Autophagy and neurodegeneration

Video Transcript

• 00:00 - 00:17: Good day. I’m David Rubinsztein from the Cambridge Institute for Medical Research and the UK
• 00:17 - 00:23: Dementia Research Institute to tell you about the autophagy lysosome pathway in neurodegenerative
• 00:23 - 00:31: diseases. I’m going to divide my talk into three sections. I’m going to start by introducing
• 00:31 - 00:38: autophagy to you and then go on to describe how autophagy compromise might contribute to
• 00:38 - 00:48: various neurodegenerative diseases. I’m going to end by telling you about some work that aims to
• 00:48 - 00:55: harness autophagy as a therapeutic strategy for various neurodegenerative conditions. Autophagy
• 00:55 - 01:02: is an intracellular protein degradation pathway that also captures various organelles.
• 01:03 - 01:09: It acts in the cytoplasm and the first recognizable component of this pathway
• 01:09 - 01:13: is this cup-shaped double membrane structure called a phagophore.
• 01:13 - 01:22: After the edges of the phagophore extend and fuse, you have an autophagosome that has captured
• 01:22 - 01:28: cytoplasmic contents. These autophagosomes are then trafficked along microtubules towards the
• 01:28 - 01:36: microtubule organizing center of the cells by dynamin machinery to bring the autophagosomes
• 01:36 - 01:42: close to the lysosomes which they eventually fuse with.
• 01:43 - 01:48: After this fusion event occurs, the contents of the autophagosome
• 01:48 - 01:52: and the inner membrane are degraded by the lysosomal hydrolases.
• 01:56 - 02:03: One of the key questions in the field is trying to understand the origin of the phagophores.
• 02:04 - 02:12: Our data over the last few years has led us to postulate that autophagosomes and phagophores
• 02:12 - 02:20: are formed on the recycling endosome membranes and cells as shown in green in this micrograph.
• 02:20 - 02:27: So the green are the recycling endosomes and LC3 which is a classic autophagy marker shown in red
• 02:27 - 02:38: and in this fluorescent confocal image you can see some autophagosomes on the recycling endosome.
• 02:38 - 02:45: You can see it here as well. This is a whole mount electron micrograph which we’ve colored in because
• 02:45 - 02:52: it performs with the immunogold and you can see again here is the recycling endosome
• 02:52 - 02:54: and the autophagosome emerging from the structure.
• 02:58 - 03:01: This picture shows a schematic of what we think is going on.
• 03:03 - 03:11: As I mentioned before, LC3 is a classic marker for autophagosomes and LC3 is actually a protein
• 03:11 - 03:20: that gets conjugated to the lipid phosphatidylethanolamine in the nascent autophagosome
• 03:20 - 03:32: membrane. A protein complex containing ATG16L1 and two other proteins ATG5 and ATG12
• 03:33 - 03:43: dictate the sites of LC3 attachment to the membranes and this is in turn determined
• 03:43 - 03:46: by an ATG16 interacting protein called WIPI2.
• 03:49 - 04:00: The sites of WIPI2 recruitment to membranes in the cells are determined by PI3P which had been long
• 04:00 - 04:07: known in the literature and we found that PI3P was not sufficient to enable this recruitment
• 04:07 - 04:15: but that this process required what is known as coincident detection of PI3P along with a protein
• 04:15 - 04:24: called RAB11. So the RAB11 provides the specificity for this as PI3P is found in many different
• 04:24 - 04:30: compartments in the cell while the RAB11 is one of the marker proteins for recycling endosomes.
• 04:31 - 04:39: So in this way the RAB11 and PI3P are recruiting the WIPI2 which in turn ultimately
• 04:39 - 04:45: determines the sites of LC3 conjugation to nascent autophagosome membranes.
• 04:49 - 04:57: In doing so, we find that not only is LC3 associated with these membranes but many of
• 04:58 - 05:03: the precursor proteins that determine autophagosome formation are similarly seen
• 05:03 - 05:13: in that location like ATG16, WIPI2 as an example. The other purpose of the slide is to highlight
• 05:13 - 05:19: another interesting finding we made when doing this work and this relates to another protein
• 05:19 - 05:25: that is seen and enriched in recycling endosomes and this is the transferrin receptor.
• 05:26 - 05:32: The transferrin receptor can start its journey on the plasma membrane where it can bind its
• 05:32 - 05:40: ligand transferrin shown here in red and after it is endocytosed the extracellular portion of the
• 05:40 - 05:49: receptor and the ligand actually are found in the lumen of endosomes and the recycling endosome
• 05:49 - 05:57: compartment and if you can imagine this recycling endosome becoming a nascent autophagosome then
• 05:57 - 06:03: you’ll have the prediction that the extracellular domains of the transferrin receptor along with
• 06:03 - 06:11: transferrin if it’s endocytosed will be found between the two membranes of the nascent autophagosome.
• 06:12 - 06:19: The other prediction is that when the autophagosome forms then as the inner membrane
• 06:19 - 06:24: is degraded by the lysosomal hydrolases so will some transferrin receptor be degraded.
• 06:26 - 06:31: And indeed we showed that the transferrin receptor could be degraded by autophagy
• 06:32 - 06:37: and we also found that the extracellular domain of the transferrin receptor
• 06:38 - 06:46: as identified by an antibody to this domain could be found between the double membranes
• 06:46 - 06:53: of the nascent autophagosomes as shown here in these various electron micrographs and indeed
• 06:53 - 07:00: we saw exactly the same when we endocytosed transferrin and pulsed it into this compartment.
