Hanson et al. combine pooled multi-donor human neural progenitor cell "villages" with Townlet, a hierarchical Dirichlet regression model, to estimate donor-specific proliferation and treatment responses from Census-seq. They identify 16p11.2 deletion–associated NPC hyperproliferation and nominate common variants near ZFHX3 for proliferation and an ARNT2-linked locus for lead (Pb) sensitivity.
0:00Welcome to Base by Base, the paper cast that brings genomics to you wherever you are. Thanks for listening, and don't forget to follow and rate us in your podcast app. Imagine for a 2nd uh, that you are trying to understand the vast, just incredible spectrum of human diversity.
0:17Oh, that's a small task. All of human diversity. Right. Just a casual Tuesday. But seriously, think about it. You want to know why some people are born with, you know, a vulnerability to mirror developmental conditions like autism?
0:31Right, or why some people can be exposed to a really toxic, heavy metal like lead and suffer severe cognitive consequences, while others just, they seem to have this invisible biological armor. Exactly.
0:43And normally, if you wanted to answer those kinds of monumental questions, you would need to run these massive observational studies or like clinical trials with thousands, maybe 10s of 1000s of human subjects over several decades.
0:55Which is agonizingly slow. Yeah, and incredibly expensive. Not to mention just full of confounding variables. I mean, you've got lifestyle choices, diet, environmental background noise. It's a mess. It's a huge mess But what if you didn't have to do that?
1:07What if, instead of testing 1000s of complex human beings just, you know, navigating the real world, you could extract their actual living brain cells and test them all together side by side in a single Petri dish?
1:22I mean, when you frame it like that, it honestly sounds like pure science fiction. You're talking about capturing and observing human genetic variation. In real time, at the fundamental cellular level.
1:34It's essentially the holy grail of molecular biology. Oh yeah, and for a long time, the technical barriers to actually doing it were just considered insurmountable, which is exactly why today's deep dive is so thrilling.
1:45It really is a huge leap forward. Today, we celebrate the work of Hansen and an extensive collaborative team across UCLA and the Broad Institute, who have really advanced our understanding of human genetic variation.
1:56Their 2026 paper in the American Journal of Human Genetics is titled, Cell Villages and a Ricklet Modeling Map Human Cell Fitness Genetics. It's a mouthful of a title, but the science is just it's brilliant.
2:08It really is. Okay, let's untack this. Because to appreciate why this new, uh, this Cell Village method is such a massive breakthrough. We 1st have to understand the deeply flawed way scientists have traditionally studied human cells in the lab.
2:24Yeah, the historical context here is super crucial. For decades, the gold standard for studying human pluripotent stem cells. And specifically progenitor cells, right? Right, right. Those are the dividing, unspecialized cells that act as the fundamental building blocks of the human fetal brain.
2:42So the standard has been what biologists call an arrayed culture format. A raid culture. Meaning they're kept completely isolated from one another. Exactly. If you walk into any standard biology lab, you'll see these rectangular plastic plates, and they have dozens of tiny little circular wells in them.
2:59Like a really tiny muffin tin. Yes, exactly like a tiny muffin tin. And the traditional method dictates that you take stem cells from one human donor, and you drop them into a single isolated well. Then you take cells from a 2nd donor, and they go into the next well.
3:12So you're studying them in, like, solitary confinement. Solitary confinement. But this old method introduces just a cascading series of severe limitations. First off, you have incredibly low throughput.
3:25You physically just cannot test that many donors at once. When every single person requires their own dedicated real estate. Right. And the more plates you have, I assume, the more room for human error.
3:36Oh, absolutely. Which leads to the biggest issue of all, which is batch to batch, or well to well technical variation. Because the micro environment of elaboratory is remarkably chaotic at the cellular scale.
3:50Yeah, I always like to think of this arrayed culture format using a baking analogy. So imagine you are trying to test 40 different bread recipes to see which dough rises the fastest. Okay, I'm with you The arrayed format is the equivalent of baking those 40 different recipes in 40 entirely separate ovens.
4:06Right. And expanding on that analogy. Think about all the invisible variables introduced by those 40 ovens. Exactly. If recipe number 12 rises the highest, you have a major scientific problem. Because you don't know why it rose.
