Resolve Biosciences at AGBT 2022 Gold sponsor workshop - Jeroen Aerts, Ph.D.
Thank you. So it's my pleasure, my honor to be here at AGBT and present on behalf of Resolve Biosciences. So my name is Jeroen Aerts and I'm a Customer Technology Advisor at Resolve Bioscience. And maybe more important, before I joined the company, I was actually a customer of Resolve Biosciences.
And just as Jason mentioned before, as a customer, I not only realized that this technology, Molecular Cartography, is extremely powerful for many biological applications. On top of that, the experience I got as a customer working with Resolve Biosciences was unprecedented. The way we at Resolve Biosciences share the idea that when we can enable our customers to deliver the best data that we can and empower them, this is also a massive success for us. So as a customer technology advisor, I really guide customers from beginning to end through their project, answering all technological questions, really making sure that we can deliver the best data and enable them answering their biological questions. So I will talk more about the technical specifications and applications of Molecular Cartography. And of course, you cannot talk about spatial biology without talking about the applications. So spatial biology is really already revolutionizing many different fields across life sciences, and you can apply spatial biology in many different contexts.
We all know that it's very powerful, very important in Tissue Atlas, but also, for example, in toxicology, patient stratification and of course biomarker discovery and target identification and validation. So through the next 20, 30 minutes, I will go through some of these applications and really show you the power of Molecular Cartography. But of course, to enable our technology to be applied in all these applications, we, of course, need to optimize our technology and make it compatible with many different tissue types. And this is just a small glimpse, a subset of some of the incredible datasets that we have generated across many different tissue types, including liver or kidney. I would say the typical mammalian tissue types that are highly important for all these different applications, but at Resolve Biosciences, we go way beyond that. So as Jason already mentioned, we are heavily invested, for example, in the Ag bio industry, enabling our technology for different plant tissues, but of course also many other highly important tissue types.
You can perform molecular cartography on that like organoids or embryos and even model organisms like Drosophila and zebrafish. And. To what does that relate? So if we have all these datasets available, all these protocols available to enable customers, so in less than one year we are we were able to complete more than 70 projects, deliver data to customers on more than 37 different tissue types. And this enabled customers to publish high-impact publications. So now in less than one year, we have six investigator-driven publications. So really driven by the user, driven by the customer publishing their work, using our technology.
And so in less than one year, this I think is a massive impact we had on life science community. And this is also illustrated as what Jason also mentioned here at LGBT. We have a very a big presence where all our customers have posters and talks, really explaining the power of our technology for their research.
So how does our technology actually work? How do we generate these beautiful datasets, these beautiful images and really powerful data sets that are highly quantifiable? Well, Molecular Cartography is a barcoded fish approach. It's based on single-molecule FISH. So what it actually looks like is we have specialized glass slides with eight different positions where you can place your tissue section on whether it be cyro or FFP. And we design transcript-specific probes against 100 genes per sample. And we have one hybridization event where all these transcripts, specific probes, several tens of these probes per target hybridize to the target. And then we have a patented process where we go in with a cyclical imaging approach, where we have a process to colorize these transcripts, specific probes, but also remove that fluorescent readout.
And that allows you to go in with a cyclical imaging approach to really generate unique barcodes for each transcript species. And that is illustrated on the right side of the slide here. So we know upfront in the way we design our probes that for example, in the first round of imaging transcripts, A in this case will light up in one of the two channels we use. So we use two channels and eight rounds of imaging. So we tracked those molecules in space and in the first round of imaging we do expect transcript A to light up in the first channel. We remove the fluorescence and we colorize the sample again and we expect it to light up again in the same channel.
We go through the process again, go through round three. We expect it to light up in the other channel, etc., etc. We go through all these eight rounds of imaging and as you can see for each individual transcript species, we're now seeing a unique barcode and the barcode is basically the sequence in which we expect these different transcripts species to light up in one of the two channels. Now, the very elegant and powerful consequence of this is that at the very end you can decode all these barcodes and annotate them to a transcript identifier.
So in this case, after the primary analysis, so decoding the barcode, we can actually track these spots in space and say, this is transcript A, B or C, and from that point in time, your data is fully, fully digitally quantifiable. So we're actually counting individual molecules. We don't have to use fluorescent intensity or optical density to really measure differences in expression. So this makes it extremely powerful in terms of quantification. So going further, what are some of the attributes of our technology? So one very important aspect of our technology is that the tissue remains completely intact throughout the workflow.