• 07:00 - 07:08: So these data support the microscopy and biochemistry data that I showed you earlier
• 07:08 - 07:17: that argued that the recycling endosome is a platform on which the autophagosomes are formed.
• 07:18 - 07:27: Now the recycling endosomes are tubular vesicular compartments and in order for autophagy to proceed
• 07:27 - 07:35: one will need some type of scission and departure of the nascent autophagosome from this compartment.
• 07:36 - 07:42: Claudia Puri who’d been working on this project found that when she stimulated autophagy
• 07:42 - 07:49: she got a change in the morphology of this compartment from a largely tubular compartment
• 07:49 - 07:55: in the basal state to a fragmented compartment in autophagy induced cells and we saw this with
• 07:55 - 08:02: three different types of autophagy inducers. When she did movies she could show the departure
• 08:02 - 08:09: of the LC3 positive structure shown here in purple from the green recycling endosome over time.
• 08:10 - 08:16: If the autophagosomes are formed on the recycling endosomes then blocking the scission of the
• 08:16 - 08:23: tubules should cause more LC3 to accumulate on the tubules and one would get a predicted
• 08:23 - 08:32: impairment of autophagic flux and in order to test this idea Claudia needed to identify
• 08:32 - 08:40: the key machinery that was responsible for this event and she found that dynamin-2 was the key
• 08:40 - 08:46: player that enabled the removal of nascent autophagosomes from the recycling endosomes.
• 08:46 - 08:53: When she knocked down dynamin-2 this is what happens. In this micrograph the RAB11
• 08:53 - 09:01: is shown in red, the LC3 the autophagosome marker is shown in green and you can see there is some
• 09:02 - 09:08: association of the LC3 with recycling endosomes but this is dramatically increased when one
• 09:09 - 09:16: knocks down dynamin-2 and one gets increased tubulation suggesting decreased scission
• 09:17 - 09:24: of the recycling endosome compartments in parallel. In dynamin-2 knockdown cells one
• 09:24 - 09:31: also has impaired autophagic flux and a reduced clearance of autophagic substrates.
• 09:33 - 09:36: This is a summary of what Claudia found.
• 09:38 - 09:48: Dynamin-2 is a protein that is not only seen as one knew before in the endocytic compartment
• 09:48 - 09:54: where it plays a role in endocytic scission from the plasma membrane but is also associated
• 09:54 - 10:01: with the recycling endosomes where it mediates the release of nascent autophagosomes from this
• 10:01 - 10:08: compartment. It’s recruited to this compartment to the recycling endosomes
• 10:08 - 10:18: via an interaction with LC3 via an LC3 interacting region. The binding of dynamin-2 to the LC3
• 10:19 - 10:26: enables the release of autophagosomes from this compartment as well as recycling endosome scission.
• 10:27 - 10:33: Dynamin-2 is mutated in Charcot-Marie-Tooth disease but also in a muscle disease called
• 10:33 - 10:41: centronuclear myopathy and we studied a particular mutation at R465W in dynamin-2
• 10:41 - 10:48: that gives this muscle disease and we found it had very interesting properties that caused us to
• 10:48 - 10:57: name it location-location-location mutation. This mutation in dynamin-2 doesn’t intrinsically
• 10:57 - 11:05: affect its ability to do its job in the scission of the recycling endosomes and the removal of
• 11:05 - 11:12: nascent autophagosomes from this compartment but instead this mutation makes the dynamin-2
• 11:12 - 11:22: bind extra well to its partner intersectin-1 in endosomes and this causes the mutant dynamin-2
• 11:22 - 11:29: to be recruited preferentially away from the recycling endosomes towards the endosomal
• 11:29 - 11:38: compartment and as a consequence one gets impaired dynamin activity here and an accumulation
• 11:39 - 11:44: of nascent autophagosomes or recycling endosomes in the exact same way as one would see with
• 11:44 - 11:52: dynamin-2 knockdown. However, one can rescue the effects of this mutation either by knocking down
• 11:52 - 12:01: its interactor in the endosomal compartment and conversely one can simply mimic the effect of
• 12:01 - 12:06: this mutation in normal cells by overexpressing intersectin-1.
• 12:09 - 12:15: In addition to working on the mechanics of early stages of autophagosome biogenesis,
• 12:15 - 12:21: my lab is also interested in understanding the key signals that regulate autophagy.
• 12:23 - 12:30: One of the primordial signals regulating this process which is conserved from yeast to people
• 12:32 - 12:39: is mediated by the mechanistic target of rapamycin complex 1. It’s a core conserved
• 12:39 - 12:47: negative regulator of autophagy and a key responder to nutrient depletion and today
• 12:47 - 12:52: when I’m talking about nutrients I’m talking about amino acids and I’m going to focus on
• 12:53 - 13:00: leucine among amino acids as this is sufficient to mediate the effects on autophagy
• 13:00 - 13:03: and mTORC1 inhibition or stimulation.
• 13:06 - 13:14: So when one starves cells one inhibits mTOR but since mTORC1 is a negative regulator of autophagy
• 13:14 - 13:22: one actually induces autophagosome formation. The prevailing view in the literature is that
• 13:22 - 13:28: the effects of leucine are mediated by various leucine sensors that have been described
• 13:28 - 13:36: particularly in HEK293 cells. The figure I’m showing here makes two points.
• 13:36 - 13:45: The first is that mTORC1 is a complex of proteins including the TOR kinase and a protein called
• 13:45 - 13:54: Raptor and Raptor is important because one of its roles is to bind to these RAG proteins
• 13:54 - 13:58: which allow the complex to be tethered to the lysosomal surface
• 13:58 - 14:04: and in doing so when mTORC1 is on the lysosomal surface it can be activated.