4:19Right. You don't actually know if it rows the most because it inherently has the best yeast to flower ratio, like the best genetics, or if oven number 12 just happens to run a half degree hotter. Or maybe there's more ambient humidity near that oven.
4:31Yeah, or maybe the lab tech opened the door to check on it one less time than the others, you know? Exactly. The environmental noise entirely masks the genetic signal. So when scientists look at cell line number 12 growing rapidly in its isolated well.
4:47They literally can't tell if it's proliferating faster because of its unique DNA or because of a microfluctuation carbon dioxide levels in that specific corner of the incubator. And because the genetic differences between humans are often quite subtle, that background noise just completely drowns out the very thing you were trying to measure.
5:06It does. So the brilliant solution that Hansen and the research team engineered is the Cell Village. The Cell Village. love that term. It's very evocative, right? Instead of solitary confinement, they pooled 12 to 39 genetically distinct human neuroprogenitor cells harvested from completely different donors into one shared environment.
5:27So they basically threw out the 40 separate ovens and baked all the dough in the exact same oven at the exact same time to see which rises best. That is exactly what they did. The conceptual elegance is undeniable.
5:41But, you know, the execution required rigorous validation. Because the immediate biological concern would be sell to sell interference, wouldn't it? Precisely. The researchers had to definitively prove that putting all these genetically distinct cells into a single shared dish didn't fundamentally alter their natural behavior.
6:01Right, because cells aren't just inanimate dough. They secrete chemicals. They communicate with each other. They're highly interactive. So if I'm an inherently slow growing cell and you drop me into a crowded dish right next to this incredibly aggressive, fast growing cell, do I speed up just to survive?
6:16Or do I get poisoned by the waste products of my neighbor? Exactly. The fear was that the village environment would force the cells to adapt, which would effectively erase their individual genetic programming.
6:28Which would defeat the whole purpose. Right. But the validation data put those fears to rest entirely. By growing cells in the village and simultaneously growing duplicate versions of those same cells in isolated wells, they could compare the growth curves, the results were staggering.
6:44The margin of error that technical noise that plagues isolated wells was virtually erased. So they successfully eliminated the wonky incubator problem? Entirely. And more importantly, the growth curves of the cells in the village mirrored their isolated counterparts almost perfectly.
7:00Oh, wow. Yeah. A cell line that grew slowly in isolation still grew slowly in the village. A fast grower in isolation remained a fast grower in the village. That is massive. The cells really just mind their own business, even when they are packed together sharing the exact same nutrient bath.
7:16They do. They stick stubbornly to their own internal genetic programming. Which is a huge triumph for the lab side of the experiment. But, I mean, solving the biological problem immediately introduced a massive mathematical obstacle, didn't it?
7:30Oh, a huge one, because you now have this microscopic soup of up to 39 different people's brain cells swimming together in a single dish. Measuring their individual growth rates requires a completely new paradigm for data analysis.
7:44Because you can't just look through a microscope and say, oh, that cell belongs to donor A, and that one belongs to donor B. They all just look like brain cells. So how do you actually extract the data?
7:54So the researchers utilize a technique called low coverage DNA sequencing, specifically a pipeline they call census sec. Census sick. Yeah. Instead of trying to sequence the entire 3000000000 letter genome of every single cell in the dish, which would be impossibly expensive and slow.
8:12They just sample tiny fractions of DNA from the dish at different time points. It's like pulling a crowd. It's exactly like polling. They know the unique genetic markers of the donors going in. So, census sick acts like a rapid scanner, checking the ID badges of a random handful of cells.
8:29Okay, that makes sense. And from that sample, it calculates the overall proportions. It tells the researchers on day three, 10% of the cells in this dish belong to donor A. 15% belong to donor B, and so on.
8:42But wait, I have to challenge this, because proportional data seems deeply flawed for measuring growth. How so? Well, if this is a 0 sum game, because a Petri dish only has 100% of its space available, right?
8:55And my piece of the pie gets smaller over time, does that mean my cells actually shrink? Or did someone else's cells just grow so unbelievably fast that they ate up the rest of the pie, making my percentage look smaller, even if my cells were dividing completely normally?
