Scientist Stories: Wei Ting Chen, Spatial Transcriptomics to Study Alzheimer's Disease

Sorry. So thank you for the introduction. Can you see the screen and hear my voice?

Yes.

Okay, perfect. And I found that there are a lot of things in between. I hope they will not be too annoying. Okay, so I will just start. So I'm happy to share our recent publication here at this symposium. My name is Weiting Chen from Professor Bart de Stroopers' lab in Belgium. So actually my work is not really focusing on the oligogenetic side, but in this presentation you will hear how we use spatial trans-omics and in-situ sequencing to unbiasedly discover a coordinated network among multiple cell types, including oligodendrocytes, their response to a certain microenvironment, the pathogenical mark of Alzheimer's disease. So we know that Alzheimer's disease, which is a major type of dementia, and in these images is the self-portrait of an artist, William, after diagnosis with Alzheimer's disease. So clearly, it's a disease which will cause gradually cognitive dysfunction and eventually death. So the central hypothesis of the disease, and there are a lot of strong genetic evidence showing that there might be a very strong etiology of the disease. However, there are several failures of the clinical trials. The anti-amyloid therapy that can actually remove the amyloid plaque from the human brain, but they cannot really stop the disease progression. So they really make the whole field really hesitate, should we still keep focusing on the amyloid? So in our working hypothesis, we have slightly modified these hypotheses. We still consider A-beta as a major trigger. It's important for the disease onset, but it's not eventually the driver of the disease, which means that A-beta overexpression or misfolding and the accumulated deposit in the brain, and that might induce abnormal cellular network around the plaque. And this cellular network that might be eventually independent of the presence of the plaque, which means that in the end stage of the disease, if you remove the plaque, they are still sitting there. And some mechanism probably happening, already happening and induced across multiple cell types, they are still there, keep driving the disease progression. So if that is the case, we are very anxious to understand what is the cellular coordinated network induced by amyloid plaque. And from multiple, lot of literature research, it's clearly that not only one cell type but multiple cell types, they have very weird morphologies around the plaque but not a cell far from the plaque. So we consider it's possible it's a network sharing by multiple cell types and they further propagate this pathology stress the signaling out of these kind of microenvironments. So then we want to understand what's the molecule, what's the cell type involved in this network. And to understand these kind of questions, initially we have immunohistochemistry, and we use the AD mouse model, which is an ABB-NRGF knocking model. So the coronal section of the mouse brain from 3 to 18 months of age. So you can see the green, that is the GFAP reactive astrocytes. So they form already kind of green envelope around the white, that is amyloid plaque, in the very initial stage of the disease. So among the disease progression, those neural networks actually are spreading over is probably depending on the presence of the plaque and also propagating to the other brain regions. But then if we want to identify cellular coordinated network, it's impossible to use IC. We need a global molecular profile with spatial information. So we use spatial transcriptomic. It's kind of a 2D RNA sequencing array. So in the old version of the spatial transcriptomic, there are 1,007 spots on top of an array. So for each array, they have 200 million of the oligos, and each oligo, they have poly-T tail and ID barcode for each spot, the spatial localization. After we permeabilise the tissue, the mRNA will leak out from the cell and attached to the poly-T tail on the array. We do the in situ reverse transcription and we collect all the cDNA afterwards. And for the library construction, then after sequencing, we can respect the spatial localization of each molecule based on the ID barcodes. So for each spot, the yellow spot shown here, they will provide us the global transcriptomic profile of of the corresponding unique tissue domain. Then the spatial resolution of each tissue domain, which is 100 micrometers in diameter of each spot, and the thickness of the tissue is 10 micrometers. So the spatial resolution is very compatible with our pathology amyloid plaque that is actually in between 10 to 100 micrometers in diameter. So this is our experimental design. The middle section we use for the RNA sequencing, and we also collect the very two adjacent sections from the cryostate we use for the immunostaining. So then if we have amyloid plaque in either two adjacent sections, Then we have very high confidence that the middle section, the corresponding spot, is collecting the transcriptomic profile in the cells in the vicinity of amyloid plaque. Because we want to link the transcriptomic profile to the abundance of the amyloid plaque, so we do the quantification of the A-beta based on the fluorescence of 6C10. So, the quantification is based on the amount of the A-beta. So, you can see that like these spots, they have very intensive fluorescence here that result in a much higher value after the quantification. And in this spot, they have diffuse plaque. Although they cover a huge area, but there is a relative small amount of A-beta deposit in this spot. So eventually they give us a relatively lower value after quantification. Then based on this experimental design, we have for each spot the full transcriptomic profile of an average 7,000 of the genes and the average corresponding level of the APL deposition of each spot. Then for this sandwich package, We carry it out on 2 genotypes and 4 time points. So eventually we have more than 10,000 of the transcriptomic profiles across 2 genotypes and 4 time points. So this is the plot that can separate based on