• 14:06 - 14:12: Sung Min Son drove the story I’m going to tell you today and it started with him trying to
• 14:12 - 14:21: understand the function of a gene called MCCC1 which stands for 3-methylcrotonyl-CoA carboxylase.
• 14:22 - 14:27: He worked on this initially because it falls in a Parkinson’s disease risk locus
• 14:28 - 14:37: and MCCC1 is an enzyme in a cascade that cells use to convert leucine to acetyl-CoA.
• 14:37 - 14:45: So Son knocked down MCCC1 with various reagents in HeLa cells, primary neurons and various other
• 14:45 - 14:51: cell types and here are the results of one of his initial experiments in HeLa cells and you can see
• 14:51 - 14:59: the MCCC1 levels, the control state with the smart pool, a panel of different individual
• 14:59 - 15:05: siRNAs and you can see that this siRNA number two hasn’t worked and so it serves in a funny way as
• 15:06 - 15:11: an additional control. He’s measured the phosphorylation status of two different
• 15:12 - 15:21: mTORC1 substrates either S6 kinase one or 4-EBP1 or the substrate of S6 kinase itself
• 15:21 - 15:30: S6 and you can see when he’s knocked down MCCC1 he gets decreased mTORC1 activity.
• 15:31 - 15:36: This is correlated with impaired mTORC1 lysosomal localization.
• 15:38 - 15:46: As predicted by the part where I’ve shown you on the previous slide, when you knock down MCCC1
• 15:46 - 15:53: you get a decrease in acetyl-CoA levels in the total cell lysate on the cytosol.
• 15:53 - 15:59: Similarly, when you deprive cells of amino acids you get a decrease in acetyl-CoA levels which are
• 15:59 - 16:06: rescued as predicted from the previous slide by the addition back of leucine or by stimulation
• 16:07 - 16:14: of an alternative pathway that forms acetyl-CoA in the cells.
• 16:14 - 16:26: So what I’ve shown you so far is that L-leucine can be converted to acetyl-CoA
• 16:26 - 16:35: and this is regulated by various enzymes including MCCC1 and our data our initial data
• 16:35 - 16:43: suggested that MCCC1 inhibited if you knock down MCCC1 you inhibit mTOR activation.
• 16:44 - 16:52: And so we hypothesized that it wasn’t necessarily leucine that was responsible for signaling to
• 16:52 - 16:58: mTORC but rather downstream metabolite like acetyl-CoA and we could test this hypothesis
• 16:58 - 17:07: in three different ways. The first prediction is that if we starve cells and then add back leucine,
• 17:08 - 17:09: KIC,
• 17:12 - 17:19: HCOA or acetyl-CoA we should be able to get a rescue with all of these metabolites and this is indeed what
• 17:19 - 17:26: we see. So in this blot you can see the mTORC1 activity again with the same markers I showed
• 17:26 - 17:33: you previously in nutrient-replete conditions you can see that it is dramatically reduced
• 17:33 - 17:42: when we remove amino acids and then when we add back either leucine, KIC, HCOA or DCA so that is
• 17:42 - 17:48: upstream or downstream metabolites we rescue the mTORC1 activity. The second prediction from our
• 17:48 - 17:55: data is that the ability of upstream metabolites to rescue the mTORC1 activity should be blocked
• 17:55 - 18:02: when we knock down MCCC1 or other steps in the pathway downstream of these metabolites.
• 18:02 - 18:08: And this is what we see in this blot so here you can see the mTORC1 activity nutrient-replete
• 18:08 - 18:15: conditions and you can see that when you remove amino acids it goes down it’s rescued by leucine
• 18:15 - 18:23: and KIC but these rescues are abrogated when we knock down MCCC1 that is blocking the conversion
• 18:23 - 18:29: of these upstream metabolites to downstream metabolites. The third prediction from our data
• 18:29 - 18:36: is that the inhibition of mTORC1 activity resulting from MCCC1 knockdown should be able
• 18:36 - 18:42: to be rescued if we take knockdown cells and add back these downstream metabolites and this is
• 18:42 - 18:48: indeed what we see so here again is the mTORC1 activity you can see how it’s reduced when you
• 18:48 - 18:56: knock down MCCC1 but that it’s rescued in the MCCC1 knockdown cells by adding back either HCOA
• 18:56 - 19:07: or acetyl-CoA. These studies and many others led us to show that in most of the cells we’ve studied
• 19:08 - 19:15: leucine doesn’t signal directly to mTORC1 but instead its metabolite does the trick.
• 19:18 - 19:24: The reason why our data are different to what has been described previously
• 19:24 - 19:28: is that almost the entire previous literature that described leucine sensors
• 19:28 - 19:35: had been performed in HEK293 cells. Indeed knockdowns of the previously described leucine
• 19:35 - 19:41: sensors have no effect on the signaling pathway that I’ve described in HeLa cells.
• 19:43 - 19:50: We found that this pathway of leucine signaling through acetyl-CoA acts in many cells including
• 19:50 - 19:55: primary cells like primary neurons and primary glia for instance.
• 19:58 - 20:03: There are cells where it doesn’t work like HEK293 cells in mouse embryonic fibroblasts
• 20:03 - 20:10: and we think that these correlations that I’ll show you on these graphs might be part of
• 20:10 - 20:16: the explanation. In the cells where leucine signals through acetyl-CoA to mTORC1 when you
• 20:16 - 20:21: have cells in nutrient-rich conditions that have reasonable levels of acetyl-CoA you get
• 20:21 - 20:28: a dramatic reduction when you remove amino acids so you get a dramatic reduction in acetyl-CoA
• 20:28 - 20:34: levels when you remove amino acids and you pretty much restore the levels when you add back leucine.
