Scientist Stories: Rebecca Feltham, PROTAC-based Targeted Validation Strategies

Thank you very much to the organizers for giving me this opportunity to present my work. So what I'm going to take you through is a recent publication which is looking at a new ProTag-based target validation technology that we've developed. And so my name is Becky and I am a laboratory head at the ubiquitin— in the ubiquitin signaling division at WEHI in Melbourne, Australia. Okay, so I wanted to start off with this, uh, review that I read which really made an impact on me. And so within this review, they mentioned that drugs fail in the clinic for two basic reasons. So they either don't work or they prove to be unsafe. And what I really liked was this quote from David Sienkowski, who was the director of biotherapeutics at Zencor. Who said that both of these reasons are often the direct result of sloppy early target validation. And I really like this because I, I agree that target validation is incredibly important. And this report here is getting a little bit old now, but I think it still was a really important study. And here they showed the probability of success of taking a drug from Phase 1 clinical trials all the way to the market. And what they showed was that the probability of success was around 9.6%, which is, which is actually really low. And they mentioned in this report that there are a number of reasons as to why this might be. So there could be internal funding decisions, internal commercial decisions, but they also mentioned that there was a clear lack of clinical validation of target proteins, and this was contributing to this reduction in the success of drugs that can actually make it to market. And so target validation is something that we really need to focus on and we really need to invest time and energy and effort into before we start investing that, that money and that effort into developing a drug against the protein of interest. And so this is where we got really excited about this, this tag technology. We got excited because this allows us to virtually remove any protein within the proteome by just attaching a small generic tag to it. And so what it involves is using a tag, and I'll discuss some of these tags in the next couple of slides, and fusing your tag to your target protein of interest, and then using a generic degrader that interacts with the tag and interacts with an E3 ligase complex to allow ubiquitylation and degradation of that fusion protein. And this gets around the need of having a degrader that interacts directly to your target, which we don't always have. Because you can use these generic ones that interact with the tag. Now what's really powerful about this technology is that we can now chemically tune destruction or degradation of virtually any protein within the proteome. We can also reversibly deplete proteins, and this is really important because this is more closely recapitulating real drug treatment scenarios, a lot more closely than, say, genetic knockouts where we knock out a gene from, from birth. And this also allows us to, unlike knockouts, to reduce the levels of a protein in an adult mouse in a particular disease situation. And then when we remove that drug, we're able to bring those protein levels back. So I think it's a really powerful system for target validation. Now there's a number of different tagged degrader systems that are available. And so we have this FKBP system which uses this small FKBP mutant, FKBP2, tag, which is based off the FKBp12 protein. And we have degraders that interact with this mutant FKBp and recruit cereblon E3 ligase complex. And just for simplicity throughout the talk, I've renamed the degraders according to the tag that they interact with and the substrate receptor that they bind. So this would be an Fc degrader. We've also got degraders for the FKBp system that recruit VHL.. And we have these FV degraders. Outside of the FKRP tag, we have other tag systems like the Halo tag system. And for the Halo tag system, we've only got degraders for VHL. So we haven't got cerebellum degraders. And I'm just mentioning here one Halo degrader that was published in this Toebel paper in 2018, because that's the one that we've used in our study. But there are other Halo protacs like Halo protac-3 that were published in 2015. Now outside of these systems, these are sort of the main systems that are available at the time we started this study. There's also the BromoTag system. And so the BromoTag system was published recently and recruits the VHL E3 ligase complex. However, as I said, the BromoTag system was published after we did this study, and so we haven't actually included this system within our study. And so when we started looking at these different tag systems, we started to look at what were the benefits and what limitations of each system when we're thinking about choosing which system we want to use to validate our protein. Now the FKBP mutant tag system is actually really good because the tag is very small, so it's only 12 kilodaltons, and we have degraders against this tag for the Recruit-Cereblon. That recruit VHL. The other nice thing about the degraders that interact with the FKBP tag is that they function substoichiometrically, and so they are catalytic degraders. But the, the problem with the