Axakova et al. produced comprehensive missense variant-effect maps for human SOD1 by assaying enzymatic activity in yeast and protein abundance in human HEK293T cells for ~86% of possible missense substitutions. The study links map patterns to sequence-structure-function features, uses kernel-density calibration to provide LLRp evidence for clinical interpretation, and shows the abundance map best discriminates pathogenic variants and supplies new evidence for many VUSs. Data and LLRp-calibrated scores are available on MaveDB and in supplemental DataS2.
0:09On the right light, on night glow, we trade one letter, watch it show. A protein tries to hold its shape, some changes face. Welcome to Base by Base, the paper cast that brings genomics to you wherever you are.
0:24Thanks for listening, and don't forget to follow and rate us in your podcast app. Imagine getting a genetic test right. And you find a mutation, but your doctor tells you they have, well, absolutely no idea if it's completely harmless, or a ticking time bomb for a severe neurodegenerative disease.
0:39Oh yeah. That is a terrifying situation. Right. What really happens when a single letter in your DNA changes, but just completely defies clinical interpretation. And, uh, how could testing 1000s of mutations before they ever show up in a patient change the way we diagnose and treat genetic disorders?
0:58I mean, it is the ultimate medical cliffhanger, really. And it plays out in neurology clinics constantly. Because, you know, we're sequencing patient genomes at this unprecedented speed now. Well, you have so much data. Exactly.
1:10Generating massive amounts of genetic data isn't the problem. The bottleneck is no longer finding the typo in the DNA. The real bottleneck is understanding what that specific typo actually does to the human body.
1:21Well, today, we celebrate the work of Anna Exakova and the extensive team of researchers at the University of Toronto, the University of Pittsburgh School of Medicine, and other collaborating institutions who have advanced our understanding of SOD one variants.
1:35Yeah, their 2025 paper in the American Journal of Human Genetics is just, it's phenomenal. They deliver this massive, unprecedented, functional map of misints variants in the human SOD one protein. And a missance variant, just to clarify.
1:50That's essentially a single letter typo in the DNA, right? Where it swaps out one amino acid for another in the final protein. That's exactly it. And to grasp why mapping every possible typo is so critical for this deep dive, we really need to look at the devastating condition tied to this specific gene, which is amiotrophic lateral sclerosis or ALS.
2:12You know, that's right. a progressive motor neuron disease. Basically, the neurons controlling voluntary muscle movement, just degenerate and die. It really is. And the statistics are pretty sobering, um, with a lifetime risk of developing it sitting at roughly one in 300.
2:26Yeah, one in 300. And while many cases appear sporadically, you know, without a clear family history, genetics actually drive a significant portion of the disease. Up to 23% of familial ALS cases are linked directly to mutations in the SOD1 Gene.
2:42Wow, 23%. So what exactly does that gene do? So SOD one is the blueprint for an enzyme called superoxide dismute taste one? You kind of think of this enzyme as this essential cellular garbage disposal for free radicals?
2:56A garbage disposal, right? It's a scavenger. It neutralizes highly toxic oxygen byproducts, specifically super oxide radicals before they can, you know, rip through your cells and damage delicate structures like your motor neurons.
3:07Okay, so if the SOD one blueprint has a mutation, the resulting enzyme might just fail to clear those toxins, right? Which allows damage to accumulate and ultimately kills the motor neuron. Exactly. That basic mechanism seems relatively straightforward.
3:20But the frustration really arises when a patient gets sequenced, a change in the SOD1 gene is identified, but the clinical databases just label it a variant of uncertain significance, or a VUS. The US, and how common is that?
3:33Currently. About 26% of known SOD1, misence variants fall into this super frustrating category. A quarter of them. I mean, a variant of uncertain significance leaves both the physician and the patient completely stranded.
3:48Oh, completely. The sequencing machine spots the swap domino acid, but the scientific literature just lacks the concrete evidence required to declare it definitively pathogenic, meaning it actually causes ALS or, you know, definitively benign.
4:02So the patient is just left in diagnostic limbo. They know they have a mutation, but they don't know its fate. But, wait, with all the incredible advances in AI and machine learning today, couldn't a computer algorithm just look at the mutation and predict the outcome?
4:16You think so, right? Yeah. I mean, to me, it feels a bit like trying to predict if a car will run just by looking at a high resolution photograph of its engine. We know what a working engine looks like, so shouldn't we be able to spot the broken parts?
4:28I love that analogy. And honestly, computational predictors are doing exactly that. They've become remarkably sophisticated at analyzing the sequence and the structural context of a protein. They scan the photograph for missing belts or cracked pistons, so to speak.
4:43Right. But I'm guessing there's a catch. A big one. The reality of biology, much like auto mechanics, is that sometimes an engine with severe cosmetic damage runs perfectly fine. Meanwhile, a pristine looking engine refuses to start because of some microscopic electrical fault deep inside the alternator.
5:02Oh, wow. Okay. Algorithms rely heavily on evolutionary conservation and static structural models. So they often fail to capture the dynamic, chaotic reality of a protein folding and interacting inside a living, crowded human cell.