• 20:37 - 20:44: However in the cells that don’t fit our model like the HEK cells you get very little reduction
• 20:44 - 20:48: in the case of mouse embryonic fibroblasts no reduction in acetyl-CoA levels
• 20:50 - 20:53: when you starve the cells of amino acids and
• 20:55 - 21:00: virtually no increase when you add back leucine to the amino acid starved cells.
• 21:03 - 21:09: Further work that Sung Min Son did led to a new mechanism for autophagy activation after
• 21:09 - 21:16: nutrient starvation that is pertinent to many cell types and it’s a situation we’ve
• 21:16 - 21:24: also got in vivo correlative support for instance in the brain. This is a mechanism where the cells
• 21:25 - 21:34: signal to mTORC1 not via leucine itself but rather via its metabolite acetyl-CoA and it’s a process
• 21:34 - 21:41: which doesn’t require the previously described sensors and this is what Sung Min Son found.
• 21:42 - 21:49: Leucine is converted by the leucine catabolic enzymes to acetyl-CoA and the acetyl-CoA
• 21:51 - 22:00: activates the acetyltransferase EP300 which leads to acetylation of lysine 1097
• 22:00 - 22:08: of this Raptor protein in the mTORC1 complex and this acetylation event is necessary
• 22:09 - 22:19: to allow the Raptor and the whole complex to bind to the RAG proteins and be associated
• 22:19 - 22:25: with lysosomal localization. This leads to mTORC1 activation and decreased autophagy.
• 22:26 - 22:33: Conversely when one starves cells and removes leucine one gets less acetyl-CoA in the cells,
• 22:34 - 22:43: inhibition of EP300, no acetylation of Raptor so the mTORC1 is not active because it doesn’t
• 22:43 - 22:48: associate with the lysosome and in the circumstance one has increased autophagy.
• 22:49 - 22:58: The next part of my talk is going to describe autophagy compromise in various neurodegenerative
• 22:58 - 23:07: diseases and this is a picture of the American folk singer Woody Guthrie who died of Huntington’s
• 23:07 - 23:17: disease. Most of the neurodegenerative diseases that afflict people manifest with the accumulation
• 23:17 - 23:23: of aggregate-prone proteins within neurons. You see this in diseases like Alzheimer’s,
• 23:23 - 23:30: Huntington’s, prion diseases, Parkinson’s and ALS and this is a picture of a traffic jam
• 23:30 - 23:37: in Naples occurring when the bin collectors go on strike and you can see the accumulation of rubbish
• 23:38 - 23:44: which is not only perturbing traffic but also hosts a whole lot of toxic matter and this is
• 23:45 - 23:51: the metaphor that I like to use when thinking about these aggregates or aggregate-prone proteins.
• 23:52 - 23:58: There’s extensive genetic and transgenic data arguing that these aggregate-prone proteins
• 23:58 - 24:04: are toxic for the cell and therefore we’ve been working over the last 20 years or so
• 24:04 - 24:09: to think if we can find ways of enhancing clearance of the mutant protein species.
• 24:09 - 24:15: We and many others initially started off by considering the ubiquitin-proteasome pathway
• 24:15 - 24:22: where target proteins are ubiquitinated by a series of enzymes to allow the attachment of
• 24:22 - 24:30: a ubiquitin chain to the target protein and the ubiquitin chain serves as a recognition signal
• 24:30 - 24:35: for getting the protein to the proteasome where it can then be degraded into peptides.
• 24:36 - 24:42: However, the entrance to the proteasome is very narrow and can only accommodate monomers
• 24:42 - 24:50: that can be unfolded so this pathway cannot be accessed by oligomeric or higher order
• 24:50 - 24:56: aggregated proteins and that’s how we started thinking about autophagy. In the early years
• 24:57 - 25:05: of the century, my lab discovered that autophagy was a key clearance pathway for many of the
• 25:05 - 25:11: proteins that are aggregate-prone and that cause neurodegenerative diseases. These include
• 25:11 - 25:18: polyglutamine-expanded proteins exemplified by Huntington’s disease, tau where wild-type tau
• 25:19 - 25:24: is an important pathogenic driver in Alzheimer’s disease, and mutants of tau
• 25:24 - 25:30: cause frontotemporal dementia, and alpha-synuclein which is a driver,
• 25:31 - 25:38: important driver of pathology in Parkinson’s disease. We found that when we impaired the
• 25:38 - 25:44: formation of autophagosomes or perturbed the fusion of autophagosomes and lysosomes,
• 25:45 - 25:50: we slowed the clearance rate of the aggregate precursors and in doing so increased the
• 25:51 - 25:58: accumulation of both soluble and aggregated species and enhanced toxicity, whether it be
• 25:58 - 26:05: in cell-based models, in neurons, or a range of in vivo systems including Drosophila, zebrafish,
• 26:05 - 26:15: and mice. Because of this, we and other labs have been looking to see whether
• 26:16 - 26:22: neurodegenerative disease or other neurodegenerative diseases are associated with
• 26:22 - 26:27: an impaired autophagy as the formation of aggregates is seen in many neurodegenerative
• 26:27 - 26:33: conditions. The important point of the slide is not the detail of all the different diseases
• 26:33 - 26:38: we have found that are associated with impaired autophagy, but rather the color code.
