Scientist Stories: Harald Hess, Super Resolution Microscopy and 3D Imaging

I now call to order the Society's 2,417th meeting in the 149th year since its founding on March 13, 1871. Good evening, everyone. My name is Larry Milstein. I am the president of PSW, one of the oldest scientific societies of Washington, D.C. Committed to providing a forum to further scientific understanding and inquiry. Welcome to our members and guests to tonight's lecture by Harold Hess in the John Wesley Powell Auditorium of the Cosmos Club in Washington, D.C., including our members and friends around the world who are watching the live stream of tonight's meeting on PSW Sciences' YouTube channel. We will begin with a few announcements, followed by a reading of the meeting minutes of the 2416th meeting and the lecture by Dean Romisch on the Argo flotilla of autonomous ocean sensors and what they tell us about the oceans and the atmosphere. We will then turn to tonight's lecture, followed by a question and answer period. When the Q&A session is done, I will present a thank-you gift to our speaker, make a few closing announcements, thank those who make PSW possible, and adjourn the meeting to the social hour. Please join me in thanking the sponsors of the— what happened here? 2019-2020 Lecture Series, the Policy Studies Organization in cooperation with the American Public University and a generous donor who has asked to remain anonymous. Also, please join me in thanking the sponsor of tonight's lecture, PSW member Tim Thomas, here in the front row. Thank you, Tim. And finally, by way of thanks, please join me Tonight, in thanking one of the members of the General Committee and the crew who keep this organization running and growing, PSW member-at-large and IT director Jared McQueen. Jared is the person who keeps the electrons humming so that PSW's website and other IT works smoothly 24/7, 365. And 366 in leap year. Thank you, Jared. We appreciate what you do. Thanks to everyone, all the volunteers who make these events possible. We just had an election, and the General Committee for 2020, per the results of the election, will be as follows: President Larry Milstein, Mark Clampin will be Vice President, Robin Taylor will serve as corresponding secretary. James Healan, recording secretary. Brett Magaram, treasurer. Member at large, Cotter Nixon. Member at large for web and IT, Jared McQueen. Member at large for social media, Anne McQueen. Membership director, Lloyd Mitchell. And special projects director, Cameo Lance. Thank you all for serving. I'm pleased to announce that the following new members have been elected to the society. Barry Bienia, associate professor at Eastern Virginia Medical School, interested in geology, paleontology, anthropology, and the cultural effects of scientific advances, who learned of PSW, quote, years ago, close quote, from a newspaper listing of community events. And Victoria Gimeno, an epidemiologist retired from a distinguished career at the Centers for Disease Control interested in diseases at the human-animal interface, environmental health services, remote sensing, geospatial analysis, and machine learning. And tonight's speaker, Harold Hess, whose background and interests will be clear to you in part from tonight's proceedings. Please join me in welcoming them to membership. There is, for all new members, a signed reprint of Volume 1 of the PSW Bulletin. If you have not received yours and you're a new member, please see me after the lecture and I will be glad to provide yours to you. And if you purchased a PSW ribbon, please see Anna McQueen at the rosette table in the back of the auditorium. Recording Secretary James Healan will now read the minutes of the 2416th meeting and the lecture by Dean Romich on the ocean drifting flotilla of autonomous sensors known as Argo. Delivered to the Society on November 15th, 2019, right here in the Powell Auditorium.

James.

Thank you. Hi, good evening everyone. On November 15th, 2019, in the John Wesley Powell Auditorium of the Cosmos Club in Washington, D.C., President Larry Milstein called the 2,416th meeting of the Society to order and at 8:01 PM. He announced the order of business, that the evening's lecture would be live streamed on the internet, and welcomed new members to the society. The minutes of the previous meeting were then read and approved. President Milstein then introduced the speaker for the evening, Dean Romich, distinguished professor of oceanography at the Scripps Institution of Oceanography, University of California, San Diego. His lecture was titled Argo, 'a flotilla of autonomous ocean sensors observing the ocean's role in climate systems.' Satellites can observe small net energy fluxes in the Earth's upper atmosphere, but those observations do not reveal the full story of net heat flux in the ocean, atmosphere, and cryosphere. Romisch said, 'The best answers to whether there is more radiation entering Earth's atmosphere than leaving it are to be found through global ocean observation.' because oceans are where net heat is predominantly stored. Global ocean observation will also provide data relating to basic research, ocean reanalysis, and weather forecasting. The first global-scale oceanography began in 1872 when the HMS Challenger, a British warship converted into a research vessel, began its 4-year-long expedition. Romish explained the methods Challenger used to gather its data, which was first to reveal basic ocean stratification, including temperature measurements down to the ocean's bottom in 300 places worldwide. Few ocean-scale expeditions followed Challenger until the 1920s, when the RV Atlantis, RV Meteor, and RRS Discovery were commissioned and made ocean-scale surveys in the Atlantic Ocean. In the following decades, scientists developed more sophisticated instruments to take more precise temperature, salinity, and depth measurements. John Swallow invented an early neutrally buoyant acoustically cracked sensor in 1955. But through the World Ocean Circulation Experiment that began in 1991, subsurface ocean observation still required the presence of a research vessel to take measurements. In the 1990s, pumps were developed to control instrument buoyancy, and profiling float sensor packages were created, making the Argo program possible. Argo floats are dispersed around the world and collect a temperature and salinity profile at a trajectory every 10 days, with data returned by satellite and made publicly available within 24 hours. The floats descend from the sea surface to a depth of 1,000 meters drift there for approximately 9.5 days, descend to 2,000 meters, and then immediately surface and report data using iridium cellular communication. The cost of an Argo profile is roughly $200, compared to a shipboard profile that costs over $10,000 when ship time is included. Argo was planned in the late 1990s to create a network of 3,300 profiling floats. ARGO reached its float target in 2007 and continued to grow. Today, there are approximately 3,850 ARGO floats, with approximately one float per every 3-degree by 3-degree square in the world's oceans. Romich described the multinational partnership vital to ARGO's success. ARGO data has contributed to over 3,800 research papers and in October 2018 celebrated its 2 millionth ocean profile. Romesh then described Argo's role in the Global Ocean Observing System. Argo data shows a global warming trend in the upper ocean, which correlates with GOESHIP deep sea data showing deep sea warming at lower rates. Argo data on interannual variability in global ocean heat GAIN also correlates with satellite radiation measurements, and Argo density data correlates with multiple sources of sea level data. By the end of 2020, Argo will have more than 150 Deep Argo floats operational at depths up to 6,000 meters. Argo is also developing biogeochemical sensors measuring 6 different parameters to inform scientists about matters including ocean acidification, deoxygenation, and ecosystem health. Romich advocates gathering as much data for as long as possible. He is optimistic the publicly— public availability of that data will lead to research and analysis not yet contemplated. After the question and answer period, President Milstein thanked the speaker, made the usual housekeeping announcements, and invited guests to join the society. At 9:59 PM, President Millstein adjourned the meeting to the social hour. Temperature 7 degrees C, weather partly cloudy, attendance in the Powell Auditorium 76, and viewing through the livestream on the PSW Science YouTube channel 24. Respectfully submitted, James Healan, Recording Secretary.

Are there any comments on the minutes, corrections? If not, I will entertain a motion from a member to accept the minutes as read. I see Bob had his hand up and Terry is second. And all members in favor?

