The Role of Microglia in Health and Disease

Part 1: Microglia States in Health and Disease

00:00:13.11 Hello, my name is Beth Stevens,
00:00:15.06 and I'm an HHMI Investigator
00:00:17.02 at Boston Children's Hospital and The Broad Institute.
00:00:19.06 And my lab is largely interested in understanding
00:00:22.09 how the immune system helps sculpt developing neural circuits.
00:00:25.25 Now, from the time I was a graduate student,
00:00:27.18 I've been fascinated by glial cells,
00:00:30.02 the non-neuronal cells of the brain.
00:00:32.20 But even as a card-carrying glial biologist,
00:00:34.20 I've largely ignored the cell I'm gonna tell you about today:
00:00:38.24 microglia, our resident immune cells.
00:00:41.17 Now, there are many reasons for that,
00:00:43.07 but it largely was in part because we really lacked the tools
00:00:45.27 with which to study these cells
00:00:47.27 until fairly recently.
00:00:49.19 We now know that microglia,
00:00:51.21 our resident phagocytes and immune cells,
00:00:54.05 play central roles in both brain development, function,
00:00:58.04 and dysfunction in disease.
00:01:00.01 They make up about 7-10% of the cells in our brain.
00:01:03.04 And despite being described over 100 ago
00:01:05.28 by Rio Hortega,
00:01:07.25 we're only now beginning to understand
00:01:10.02 their dynamic functions in the brain.
00:01:12.02 Now, we've known for quite some time
00:01:14.12 that they play a key role when the brain is injured,
00:01:17.00 or in the context of disease.
00:01:19.00 Microglia change dramatically
00:01:22.06 in the context of even local perturbations in the brain.
00:01:25.04 They largely are these beautiful cells
00:01:28.01 that have these dynamic processes
00:01:30.17 that you can't quite appreciate from this image.
00:01:32.22 But under the context of injury and disease,
00:01:35.21 they can dramatically change shape.
00:01:37.11 They become more phagocytic in many cases.
00:01:40.01 And they are known to play key roles in neuroinflammation,
00:01:43.18 by releasing molecules like cytokines into the brain.
00:01:47.10 They're also known, as our resident phagocytes,
00:01:49.23 to be very good at clearing pathogens and debris.
00:01:53.06 And moreover, they can remove local
00:01:56.23 and toxic proteins in the brain.
00:01:58.24 Now, these functions,
00:02:00.14 which have been described for many years,
00:02:02.07 are both detrimental and beneficial,
00:02:04.14 depending on the context.
00:02:06.03 But what we're now appreciating
00:02:08.06 is the fact that these cells play far more diverse roles
00:02:11.13 than we previously appreciated.
00:02:13.20 And this, again, is because until recently
00:02:17.00 we lacked tools to study them in detail.
00:02:20.02 One of the things that really has cracked open the field
00:02:23.00 is new reporter transgenic lines,
00:02:24.28 like this one I'm showing you that was developed
00:02:27.20 by Steffen Jung a number of years ago.
00:02:29.16 This is a mouse that's been genetically engineered
00:02:32.10 to expressed EGFP -- this fluorescent green protein --
00:02:34.16 in all the microglia and myeloid cells.
00:02:36.21 And using this tool,
00:02:39.10 researchers can look at microglia
00:02:41.21 in the developing, or adult,
00:02:43.26 or even diseased brain
00:02:45.16 and watch these cells in action.
00:02:48.27 And this is in large part due to the fact
00:02:50.27 that we can do two-photon imaging
00:02:53.00 by making these thin cranial windows
00:02:55.02 and put a microscope over these fluorescent green mice.
00:02:58.06 And using these tools,
00:03:00.20 a number of pioneers in the field
00:03:03.12 really were the first to start to make some of these really fundamental
00:03:06.13 new observations about microglia
00:03:08.11 that certainly changed the way I've been thinking about them.
00:03:11.11 So, Axel Nimmerjahn and Dimitrios Davalos and Wenbiao Gan
00:03:15.25 were some of these pioneers
00:03:17.13 that started to do imaging experiments like I just described.
00:03:20.04 So, as I mentioned,
00:03:21.29 microglia are really good at eating or engulfing things.
00:03:24.27 They can rapidly clear debris
00:03:27.05 or dead cells in the brain.
00:03:29.17 And they respond pretty dramatically to injury.
00:03:32.04 And this is illustrated by an experiment
00:03:34.11 that Axel Nimmerjahn did a number of years ago.
00:03:36.03 And this was enabled by the fact that...
00:03:38.11 again, using this reporter line, this green...
00:03:40.13 GFP fractalkine receptor mouse
00:03:42.24 that was crossed to a neuronal reporter line,
00:03:45.12 where the neurons are labeled in red...
00:03:48.02 what Axel did is in the Purkinje cells in the cerebellum,
00:03:52.01 he made a little laser injury to that red neuron.
00:03:54.18 And what you can see from this movie
00:03:56.05 is that the microglia were dramatically responding
00:03:58.19 to that local injury and engulfing,
00:04:00.20 or pulling up the parts of that neuron,
00:04:03.11 after the injury.
00:04:04.27 Now, this is a very dramatic example,
00:04:06.26 but this just illustrates their ability
00:04:09.23 to phagocytose through the expression
00:04:12.02 of a lot of different phagocytic receptors.
00:04:14.23 Now, in addition to phagocytosis,
00:04:17.10 as I mentioned,
00:04:19.10 microglia are incredibly dynamic.
00:04:21.24 The processes in particular
00:04:24.00 are constantly surveying the brain environment
00:04:25.29 and the brain parenchyma.
00:04:27.24 And these experiments, again,
00:04:29.13 that were first done by these pioneers,
00:04:32.03 showed that by just looking at microglia
00:04:34.18 under normal conditions in the brain
00:04:36.20 you can see their fine processes moving.
00:04:38.19 But if you were to make an injury
00:04:40.21 -- in this case, with a laser --
00:04:42.23 the microglia dramatically move and extend their processes
00:04:45.14 toward the injury, as shown here.
00:04:47.15 Now, this is really rapidly happening,
00:04:51.01 although it's quite sped up in this movie.
00:04:53.08 But you can see the fine processes
00:04:55.28 chemoattracting to the injury site.
00:04:58.10 And it's been shown, now,
00:05:01.04 that one of the signals released
00:05:02.15 that actually brings the microglia to the site of injury
00:05:05.16 is ATP or puri... they have purinergic receptors that recognize this signal.
00:05:09.01 But this is just one example of under...
00:05:11.15 under a real acute injury what microglia look like.
00:05:14.04 Okay, so microglia not only a respond
00:05:17.08 when things go wrong,
00:05:19.06 but two-photon imaging like this has revealed that
00:05:21.26 microglia are constantly surveying the brain environment
00:05:23.29 with their fine processes.
00:05:26.05 So, the question of course is,
00:05:28.15 what is it that they're surveying?
00:05:30.07 What are the signals that are regulating their motility?
00:05:33.06 And although you can't see the cells underneath there,
00:05:35.21 what signals are they using to communicate
00:05:38.10 with neurons and other cells
00:05:40.13 to regulate synapse and brain development?
00:05:43.04 So, microglia are indeed integral parts of brain circuitry.
00:05:46.18 Over the last many years,
00:05:49.00 more and more evidence is suggesting that microglia
00:05:51.12 actively signal with neurons and synapses,
00:05:54.03 but also with other glial cells,
00:05:56.00 where they play critical roles in brain development
00:05:58.17 and plasticity and function.
00:06:02.12 Now, what we also appreciate is that microglia
00:06:05.09 are the only cell not born in the brain.
00:06:07.15 Recent work...
00:06:09.20 a fate mapping study by Miriam Merad's lab
00:06:12.06 and Florent Ginhoux a number of years ago...
00:06:14.13 did some really important fate mapping studies
00:06:17.00 and showed that microglia
00:06:19.29 actually derived from the yolk sac
00:06:22.08 as myeloid progenitor cells.
00:06:24.14 They enter the brain quite early in development.
00:06:26.28 And then, once they're in the brain
00:06:29.20 and they're exposed to brain signals,
00:06:31.09 they become known as what we call resident microglial cells,
00:06:33.28 and they differentiate into these cells
00:06:36.24 that I just mentioned -- these surveying microglia
00:06:39.12 that mature over the first several weeks of development.
00:06:41.08 Now, we know that these microglia
00:06:43.21 have many different functions,
00:06:46.00 many more that I'm probably gonna tell you about today
00:06:47.29 that we're just beginning to unearth.
00:06:50.19 But in particular, even before birth,
00:06:52.25 there's been evidence to suggest that microglia
00:06:55.01 are critically important in neurogenesis.
00:06:57.13 For example, as resident phagocytes,
00:06:59.09 they actually help to clear and phagocytose
00:07:02.01 newborn neurons,
00:07:03.09 and there's actually even some evidence
00:07:05.10 that they can engulf neurons
00:07:07.26 without clear evidence of apoptosis.
00:07:09.26 So, some of those signals have been identified.
00:07:11.28 There's also evidence to suggest
00:07:15.10 that microglia release signals such as chemokines and cytokines
00:07:17.28 that are also important in regulating
00:07:20.10 not just neurogenesis but different aspects
00:07:22.26 of neuronal differentiation during development.
00:07:25.24 And there's also evidence to suggest that axons...
00:07:29.15 that microglia are also dynamically regulating
00:07:33.05 the refinement and pruning back
00:07:36.02 of these developing axon tracts before birth,
00:07:38.19 and that they can also regulate the fate,
00:07:41.10 and the outcome,
00:07:44.06 of certain subtypes of interneurons in the adult brain.
00:07:47.18 So, this is all happening during embryonic development,
00:07:50.18 but of course, after birth,
00:07:53.01 more and more evidence points to the fact
00:07:54.22 that these cells are playing key roles
00:07:57.05 in myelination...
00:07:59.01 so, they can regulate the process of myelination
00:08:00.27 but also the process of oligodendrocyte differentiation.
00:08:04.26 They're involved in synaptic development, synaptic maturation,
00:08:08.11 and synaptic plasticity.
00:08:10.15 And they're also key regulators
00:08:12.29 in synaptic remodeling, or synaptic pruning,
00:08:14.23 which I'll tell you more about today.
00:08:16.27 So, this suggests that microglia
00:08:19.00 are existing in specialized states.
00:08:20.26 And we... as I mentioned before,
00:08:23.26 these cells are obviously dynamically associating
00:08:26.28 with synapses.
00:08:28.12 And this is now illustrated in this movie,
00:08:30.12 where we can overlay a microglia, in green,
00:08:33.11 with neurons, which are labeled in red,
00:08:36.17 and this led to the observation
00:08:39.07 that these microglia are dynamically interacting
00:08:41.23 with subsets of synapses -- spines and axons --
00:08:43.21 especially during developmental critical periods.
00:08:47.01 These cells are incredibly dynamic,
00:08:50.10 and they're, as I'll tell you,
00:08:50.27 particularly phagocytic in times
00:08:55.11 and places of synaptic remodeling,
00:08:57.00 which is a normal developmental process
00:08:59.00 that's happening over the first weeks of life.
00:09:02.16 And so, the question is,
00:09:04.04 could microglia be involved in this process
00:09:06.25 of synaptic refinement or synaptic pruning?
00:09:09.17 Now, synaptic pruning is, as I mentioned,
00:09:12.06 a normal developmental process
00:09:13.25 in which excess synapses or connections
00:09:15.19 are permanently removed
00:09:18.06 while others get strengthened and maintained.
00:09:20.15 They happen over... this process happens over different critical periods,
00:09:22.23 so there are parts of your brain, like sensory systems,
00:09:26.01 that are... that are pruned relatively early in development,
00:09:28.13 but there are some parts of our brain,
00:09:30.07 like the frontal cortex,
00:09:32.04 that continue to prune and remodel
00:09:33.29 into early adulthood.
00:09:36.07 And so, the question is,
00:09:38.01 you know, could microglia play a role in this process?
00:09:39.25 And when you look at microglia morphology
00:09:42.02 and you start to characterize
00:09:44.01 some of the markers they express,
00:09:45.16 we noticed pretty early on that
00:09:49.14 during this developmental critical period of pruning,
00:09:51.10 microglia were particularly phagocytic,
00:09:54.14 meaning they expressed a lot of phagocytic receptors
00:09:56.05 and they had more lysosomal activity at these periods
00:09:58.16 where they were associating with synapses.
00:10:01.06 And in the mouse visual system,
00:10:03.06 the model system that we've been using,
00:10:05.04 we provided evidence that microglia are in fact engulfing
00:10:09.20 some of these axons, or inputs,
00:10:11.17 during development.
00:10:13.06 This is a 3D surface rendering of a microglia
00:10:15.17 from the visual thalamus, for example,
00:10:18.14 during its critical period of remodeling.
00:10:20.20 And Dorothy Schafer in my lab
00:10:22.25 showed that most of the microglia at this time
00:10:25.12 of eye segregation in the visual thalamus
00:10:27.29 were engulfing large numbers of these presynaptic inputs,
00:10:30.19 which could also be shown by EM.
00:10:34.29 And it was happening, again, during this window of development,
00:10:37.08 and they largely decreased their ability to engulf synapses
00:10:39.24 after that critical period, raising the question about...
00:10:41.26 what are the signals that are instructing microglia
00:10:44.18 to engulf synapses,
00:10:47.01 and are there even... more specifically,
00:10:49.09 signals that are telling microglia to engulf specific synapses,
