MAPPING THE PARKINSONS BRAIN IN SPACE AND TIME

Transcript

• 00:00It's my pleasure to introduce
• 00:03doctor Sanjun Dong. He's an
• 00:05associate professor in neurology
• 00:07and in,
• 00:08biomedical
• 00:09informatics and data science.
• 00:12He he I was fortunate
• 00:13to recruit
• 00:14him from Harvard where he
• 00:16was directing,
• 00:17the genomics and bioinformatics
• 00:19hub,
• 00:20and, I'm thrilled that you're
• 00:23here. Actually, your lab is
• 00:26now fifteen people already or
• 00:27something. It's really, amazing, and
• 00:29I can't can't wait
• 00:31to to hear more about
• 00:32your work. Thank you, Clemens.
• 00:35So honored to be here.
• 00:36Last time I was so
• 00:37excited and nervous was when
• 00:39I was giving a speech
• 00:39on my wedding, actually.
• 00:41But, let's see. So,
• 00:45good afternoon, everybody
• 00:46everybody. And, at first, I'd
• 00:48like to, welcome and, of
• 00:50course, congratulations to Clemens for
• 00:52the Adam Center. Welcome here.
• 00:54Ladies and gentlemen, and, today,
• 00:56I'm going to report you
• 00:58what one of the project
• 00:59that, we are working here
• 01:01with Clemens about the spatial
• 01:02transformers of human brain.
• 01:06As we know,
• 01:07Clemens already give good enough,
• 01:09enough intro introduction about the
• 01:11the disease of progression when
• 01:14the Parkinson progress
• 01:15and basically
• 01:16moving start with, like, a
• 01:18a non motor symptom and
• 01:19then motor symptom and then
• 01:20later dimensional, like, cognitive impairment.
• 01:23But, that's clinically happened. But
• 01:25pathologically,
• 01:26when you look into the
• 01:27brain,
• 01:28the Lewy body, which is
• 01:29a whole marker for,
• 01:30pathologic whole marker of Parkinson's
• 01:32is aggregate of alpha synuclein
• 01:34protein, also spread across the
• 01:36brain from the,
• 01:38olfactory bulb to the stem
• 01:40brain stem to, limbic system
• 01:42and then eventually go to
• 01:43the cortex areas.
• 01:45So,
• 01:48how this has happened,
• 01:50how this happening,
• 01:52this kind of transition,
• 01:54clinically and, pathologically
• 01:56are whether they are driven
• 01:58by any marker or or
• 01:59driver molecularly.
• 02:01So that's a question we
• 02:02want to understand. So few
• 02:03years back when I was
• 02:04working with Clemens, in Harvard,
• 02:07we look at this human
• 02:08postmortem brain,
• 02:09a one hundred nineteen brain,
• 02:11and we dive into the
• 02:12specific brain regions and use
• 02:14a laser capture microdissection,
• 02:16which is, this this technology
• 02:18become available before the single
• 02:19cell is available there. So
• 02:21we manually do this capture
• 02:22single cell and specific brain
• 02:24region
• 02:25and, target to three brain
• 02:27regions like a middle brain
• 02:28for dopant
• 02:30dopant neuron and the corticoster
• 02:32regions for upper middle neurons.
• 02:34And, then we pull the
• 02:35neuron together, each type of
• 02:37neuron together and do the
• 02:38total RNA and deep RNA
• 02:39sequencing.
• 02:41And there are several interesting
• 02:42messages from the work. One
• 02:43of the main message is,
• 02:45we found many,
• 02:47known,
• 02:48polycholine gene and link RNA
• 02:50or non coding RNA expressed
• 02:51in the in the brain,
• 02:52of course. For example, seventeen
• 02:53thousand of polygonal gene and
• 02:55seven thousand known non coding
• 02:57RNA expressed. But we also
• 02:58identified many novel
• 03:00non coding RNA that nobody
• 03:01had found them before also
• 03:03highly expressed in document neuron.
• 03:05For example, we found, like,
• 03:06around twenty six thousand of
• 03:07enhanced RNA and around ten,
• 03:10eleven thousand
• 03:11of circular RNA in the
• 03:13in the brain.
• 03:14So I want to spend
• 03:15a minute here to talk
• 03:16about one of the RNA,
• 03:17which is circular RNA.