All members opposed? The minutes are accepted as read and will be posted to the website in due course. The video of the Romich Lecture has already been posted to the PSW Science YouTube channel and the PSW Science Vimeo channel and can be accessed directly there or via the PSW Science website. We do have a special announcement tonight by request to say happy birthday to Carl Merrill.

Happy birthday, Carl. Did you bring a cake, Alice? We now turn to tonight's lecture on super-resolution and 3D imaging, brain circuits to cell ultrastructures. It's my pleasure to introduce tonight's speaker, Harold Hess. Harold is senior group leader at the Howard Hughes Medical Institute Janelia Research Campus. Previously, he was group leader there, and for several years before that, he was director of the Applied Physics and Instrumentation Group. Harold invents new microscopes with the potential to impact biological research. Currently, he is working on interferometric photoactivated localization microscopy, more easily known as PALM. To produce 3D super-resolution images, 3D electron microscopy using a variety of high-throughput and sectioning approaches, and correlative tools to bridge optical and electron microscopy. His recent work has extended PALM to 3D super-resolution microscopy and its application to research in cell biology. He also is developing 3D electron microscopy techniques for volume imaging of cells and neural tissue. And he recently combined these techniques with cryogenic PALM to produce images correlating information from both techniques, which you will hear about tonight, I believe. Previously, he worked on hydrogen atom trapping and Bose-Einstein condensates and evaporative cooling to achieve BECs in the atomic gases. He developed low-temperature scanning probe microscopes to visualize atomic and subatomic physics phenomena, including, for example, vortices in superconductors. And he worked in industry developing advanced experiment—equipment for hard disk drive and semiconductor inspection and production. In 2005, he and a colleague, Eric Betzig, learned about photoactivatable fluorescent proteins. And using their expertise in inspection technology, invented the method of photoactivated localization microscopy, PALM, to reveal details of cell structure beyond the diffraction limit. It was built in his La Jolla condo, tested at the National Institutes of Health, and contributed to the 2014 Nobel Prize in Chemistry. Harold is a member of the National Academy of Sciences and a fellow of the American Physical Society and the American Association for the Advancement of Science. He earned a PhD in physics at Princeton University.