00:10:52.06 for example the less active inputs during development?
00:10:55.13 And as I mentioned, you know,
00:10:57.08 this is an activity-dependent process.
00:10:58.28 How do microglia know which synapse to prune
00:11:01.19 and which synapse to lea... to leave alone?
00:11:03.29 Raising the question that there may be some sort of cues or instructive signals
00:11:07.26 that are recruiting microglia to certain synapses or inputs,
00:11:10.19 and there may also be signals that are telling microglia
00:11:14.01 not to engulf certain synapses or inputs.
00:11:16.26 So, one of the signals that we have been focused on,
00:11:19.12 and that we identified a number of years ago,
00:11:20.29 was a group of molecules called complement,
00:11:23.20 which was a group of innate immune molecules --
00:11:26.26 well studied in the innate immune system in the periphery.
00:11:29.21 These are "eat me" signals
00:11:31.27 that are known to tag apoptotic cells or debris...
00:11:35.06 and that they are a way that circulating macrophages in the periphery
00:11:41.10 recognize these apoptotic cells.
00:11:42.29 So essentially, these proteins bind or opsonize
00:11:45.12 the surface of these... the surface...
00:11:47.17 the membrane of these cells
00:11:49.07 and then tell the macrophages to engulf them.
00:11:51.25 Now, in the brain,
00:11:53.23 what we discovered when I was a postdoc
00:11:55.21 in Ben Barres' lab
00:11:57.16 is that a large number of these secreted complement molecules
00:11:59.15 were actually expressed in the healthy brain
00:12:01.16 by neurons and glial cells,
00:12:03.13 and especially, the microglia were making a lot of C1q, for example,
00:12:07.13 that initiating protein of the cascade,
00:12:09.09 and that many of these molecules
00:12:11.00 were actually binding to subsets
00:12:14.02 of these developing synapses.
00:12:15.23 And importantly, microglia are the only resident cell
00:12:18.14 to express the receptors for these complement molecules,
00:12:21.06 or CR3, for example.
00:12:23.10 So, in the immune system, as I mentioned,
00:12:26.16 complement is tagging a bacterial cell for elimination.
00:12:28.12 That's recruiting macrophages,
00:12:30.08 which express complement receptors.
00:12:31.26 And what we discovered is that the brain...
00:12:35.11 in the brain -- especially in the healthy brain --
00:12:37.08 complement molecules are binding to or localizing to
00:12:40.05 synapses and axons,
00:12:41.29 and this is mediating the engulfment of those inputs
00:12:45.15 through complement receptors.
00:12:47.09 And so, what we went on to show
00:12:50.05 was that if you knock out either the molecules
00:12:52.15 -- complement, the secreted molecules --
00:12:54.11 or the receptor on microglia,
00:12:55.22 there were defects in synapse elimination,
00:12:57.23 and that microglia, at least in the visual system,
00:13:00.13 were only about half as good at engulfing these synapses.
00:13:03.29 Now, this is just one example of a group of molecules that we...
00:13:07.10 that are instructive in synaptic pruning in this circuit,
00:13:10.23 but work over the last many years
00:13:13.14 by many labs
00:13:15.09 have identified more and more signals
00:13:17.12 -- some secreted, some surface-bound molecules,
00:13:20.11 both by microglia and the neurons
00:13:22.08 and other glial cells --
00:13:23.29 that are contributing to different aspects of both brain development
00:13:26.20 and synaptic refinement.
00:13:28.09 I just illustrated a few here.
00:13:30.26 And more importantly... which I want to touch on briefly,
00:13:34.25 it's... it raises also the question of whether or not,
00:13:38.19 in the context of disease,
00:13:40.16 where there's aberrant synaptic loss and an activation of microglia
00:13:43.12 and other immune-related pathways...
00:13:47.05 could some of these same pathways and mechanisms
00:13:49.27 that normally function to regulate brain development...
00:13:52.18 could they go awry and become dysregulated
00:13:55.02 in the context of brain injury and disease?
00:13:57.00 And indeed, many of these molecules, including TREM2, APOE...
00:14:01.06 these are actually molecules or proteins
00:14:03.29 that are also risk genes for diseases like Alzheimer's disease,
00:14:07.13 and we now appreciate from emerging genetics
00:14:09.24 that a large number of the late-onset Alzheimer's risk genes
00:14:14.15 are expressed or enriched in
00:14:17.22 myeloid or myeloid cells.
00:14:19.14 And that really does change the way we're thinking about these cells
00:14:22.02 in the context of these diseases --
00:14:24.01 not just as playing a secondary role,
00:14:26.06 but actually may even be driving some of the disease
00:14:30.22 in the context of at least Alzheimer's disease --
00:14:34.11 for example, TREM2 and some of the complement genes as well.
00:14:37.06 So, this raises the question,
00:14:40.16 how do microglia contribute to disease?
00:14:42.10 When do they contribute?
00:14:43.23 And these are really challenging questions to address,
00:14:45.08 obviously, in human disease,
00:14:47.00 because we really do lack the tools to be able to
00:14:49.11 both manipulate and label
00:14:51.19 and track microglial changes over the course of disease,
00:14:54.06 especially non-invasively,
00:14:56.07 but even in fixed tissue we don't have very good markers
00:14:58.05 that denote all these different functional states
00:15:00.23 that I just told you about.
00:15:02.11 So, imagine if we had a signature
00:15:04.22 that denoted a microglia
00:15:06.26 that was undergoing phagocytosis
00:15:10.00 during neurogenesis or synaptic pruning,
00:15:12.09 and differentiating it from a microglia
00:15:15.05 that was pruning a plaque versus aberrantly pruning synapses.
00:15:18.10 Right now, when these microglia change state,
00:15:21.09 we don't really have markers
00:15:23.03 that denote the beneficial
00:15:25.18 versus the detrimental states in the brain...
00:15:28.04 even inflammation... we call these inflammatory states,
00:15:30.28 but more than likely this represents
00:15:34.10 many different substates or subpopulations,
00:15:36.14 and even then, inflammation could be both beneficial and detrimental
00:15:40.14 depending upon the context and the cues.
00:15:42.23 So, this really is raising the...
00:15:46.00 both the challenge but also the fact
00:15:47.28 that we now are in a position to start to develop new tools
00:15:50.15 to be able to start to track these different states.
00:15:52.28 How heterogeneous are microglia?
00:15:54.14 How diverse are they?
00:15:56.08 How many states really even exist?
00:15:57.26 Before we can understand their role in Alzheimer's disease,
00:16:00.15 we thought maybe we should step back
00:16:03.10 and start to ask,
00:16:05.05 how many states exist under normal development across the lifespan of a mouse,
00:16:07.26 and what can we infer about their function,
00:16:10.16 and how do these states...
00:16:12.10 these microglial states change in the context of disease?
00:16:15.25 So, to address this question,
00:16:17.17 we've been applying a new technology that many in the field of course are using
00:16:21.09 called single-cell sequencing,
00:16:23.17 initially developed by Steve McCarroll, my collaborator,
00:16:26.01 and Aviv Regev and Evan Macosko.
00:16:27.18 Now it's a technology that many are using routinely,
00:16:31.01 but we applied this a number of years ago,
00:16:33.13 and I should say Tim Hammond and Connor in my lab,
00:16:35.19 in collaboration with Steve's lab,
00:16:37.23 decided to use single-cell sequencing
00:16:40.08 to assess microglia diversity
00:16:42.02 across the lifespan of a mouse.
00:16:43.22 And to do that...
00:16:45.10 and you may recall that microglia only represents
00:16:47.21 between 7 and 10 percent of the cells in the brain,
00:16:49.29 so that also involved having to develop
00:16:52.11 new methods to both dissociate and purify,
00:16:56.26 and isolate,
00:16:59.11 microglia only from the brains,
00:17:01.02 from many, many mice.
00:17:02.22 In fact, they ended up profiling
00:17:05.08 78,000 microglia across 41 mice.
00:17:07.14 And using these approaches
00:17:10.15 that Tim and Connor developed to purify the microglia,
00:17:13.09 we put them through a 10X or single-cell sequencing,
00:17:15.28 and in this way,
00:17:17.17 we could start to ask,
00:17:19.26 how diverse are they?
00:17:21.15 And the take home of that... this is a summary of Tim's paper,
00:17:23.15 but also a...
00:17:26.03 very consistent results were observed in a study
00:17:28.23 by Ben Barres' lab,
00:17:30.14 which is that there's a lot of diversity during development.
00:17:32.21 By about one month, at least in a mouse,
00:17:35.24 they're largely homogeneous --
00:17:37.19 one big population.
00:17:39.04 But then in the context of aging, pathology...
00:17:41.29 there's a shift in some of these states.
00:17:43.24 And this is the data
00:17:45.18 -- just to illustrate an example of that --
00:17:47.06 where you can actually...
00:17:49.09 we looked at all of the data put together
00:17:51.08 in these t-SNE plots,
00:17:53.14 and we started looking at what the different clusters
00:17:55.24 or the different populations of microglia were.
00:17:57.14 What were the genes in those different clusters?
00:18:00.02 And in particular, cluster 4 emerged
00:18:03.21 only during development,
00:18:05.12 during this postnatal period,
00:18:06.28 and this particular cluster, or state of microglia,
00:18:10.21 expressed genes like osteopontin or Spp1, Igf1, Lpl, and others.
00:18:16.14 And these are interesting because they're the same signatures
00:18:18.23 that are coming up in some of these disease-associated states
00:18:21.17 in Alzheimer's and other diseases,
00:18:23.14 yet we were seeing them in a small population of microglia
00:18:26.03 only during development.
00:18:28.02 And using... and interestingly,
00:18:30.01 these and other genes in this cluster
00:18:32.20 suggested immune recognition, lysosomal activity,
00:18:34.28 phagocytosis.
00:18:36.07 And so, to ask, where do these microglia exist
00:18:40.01 in the brain,
00:18:42.08 Connor and Tim did spatial mapping and multiplex in situ hybridization
00:18:45.18 -- they made probes for these markers --
00:18:48.07 and then found, quite excitingly,
00:18:50.07 that the... these microglia in cluster 4
00:18:53.16 were largely localized to developing white matter tracks,
00:18:57.16 and nowhere else.
00:18:59.15 And that of course opened up lots of interesting questions.
00:19:03.01 Could these special axon-associated microglia
00:19:05.14 be important in myelination, axonal refinement,
00:19:09.01 other aspects of refinement of these white matter tracts
00:19:12.15 that we were not yet thinking about?
00:19:14.20 And so, using some of these tools,
00:19:16.25 we're now in a position to start to manipulate
00:19:19.22 and deplete and knock out just these specific
00:19:22.06 cluster 4 microglia
00:19:24.15 to start to interrogate the function of these cells.
00:19:26.27 And we're taking a similar approach and...
00:19:28.25 as are others in the field,
00:19:30.13 to start to use the single-cell data to try...
00:19:32.07 to try to link changes in transcription
00:19:35.05 to specific functions of microglia.
00:19:37.16 I gave you one example:
00:19:39.21 these axon tract-associated microglia.
00:19:41.17 I don't think we necessarily would have found this particular population
00:19:44.21 had we not done the single-cell sequencing,
00:19:46.26 but this is now giving us a signature and markers
00:19:49.27 to start to use as tools to start
00:19:51.20 to manipulate specific states of microglia.
00:19:54.12 And I think this approach,
00:19:57.14 by combining single-cell and spatial mapping
00:19:59.21 and functional studies...
00:20:01.07 are gonna enable the field
00:20:03.14 to ask a number of fundamental questions.
00:20:05.10 What are the functions of these different populations
00:20:07.03 in different contexts?
00:20:08.18 Can we start to model...
00:20:10.22 faithfully model microglia?
00:20:12.12 Can we... how do the genetic mutations
00:20:14.21 and variance influence microglia states?
00:20:18.03 How do we link genetics to function?
00:20:20.22 And ultimately, can we use this information
00:20:23.03 to target and track microglia
00:20:25.01 across disease progression?
00:20:27.16 And I think this really is sort of where I want to wrap up... is that...
00:20:31.03 I do think that we're at an exciting time in the field.
00:20:33.25 But again, without understanding the normal functions of these cells,
00:20:37.01 it's gonna be really challenging to try to understand
00:20:40.04 how they go awry in these complex disorders,
00:20:42.05 especially given the diversity and heterogeneity
00:20:45.07 of these microglia in the brain.
00:20:47.05 So, we want to be able to develop new tools
00:20:49.21 to identify and track microglia populations.
00:20:51.29 We want to be able to use this information
00:20:54.11 to target the detrimental states, for example,
00:20:57.01 of microglia in different diseases.
00:20:59.00 And we ultimately hope that, maybe someday,
00:21:01.11 digging deeply into the biology
00:21:03.29 could unveil new biomarkers,
00:21:06.04 or maybe even new therapeutic targets,
00:21:08.06 that we could... we could use to start to manipulate microglia.
00:21:11.15 So, I'll end there by thanking the human microglia in my lab
00:21:14.27 that made a lot of the work I told you about today possible,
00:21:17.10 and also a lot of really terrific collaborators
00:21:20.16 and lab members,
00:21:22.00 both past and present.
00:21:23.12 So, thanks very much.

