This paper presents BALI, a light-driven method that writes combinatorial DNA spatial barcodes directly onto biomolecules in tissue by iterative photocleavage and ligation, enabling user-defined, scalable spatial profiling of RNA, chromatin accessibility, or both from the same section and automation via a LightScribe instrument.
0:00Welcome to Base by Base, the papercast that brings genomics to you wherever you are. Thanks for listening, and don't forget to follow and rate us in your podcast app. So, imagine for a 2nd that you were trying to understand exactly how a massive bustling city works.
0:15Okay I'm picturing it. Right. So you want to understand the economy, the culture, the daily routines of the 1000000s of people who live there. But you're forced to choose between 2 completely inadequate maps.
0:27So Map A gives you the exact GPS coordinates of every single person in the city, but tells you, like absolutely nothing about who they are or what they're doing, you just see these blank dots. Oh, wow.
0:39Yeah, not very helpful. Exactly. And Matt B tells you everything about what everyone is doing, baking bread, you know, performing open heart surgery, writing software, but it completely blurs their locations.
0:50Ah, right. So you have no idea if the baker is like standing inside a commercial kitchen or just sitting in the middle of a busy highway. Yeah, it sounds like a ridiculous way to study a city, right? It does, but, I mean, for a very long time, that has been the exact dilemma biologists have faced when trying to map out a living tissue sample.
1:08Because it's a massive tradeoff. Totally. You could either look through a microscope to see the precise physical structure of the cells, or you could basically grind the tissue up into a molecular smoothie to read the complex genetic data.
1:23Right, but combining those 2 things, like getting the deep molecular data, well, actually keeping the physical map intact, that has been, well, one of the biggest roadblocks in modern biology. It really has.
1:35So today we celebrate the work of the IMAXT Cancer Grand Challenge consortium, including researchers like Gregory J. Hannan and Dario Bresson at the University of Cambridge, who have advanced our understanding of spatial multi-omic sequencing.
1:49It's an incredible body of work. And I have to say, just the phrase cancer grand challenge carries so much weight. I mean, it sounds like the scientific equivalent of a moonshot project. The scale is certainly comparable, yeah.
1:59The cancer grain challenges are these massive, globally funded initiatives designed to tackle, you know, the absolute toughest, most stubborn questions in oncology. The ones that just refuse to be solved.
2:10Exactly. And one of the most stubborn historical challenges has been exactly what we are exploring in our deep dive today, which is tumor mapping. Right. Because, uh, decades ago, we kind of thought of a tumor as just this lump of identical rogue cancer cells.
2:26Just a uniform bad mass. Right, but today we know that is completely false. A tumor is this complex, terrifyingly efficient ecosystem. Yeah, I mean, it has its own infrastructure. You know, you have immune cells trying to fight it.
2:39Blood vessels basically being hijacked, defeated, structural cells, building, scaffolding for it. Exactly, which means if you want to truly understand how cancer grows or why a certain tumor suddenly just resists chemotherapy, you cannot just grind it up.
2:54Right, the smoothie approach doesn't work. No, it doesn't. You need to map that entire micro environment at a molecular level, cell by cell, to see exactly who is talking to whom. Okay, so let's focus on that biological context for a moment.
3:06Because for anyone listening who might be wondering why physical position is so critical in the 1st place. I mean, a sales DNA manual is the same, regardless of where it is in the body, right? Right. Like the DNA in my skin cell is exactly the same as the DNA in my heart cell.
3:22The manual is the same, yes. But the location dictates which pages of the manual are actually being read at any given moment. Oh, okay. Think about an embryo forming. Right. How does a single clump of cells know where to put a head and where to put a tail?
3:39Right, a single cell doesn't have a brain of its own to figure out its coordinates. Exactly. It relies entirely on chemical signals from the specific cells, physically touching it. So if it's sitting in a neighborhood bathed in specific chemical gradients, it turns on the genes to become, say, part of the spinal cord.
3:55And if it's somewhere else, it becomes skin. You got it. Or, uh, think about the brain. Like, if you've ever wondered how your brain builds new connections or generates new neurons as an adult, it often comes down to physical touch.
