Vincere's Technology Platform for Conquering Age-Related Decline

Vincere's Technology Platform for Conquering Age-Related Decline

Show Video

At Vincere Biosciences we are on a mission to conquer age related decline and we are starting with Parkinson's disease where we know that aging is the biggest risk factor. Today, I'm going to talk about the technology platform that powers our repeatable asset invention. I'm going to walk through four tools today that we have built in in use. The first to prioritize targets, the second how we validate these targets for human disease. Third, how we select molecule hits that can modulate these targets.

And finally how we select the patients to are most likely to respond to these drugs that we're developing. So I'm going to use Parkinson's disease as a use case throughout this presentation and the question we want to know is, what do the genes that cause Parkinson's disease have in common. We know there are a dozen or so genes with Mendelian inheritance that seemed to be across a variety of pathways and processes.Well if you

look here at this schematic and you think about each of those proteins depicted in red having a community of other proteins that they interact with inside the cellular system, you can then start to get an idea of the process and what behavior inside the cell that protein is responsible for When you put those together you can start to identify points of overlap where there are shared interactors of two or more of the genes that cause Parkinson's disease. If we then rank order by the genes that are members of the most communities of interactors, the ones that share the most overlap with all the forms of genetic Parkinson, we think that's a really good place to look for what's really going wrong in the disease and where the most likely effective targets will be. We know these genetic forms, while they only represent a fairly small percentage of Parkinson's patients are the only place where we really have cause and effect relationship, where we know this single point genetic problem leads to Parkinson's disease and all of these forms share the same selective loss of nigrostriatal dopamine neurons and the same accumulation of protein aggregates and motor symptoms and all of the other things that go along the disease. So when we do that

for Parkinson's, we can then, we come up with this new set of genes that is much larger than just the dozen or so genes that we have with purely human genetics and so we can say what are the biological processes that are impacted by that. When we do that for these twelve Mendelian genes we find that mitochondrial quality control is by far the most highly over represented biological process. All of these forms converge on disrupting mitochondrial quality control. There are several other ancillary mitochondrial processes that rise to the top here. Oxidative stress apoptotic processes, electron transport chain and several others along with some of the things you might expect including inflammation and ubiquitination and trafficking ,things that are more directly related to some of the other forms.

So once we have this idea of, we know mitochondrial quality control is a central process that is disrupted in Parkinson's and we also have this genetic network of things that are related to all of the forms of disease we can zoom back in on the overlap there of which of the mitochondrial quality control genes are also within this overlap of disease and we can get an idea of which of those are the most heavily connected. We think these ones that share a lot of interactors within this are probably more influential. They may be places where you have points of convergence where a modulation might have a more broad effect and so we can think about those as maybe nodes of influence that we might want to test for therapeutic benefit.

So the next question we ask is alright, how can we validate that this query into the genetic network has provided things that are really relevant to Parkinson's disease. The next perspective we looked at is are any of these changed in the brains of Parkinson's patients. So here we took laser captured dopamine neurons from idiopathic Parkinson's disease patients and looked at whether these proteins are increased or decreased in those post-mortem brains and what we find is that indeed many of these are modulated, some increased and some decreased. So this gives us pretty

good confidence that what the network produced from this first approach really did point us in a direction that is relevant not just for genetic Parkinson's but also for sporadic Parkinson's. We can then start to look at those and see which of these are potentially druggable. Are they classes that have had therapeutic success before or are there any initial starting chemical matter, has anybody tested these for knockout or over-expression studies and what we do when we look within that are two primary programs. USP30 and Parkin arise

as good druggable targets within this. Now within the normal cellular process this is all of the cells in your body there are mitochondria which you may know are the powerhouse of the cell but they also do quite a few other functions around cellular stress response and some signaling pathways, well the mitochondria normally live for about thirty or forty days and when they become damaged they have to be recycled through a process called mitophagy and normally when this happens the damaged mitochondria signal for the recruitment of parkin which is E3 ligase that adds ubiquitin chains to the outside of the damaged mitochondria. When it does this, that then recruits the autophagosome machinery that carries that off to the lysosome where it's broken down into it's components which can be recycled for generation of new mitochondria. That is normally a balanced homeostatic system but in Parkinson's disease and aging we see that that becomes unbalanced and there starts to be a much larger number of damaged mitochondria compared to the healthy mitochondria. So if you