Part 2: How Microglia Sculpt Brain Circuitry in Health and Disease

00:00:13.05 Hi, my name is Beth Stevens.
00:00:14.27 I'm an HHMI investigator
00:00:16.22 at Boston Children's Hospital and The Broad Institute.
00:00:18.18 And our research is focused on understanding
00:00:21.06 how the immune system helps sculpt
00:00:22.21 developing neural circuits.
00:00:24.17 And in particular, we've been focusing on
00:00:26.15 this resident immune cell called microglia
00:00:28.20 that I'm gonna tell you about today.
00:00:30.19 So, I think it's become increasingly clear
00:00:32.21 that a large number of molecules
00:00:34.27 traditionally associated with the adaptive
00:00:37.14 and the innate immune system
00:00:39.07 are actually expressed normally in the healthy brain,
00:00:41.08 especially during development,
00:00:43.06 when the brain is wiring up.
00:00:44.29 So, neurons and glial cells make a lot of these molecules.
00:00:47.15 And the more and more that we look,
00:00:49.17 the more we realize that we've learned a lot
00:00:53.06 about how these molecules function
00:00:55.04 by going back and learning what they do in the immune system,
00:00:57.24 and I'll give you a couple examples of that today.
00:01:00.07 So, fundamentally, one of the questions that my lab
00:01:02.03 has been focused in on
00:01:04.03 is how it is that specific synapses,
00:01:07.13 the connections between neurons, are eliminated,
00:01:09.10 both in health and disease.
00:01:11.03 Now, during development, synapse elimination,
00:01:13.10 or synaptic pruning,
00:01:15.05 is a normal developmental process
00:01:17.05 in which excess synapses or excess connections are permanently pruned away,
00:01:21.02 while others get strengthened and maintained.
00:01:24.04 Now, this is illustrated here in a...
00:01:26.12 in a series of images
00:01:28.08 -- Golgi stains from the human cortex --
00:01:30.02 that just shows that there's a process
00:01:32.10 in which synapses are...
00:01:34.04 you know, there's an exuberant synaptic connectivity,
00:01:36.16 in this case at six years old, in the cortex,
00:01:39.03 but by 14 years old in the cortex,
00:01:41.06 you can see sort of a decrease in the connectivity.
00:01:43.28 This is what we would call pruning.
00:01:46.10 Now, this is not an all-or-nothing process.
00:01:48.23 This happens in different critical periods at different times.
00:01:51.05 In areas like the prefrontal cortex,
00:01:53.17 that are involved in executive function
00:01:55.17 and working memory,
00:01:57.12 they actually continue to prune pretty late
00:01:59.19 -- into late adolescence and early adulthood --
00:02:01.21 but areas like the visual system and sensory systems
00:02:04.10 tend to prune or remodel relatively early.
00:02:06.14 So, this is a key process
00:02:09.00 that's necessary for precise brain wiring,
00:02:10.29 and defects in this process of pruning
00:02:13.06 are thought to underlie neurodevelopmental
00:02:15.18 and neuropsychiatric disorders,
00:02:17.12 including schizophrenia and autism.
00:02:19.07 But it's been pretty hard to test in people,
00:02:21.03 in part because we don't have a way
00:02:23.18 or a biomarker to be able to go back
00:02:25.22 and study pruning in people over time,
00:02:27.25 which is why, fortunately,
00:02:31.08 we have animal models that have been able to provide us
00:02:33.10 with some of this insight.
00:02:35.05 In addition to the neurodevelopmental and neuropsychiatric disorders,
00:02:37.09 it's becoming increasingly clear that
00:02:40.05 synaptic loss and dysfunction is a hallmark of, unfortunately,
00:02:42.23 normal aging
00:02:45.05 but also of many CNS neurodegenerative diseases,
00:02:47.17 including Alzheimer's disease.
00:02:49.16 So, while each of these diseases listed here
00:02:52.13 is quite different with respect to the age of onset
00:02:55.14 and the clinical symptoms,
00:02:57.12 even the circuits that they affect,
00:02:59.15 could it be that there could be some common mechanisms
00:03:01.27 or common pathways that converge to enable...
00:03:05.00 to drive this synaptic vulnerability.
00:03:07.13 And this is the question at we've been really interested in in the lab,
00:03:10.22 which is why we think, or hypothesize,
00:03:13.04 that understanding how synapses are normally pruned
00:03:16.14 during development
00:03:18.07 could provide new insight.
00:03:20.03 Now, as I mentioned, synaptic elimination
00:03:22.26 is a normal developmental process.
00:03:24.18 It's been studied by many investigators over the years
00:03:27.00 in many different circuits.
00:03:28.25 This is just an oversimplification of what I mean when I say pruning.
00:03:31.20 This is a typical postsynaptic neuron,
00:03:33.29 initially innervated by lots of inputs or axons
00:03:36.29 that are relatively weak starting out,
00:03:38.28 and then, through this process called pruning,
00:03:41.14 you can see some of these inputs
00:03:43.23 -- the blue ones --
00:03:45.18 get pruned away while the red ones get strengthened and maintained.
00:03:48.16 This idea of use it or lose it...
00:03:51.03 this was coined by Carla Shatz...
00:03:53.27 and we know that this process is regulated by neuronal activity,
00:03:57.19 meaning the axons, or the inputs,
00:04:00.19 that are able to drive the postsynaptic cell more effectively
00:04:03.23 tend to win the competition
00:04:05.21 at the expense of its less active nearby neighbor,
00:04:07.25 in this case the blue one.
00:04:09.19 So, we know it's activity-dependent.
00:04:11.09 Much of the work in the field has largely focused, until recently,
00:04:14.20 on neuronal signals that regulate this process --
00:04:17.02 and there's a large number of those mechanisms
00:04:19.05 that I'm not gonna talk about today --
00:04:21.04 but more and more evidence over the last decade or so
00:04:24.11 is starting to implicate glia,
00:04:25.28 the non-neuronal cells of the brain,
00:04:28.00 including microglial cells.
00:04:30.12 And also astrocytes, which I'm not going to focus on today...
00:04:32.18 these are cells that are also critically important
00:04:35.02 in synapse development and function.
00:04:37.02 But today I'm gonna focus on microglia.
00:04:39.10 And what you can appreciate from this cartoon
00:04:42.06 is how dynamic these cells are in the brain.
00:04:45.16 So, microglia make up about 7-10%
00:04:47.29 of the cells in our brain.
00:04:49.18 And unlike other glia and other brain cells,
00:04:51.29 they're the only cell not born in our brains.
00:04:54.20 And so, fate mapping studies by Miriam Merad's lab
00:04:57.08 a number of years ago
00:04:59.15 showed that microglia actually derive from the neural...
00:05:02.05 from yolk sac progenitors --
00:05:04.13 myeloid cells
00:05:06.18 -- that enter the brain quite early in development,
00:05:09.04 and that these cells actually become
00:05:11.26 what we now know as microglia
00:05:13.21 when they're exposed to brain signals.
00:05:15.24 And when you can look into the brain
00:05:17.16 and image them,
00:05:19.24 which is what we've been doing and the field has been doing,
00:05:21.25 you can start to really appreciate how dynamic these cells are
00:05:24.02 and how much they're really interacting with synapses.
00:05:27.21 So, thanks to some tools that were developed by Steffen Jung's lab, initially,
00:05:32.03 that can use genetic engineering to label or to...
00:05:36.14 label microglia with a GFP...
00:05:39.16 this is a fractalkine receptor GFP mouse
00:05:41.17 crossed to a tdTomato mouse
00:05:44.14 that's enabled you to label the neurons in red...
00:05:47.00 this is a two-photon imaging experiment
00:05:49.12 by Jenelle Wallace in our lab.
00:05:52.18 And Jenelle is actually... as you can see here,
00:05:56.05 these microglia's processes
00:05:58.16 are dynamically associating with these synapses.
00:06:00.20 And the other thing to mention
00:06:03.24 is that microglia are particularly phagocytic during development,
00:06:06.15 and it looked, to our eye,
00:06:08.13 that microglia might be engulfing
00:06:10.29 some of these inputs.
00:06:12.23 And others have made similar observations over the years,
00:06:14.27 but we needed a way to be able to study this
00:06:16.25 and test this more directly.
00:06:18.16 And to address this question,
00:06:20.03 we decided to turn to a circuit
00:06:23.23 in which pruning has really been well studied:
00:06:26.10 the developing visual system of the mouse.
00:06:28.11 And in particular, we're studying the circuit
00:06:31.01 between the retina and the visual thalamus:
00:06:33.10 the retinogeniculate circuit.
00:06:35.10 Now, this is a circuit that's really been well studied
00:06:38.12 thanks to pioneering work by Carla Shatz
00:06:41.04 and many others in the field
00:06:43.00 -- Chinfei Chen --
00:06:45.16 where this circuit can be both labeled
00:06:46.28 and studied anatomically using tracers.
00:06:48.28 So, for example, we can put cholera toxin
00:06:51.04 or anterograde tracers
00:06:53.07 into the eyes of a mouse
00:06:56.11 that enables you to track the projections into the LGN here.
00:06:58.06 And what you can see is that initially,
00:07:00.02 early in development,