genotype, time points, and the brain regions. So we download the Allen Brain Atlas and do the image alignment so you will know which region the spot coming from. So they can even separate and refine spatial localization of the spot in the somatic layer of the hippocampus. So like very small 3 groups here in the red. This is the somatic layer of the dentate gyrus. And this one is the CA3, and this is CA1. So after those overview of the database, we have kind of confidence it's biological meaningful. And we want to understand whether this database can cluster the gene co-expression network driven by certain biological factor. For example, if there's a landmark of the dentate gyrus, then that will only co-express and express especially from the spot from the dentate gyrus. Then it's also possible that some genes, they're only showing and highly expressed in the presence of the amyloid plaque, then they might coexpress, and their expression level might depend on the level of the plaque. So to do this kind of gene coexpression network analysis, we use a very well-established algorithm, the WGZNA, which has been shown in multiple publications to show they can identify genes coexpressed together. Normally they're functioning together. So the more complex your database is, then they will have a higher chance you can find a refined network driven by a particular biological factor. So here we do it unbiasedly across our full database. So it's like more than 10,000 of their profiles. We don't put any assumption for analysis. It could be driven by the age, genotype, brain region, or the level of A-beta, or brain regions, or cell types. So to understand what's the biological driven of the co-expression of each network, we want to understand the differential expression of the network to the AD versus wild-type mouse model. So we plot the differential gene expression of each network, whether that is up- or downregulated in AD versus wild-type, or whether that is up- or downregulated under the A beta stress. So the plot showing here, that is the mean of the log-fold change of the genes within each network. And the yellow spot is the result, the gene alteration from the 3 months, and the blue is the gene alteration in the 18 months. Then you can see that there is an outlier, the purple network, which is upregulated at the 18 months. Under the genotype axis and A-beta axis. But this network is kind of silent at 3 months. So the yellow spot here is not too much change. And the other interesting network is the red network, which is upregulated at 3 months but downregulated at 18 months in the A-beta axis. So we particularly look into more detail of those two networks. We also do a lot of characterisation to estimate what's the cellular composition of each network and which brain region they express. So eventually we named this purple network, which is composed by 57 of the genes, as blood-induced genes network, the PIGS network. And the red network is really enriched with a lot of myelin-related transcripts. We call it oligodendrocyte network. So here I'm going to show you the detail of each gene within each network. The red is the oligodendrocyte. So they are upregulated in the plaque axis, in the plaque at 3 months, but they are depleted at 18 months. In these microenvironments. And the PICS is silent at 3 months, and most of the components of the PICS, they are upregulated in AD, especially in the plaque niche. So then I'm going to show you detail about the PICS. It's composed by 57 of the genes. Genes. So the plot shown here, the green indicates the strength of the connectivity score of the gene pairs of the pigs. The darker the color indicates the higher co-expression of the gene pairs within the selected transcriptome profile. And the picture showing here is the transcriptome profile selected only from the wild-type mice sperm. And you can see it's not really co-expressed among all the gene pairs. They split into 3 clusters. The green cluster, they have very strong internal connectivity, and many of them, you can— they are kind of well known. They are expressed by astrocytes. And the blue and the orange cluster, Many of them are expressed by microglia. But this fixed network, when we start to look into the transcriptomic profile in the AD brain— so now I only select the AD transcriptomic profile with lowest A-beta deposition— they already start to build on this kind of connectivity between the gene pairs and between the clusters. And if we gradually increase the amount of the A-beta, you can see that the connectivity score is going to be more and more stronger, so which indicates this network is gradually built up based on the amount of the A-beta deposition in each tissue domain. And especially this orange network, This orange cluster, they start to bring into the network, especially via the TRIM2 and TyroBP, with their projection, their connectivity to the ApoE and the C1qAbc. And we are kind of surprised because this unbiased analysis, which actually point out one of the most important molecule in Alzheimer's disease, And based on the gene ontology analysis, it indicates that this network is overactivation of the classical complement cascade. For example, they have C1qAbc and C4ab and Clu. They also have a lot of downstream effectors of the complement cascade. For example, lysosomal dysfunction, antigen processing and presentation, oxidative stress, immune response. So now we find a network which corresponding to the plaque deposition, and we know the mechanism, we know the molecule, and we also know what's the potential functioning mechanism of this network. But we don't know what's the cell which represent and express and driving this network. So we overlap our network together with previously identified marker gene of the disease-associated astrocytes and microglia. So it's clearly not only from one cell type. We also have some genes which is not identified as those kind of cell markers. So to get the real experiment and single-cell resolution, we incorporate the other orthogonal technology in situ sequencing with the collaboration with Catana. The technology enabled us to get a gene