Okay. Thank you, yeah, very much. Thank you very much for the introduction and it's a real privilege to be invited here and an honor to speak in front of all of you. And so tonight I would like to discuss various forms of microscopy. And to put this into a— and I'll give it a bit with a historical perspective to add some information and also weave in my personal story at the same time and come into, you know, current topics of research in a variety of areas of biology and neuroscience. Before I launch into this, I'd like you to sort of think of the rest of the talk, and there may be 3 themes in microscopy. First of all, Cole, is the equipment itself, driven a lot by optics technology over the years. And especially recently, the data acquisition technology has become more important. It's very similar to telescopes, right? The optics technology and data acquisitions are just leading to great progress in in both fields. Now, microscopy has something that the astronomers cannot do. We can stain our sample. And that gives us very powerful, unique advantages, which you'll see how they play out in cheating what might normally be, you'd think, the limits of physics. So let me just go back about 400 years or, And at the birth of microscopy, it's basically some spectacle makers were having some fun with their glass trying to look at small things like bugs. You know, Janssen invented it. Robert Hooke sort of refined it, and van Leeuwenhoek also sort of developed even better lenses for it. And, you know, they learned— yeah, this is the pointer here. Yeah. And they learned They can not only see bugs in greater detail, they discovered cells, which was transformative to our understanding where life is coming from. They saw small objects swimming around cells, bacteria, which is completely new and novel. They even sort of saw the first examples of neural tissue way back hundreds of years ago. And over the next few centuries, things gradually improved. They learned a little bit more about better lens making, able to deal with spherical and chromatic aberrations, getting to ever finer focus throughout the 1800s. And then finally, toward the end of the 1800s, this person, sort of a physicist, understood that light is made out of waves. Waves can be focused only to a certain extent called the diffraction limit. So if you focus light with a wave, you can basically get a— whoops— a bump, let's say about this big. And if light wavelength is about 500 nanometers, it turns out you run through their equation, you can maybe see about 200 nanometers. That's sort of like the finest you can see with light, and that's considered the diffraction limit. This is an important number for all light microscopy. Diffraction limit is defined by what it takes to distinguish two close point objects. If they're a little bit too close, you can't tell them apart. So again, considering these three, this triad of three things, people started playing with the new microscope, which was diffraction-limited, and developed stains developed by Golgi. And this person, Cajal, began to look at brain tissue and could begin to see rather new things. And the brain really is not just some mass, but actually has this fine structure inside. And this was using at the time this camera, Lucida, and could sort of draw while imaging at the same time and just made hundreds of these pictures and sort of ended up being the birth of neurobiology— neuroanatomy. At another level, this person, Robert Koch, was concerned a little bit more about human health. And again, using these new diffraction-limited microscopes, played with staining extensively and looking at the sputum of people dying from TB, which was like a great problem, a devastating problem for a lot of people in the advanced countries at that time even. And was able to identify the TB virus and developed a new image acquisition technique, or adopted it rather, which is photography. And of course extended it to other bacteria. Infectious agents, cholera, anthrax, and I think it's had big impact just for the world as a whole. Going forward again, this person, Ernst Ruska, started playing with electrons. And the potential resolution there is a completely different story. You can go down well below this limit of 200 nanometers before, down to about a nanometer or maybe even beyond. That's more or less at the atomic level. But again, associated with this issue of staining, a whole new methodology had to be developed to prepare samples so that they could be visible. So here, for example, people learned how to stain the sample with heavy metals that made them light up for electrons in a way which was unique. They replaced the water with epoxy because it had to go in a vacuum, and they turned it into plastic as one example, and used photography to take the data. So there's a whole fancy sample preparation technique which we're even playing with today a little bit. But in the end, then you could see even finer details of cells. This, for example, is the Golgi apparatus. Who you've heard about just a second ago, mitochondria. Essentially everything which is a lipid bilayer membrane gets stained very well with these osmium metal atoms, and they just light up and define just a beautiful global context of what's going on. It's a very powerful tool and really allows us to peer inside these cells. So, just to briefly summarize, In the beginning, we had eyes, right? And then microscopes came along and they gradually improved until finally we got to the diffraction limit, barely able to see bacteria and maybe some of the bigger viruses. And then electron microscopes just sort of blew right past that. And so now there are other dimensions that begin to, that one can look into. Going into 3D. Biological materials aren't 2D. They don't fit on a sheet of paper necessarily. And one way to do this is simply to take a sequence of pictures at different depths. And then this was done initially in 1958 by Ströström, just trying to see the structure of a synapse which communicates between two nerves. This is a little video which just sort of gets the idea. Diamond knives were invented in 1955 as a preferred way to cut thin sections. This is a piece of plastic or plastic-prepared sample with metal staining that lights up all the membranes. You shoot electrons through, you get these pictures. That's neural tissue. And then you can sort of color in the boundaries of all the, you know, what's inside the nerve cells. Then you just reassemble and draw by hand in these early days, and you can have the beginnings of what could look like a 3D image. That principle got extended even further with effort by Sidney Brenner, when they thought, let's get the first real significant organism, a worm, C. elegans. And decided that time also to take a series of pictures. You can see the film roll here in the back. And here you're just literally painting in by hand each little part. And over a period of 15 years of many people, you eventually can write a paper which begins to show what the neural circuitry looks like. And it looks like that. I'll talk a little bit more about that later. But essentially, you can sort of map various inputs to outputs. And what is it going through in between? You know, there are sensory inputs, there are motor outputs to either the, you know, to the movement or to, you know, the ingestion of food and different aspects like that. This animal has only 300 nerves. So everything is highly, highly specialized. So let's just take a look at this and try to think, well, how much data is that? And it's actually quite a bit. If you have 70 microns by 70 microns, that's the size of the worm, 200 microns long, and you're looking at it at a tiny little resolution of about, let's say, 8 nanometers, which is What you need to really see it, you're probably talking close to the order of a teravoxel of data. A lot of data comes in very, very, very fast when you begin to think what are the challenges to really see biology down at this scale in 3D. Here's another way to put this in perspective. This is an important chart, so let me just spend a minute on it. The horizontal axis down below, Whoops, I didn't mean to do that. Let me try that. Okay. Horizontal axis down below tells how big the samples are. Up here, we're talking about a cubic millimeter. Down here, a cubic micron. So this is going by orders of magnitude very, very quickly. The vertical axis is resolution. Down here, it's really detailed, fine resolution, 1 nanometer. Up here, we're talking about a much larger scale, maybe 1 micron. So if you want fine resolution in a large object, you're literally operating over here. And you're probably talking about, you know, petabytes, you know, just huge amount of information. The C. elegans, as I said, sort of falls right on this one corner right here, about a terabyte of data. Fly brains, a little bit more. If you go a little bit further, mouse brain is out here. It's like about an exabyte. Talk about a human brain, maybe you're about a zettabyte. This is just going 3 orders of magnitude at a time. It goes very quickly. So, and at about, in the late 1990s, I was actually in this industry, this data storage industry. And it was just amazing to see the level of progress. Which has taken place over the past, you know, what is this, about 50 years. At about this time, compound annual growth around 2000 was about a 60%, escalating to about 100% improvement every year. And here we've gone up about, on the vertical scale here, that's almost 10 orders of magnitude. In capability, and that really has profound impact on what you can do with the data. Now, also what happened is that prices have come down in a huge way. Again, orders of magnitude. In fact, even here, this is coming down even faster than the rate of generating more capacity. So this is a great bonus for people who want to use these hard disk drives. They're just becoming cheaper. There's overcapacity in the industry. And so, you know, and price dropping this fast also means a lot of economic stress, companies on the verge of bankruptcy., and I lost my job. This is also a point where, just to point this out, prices have gone below the price of paper and film as the preferred storage media. So that was both a blessing and a little curse in disguise. I think more of a blessing in the end. So faced with this question, now what? I was a little bit too far away from my former research directions to really have an easy access to any new job at a university. But I did have a colleague who I knew from Bell Labs, Eric Betzig, who had done other experiments with near-field microscopy in a research environment. And he also had ventured into the, you know, into the business world trying to produce new equipment to save the American auto industry. And it didn't quite work out well for him. So he encouraged me to join, and we were trying to think, well, where do we go next? And we sort of read up on literature, sort of came back a little bit to the theme of microscopy. And we hid ourselves behind these little shell companies. It was just really ourselves. There really wasn't any company behind it. Called New Millennium Research and New Quest Research, trying to figure out where we go. One of the things that we learned about was a new stain, the green fluorescent protein. It was invented by, or discovered by a number of people, Roger Tsien included, where you can actually label the cell genetically. You can go into the gene, put a little code in, for this particular molecule, and if you shine light on it, it can glow. And the beauty is that it's highly specific. You can attach it to a protein of your choice. You can make the mitochondria light up. You can make— whoops— you can make the, you know, the ER light up or any other