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. What if the very act of a cell reading your DNA during the 1st few days of your existence actually rewrote the code?
0:15It's a pretty wild concept to think about. Right. I mean, think about that for a second. We put so much faith in the stability of our genome. We treat it like this locked, immutable vault of information.
0:27Exactly. Like the text is written in stone. Yeah, but what if simply accessing those instructions causes the text to fundamentally mutate? And what if standard genetic tests taken by families around the world are, you know, systematically blind to a massive hotspot of these genetic mutations?
0:45Which is terrifying when you realize those mutations could explain a host of mysterious inherited diseases. Exactly. It makes you ask, how could this change everything we need about genetic screening? Well, I mean, it forces us to completely reevaluate how we think about the transmission of our genetic code from one generation to the next.
1:01Because we are essentially talking about a fundamental biological blind spot that has just been, well, sitting right in front of us. Right, hidden in plain sight. Today we celebrate the work of the Center for Genomic Regulation, Universe Tap Pompo Fabra, and Harvard Medical School, who have advanced our understanding of germline mutagenesis and transcription start sites.
1:22It's really groundbreaking work. It really is. And the mission for you, the listener in this deep dive today, is to unpack exactly how early embryonic development shapes our genetic legacy. We are going to explore the hidden mechanics of your genome.
1:37And, uh, to really understand the gravity of this discovery, I think we 1st need to establish how the scientific community traditionally views genetic mutations. Because mutations are, you know, the ultimate engine of evolution and genetic diversity.
1:52Without them, there's no adaptation at all. But they are distributed randomly across your DNA, like say, rain falling on a sidewalk. Right. The damage tends to cluster in specific areas based on what the cell is actually experiencing at the time.
2:05Precisely. I mean, we've known for decades that environmental factors and even the physical act of transcription, which is when the cellular machinery unzips and reads the genes that can drastically increase mutation rates in somatic cells.
2:18And just to clarify, somatic cells are the mature cells that make up your body, right? Like your skin, your liver, your lungs. Exactly. And we see this mechanical damage very clearly when we sequence cancer genomes.
2:29It's well documented there. But the massive ongoing debate in genomics has been about the germ line. Meaning the DNA in the sperm and egg cells or the very early embryo, the code that actually gets passed from parent to child.
2:44The big question is whether the physical act of reading a gene in those foundational early cells also causes permanent heritable mutations. Okay, let's unpack this with a visual. Imagine your genome is a rare ancient book, and every gene is a different chapter.
2:57like that. And the transcription start site, which is what we're focusing on. That's the very 1st page of a chapter. Yes, the exact sequence of base pairs where the machinery attaches. Right. So the question these researchers are asking is, does the simple act of opening the book to that 1st page over and over again physically damage the binding?
3:16That analogy captures the molecular mechanics perfectly honestly. They're looking for wear and tear at that specific binding. But um, to prove this wear and tear is happening across the human species, they couldn't just sequence a dozen people in a lab.
3:30Yeah, to find incredibly rare typos that happen at the very start of a chapter, you'd need an absurd amount of data. You need immense population level scale. So to solve this debate, They analyze what we call extremely rare variants or ERVs.
3:47Meaning mutations that are just super new. Exactly. Mutations so incredibly new in the human population that evolution literally hasn't had the time to weed them out yet. They pull data from 70,000 individuals in the Nome database.
4:00Oh, wow. And another 150,000 individuals from the UK biobank. Wait, so they are cross referencing the genetic code of 220,000 different people. Yeah, it is a staggering data set. And they anchored their entire analysis, right at the transcription start site, or the TSS of over 14,000 different protein coating genes.
4:20That is massive. It is. They looked at tiny windows of 100 to 1000 base pairs right around that start line. And crucially, they compared these incredibly rare population variants against de Novo mutations, which are brand new mutations found in standard family sequencing and also against somatic mutations from pan cancer genomes.
4:40Wait, I need to pause on the methodology here because the human genome is, what, 3000000000 letters long? Give or take, yeah. So if you are looking at incredibly rare single letter typos across 100s of people, just at the capital letter of every genetic sentence, how do you separate the noise from the signal?
4:57That is the $10000 question in Genoics. Because my understanding is that certain sequences of DNA are just naturally more fragile and prone to mutating than others, right? That is the biggest hurdle. And it's where the rigor of their methodology really shines.
