use cell-specific by using neuronal promoter or glial promoter accomplish cell specific single transduction. The beauty of the AAV is a single injection can maintain long time, this is ubiquitous promoter expressed, which is equal in monkey muscles for eleven years. However you could also add a regulated promoter, in this case erythropoietin you can have inducible promoter as the level as well as the intervals you want.

Another engineering you can do is post-transcriptional regulation.  In this case are you can add a microRNA binding site to the three prong UTR when you are getting the tissue down without this microRNA binding site and you will have a beautiful expression in CNS and peripheral tissue. If you get into tissue that express the microRNA binding site, what happens is the messenger will be CLIP and you only express in the CNS. Another engineering can be done on the cDNA the other transcriptional accessory, this can be a Kozak sequence and the transgene code optimization to generate a much potent cDNA cassette and you can improve intron to get a higher expression. 

However, I have to warn you that when you are engineering cDNA, you want to sustain strong expression, you need to balance the CPG content and the codon optimization. This is because you have two opposite facts. That is, you have balance between expression and the DNA sending as well as stability of the expression when you have a reduction of the CpG islands this stability will be enhanced.  And however, when you have completed codon optimization, you will have the DNA sensing and your stability and your expression goes up but your stability will be reduced. The balancing is the key to accomplish high level but stable expression.

That is the basics of AAV. I want to give you overview of the current landscape of rAAV gene therapy. We have done survey and found currently there are 13 different diseases and tissue targets. Among them, as you can see, CNS is the hottest target as you can see here. As the gene therapy target CNS has when we do gene therapy, we have several factors we need to be considered for gene therapy. First when you think about distribution and molecular mechanism of disease pathology, that include localized versus global pan-CNS pathology. When you think about genetics versus epigenetics, gain vs loss of function vs haploinsufficiencies. Second factor is a cellular target. This could include neuronal such as neurodevelopmental, psychiatric and neurodegenerative disease vs glial such as leukodystrophy. Blood brain barrier is the key delivery barrier and should have a different approach to overcome. First is direct CNS injection, as you can see, we have many different ways to deliver genes in CNS by direct injections. You could also use AAVs that can cross the blood brain barrier by intravascular/systemic injection. 

One thing I want to bring to your attention is haploinsufficiency because this is a common genetic cause for neurodevelopmental and psychiatric disorders. And the potential for approach for this therapy of this type of disease is first we can deliver cDNA gene augmentation. We can also repair mutant allele by DNA/RNA editing, we can also boost expression of the mutant free allele through either gene activation through CRISPRa activation or by ASO through productive transcripts and also enhancing transcript stability by ASO.

Another approach I want to indicate is readthrough therapy of nonsense mutation. As you may know, 11% of all pathogenic mutations are nonsense mutation. We could use small molecule drugs, such as G48 and ACT128, but we could also use suppressor tRNA. And the factors to be considered in this haploinsufficiency gene therapy is first need tight control of gene expression.  This include level expression as we know many cases gene replication is toxic. Second it's anatomic regions and tissue cell type specific. The third window because the developmental disease you miss the window and you may have less therapeutic effect. The second large category is promoter, self regulation and negative feedback and it includes microRNA regulated feedback. The third is sometimes mutant allele is not purely loss of function. In this case knock down mutant allele and the same time augment the normal gene.

Actually I want to introduce some concept at work by my colleague professor Dan Wang and his group. What he was trying to do is use AAV suppresser RNA for readthrough therapy. You have regular tRNA and regular stop codon but also could have premature termination codon. And suppresser tRNA only have one difference from the natural tRNA.  So the group pick up this old concept pick up for new opportunities. We are doing research for treating Rett Syndrome, FoxG1, CDKL5, Dravet Syndrome and others. The long term delivery is AAV advantage. Of course one concern is global readthrough and perturbating elongation and they are doing tRNA sequencing ribosome profiling. The reason Dan pick up this approach includes you can use one stone kill experts because a single suppresser tRNA can maybe attack different mutations and different target genes. Now you put a multiple expression cassette of a higher potency, you can see there's no exogenous protein or transgene immunity. And this is under control of the transgene regulation, no over expression. And this amenable for engineering and transient delivery.  And so –

Mustafa Sahin: Two minutes.

Guangping Gao: Yes. AAV by this method, Dan used a hurler model for concept. You can see through systemic delivery, in four weeks you can see partial restoration of IDUA and also normalization in 10 weeks of the urine GAG substrate. If you look at individual tissues, you can see you have even higher expression or so you have almost complete normalization of the GAG.  If you target the brain by systemic injection, this dose is not very effective. But if you do intracranial injection you can see you can restore partially brain activity. But that's showing the StRNA is quite effective. But whether this is safe in treating disease?

They also analyze whether this is safe by ribosomal finding. And in this case they show in patient fibroblast, G418 caused perturbation off target global rate, is true but not by StRNA. Also in mouse liver it's the same situation. If you look at also the elongation that G418 cause disturbance of elongation but not by this StRNA construct, in vitro patient fibroblast and in mouse liver. 

Thank you very much. I want to end here. I want to thank my colleagues. My group is a big group directed by four PIs, includes Dr. Xia and Wang and Tai. I want to thank our vector core.  Thank you very much for the opportunity to present. Our next speaker is Dr. Sahin from Boston Children's Hospital. He will give an overview of a rare genetic disease and neurodevelopmental disorder.  Thank you.

Mustafa Sahin: Thank you. My task is to give you an overview of rare genetic variants associated with neurodevelopmental disorders. I will start by acknowledging my disclosures because of my role at the translational neuroscience center in their hospital I work with a number of pharmaceutical and biotech companies. 

Let me start by defining rare diseases. In the United States rare diseases are defined as conditions affecting fewer than 200,000 individuals at a given time. Taken together an estimated 300 million people are affected with rare diseases. They used to be referred to as orphan diseases. Due to the orphan drug act and better understanding of the biological mechanisms, many companies have been interested in these rare diseases over the last few decades.

Here is a visual representation of the number of rare diseases increasing over the years as we identify genetic causes. As you can see, in 2020 now we're over 6700 rare genetic disorders identified and only about 500 of them have any type of therapy. Importantly, for the purpose of this workshop, roughly 75% of rare genetic diseases affect the central nervous system.

There have been a number of successes in this field, as mentioned earlier. One of the major success stories is the FDA approval of drugs for spinal muscular atrophy. This is an announcement from the FDA around Christmas of 2016 when the first drugs was approved.  Since then two other drugs have been approved and they have completely the changed the natural history of this disorder in child neurology. We’re going to hear much more about that from Jerry Mendell later today.

While these success stories are occurring, there's a huge unmet need in the field, especially in mental health disorders. Here's another FDA announcement. This time warning about potentially dangerous therapies and products in the field of autism. And this is due to the dire need of families dealing with this disorder. And since there are a limited number of scientifically proven therapies in this field, they are reaching out to therapies that may not have scientific basis and that might be potentially dangerous. So how do we start to address this huge unmet need?

One aspect that has come to the forefront is the role of genetics. Here are the heritability estimates for neuropsychiatric disorders, especially for disorders such as autism spectrum Disorder, the heritability is remarkably high. Here's a slide from NHGRI workshop back in 2009. This summarizes the role of rare and common genetic variants in the field of complex disorders.  For the purposes of neuropsychiatric disorders we really fall into the diagonal for the most part.  There's a role in the population for common variants, but these common variants have a small effect size per individual. On the other side we have rare alleles, what we would call Mendelian diseases where a variant of this sort, while rare, would have a high effect size. And for the most part in the field of autism spectrum disorder, the main progress has been made in identification of rare alleles associated with autism spectrum disorder. 

Here's the progress that's been made over the last several decades from a slide from Tom Bergeron's paper. This has happened thanks to advances in the technology we're using, we're moving from karyotype to chromosome microarray to identify deletions and replications to finally exome and genome sequencing that allowed us to identify sequence variants. 

And the variants that are associated with autism in terms of the type of variant really take place in the whole spectrum. There are structural variants with a large chromosome deletions or duplications that can encompass several genes. In several cases, either deletion or duplication of that chromosomal region can be associated with neuropsychiatric symptoms. Importantly, in these disorders which encompass several genes, often it is the case that it is hard to identify a single gene that's predominantly responsible for the phenotype. It might be a combination of oligogenic effects. There's also trinucleotide groupings, missense mutations, nonsense mutations, indel splice site and possibly and likely intronic and intergenic variants as well.

There was an NIMH organized workshop in 2017 that discussed the role of rare genetic disorders in neuropsychiatry and I won't have time to go into the details of this discussions but I urge you to take a look at this paper led by Stephan Sanders for some of the highlights of that discussion. What I'm going to do today is take a much more simplistic approach and focus on the heterogeneity of rare genetic variants associated with Autism Spectrum Disorder. In the field the estimates range from 400 to a thousand susceptibility genes. How do we approach treatments for this heterogenous condition?

There might be a couple of scenarios. On one side one can think of a single broad spectrum treatment that works for all types of genetic variants. But as we start to understand the genetic code of autism, it is becoming clear that certain genetic causes result in diametrically opposite changes in synaptic function, such that a treatment that works for one form of genetic cause of autism may actually worsen other genetic cause of autism. On the other side we may have to think of a scenario where we have to develop a treatment for every gene or every variant. This is very much the topic of our discussion that's going to take part today. I should highlight that several of us in the field believe that there might be subcategories of autism that share convergence either at the cellular, circuit level or maybe at the level of platforms used for interventions. And learning from one single variant or one single gene may have implications for related disorders.

So with this landscape at our institution and many other institutions around the world we have been taking this approach to try to develop new therapeutics for autism spectrum disorder, starting with our genetically defined autism patients at specialty clinics, we try to do phenotypic characterization of individuals affected if there is neurodevelopmental disorders as well as biomarker identification. In many cases we develop cellular and animal models to test for both efficacy and toxicity.  And we finally move into clinical treatment trials. What I'm going to do in the next portion of the talk is to give examples from some genetic disorders that we have tried to approach from a mechanistic perspective and tried to address some of the issues that have come up as potential discussion points for the rest of the workshop.

The disorder I'm going to focus on is tuberous sclerosis which I have been working on for a number of years. There are a number of reasons we chose this disorder to focus on, roughly half of the patients with TSC are affected with autism spectrum disorder. But importantly, any of the specialty patients will be diagnosed very early in life, and sometimes even before birth. The cellular mechanisms of TSC have been studied from a number of different perspective from a large number of labs around the world. And importantly there are FDA approved specific inhibitors of this pathway that can be repurposed to test hypotheses in clinical trials. The combination of these factors makes tuberous sclerosis a good model to study in neurodevelopmental disorder space.

This is a simplistic point of view of what TSC1 and TSC2 proteins that are the causal genes in this disorder do in the cell. These proteins, encoded by TSC1 and TSC2, form a complex. When this complex is functional, it puts a brake on a protein called mTOR which stands for Mechanistic Target of Rapamycin, which controls protein synthesis and cell growth.  Importantly, if you missing TSC1 or 2, you have up regulation of this pathway resulting in this disease. A corollary of the understanding of the genesis is if you over express TSC1 or TSC2 alone, you actually don’t see much of an effect. The complex is such that you have to over express both of these genes to actually see a bio chemical effect. In terms of dosage there is a safety factor here that's inherent to the biology of the disorder. 

One of the most important aspects of this interaction between TSC genes and the mTOR pathway has been the recognition that we have naturally occurring inhibitors of the mTOR pathway that have already been in the clinic for various indications. These are molecules such as rapamycin that we've known for several decades. I'll give you examples of how rapamycin and similar molecules such as sirolimus have been used in the clinic and clinical trials to address symptoms associated with tuberous sclerosis. I'm going to talk about three prelaminated symptoms of tuberous sclerosis, one is benign astrocytoma, the second one is epilepsy, and the third is autism and related neurodevelopmental deficits. First with astrocytomas, here's a first study from David Franz, and colleagues at Children's Hospital, when he took a handful of patients with these benign astrocytomas shown here and treated with rapamycin, what he showed is when the patient is taking the drug, the tumor shrinks, if the patient is off the drug, the tumor comes back, when you put the patient back on the drug, the tumor shrinks again. This was a very strong effect, easily quantifiable. And this led to a Phase 2 trial of 28 patients led by Darcy Krueger and the effect was reconfirmed in this study and published in the New England journal that everolimus was effective towards this type of tumors in tuberous sclerosis. This led to approval of this drug a month later by FDA. 

The next aspect I want to talk about is epilepsy which is seen in about 90% of the patients with tuberous sclerosis. In this paper by Elizabeth Thiele and colleagues at MGH, they show that most of the diseases with tuberous sclerosis start early in life and it's thought these seizures have an important impact on neurocognition and development of the child. There have been a number of mouse models developed in various labs around the world. Here's an example from Michael Wang's lab at Washington University. When the animals are developed these knockout mice have seizures starting early in life. If you treat the mice with rapamycin early in life, you can stop the development of seizures completely. And you can also increase survival markedly. These are the vehicle treated knockout mice, dying by about 16 weeks, and these are the mice treated by  rapamycin. Importantly, almost all the mouse of tuberous sclerosis developed in different labs show this very strong effect. 

We have been working with iPSC-derived human in the lab. We see a similar hyperexcitability phenotype that could be associated with epilepsy. These black lines are showing firing of human neurons in culture. This is a wild type, this is a heterozygous TSC2, and this is a homozygous deletion of TSC2. As you can see, the cells that are missing TSC2 are firing much more than the other two genotypes, and importantly you can treat these cells in culture with rapamycin, you can significantly reduce the hyperexcitability phenotype. These type of evidence from mouse and cell culture led to the hypothesis that you could use rapamycin type drugs in the caudate to reduce epilepsy in patients affected with tuberous sclerosis. 

Here's a Phase 3 trial that took place with individuals between the ages of 2 and 65 with refractory seizures. This enrolled 345 patients and the primary outcome measure was the frequency and severity of seizures in the placebo group versus the everolimus groups. I'm showing you the results from this study. The response was defined as greater than 50% of reduction in the number of seizures from baseline. There was a 15% response rate in the placebo group, 30% response rate in the low everolimus group and a 40% response rate in the high everolimus group. This was not 100% response rate but this was sufficiently successful to lead to approval of this drug by the FDA in 2018. 

The third aspect is the neurodevelopmental and neuropsychiatric symptoms associated with tuberous sclerosis. We took the same approach and used Everolimus in a Phase 2 trial in patients with tuberous sclerosis between the ages of 6 and 21. This was a randomized placebo controlled double blind study with two sites: 47 patients in all. We did neurocognitive testing at baseline, three months and six months and also looked at autism symptoms. We published it a few years ago, and the basic finding was that there was no significant signal in the arm tree of the everolimus compared to the arm tree of the  placebo. There might be a number of reasons why we may not see an effect. Outcome measures is obviously a significant concern in this field.  Another concern is the age of treatment. By the time a patient is 10, 15, 20 years old with these symptoms is a short treatment of six months sufficient to see the effect. We have turned to the preclinical model models and Peter Tsai in my lab had developed this mouse model of tuberous sclerosis by deleting the TSC gene in the cerebellum and these mice have both social interactive deficits and repetitive behaviors and cognitive inflexibility. If you treat these mice at one week of age with rapamycin, you can rescue all of these deficits. If you wait six weeks to start treatment, you can only rescue the social interactive deficits but no longer can rescue the grooming and cognitive inflexibility. If you wait 10 weeks to start treatment, you cannot rescue any of these deficits. These types of sensitive periods have been now demonstrated in a number of animal models of neurodevelopmental disorders.

As mentioned earlier, PHC patients can be diagnosed very early in life with these cardiac tumors. We've actually formed a consortium to look at any biomarkers that might be associated with development of autism and epilepsy in babies born with tuberous sclerosis, this is a five site consortia that's been funded by the NIH. And I'll show you the findings. We did prospective  EEG of babies born with tuberous sclerosis and we looked at the onset of abnormalities on EEG and onset of clinical seizures. To summarize this work, the onset of epileptic discharges on the EEG was around 4 months. The onset of clinical seizures were around 6 and a half to seven months. There's a roughly two to three month period where the EEG is abnormal and the clinical seizures have not started. 

Based on that, we have initiated a prevention trial. It's called Preventing Epilepsy Using Vigabatrin In Infants with Tuberous Scherosis Complex (PREVeNT trial). Vigabatrin is an antiseizure medication that works on the GABA pathway and we're looking at the developmental impact of early versus delayed treatment and the results should be obtained within about 12 months. 

Now I want to briefly mention that TSC is not the only disorder where mechanistic have been tested. Fragile X is another cause of intellectual disability in autism. The metabotropic glutamate receptor (mGluR) theory puts inhibition of metabotropic glutamate receptors should reduce these symptoms and in fact this works very well in animal models. There's been a couple of studies in the clinic and they have not shown a significant efficacy. This important paper highlight that the clinical trials in the future should consider initiating treatment in a younger population if longer treatment duration, longer placebo run ins and identifying new markers to better assess behavioral and cognitive benefits.

So I've talked about fragile X and tuberous sclerosis. There's another large number of genes that are associated with neurodevelopmental disorders. How can we start to address those?

One of the potentials is identifying these convergent mechanisms and one mechanism we focused on is the mTOR pathway activation. We formed a consortium to study three gene related disorders in the clinic. With funding from the rare diseases clinical research network, funded by these four NIH Institutes. This is a multi center study where we’re performing detailed neurotological and behavioral assessment of these patients longitudinally in the natural history setting. We also have a pilot project looking at the effect of [away from mic] in Phelan-McDermid Syndrome patients. 

I want to summarize what I talked about. Pre clinical studies are leading to clinical trials in rare genetic variants associated with neurodevelopmental disorders. Trials are currently taking place looking at symptoms. Prevention trials are being launched and I think are going to play an important role going forward. And biomarkers improve outcome measures and in depth natural history studies can accelerate these trials. I want to highlight that timing will be critical. The distribution of the therapy will be crucial, especially for cell autonomous treatments and the dosage of gene expression may be essentially for these disorders in which the range is narrow. I want to acknowledge in addition to the many of the collaborators we have in this field and acknowledge our funding sources. Thank you very much.

Geetha Senthil: Thank you, Mustafa. Do you want to introduce the next speaker?

Mustafa Sahin: Thank you. Our next speaker is Dr. PJ Brooks and he's going to give DNA targeted therapeutic platforms for rare genetic diseases.

PJ Brooks: Great. Thanks. So thank you for the opportunity to talk to you today about the different NIH programs. Very much looking forward to doing that. I'll actually go through three of them. Platform vector gene therapies, bespoke gene therapy consortium, and the NIH common fund somatic cell genome editing consortium. Then I'll hit a couple key questions I think that are relevant for further discussion about genetic therapy development.

This is kind of the slide that most of have already showed you. We're developing and identifying many disorders with a known molecular basis but the rate at which we're developing therapies is far too slow, particularly when we're talking about disease where we know the molecular basis.  We ought to be able to do better. One way we think about this is to stop thinking about one disease at a time and thinking about platform approaches to multiple diseases and certainly genetic therapies are a great example of that.

