Dual-View Inverted Selective Plane Illumination (diSPIM)

00:00:11.18 Hi. My name is Hari Shroff and I'm an investigator
00:00:15.12 at the National Institute of Biomedical Imaging and Bioengineering at the NIH.
00:00:19.16 My lab designs microscopes and today I'd like
00:00:22.09 to tell you about a tool we developed for studying
00:00:24.26 life at high spatial, temporal resolution.
00:00:26.27 Before I get into the details of the microscope
00:00:29.10 I'd like to describe one of the problems
00:00:31.09 that we are using the microscope to study.
00:00:32.29 One of the main goals of my lab is to
00:00:35.26 uncover the basic rules that govern neurodevelopment.
00:00:39.03 So if we consider our own brains as a fantastically
00:00:41.15 complicated organ, we have hundreds of billions of
00:00:44.12 neurons and orders of magnitude more synapses
00:00:46.28 yet despite this large parts number, the parts come together
00:00:50.17 with remarkable precision.
00:00:51.22 How this is orchestrated in the vivo is not very well understood
00:00:55.02 but interestingly, although there are hundreds of billions of
00:00:58.03 neurons, there are only about 25000 genes that give
00:01:00.28 rise to this complexity.
00:01:02.08 This mismatch in number suggests that merely identifying
00:01:06.01 or cataloging the names of the genes is not enough to
00:01:09.00 explain how the system comes together.
00:01:11.04 What you'd really like is a movie of neural development
00:01:13.11 in vivo but for various technical and ethical reasons,
00:01:16.04 this is still impossible in human embryos.
00:01:18.10 However, if we jump down a few orders of magnitude
00:01:22.06 in complexity to the C.elegans worm embryo
00:01:25.10 this is a system that is much simpler.
00:01:29.09 Now there are only 302 neurons, the worm is transparent,
00:01:32.22 so we can use GFP to visualize where the neurons are
00:01:36.15 inside the embryo, there are about 5000 chemicals
00:01:39.14 synapses and about 20,000 gene products.
00:01:42.02 This is a simple enough system that we can use a
00:01:45.04 microscope to look at where the genes products are
00:01:47.28 and see how the neurons form the brain in a living embryo.
00:01:51.21 So we can ask questions like
00:01:53.03 where the neurons are and how got there.
00:01:55.00 So one of the goals then, one of the major goals of my lab,
00:01:59.11 in collaboration with these other PIs,
00:02:02.08 is to build a four dimensional digital atlas of
00:02:04.22 neurodevelopment in the worm embryo.
00:02:06.23 That is to say, we would like to use microscopes
00:02:08.24 to follow the cell migrations and outgrowths
00:02:11.21 of all 222 neurons in the worm embryo.
00:02:14.11 Now since C.elegans is stereotyped,
00:02:16.27 we can look at hundreds or thousands of embryos,
00:02:19.19 each with different sets of neurons that are labeled with GFP
00:02:22.25 and then reconstruct a digital representation
00:02:25.20 from all of these datasets in order to see how
00:02:28.04 the brain forms volumetrically and across time.
00:02:31.02 After we have such a dataset, we
00:02:33.09 plan on releasing this resource to the community
00:02:36.00 and if you'd like to learn more or follow us,
00:02:38.05 you can do so at wormguides.org.
00:02:40.10 Now my interest in this problem is also technical.
00:02:44.15 That is to say, how would you build a microscope
00:02:46.28 that can visualize an embryo without killing it
00:02:49.11 or for about 7 or 8 hours of neurodevelopment during embryogenesis?
00:02:54.17 If we consider a C.elegan's embryo in cross-section
00:02:57.17 it's about the size of a large cultured cell
00:03:00.08 that is about 50 microns laterally and
00:03:02.17 and about 20 or 30 microns axially and our task as microscopists
00:03:07.17 is to carve up this volume into lots and lots of
00:03:11.05 little diffraction limited voxels.
00:03:13.14 Now although the size of the embryo is not that much bigger
00:03:17.02 than a large cultured cell,
00:03:18.04 the size of a diffraction limited voxel is much much smaller
00:03:21.16 and so building a microscope that can section this embryo into
00:03:25.17 voxels of this size without killing it over
00:03:28.06 the many hours we need to investigate
