Hello, everyone. I am Anne Bertolotti. I’m a Program Leader at the MRC Laboratory of Molecular Biology in Cambridge in the UK. And this talk is the third of a set of talks I’ve given on protein phosphatases. And in this talk, I will share with you some assays that we’ve developed that can enable the study of protein phosphatases, as well as enable the discovery of selective phosphatase inhibitors. But before I go there, I’d like to give you a bit of background to tell you how we became interested in phosphatases in the first place. So, my lab has been interested in misfolded proteins for many years, because these proteins, when they accumulate in the form of insoluble aggregates, represent a huge problem for cells and organisms. And this problem is the molecular basis of a broad range of human diseases, including the devastating neurodegenerative diseases Alzheimer’s, Parkinson’s, amyotrophic lateral sclerosis or Lou Gehrig’s disease, Huntington’s disease, and so on. So, these are devastating diseases that affect an increasing number of individuals in our aging societies. They are very different clinical disorders, but at the origin what we have is a common problem, which begins when cells fail to handle these misfolded proteins that are normally produced at all times. It turns out that we have mechanisms built in — self-defense mechanisms — to neutralize these aggregation-prone proteins. And these mechanisms work really, really well for many decades of our lives. And so one idea we’ve developed in my lab is to try and boost these natural defense mechanisms in order to try to find treatments that could perhaps benefit these diverse neurodegenerative diseases in the long run. So, today I’ll show you one approach that we’ve developed, that enables cells to increase their self-defense capacity against misfolded proteins. So, a natural defense mechanism against many forms of stresses consists of phosphorylating a protein which is called eIF2-alpha. This is a translation initiation factor. And when eIF2-alpha is phosphorylated, this results in a decrease in the rate of protein synthesis, because eIF2-alpha is a translation initiation factor, and when it’s phosphorylated it no longer functions properly in initiating protein synthesis. So, the benefit of this, the decrease in protein synthesis, is to increase the protein quality control capacity of cells. Because under normal circumstances, the vast majority of cellular resources are engaged in mass protein production. And so if you decrease protein synthesis, then, as a result, it spares the existing resources to handle the damage. And here the damage is about misfolded proteins. So, this is one way this signaling event is protective. Now, the activity of the kinases that phosphorylate eIF2-alpha is antagonized by phosphatases. And we mammals have evolved with two selective eIF2-alpha phosphatases. These enzymes belong to the family of PP1 phosphatases. They have… they are peculiar enzymes in the sense that they are split enzymes. They are composed of two components that need to be brought together for this enzyme to function. So, the eIF2-alpha phosphatases have a common catalytic subunit, PP1, that they actually share with about 200 other phosphatases in the cell. But what dictates their selectivity is the regulatory subunit, the non-catalytic subunit. So, R15A is one of them; the other is R15B. These are related proteins, functionally, but they are different. They are encoded by two different genes. And it turns out that R15A is stress-inducible. It’s selectively translated when eIF2-alpha is phosphorylated, and in this way it comes up as a response to stress, to enable the rapid dephosphorylation of eIF2-alpha. And R15B, on the other hand, does the same job. But it’s constitutively expressed in all cells and all tissues, as far as we can tell, to maintain low levels of eIF2-alpha phosphorylation. So, we’ve discovered, and I’ve discussed this in the second talk of this series… we’ve discovered small molecule inhibitors of the stress-inducible eIF2-alpha phosphatases… sorry, the stress-inducible eIF2-alpha phosphatase. So, these inhibitors, guanabenz and Sephin, selectively bind to R15A. And as a result, they inhibit this phosphatase to prolong the duration of this naturally occurring self-defense mechanism against misfolded proteins. And to illustrate how this works, I’m gonna share with you a small video that actually summarizes ten years of working in about one minute. Under normal circumstances, our cells are busy making thousands of proteins required to execute all the cellular functions. And this process requires chaperones — these little hands in orange — that make sure proteins fold properly, that they don’t aggregate. We make errors. We make mistakes. We sometimes produce bad proteins. But we are able to target them to degradation. Here I’m showing the proteasome, which is… which is not only important in degrading proteins but also vital to recycling amino acids. And this all goes well for years, for decades, up to a point. As we age, we accumulate… or in diseases, we accumulate these bad proteins. And one way to defend ourselves against this is to phosphorylate eIF2-alpha, to slow down protein synthesis. And this in turn enables the chaperones that are normally busy