5:11I mean, you can't simply count up the number of mutations and declare you found a hot spot. Because it could just be a naturally fragile area. Exactly. You have to establish a mathematical baseline expectation.
5:22And to do that, they used a rigorous 5 mer sequence context model. Okay, explain the fiber model. What does that actually mean in practice? So the term myr refers to parts? A 5 mirr is a pent nucleotide sequence context.
5:36Right. Basically, instead of just looking at a single letter that changed, say, a cytosine turning into a thymine, the model looks at the 2 letters immediately preceding it and the 2 letters immediately following it.
5:48Ah, so the local neighborhood? Yes, the local neighborhood of those 5 specific letters heavily dictates the natural background mutation rate. So you are essentially calculating a baseline odds ratio for every single neighborhood of letters, and then looking for places where the actual observed mutations completely blow past that baseline.
6:06You got it. You build a predictive model that says, given this specific sequence of letters, we expect to see exactly this many mutations by pure random chance. If the observe mutations in the patient data are significantly higher than your model's expectation.
6:22Well, then you've isolated a true signal. You have found a genuine mutational hotspot. And when they applied that 5mer model to the transcription start sites of those 14,000 genes, What did the data actually show?
6:35It revealed a massive spike in mutations. Really? Yeah. When they looked at those extremely rare variants in the population data, they found a pronounced mucational hotspot right at the transcription start site.
6:48Zooming into the 100-based pair scale exactly where transcription begins. There is up to a 35% excess of mutations compared to the background expectation. A 35% excess right at the start line. So that proves the engine book analogy.
7:01The binding isn't just fraying. It is taking heavy, consistent damage every single time the cell opens the chapter. The mechanical damage is undeniable. And they saw a similar, though slightly narrower hotspot in somatic cancer mutations, which makes sense given the aggressive cell division in cancers like, you know, bladder, breast, lung cancers.
7:22Sure. But the germline hotspot, the inherited mutations passed down through the population that was incredibly pronounced, and stretched several 100 base pairs in both directions from the start site. But in science, finding the signal is only half the story, right?
7:37The other half is validating it. Which brings us to the mystery of the missing data. Here's where it gets really interesting. It does. So the researchers wanted to independently confirm this massive population hotspot by looking at do novo mutations.
7:50These are the gold standard for tracking brand new germ line mutations in real time. Right. The standard process is trio sequencing where you sequence the mom, you sequence the dad, and then you sequence the child.
8:00Exactly. And if the child has a genetic variant that neither parent carries in their blood, you have documented a de novo mutation. Let me trace this out logically then. If this mutational hotspot is a fundamental biological process happening in human embryos, it should be glaringly obvious when we sequence those family trios.
8:21Are you saying they looked at that standard family data and found nothing? The hotspot vanished entirely. Wait, what? It completely disappeared from the family sequencing data. It wasn't there. But that seems deeply contradictory.
8:33You have a massive mutational spike in a population of 220,000 people, but when you zoom in on individual family units, the signal just evaporates. If it is in the population, it has to come from the families.
8:46How do you reconcile that disappearing act? By realizing that the hotspot wasn't missing from the human biology. He was missing from the software. Oh, wow. Yeah. Standard family sequencing pipelines are designed by bioinformaticians to act as ruthless editors.
9:01They filter out anything that looks like a machine error or background noise. Okay. And one of the specific things those algorithms are programmed to throw in the trash are early mosaic variants. We need to break down what a mosaic variant actually is for the listener, because this feels like the crux of the entire diagnostic problem.
9:19Absolutely. So imagine a developing embryo, just a couple of days after fertilization. It has only undergone a few divisions. It might be 4 cells or maybe 8 cells total. Tiny. Right. If a mutation happens in just one of those 8 cells.
9:35As that embryo continues to grow into a full adult human, only a fraction of their body cells will actually carry that specific mutation. They become a genetic mosaic. So if you take a blood sample from that person 20 years later and put it into a sequencing machine, the mutation won't show up in every single cell.
9:53Exactly. It might only show up in, say, 15% of the rees generated by the machine, rather than the neat 50% you would expect from a standard trait inherited from one parent. And the software algorithms, look at that 15% read.
