Roger Y. Tsien (2015) - Molecules Against Cancer or for Long-Term Memory Storage

So I'd like to tell you this morning about the two different stories. One is our recent molecules that we have developed against cancer and particularly they are called activatable cell penetrating peptides, ACPP's. The primary use, the first use is for florescence imaging of matrix metalloproteinase activities or actually they can be used for targeting chemotherapy and radiotherapy. But I don't have time to discuss that sort of work. The people in my lab and our outside collaborators are listed here, but I have a conflict of interest disclosure. A new company, Avelas Biosciences, has been started to commercialize these synthetic probes, and I and some of the members of my lab, our advisors, collaborators, have some stock in the company. So you have to bear that conflict of interest in mind. Also, it opens the possibility that I will discuss at the last slide of this half about the long term prospects. Then the other story is molecules that nature uses to store memories in our brain and this is the neurobiology to perineuronal net. The people involved are Varda Lev-Ram, Sakina Palida with some help from mass spectrometer colleagues. This is a very different story. Bunch of more basic research instead of applied research. They're very seemingly very different, but at the end I will try to show you that there is a, surprising similarity, certain surprising similarities as well as the contrast. First, I have to explain why I got the Nobel Prize and why I don't work on these subjects any more. Some of the fluorescent proteins that were available in 2004, in this famous one slide summary, encode pretty colours across the rainbow. Each of these tubes contain the fluorescent protein and each fluorescent protein is encoded by DNA. Non-toxic, synthesizable by any aerobic organism. And by the way if you’re a chemist, I, you know, got a chemistry prize, you have to mention that they are synthesized with 230 condensations of cyclization and whatever oxidation, in 100% yield, in aqueous medium, and only a little bit of hydrogen peroxide as a byproduct. That just shows how good a chemist nature is compared to us as ordinary chemists. But, fluorescent proteins have advantages and limitations. The good side is that fluorescent proteins have had a major impact on basic biomedical science. Most important proteins and key biochemical steps. Eventually we were able to image them in live cells or model organisms of yeast, worms, flies, mice, etc. Other people have gone on to, many super-resolution Nobel Prizes, derived some benefit from our fluorescent proteins, etc. But, I felt that they were unfortunately weak, fluorescent proteins were weak because they could not be applied very well to people. Humans are too opaque, thick and opaque for a whole body florescence. We're not the size of a mouse and the light doesn't penetrate more than a millimeter or two throughout our tissue. The ugly aspect is that the fluorescent proteins have to be introduced by genetic engineering. That's the whole point. They were the first pigments that could be introduced by genetic engineering and only on people, gene therapy is not sufficiently advanced. You can do gene therapy on a mouse, but for ethical reasons, you don't want to do gene therapy on people. At least not in it's current form, not very commonly. In the introduction of fluorescent proteins generally would not help the patient and that is what would be necessary. So, for clinical applications we need synthetic molecules to do the equivalent of GFP but by other means, more traditional means. That would light up and treat disease tissue and especially cancer. I focused on cancer because my father died of pancreatic cancer. Therefore, one of the strong motivations for me, to work on cancer. There were other motivations, of course. So, why study proteases? As it turned out we studied the proteases and the reason was that most cancer deaths occur due to metastasis, that is the spread of the cancer. Instead of just a primary tumour, which is some symbolized here, you are killed by the cancer when the cancer is able to spread uncontrollably to secondary distant sites. Most of the work on basic biology of cancer is done on primary tumours but the metastasis, which actually kills you, is when the tissue, the tumor goes beyond the primary tumor. Cells have to break through a lot of healthy tissue, basement membranes, extracellular matrix, and so on. And cut its way, like with machetes, waving machetes to carve its way through the jungle. Then reaches the blood stream and crawls out and then crawl through the blood stream or lymphatic organ stream and crawls out at the other end. All of this requires over-expression activation of enzymes that cut proteins because you have to cut through this matrix. And those are proteases. Two of the most famous, but not the only ones, obviously, but two of the most famous are matrix metalloproteinase-2 and MMP-9. They're only just the things for which we have the best reagents. In fact, this is stolen from a commercial website, which advertises reagents in vitro for checking on MMP-2 and 9. But we thought that if we could see them in vivo and eventually in a patient, we would have better diagnosis and treatment options for them. The biochemical mechanism by which we were going to sense, we planned to sense the MMP-2 and 9 activities. It starts with this basic biochemistry fact, which is not due to me, that if you have a payload a and it is stuck onto a polycation peptide, such as roughly eight or 10 positive charges in a row. when it put next to cells and here's a cell, first sticks to the outside of the cell by electrostatic interaction, and then gets endocytosed or taken up inside the host, target cell. A little bit can even get into the nucleus. This is a well-known phenomenon that was already discovered but it is rather