expression profile for more than 100 targets together on one section in microscope resolution. So then the picture showing here is the puncta of the PRAX1, which is the landmark of the dentate gyrus. And I'm going to show you the expression of the cell type markers. The red puncta is from the neuron D6 or synaptophysin. And the blue is from oligodendrocyte markers, and the green from astrocyte markers, and the yellow from microglial markers from those genes. And you can already see a cluster of green and yellow And that's actually the place of amyloid plaque. So we can do the immunostaining after the in situ sequencing on top of the same section. So then we can try to use this technology to understand what the cell express or picks in the mouse brain. We do this experiment in two wild type and two AD brains at 18 months of age. So first of all, we try to see whether the peaks, they are enriched in the plaque. And to our surprise, the two orthogonal technologies, they give us very highly consistent results. So the y-axis is the local change of the genes to the A-beta level based on spatial transcriptomic. And the x-axis is the log odds ratio, the enrichment of the genes in the RANK1. And the plaque based on in situ sequencing analysis. So both of them, they all showed that the pigs is significantly upregulated around the plaque. And we will try to do the cell typing of each gene within our pigs. For example, we start from looking at the green subcluster of the pigs. There are many bubbles showing here. The bigger the bubble indicates the higher log odds ratio of the enrichment of that gene correlated with the cell type marker of certain cell type. So then you can see most of them are expressed majorly by the astrocyte. But the interesting stuff, you can see the ApoE normally expressed by astrocytes in wild type in AD models, but they start to express by microglia around the amyloid plaque. And it's also interesting for C4a and b. So it's normally expressed by astrocytes, but they start to express by oligodendrocytes in the AD models, especially around amyloid plaque. So this indicates that this microenvironment that triggered the activation of microglia or oligodendrocytes to start to express ApoE and C4qMB involving this network. We also do the same for the blue and orange cluster. And it seems that the majority of them are expressed by microglia, but neurons, astrocytes, oligodendrocytes are sometimes also involved. We also do the similar practice in the human brain. So not only the pigs, we also do the in situ sequencing for the oligonetwork and the superior frontal gyrus of postmortem human brain tissue. So together we do the 220 targets, 380 and 3 control. And the image showing here is the white amyloid plaque is colocalized with the red, the neurons. They are mostly in the gray matter. And the blue is oligodendrocytes, mostly in the white matter. And the purple image is showing here is the expression of RPEX. It's colocalized with— it's more enriched with the plaque and the gray matter. And for the oligodendrocytes, During the cell typing, it's really purely from the oligonetwork, really purely from the oligodendrocytes. So that's why we named it as an oligonetwork. So I make a short summary of the PICS network. So we consider it's a gradually built-up network between especially microglia and astrocytes and the vicinity of amyloid plaque. So this is the mean z-score of the gene expression level of the pigs at 3 months of age in different conditions. For example, in wild-type mice, or AD models, depending on different level of the A-beta, it's kind of silenced at 3 months. But the pigs, they start to increase the expression based on the level of the A-beta deposition. In the AD brains. However, this response, they don't really have any unique regional specificity. So the more red, that is more upregulated in the plaque niche, and the blue is depletion in the plaque. And with the mesh that indicates it's not significant of up or down regulation, so at 3 months, it's kind of silent. At 18 months, as long as they showed a significance that's highly overexpressed around the plaque, but kind of all over the brain. And for the oligodendrocyte network, it's kind of interesting. Their expression level at 3 months is— during the mild amyloid deposition, their expression level gradually goes up.. But if there are humongous ABL deposition, such as plaque in the brain, so even at 3 months, this network, they disrupt, they reduce the expression level. And for 18 months, the olfactory dendritic cell network in the mild ABL deposition, they already go up compared to the wild type, but they gradually decrease when we increase the A-beta per microenvironment. So we consider the ollic network as an early reaction to the mild A-beta deposition, but they eventually break down if there is severe A-beta deposition in the brain. And this reaction actually halved the regional diversity. For example, if you look at the 3 months of age, So this network is highly enriched in the plaque, especially in the fiber tract thalamus and hypothalamus, where are the brain regions enriched with the myelin and oligodendrocytes. And for the 18 months of age, that is not very clear about the depletions in certain brain regions. So it's more like a early reaction in the oligodendrocyte-enriched area that have higher response to the plaque. And so I would like to especially appreciate several collaborators. It's a humongous work across multiple colleges, especially Ashley and Mark. They are bioinformaticians and they carry out majority of the analysis of this paper. And many people from Barty Shoopers' lab and people from Image Core in the IBK Ruven. So the spatial transcriptomic, which we have the collaboration with Joachim Lemberg from Cornell Liska. And now the spatial transcriptomic has been bought by 10x Genomics. It's available. It's called like 10x Visium slides. Work together with Cortana is now also being bought by Tengenomics as well. So they are also available from that company. And mice is from RIKEN, and the human brain sample is from the Allen Brain Bank. Yeah, it's now open for questions.

