Webinar: Webinar: Autophagy & Neurodegeneration with David Rubinsztein

Webinar Transcript:

Ray Chan: Hello, and welcome to the webinar today entitled Autophagy and Neurodegeneration.

My name is Ray Chan, I’m a product manager of Bio-Techne, and I’ll be your moderator for today’s event. I’d like to thank everyone for attending today, and I’m happy to welcome our featured speaker, Professor David Rubinsztein.

This webinar is supported by Bio-Techne. Bio-Techne brings together the prestigious life science research brands R&D Systems, Novus Biologicals, Tocris Bioscience, ProteinSimple, and Advanced Cell Diagnostics to provide the scientific research community with a comprehensive and well-classed portfolio of antibodies, proteins, small molecules, assays, and instruments. Novus is an industry leader in the development of autophagy research tools.

Before we begin, I would like to briefly highlight the range of autophagy research tools available from Bio-Techne.

Bio-Techne provides a wide range of antibodies, proteins and enzymes, small molecule autophagy inhibitors and activators, and autophagy-inducing peptides. You can explore more with our interactive pathway for autophagy research in the online Novus pathway research pages. Click on the key targets and interactive pathway to discover available products.

Novus Biological’s knockout validated LC3B antibody NB100-2220 is the gold standard to monitor autophagosome information and autophagy induction. With over 770 citations it has been recognized by CiteAb as the most cited antibody for autophagy research, and it has been extensively validated by leading autophagy research groups around the world. Bio-Techne antibodies have been extensively validated with different validation pillars, including genetic, biological, orthogonal, and independent antibody strategies.

Novus has released three new highly specific and sensitive rabbit recombinant monoclonal antibodies. These have been validated by both CRISPR knockout and biological validation strategies, and across a wide range of applications.

In the upper right-hand corner of the screen, you will see a tab called Resources. In that tab you’ll find extra content that relates to today’s webinar. An autophagy handbook is available from Novus that gives an overview of the key molecular players and regulatory mechanisms, as well as helpful protocols and assays to measure autophagy.

This white paper from Novus provides an overview of autophagy in neurodegenerative diseases, focusing on dysfunctional autophagy and the key molecular factors that link the two.

In this highly cited and detailed review, Menzies and colleagues discuss the importance of autophagy function for brain health, outlining connections between autophagy dysfunction and neurodegenerative disorders. The potential for autophagy as a therapeutic strategy is discussed.

Before handing over control to our speaker, I have just a few housekeeping items to address. Throughout the webinar we invite you to submit any questions you have for David using the Ask A Question box just below the presentation screen. These questions will be asked of David directly following the presentation.

And now I’d like to introduce our speaker for today, Professor David Rubinsztein. David is a UK Dementia Research Institute Professor, Professor of Molecular Neurogenetics at the University of Cambridge, Deputy Director of the Cambridge Institute for Medical Research, and Academic Lead of the Alzheimer’s Research Cambridge Drug Discovery Institute.

He has been a leader in the field of autophagy particularly in the context of neurodegenerative diseases. His laboratory has pioneered the strategy of autophagy upregulation as a possible therapeutic approach in various neurodegenerative diseases, and he has identified drugs and novel pathways that may be exploited. He has made key contributions to illuminating the relevance of autophagy defects as a disease mechanism and to basic cell biology.

His laboratory has also identified drug wall pathways independent of autophagy that may be relevant to diseases caused by aggregate-prone proteins. These insights open novel avenues for developing potential therapies.

David has been elected as a Fellow of the Royal Society, the Academy of Medical Sciences, and as an EMBO member. He has won numerous awards for his contributions, most recently the Roger de Spoelberch Prize 2018, recognizing his work in autophagy and neurodegeneration.

And with that, I’d like to hand over control to David.

David Rubinsztein: Thanks, Ray. This afternoon I’m going to tell you about the autophagy-lysosome pathway in neurodegenerative diseases.

This slide shows the brain of somebody who’s died of Huntington’s disease compared to an aged-back normal brain. You can see in the Huntington’s brain there’s extensive loss of the striatum giving rise to these large ventricles, as well as marked loss of the cortex compared to the normal brain. These pathological changes give rise to the hallmark features of Huntington’s disease – abnormal movements, cognitive deterioration, and psychiatric symptomatology.

One of the dreadful features of this autosomal-dominant disease is that it has a median age of onset around 40, and that’s when has the situation where many people find out they’ve got the diagnosis of Huntington’s disease after they’ve had children, and this is indeed what you see in this slide. This lady has Huntington’s disease, and this is her daughter. Her daughter, unfortunately, carries the mutant huntingtin chromosome, however at the moment she’s entirely normal; she’s a Cambridge graduate and an Olympic athlete, so she’s actually a bit more than normal.

One of our key objectives is to see how we can find ways to buy this young lady as many years of normal disease-free life as possible, so can we find ways of delaying the onset of the disease. In addition, we’re keen to develop strategies that might ameliorate the condition in her mother.

We work on Huntington’s disease not only because it’s a terrible condition in its own right, but because it shares key features with most neurodegenerative diseases that afflict people, and that is the formation of inclusions within neurons comprising the aggregate-prone protein particular to that disease. So here you see these inclusions or aggregates within the neurons in Huntington’s disease, you see this with tau in Alzheimer’s disease, you see it with prions in prion diseases, you see it with α-synuclein in Parkinson’s disease, various proteins in motor neuron disease.

