• Skip to primary navigation
  • Skip to main content
  • Skip to primary sidebar
  • Skip to footer

iBiology

Bringing the World's Best Biology to You

  • Start Here
    • Explore iBiology
    • All Playlists
    • All Speakers
    • All Talks
    • What’s new at iBiology?
  • Research Talks
    • Talks by Topic
      • Biochemistry
      • Bioengineering
      • Biophysics
      • Cell Biology
      • Development and Stem Cells
      • Ecology
      • Evolution
      • Genetics and Gene Regulation
      • Human Disease
      • Immunology
      • Microbiology
      • Neuroscience
      • Plant Biology
      • Techniques
      • Archive
    • Talks by Series
      • Bench to Bedside
      • Famous Discoveries
      • Great Unanswered Questions
      • Microscopy Series
      • Share Your Research Series
  • Stories
    • Background to Breakthrough
    • Interviews and Profiles
    • Science and Society
  • Courses
  • Careers
    • Professional Development Talks
    • Professional Development Courses
    • Career Exploration
    • NRMN Resources
    • Biomedical Workforce
  • Educators
  • About
    • Mission
    • Commitment to Inclusion
    • iBiology Team
    • Board of Directors
    • iBiology Funders and Partnerships
    • Contact Us
  • Donate
Home » Courses » Microscopy Series » Specialized Microscopy

Array Tomography

  • Duration: 38:55
  • Downloads
    • Hi-Res
    • Low-Res
  • Subtitles
    • English
  • Transcript