• 03:19Usually, the DNA can be
• 03:21transcribed linearly to linear RNA
• 03:23like this happened. But they
• 03:24also,
• 03:25show that,
• 03:27since, around two thousand fourteen,
• 03:29two found there's the RNA
• 03:30can also,
• 03:31form circular RNA through back
• 03:33splicing as you can see
• 03:35here. And this circular RNA,
• 03:36because their shape is circular,
• 03:38there's no free end. They
• 03:39are more stable. The half
• 03:40life is ten times longer
• 03:41than the linear RNA. So
• 03:43they're perfect candidate for biomarker.
• 03:45And this circular RNA, very
• 03:46interesting to show they're dominant
• 03:48in brain versus the other,
• 03:50tissues. And more interestingly, as
• 03:53you can see in the
• 03:54bottom right corner here,
• 03:55comparing the synapsosome versus neuron
• 03:58soma, this sarcoid RNA is
• 03:59more enriched in the synapses
• 04:02which is, kind of puzzling
• 04:04too.
• 04:05And in our data, we
• 04:07look at this, eleven southern
• 04:09circle RNA. We found many
• 04:10of them actually transcribed from,
• 04:12Parkinson risk gene. For example,
• 04:13snoopling GBA log two and
• 04:16the rims two, rims one,
• 04:18and, VPS thirteen c, for
• 04:20example.
• 04:21And, also, many of them
• 04:22also transcribe from a synaptic
• 04:23genes. Like, out of, eleven
• 04:26hundred synaptic gene,
• 04:28nine hundred of them transcribed
• 04:29in, circular RNA.
• 04:31And many circular RNA actually
• 04:33when we look at this,
• 04:34brain sample between different stage
• 04:36of Parkinson's, like, control versus
• 04:38late stage and early stage,
• 04:40some sarcoRNA even start to
• 04:41show
• 04:42changes
• 04:43even in early stage of
• 04:44prodromal stage of PD.
• 04:46And one of the targets
• 04:47one of the candidates from,
• 04:49this study is,
• 04:52is DNA j c six.
• 04:53We are quite interested in.
• 04:55As you can see here,
• 04:56the DNA j c six,
• 04:57the messenger RNA does not
• 04:59change between PD control, but
• 05:01the circular RNA as the
• 05:02the
• 05:03form, it it changes significantly
• 05:05and drop a lot in
• 05:06PD. And DNA j c
• 05:07six, I mean, there's the
• 05:09following talk talking about this
• 05:11two but, DNA j c
• 05:13six is a known, a
• 05:14protein also known as auxilin.
• 05:16It's a very important,
• 05:18player in the,
• 05:19encoding of the cholesterol mediate
• 05:22endocytosis.
• 05:23So this, Vasco in the
• 05:25snatches need to be encoded
• 05:27and then be reused, recycled.
• 05:28If this encoding is not
• 05:30working,
• 05:30then there's a lot of
• 05:31buildup or accumulation of coated
• 05:34vesicle and then then a
• 05:35neurotransmitter like dopamine cannot be
• 05:37released timely. So that's a
• 05:38very important player there. Of
• 05:40course, the protein itself. But
• 05:41whether circular RNA play a
• 05:43function in that step of,
• 05:44for example, synaptic loss in
• 05:46early stage of PD that's
• 05:47still a question mark. So
• 05:49in my lab we follow-up
• 05:50this
• 05:51discovery. Now we go to
• 05:52the white lab. I mean
• 05:53this is something really I
• 05:54start to learn from after
• 05:56moving here with using a
• 05:58build organoid model.
• 06:00We want to,
• 06:01inject or or treat this
• 06:03organoid p d organoid model
• 06:04with DNA j c c
• 06:05circular RNA to see whether
• 06:07they can rescue the p
• 06:08d symptom. This is really
• 06:10pioneered by, one of the
• 06:12Thailand poster in, Maria in
• 06:13my lab, and, we are
• 06:14testing different way of deliver
• 06:17circular RNA. For example, using
• 06:18EV to wrap into the
• 06:20axon extracellular vesicle and, put
• 06:22a surface marker to make
• 06:24the brain target specific and
• 06:25also try another way, the
• 06:26lipid nanoparticle there is too.
• 06:28So please stay in tune
• 06:30on this, research and there's
• 06:31really a lot of risk
• 06:33on the on the way
• 06:34I can see. Okay. Back
• 06:35to today's focus.