00:03:30.09 neurodevelopment is a challenge.
00:03:32.05 How might be address this challenge?
00:03:34.25 The simplest microscope that we might use
00:03:37.28 is a widefield or epifluorescence microscope
00:03:40.18 and in this microscope, we shine the light
00:03:43.06 throughout the entire embryo volume
00:03:45.04 while recording fluorescence from one plane at a time.
00:03:48.13 In order to build up a volume, we move the embryo
00:03:51.25 relative to the microscope objective or move
00:03:53.17 the objective relative to the embryo but the problem is
00:03:56.11 that we are only recording from one plane at a time
00:03:58.24 while the embryo is getting dozed
00:04:00.19 throughout its volume with light.
00:04:02.12 So this is a very inefficient type of microscope
00:04:04.26 and there is no rejection of out-of-focus light.
00:04:07.11 What this means is that we cannot continuously monitor
00:04:10.29 the embryo or for the many hours we
00:04:13.03 need without killing or arresting the embryo.
00:04:15.04 Now the next most sophisticated tool you might choose
00:04:19.01 is the confocal microscope and in this microscope
00:04:21.19 we focus the laser beam to a diffraction limited spot
00:04:25.11 and we scan this diffraction limited spot in all three dimensions
00:04:28.25 throughout the embryo.
00:04:30.02 The useful information generated by a confocal microscope
00:04:33.22 only at the focus of the beam.
00:04:35.24 All of the rest of the light is thrown away by a pinhole.
00:04:38.29 So although this microscope, unlike the
00:04:41.05 widefield microscope, provides us with optical sectioning
00:04:44.14 and rejection of out of focus light,
00:04:46.09 it suffers the same problems as the widefield microscope
00:04:49.10 in terms of out of focus light
00:04:50.28 and out of focus bleaching of the fluorophore.
00:04:53.11 So this also is not an ideal tool for studying
00:04:57.01 embryo genesis over many hours.
00:04:59.19 A conceptually much cleaner approach would be to
00:05:03.12 focus the light into a single plane or a light sheet,
00:05:06.13 as I am showing over here.
00:05:08.04 If we can tailor the light sheet so only
00:05:11.08 the plane that the object is looking at is illuminated
00:05:14.07 we can drastically cut down on out of focus
00:05:18.21 light and out of focus photo damage
00:05:21.06 and photo bleaching.
00:05:22.07 This is also much faster than a confocal microscope
00:05:25.17 because in order to build up a three dimensional volume
00:05:28.11 we only have to scan this plane in one dimension
00:05:31.23 while recording each plane with a camera in order to bulid up a volume.
00:05:35.01 Now the idea of using light sheets in microscopy
00:05:38.09 is not a new idea.
00:05:39.05 In fact it is more than a 100 years old.
00:05:41.10 The idea of using them to study development
00:05:43.25 is relatively new and and there's a whole family of these
00:05:47.09 techniques that are broadly known as
00:05:49.11 Selective Plane Illumination Microscopes or SPIM microscopes.
00:05:53.07 One of the early examples of selective plane illumination microscopy
00:05:57.20 is shown in this cartoon over here from a paper
00:06:00.25 almost a decade ago.
00:06:01.25 Now in this SPIM microscope, a fish embryo is
00:06:05.03 embedded in a cylinder of agrose.
00:06:07.06 The fish plus the agrose is embedded in water filled chamber
00:06:11.21 and light is brought in from the slide in the form of sheet.
00:06:15.04 Fluorescence detection occurs in a direction perpendicular
00:06:18.20 to the light sheet and so by moving the
00:06:21.00 embryo though the light sheet while recording an image
00:06:23.18 for every position of the embryo,
00:06:25.14 we can build up a beautifully clear image of the embryo,
00:06:30.02 as is shown in this middle panel over here.
00:06:31.28 Now note that if were to bring in the light volumetrically
00:06:35.19 as an epifluorescence, the contrast in the embryo
00:06:38.23 is much reduced relative to the case of the sheet.
00:06:41.14 So this technique is now, I would say, established.
00:06:45.10 There are several labs around the world
00:06:47.06 that use it and it has been commercialized.
00:06:49.05 The problem is it is a pain the butt to build