can increase the activity of Parkin you can add those tags more quickly, clear those damaged mitochondria more quickly and restore the homeostasis bringing the cellular system back to health. USP30 is the deubiquitinating enzyme that chops those tags off. For every process within the cell there's a yin and yang. There's you know, if parkin is the accelerator that speeds up mitophagy, USP30 is the inhibitor that, the brakes that hold that back so that you don't get runaway processes. So by either activating the activator in parkin or inhibiting the inhibitor in USP30, we can restore this process back to homeostasis and bring the cell back to the healthy system.

So the next question is, okay we've identified a couple of targets that we think make mechanistic sense. We can make a good story of the way these interactions work. The question is will these overcome deficits that we see in Parkinson's disease. So the next method that I'm going to talk about is our simulation of biological systems and this is a physics based simulation where we can take omics profiles from patients that tell us how many proteins and lipids and other things that are inside the cell are in each the particular cell type. We can then use particle

physics to simulate that a virtual cell in the computer and along with a library of biological interaction rules that define what happens when two proteins come in contact in this cellular environment, we can track the state changes and positions and behavior of the cell and then start to compare different cell types. We can compare a Parkinson's patient brain cell to a healthy human. We can compare a brain cell to a mouse brain cell all to a human iPSC and so this allows us to do some really interesting target validation work that just isn't possible otherwise. We can run experiments on human substantia nigra cells which can't be done in the real world. We can also extrapolate from what we see in a model system like a mouse or or an iPSC to what we would expect to see in the patient brain and we think that's a really powerful way to set your hypotheses around what level of modulation you would need for a therapeutic effect and whether a target is actually likely to be effective in your target tissue, in this case the substantia nigra of Parkinson's disease patients.

So this is a new method that we've developed in-house. We actually hold the US patent on the simulation platform and have quite a bit of proprietary data that drives the system. So the way that we do this is to compare to the real world. So we started

by doing simulations of things that had been published and then see do our simulations match the previously published work and then the next step was can we take simulation predictions and then find things that have never been shown in the real world, test them in animal and cell models and show whether our predictions are right. So we did a couple of projects funded by the Michael J. Fox foundation, one with the Mayo Clinic where we did simulations of transgenic mouse brain cells and then they validated those in their G2019S LRRK2 transgenic mouse lines, this is a gene that causes human Parkinson's disease and then with the Mayo Clinic or with Oxford rather, we did a collaboration where we took omics profiles from GBA patient derived iPSC cells and then simulated those GBA iPSC cells in the computer, made predictions of what would change there and then went back and did profiling in those iPSC's so between those projects we see a really good predictability. If you look at the R², the pearson correlation coefficient of the predicted data versus the data from the laboratory system, we see an R² .8, one would be a perfect correlation, zero would be none. So this is very close to a perfect correlation, very good for a biological simulation system. We're quite proud

and happy about that level of of validation. So now I'm going to step through some of the data that goes along with this, that has come out of that. So we looked at in this set of experiments idiopathic Parkinson's disease, substantia nigra versus healthy control substantia nigra, each match controls and what you see normally there is very little Parkin enzyme on the outer mitochondrial membrane but when the mitochondria become damaged you get this really robust up-regulation of Parkin. Parkin comes to the most mitochondrial membrane to start to kick off that mitophagy process. Parkinson's disease patients and these are idiopathic Parkinson's patients brains without one of the mutations that directly link to this. They have a