00:00:11.20 Hello, my name is Stephen Smith. I am a neuroscientist
00:00:16.07 and professor of physiology at Stanford University.
00:00:19.19 And I am very happy to have this chance to talk to you
00:00:24.03 today about array tomography. Array tomography is
00:00:29.12 a new microscopy method developed in my laboratory
00:00:33.12 initially for neuroscience applications. Lately we've been seeing
00:00:39.00 that it may be useful in other areas in cell and tissue
00:00:42.10 biology. But in this lecture, we will concentrate specifically
00:00:46.10 on neuroscience applications from my own Stanford
00:00:50.10 laboratory. So what is array tomography?
00:00:54.05 It's a microscopy method that allows for combinations
00:00:58.15 of fluorescence microscopy and electron microscopy.
00:01:02.21 In my lecture today, I'm going to talk mainly about the
00:01:06.09 fluorescence microscopy forms of array tomography,
00:01:09.12 but I'll give you a little taste of the electron microscopy
00:01:13.09 capabilities, and the way that the fluorescence and the electron
00:01:16.03 microscopy can be used together. In both forms,
00:01:20.18 array tomography is a volumetric imaging method.
00:01:24.09 That is, it lets you image a 3-dimensional volume in
00:01:27.26 a specimen in the form of a digital image representing
00:01:32.11 three spatial dimensions. It is proteometric, maybe a new word
00:01:38.21 to some of you, the idea here is array tomography is capable
00:01:42.04 of measuring many different proteins within a tissue.
00:01:45.19 Not necessarily the whole entire proteome as the name
00:01:50.04 might be applied, but a good start at dozens of proteins.
00:01:54.03 Last but not least, it is an ultra-high-resolution method.
00:01:59.12 And I will illustrate more of what I mean by that as we
00:02:03.10 go along. So my lecture today is really going to be in three parts.
00:02:08.12 In the first part, I will give you a very brief description of
00:02:12.19 what array tomography is, how it's done, a very bare-bones
00:02:16.03 explanation. Then I'm going to run through a little gallery
00:02:20.22 of images from the neuroscience applications of array
00:02:25.00 tomography that we have pursued so far in my own
00:02:27.24 laboratory. And then in the final part of the lecture,
00:02:32.02 I'll come back to more of the details about exactly how
00:02:36.04 array tomography is done. So let's get started.
00:02:40.14 The basic form of array tomography is illustrated
00:02:46.14 in this little cyclical slide. There are a number of steps.
00:02:51.09 Array tomography is more of a process than it is something
00:02:55.08 you do with a particular unique tool. So the method is
00:03:00.21 essentially a histology. In other words, you have to begin
00:03:04.24 by fixing a piece of tissue. So this is not one of these wonderful
00:03:08.22 kinds of microscopy that you can use to look at the
00:03:11.13 dynamics of life. But it is a very wonderful complement
00:03:14.21 to those kinds of microscopy. As a matter of fact, we're using
00:03:17.29 in vivo imaging and array tomography as a histology
00:03:21.10 in combinations. So the first step, for better or worse, is
00:03:25.27 to fix the tissue. And I've indicated here very briefly
00:03:28.24 that the tissue is embedded in plastic, it is sliced,
00:03:32.24 mounted on a precision glass coverslip, stained,
00:03:36.17 imaged, stitched together by digital computation, and finally,
00:03:40.15 stored on a desk for later appreciation or analysis.
00:03:44.00 So this boils down into the three steps listed in the
00:03:48.24 center of this little cycle. First, these 3-dimensional
00:03:53.06 specimen are sliced into a 2-dimensional array of sections.
00:03:58.21 Then, all the actual staining and imaging is done on
00:04:03.13 those sections in the form of the 2D array. The acquired
00:04:09.21 2D images are then reconstructed into the form of a
00:04:13.07 3D image in the computer. Now, one of the really
00:04:20.29 special features of array tomography and the thing that
00:04:23.22 opens the door to proteometric imaging, as I highlighted
00:04:28.25 in my title slide, is that it is possible to reiterate a couple
00:04:35.07 of those steps in array tomography in a way that allows
00:04:38.15 for greatly increasingly multiplexing our multiplicity of proteins
00:04:42.23 that can be imaged in a given specimen.
00:04:45.24 That's indicated on this slide here. So the idea
00:04:49.06 is that there is a repetition of cycles of staining the array,
00:04:55.21 imaging it, wash the first round of stains off, repeat
00:05:00.04 the process. So this is one of the really key features
00:05:04.13 that has kept us going with the development with this
00:05:07.09 new imaging technology. And then last but not least,
00:05:12.06 certainly, a unique feature is that after fluorescence imaging,
00:05:16.16 as indicated in the previous two slides, it's possible to
00:05:20.26 restain the specimen for electron microscopy and to
00:05:24.10 image it with a field emission scanning electron microscope.
00:05:28.00 Which allows for perfectly correlated in the conjugate
00:05:32.23 electron microscopic images to go along with the proteomic
00:05:37.29 fluorescence images. There's a little inset here from which you may be
00:05:42.18 able to glean the fact that scanning electron microscope
00:05:47.15 images of array sections look very much like the
00:05:52.06 traditional transmission electron microscopes of
00:05:55.12 ultra thin sections that you may be used to from many
00:05:59.03 other areas of cell biology that relied heavily on over the decades
00:06:02.15 on electron microscopy. Okay, so that is the outline
00:06:08.15 of how array tomography is carried out. Now I want to
00:06:13.28 give you a little gallery of examples to further elicit
00:06:19.08 your interest for the long haul of going through the
00:06:22.12 details of -- the important details of how array tomography
00:06:27.04 is done. So these examples will all be drawn from research
00:06:31.09 in my Stanford University laboratory. And there's just a little
