 Today's lecture will be conducted by Dr. Prasanna Venkatraman who is principal investigator and scientist at Actrack, Mumbai. Dr. Prasanna Zab is interested in general understanding of mechanism of cellular homeostasis both in health and disease. Her group is trying to develop a model system to navigate through the various steps involved in proteosomal degradation. She is also interested to understand how the communication between substrates and the proteosome then translates into downstream events like unwinding of the polypeptide chain and its subsequent degradation. In today's lecture, Dr. Prasanna will further explain the surface plasma resonance SPR technology to know why and when it is applied to study the protein-protein interactions. So, let me welcome Dr. Prasanna for today's lecture. Hi, very good morning. This is the title use of SPR in under-revelling domain mode of interactions of proteosomal assembly chaperones. But before directly going into SPR, I wanted to give the philosophy of what this domain mode of interaction is, maybe not all of you are familiar with it. And especially in the context of a non-coprotein, where instead of one-on-one protein interactions, the protein interaction exists as a network which changes from the normal to the cancer cells. You need to understand how would one look at from a network perspective. And even if you looked at the network perspective, at the end of the day, you want to inhibit these protein-protein interactions with small molecules. So how do you get to the details of it from a global perspective and you narrow down it to single protein-protein interactions or destabilize the network of interactions because you have a common principle that drives this interaction. And that is basically the domain mode of interaction concept that can be extrapolated in the context of a protein interaction network. I'll show you one example. I don't have time to go to both. And then in the end, I'll tell you how having done different techniques, I'll ask you a few questions of how SPR is going to help us in taking it to the next step. So these are some of the structures of protein-protein, two proteins interacting. It doesn't matter what they are. And those of you who are very familiar with small molecule protein interactions in terms of a substrate enzyme interactions, you know that they will occupy a very small site, active site is very small, and the molecules that bind to the active site are very small. So you bury only a small surface region, like a 600 Armstrong square region when the small molecule binds with a protein. Whereas if another protein interacts with a protein, then they bury really large surface areas which can go up to 4,500 Armstrong square of buried surface region. That's what contributes to high affinity of interaction. And because you have a large surface area that covers the interface, you can't find a principle that will inhibit this interaction, right? So this seems so complex. And they are enveloped against each other over a very large surface area. For you to pinpoint that there is a specific region that I would like to target with a small molecule becomes difficult. And that is why people say these are unruggable. Now if they are unruggable, are they really so? And can you convert them into druggable targets? Using principles of biophysics and biochemistry. Now long ago, people worked on these protein-routine interaction, dissected them. And this is your reductionist picture of a protein-routine interaction. Even though they envelope over really large surface area, you can see this protrusion that is a knob that is sitting into this hole, right? And if you mutate any of these residues within this region, which is normally one or two, this will fall apart. So despite a large surface area that dictates the interaction, the key interactions are driven by very few residues at the interface, and these are called as hotspots. In many cases, it can be a single amino acid that is dictating this bulk of the binding energy. And if you make the mutation in this residue, you will weaken this interaction that they will not be able to associate with each other. So this is a concept of a hotspot interaction, and this has been increasingly shown to be true in a number of interactions where you are able to identify such a key residue that is responsible for interaction. It seems easy. It is not trivial, but it is being increasingly observed that this can be true for many, many protein interactions. So I said that no longer can we think of single protein-protein interaction. We love to view protein interactions as a network, where there are different hubs with their own islands of interactions. There are unique edges, and these edges are connecting different hubs as well. So this is a kind of interaction that exists in a normal cell. And if you are on a Cogene or a tumor suppressor, because of their levels, mutations, what they can do is that they can rewire the network, because protein interactions are primarily determined by affinity. So if you have overexpression or a mutation