4:08Notch signaling, yeah. Right? Cells literally whisper to each other through direct physical contact. A cell surface protein on one neuron will physically bind to a receptor on its neighbor, and that, like, that mechanical handshake triggers a massive cascade of gene expression inside the cell.
4:27And if you pull the cells apart? The signal stops, the biology fundamentally changes. So spatial context is everything, period. And to solve this, the scientific community went from having a complete technological void to suddenly having a very crowded market of what we call spatial omex solutions in just like a few short years.
4:46There's a catch. There is a huge catch. The problem is all of these current methods suffer from a massive trade-off triangle. The dreaded triangle. Yeah. The 3 points of the triangle are spatial resolution, depth, and scalability.
4:58Okay, so let's quickly define those terms for the listener. So resolution is how far you can zoom in. Like, can I see an individual cell or just a blurry blob of 10 cells? Right. And depth is how many distinct molecules or genes I can actually read at once.
5:13And scalability is basically the real estate. How large of a tissue slice can I actually afford to map? Perfect. And historically, you know, you can pick two, but you absolutely cannot have all three. Like the old project management joke.
5:24Cheap, fast, good pick two. Exactly like that. If you want high resolution and incredible depth, you can really only afford to look at a microscopic speck of tissue. But if you want to look at a whole human brain slice.
5:37You have to sacrifice the zoom or the depth. So many current next generation sequencing methods try to bypass this by basically using a grid system. Okay, how does that work? They manufacture a glass slide with a predefined grid of microscopic DNA bar codes on it, and they just slap a slice of tissue over the top.
5:57Wait, so if throwing a rigid net over the tissue is the standard. Why is that failing us? I mean, we use grids for latitude and longitude. We use grids for pixels on a 4K TV. Why doesn't a grid work for biology?
6:09Because human tissue does not grow in perfect little squares. Right. I mean, it's like dropping a giant ridgy checkerboard over a map of Paris and trying to define the complex winding neighborhoods based solely on the squares.
6:21That's a great way to look at it. You end up slicing historic buildings in half, you accidentally lump the local bakery in with a massive multi-lane highway, and worst of all, you end up wasting a ton of money, analyzing the empty space in the middle of the river sane, just because a square happened to fall there.
6:39Yes. And that waste is exactly why current methods max out their budget so quickly. When you use a grid slide, you are paying for expensive sequencing regions across the entire checkerboard. Regardless of what's underneath.
6:52Exactly. If your grid lands on empty space between cells or on a giant patch of structural tissue that you don't even care about, you are bleeding your sequencing budget dry to sequence nothing. You're casting a blind net.
7:03Which brings us to the breakthrough from the Cambridge team and the core of our deep dive today. If throwing a rigid net over Paris is a massive waste of time and money, how do we actually build a smarter map?
7:15Right, because clearly we need a way to draw our own custom borders. Exactly. So the team developed a method called B-A-I. Yes, B-A-I, which stands for barcoding by activated linkage of indexes. They usually are.
7:30But the core innovation here is that it completely abandons the rigid grid. Instead, it uses highly targeted pulses of ultraviolet light to write combinatorial spatial barcodes directly onto molecules inside the intact tissue.
7:45Okay, I want to stop right there and put my skeptic cat on for a second. Okay, go for it. We are talking about using ultraviolet light to write barcodes directly onto delicate biological cells. Right. UV light causes sunburns.
7:58It notoriously causes skin cancer because it physically shatters and mutates DNA. Yeah that's a very fair point. How do they shine UV light onto a tissue sample without just frying the delicate genetic material they are trying to read?
8:11It is a critical distinction. We are not blasting the tissue with the harsh, deep UV radiation that causes extreme cellular damage. Okay, so what is it then? The team is using a very specific wavelength of near UV light.
8:23usually around 365 nanometers, delivered in highly controlled microscopic, low power pulses. So it's gentle. Exactly. It is just enough energy to break one specific engineered chemical bond that the researchers put there without having the power or the duration to shred the native genomic DNA underneath.
8:43Okay, so the tissue survives the light, that makes sense. Let's walk through the actual chemistry of how this mapping happens. Sure. Let's start with a standard piece of tissue on a glass slide. The 1st step is to attach what the researchers call a ligation route to the molecules you want to study.