So thinking first about AAV, which we heard about from Guangping Gao, there's really good news. There's excellent safety record, recent clinical success stories, two approved products in the U.S., and really a lot of pre clinical success stories. We're really great at treating and curing diseases in mice and animal models. The real problem is getting to gene therapy clinical trials, which again is typically done one disease at a time, which is slow and inefficient and doesn't really take into account the potential for learning and the platform capacity of AAV. And also to the extent there's a commercial aspect to this, there's an obvious bias towards the most common rare diseases. But it kind of makes sense. One of the things we're thinking about is if we design clinical trials for multiple diseases at a time, really taking advantage of the platform capacity of AAV vectors, we can increase the efficiency and reduce the time of clinical trial startup.

At a very rough level, this is kind of the representation of AAV. There's a Capsid with a viral DNA replaced by the therapeutic human DNA and depending on the therapeutic human DNA you put in there, you can generate therapies for different diseases simply by swapping it out. You can almost think of AAV as a delivery vehicle like a delivery box for example, in fact one that sort of pre addressed if you will to certain specific organs and cell types. And given this obvious platform capacity, you do wonder if we do multiple diseases, are there not steps that could be taken to increase the efficiency. Or to say it another way, do you really have to repeat all of the steps in the pre clinical development every time, even if you're just swapping out one therapeutic gene?

So the goal of PaVe-GT is to put that question to the test. We see this as an experimental pilot product. It's really a translational science experiment. And the idea is, we think of it as a sort of public AAV gene therapy development approach, where we’re going to be going forward for clinical trials for rare diseases, four different rare diseases under study by intramural investigators at the NIH clinical center. These diseases, at least at the time we began, were of no known commercial interest. And all four diseases we'll be using the same viral vector, in this case AAV9, the same route of administration. Same production and purification methods.  As they say at the FDA, the process is the product. And we are going to keep the whole process the same and simply change out the therapeutic constructs. 

And then I think the different thing about this as we go forward through the process and to the FDA, we will be making all of that information public, including the methods, the protocols, all the regulating documents, including our discussion through the FDA, and ultimately improved INDs and make them available to the website so other people can use them and hopefully use them in a cut and paste approach. 

So as another illustration of this, at the top you can think of what you might consider to be the null hypothesis which is we have got four different diseases here. On the left we have got two organic acidemias, PCCA deficiency and MMAB deficiency to on the right to neuromuscular diseases, DOK7. And the null hypothesis is that for each one of these, we have to do everything in parallel, all of these steps, proof of concept, CMC, biodistribution, toxicology, etc. But the question we’re are going to ask in PaVe-GT is since we are using the same platform, maybe we don't have to do all these things. Maybe we can reduce the number of biodistribution studies or toxicology studies and also find ways to increase the efficiency and streamline the CMC process.  These are the kinds of questions we'll be asking as we go along to the FDA. We'll get the answers back and make those answers publicly available to make the process transparent and hopefully help everybody understand how the clinical trials are developed and make them more efficient with the goal of benefiting all stakeholders with the particular focus on those focused on diseases of no commercial interest.

So the project is ongoing. This is the team. It's a great team effort at the NIH. And everybody involved is at the NIH, I particularly wanted to highlight my colleagues Donald Lo and Elizabeth Ottinger, and Anne Pariser, the investigators involved, Chuck Venditti and Carsten Bonnemann from NINDS. There's the websites you can follow on the progress and we also have a recent publication going into PaVe-GT in more detail.

Moving on to the next project, which is the bespoke gene therapy consortium, which is a public/private partnership, again focused on making AAV gene therapy a reality for genetic diseases affecting populations too small to be viable from the current commercial perspective. This is a public/private partnership organized by the foundation for NIH, which is not part of the NIH, by the way, not part of the Federal Government. The FNIH, the FDA Center for biologics, NCATS taken kind of a leadership role but also involving many other NIH colleagues as I'll highlight later.

So the BGTC really has two different components to it. One component is focused on the basic biology of AAV as it relates to translation and the production of recombinant AAV vectors for gene therapy. So optimizing and better understanding of the basic biology of how we make vectors in manufacturing facilities and then also can we enhance therapeutic gene expression once these vectors go into patients.

And then the larger component of the BGT is actually supporting some clinical trials. And in the process, streamlining the manufacturing and analytics for the vectors involved in the trials and the clinical development process.

So in terms of the clinical component, we envision here a pilot project involving perhaps five to six diseases still to be determined what those diseases are. And by doing this in a concerted way, to streamline the process for going through the idea of gene therapy all the way out to the clinical trial and standardizing things like the vectors available, the process in toxicology and testing and limited number of delivery methods and doing this in a consortium so all the things we learn from the process are reported back to the consortium to support iterative learning. And ultimately all of the work that comes out of this, we plan to put into the public domain as well so it can benefit the whole community and all stakeholders.

Where we are here, this has been a fairly long development process beginning with the concept evaluation, well over about nine months ago, initial concept was approved. We're now in the process of getting and finalizing the support from different stakeholders, including the government, private companies and nonprofits and are working on the finalized detailed research plan and again, want to thank my colleagues at NIH, Patina from NIH and Chris working with this on this effort along with the FDA Center for Biologics, private sector partners and others, it is very much a collaborative effort. Once we get all the funding agreements in place, we anticipate launching the program hopefully in the first quarter of 2021, like many other efforts, this one has also been slowed early on by the COVID pandemic.

Then the final program I want to talk about is the NIH somatic cell genome editing consortium (SCGE). This is an effort supported by the NIH common fund to the Office of the NIH director and involves participation, again, many different NIH Institutes and centers. NCATS is taking a lead role but it's very much a collaborative effort as well. The focus of the SCGE is to lower the barriers for new genome editing therapies by a variety of initiatives including better animal models and animal systems to test genome editing reagents and delivery systems. These are animals that have reported genes in them so we'll be able to assess editing in all cells and tissues in these animals, including rodents, pigs and non human primates. Another component is focused on assessing unintended biological effects of genome editors. Not just looking at off target sequencing or off target genetic effects but going more forward and asking what are the biological consequences of some of the editing effects. And all these studies are carried out in human cell systems with a goal of ultimately reducing some of the animal usage in the regulatory process. That's an aspirational goal of the SCGE long term. 

There's also a component involved looking at better ways to be able to monitor genome edited cells in vivo. The biggest component of the program is focused on delivery of genome editing machinery, given that's one of the big needs in the field. We have a smaller component on expanding the human genome editing repertoire, including developing of new editors. We didn't want to put a lot of effort or funding into that because it's a very hot area to begin with, but we felt it was important to have some of that in our consortium. We'll also have a coordinating center, a dissemination coordinator center which will be primarily responsible for generating a toolkit that will be publicly available and allow interested stakeholders to take a look at what has been produced by the SCGE and see the data upon which the results are based.

And this slide kind of shows where we see the SCGE in terms of the developmental process of new therapies. We are not supporting any direct clinical trials with the SCGE. That's beyond the scope. But the way we see it is some of the tools we're developing would be filling gaps in the process of getting do an IND. So you might imagine that for certain diseases there would be a need for the ability to deliver a genome editor to some specific cell type which doesn't exist at present and through the efforts of the SCGE investigators, we could provide some of the delivery methods that could then be used by patient advocate groups or small business entities, et cetera, to develop an IND. That's the way we see it is being an IND enabling effort. 

We would like to think as an aspirational goal that some of the biological systems we're developing here might fill some gaps in the regulatory process and streamline that as well, getting around the need for testing in animals, at least where that would be possible. Because that's obviously a great cost and expensive aspect of the regulatory pathway. 

So this is the different give you a sense of different components. You can see on the right when people say it's all about delivery, we took that very seriously.  We have a total of 20 different grants focused on different approaches to deliver genome editors including modified AAVs.  And one of those investigators is Guagnping Gao. We have got several working on nanoparticles, modified viruses, basically synthetic viral particles, and also one on adenovirus. Quite a diverse group of awards. I believe six or seven of them are focused on delivery to the nervous system, at least to some degree. And below is the website so you can see more about the program.

Listening to Mustafa talk about the idea of platforms that would be appropriate for use in multiple diseases and also Guangping talking about that with the delivery of modified tRNAs, I think one of the things that really gets me most excited about genome editing is the idea that you could have ultimately a single biologic that would be applicable to multiple diseases and that would be a single editor. And one of particular interest was developed in part in support by the SCGE, the prime editor developed by David Liu's lab.  You can see based on the way it's designed, it is anticipated that this single editor could in principle at least be applicable to almost 90% of all genetic diseases. And if this works as anticipated and ultimately gets into clinical trials, you could imagine then that applying this to different diseases would simply be a matter of changing the sequence of the guide RNA to direct it to different locations within the genome. I think these kind of ideas have really huge implications for getting past this one disease at a time approach and getting to a lot of diseases in an efficient manner.

Then I thought at the end to kind of, because we at NCATS think a lot about generalized approaches rather than specific diseases, we get a lot of questions, people come to us saying they're thinking about developing a gene therapy for the disease. I find myself always thinking about the same questions as I have those conversations. I thought I would raise some of them here for the sake of discussion. Assuming one is interested in genetic therapy and you can develop a genetic therapy for a disease, some of the key questions you ultimately would have to answer or at least address are which cell type or cell types and organs and cell types are you going to need to target to have a therapeutic effect. And amongst those cell types, how many of those cells do you need to correct to have a therapeutic effect? Do you know the answer to these questions?

I think for many cases it's going to be very important to know the answers to the questions.  Because the next one is assuming you know those cells, the question would be do you have a way to deliver your genetic therapy to those cells. And of course that depends very much on what the therapy is. We've heard about AAV has a tremendous capacity to develop therapeutic genes to certain cell types but as yet we don't have one that can go to basically all cell types in an organ and in some cases they're very much cell type specific.

Whereas some other approaches such as antisense oligonucleotides could potentially go to a much greater number of cell types in the brain. If you have got drugs like that that can get to essentially all the different cell types in an organ, you may in fact not need to know the cell types you have to target because you'll be targeting them anyway. Then of course there are constraints based on the time of the course of the disease. As was mentioned, this can be a particular issue in the neurodevelopmental disorders and also neurodegenerative disorders, where if you're looking at delivering therapies to a cell population that's dying off, you have to consider how many cells are still there by the time you develop the therapy.

Geetha Senthil: Two minutes.

PJ Brooks: Yep, got it. Finally, is clinical trial endpoints. If you're going to do a clinical trial you have to have something to measure that's acceptable to the FDA and optimally this would be based on Natural History data such as may come from Natural History studies like the one that Mustafa mentioned as well. These are some general thoughts that might come up later in the meeting.  Just want to thank my other colleagues at the Office of Rare diseases research. We have got a lot of other programs I didn't have time to talk about. I'll stop there. Thank you.

Guangping Gao: Thank you, PJ. Our next speaker, it's our fearless leader in the field, president elect of ASGCT Beverly Davidson. Please, Beverly.

Beverly Davidson: Can you see my slides?

Guangping Gao: Yes.

Beverly Davidson: Okay. Fantastic. First of all, I also want to thank NIMH and Guangping and Mustafa and Sitra [video cut out] ... There was enough presented that we thought we would probably span five days and to get it into just a couple days is really impressive.

I'm going to bring up something that I think the field has been looking for, for some time. This was also brought up in the chatbox after Guangping's presentation and that is how do we think about really refining control of gene expression for a number of applications?

In my laboratory, these are my disclosures, in my laboratory, we work on repeat expansion disease. This is just a cartoon showing you an example of where many of these repeats can occur within the genome. And my lab focuses on spinocerebellar ataxia … [video cut out] Huntington's disease, due to a polyglutamine repeat expansion in exon 1, the HTT gene. If we think about targeting dominant alleles, you've heard both from PJ and from Guangping that there's a number of approaches for these disorders and folks are looking at going after the protein using various approaches. Of course, there are antisense nucleotide approaches to reduce the RNA that's encoding the mutant Huntington, those are in the clinic and moving along in clinical trials and very exciting. 

There are also efforts to induce splicing changes and also AAV delivery of RNA interference expressing vectors for therapy for Huntington's. Some of these can be allele specific. We've also heard about approaches targeting reduction of both mutant allele as well as the wild type allele. Of course we’ve also heard about approaches to target DNA and that was brought up in one of PJ's last slides, and that was to look at editing machinery. That's really where I want to focus on here. And the work I want to present today was largely spear headed by Alex Mas Monteys in the lab, who is an assistant professor at the University of Pennsylvania. 

For Huntington's disease we have to think critically about what part of the gene to target. Our RNA interference approaches and ASO-based approaches are really targeted to any part of the gene. And the impact is a reduction of both alleles to some degree but not … [video cut out]  a little bit more careful because we are essentially going to reduce the expression of the gene in that cell to nothing. There's some recent papers that have come out of Joe Bates and David Hausman's labs, shows if you do target downstream with exon one, you’re still going to get aberrant splicing of exon one, which is where that CAG expansion occurs. So you may not be doing any good by editing downstream of exon one. More recently also Laura Ranum's group has shown that there may be transcription off the CAG repeat, both strands, with subsequent translation of those repeat consequences leading to additional forms of toxicity. Our idea was let's get rid of exon one and do it in an allele specific way.

… [video cut out] polymorphisms that insist with the mutation on the mutant allele. For a fully editable Cas9 approach we would take advantage of a snip present on the mutant allele along with a common neutronic slip to allow for a deletion of exon one and the approach would essentially leave the non mutated allele intact. When we tested this initially in cells that express the snips on both alleles, you could see that we could effectively get very robust knock down looking at Huntington expression here at the RNA level and here at the protein level. And you can see that these snips effectively removed exon one, the most effective … [video cut out] we next moved into Huntington cells.  We worked in both cells that had a PAM on the normal allele as well as those with PAMs on the mutant allele. In this data I'll show you a little bit of work from this. You can see that we have knock down with the guide sequences targeting upstream of exon one with common neutronic sequence, we get about 50% reduction of exon one. Can you all still see my slides?

[Male speaker]: Yes, we can see the slides.  We just turned off the video to improve the sound.

Beverly Davidson: Okay. Thank you. You can see we can reduce the expression of the mutant Huntington. This is a Western blot of a relatively small compared to what we're used to in the mouse tissues. You can see a relative reduction of the protein levels here that correspond again to about the same levels of knock down at the RNA level. 

We next moved this into in vivo using a BACHD mouse model that expresses the full length human Huntington's transgene and also has this common snip in the 5'UTR that allowed us to test for the effectiveness of this approach. We developed AAVs that expressed Cas9 and the guides and introduced those into one side of the rodent brain. When we looked BIO-PCR assay for cleavage between the injected and not injected side, you can see robust editing and the tail here was a control, this was a direct injection of parenchyma with no systemic deliver. … [sound cut out]

And this is all well and good. We can reduce the levels of Huntington and leaves the normal allele intact. We need to think of moving this forward and improving the safety of this. The nuclease are foreign entities and elicit immune responses in cells so how can we make this safer.  Guangping mentioned various ways for AAV targeting so that you’re only going to deliver the vector to the right cell, you can use microRNA targeting sequences in the three prime UTR. The other ways to improve the safety for AAV delivery is to regulate the expression for a short burst that's sufficient for editing. I think this is going to be critical as we move these foreign entities into cells.

So I'm going to side step and talk about how we reintroduce for you the gene that when mutated causes spinal muscular atrophy. You'll hear very beautiful talk later from Jerry Mendell's group.  Just to remind you, in healthy individuals we have a normal SMN1 gene and various copies of SMN2 pseudogene. And this SMN2 pseudogene, sometimes exon 7 is skipped and sometimes exon 7 is included. The disease severity of SMN2 has to do with how many copies of this pseudogene you have and how much functioning full length SMN2 can complement the mutated SMN1 in SMA patients.  And fortunately there are two drugs that can induce or enhance that exon skipping. One is … [sound cut out] the other is approved for use in Europe. What both of these small molecules do is essentially promote exon 7 inclusion. We reasoned this is great for SMN2 but can we take advantage of the splicing phenomena and these orally bioavailable molecules for regulation of gene expression. The idea is really quite simple. You take this cassette from SMN2E6, E7, E8, and infuse gene of interest, in general you get very low expression of gene of interest and with a drug you would get protein synthesis. So we took this and engineered a report of vectors initially and then we also altered the exon splicing junctions here. We could make it constitutively active or … [sound cut out] shown by the splicing assay here. Here’s the wild type construct and you can see it very much mirrors what we see in patients, about a 10% exon 7 spliced in. If we modulate these sequences here we can get it down to less than 1% and this is the constitutively active.

Does this work for gene expression? Indeed it is. This is an example of luciferase. When we use this system with a drug, we get about a 10 fold induction of gene expression, which may be fine for some cases, but possibly not enough for others. With that low level of induction we decided to go back and treat cells with very low dose drug, a couple of orders of magnitude lower, and screen … [sound cut out] responded low dose drug. This was work done by Paul Ranum, post doc in the lab together with Alex Mas Monteys. And when they looked at cells that were treated at low dosage or human cells, this is an example of one of the genes in which in the presence of drug you get this novel splicing event. We evaluated these five candidates here for their inducibility in the system that I just showed. So we tested all these non SMN2 mini gene candidates.  In the absence of drug, they're essentially off. In the presence of drug, now you're seeing we're getting roughly with the first gene candidate a 200-fold induction. This is much more sensitive than our previous SMN2 mini gene cassette. This gives you an idea of how on and off it is just using GFP as a reporter. These are tissue culture cells either treated with DMSO or the drug. In this case it's LMI070 this system also works with [away from mic] using other cassettes. You can see expression control.

We also tested how responsive this was with regards to promoters. This is just an example where we used RSV promoter, PGK promoter and minimal mCMV, which is a weak promoter because it's stripped of enhancer elements. Just looking at the levels of induction we could achieve with this, you can see we can control by low dose drug or by using the different promoters to get the level of gene expression that we think we need. And importantly this drug, again I told you is orally bioavailable and brain penetrable and the beauty of this is there's a relatively rapid wash out, so you could dose an animal once or maybe once or twice a week, or you could dose it at low dose more often, depending what levels of expression you were trying to achieve. This just shows you the dose responsiveness of the splicing here using LMI070 as an example, and this is just the folded induction which we were seeing. So we can have relatively, I'm sorry, this is the actual exon inclusion ratio. This we've seen. 

What about transgene expression in vivo? We generated a series of AAVs for liver targeting or brain targeting initially using some very weak promoters. … [sound cut out] intervene to mice. You can see that we have a robust expression after a single dose of drug. When we do Western block for GFP you can see only in the presence of the drug do we have GFP expression. This is showing you a very long exposure of that western so you can see this is pretty real and pretty tight. And similarly if we move that cassette into PHBeB, again using a relatively weak promoter, in the absence or when we treat the animals with vehicle, there's essentially no expression, but when we give low dose drug, you can begin to start to see expression showing the cortex and the hippocampus as an example. This is a single higher dose where you can see again induction of expression.

So I started out by, and this is just giving you an [sound cut out]I think in the hippocampus here. This is looking at the CT values and the fold induction of expression. At the low dose it's about 180 fold and at the high dose it’s 1500 fold induction. What about using the system for editing control? Instead of having GFP or luciferase here we engineered in Cas9. The idea would be to essentially use this as a way to make or improve the utility of Cas9 and certainly the safety of Cas9 as an editor. Again, we have very tight expression with, so this is an example of constitutive expression of SaCas9 as a vehicle treatment, and you can see increasing doses of drug, we have increasing levels of Cas9 expression. And this is particularly important if you [sound cut out] efficient than the SpCas9.