non-selective, this uptake is. We made it a little bit more selective by the following procedure. We attached two more parts to this polycationic peptide, by the way, that's called a cell penetrating peptide. We attached a polyanion which had equal number of negative charges. We attached by means of a cleavable leash or linker. By putting on a matching number of negative charges to the positive charges that were already there, the cell uptake fortunately was largely blocked because the pluses and minuses neutralized each other internally, rather than trying to bind to the negative surface charge of the cell. This is called an activatable cell penetrating peptide. In this form, so far, it doesn't do anything but it just waits. It's got this cleavable portion and it can be cut by the proteases we care about. And when the proteases, such as MMP-2 and MMP-9 cut the sequence that is built into this protease, this cleavage sequence, then they cut, snip, they snip it apart and the halves drift apart and you get a couple of copies of polyanion which you then throw away. Result is also the freeing up of the original polycation payload, the CPP, which then does its normal thing. Wherever the protease activity resides it's like getting a little microinjection, not a physical injection but a microscopic injection of the polycation payload. So, I'll cut through all the, I'll just try to show you how it actually works, in principle. So can you see the tumor? This is in the middle of an operation. We're carving up the tissue. It happens to be a mouse. Obviously, under anesthesia. But, I challenge you to tell what is the tumor and what is not. The surgeons have been using, having difficulty for years and years, for thousands of years, telling cancerous tissue from healthy tissue. Now we turn on the fluorescent light and we use fluorescence imaging and a bit of an image processing. And now we can clearly see, not only even where the tumor is but how much tumor but how much tumor is at the edges, it's a mixture of tumour and normal and how active is the MMP enzymes by the colour coding built in. So now you can see by the help of the fluorescence imaging all the... obviously makes it much easier to cut the tumour accurately. By comparing fluorescence guided surgery with the lights on versus white light reflectants, which is your ordinary naked colour vision, using just white light, you can see in these so called Kaplan-Meier plots. These are the plots of time versus the operation time since the operation, as for example here, twenty five weeks after the operation. As you successfully go on the tumour recurrence, because of not having accurate enough surgery initially gives you this white light reflectants. This is the performance curve of the traditional method, but FGS does much better. In other words, in this case a great many more mice were still cured. Not all of them but many more. The same is true in a pancreatic model or in a parotid model of cancer. We can go on and on. Even, just in the pancreatic model, white light reflectants leaves a good deal of tumour, whereas the fluorescent guided surgery the tumour can be verified to have been left behind is almost eliminated. This is a complicated story but the same mechanism of cleavage-activated uptake, I have shown you, can deliver a contrast agent for optical imaging. Not shown but also there, is magnetic resonance imaging and actually delivery of drugs and radiation synthesizers, but there has not been time to show you. The linkers can be cleaved by proteases. It offers enzymatic application because one molecule of protease can cause many molecules to be taken up. Protease activities are good markers, not only for metastatic ability, that's the MMP-2 and MMP-9 that I talked about, but also for inflammation, thrombosis, ischemic cascades, and so on. For more cardiovascular type of applications, but you have to change the focus to sense thrombin instead of MMPs. I don't have time to show you that. Earliest clinical judgement application is an image guided surgery. I tried to show you how you could help. Drug delivery will be more challenging because because background labelling cannot be filtered by clinical judgment. It is good, I think, to help the surgeon, because tumours cannot acquire resistance to surgery. That's the one very big advantage of surgery. If you actually cut it out, the tumour, and drop it in formaldehyde. I don't care how much it wants to grow back or how rugged the cells are, they're dead. Okay, if tumour is in non-essential tissue and it can be detected early enough, then surgery usually offers a complete cure. It was important to work in vivo, we found, and directly with clinicians obviously. I can't cut open a mouse let alone a human being. The future success or failure obviously will hinge on clinical trials because in my lab we could demonstrate the workings on a mouse but who cares about them. It's whether it will really, really, really work on people, work in people. And fortunately just about a week or two ago the small company of Avelas Biosciences treated its first patient. So finally, at long last, it is put into humans and the patient, so far, did perfectly well. But there is a very long road of development ahead of us, before we know whether it's truly valuable. Now, I'd like to switch to the neurobiological story, life-long memories. We are hypothesizing our story as a pattern of holes in the perineuronal net, the PNN. The PNN is a specialized form of extracellular matrix deposited around neurons during critical periods in development in the brain. The PNN is interrupted by holes. The holes are the places that synapses occur. Large holes in the insulation material, that is in the insulation caused by the PNN, correspond to large synaptic context. That can be strong. Small holes in the PNN give a weak synapses, no holes, gives no synapses, and no connections. PNN is there for like an enormous