Jack Tuszyński

Transcription

>> The next speaker is going to be Jack Tuszynski from the University of Alberta. He's going to talk about cytoskeletal information processing, intracellular information processing. >> TUSZYNSKI: Thank you very much for the invitation. I'd like to start by thanking the organizers for putting this together. It's a great workshop and a great place, very inspirational for all of us. I'm a theoretical physicist who is now a professor of experimental oncology, so I guess I've reached my level of incompetence. But I'll be talking about something else today. First of all, I acknowledge the sources of funding and, in particular, technology innovations with Mike Weiner who has been faithfully supporting my group for a number of years. But there's a lot of other sources of funding. Most of it, by the way, is for a different purpose, the--and collaborators. I want to acknowledge collaborators with Stuart and Vahid Rezania and Holly Freedman and Avner Priel, Horacio Cantiello and John Dixon, Miljko Sataric. And being here, I'm not going to go and give any references whatsoever. Just Google "Tuszynski tubulin" and you'll all get the references so I'm not going to waste any information on these slides. Okay. My day job is actually computational drug design and this is where most of the funding is coming from. That's a fraction of my group. I don't have kind of pictures available. We do computer design and testing and silico searches for novel cancer chemotherapy drugs. So that's basically what we do for a living. But because microtubules are everywhere, microtubules play a role in both the oncology aspects as well as the neuroscience aspect, which is more of a longstanding interest. And I'm going to try and follow the best I can the tough acts that we heard today, this morning, this afternoon. It's not easy and especially not with 5 o'clock, and for some of you, probably it's early morning. I want to start by saying that--and integrating some of these observations that were made today that in my opinion, the human brain is the world's fastest, most portable supercomputer cluster. And every word has a meaning here. And I think this is quantum biology. I will not say too much about quantum, being a quantum physicist, and overstepping my competence already. But on the issue of human brain and some of these capabilities that were mentioned, there are some numbers to quote. So we have 10 to the 10th neurons, plus-minus some percentage. Roughly 10 to the 15th synapses, so there's a hundred thousand synapses per neuron. And synapses operate at about 10 impulses per second, so that's 10 to the 14th FLOPS--sorry, 10 to the 16th synapse operations per second, which compares to 10 to the 14th FLOPS of the Blue Genes. So even at the most crude, most classical level, the brain is better than the Blue Gene. But I'm going to try and convince you that actually the brain is orders of magnitude better than the Blue Gene, maybe all computers in the world put together. A single brain. And of course, much more energy efficient as well because it only consumes 25 watts as opposed to 1.5 megawatts for the Blue Gene. So, I said it's a supercomputer cluster. So what are the little computers inside the brain? So is there anything inside? So, I'm giving you the--I'm also a bit tired so cut me some slack, please. Each computer is a neuron and each neuron has its own processing units, and you probably guessed which ones are these. But as a physicist, I also want to understand what sort of physics to apply to different spatial ranges or dimensions. So typically, we would apply thermodynamics to the level of an organism. And at the mesoscale, now we are grappling with nanotechnology, we don't really know what it is. Is it quantum, is it classical? So this is where cells and sub-cellular structures operate at. And biomolecules with 10ł atoms, we are stretching into the range of quantum chemistry and eventually, the single level, the single atom or tens of atoms, quantum physics. It's important also to keep in mind the energy scale in biology. These are all the important biological distances and energy scales and also affinity. So it's all biochemistry, reactions and how fast they precede their questions about time scales. I really want to draw your attention to the bottom of this scale, the GDP/ATP hydrolysis. And that's probably the only time I'll mention quantum today--the energy quantum, biological energy quantum. GTP is a bit smaller in energy than ATP--and I don't know what happened with the number there--it's between 2 and 10 KTI. I don't know; it got deleted, maybe it's below the frame. So, this is above the threshold of noise but not really incredibly much above the threshold of noise. All the quantum--all the biological processes use either GTP or ATP energy to continue. Therefore, it's important to keep this in mind. Different computers--again, this is cutting things off. And again, for the third time today, I want to mention Guenter Albrecht-Buehler. And you can Google "cell intelligence" and you'll find all about him and what he's been doing for the past 30 years. What actually Albrecht-Buehler demonstrated is--and this is the premise of my hierarchical scale in the organization of this supercomputer cluster. He demonstrated actually that a single cell is intelligent. And when you go to his website, you will see these hundreds of pages of illustrations, and this is one of the videos I'll try to show you. And it's actually--the cell is perceiving infrared signals. So it's electromagnetic signaling. Of course, it also perceives mechanical signals and chemical signals, which everybody knows about. But the fact that electromagnetic energy is received and analyzed by the cell, by a