Does this work using that same approach I opened up with using snip base targeting for deleting the mutant Huntington allele. Yes, indeed it does. This is the constitutive expression in the absence of a regulated cassette and these are the cells that were treated with the SaCas9 exon cassette. So a control guide, there's no silencing, presence of drug with a control guide, no silencing, absence of drug with a cassette, no silencing, and about 50% silencing of that mutant allele in the presence of drug and in the proper control of the guides.

[Male speaker]: Two more minutes.

Beverly Davidson: Okay. The challenges moving forward, I think regulation … [sound cut out] where we have very fine control of gene expression that can be controlled with a brain penetrable orally bioavailable drug and it's a very exciting tool for us to move forward because it's been in people, it's been in children. For AAVs the Capsid, you heard beautiful work from Guangping's lab and my lab also has ongoing engineering efforts in NHPs using different methodologies to evaluate tropism for cells that have been somewhat refractory for AAV transduction.

And just to summarize as this relates to HD, I think we can achieve allele specific editing. We have advances in regulating these knock down approaches particularly with something immunogenic like the bacterial nucleases, we can use drugs to regulate them or disease cells … [sound cut out] tools to assess the footprint that is required for transduction from these novel vectors. I would like to acknowledge my lab. It's a relatively big group. Former members of my lab that have really also contributed to all the science I presented today and are funding NINDS and foundations. With that, I thank you.

Geetha Senthil: Thank you for a fascinating talk. We'll go on a 10 minute break. Please stay on the line and Zoom, mute yourself, and we'll join back at 12:41. See you all.

[after break]

Mustafa Sahin: Should we get going? Welcome back everybody. Our next speaker is Dr. Mendell, neurologist at Nationwide Children's Hospital and he's going to talk about spinal muscular atrophy, from benchtop to bedside and the market.


Jerry Mendell: SMA is one of the stories where everything came together: The vector, the transgene, the adverse events and here's the story that I'll tell about that if I can get… So these are my disclosures .I'll spend little time on that. These are the topics for discussion. I also won't list each one of them, but I followed Guangping's directions in terms of talking about the history and some of the features of SMA leading to clinical approval. 

So I like this slide because it shows just how long SMA has been in the clinical picture. Werdnig and Hoffman described the disease in the 90s and then really stayed. I trained somewhere around the 60s and we still called it Werdnig-Hoffman disease. We had recognized at the time that there was a milder form called Kugelberg-Welander, but it was not until the 90s that we really classified the disease in terms of three different entities, one where the infants never learn to sit as Type I; Type II were patients who could sit and not stand; and then finally, the Type III patients who could walk. The real essence of the disease that came out was an effort to link the SMN survival motor neuron gene on chromosome 5Q13. A lot of this was Judith Milke's work and identifying there was a copy indication, a paralog to SMN1 which was the main disease that expressed the disease. And our clinical trials took place and were reported in 2017 in the New England journal that led to approval of this. This was really the first systemic AAV gene therapy that was approved for clinical use. The molecular basis of SMA is really fascinating because this is the normal appearance. We have SMN1, which is telomeric and SMN2 which is centromeric, this is paralog and this is duplicated and has a deletion that in exon 7 it only allows for minimal gene expression and in SMA, the disease SMA now, we have loss of SMN1 expression that again is mainly due to an exon 7 mutation but can also be due to a stop codon. And the disease can be modified then by how many copies of SMN2 there are and we'll see how SMN2 copies lead to different clinical diseases, clinical types of the disease that I mentioned. 

This is a very large study from Spain which I think illustrates the points very well. Type I patients who never sit and have onset of the disease in the first six months have two copies, typically, of SMN2. They can have more, but this is the typical appearance. Then for the milder patients who can sit but not stand and have onset of disease typically after 6 months, somewhere between 7 and 18 months, they typically have three copies of SMN2 and then we have patients who have milder disease who can have four copies and even more, and these are the patients who stand and walk. This is a typical appearance of an SMA type I child, which this presentation focuses on. Onset before six months, never sit, die by two years, hypotonic, have difficulty swallowing, tongue fasciculations, have trouble breathing, and die by two years of age. This is an elegant study run by Richard Finkel and it involved the pediatric neuromuscular clinical network for SMA showing the Natural History of the disease. And the big point that only 8% survive by 20 months of age.  By 13 months only 25% survival. And by 10 months there was a 50% survival.

So the other thing that came clear at the time, and Guangping mentioned this was, were the labs that showed that AAV9 could cross into the central nervous system, cross the blood brain barrier and target motor neurons, they targeted other cells as well, astrocytes but they were done particularly by the Barkats lab and also at Nationwide Children's by Kevin Foust working in Brian Casbar’s lab and this really set the stage for us to move ahead with SMA gene therapy. It was a ubiquitous promoter and a CMV enhancer. The full SMN gene or cDNA. And the rescue of the delta 7 mouse which was also part of the discovery at the Ohio State University and Nationwide Children's as part of that complex, Arthur Burgess described the delta 7 mouse. And the mouse could be rescued if it was treated early, and this has great implications for the clinical trial. So complete rescue if the self complementary AAV9 is delivered by day two  And virtually no rescue if it's delivered by day 10. Obviously, clinical implications. 

It was about that time that I went to the CEO of Nationwide Children's Hospital and said if we are going to get seriously into the business of AAV gene therapy we need to build our own vector manufacturing facility. We had done several trials before that but they were mostly intramuscular. Now we needed a lot of virus and needed our own vector manufacturing facility.  We couldn't understand afford what was available in the commercial market and developed the clinical protocol. The clinical protocol was what I'll show you here. We get questions about this.  We deliberately made inclusion criteria up to nine months even though we targeted SMA1 with onset at six months. That was because we knew little about what we could rescue at the time. They had to have bi-allelic SMN1 gene mutation and two copies of SMN2 and obviously were symptomatic. They couldn't be on a ventilator, couldn't have binding antibodies to AAV9 and couldn't have any genetic modifiers. The primary outcome was safety and tolerability, as this was a Phase 1 study. And the secondary end point was patients who were dying at the 20 month time point. With other exploratory outcomes like the CHOP INTEND which turned out to be extremely important. 

One thing I like to point out because this is important is that if we look at cohort one, and I didn't mention it this, there were two cohorts for this study that I'll show you in the next slide after this. The two cohorts received different amounts of, it was a dose ascending schedule. In the beginning we could only recruit, there was a great deal of fear about gene therapy back in 2014 when we started this clinical trial, especially given IV gene delivery. And the older patients came for treatment and these were 6.3 mean age in years. As the trial got moving along, we got younger and younger patients and this turned out to be what was the determining factor in the success of this trial. Notice also that the onset of symptoms were about the same, but it was when the patients were treated in this clinical trial that made a difference.

Totally, there were 15 patients in the trial and the age between four weeks and 7.9 months. This was a dose ascending schedule. Toxicity showed that about 30% toxicity with the lower dose of 6.713 vector genomes. And the high dose was 2E to the 14th. We had good toxicity data for both the 6.7, I meant 6.7 gave us about 30% efficacy in the mouse. This gave us 100% efficacy in the mouse.

Now, this slide really frames this whole idea of prophylactic immunosuppression. In the very first patient we did at the low dose, 6.13E to the 13th, patient one had a 31 times normal elevation of AST, indicative of liver enzyme elevation. I went to the FDA at that time and discussed with the clinical reviewer, we discussed using Prednisone and when we had the very high AST and ALT, we were able to suppress it easily with Prednisone. And from that point forward, we had decided and got approval for using immunosuppression one day before gene expression. This is something we carried on for all of our clinical trials since then and many of the other clinical trials have used that as a model. Also, because of the efficacy, this dose of virus, 2E to the 14th vector genomes, which was the highest dose of virus AAV that had ever been delivered before has now been adopted for many subsequent clinical trials. I like to show this slide because in babies, when you give the virus, you have virtually no reaction. We have no AEs that we've seen with delivery on the day of delivery. It's very well tolerated despite the high dose. Now, this frames the clinical trial in two cohorts. Here on top you see the three patients in cohort one was 6.7E to the 13th and cohort two with 2E to the 14th. This is 12 patients, this is three. All 15 patients in the trial survived beyond the 8% predicted vent free survival.

If we look at the two cohorts, now we have very good survival. This is the primary outcome. At the time we reported the results in 2017, cohort had all survived to over 30 months and cohort was beyond the 8% 20 months survival and survived to over 25 months. One patient in this trial did use, in the older cohort, went on ventilatory support at 29 months. 

This is another very important slide framing the clinical trial. I show this one. The other Natural History study was done actually by Steven Kolb at OSU showing that even worse survival in their Natural History study with 50% only surviving eight months. Now, the patients who were the earliest in this trial, one at one month or .9 months and the other at 1.9 months, two months, they basically returned to normal. And that you see here.  All the other patients, this is cohort 2, reached beyond the predicted functional result using the CHOP INTEND model, they were all above the 40 CHOP INTEND. These reached 64, and most of these have now even improved since that time of reporting in 2017. If patients are treated at an early time point and they have a low CHOP INTEND, so they have a more severe disease, they don't accelerate up to this normal level. And further emphasizing the time of delivery, one patient in this first cohort was treated at 7.9 months and that patient never improved to any significant degree.

The response was also remarkable when we were watching these patients. By a month we already had significant improvement. This was an open label trial. By three months, we had improvement in CHOP INTEND by 15 points. This is a cohort 2, which shows the low, this is the 6.7E to the 13th vector genome delivery cohort and they basically stayed the same, improved some. Actually, out here as we're following these patients, they're gradually improving. This slide just frames what they were able to do at the time we reported this in 2017. Here you see the patient. This is the patient who was treated late at 7.9 month, really never gained any significant milestones. The two early ones achieved everything there in blue. Two months and one month.  And then there was varying degrees of improvement, mostly significantly improved depending on how long they were in the trial and whether they started by November 5, when we reported the initial results.

Here's an overall framing of it. 92% of patients in cohort 2, the 2E to the 14th, were fed orally, 92% were speaking. Only 66% were having to use BiPAP and none on an ongoing basis and there were two patients who were standing alone and two patients who were walking independently.

Now, this was one of the most remarkable things about this trial and it's very important in thinking about clinical trials in terms of age. We had very few side effects adverse events in this first trial, and that's more or less been validated. Only two patients had SAEs with liver function elevation. That meant they reached 10 times elevation of the liver function test. And then there were an array of clinical trials related to their disease but not related to gene delivery. So this trial went extremely well in terms of safety. And just to show you a few pictures of what things were like, these are patients who, this is essentially a normal child function and this occurred four months, he was treated at 27 days and by four months was lifting his head. Another patient who was able to use a bottle six weeks post gene therapy. This patient had achieved good truncal control six months post gene therapy. And the truncal control is also illustrated hereby able to maintain truncal control and legs extended.

Then here's a patient who's able to, sorry. That was one who was able to sit independently.  There it is. And then on the next, here's a patient, who's cruising at 12 months post gene therapy.  And another patient who is able to walk. So we're really following normal development. We have patients with head control, patients who are sitting, patients who are cruising, patients who are walking, and then finally patients who gain good muscle strength. This is a patient who was treated early and was able to really achieve independent function. So the trial went very well. 

One of the most important things in terms of impact of this trial was it led to newborn screening for SMA. And now there are 33 states that have newborn screening for SMA. In Ohio we have reported the first cohort of patients who were screened and that is 21 patients who were reported, 17 improved, 2 stabilized. And this was recently reported in pediatrics. So treating very early can essentially rescue most of the patients. These patients were not all treated at this early time point, but the approval for SMA by the FDA was up to two years. And that up to two years, so they could be treated commercially anytime in the first two years. And that reflects the two year death time point for patients with SMA.

So this is very gratifying. This was a report by Ann Connelly and Meghan Waldrop at Nationwide Children's Hospital. Now, Novartis acquired the original AVEXIS license for carrying the gene forward. They have now put in a program that really encompasses the whole disease.  The first I'll talk about because these two have now been completed. From the start trial, that was the initial trial of the 15 patients, this is now extended out. The long term follow up has now been reported as of December 31, 2019, and the up for that has now been accepted for publication. That's in JAMA neurology. And this is the data back in December 2019 for embargo purposes.  At the beginning of 2020, patients were living a mean age of 6 years and a meantime since dosing of 5.5 years. As far as the therapeutic effect for long term, basically all patients were staying the same and continued to improve. Here are the two patients that were walking.  They lost no function. All the patients now were sitting. And two of the patients were standing with assistance. 

The other part of the trial was the Phase 3 trial that started shortly after the 2017 paper was published in the New England journal. This was called the STR1VE study. It was the Phase 3.  The reason I like to show this is because the START trial was no accident. The STR1VE trial, the Phase 3 trial which now actually narrowed infant enrollment down to six months, not that nine month window that we had for the first trial, but this had two endpoints: Survival free permanent ventilation at 14 months, even narrowed it further, and independent sitting over 30 seconds at 18 months of age. This is the age of when then these 22 patients were enrolled in Phase 3. The PI on this study was John Day at Stanford. This is a pretty good ratio between males and females. None were ventilator dependent here. And the primary outcome was ventilator free survival at 14 months and 20 of 22 patients survived not only the 14 months but now out to 18 months. And the majority of patients had achieved independent sitting at 18 months. These are the scores at CHOP INTEND in Phase 3, over 95% of patients were over 40 which is never seen in the natural history of the disease. 60% of patients reached 50, and five patients now were back to normal. So patients also in this trial improved as early as one month.  Now, where is this going from here?

I didn't put in the results here because they're ongoing trials, although the SPR1NT trial now is being reported. This is one that was Novartis put into place that would be the treatment of infants less than six weeks of age. So this is essentially from the newborn screening that patients are now identified at this very early age and the results of this trial have just been submitted for publication and is also equally effective.

Then the one trial that has become somewhat to the attention of many of the clinicians and translational scientists is intrathecal administration. There were attempts to circumvent these very large doses IV by putting the gene into the spinal fluid directly. The results have been good.  I don't have them reported here because this trial is currently still underway and there was in experimental studies in non human primates, there was a question of whether there was an inflammatory response in the dorsal root ganglion. The trials that have followed what we did originally of large dose intravenous delivery of the virus with AAV9, the SMN1 gene that was codon optimized that was delivered with ubiquitous promoter has gone very well. 

So I'll stop there and we have, I put mostly here people at Nationwide Children's, but now the trials have extended way beyond that. Industry partners in the beginning, it's interesting to me as an academic neurologist and translational clinician that this AVEXIS started out with five people that I got very close to in developing the clinical product. And then by the time we published the results in 2017, they had expanded to 200 or more. And then licensed the product to Novartis who are really carrying this to a new level. I'll stop there. I don't know how much time I've used but obviously I haven't gotten any warnings yet so I'm happy about that. I'll stop there.

Mustafa Sahin: Thank you, Jerry.

Guangpin Gao: Our next speaker is actually the Rett Syndrome and GAN, this is by Dr. Steven Gray from University Texas Southwest. Thank you.

Steven Gray: Yes, it's a pleasure to have been invited. Glad to share our work on giant axonal neuropathy. This is a talk really, a little bit about gene therapy for Rett’s Syndrome, I think the main purpose of this is thinking about how to regulate dosage of genes. I do have some competing interests to disclose because I do work with industry to translate discoveries out of my lab. And I want to say really quick that we have got a long time collaboration with a group of researchers led by Stewart Cobb at the University of Edinburgh that contributed to work over the decade, and Sarah Sinnett at UT southwestern that’s leading most of this work now. 

Rett Syndrome is a disease due to deficiency in the MeCP2 gene. MeCP2 is a X-linked gene. A complete loss of MeCP2 is typically fatal in boys, but girls do survive due to random X inactivation. So basically about half of their cells will have the defective gene expressed and then half will have the wild type gene expressed. So this ends up affecting about one in 10,000 girls. There's neuromuscular defects, developmental delays, mental retardation, seizure, breathing problems. There's a variety of complex symptoms with these girls. Their median survival is about 40 years. The complications of trying to treat Rett syndrome are that MeCP2 is a broad chromatin binding factor that regulates the expression of many, many additional genes, so trying to treat the downstream symptoms of this can be quite complex. So gene therapy, if you could restore MeCP2 had the promise and prospect of providing more comprehensive rescue treating this at the root of the problem.

Excitement about the prospects of gene therapy really came about from a paper published in Science in 2007 from Adrian Bird’s lab that showed that if you genetically restored MeCP2 expression in mice you could alleviate most of the symptoms. It gave a lot of encouragement to the field where there were questions about whether this disease had any type of reversibility.

If you think about this scenario in terms of a gene rescue approach, a gene approach, you're in a situation where you have got in these patients they're essentially chimeras, so half of their cells will have wild type MeCP2 levels and half will have dysfunctional MeCP2 and it's this mixture of this throughout all their tissues. There's also a MeCP2 duplication syndrome that Dr. Zoghbi will talk about in a little bit, where some patients are born with duplications of the MeCP2 gene and they have a condition that is essentially as severe as Rett syndrome or similar in severity to Rett syndrome.  Our goal in terms of the MeCP2 gene transfer approach would be to get just the right levels of MeCP2 in every cell, not too little, not too much, just right. 

Of course from a gene transfer perspective, hypothetically if we could deliver the gene at the same level and express it to the same extent in every cell, then we would take these previously normal cells and induce essentially a duplication or MeCP2 over expression phenotype. Whereas we would take the previously dysfunctional cells and rescue them and make them normal. So this creates a pretty difficult prospect of how do you treat this disease because the reality of the situation is you're not going to deliver one copy to every cell. You might have one cell that get one copy, you might have some cells that get 10 copies of the transgene. 

So there's a long history from my lab and other labs trying to develop gene therapy approaches for Rett syndrome. There's been some encouraging pre clinical data published in terms of some improvements that have been reported in the mouse model, but there's also some significant side effects including death that have been reported when MeCP2 levels get too high. This is just a history at least of the vector designs that have come out of my lab, where we started off with a single stranded AAV vector with a strong ubiquitous promoter, a very basic generic construct, and then later moved into a self complementary design that would be more efficient and incorporating fragments of the endogenous MeCP2 promoter and an additional regulatory element in the polyadenylation signal, and that is 3' ETR. And kind of what this looks like and why some of the changes happen was it was identified that this initial core MeCP2 promoter was missing an inhibitory element that was included when it was expanded slightly. And we published that that inhibitory element could confer regulatory control, additional safety. And then this polyadenylation, it was a fragment from the endogenous 3' UTR that incorporated some microRNA binding sites that might limit expression in the liver and potentially other cell types where off target expression would be undesirable. Moving into the efficacy of this, if you were using this AAV9 vector, which Dr. Gao talked about in the beginning that's kind of become the favored vector for CNS gene transfer, if MeCP2 is packaged within AAV9 and injected into the sterna magna, into the cerebral spinal fluid for broad CNS distribution, then in a dose dependent fashion, you could increase the survival of these mice. However, if you look at this the severity score, a Rett specific severity scale looking at Rett symptoms, if you follow these mice post treatment, you'll see untreated mice develop Rett like symptoms in the gray line. The low dose doesn't really change this and the high dose actually makes this worse. Even though the mice are surviving longer, we're making their symptoms more severe. This again touches back to these problems of the Goldie Locks scenario of MeCP2 about trying to get just the right level.