00:06:34.23 listing of some of the topics that we work on here.
00:06:38.02 For the largest context, we're interested in exploring
00:06:42.09 and better understanding the molecular architecture
00:06:45.14 of neural circuits. Within that area, we're extremely
00:06:49.11 interested in the molecular diversity of the myriad synapses
00:06:53.19 that comprise the major signaling organelles of the brain.
00:06:58.20 We're interested in mechanisms of synaptic plasticity,
00:07:02.20 and last but not least, we're interested in better
00:07:05.23 understanding and hopefully helping to develop new
00:07:08.26 therapies for neurodegenerative diseases. And all of these
00:07:12.22 project areas today, array tomography figures prominently.
00:07:16.19 So unfortunately, I don't have time today to talk much about
00:07:20.19 the logic or the meaning of any of these projects, but
00:07:23.10 I do want to show you some examples of the array
00:07:26.02 tomographic images that we have produced.
00:07:29.21 Okay, we're going to start out with a very low magnification
00:07:32.27 array tomography view of mouse cortex. What I'd like to do
00:07:38.19 now is zoom in successively to show you how much
00:07:43.08 detail it is possible to capture with array tomography.
00:07:47.00 And in this case, we will be looking at a single fluorescence
00:07:50.14 signal. Namely the fluorescence of YFP that is expressed
00:07:55.06 in a small subset of neurons, in particular, transgenic
00:07:58.16 mouse that this image was made from. So here's
00:08:02.24 a section representing all six layers of the cortex, and
00:08:06.06 then some. The little magenta rectangle shows the
00:08:10.05 area we'll zoom into on the next slide. There it is, and now
00:08:14.10 you can start to see clearly the individual neurons. We're
00:08:17.25 in layer five now, layer 5A at the top, 5B at the bottom.
00:08:21.14 You can, I hope, see the dendrites, the spines on the
00:08:27.23 branches. And if you get up real close, you might see
00:08:30.20 some beautiful little axons threading through that volume.
00:08:35.03 Now we're going to zoom in even further within the area
00:08:38.22 of this magenta rectangle. And now I've prepared a little
00:08:42.29 animation, so you can better appreciate the 3-dimensional
00:08:46.04 nature of the image we are rendering. And so we're going to
00:08:51.14 revolve it through 360 degrees, but I'll probably get
00:08:57.12 bored and move onto the next slide before we get all the way
00:08:59.28 around. But, I hope you can see very clearly here that this
00:09:05.02 is a very detailed 3-dimensional rendering and perhaps
00:09:09.15 you're already beginning to appreciate the really superlative
00:09:12.21 resolution in detail that array tomography is capable
00:09:17.04 of delivering. And keep in mind, the level of detail that
00:09:20.16 we see in this zoom in is present in that entire slab
00:09:23.20 representing the entirety of the cortical layer structure.
00:09:27.09 Okay, so let's look at some other pictures. Here's
00:09:33.05 similar volume, but now in addition to fluorescence
00:09:37.05 signals derived from the YFP, we've added fluorescence
00:09:41.02 from the microtubule bundles present in the axons and
00:09:46.03 dendrites of non-expressing neurons. And we've
00:09:49.24 filled in all of the synapses. The synapses have been
00:09:54.18 colored in randomly, just to make the point that array
00:09:59.09 tomography is capable of fully resolving the individual
00:10:02.21 synapses in three dimensions. And that capability
00:10:05.23 is really a new one that was not shared by previous
00:10:08.23 fluorescence microscopy methods. And this one is really the
00:10:12.01 core of our research program on synapse molecular
00:10:16.18 architecture and synapse diversity. Now we're going to look at
00:10:21.17 a set of images that represent the ability of array
00:10:27.02 tomography to capture lots of distinct protein channels, so I'm not
00:10:31.19 going to really explain what's going on here. There are
00:10:34.23 titles on this slide with the names of the protein
00:10:38.00 antigens to which the antibodies that were used in this
00:10:41.07 experiment are directed. I'm going to flip through the slides
00:10:45.12 and what I want you to grasp is that these four slides
00:10:50.27 represent the exact same volumes. If I flip back and forth,
00:10:56.02 you can kind of see that those DAPI nuclei are sitting
00:10:59.22 in one place. And now we're beginning to get a sense
00:11:04.12 of what I mean when I say the molecular architecture
00:11:08.08 of the brain. Maybe in your spare time, you can take a look
00:11:12.21 at these and think about what it might all mean.
00:11:15.28 We think it means quite a lot, but we're quite a ways
00:11:19.16 from understanding it. I wish I had time to talk about
00:11:23.09 this work a lot more now, but I don't.
00:11:26.03 Now here's another slide that represents another
00:11:29.16 aspect. I'll step out of the picture here so you can
00:11:32.16 see the whole thing. These are synaptic markers,
00:11:36.11 in a piece of neuropil, similar to that that's been illustrated
00:11:41.04 in the previous slides. And these are outtakes from our
00:11:44.29 work aimed at better understanding the molecular
00:11:48.14 diversity of synapses. So here we see markers that indicate
00:11:51.26 that some of these synapses are glutamatergic
00:11:55.06 excitatory, some are GABAergic inhibitory, some
00:11:58.14 have neuroligin2, which is a particular molecule
00:12:02.28 associated with a subset of inhibitory neurons, and
00:12:06.02 so on. And again, I'm sorry I don't have time to go into
00:12:09.08 the very exciting science related to this slide, but this is just
00:12:13.18 to give you a better sense of what array tomography
00:12:16.25 data looks like. Here is a way of representing the high
00:12:23.04 dimensional proteometric data that may not be so
00:12:26.18 pretty. It's certainly not so colorful, but we use this
00:12:29.24 for the more careful analysis in my laboratory. So this
00:12:33.15 is what we'd call a synaptogram. There are nine columns