• 06:38We more previously, we focused
• 06:39on specific brain regions and
• 06:41like middle brain or cortex,
• 06:43but the brain region is
• 06:44not just one piece of
• 06:45block. Right? If you look
• 06:46at this, very famous, figure
• 06:49from Santiago,
• 06:51Ramani
• 06:53Haaha's figure is drawing.
• 06:55And you see the the
• 06:56the cortex region is actually
• 06:57formed by different layers or
• 06:59sub layers. They are not
• 07:00exactly the same uniform. Right?
• 07:02And, so that's why we
• 07:04studied spatial transformics.
• 07:06And spatial transform is we
• 07:07use technology called ten x
• 07:08vision and cut the brain
• 07:10slice and put barcode on
• 07:11each spot of the x
• 07:13y coordinate of the of
• 07:14the slice and it measures
• 07:15RNA each coordinate. And then
• 07:17and then meanwhile, we do
• 07:18single cell RNA seek
• 07:20to to match up the
• 07:21same data to see whether
• 07:22the cell type deconvolution there's.
• 07:24So for those that are
• 07:25less familiar with single cell
• 07:26and and spatial,
• 07:28I borrowed this analog for
• 07:29you to understand. Basically, if
• 07:31you do the bulk is
• 07:32you make it a smoothie
• 07:33for everything together. A single
• 07:35cell, you mix the different
• 07:36things together, but they are
• 07:37still like a mix but
• 07:38individually.
• 07:39And spatial, you form a
• 07:40pattern there based on the
• 07:42the location of of the,
• 07:44of the cells and the
• 07:45expression levels there.
• 07:48So we start with around
• 07:49hundred brains, and this brain
• 07:52are very well characterized clinically
• 07:54and pathologically.
• 07:55And they can span the
• 07:57whole ten post,
• 07:58trajectory of Parkinson's, development
• 08:01from healthy control
• 08:03to the late stage PD
• 08:04and also some, one third
• 08:06of sample, we have also
• 08:07called instant Lewy body, which
• 08:09means they are clinically healthy.
• 08:10No tremor, no, dementia. But
• 08:13pathologically, there's brain already had
• 08:14Lewy body found in the
• 08:15brain. So this is very
• 08:16unique sample to help us
• 08:18identify those markers on early
• 08:19PD.
• 08:22So j a is a
• 08:23very talented post scientist
• 08:25in my lab. He developed
• 08:26this computational advanced computational,
• 08:28algorithm or pipelines
• 08:30to look into those spatial
• 08:31data, and we identify seven
• 08:34clusters in the space, based
• 08:36on the spot data. And
• 08:37these seven cluster align pretty
• 08:38well to the seven layers
• 08:40of cortical like layer one
• 08:41to six and why matters.
• 08:43And we also benchmark with
• 08:44another data set which is
• 08:45manually annotated by a pathologist
• 08:48and our layer marker gene
• 08:50highly correlated with what they
• 08:51found in their manual annotations.
• 08:54And,
• 08:55we also, he also developed
• 08:57a a machine learning based
• 08:58tool called layer smoother. As
• 09:00you can see, the original
• 09:01data looks
• 09:02sparse and and very,
• 09:04kind of noisy noisy in
• 09:06some layer boundary.
• 09:07And with this layer smoother,
• 09:09the border is much cleaner
• 09:10and, data is cleaner too.
• 09:12So we enhance the data
• 09:13a lot based on this,
• 09:14machine learning algorithm.
• 09:17And it also show this
• 09:18algorithm works pretty well when
• 09:20we compare their benchmark or
• 09:21ground truth tools
• 09:23ground truth annotation based on
• 09:25the based on the from
• 09:26the pathologist
• 09:28our layer smoother actually improved
• 09:31the original clustering a lot.
• 09:33And also, we also identify
• 09:35some of the,
• 09:36regions. For example, here is
• 09:38not really annotated by the
• 09:39even by the pathologist, but
• 09:40we identified those substructure stairs
• 09:43based on our laser smoother.
• 09:46So next question we ask
• 09:47is what kind of cells
• 09:49presented in each layers. Right?
• 09:51And whether, they're uniform distributed
• 09:54or some cells are more,
• 09:57presented in, certain layers. So
• 09:59this joint work between
• 10:01my lab and Clem's lab
• 10:03and Jacob and Jay, a
• 10:04team of very beautifully to
• 10:06get this
• 10:07cell type convolution
• 10:09in a spatial wise and
• 10:10we see each major cell
• 10:11types
• 10:12they, their relative location in
• 10:15each layers. And one of
• 10:16the take home message here
• 10:17is when you look at
• 10:18the glu glutamatergic
• 10:20neurons and the glutamatergic neuron
• 10:23have show very strong
• 10:25layer specificity.