single cell, is incredibly important in this context, because the cell is already a computer in its own right and a very powerful one at that. Okay, so we'll just go with the flow. So, some of it was covered, so I'll try to quickly go through it. Microtubules and mitosis in dividing cells and non-dividing cells, different roles, different architectures, they definitely play a central role in every cell. Every eukaryotic cell has microtubules. There is no time to discuss the variety of the building blocks, which is incredibly interesting; the two billion years of evolution, this experiment on our planet in terms of testing. It's a lab, changing, mutating and finding out what is the best for what different cells. Use different building blocks. Tubulin is expressed by 20 genes in the human and different variants are used for different cells. Cancer also makes choices, which is incidental to this talk. In nerve cells, you have microtubules packing densely the axon and also present in dendrites. So I don't know if anything will work today. No? Okay. So, I just wanted to show this dynamic instability. There's a video that doesn't show, but microtubules are very dynamically unstable. They grow and shrink. It's a very different polymer from anything else in the body. DNA actin filaments, intermediate filaments, they don't behave like microtubules. So there's something special about it biologically and we would like to know what is special about microtubules physically or biochemically. And that's another cartoon which I stole from Stuart Hameroff in this particular case. So they self-assemble and disassemble. So I want to add another buzz word to it: evolvable computer, a recyclable computer. If we can capture this power, that will be incredible. And talking microtubules interacting with actin filaments with ion channels, this is documented in the literature. Now the challenge is the integration of these various levels in a hierarchy. And bottom up or top down, I don't care. We will try to look at it from the bottom and create models that stretch different orders of magnitude. So that's shown here in a schematic with a tubulin dimer and microtubule with maps and a bundle of microtubules inside the dendrite. And probably none of this will work either. And this will not work. And this will not work. So, sorry about this, but... >> [INDISTINCT] the directory, play it directly from the directory. You can do that. >> TUSZYNSKI: Maybe. Everything is skipping it. So, Anne O'Brian talked about the geometry, the topology, and beautiful crystal structures. I have some--what I'll show you is some computer reconstructions, an in silico understanding of these structures built together atom by atom. And this is incredibly powerful today. You can use computational resources to reconstruct biological structures that otherwise you can't see. Crystallography gives you fixed images at some level of resolution. So this is a video from actual Berkeley from Eva Nogales and her group, showing the formation of microtubules from tubulin dimers and the structure of dimers in great detailed ribbons, the helices and these sheets, and how they fit into this cylindrical structure, how depolarization is asymmetric. It's different from polymerization, the rings. And actually, gamma tubulin is used to nucleate microtubules, so it's a different building block forming rings from which alpha and beta-tubulin dimers grow into cylinders. I will stop this moving in a second because we don't want to fall asleep. But Stuart mentioned GTP and GDP hydrolysis, so beta-tubulin has exchangeable binding site for GTP and that is one of the possible modes of behavior. Conformational change is caused by hydrolysis of GTP. That's, again, the quantum of energy, just about two and a half KT, just above noise level, enough to cause these instabilities to take place. So it's a very powerful nanomachinery if you want to look at it this way. All right, I'll stop it now and go back to the slideshow. One of the structures that was not mentioned today and plays an incredibly important role in microtubule biology that we know of and, in my opinion, also in all the processes that were mentioned today is C-termini. C-termini of tubilin are not crystal or graphically-resolved. These are little tails, whiskers, that decorate the surface of a microtubule and they are very dynamic. They contain forty percent of the electro-static charge of the protein, being only a very small fraction of the mass, and they're involved in conformational dance that is highly dependent on the sequences. So these C-termini have sequence variability which is completely unique to each species and each form of tubilin in the human body. And I think that's where a lot of research will go into. And a word of caution if you use laboratory techniques, Anne O'Brien and colleagues, they have to be very careful what sort of tubulin they are using because each tubulin has different C-termini that are absolutely crucial to their functioning. Here's a reconstruction of the microtubule in silico. That's the entire--its millions of atoms here actually. They went into the computer program that created this image and you have C-termini that I mentioned decorating the surface. They have dynamical structures and I have some videos to show you how they have different modes of oscillations. So we've talked today about eight megahertz, twelve, eleven megahertz. Some of it may have to do actually with the dynamics of C-termini. And they are shown here prominently in a reconstruction of the tubulin dimers. So these are the protruding, in this particular case, straight up but that's not necessarily the confirmation that they always adopt. It was just for the purpose of simulation. Okay, so