00:06:51.27 for your average biologist and it is not very suitable if
00:06:55.16 what you would like to study is samples on a glass coverslip.
00:06:58.21 And samples on glass coverslips might include cells
00:07:01.17 in culture or in my case, C.elegan embryos.
00:07:05.00 In order to better image C.elegans for this neurodevelopment project
00:07:09.22 we have redesigned a selective plane illumination microscope
00:07:12.19 so it can be mounted on a conventional inverted
00:07:15.06 microscope base and we call this
00:07:17.12 iSPIM for inverted microscope based SPIM.
00:07:20.07 The design is as follows:
00:07:21.28 We have two objectives orientated at 90 degrees.
00:07:25.00 One of these objectives brings in a light sheet
00:07:27.15 and the other one collects the fluorescence at 90 degrees.
00:07:30.04 By sweeping the light sheet through the sample
00:07:32.23 with a galvanometer mirror and following the way the
00:07:36.16 light sheet moves with our detection objective
00:07:38.22 we can rapidly acquire a volume that is illuminated with a light sheet.
00:07:43.05 Now there are a couple of examples of
00:07:45.26 this sheet that I've showed over here.
00:07:47.21 Here is a cartoon of showing the two objectives
00:07:49.12 and here is a photograph showing the first
00:07:51.18 implementation of the light sheet where we've
00:07:53.12 merely replaced the bright field illumination pillar of the
00:07:56.24 microscope with this mount containing these
00:07:59.06 two perpendicular objectives and we have
00:08:01.20 a sensitive camera that we use to detect the fluorescence.
00:08:04.22 Having built this microscope we wanted to of course test
00:08:08.00 it to see if we could use it to follow embryogenesis
00:08:10.12 in C.elegans without killing the embryo.
00:08:12.22 And so we first did proof of concept experiment
00:08:16.06 where we labeled all the nuclei in the C.elegans' embryo
00:08:18.22 with a GFP histone marker and what I am showing you over here
00:08:22.00 is the single dataset that has been broken up into
00:08:24.17 two halves but it's from the same embryo.
00:08:27.15 Every frame in this movie is a maximum intensity
00:08:30.25 projection from a volume,
00:08:32.03 every volume is acquired in two seconds,
00:08:34.18 and there are about 25000 such volumes in this dataset.
00:08:37.25 And there are a few things to note.
00:08:39.12 One is the cell divisions all occur on time.
00:08:42.12 So to our knowledge we are not perturbing the embryo
00:08:45.11 by introducing the light sheet and acquiring it at such
00:08:47.25 high speed and over such an extended duration.
00:08:50.12 The other thing to note is that even when
00:08:52.10 when the embryo starts to twitch
00:08:53.20 after the muscles form, our imaging is still
00:08:56.14 fast enough that we can freeze the embryo
00:08:58.23 without significant motion blur from any of these volumes
00:09:01.11 and the fact that we are using the light sheet
00:09:03.26 makes us about 30 times faster than confocal spinning disk microscopy
00:09:07.18 and roughly the same signal to noise ratio and resolution.
00:09:12.19 Having tested this on GFP histones, we next wanted to
00:09:17.10 test it on a few neurons and so we built another strain
00:09:20.26 where we expressed GFP in just a few neurons and looked at the
00:09:24.17 behavior of these neurons after they were born
00:09:26.13 and through embryogenesis.
00:09:27.24 Now in this particular dataset, it is again data from a single embryo
00:09:31.26 that we have broken up into 4 different sets for clarity.
00:09:35.03 You can see that these neurons start out at the
00:09:38.27 anterior end of the embryo and they slowly migrate
00:09:41.14 to the posterior end.
00:09:43.14 After twitching starts, it becomes a little bit more difficult to
00:09:46.17 see what's going on but note that again our imaging is
00:09:49.19 fast enough that we can easily see individual cell bodies.
00:09:53.10 We can see long axons that grow from individual cell bodies
00:09:56.22 and our resolution is good enough to see
00:09:58.17 the growth cones at the ends of the axons.
00:10:00.28 And again, this all can be done non-invasively
00:10:03.15 because the embryo hatches at the end of this 7 hour movie dataset.