• 10:26Meaning,
• 10:27the subtype of glutaminergic neurons,
• 10:30they can present in certain
• 10:31layers.
• 10:32On the other hand, if
• 10:33you look at the GABAergic
• 10:34neuron, they don't have so
• 10:36strong layer specificity as glutaminergic
• 10:38neurons.
• 10:40Maybe this figure is not
• 10:41so, intuitive. Let's look at
• 10:43the movie here.
• 10:45This is one of the
• 10:46sample we found and you
• 10:47can still see the layer
• 10:49marker gene for certain,
• 10:51sub cell types. They are
• 10:53highly
• 10:54presented. I wanna show this
• 10:56again but it's kind of
• 10:57positive and
• 10:58you can imagine the the
• 11:00different
• 11:01subtype of marker gene. They're
• 11:02highly correlated or highly specific
• 11:04in certain layers.
• 11:07And there are more samples
• 11:08there and we have a
• 11:10hundred samples but there's a
• 11:11few other samples we show
• 11:13the glucanergic neuron show very
• 11:14strong layer specific copper neuron
• 11:17more
• 11:17uniform relatively and other glial
• 11:20cells also show there's their
• 11:22own, cell layer specificity there.
• 11:25Now next question is, of
• 11:26course, we wanna study Parkinson.
• 11:27Right? We may want to
• 11:29understand how this location or
• 11:31this, marker molecular signal change
• 11:34along the Parkinson progressions.
• 11:36So we viewed a, using
• 11:38the a metric binomial model
• 11:40to test each gene and
• 11:41each layers and see whether
• 11:43any gene show
• 11:45a link to the Parkinson
• 11:46progression by regress with the
• 11:48Lewy body scores and also
• 11:50with just other covariance here
• 11:51as you can see.
• 11:53So what we found very
• 11:54we found a lot of
• 11:55gene, but this gene formed
• 11:56very interesting pathway. So this
• 11:58figure I'm sorry. I apologize
• 11:59for the busy figure, but,
• 12:01the the the figure is
• 12:03a circle plot. Each layer
• 12:05is a,
• 12:07each circle is a layer.
• 12:08So layer one to to
• 12:09six and then y matters.
• 12:11And the pathway if the
• 12:12pathway
• 12:13is operate regulated
• 12:15in, that layer along the
• 12:17Parkinson's progression, it will show
• 12:19a purple color. If it's
• 12:20down regulated
• 12:21along the progression, it show
• 12:23the the yellow color. As
• 12:24you can see, this pathway
• 12:26are clustered based on their
• 12:28similarity of the leading adenines
• 12:29and we show some group
• 12:31are are highly significant across
• 12:33multiple multiple region. For example,
• 12:35this is octave, oxidative phosphorylation
• 12:38is a critical role for
• 12:40mitochondrial energy,
• 12:41generation
• 12:42and for ATP
• 12:43metabolic. And see this pathway
• 12:45kind
• 12:47of, getting,
• 12:49downregulated
• 12:50in along Parkinson reg regression.
• 12:53There's other pathway linked to
• 12:55this, energy
• 12:56generation.
• 12:57For example, the the protein
• 12:58synthesis is a ribosome,
• 13:00biogenesis.
• 13:01They also get down spatially
• 13:03in Parkinson across multiple layers.
• 13:06And
• 13:07meanwhile, we see other pathway
• 13:09which is for example, immune
• 13:11related pathway like t t
• 13:12seventeen activation
• 13:14and immune response and and
• 13:16t cell regulation and and
• 13:17also b cell top six
• 13:19path response
• 13:20they also get upregulated
• 13:22in the along the pro
• 13:23three d progression.
• 13:25But we also see some
• 13:27kind of,
• 13:28hard to explain,
• 13:29since, like, a neutrophil mediated
• 13:31immunity
• 13:32pathway that actually get down
• 13:34here. So this is kind
• 13:35of a little bit puzzling
• 13:35for us too.
• 13:38Another thing we see many
• 13:40parts that relate to cell
• 13:41response to the environment to
• 13:43the, external stimuli
• 13:45or environment of surveillance, they
• 13:47also get go up,
• 13:50including a drug response to
• 13:52virus infection
• 13:55to a biotic
• 13:57infection there's this this pathway
• 13:59the purple color show they
• 14:00go up too.