8:58Like what kinds of molecules? Let's say we were looking at the RNA transcripts floating around inside the cell. This root is basically a tiny synthetic DNA adapter, but the clever part is that the root has a photo caged blocker attached to its end.
9:11A photo cage blocker. Let's visualize that. It's almost like putting a tightly fitting plastic safety cap over a bottle of super glue. I like that. Like, the glue is there. It's ready to work, but absolutely nothing can attach to it as long as that cap is sitting on top.
9:26Perfect analogy. Now, a researcher sits down, looks at a high resolution microscopic image of that tissue on a computer screen, and literally uses their mouse to draw custom borders around the specific regions they care about.
9:39Just drawing right on the screen. Yep. You want to map just the immune cells infiltrating the tumor edge, draw a line around them, and you can define these digital shapes down to a tiny 5 micrometer scale, which is roughly the size of a single human cell.
9:53Wait, 5 micrometers. So instead of accepting whatever random cells fall into a predefined checkerboard square, I'm essentially using a digital lasso to rope off only the real estate I want to investigate.
10:04That's exactly it And once your map is drawn, the system takes over. It shines that safe, low power UV light only onto the specific regions you circled. Ah, and the light pops the cap off. Exactly. The light acts as a key, snapping off that chemical safety cap.
10:20The area that was hit by the light is now active and chemically sticky, and the rest of the tissue remains completely capped and protected. So what happens to the sticky part? Well, then you flood the entire slide with a liquid containing a short DNA index, an enzyme called a Ligus, which is essentially nature's molecular sewing machine stitches that new DNA index onto the uncaged roots.
10:43But the Legus only sows it on where the light hit, because everywhere else, the safety caps are still in place. You've got it. And the real genius is what happens next. There's more Oh, yeah. The new DNA index you just glued on brings its own photo caged blocker.
10:58Wait, really? Yep, it comes with a brand new safety cap. So the area you just target is instantly protected again, and the system is basically primed for the next round of light, targeting a completely different region.
11:09Wow. So you can just build these microscopic DNA barcodes layer by layer, digit by digit by pulsing light in different shapes over and over again. And the math on this must scale up exponentially. Like if you're listening to this and trying to do the math in your head.
11:23Don't worry, I actually had to write it down earlier. gets big fast. It really does. Let's say our barcode is going to have 10 digits. Meaning we do 10 cycles of pulsing light and gluing on a new DNA piece.
11:35And let's say we only use 4 different DNA variants for each digit. Four to the power of 10 means we can uniquely label over 1000000 distinct microscopic areas on that slide. I mean, that level of combinatorial math is wild.
11:50It is. You are multiplying your spatial possibilities with every single chemical cycle. You are no longer limited by how many unique pre-printed barcodes a company can cram onto a glass slide. You just make them on the flight.
12:03Exactly. However, you can imagine that pulsing tiny beams of light 1000s of times by hand under a microscope would, you know, take a lifetime. Yeah, nobody has time for that. They needed to automate it.
12:13which is where the lightscribe device comes in, right? Yeah, right. The team built an automated custom microscope system called the light scribe, and it relies on digital micro mirror devices or DMDs. Which are? Well, if you've ever been to an IMAX movie theater, you've actually seen DMDs in action.
12:29Oh really? Yeah, they are microchips covered in 1000000s of microscopic, individually controllable mirrors. The light scribe uses these mirrors to perfectly pattern the UV light over the tissue, matching the exact custom borders the researcher drew on their computer.
12:43That's brilliant And it also automates all the fluid washes to deliver the enzymes and the DNA indexes. Okay, so the theory is beautiful and the automation sounds like science fiction, but biology is incredibly messy.
12:56It definitely is. doing robust chemistry inside a living or, well, formerly living tissue slice is notoriously finicky. So what did the data actually show? Does BLI work efficiently? It does. The chemical ligation efficiency is remarkably robust.
13:11When doing complex institute chemistry, meaning chemistry inside the actual tissue rather than in a pristine test tube? You worry the enzymes won't work or the light won't uncage the molecules properly.