If we do a dose response study in wild type mice where these mice already have wild level MeCP2 in the cell and we do the cisterna magna injection, we actually see a dose dependent early mortality in those mice. Again, looking at the severity score of Rett like symptoms, in a dose dependent fashion, we reduce Rett like symptoms in previously normal mice. I hope this really confers that the problems and the difficulties of gene therapy for Rett syndrome where there's a very high capacity to actually make things worse instead of make it better.

Really the topic, that's to set the background of this. I think looking forward what we think is an innovative strategy that we embarked upon with Rett syndrome that might be applicable to other dose sensitive neurodevelopmental genes is the creation of this miRARE element. This is a regulatory element that we're calling a microRNA responsive auto regulatory element.

And so it is essentially a panel of microRNA binding sites. It's not expressed in any microRNAs, it's not expressing any polypeptides, it's just binding sites for various microRNAs. And microRNA regulation has been used in the past, it's published from a number of labs, as a way to, say, limit off target expression in the liver. You might include a liver specific microRNA.  This is a little bit different because it was selection of microRNAs that would actually create a feedback loop of gene regulation. And so the way that this works is we have this miRARE element, microRNA binding sites, and the 3’ETR of our genome. If the genome goes into the cell, we start expressing mRNA, we express MeCP2 and for whatever reason, either we go into a wild type cell or get too many copies of the genome per cell then we might drive overexpression of the MeCP2 protein. Through some separate work that we did, we identified a set of microRNAs that just naturally in the cell are responsive to MeCP2 expression levels. So essentially if MeCP2 levels in a cell go up, these microRNAs from the cell also go up. These are the microRNAs that we built in targets for. Again, MeCP2 levels in the cell go up, these endogenous microRNAs from the cell also go up. And then those in turn feedback to bind to the transgene mRNA and silence it. You can imagine this would reach some kind of steady state level of expression where some amount of MeCP2 is tolerated but not too much.

And so just some quick images of this. If we have a non regulated MeCP2, it's over here in these panels on the left where you see expression in green. And then the regulated MeCP2 we get much diminished expression here on the right where you see very low levels expression in green.  We haven't silenced it completely. If we look at the gain settings on the microscope, we can see we have positive cells, it's just the expression is attenuated and we're basically not seeing any instances of MeCP2 over expression in any cell.

So the functional consequences of this are this is a survival curve in wild type mice.  So the mice should not be dying. But if we inject a non regulated vector shown in blue and shown in dark green here, the non regulated vector, just like I showed you before, can induce some levels of mortality, early mortality, about 40% of the mice die. Whereas our regulated vector that's here in orange and red, this is safe in wild type mice.

And the same is true if we look at things like severe gait and severe clasping, which would be side effects of MeCP2 over expression, again this in wild type mice, it's essentially a safety study, then the non regulated vectors induce severe clasping, induce severe gait abnormalities. Whereas regulated MI rare vector is well tolerated and doesn't induce these side effects, at least not nearly as much as a non regulated vector.

And finally, if we look at this aggregate score of Rett like behaviors, like I showed you before, similarly here, this is in wild type mice, these are normal mice that shouldn't be developing Rett like symptoms. We have vehicle injected mice in black, we have our non regulated vectors in green and blue that induce Rett like behaviors, and then our regulated vector, high doses in red here is showing safe.

And I think my last data slide here is we could also inject water and it would be safe, but we show here just early experiment in the knockout mice with Rett syndrome where the regulated vector, which is here in red, does extend survival of these mice, it does confer a therapeutic benefit and the therapeutic benefit is actually equivalent to the non regulated vector in blue, but it has a considerable safety advantage. Next steps would be to further evaluate this for efficacy in Rett mice and further evaluate this as a potential treatment for Rett syndrome for safety in small and large animal models and also to see if this MI rare regulation platform could extend to other neurodose sensitive genes. I'll end off there and I can pass off to Dr. Bonnemann.

Guangping Gao: Thank you Steve. Dr. Bonnemann will give a talk on the gene therapy clinical development. Thank you.

Carsten Bönnemann: Thank you very much for allowing me to tag team on with Steve. We continue with Steve Gray initiated project. This is the GAN (giant axonal neuropathy) trial obviously in the clinic. This is what the GAN trial is in a nutshell. It is a relentlessly progressive early childhood neurodegenerative disease. It is a single center study, intrathecal dosing with ascending doses, per patient dosing. It includes patients who are NAB positive for AAV9 and AAV9 antibodies. And it includes CRIM positive and CRIM negative patients. It allows for preliminary efficacy analysis based on natural history data.

What I tried to do here is because we are talking about CNS disorders of course I'm trying to extract from this the relevant general learning points that are useful perhaps to this group, which is a collaborative single site model as a possibility, intrathecal dosing is feasible and can be done safely, preexisting NAB do not preclude IT dosing, immunomodulation can be adjusted to needs and biodistribution with intrathecal special dosing is a gradient, and the importance of natural history data will be highlighted.

Very briefly, giant axonal neuropathy is an early onset neurodegenerative disease of children. It is affecting initially the dorsal root ganglion and sensory neurons and then motor nerves, but as the disease relentlessly progresses, it also effects spinal cord eye and brain becomes a global neurodegenerative disorder. The gene is a small gene that is involved in ubiquitination of intermediate filament and the mechanism is a loss of function mechanism. So this is a staging of the disease. It starts very early on in two to three years of age really it becomes evident there's a gait abnormality which then progresses along this neurodegenerative course affecting motor functions increasingly as well as eye and central nervous system function. And as we see the feasibility for rescue of a disease like this may decrease as the disease progresses.

So let's start with the point that I outline that I would like to highlight in this talk is that we really got this going as a collaborative effort between patient advocacy, academia, NIH and industry. In a single site as perhaps a model how to make an ultra-rare disease therapeutic situation feasible and happen. With that, I would like to introduce the study team that's involved in that. In particular, clinically at our side, Diana Bharucha-Goebel and Dimah Saade as the lead clinicians. And then of course Steven Gray is our initiator who invented this trial. And then HHF is the patient advocacy who really got this off the ground. Industry partners at bamboo and Pfizer and the immunology core.

There is the conceptual overview of the trial. It's a small coding region allows for safe complementary design. It's ubiquitous and weakish JeT promoter that uses AAV9 as a Capsid code optimized transgene and the dose escalation is shown here. We started at 3.5 E13 genome from entire patient, and now at the 10 fold higher dose of 3.5 E14. Because we're targeting spinal cord structures initially via the motor neuron, the intrathecal approach was chosen and this was done in a Trendelenburg position to allow the percolation of this vector along the neural axis so that's kept an hour after dosing in the Trendelenburg position with a goal of also targeting of course higher structures such as the optic nerve and the brain eventually. 14 subjects have been injected thus far including four that are CRIM negative and three at the highest dose. CRIM negative means by genetic analysis these patients are expected to have no baseline protein expression of neural exon and so that should be immune-naive, meaning if you give a transgene encoded approach in the immune system would recognize it as foreign and that requires special considerations as you will see.

First about the neutralizing antibodies, in this trial pre existing neutralizing antibodies against AAV9 do not preclude IT dosing, not an exclusion criteria, we saw is a faster mnestic anti AAV9 response neutralizing antibodies and of course appearance of NABs in the CSF postdosing, but there was no efficacy or safety signal associated with this. This of course based on work also from Steve Gray and others that shows intrathecal delivery can circumvent systemic neutralizing antibodies against AAV9. And this is a core here can you can see 22% of our patients were completely seronegative and 21 are borderline.

Just as one example, what happens here, I mentioned the mnestic response, so the rise of neutralizing antibodies in the periphery is high in everyone but is faster in the patient who are positive at baseline, so they have a fast amnestic response. The patients negative in the CSF for neutralizing antibodies but our patients over time developed neutralizing antibodies in the CSF and that again is I think relevant to consider when redosing considerations are brought into play, this will be expected to impede redosing by the same route of administration even though at baseline they were negative.

Immune-modulation is the other aspect of this. This is of course a complicated equation because it has to take into account immune reaction against the Capsid as well as against the transgene. And so in our trial I think we are showing that dosing of CRIM negative patients, so patients with no immune knowledge of the transgene, that is possible and we're hoping that a protocol may also allow development of tolerance. 30% of our patients were CRIM negative in this trial and they received the additional immune modulation. This is a diagram of the different phases of a potential immune complications. The gene transfer happens here. There may be immediately innate immune complications, the anti Capsid and the transgene is turned on, leading to adaptive immunity and hopefully in the long run for tolerance. Every patient receives an IV Methylpred at the time of dosing to dampen innate immunity and then gets an extended prednisone systemic regime for four months, you’ll see why in a second. And then patients who are CRIM negative, we have to encounter the possibility that there may be anti transgene T cell mediated immunity that could impair the effect of gene therapy or even be a safety issue causing TMS inflammation, kept on this modified T cell active protocol whether rapamycin as well as tacrolimus as a T cell directed immune modulation protocol. And at the highest dose we’re currently giving of 3.5E14 for patient, we’re actually keeping all patients whether they are CRIM positive or CRIM negative on standing rapamycin with the hope of also inducing tolerance in the long term, although we’ve not been really monitor that effectively, but it's been safe and effective thus far. This is for the CRIM negatives and this is for all patients involved in this trial.

So the next point of perhaps to take home here is thus far intrathecal dosing appears to be not only feasible but also safe in our trial that we've done. I point particular on phenomena of delayed inflammation in the CSF that can be suppressed with steroids but is clinically silent. And then I would also touch briefly on the issue of DRG toxicity because that's been up in the forefront of the field now that we have data points that have been supplied pre clinically as well as clinically.

So this is about the pleocytosis. In patients with a dose dependent way, so the higher the dose the more it seems to be appearing. Three months delayed pleocytosis in the CSF, there is some elevation as well that is clinically asymptomatic but it is an inflammatory response within the CSF. That can be suppressed with a prolonged steroid regimen. So steroids now, this is steroid response, you suppress this inflammation, it will eventually return back to normal. The driver of this inflammation is not clear to us, but it could be Capsid related. Again, it's manageable and clinically silent.

The DRG toxicity is an interesting issue for the GAN trial because the root ganglion causes the first side of the disease process is the structure that we need to target the most with our gene therapy as part of the therapeutic approach. Also the disease destroys the integrity of the DRG earliest in the disease course and that of course complicates our monitoring of DRG toxicity. One transgene transfer autopsy case at the very lowest dose we had, eight months after delivery of the gene therapy in a patient already quite advanced in the disease. We saw a neuronal loss consistent with the disease but no additional inflammation. In the biodistribution data from that autopsy was in keeping the loss of the substrate rather than with inflammation of the DRG. This is some of the biodistribution data here, you can see along the spinal cord expressing that the genomes, that there was eventually good transduction at the low dose along spinal cord, but the DRGs, being the first disease target in the spinal cord, were the lowest transduction here. We feel that's a reflection of the disease process rather than of the DRG toxicity. In support of that, when we look at our younger patients now dosed at an earlier disease stage and at higher doses we have no clinical inclination of DRG toxicity thus far. In fact, one of the patients at the now middle dose regained sensing of action potential two years for the gene therapy so that is a reflection of more healthy DRG post gene transfer. But the final word on that of course is not out. Our patient population, as I said, is difficult to use for monitoring for DRG toxicity.

The biodistribution is important to consider with intrathecal dosing that we do here. It is a gradient. Dose lumbar per lumbar access and gradient of course to supratentorial structures is further away from where you deliver the gene therapy product. This is the more complete biodistribution data from the initial patient that I mentioned. One thing that is clear is the spinal cord at this dose level is better transduced and higher auto structures with the exception interestingly of the cortex, which may be lucky sampling but it's interesting finding. There is a gradient. And if we saturate the spinal cord at a certain dose level, we probably are not yet saturating the brain. The good news is pre clinical, non human and rodent data does predict the human biodistribution quite fairly so that can be used.

And then lastly, the last point I want to make is that strong natural history data and individual one in data helps to power analysis efficacy in super rare diseases such as GAN that has implications for disease planning as well as trial intervention. This is our Natural History data on the motor outcome measure we use, the MFM32. You can see a very solid and very tight 8 point loss per year in this patient population. If you overlay your post-dosing data on that, you can see how it diverts from the Natural History data that's been collected before. That can be the basis of evasion type analysis where you calculate the posterior probability of halting the disease in comparison to natural history and individual running data.

Geetha Sentil: Three minutes.

Cartem Bömmermann: Thank you. And dose dependent manner this now shifts towards halting the disease, peak probability is no longer minus 8 but is minus 6.1, as I said, a potential efficacy marker. You can see schematically, this is a non loss here, the MCID for MFM32 is now at 32 and our analysis suggests slowing and of course as you go into younger patients in higher doses and intervene earlier, it's clearly likely that you have more efficacy from that.  Our current trial patients are about here. Current trial patients are about here, and the next phase of the trial at the higher dose, we're moving into younger patients. So dose and disease stage I think are determining feasibility for rescue in a disease like this. With that I would like to close out my presentation here and just really highlight the GAN study team one more time who made this trial really possible from pre clinical inception, to planning, to execution at the NIH Clinical Center. Thank you very much.

Mustafa Sahin: Thank you, Carsten and Steve. Our next presenter is Dr. Huda Zoghbi, professor and HHMI investigator at Baylor College of Medicine. She's going to talk about MeCP2 duplication.

Huda Zoghbi: Thank you Mustafa and Guangping for asking me to join this wonderful meeting. This afternoon I'll share with you our work that provided the proof of concept data to proceed with the clinical path our translation and hopefully we will address some of the things he charged us with as we contemplate such studies. I would like to disclose this work is collaborative with IONIS pharmaceutical and also collaborate with USB pharma and a couple of advisory boards.  You've already heard from Steven that MeCP2 is an X linked gene that encodes the methyl CPG binding protein 2 and it is a protein expressed in all brain cells, very abundant in neurons and as you know back in 1999 we learned that loss of function of mutation in this gene cause Rett syndrome which really much affect all parts of the brain and nervous system and causes all sorts of autonomic dysfunction, typically seen in girls because they're mosaic, they have 50% of their cells with the healthy X chromosome and the other 50% with the mutation.

What we also learned in 2005 and this was nice work published by Van Esch and colleagues is that doubling this gene in males can cause similar progressive neurological disorder. Slightly different features, recurrent infections in the early hypotonia is unique to boys, but otherwise there are many overlapping features. I should mention that the duplication is typically seen in boys because most of the carrier females, the mothers, have cells that preferentially survive with the healthy X chromosomes. Cells with a duplicated X chromosome are very few in these females. However, when you have a female that's mosaic, you will also get the progressive neurological disorders you see in this syndrome.

Because many of the autism and neuropsychiatric disorders we speak about are caused by  considerable variation, it's always asked how do you know which gene to target when you have so many genes duplicated. So I want to share with you here how we narrowed down our studies to the MeCP2 gene itself. First you'll notice that most people with MeCP2 duplication, this is work from the Lupski lab and Carvalho the very detailed work in the area where they mapped structurally the copy number variations, you'll notice they start and end in different places. So this is not the same copy number variation, there are multiple different ones and sometimes you have triplication, where the dosage of the protein is higher and the disease will be more severe. The only region of overlap among all these patients spares the MeCP2 gene and  kinase IRAK1 that for now we don't really know its function in the nervous system but it has function in the immune system.

That narrows down at least many of the phenotypes that are overproducing in patients to these two genes. However, serendipitously and actually before the discovery of the human disorder, we created a mouse that expressed the human gene with all of its regulatory element. And these mice had two copies of MeCP2: the endogenous mouse copy and human copy. And to our surprise, we found they develop a progressive neurological disorder with all the following phenotypes including premature death about 20 year of age in these mice on an FAB background. And shortly after the MeCP2 duplication was discovered and if you look at the phenotypes, they're really identical. While in humans we see respiratory infections, we know their T helper cell abnormalities in these mice, we don't see infection because they're probably in barrier facility, but importantly the neurological features are identical between the humans and the mouse model, at least telling us that quite a bit of the neurological symptoms in this disorder are due to the MeCP2 gene.

So what I would like to share with you now is the work we've done to figure out how can we intervene to help this disorder. And this is work that was started by Hezi Sztainberg when he was a post doc in the lab and completed and expanded by Yingyao Shao, who is currently defending next week her PhD thesis. We’ve done experiments genetically and we’ve showed we can reverse the phenotype but I'm going to share with you the studies using antisense oligonucleotides. We use the model which has human gene and mouse gene. We targeted the human allele and we looked at the following phenotypes as we see in the mice. I'll show just a couple of examples of the data.

In this case the treatment started in two months of age. We heard that timing was important and these animals develop symptoms of six weeks of age, and we waited until they developed symptoms. Here you see activity levels rescued. These are the cinegenic mice, they have lower activity level, lower rearing episodes, lower entry to the center. And you'll see on one mouse and you'll see how all of that rescued by normalizing levels of MeCP2 targeting the human transgene. But as a neurologist, I was worried that maybe two months was still early and I wanted to look at a much later time point. In this case we waited until the mice were six to eight months old. On the background they have constant seizures and will die by about 20 year of age and you can monitor these by EEG as well as video. We started the treatment somewhere between 6 and eight months in these animals. After normalizing MeCP2 levels, both the electric seizures measured by EEG as well as the clinical seizures disappeared in these animals. So at least this was quite reassuring that at least we know behavior rescued after two months of age treatment and seizures even as late as 8 or 9 months were reversible.

However, in humans you don't have the protective effect of the mouse endogenous allele; you only have two identical genes, two human copies of the genes two or three. For that we needed to create a new mouse model, which is a humanized form where both alleles are the human gene.  The first question we wanted to ask: Can you use the ASO, can you titrate the dose so you can safely lower MeCP2 levels? And, Yingyao and Hazi generated this new mouse using two alleles, the original one and new one tagged with GFP, they showed these mice make twice the normal levels of the MeCP2 RNA and protein and they were able to show in a dose dependent manner, based on the amount of the ASO, you can actually titrate the level of the mRNA and the protein.  You'll see the duplication model and you'll see based on the dose you can get it normalized.  With that, we proceeded in this mouse model with studies for pharmacodynamics of the ASO treatment, what happens to the MeCP2 protein levels and RNA and what are the changes, when does the behavior rescue.

What you will notice is that we put everything at one, this is the mouse with the duplication against treated at two months of age with 500 micrograms of the ASO. You'll notice that the RNA level go down very quickly, by about one to two weeks, we see reduction in the mRNA level and then it gradual rises back again by 16 weeks it's almost reached the normal level. When we look at the protein, the protein lags behind. You'll notice here that by one week the protein is used by 20% and we reach the desired level of 1X because this is 2X by about two to five weeks. And then gradually again the protein lags behind the mRNA and you'll see the protein rising again.

However, genes that are downstream of MeCP2, when do these correct? We found that actually those begin correcting very quickly, within one to two weeks we begin to see correction of these genes and that correction is more complete by 16 weeks.  So although the protein level, I'll go back one slide to show you, and the RNA started to go back up, what's interesting is the benefit lasts, lasts a little bit longer than we had anticipated. So how about the behavior rescue and when do we see that?