9:09That's great point. Like, how can you actually trust a measurement that only gives you percentages? What's fascinating here is that you've intuitively hit upon the exact trap of what statisticians call compositional data.
9:20Compositional data. Yeah. By definition, compositional data must add up to 100%. If one slice of the pie expands, another slice is mathematically forced to shrink, regardless of what the underlying biology is actually doing.
9:34So all the standard off the shelf math that biologists usually use is effectively broken for this kind of village experiment. Completely broken. Standard linear regression models assume that all data points are independent variables, but in a 0 sum pi, they are highly dependent.
9:51Because if donor egg grows, donor B's percentage has to go down. Exactly. So the researchers had to custom build a mathematical solution. They developed a sophisticated computational tool called a hierarchical directlit regression model.
10:05Okay, that sounds intense. It is, but they affectionately named it Townlet. Townlet. For the cell village, I love a good naming convention. But how does townlet actually solve the pie problem? Well, without getting bogged down in the deep calculus, A durical distribution is a specialized mathematical concept designed specifically to handle multivariate outcomes that must sum to one.
10:26Calett takes the proportional data from censusec. But instead of just looking at one snapshot in time, it analyzes the continuous trajectory of those proportions across multiple time points, say day one, day three, day 5, and day 7, and across multiple parallel village replicates.
10:43Ah, so it's looking at the velocity of change, not just the final size of the slice. Precisely. By analyzing how the proportions shift dynamically over time. Townlet's algorithm can mathematically decouple the data.
10:56Oh, that's cool. It really is. It calculates the true absolute proliferation rate of each donor relative to the baseline. It is smart enough to tease apart, who is actually growing vigorously from who is just growing normally, but being statistically squeezed out by a hyperactive neighbor.
11:13So we now have the biological hardware of the village, which removes all the environmental noise. And we have the software to read it townlet, which solves the 0 sum math problem. That's the 12 punch. Once they had those 2 tools working in tandem, they immediately put them to the test on a real world medical mystery, right?
11:28And a highly controversial one of that, they targeted a specific neurodevelopmental condition. Associated with something called the 16 P 11.2 deletion. 16 P 11.2. Yeah. This is a rare genetic copy number variant.
11:42To visualize it, imagine the human genome as an encyclopedia. In these patients, one tiny paragraph on chromosome 16 has been entirely ripped out. Wow. And what are the physical consequences of missing that tiny paragraph of DNA?
11:58It significantly increases an individual's risk for autism spectrum disorder, intellectual disability and a distinct physical condition called macrocephaly, which is an enlarged cortical surface area. Essentially, the brain grows too large during early development.
12:11Exactly. But the underlying cellular mechanism, like the why behind that overgrowth has been heavily debated, hasn't it? Heavily debated, mostly because the previous scientific literature was just a mess.
12:21Prior in vitro studies on this exact 16 P 11.2 deletion were wildly inconsistent. Yeah. Some labs reported that the neural progenitor cells from these patients hyper-proliferated that they grew way too fast, but other highly respected labs ran the exact same experiments and found no difference in growth rates whatsoever compared to neurotypical cells.
12:42And let me guess, the reason for that contradiction comes right back to the solitary confinement problem. You nailed it. Because this deletion is a rare genetic condition, most labs only had access to a tiny sample size, maybe 3 or 4 donor lines total.
12:57Oh, that's nothing. Right, and they were running these precious few cells in the old arrayed format. When you have an incredibly small sample size combined with the massive well-to-well environmental noise of different incubators, it just completely swallows the biological signal.
13:12So the effects of the genetic deletion are subtle early on, and they were just getting lost in the background static of the isolated wells. So the Hansen team completely bypasses all that historical static.
13:23They built a single cell village out of 23 different human donors. 11 were neurotypical controls and 12 were bands individuals who actually carry the 16 p 11.2 deletion. They put all of them into the exact same dish, they bake them all in the exact same oven.
13:41Here's where it gets really interesting. It really does. Towlet ran the trajectory numbers on the census sec data, and the results were unmistakable. It detected a highly significant, consistent increase in the proliferation of the neuroprogenitor cells, carrying the 16 p 11.2 deletion, compared directly to the neurotypical controls sharing their dish.