13:22Right. But the team achieved about 86% efficiency per cycle. That's high. It is. And when you compound that over multiple cycles, they can build a complex multi-digit barcode across 1000s of locations with roughly 48% overall efficiency.
13:39In the world of spatial biology that is a massive success. Okay, so we know the barcode sticks. But does the biological data we get back actually tell the truth? Like, how did they validate the accuracy?
13:50They turn to the mouse embryonic brain. The brain is fantastic for testing spatial technologies because it is highly structured. Right, clear neighborhoods. Exactly. They specifically mapped out 2 distinct, tightly packed regions called the dentate gyrus and the CA1 region.
14:06And to prove BLI was mapping these areas accurately, they had to compare it against a gold standard. I understand they use something called laser capture microdissection or LCM. Let's explain what that physically involves because it sounds intense.
14:18It is intense. LCM is basically microscopic surgery. You literally use a powerful laser to cut the specific tissue region out of the slide. Yeah, you pick up that tiny piece of cut tissue with tweezers, drop it into a tube, dissolve the cells, extract the RNA, and sequence it the old fashioned way.
14:34sounds miserable to do. It is incredibly accurate because you physically isolated the cells, but it is painstakingly slow and completely unscalable. But it's the ultimate control group. Yes. And so when they compare the genetic data from the laser cut tissue against the genetic data that BLI captured just by using light.
14:55The gene expression profiles matched perfectly. Amaing. B-A-L-I accurately captured the specific biology of the dentate gyrus without ever having to physically cut the tissue apart. But the paper's title points to an even bigger achievement, multi-omic sequencing.
15:10Ah, multi-omic. Because Almic usually implies we were looking at more than just one layer of biological data. Up to this point, we've only been talking about reading RNA transcripts, which is the output of the genome.
15:22But they mapped a 2nd layer simultaneously, didn't they? They did. They mapped chromatin accessibility. Okay, for our listeners who haven't taken a molecular biology class recently, Chromatin accessibility is essentially a measure of how tightly pack your DNA is.
15:37Exactly. If a section of DNA is wound up tight like a closed, locked book, The genes inside it just can't be read. If the DNA is physically open and loose, those genes are active. Right. So measuring chromatin accessibility tells you the structural state of the genome.
15:53It tells you what a cell is capable of doing, whereas RNA tells you what the cell is currently doing. But wait, how does BLI read both at the same time? Like, if I shine a light on a cell? How does the system know whether it is tagging an open piece of DNA or a floating piece of RNA?
16:09It comes down to how you place those initial ligation routes we talked about earlier. Oh, the safety cap bases. Yep. Before the light ever touches the slide, you treat the tissue with 2 different processes.
16:19First, you use targeted probes to attach roots to the RNA. Then, you use a special enzyme, like a transpose that physically slips into the open, accessible parts of the DNA and drops a slightly different route there.
16:31Ah, so the roots themselves have, like, a molecular signature that says, I am an RNA root, or I am a DNA root. Yes, exactly. So when the UV light hits that custom drawn area, it uncages both types of roots simultaneously.
16:45The automated system glues the same spatial barcode onto everything in that circle. And later, when you grind up the tissue and sequence the molecules, the attached root tells you whether you are looking at RNA or open DNA, and the barcode tells you exactly where in the tissue that molecule came from.
17:02So instead of having to look at one tissue slice to see the structural blueprints of the DNA, and then a completely different neighboring slice to see the RNA output. You're the same house. BAL lets us walk into the exact same microscopic house and look at both at the exact same time.
17:18It really is the holy grail of spatial biology. It provides a comprehensive characterization of cellular identity that we simply haven't had access to before. And they prove the light patterning automation works flawlessly, writing 256 unique barcodes automatically without human intervention.
17:34Okay, this sounds incredibly powerful for a specialized lab in Cambridge, but zooming out to the broader scientific community. Is this flexible and cost effective enough to actually become a standard tool?
17:47Cost effectiveness is perhaps its most disruptive feature. Think back to the Paris checkerboard analogy we used, because the user explicitly draws the borders, choosing exactly where to write the bar codes, you completely eliminate the waste.