very much reduced ability of Parkin to be recruited. Reduced recruitment of Parkin to the membrane. You see then the same deficit cascade into the substrates of Parkin as you might expect, so Parkin adds ubiquitin tags to membrane proteins like VTACs and mitofusins and we see that those are quite a bit lower levels of ubiquitination in the simulated cells. So having this deficit now in the virtual cell well we wanted to see can we overcome that so we started looking at testing some of our hypothesized targets like USP30 and Parkin to see can inhibiting USP30 or activating Parkin overcome this deficit and if so how much modulation would you need and then we see that indeed, fifty percent inhibition of USP30 or thirty per cent activation of Parkin are sufficient to reverse that deficit in the idiopathic Parkinson's brain cells. So this is great. This is all

work that we did several years ago in the computer to validate that mitophagy is a key process that has disrupted in Parkinson's and that USP30 and Parkin would be great targets to modulate. The next challenge is can we get chemicals that can modulate these targets. Now Parkin's been quite a while validated target in Parkinson's for quite a while as the name suggests it's clearly related to Parkinson's disease.

The challenge is you're trying to activate, increase the activity of an enzyme which is incredibly difficult with a small molecule to do. We do think that there's some reason to believe it can be done because Parkin is normally in an autoinhibited state so if you can stick a compound to it in the right place you may be able to kick it into an active form more quickly or to hold it into an active form. USP30 is an emerging target that has less published evidence in animal models of the disease but that you're inhibiting an enzyme that is much more tractable from a small molecule perspective.It's pretty easy to stick something to an enzyme and have it stop doing what it normally does. So we took a couple of approaches with those and i'm gonna talk about with Parkin , a virtual screening approach that we used where we screened a series of small molecules , simulated chemical structures against a pocket in the crystal structure of the 3D structure of the protein so if you take the Parkin protein and create a 3D model of how it's folded and it's shape and that's been collected through some crystallography experiments that other academic groups have done that of sure shown the shape that Parkinson when when it's fully autoinhibited, the shape that it turns into once it's been partially activated and then the shape it's in finally when it's fully active and so from that we were able to identify a pocket that we believe if you stick a compound would increase the activity of Parkin. Then we curated a library of over ten million small molecule compounds that are commercially available and fit a good drug like profile they they have all of the properties that would make them likely to be druggable.They don't have any

structural alerts, they're not known to be toxic things like that, and then we ran a simulation where we tried many iterations of different confirmations of sticking each of those ten million compounds into the pocket that we predict would be effective on Parkin. When we did that we had eighty five hundred small molecules that looked like they would bind to that pocket. We had a team of chemists that looked at those and voted to pick two hundred and fifty that they thought were the most likely good drug starting points that had a good diversity of structures and ordered those and tested them in the laboratory in an enzymatic test where we could see the normal levels of Parkin enzyme in transferring ubiquitin and then whether putting our compounds these predicted compounds into that test-tube if it changes the activity level of Parkin. What we see there, we're very excited to find that indeed there are we came up with multiple modulators of this.

One of the things with a virtual screening you can predict that a compound will stick but you can't predict what the enzymatic activity change would be. We did identify multiple activators of Parkin that came out of that which was our hypothesis of what happens at target. We also identified several inhibitors of Parkin that reduce the activity so those obviously won't be relevant for Parkinson's disease but they are good validation that the simulation approach worked to find things that can modulate, that can bind to our target enzyme. These also may have

applicability to other diseases outside of Parkinson's. An emerging area is potentially considering them for other uses outside of direct modulation. So once we have small molecules that can modulate a target of interest then the next big challenge is how do we know which patients to test these in, and this is an area that we're just starting to do some development that's pretty exciting. One of the things that we think is important is to start to look at multiple types of data together and we've begun to gather the genetics of disease along with the protein interactome along with patient data including the transcriptomics and proteomics that tell you how many of the proteins are there, lipidomics that tell you the different balances of lipids that are on the membranes and within this system methylomics that tell you how the DNA is marked up, this is epigenetics that you see changing throughout the age of a person and with all of that together we can start to drive AI and machine learning models that can classify different types of patients out of that and also look at progression a long time and so as a proof of concept using these types of models we started working with some of the biological clocks that have been emerging over the last ten years or so really spearheaded by the work of Steve Horvath with the methylation clocks that can based on the tagging of your dna at these methylation sites you can identify a small number, a few hundred methylation sites that if you look at those together can predict a person's age. It's able to based off of a saliva or blood sample predict your age within a range of three or four years. We've been able to take that, improve on it with more cutting edge, state of the art deep neural techniques and identify now some great ways to identify better sets of of markers that correlate to the disease or to the progression of the state and in our proof of concept looking at biological age clocks and then now looking at Parkinson's disease where we've imported tens of terabytes of data from the PPMI Michael J. Fox foundation's Parkinson's