14:01So there was no ambiguity this time the village proved it. Signal was undeniable. This definitively proves the hypothesis that this specific genetic deletion causes hyperproliferation in early fetal neural pools.
14:13No. And think about the mechanics of brain development. Yet the foundational stem cells divide too rapidly too many times before they finally mature into actual neurons. The brain physically ends up with too much raw material.
14:24Which directly explains the enlarged cortical surface area, the macrocephaly, observed in these patients. Exactly. Because putting all the cells in the same room, they cut through the noise and solved a long-standing controversy in developmental biology.
14:38That alone is an incredible leap forward for studying rare genetic diseases. But the researchers didn't want to just stop at ramutations, did they? They wanted to know if this village system could map the everyday, common genetic differences between all of us, not just deletions, but the normal, subtle variations that make my brain slightly different from your brain.
14:58Which represents a massive shift in scale because moving from rare disease modeling to mapping common traits means running a genome wide association study or GWAs. And typically, to find a common genetic variant linked to a subtle physical trait, you need to recruit 10s of 1000s of human subjects to get enough statistical power.
15:19But they performed a G-Woss in a dish. A G-Woss and a dish. Yeah, they constructed a massive 35 donor village. And these were all neurotypical cells. No rare deletions, no known developmental disorders, just standard natural human genetic variation.
15:35Okay. They place them in the village and track their baseline growth rates over 10 days. Wait, are you saying that just by looking at this one dish, we can see if everyday people have inherently different genetically programmed speed limits for building their brains?
15:49If we connect this to the broader understanding of human development, Yes, absolutely. Our unique genetics, constantly tweet the speed at which our neural stem cells divide. The X wild. And the village GD has proved it.
16:01In just that single 35 donor dish, they found 21 independent genetic low size specific neighborhood regions across the genome, there were statistically linked to how fast or slow a donor cells grew. Okay, so let's talk about the biggest signal they found, the heavyweight champion of these speed limit genes.
16:19The most powerful biological signal originated from a region on chromosome 16, sitting right next to a gene known as ZFHX3. Z-F-H-X3. What is the day job of that specific gene? How does it control the speed limit?
16:34So mechanistically, ZFHX 3 acts as a vital cell cycle inhibitor during neuronal differentiation. Okay, cell cycle and him. Yeah. To use a factory analogy. Imagine the stem cells are on an assembly line loop constantly dividing to produce more raw material.
16:48ZFHX3 is the brake pedal. Ah, I see. When the time is right, it steps in, halts the division cycle, and pushes the cells off the assembly line so they can start specializing into mature functioning neurons.
17:00Okay, so if ZFHX 3 is the brake pedal, what kind of genetic variation did they find in the donors? Did some people have like faulty brakes? Well, they isolated the effect down to a single nucleotide polymorphism, ASS&P.
17:13Basically, a single letter typo in the 1000000000s of letters of DNA, located near this brake pedal gene. Just one letter. Just one. They discovered that donors who inherited the baseline version of this genetic sequence had cells that divided at a normal regulated pace.
17:27But donors who inherited the typo version of this sequence had neural cells that proliferated significantly faster. So that one tiny single letter typo changes how effectively the brakes can be applied during brain development.
17:40How much of the actual growth difference between the 35 people did that one typo explain? This is a truly staggering statistic. That single, highly common genetic variation explained nearly 50% of the natural variation in neuroprogenitor cell proliferation between those donors, almost half.
17:58Half of the entire difference in growth speed across 35 randomly selected people came down to this one tiny genetic switch near the brake pedal. That completely blows my mind. Our brains are literally assembling themselves at different biological speeds based on these tiny, everyday genetic variations.
18:18incredible to think about. But, you know, natural development is really only half the story of a human life. Genetics dictate how we grow internally, but what happens when those cells are attacked from the outside?
18:27We eventually transition from natural growth to environmental survival. And that brings us to the exposome. Yes, the exposome is a crucial concept here. It represents the totality of all environmental exposures and individual encounters throughout their life, you know, diet, stress, pollutants, toxins.
18:46And some of those exposures are profoundly harmful to our cellular machinery. Very harmful. To test how our baseline genetic variation either protects us or fails to protect us from the exposed home, the team exposed a massive 39 donor cell village to a notorious narotoxin, lead.