protein that you're curious about. You can just see where it is. The specificity is amazing. And it even works on live cells. And in our— let's see, and so just to put this into one more perspective, this molecule is small. There on the right side, that's a diffraction-limited optical spot. In the middle is about the size of the molecule. What you can image way bigger than the size of the molecule, 200 nanometers versus a couple of nanometers in size. Eric previously was doing some single molecule imaging. It was known that individual molecules can be imaged, and that was all cute. But in our wanderings, we actually learned about something brand new. Some of these molecules can switch. They can turn on and they can turn off. And they can be dark and maybe a few can be on. So, So that was sort of like the magic key to a missing puzzle for us. Let's go back to Abe. And here he could observe only at the diffraction limit, let's say 2 particles right here. So let's say you turn on only 1. You can look at that fuzzy spot and you say, well, gosh, the molecule, that really must be in the center. It's really a much smaller region where it's probably located. And now you turn on another molecule and the first one goes dark and you say, oh, I know where that position is too, much more precisely than what I see in that fuzzy spot. It's there. You can add the two together and now you can resolve the two molecules, whereas before you would consider this is not resolvable. So just by the magic of these molecules being able to switch, you can sort of exceed and transcend this diffraction limit. So, we said, "Wow, we got a cool idea." And there's my living room. And we set to work. And we knew this was easy. We didn't want to go run around for grant money or anything. You know, for the price of remodeling a bathroom, you can do this yourself. And this is the idea. Again, over here, if all of these molecules would be on at once, you would see a little diffraction-limited blur like that. Your ability to sort of turn them on one at a time allows you to localize it, redraw the lines, and you can get super resolution. And we applied it, went to NIH, teamed up with some other people there. Here you can see the molecules blinking. And over on the left side, you can see the resulting image, which is super resolved. Essentially, what we did is we've taken this diffraction limit, which was the limit of optical microscopy for about a century, and just sort of scooted over a little bit by about an order of magnitude. Here's another example. You could extend this principle into the third dimension, and you can go what previously would have been an optically diffraction-limited image like this from conventional microscopes to something of this order. So, at that time, there was a new place which was just being built a little bit on the other side of the Potomac River called Janelia Research Labs. At the time, it was called Janelia Farm Research Campus. Their mission was to do neural circuits and imaging, And as an aside, flies were considered a cool organism. I didn't really know anything about neural circuits and I still don't know that much. Flies, they're very different, but I did know, have a little confidence in the imaging side. And lucky enough, they hired me. And so I've been there ever since working on these projects. And I might add, Eric Betzig also was sort of working at this place for a while and was later well recognized for this work too. So why the fly? The fly is, this is what other people say, it has a very rich behavior. It can navigate, it knows where it is, it has memory of places. It has complicated courtship rituals. It can get addicted. Frustrated flies can drink. They groom, they sleep. There are a lot of attributes which are very much what, you know, we're all about too. So they're amazing little organisms. And the nice thing is there are a lot of genetic tools available too, which is a big bonus for researching it. And for me, when I sort of just look at size, they're not too small like the C. elegans, which has only 300 neurons. It's not too scary big like a human, which would be just technically impossible to fathom, but about just in the right sweet spot in size to look at its brain. That was the challenge. Its volume is about 40 million cubic microns just for the top brain part. In order to resolve it, we need to do these 8-nanometer voxels in size. And if you add up how long does it take to do that imaging and we sample at 1 megahertz, it's like about 8 years of operation. That amount of data is a lot of data. And what's even harder is that the data has to be perfect. You're looking at wiring. If there's a little gap from the left side to the right side, You can't tell where these tiny little wires, nerves go at all. And what might be like a few years worth of effort could just be totally then wasted. So there's an extreme demand on perfection for that. So here we are again, fly brain in that region. And so we were exploring initially how to go about it. There's been the traditional methods of sectioning, which I showed earlier. You just take a diamond knife like this and cut, and that's actually been used to section up a fly, but has a little bit issue related to the thickness of the section, which I'll talk about in a second. This diamond knife cutting can also be put on sheets which are put on a flat surface, and electrons can scatter back up instead of going through. That's another approach. Or you can serve just take the object itself, cut it, and look at the block face itself. You don't have to deal with distortions of the section. We took a different approach, and this is something which actually is done in the semiconductor industry if they want to reverse engineer or look at failure modes in a circuit. You, instead of using diamond knife, use an atomic beam, actually an ion beam called a focused ion beam of gallium atoms, and that can just sort of sandblast right across the surface. And now you can remove tiny amounts of surface, nanometer-level amounts of the surface. And so here's a little quick overview. You scan the surface and then you blast away and see a new, expose a new surface a little bit lower. You scan that, get another image, blast away and you get a new surface and you iterate. You do this about once every minute. And here's the machine. There's a focused ion beam put here on the side. The SEM goes down and all this activity happens right here in the middle. And this is the kind of data that you get. This is a piece of neural tissue looking at it in all the three different directions of a cube. From this focused ion beam SEM approach. It's an isotropic pixel. No matter which way you look at it, it looks the same. Whereas if you have a section-based approach, you can't cut smaller than about 30 nanometers. It just becomes unstable and annoyingly difficult. So, because all of the processes in these flies are very fine, we decided to go a little bit the odd route. Most people are taking diamond cutting approaches and we just went with the FIBSEM. So we bought one of these machines and we turned it on and said, okay, let's go to it. And we're only— machine only lasted about 3 days and then things give out and you have to restart it. So the grand volume that we could achieve was only about this big. You know, we're like orders of magnitude short of what we really needed to do. So we had a problem. Just spent $1 million and, you know, a little bit far away from the goal. So we learned the FIB-SEM is not reliable more than a few days. So we needed a solution. And so we had to think creatively about it. And so there it is. We just have to make it more reliable. And so we spent a few years, I should say Sean, who's actually here in the audience, spent a few years really trying to perfect this on taming this focused ion beam. That beam can cut and any cutting is good if it's controlled. If it's uncontrolled, it'll destroy the sample and basically we'll lose our data or ability to get larger samples. So, you know, plus we wanted to minimize any interruptions that happened. Temperature variations, we're aiming this, you know, with a few nanometers accuracy. If temperature changes in the room, the beam can plow down. We had feedback techniques in there. If something runs amiss, the focus goes out, you know, we set things up so it can pause or send out warnings or emails just to make sure, absolutely sure, that things don't get damaged. And if something bad happens, you know, some component explodes in the microscope, then we can go in, overhaul it, repair it, and try to re-engage and make a graceful start right at that same nanometer level as if nothing ever happened. And so that took many years, actually probably even more than 5 years to really perfect that. But we did in the end. So initially, we then were aiming at— this here is a several-month run, 3-month run, looking at part of the optic lobe. That's where the visual processing takes place of a fly. This is what the data looks like. Here's a large part here. These are all cables, literally. It's called a chiasm, where nerves from one module go to another module. Here you can see the nucleus. Now, let's see, is this— no, you can't see nucleus anymore. These are mitochondria in here. These little tiny little dark spots right in between there or right there, those are the synapses. Those are where the nerves connect to each other and talk to each other. And so there's a lot of wiring. And you notice some of them are just incredibly tiny too. If you look at the details. So we now got a slightly bigger piece, but then we ran into yet another model problem. This FIB SEM can't run more than, or mill deeper than about 50 microns. So if we're here, you know, shooting our little gallium ion beam over the surface of the sample and we expose it, we actually get waves. It's like wind going on the surface of the pond. The milling and the sandblasting becomes a little bit uneven the further downstream you go. So, and the fly brain is way bigger than 50 microns. So again, we needed a— we had a problem. We needed a solution. And think about it. So, okay. Just cut the sample into pieces less than 50 microns thick without loss. And that's what we're doing here. Here you take your fly brain, imagine it like a loaf of bread. You're going to chop it up into about 30 pieces, each about 20 or 25 microns in size. You sort of lay them all out like that. You laminate them like a little credit card or something like that. Now you have a little bank of fly pieces that are ready for the FIB-SEM machines. Nice thing is it's parallel. So now you can get several machines, not just one. And you can begin to reduce the time that it takes to get the whole thing done. But you still need the reliability. You don't want to fail on any one of these things. So what do we do with this? In an initial experiment, we went after something that we call the hemi-brain. If you look in here, this is a frontal view here of a fly brain. The fly brain is actually sort of— has a lot of modules inside, little very visible components. Out here, this is part of the optic lobe. That's where visual processing takes place. And there's several layers of visual processing. This is the antenna lobe. That's sort of the earliest input for smell. Other things called mushroom bodies, that's just sort of descriptive names based on their shape, where the olfactory memory can be located or what decides olfactory response. And there are other parts inside. I'll mention one for navigation. You'll see that in a second, called the ellipsoid body. It looks like a donut. And if people do imaging of that while the fly thinks it's in a virtual environment that's rotating, it tracks the rotation by the neural activity being in