So we don't use any clearing agents, we don't use proteinase, for example, to remove protein information. And this allows you to have quite some flexibility in your experimental design where you first capture the full transcriptomic information. And then you can go in downstream and, for example, perform immunohistochemistry to get protein information or DNA information for that matter.
So all the analytes remain in the tissue. So we can do 100 genes per sample and as Jason mentioned, fully customizable. So we tailor our technology to your biological questions. And we also have a roadmap to increase that multiplexing capacity to go even beyond that and really capture much more transcriptomic information per cell.
An Important attributes or specification in our technology is that we capture the full z-axis throughout our tissue. So we have subcellular single-molecule resolution in all three dimensions, and this is illustrated in this figure. So this is the same data set that Jason showed. So the hemisphere of the mouse brain, but in the same data set, you can actually just zoom in all the way to the single-molecule level. And this is an illustration of somatostatin mRNA in an internal neuron of the cortex in a mouse.
And as you can see, because we captured the full 3D information, you can actually map it out in 3D. And each individual molecule here represents an individual transcripts. And so if you need for your research, for example, or are interested in quantifying the subcellular localization differences between individual cells, this is definitely imperfectly possible with our technology because we offer that resolution and also because we capture the full Z information and ten-micrometer volume. This also increases our spot calling accuracy and increases our detection efficiency. So our detection efficiency, because we based it on single molecule fish, we don't use any amplification. We really stick to the core principles of single-molecule FISH.
Our detection efficiency is extremely high. So you capture basically all the molecules that are in your tissue. And this also relates to our false positive rate, which is extremely low, so well below 1%. So all the spots that you capture in your tissue, you're sure that this is the information you need.
And as I mentioned before, digital quantification is extremely powerful. In spatial biology. You're actually able to count individual transcripts and compare regions and cells. And with Molecular Cartography, you're not only looking at mRNA, we can do many more things. We have data sets on circular RNA or even custom sequences, which is very interesting.
If you think about mRNA vaccines, we have data sets where we actually track the custom sequence of an mRNA vaccine and how that mRNA is distributed across the tissue. So we're very accommodating to all the different biological questions our customers need, and we really try to enable and empower them to our technology. And this goes again into the tissue flexibility which fuels the application plexity and also having fresh frozen, FFPE, cell culture and many different tissue types and sample types compatible with our technology. So I will go through some of the applications that will tell you more about how to think about spatial biology in general, about our type of data and how you can incorporate that into your biological question. So first of all, liver, extremely important tissue, of course, spatially, extremely interesting. And so we have many different applications using liver tissues, be it Atlassian or biomarker discovery.
But at Resolve, we have plenty of experience with both mouse and human liver tissues. So this is an example data set showcasing a little bit the structure of a liver. So here in the sea of red, these are all hepatocytes, all liver cells. You can actually see some zonation patterns.
But if you really zoom in and this is really the power of subcellular single molecule spatial transcriptomics that Molecular Cartography provides, you can actually quite easily identify these rare cell types like Kupffer cell cells or T cells just based on a couple of markers. If you zoom in on the central vein, you can actually see a nice single-cell layer identified by a single marker VWF, that marks the endothelial cells a bit further away from that. So we have central vein and portal vein. The portal vein you can actually again see these endothelial cells, but also interestingly quite easily detectable are these cholangiocytes that actually produced bile in the bile ducts. So what else can you do with this type of dataset? So with every spatial dataset, you can actually go back and treat it as a single-cell dataset as well.
You can segment out each individual cell and determine its cellular phenotype based on the transcripts that it expresses. And when we do that with this type of data set, we can actually identify different subsets of hepatocytes. And when mapping those cell types back onto the tissue, you can actually see very beautifully the zonation within the liver where you have different hepatocyte subtypes going from the central vein all the way to the portal vein. And that pattern repeats itself across the liver. So it doesn't only generate beautiful images, it's really highly meaningful for how they function, how the liver functions actually need zonation to properly function. And this brings me to a recent study that is available on Biorxiv as a preprint.
So this is from the Pittsburgh Liver Research Center, from the Director in his group. So this is Professor Paul Monga and he's very much interested in liver regeneration and liver repair after resection. And specifically he used Molecular Cartography and other spatial technologies to really look deep into the underlying molecular mechanisms that initiate a regeneration. So in controlled liver tissues he actually saw indeed with the general zonation markets beautiful zonation as you would expect.