assisting the folding of newly synthesized proteins to become available to handle the misfolded ones. And this is one way this pathway is protective. So, in this work, we realized that we can actually selectively inhibit a phosphatase by targeting the regulatory subunit. And this was exciting news, because phosphatases were thought to be undruggable, largely because they share a catalytic subunit. So, if you inhibit the catalytic subunit, in the way people usually classically think about enzyme inhibitors, here… if we inhibit PP1, it’s not one enzyme that we inhibit but hundreds. And that’s really not useful, because this actually is toxic and kills cells. So, having discovered that we can inhibit one phosphatase, we realized also that perhaps we can inhibit many more phosphatases using the same principle — targeting the regulatory subunit. But for that, we really needed to develop methods and assays to do so, because the first phosphatase inhibitors… selective phosphatase inhibitors we had discovered were discovered through a phenotypic screen. So, to move forward, we thought we absolutely need to reconstitute the function and the selectivity of these phosphatases with recombinant proteins. And this was challenging for many reasons. But I’ll only summarize this piece of work, which took many years to develop in my lab. But before I get there, let me summarize for you where we were when we started this work. So, for many years, as I explained in the first talk of this series, people had been working with the highly purified phosphatase, which is the enzyme removed from its cellular regulators, these non-catalytic subunits that normally confer selectivity to the enzyme. And as a result, the phosphatase, highly purified in this way, was found to be non-selective, because it can dephosphorylate pretty much any substrate you will give to it, particularly in a test tube, when it’s used at high concentration. And the non-catalytic subunits of phosphatases were known, but they were largely described as inhibitors of PP1. Because when bound to PP1, they inhibit the dephosphorylation, in vitro, in the test tube, of this classical… the historical substrate that was… that led Fischer and Krebs to discover the role of protein phosphorylation in controlling the activity of proteins. So, non-catalytic subunits of PP1 were largely described as inhibitors in the literature. And our favorite at the time did not escape this. R15A was reported by the lab of Shirish Shenolikar in 2001 to inhibit the dephosphorylation of phosphorylase a in a concentration-dependent manner. And intriguingly, Shirish Shenolikar’s group noted that R15A, when added to a dephosphorylation reaction with eIF2-alpha phosphorylated as a substrate, did nothing at all. So, we started recapitulating what was known. And so we started by reproducing what was in the literature. So, Marta Carrara, a talented biochemist, joined the lab, purified PP1, purified R15A, R15B, prepared some phosphorylase a, phosphorylated it. And she found, which was no surprise, right?, that PP1 could dephosphorylate the phosphorylase a. But when she added R15A or R15B, this completely blocked the dephosphorylation. So, that was fine. But remember, we are interested in the dephosphorylation of eIF2-alpha, not so much in the dephosphorylation of phosphorylase a. And by the way, not surprisingly, our inhibitors had no effect whatsoever in the assay, here, using phosphorylase a as a substrate. So, we really needed to develop assays where we can study dephosphorylation of eIF2-alpha by its cognate holophosphatases, so PP1-R15A and PP1-R15B. So, the first thing Marta did was to develop the assay along the line of the classic assays that had been used in the field, using phosphorylase a as a substrate. But here she replaced it with eIF2-alpha. And in the same condition, what she observed was, indeed, PP1 could dephosphorylate eIF2-alpha. But when she added the non-catalytic subunits, R15A and R15B, she saw nothing at all. Nothing happened. And here, I can’t list all the possible reasons why an assay like this might have failed. The proteins we produced may not be active, and we know that these proteins are thought to be natively unstructured, so maybe they are not functional. Or there could be a million other reasons… the assay is not performed in the right conditions… where to begin we didn’t quite know. So, we went back to our computers and read the literature, again, thoroughly. And here we realized that actually the concentration of PP1 in the cell is estimated to be 0.2 micromolar. So, knowing that PP1 is not present in isolation in the cell but is a subunit of more than 200 protein complexes, then we did a simple mathematical calculation and realized that any given holoenzyme had to be active in a cell at a nanomolar concentration. So, then we did something very trivial. We titrated PP1 and looked at its ability to dephosphorylate eIF2-alpha. And sure enough, the less you add, the less activity. And we realized that at 10 nanomolar, with this preparation of PP1… from 10… 30 nanomolar and below, there was no dephosphorylation of eIF2-alpha. I must stress that for those of you who want to follow these assays, it is really important to titrate every preparation of PP1. Consistently with this procedure, we