00:12:38.18 and something like 20 rows. All of these represent the same
00:12:43.27 1 cubic micron, so there's a number of serial sections
00:12:47.09 lined up. We have described synaptograms in several
00:12:52.00 publications. I hope you'll find them and enjoy them, I don't
00:12:55.13 have time to talk about them now. But I'm showing this
00:12:58.02 slide to indicate that it is both necessary and important
00:13:05.08 to go beyond staring at millions of beautiful colors
00:13:09.08 when you're trying to understand the messages behind
00:13:14.05 array tomography. So this is a bit of a more analytical
00:13:16.13 approach, just a taste. Based on that approach,
00:13:20.08 we have developed machine learning tools that can
00:13:24.10 analyze raw data that is extraordinarily complex, like what you're
00:13:29.06 seeing here. This is just four fluorescence channels, but
00:13:33.02 this is a subset of an experiment that has about 20
00:13:36.13 channels. We use the analysis, like the synaptogram
00:13:42.22 in that slide, to craft machine learning tools that are
00:13:47.12 capable of plowing through large data objects of
00:13:51.20 this type and analyzing them, classifying them into
00:13:56.01 subsets of synapses. So here's a classification into two
00:13:59.01 subsets, rendered in two different colors. Again, just
00:14:03.08 as an example. So here's the results of an analysis
00:14:07.05 based on a classification scheme like the one I just
00:14:10.03 described. And this simply reports numbers reflecting
00:14:14.19 the densities of different types of synapses at different
00:14:18.02 depths within the cortex. These curves are actually based
00:14:23.18 on the analysis of a million or so synapses. So it's a good
00:14:28.26 thing we have computer algorithms to help us go through
00:14:33.11 and count and classify and analyze all of those synapses.
00:14:36.10 Here is a little bigger version of an electron micrograph,
00:14:44.21 made on a tomography array specimen. So I indicated
00:14:48.03 earlier on in my brief once-over that it was possible to
00:14:51.15 obtain electron microscopic views of our tissue and
00:14:55.24 in this case, mouse cortex. And maybe in this larger view,
00:15:00.13 you can better appreciate how really nice these micrographs
00:15:05.06 are. This can pass as a pretty good transmission EM,
00:15:09.27 even though it's made in a completely different way.
00:15:12.25 And in the last part of my talk, I'll come back to how we can
00:15:17.07 make this particularly potent by combining it with the
00:15:20.09 fluorescence imaging. And as a final example in our
00:15:25.10 little stroll through the array tomography gallery,
00:15:28.07 here is a rendering, a really beautiful rendering
00:15:32.06 of a really ugly thing. This is an image in 6 colors of
00:15:39.05 Alzheimer's disease plaque that was imaged from
00:15:44.11 an autopsy specimen from a human suffering from
00:15:48.20 Alzheimer's disease. So this is a little icon of our
00:15:51.13 neurodegenerative disease program.
00:15:55.09 Okay, we're going to go through each step, one at a
00:15:57.26 time now. The first step is a chemical fixation and solvent
00:16:04.04 replacement step. Where the idea is to preserve your
00:16:09.10 piece of tissue with all of the biomolecules right where they
00:16:12.23 started out, but to replace all of the water with an
00:16:16.03 organic solvent compatible with the next step that we'll
00:16:19.23 come to in just a second. And the procedures here are
00:16:23.13 really nothing new. These are the procedures that have been
00:16:26.21 developed by electron microscopists over the last few decades
00:16:30.01 to perfect the arts of fixation and embedment for electron
00:16:34.00 microscopy. Next we come to the embedment step, and here
00:16:39.04 one of the special tricks in array tomography is to use
00:16:44.23 one particular class of resin for embedding. And that would be
00:16:49.18 an acrylic resin. Acrylic resins are better for post-embedding
00:16:53.04 immunostaining, which is the lifeblood of fluorescence
00:16:56.14 array tomography. So there's a number of acrylic resins
00:17:00.07 that can be used. The point is this is not the epoxy
00:17:03.19 resin that is used traditionally for electron microscopy.
00:17:06.19 Okay, after infiltrating the resin into the dehydrated
00:17:11.14 specimen, the resin is polymerized and now you've got your
00:17:16.10 specimen in a solid block of plastic with ideally all
00:17:19.26 of the biomolecules right where they were in that tissue
00:17:22.29 in its last moment of life. But now, the tissue is embedded
00:17:26.25 in a very hard piece of plastic. That piece of plastic can then
00:17:31.14 be trimmed and mounted and prepared for sectioning on
00:17:35.11 an ultramicrotome. So perhaps you've seen an ultramicrotome
00:17:40.15 around. There's one in the picture in the center of this slide,
00:17:43.23 along with our wonderful technician, Nafisa, who is our expert
00:17:48.22 at working this particular tool. It is a tool that electron
00:17:52.23 microscopists have been evolving for over half
00:17:56.05 a century and is really quite spectacularly refined
00:18:00.26 and capable in its present form. It is capable of cutting
00:18:05.24 a block of tissue embedded in resin in sections that are
00:18:12.06 as thin as 40 or even 30 nanometers. These are very
00:18:18.11 very thin sections. And that's really where a lot of the magic in
00:18:22.26 array tomography comes from. Okay, digging in a little
00:18:26.27 deeper, zooming in a bit. This is the diamond knife
00:18:31.15 boat, as it's called, mounted on the stage of the ultramicrotome.
00:18:35.18 The tissue itself is right where my fingertip is pointing, and maybe
00:18:44.16 you can catch the glint of transparent plastic there.
00:18:48.04 The blue item is a boat full of water, the water is
00:18:53.09 actually critical. And at the point where the boat meets the
00:18:56.17 specimen, there is a diamond knife blade. This is a very, very
00:19:00.21 sharp knife. This is a big piece of the trick of cutting