punch card in 3-D. The one problem is I recognize that all of you people, these young students, have probably never even seen a punch card. That was very common in my era but at least look up what it looks like. The crucial idea is that the stability of synaptic strengths and therefore memories, we respect Donald Hebb here, stems from the stability of the PNN, whose molecules can individually last a lifetime. Single molecules of extracellular matrix. Whereas, the molecules that everyone had thought of that were actively retaining memory, such as the actual macromolecules that make up the synapses themselves, turn over in just a few days, and just turn over continually. For your lifetime. We don't need to postulate a copying mechanism that preserves information faithfully over hundreds or thousands of generations. If you believe that macromolecules, synaptic macromolecules are the source site of memory, you then are trying to stabilize a picture in writing on water. You’re trying to, you will have to have a copying mechanism that preserves information even though the individual molecules are turning over. How does this happen? Intracellular molecules always have an automatic turn-over mechanism which is the proteasome and so on. The phenomena discovered by Hershko and Chiechanover, who got Nobel Prizes and are talking at this meeting. Fortunately, extracellular matrix lacks an automatic turn-over. When you need, want to turn over the extracellular matrix and remodel it, there are ways but you have to have specific proteases. This is a little bit closer look at the perineuronal net and it consists of chondroitin sulfate proteoglycans, CSPG's, with these sorts of names, Aggrecan, Versica, and so on. Held together by carbohydrate chains anchored into the plasma membrane. By the way, morphologically as opposed to biochemically they were already known quite a long time ago. The great pioneer Ramon y Cajal who got, I believe he got his own Nobel Prize in the early part of the century. He could see them. In a more modern way of looking at them we visualize green, the chondroitin sulfate proteoglycans versus in purple, prosynaptic density, protein 95, a traditional marker of this synapses. You see some of the relationship between the two. One of them is, they interweave but don't overlap. That is the PNN versus the synapsis. We can also look at the hyaluronan or some of the link proteins. The same story we can see. That was in culture. We believed that during the critical period during which the PNN is first laid down, in mice that is predominately about four weeks old. Four weeks of age and the period from four weeks to six weeks is the time that the PNN is laid down. Then after that if learning is to occur the very, very locally and microscopically, the PNN is further degraded at defined sites and creates a hole. And this allows the synapses to go from this state, which is no synapse to a strong synapse when the PNN has been destroyed at the position of the synaptic cleft. The negative image of this PNN is formed and permits the synapses to poke it's head through. This is a more modern electron micrograph and in this here's the whole cell enlargement here. You can see a presynaptic of the upstream portion of the neuron, and the downstream portion, and this is a synaptic communication. Everywhere else, but the synapses is the jacket of PNN which preserves the position of the synapses, we believe. The half-life of the neuronal intracellular proteins is, as I said, then known to be rather short. Some three or four days half-life. If you try to store memory, as I said into something in which only last three or four days and yet the memory last for 80 years say, as for an age of a human being, you would have to have about 8,000 generations of copying and recopying and recopying. We tried to test this by what is the actual length of time of the perineuronal net, which had never been looked at. How stable are those proteins? We postulate today that they are stable but is it proven? Well, the way we did is take nitrogen, normal mouse, and they normally have essentially 100% nitrogen 14, the normal stable isotope of nitrogen. Then we started feeding them for ten weeks with nitrogen 15. This is a rather experiment, you have to by a lot of mouse chow, expensive mouse chow, which is specially modified so that only the nitrogen is nitrogen 15. Then after ten weeks we gave them one week of happy time with the mating, the male mice could fortunately be with back to the 14. Three weeks while they gestated, and then finally they gave birth. Then the next generation we kept them to day 45 or six weeks. We had in essence twenty weeks of labelling. That is essential because the brain very, very slowly does it equilibrate with the food that it eats. It's not every day. It's the brain recycles it's aminoacids so you have to do extra hard work. When we sacrifice the cohort which we placed back in nitrogen 14 food, the most meaningful was six months feeding. After 180 days continued feeding of nitrogen 14, after 20 weeks initially, all the traditional candidates are already back to nitrogen 14, because they've had many chances to wash out the nitrogen 15. But the perineuronal net, as shown by Versicans and the hyaluronan link proteins have retained the most nitrogen of nitrogen 15. Not all but a large amount retains quite as well as myelin basic protein, which is one of the most stable proteins in the brain. This we consider constitutes evidence that nitrogen, that the perineuronal net is indeed can be stable for long periods of time. Much more stable than the synapses. So proposed molecular memory mechanisms have been several in the literature. CaM kinase 2 was proposed by John Lisman for the site of memory. Todd Sacktor put forward PKMzeta. Eric Kandel, a famous neuroscientist, who's a Nobel Prize winner himself, proposed CPEB, translation factor. All of these are intracellular