• 14:02And lastly, we see the
• 14:03cell cell communication, synaptic activity,
• 14:06those pathway go up along
• 14:07progression, especially in the white
• 14:09matter. This is likely to
• 14:11do a, compensatory
• 14:13response of neural degeneration that
• 14:15the neural neural connection need
• 14:17to go up at, along
• 14:18the progression.
• 14:21So last question we'll try
• 14:22to address here is, do
• 14:24those pathway change already early
• 14:27of the PD or they
• 14:28happen later?
• 14:29So because we have well
• 14:31characterized a sample in different
• 14:32stage of Parkinson's,
• 14:34we specifically look at the
• 14:35early PD versus control.
• 14:37And this group, as you
• 14:38can see,
• 14:39we when we,
• 14:42when we look at late
• 14:43PD and, versus control,
• 14:45the the pathway looks pretty
• 14:48mimic well with the Lewy
• 14:50body progression.
• 14:51But if you look at
• 14:52the early PD
• 14:53and some of the pathway
• 14:55actually flipped the direction
• 14:57as for example, in this
• 14:58category highlight here, these are
• 15:00energy generation like mitochondrial,
• 15:02like,
• 15:03oxygen phosphorylation.
• 15:04They they speed up first
• 15:07and pump up energy, but
• 15:08later they get exhausted and
• 15:10then they get, power off
• 15:11globally. That's what we see
• 15:13from this,
• 15:15example.
• 15:17And we also give live
• 15:19a, this is ongoing work,
• 15:21actually. This is a prototype.
• 15:22We've had to build a
• 15:23online open source web portal
• 15:25to integrate all this data
• 15:27to so biologists and neurologists
• 15:28can
• 15:29look into this system and
• 15:31type their favorite gene and
• 15:32check our sample and and
• 15:33their how the gene change.
• 15:35And this work is really
• 15:36ongoing. I think this is
• 15:37going to build a foundation
• 15:38for for the AI model
• 15:40in the future too.
• 15:43So by that, I want
• 15:43to wrap up my, presentation
• 15:45today. PD five d provide
• 15:47a unique, resources
• 15:49and also hopefully build a
• 15:50foundation
• 15:51for precision medicine in the
• 15:53future. And we look at
• 15:54this Parkinson brain in multi
• 15:56omics view and, glucanergic
• 15:58neuron show very strong layer
• 16:00specificity where GABAergic neuron is
• 16:02more, relatively uniform, spatially distributed.
• 16:05And as PD progress, we
• 16:07see many interesting pathway
• 16:08either go up down,
• 16:10some will go up or
• 16:11go up like synapses or
• 16:13neurotransmission
• 16:14communication or vesicle trafficking and
• 16:16neuroinflammation
• 16:17and the cell response to
• 16:18external stimuli. Some go down
• 16:20for some energy generations and
• 16:22protein synthesis and metabolisms
• 16:24and neutral field pathway too.
• 16:26And we only see very
• 16:27marginal cell composition change along
• 16:29PD progression. This data is
• 16:31not included in this slide.
• 16:32And, energy related pathway appear,
• 16:36activating in early stage of
• 16:38PD but later become depleted
• 16:39in, in the PD.
• 16:41So this hopefully,
• 16:43I can,
• 16:44just show this, data as,
• 16:47convince you really build up
• 16:48data and get knowledge there
• 16:50as a training center for
• 16:51next phase of AI based
• 16:53modeling there.
• 16:54So lastly, I want to
• 16:55thank everybody here to hear
• 16:57my talk and, also like
• 16:59to thank all the members
• 17:00in my lab. This work
• 17:02largely driven by, done by
• 17:04by j in my lab
• 17:05and a few other member
• 17:06working on different aspect ASO
• 17:08project. I also like especially
• 17:09like to thank Clemens to
• 17:10bring me to the stage
• 17:11and to the field. This
• 17:13is really great to follow
• 17:14the whole scenes that you
• 17:15lead in the years and
• 17:16the many collaborator in ASOP,
• 17:18also Brainco project.
• 17:21And, and thank all the
• 17:22sponsors here too including APDA.
• 17:25I see Rebecca here at
• 17:27is, where I gave my
• 17:28first grant to start my
• 17:29lab. Thank you so much.