I like to make the analogy that these aggregate-prone proteins, these depositions, are like the rubbish accumulating when it hasn’t been cleared properly in Naples. The rubbish itself is poisonous and can cause traffic jams, and you can get the same in neurons if you get the accumulation of these proteins in axons. And when a situation where the inclusion is larger than the diameter of an axon, it will clearly block the traffic. That’s one of our major objectives over the last two decades or so is to see if we can find ways of enhancing the clearance of the mutant protein relative to the wild-type counterpart.

When we started working on this problem, we thought about the ubiquitin proteasome pathway. In this pathway target proteins are ubiquitinated by a series of enzymes that attach the small protein ubiquitin to lysines on the substrate. The ubiquitins are themselves ubiquitinated, leading to a ladder of ubiquitins that serve as a recognition signal to get the target protein to the proteasome for degradation. At the proteasome the protein needs to be deubiquitinated, unfolded, and monomeric to thread through the narrow entrance of the proteasome. In the proteasome it’s degraded into peptides which are further degraded into the cytoplasm or nucleus, turning into amino acid by peptidases.

One of the constraints of degradation through this route is the narrow entrance of the proteasome, which means that this pathway can only degrade monomeric forms of proteins and not higher order oligomeric structures that occur with some of the population of the proteins that are causing these neurodegenerative diseases.

For this reason we’ve looked at the autophagy-lysosome pathway. In this pathway the first morphologically recognizable component is this cup-shaped double membrane structure called a phagophore. These form more or less randomly in the cytoplasm of most mammalian cells, although this process is more polarized in neurons. After the edges of the phagophore are extended and fused, you have an autophagosome which has engulfed a portion of cytoplasm. The autophagosomes are then trafficked along microtubules to the microtubule organizing center of the cell where the lysosomes are concentrated to facilitate autophagosome-lysosome fusion, after which the lysosomal hydrolases degrade the autophagic contents.

This slide shows a cell with the microtubules in red and the autophagosomes in green, and these will be moving in this direction, in a net sense, towards the microtubule organizing center here where the lysosomes are concentrated.

On the previous slide, there were question marks upstream of the phagophore, one of the key questions in the field relates to understanding the membrane origins of the phagophores. Many membrane origins likely operate including the endoplasmic reticulum, endoplasmic reticulum mitochondrial contact sites, mitochondria of the Golgi, the plasma membrane, etc. Recently, our data suggests that an important component for generating the autophagosomes is the recycling endosomes. Indeed, our data suggest that most of the membrane of the autophagosome actually evolves from a portion of the recycling endosomes.

This slide shows RAB11A, a marker recycling endosomes, in green, and autophagosomes in red marked by LC3, and you can see significant overlap. This is a whole mount electron micrograph showing, again, the recycling endosomes in green, and the emerging autophagosome in red.

The reason we’ve worked so much on trying to understand how autophagosomes are formed is that in the early years of the century our laboratory discovered that autophagosomes were important rubbish tracts in the cells for getting rid of aggregate-prone proteins. We first did experiments in cell-based models of Huntington’s disease and discovered that if we decreased the formation of autophagosomes or impaired the fusion of autophagosomes with lysosomes, then one retarded the clearance of the aggregate precursors you see in such cell models of Huntington’s disease, and in doing so we increased the accumulation of both soluble and aggregated species and enhanced toxicity. We’ve shown the same subsequently in Drosophila, zebrafish, and mouse models.

Perhaps more importantly, we found that if we enhanced autophagy, we had beneficial effects. At the time we did the experiments, the only drugs we knew about that were used chronically in people that were predicted to induce autophagy were rapamycins that act by inhibiting the mammalian target of rapamycin complex 1. And we showed that rapamycins increased the clearance of mutant huntingtin in cells of Drosophila, subsequently zebrafish and mouse models, and they decreased toxicity as a function of reducing the accumulation of the toxic proteins.

We then extended this phenomenon beyond Huntington’s disease. Huntington’s, as I will tell you much more about later, is the most common of the nine conditions caused by expanded polyglutamine tract mutations. The second most common disease in this group is spinocerebellar ataxia type 3, and we showed that the mutant protein in this disease was an autophagy substrate, and if we induced autophagy in a mouse model expressing this mutant protein, we could virtually normalize its phenotype.

We also studied isolated polyglutamine tracts in cells and in Drosophila, in order to try to generalize the process across this whole class of diseases and showed that autophagy induction could protect against isolated polyglutamine tracts.

But perhaps the most important diseases on this slide from a numerical standpoint are those where tau is involved. Wild-type tau is believed to be an important effect of pathology in Alzheimer’s disease, and point mutations in tau cause frontotemporal dementias. We showed that both of these forms of tau in cells of Drosophila were autophagy substrates, and that we could alleviate its toxicity by upregulating this process.

And, finally, we showed that α-synuclein, which is the hallmark protein that accumulates in Parkinson’s disease, is also an autophagy substrate.

With that knowledge in mind, we started thinking about therapeutic strategies. I like to make the analogy that one might be able to consider treating with autophagy upregulation is similar to treating high blood cholesterol levels with statins. Statins reduce the risk of heart disease by reducing cholesterol levels in people from a young age, far younger than the age of onset of heart disease. Potentially, one might be able to have a neurostatin analogy where autophagy is reducing the levels of aggregate-prone proteins in the brain in individuals from a young age, far younger than the onset of neurogenerative conditions, and in this way reducing the risk of these dreadful diseases.

Accordingly, we’ve undertaken many different screening approaches to identify autophagy-inducing drugs, with the focus on those that act independently of the mammalian target of rapamycin, since rapamycins are very large molecules that don’t cross into the brain very well, and in order to get an effect in the brain one needs a large peripheral dose with some side effects.