that can affect this binding, you will deregulate this network. Some of these interactions may get strengthened, the edges may be stronger, some of these interactions get weaker, and that is how you rewire the network. And that is the difference between a normal and a cancer cell. Now how do you view this? How do you view this in the concept of protein-protein interaction? How do you study it using biophysical techniques such as SPR? So as I said, this is the network, this is made of multiple interactions. So what you try to do is that, can I find a subnetwork within this that I can describe in terms of known structural and biophysical principles? So if you consider that there is a domain and a motive, and these are interacting, and this interaction is a conserved interface, what most of the time happens is that hubs generally borrow this common interface. For example, I have residue something like EEVD, which will come in the context later. There are four residues that are at the interface in this complex. The hub will borrow this interaction interface, use its site to interact with multiple proteins which carry the same motive. Now these proteins need not be homologous, they not carry any sequence similarity, but they happen to carry the short sequence within them, what are called as short linear sequence motive. And if you have this conserved across multiple proteins, these would interact with this hub. So now you have a hub-centric network dictated by a single motive. And say for example, I understood what is the residue that is contributing to this interaction among the EEVD, or maybe just these three residues are really, really important for interaction. I mutate it, I destabilize. So I destabilize not just a single interaction, I destabilize multiple interactions. Got the concept? So this is how you go from principles of domain mode of interaction to the network, and then you come to the reductionist approach of destabilizing the network using the very same principles. That's what we had employed in our study to look at Gankerin. It's a proteasomal chaperon, assembly chaperon as well as a non-coprotein. Then what you could do is that to know which one of them may be functionally relevant, you can couple it to a genomic approach where you use SIR and A to knock out each of these genes, and see, pick up really the functionally relevant ones, or which one actually dictates cell fate, malignancy, metastasis. And then you could go for that particular interaction in the cancer type that you have. So this is the global picture of what our concept is, our philosophy is, and we call this as trying to find out the Achilles' heel in cancer. So now let's go directly to the example. So this is PSM-D-10 or Gankerin, it is a proteasomal assembly chaperon. Don't worry too much about it. It's a huge machinery for it to work. It needs to be assembled from different protein subunits, and these chaperons help in the assembly. And what happens is that Gankerin also turned out to be a non-coprotein which is over-expressed in multiple cancers. And it is, as you can see in this, it is involved in a plethora of cancers and plethora of Halmar cancer properties. What was available to us was a crystal structure of Gankerin and one of the assembly component which is the ATPase of the proteasome. And you see that there is an interaction between Gankerin and S6ATPase through similar to the knob and the protrusion that I showed you earlier. So we recognize this as a hot spot, the crystal structure is available, but we recognize this as a potential hot spot. We looked at residues that are there and we found that EEVD from the S6ATPase protrudes into the Gankerin surface. And if there are proteins which have EEVD on the surface, on a surface they are likely to interact with Gankerin. That is the prediction, a bioinformatics prediction. And we had some 32 proteins for which the crystal structures were available. So we know that these are in the accessible regions of the protein and we asked, tested eight of them as to how many of them would interact. And then except for this one protein, we found all of these that we predicted interact with Gankerin experimentally. And if you mutate the EEVD here, you lose interaction. This is already published. And as you can see that seven of them drive this interaction through EEVD. So this is a subnetwork within the Gankerin EEVD interaction. Now is it functionally relevant? So you have to demonstrate that this physical interaction makes sense. We did a series of experiments where you knock out one of the interacting partner and then you over express the wild type of the mutant. The wild type is able to rescue the phenotype. The mutant is unable to rescue the phenotype. The peptide alone, a short EEVD peptide, is able to inhibit this interaction. And that can be quantitated and you can get an IC50 of about 50 micromolar. So that is important, right? When you're saying this peptide is responsible, mutation is one way of doing it. And throwing in the peptide to inhibit the interaction is another way of establishing