10:06Assume it's just a smudge on the lens or a chemical error in the sequencing process, and automatically delete it. The pipeline literally throws the data in the trash before the clinician ever even sees it.
10:16That is wild. Knowing this, the researchers went back and specifically compiled the raw data on those discarded early mosaic mutations. Once they bypass the software filters, they uncovered a massive 52% excess of early mosaic mutations sitting right after the transcription start site.
10:35So you're saying a patient could have a severe genetic anomaly, undergo advanced family sequencing and be told their genes are completely normal, all because a computer algorithm decided the real cause was just background noise.
10:48That is the staggering reality of clinical genomics right now. Our algorithms were programmed to throw it in the trash. Our tools have inadvertently blinded us to biological reality. Wow. The hotspot is profoundly real.
11:01And it is happening during the very earliest stages of embryonic cell division. The researchers even correlated the intensity of this mutogenesis with the major wave of zygotic gene activation. Which is what, exactly?
11:12It's the moment when the early embryo boots up its own genes for the 1st time, which happens right between the 4 cell and 8 cell stage of human embryo. Okay, I get that the software hides the data because the mutation is only present in a fraction of the cells.
11:25But that doesn't explain what is physically causing this. I mean, an embryo that is just a few days old hasn't been exposed to environmental damage like UV light or toxic chemicals. So what physical force is tearing the DNA apart at these start sites?
11:39The researchers dug deep into the molecular mechanisms, and it comes down to the extreme physical stress of early embryonic transcription. The cellular machinery is just working in overdrive. They found that this mutational hotspot strongly correlates with RNA polymers the 2nd stalling.
11:57Or Naplimmerus II is the actual microscopic machine that unzips and reads the DNA, right? Exactly. Why is it stalling? Pick a zipper getting caught in a loose thread. When the polymerase stalls right after initiating the reading process, it creates a physical traffic jam on the DNA strand.
12:12And this stalling is directly associated with the formation of structures called R Loops. Wait, what is an R loop? That sounds like a structural knot or something. It is essentially a microscopic knot.
12:23And our loop is a tangled, 3 stranded DNA RNA hybrid where the newly manufactured RNA thread accidentally sticks back onto the open DNA, physically displacing the other half of the DNA strand. Well, exposed tangled DNA sounds like highly vulnerable DNA.
12:39It is incredibly fragile. The tension leads to mitotic double strand brakes. Meaning? The DNA helix literally snaps in half. Oh man. Yeah, and the cell senses is catastrophic damage and has to rush in with emergency repair crews to fix the brake before the cell divides again.
12:55The researchers actually confirm this chain of events by matching it to specific somatic mutational signatures. Our mutational signatures like uh, chemical fingerprints left behind by a specific type of damage or repair process.
13:08That is a perfect way to describe them. They scaned the data for these fingerprints and found high levels of a signature called SBS 3. And Geno makes SBS 3 is the known hallmark of alternative error prone repair of double strand brakes.
13:20So the cell is patching the snap DNA in a panic, and it makes typos. Exactly. They also found signature SBS 39, which is linked to maternal mutation clusters during those frantic early cell divisions. It is just a perfect storm of mechanical stress, tangled structures, and frantic cellular repair.
13:37Okay, if a single RNA reader getting stuck causes a one-way traffic jam, what happens in those regions of the BNA where the reading machinery operates in both directions at once? Ah, you were talking about divergent, long, non-coding RNAs or LNC RNAs.
13:53The researchers specifically analyzed genes where transcription initiates from a shared promoter region, but travels in opposite directions simultaneously. It's like a two-way street with no stoplights.
14:04Exactly. And the bi-directionality inherently causes immense torsional stress on the DNA molecule. There is more stalling, more R loop formation, and significantly more double strand brakes. So the damage is worse.
14:15Much worse. At these divergent start sites, the hotspot was even larger showing of 47% excess of mutations. It is a literal physical pileup on the molecular highway, causing these genetic pileups. This raises a massive evolutionary question for me, though.
14:29If our genomes are taking this much mechanical damage at the most critical control switches of our genes during the 1st few days of life. Why hasn't this destroyed the human species? I mean, if up to 50% more mutations are piling up at the start sites, shouldn't our genetic code dissolve into chaos over 1000000s of years?