• 26:39 - 26:44: The diseases in blue represent scenarios where the formation of autophagosomes
• 26:45 - 26:50: is impaired by the relevant mutation, for instance excess alpha-synuclein
• 26:50 - 26:55: in Parkinson’s disease caused by alpha-synuclein gene duplications.
• 26:57 - 27:02: The red scenario is a situation where the autophagosomes are formed properly,
• 27:03 - 27:07: but they’re not properly delivered to the lysosomes and this can occur
• 27:07 - 27:11: if you’ve got mutations that affect the dynein machinery as you see in forms of motor neuron
• 27:11 - 27:18: disease and Parkinson’s. Finally, in order for the autophagy process to be completed,
• 27:18 - 27:24: the lysosomes need to work, and if there are defects at the level of the lysosome,
• 27:24 - 27:35: one can get impaired degradation of autophagic substrates or a perturbation in autophagosome lysosome
• 27:35 - 27:45: fusion. Indeed, there are defects in autophagic flux in lysosomal storage diseases, which is a
• 27:45 - 27:50: project we worked on originally with Andrea Ballabio’s lab, where there are defects in
• 27:50 - 27:56: autophagosome-lysosome fusion, and it’s worth pointing out that these are the most common
• 27:56 - 28:03: neurodegenerative diseases of childhood, as well as variants that give you Parkinson’s disease,
• 28:03 - 28:12: neuropathy, etc. I’m going to focus my discussion today on polyglutamine expansion diseases as this
• 28:12 - 28:19: project reveals a whole group of diseases that have a defect in autophagy due to a common
• 28:19 - 28:27: mechanism. There are nine diseases caused by polyglutamine expansion mutations. In all cases,
• 28:27 - 28:34: the polyglutamine expansion is encoded by an abnormally long CAG trinucleotide tract
• 28:35 - 28:42: in the relevant gene. In all of these diseases, and I’m just showing five of them,
• 28:42 - 28:48: there’s an inverse relationship between the length of the polyglutamine tract
• 28:49 - 28:53: and the age of onset of disease. So, the longer the tract, the earlier the onset,
• 28:54 - 29:02: and of course, there’s a threshold at which the length of the tract becomes mutant versus wild
• 29:02 - 29:10: type, and the length of the tract where this threshold occurs is roughly in the same type
• 29:10 - 29:17: of length across many of these diseases. This raises the question whether there’s a common
• 29:17 - 29:23: mechanism for all of these diseases, and one of the early mechanisms that was proposed was
• 29:23 - 29:29: the fact that these proteins generally form aggregates in neurons in disease.
• 29:30 - 29:38: The idea that the aggregates themselves were the toxic entity became complicated when Steve
• 29:38 - 29:43: Finkbeiner’s lab did elegant experiments in cell-based models of Huntington’s disease
• 29:43 - 29:50: and showed that the cells that formed aggregates actually had a lower propensity to die compared
• 29:50 - 29:56: to those where the protein stayed apparently soluble as determined by light microscopy.
• 29:58 - 30:06: So, the minimum interpretation of those data is that the protein can be toxic in the absence
• 30:06 - 30:11: of visible aggregate formation, and therefore, we and many others have been trying to understand
• 30:11 - 30:16: if there are generic pathogenic consequences of expanded polyglutamine tracts in different
• 30:16 - 30:25: disease proteins beyond the aggregation. The second question that many of us have been
• 30:25 - 30:31: wondering about is what are the roles of normal polyglutamine stretches in wild type proteins,
• 30:31 - 30:36: as these are very well conserved through many species. Indeed, some of the earlier studies
• 30:36 - 30:43: I did when I came to Cambridge was to show that these polyglutamine stretches are conserved across
• 30:43 - 30:49: many species in many of these proteins. So, Avi Ashkenazi in the lab started off by studying
• 30:49 - 30:55: wild type ataxin-3, which when expanded gives you spinocerebellar ataxia type 3,
• 30:55 - 31:04: otherwise known as Machado-Joseph disease. And this is what Avi found. Ataxin-3 is a
• 31:04 - 31:11: deubiquitinase, and it’s a deubiquitinase for a protein called Beclin-1. Beclin-1 normally
• 31:11 - 31:20: is part of a complex that forms PI3P, which I showed you earlier was important for autophagy
• 31:20 - 31:25: by recruiting WIPI2 to sites of nascent autophagosome biogenesis.
• 31:25 - 31:35: Normally, Beclin can get ubiquitinated, but ataxin-3 will remove this degradation signal.
• 31:35 - 31:44: When ataxin-3 function is defective, then the ubiquitins are not removed, and the ubiquitins
• 31:44 - 31:49: on Beclin-1 serve as a degradation signal, leading to enhanced Beclin degradation through the
• 31:50 - 31:58: proteasome. And the attendant loss of function results in impaired autophagosome formation.
• 32:00 - 32:09: The ability of ataxin-3 to act as a deubiquitinase for Beclin-1 is, as one would expect,
• 32:10 - 32:14: regulated by the ability of ataxin-3 to bind to Beclin-1.
• 32:14 - 32:23: This is enabled by the wild-type polyglutamine stretch in ataxin-3, and if you remove the
• 32:23 - 32:29: wild-type ataxin-3 polyglutamine stretch, so you have an otherwise normal protein, but you just
• 32:29 - 32:36: don’t have any glutamines in it at this site, then the ataxin-3 doesn’t bind, and it is functionally
• 32:36 - 32:40: dead with regard to acting as a deubiquitinase for Beclin-1.
• 32:44 - 32:51: So, that reveals a normal function for this wild-type polyglutamine stretch in ataxin-3.