that. So this we did. And we know that we can go this route to find the Achilles heel. And you can see that many of these proteins are involved in many cellular activities like angiogenesis, apoptosis, proliferation and metastasis. So we are looking at the functional network of Gankerin through this domain mode of interaction and a conserved hotspot site surface. Why is it important? Because now you have reduced the surface like a small molecule. EEVD is now like a small molecule. You can perturb it by peptide derivatives or small molecules. You are converting a large surface area now into something that will mimic a small molecule protein interaction. So what was non-drugable earlier is now becomes drugable because you are looking at the key interactions that stabilize the complex. And therefore you can perturb them with small molecules and peptide derivatives. So now does the peptide directly interact with the protein? You need to do a series of experiments to establish that the peptide directly interacts. Of course, it interacts because it inhibits the complex, but does it directly interact? We did a series of experiments where there was a thermofloor assay. And there was also a DanoDSF. I think you are familiar with it now that the courses have been conducted. So you can look at these and then we established that there is a direct interaction. We went back to our old technique of ELISA where you can look at peptide protein interaction. Only that the peptide is now labeled with a biotin so that you can pick it up with a streptavidin alkaline phosphatase. You can see this is the wild type peptide interaction. It hasn't gone to the saturation. This is still in process. We probably have to get new peptide preparation. This is hot from the oven. And then you can mutate the same residue that we think is important for interaction from the Gankerin interface because you saw that the S680PS comes with EEVD. Other proteins are coming with EEVD in their sequence. Gankerin uses lysine and lysine 116 and R41 to interact with the S680PS. We believe that it will use the same residues to interact with other proteins as well as with the peptide. Are we right? Here is an interaction with CLIC1, one of the interacting partners which we suspected would interact with EEVD. Here you can see when we make a lysine or an arginine mutation, you abrogate interaction. You do that similarly for the peptide and I show direct binding of the peptide to the protein from the other assays. So we have now narrowed down to the peptide protein interaction using this short motif. Now that we have all these answers, we know the peptide binds. If peptide binds at the intended place, it is able to compete with the protein. So why should we do SPR? That is a question to you. Why do we need SPR? We also have a crystal structure of the protein by the way, it is published but we can solve the structure. We are soaking it with the peptide to determine the structure that is different. So if we can go to that extent, why do I need SPR? Kinetics. Kinetics. Very good. What is kinetics? How is kinetics going to help me? Why should it help me? Why should I bother to study kinetics? I should compare the on and off rate. Very nice. But why should I compare on and off rate? Why am I not satisfied with dissociation constant, equilibrium dissociation constant, KD. What is the relationship between KD and on and off rate? KD is equal to K off by K on. These are rate constants, right? They are equated to the equilibrium rate constant. But equilibrium dissociation constant. But why? So what? So once you have the ratio, what does it mean? The ratio will determine the equilibrium dissociation. What does it mean then? So if I have the same KD, what do you think will happen to the on and off rate? I have so four molecules, I have the same KD for them. Very good, that is the reason to do SPR. But still even if I knew that the on and off rates are different, how does it help me in the next step? How can I use this? How can I apply the SPR derived information which can be given by no other technique, right? Real-time kinetics of interaction, is there any other technique that can truly give you? The label-free SPR is the one that will give you. So why should I do it? What am I going to get out of knowing what is a K on and K off, okay, correct? So how do I engineer it now? Suppose I want to make it more, I want to dissociate it slowly or I want to have a rapid association. So most likely you want to slow the dissociation of a drug, right? And you don't want it too slow because you don't, exactly. So what would I do when I look at the structure? Can it help me in any way? Can structure help me in any way? So independently K on and K off I have and I want to improvise this in producing a drug that will compete with this peptide or it will compete with the protein-protein interaction. I want a drug and I want to get the drug, small molecule inhibitor which will go into a drug with all the properties that are necessary. So a simple on and off rate will tell me or what should I do to, so how do I