But specifically, he was interested in WNT signaling and beta-catenin pathway molecules and looking at all the different WNT molecules. He was particularly interested in 9B and two that are very much concentrated around the central vein. And they knew that these were important in liver regeneration and really establishing the zonation, the liver function. And when knocking out those two targets, they actually clearly saw a complete disorganization, a loss of zonations.
So when you see here control versus the double knockout, you can actually clearly see that zonation around the central vein is completely lost while the zonation around the portal vein is still quite intact and segmenting out each individual cell based on DAPI signal, they extracted the single-cell information and they could also see it very nicely in these UMAP clusters where these pericentral signatures. So pericentral cells were lost when you knocked out these two targets and you would think, why would you not be able to do that with single-cell RNAseq? Well, with a spatial dimension in play here, if you would have a loss of zonation, it could also be that you just have a general change in phenotype of these hepatocytes, and this would mean a different extraction when you do single-cell dissociation. So I think going in in situ with a spatial technology like Molecular Cartography is much more powerful in terms of checking donation and what actually happens within the tissue.
And so going beyond that, so in this beautiful publication. So. Feel free to look into that in BioRxiv. It has many different spatial data sets and mainly Molecular Cartography was able to actually rescue this phenotype with a certain drug and antibody for six where when treating these double knockout mice, it actually initiates the beta-catenin pathway and restores the zonation specifically within the central vein location but not in the portal vein.
This is also nicely quantified on a single molecule level when quantifying the individual transcripts in the spatial dimension. So pericentral versus periportal. So this is an example of target identification, validation, even biomarker discovery, because you can include many different other genes in your in your gene list and look into pathway analysis and things like that. This also brings me to another study that was recently published in Cell, and this is coming from Professor Charlotte Scott and Martin Guilliams from VIB Ghent University in Belgium.
And so they are single-cell experts, liver experts. And what they did was they optimize many different protocols to get all the cells out of both mouse and human and control and disease conditions. And they needed to optimize all these digestion protocols to extract the necessary single cells or single nuclei to really map out a complete tissue atlas of the mouse and human liver. They also supplemented that with CITE-seq antibodies up to 161 really enabling or establishing a complete proteome, genomic single-cell atlas. And on top of that, they, of course, went into the spatial component to really understand how these cells are interrelated to each other. And so they first started off with the first generation of spatial technologies such as Visium. So highly powerful technology sequencing-based where you can have a very big field of view and they used Visium in combination with their single-cell RNAseq data set to fine-tune the 100 plex gene list that they used for Molecular Cartography.
But importantly, when looking at Visium, they could use it to really tailor these gene lists. But it was still lacking the resolution and sensitivity to really map out all these rare cell types. And that's where -Molecular Cartography comes into play. So they were able to detect these very rare cell types like these dendritic cells in bluish pinkish color here quite easily, based on a couple of markers just surrounding the central vein and portal vein. They were not able to detect them with either Visium or also with a highly multiplexed antibody-based technology, because it can be quite difficult to optimize all these or get sufficiently specific antibodies to use these technologies.
So this really puts Molecular Cartography at the very important position when you think about Atlasing. So very high resolution, highly sensitive technologies, Molecular Cartography can serve this need within the Atlasing community. And important to mention here is you can access all this data, not only Molecular Cartography, but also the other data sets. But if you want to play around with our Molecular Cartography data, you can download this data set from livercellatlas.org. So feel free to have a look and play around with our data.
Next up is Brain. Of course, we already saw a little bit of brain data here, but of course, brain, we have many different applications in terms of target identification, validation, or pathway analysis. And I will give you one case study that we did together with a major pharmaceutical company, and they were interested in disease-associated microglia. So these are specialized immune cells that are highly correlated with the amyloid pathology that is present in Alzheimer's disease. So what we did was we compared wild-type mouse cortex with an Alzheimer mouse model.
And just by mapping out the different major cell types based on some markers like snap 25 and things like that, where you can actually clearly see these neurons, microglia and astrocytes in the Alzheimer's condition. What we did was because the tissue remains intact throughout the assay, we were able to label these amyloid plaques or stain these amyloid plaques after we captured the transcriptomic information using Molecular Cartography. So these amyloid plaques were stained afterward, and you can see them here in the bright yellow present in the Alzheimer's disease condition.