again, more--for the physicists actually, electrostatics plays a crucial role here because the way the structure is put together in this beautiful lattice, including the ring, reflects the attraction between the negative and positive charges as they are expressed on the surface of tubulin. That shows you how the ring is made for example. And this is the formation of the microtubule from dimers also following the principles of electrostatics. This has not been completely understood to the greatest detail and the energy involved in the formation but definitely the principle is there. And that's a reconstruction with an electrostatic surface rendition. So the red is negative charge, blue is positive. It has actually a very important meaning because microtubules have a net polarity. They are not--it's not arbitrary, which is the +end and the -end. Both from the biological point of view, one is more actively growing than, plus, than the other, minus, but also they have electrostatic polarity. So there's an electric feed along the microtubule which would explain some of the effects that we're talking about today and also some of the experiments that I participated in with Doctor Cantiello from Harvard. This slide also shows you the values of these charges which maybe, in view of these technical difficulties, I'll just skip. What I want to spend a few minutes on is C-termini and their dynamical states. Those of you who are physicists in the audience will right away hopefully think about spin-up, spin-down. Actually, the C-termini have two stable states that we demonstrated by molecular dynamic simulations. One is up and one is down. In fact, it makes contact with a positively charged patch on the opposite dimer. So a beta-tubulin C-terminus touches the alpha-tubulin positive dimer and vice versa. This actually also causes conformational changes, work that is currently submitted for publication. So at least four states, at least four, because for the dimer, up and down for beta, up and down for alpha. So 2x2 is four. And this is another simulation that I dare not touch that shows you how actually dynamically these states evolve. You can start with a down-state or bent C-terminus and run the simulation for a few nanoseconds and then they'll be--you'll see it unbind and vice versa. So these things oscillate. There are natural oscillations between these two states. I'll give up on all the videos, by the way, so. And these are the ones that I maybe I would like to show but maybe at the end. We'll come back to it, time permitting. What I'm showing you here, and we've classified them completely, it's almost like a library of dynamical states. So I'll step away from the microphone and talk like this for a second so everybody understands the message. We have these C-termini states that may go down or up. But they also depend very strongly on where they--what kind of tubulin they come from. So it's bovine tubulin or if you don't like it--or human tubulin. In the lingo of tubulin microtubules, it's alpha one, two, three, four, five. They are isoforms, isotypes of tubulin and beta one, two, three, four, five, six, eight. Each one has a different mode, has a characteristic dynamical mode. It's like a dancer. Somebody is dancing the waltz, the other is dancing the tango, somebody is cha-cha; they have their own little dancers. And I think this is very much in terms of resonance. They--so if you can now imagine the microtubule decorated with these dancers and it's made--one cell is making dancers from alpha-6 and the other from beta-4, they will be engaged in different dances and therefore, different resonant frequencies. I hope this message is getting through. And I'll show you these dancers. I have all of them but assembled--okay. So that's under the topic of C-terminal tail dynamics. And they--we have estimated that they oscillate on the frequency of gigahertz. So, maybe not eight megahertz but much faster than that. And this is now the experiment that was performed at Harvard with Horacio Cantiello. Horacio is a very skilled biochemist-biophysicist who is, to my knowledge, the only person in the world maybe outside of Tacuba who can actually grab microtubules with micropipets and manipulate them. He used micropipets which were electrodes also. So he sent electric pulses from one pipet, micropipette, to the other, measured the signal, and to his astonishment and mine also, this is what they found. Actually, these electrical pulses were amplified. So if--so the blue one is the initial pulse at the first end and the red is at the receptor end of the transmitter and receptor. And the amplitude is increased between, depending on the simulation, between two and five times. So this is a big puzzle and not--in addition to all these crazy things that Stuart talked about and Anne O'Brian mentioned, this is in a published experiment in the respectable journals, biophysical journals, reporting an amplification of electric pulses along the microtubule surface. We have published papers also theoretically explaining what's happening here and they involve actual ionic. So now, to some physical concepts and explanations. Ionic waves which were sent by the--one of the electrodes sends an ionic cloud which senses the surface of it, of the microtubule. The microtubule is actually acting to attract ions because it's negatively charged. We have positive ions, sodium-potassium, and they congregate around it and they feel it. So on one level, microtubule surface is a capacitor. On another, water is a resistor. Ions and water resist the flow because of the viscosity, so we have two of these. And there is also helicity that Anne O'Brien mentioned. So it also can involve inductions. And we created the RLC model that--sorry--that