Thank you. Thank you so much for the beautiful talk. While we are waiting for the questions to pour in, I personally have a question because our lab is interested in multiple sclerosis and then normal-appearing white matter. We're always curious how far from the lesion that is considered normal. I guess in your system you have the chance to really see how far away from the plaque like the oligo modules, like genes respond, early responses, like back to like homeostatic states? Like, yeah, like I said, my question is like, have you looked into that?

You mean that did I look into the homeostasis of, sorry.

Yeah, like from the plaque, the center of the plaque, I assume that surrounding the oligo will respond to the plaque plaque and then by upregulation of like the myelin genes. And then I wonder like how far, like maybe like 5 microns or 100 microns from the plaque, then you no longer see this upregulation.

I wonder, have you checked that? Yes. So from the spatial transcriptomic, we do the analysis in a different way, but from the in situ sequencing, it's possible to understand the distance from the plaque. So here you can see that we drew the rings based on the distance from the plaque. The white is the 60/10 staining. So we expand the mask of this kind of amyloid plaque by 10 micrometers in radius as the ring 1. And then we expand by 20 micrometers as a radius for ring 2, ring 3, 4, 5. So then oligodendrocytes we see the enrichment of the C4a and C4b in oligodendrocytes, especially in ring 1, and that gradually decreases from ring 1 to ring 5. So, actually, after ring 2, it's already almost going to the background levels. So the enrichment of the C4b in oligodendrocytes is like this, but for the enrichment in the astrocyte is more like in ring 2 or ring 3, and the reaction of the microglia is more like within the ring 1. Yes, that's the result from the in situ sequencing. But for the spatial transcriptomic, unfortunately we do the quantification in the other way around, so we don't have like those rings because the those spots are physically fixed by their spatial localization. Yeah.

Thank you. Okay. We have a question from Nathan Pomper. Great talk and work. Based on your staining for PLP1 and MBP, it suggests that your work focused on myelinating oligodendrocytes. Did you see any interesting phenotypes with regard to OPCs or newly born oligodendrocytes? In many other disease models, OPCs demonstrate increased proliferation, whereas new oligodendrocytes degenerate?

It's a really nice question. So, unfortunately, during our experimental design, we did not really focus on different lineages of oligodendrocytes. So, eventually, we don't have a reliable marker to indicate whether the OPC is involved in the PIKs or the OLEPs network. So, we— I cannot provide any answer with confidence. But we just aware this kind of important question after we generate all the data. And there are several interesting paper recently. They also find the OPC, they start to go into senescence and the plaque niche. And they start to go like differentiate into the myelinated oligodendrocyte. And they instead, they release inflammatory cytokines. And they become to involving this kind of inflammatory network together with the microglia around amyloid plaque. But it's kind of pity here that our project cannot answer the question about OPC. Yeah. Okay.

Well, quickly, next question from Brian. Amazing data in the presentation. What is the expression of the complement molecules in oligodendrocytes at later stage in your AD model, like your second point of analysis?

18 months? Sorry, the question is—

Could you repeat again?

Sorry.

What is the expression of the complement molecules in the oligodendrocytes at a later stage? When? I guess is about the species, but what is the expression of the complement molecules?

What's the expression? So the complement, so the peaks, the expression level, they are silent at 3 months and also 6 months, but then they are dramatically increased between 6 and 12 months. And after 12 months, those peaks and complement molecules, they are kind of saturated after 12 months. And for their expression in the oligodendrocyte, the species, sorry. So the question is about whether they also express, how they express in the human oligodendrocyte?

Oh, I got a clarification that I meant the expression level, like, yeah. Yeah, I think it can like further maybe reach out to Wei Ting to further address the question. For now, like, we thank Wei Ting for a very—

Okay, no problem.