different parts of the donut, like a little compass. And there are sleep circuits in there. There's a lot of stuff inside that head. These are the sections all cut up 25 microns thick. You can see here is this ellipsoid body, out here fan-shaped body, and you start to see a little bit the different modules here. Now we just zoom through and you can see just a lot of detail. Zooming really in at a very fine scale, You can see synapses, mitochondria, et cetera. Synapses, we had to sort of first find and automatically predict these guys. And often you have fan-in and fan-out kind of synapse structures. It's not just a one-to-one. So here one nerve goes into sometimes 4 or 5 other nerves. We have to predict that, try to do that automatically. And then the second challenge is to segment. At one level, you can imagine, okay, we can just sort of color in each one of these little neurons, assign it a different color. But now we have to trace that through the whole Z-stack. And trying to do that reliably, particularly when things get small, is a major challenge. And the data is huge. You know, it's not an easy processing task. And so we partnered up with, of all things, Google. They have computers and a lot of computing power. And they developed initially this flood fill algorithm that literally takes each nerve and fills and traces it out and does that throughout the whole volume. And at first pass, you can actually get rather remarkable shapes. Of what each one of these nerves look like. And this here is just a few nerves from the ellipsoid body where the navigation is centered. Data is not perfect. Computers make mistakes. So here, for example, two neurons are merged. They're the same color. It's like saying, okay, you have two trees, but the computer connected the twigs at the top. Some human is still needed to try to separate out, cleave it, and because we have a little bit better knowledge than just what this neural network was able to decipher. We have a team of about 50 annotators and proofreaders who can just go through the data, correct this, and make sure the data is reliable. That's a bigger project effort that that's anelia. And in the end, you get final set data like this. Here again is that ellipsoid body. The little handlebar thing in the backside, that's called the protocerebral bridge. It also has different little, about 12 different little modules that all talk to each other. There's one neuron, we'll zoom in, show where the synapse is. That's the raw data where it's coming from. And that's synapsing to another neuron over there. That's the raw data. And you can see the synapse there. And you can't do it just for one, got to do it for a lot. And thank goodness there are computers and storage to do these things for you. And so this is just one module out of that fly brain. Turns out there, even this navigation area has lots. So the whole brain, this hemi-brain, it's about half of it. We capture about 25,000 neurons. I think it's around 10 million synapses. And we can just literally map where do all these neurons project out from the different compartments. This is the central complex associated with navigation. Now we're in the olfactory. There's parts of the antennal lobe which does the smelling, and they project out to other parts associated with vision or with navigation. It's an amazing amount of wiring that's going on in such a tiny object. Then higher order, I'm not sure what that means. Yeah, maybe this is where the spirit resides. So, and then now, okay, once you have the data, there's a little bit of question of, well, what on earth do you do with it? At one level, you can say, oh gosh, which nerve is connected to which? And you try to tabulate it, make tools for browsing it, searching it, and try to make networks, graphs of what is working together. And you might end up with something like this. If you ask one particular question, it can generate that pretty quickly. So I'm not sure if you understand that. I don't understand it. But in here, there are, you know, questions of how certain neurons from the olfactory part map onto the fan-shaped body which is associated with navigation. You really need to get more downstream data to really answer these questions. But if you form graphs, one colleague, I found it confusing, but he sort of explained it in it this way, which I think is a very compelling explanation. Imagine you're a driver of a car and the car has several outputs and inputs, like the throttle, the radio, the brakes, the steering, headlights, and things like that. In a simple car, the driver controls everything. Now let's add a little bit of neural circuitry to it to make the car a little bit smarter. You don't want to brake and throttle at the same time. That's a waste of energy. If you go a little faster, you might want to automatically turn up the radio so you can hear it better. If you go faster, you might also want to maybe turn the air conditioner off because you want extra acceleration in an emergency. Or if the radio is too high, you might want to change the sensitivity of throttle because you might have a little different, you know, sensitivity to your foot. If the heat and the fan is blowing, you might want to turn up the radio again. If the windshield wiper is on again, that might mean there's humidity in the air and you might need the defrost on automatically. And at the same time, you might want to slow down a little bit and you might want to turn the anti-lock brake systems on if it's slippery on the roads. Same time, headlights have to go on by law. And with headlights on, maybe you want to dim the lights on the radio so it doesn't blind you when you're driving. And headlights might, I forgot which one is which. But anyway, you get the idea. There's a lot of complexity in there. What happens, I think a lot of these neural networks, or neural networks, they're defined by evolution. And they're there to solve a task. And you just sort of end up with multiple ways sometimes to address a problem and very complex problems. And you get just very complex graphs. So, it's just a little note. This connectome, this is something that Jeanine Lee is going to be releasing in a couple of months. It has 25,000 neurons, tens of millions of synapses, and maybe 60 million postsynaptic densities. A lot of distinct compartments being annotated and the whole environment to browse and make sense of it all. So what else can you do? This is for another collaborator. You can actually image neural activity. And this is an example of a fly larva. You can put special calcium indicators into individual nerves, and you can see the cell bodies light up. You can't see the individual fine dendrites necessarily, but that data can be combined with this 3-dimensional electron microscope data, and we can identify which— what the circuitry context is at the same time to get a little additional layer of understanding into what's otherwise just a static circuitry pattern. What might influence what in what order? Get a causality in there. So, if we look at that particular region, I'll just sort of zoom in on it a little bit. What we've done with this technology, we've essentially made a new kind of microscopy capability that addresses a region that's otherwise not accessible. There are techniques for doing 3-dimensional electron microscopy. They in the past have been section-based. They can get very good molecular resolution in some cases. But now if you have larger objects with this kind of capability, we can now look at larger objects in their entirety with a certain resolution level. Let me just back up briefly. And then of course, with an eye to the future, we're hoping eventually to maybe gain another factor of 30 or so in speed, which I can talk about later. So, so far our drive has been to image a whole fly brain. And that has an emphasis on throughput. Maybe at the expense of resolution. We can drop the electron beam current so that there's, you know, obviously then we have to image longer to get the same quality image. But the benefit is we have maybe twice the resolution. So there's a little bit of a new opportunity here. Put back in the context of this other graph, we can look at smaller objects with a little bit better resolution. But still, it's a new domain that's not really explored or served by any other technology out there. Here's an example of what the resolution difference might really look like. On the left is the traditional way of looking at it, at our high-throughput mode. On the right is the high-resolution version. And here's just an example again, looking at a part of the fly brain. And now what you can appreciate is we're looking at the subcellular components with a lot more fidelity, all the organelles. We can even see the ribosomes and the mitochondria, you know, in great detail and, you know, in completeness also. And that allows us to go in there segment out, we can, you know, map out each one of these neural processes. We can find the endoplasmic reticulum inside each one of these neurons. It's just shown up in yellow. We can define the mitochondria, pull out the vesicles, find contact sites, which are all relevant to the various biological processes. Here's another example. On the left, you're going to see the flagella of a little single-celled organism. See, and that's a little molecular motor down there. The same, there's a Chlamydomonas on the right, is a view through at its nucleus. And if you look around the nucleus, this is actually just very densely packed with all sorts of data. It's hard to really see anything. If you just take the outer envelope of that nucleus, crop away everything else, and see what's left over, it'll look like this. That's the nuclear envelope, and on the nuclear envelope there are polyribosomes, little string, little string of pearls. Each one of the pearls is a ribosome, and connecting it is a, you know, little RNA molecule. In between, let's say like right here, that's a nuclear pore. On the nucleus. This has been seen before. It's nothing new. For example, down here, this is a TEM image, but you have to get lucky. You have to cut that section just right to catch it. Here, with this technique, you just see the whole thing at once in its entirety, and you can get sort of comprehensive feeling or statistics of what's going on. So, Just to set a standard, we said, well, okay, let's take a standard laboratory cell. The HeLa cell is the standard one. Go at the high-resolution imaging mode, look at it, spend 1 month imaging, and try to just get a, you know, a good reference atlas-type cell. And there you see the nucleus. You can see the little polyribosomes decorating the outside, mitochondria. And the like. Now, that was the whole data. If you zoom in, you can see a lot more detail. There's the nuclear envelope going across the middle of the picture. That was a centrosome that just sort of went there. You can see another one just sort of floating up right there. Golgi apparatus, mitochondria, vesicles, you know, just everywhere. And there's down below, you can see the different parts of the DNA, the heterochromatin, the euchromatin. You can almost see the histones themselves, but they're a little bit blurry to try to distinguish them one at a time. Here's one example of a still image. You can see this little tube that goes across. That's a microtubule. It's like a little structural element inside the cells. It's about 25 nanometers in diameter, and a lot of transport of different cargo goes along these microtubules throughout the cell. So that data, it was just too massive for us, and we passed it on to another group to go in there and manually annotate just 1 cubic micron of it by hand. And this took several months of a few people to do that, coloring it in and