00:10:08.13 Now, in doing these experiments, we realized the light sheet illumination
00:10:12.28 was great in terms of gentleness and in terms of
00:10:16.00 absolute lack of photobleaching for this sample
00:10:18.27 but there is one area we felt we could still improve upon
00:10:21.26 and that is axially resolution.
00:10:23.24 All of the movies I've shown you so far of these datasets
00:10:27.15 acquired in our iSPIM have been projections that have been
00:10:31.09 XY maximum intensity projections so the gene information
00:10:34.09 is not explicit in these movies.
00:10:36.17 And there's a reason for that.
00:10:37.27 And that is that in any single view microscope
00:10:40.22 the axial resolution is always worse than the lateral resolution
00:10:44.15 by at least 2 to 3 fold and this is because
00:10:47.19 any single objective lens collects fewer angles
00:10:50.28 along the optic axis than it does perpendicular to the optic axis.
00:10:54.26 What does this mean for light sheet microscopy?
00:10:57.10 What it means is that objects that appear in the plane of the sheet
00:11:00.18 appear sharper than objects perpendicular to the sheet.
00:11:04.04 And so if we think about our C.elegan embryos again,
00:11:06.28 if you look at one of these volumes of nuclei
00:11:09.19 as we scroll through in Z, the nuclei appear clearly separated
00:11:14.10 and we can resolve them.
00:11:15.14 If you take that same volume and rotate it 90 degrees
00:11:18.21 notice that the nuclei look distorted and elongated
00:11:22.03 and that is because the loss of axial resolution in this microscope.
00:11:25.19 So this got us to sort of thinking about designing microscopes
00:11:29.23 that could be used to produce isotropic resolution.
00:11:33.16 That is, resolution that is the same in any direction you look at it.
00:11:36.27 How do we do this?
00:11:38.17 Well, the perpendicular geometry of SPIM suggests
00:11:42.26 an almost trivial way of creating a volume with isotropic resolution.
00:11:48.09 If we take that cartoon in the previous slide
00:11:51.02 and rotate it 90 degrees, notice that we flipped the
00:11:54.18 directions of good and bad resolution.
00:11:56.14 If we orientate the light sheet now in the up down direction
00:12:00.04 objects in the up down direction now appear sharp
00:12:04.12 whereas objects in the X direction appear worse resolution.
00:12:08.14 So we reasoned if we could acquire 2 volumes
00:12:11.09 from perpendicular directions and fuse the results together
00:12:14.14 cleverly, we might be able to obtain a volume
00:12:17.08 with equally good resolution in an X, Y and Z.
00:12:19.20 The modification to the original iSPIM in order to
00:12:24.12 acquire the secondary view is actually the easy part
00:12:26.27 because we use the same perpendicular objectives.
00:12:29.20 The first objective brings in the light sheet
00:12:31.29 and the second object collects the fluorescence under a camera
00:12:34.24 but then we just reverse this process.
00:12:36.27 So we just have the second objective
00:12:38.11 to bring in the light sheet and the first objective
00:12:40.13 collect the fluorescence.
00:12:41.23 So in essence, the main modification to the instrumentation
00:12:45.07 is a second camera.
00:12:46.12 And I think what I would like to do now is show
00:12:49.17 you the instrument in our lab.
00:12:51.14 What we are looking at here is diSPIM as I discussed in my talk.
00:12:56.11 This particular diSPIM is the latest incarnation of the
00:12:59.16 device that has been built by 2 postdocs in my lab,
00:13:02.17 Yicong Wu and Abhishek Kumar.
00:13:05.00 And so what I would like to do is just walk you
00:13:07.13 through the different components of this instrument.
00:13:09.10 So we have kind of off screen over here,
00:13:12.16 a laser sled that provides illumination to the diSPIM
00:13:16.13 and one of the nice things about this incarnation
00:13:19.01 of the device is that the lasers coupled
00:13:21.21 via fibers, for example over here and at the other end
00:13:25.23 and that's what brings the laser light into this assembly.
00:13:29.04 The two perpendicular objectives I mentioned in my talk
00:13:33.05 are mounted on top of this frame.
00:13:35.14 This is an inverted microscope based frame
00:13:38.12 that is commercially available and so