These screens have been useful and have identified many drugs that can induce autophagy independently of the target rapamycin and have given clues about new pathways regulating the process. These pathways can clear mutant aggregate-prone proteins like mutant huntingtin or mutant α-synuclein or mutant tau, and can be protective in cell, Drosophila, zebrafish, and mouse models of Huntington’s disease, and more recent data show that they can be protective in mouse models of Parkinson’s disease caused by α-synuclein as well as tauopathies.

One of our most successful screens was one that used a repurposing library. So we took drugs that had been used for other indications in people and looked to see which of those induced autophagy and identified many of the compounds shown on this slide. We’ve already started trying to take some of these which we’ve got successful data in mice into people, and last year we completed a safety trial in early Huntington’s patients with rilmenidine which was driven by my colleague Roger Barker in Clinical Neurosciences in Cambridge.

I hope at this stage of the talk I’ve conveyed my enthusiasm for the idea that autophagy upregulation might be a rational therapeutic strategy for many neurodegenerative diseases. I’d like to now switch tack and discuss what happens when neuronal autophagy is impaired.

This question was attacked very elegantly by Noboru Mizushima and Masaaki Komatsu, two different Japanese group leaders who conditionally knocked out different autophagy genes in the neurons of mice. They found that if you knocked out autophagy in these neurons you got the accumulation of these inclusions that were not seen in the wild-type mice, and they found out these inclusions comprise p62, an autophagy substrate, and ubiquitin, among other proteins probably.

They also showed that this protein aggregation was accompanied by cell death. So this is the Purkinje cell layer of the cerebellum in normal mice, and in autophagy-null neuron mice you can see clear cell loss in the cerebellum.

This raises the question whether there is an impairment of neuronal autophagy in other neurodegenerative conditions which might contribute to the aggregate-formation and cell stress we see.

This is a rather busy slide that illustrates some of the diseases we’ve worked on in my lab where we’ve identified impaired autophagy as a consequence of the genetic lesion, or the chemical lesion in one case.

What is important on this slide is not the detail but the color code. We have some diseases in blue where there’s a deficit in the generation of autophagosomes, we have some diseases in red where the autophagosomes are made normally but are trafficked to the part of the cell where the lysosomes are concentrated properly, and, finally, we have in green diseases where there’s a deficit of lysosomal function which prevents the degradation of the autophagic contents. Indeed, the most common neurodegenerative diseases of childhood are lysosomal storage diseases which fit into this category.

The next slide now takes diseases which can manifest as Parkinson’s disease and shows that these can impact different stages of the autophagic itinerary. We have some diseases in blue; for instance, if one’s got an excess of α-synuclein, one then impairs the biogenesis of autophagosomes. The diseases in blue might be remediable by increasing autophagosome biogenesis. However, there’s a condition called Perry syndrome where there’s a defect in the machinery that gets autophagosomes to the lysosomes, and glucocerebrosidase heterozygotes are the most common genetic influence on Parkinson’s disease and results in impaired lysosomal function.

In the situation of the red and green diseases, the autophagosomes are made normally but they’re not removed normally. In these circumstances increasing autophagosome biogenesis might not be safe because the autophagosomes will not be properly removed or not removed at the correct rate, and might indeed accumulate with deleterious consequences, and indeed this is an area that needs further investigation.

I’m going to focus most of the data in this talk around the idea that polyglutamine expansion diseases might impair autophagy. As I mentioned at the beginning, there are nine diseases that are caused by polyglutamine expansions, however these diseases all share a number of key features. First, all of us have normal alleles for the genes that ultimately can become mutated, and these normal alleles will have shorter polyglutamine stretches. As an example, the median number of glutamine repeats in normal huntingtin alleles is 17, 18, 19 repeats.

When the number of successive glutamines exceeds a certain number, one then gets a disease chromosome. In the situation with Huntington’s disease, this happens when you have 38 or more uninterrupted glutamines.

All of these nine diseases have a roughly similar, although not identical, threshold for moving from a wild-type to a disease allele, so that’s feature number one. Feature number two is shown by five of the diseases, but it’s seen in all nine, and that is that there’s an inverse relationship between the number of repeats and the age of onset of disease. So I told you earlier than the median onset of Huntington’s disease is 40, but you can get Huntington’s disease in very early childhood if you have more than 100 repeats.

This begs the question whether there’s a common pathogenic mechanism across all of these nine diseases. The first clue of such a mechanism is shown here in this panel from a seminal paper from Marian DiFiglia. And here her group have looked at the brains of Huntington’s disease patients and they showed in the rare juvenile-onset forms of Huntington’s disease you get aggregates of mutant huntingtin in the nucleus; however, you see them outside the nucleus in typical adult-onset Huntington’s disease.

And, indeed, for many years it was thought that the aggregates seen in these different diseases might be the toxic entity that is shared across all nine conditions. However, in 2004, Steve Finkbeiner’s lab published a very interesting paper. In this study, they analyzed cells expressing wild-type and mutant huntingtin, they analyzed the appearance of aggregates in the mutant cells and related this to the propensity to cell death, and remarkably they showed that the cells that developed aggregates had a lower rate of cell death compared to those that didn’t.

The most conservative interpretation of the data is that cells that don’t form aggregates but express mutant huntingtin – and when I say don’t form aggregates, I mean do not form aggregates visible by light microscopy – those cells have toxicity, and one does not need to form an aggregate to see toxicity, to manifest toxicity. So this raises the question whether there are generic pathogenic consequences of expanded polyglutamine tracts in these different diseases that are distinct from the aggregates.