defining it depending on whether it's Golgi, nuclear pores, nuclear membrane. And the like, and that's what it looks like. But that's 1 cubic micron. Our cell is orders of magnitude bigger than that. So, the whole point is that having this data can be used to train a neural network to apply that to the larger cell and see it in its entirety. Here's an example of the same cell now, and we can segment out different parts of it. Now, I get a little video zooming in on it. And here we're sort of first looking at the outer plasma membrane. This is still the outer plasma rescaled. And if we take the plasma membrane away, then, which will happen in a second, you'll see the endoplasmic reticulum, which is a fine network of fine tubes and sheets. That surrounds the nucleus, and that's the ER. And we can also look at all the ribosomes. They're tiny little dots scattered around throughout the whole cell. And we can subdivide the ribosomes into free ribosomes that are floating around, or ones which are attached to membranes, which many are attached to the endoplasmic reticulum or the nuclear envelope. In some cases, you can see these string of pearl kind of structures in there. Then this goes back. Here's another version, same dataset. There's nucleus initially, and now we light up a few other structures. Again, here's the endoplasmic reticulum, and these are the mitochondria. And just the various organelles here just going through in different sequence. They made about, for this cell, about 30 different categories of different parts of the cell to try to reference and analyze. This kind of technology sort of inspires a lot of people just to explore other samples. Here, for example, is a mouse liver sample, and people are interested how do the mitochondria, which you can see right in here, connect to the ER. Depending on the glucose or the metabolic state, you can get very different formations of mitochondria and the ER in the contact, and this is just a way to really quantify that, you know, in its entirety for a whole cell or maybe for several cells. And so I think there's a lot of promise for that. Okay, so finally, I talked briefly at the beginning about fluorescent images. And so far I just talked about electron microscope images. Those usually give a generalized context. They stain lipid bilayers. They're very specific. They give a lot of information. They give global context. But they don't tell you that much about which specific protein there is. These other techniques, you know, in fluorescence microscopy in particular, they can be highly specific to one particular protein. In fact, there are about 10,000 different proteins in a cell. And so it's a very daunting challenge, but we can try to get one or two of these proteins and maybe overlay that on top of the other image. The fluorescent images by themselves sometimes lack context because you're only looking at one out of thousands of different proteins, and you don't really know what's nearby. So, this is correlative light and electron microscopy. So, we're trying to add color to this little area in there. We only have a couple of colors. Out of the 10,000 colors, in other words, 10,000 different proteins, we're a little bit limited here. So a lot to work on. Traditionally, this is nothing new, people have tried it. Usually people use chemical fixation, and you end up in a dilemma where either you get like very good electron microscope images, but the way you prepare the sample destroys all the fluorescence on it, and it just doesn't look very good. So you end up in a good with a bad fluoresce— good EM and bad fluorescence. Or you can try to prepare it really well for fluorescence, which fixes it strongly, but it makes— sorry, fixes it weakly, preserves the fluorescence, but then kills the electron microscope ultrastructure. So there really wasn't a good solution for this, but one option is to do cryofixation. And we found that actually it works well in both cases. So just to briefly explain, if we prepare a sample for electron microscopy, one way to prepare the sample is to first grow it on coverslip and then freeze it really fast. You vitrify it. And then you turn it into plastic, you pull the water out, and you metal stain it, you cut it up, and you put it in this FIB SEM, and you get this picture. Now, while it's frozen, there's an opportunity. Do these fluorescent molecules, if they're there, would they still glow? Would they blink? We tested it, and sure enough, they do. So, this whole technique of super-resolution with the PALM approach, you know, is going to work. And so that simply means adding one extra little step right in there where we just do the fluorescence microscopy at low temperature. And we go a little bit above helium temperature, about 10 degrees Kelvin. And now you have two images and they can be merged and that forms the correlative microscope image. To be honest here, we're not really getting 4-nanometer resolution on the optic side. We get that maybe with the electrons. So really the actual optical resolution is a little bit higher up there. So here's some example images. The yellow is this endoplasmic reticulum network. The blue are the mitochondria in fluorescence. And then here they are in the EM. So you can see right, let me see, right here, here's the mitochondria, the outside of it. Over here too. And if you overlay it, it goes nicely over it. And even some of the finer details, you know, little varicosities in the ER can be seen. Now, even with a simple sample, which was really meant as a proof of principle, you can begin to see the power of it. So in one case here, we look at area and we don't really see endoplasmic reticulum, but we see like one little spot seeming to float in there of ER protein, and over here we see another little spot of the mitochondrial protein. If you look at the corresponding image, and I'm not sure if you can see it from the back, but there is a very tiny little vesicle, tiny little bubble right there, otherwise indistinguishable from this little object right here in the middle there. But they are distinguishable because one is a carrier of one protein derived from mitochondria and the other's dried from the ER. So now there are— cells have lots of little bags, lipid bilayer bags of different components, and now one has a chance to go beyond just the limitations of EM and really tag all of that. Proxysomes are another rather nondescript bag that might be existing throughout the cell, and we can easily identify it. Here you can see what it looks like in EM. This is this almost circular object with a little hole in the middle. It contains this protein associated with proxisomes, lies right on top. And here are just a few examples. They can sort of run through. Down here you can again see it here. And now, since you did this for the whole cell, you can say, okay, what's the distribution of sizes and shapes of proxysomes, and just catalog and get sort of comprehensive statistics of what a cell really looks like. And here's yet another example. These are two cells. These are two neural cells which are touching each other right along a boundary right here. Here's the EM version, and here's the combined correlative version right over here. And so this is a membrane protein which is known to sort of exist and particularly might form junctions between the nerve cells. And what's the structure? Is it sort of distributed uniformly, or what does it do? Let's look very carefully at that membrane. There's a cell here on the right and on the left, and in between here you see a dark line. It's sort of like dark, a little bit squiggly, dark, and a little bit squiggly again, and dark again. And you can just by hand sort of segment that out with a pencil and just distinguish the two different regions. Now you can also label it with a protein, and you see that this particular protein resides in the dark regions. And it's associated with a certain kind of binding between these cells. If you look at it, this surface in a full cross-section, not just in one cross-sectional plane like that, you see it has these little squiggly regions, solid regions in a pattern where the EM matches the optics just as before. So, we can really see the finer distribution of these proteins throughout a cell. So, here's just one final video which shows, you know, these cells. These are two cells. This is seen image as rendered by the electron microscope. And in the detail, this is the bottom surface. A lot of nuclear pores that you can sort of see here at the edge. Mitochondria everywhere. And we're just seeing this in its entirety. And again, you can see the overlay and the quality of the overlay across different regions. And I'll zoom in again in just one little area. There you can see that little vesicle that's similar to the one I mentioned earlier. You put the overlay of the ER and it sort of matches very nicely. Yeah, a lot of detail. Another example will sort of show up, I think, showing the details of a mitochondria. Or another vesicle, I think, in the middle. Oh, yeah, mitochondria down there. And again, a cross-section of it in the two different imaging modalities. So right now, this is a very general tool. It can be applied to, you know, a number of different biological systems. And, you know, I think just has, you know, great, great promise for the future. So here's one final example, just something we were playing with recently. On the top, you're seeing a killer T cell. Let me just sort of play it out again. And on the other side, it's a cancer cell. And the red parts, those are lytic granules that the killer T cell has within it. If it smells a target, It then moves all of them forward and is ready to deploy that against the target to kill it. So if we have a bunch of these crawling around on the coverslip and then we flash freeze them quickly, as we do right here, we might be able to catch this in action. And this is actually one of the things we did. Here's a coverslip, and you can see lots of blue and green ones. And here is a killer T cell on the bottom. And there's a cancer cell there on top. And now we can go in there with the extra resolution of the electron microscope and really see what's going on. Here you can see the two different nuclei have different contrast. The really dark little spots that you see down in here, those are the lytic granules. And so what happens is the killer T cell forms something called an immunological synapse against the cancer cell and tries to inject these lytic granules. And you can almost just see the damage that's happening in place. The nuclear envelopes retracting back here. The mitochondria are turning into black crisp things. They're dying. In here, there's basically a battleground going on where the one cell is trying to throw membranes over what's happening and prevent it from entering, and while the T cell is still trying to push more of these lytic granules into the cell. So this is not correlative yet. I mean, there, you can just easily imagine experiments to really understand what is this molecular warfare going on in this area with all the tools that are available. I think there's just an amazing amount still to be done. So at a certain level, I think you could say we're still back and share the same basic questions that we've had maybe about hundreds of years ago, like what are the details of how life really works and where can this go to help, you know, maybe help, and maybe even in trying to understand the mind, you know, we have new tools maybe to understand, you know, how that works too. But it's an amazing challenge and I think we'll see a lot happening in the next few years. It's a wonderful time to be having fun with microscopes and in biology. Thanks for your attention. Thank you.