14:50Well, it would if evolution didn't act as a ruthless filter. The mechanism saving us is a process called purifying selection. Over vast scales of evolutionary time, nature actively scrubs these harmful mutations out of the population.
15:04And the researchers prove this by comparing the extremely rare, brand new variants to older, common genetic variants that have been circulating in humanity for 10s of 1000s of years. How do you mathematically measure evolutions scrubbing something out?
15:17That sounds impossible. It's clever. They calculate a metric called a DNDS ratio. This ratio compares the rate of non-synonymous mutations, which are typos that actually change the resulting protein against synonymous mutations.
15:29which are silent typos that don't change the protein. So a ratio of one would mean mutations are just piling up neutrally and nobody's checking the work. Correct. But when they analyze these regions, they found a DNDS ratio of .82 for the rare new variants.
15:44And even more telling, they found an incredibly low ratio of .35 for the common older variants in these exact same start site regions. Wow. That drop from .82 into .35 means nature is actively identifying and deleting these mutations from the gene pool over 1000s of generations.
16:04Exactly, because mutations at the transcription start site are highly disruptive. They can break the promoter, also how much of the gene is expressed, or mutate the protein itself. Individuals carrying these mutations often suffer from lower evolutionary fitness.
16:18They are simply less likely to survive and pass on the trait. Right. So over 1000s of years, the population record is wiped clean of the damage. But evolution might clean up the species eventually. That doesn't really help the individual patient sitting in a doctor's office today carrying one of these brand new typos.
16:33What is the clinical relevance for the listener? Why does this matter to you? It matters, because the clinical relevance touches almost every branch of medicine, because these mutations are constantly being generated anew in every single developing embryo, they heavily impact disease associated genes in modern patients.
16:51Right. The researchers actually mapped this newly discovered mutational hotspot against the human phenotype ontology database, which is basically the master index linking specific genes to human diseases.
17:04What kind of diseases are we talking about here? They found that genes associated with over 20 different neoplasms and carcinomas are heavily impacted by this start site metogenesis. Wow. Furthermore, genes linked to mitochondrial disorders and genes tied to defective limb development, all overlap significantly with this hotspot.
17:22So that covers oncology, cellular energy, and developmental biology. What about neurological conditions? Neurological and developmental phenotypes are strongly highlighted in the data. And if we connect this back to our earlier discussion about the bioinformatic software, the diagnostic tragedy becomes super clear.
17:38Right. Imagine a family seeking answers for a child with a severe, unexplained developmental defect or mental health disorder. They would undergo the standard trio sequencing we talked about earlier to find the genetic cause.
17:51And the software might return a completely normal result, because the actual culprit, a mosaic mutation sitting right at the transcription start site of a critical neurological gene, caused by polymerase stalling when that child was just an 8 cell embryo, well, it was flagged as noise.
18:08It was discarded by the algorithm. The algorithm threw away the diagnosis. Patients are slipping through the cracks of our medical system because we didn't understand the fundamental physics of how early embryonic transcription damages the DNA.
18:20Yes. The urgency of this paper really cannot be overstated. We have to fundamentally update our bioinformatics pipelines. We must stop filtering out the mosaic variants at the start sides of genes because that is exactly where biology is the most fragile and where the cellular machinery is the most fiercely active.
18:38To bring all these complex threads together for you. The mechanical act of initiating transcription during early embryonic development creates a massive, previously hidden hotspot of genetic mutations right at the start of our genes.
18:54Exactly. Because standard genetic screening pipelines actively filter out these early mosaic variants as background noise, science has been completely overlooking a crucial source of genetic diversity and severe disease risk.
19:07That perfectly synthesizes the stakes of this research. It demands a paradigm shift in how we analyze human genomes. What does this mean for the future of genetic screening and our understanding of human evolution?
19:20It's a question that's going to drive the next decade of research, I think. It really makes you wonder what clinical medicine will look like in 10 years. Could we eventually map every single stalled zipper across the entire human genome?
19:32would be incredible. Right. Could we predict exactly which complex diseases a family is predisposed to simply by modeling the physical mechanics of how their early embryonic cells will read their DNA? It changes the genome from a static reference book into a living, breathing, and sometimes breaking physical machine.
19:50really does. This episode was based on an open access article under the CCBY 4 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.
20:05If 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.
20:14Thanks for listening, and join us next time as we explore more science base by base.