00:19:03.10 tissue as thin as tens of nanometers. The next slide is a little
00:19:10.00 movie. So now we've zoomed in closer still, and you can see
00:19:13.27 the microtome in action, cutting off its thin sections.
00:19:19.18 The tissue block, the resin block, is that hemispherical looking
00:19:27.05 thing sitting in the steel chuck of the ultramicrotome.
00:19:30.11 You can see the glint of the diamond knife blade, and
00:19:33.19 then as the cutting action of the microtome moves
00:19:40.02 the resin in a very precise way over the diamond
00:19:43.17 knife, it is cutting one thin section of the specially
00:19:47.21 trimmed block at a time. And each new section is
00:19:53.05 neatly spliced onto the tail of the previous section to
00:19:58.06 produce a ribbon that extends along the surface of the water.
00:20:02.00 And if you look carefully, maybe you can see the incremental
00:20:06.04 advance of a ribbon on the surface of the water. It all depends on
00:20:11.15 how big the screen you're watching this on is.
00:20:14.22 What's actually going on there, here's a little cartoon
00:20:18.28 I've prepared. So the resin block is very precisely stroked down
00:20:27.01 over the very sharp diamond knife, peeling off the
00:20:31.05 ultra-thin section. The sections glue themselves together
00:20:36.04 to form that ribbon that was floating on the surface of
00:20:39.28 the water, because a thin layer of a tacky adhesive
00:20:43.21 has been applied to the top and bottom of the block.
00:20:46.20 So with each new stroke of the cutting action of the
00:20:50.17 microtome, the leading edge of the section about to be
00:20:53.26 cut is neatly spliced to the trailing edge of the preceding
00:20:59.11 section. So this is a rather exotic sounding trick, it's
00:21:04.21 actually been around for decades. And has been used
00:21:07.26 by many electron microscopists for serial sectioning.
00:21:11.22 We've kind of supped it up with better glues and so on.
00:21:15.11 And the big difference here is that we're going to make much
00:21:18.08 longer ribbons, whereas traditional electron microscopists
00:21:21.10 make ribbons that are 2mm long and put them on
00:21:24.18 tiny little EM grids, we do something a little different which
00:21:30.11 you can maybe get an idea from the next slide.
00:21:32.20 We grow those ribbons to be 40 or 50 mm long.
00:21:36.13 And pick them up on specially prepared precision
00:21:40.27 optical coverslips. We do this by reaching the coverslip
00:21:46.01 down into the water of that little blue boat you saw.
00:21:48.29 And the result is indicated in this cartoon. There's a coverslip
00:21:52.20 with a neat row of something on the order of 100 serial
00:21:57.06 sections sitting on that surface. And because of the treatment
00:22:01.14 of the coverslip surface and the drying process we use,
00:22:05.03 a permanent adhesive bond is formed between the
00:22:08.16 ribbons, all of the individual sections, and the coverslip.
00:22:12.16 So that is the array that gives its name to array tomography.
00:22:18.04 here is a close up of sorts of one of those little ribbons.
00:22:24.22 It's just a section of seven out of a hundred or so
00:22:28.05 serial sections on the ribbon, as you might seem them if you
00:22:32.29 popped that array under a low-powered light microscope.
00:22:37.06 Then, the array and its coverslip are mounted on some kind
00:22:43.08 of a slide carrier and coupled to some kind of an apparatus
00:22:47.05 for fluid staining of that coverslip surface. So you see
00:22:52.06 one such apparatus in the center of the field here.
00:22:56.08 And the idea is that we can now apply aqueous
00:23:00.17 fluid stains to the surface of the array, and typically
00:23:05.23 the stains we use are antibodies and they stain proteins
00:23:10.13 embedded in the array plastic. The same way you might
00:23:13.04 be used to studying them by staining proteins in a fixed
00:23:18.19 cell in a dish. In a typical experiment, we might use
00:23:23.25 three or four colors of antibodies in each cycle of
00:23:28.10 staining. So this would correspond to the standard 3-4
00:23:32.16 channel indirect immunofluorescence that you
00:23:35.01 might be familiar with from other lectures or from doing
00:23:38.16 it yourself. Then, after staining the array, we put it on the stage of
00:23:46.22 an automated fluorescence microscope with a digital
00:23:52.13 camera. The automation controls motors that move
00:23:56.26 the stage in X and Y and focus the microscope in Z, and
00:24:01.13 the graphics on this slide just indicate a couple of
00:24:05.03 the extremes of the range of possibilities we can use to
00:24:09.16 image the serial sections. In the simplest kind of experiment,
00:24:12.25 we might just image a given spot. One spot within a section
00:24:18.21 then, and try to find the same spot in the next section
00:24:21.29 and move on, and repeat that for each section for each
00:24:26.01 of the hundred or so sections. And in that way, acquire
00:24:31.01 serial section images throughout the ribbon. Another
00:24:35.18 possibility is to make a mosaic, as indicated in the
00:24:39.05 lower field here, where we tile together enough
00:24:43.27 individual high magnification microscope fields to
00:24:47.18 capture images of the entire section. And that's what was done
00:24:51.01 to acquire that large image that I showed you at the beginning
00:24:55.10 in our little walkthrough of the gallery, where we zoomed in and in and in.
00:24:58.21 We were able to do that because tiled together a very large
00:25:02.15 field of view using the infinite patience of the automated
00:25:07.10 motorized microscope to take in that case, about
00:25:10.25 1200 individual pictures while we slept and played and
00:25:15.15 had fun. Okay, so here is a picture in real time of
00:25:23.11 the microscope at work. You can't see much here. You can see
00:25:26.25 the flash of the fluorescence illumination during each of
00:25:31.07 those illumination events, the digital camera is snapping a
00:25:35.05 picture. If you watch carefully, you'll see there are four different