• 32:52 - 33:01: When you have a cell with a mutant polyglutamine protein in it, like mutant huntingtin, then it can
• 33:01 - 33:08: act in trance in this pathway, and the mutant polyglutamine stretch ends up interfering with
• 33:08 - 33:15: this reaction, because the long polyglutamine stretch binds more tightly to Beclin-1 than the
• 33:15 - 33:23: wild-type shorter stretch in ataxin-3, and in doing so, it displaces ataxin-3 from binding to
• 33:23 - 33:31: Beclin-1 and results in a loss of activity and more rapid Beclin-1 degradation. Interestingly,
• 33:31 - 33:37: when one has an abnormally long polyglutamine stretch in ataxin-3 itself, it binds better,
• 33:37 - 33:44: as one would expect, to Beclin-1, but it has reduced deubiquitinase activity, and the reason
• 33:44 - 33:51: for this, we think, is that in practice, in order for a protein to act as a deubiquitinase,
• 33:52 - 33:59: it needs to be able to bind, deubiquitinate, and then come off and bind to other substrates,
• 33:59 - 34:03: and if it’s sticking too long because of this long polyglutamine stretch,
• 34:03 - 34:08: then it’s not going to be able to service other substrates effectively.
• 34:11 - 34:17: So, these data provide a function for the wild-type polyglutamine stretch in ataxin-3.
• 34:19 - 34:23: This is real, doesn’t provide a function of the wild-type polyglutamine stretches
• 34:23 - 34:27: in other proteins, but it’s a start. It tells you about one of the proteins.
• 34:28 - 34:36: It also can provide an explanation for how mutant soluble polyglutamine stretches might cause
• 34:36 - 34:44: toxicity. I should point out at this juncture that we did many of our experiments in conditions
• 34:44 - 34:52: where these proteins do not form aggregates, and so here we’ve got a situation where the mutant
• 34:52 - 34:58: soluble proteins are impairing the wild-type ataxin-3-Beclin-1 interaction,
• 34:58 - 35:05: leading to destabilization of Beclin-1, a consequent impaired formation of PI3P,
• 35:05 - 35:13: and reduced autophagosome biogenesis. It’s interesting to hypothesize that there might
• 35:13 - 35:19: be a feedback loop operating here, because the mutant polyglutamine stretches themselves
• 35:19 - 35:25: are autophagy substrates. So, imagine a situation where, at the start,
• 35:25 - 35:33: one has modest levels of huntingtin and good functioning autophagy. Perhaps as one ages,
• 35:33 - 35:38: the function of autophagy decreases when one passes a certain threshold, and this leads
• 35:38 - 35:47: to some accumulation of huntingtin. This will end up compromising autophagy, and as a result,
• 35:47 - 35:54: one will get even more huntingtin and even more compromise of autophagy, and this might account
• 35:54 - 35:59: for a rapid runaway of pathology once this type of loop kicks in.
• 36:04 - 36:10: I hope to end on some good news, and this is the possibility that one might be able to
• 36:10 - 36:17: upregulate brain autophagy for therapeutic ends, and this musician is a man called Albert Sammons,
• 36:18 - 36:23: who was one of the greatest British violinists of the early part of the 20th century,
• 36:24 - 36:30: and he developed Parkinson’s disease, which must be devastating for someone in that profession,
• 36:30 - 36:35: and so he represents an emblem to search for therapeutic.
• 36:36 - 36:40: When we did our earliest experiments on Huntington’s disease and other conditions,
• 36:40 - 36:46: we found that these proteins that were autophagy substrates,
• 36:47 - 36:55: that their toxicity and accumulation could be enhanced by upregulating the process.
• 36:55 - 37:03: So, if we increased autophagosome biogenesis, we could enhance the clearance of mutant polyglutamine
• 37:04 - 37:12: proteins, tau and alpha-synuclein in cells, Drosophila, zebrafish, and mice, and in doing so,
• 37:12 - 37:19: decrease the accumulation of both the aggregated and soluble species and alleviate toxicity.
• 37:22 - 37:27: At the time we did these experiments, we used rapamycins, which inhibit the mammalian target
• 37:27 - 37:32: of rapamycin complex 1, mTORC1, that I’ve talked about earlier, and we did this because they
• 37:33 - 37:39: were the only drugs known at the time that we used chronically in people that would be predicted
• 37:39 - 37:50: to induce autophagy. Our earlier data and those of others have led us to postulate
• 37:50 - 37:55: a neurostatin model, as I call it, for managing neurodegenerative diseases.
• 37:55 - 38:03: In the context of coronary artery diseases, statins reduce cholesterol in the circulation
• 38:03 - 38:10: of individuals from a young age to protect against heart disease in middle age and later.
• 38:11 - 38:17: I think we should maybe consider the analogous situation for autophagy, where it has the
• 38:18 - 38:25: potential to reduce the aggregate-prone proteins in the brain of individuals by starting at a
• 38:25 - 38:32: young age and thereby protect against neurodegeneration in more advanced age.
• 38:36 - 38:43: A note of warning, though. The strategies I’m going to talk about focus on the induction
• 38:43 - 38:49: of autophagosome formation. While this might be advantageous in normal situations where
• 38:49 - 38:56: autophagy is not perturbed, where autophagosome biogenesis is impaired, as shown by these two
• 38:56 - 38:59: lesions that can cause Parkinson’s disease,
• 39:02 - 39:08: the acceleration of autophagosome formation might not be beneficial when one has defects.