I'm just going to show you, I'm not going to show you all the behaviors that we've done. In the previous study, the first study that we published, we used pumps and used the intraventricularly in all the behavior rescue. In this case we gave one intrathecal dose in the ventricle. One of the phenotypes these animals have, they stay much longer on the rotarod, and you’ll see the low dose and the high dose corrected that abnormality. This is a learning memory phenotype where these mice tend to freeze more, have a lot more freezing intellectual fear assay, and the low dose did not seem to have as big a benefit but the higher dose you'll see here corrected that phenotype. This is the dose that bring back protein to normal level. This already shows you that this dose rescues at least these two independent behavioral phenotype, but the rotarod seems to be very easy rescuable, even slight reduction of MeCP2 levels can give a benefit.

What's interesting, again, all the other features were rescued but what's interesting is the anxiety phenotype using this method of delivery and that short period which we followed this mice, about 8 weeks from the time, well 16 weeks from the time of the first treatment, the anxiety did not rescue. This may require multiple treatment to see benefit. We did not see a rescue for the anxiety, although the protein was expressed. If I was to summarize all the data for you, we started with twice the normal level. We see the mRNA to be the first thing that normalizes. And then the protein follows that. That's what normalizes next. The targets also are normalized early. This is weeks after therapy. But the behavioral rescue is still not normal. We see the behavioral rescue happening almost weeks later, 8 to 10 weeks, if you will, after the initiation of therapy, even though by them the RNA and protein level is coming back to normal. This is really important.

I want to maybe sit for a minute here on this slide and have a very brief discussion of some important things to note. It is good that we see molecular changes very early on because that means we can follow these molecular changes and we know we're hitting our target and hitting the target to the range we want. And you can come back with anti sensing kits, you're overcorrecting in that you don't get behavior or clinical changes until much later. That's one thing that's really important to learn from the study. The second thing that's really important to learn from this study is that even in mice, to me, they rescue easily, I would think, compared to a human probably because of the shorter span and so on. Even in mice it does take time to get functional behavioral rescue. We also did additional synaptic plasticities and in all of the studies have been about the same time. So we don't know yet if the physiology happens earlier so we're doing the studies to see what's the earliest physiology might happen with normalizing MeCP2 level, but what's really important if we are to do a clinical trial, we have to really consider time as we think when we would see a benefit. And I think Mustafa alluded to that in his talk with the TSC studies that the short period may not be enough.

The steps we've taken to translate this approach to therapy is created humanized model. We validated the model, we did show effective distribution, we showed efficacy, the dosage, the dynamics of the levels of the protein, and I didn't show you the gene expression rescue, but we got a very nice gene expression rescue as well. What we really need at this point, what's still in the work and we're working very hard on that is a biomarker that can be helpful at the single patient level. You heard that this is a Goldie Locks protein. You have to get it just right. Too much can cause a duplication, too little can cause Rett syndrome. Ideally what we would like to do is get it just right. I show you here some examples. If you lack it in 50% of the cells, give you the Rett syndrome in females, of course total absence is very severe in males. When you have 50% reduction, you have a milder phenotype but still you get symptoms and the duplication, triplication is high. Ideally we want to get to this spot and biomarker to the surrogate levels of MeCP2 that's faster than clinical features so that when we treat we monitor that and we know in the right ballpark.

With that, however, we still wanted to know what would happen if you treat with an ASO and you over shoot. What might happen? Are the consequences reversible.  For that, Yingyao injected wild type mice with ASOs that only have one copy of MeCP2, and pretty much depleted these wild type mice from their MeCP2, made them essentially null as adult mice and you can see that here. She collected tissue in different weeks after the injection and did behavioral studies and followed them. You will notice here that these animals that were injected with doses of lowered MeCP2 that they did have a Rett like phenotype, if you will, decreased motor performance on the rotarod, that’s the opposite of the duplication phenotype. I should mention that phenotype appeared nine weeks after the ASO injection. It takes time from lowering the protein, you have time, even when you have toxicity, it takes almost two months to have phenotype in the mouse, what you see here is that phenotype corrected by 22 weeks of age. She looked at hypoactivity which also occurred at nine weeks of age. That again persisted at 13 weeks and 21 weeks but by 25 weeks corrected.

I would summarize here, I'm not going to show you all the behavioral data she's done. She's taken healthy adult animal, depleted of MeCP2. And those animals remain normal as part of the depletion until about nine weeks after the depletion when they began to show anxiety, motor coordination and hypoactivity and those rescued in different periods of time, anywhere from almost 7 weeks after the appearance of symptoms to 22 weeks or so.

So again, this tells us that we have a period to monitor the level of protein or surrogated biomarker for this protein before we encounter any symptoms. If the treatment is overshot and even in the worst case scenario it's reversible. I just want to take two minutes to tell you new data that we learned about the MeCP2 regulation that also gives us some clinical insight about treatment and what we should aim for. One thing we were interested in is understanding the regulation of MeCP2 level. We know the promoter, it's right here, but we sense there should be other regulatory elements we should go after. So Yingyao collaborated with Josh Whythe, who was doing ATAC-seq in the brain in both during development as well as in adulthood and also in other peripheral tissues. And we focused on ataxic peaks that indicate open chromatin particularly in the mature brain and identified these six including the promoter.

Since the promoter is known, we decided to study the other five in vivo by deleting each one of them and assess what does it do to MeCP2 RNA. You'll notice here these deletions, the first peak did not have much effect, but three of the peaks are clearly importance enhancers, they lower the MeCP2, and one of them caused an increase in the MeCP2 RNA. At the protein level the effects were most dramatic for peak two where you see in this case reduction of protein with very clear in the brain and peak six that causes increase in the protein and the brain tissue. So we've identified two peaks in the mouse that reduce MeCP2 or increase MeCP2.  Then we asked are these conserved in the human. And based on [word] data in human brain they are. But we went to iNeurons and asked what do they do if we deplete them, if we remove these peaks from human neurons. You notice lots of peak two reduces the protein whereas loss of peak six increases the protein. We have, if you will, a mild reduction of MeCP2 protein levels and mild increase in MeCP2 protein levels. So what are the phenotypic sequences?

You would imagine this should look like RTT, and this should look a little bit like the duplication. This is exactly what we learned. You'll notice here 100% is normal and in Rett we see all these features in mice. In these animals the peak two depletion causes about 30% reduction of the protein. We only get partial phenotype, the hyperactivity, anxiety and social deficits, that subset of the full picture. On the other hand, if the duplication animals have 200%, the 50% increase gives us only partial features of the duplication. These are all the features of the duplication and with just this 50% increase, you're getting anxiety, learning deficit and hypoactivity.

There's two important things to really learn from this. The first thing is that we imagine there are going to be some mutations that partially inactivate the protein or maybe just need to 30% reduction that may give Rhett-like phenotype but much milder, maybe more autism like phenotype, and some that may cause just a slight increase in the protein and intellectual disability, mild hypoactivity and anxiety. But from the other side, from a therapeutic standpoint, it tells you if you don't correct back to 100% normal. If we don't correct to 100% normal, we will still get, even if we corrected to 50%, if we decrease from 200 to 150 with ASO therapy, we are going to get benefit on some behavior, motor function and epilepsy. And the same with Rett. Even if we boosted the protein, if we were to express at 70% at the normal level in every cell, at that point you may be only increasing it some in the wild type cell but at least you will correct quite a bit of the features of the disease, and I view that as really quite encouraging and helpful. 

So this now shows us that there's a really very nice spectrum of the levels of the protein and the gradation of the phenotypes that we see with of course massive increase and decrease being fatal and very severe, but in between you have opportunities now to treat and improve multiple phenotypes and of course I guess there's still mutations to be discovered that cause partial loss and partial gain of the protein.

So in summary, I hope I convinced you that mouse models can help us pinpoint which gene or genes drive disease in a copy number variation. This is important because quite a bit of newer psychiatric disorders are caused by a copy number variation and that humanized mice are really important when you want to target RNA and know exactly the sequence is, putting the human gene in place of the mouse gene is helpful. I hope I've convinced you that one can use ASO and can reduce MeCP2 levels in a dose dependent manner and rescue most of the behavioral of phenotypes as well as the synaptic plasticity and molecular, which I did not show. In the case that the ASO therapy leads to excessive lowering of MeCP2, the effect is reversible. And even small corrections in MeCP2 levels in the 30 to 50% range in either direction will benefit some phenotypes. This is where we are today and I hope in the coming year with the discovery of reliable biomarkers we can hopefully continue to move these studies forward. I just would like to acknowledge a few people. I always acknowledge Ruthi Amir because her determination is what got us the Rett syndrome gene in 1990. And Hezi and Yingyao did all the promoter studies and the safety studies that I shared with you. And Jianrong the EEG studies and our collaborators on the enhancer and computational and of course Ionis for all the ASO work. And I’m deeply grateful to all the Rett Syndrome families and our funders. Thank you.

Guangping Gao: Thank you. I would like to move to the next talk. That is by Elizabeth Berry-Kravis as well as Frank Bennett from Ionis Pharmaceuticals. Thank you very much.

Frank Bennett: I was going to start the presentation, give me a second.  Are you seeing my screen?

Guangping Gao: Yes, perfectly.  Ions.

Frank Bennett: Thank you for inviting me. What I wanted to do was pick up little bit of what Huda introduced and maybe describe in a little bit more detail about antisense oligonucleotides. And then hand the second half of the talk over to Dr. Barry Kravis who will talk about some clinical and exciting clinical data in Angelman Syndrome. And, first I’d just like to acknowledge our collaborators who have been working with us on Angelman Syndrome, starting with Art Beaudet and Linyan Meng at Baylor, Stormy Chamberlain has been working with us and our partners at Biogen, and my colleagues at Ionis that have been really dedicated to moving this project forward.

So, just to begin with, I wanted to highlight that antisense oligonucleotides in principle is a fairly easy concept to understand and what we’re doing is designing synthetic oligonucleotides to bind to RNA through Watson-Crick base pairing. Once they bind through RNA they can invoke a number of different effects on RNA. Perhaps one of the best characterized effects is because degradation of the RNA, either through endogenous RNase H1 that occurs both the cell nucleus as well as in the cytoplasm, and so we promote the degradation. These are single strand oligonucleotides that have some DNA in them to support the RNase H1 mechanism in action. You can also use oligonucleotides to invoke degradation through a different mechanism that these are siRNAs that work through Ago-2, it’s a different enzyme than RNase H1but that is very similar. In addition to causing degradation in the RNA, you can interfere with RNA intermediate metabolism. So one of the areas that we use, and are quite interested in this to modulate splicing, where you can design ASOs that would bind to the mRNA and either cause exon inclusion, as is the case for SMA, or you can cause exon skipping, as per the dystrophin protein muscular dystrophy drugs. And so it really comes down to where on the transcription you design the ASO to bind, would result in either exon inclusion or exon skipping. You can also use oligonucleotides to modulate polyadenylation site skipping, and this is still not widely used, but I still think it is going to be an important mechanism to modulate where transcripts could polyadenylate and maybe eliminate certain regulatory sequences on the RNA. And then finally, you can use ASOs to inhibit translation by blocking either the AUG codon or upstream regulatory element. And more recently we’ve been interested in ways we can use ASOs to enhance translation, either by binding to regulatory elements such as uORFs in the [word] region, or by antagonizing microRNAs, either by binding directly to the transcriptor or binding to the microRNA, depending on the specific use that you want. And so mainly wanted to leave you with there are a variety of different ways that we can use ASOs to modulate RNA function in cells and depending on the particular molecular defect that you’re trying to correct, one or another mechanism may be more appropriate.

So, some of the attributes of antisense technologies is the direct translation of genomic information to a drug. We use a primary sequence as a ways to design our drugs. You can target any RNA coding or noncoding. It’s a very efficient drug discovery process compared to more traditional small molecule type strategies. As Huda highlighted, the effects are titratable, so the more oligo you add, the more effect that you see. And she highlighted that they are reversible. So if by chance you over-suppress a target gene, you can just allow the effects to wear off over time.

Today there are eight approved drugs that utilize antisense technology. All of them are for rare diseases, not to say that antisense is only usable for rare diseases, it just happens to be that first eight drugs that have been approved have been for rare diseases. And there are more 50 antisense drugs currently in clinical trials.

So for treating CNS diseases, we recognized early on that ASOs to not distribute across the blood brain barrier. Most of the ASO chemistry that we use has a phosphate-backbone, so they are anionic, and so as you would expect, anions don’t cross the blood brain barrier. There are also some neutral backbone molecules that, because of size or other issues, they also don’t cross the blood brain barrier. So what we identified is that if you do an intrathecal injection into the CSF that surrounds the spinal cord and the brain, you get very robust distribution into CSF tissues. We have a broad experience with antisense, with over 11,000 patients treated by intrathecal injection. The majority of these patients are from our SMA work, but we have a growing body of patients that are being treated for ALS, Huntington’s disease, Alzheimer’s, Parkinson’s. So it is an expanding safety as well as efficacy dataset that we are getting with the technology.

I did want to highlight that antisense drugs are being developed for a broad range of neurological diseases. Currently in clinical development, there’s antisense for Huntington’s disease, four different programs for ALS that target both genetic forms of the disease as well as nongenetic. Drugs in development for Alzheimer’s, FDD, PSP, Parkinson’s, MSA, spinocerebellar ataxia, and as you’ll hear momentarily, on Angelman syndrome. Again, we’re just at the tip of the iceberg as far as using the technology and there’s a number of different additional indications that are being pursued that should be in the clinic fairly shortly.

So, one thing I did want to share with you was the distribution of antisense oligonucleotides into the brain. And by both immunochemistry staining for the presence of the oligonucleotide, as well as quantitating by mass spec methods. We see very broad distribution, sort of the key question we wanted to address was pharmacologically, what are the cell types in brain regions that we can affect the expression in.

So this is a study that we did in non-human primates, we were targeting a nuclear retained long-non-coding RNA called MALAT1, so it’s our equivalent to a RFP type study that you see other people use for reporter assays. In this case, we’re targeting an endogenous noncoding RNA that is widely expressed in all cell types. And so these were monkeys that were given intrathecal injection, and then after a few weeks, we sacrificed the animal and dissected a variety of different brain regions. As you would expect after intrathecal injection, we’re getting very robust knockdown in the spinal cord, about 90% of MALAT1 expression is depleted. But surprisingly if you look in cortex, you see we get distribution of the drug from the spinal cord up into the cortex and get very robust knockdown in very different cortical regions. That extends throughout the neural access. Hippocampus is an area we see very good knock down in, the one area where we get less effect are in deeper brain structures such as striatum. We still see effects, but the effect is less than we see in other brain regions. Importantly in the cerebellum we get very good knock down as well. So it gives us a lot of confidence that we are broadly getting distribution of the ASO throughout the neural axis. That’s also been supported in some clinical studies, unfortunately in our SMA program we had a few patients who died from their disease, that were gracious enough to allow us to analyze for drug effects in different brain regions. And it very nicely mirrors what we’re seeing in the non-human primates.

The second question is what cell types are we affecting target expression in. To do this, we did an in situ approach, where we developed an in situ probe for MALAT1 and a second probe for cell type specific markers. This is just an example that you can see in the bottom that the dose response curve that we saw for knock down at MALAT1 in different cell populations, and you see that essentially all the major cell populations in the CNS astrocytes, oligodendrocytes, microglia cells and neurons are sensitive to ASOs. With oligodendrocytes being a little bit more sensitive than the other cell populations. And so it was gratifying to see that we can effect expression in all cell populations in CNS.

With that, I’d just like to highlight some work that we’ve done to really establish the utility of ASO approach for treatment of Angelman’s syndrome. This was a collaboration that was done with Dr. [name’s] lab at Baylor. And just as a brief introduction, Angelman’s syndrome is a disease due to loss of the UBE3A gene. UBE3A is an imprinted gene, so that the normal individual, the paternal copy is silenced in neurons and then the maternal copy is what produces the UBE3A protein. So if you have a mutation in the maternal copy of the UBE3A protein you have a loss of UBE3A neurons. And so what we envision is based on some other studies that we could use ASOs to knockdown this long noncoding RNA that is in the antisense orientation to the UBE3A gene, and by doing so we would up-regulate the UBE3A gene and cells.

So that was confirmed in cell culture, where we used ASOs in a very nice dose-dependent manner showed knockdown of the antisense transcript, this long noncoding RNA, and reactivate the expression of the UBE3A RNA as well as proteins shown at the bottom. That was translated to mice where we were able to use reporter mice that expresses YFP UBE3A fusion protein that’s either maternally encoded or paternally encoded. And you see that in Angelman mice that there is loss the paternal copy so there’s no YFP expression, whereas if we treat with an ASO you see reactivation of YFP expression, or the UBE3A. And by staining for the ASO we can colocalize that the cells we get high concentration of ASO in, we also see reactivation of the UBE3A in these cell types.

And then finally, we were able to show increase in protein levels in the cells. And that translated to some movements and behavior and I’m showing one of the experiments that was done, used to look at contextual fear models in which animals were treated with the ASO and we could reverse the loss of contextual fear that was seen in the Angleman mice by treating with the ASO. Based on this result, actually we created some competition for ourself, there’s actually three programs now that are ongoing using antisense oligonucleotides. The most advanced is the GeneTx & UltraGenyx study that’s currently in a phase one-two study and Dr. Berry-Kravitz will describe that study in a little more detail. Second study that was initiated by Roche, again using antisense oligonucleotides to treat Angelman syndrome that’s in an early phase one-two study as well. And then we have a program partnered with Biogen that we plan on starting a study later this year. And so, with that is a background, and I’ll turn the talk over to Dr. Berry-Kravitz.

Elizabeth Berry-Kravitz: Can everyone see my screen?

[Male Speaker]: Yes.

Elizabeth Berry-Kravitz: So thanks for inviting me to carry Frank's discussion one step further and talk about the key questions, will ASOs work in kids with Angelman syndrome and how do you translate them to actually show the clinical effect in a severe neurodevelopmental disorder. To do that I'm going to talk about our experience treating the first five kids with Angelman syndrome with the GTX-102 ASO to activate the paternal gene as Frank described. These are the only five kids in the world that have been treated, well the first five kids in the world that had been treated. The Roche study is now ongoing but we can't talk about that.

This is a collaboration between at the point of these five patients, Roche, Ultragenyx and Genetyx and I have funding from lots of companies to do work on neurodevelopmental and neurodegenerative disorders but do not keep any of the money.  Okay, so GTX-102 targeted the cluster 2 of SNORD115, and the reasons for choosing the region was because it was at the beginning of the UBE3A-AS transcript, and cluster 2 is genetically and structurally different than cluster 1, implying a differential function. It has highly evolutionarily constrained in exons and splicing motifs, implying regulatory function. And several exons are identical across humans and nonhuman primates, thus allowing PD studies in the nonhuman primate model. And it was thought that targeting exons would obviously increase the ASOs chance of terminating transcription since introns would be spliced out during transcription.