The second is whether there are roles for the normal polyglutamine stretches in the wild-type counterparts of the disease proteins, as the polyglutamine stretches in the wild-type proteins are very well conserved across evolution. To initiate this, Avi Ashkenazi, when he was in my lab, studied ataxin-3, and he started off analyzing the wild-type protein. We studied ataxin-3, as this is a deubiquitinase, so it typically is thought to remove ubiquitins from proteins that are ubiquitinated and thereby protect them from degradation through the ubiquitin proteasome pathway.

First, Avi knocked down ataxin-3 in a range of different cell lines and showed that the ataxin-3 knockdown cells had few autophagosomes and few autolysosomes.

We analyzed autophagy with multiple different assays, and here I’m going to show you the classical LC3-II assay where LC3-II levels as a function of actin correlate with the volume of autophagosomes in cells. And you can see neurons when you compare the control knockdown cells to cells knocked down with two different reagents, the knockdown cells appear to have less LC3-II. And in order to see whether the decrease in the volume of autophagosomes is a consequence of decreased formation of autophagosomes versus increased consumption of the autophagosomes, we then clumped the degradation of autophagosomes using a lysosomal poison called bafilomycin A1. And in the bafilomycin A1-treated cells, you can see that the ataxin-3 knockdown dramatically reduces the amounts of LC3-II, and we can then infer that this is a consequence of decreased LC3-II biogenesis, decreased autophagosome biogenesis.

We then interrogated different stages of the autophagy pathway to identify which stage might be perturbed, and we found that wild-type ataxin-3 regulated the levels of phosphatidylinositol-3 phosphate, PI(3)P.

PI(3)P is generated by an enzyme complex comprising a number of proteins, including the kinase VPS34 and the key autophagy protein Beclin 1, and it’s seen on the recycling endosomes. PI(3)P recruits key proteins regulating autophagosome formation, and this is its key role in the process.

In normal cells the staining for PI(3)P is rather low, however if you induce autophagy by starving the cells, then the PI(3) staining becomes quite dramatic. This is what happens in control knockdown cells, however if you knock down ataxin-3 almost nothing happens, there’s almost no visible increase in the PI(3)P staining.

We then interrogated different components of the PI(3)P regulatory machinery and found that ataxin-3 knockdown regulated Beclin levels. So you can see here this is the control knockdown cells, and here are two different shRNAs targeting ataxin-3 and their result in decreased Beclin levels.

We then wanted to see whether ataxin-3 regulated Beclin 1 degradation and did cycloheximide chase experiments for 8 hours, and in the presence of cycloheximide the protein synthesis is blocked, and this allows you to infer degradation rates. So in the control cells, this is what the level of Beclin 1 is at time 0 and this is the level at time 8 hours, while in the ataxin-3 knockdown cells this is the level at time 0 and this is the level at time 8 hours. And when one does the calculation, you can see there’s significantly more rapid degradation of Beclin when you do the ataxin-3 knockdown.

This degradation is likely proteasome-dependent. Here’s an experiment where we got control cells measuring Beclin, cells with ataxin-3 knockdown where Beclin levels are reduced, and you can see the Beclin levels are rescued when one treats with a proteasome inhibitor.

We thus thought that this is likely due to ataxin-3 deubiquitinase activity. We tested this in the following ways: Here again are Beclin levels in control cells, Beclin levels in ataxin-3 knockdown cells. And if you take the ataxin-3 knockdown cells and reconstitute them with wild-type ataxin-3, then you can rescue the Beclin levels to normality. However, if you reconstitute them with ataxin-3 which is dead in terms of its deubiquitinase activity, then nothing happens.

Our idea that ataxin-3 deubiquitinase activity might be important in terms of Beclin 1 regulation was strengthened when we looked at the levels of ubiquitinated Beclin 1. In control cells, we found that it was very low, but when you looked at the levels of ubiquitinated Beclin 1 in ataxin-3 knockdown cells it was dramatically increased, again consistent with our idea that the ability of ataxin-3 to remove the ubiquitins from Beclin 1 might be relevant to this biology.

We then showed that ataxin-3 interacted with Beclin 1 and did an in vitro deubiquitination assay to confirm the idea that ataxin-3 was a deubiquitinase for Beclin 1. So in this lane we’ve got Beclin, we’ve got ubiquitinated Beclin in the control situation in vitro, and you can see when you add wild-type ataxin-3 you remove the ubiquitin from Beclin 1, however when you add a deubiquitinase dead form of ataxin-3 you get no such effect. So these data suggested that ataxin-3 is a deubiquitinase for Beclin 1, it thereby protects Beclin 1 from degradation and thereby protects the autophagy integrity of the cell.

We then wanted to identify the lysine on Beclin 1 that was ubiquitinated and identified lysine-402 as such a candidate for mass spec studies. Then Avi took cells he’d knocked down Beclin 1 and replaced it either with wild-type Beclin 1 which he showed was ubiquitinated and where the ubiquitination could be increased by inhibiting ataxin-3, or he had it in a form of Beclin 1 where he mutated this lysine-402 to an arginine so it can’t be ubiquitinated and showed that the mutated Beclin was hardly ubiquitinated either in control situations or ataxin-3 knockdown situations.

Our data suggested that the ability of ataxin-3 to protect Beclin 1 from degradation was a consequence of ataxin-3 binding to Beclin 1, the binding was required to enable ataxin-3 to deubiquitate Beclin 1, and this binding was occurring between the evolutionarily conserved domain of Beclin and the wild-type polyglutamine stretch in ataxin-3. So if you remove the wild-type polyglutamine stretch from ataxin-3, it doesn’t bind and doesn’t serve as a deubiquitinate for Beclin 1.