00:07:02.04 a postsynaptic neuron in the lateral geniculate nucleus
00:07:06.20 -- the relay station here in the thalamus --
00:07:09.01 is initially innervated by the axons
00:07:11.05 from both left and right eye retinal ganglion cells,
00:07:14.04 as you can see here.
00:07:16.01 So, they're binocularly innervated.
00:07:17.25 But over these next two weeks,
00:07:19.19 you can see the segregation, or this refinement,
00:07:22.09 such that, in the mature LGN,
00:07:24.04 there's no overlap,
00:07:26.17 and then there's a shift from binocular to monocular innervation.
00:07:30.01 Now, again, this has been well studied,
00:07:32.21 and we know that this process is regulated by neuronal activity.
00:07:35.01 So, the question then is,
00:07:37.07 are microglial cells playing a role in this refinement process?
00:07:41.15 And if so, what are the mechanisms
00:07:43.17 that are regulating it?
00:07:44.12 So, if you look in the LGN
00:07:46.27 in a mouse in which the inputs have been labeled
00:07:49.09 with cholera toxin,
00:07:51.02 what you can see is a microglia -- shown here zoomed in --
00:07:53.27 interacting very intimately
00:07:56.20 with the axons from both the left and the right eye.
00:07:58.15 So, the question was, are they actually phagocytosing or engulfing
00:08:01.25 these inputs?
00:08:03.12 So, to address this question,
00:08:05.13 Dori Schafer in my lab actually did an experiment
00:08:08.04 where she measured the engulfment,
00:08:11.03 did some 3D imaging,
00:08:13.04 and could then remove the inputs
00:08:15.10 that weren't within the microglia.
00:08:16.27 And through 3D rendering... surface rendering,
00:08:19.03 what we observed is that almost every microglia
00:08:22.05 throughout the LGN
00:08:23.26 during this process of eye segregation and pruning
00:08:26.05 were engulfing both left and right eye inputs.
00:08:29.15 And you can even have... we showed evidence of that, actually,
00:08:32.24 engulfment into their lysosomes as well.
00:08:35.13 And more interestingly, even by EM,
00:08:37.17 we can show that they seem to be sort of nibbling off
00:08:40.23 or phagocytosing
00:08:42.13 the presynaptic terminals of these retinoganglion cell axons.
00:08:44.28 And it's happening in a developmentally regulated way,
00:08:47.08 meaning between postnatal day 5
00:08:50.05 and about postnatal day 10
00:08:51.25 is when we see a lot of this engulfment.
00:08:53.21 But after about P10, this really diminishes significantly.
00:08:57.26 Now, the big question now,
00:08:59.27 since I told you already that this process of pruning,
00:09:02.00 especially in this circuit, is activity-dependent,
00:09:05.13 we wondered, could microglia be responding
00:09:08.22 to activity-dependent cues?
00:09:10.20 Could they be engulfing less active inputs preferentially,
00:09:13.13 for example?
00:09:15.07 So, to address this question,
00:09:17.15 again Dori labeled the inputs with cholera toxin, shown here in left...
00:09:21.00 in red and blue.
00:09:22.25 And then she used tetrodotoxin,
00:09:25.06 which blocks sodium-dependent action potentials...
00:09:27.06 injected that into one eye to drive this activity-dependent competition,
00:09:29.25 and then asked the question,
00:09:32.11 do microglia preferentially engulf the...
00:09:34.21 the weaker, or the less active, inputs?
00:09:36.29 And she did the experiments a couple of different ways,
00:09:39.13 but in both cases we saw evidence of an activity-dependent
00:09:43.16 or preferential engulfment of the less active eye inputs.
00:09:46.15 Now, this was an important experiment
00:09:48.27 because it shows, or it suggests,
00:09:50.21 that microglia aren't just acting as the cleanup crew here
00:09:53.14 but that they were actually responding to
00:09:56.08 or were being instructed by some sort of activity-dependent signals.
00:10:00.04 And then the question, then, was, what are those signals?
00:10:03.06 We think there are probably many signals,
00:10:05.05 but I'm gonna tell you about one of them,
00:10:07.09 at least initially.
00:10:08.28 So, thinking about how this might work
00:10:11.09 in the immune system
00:10:12.25 raised the question of, could there be, actually,
00:10:16.02 cues that were... that were instructing microglia
00:10:18.11 to engulf the less active inputs,
00:10:20.10 and what might those cues be?
00:10:22.10 So, when I was a postdoctoral fellow in...
00:10:24.25 with Ben Barres at Stanford,
00:10:27.16 we unexpectedly discovered a role for the classical complement cascade.
00:10:31.17 These is a group of innate immune molecules
00:10:35.23 traditionally associated with clearance of pathogens
00:10:37.24 or debris in the periphery.
00:10:39.15 We showed that some of these molecules,
00:10:41.15 including C1q and, downstream, C3...
00:10:43.18 this is a part of the classical complement cascade...
00:10:46.01 that these molecules, both at the protein and the RNA level,
00:10:48.26 were expressed in the healthy developing brain
00:10:51.06 and, quite interestingly,
00:10:53.26 subsets of synapses in the LGN and the visual system,
00:10:56.27 but also throughout the brain,
00:10:58.17 were colocalizing with these complement molecules.
00:11:02.17 And what we went on to show...
00:11:04.29 and this is just shown here in this high-resolution imaging...
00:11:06.28 you can see subsets of synapses
00:11:09.07 labeled here in blue and red...
00:11:11.01 these are pre and postsynaptic markers...
00:11:12.25 you can see subsets of these synapses
00:11:15.01 were also colocalizing with C1q, in green.
00:11:17.17 And we went on to show that if you blocked
00:11:20.07 -- or genetically blocked using knockout mice --
00:11:23.01 either the complement molecule C1q or C3,
00:11:26.14 or the receptor on microglia...
00:11:28.13 and it turns out that the only resident cell in the brain
00:11:31.21 to express complement receptors in the healthy brain,
00:11:34.03 at least, is microglia...
00:11:35.29 when we blocked either the receptor or the molecules,
00:11:38.15 these mice had defects in eye-specific segregation
00:11:42.00
00:11:43.19 or synaptic refinement, and microglia were only about half as good
00:11:46.02 at eating these synapses,
00:11:48.00 which identified complement as one of the molecules,
00:11:50.10 or the instructive signals,
00:11:51.27 that was regulating this process of pruning.
00:11:54.00 So, this raised a number of questions
00:11:55.19 and also proposed a model that we're still working on,
00:11:59.04 but I wanted to take you through, kind of,
00:12:01.06 what we're thinking about at this stage.
00:12:03.02 You can imagine a system where,
00:12:05.09 during development, some of these inputs
00:12:07.03 -- in this case, the blue ones --
00:12:08.26 are targeted or tagged by complement molecules.
00:12:11.09 And that that's an instructive signal
00:12:13.25 that's telling microglia, engulf those inputs.
00:12:16.29 But it also raises the question of why microglia
00:12:19.18 aren't engulfing the red inputs,
00:12:21.13 or for that matter the whole axon.
00:12:23.02 Why might there be specificity?
00:12:25.03 And what molecules could be regulating this process?
00:12:27.19 Could there, for example,
00:12:29.19 be molecules that are also saying to the microglia,
00:12:31.17 don't eat me, or are essentially protective molecules
00:12:33.29 that prevent microglia from engulfing.
00:12:36.15 So, again, we want back and we learned from the immunologists,
00:12:39.10 and we thought, well,
00:12:41.15 what molecules like this actually prevent phagocytosis in the immune system?
00:12:46.12 And it turns out that there's a large number of these molecules
00:12:48.15 in the immune system
00:12:50.06 that instruct macrophages not to engulf.
00:12:52.14 These are both complement inhibitors
00:12:54.15 -- there's a whole group of these molecules
00:12:56.10 that regulate the classical complement cascade at different levels --
00:12:59.22 but there's also a group of molecules
00:13:02.09 not so creatively called "don't eat me" signals
00:13:04.18 in the immune system.
00:13:06.14 And so, Emily Lehrman, a graduate student in my lab,
00:13:08.15 decided to ask, well,
00:13:11.03 are some of these molecules, especially the "don't eat me" signals,
00:13:12.27 actually expressed in the developing healthy brain?
00:13:15.11 And could they actually be some of the signals
00:13:17.06 that are working in concert with complement to regulate pruning?
00:13:20.03 So, in the immune system, as I mentioned,
00:13:23.23 a healthy cell, or a self cell,
00:13:25.19 has a large number of these "don't eat me" signals
00:13:27.20 that label their surfaces.
00:13:29.19 And essentially, they act as stop signals
00:13:31.17 for circulating macrophages,
00:13:33.11 which express receptors that recognize these "don't eat me" signals.
00:13:36.10 Essentially, they're stop signals,
00:13:38.15 or sort of protective molecules,
00:13:40.25 that even if complement was all over the cell,
00:13:44.05 it would prevent phagocytosis.
00:13:46.20 In the context of disease or apoptosis,
00:13:48.09 it turns out a lot of these "don't eat me" signals
00:13:50.23 become downregulated, and then molecules like complement