Thank you. So we have time for some questions. The procedure is that there are microphone runners with microphones that have a foam top like this one with with different colors, and I'll call on you in the order of those colors. So we have a black, a red, and a pink. We'll start out with the birthday boy. He has a black microphone, and given that he has a cane, we will allow him to not follow the procedure of standing up, telling us his name, and asking a question, not giving a speech. So Carl Merrill. He is a member and he's going to ask a question, but he doesn't have to stand up. Thank you.

I'm Carl Merrill, a member.

I know that where you're at is a hotbed for research with the flies and genetic research. So the question is, I mean, once you start doing genetic modification and looking at what areas of brain are affected, that must be just spectacular.

So that must have already been begun or— Yeah. I think as far as actual experiments, I think they're not that much. I mean, some level of experiments have been begun using maybe other techniques. Right now we're at a stage where we're just making this kind of technology available. People can easily think of getting different mutant flies, looking at different compartments, different growth or background environmental genetic conditions to really see differences.

But it just gives you a taste of what's going to happen.

It's going to be totally revolutionary.

Yes, thank you. Pink microphone. Pink microphone.

My name is Scott Matthews and I'm a member. Um, first let me just say this work is absolutely mind-boggling, and the beauty of some of the images that you showed is just really staggering. And if you would permit me two questions, My first question is, when you image in two different modalities, how do you register the images from two different modalities? And then my other question is, for the palm, when you're blinking these molecules on and off, it seems to me that two adjacent molecules have to be blinking sort of out of phase or incoherently with one another. How do you manage that? Okay. Yeah, both good questions. The registration is— there's an extra part I left off. When we look at it optically, it's frozen in vitrified material. It's good to use reference marks, fiducials, to sort of connect the one image to the other. So you could use some of these structures like a mitochondria. The edge of a mitochondria, and there are lots of them throughout the image, and then try to match them with similar structures on the EM image. And there's a little bit of distortion that unfortunately happens when you go to this EM sample preparation step. So you do need to do a finer mesh registration beyond just sort of basic translation and magnification. But I think we have It can be done just by having an extra channel devoted to registration on the fluorescent side. For the palm, the blinking, the idea is that the blinking has to be very short on and long off. This is a statistical process. And it's stochastic. So, you know, 1 in 100 might turn on, it glows, and then it bleaches and goes off. And then we do a little activation, and 1 in 100 of some remaining molecules might catch it. It could be the same one, but a very small chance. And so, you have to operate in a mode exactly as you say, the molecule should should for the most part be far apart so they don't interfere with each other. The photophysical properties of the molecule blinking is a key parameter in deciding whether this is the right floor for it to use or not.

Red then black.

Did you consider doing it on a maggot, just the same species, same, you know, looking at the brain of a maggot, different developmental states? Yes, yes. And in fact, the early parts of this work really were on larva, on maggots, you know, at different stages after hatching. They're quite a bit smaller, much easier to image, and I did show briefly some one video of neural activity that was things were nerves were flashing on and off that actually was of a maggot. Not the imago. Pardon? Not the imago. Oh, imago, maybe I misunderstood. Imago is the adult. Oh, yeah, not the full, it was, yeah, it was the larva, right, right. Black.

I'm Federica de Rima.

Could you hold it up to your mouth so we can hear you?

I'm Frederica Deryma, and I'm a member. Thank you very much. Very nice presentation. Very nice tools. Thank you. I actually have a question related to the example of the maggot that you showed. You showed the ability to create a network of all the neurons in primitive, and then you showed the imaging of the dynamics of the— and actually the way depicted there was not a uniform kind of propagation. There was some kind of a collective excitation of the neurons.

Right.

So the next step would be to overlay that with some kind of a neural network or a Kirchhoff network, really, and model and see how the model predicts. And so it gives us a little bit more insight of what is the collective excitation. Is there anybody doing this?

That's a little bit the goal of this is to make that connection. Right. What that actually was, that was the brain, but it was in a dish and that was sort of fictive motion. It was trying to excite the muscular neurons and you can see the waves going down. And one can begin to play experiments exciting different parts, seeing the the response and the same sample, and that's key, the same sample can then be fixed, processed, and then we can try to identify, make a match between the optical image and the EM image and get the connective activities and the positive activity.

But, the next step would also be to overlay a model, right, of the electrical kind of circuit. And you can, and you have potentials, you can in a sense create a multi-objective optimization kind of F-based understanding of the system there.

Right, right. Exactly. That would be the end goal.

Could we save the discussion for the social hour and let's have some other questions? Thanks. Tim, Fred. Timothy Thomas, I'm a member. In your analogy of the car and driver to illustrate the connection between the modules, your driver module, are you implying that there's a module that tells the fly what to do and where to go?

Oh, its soul essentially or?

Just what the fly's doing or is it all just emerging from the case?

I think it's sort of, you could say it's emergent. I don't want to sort of claim full accuracy that there is going to be a driver. Necessarily.

You don't have a driver module.

You didn't discover one. Not explicitly. But there will be maybe some higher-order circuitry that might be based on memory or some other stored stuff and not an immediate sensory-to-motor output. And that has more layers of neurons in depth. So, a lot could be inferred into a driver module, but it's just sort of a covering description of it. Pink microphone.

Bob?

Oops.

Bob Hershey.

I'm a member. What kinds of changes do you expect to see with development of a brain or with a disease? Oh, gosh. With development, I think that's a very interesting area too. Nerves are growing and they have— and it's just amazing properties to see how does a nerve grow, form that pattern to go so many layers up, turn right, branch so far, make another turn to the left and branch over here and touch that nerve. That is a field in itself to trying to understand that.. And, you know, and will require, you know, when this goes into a pupa stage, you know, things are happening there. Is activity part of the development and synapsing? All I can say, these are just all good open questions. And then, of course, once you understand what it's doing when it works, then obviously disease is a question of what's going wrong. And maybe it for example, sleep or addiction or other things like that, there are completely some cases, some failures going on. Black microphone.

John Merrill, son of the birthday boy. So my question is, I applaud the excruciating detail, the application of the semiconductor to biological systems. It's fantastic. But you did mention with one of the other questions that you sort of have to register the data, you know, to be able to create these things. My question is, you know, there's functional MR, you know, where you could potentially take— is there functional MR of flies? Could you do that? Is that even done? And then my other part to that question, because you could overlay it even though it's not exact to kind of address some of the questions that have come up. And then the other question around that is, Do you see application of some of the techniques to something that's totally different like MRI to get high resolution?