is sketched here. Its electrical circuit model which--with parameter values that were estimated from molecular biophysics. And one complication that is interesting and important to observe, especially in the context of the work from Tacuba, is that a microtubule is not a solid surface. It has pores, nanopores, periodically located between alpha and beta dimers. So ions actually are drawn inside the lumen because of the potential difference between the outside surface which is more negative than the inside surface which is more positive. So we've created a three-dimensional model that includes flow along, inside, and through the lumen. And this is published in this year's Physical Review. Again, Google it. And we found actually that it explains asymmetry--this is an I-V characteristic--asymmetry of the microtubule's conductive property. So Stuart mentioned this and now let me try and summarize some numbers. I started out by saying that the brain is probably more powerful than the most powerful supercomputer. I think we are scratching the surface. And here's a conservative estimate in terms of if you take into account the number of microtubules, the number of neurons in the brain, and the number of states that each dimer can encode--this is actually before we do the phosphor-relation calculation--C-termini states at least a four-pair dimer. Electron hopping, they're at--I didn't have the time here to show you that electrons also have double-well states in which they can hop, so let's say four. And conformational changes, this is documented experimentally, two per dimer. So conservatively speaking, you have 32 states per dimer for 1 gigabyte of processing power per neuron with a 100 billion neurons in the brain if you use your brain to full capacity. Einstein said that the only 10 percent and even, I think, he would be in this category. So let's, say, put these numbers together. That's 10 to the 20th bits per brain is my estimate of the computational capability of the average human brain. And if these transitions can occur at nanosecond scale, then you have 10 to the 28th FLOPS. Actually, some of you may be very skeptical of this estimate and will laugh at me. Let me just--don't laugh, because Von Neumann in 1950 came up with the same number, 10 to the 20th. Subsequently, we became more and more cynical about our ability to think. And you have the--you know, some of the gurus of modern computational science estimating only ten--one gigabyte--one gigabit per brain with two bits per second of visual, verbal, tactile, musical memory, and a human lifetime of two and a half billion seconds. This is incredibly depressing. How does it compare? My estimate is in red, and you can challenge me on this, 10 to the 19th if we used bytes. Okay, so that's still 100%. The total number of the data on the Web is 10 to the 15th, probably growing by one percent every minute. The Library of Congress, three times 10 to the 15th. So, I think we are definitely capable of incredible amounts of information storage. The question of how do you read and--write and read is of course of absolute importance. And with Stuart's explanation of how we see phosphorylation sites on tubulin as potential memory storage, we're getting closer and closer. I want to say this is not--tubulin and using proteins as memory chips is not our original idea. This has been around and I want to give credit to David Wishart from the University of Alberta who has been thinking about this for a long time. And perhaps, there can be even a greater capacity than what I mentioned here. But what I want to--before I forget and before it's--the time is over, I want to say something that may be a little bit on the speculative side. You must have heard about near-death experiences and people going through this life-history in their mind and everything is encoded upon. If we believe these stories, then if you want to store all this information that you ever received through sensory perception, this will be an incredible amount of information. And these numbers that Landauer and others are talking about are insufficient for that, so you really have to look deeper. And the second possibility is hypnosis. Also, under hypnosis, people recall a lot of events that normally you don't have access to, so they may be stored quite deeply in the brain. And this--just one more addition to Stuart's description of our calmodulin kinase. The original idea for this actually came recently two years ago through the experiment report in Science Daily and, I think, subsequently in science of PNAS--I remember the journal--where actually calmodulin kinase was inhibited in mice and the mice lost the memory of training in the previous half an hour. So that's how it kind of started the search for how calmodulin kinase is involved in memory and we linked it to phosphorylation sites on tubulin. Remember these pictures. So, my estimate regarding the capacity of the human brain through microtubules did not involve this aspect at all. That's another means of storing information. Now, we did something more that Stuart didn't mention. Travis Craddock, my student, asked if--the following question following after, actually, the advice from Stuart; where do anesthetics really bind? Only to receptors and synapses or is the possible--any possibility of anesthetics binding to tubulin, and if so, where? So, this is unpublished work but finished calculations. And I'm going to give you the sites of anesthetic binding on tubulin that we discovered through computational modeling. And it's very telling. So, I've highlighted in red some sites for anesthetic binding. Halothane was one of these, the volatile anesthetics that we tested. They bind in places where oncologic agents, the