found that below 30 nanomolar PP1 is no longer active. But since then, we’ve improved the purification of PP1 further, and we now work with a… in our assays with 1 nanomolar of PP1. So, it’s important to do this titration with any preparation of PP1. So, here we used the same assay as I’ve shown you before, but we used much less PP1. And with 1 micromolar of substrate and 10 nanomolar of PP1, we see PP1 does not dephosphorylate eIF2-alpha. But here, when we add the non-catalytic subunit… bingo. We found that this converted this otherwise inactive PP1 in this assay condition into a very active enzyme. And we found the very same with R15B. And this was very, very exciting for us, because this provided us with an assay, biochemically defined, with a small number of components — the catalytic subunit, the non-catalytic subunit, and the substrate — and this assay recapitulated not only the function… so, our eIF2-alpha holophosphatases were able to dephosphorylate their cognate substrate, eIF2-alpha, but importantly also this assay recapitulated the selectivity of holophosphatases. Because the eIF2-alpha phosphatases, R15A-PP1 and R15B-PP1, were unable to dephosphorylate the phosphorylase a. That was really exciting, because for the first time we had an assay, a simple biochemical assay, that enabled us to study the function at the molecular level of these non-catalytic subunits, that we had known for more than 20 years, that were important to help the dephosphorylation of eIF2-alpha. We began, again, by looking at what was known. And it was known from many, many groups that I can’t all cite here that the carboxy-terminal region of R15A and R15B contains a binding site to PP1, and is important to recruit PP1. But that is not sufficient to reconstitute a functional holophosphatase. To make a functional holophosphatase, one needs… in addition to the carboxy-terminal region of the R15 proteins that bind PP1, we need a large fraction of… a large fragment of the amino-terminal region of R15. And I’ll tell you why this region is important. We found, or rather, Marta found, that the amino-terminal region of R15B, as well as R15A, serves as a really high affinity binding site for the substrate eIF2-alpha. And this is where selectivity is encoded. If we take a non-related non-catalytic subunit, we find no binding at al l to the substrate eIF2-alpha. So, from this, we’ve proposed that the non-catalytic subunits of phosphatases, particularly these, are modular proteins. They contain a binding site to PP1, which is important to recruit the catalytic subunit, but also they contain a binding site for the substrate. And this is important to… for the holophosphatase to function. And this led me to propose that actually phosphatases are split enzymes. They’re composed of two modules that absolutely need to be together in the cell for the phosphatase to work. The catalytic subunit is there and is important to cleave the phosphate off proteins. But this action only occurs if a non-catalytic subunit has brought the substrate for the catalytic subunit to dephosphorylate this substrate. And so, from that I think… knowing that we have 200, or perhaps even more, holophosphatases in the cell, this brings the notion that we have actually a split protein phosphatase system. And I’d like to think about it as a sort of key-lock system, where only the right key can open the right lock, and only the right, the cognate, regulatory… formerly called regulatory subunit that I’d like now to call substrate receptor… only the cognate substrate receptor brings the cognate substrate to PP1 to enable productive dephosphorylation. And this simple model actually explains a very old conundrum in phosphatase biology, which is how these non-catalytic subunits, remember, initially were described as inhibitors of phosphorylase a dephosphorylation. Now obviously, we didn’t evolve 200 enzymes… 200 proteins, sorry, to block the dephosphorylation of phosphorylase a. These non-catalytic subunits of phosphatases on one hand serve as a high affinity receptor for the substrate, but in this way also prevent PP1 from non-selective dephosphorylation. So, a dual function in this way. With the knowledge we had acquired, we could then go back to our initial question, which was trying to understand how these small molecule inhibitors, guanabenz and Sephin, that we had discovered selectively inhibited R15A. And I’m gonna summarize our findings for you. We found that these inhibitors bind specifically to a region in the amino-terminal part of R15A. And following binding, they induce a conformational change, which then perturbs the substrate-recruiting activity of R15A. And this is how these small molecules prevent dephosphorylation of eIF2-alpha. So, this was a summary of many, many years of work. So, here… starting from our interest in protein misfolding, we discovered a protein that we can selectively inhibit, R15A. It’s a non-catalytic subunit of a phosphatase. And I’ve shown you in talk number two the benefit of this selective inhibition of R15A to improve a rare disease caused by an accumulation of misfolded proteins in the endoplasmic reticulum. And the disease is called Charcot-Marie-Tooth 1B. So, this is exciting. But if you remember from my first slide, I mentioned a broad range of