So, this study was initiated last February, it was the first in human study, it was a study of four monthly doses of the ASO, and the patients were divided into cohorts so that the starting dose level would escalate between cohorts but also escalate within-person within the cohort, and the levels were chosen based on experiments in the nonhuman primates. There’s no placebo control group because it was a very early study, and the ASO was given intrathecally by lumbar puncture.

So the primary outcome is the safety of GTX-102, secondary is pharmacokinetics, and exploratory outcomes were also being administered in multiple domains of AS to try and determine whether these will these measures really work, to look at communication, sleep, behavior, gross motor, fine motor, seizures, all problems in Angelman syndrome, can we measure change in early study in those kinds of domains in terms of choosing outcome measures for the future. So patients could get in the study if they had a full maternal UBE3A deletion in the correct region.  They had to be between four and 17, have stable seizure control, normal kidney and liver, no bleeding disorders, and had to be able to tolerate anesthesia without intubation and they could not have made any changes in their medications or diets that were being used to treat symptoms in the prior three months before starting the study and they had to be ambulatory. So we were trying to pick a population that could be relatively easy to assess and was stable.

From a study design conduct standpoint, this is a complex study. It was divided initially into five cohorts but three cohorts got through dosing before we had to pause for an SAE. The dose was increased one step at each dose administered. Cohort one went in and had a 3.3mg, then 10, then 20, then 36. And then once the first two doses had been administered, there was a DSMB review, and we could start cohort two, which started 10mg, 20, 36. And then cohort three started at 20 mg. So the dose escalated between cohorts within patient.

So the patients who were enrolled in this study displayed a wide range of effectiveness with respect to Angelman’s syndrome. So this is a CGI rating of severity, which describes the clinician’s global impression of how severe the patient is. Now this is, also I have to say anchored for Angelman’s syndrome, so mildly effective means mild for Angelman’s syndrome, not mild for normal. And it avoids all the Angelman syndrome patients basically being six or seven as they are so severely effected in these domains. So you can see we have a range. For sleep and behavior, normal is actually normal, so it’s not a relatively appropriate scale, and we had a wide range of patients from slightly disturbed sleep to severely disturbed sleep. Behavior ranged. And then we had more consistency with these core features of communication, gross motor, fine motor, and then this is global impression. The reason for this is these are just symptoms that all the patients have. But there was some range in term of how much problem patients had, for instance walking or with communication.

So five patients total were treated to date. They were all treated at Rache only because we were open right before the COVID pandemic, and then no one else could get open during the COVID pandemic. So it turned out then, we had a serious adverse event and had to halt the study before any other sites got open. So it’s an 11-fold difference from the lowest to the highest doses tested. So cohort one went from 3.3 to 36mg, and one patient even got a bridging dose through the first four doses. And then got one bridging dose, which was the plan before the patients went into an open-label extension where they were treated every month. We had another patient get through, the other patient in cohort one went all the way through. Patients in cohort two got to 36mg, but then the SAE (safety adverse event) occurred and dosing had to stop. And then we enrolled one person in cohort three who just got the first dose, which was 20mg.

So the pharmacokinetics was pretty linear. These two lines down here are the first dose, the 3.3mg dose, these two in the middle are 10mg, and these two are the 20mg dose for patient 5. So it looks like that is pretty linear and consistent and the level falls off and is not detectable at all 28 days later when you get to the next dose.

So we did have some safety issues. The adverse events were transient ataxia and fatigue. It seemed dose related and primarily observed at 20mg or higher, there was variability between patients and how severe this way. It would start like 2 – 6 hours after the injection and then lasted a couple days, sometimes really just barely a day. The patients would be back to normal when it resolved and they would be able to eat and interact normally they were just more off-balance than they usually are. There were other mild events like headaches, childhood infections and UTI which we expect in any study population. And the patients all tolerated the LPs and anesthesia well.

So there was a serious adverse event that we had that was what ultimately stopped the study. This was presented as lower-extremity weakness that ascended from the foot. It looked a little bit like Guillian-Barre would look. So this turned out to be mild to moderate severity of lower extremity weakness. Three patients were weak and kind of grabbing onto things and had trouble stooping, and two were actually unable to walk, and they were the youngest patients. And this occurred about 6 – 30 days after the prior infusion, but probably represents a cumulative toxicity. Four patients had this after the highest dose, one patient had it after the second dose, and that was the second highest dose of 20mg, that was the smallest patient. Dosing was paused for all patients after the first case was observed. And this actually resolved on its own within 19 to 86 days. But the clinical improvement that we saw in the patient were sustained longer than the SAE.

So this is just an MRI of what the SAE looks like. It’s really an inflammatory polyradiculopathy and you see this inflammation along the spinal cord and then these little dots here are the inflammation on the infused scan of the nerve roots and the lumbar images. The CSF protein increased during this, kind of like Guillian-Barre, we treated these patients with IVIG and actually steroids to make sure we were doing anything possible for them at the time. And it did not turn out when the patient who had the more severe ataxia acutely after their infusions really did not correlate with the severity of this problem at all.

So we were planning to assess efficacy as exploratory endpoints and really looking at things that might assess different domains of Angelman’s syndrome and actually show a result. So for sleep we had an ActiGraph, a diary, and then an EEG during sleep. For seizures we had an EEG and a diary and there are a number of good EEG markers in Angelman’s syndrome we’re also looking at. For communication we had the ORCA which is really a good example of needing an outcome matched to the disease, and most of the standard communication measures patients kind of floor on, so this is a measure that was developed that takes into account patient symptoms and what families thought was meaningful in terms of an improvement. And so we were exploring the use of this in this trial. And then the Bayley-4, these are developmental, the Vineland-3 is an adaptive behavioral scale, and the Bayley is a developmental test for preschoolers and we need to use it in Angelman’s syndrome, cause that’s where their level of cognitive function is. Cognition was the Bayley-4 and the Vineland-3, and then we had an ActiMyo that was a device the patient wore on their ankle to measure walking and motor function, and then the Bayley-4 and the Vineland have motor domains. For behavior we have the ABC (aberrant behavior checklist) and the Vineland-3.

So this is just what happened with the CGI-I. So what is the CGI-I? The CGI-I is the clinician global impression of improvement and it’s a way of having the clinician rate the patient as if they were in clinic. Like when you put a patient on a medication, you try to decide, are they improved enough that I want to continue this medication. And so mildly improved you might think well, let’s wait longer and see if this is a good treatment or not. Much improved is usually meaningful improvement in at least two areas, or a huge improvement in one area, and that’s a kid in clinic you put on medication and say, I’m definitely gonna keep this kid on this medication cause it’s really helping them. And then for very much improved, it’s almost like, that’s like a seizure patient that’s already been five or six medicines and they are still having one seizure a day, and then you start them on a new medicine and their seizures go completely away and knock your socks off. I didn’t really think this could happen, kind of improvement. And then the same is for worsening. So these are the five patients in clusters, and the colors are the different domains that we were assessing. The red is the global, this green is fine motor, orange is gross motor, blue is communication and grey is behavior, and darker grey is sleep. And you can see there was improvement in multiple domains, really in every patients, and many of them were much improved or even very much improved. And I rated these kids and honestly, I was quite surprised at what I saw in this study. We had kids who were really making new gains that we just don’t expect to occur in Angelman’s syndrome. So just as a time course description, this was the SAE time course, which happened kind of after the main dosing occurred. And we started to see improvement in some of the young kids, even after the 3.3mg dose, and that was surprising cause that was supposed to be subtherapeutic. And then the SAE started after multiple doses and it went away, but we’re still seeing clinical improvement in the patients after the SAE disappeared.

So what kind of things did we see? Acquisition or words, patients who had no words started to say words. Acquisitions of signs and gestures, ability to use their augmentative communication device when they weren’t using it previously, ability to respond to their name, follow commands and stay focused on their Zoom school. And this is probably the only study in a developmental disorder that will ever be conducted without therapy and school because it was in the middle of COVID and they were in Zoom school, which they really can’t do basically. So everything that was happening in this study was happening without therapy. Ability to self-feed with a fork, point better, engage in physical activities like catching and throwing a ball, which they couldn’t do before. Improved gait and posture, able to go up hills, able to walk on sand, which is uneven, without falling, able to step on curbs, and they really just looked more stable, many of them, with a faster gait and they were kind of walking farther. And then the one kid that had horrible sleep actually started to sleep normally and the parents had three weeks of normal sleep for the first time in their lives since this kid was born. And they were like, we don’t even know what to do with ourselves. So there were dramatic areas. There was also improved behavior and social engagement and ability to play games. So there was improvement across multiple domains, which was really striking to me because I did not expect them to improve this fast, I thought well this is cognition, this is going to be a long study, we’re gonna have to wait a year to see anything, but we really did see actually see these things. And I actually have video of them.

So on the Bayley, when we tested them on the Bayley, when they came back in after the four, but in some cases two, or three, or one infusion because of the study stopping, there was more improvement, and this is the different domains plotted out in clusters. And there was the most improvement on receptive and expressive communication. And this is the development of Angelman’s syndrome with expressive communication and receptive communication. You can see the progress is very slow and even flat with expressive communication, they just don’t do any development after age two. So amount of improvement on growth scale values on the Bayley was really different from what Angelman’s syndrome ever does. We did see improvement in gross motor in one patient, and some of these areas of deterioration were actually because we reassessed the patients after they were in the middle of acute inflammatory polyradiculopathy so they weren’t walking, so they had lost some motor skills at that point. But they have all since recovered and are actually better than baseline with respect for gait in every patient, even still to the present at 7 months post our last dose.

This is just the Bayley showed very nicely in this patient, the deterioration related to the serious adverse event, we were able to detect that with the Bayley. And then this is the ORCA, that was developed for Angelman’s syndrome, and it shows a significant, they did a whole analysis on what was a clinically significant benefit and it’s more than a five point change, so three of the patients had more than a five point change, and the other two had nonsignificant changes. This is the EEG which is a really nice quantitative, objective measure, and we had two blinded epileptologists review the EEGs and it was randomized so they did not know which was the first one and which one was the one at the end of the study. So three patients had a decrease in this notched delta, which is a common, classical EEG finding of Angelman syndrome. Three of them had decrease in epileptiform discharges that go on in these patient’s EEG forms all the time, and two of them that are analyzed, we don’t have the other two, have a decrease in this delta power abnormality that we see in all the Angelman patients. So this is three EEG parameters that look like they are getting better and one of the patients couldn’t complete the second EEG because she was just a nightmare in the hospital, they were trying to do fourteen hour EEGs and we couldn’t get the leads on here.

This is the ActiMyo, which is that little thing around the ankle, and it measures gait. So you can see some of these patients really did start to increase the amount of walking they were doing. The gray is the normal control, range of low-to-high for normal controls, and you can see some of these patients really going up, especially with the 95th percentile differences. So this is a promising measure.

So in summary all five patients that we treated had clinical improvements observed, with a big mean CGI-AS change. All patients improved on at least two domains that were much improved or did better, and even had some things that were very much improved. The youngest two subjects improved the most and also had the most severe side effect. The improvements were confirmed on a variety of measures including patient diaries, the Bayley, the ORCA, the EEG, and potentially the ActiMyo, so all these things are moving in concerts that really looks like there is an effect here. Improvements began at the lowest two doses or 3.3mg and 10mg. The SAE occurred at the highest two doses that we used, and the most impaired were the young patients. So this really speaks back to the idea that if you go younger and younger, you may see earlier and better responses in younger patients with a neurodevelopmental disability. And if we went back to age zero, we might have a much greater opportunity to change the brain with something like this. So the SAE impact we can document on measures. The key questions really are can we lower the dose if we apply it repeatedly, can we accumulate enough drug to gain the improvements while avoiding the SAE? And are there adjustments in dosing administration that would reduce the local exposure at the lumbar nerves and improve delivery to the brain.

We are going to try and address these, GTX-102 appears to be clinically active at lower doses, and it seems more recent nonhuman primate studies seem to indicate that lower doses produce good pharmacodynamic effects in terms of reduction of the antisense RNA. We have more problems with the SAE at higher doses. And we need to get the drug away from the lumbosacral spinal nerves, because that’s where it is causing the toxicity. But we have seen clinical impact on communication and other measure that lasts at least 3-5 months, and I think has lasted longer than we expected. All of the kids have regressed a little bit, but not in every single thing, and some of them are still learning unexpected things. So there’s an effect on neuronal connection that is longer than we thought it would be, and a lot earlier than I thought it would be. The plan is to amend the protocol for dose and administration, lower the starting dose depending on the age of the individual, do an individualized dose titration where we increase the dose if the patient is not responding, but cap the dose at 16mg, and then modify the administration procedure to use a flush and Trendelenburg, which since the drug is isotonic, it is questionable how much the Trendelenburg will help, and I know that Frank likes to get the patients up and walking around to get the drug moving, but these kids are under anesthesia so we can’t do that. The protocols are at the FDA and hopefully soon they will let us restart the study so there is more to come. So that’s all I’m going to say, but this was really a striking study, this is really the most striking study I’ve ever seen. I literally watched day to day videos on my phone of a kid go from not being to able hold their breath, to gradually learn how to dive underwater and get stuff on the floor of the pool and swim across the pool. And that kind of motor learning, I have never seen before in anything. So this is pretty impressive and I’m really looking forward to all of the studies and to getting back going with this one.

Mustafa Sahin: Thank you, Liz and Frank. We are going to turn now to Dr. Heather Gray-Edwards from University of Massachusetts. She's going to talk about large animal sheep model Tay Sachs cognitive dysfunction.

Heather Gray-Edwards Thank you so much.  Can you guys see the screen okay?

Mustafa Sahin: Not yet.

Heather Gray-Edwards: Hold on one second. How about now?

Mustafa Sahin: We can see it now.

Heather Gray-Edwards: I really want to thank the organizers. I think this is a really great idea to have this workshop. I feel this is really important and I'm excited to share some of my experiences with you all.

So one of the first things to talk about, I feel like, is when should one consider a large animal model. Large animal models are certainly not necessary for every disease and there's a lot of examples where animal testing of rodents has gone to [word] and shown some pretty profound success but there are some instances where a large animal model is a good idea. When the mouse does not replicate the human disease, which we know can happen more than occasionally in mouse models, either a biochemical or phenotypic recapitulation. Sometimes you have requirements for tightly related and/or cell-specific expression that you are trying to target and using a larger more human like brain is often a good idea in those cases. An example also is if you have some unexplained toxicity and you need to dive deeper into what might be going on.

In the case of Tay Sachs disease it was the case where the mouse model didn't recapitulate the human disease. Mice with Tay-Sachs disease were completely asymptomatic, unless there were some instances of maybe some neuropath signs in older breeding females after many many litters is one instances, but largely the mouse model wasn't usable for this. This is because mice had a Sialidase enzyme that was able to cleave the storage product in GM2 ganglioside, and in that case we couldn't use the mouse model. As done for other diseases, what they ended up doing is making a double mutant for Sialidase and for HexA to study the disease, which is not ideal because now you're not just dealing with Tay-Sachs but you are dealing with another mutation as well. That is why we selected to do studies in Tay Sachs sheep and they were naturally occurring model so we could do that.

At one point in time we were all limited to naturally occurring models of human genetic diseases and that is no longer the case. This is data from my lab where we're generating large animal models of human genetic diseases and basically the steps are you collect embryos from these animals, using CRISPR technology, design a guide, select a guide, and you can edit these embryos and determine your editing efficiency in the embryos and once you have a desired editing efficiency you can plant these embryos into recipient dams and grow essentially your model of interest. So down here is an example of the type of mutations that we have been making which include deletions and/or point mutations that end up resulting in genes being non functional. We've had pretty good luck. It took us a long time optimizing this. We've had pretty good luck of late. And I think this is definitely an option for those diseases where it's necessary. And we are not the only ones doing this. There are companies doing it. There are other groups doing it, and there's a recent paper coming out for CLN1 where they need a CRISPR sheet that could be used for gene therapy testing. So we’re really excited to see that this is coming up and may be able to use for some of these more complicated neurodevelopmental psychiatric disorders we’re discussing today.

Tay Sachs disease is a disease of like many other diseases where there's different forms, infantile, juvenile and adult onset forms. For the case today we are going to focus on the adult onset forms of Tay-Sachs disease. Tay-Sachs is caused by a mutation in an enzyme hexosaminidase (Hex A) which is consisted of an alpha and a beta subunit. The combination, the isozyme that has both the alpha and beta subunit is the one that cleaves GM2 ganglioside, and GM2 ganglioside is the storage product in this particular disease that results in the nerve degeneration. Late onset Tay Sachs patients, their onset is usually early in adulthood. These patients will start off being misdiagnosed, they will be clumsy kids, have some interesting weakness in their quadriceps and triceps muscles, they will develop dysarthria and often misdiagnosed for a decade or more. They do develop very commonly psychosis and decreased cognition in addition to some other clinical phenotypes, but overall they do have a pretty decent quality of life. That makes gene therapy in these patients, you want to make sure that you're not going to make things worse for them because their quality of life is so much better as compared to some of the younger patients. We've focused really hard for treatment for late onset Tay Sachs trying to really characterize what we're doing in these animals and how can we make a therapeutic for them that will be considered to be safe and efficacious.

As I said, late onset Tay Sachs can be characterized as a neurodevelopmental psychiatric disease. There's a lot of reports in the literature of this. The disease can be completely debilitating where these patients are institutionalized with severe psychosis. It can also be mild in some of the cases where they can live with it. Pretty much in patients with late onset Tay Sachs disease psychiatric component is a large component of the disease. In addition to that, it's really important that these patients get diagnosed because a lot of the drugs that can be used to treat neuropsychiatric disease phenotypes actually can worsen the symptoms. A lot of times what happens is these patients get treated for that, they get worse and they have to come off the drug and they don't understand why. Definitely an area that falls into the category of neurodevelopmental psychiatric disease.

The late onset Tay Sachs or Tay Sachs in sheep does recapitulate late onset Tay Sachs disease to some degree. The sheep do live into sheep adulthood. Sheep reach sexual maturity at about five to six months of age so nine months of age they're considered to be adults. The one thing about Tay Sachs in sheep that was troubling was not all the sheep developed the same clinical signs which means some sheep would develop seizures like this one on the video shown here. In addition to that, some fetlock hyperextension, spontaneous knuckling as shown above here on the top left which is a proprioceptive deficit, basically saying they don’t know where their legs are in space. Some will develop tremors. The onset of which time they have these clinical signs also changes, which made it really challenging to evaluate whether or not your therapeutic was working because like shown before for some of the patients, they didn't experience some of the clinical signs you could say maybe they were never going to get that clinical sign. It was a challenge. And one of the things we did early on was cognitive testing, just to give you a quick video of what a normal sheep looks like here walking down the lane versus an untreated Tay Sachs sheep. You can appreciate the Tay Sachs sheep is slower. It has a truncal sway, with is consistent with cerebellar disease, it has a proprioceptive deficits which is consistent with some other areas localization. You can see it moves slower, it stumbles some when it tries to turn. Overall, these sheep are ambulatory up until near the endpoint, then they rapidly decline. That’s kind of just a sheep thing.