FKBP tag is that there's really not many other tools available for it. And if we're going to go to the effort of knocking the tag into the genomic locus of our protein of interest, or potentially generate an animal that stably expresses this tagged form of the, of our protein of interest, we might want to do more than just degrade the protein. And so having other tools that are available that allow us to interrogate the biology of our protein with the tag that we've chosen can actually be quite important. Now the HaloTag is a little bit bigger, so it's about 33 kilodaltons, around about the same size as GFP. However, we only have VHL degraders, so we don't have seroprotein degraders. So that's one potential limitation. The other limitation is that these are non-catalytic degraders, and I'll explain a little bit about this on the next slide. But what's nice about the HaloTag is that we have lots of different tools available through Promega that you can buy commercially to interrogate the function of your, of your protein, to look at the localization of your protein, just explore the biology a little bit more. And at the same time, you have degraders where you can induce degradation. So given that these were the only two systems, sort of ProTag-based systems that were available at the time, we looked at other tags that might actually be quite beneficial to have degraders against. And the Nanolook tag was interesting to us because the Nanolook tag is quite small, so it's around 19 kilodaltons, and the Nanolook tag has intrinsic luminescence properties which can be quite a powerful technology. Now at the time there were no Nanolook degraders available, so there was an opportunity there. But nicely similar to the HaloTag, because Nanolook is a trademark of Promega, there's lots of different tools available for the Nanolook tag. So you've kind of got that flexibility. If we did have degraders where you've got the luminescence capability, you could degrade the protein but also study the biology of the protein. And so what I mentioned before, that the FKBP degraders function substrachometrically, this is because these degraders can bind the FKBP tag reversibly. And so they can interact with the tagged fusion protein, trigger ubiquitylation of the fusion protein, and then as that protein heads off to the proteasome to be degraded, this PROTAC can then re-engage more substrate. Whereas the problem with the HaloTag system is that there's a covalent link that's formed between the HaloTag protec and the HaloTag. And so what actually happens is that this HaloTag protein and the compound is actually consumed in the reaction. And so you need to maintain a 1:1 stoichiometry in this system to be able to get efficient degradation. Now when we looked at Nanolook binders, we were actually quite pleased to see that Nanolook binders had been published. And this is a publication that came out of ProOmega, and the lead chemist at ProOmega on the Nanolook project was, was Joel Walker. And so we reached out to Joel and told him about our idea to generate Nanolook protacs, and he was quite willing to collaborate, which was which was really great. And what's good actually as well is that these Nanoluk inhibitors were reversible inhibitors of Nanoluk. So that gave us the opportunity to produce these substoichiometric degraders. And so through a really great collaboration internally with the chemistry biology division, but also externally through, through Promega, we were able to generate a series of Nanolook cereblon degraders. And these were produced by the lead chemist Christoph Groman on this, on this project. And we also at the same time produced a series of Nanolook VHL protacs. The Nanolook VHL protacs though didn't work. So I'm not going to take you through any of that data. But if you're interested, it's in the paper. And I'm just going to focus here on these Nanolook Cereblon protacs that we made. Now ultimately, what we wanted to do was produce a new protac, but also we wanted to be able to benchmark and compare all the different tag systems against each other to see which one was the best. And so we produced this reporter protein here that contains the different tags we're interested in. So it's got a Halo tag, it's got a mutant FKBP tag, and it's got an analog. Then we've also added this reporter protein in here, Firefly. We then made a very similar reporter protein which instead of Firefly contains EGFP, and we use them for the different assays. Okay, so just testing our Nanolook Cereblon degrader series first. What we used was our Firefly reporter, and here we're reading out Firefly luminescence. These are stable cells that express this reporter protein, and everything is under doxycycline induction. So when we induce with doxycycline here, we can see really nice firefly luminescence. And when we treat with our different degraders, we can see here with Nanoluxera-14, NC4, that we see a really nice reduction in that firefly luminescence, almost back down to the baseline levels. So that was really encouraging. Now just to make sure that this is actually degradation, we did a Western blot. And we can see that again, we see a very similar pattern to what we saw with the luminescence. We can see a really beautiful degradation of this construct here with Nanolox cerebellum-4. And this, just