And as you can see, you can see very big differences between the wild-type and disease condition in terms of loss of neurons and also an increase in these microglia. And one other type of analysis that you can do with a spatial dataset like this is doing differential gene expression and mapping out which transcripts are upregulated, but with the spatial component available. And then you can see when you map out which transcripts are upregulated, you can clearly see that they're very tightly organized around the amyloid plaque pathology, which makes it very interesting to go in and have a look which genes are these and which cell types are behind them.
So again, what we can do is extract the single-cell information, so segment out each individual cell. And here you can actually see very beautifully that we're able to identify all these different major cell types like GABAergic, interneurons, microglia, etc. And of course, you can compare wild-type versus diseased condition.
This is very reminiscent of what we've known for a long time. I think the first publication was in 2017 where they identified a unique subset of microglia that were associated with the disease condition. And again, here you can actually clearly see that we nicely recapitulate those findings with our spatial dataset.
But for the first time, what you can actually do with the spatial dataset is map spatial component on top of these UMAP clusters. And when you map the distance towards the center of the plaque, you can actually see the closer you get to the center of the plaque, the more you go into the diseased state of these microglia. And of course, you can map this out in a UMAP cluster, having the single cell information. But what you can also do is quantify it on a single-molecule level. I mean, we have shown you a lot of beautiful pictures, but what does it actually mean to quantify these things with a high resolution, highly sensitive technology like Molecular Cartography? So what we did was we drew concentric circles around all these individual plaques, more than 600 plaques across multiple animals. Really zooming in on these specific markers for these disease-associated microglia that we included.
You can actually count these individual transcripts and quantify them in terms of distance toward the amyloid plaque. So really taking advantage of that spatial component. And as you can see, the closer you get to the amyloid plaque, the more upregulated or the more transcripts you actually quantify of these disease-associated microglial markers and the other way around for the different neuronal markers. And the same goes for when you extract the single cell information from that data set and map it back and quantify it similarly that you actually see an increase in the number of microglia, the closer you get to the amyloid plaques.
So this is just an example of what you can do with this type of data set, but you can actually think way beyond that. You can actually look at this more from a perspective of pathway analysis, for example, where you would include many more genes that are related to the different pathways you're interested in underlying these microglial phenotypes and really quantifying which pathways are activated, which are downregulated in the spatial context, further away or closer to the amyloid plaque niche. So the last example I want to present today is an example of skin at the molecular.
At Resolve Biosciences, we have a lot of experience in both mouse and human skin. There will also be some talks in this conference on skin. And this is an example of human skin, where together with a major pharmaceutical company, we did a pilot experiment where they were interested in identifying many different targets related to immune cells and also epidermal cell phenotypes both in healthy skin. So this is a beautiful image of a human skin depicting all the transcripts present from all these genes. But they were mainly interested in how this compared to psoriasis.
So as you probably know, psoriasis is really an very inflammatory disease where the epidermis becomes thicker, you get these more psoriasis plaques on top of the tissue and you get a lot of inflammatory reactions. So going through this data set, you can actually see again all these same transcripts that I showed you just in the control tissue and you can see many more of these transcripts actually located in the thick epidermal structure. So the skin is really much more thicker than the control or healthy condition. And here you can see how big the difference actually is. Now, it becomes interesting when you then quantify all the differences you can actually, as I mentioned before, do transcript counting really count the total number of transcripts in this ROI and compared the healthy versus the disease condition. And this is just a proof of concept of first initiation of this experiment.
And the next step would be to really zoom in on these different marker genes, really looking into which inflammatory cell types are related in a spatial context, how they interact within these in the thick, thick skin here. Now more practically, how can you access our technology? Jason already hinted towards the fact that we have many workflows in the field and the workflows consist of an imager and a sample preparation unit. So you can do the sample preparation, you load in your slides, you load in the reagents, and it's a fully walk-away solution. So you basically press play and the workflow will generate the data automatically.
And on the other hand, we have the end-to-end service. So the major difference here is that instead of running the samples yourself on the workflow, you basically ship the samples to us either in the US or in Germany. And how it works is you can design all the probes and the probe sets for all the genes you have in Silico through our probe design portal, we ship you specialized glass slides and you mount the sections onto those glass slides. You ship them back to us and we do everything else.
We run the entire assay, we generate the data, and we schedule a meeting with you to go over the data, all the QC metrics and also support you in further data analysis. So maybe more, a bit more on, on sample preparation. It's very important to realize that because we adhere to the core principles of single-molecule fish, we have a very straightforward sample preparation procedure and also the protocols are very straightforward. So this is captured from tutorial videos that we brought out. So they're freely accessible on YouTube.