00:25:39.05 colors here. When it's done with all four colors at one spot,
00:25:43.14 it moves over a tiny bit to work on acquiring the next spot
00:25:47.28 according to one of the patterns I described in a previous slide.
00:25:51.29 Okay, now I'll remind you once more of really the special magic
00:25:57.26 of array tomography. The thing that expands its
00:26:01.11 multiplexing capability to a truly proteomic scale. And this is
00:26:05.28 the fact that we are, after we've imaged one set of
00:26:10.11 three or four immunostains, we can wash those antibodies
00:26:15.11 off. All it takes is a change of pH by 2 or 3 units, either up or
00:26:20.20 down, that disrupts the fit of the antibodies to their
00:26:23.20 antigen. They wash off, and then we restain the array.
00:26:27.26 If we restained it with the same four antibodies and
00:26:31.28 imaged it again, we'd get the same answer. We have proved
00:26:34.29 that to ourselves over and over. Now, the point is to
00:26:38.02 restain it with three or four different antibodies each time.
00:26:42.20 And that's how we get up to the high order multiplicity, the
00:26:45.12 10, 20, 50, is just around the corner, proteins per specimen.
00:26:50.28 Then, because the microscope doesn't do a perfect job
00:26:58.20 of photographing each serial section exactly the same
00:27:02.15 way, and because there's some dimensional warping and
00:27:07.24 distortion that occurs during the cutting process,
00:27:09.21 we finally need to use a computational registration process.
00:27:14.02 In the case of stitching together large mosaics, we also have to
00:27:17.16 computationally stitch the large mosaics together. The fact is
00:27:20.28 that algorithms for performing this work nearly flawlessly
00:27:26.16 and highly efficiently are already available in the public
00:27:30.07 domain. So, a lot of computer power goes into this, but
00:27:34.07 in general, not too much human effort. And the results
00:27:38.13 are quite nice. The result is a beautifully registered
00:27:43.05 stack of images that is indicated in my little cartoon in the
00:27:51.01 left here. And amounts to a 3-dimensional image, or an image
00:27:59.11 representing three spatial dimensions, even though it may have
00:28:02.15 20 dimensions of color channels. And then finally, those
00:28:06.03 images are stored on a hard drive. And as you might have
00:28:10.28 figured out by now, this tends to lead to what is coming to be
00:28:15.15 known as big data. Gigabytes, terabytes worth of data.
00:28:20.03 This is a good thing, really. If you're trying to characterize
00:28:25.10 a very intricate and elaborate tissue, such as brain,
00:28:29.10 but you've got the capability of acquiring large volumes
00:28:33.16 in many channels, and we like tiny voxels. That's part of the game
00:28:39.07 of high resolution. So you put those together in a little
00:28:42.12 formula, and you end up with big data. And this is
00:28:47.10 opening the door to some great new fun for bioinformaticians.
00:28:52.21 I gave you a little taste in our gallery walkthrough of some
00:28:56.15 of the bioinformatics analyses we're developing. But with
00:29:01.12 big data like this, you don't want to have to spend too much
00:29:04.04 time analyzing it by hand. Okay, so I've described
00:29:09.03 a procedure to you, I have claimed that it is extremely
00:29:14.11 powerful, and I'm meaning to imply that it is worth
00:29:19.24 all the trouble I've described. I've already indicated some
00:29:23.11 of its singular advantages in terms of the high order
00:29:26.20 multiplexing and so on. On the other hand, many of you have
00:29:32.21 worked with confocal microscopes. You're used to image stacks,
00:29:35.29 some of you may be very good at working with this kind of
00:29:40.22 data. And you may have been wondering how array
00:29:43.24 tomography compares to confocal microscopy. And
00:29:47.13 it's not a simple comparison. Confocal microscopy
00:29:50.00 can be used on living specimens, array tomography cannot.
00:29:54.22 So confocal microscopy has its areas of application.
00:30:00.15 But when it comes to high resolution and quantitative
00:30:04.22 interpretation, we need to ask questions like the one
00:30:09.17 I've indicated on this particular slide. So just as
00:30:12.04 one example. For quantitative interpretation of specimens
00:30:18.08 like synapses in the brain, it's important to have high
00:30:22.26 resolution on all three axes, and it's important to have
00:30:26.19 results that are of similar quantitative value regardless
00:30:33.13 of where they are within the specimen. Okay, so
00:30:37.03 here's a little comparison of the best we could do with
00:30:40.27 confocal microscopy in an average day at work with
00:30:47.03 array tomography. So in this particular comparison, we're
00:30:50.12 just looking at one color. This happens to be an immunostain
00:30:54.14 that stains all of the synapses in little pieces of mouse brain.
00:30:59.23 And what we've done here is image two very similar volumes.
00:31:04.15 One with array tomography, one with an excellent laser
00:31:08.05 scanning confocal microscope. And when we reconstruct
00:31:14.14 these volumes in their natural x-y plane, from the top we can see
00:31:18.28 that they're somewhat similar looking. But if we check out the
00:31:24.09 array tomography volume from the side, we can
00:31:29.00 see that there is still very crisp resolution along the
00:31:35.03 z-axis, along the focal axis. And we can see that the density
00:31:39.15 and character of the spots is uniform throughout the depth of
00:31:43.04 the specimen. So array tomography scores well on the important
00:31:48.28 criteria of having isotropic high resolution, that is high resolution
00:31:52.23 on all three axes. And for having a sensitivity that is
00:31:57.18 independent of depth within the specimen. So we'll give
00:32:00.28 array tomography a big yes. Now let's take a closer look
00:32:04.06 at the laser point confocal volume. And what you can see here
00:32:07.23 is that even though it looks pretty nice from the top,