5:15Basically, you don't actually know if the car works until you turn the key. Exactly. And clinical diagnostic guidelines, like the ones from the American College, Medical Genetics and Genomics. They reflect that exact principle.
5:26They require validated functional biological assays to provide strong evidence for variance impact. But historically, getting that biological evidence has been super reactive, hasn't it? Like, a patient develops ALS, a novel mutation is discovered.
5:41And then a research lab spends months or years developing a test just for that one specific mutation. Exactly. very reactive. But this research team flipped that model entirely. They engineered a proactive approach.
5:54So instead of waiting decades for patients to show up with rare mutations. They use a technique called saturation mutagenesis. Saturation mutagenesis. That sounds intense. It is. They systematically created over 2000 possible amino acid substitutions in the SOD1 gene inside the lab.
6:12They basically synthesize nearly every possible misense typo that can theoretically exist. That is an incredible feat of bioengineering. But just having the variant sitting in a test tube doesn't actually tell you how they behave, right?
6:23Because proteins can fail in different ways. Right, right. So the researchers deployed 2 parallel biological assays to capture different modes of failure. The 1st assay measured total enzymatic activity.
6:34And for this, they turn to yeast cells as their living test tubes. Yeast cells. Yeah, they genetically deleted the yeast's natural sawd one gene and inserted the mutated human versions. And to really force the yeast to rely on the human enzyme, they grew them at a restrictive 38 degrees Celsius.
6:51Wait, why does turning up the heat matter? Heat induces severe oxidative stress in yeast. At normal room temperatures, a yeast cell might actually scrape by without a perfectly functioning SOD one enzyme.
7:04But at 38 degrees, the cellular environment becomes incredibly hostile, just flooded with toxic superoxide radicals. Ah, so having a functional SOD one enzyme becomes an absolute matter of life and death for the yeast.
7:17Precisely. If the yeast colony thrives and grows, the inserted human variant is actively scavenging the radicals. If the yeast colony dies, the variant is functionally useless. Okay, so the yeast essay checks, if the enzyme can still perform its chemical job, but there's a second test, right?
7:33And that measured something fundamentally different, protein abundance. Yes, abundance. For this one, they moved away from yeast entirely and used human HEK293T cells, which is just a standard, really robust cell line, used all over molecular biology.
7:46And they utilize this high throughput screening method called VMP sac. That stands for variant abundance by massively parallel sequencing. Basically, they physically attached a glowing green fluorescent protein, or GFP, to every single SOD one variant.
8:01Oh, that's clever. That elegantly transforms a complex biochemical measurement into a simple visual readout, doesn't it? It does. See, the cell has a natural quality control system. primarily the produceum, which acts as a sort of cellular shredder.
8:15If a mutated SOD1 protein folds correctly and remains stable, it evades the shredder, accumulates in the cell, and the attached GFP makes the entire cell glow a bright vibrant green. But if the mutation destabilizes the physical structure, the quality control system spots the misfolded protein and just shreds it, right?
8:32Right, destroying the GFP tag along with it. So the resulting cell is extremely dim. The researchers then use these automated cell sorting machines with lasers to physically separate the bright, healthy cells containing stable variants from the dim cells with degraded ones.
8:44That perfectly isolates 2 totally different points of failure. Going back to my analogy, the yeast assay checks if the wrench can actually turn the bolt. But the human cell abundance essay checks if the wrench simply dissolved into rust before you even open the toolbox.
9:01I love the rust analogy. It spot on. Combining these 2 assays allowed the team to successfully measure the functional impact of about 86% of all possible misence variants in the SOD1 gene. 86%. That's massive.
9:16It is. They generated 1000s of data points, mapping both activity and abundance long before those specific mutations are ever encountered in a clinic. So when you plot all those 1000s of data points on a graph like activity on one access, abundance on the other.
9:30You get 4 distinct quadrants, right? Yeah exactly. And the data showed that about 54% of all possible mutations were completely neutral. They were perfectly in the yeast and they glowed brightly in the human cells, and on the extreme other end, I think 12% were just devastatingly bad across the board.
9:44Right. They degraded in the human cells and totally failed to prick the yeast. But honestly, the variants that fall into the discordant quadrants, the ones that failed one test, but past the other. Those reveal the most profound insights into structural biology.
9:58Like the stable but inactive variants. Yes. Think about those. These are the proteins that glowed incredibly bright green in the human cells, meaning they were perfectly stable and abundant. Yet, when placed in the heated yeast model, the yeast died.
10:14So there are the pristine wrenches that completely fail to turn the bolt. Exactly. And when the researchers map the locations of these specific stable but inactive mutations onto the 3D architecture of the SOD1 protein, a really striking pattern emerged.
10:28You see, SOD1 is an enzyme that requires atoms of copper and zinc to perform its chemical reaction. Oh, right. The stable but inactive mutations clustered almost exclusively around the specific amino acid residues responsible for binding those metals, specifically locations like histidine 47 and histidine 121.