progression marker initiative to to look at multi-omics, kind of ensemble predictors of disease state and so what we're looking to do at this point are to find signatures of patients that have deficits in mitochondrial quality specifically, so we should be able to pick out the patients off of a blood test that are most likely to respond to Parkin activators or USP30 inhibitors that we have moving close to the clinic and this is really enabled by our data that describes the biological system and identifies which markers are relevant to this pathway and the function of these drugs and then with the longitudinal data on patients that we have where we can start to look for which of those patients have disruptions in the enzymes in markers that are related to the enzymes that that are around are targets of interest we can look at things like looking for the patterns that emerge within the PINK1 and Parkin and mutation carriers and then see if those same patterns exist within the larger idiopathic or other genetic forms of patients. So I'm very excited about that work and we think it's critically important to have good mechanistic markers that can define which patients are most likely to respond and then ways to look at those same markers to validate that your drugs are having an effect. So with all of that we now have this platform for drug discovery that has computational technologies at almost every step of the process from hypothesis generation through to IND-enabling studies and clinical trial design coupled with laboratory methods that go back into the real world for enzymatic assays and synthetic chemistry and cellular assays of disease, animal models of disease and target engagement safety profiling, and PK, and so on. With that we now have assets that are in development for USP30 inhibitors targeting both Parkinson's disease and multiple peripheral indications. We have a Parkin activators that we're developing for Parkinson's disease. We have a fourth target that's completely unrelated that we are in the earlier hit validation stage around and then a whole pipeline of targets that have been identified from the computational platform that we think have good rationale but we've not yet moved into the chemistry stages. Our plans are

to expand out from here into more general autophagy and then also back into other areas of mitochondrial health beyond mitochondrial quality control and clearance. So, finally I want to touch on kind of the fruits of our labor and look at our lead programs we now have small molecules that are highly potent in the lone animal arrange and highly selective for USP30 compared to other DiVA coordinating enzymes. We have demonstrated that these compounds are able to to increase mitophagy in human neural cells in a petri dish and importantly they very selectively boost mitophagy of damaged mitochondria so once you add a stressor like a oligomycin or antimycin that depolarize mitochondria membranes they, our compounds increase mitophagy very robustly in that case with only a subtle increase of mitophagy in the normal baseline system without a stressor. That's really important for the safety profile of these compounds. We have now shown that these compounds are also able to increase mitophagy in rodent brains so in this case we dosed mice for five days, also a great signal of the safety profile, at least the initial safety profile of these compounds and these were using the transgenic mt-Keima mice developed by Nuo Sun and Toren Finkel's group at NIH. Nuo's now at the - Dr. Sun is now at the Ohio State University and these mice have a transgenic reporter or have a fluorescent reporter on the mitochondria that changes to a different wavelength and the acidic lysosome so you can actually measure mitophagy in the in vivo system so using this we were able to demonstrate that a basil mitophagy has increased in the rodent brain after the dosing with our compounds.

Again, it's a fairly subtle increase of the basil mitophagy that we're now looking forward to repeating this pilot experiment with larger groups and also moving on to testing the compounds in models of cellular stress such as MPTP, rotenone, and other other models where we know there's mitochondrial damage happening in vivo So with that I would like to thank you. My contact information is here if you have any questions and would like to follow up, I'm always happy to chat. Thank you.

2022-08-13 07:34

Show Video

Other news