This stimulated the idea if one had a cell that contained a protein like mutant huntingtin with an expanded polyglutamine tract, whether the expanded polyglutamine tract would bind more effectively to Beclin and displace off the shorter polyglutamine stretch in ataxin-3, and in this way the hypothesis was that if you had a protein like mutant huntingtin in the cell, you’d get less ataxin-3 binding to Beclin 1, more rapid Beclin 1 degradation, and less autophagy. And this is, indeed, what we set out to test.

We first used a model system here where we had cells that either expressed empty GFP, or GFP tagged to 35-glutamines, or GFP tagged with 81-glutamines. Importantly in these experiments, the 35-glutamines do not form visible aggregates by light microscopy. The prediction from studies with this reductionistic system was that 35- and 81-glutamines should bind to Beclin, and 35 should bind less than 81. And this is, indeed, the case. Empty GFP doesn’t bind, 35 binds, and 81 binds more strongly.

As a consequence of these glutamines binding to Beclin, the prediction was they should displace off the Beclin/ataxin interaction, and indeed this is occurring. This is the strength of the ataxin-3/Beclin interaction with GFP, with GFP with 35-glutamines, and with GFP with 81-glutamines.

As a consequence of disruption of this ataxin-3/Beclin interaction, the prediction was that Beclin should be more ubiquitinated with 35 and 81 repeats, and that’s the case so this is the deubiquitination of Beclin with GFP, with GFP with 35-, and GFP with 81-glutamines.

And, finally, as a consequence of the increased ubiquitination, the prediction is that the steady state levels of ataxin-3 should be reduced in cells expressing 35- and 81-glutamines. And that’s indeed the case, this is GFP 35- and 81-glutamines, and these are the Beclin levels.

This experiment now makes, I think, quite an important point, and we’ve done it again just with the 35-glutamines attached to GFP that don’t form aggregates. Lane 1 is the control, and you can see the Beclin levels and the LC3-II levels marking the volume of autophagosomes in the cell. These are reduced when you overexpress GFP tied to 35-glutamines; Beclin goes down and in parallel the amount of LC3-II goes down. But if you take the cells overexpressing the 35-glutamines and you simultaneously overexpress wild-type ataxin-3, then the situation is rescued; Beclin 1 is normalized and LC3-II is normalized.

And this is consistent with the model we proposed that the longer glutamine stretches in the protein in-trans is competing with the shorter glutamine stretches in wild-type ataxin-3 for binding to Beclin 1.

We then wanted to examine this phenomenon in the context of proper disease-causing proteins, and I’m not going to show you all the data but I’m going to focus mainly on the huntingtin experiments which I think give a clear example of what’s going on.

We showed that wild-type huntingtin when overexpressed could bind to Beclin 1, and this was a function of its polyglutamine stretch. So this was wild-type huntingtin with 17-glutamines, but if you removed the glutamine from wild-type huntingtin the interaction is dramatically impaired. We showed that mutant huntingtin could interact with Beclin 1, and when you have mutant huntingtin in the mix the Beclin 1 interaction with ataxin-3 is reduced, consistent with what I showed you with the model protein GFP in the previous slide.

We then wanted to test whether mutant huntingtin would inhibit starvation-induced autophagy, which ends up being a particularly suitable readout for impaired VPS34 or Beclin function. First, we used stable, inducible cell lines that we’d made that express either wild-type huntingtin exon 1 with 23-glutamines or mutant exon 1 with 74-glutamines. In the first lane you’ve got the wild-type-expressing cells uninduced with no expression of the protein, and when you can switch on the expression of the protein with doxycycline you get a nonsignificant reduction in the levels of Beclin 1, accompanied with a significant reduction in the number of autophagosomes in the cells. However, when you switch on the mutant proteins – so this is the mutant cells off, the mutant cells on – the levels of Beclin 1 plummets to very low levels, as well as the numbers of autophagosomes being reduced to very low levels.

We then looked at cells derived from the part of the brain which is most sensitive to the huntingtin’s mutation, the striatum, and we studied cells from a knock-in mouse model which is basically humanized for exon 1 of the huntingtin gene; so the exon 1 of the human huntingtin gene has been inserted in place of the mouse gene, and this was a model that was made originally by Marcy MacDonald’s lab.

If one takes the cells from these knock-in mice expressing wild-type huntingtin and compares them to mutant huntingtin, you can see the mutant-expressing cells have decreased Beclin levels and have decreased starvation-induced autophagy.

We then studied brains from exon 1 huntingtin transgenic mice and compared them to controls, and again we saw that the transgenic mice had decreased levels of Beclin 1, particularly in the starved mice, and this was accompanied by a significant decrease in autophagosome load in the fasting mice with the mutant huntingtin transgene compared to the wild-type controls.

Finally, we looked at patient fibroblasts and compared them to control fibroblasts, and if one looks at LC3-II levels in the controls you can see that they are much higher than the LC3-II formation levels in the huntingtin patients, suggesting that this phenomenon occurs in the fibroblasts with endogenous levels of expression of the huntingtin protein.

We saw similar trends in fibroblasts from a range of other polyglutamine diseases, and also saw similar trends and similar biology when we overexpressed constructs for wild-type in mutant polyglutamine proteins more or less across the group of diseases. When this occurred, we also saw decreased levels of Beclin 1 consistent with our model.

The examples and the scenarios I’ve described so far indicate what we think is happening where we’ve got a protein with the polyglutamine expansion in-trans with wild-type ataxin-3 with the short polyglutamine stretch.

We next wanted to understand what happened if you have a polyglutamine stretch expansion in ataxin-3 itself, as this causes the disease spinocerebellar ataxia type 3. When we did this analysis, we found that wild-type ataxin-3 bound Beclin less effectively than the mutant, the mutant binds much more strongly to Beclin 1.