00:13:41.14 we have the laser light and we have the two objectives
00:13:44.24 and then everything in between the fiber and the
00:13:47.13 the objectives are optics and mirrors in order to
00:13:50.22 both create the light sheet and to scan the light sheet
00:13:53.26 in a volume fashion.
00:13:55.17 So just to give you some sense of what's involved here
00:13:58.17 we have a fiber and a scan head.
00:14:01.02 This is a micromirror that moves in two dimensions
00:14:04.26 to create the light sheets and to create the volume
00:14:07.14 and then we image that micromirror to the back focal plane
00:14:11.25 of this objective lens via some other lenses
00:14:15.13 that are in this arm.
00:14:17.06 Now one of the unique things about the diSPIM
00:14:19.13 is that it is symmetric so we can take everything I just described
00:14:23.09 and apply it over here to this side.
00:14:26.02 So there are two micromirrors, two sets of lenses,
00:14:29.11 and two perpendicular objectives.
00:14:31.03 The other important part of the diSPIM is
00:14:34.00 of course the sample and so this microscope
00:14:36.22 base also has a stage.
00:14:38.25 The stage has a chamber that contains some embryos
00:14:42.02 and liquid that covers the embryos and these two objectives
00:14:45.25 can be brought down to focus on the embryos
00:14:49.16 and maybe I can show you that process just now.
00:14:53.08 So when we are about to take an experiment
00:14:55.23 we load our sample in the stage, we turn on the laser beam
00:15:00.00 and you can kind of see there is light coming through both
00:15:03.20 objects and then we lower the two objectives so that they
00:15:07.23 both focus on the sample.
00:15:09.01 So this mount over here is on a mechanical translation stage
00:15:14.12 that Abhishek is controlling to bring the objectives down.
00:15:18.05 I should also say these objectives are water dipping objectives.
00:15:22.11 So we actually have liquid in this chamber and the reason we
00:15:26.16 use water dipping objectives instead of an air objective
00:15:29.05 or an oil objective is that these minimize spherical aberrations.
00:15:32.21 And so if we kind of zoom in here, you can see that
00:15:36.00 these objectives are really very close.
00:15:38.07 They are not quite touching but in order to have two objectives
00:15:42.08 at 90 degrees and to both focus on the same spot in the
00:15:45.24 sample, we need very long working distance objectives.
00:15:48.17 So these are commercially available but specially
00:15:51.23 chosen to have the largest numerical aperture
00:15:54.02 they can while fitting into this tight geometry.
00:15:56.29 And what I would like to do next is actually show
00:15:59.13 you some raw data of this system as it acquires
00:16:03.17 images of embryos that are mounted on this coverslip.
00:16:07.05 Ok, to give you an example of the light sheet sectioning
00:16:10.25 in action we have an embryo mounted on the plate,
00:16:13.05 on the glass coverslip and what we are going to do
00:16:16.10 is scan the position of the embryo through the light sheet.
00:16:20.11 So right now, what you can see is a cross section
00:16:22.25 through the embryo with very little out of focus light
00:16:26.02 and as I move the embryo through the light sheet
00:16:30.20 notice that we can focus on the different segments
00:16:33.02 of the embryo. So we can go through focus
00:16:35.24 by translating the specimen through the sheets.
00:16:38.16 Now what we do an actual experiment
00:16:41.00 we leave the embryo stationary and then sweep
00:16:44.07 the sheet though the sample.
00:16:45.06 Here are the GFP histones again in a growing
00:16:49.09 C.elegan embryo but now viewed with both
00:16:51.18 iSPIM and diSPIM and you can see that the diSPIM,
00:16:55.02 in particular in the YZ view at the bottom,
00:16:57.28 offers a much higher resolution than the iSPIM.
00:17:00.11 The nuclei no longer look distorted, they look circular
00:17:03.18 as they should and we have been able to maintain
00:17:05.22 the high imaging speed of the iSPIM because we've
00:17:08.26 just added one extra view of the sample.
00:17:10.23 The embryo is still stationary and we can acquire
00:17:13.10 these views rapidly, both views in about half a second.