improvise on the drug? How do I engineer the drug? Now you alter the pharma course of these, right? You look at the interactions with the protein and then you find out what are the molecules, what are the residues that are interacting, whether it is a hydrogen bond, whether it is a salt bridge, whether it is a hydrophobic interaction. And knowing the residues that are interacting, can I now perturb these and can I engineer a better drug? Or if I engineer a better drug, I'm thinking I'm engineering a better drug and I come and look at the SPR, maybe I did not achieve the goal. I actually made it into a poor, it had a poor outcome as compared to what I expected, right? So you couple these structure-guided drug design with the SPR on and off rate, you can actually begin to look at how to get the best drug that is possible, engineer it with the properties that are favorable for a drug protein interaction that will displace the protein. So this is an example where you had the same KD but different on and off and as you guys rightly pointed out, since the ratio is going to change, you can have the difference. You can find out the molecules that are different. So you should look at the BioRAT bulletin and the GE bulletin to understand these things. So we started off trying to now, we had a fairly good idea of how the peptide is interacting with the protein and then in the crystals that we soak with the peptide, we do get the crystals but all the crystals that we have diffracted still now have a vacant binding site. We haven't found the peptide there but the process is ongoing. So here you can see that we tried the, have you, do you know about the SCM-5 chips? CM3 chips, yeah? So these are chips with the dextran molecules to which the protein binds, you can immobilize them, yeah? So we tried the CM5 chips here which is normally used for these kind of studies, protein peptide interaction and we also tried the histidine capture chemistry by having the nickel NTA there on this. In all the cases what we found is that there is a non-specific binding to the reference cell. So once you subtract, you don't get any interaction and then non-specific again in the reference cell even after blocking with the PSA and we have not been able to successfully get the CM5 working. Then we went to CM3, one of the suggestions because what is the difference between a CM5 and the CM3 dextran chips? Exactly. They do not, so they have a lesser surface to bind and therefore you can control your binding, your occupancy better and therefore you are going to avoid this non-specific binding. That's the logic. We did some improvement to this. You can see this is the wild type protein and the peptide interaction and in each case you see that we have done a duplicate run with the peptide and then you have the R41 and K116A mutants because we know that interferes with the interaction. You can see they do interfere with the interactions. So what is your take on these SPR sensor grams? Do you like it? Do you not like it? Is it ideal? Is it interpretable? What are the features that you see? What are the features you should see and what are the features that you see? So do you know how to interpret a sensor gram? What is in the x-axis? Time. What is in the y-axis? RU units. RU units. Okay. And then what is the sensor gram telling us? There is an association phase and the dissociation phase. Great. And what is in between the association and the dissociation phase? So when do you start dissociating? How do you determine for what time that I need to run the analyte? When do I stop and start dissociation? Very critical. Can I say arbitrarily I will run it for 10 seconds and I will start wash it off? No. So what determines it? Saturation. Do we call it a saturation? Saturation by definition is what? Saturation by definition is all the sites are occupied. Will that happen when you have lower concentration than Katie? It won't. So what do we call that as? Equilibrium. Yeah? So you need to achieve equilibrium. Have we achieved equilibrium here? Maybe. Maybe not. Maybe. But we should try and extend the time for equilibration. This you learn by trial and error. You fix some time for the different concentrations and as the time goes you improvise on that. But this is not too bad. And what is happening with the dissociation? Is dissociation critical? I already got this. Why do I worry about the dissociation? I just flush get all the analyte out of it. Should I wait for the dissociation to happen? Why is dissociation important? Very nice. So what does it determine? What does dissociation determine here? Exactly the off rate right? That is so what is so in something important about the off rate? Why should I, if I didn't have off rate in SPR, I will convert this into a normal equilibrium study where I get the KD values. It makes no fun unless I get the K off. And that comes from the dissociation phase. And how should the dissociation be? Are you happy with this dissociation? Criticize. You are not happy with the dissociation. Why is that? Very nice. It is very slow right? So if I, if then