neurodegenerative diseases — Alzheimer’s, Parkinson’s, ALS, Huntington’s, and so on. And I’ve often been asked, after the discoveries of guanabenz and Sephin, do you think you can treat all these diseases with this inhibitor? Well, disappointingly, the answer to that question came very rapidly. And the answer is no. Because simply, in these diseases we don’t think that R15A is expressed. And this is because R15A is inducible by a subset of condition… of conditions, particularly conditions causing accumulation of misfolded proteins in the endoplasmic reticulum. This is not the case in these other diseases, and sort of stating the obvious, if R15A is not expressed then R15A inhibitors are not going to be of much use in these contexts. So, that led us to become interested in the functionally related protein R15B. And the idea was simple. R15B is expressed everywhere, so perhaps we could achieve the same benefit that I’ve explained at the beginning of this talk — reduction of protein synthesis to increase protein quality control capacity — by targeting R15B, alleviating the limitation of R15A, which is the fact that R15A is only expressed in a small subset of diseases. So, that was our next idea. And the issue was, well, we still didn’t have any assay to enable rational discovery of selective phosphatase inhibitors. So, Anna Sigurdardottir, a talented postdoc, joined my lab to develop a screening method with the recombinant and functional holoenzymes that we had characterized before. So, what Anna did was she reconstituted these functional enzymes on a chip that we then used for surface plasmon resonance measurements. So, this is a method that enables the measurement of the binding of molecules to each other. So, of course, the first thing we wanted to know is whether this method was good enough to detect the binding of the known inhibitors of R15A. And this actually worked remarkably well. Anna found that guanabenz binds selectively to R15A-PP1. And it does so with a submicromolar affinity, which is compatible with the potency of the compound we had observed in cells. And guanabenz, as we had seen through many other cell-based assays before, doesn’t affect the related phosphatase R15B-PP1. So, that gave us confidence that the assay was relevant, and so we went on and characterized Sephin in the assay. And interestingly, we also found that Sephin is selectively binding to R15A-PP1 with an affinity that is 30 times higher than for R15B-PP1. So, all this gave us much confidence that the assay was relevant, and we could use it to discover new inhibitors. So, there was something interesting here, which came from the observation that by making Sephin, this derivative of guanabenz, we actually created a low-affinity binding site for the compound to R15B-PP1. And that led us to suspect that perhaps in the same chemical space as guanabenz and Sephin we could perhaps identify some small molecule inhibitors that could selectively inhibit R15B. So, following this, I then designed a small library of molecules, and Anna used those in a screen for binders to are R15B-PP1, and she counterscreened on the related phosphatase, R15A-PP1, and we also counterscreened on PP1, because we don’t want at all to inhibit PP1. Remember, that’s not selective at all. So, she identified in this way a number of tight binders to R15B-PP1. And I’m gonna discuss during the rest of this talk a molecule that we called Raphin. Raphin stands for rationally discovered phosphatase inhibitor, because this is the first time we’ve been able to rationally identify a selective inhibitor of a phosphatase, here targeting R15B. So, Raphin binds selectively to R15B. And following binding, it actually alters the conformation of R15B. And this prevents substrate recruitment and inhibits the enzyme in the in vitro assay that I’ve described earlier. Interestingly, when this happens in cells… so, Raphin can enter cells, and inhibits R15B in cells as well… but following the binding of Raphin to R15B in cells, the conformation of R15B is altered, as we had seen in vitro, but this insult is recognized by a protein quality control system in such a way that R15B bound to Raphin in this altered conformation is then targeted to the degradation machinery by a process that involves a chaperone, p97, which is important in recognizing this abnormal conformation of R15B. So, in this way, Raphin inhibits R15B and targets it to degradation. The consequence of this is to increase the phosphorylation of eIF2-alpha in cells. And this leads to attenuation of protein synthesis. That was anticipated. But interestingly, this attenuation of protein synthesis is actually transient. And this is because when eIF2-alpha is phosphorylated this also leads to a selective translation of R15A, which then comes up to dephosphorylate eIF2-alpha. So, inhibition of R15B leads to a transient phosphorylation of eIF2-alpha. And again, this is because the compound is selective. Naturally, when we add an R15A inhibitor together with the R15B inhibitor, we convert this transient attenuation of protein synthesis into a persistent one, because we inhibit this negative feedback loop involving R15A. So, that was quite exciting. We thought, here we have a wonderful way to presumably safely increase protein quality control capacity. Safely