19:03Lead, heavy metal poisoning, a chemical known to cause devastating cognitive issues, especially in developing brains. Exactly. They introduce varying concentrations of lead into the shared village environment zero.
19:15A low dose of 3 micromolar, and a high dose of 10 micromolar, and then they track the survival of the cells over 7 days using the census sec and townlet models. Now, naive logic would suggest that led just poisons everything equally, right?
19:28Like it's toxic to human biology. So it should indiscriminately kill all the cells in the dish at roughly the same rate. Which couldn't be further from the truth. The heterogeneity. The massive variation in how different people's cells responded to the exact same dose of lead was shocking.
19:43Really? Yeah. Viability varied wildly, strictly based on the donor's underlying genetics. After a week of exposure to the high dose of lead, some donor cell lines were absolutely decimated, plummeting to a mere 15% survival rate.
19:57Wow. But other donor cell lines sitting in the exact same toxic bath, maintained a survival rate of nearly 80%. 80%. It's like having a biological shield. Some people's cells seem to have this built-in genetic force field that activates against heavy metals, while others essentially have the blast doors left wide open, letting the toxin just wreak havoc.
20:17And because they had the village system running, they didn't just observe the shield. They ran another GYs to find out exactly what the shield was made of. They searched the genome to find the genetics of lead resistance.
20:28What did they find? Where does the force field come from? They pinpointed a genetic variant located right next to the RNT2 gene. ARNT 2 is a known neuroprotective transcription factor. Let's break that mechanism down.
20:41How does a transcription factor act as a shield? Okay, so when lead enters a cell, it acts like a microscopic wrecking ball, stripping electrons, and creating massive amounts of oxidative stress, and free radicals that tear apart the cell's internal machinery.
20:55Sounds bad. Very bad. But ARNT 2 acts as the emergency response commander. It detects that severe oxidative stress, rushes into the nucleus of the cell, and activates a whole suite of defensive genes designed to repair the damage and keep the cell alive.
21:11Oh, that's fascinating. So how did the genetics of the donors dictate how well that emergency commander functioned? They found another single letter variation. Donors who possessed the robust baseline version of this genetic sequence were highly resistant to the lead.
21:25They had the force field powered up. The ARNT commander responded perfectly. And the others. Donors who were heterozygous, meaning they inherited one good copy and one faulty copy of the sequence, were somewhat vulnerable, but donors who inherited 2 copies of the faulty sequence were highly vulnerable.
21:41Their emergency response system essentially failed. And their cells experienced massive dieoffs. And how much of the survival variation did this single AR and T 2 difference explain? Having a functioning shield versus open blast doors explained over 55% of the variation in cellular lead toxicity across all the donors.
22:01Over half. Over half the variation in surviving a toxic attack comes down to one genetic mechanism that is profound. So what does this all mean? It means that the synergistic combination of cell villages and the townlet directlit modeling provides a revolutionary, scalable way to map human genetic variation in a simple petri dish.
22:21We are no longer blinded by the technical noise of isolated wells. We can now accurately measure in real time, how our unique DNA drives both our natural development, like how fast our fetal brain cells divide, and our cellular resilience to the toxic environments we live in.
22:35Which leaves me with a thought, I really want you, the listener, to mull over today. If our underlying genetics dictate wildly different cellular sensitivities to common. Toxic chemicals like lead does a universal safe exposure limit even exist.
22:49That is the $1000000 question. Right now, public health agency's worldwide set a single blanket regulatory limit for what is considered a safe amount of a chemical in our drinking water or our air. But if someone sells have the blast doors completely open, while someone else has an ARNT 2 force field engaged, that single universal limit might be completely failing a huge portion of the population.
23:12It's a sobering thought. Are the discoveries from these cell villages, pushing us toward a future where environmental safety standards and toxicology must be entirely personalized based on our individual DNA.
23:23Think about what a genetically personalized exposone would mean for public health for industrial regulations and for how we identify and protect the most biologically vulnerable among us. This episode is based on an open access article under the CCB Y4.0 license, you can find a direct link to the paper and the license in our episode description.
23:43If you enjoyed this, follow or subscribe in your podcast app and leave a 5 star rating. If 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.
23:57Thanks for listening and join us next time as we explore more science base by base.