And this shows you that our glass slide we have, they have eight different positions. Each position is around a square centimeter. So you can place your tissue section onto that. And these glass slides have exactly the same dimensions as a normal microscopy glass slides. So after Molecular Cartography, after downstream analysis, you can still take that glass slide and put it in any microscope for any other application that you may want. And also in terms of protocols, as I mentioned, because we adhere to the traditional single-molecule fish type of chemistry, we have a very straightforward, customer-friendly protocol with a typical different steps that you would expect from a protocol like this with post-fixation.
So after we thaw the sample, you do post-fixation, you rehydrate it, you counter-stain it, you block it, you hybridize it, and then you can put it on the workflow. So because we don't use any clearing steps or any other manipulations to the tissue, we increase the customer experience in how to treat the samples, how to actually work your way through the protocols, which is quite straightforward. So after the experiment is completed, how does the data look? It's very important to realize that because we use a barcoded fish approach, the data output that you get to analyze your tissue is a count matrix. It's basically a text file with X, Y, Z, coordinates and transcript identifiers.
So you don't need to deal with terabytes of imaging data. But what you get is just in the order of a couple of gigabytes to maybe ten gigabytes per data set. And these text files, these count matrices are easily you're very easy. You can manipulate them very easily because these are just text files. It's very open. And what we also provide you is DAPI images and raw images, brightfield images of these experiments, on the other hand, which I think is quite unique in our company, is that we also provide a full data report with interesting spatially distributed transcripts with a lot of QC metrics in terms of quality scores, average counts, false positive rates. And this is really helpful for customers to have a notion of how the experiment performed, where should they look first, in which genes they might be interested? Does it look okay? And we also offer further support in bioinformatics and in cell segmentation in really enabling our customers to get the best out of their data.
So aside from the data that you get, we also offer tools to analyze the data yourself. So on the one hand, we have a tool called Polylux. It's actually a plugin of ImageJ and it's really designed to be very open. As I mentioned, the count matrix is a text file. You can do with it whatever you want. You can also load it into the Polylux tool.
And what happens is in the Polylux tool, you will see then all the genes that you have selected appearing in the user interface, you can select any gene that you like, give it any color that you like. You press the update button and it will visualize it on the image or the raw image that was also provided in the dataset. And then you can use the power of ImageJ to your advantage so you can select regions, compare counts between different regions or coupled to other tools like open source tools for cell segmentation. Or you can make the connection with single cell tools like Seurat for UMAP clustering. So you can really position this tool as both being very open and central between cell segmentation and single-cell clustering.
And you can also map those single cells back onto the tissue using our ImageJ plugin. On the other hand, we have a web-based portal, a web-based tool called Recognize, and this is a beautiful visualization software where you can actually visualize all the different transcripts and fly through the tissue fully in 3D. And on top of that, it basically has Seurat running in the background. So when you import ROI from cell segmentation, it automatically generates 3D UMAP clusters. Based on the settings that you provide, you can select one of these clusters in 3D and automatically it will map those cells back onto the tissue. So this is a more streamlined solution for people that are interested in an easy workflow, where they're interested in mapping different cell types on onto the tissue.
And this brings me to a final overview slide of what I just presented today. So Molecular Cartography is extremely versatile, extremely powerful and proven to work in many different tissue types, many different species. We can do 100 genes fully customizable for your project needs. And we capture we capture the full set access with a single molecule resolution. As I mentioned just now, we have bioinformatics solutions, so we offer the Polylux tool and Recognize tool, and we also offer additional bioinformatics support. We have a dedicated bioinformatics team that is happy to support and has a lot of experience also with difficult topics like cell segmentation, for example, and important, as I mentioned before, the true power of a technology like Molecular Cartography is in the digital quantification.
You can actually count each individual transcript within your tissue. Because the tissue remains intact, you can do downstream analysis or downstream gaining more information from p rotein or DNA or use any other assay that you might, might want for your experiment. We can process up to 24 samples in a single run, as Jason also mentioned.
And of course, the sensitivity is extremely high because we adhere to the core principles of single molecule FISH without the need for amplification or using any enzymatic reactions within the tissue. We assure that we capture the most amount of molecules in a given tissue section. That brings me to the end.
So thanks a lot. And you're all welcome in the Madison Suite. We're very happy to get to know you and if you're interested in running projects with us, we're very happy to support and initiate these projects. Thank you very much.