proteins. They have up the problem and very elaborate mechanism that you have to postulate for how the memory could be a single train molecule, can passage message to a second progeny molecule. Whereas, we believe that none is required. The molecules themselves don't have to be copied. We think that the mechanism activation will be proteolytic attack by the local degradation and formation of which allows the strengthening of synapses in the vacant space. The read out mechanism is this competition with synapses for space. You either have PNN or the synapses. The PNN mechanism will give you a very low vulnerability to metabolic interruption. If you have a coma, for example, or cool the brain, all these intrasynaptic mechanisms will be vulnerable to metabolic interruption. The perineuronal net is a stable and it is non energy requiring to survive and it's just going to give you a much lower vulnerability. Also, it can explain critical periods. Now, what about the literature evidence? Literature evidence already says that there is a good deal, that the PNN can act as, can be served as proteolytic degradation such as the protease MMP-9. If you check they are locally translated and secreted to response to synaptic activity. If you puff and dodge exogenous MMP 9 from a bottle, if you puff it on locally you can provoke spine enlargement synaptic potentiation. Here's an experiment which is done on a whole animal. Global disruption of the PNN, for example, the special enzyme can actually reopen the critical periods and it call allow animals which have undergone monocular deprivation. Which are normally then stable but after this experimental procedure injection of enzyme into the brain to weaken the perineuronal net again, you can get a second round of training which reverses monocular deprivation. Then similarly you can wipe out fear memories and make them susceptible to erasure. Inhibitors have been claimed to cause late phase synaptic potentiation. Other enzymes, by the way, are very likely to be important, but we just don't know enough about these enzymes. The MMP's, you can buy them, EMP Sciences makes them and makes reagents, but the other enzymes that are also important. You can't buy so the emphasis, we have deliberately emphasized, such as MMP-2 and 9. Now convincing direct truth would be implantation and deletion of specific memories, but this would require knowing how synaptic activity encodes experiences. If any were in the audience for Professor Tony Gabel's talk early this morning you know that we're very, very close or even possibly there at this challenge. The one experiment and the final experiment that I will describe to you is we tried to inhibit MMP-9 and we said it's known to prevent synaptic growth but we wanted to check whether it affects the mice behavior. Therefore we did an experiment of this sort. We compared their ability to recall within only 24 hours, which is a long period by ordinary standards but for us it's a short time. We compared it to a testing them fully one month later, 30 days later. We either injected nothing or DMSO, inert substance, or the drug Prinomastat which is an inhibitor of matrix metalloproteinases, or we used transgenic mice. After 24 hour cue test not much was changed. Depending on whether we interfered with the PNN proteases or not. If after 30 days and testing them with the cue of the sound pulse about half the memory was gone, was half as strong. If we used Prinomastat relative to control or we did the genetic knock out with either, obviously once you've knocked it out genetically you don't care whether you further give it Prinomastat. In neither of these conditions by inhibition of MMP-9 we blocked about half of the memory. Now, the problem of course, is that MMP-9 is only one enzyme and we don't think that it's responsible for all the enzymes but it may be responsible for half. My lessons from these two topics. First thing is that their surprising role of extracellular matrix and MMPs. I take this as rather significant because all my original career and all the work of GFP was intended at strengthening intracellular mechanism and our ability to watch intracellular signal induction. Here am I being forced to learn about extracellular matrix. Basic and applied research, applied research for the cancer, basic research for the memory can be of equal value and difficulty. Finally, there is a genuine likelihood of failure in these examples that I've given. These are very recent. If I gave you reminiscences about GFP or take an example, if you go and watch most movies, say you go and watch a James Bond film. You know that there is success at the end. The hero lives to fight another day. No matter what risk he takes he survives and triumphs. James Bond always triumphs. Similarly across many, many movies. Of course, here at Lindau, every time you hear Nobel Laureate do retrospective of, "How I got the Nobel Prize," that has no likelihood of failure. This is still got plenty of chances to fail. Most preclinical avenues don't reach clinical trials. Now that we have reached clinical trials most of them will fail. Chemotherapeutic and genomic approaches are far more popular in cancer research than helping surgeons. Turning down neurobiology, the PNN hypothesis is far from proof of acceptance. Very, very sketchy evidence right now despite my always talking about postulate. Practically no one works on it, of the mechanism of the longest term learning or memory. It's admitted to being of philosophical importance but basically no body works on it. Finally, anonymous refereed funding and publications are increasingly hard to get. Whereas others happily tell you about the 10 billion dollars that they spent on their experiment. We're finding it difficult. There's distinct possibility yet exists, that I am going senile, as has happened to other prize winners. So take everything you heard with a grain of salt at Lindau. Thank you.