00:17:16.20 Furthermore as you can see, the embryo hatches on time.
00:17:20.09 Having verified the performance of the diSPIM on nuclei,
00:17:26.13 we again started to look at neurons and once more
00:17:28.24 we noticed that the improvement in axial resolution
00:17:31.26 is significant for very fine processes
00:17:34.15 that grow out of the neurons.
00:17:36.04 Although we can easily single cell bodies in these single
00:17:40.08 view iSPIM, if you want to look at the very fine
00:17:42.23 projections that emerge from each of these neurons,
00:17:45.17 they are much more clearly resolved in the diSPIM than in the iSPIM.
00:17:49.03 I'd like to finish by pointing out the although we now
00:17:55.06 have a microscope that capable of imaging
00:17:57.06 C.elegans’ embryos at extremely high spatial and temporal
00:18:01.26 resolution and so our goal of imaging all the neurons
00:18:05.22 in C.elegans’ embryo can now proceed,
00:18:08.08 they system can also be used to look at many other
00:18:11.03 specimens, two of which have highlighted here.
00:18:12.21 In the upper example, we can follow GFP-EB3
00:18:17.11 labeled microtubules. So you see that the tips of the microtubules
00:18:20.20 can be visualized easily in both lateral and axial views.
00:18:24.20 For the first time, the diSPIM enables us to track
00:18:27.23 the tips of the microtubules in all three dimensions.
00:18:30.18 If you try and look at the same proteins in spinning disk confocal
00:18:35.07 microscopy, although we get good lateral views,
00:18:37.26 the axial resolution in the spinning disk is much worse
00:18:40.16 than in the diSPIM. Furthermore, we also are barely bleaching
00:18:44.05 these samples at all in the diSPIM whereas in a
00:18:46.23 spinning disk confocal microscope, we see rapid bleaching.
00:18:49.27 As a final example, these are GFP labeled
00:18:53.00 caveolae in a cultured cell and again, this particular movie
00:18:57.13 proceeds for a long time.
00:18:59.07 I'm showing you one cross section in the axial, XZ, plane.
00:19:03.06 The caveolae well dissolved if there is no axial distortion
00:19:07.14 and this movie goes on for about half an hour.
00:19:09.22 So there is a ton of data that can be collected
00:19:12.03 at very high resolution over an extended period of time
00:19:14.27 and for the first time, we can think about computationally
00:19:17.24 following the positions of these small objects
00:19:20.20 as they traverse the cell.
00:19:21.20 I would just like to end by thanking all of my collaborations,
00:19:25.22 both within the NIH and without.
00:19:28.07 In particular, all of the work that I have described here,
00:19:31.05 namely the building of the iSPIM and the diSPIM,
00:19:33.21 has been conducted by Yicong Wu, a very talented
00:19:36.21 staff scientist. Abhishek Kumar has been continuing the work
00:19:40.05 and has been building the microscope that we've seen
00:19:43.05 in this presentation.
00:19:44.06 Thank you very much.

This Talk

Speaker: Hari Shroff

Audience:

Researcher

Recorded: July 2013

Talk Overview

Selective Plane Illumination Microscopy (SPIM) greatly reduces phototoxicity in comparison with other fluorescent imaging modalities and makes it possible to image living small animals in 3D over extended periods of time. This talk describes an extension of SPIM such that it can be used with specimens on a coverslip rather than in a capillary (inverted SPIM or iSPIM), and a second modification that images the specimen using two perpendicular light sheets (Dual-View iSPIM or diSPIM), resulting in 3D datasets with the same resolution in X, Y, and Z (isotropic resolution).

Questions

What distinguishes iSPIM from SPIM?
What is the advantage of diSPIM over iSPIM?
What types of biological objects can be images with diSPIM?

Answers

Speaker Bio

Hari Shroff

Hari Shroff

Hari Shroff is an Investigator at the National Institute of Biomedical Imaging and Bioengineering at the National Institutes of Health. During his post-doc, which was with Eric Betzig at HHMI’s Janelia Farm Research Campus, he focused on the development of PALM (photoactivated localization microscopy) microscopy. Since then he has developed several microscopy techniques, such as… Continue Reading