if I want to keep, do you think I'll ever achieve complete dissociation in this case? Why not? So I'll keep it for three days, no? Okay. Yeah, but then imagine over the days it will come like this, okay? And definitely come. But that's not the point. So how do I now, so what should I do now? I won't get a good off rate from this. So what should I do? And without off rate my kinetic estimations are not going to be right. Why is it, why is it K off so important? Why can't I look at K a which seems very good and if I achieve the equilibrium I should get the K a. What is so unique about K off? K on is diffusion control. If it is, it is diffusion control and there is also it is controlled by another parameter. What is the difference between K on and K off? Excellent. The first one is concentration dependent K off is independent of the concentration because now I'm looking at a complex A B. It is there, it is not determined by concentration. So the inherent property of the binding comes from this independent variable that is K off, independent constant which is K off and if I can determine that all my kinetic estimates are going to be more or less accurate, right? And imagine I have the right K off. I have Kd which is a equilibrium constant I determined by n number of methods, right? What can I get? I know K off, I know the equilibrium dissociation, I can get K on, right? I can get a rate constant for association independent of the instrument and can I go back and verify whether I got the same K on or not? Yes, right? And do you think the Kd that I determined by SPR that is the equilibrium dissociation constant is going to be very different from the ones that I determined elsewhere? Say for example, analyzer it shouldn't, right? And how do I get the equilibrium constant from the SPR data? Steady state measurement. How do I do a steady state measurement? Different concentration and what do I plot? How does it look like? Very nice. So you guys are experts here, huh? So then there do I need equilibrium or do I need saturation? Do I need saturation? Very nice. So equilibrium whether it is a kinetic measurement or equilibrium measurement, you need equilibrium and you need saturation which is determined by what in SPR? What is a unit? Saturation. The first and foremost thing that you do before you start the SPR, there you go, R max, right? So why should we determine on R max? That is a unique property of R max that allows you to interpret your SPR data faithfully. So suppose I said I want 400 R u as R max, right? I want 400 R u as the R max value. I know the molecular weight of the ligand, I know the molecular weight of the analyte and what is another parameter that is in the equation? Okay. Now tell me why should I go back and calculate what the R max is? Exactly. If I got something weird than what I expected, there are two things that happen, right? One, there is some weird non-specific binding going on or two, the estimated stoichiometry is not correct. The second you come, you interpret later. The first thing that you should worry when you get a more R max than what you expected because you know what you immobilized, right? And if it is one is one, no matter what you do, those mathematical equations have to be satisfied. So whether you understand the kinetics, whether you fit it into a one is to one binding or more complex binding isotherms, the first level of checkpoints have to be done, okay? So when you are doing SPR, you be careful about all these things, your calculations, you first do a predetermined calculations, you go back and look at your sensorgrams and see whether you have got all these values correct, you estimate the equilibrium dissociation constant by, even if you cannot get the kinetics, estimate the equilibrium dissociation, find that your binding interactions are very similar to what you have obtained by other studies, it is better to do SPR after doing some kind of measurements that tell you the equilibrium constant. Is there any other technique that allows you to look at equilibrium constant? Any biophysical technique? So you do ITC, you will get the stoichiometry there, yeah? It is very good for stoichiometric measurements. And then you get the KD there as well, and then you combine it with KON and KOFF, you get a delta S, delta G, delta H, the stoichiometry, KD, on and off, you are done, right? So that is a way to probe the protein-protein interactions in depth. And if you have a select example that you started with the hypothesis and it has ended up giving you the expected results, then you try and do all these parameters and see whether either by screenings, molecules to find the inhibitor, or you start with the structure-guided design by docking, or you start with already known peptide that it binds and then you begin modifying the peptide using chemist help and then designing the molecules so that you can get better and better inhibitor. This is the reason why we are doing SPR, and so far it seems you guys rightly pointed out what are the problems, but what is the, what is the positive aspect of this data? What