• 39:08 - 39:14: Again, I’ve shown you situations in Parkinson’s disease, either in the dynamin machinery or at
• 39:14 - 39:21: the level of the lysosomes, where the autophagosomes are formed properly. Because if you’re
• 39:21 - 39:28: not removing them properly, one might have a deleterious and unproductive accumulation of
• 39:28 - 39:34: autophagosomes in the face of a drug that’s aiming to increase autophagosome biogenesis.
• 39:35 - 39:40: The point I’m trying to make on this slide is that we’re now in an era where we can consider
• 39:40 - 39:48: personalized medicine and tailor a specific therapy to the genotype or the underlying
• 39:48 - 39:54: disease pathogenesis that one is dealing with. The conditions I’m going to discuss today are
• 39:54 - 40:04: scenarios where one’s got cellular makeup that is amenable to increasing autophagosome biogenesis.
• 40:04 - 40:13: To get to this point, we’ve been undertaking a number of different screens, as have others,
• 40:13 - 40:22: to identify autophagy-inducing drugs and have focused on mTOR-independent pathways because we
• 40:22 - 40:30: think that they might be safer and better tolerated compared to drugs that hit the
• 40:30 - 40:37: mammalian target of rapamycin and therefore potentially be better suited for long-term
• 40:37 - 40:44: therapy to slow the onset of disease. These screens have identified many different compounds that
• 40:45 - 40:50: act independently of the mammalian target of rapamycin to induce autophagy. They clear
• 40:50 - 40:55: aggregate-prone proteins like mutant huntingtin or alpha-synuclein and are protected in cell
• 40:55 - 41:02: Drosophila zebrafish mouse models of Huntington’s disease. One of these, a compound called ranitidine,
• 41:02 - 41:08: as shown here, we’ve even taken into safety trials with Roger Barker in Cambridge
• 41:08 - 41:15: in early Huntington’s patients. Today, I’m going to focus on one of the hits from
• 41:15 - 41:22: one of our early screens done by Andrew Williams, which are L-type calcium blocker drugs.
• 41:26 - 41:33: Andrea’s work led us to suggest that L-type calcium channel blockers can induce autophagy
• 41:33 - 41:41: and we went on to show that they do this in primary neurons too. We validated that the target
• 41:41 - 41:46: of one of these, verapamil, was regulating autophagy because it was acting through L-type
• 41:46 - 41:53: calcium channels. But unfortunately, verapamil, which was the compound that we started working
• 41:53 - 42:03: with, does not cross the blood-brain barrier. So, Faris Siddiqui did a small experiment looking
• 42:03 - 42:08: at a number of different L-type calcium channel blockers that do get into the brain
• 42:08 - 42:16: and compared them to verapamil and assessed autophagy using primary neurons from a transgenic
• 42:16 - 42:20: mouse model that we’ve made that allows us to quantify the number of autolysosomes and the
• 42:20 - 42:26: number of autophagosomes and found that all these L-type calcium channel blockers, as did
• 42:26 - 42:32: verapamil, increased the number of autolysosomes, suggesting those more autophagic flux. She then
• 42:32 - 42:39: transfected primary neurons with mutant huntingtin and scored the number of aggregates in these
• 42:39 - 42:45: neurons and compared them to positive control compounds here, verapamil, rapamycin, and
• 42:45 - 42:51: another autophagy-inducing trehalose and found that all these L-type calcium channel blockers
• 42:51 - 42:57: similarly reduced the number of aggregates compared to the control state. So, emboldened
• 42:57 - 43:02: by these data in primary neurons with these L-type calcium channel blockers that get into the brain,
• 43:04 - 43:12: Ana Lopez in the lab tested for felodipine and verapamil in zebrafish tauopathy models
• 43:12 - 43:17: and found that these L-type calcium channel blockers reduced the insoluble tau levels in
• 43:17 - 43:26: the zebrafish and protected against the tau toxicity. This worked when fish had a wild-type
• 43:26 - 43:32: autophagy background, but not when we knocked out a key autophagy gene, suggesting that these effects
• 43:32 - 43:42: were autophagy-dependent. Based on that, Farah then proceeded in mouse studies to assess the
• 43:42 - 43:49: dose response by treating mice with felodipine, which was the best performing of the drugs that
• 43:49 - 43:56: we saw on the previous page, and did initial intraperitoneal injections and acutely measured
• 43:56 - 44:03: autophagy in the brain with this autophagy reporter a few hours thereafter. She found
• 44:03 - 44:09: that 5 mg per kilogram had the best induction of autophagy into different parts of the brain.
• 44:10 - 44:17: On the basis of that, she used intraperitoneal injections of this drug to treat a Huntington’s
• 44:17 - 44:23: mouse model for six weeks. Then she scored the number of aggregates in the brain as a measure
• 44:23 - 44:29: of autophagic substrate clearance. She found, again, that 5 mg per kilogram resulted in the
• 44:29 - 44:35: greatest removal of the aggregates. On the basis of that, she selected 5 mg per kilogram
• 44:36 - 44:43: for an efficacy study in mouse, where she measured four different behavioral parameters
• 44:43 - 44:50: in these Huntington’s mice, grip strength, tremor, a Y-maze task, and rotarod.
• 44:50 - 44:56: And she found that all four of these behavioral tasks were improved in the Huntington’s mice
• 44:57 - 45:02: by the felodipine. And it’s important to point out that this drug had no effect
• 45:02 - 45:06: on any of these parameters in wild-type mice.