Mm-hmm.

Okay.

Yeah, that's— that would be an interesting challenge. And we're trying in that case to go from a very microscopic view on rather small things to something that works on a scale which is about 2 orders of magnitude larger. Right now, I'd say that's That's a little bit of a stretch goal, trying to get either push the MRI down that far or push the circuitry resolution up that far is maybe not now, but maybe in a decade, one begin to think of it. Or an alternative to MRI to see activity or what is really being used here is trying to do some calcium imaging or neural activity imaging. Imaging, and that can sort of be the, you know, the alternative replacement to see, you know, nerve activity and map that on in this context, and it'll be compatible with the microscopes.

Magenta microphone.

Robert Thompson, I am a member. I just have a—

well, thank you for your talk, it was very interesting. I just have a sort of general microscopy question. The Compton wavelength of a muon is about 200 times smaller than the wavelength of the electron.

I'm just wondering if anybody has tried to build a muon microscope and what the challenges might be for that. I remember when I was a student, people were trying to do proton microscopes and imaging. I think the— Yeah, I mean, you have questions of sources, obviously, which would be equivalent to getting it from a needle tip. I think it's interesting to think about. It's probably have better penetration than protons. People are doing helium atom ion microscopes. It's very surface sensitive. They're cool things to contemplate, but I think the first thing is figure out, okay, can one make a nice, you know, source and be efficient in, you know, and sort of harness that, and also with enough flux to be able to, you know, to get the data, you know, like, like this.

Any more questions?

If not, let's thank our—

oh, Will Angel.

I am a member. I guess two brief questions. Sure. One, what is the rough cost of getting a sample now that you've got these machines? And then two, when inventing Palm, did you have— did you run into issues by doing this in your home that you would not have encountered in a lab? Like, did the carpet get in the way? Oh, okay. Rough cost, I'm not sure I understood the cost. This is the cost of the samples The machines? Oh, the samples. It's largely labor, you know, some student or, you know, expert researcher, you know, putting time and effort into harvesting the fly. I— to be honest, the flies that we have here, we have one expert who's been doing this for about 10 years and has perfected the staining in a level that I think makes— it gives us a little bit of a unique capability. Extra contrast pays huge dividends here. And let's see. Yeah, so sample preparation. Sorry, the other question was? The carpet. Oh.

Oh, carbon.

How do you do it at home? Oh, yeah. To be honest, we didn't get too far in doing it at home. We basically built the apparatus. Put it in a suitcase, flew over to NIH to somebody who knew what biology was. And they had cells. We didn't know what a cell was, barely. And they knew how to transfect, do the fluorescent labeling. And I'm fortunate, you know, at Janelia too, to sort of be able to sit back on the biological side and let the biologists do their biology.

Okay, well, let's thank our speaker. Before you go, we have two small gifts for you in appreciation for giving this talk. One is a framed copy of the announcement of your talk, and the other is a signed copy of Volume 1 of the Bulletin of the Philosophical Society of Washington covering the period from March 1871 to June 1874, which includes the reasons it was founded, who the original founders were, why they foolishly called it the Philosophical Society of Washington. No, I'm just kidding. And there are a number of interesting summaries of presentations that were made at the early meetings, including a presentation on how to calculate the value of pi to 31 decimal places.

Well, thank you very much.

Before we go, we have a few housekeeping announcements, if my microphone holds up. So as you know, The dues payments were due. If you're a member and you haven't paid your dues yet, please do so.

And if you haven't—

if you have paid your dues, or even if you haven't paid your dues and you're going to, please consider giving us an additional donation or sponsoring a lecture. If you want to join, it's pretty easy to do.

You just go to the website.

On the home page and look up in the upper right-hand corner, and there is a blue button. And when you hit on that hyperlink, it brings up the membership page to tell you some of the advantages of membership and gives you a hyperlink to the application, which will pull up the application form. And you fill it out and fill in all the required fees and press the submit button, and it will ask you to pay your dues. And once you've done that, it will be forwarded to the committee to consider, and you should hear back in a week. We have one very important requirement, and that is that you have a genuine interest in science. PSW is a nonprofit educational organization, tax-exempt under Section 501(c)(3) of the Internal Revenue Code, and contributions are tax deductible to the extent permitted by law and the IRS. All PSW members in good standing may wear the PSW rosette. If you wish to purchase one, $15 plus tax, you may do so at the rosette table at the back of the auditorium, and McQueen will be happy to help you with that. If you have any questions about membership, please see our membership director Lloyd Mitchell or corresponding secretary Robin Taylor. Who's hiding behind the audio rack, or Anne McQueen. Our next lecture will be the 2418th meeting of the Society. It will take place on January 10th, 2019 here in the Powell Auditorium. And we don't know who the speaker will be just yet. Ordinarily, the first lecture of the year is the president's lecture, but not this year. Instead, the lecture, the president's lecture will be on February 21st. On January 21st, we'll have a lecture on Venus by Ellen Stofan, the Mars director of the Smithsonian's National Air and Space Museum. On February 7th, we will have a lecture by Jack Gilbert, who is a professor in the Department of Pediatrics and at the Scripps Institution of Oceanography.. And he'll be talking on the World Microbiome Project and integrated microbiomics. As I mentioned before, the President's Lecture will be on February 21st, 2020. And the speaker will be Joachim Frank, who is Professor of Biochemistry and Molecular Biophysics and Distinguished Professor of the State University of New York at Albany. Frank is the inventor and developer of single-particle cryo-electron microscopy and methods for cryo-EM determination of molecular structures at atomic resolution, which has been a revolution in the determination of large protein complexes at atomic resolution. He was awarded the 2017 Nobel Prize in Chemistry for this work together with Jacques Dubochet and Richard Henderson. And he will be speaking on single particle cryo-electron microscopy. On March 6th, we have a talk by Rajesh Rao, who is the Yoon Wang Professor and Director of the Neural Systems Laboratory in the Department of Electrical and Computer Engineering at the University of Washington, Seattle. He is an expert on brain-computer interfaces. We're not talking GUIs here. These are direct interfaces between brains and computers, and that will be the subject of his talk. On March 20th, we'll have a talk on robotics by Henrik Christensen, who is the director of the Contextual Robotics Institute at the University of California, San Diego. And I'll mention one other of the dates, which is May 15th, which I think is going to be the Henry. It's our annual Joseph Henry lecture. And the speaker will be Shep Doeleman, who is the lead scientist on the Event Horizon Telescope Consortium, and not too long ago reported the historically significant first direct image of a black hole.

And with that, let me say that PSW is run entirely by volunteers, and I'd like to thank tonight's volunteers: James Heelan, who read the minutes, and McQueen, who took care of the live chat and rosettes, Robin Taylor, who did the video setup and will be working on the breakdown, took care of the live stream, and served as video director, Jared McQueen and Connor Nixon who ran our cameras, Cameo Lance who did the room management, Noel Block, our best boy, our mic runners, and our post-lecture wrangler Lloyd Mitchell. Please join me in thanking them all for their work. The social hour ends promptly at 10:30, and I will now accept a motion for adjournment of the 2,417th meeting of the Society to the social hour. Do I have a second? All members in favor?

All members opposed?

Oh, I guess they're unanimously approved minus one, and the meeting is adjourned.