chemotherapy agents bind; taxol, vinblastine, colchicine, and also where GDP binds. So, this is a testable prediction that we'll be confirming in the lab. In other words, if this is so, anesthetics will be inhibitors of things like taxol. Another is they bind to a lot of tryptophans and interestingly, the serines and threonines in C-terminal region. This is exactly where calmodulin kinase binds. These pictures that Stuart and I just a second ago showed, the--what do you call it? The poodle? Right, the nano-poodle. The nano-poodle binds exactly where anesthetics binds; they're one of the sites. So it makes sense that you'll lose memory because the nano-poodle doesn't work. And finally, also bind in the positive patches where C-termini binds. I told you that C-termini bend over and bind to the respective part--positive patches on the beta or alpha tubulin. They also bind there. So, there is a lot of interesting information, I guess, to process. We have the skeleton of the model. And just to add to the complexity, you can have--in addition to these rather fast changes and information storage capabilities, you may also have in the brain long term effects of learning and sometimes, pathological developments. Stuart mentioned Alzheimer's disease and decoupling of TAO maps--map TAOS from microtubules. You can also have strengthening of these connections through enforcement--through reinforcement through learning and repetition. So this, again, a video clip that I dare not touch, shows you how these map interconnections dynamically change architecture. So that's another level of--and I guess I'm done. Allow me just to show one of these dancing--dances maybe with--after I go through the conclusions. The human brain, in my opinion, has an amazing computational potential. I call it supercomputer cluster. The neuron is a single computational unit. A cell, like any other cell, can exist on their own and performs complex computational processes at the level of quantum of biological energy. I dare not say quantum of physics but quantum of biological energy. If--it's an evolvable--itself, the neuron is an evolvable computer which communicates with up to 100,000 nearby computers. So, think about these communicating networks inside the brain. And within each computer, you have micro chips, which we call microtubules, as elementary computational elements with approximately one gigabyte of memory or random access memory or whatever else there may be there. And this is all I prepared for today. And I apologize for all these technical difficulties. Thank you very much. >> Okay. And we have some time for questions for Jack. Any questions? Okay, if there are no questions, I'll--oh, I'm sorry. >> Yeah, great talk. Thank you very much. I wanted to thank you especially for bringing up Albrecht-Buehler's work. And I wanted to highlight one other thing that you didn't bring out. And Albrecht-Buehler has an area he refers to as functional anarchy of the genome. I want to comment and tie this a little bit back into what Elizabeth talked about earlier. The first stake through the heart of the central dogma of biology was really when Barbara McClintock won the Nobel Prize for the notion of jumping genes. And in her acceptance speech, she said we will be surprised to find out the degree to which the genome senses its environment and shapes itself in response to that. Now, if that doesn't make Schrodinger's cat start to howl, the idea of a sensor and a shaper, it did to me when I was first looking at some random mutations. So I would be happy to direct anybody into the genomic literature. The second stake through the heart of the central dogma of biology now is with the sequencing of the genome, less than two percent of the human genome codes for any protein. And people who look at that say, "How can you generate something as complex as a human being from such a small number of genes?" They--what I think they haven't come to terms with yet are the non-coding portion of the genome. Forty percent of the human genome is mobile. It gets up and it moves around and it rearranges itself. And Albrecht-Buehler talks about this in the functional anarchy of genome. So, if we can talk some more, fine. >> TUSZYNSKI: Thank you very much. It's a very good comment, actually. And I think there is also a talkback to the gene. So, there's two-way communication. It's not just instruction; it's actual information flowing back in. And I wouldn't blame microtubules for everything but they may also play a role in this. >> Okay. Jack, you want to describe your video? >> TUSZYNSKI: So this is one of the C-termini. It's from beta-tubulin two. I'll show you--so that's basically--you remember these whiskers? They're not just randomly flailing around subject to thermal wind. When you stimulate this in the computer, you'll see that each one of them has a slightly different mode of vibration or oscillation. I'm going to show--I have all of them classified. We've done extensive simulations with my former student, Tyler Luchko who's now at Rutgers. And this is--you see quite different. And each of these modes has a different frequency so you can communicate with them. That could be another dynamical way of encoding. That's--this address, you know, this--the shape that I'm showing you is encoded by the gene. So it's in the genome, by the way, so it's beta 3 tubulin, beta 2 tubulin. So a different gene and every organism will have a different one. So if you want to cure cancer in mice, maybe you'll need a different frequency. And that's another one. So, I have many of these. But just to illustrate the point that they complete a--they execute completely different motions. So, that's the explanation. >> Okay, thank you. Let's thank the speaker last time.