00:13:53.17 then can enable the engulfment of those...
00:13:55.20 of those either apoptotic or non-self cells.
00:13:58.17 Now, could similar molecules
00:14:00.28 actually be labeling synapses?
00:14:02.15 And so, Emily in particular focused on a molecule
00:14:04.27 called CD47,
00:14:07.16 which is a well-studied "don't eat me" signal
00:14:09.10 in the immune system.
00:14:10.29 And one of the reasons why she did that
00:14:13.05 is because it turns out that CD47
00:14:15.10 was highly expressed, or is highly expressed,
00:14:17.05 in retinoganglion cells,
00:14:18.21 but in many neurons throughout normal healthy development
00:14:20.27 and into adulthood.
00:14:22.22 And microglia express a receptor
00:14:25.06 called SIRP alpha at high levels,
00:14:27.03 especially, again, during these developmental pruning periods.
00:14:30.13 So, just as an example...
00:14:32.05 I just wanted to sort of give you a sense of some of this work,
00:14:35.21 still ongoing...
00:14:38.02 when we looked at mice that lacked either CD47
00:14:42.13 or the receptor SIRP alpha,
00:14:44.25 these actually... these microglia were overpruning.
00:14:49.03 So, in other words, they were engulfing too many synapses
00:14:51.22 or the wrong number of synapses.
00:14:53.21 And so, loss of function of either the molecule CD47
00:14:56.29 or the receptor on microglia
00:14:59.24 led to increased engulfment of or by microglia in the...
00:15:02.22 in the same system I just told you about, the LGN.
00:15:05.29 And these mice also had enhanced pruning, excess...
00:15:09.23 fewer numbers of synapses,
00:15:11.16 and we actually think they also have some behavioral defects as well.
00:15:15.10 So, we have much to do to try to understand mechanism,
00:15:17.22 but this is just another example
00:15:20.06 of a molecule in this "don't eat me" signal family
00:15:23.11 that we think might be acting in concert
00:15:25.18 with complement
00:15:27.10 to give specificity to this circuit.
00:15:28.28 So, if you look at this story put together,
00:15:31.04 you can imagine during this activity-dependent remodeling process,
00:15:34.20 the working model is that microglia
00:15:37.21 have a lot of these receptors,
00:15:39.12 and because they're dynamically surveying the environment,
00:15:41.19 they can essentially read out the combination of cues
00:15:45.14 on those inputs or synapses.
00:15:47.18 So, synapses that have complement
00:15:50.09 but lack the "don't eat me" signals
00:15:51.27 might be preferentially eliminated.
00:15:53.21 And work now in the lab
00:15:55.21 is designing tools and new model systems
00:15:58.06 to be able to interrogate this
00:16:00.04 and even be able to watch these molecules
00:16:02.03 interacting in real time
00:16:04.08 during this process of activity-dependent pruning.
00:16:06.06 This also begs the question,
00:16:08.02 what is complement binding to at synapses?
00:16:10.01 These are secreted molecules,
00:16:12.02 but yet we're seeing some sort of specificity
00:16:14.28 in terms of some synapses that are labeled with complement
00:16:19.08 and others that are not.
00:16:21.00 What are the receptors or binding proteins
00:16:23.08 that might recruit complement to those synapses but not others?
00:16:25.09 And again, how does activity regulate this process
00:16:28.01 of complement binding
00:16:30.00 or complement expression in the brain?
00:16:31.21 What are some of the upstream signals in neurons
00:16:35.04 that are leading to this process?
00:16:36.23 And could the same mechanisms
00:16:39.11 that I just told you about today...
00:16:41.04 could they become dysregulated
00:16:43.01 in development or in the adult brain?
00:16:45.01 And could this lead to pathological synapse loss,
00:16:47.00 which again is a hallmark of a lot of these disorders
00:16:49.17 that I... that I mentioned earlier?
00:16:52.05 And so, this sort of suggests a model
00:16:54.01 that maybe some of the same molecules and mechanisms,
00:16:57.05 especially relating to the immune system,
00:17:00.04 that are normally helping to prune synapses
00:17:03.05 during normal development
00:17:05.05 could become dysregulated,
00:17:06.29 either through genetic or environmental perturbations,
00:17:09.29 that could then lead to
00:17:13.09 aberrant or pathological synapse loss.
00:17:15.00 Now, this is a hypothesis,
00:17:17.04 but work over the last many years by my lab,
00:17:18.27 and now many other labs,
00:17:20.21 has provided some evidence to support this,
00:17:23.01 at least in animal models -- in Alzheimer's disease, glaucoma,
00:17:25.26 frontal temporal dementia models, and others.
00:17:29.22 Now, I'm gonna tell you, now, in the second part of my talk,
00:17:32.03 about some work to support this hypothesis
00:17:34.26 in Alzheimer's disease.
00:17:36.10 And there's a number of reasons why we decided
00:17:38.09 to focus on Alzheimer's disease first.
00:17:40.06 One is because of the fact that there's already evidence from humans,
00:17:44.01 and in animal models,
00:17:45.19 that synapse loss does happen,
00:17:47.15 and it's a relatively early event.
00:17:49.22 In fact, it's predicted that synaptic dysfunction
00:17:52.01 and synapse loss
00:17:53.25 is happening in the preclinical phase of Alzheimer's disease,
00:17:56.01 and there's evidence to suggest
00:17:58.27 this is also happening in various models of Alzheimer's disease as well.
00:18:02.09 So, synapse loss is also the strongest correlate
00:18:05.17 of cognitive decline in Alzheimer's disease,
00:18:07.14 yet we still don't have a lot of work in the field,
00:18:10.02 or evidence,
00:18:12.02 that really points to specific mechanisms
00:18:15.05 that drive this synaptic vulnerability.
00:18:17.02 So, understanding that mechanism
00:18:18.26 I think is gonna be critically important,
00:18:20.14 and we wondered, could some of the same pathways
00:18:22.27 we've been studying during development
00:18:24.20 be playing a role in this early synapse loss.
00:18:27.01 The other thing to support a potential role for complement...
00:18:31.06 if you look in the literature,
00:18:32.26 it's long known the complement is upregulated in Alzheimer's,
00:18:35.24 in humans and in animal models.
00:18:37.22 In fact, you can see from these studies,
00:18:39.18 complement not only becomes upregulated
00:18:43.11 but it also colocalizes with amyloid plaques in the...
00:18:46.08 in these brains,
00:18:48.03 so that's been well known.
00:18:49.26 But it's always been thought to be sort of in late stages
00:18:52.00 and maybe secondary to neuroinflammation
00:18:54.16 and disease and damage.
00:18:56.09 And we also know, of course,
00:18:58.08 that microglia play critical roles in Alzheimer's disease.
00:19:01.08 They are not only important in neuroinflammatory events,
00:19:04.17 but they're also important in phagocytosing,
00:19:08.00 or clearing, those amyloid plaques.
00:19:10.04 So, the question is, you know,
00:19:12.15 putting all this together,
00:19:14.01 could microglia be actually mediating
00:19:18.06 some of this synapse loss
00:19:20.00 through some of these mechanisms,
00:19:21.18 including the complement cascade?
00:19:23.04 Now, it's well known that microglia
00:19:25.19 dramatically alter both their morphology and function
00:19:27.19 in the context of diseases like AD.
00:19:29.25 They go from this really beautifully ramified microglia
00:19:33.12 that I told you about in the first part of my talk,
00:19:35.10 and they become quite different,
00:19:37.11 and, quote unquote, "activated".
00:19:39.13 They increase a lot of phagocytic receptors.
00:19:42.10 They become more lysosomal --
00:19:45.00 there's more lysosomal activity.
00:19:46.25 But the problem with this sort of shift in morphology
00:19:49.02 is it really doesn't tell us when these cells changed
00:19:52.09 or, at least by morphology
00:19:54.19 and with this handful of markers,
00:19:56.12 whether they're actually contributing
00:19:59.01 to the detrimental or specific functions in Alzheimer's disease.
00:20:02.08 Now, it's been thought for the most part, until recently,
00:20:06.20 that microglia dysfunction
00:20:08.22 was sort of a secondary event.
00:20:10.20 But emerging genetic studies,
00:20:13.05 especially in late-onset Alzheimer's disease,
00:20:15.12 more and more are implicating microglia and myeloid cells
00:20:19.12 as being key drivers in Alzheimer's disease,
00:20:22.24 at least late-onset Alzheimer's disease.
00:20:25.15 In fact, almost half of the late-onset AD risk genes,
00:20:28.19 which are over here
00:20:30.23 -- a lot of these are common variants --
00:20:32.24 are actually expressed in or enriched in myeloid cells.
00:20:35.22 And even APOE,
00:20:37.29 which we know is a critical risk factor in Alzheimer's disease
00:20:40.03 is indeed upregulated and expressed in microglia,
00:20:42.24 both in AD... human AD brains
00:20:45.18 as well as mouse models.
00:20:47.07 So, putting all of this together,
00:20:50.01 this is a sort of chance to change the way we've been thinking about microglia
00:20:52.08 in the context of Alzheimer's disease.
00:20:54.06 And so, the big question that we've also been focused on is,