This is associated with dramatically increased ubiquitination of Beclin 1, and this can be explained by the fact that in the in vitro deubiquitination assay this is the control situation, the wild-type ataxin through with the short polyglutamine stretch deubiquitinates very effectively while deubiquitinates function of ataxin-3 with the long polyglutamine stretch is compromised.

Importantly, one must realize that this is an autosomal dominant disease, and if one’s got a sticky mutant ataxin-3 which doesn’t deubiquitinase properly it’s going to compete out the less sticky wild-type ataxin-3.

Our view is that the decreased deubiquitinase activity of the mutant ataxin-3 is most likely a consequence of it just being too sticky and not letting go of its substrate effectively enough to be able to cycle effectively and attack the whole pool of the protein.

This panel in the corner, I think, summarizes much of what I’ve told you today. In control cells the levels of autophagy, the number of autophagosomes are set here at 100%. They are dramatically reduced when one knocks down ataxin-3. If one takes the ataxin-3 knockdown cells and reconstitutes them with wild-type ataxin-3, then one normalizes the number of autophagosomes. If you reconstitute with mutant ataxin-3 with polyglutamine expansion, one gets only a partial rescue consistent with the deubiquitination data I’ve shown you above. But if you reconstitute with the form of ataxin-3 that has no glutamines, it’s like putting nothing back in, it’s completely dead.

We were delighted to see some previous papers that were consistent with our data, so groups that reported that the decreased Beclin 1 and autophagosome biogenesis data in fibroblasts from patients with Machado-Joseph disease, which is another name for the disease spinocerebellar ataxia type 3 and delighted to see that in rodent models of spinocerebellar ataxia type 3, that Beclin 1 overexpression could rescue the phenotypes.

So in summary, our data provided normal function for the polyglutamine stretches in ataxin-3, because these mediated the interaction of ataxin-3 and Beclin 1 and allow ataxin-3 to deubiquitinate Beclin 1 and protect it from degradation, and thus enable functioning autophagy in the cells.

However, when one has a mutant protein with long polyglutamine stretches in-trans in the cell, for instance with mutant huntingtin, these long polyglutamine stretches out-competes wild-type ataxin-3 from binding Beclin 1, and this results in impaired deubiquitination of Beclin 1; so increased ubiquitination, increased degradation of Beclin 1, and decreased autophagy.

One sees a very similar situation when the polyglutamine stretch in ataxin-3 itself is expanded, as this binds more strongly to Beclin 1 but has reduced deubiquitinase activity.

So we have a function for wild-type polyglutamine tracts in at least one of these proteins. In one of these proteins, ataxin-3, we have a toxic function for soluble mutant polyglutamines across many different polyglutamine disease proteins. I showed you data with a Q35 tract, but many of our experiments were performed under conditions where the mutant proteins we were studying did not form visible aggregates by light microscopy.

It is interesting to speculate that this scenario might be subject to some type of a positive feedback loop. For instance, if one has cells with modest levels of huntingtin and good functioning autophagy, the autophagy will be impaired to some extent by the low levels or the modest levels of huntingtin. This will result in decreased autophagy and increased levels of huntingtin, and an increase in the levels of huntingtin because mutant huntingtin itself is an autophagy substrate. And as the levels of mutant huntingtin go up, so does one get a further decrease in the autophagic capacity of the cells which would lead then to an even further increase in the levels of the mutant protein.

I think it’s important to consider that this situation might be nuanced in the disease context and in a real diseased brain, and might depend on the levels of a wild-type ataxin-3 versus the levels of the particular disease protein with the polyglutamine expansion like mutant huntingtin versus the levels of Beclin 1, and these might vary across different cell types, different neuronal types in the brain, and this might indeed account for some of the cell-type vulnerability differences across this group of nine diseases.

Finally, in the patient him- or herself, the different repeat lengths in wild-type ataxin-3 versus the disease gene – that’s why I was concerned – might also in some way modulate the defect that we see.

I don’t want to leave with bad news today, I just want to tell you one story which describes a particular phenomenon that one might be able to use to fool the disease biology, to bypass the disease biology.

I mentioned to you earlier that the complex including VPS34 and Beclin 1 generate PI(3)P recycling endosomes, and the PI(3)P on these structures recruits proteins that dictate where and when autophagosomes are going to form. And the first protein that is involved in this cascade is a protein called WIPI2 which binds to PI(3)P, and WIPI2 then, in turn, recruits machinery which dictates where the autophagosomes are going to form.

However, a number of studies have observed that you might be able to get autophagy in the absence of VPS34 activity or possibly PI(3)P activity, and so Mariella Vicinanza in my lab worked on the idea that there might indeed be PI(3)P-independent autophagy if the PI(3)P and PI(5)P, another phosphatidylinositol phosphate were mirror images and worked in the same way. So these are two paintings that Mariella produced to illustrate the hypothesis showing the mirror image phosphoinositides.

This is a summary of some of the key data that she produced. This is a simplified diagram showing the biogenesis of these different PI(3)Ps and PI(5)Ps from PI – so this is PI(3)P, this is PI(5)P – and this shows that both can be found on the same membranes. This shows that we found that if you deplete PI(3)P levels in cells you impair autophagosome biogenesis, and you can rescue that defect by elevating PI(5)P levels. And conversely, if you deplete PI(5)P levels you impair autophagosome biogenesis, and you can rescue that by elevating PI(3)P levels. So the two PIPs are interchangeable.

Mariella found that both PI(5)P and PI(3)P bind WIPI2, and indeed in in vitro assays we found that they both can compete for binding to WIPI2 and bind to the same domain of WIPI2. So this suggests that they have the same biochemical mode of action, and indeed this suggests that they both work in the conventional way by recruiting WIPI2 and bringing along with it co-machinery that dictate where the autophagosomes are going to form on the cells.