00:32:10.15 it looks quite a bit different from the array tomography
00:32:14.29 when you look at it from the side in this volume rendering.
00:32:18.03 And you're just going to have to believe me that the array
00:32:20.09 tomography is the one that's giving us the right answer.
00:32:23.21 What you can see in the confocal volume, is for one thing,
00:32:26.23 all of the spots are stretched out drastically along the z
00:32:30.07 axis. That's because confocal microscopes, and really all
00:32:35.01 microscopes that work with focused cones of light,
00:32:37.11 have much worse resolution along the z-axis than they do
00:32:42.02 in their xy axes. You may have learned about that in one of
00:32:45.19 these lectures. So, the elongated shape of those particles
00:32:53.24 along the z-axis is an artifact. And the other really conspicuous
00:32:58.03 difference you can see is the apparent density
00:33:02.17 or the brightness of those synapses falls off drastically
00:33:07.00 within the 7 micron depth of that specimen. There's at least
00:33:11.15 two broad classes of reasons for that. One is that
00:33:15.05 with the confocal, the antibody stain has to be applied
00:33:19.00 and diffused through the depth of the specimen. And
00:33:22.26 even with detergent permeabilization techniques,
00:33:25.24 and even with multi-day incubation in antibodies, that
00:33:30.18 process itself, the staining process is highly depth dependent.
00:33:34.09 And then there are imaging losses in sensitivity due to
00:33:38.12 light scattering and spherical aberrations accumulate
00:33:43.03 rapidly with depth in the confocal microscope. And
00:33:46.08 these result in a loss in sensitivity and imaging deeper
00:33:50.00 parts of the specimen. So we're unfortunately going to have to
00:33:53.12 give the confocal a great big no on these important
00:33:57.09 criteria of imaging quality. Okay, so what accounts for
00:34:03.23 those advantages? These are both fluorescent images,
00:34:07.07 both made with fine high numerical aperture microscope
00:34:11.05 objectives of the appropriate kind. One major reason for
00:34:17.06 the greatly enhanced resolution of the confocal microscope
00:34:22.25 and greatly enhanced sensitivity for small objects,
00:34:26.12 is that all imaging is done right at the surface of the
00:34:29.23 coverslip. And high numerical aperture objectives
00:34:34.05 in general, are designed to do their best only when the
00:34:39.12 specimen is very, very close to the coverslip. And when you have
00:34:42.22 to focus down, even a few microns into an inhomogeneous
00:34:46.24 specimen, as you do with a confocal microscope, there's
00:34:49.25 a pretty disastrous accumulation of spherical aberrations
00:34:53.10 and wavefront distortions that greatly perturb the imaging
00:34:57.05 quality. So that's the first big advantage of array tomography.
00:35:00.14 next big advantage is that the physical sections where we cut
00:35:04.19 on the ultramicrotome, or Nafisa cuts on our ultramicrotome,
00:35:08.11 are very much thinner than the thinnest confocal optical
00:35:12.15 sections. Using practical matter, they are generally
00:35:15.16 less than 1/10 the thickness of the optical section.
00:35:19.06 Therefore, an immediate 10x boost in resolution along the
00:35:23.16 otherwise troublesome z-axis. With array tomography,
00:35:28.11 there are no out of focus photons. The specimen is so
00:35:33.03 thin that it lies entirely on the depth of focus on the microscope.
00:35:37.12 So there are no out of focus photons to eliminate with a
00:35:42.27 pinhole, or to add noise to the image. So these are
00:35:46.21 a major limitation to image quality, and they simply
00:35:50.07 don't exist with array tomography because of the
00:35:53.10 physical sectioning. And because no part of the
00:35:56.00 specimen is out of the focal zone of the microscope.
00:36:00.26 Part of the reason for the depth dependence of array
00:36:06.03 tomography is that depth on the 2-dimensional array
00:36:10.02 is virtual. That every aspect of stain-diffusion and the
00:36:14.15 imaging are completely depth independent, because
00:36:17.09 every volume element of the specimen is imaged exactly
00:36:20.27 the same way. It's right on the surface of the coverslip,
00:36:24.07 it is within nanometers of the stain, in contrast to the
00:36:28.02 situation working with the whole mount. And then
00:36:33.05 finally, the runaway advantage of array tomography is that
00:36:37.29 it permits that iterative cyclic multiplexing I've already
00:36:43.04 described. That strategy has been tried repeatedly with
00:36:48.10 confocal microscopes, with cells, with tissues, and it
00:36:51.17 doesn't work. You need the resin embedment to
00:36:55.01 stabilize the tissue enough to withstand the effects of
00:36:59.27 the antibody stripping wash. Last but not least,
00:37:08.12 I will point out and show a few preliminary examples
00:37:12.22 here of the friendliness of array tomography to super
00:37:16.20 resolution microscopy. So, I think you've probably
00:37:21.11 had access to other lectures about super resolution
00:37:24.00 microscopy methods like STORM and Mats Gustafsson's SIM,
00:37:29.07 and STED. And another one here, which is peculiar to
00:37:36.01 array tomography, and we like a lot, is the use of
00:37:39.07 the astronomus 2D deconvolution algorithm called
00:37:43.09 Richardson-Lucy. All of these methods allow very
00:37:47.29 substantial improvements in resolution beyond the
00:37:52.21 typical Abbe limit. And these work together with the extraordinarily
00:37:56.21 good axial resolution we obtain by the physical
00:38:01.09 sectioning process to achieve really high resolution.
00:38:07.05 I want to conclude by giving credit to the many people
00:38:12.04 in my laboratory at Stanford, and people that I have worked
00:38:15.17 with really all over the world, to develop this technology
00:38:20.00 and some of the results that I have shared with you today.
00:38:22.23 Please take a minute to read the names of the fine young
00:38:26.29 people whose names are listed on this slide.