00:20:57.19 do microglia contribute to synaptic and cognitive impairment
00:21:01.10 in Alzheimer's?
00:21:02.23 And if so, how?
00:21:04.06 What are the mechanisms?
00:21:06.00 So, going back to what I told you about in the first part,
00:21:08.04 we specifically asked,
00:21:10.15 could microglia and the complement cascade
00:21:13.12 be contributing to this early synapse loss?
00:21:16.07 And this is actually important because,
00:21:18.11 at least in animal models,
00:21:19.28 unlike in humans,
00:21:21.10 you can kind of rewind the clock to the beginning,
00:21:23.11 before this pathology,
00:21:24.26 and you can ask, when do these things start to change?
00:21:27.01 And there are certain parts of the brain,
00:21:29.11 including the hippocampus,
00:21:31.12 that are... that are vulnerable,
00:21:33.05 and they're vulnerable to early synapse loss.
00:21:34.25 So, putting this together,
00:21:36.24 my postdoctoral fellow at the time,
00:21:38.19 now a PI of her own lab at UCL,
00:21:40.23 Soyon Hong,
00:21:42.23 in collaboration with Dennis Selkoe and Cynthia Lemere...
00:21:44.22 we set out to ask the following question:
00:21:47.10 is early synapse loss mediated by microglia...
00:21:50.23 aberrant reactivation of this microglia pruning?
00:21:52.28 And so, using established genetic Alzheimer's mouse models,
00:21:56.10 we set out to test this question.
00:21:58.14 And the first thing that Soyon did
00:22:00.17 is she looked in the brains of these Alzheimer's mice,
00:22:02.23 asked, you know,
00:22:05.02 how soon does synapse loss happen?
00:22:07.00 Established that window.
00:22:08.23 And she showed, using both anatomical and immunohistochemical approaches,
00:22:11.13 and high-resolution imaging,
00:22:13.08 by staining these mouse brains
00:22:15.27 that are Alzheimer's versus control
00:22:19.17 with antibodies for pre and postsynaptic markers,
00:22:21.27 in red and green,
00:22:23.24 you can... you can... you can actually quantify the number of synapses this way.
00:22:26.14 And in doing so,
00:22:28.23 she found that there's a quite an early loss of synapses
00:22:30.23 that happened prior to amyloid deposition
00:22:33.15 in these amyloid or Alzheimer's mouse models.
00:22:37.02 And moreover, when she looked at the hippocampus
00:22:40.00 and the frontal cortex of these mice,
00:22:41.27 there was a significant increase
00:22:44.09 in the expression and localization of complement proteins,
00:22:46.04 including C1q.
00:22:47.23 What you can hopefully see here...
00:22:49.19 that if you zoom in on the hippocampus,
00:22:51.14 there's an increased expression and localization of C1q,
00:22:54.23 shown here in green.
00:22:56.18 And interestingly, this was happening, in at least early stages,
00:23:00.04 in vulnerable brain regions like the hippocampus,
00:23:02.09 where we didn't really see this in the cerebellum
00:23:04.16 or the striatum, for example.
00:23:06.08 And this is just quantified here.
00:23:07.28 So, this is interesting because it's showing
00:23:10.05 an early increase in complement expression and localization
00:23:14.18 to vulnerable brain regions
00:23:16.18 in two different amyloid mouse models.
00:23:18.11 Moreover, if we zoom in and do high-resolution imaging,
00:23:20.10 much the way we do in development,
00:23:22.04 Soyon observed that a large number of these synapses
00:23:25.20 in the hippocampus
00:23:27.25 were colocalizing with these complement proteins,
00:23:30.05 C1q and C3...
00:23:32.05 as shown here.
00:23:34.07 So, what are the consequences
00:23:37.04 of having complement binding to synapses?
00:23:38.16 Is this... is this detrimental or beneficial?
00:23:40.13 We hypothesized that blocking complement
00:23:42.22 or removing C1q or C3
00:23:45.26 in these animal models
00:23:47.22 could be protective of synapse loss and cognitive decline.
00:23:49.20 So, to address this question,
00:23:51.07 we crossed amyloid or AD mouse models
00:23:54.18 to C3 knockout mice
00:23:57.10 -- so, these are mice that lack that downstream protein, C3,
00:24:00.10 that tags the synapses --
00:24:02.01 and then we compared synapse loss
00:24:04.07 and cognitive function
00:24:06.05 in these different genotypes:
00:24:07.28 those APP models that have C3
00:24:09.21 and those that lack C3.
00:24:11.15 And what we found is more synapses in the mice
00:24:13.28 that lack C3.
00:24:15.20 This was actually in 3-4-month-old animals,
00:24:18.11 as shown here,
00:24:20.13 but quite interestingly,
00:24:22.17 when we then looked many months later
00:24:24.19 -- so, this is just showing significantly more synapses
00:24:27.01 quantified the way I told you before,
00:24:29.03 by looking at colocalized PSD95 and synapsin --
00:24:32.22 we could see significantly more synapses
00:24:35.02 in the C3-deficient APP presenilin mice.
00:24:39.27 And even in the aged AD mice,
00:24:43.04 after a year,
00:24:45.01 we could still see significant more...
00:24:46.23 significantly more synapses,
00:24:48.28 which is important because it's saying
00:24:50.09 that this is happening early,
00:24:52.07 and it's also being observed at late stages of the disease.
00:24:54.12 So, what about functional and behavioral consequences?
00:24:58.23 So, in collaboration with Cynthia Lemere's lab,
00:25:01.06 she carried out some behavioral and cognitive testing
00:25:04.26 in these mouse models,
00:25:06.16 and showed, interestingly and importantly,
00:25:08.20 that C3 deficiency from these amyloid models
00:25:11.10 was protective against cognitive defects,
00:25:14.11 despite the fact that these animals had more amyloid --
00:25:19.02 increased amyloid plaques.
00:25:20.28 And I think this is a really important point
00:25:22.16 because it's saying that...
00:25:24.08 and it's consistent with the fact,
00:25:26.10 that complement is involved in plaque clearance...
00:25:29.25 that's been shown by others in the field...
00:25:31.21 this is saying that even though these mice
00:25:34.05 have more plaques, they're protective...
00:25:36.01 they're protected from some of their cognitive impairment,
00:25:38.27 and that also relates to their protection
00:25:41.17 with synapses.
00:25:43.08 So, this is sort of suggesting two different processes, here,
00:25:45.25 that may be independent.
00:25:47.23 It's also raising the question that this...
00:25:50.19 if you're starting to think about when
00:25:54.22 you would want to target some of these pathways...
00:25:56.15 that context matters.
00:25:58.01 So, early on in the process,
00:26:00.01 blocking complement, especially before you start to get plaques,
00:26:02.04 seems to be beneficial.
00:26:05.01 But then we also know that later stages in the disease,
00:26:09.01 when there's a lot of amyloid plaques,
00:26:11.07 you know, you want to actually have that complement
00:26:13.20 and microglia clearing the plaques.
00:26:14.27 And this is really, now,
00:26:17.03 starting to... wanting us to dive deeper into understanding
00:26:19.14 what the mechanisms are
00:26:22.04 that regulate this process at different stages.
00:26:23.26 And in particular,
00:26:25.00 we want to understand,
00:26:26.12 even more so than ever before,
00:26:28.16 what complement is binding to at these synapses,
00:26:31.06 and we want to know how complement
00:26:33.20 is interacting with molecules like amyloid and activin
00:26:37.07 to regulate this process.
00:26:38.26 Now, the question is,
00:26:40.21 is microglia part of the story?
00:26:42.07 Is microglia mediating the synapse loss?
00:26:43.25 And much like we showed in development,
00:26:45.24 what Soyon showed is if we block the receptor on microglia,
00:26:49.16 this complement receptor,
00:26:51.28 we also protected the synapse loss
00:26:54.18 and also showed some benefits in this model as well.
00:26:57.06 So, this really together suggests that the same pathway
00:27:00.18 that's normally regulating synaptic pruning during development
00:27:03.17 can become aberrantly activated
00:27:05.16 to contribute to aberrant or pathological synapse loss,
00:27:08.04 at least in animal models,
00:27:10.09 and we're now looking in human Alzheimer's brain,
00:27:13.07 especially at the earlier stages,
00:27:15.00 to see if we can see evidence of changes
00:27:17.07 in some of these pathways,
00:27:18.23 both in the brains
00:27:20.14 and in the cerebral spinal fluid, or CSF,
00:27:22.20 of Alzheimer's patients
00:27:24.15 to try to see if we can validate this mechanism
00:27:27.02 in humans.
00:27:28.26 But clearly, microglia have complex and diverse roles.
00:27:30.28 I told you only about one of them today,
00:27:32.14 which is synaptic pruning,
00:27:34.00 and I largely focused on one pathway
00:27:35.17 -- complement --
00:27:36.28 but we know from the work from many in the field
00:27:40.02 that many different mechanisms and signals
00:27:42.14 are likely mediating different aspects of this process.
00:27:44.05 And of course, while that's quite interesting biologically,
00:27:47.11 it also is a challenge for starting to think about
00:27:51.11 how we might target both microglia
00:27:53.08 and the complement system,
00:27:54.21 and some of these molecules,
00:27:56.19 in different stages of disease progression.