• 45:09 - 45:15: We then did pharmacokinetics experiments of felodipine and showed that when we gave the
• 45:15 - 45:23: drug intraperitoneally to mice or orally to mini pigs, we got an accumulation in the brain
• 45:23 - 45:29: compared to the plasma. And that was the good news. However, the bad news is that felodipine
• 45:29 - 45:38: is normally dosed to people for hypertension at 2.5 mg to 20 mg per day. And if one takes a 10
• 45:38 - 45:47: milligram daily dose, the sort of typical dose that people would get, then the previous studies
• 45:47 - 45:52: had suggested in the literature that this would translate to a peak concentration of about
• 45:53 - 46:00: 11 or 12 nanograms per ml. However, the 5 milligram per kilogram IP dose
• 46:00 - 46:04: gave a maximum concentration of 100 times higher in the mouse model.
• 46:05 - 46:10: And so we were concerned that if the concentration was so high in the mice,
• 46:11 - 46:16: this would risk off-target effects, which would make these studies irrelevant to autophagy
• 46:17 - 46:22: or might cause side effects. And this would make the drug unsuitable for direct repurposing.
• 46:23 - 46:25: So in order to pursue this, FARA designed
• 46:28 - 46:31: a mini pump protocol with subcutaneous mini pumps
• 46:31 - 46:39: to achieve plasma concentrations in the mice that were similar to those I reported in people
• 46:39 - 46:45: taking the drug for hypertension in the previous patient. So she got this right, so she could
• 46:47 - 46:52: dose the mice in a way that she was mimicking the plasma concentrations that you’d seen people
• 46:52 - 46:58: taking the drug. And when we measured the concentration in these mice, in their brains,
• 46:58 - 47:02: we found that they had a brain concentration of about 105 nanomolar.
• 47:04 - 47:12: 105 nanomolar felodipine induces autophagic flux in primary neurons,
• 47:12 - 47:18: reduces mutant huntingtin aggregates in primary neurons, and reduces the accumulation of another
• 47:18 - 47:23: autophagy substrate, the mutant form of alpha-synuclein A53T, which gives you forms
• 47:23 - 47:33: of Parkinson’s disease, which we studied in iPSC-derived neurons, which would be a human
• 47:33 - 47:45: neuron model. When she dosed mice with this human-like dosing protocol, she found that
• 47:45 - 47:50: in the Huntington’s mouse model, that the felodipine significantly reduced the number
• 47:50 - 47:56: of aggregates in two different parts of the brain. Unfortunately, when we use these mini-pumps,
• 47:56 - 48:02: we can only treat the mice for a month. So we needed then to find a mouse model
• 48:03 - 48:08: where one month’s treatment would be sufficient to allow us to see if there were any effects
• 48:08 - 48:15: on neurodegeneration. And so by chance, John Sukaribik in the lab had characterized a mouse
• 48:15 - 48:22: model overexpressing this A53T alpha-synuclein mutation, and she found that it got sick for
• 48:22 - 48:28: about six months of age and showed clear progression from six to seven months. So Farah
• 48:28 - 48:35: dosed the felodipine using the human-like protocol from six to seven months in these mice, and she
• 48:35 - 48:40: showed that when she did this, she could reduce the levels of insoluble alpha-synuclein in different
• 48:40 - 48:48: parts of the brain. In doing so, she could rescue the neurodegeneration as scored by the number of
• 48:48 - 48:53: tyrosine hydroxylase positive neurons and improve motor performance in these mice.
• 48:53 - 49:00: So we’re very excited because these data suggest that felodipine is inducing autophagy and is
• 49:00 - 49:05: neuroprotective at plasma concentrations in mice equivalent to those that would be seen in people
• 49:05 - 49:11: taking the drug with hypertension. We’re even more excited when we went back and carefully
• 49:11 - 49:17: looked at the literature because there are a number of studies that have suggested in
• 49:17 - 49:24: epidemiological analyses that prior use of L-type calcium channel blockers that get into the brain
• 49:24 - 49:30: like felodipine reduce subsequent Parkinson’s disease risk. So on the basis of this, we are
• 49:30 - 49:39: planning with Roger Barker some experimental medicine studies in patients next year in order
• 49:39 - 49:46: to take this further. So I’ve told you many things today, and I’d just like to leave you with my main
• 49:46 - 49:52: conclusions. First, that autophagy upregulation might be a rational therapeutic strategy for
• 49:52 - 49:58: many neurodegenerative diseases because it removes toxic aggregate-prone proteins. I haven’t reviewed
• 49:58 - 50:04: the literature, but we and others have found that autophagy protects against various forms of cell
• 50:04 - 50:10: death. I’ve shown you data suggesting that it’s protective in various disease models, and these
• 50:11 - 50:18: benefits can be achieved by drugs that target the mammalian target of rapamycin and mTOR independent
• 50:18 - 50:25: pathways. On the other hand, autophagy compromise is frequently seen in neurodegenerative diseases,
• 50:26 - 50:34: and if this occurs, was predicted to increase aggregate formation and enhance cell stress.
• 50:35 - 50:40: I think it’s important to understand the biology of the disease in order to see whether it’s going
• 50:40 - 50:47: to be suitable for autophagy upregulation therapies, and understanding biology might
• 50:47 - 50:52: also inform where best to intervene in the autophagy pathway itself.
• 50:55 - 51:00: Finally, I’d like to acknowledge that this is work done over many years by many people in my lab,
• 51:00 - 51:06: and I’m very lucky to have a fabulous group of people working with me. I’ve had great collaborators,
• 51:06 - 51:10: and of course, I couldn’t have done this work without the generous funding we’ve had
• 51:10 - 51:20: over the years. Thank you very much for listening.