to make a note here, this Nanolox cerebellum-star is our inactive compound, and this is not able to induce the degradation of the construct. So That's nice to see. Now just one comment on linker length. Now I know we've seen in the literature quite a lot that linker length can make a huge difference, and it does make a huge difference here as well with our nanotags. Now I'm not a chemist at all, I'm a molecular biologist, but I actually find this really remarkable. And so the only difference between Nanolook or NC4, NC3, and NC5 is this little region here within the linker. And the difference between NC4 and NC3 is that we have an extra 2 carbons, and NC4 and NC5 is that we have one less carbon. And so when we look at how that actually impacts degradation, we can see that here NC4, which is our best compound, triggers nice degradation. But if we remove one carbon within that area of the linker, we almost completely abrogate the degradation of that construct. And we see this with our firefly construct and also with our EGFP construct where nanoluc serilon-5 doesn't degrade nearly as well as nanoluc serilon-4. And this to me is really interesting, and it can be due to steric clashes that can occur between the substrate and the E3 ligase. And we see this for a number of different protacs. So with halo linkers— sorry, with halo protacs, we also see that longer linkers are actually more favorable, similar to the BTK cereblon degrader. Whereas also it's important to acknowledge that in some situations shorter linkers can actually be more favorable. But yeah, I actually find that this is quite interesting that you can take out one carbon within the linker and completely abrogate the degradation of the substrate. Okay, so focusing now on Nanolox Cereblon4, we did a number of characterizations. I won't take you through all of them because I want to move on to the comparisons of the different systems. And so what we did was we took the different systems that were available to us at the time. So the Nanolox system that we just generated, the two FKBP systems which recruit VHL or cereblon. And for the cereblon system, we actually had two degraders, FC1 and FC2. And the Halo-VHL system. And again, for this Halo-VHL system, we had two degraders, HV1 and HV2. And we're really lucky through a collaboration with Alessio, we were able to obtain both of these degraders for this study. And so using our reporter protein, we're able to compare the degradation of each tag system against the same reporter quarter protein. And so what we can see here again, we're looking at firefly luminescence on the axis. When we add doxycycline and induce the expression, we get a nice induction of the luminescence. And then if we compare treatment with our different degraders, so I've just color-coded everything here. So we've got our FKBP degraders in pink, our HALO degraders in green, and our Nanolab degrader here in gold. We can see over concentration different concentrations that we can see a really nice titratable degradation of that construct. And what was encouraging was that our Nanolook system here does compare to the other systems. Overall, I would say that the FKBP system is superior to the other systems, but it was nice to see that our Nanolook system was actually functioning quite similarly. We also see this characteristic hook effect here with an analog cereblon compound where we start to lose degradation at the high concentrations. And interestingly, we don't see that hook effect, or we don't see as much of a hook effect with these FKBP systems. And this could be due to the FKBP system forming a more cooperative complex with this particular substrate, which overcomes that hook effect. Okay, so just to make sure that the degradation that we were seeing was actually mediated by the proteasome, we used a proteasome inhibitor MG132 to see if we could block the degradation. And so again, using our reporter system, if we treat with our different degraders, we can see a really clear reduction in the luminescence with all of our different degraders. If we then co-treat with the proteasome inhibitor MG132, in all of the cases, we are able to rescue the degradation to some degree. So that was encouraging. When we submitted the paper, the reviewers asked us to look at the durability of each tag degrader upon washout. And so we did this study. And basically what we did was again taking our reporter protein, this time using the EGFP version. We treated with each compound for 5 hours, and then we removed the compound and replaced media and cultured the cells for either 24 or 48 hours to see how long we could sustain the loss of that protein over time. And so the FKBP degraders actually sustained the degradation quite well. And so you can see we get, especially with the FKBP-VHL and FC1 degrader, we can get nice sustained degradation to maybe around 50% even after 48 hours. Whereas with our NC4 compound, even though we get really nice degradation at 5 hours, we get almost complete recovery of that after 24 hours. But this is also sort of an important consideration, because if you wanted a highly reversible system, you might choose something like the Nanolook system, whereas if you wanted something that sustains knockdown for an extended period of time, you might choose the FKB. System. Now there is