is the positive aspect of the data? Very good, you see some binding. And then what happens with the mutants? The mutants behave like they are expected to behave, they do not bind very well, right? Now you see here, it is not so bad, we still fitted these binding kinetics, and why do I say it is not so bad, chi-square, and why is chi-square important? What is it that you look for when you look at the chi-square, expected and observed and what should it look like? Residuals, right? So it is the deviation from the residual that gives you the chi-square value, this is pretty tight, so you can believe all these, right? So you look at the Gamkaren and we get Kd is around 12 micromolar, is it right? This is something that we expect around 12 to 50 micromolar is what we expect, and the Ka and Kd and you look at the mutants, okay? It seems to suggest that the mutants have better on rate, right? And very similar or slightly faster of, slower of rate than Gamkaren, the wild type, and you look at all these Kd measurements and look at the Rmax, despite poor binding, these seem to be behaving pretty well in terms of the kinetic constant. We do not understand this, I am showing you what the graph values are, but we do not understand this, yeah? We are trying to interpret this as to what may be the problem, and definitely we are not happy because if you look at, it is not very clear to you from here, it is, we are achieving up to 15 RUs on the binding and which is not very good, and we are not able to achieve Rmax because the chip whatever we have used, whether it is histag NTA chips or we have tried direct immobilization, we are having problems, and non-specific binding to the flow cell is seen, and some of the times, many times some of the flow cells do not work, and what we have come to conclude by looking at many, many interactions of this kind, changing the chemistry, you definitely need to change the chemistry, either it is a thiol or it is a biotin streptavidine or direct immobilization or a capture, you need to do multiple things to confirm that these are behaving the way you should. So what we have understood now is that the IFC, that the fluidic cell itself is a problem, and that creates a replacement every 6 months, costs about 6 lakhs, but we are also trying alternative chemistry as well on this, and trying to reverse. So we took the streptavidine chip and tried to bind the biotin peptide and then come up with the protein. What is the advantage that I have in that case? The annihilators bound now, that is a biotin-related peptide to the streptavidine chip, and I am passing the ligand instead of having the ligand immobilized and I pass the annihilate, which is a short peptide, which is better, bigger molecule on the chip. How is the, how is the, are you generated? Why is it reverse? You finish your thought, exactly. So when you design experiments that you want to, so here is where the sensitivity, question of sensitivity comes in, and if you are looking at very small molecules, it is better to have the ligand, and then you bind the annihilate, but G Bose of T 200 to be able to capable of detecting small molecules, but once probably we get these things going straight, and especially the fluidic cell, we should be able to see these things happening, but yes that is a main point. And these are, before you start an SPR experiment, just keep these things in mind, you calculate, you find the expected values, go back, interpret your sensor gram, do not try to fit the data before understanding the sensor gram, you can fit it to any equation that you want, you will get a value, you will get a chi square, never ever do that. Most often, it is the simple one is one binding that is happening, just because your sensor, you did not do the experiment right, a second order fit or a third order fit might give you a very, very good fit, because now you are increasing the parameters to fit, so the, your sensor grams will look nicely fitted, but that may be a wrong interpretation. Most often it is the first one is to binding that you should try, if there are, after doing many, many experiments, you think that things are not explained by one is to one binding, then you begin to use other equations of this. So, these are some things that definitely you have to look at, you have to look at equilibrium, try to look at, and how do I, how do I enhance dissociation, can I enhance dissociation, I can change the pH, can I change anything else, flow rate very nice, anything else, why should I change pH, what happens if I change pH, why should the interaction get weak, I change the protonations right. So, what is the other way of doing it, dissociating to proteins, salt, ok, then I am speaking to experts, that was my last slide, thank you very much. So, I hope now you are convinced that SPR based systems is a very powerful platform to generate high quality data for biomolecular interactions, especially to obtain the dissociation constant the KD values on rate of rate, the kinetic data which could provide you very quantitative valuable information, thank you.