I’m telling you this story because I think it provides an opportunity for a therapeutic intervention. PI(5)P can be metabolized to PI(4,5)P2 by a series of three isoforms of PI(5)P4 kinases. When one perturbs the levels of these kinases, one has a dramatic effect on PI(5)P level but very little effect on (4,5)P2 levels because most of the (4,5)P2 pool which is much larger than PI(5)P pool, but most of the (4,5)P2 pool is not generated by this conversion, but instead generated by conversion of PI(4)P to (4,5)P2.

If one overexpresses any of the isoforms of this kinase, one depletes PI(5)P, impairs autophagosome biogenesis, and if one has this scenario and one is expressing mutant huntingtin in cells – these are the number of cells with mutant huntingtin aggregates – these are dramatically increased when one is depleting PI(5)P by overexpressing these kinases. In fact, these first three lanes reflect autophagy-competent cells, the last three lanes reflect autophagy-null cells, and you can see that when you overexpress these kinases you have about the same number of cells with aggregates as in the autophagy-null state. Indeed, when you take autophagy-null states and you overexpress these kinases, you have no effect, suggesting that the effects of these kinases on this aggregate phenotype is autophagy-dependent.

More importantly, if you genetically deplete these kinases, you increase the levels of PI(5)P and you drive an increase in autophagosome biogenesis. And again if you’ve got cells overexpressing mutant huntingtin, then you get a dramatic decrease in the number of cells that have aggregates when you impair these enzymes, and you don’t see any such effect in autophagy-null cells. So this suggests that impairing activity of these kinases might be a therapeutic target.

And, indeed, in Cambridge in our Alzheimer’s Research UK-funded Drug Discovery Institute, we are trying to make chemical inhibitors of these kinases to that end.

So I’ve told you a lot of things today, and I’d just like to leave you with two bottom lines. The first is that I’m keen on the idea that autophagy upregulation might be a rational therapeutic strategy for many neurodegenerative diseases by removing toxic intracytoplasmic aggregate-prone proteins. I haven’t shown you, but we and others have shown that this protects cells against cell death.

We and others have shown that autophagy upregulation is protective in a wide range of different neurodegenerative disease models – Huntington’s Parkinson’s, tauopathy, etc. – and these benefits can be mediated by perturbing the mammalian target of rapamycin, but also by impairing or modulating the activities of a range of mTOR-independent pathways.

On the other hand, autophagy compromise is frequently seen in neurodegenerative disease conditions and might enhance the aggregate-formation and cell stress. I wouldn’t want to pretend that the genetic lesions causing autophagy compromise are causing the disease only through this mechanism, indeed it’s very likely that they’re having multiple deleterious effects, but I think it’s fair to say that I believe that the autophagy compromise is contributing to the effects and it’s particularly important when the relevant aggregate-prone protein is itself an autophagy substrate, like mutant huntingtin or mutant α-synuclein.

We need to understand if a disease is suitable for autophagy upregulation therapies by understanding where in the autophagy itinerary the lesion is impacting, and that might influence where one wants to intervene and what particular stage of the autophagic process one wants to upregulate. And, finally, I think that if one understands the disease biology, this might inform the most attractive targets to drug.

Finally, I’ve talked about work that was done over many years by many people in my lab. I’ve had great collaborators over the years, and this work wouldn’t have been possible without the generous funding we’ve had. Thank you very much.

Chan: Thank you, David, for that very informative presentation. Before we begin the Q&A, I would like to remind the attendees that they can submit questions to David using the Ask A Question box just below the presentation screen.

So the first question is if targeting autophagy in the CNS, which cell type is most targeted?

Rubinsztein: I think that’s an interesting question, and I’m not sure anybody’s done the experiments that will tell you which cell type is the most targeted. However, the mouse model of Huntington’s disease that we’ve studied expresses the mutant protein almost exclusively in neurons, so there’s no detectable expression in glia, and when we induce autophagy, we reduce the levels of the transgene product in neurons and ameliorate the disease in these models. So I think at least in those models we can say that it is almost certainly working on the neurons themselves.

However, that does not mean to say that in models that have other types of pathology, that one wouldn’t target the glia and that wouldn’t have some type of benefit. Parenthetically, in that model then there’s no glial response to the transgene product. So I think that there it’s simplistic, but I think that one might have benefits targeting glia as well.

Chan: Lots of questions coming in, so next one is if autophagy is involved an extracellular amyloid beta, what specifically helps link to your phagocytosis and engulfment by microglia, and do microglia have possible defects in degrading amyloid beta after phagocytosis?

Rubinsztein: So I’m not sure that anybody knows whether autophagy is actually removing the extracellular amyloid beta. Certainly in some experiments in mouse models when autophagy is induced, one reduces the overall load of amyloid beta, but it’s possible that that is driven by reducing the intracellular pool. I think the point you make about phagocytosis and engulfment is quite possible, and there is a form of phagocytosis called LC3-associated phagocytosis which employs many of the components of the autophagy machinery, and I think it’s plausible that that process might be playing a role in the type of microglial engulfment that you’re proposing. I’m not sure that that’s been formally worked through, but I think that’s an interesting idea that would be fuel for further experiments.

Chan: Thank you, David. A question about how will it be possible to target autophagy in diseases where there are multiple steps that are affected, such as Parkinson’s?

Rubinsztein: So, I think that’s a complicated question, because many of the examples I gave you in Parkinson’s disease described monogenic forms of the disease which might, in some cases, affect different steps of the pathway. So I think in the monogenic scenarios one would try to tailor the type of therapy to where autophagy is impaired, as I described in the talk.