a limitation to the Nanolook SERAB1 system that we generated, and that's that the Nanolook inhibitors are actually just that— they're Nanolook inhibitors. And so they inhibit the ability of the Nanolook enzyme to convert the substrate to produce luminescence. And you can see that demonstrated here. So if we compare our Nanolook SERAB1-4 compound to our inactive compound, this star compound. And here we look at nanoluke luminescence, so not firefly but nanoluke luminescence. We can see that we reduce the luminescence with our, with our nanoluke Sermon 4 compound, but with the star compound we also reduce the luminescence. And so that's really problematic if you want to use nanoluke luminescence to measure degradation because the protax although they do trigger degradation, they're also inhibiting the ability of the enzyme to trigger, to convert that substrate. And so we wanted to make this a little bit more accessible and see whether we could find a concentration and a window where you could actually use nanoluminescence to monitor degradation. And so for this, we titrated down and we reduced the concentration of the degraders to try and find a concentration where our NC4 compound would trigger reduction and that would represent degradation, but our control compound wouldn't trigger a reduction because it wasn't at a high enough concentration to inhibit the enzyme. And so if we use it at a low concentration like 15.6 nanomolar, you can see that we do actually reduce the levels of the reporter protein to quite significantly. Whereas if we compare this to our inactive compound here, we don't see degradation, we don't see a huge reduction in that luminescence. But what's really important is not just to compare the Nanolux-714 compound to the star compound, it's also to compare minus plus rescue with MG132, or in this case we did also the MLN inhibitor.. And so here you can see that when we treat with our NC4 compound, we get reduction in luminescence. But when we co-treat with these inhibitors, we're able to reduce that luminescence. And that, so that does show you that you can use nanolipoluminescence to monitor degradation, but you do just need to be a little bit mindful that it's important to add these controls. So it's important to add in the star compound, which is the inactive compound to make sure that you're not just inhibiting the luminescence from the enzyme, and also to make sure that you are reliably reading out degradation by adding in these inhibitors. Okay, so next we wanted to compare all of these systems using a physiological situation, using a biological readout. And so something that we were quite familiar with was a protein called MLKL. And MLKL is a protein that activates a form of cell death called necroptosis. And so we tagged MLKL with our three different tags, so the HaloTag, FKBP, or Nanoluc, and we also produced an untagged version of MLKL. And we expressed these different versions of MLKL in MLKL knockout cells. And we can see here that this is— everything's done in inducible system. So when we induce expression, we can see nice expression of the wild type untagged MLKL. We can also see clear expression of the halo-FKBP and also Nanoluc. However, you can tell immediately that the tagged forms do impact the expression of the protein, and that's just something that you need to consider. However, what I'll show you in the next slide is that this reduction in expression doesn't impact the ability of MLKL to drive cell death comparably to the wild type. Okay, so when we induce necroptosis in cells, in necroptosis-competent cells like HT-29s, what we usually do is add this cocktail to the cells. And so this cocktail contains the ligand TNF and then two inhibitors, Smacomedic and ZVAD. And what this cocktail does is it activates the TNF signaling pathway, activates MLKL, and triggers necroptosis. And so just asking the very first question, can each of our tag proteins function? Can they trigger cell death? Yeah, can they, can they, do they not impact the function of MLKL? Because this is really important when you use these systems, you want to make sure that the tag that you're adding to your protein doesn't impact the function. And so here we're looking at live cells. And here we've got, if we just focus here on this induced state in the presence of our cocktail, in our MLKL knockout cells, we have lots of live cells, which we would expect because they're MLKL knockout, they can't trigger cell death. When we reconstitute wild-type MLKL, we can see that we reduce that cell viability. So that's what we would expect. But what was also really nice to see was that when we reconstitute with the halo-FKBP or Nanolook versions of MLKL, we can equally induce that cell killing. And so that shows us that all of our different tagged forms of MLKL are actually functional, which is, like I said, really important. Okay, so we then compared the ability of each of the degraders to reduce this cell death by degrading MLKL. And I'm going to just focus on one first and take you through this, and then I'll show you the results for all of them. Okay, so if we start here, basically what we have is again live cells on this axis. We're looking at MLKL knockouts reconstituted with either untagged MLKL MLKL or FKBP-tagged MLKL, and everything is treated with our