So we had a hard time determining whether or not our therapeutic was working initially. One of the things that I decided to try was maze testing in sheep. This came from a very obscure publication out of New Zealand where they did maze testing in some normal sheep to show cognition. I thought why not, give it a try. Some people said they were going to popcorn and come watch the Tay Sachs maze testing. Basically what we did is take a sheep near the entrance up at the top screen. The sheep goes in the entrance and the flock is located at the end of the maze where they cannot see it. They can hear and smell the sheep but not see it. You can see, this is me letting in a normal sheep into the maze. And it's never been in this maze before, you can see it walks through, tries to find its flock mates, eventually is able to navigate the maze and get through. You can kind of see at the end, there's its herd mates at the end. Sheep have this natural flocking where they want to be with their herd mates so they have this natural desire to want to navigate the maze. Then so if you look at the other video, you can see the Tay Sachs sheep navigating the maze. I hope you can appreciate this sheep is affected but is still ambulatory. It has some ataxia and some things in the first video that I showed you but it is able to ambulate and see the maze and walk through the maze. One thing that's important to note is that this sheep never figured out how to navigate the maze. It looked and looked and looked and looked. But this is a 15 minute long video. I'll skip to the high points. Continues to try to find its herd mates but is not able to do so. These sheep can hear. Their auditory responses are normal and they are visual at this time. It's not that they can't hear the flock.

Moving on, what do we see for this? The first time we did the maze testing, we did all of the animals together at one big lump. You can see here are the Tay Sachs animals versus the normal animals. Basically we hit the maze at one point during the study and the Tay Sachs are around six months of age and the normal sheep were age matched. I hope you can appreciate the normal sheep were able to navigate the maze pretty efficiently, and the Tay-Sachs sheep, while there were a couple of them that did okay, largely most of them were delayed in navigating the maze and some of them had a very hard time getting around. Interestingly enough when we put the Tay Sachs sheep in the maze starting at one month of age when they are asymptomatic and following to seven months of age, it appears that sheep are able to remember the maze and able to navigate it, albeit slightly slower, this may reflect their ability to remember how to get through it.

So this sort of testing has been expanded quite dramatically by a woman named Nadia Mitchell in New Zealand with CLN5 sheep. She took the maze and made it way better where she had alternating gates and different mazes to really challenge these animals. I hope you can appreciate in her gene therapy study she was able to show pretty profound improvement in some of the maze navigation of the treated animals.

So our gene therapy study, we tested two groups. We did intravenous gene therapy and also CSF delivery by either four sites so delivery into the CSF either by bilateral ICV injection combined with lumber intrathecal or just at the cisterna magna level to try to get an idea of how CSF delivery can be done and these were the cohort intravenous delivery. The dosing is equivalent roughly across the groups. You can appreciate here that it was a lot of virus that we had to make to do this study. What we found in these guys, to remind you of the Tay Sachs sheep up top, we'll play that video. And also the IV treated animals and the CSF treated animals at different time points. The IV treated animals on the left about 12 months of age and the other animals about 18 months of age. They are able to after a time navigate the maze more like normal than untreated Tay Sachs sheep. This very interesting thing about this is this pointed in time we weren't able to tell the difference between the groups and we learned a lot about maze testing early on. I'm really glad we decided to do this and look forward to applying things like this and even making it better with some GPS tracking and some more challenges and some other things to really learn more about how we can exploit natural animal behaviors to understand the disease phenotypes in those species. This is just a summary of the data showing that the IV group and the CSF group were improved over the untreated Tay Sachs group. That is an example of cognitive testing that can be done in sheep models of human genetic diseases.

Other things we look at are MRIs. MRIs are a great way to evaluate neurologic disease. I hope you can appreciate the untreated Tay Sachs sheep at about 9 months of age has a pretty different architecture than the normal sheep. There's an overall isointensity or overall gray appearance in the brain of the sheep compared to the normal. Then after treatment with the IV treatment or the CSF you see a normalization of those structures. We were able to perform MR Spectroscopy, which gives you information about neurochemistry in the brain. This is a very complicated figure and I don't mean to get too hung up on it. Basically, it can tell you about gliosis through myo-inositol, N-acetylaspartate and N-acetylglutamate which marks the neuroaxonal health of neurons in brain. Glycerol phosphorylcholine and phosphorylcholine inform a demyelination in creatine and phosphocreatine on metabolic rate and then glutamine and glutamate inform on the glutaminergic systems. In untreated Tay Sachs animals there is a drastic reduction in N-acetylaspartate (NAA) here, an increase in glycerol phosphorylcholine and phosphorylcholine which is indicative of demyelination and reduction in N-acetylaspartate and N-acetylglutamate and increase in creatine and a reduction in the glutamine and glutamate. What we do see after treatment in gene therapy, by the CSF route and not the IV route, we see normalization of a lot of these neurochemical markers which is really important to note that there does seem to be a pretty big difference in the neurochemistry. These MRI and MR Spectroscopy are clinically routinely done in patients and done in patients in ongoing Natural History studies for Tay Sachs disease getting ready for this clinical trial. We're hopeful that by performing these studies in the animal models it will better inform the clinicians that will be conducting future and clinical trials.

Other things that are important are biomarkers, fluid based biomarkers. This is cerebral spinal fluid from the Tay-Sachs sheep at their humane endpoint versus normal animals. This is GM2 ganglioside and the CSF are two different forms. What we see in the data is the animals treated by intravenous administration, there is a little bit of an intermediate of normal and untreated levels of GM2 ganglioside in the CSF. But interestingly after the CSF administration with the bilateral ICV, the cisterna magna and the lumbar intrathecal delivery combined, what we see there is a normalization of GM2 ganglioside in the CSF of these animals, which is encouraging. Then we look at the post-mortem assessments of these animals, what we do see a HexA activity is very low, and this is only the IV treated animals because the CF treated animals are all still ongoing. But the IV treated animals we see very little HexA activity in the brain. We see some in the cerebellum, some in the brain stem, and quite a bit more in the cervical and lumbar stenosis of the spinal cord. That also is corroborated largely with the GM2 ganglioside storage levels. These data suggest that IV delivery which is kind of been seen over and over again in animal models and otherwise that there is definitely a tropism with intravenous gene therapy for the spinal cord and caudal aspects of the brain but not so much the forebrain.

Overall we see quite a big dramatic increase in survival. The IV treated cohort has reached end point. Those animals are all done with the survival, about 19, 20 months of age. What you can see is with the CSF treated animals, by cisterna magna and also lumbar intrathecal, the combination therapy also is markedly improved. Begs the question, what kind of criteria do you really need to feel confident in a human, and if possible, some of the biomarkers and obviously survival is a great marker. A lot of people contribute to this work. Large animal work is always complicated. I just put this up there to thank all of my people and acknowledge my Tay Sachs lambs that have a pretty great quality of life, I must say. Thank you guys very much.

Guangping Gao: Thank you, Heather. Our Next Talk is from Guoping Feng from MIT.

Guoping Feng: Can you see me? Thank you. I would like to thank Mustafa and Guangping to organize such a wonderful workshop. I have been learning all day. I'm going to present a little bit of one of our main studies in the lab.

So with the large genetic studies it's become very clear that a synaptic dysfunction is a key mechanism or key pathology in ASD and many other brain disorders. My lab, like many other labs, are very interested in trying to understand not only the development and function of synapse but also how this function of synapse may contribute to neurodevelopmental and psychiatric disorders. One of the important protein synapse is a group of proteins we call scaffolding proteins. My lab did a lot of study on one of these scaffolding proteins called Shank3. Shank 3 is a scaffolding protein making connected with the PSD95 SAPAP proteins form a larger scaffold on the postsynaptic synapses. They play a very important role in the scaffold, in anchoring and trafficking of synaptic proteins, ion receptors and channels and through synapse also dynamically regulated function and also plasticity of both molecular plasticity and structural plasticity of synapses.

We, like many other labs, have studied the Shank3 extensively, because Shank3 mutation now is strong link to severe neurodevelopmental disorder called Phelan McDermid Syndrome but also sporadic ASD. So this is basically a monogenic cause of severe development disorder and ASD. So many labs including my lab and many other labs have been using mouse models and also some IPSL to study the function of the Shank3 but also how Shank3 mutations affect the neuron function at multiple levels, at the molecular level and behavioral level. All these studies in the past decade basically made it very clear that although this is a monogenic ASD model, however, there are many, many neurobiological defect in many different cell types and many different circuits. They all link to a variety of different behavior abnormalities. So this is only one model and there are hundreds of genes, either causative genes or risk genes for many different molecular defect.  So that provide a very big challenge for develop therapeutics. So we have been thinking like many others that there are several different levels of therapeutics we could develop for monogenic, focus on gene therapy, for others we could identify early cellular defect, early development and to prevent or minimize the circuit malformation but for some other much more polygenic we don't know the cause, then the end result is maybe we can identify the circuits and the line, debilitating symptoms and the modulate circuit to relieve the symptoms.

For gene therapy, one of our earlier questions was it's such a severe neurodevelopmental disorder, at what time and what development of time are these things reversible. So we develop a Shank3 mouse model to use genetic approach to see whether if we induce a Shank3 expression in adult or different stages, how much is reversible, at a molecular, synaptic and behavioral level.  We use tamoxifen-inducible CreER to induce expression in Shank3 in their native endogenous locus, we flip the gene so the mice will grow up as knockout and whenever you induce the Cre expression, you will have function and normalized expression without over expression.

I'm going to summarize that it turn out that if you let the mouse grow up to three or four months old, then you re-express, you induce the Shank3 expression, many things are reversible, quite a few things, including at a molecular, synapse, spine density in the striatum, also cortical-striatal synaptic neurotransmission, and behavior level at least in the pet grooming their skin off, but it will heal if you treat with Tamoxifen and hair will grow back. Surprisingly also social interaction in mice these are reversible. There are many things we found if you reintroduce and reexpress Shank3 in adult, many things you cannot reverse, including motor defect, anxiety like behavior, motor learning and also oversensitivity. There are many things that are not reversible in adult.  However, if we reinduce the Shank3 early enough, between postnatal P10 to P21, actually almost all the phenotype can be pre corrected or significantly improved. That tell us although they are adult, plasticity in this mouse model, in certain brain regions, mostly in the striatal-related circuits, there are critical development windows that you have to re-express Shank3 if you do gene therapy before that critical or window closed and will have the maximum benefit. So at least give hope there's a postnatal window. If we can define that, that may be dramatically beneficial to patients.

Maybe easiest the way to test is use gene placement. However, Shank3 is quite a bit 5.2KB in cDNA, so it's exceeding the package capacity for the AAV. We develop based on function window and many other people have published, we developed a mini gene and tried to keep as much we think important function of the Shank3. And using AAV, we can cast whether they go to synapse or restore synaptic functions in vitro, then we can test whether it would be functional in vivo. Most important thing of course it has to go to synapse in PSD, so we infuse with GFP and show if you transvect into culture neurons, these are cortico-striatal neurons and in the cortical neurons and cortico-striatal synapse, you can see the GFP-tagged Shank3 goes to the synapse, and it matches with the marker PSD95 very well. That's one good sign. Then we took advantage of our knockout mice and also advantage of the PHPB vector Capsid developed by [name] and Viviana Gradinaru at Cal tech and used systemic injection AAV expressed the GFP-tagged mini Shank3 into the mouse model. We tested different stages at P0, P7 and P28. Then we after the injection, we waited for two months to see whether we see a change at molecular, cellular and behavioral. What we found actually is that at a behavioral level, this just show you P zero but similar results from P7, we're still working on P28. At P7 it basically rescue all the behavior phenotypes we found and tested in the Shank3 mice, including social interaction deficits, anxiety like behavior, locomotion, repetitive behavior, motor learning and many other different things. If we use a slice phenology to look at whether this can also rescue the cortical synapse transmission, it's also very clear if we injecting P0 after all the behavior, this is usually four to six months later, you can see very clearly the induction of the expression of the mini shank actually rescue the cortical synaptic defect.

So that tell us at least the mouse model this approach, although it's a mini shank very small, it can appreciate or restore most of the function of Shank3 added synapse. The key question, if we want to really move forward and there are quite a few things to address. One is identify and characterize primate-specific Shank3 promoters and enhancers because smartphone 3 are not expressing all cell types, and also, now based on single cell data, we know Shank3 expression cell types in mice and monkeys are quite different. We also need to determine what is the latest environmental stage we can fully rescue to have a very significant beneficial effect in ideally mice and human primates because mouse development is much shorter and compressed, it's harder to extrapolate into the humans. Most importantly right now there's not an easy way to really getting AAV or anything into the brain, into all the neurons if we want to. In mice this works really well because of the [word] vector, but in humans, you have probably see some of the publication in non human primates and many other labs have tested them. Actually the infection rate is quite low with the AAV vector. We need to know what is the percentage correct in neuron you need to really have effective gene therapy particularly for this gene. Every gene will be very different, so this will be, since mouse brain is much easier to be infected, it's such a small, a macaque brain will be much more scalable to humans, so we're hoping to understand that. And in the end, we also needed to evaluate vectors which we benefit from many others, what is the best way to getting this protein, this gene into the different type of neurons in the brain.

So for all these purposes we have been working on generally large animal models. Because of the cognitive function and other things, we choose to use a non human primate as a model system to generally model and test all these possibilities. So we work on two different type of non human primate. One is macaque cyno and the other is marmoset. With the macaque, a group of expert in China, they were first to generate a macaque model and use CRISPR technology. We're very lucky we found a group of people very interested in helping us and working with us together.

This data I'm showing first generation model. So we give like to humans, we give activity watch wearing and monitor them for a week or two, take them off and download data to look at them.  And the blue here is the control. You can see the monkeys travel during the light cycle. You can see during the day, very active. During night they have very little activity, sleep really well. But the heterozygous Shank3, you already see, by the way, all humans are heterozygous so far, so you can see daily activity is reduced, but night increase, they have disrupted sleep. And homozygous, which we don't have anymore. These were generated by accident, not our intention, you can see the data is very low in the night activity and day activity is even smaller.  We think this will be much more stable or robust phenotypes that we can test the human rescue, all these things.  Of course, one of the most interesting things we are interesting in and probably many others are is whether they have social deficit, which is more or less mimicked human better than the mouse models. So this is the test that we usually do. We put into two different separate cages. The middle there's a divider. You can lower the divider and they can interact with each other. However, this monkey never see each other. For each control or mutant monkey, we basically use ten prober monkeys, all the probe monkeys are Y-type, to let them interact with the mutant, they never saw each other before. Then we average how this control our mutant monkey interact with the ten different probe monkeys. Monkeys sometimes have personality they don't like one monkey so if you use one probe monkey, your data may not be reliable. The probe monkey has a green collar on. This is one Y-type prober monkey with one of the Y-type controls, so this person is coming to release the divider, so the two monkeys can interact. They never saw each other before, so this, this one has the green collar. You can see they are interested in each other, they are following, they are engaged, you can tell. Sometimes they look at each other. This is a very typical Y-type monkey interactions. Next video is exactly the same except that the other monkey is the Shank3 monkey.

So first you can see this monkey has repetitive flipping, they have a lot of repetitive behavior. And this green is the prober monkey. So now this person coming to release the divider. You can see the prober monkey, and this mutant monkey goes to the other side. This mutant monkey is interested in something, looks around the environment, but it never look at that other monkey. So even very close, it never looks. Just completely ignore the presence of the other monkey. This is kind of social behaving and interaction we can quantify. We're hoping this is more close to understand this behavior help us to understand how the social interaction deficit generated in the brain. You can see, always look around, but very different than the control monkeys.

Because the brain structure is very similar to humans, so we can do functional MRI. In mice we can do slice physiology to look at synaptic defect. In humans in the future would be impossible but we can do function of Y. If can detect the defect with the function of Y, we can possibly do human to see whether you see this kind of thing in the human, whether gene therapy and [cross talk] this is resting state connectivity. You can see they are very different from Y type and mutant groups and some regions very highly dramatically reduced functional connectivity but sensory cortex is highly increased. These are things we now try to see with the gene therapy in the future you can rescue.

And at MIT we also have large effort to use marmoset as a model. Our hope is we can develop a large set of assays for gene testing and drug testing that touch screen based cognitive testing, neurophysiological EEGs and MRIs. This can help us test the gene therapy I mentioned with efficacy or the effective age and dose and also testing in the future circuits based on drug candidate. We don't know whether monkey will be a better model but we're hoping it's a better model for translational. It will take a long time to prove that. It's our hope and hypothesis. I'm going to stop here. This is a very large collaboration with many, many labs at MIT, Broad, Harvard and NIH like Jim Pickel, Alphonso Silva in Japan, Okano, Dr. Sasacky, also a group of people, our colleagues in China. I'm also grateful to all the different donors in supporting us and also now from Brain Research Foundation support our effort. Thank you.

Mustafa Sahin: We can stop sharing and go to the questions now. I want to thank all the speakers. As the organizing committee we were aiming for a survey of the state of the art in terms of gene based therapies in the central nervous system. I think this was a beautiful survey from each of the different diverse perspectives.

I would urge the attendees to keep typing their questions in the Q&A so we can pick them. There are some questions and we'll start with those. I will try to address members of the panel for the various questions.

For Guangping, there was a question about AAV for astrocytes, they are interested in astrocytes in the brain and going to use an AAV vector GFAP promoter, the person is asking I wonder which AAV vector, AAV5, 9, other vectors could be used for a better transfection rate?

Guangping Gao: This is for Guangping or Guoping?

Mustafa Sahin: I think for you.

Guangping Gao: Thank you. Yeah, that's actually very interesting target. I must tell you, you will be happy, most are AAV that can cross blood brain barrier and they will first hit astrocytes. Many AAV, such as AA9, AAV8, all those ones can transduce astrocytes quite efficiently, particularly when you use GFAP promoters and that will be quite efficient to accomplish transduction. For diseases for such as Alexander disease and others that affect astrocytes and many AAV vectors can accomplish that.

Mustafa Sahin: Great. A couple of questions for PJ Brooks. One is, how to know about different gene constructions available at the NIH, and another question is what the process will be for selecting genetic disorders for inclusion in the Bespoke Gene Therapy Program. PJ? Is PJ available? I think you're on mute.

PJ Brooks: Yeah, I'm not sure I understand the first question. I think if you're talking about specific constructs made by different investigators, it would probably be better to direct the question to them. I think for the second question, what we're anticipating at the present time is some sort of funding opportunity announcement would be asking people to propose their specific disease for the BGTC. That's what we're thinking at the moment. I do want to just make clear too though that we're still in the final stages of fundraising for this. So this is of course dependent on getting the appropriate funding.

Mustafa Sahin: All right. Thank you. There was a couple of questions about the timing of treatment. Some of them I think have been addressed online, but I want to raise them again. One of the questions was about sensitive periods, early intervention is likely to be more effective.  However I wonder if there might be partial improvement even when treatment is provided to adults. Given the severity of some patients with neurodevelopmental disorders even partial cognitive improvement might make a large difference in the activities of daily living. You already wrote a response to this online  I was wondering if you could expand on that.