This Talk
Speaker: Stephen Smith
Audience:
  • Researcher
Recorded: August 2012
More Talks in Microscopy Series
  • Mark Schnitzer
    Miniature Microscopes for Deep Tissue Imaging
  • Optogenetics (Karl Deisseroth
    Optogenetics: The Use of Opsin Genes in Technique Development
  • Daniel Fletcher Microscopy
    Mobile Microscopy
All Talks in Microscopy Series
Share

Talk Overview

Array tomography was developed in the Smith laboratory for neuroscience research. In this lecture, Smith talks mainly about the fluorescent microscopy forms of array tomography.

Questions

  1. In array tomography, the embedding step is modified compared to electron microscopy. Explain how.
  2. In array tomography, the sectioning step is key to the success of the technique. What is the product of the sectioning step?
  3. Comparing confocal microscopy to array tomography, which of the following statement(s) is true (select all that apply)?
    1. Confocal microscopy provides better isotropic resolution than array tomography
    2. Array tomography has uniform resolution regardless of the depth of the specimen
    3. Confocal microscopy provides better resolution on the z axis
    4. Confocal microscopy can be used to image live samples, whereas array tomography cannot
    5. None of the above
  4. Imagine you are planning an experiment involving imaging of the mouse cortex using in array tomography. Because you do not have access to an ultramicrotome, you are considering using a regular microtome, which would result in much thicker sections than an ultramicrotome. Using thicker sections in array tomography may… (select true statements):
    1. Impair the diffusion of antibodies inside the tissue
    2. Reduce background signal
    3. Decrease spherical distortions
    4. Result in lower isotropic resolution
    5. None of the above

Answers

View Answers
  1. Acrylic resin is used instead of epoxy resin
  2. using an ultramicrotome, the block is sectioned into ultrathin 30-40 nm tissue sections, in extra long ribbons of 100 sections (40-50 mm)
  3. B, D
    1. Incorrect: array tomography provides better isotropic resolution than confocal microscopy. Isotropic resolution refers to resolution in all three axes (x, y, z).
    2. Correct: Array tomography has uniform resolution regardless of the depth of the specimen because tissue sections are extremely thin.
    3. Incorrect: Array tomography provides better resolution on the z axis because tissue sections are extremely thin (this statement is a variation of b)
    4. Correct: Confocal microscopy can be used to image live samples, whereas array tomography cannot)
  4. A, D
    1. Correct: Thicker sections may impair the diffusion of antibodies inside the tissue
    2. Incorrect: Thicker sections may increase background signal by increasing out-of-focus light coming from deeper areas in the tissue
    3. Incorrect: Thicker sections may increase spherical distortions as one gets further from the coverslip surface
    4. Correct: Thicker sections may result in lower isotropic resolution, as the resolution on the z axis decreases with depth)

Speaker Bio

Stephen Smith

Stephen Smith

Dr. Smith is a Professor at Stanford University School of Medicine where his lab pioneers new methods to map connections in the brain.  Smith and his co-workers developed array tomography, a high throughput imaging method that allows visualization of the molecular architecture of the brain. Continue Reading

Playlist: Microscopy Series

  • High Throughput Microscopy (Jan Ellenberg)
    High Throughput Microscopy
  • Mark Schnitzer
    Miniature Microscopes for Deep Tissue Imaging
  • Optogenetics (Karl Deisseroth
    Optogenetics: The Use of Opsin Genes in Technique Development
  • Daniel Fletcher Microscopy
    Mobile Microscopy

Reader Interactions

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

Primary Sidebar

Sign up for the Science Communication Lab education newsletter

  • Sign up Newsletter
  • This field is for validation purposes and should be left unchanged.

Privacy Policy

Help us keep bringing the world’s best biology to you!

Footer

Funders

NSF
NIGMS
Lasker
Rita Allen

Start Here

  • Talks for Everyone
  • Talks for Students
  • Talks for Research
  • Talks for Educators

Explore

  • Explore
  • All Playlists
  • All Speakers
  • All Talks

Talks By Topic

  • Biochemistry
  • Bioengineering
  • Biophysics
  • Cell Biology
  • Development and Stem Cells
  • Ecology
  • Genetics and Gene Regulation
  • Human Disease
  • Immunology
  • Microbiology
  • Neuroscience
  • Plant Biology
  • Techniques

Talks by Series

  • Bench to Bedside
  • Famous Discoveries
  • Great Questions
  • Share Your Research Series

Career

  • Professional Development
  • Career Exploration
  • NRMN Resources
  • Biomedical Workforce

Courses

  • Microscopy Series
  • Short Microscopy Series
  • Open edX Courses
  • Cell Biology Flipped Course
  • Engineering Life Flipped Course
  • Evolution Flipped Course

Educator

  • Educator Registration
  • Educator Resources
  • Log In

About Us

  • About Us
  • iBiology Team
  • Wonder Collaborative
  • Contact Us
  • Mission
  • Privacy Policy
  • SCL Financial Conflict of Interest Policy

This material is based upon work supported by the National Science Foundation and the National Institute of General Medical Sciences under Grant No. 2122350 and 1 R25 GM139147. Any opinion, finding, conclusion, or recommendation expressed in these videos are solely those of the speakers and do not necessarily represent the views of the Science Communication Lab/iBiology, the National Science Foundation, the National Institutes of Health, or other Science Communication Lab funders.

© 2023 - 2006 iBiology · All content under CC BY-NC-ND 3.0 license · Privacy Policy · Terms of Use · Usage Policy
 

Power by iBiology