because the attenuation of protein synthesis is transient. And we want to use that to see whether we can ameliorate diverse diseases associated with the misfolding of proteins, and particularly diseases where R15A is not expressed. So, first let me tell you that Raphin has suitable properties for in vivo studies. It’s orally available. It crosses the blood-brain barrier. And we’ve performed tolerability studies and found that even at high doses the compound is actually safe. And we think the safety comes with the fact that we have this negative feedback loop involving R15A which prevents phosphorylation from being excessive… phosphorylation of eIF2-alpha from being excessive. So, in this way, we… the system is safeguarded against an excessive phosphorylation of eIF2-alpha. So, in pilot studies, we’ve started treating models… mouse models of Huntington’s disease, because we didn’t find R15A in these mice, and therefore R15A is not a target for this. So, we treated mice expressing mutant huntingtin, and these mice, as is the case in the human disease, fail to gain weight over time, so weight loss is a prominent characteristic of Huntington’s disease. And you can see that this is neutralized with treatment with Raphin. And interestingly, this comes about because Raphin decreases the accumulation of the huntingtin inclusions that cause the disease. And we can see that by many different approaches. Biochemically, we look at these aggregates, running gels as shown here. But we can also see the same benefit by looking at inclusions by immunohistostaining, and quantify this in a double-blind way. You can see that the effect of the compound is extremely robust, considering the fact that this model is a very aggressive model of Huntington’s disease. So, this is very exciting. And all in all, this work really highlights the beauty, the power, and the benefit of inhibiting a phosphatase by targeting the regulatory subunit. Here, I’ve shown you two examples. We’ve identified two targets that are functionally related, R15A and R15B. We can inhibit, selectively, one or the other to increase protein quality control capacity in cells. And this translates, at least in mice, into therapeutic benefit. R15A inhibition, we think, is useful for diseases where the misfolding pathology is in the endoplasmic reticulum, whilst R15B inhibition will be useful for diseases where the accumulation of misfolded proteins occurs in the cytosol or in the nucleus. And that’s the case for the common neurodegenerative diseases I’ve shown you, for example in Huntington’s, but we are thinking that this might be applicable to other protein misfolding diseases as well. So, in the background of this, we’ve created a platform to enable the rational discovery of selective phosphatase inhibitors. The platform that I’ve introduced to you, starting with the biophysical screen that Anna has developed, can be used with any phosphatase, with a positive screen to identify selective binders to a given phosphatase and a counterscreen with an unrelated phosphatase, as well as an isolated catalytic subunit. And then we follow up with a cell-based assay where we measure target engagement in cells, and we also filter off-target compounds by testing the effect of the compound in cells knocked out for the target. And we can also look at the mechanism of action of these inhibitors using the in vitro assay that I’ve shown you. So, I think it’s fair to say that phosphatases are no longer undruggable. I’ve reported to you how we discovered that we can selectively inhibit a phosphatase. And this discovery came up through a phenotypic screen. But following that, we were motivated to study phosphatase function, and we developed a series of assays that then enabled us not only to understand the function of phosphatases and how their selectivity is encoded but also to design assays that we can use, now, to selectively inhibit diverse phosphatases. And that’s exciting for many reasons. These phosphatase inhibitors will become tools to study phosphatase biology, and that’s very much needed, but there’s also something very interesting that I’d like to bring to your attention. I think that phosphatase inhibition may represent a really attractive new therapeutic modality for the following reasons. As you may know, many signaling pathways operate on the same principle, with a kinase in the activation phase of a signaling event, followed by a phosphatase involved in terminating this signaling event. And I’ve shown you in two examples that by inhibiting a phosphatase we prolong the duration of a naturally occurring signaling event. And in this way, I think we might be able to deliver new and safer medicines that will enhance our self-defense mechanisms. So, that’s an attractive possibility for the… for the future. But certainly, we’ve brought here assays that enable us not only to study phosphatases but also to come up with selective inhibitors. And I will close here on this speculative note, and contemplating the future of phosphatase research, but also acknowledging all the fantastic people that have joined my lab and embarked on really challenging pieces of work, and joined me in doing what we do in my lab, which is studying the unknown. And that’s always been great fun. Thank you very much.