00:27:58.08 So, this really goes back to the question
00:28:00.17 I led with earlier,
00:28:02.11 which is,
00:28:05.06 how and when do microglia contribute to Alzheimer's disease,
00:28:07.02 especially now that we know, from emerging genetics,
00:28:08.26 that microglia really are
00:28:11.10 not just responding to neuronal dysfunction,
00:28:13.02 but they may also be drivers in the disease?
00:28:16.25 And even though much of the field
00:28:18.29 is focused on the plaques and amyloid,
00:28:20.13 which is critically important,
00:28:22.07 the idea that these genes that have been identified by genetics...
00:28:24.24 many of which,
00:28:27.17 we don't even know their normal biology, their normal function...
00:28:30.17 we need to better understand how these cells
00:28:33.04 and these pathways
00:28:34.24 are normally functioning in microglia
00:28:36.10 before we can really understand how it goes awry in disease.
00:28:39.08 And again, this is really...
00:28:41.21 brings back the fact that it's not as simple as it seems,
00:28:45.17 because microglia
00:28:48.05 actually have many and diverse roles,
00:28:51.16 both in normal health, normal development,
00:28:53.18 and in disease.
00:28:55.09 And more and more evidence is suggesting
00:28:58.00 that they exist in different states
00:29:00.01 throughout a lifespan of a mouse.
00:29:01.27 I just told you about one function -- pruning --
00:29:05.04 but microglia also play roles in synaptic plasticity,
00:29:08.15 they're critically important in synapse function,
00:29:11.04 but also other functions in the brain.
00:29:13.17 And we also know that when there's a challenge,
00:29:15.19 whether that's a local challenge or a larger...
00:29:18.15 more large-scale genetic challenge,
00:29:20.20 microglia can change state.
00:29:23.08 They can go from this homeostatic, surveying microglia
00:29:26.18 that I just told you about in the first part of my talk,
00:29:28.15 and they can shift into this pathological state.
00:29:32.04 For ex... they can also move into an inflammatory state,
00:29:35.24 but we appreciate that inflammation
00:29:38.08 can be both detrimental and beneficial,
00:29:40.14 depending on context.
00:29:42.22 They can shift into a more phagocytic state,
00:29:44.12 but that again can be both beneficial,
00:29:46.26 in the context of clearing a plaque,
00:29:48.27 or detrimental, in the context of clearing a synapse aberrantly.
00:29:52.19 So, how do we get at this question?
00:29:54.29 And a challenge has been that the field
00:29:57.11 still lacks tools that enable us
00:30:00.18 to both track these state changes,
00:30:02.11 especially across disease or development,
00:30:04.19 and manipulate and target specific populations
00:30:08.01 or states of microglia,
00:30:09.18 which is ultimately what we want to be able to do.
00:30:12.09 So, how do we go about this?
00:30:14.05 And I think this is where emerging technology,
00:30:17.01 including single-cell sequencing,
00:30:18.22 has really opened up the field,
00:30:21.03 especially when you combine that with genetics,
00:30:23.12 which is now kind of laying the groundwork
00:30:26.12 for some of the pathways and genes
00:30:28.15 that we can start to focus in on.
00:30:30.18 And pioneering work that began
00:30:33.04 with Ido Amit's first paper
00:30:35.22 that applied single-cell sequencing to Alzheimer's mouse models
00:30:38.24 and human brain
00:30:40.15 showed that, indeed, when you sequence
00:30:42.23 and break the cells into individual cells
00:30:44.12 and readout RNA using single-cell sequencing,
00:30:47.03 there is indeed specific disease-associated populations,
00:30:50.16 or states, of microglia
00:30:52.29 that can be observed in the brains of Alzheimer's patients
00:30:56.03 and in Alzheimer's mouse models,
00:30:58.16 but even beyond Alzheimer's,
00:31:00.06 there's evidence that this is happening
00:31:02.21 across multiple diseases.
00:31:04.15 And so, this is indicating and identifying
00:31:07.26 markers or signatures that denote disease-associated states in microglia.
00:31:12.16 So, the question now is,
00:31:14.13 how do we understand what these states are doing?
00:31:16.15 One of these disease-associated states,
00:31:18.13 based on the markers that came out of the single cell,
00:31:20.21 could be used to spatially map these populations,
00:31:23.13 identify where in the brain they exist.
00:31:25.09 Many of them are actually localizing
00:31:28.12 to the microglia that surround the plaques,
00:31:30.04 which have been observed anatomically for years,
00:31:32.15 but now we actually have markers
00:31:34.11 that can actually identify those cells.
00:31:36.10 And of course, this also opens up
00:31:39.14 a lot of critical questions in the field.
00:31:41.21 Which is, how dynamic are these states?
00:31:43.24 Are they reversible, for example?
00:31:45.21 Once a microglia shifts
00:31:48.16 into this disease-associated state,
00:31:50.02 do they have the ability to shift back?
00:31:51.28 How plastic are they?
00:31:53.12 What functions do each of these different populations have?
00:31:56.20 How do we link changes in transcriptional state
00:32:00.17 to specific functions of microglia?
00:32:02.16 Can we use the information gleaned from single-cell sequencing
00:32:04.25 to develop new tools and probes
00:32:07.02 to be able to track and target specific populations,
00:32:10.01 especially non-invasively,
00:32:11.21 where you can imagine trying to track this in...
00:32:14.00 over time, across disease progression?
00:32:17.03 What's the impact of environmental challenge, or aging,
00:32:21.01 in the context of microglia state changes?
00:32:23.28 And of course, what is the impact
00:32:26.05 of specific genetic variants or mutations
00:32:28.08 that are being identified?
00:32:29.25 And although I focused today on Alzheimer's disease,
00:32:34.03 more and more work is implicating microglia
00:32:35.23 in other neurodegenerative diseases as well,
00:32:37.26 including Parkinson's disease, ALS, and others.
00:32:39.20 So, really, we're in a position
00:32:41.08 -- and it's an exciting time in the field --
00:32:44.08 where there probably are, now,
00:32:46.09 many potential paths to therapeutics
00:32:48.22 that we didn't have before.
00:32:50.09 The old approach might be,
00:32:52.00 oh, let's just block inflammation,
00:32:53.26 or let's block all microglia.
00:32:55.11 Clearly, that's not going to work,
00:32:57.09 because there's lots of diversity in states,
00:32:59.10 so we really want to start to think about
00:33:01.16 how to target the detrimental but not the beneficial states in microglia.
00:33:04.10 And the idea that you can use
00:33:07.08 some of the single-cell sequencing
00:33:08.28 that is identifying these markers
00:33:10.28 we could use to then target, and then manipulate,
00:33:13.29 just those populations,
00:33:16.05 and then hopefully be able to shift them back
00:33:18.09 into a more beneficial state.
00:33:19.28 So, there's much to do in the field.
00:33:21.18 And I wanted to just end with the fact
00:33:24.10 that we also have really lacked biomarkers
00:33:27.25 for microglia.
00:33:29.08 And in particular,
00:33:31.15 I think it would be a real game-changer
00:33:33.19 if we had ways to be able to non-invasively
00:33:36.28 track microglia changes in the brain
00:33:38.21 across disease progression.
00:33:40.14 There are some PET ligands or imaging biomarkers
00:33:42.19 that are suggestive of microglia.
00:33:44.27 They seem to be more a marker of inflammation
00:33:47.19 of some sort --
00:33:50.00 this TSPO PET ligand for example --
00:33:51.24 but it's not specific enough for microglia.
00:33:54.27 So, perhaps some of the molecules
00:33:58.12 and markers that we're identifying in the single-cell data
00:34:00.27 might identify new markers that we could use
00:34:03.17 to develop new PET ligands.
00:34:05.18 And we specifically want to start to develop biomarkers
00:34:08.09 --- both imaging and also cerebral spinal fluid,
00:34:11.18 or CSF, biomarkers --
00:34:13.16 that can enable us to be able to track microglia
00:34:16.05 and myeloid cell state changes
00:34:18.19 across disease progression,
00:34:20.09 especially early, when synapses and cognitive impairment
00:34:22.28 is beginning.
00:34:24.27 So, that's an example of how understanding
00:34:27.12 really basic functions of microglia,
00:34:29.05 that for us began in development,
00:34:31.07 is providing new insight into synaptic vulnerability
00:34:34.14 in the context of disease.
00:34:35.28 And how, then, using unbiased
00:34:38.24 and other new technologies and tools
00:34:40.19 could then enable us to try to better understand
00:34:42.23 and validate these mechanisms
00:34:46.02 across multiple species, including human.
00:34:47.20 And so, I'd like to end by thanking
00:34:49.25 all of the amazing collaborators and lab members,
00:34:51.12 both past and present,
00:34:52.28 some key collaborators,
00:34:54.18 generous funding,
00:34:56.07 and I want to end with my human microglia
00:34:58.29 in my lab
00:35:00.19 that have really made a lot of the work I told you about possible.
00:35:02.18 So, thanks very much.