I think that if one is dealing with the disease where there might be defects in the pathway at different stages, for instance at the level of autophagosome biogenesis as well as with autophagosome lysosome fusion, then one has a greater difficulty and I’m not sure that there’s straightforward answers to that.

I think in stories where there are defects in the itinerary subsequent to the completion of autophagosome formation, one would probably want to do experiments in animal models trying to see whether one could bypass the partial block in the pathway with autophagy-inducing drugs, but I think one would need to be very cautious, and there are going to be some instances where autophagosome biogenesis induction is certainly not going to be the ideal strategy.

Chan: Thank you. I guess a question about clarifying what’s the role of ubiquitin-mediated proteasome pathway, is it a major player in neurogenerative pathology?

Rubinsztein: So I think it is a major player and I’d refer people to a very nice paper from Karen Duff and Fred Goldberg in Nature Medicine where they showed that tau mutations impede flux through the proteasome, so cause a block in the ubiquitin proteasome system, and that’s something that we recapitulated subsequently in a zebrafish model. So I think at least in that context it’s important.

And what we found actually, it’s quite interesting, is if you have such tauopathy models, at least in the zebrafish that we worked with, if you induce autophagy you can ameliorate the problem by upregulating the alternative degradation routes.

Chan: Another question related to that was how does overexpressing tau, either wild-type or mutant or even aggregate, affect in the autophagy pathway; is it autophagy initiation, flux, or later steps?

Rubinsztein: We haven’t done experiments with overexpression of tau and autophagy, or let’s put it this way, the experiments we’ve done have not given us any clear-cut answers. So I suspect others have maybe worked in that field, but in our hands, we didn’t see any gross changes.

Chan: Lots of questions. We will followup with all the questions that we cannot cover. Can you clarify why in patients who have GBA mutations who are heterozygotes, that inducing autophagy could be deleterious from α-synuclein buildup in these patients rather than an improvement?

Rubinsztein: Well, I think the problem is if one assumes that the GBA heterozygotes have a level of lysosomal dysfunction, and I think there is a fair amount of data suggesting that is the case, then they’re not going to degrade autophagosomes at the normal rate, and therefore to induce autophagy might lead to an aberrant buildup of autophagosomes. I think this is, again, one of those scenarios where if one wants to investigate the strategy of inducing autophagosome biogenesis in the GBA heterozygotes, I would do experiments in animal and cell models before I did any experiments in people, and I think one would look very carefully at removal of autophagy substrates and accumulation of autophagosomes. So I think that it’s a set of questions that are begging to be done, but to answer it experimentally.

Chan: Thank you. Question about huntingtin. I would like to know what is the function of wild-type huntingtin, and what is relevant for this interaction of Beclin 1?

Rubinsztein: So Ana Maria Cuervo’s done quite a lot of work on the functions of wild-type huntingtin, and I’d suggest you try to look through all those papers, and she found roles particularly at the C-terminus of wild-type huntingtin. I think that in the context of what we found with Beclin 1, we could show that the pathway we were describing was independent of roles that had been described previously because they could be attributed specifically to the expanded polyglutamine tract in the protein. So the wild-type protein doesn’t have an expanded polyglutamine tract and probably does not significantly impede general autophagy.

It might have some regulatory effect, but it doesn’t significantly impede the wild-type polyglutamine stretch. But when you have a long polyglutamine stretch, then one gets the competition between the long polyglutamine stretch and huntingtin, or for that matter many of the other polyglutamine-expanded proteins that cause disease and the short polyglutamine stretch in ataxin-3, and that basically pushes ataxin-3 away from its interaction with Beclin and leads to more rapid degradation of Beclin and impaired autophagy.

I think it’s important to stress that the mechanism we were proposing for the polyglutamine toxicity is almost independent of the protein that is hosting the polyglutamine expansion – almost.

Chan: Thank you. Got some questions about drugs in clinical trials. In what stage of Alzheimer’s disease patients are autophagy-inducing drugs effective?

Rubinsztein: So nobody’s done that experiment. People have done that experiment in mouse models, overexpressing APP, and the data in those studies suggests that if you go early you have benefits, however if you go late you don’t. And I think that probably is going to be a general rule across diseases, and one of the challenges is to try to understand what early and late means in the context of a human disease as opposed to a mouse model. But that’s a clear challenge.

Chan: Thank you. Can you just give some general comments about the current progress of some of the trials you’ve done in Cambridge; a lot of the drugs are modulating autophagy, what’s the general situation right now for treating neurodegenerative diseases?

Rubinsztein: I’m not sure what you’re asking. I mean, I think that what we’re trying to do here is develop compounds that induce autophagy, provide the strongest preclinical data we can for their relevance, and I think we’re learning as time passes how to do that better than we did it previously. And then do safety studies to test that the drugs are tolerated in people and try to build a case with the preclinical and clinical data over time that will allow us to be able to motivate for a suitably powered proof of concept study in an appropriate disease. And I won’t tend to debate what the one appropriate disease might be, but we think Huntington’s disease is a good group to try to develop proof of concept for the hypothesis. So that is what the overall aim is, and I think we’re trying to develop drugs or repurpose drugs as well as we can to get towards that objective.

Chan: Terrific, thank you very much. Unfortunately, we’ve kind of run out of time, but we do have some questions and we’ll post these to David, and then we will try our best to get back to everyone.

I would just like to thank David, again, for a wonderful presentation, and thank everyone for joining us for today’s webinar. And I’d just like to wish everyone a great day and thank you very much.