18:00Every single sequencing read you pay for is providing deep data on a cell you actually care about. You aren't wasting 1000s of dollars sequencing the empty space in the River Seine. Exactly. It's precision budging for genomics.
18:12It truly is. Furthermore, BALI uses standard, inexpensive glass histology slides. You do not need to purchase exotic, custom manufactured flow cells with proprietary grids printed on them. That's huge.
18:25It is. Tiling the UV light masks allows you to scale up to massive human tissue sections very affordably. This suddenly makes large patient cohort studies financially viable. Instead of analyzing a tumor from one single patient, a research hospital could analyze tumors from 100s of patients.
18:41I do have to push back on the data side of this, though. Okay, what's your concern? If we can suddenly map 1000000s of customized microscopic regions across 100s of human tissue samples. Aren't we just trading a chemistry problem for a massive data bottleneck?
18:58Like, can an average lab actually handle the computational load of drawing 1000000s of microscopic borders around individual cells? Well, it is not entirely trivial, but it is no longer the massive roadblock.
19:11It was even 5 years ago. Tissue segmentation, teaching a computer where one cell ends and another begins, is computationally intense. Yes. However, modern machine learning algorithms and convolutional neural networks have fundamentally changed the landscape.
19:24Software can now look at a microscopic image of a tissue and learn to recognize the borders of individual cells, almost like facial recognition software identifies a specific face in a dense crowd. Oh, interesting.
19:36So the AI can handle the digital lassuing. It can. So what is the real limitation holding this back from being in every hospital tomorrow? The paper note is a significant hurdle. The team is actively working to overcome, which is optimizing BLALI for archival FFPE samples.
19:52FFPE. Formal and fixed paraffin embedded. Right. Think about anyone you know who has had a clinical biopsy taken at a hospital. That tissue is preserved in a very specific way. It gets soaked in formaldehyde and encased in a block of wax, and it usually just sits in a hospital basement archive for decades.
20:12Yeah, and the formal and fixation process is great for preserving the shape of the tissue, but it cross-links everything. It just glues it all together. Exactly. It turns the inside of the cell into this tangled chemical statue, and it severely degrades the nucleic acids, performing elegant, light driven chemistry like BLI on tissue that has basically been chemically turned to stone is incredibly difficult.
20:33But if they can crack that code. Like, if they can optimize the chemistry to work on those wax blocks. It unlocks decades of biobanked clinical history. We could pull a tumor sample from a patient who was treated, say, 20 years ago where we already know the long-term clinical outcome.
20:48Because we know exactly which drugs worked and which ones fail for them. Exactly. We could retroactively apply BLI to map their exact cellular neighborhoods. We could finally see the spatial differences between a tumor that responded to immunotherapy and one that didn't using historical data.
21:04It would be an unprecedented treasure trove of correlative biological data. Man, that's exciting. So let's summarize the immense ground we've covered today. BLA is a highly adaptable, light-driven spatial multi-omix technology.
21:19It abandons the rigid, wasteful grid systems of the past in favor of histology aware, custom drawn boundaries. Completely. And by using tiny mirrors and targeted UV light to write complex DNA barcodes directly onto tissues.
21:33It enables the simultaneous mapping of gen expression and DNA structure at an unprecedented scale without destroying a lab's budget. It represents a paradigm shift. I mean, we are finally bending the technology to fit the complex biological questions we want to ask rather than forcing the biology to fit within the rigid limits of our tools.
21:50Which leaves us with a truly mind bending implication to ponder. If this technology allows us to perfectly map the exact micro neighborhood of a cancer cell today, capturing both its closed off potential and its active reality.
22:02How long until we can use these intricate spatial maps, not just to diagnose what a cell is doing, but to accurately predict exactly what that cell will do tomorrow? Will spatialomics allow us to forecast the spread of disease before the cell even takes its 1st step?
22:17This episode was based on an open access article under the CCBY 4.0 license. You can find a direct link to the paper and the license in our episode description. If you enjoyed this, follow or subscribe in your podcast app and leave a 5 star rating.
22:31If you'd like to support our work, use the donation link in the description. Now stay with us for an original track created especially for this episode and inspired by the article you've just heard about.
22:40Thanks for listening and join us next time as we explore more science, base by base.