Elizabeth Berry-Kravitz: Yeah, we have been thinking about this issue in fragile X for 20 years now and I'm not sure we got all the answers in 20 years or even any of them. I think it stands to reason that we are going to see better effects in younger patients. That's the whole premise of our Neuro X trial and Fragile X trial where we're treating the kids with McGluR blocker that Mustafa mentioned. There are three to six year olds and doing language intervention and try to push learning and see if the drug helps to push learning when you work on learning. We had such a quick effect in this Angelman syndrome study that it makes me think there's some kind of acute effect in UBEA3A that's promoting wiring maybe on a shorter time frame than we really expected. Given that, it may be that we see effects in adults. We did have two kids that were like five and six and the other ones were 11 to 15. It was very clear that the little kids showed effects much more rapidly than the older ones did, but by four or five months, we saw stuff in everyone.  I would think you could get benefits in adults. And sometimes the benefit in adults may not be as much cognitive but might be like these adults have horrible problems with things like non epileptic myoclonus and we have nothing to treat that, and if it would just change the disease enough that we could get rid of some of the ancillary symptoms and maybe make them a little bit more interactive, that would be a huge benefit. So I would not rule out treating adults. I would just not maybe expect as much as fast but I do think that the book is still waiting to be written for diseases like Angelman syndrome and Fragile X and TS and all the other diseases we work with.

Mustafa Sahin: Are there any other panelists that want to add on?

Jerry Mendell: In the muscular dystrophy world, we've treated 13 and 14 year olds with limb girdle dystrophy and had a good response. To some extent that may depend on the particular disease, the evolution of the disease in terms of muscle loss, but if we have good preservation, a good substrate, I think we can get a response. The tradeoff is we'll have more adverse events in the older patients because of the higher viral load.

Mustafa Sahin: Any other perspectives on this?

Carsten Bönemann: I just wanted to reiterate what Jerry said. I think the integrity of the substrate sounds simplistic but is absolutely crucial in that. In a neurodegenerative disease where you lose neurons and they are not recoverable, so when you have a degenerative disease like SMA it really is absolutely critical to hit that window of opportunity where the substrate is still there. In a neurodevelopmental disease it becomes much more interesting, if you will, much more nuanced, because what is the substrate, yes neurons are the substrate but it's also synaptic chemistry and function. It will be interesting to figure out and correlate how you can study the integrity of the substrate for rescue and therefore determine what your window of opportunity is. If the substrate is there, I would argue with the others that the window of opportunity may be much wider than we assume from these very neurodegenerative diseases.

Mustafa Sahin: Thank you. There are a couple of questions about AAV vectors. Maybe Bev or Guangping, you could take these questions. From Sam Young to the panel. Is NIMH solely interested in AAV vectors. There are certain genetic disorders that encode large cDNAs that the AAVs may be effective as a treatment while split AAV vectors have been proposed and discovered 20 years ago it appears that efficacy is limited.

Beverley Davidson: I can go ahead and start. I can't enable my video. They've cut me off at the office. But with regards to the large gene products, for some of them I think there's been some developments in the NTN space, there have been developments in something called the over packed AAV, and also some AAV or other viral derivatives that are being developed to package larger genomes than the current five-ish KB genome. In addition to that, you bring up a good point. There's been quite a bit of development in lentiviral viral space, right now I think biodistribution is just not as broad as what we see with the mini AAVs, it's also a little bit harder to pseudotype them and get the alternative tropisms that one gets by evolving or selecting new AAV serotypes.

Guangping Gao: That's perfect, Beverly.  That addressed the question very well.  Thank you.

Mustafa Sahin: There's a question to Guoping Feng. Could you please elaborate on what you meant by circuit based drug candidates?

Guoping Feng: So the neuroscience in the last 15 years thinks the optogenetics and many other new technologies have a lot of what are neurocircuits are responsible for what behavior. So with optogenetics and other technologies. Now in adult when development is done, let's say you have sleeping problem. We now know for example, what neurocircuits control sleep. Now we have single cell ionic sync technology, you can basically look at what are particular circuits and what cell types are controlling this circuit's function and you can identify druggable targets. So the future drug for the circuit modulation will be much more specific. They are specific targeting sleep.  So for example, [word] sleep is GABA agonist, it's basically suppressing whole brain activity by activating inhibitory neurons. But in the future, you have a lot of side effect, if you overdose, you will die. That's one of the major issues. The future will be not really putting to sleep. When you sleep, it will improve your sleep quality, help you falling to sleep and sleep quality. When you wake up, the same circuits, for example, also improve attention. So you basically have a drug that will during the night it will help you sleep and during the day it will help up increase the attention and more focus. That's a future drug thinking it will be a very big area to develop actually.

Mustafa Sahin: Thank you. The same question, asking to panelists, could you speculate on the utility of gene therapy for polygenic less heritable psychiatric disorders such as major depression? Perhaps via targeting epigenomic modifications resulting from chronic stress and abuse. Anyone want to take this thread?

Guangping Gao: You or Beverly can do this well.

Beverly Davidson: Go ahead, Mustafa.  I'm still trying to get someone to unlock my video.

Mustafa Sahin: Yeah, I think one of the thoughts on our NIMH workshop on rare diseases was that the rare diseases could provide insight to disorders that are not Mendellian that might have small effect common genetic effects. I think those are going to be harder to address as a first pass, but if we have successes in these single gene disorders that will enable us to find a timing, enable us to find the groups of neurons or parts of the brain to focus on and potentially stimulate or change gene expression in those areas successfully, then future studies, multiple geneses are involved or even non genetic psychiatric mental health disorders might be addressable going forward. I see that as version 2.0 or 3.0. I would love to hear from the rest of the panel. Huda, would you like to add on.

Huda Zoghbi: I think everything you said was true, but I also would say for some of the genes involved in childhood neurodevelopmental and autism like disorders, milder mutations, much milder, that even some of them occur in the general population may become actually the predisposing factor for later onset psychiatric disease. It could be that two hypermorphic mutations, two different genes that are critical for neurodevelopmental disorders but much mild together might give someone susceptibility to schizophrenia, whereas in the general population, without an environmental trigger or maybe only one of those alleles, you may not see it. My prediction is we are going to see a lot more of that. That's harder to nail so we need much more robust functions to nail such mutations.

Mustafa Sahin: I can also pick on Joe Buxbaum to see if he has any other additional thoughts on this from his collaborative studies.

Joe Buxbaum: About common variation?

Mustafa Sahin: Yeah, or diseases that are not monogenic, per se.

Joe Buxbaum: I think we all agree that we're somewhat far away from that at this point. The more severe the disorder, the more likely we are going to take profound interventional steps like intrathecal injections, so obviously these rare disorders that are quite severe are a natural place to start. As you go to common variation and variants of weak effect, in the polygenic model there's hundreds of variants working together and there's not an obvious way at this point to really understand how we can simultaneously manipulate all these variants especially because in each person the number of interacting variants, the specific variants are very different. There's interesting work by Kristin Bernan looking at some of the top top variants in schizophrenia in stem cells and showing there's some kind of epistatic interaction. But that's choosing variants that are really unusual in the sense they have a reasonable effect size and more importantly there's a very clear impact on the polymorphism on expression of a single protein so you can make a one to one prediction on this risk variant will change the expression of that protein. And for most polygenic loci we don't have that. I could go on and on.

Mustafa Sahin: I'm sure you can, Joe. The follow up question is from Brandon Lee. How important is an individual's background of common genetic variance determine how effects of rare genetic variants are manifested?

[PJ Brooks] I think it's a great question. I'm not sure we have a good answer for it.  The idea for modifier genes is something of great interest. But I think difficult to study in many cases because of small sample sizes.

Joe Buxbaum: Although I will say that we do have a pretty good estimate for some of the most significant genes that we've already talked about, that most of the risk is attributed to the rare variant, right? The common variation risk is in these disorders is actually quite a bit smaller than idiopathic autism, as an example, and it might be much more involved in what is the actual full manifestation of the disorder rather than the actual presence or absence of the disorder, if that makes sense. For 20qN 1 there's some very elegant studies that show that the common variation might increase or decrease by some amount, just like with the regular population. Your risk for schizophrenia but most of the risk is attributed to 20qN 1. We have exactly the same thing with other disorders with genes and major effect.

Mustafa Sahin: Any additional comments on that? Liz, there's a question on many of these disorders Rett, Tay Sachs, Angelman, is there a sense of percent transduction of various cell types, neurons, etc. be required the show effect?

Elizabeth Berry-Kravitz: Frank might know that better because they've done a lot of studies on where the ASOs go and how many are in the brain. I don't think we know that. In Angelmans syndrome we know that the patients who don't have deletions and have two copies of UBE3A and therefore might have a little more leak through stuff or might not have the other genes deleted are higher functioning than the patients who have the deletions. They are these mosaics where we sort of know that there's, you do see higher function with some UBE3A. So you don't need to get it to 100% to see improvement. And I don't think we totally know what that curve looks exactly like. There's a thinking that patients who have some UBE3A, like even 30 to 50%, that would be a huge difference between them and a person with Angelman’s syndrome. You don't have to get it into every cell, maybe. And we maybe don't have to have completely normal levels in every cell, which is a different thing. There's how many cells you get it in and how well it makes the protein in each cell. I don't think we have a good idea of what ASOs are doing with either of those. But the good thing is we might not have to max it out.

Frank Bennett: I think each disease is going to be unique in that regard. There may be some where you want to get more even distribution. The other ones, as you describe, a partial effect can have a very big effect size on the patient. I still don't see there's one answer for everything there.

Elizabeth Berry-Kravitz: Yeah.

Heather Gray-Edwards: I also think there's a significant component to think about with delivery. You also think about your disease, where is your disease and how do you need to best treat it. That's not a one size fits all either. Specifically, for like late onset Tay Sachs, we know that the patients most effected in the cerebellum and the spinal cord. Based on that knowledge, we can tailor a delivery method to try and target those cells. With the response to the percent of cells transduced, I think in the deep brain structure it's very low in the AAV therapy but we see quite a few in some of the cortical structures after CSF delivery, so it doesn't have to be all of them, especially if it's a secreted protein. Definitely the cells can help each other out.

Mustafa Sahin: There's a couple of questions, and maybe you and Frank can address these together. In retrospect, was there any signal that an inflammatory polyradiculopathy would develop elevated CSF protein for cells?

Elizabeth Berry-Kravitz: Yeah, I think I had one slide, the CSF protein. It turned out that the CSF protein did go up a touch. It started at 15 and went up to 25 or 28 which is in the normal range. We didn't make too much of that especially since there are numerous papers, there’s numerous evidence that CSF protein does go up in other ASO trials, so we weren't really that concerned about that. But then that was the harbinger of the next LP having a protein of 150 or 200, which was clearly way abnormal. But the patients didn't have clinical symptoms yet but then they did develop symptoms shortly after they got the high proteins. So the little kids seemed to get to very high proteins before they developed symptoms. The oldest two patients actually had some clinical symptoms before we could appreciably see their CSF protein going up. There was something going on there. The protein is a good marker but it's a good marker when you see it.  It's not a good marker for everyone. There were definitely, and initially, the first patient was actually diagnosed Gambray/Guillain-Barre.  Because of the high protein and the radiculopathy but it looked different from Gambray over the time because she was the only one with dropped reflections. It was clearly a different entity from Gambray. It only ascended in the legs. We never saw any symptoms referable to respiratory or arm function or anything. We put them in the hospital and watched them because we were pretty worried. We did not see anything above the hips.

Frank Bennett: I would add to that, that the CSF protein is a blunt instrument. There are multiple ways to get an increase in CSF protein. We've seen it in other trials but yet don't see the radiculopathy or other signs of inflammation. In some pre clinical studies we'll see CSF proteins go up a cell count so there's some going on. I caution that CSF protein is something to pay attention to, but it's a pretty blunt instrument and may not, have a mechanistic standpoint, all point to the same direction.

Mustafa Sahin: Just to follow up on that, you had I think a comment about whether the side effects could be related to sequence or the chemistry. Could you expand on that a little bit more?

Frank Bennett: Yeah, we've tested hundreds of thousands of ASOs in pre clinical species as part of our screening process. We have found that both the chemistry that's used in the ASO as well as the sequence of the ASO can cause, particularly the acute transient effects that Liz was talking about. We could separate those out so you could screen out those kinds of activities. The inflammation, again, there is sequence dependencies to it. A lot of ASOs will cause pro-inflammatory effects but they are dose dependent and some sequences and chemistries are sort of on the bell shape curve, some are on the left side of the curve and similar on the right as far as those that cause these effects. So we try to identify ASOs that are on the right side of the curve where it takes extremely high doses to start producing the inflammatory effects that were described. And that is clearly both sequence and chemistry. So with the same chemistry, if you change the sequence, that will go away. If you have the same sequence that causes it, you can use chemistry and make it go away. Both of them seem to have an effect.

Mustafa Sahin: Are they predictable or some way you could model them?

Elizabeth Berry-Kravitz: What they did was they used genetics gave 10 times the dose to the non human primate model and it had no signs of this until you got to humongous doses. So they try to model it in nonhuman primates, but we think Angelman’s syndrome may be a special case because they have less myelin, so there may be protection there, from having normal amounts of myelin. I mean, we don’t know that, it certainly seems like it could be a susceptibility factor.

Frank Bennett: I don't want to get into argument, but we have looked in various diseases of myelin deposition and don't really seem to see an enhanced sensitivity in those diseases. These are all mouse models, so it's not the same. I just throw that out for consideration.

Mustafa Sahin: And is it species specific? That sounds like, from Liz's comment, there might be different inflammatory responses in different species?

Frank Bennett: There can be. Definitely can be. Generally at least the acute effects that Liz was describing, that seems to translate across mice, rats and monkeys. We haven't, so that effect seems to be predictable based on species. Some of the inflammatory effects are not predictable because the innate immune system responds differently in mouse and humans or mouse and primates. That's a little bit more trial and error.

Mustafa Sahin: Guangping and Bev, if you have other questions you want to pick on, please go ahead. There's one question about lentivirus. The use of lentivirus for CNS.

Beverly Davidson: Sure. That is being used for CNS. Most of the work for CNS gene therapy using lentiviruses have been done with ex vivo approaches where they place the gene product, and the example of this is the gene therapy trials for metachromatic leukodystrophy. They perform ex vivo gene transfer in stem cells and reimplant those cells into children after they partially myeloablate and there's very good reconstitution and protection from disease onset in those children. That's the most that lentivirus has been used in a clinical trial. It's also being used for direct brain injection. But as I mentioned earlier, it just doesn't quite from the broad biodistribution that AAVs provide and it's also harder to develop for cell type specific targeting. As we know, it can handle a larger payload, so advancements in lentiviral manufacturing and lentiviral technology would certainly be helpful in the field.

Mustafa Sahin: One other thing I want to pick up on was you mentioned use of small molecules to reduce or regulate expression. There were other talks about using microRNAs. I was wondering if we could start a discussion on the use of those technologies and advances in the field.

Beverly Davidson: Yeah, we and others, and as you saw presented here, you know Guangping and Steve both presented ways to regulate transcript expression using microRNA binding sites first presented by Brian Brown in the late 2000s, actually in the context of a lentivirus, funny enough. I think those are wonderful approaches, the microRNA based approaches. For expressing products that might be immunogenic, you may need to actually keep the regulation or expression of that protein to a minimum or even for only as long as you need it. A good case in point is the expression of editing machinery. You certainly don't need editing machinery to be in cells for the life of that cell. You consider the life of a healthy neuron, that can be quite some time. I think the regulated expression system where you would take a drug, turn it on, do the editing, let the drug wash out, which can happen in a matter of hours to days. Then you're essentially done with it and you'll have no more expression even though the AAV payload will still be there, you'll have no consequence of prolonged expression of an immunogenetic protein, so, there’s promoters to regulate expression, I think using a regulatable switch … [audio cut out] we were shown that the rapamycin induced constructs is that this doesn't include anything foreign. And again, you don't want to have to express a foreign depressor and turn off that foreign depressor. It's replacing something immunogenic with something immunogenic. Using splicing as your control mechanism I think provides a new opportunity for regulation. Maybe Steven or Guangping, you want to comment further.

Guangping Gao: I thought what you mention is very interesting, it's this new riboswitch strategy. It may have a great impact in gene regulation, because as you said, there's no floor in protein. Just small ASOs can accomplish induction of pol II transcripts.

Beverly Davidson: Right. The ASO approach is a little different because as you can see from Liz's data and Huda's data, the wash out is months, not hours or days. The regulation is not quite as on and off. It's more on and then slowly waning as opposed to drug induced responses.

PJ Brooks: Actually, could I comment on that one, Mustafa? At the risk of offending my AAV generating colleagues, I do think for genome editing one of the technologies that's of most interest would be to deliver these, I would say to deliver the merge RNAs and coding the editors along with the guide RNAs with nanoparticles. Then you get expression of the editor. And then it doesn't sting and it's gone.

Beverly Davidson: [cross-talk] Absolutely. I think, yeah. Completely agree.

Guangping Gao: No offending. Completely cool.

PJ Brooks: Okay. I think right now, if I needed to do gene therapy on somebody, I think AAV for many diseases really is obviously attractive. We have many efforts going on to support it, but I think in the future that may be the way to go.

Beverly Davidson: Yeah, I think we all see that five or 10 years down the road when a nanoparticle can provide the sort of spectrum of coverage that we've getting from the AAVs, I agree. Right now it's just not quite there.

PJ Brooks: Yeah.

Steven Gray: And I wanted to jump in and follow up on this discussion when we talk about gene regulation because there's two things to think about with these dose sensitive genes. We're never going to get broad to completely even distribution, we're going to have hotspots and low areas of transduction. Any kind of global drug-based or ASO-based regulation of a transgene, if it's raising all of them up or all of them down in every cell to the same extent, then you can still end up having over expression issues due to the hotspots. That was where our attempt right now, that we're excited about but are still exploring, is really to have a true feedback mechanism of regulation on a cell by cell basis. The other thing to that is I was struck with Dr. Zoghbi's talk about ASO for MEC duplication syndrome, that could end up being viewed like an emergency antidote if there was a gene transfer study that basically got it wrong and induced an over expression phenotype in the Rett girls. A lot of these developmental genes have duplication syndromes. So if they were ASO or knock down approaches being developed for the duplication, this could have a secondary benefit of being a fail-safe against a bad gene therapy.

Huda Zoghbi: I agree 100%.  And I think the other thing, Steve, by the way I should say I really appreciate the thoroughness with which you have done this study on Rett. Thank you. I think we also should consider modest goals, right? That's why I showed the last bit of data about the promoters and enhancers where even it changed by just a little bit so you may not leave the toxicity but get enough benefit to take off the worst symptoms, you'll end up with slightly more social anxiety like symptoms. These are the two things really. Because most of these neurodevelopmental genes are dose sensitive. For many of them we see deletions and duplications. We are going to have to think about this in a much more modest way or strategies.

Mustafa Sahin: This may be a good place to stop. We've covered I think most of the questions on the Q&A. I want to, on behalf of all the organizers and chairs, thank all the speakers. I think this was a stimulating day, definitely beginning of many more conversations coming out of this. I will turn it back to Geetha for some closing comments.

Geetha Sentil: Thank you, Mustafa and Guangping and the chairs of the workshop and the speakers. It was an awesome, really enlightening, informative talks. We learned a lot. I learned a lot. Thank you for your time and we have a day two meeting on January 29. This is a closed meeting so we'll post a summary of that meeting following the discussions we have on 29th. Thank you everyone for joining and no one has any final comments from any of the chairs and committee members, we are ready to adjourn.