Videos in this Talk

Part 1: Microglia States in Health and Disease
Audience:
- Student
- Researcher
- Educators
- Educators of Adv. Undergrad / Grad
Part 2: How Microglia Sculpt Brain Circuitry in Health and Disease
Audience:
- Student
- Researcher
- Educators
- Educators of Adv. Undergrad / Grad

Speaker: Beth Stevens

Total Duration: 0:57:02

Recorded: February 2020

All Talks in Neuroscience

Talk Overview

Microglia are the primary immune cells in the central nervous system. In the brain, they play central roles in proper development and function, as well as dysfunction and disease. In her first talk, Dr. Beth Stevens provides an overview of the many ways microglia cells operate, and how they can both harm and protect the brain. Fairly recent advances in the study of microglia through imaging have allowed researchers to identify different microglia states and study their dynamic roles at different stages of development.

In her second second talk, Dr. Stevens dives deeper into the mechanisms that allow microglia to shape the network of connections between neurons in the brain. She provides an introduction to the role of microglia in synaptic pruning, the process of eliminating extra synapses in healthy developing brains. She then goes on to explain how the reactivation of this process affects aging and diseased brains.

Speaker Bio

Beth Stevens

Beth Stevens is a neurobiologist at Boston Children’s Hospital, the Broad Institute, and the Howard Hughes Medical Institute. Her lab studies neuro-immune interactions in developing and diseased brains. She has made notable discoveries about the role of microglia cells in developmental disorders and neurodegenerative disease. Stevens earned her B.S. from Northwestern University, her Ph.D. from… Continue Reading