cocktail. Now, when we— if we just compare minus plus induction, when we induce untagged MLKL or FKBP-tagged MLKL, we can see in both situations we're able to trigger necroptosis. We're able to reduce cell viability. But when we treat with our FKBP VHL degrader, in the untagged situation, obviously because there's no tag, we don't trigger degradation. We still induce necroptosis and we can see that really clearly here. When we treat our tagged cells though with our FKB-PBHL degrader and we recruit BHL and ubiquitylate this tagged fusion, we degrade MLKL and therefore the cells should survive. And we see that demonstrated really clearly here over a wide concentration range. And this was actually really remarkable when I saw this under the microscope because I wasn't expecting these degraders to be so proficient at degrading this protein and being able to protect against cell death. And so we can really see that we see almost complete protection against cell death right up to the same level as the knockout. And so if we then compare the results across all of the different degraders, so here we've got our VHL degrader. This was the data that I showed you in the previous slide versus our two FKBP cereblon degraders. We can see that again in both situations we are able to really nicely reduce the cell killing, and this is because we're degrading MLKL. Although the cereblon degraders do drop drop off a little bit at these lower concentrations. So they're not functioning as well at lower concentrations as the VHL version. What was interesting though was with the halodegraders. So with the halodegraders, we, we do again protect at these higher concentrations right up to the same levels as the knockout. So they are functioning quite well, but we see this really sharp drop off of in the ability of these degraders to protect against cell death. And I think that's due to that, these degraders not functioning substoichiometrically. What was quite encouraging to see was that our Nanolook cereblon degrader is actually able to reduce MLKL Nanolook to a level that can protect against cell death. So we can actually have a protection against physiological readout here, which was really encouraging to see. And so using these systems, we can do some really nice comparisons. So we can have a look at whether or not the degradation that we see does mimic the degradation, that it does mimic the protection that we see in cell death. And that's exactly what we see here. So here we can use the same E3 ligase complex. So in both situations, we're recruit VHL. But in each situation we have a different tag on the same substrate protein. And with the FKBP system, you can see that we see really beautiful robust degradation over a wide concentration range, and this reflects the protection in cell death that we're able to see. With the halo-VHL degrader though, we're not able to degrade MLKL halo as much as we do within the FKBP system. And then this really clearly recapitulates the lack of protection that we see at these lower concentrations. And so I guess what's really important here when you're choosing one of these systems is to just make sure that you are comparing the different systems with your substrate protein to choose the system that's the most appropriate for your particular protein of interest. Now I just wanted to finish with some data that we have looking at our NC4 compound in vivo, because I think this is really exciting and really encouraging. And so we have a mouse that expresses a construct that contains EGFP and Nanoluke, and we've taken a retroorbital bleed from that mouse and then just done a crude white blood cell prep. And this is just a control control here on the left-hand side to show that this mouse does express what it should express. We can see a nice clear shift in the EGFP peak showing that over control mice we see that our EGFP nanolock is expressed within these mice. But what we can then do is we can then treat with either a vehicle control or our NC4 degrader compound. And here we've used 30 mg/kg. In the vehicle situation, you can you can see everything here is EGFP positive. And when we treat for 4 hours or 8 hours, everything remains EGFP positive. But when we treat with our NC4 degrader, you can see we see this really nice shift back into this EGFP negative population after 4 hours, and it's sustained at 8 hours, showing us that this construct is able to be degraded by our NC4 compound in vivo. And so we have more studies that we're following up now looking at how this protein works, which organs and tissues this— sorry, how this degrader works and which organs and tissues this degrader can work within. So that will be a follow-up study. And so I just wanted to finish with acknowledging everyone who has been involved within the study. I did mention people throughout, but a huge thank you to Christoph Groman, who I said was the lead chemist on this project. Also Joel Walker, who provided all the nanolock inhibitors to allow us to make these degraders. Obviously my lab, and I've got just a few images of my lab here, but the main person on this study was Charlene. Charlene did basically all of the assays that I've shown you today. So yeah, thank you everyone, and I'm really happy to take any questions, or obviously feel free to email me your questions as well. Thank you.