10:46So altering those precise coordinates destroys the enzyme's ability to hold onto the copper and zinc. Right. The physical scaffold of the protein remains totally intact and highly stable, but the chemical factory inside is permanently shut down.
11:00The precision of that failure is fascinating. But then the exact opposite scenario happens with the structural failures, right? These are variants that failed the human cell abundance test. The protein simply degraded, and the cells went dim, regardless of whether the active site was intact.
11:16Right. And mapping those structural failures onto the 3D model revealed they occur primarily in the buried hydrophobic core of the protein or along the critical edges of its beta sheets. Hydrophobic meat and water repelling, right?
11:30Yes. Proteins fold into specific shapes by burying those water repelling amino acids deep inside, away from the watery environment of the cell. If a mutation introduces a bulkier charged amino acid into that tightly packed hidden core, it acts like a localized earthquake.
11:45Oh, wow. The internal structural pillars buckle. The entire architecture collapses. and the cell's shredder eliminates the unstable mess. But the researchers actually dug even deeper than just the obvious metal binding sites and buried cores, didn't they?
11:59They identified variants that were inactive but stable, yet located entirely outside of the main middle binding pocket. Yeah, they called these 2nd shell residues. They highlighted one specific amino acid in particular, asparagene 87. or as any seven.
12:14As 87. What makes that one so special? Well, it's a brilliant case study in protein dynamics. It sits several nanometers away from the active site where the copper and zinc are actually held. So to understand why a mutation at a distance causes the entire enzyme to fail, they ran 400 nanosecond molecular dynamics simulations.
12:32400 nanoseconds. They basically built an atom by atom digital replica of the mutated protein. Exactly. And they use supercomputers to just watch how it physically moved over time. They observed that as 87 normally forms a salt bridge, a strong electrostatic bond, with another residue, ask 125.
12:51Okay, assault bridge. And that specific bond locks a large, flexible part of the protein called the electrostatic loop into the exact position required to guide the metal atoms into the active site. Ah, I see.
13:01It functions exactly like attention wire on a suspension bridge. If the tension wire snaps, the structure meant to catch and hold the metal simply sags out of alignment and fails. The central tower of the bridge might be completely fine, but the road gives way because a secondary anchor point snapped.
13:17That is exactly what the simulation showed. Mutating as an 87 breaks that salt bridge. The electrostatic loop becomes wildly unstable and just flops around, allowing water molecules to rush into the active site.
13:31Which means the essential zinc and copper atoms can't bind securely anymore. Right, and the enzyme's ability to scavenge free radicals drops to zero. This level of detail is amazing. It really brings us back to the computer algorithms we discussed earlier.
13:44The researchers directly compared the performance of their biological abundance map against massive AI predictors like alpha missants, right? They did. Alphemisence uses deep learning models similar to Alpha Fold.
13:56The AI was trying to guess pathogenicity based on sequence patterns, while the biological assay was measuring physical degradation. And the biological essay won. The biological cell abundance as a massively outperformed the artificial intelligence.
14:10You see, alphemisens and similar AI tools are highly trained on evolutionary conservation. They know which amino acids haven't changed over 1000000s of years of evolution. So they can predict the static 3D structure really well.
14:24With shocking accuracy. Yeah. But what they struggle to predict is the dynamic half-life of a protein. They cannot reliably calculate how quickly a slightly misshapen protein will be recognized and shredded by the produce sum in a living human cell.
14:38And the VMP sec assay measures that exact degradation dynamic directly. Precisely. It provided a ground truth measurement of cellular stability that the AI just couldn't simulate. Which translates into a massive clinical bythrough.
14:51I mean, by relying on that ground truth data, the researchers generated new, highly accurate functional evidence for 41% of all the existing SOD1 variants of uncertain significance currently sitting in clinical databases.
15:03It's huge. Almost half of the patient's trapped in diagnostic limbo could potentially receive a definitive answer based on this single proactive data set. But wait, I have a question about the disease mechanism.
15:14The protein is degrading and disappearing. How does that cause the toxic clumps we see in ALS? We know that ALS is fundamentally driven by a toxic gain of function, right? The mutated SOD1 protein misfolds and forms sticky aggregates that physically choke and poison the motor neurons, but shouldn't a degraded protein just be gone, preventing those clumps from forming in the first place?
15:37That is such a crucial question. The biochemical reality of neurogeneration involves the concept called a toxic oligomer. When a variant scores low on the abundance map, it means the protein is fundamentally unstable.
15:51But the cell's quality control machinery, the prodisome isn't instantaneous. Oh, so there's a delay. Exactly. There is a critical window of time between when the mutated protein is synthesized and when it's finally shredded.
16:04During that window, the unstable protein struggles to fold. It briefly exposes its hydrophobic core, those sticky water repelling amino acids we talked about earlier. Ones that are supposed to be safely buried inside.
16:15Right. And in the watery environment of the cell sideoplasm, those exposed hydrophobic patches, desperately seek out other exposed hydrophobic patches to avoid the water. So they begin sticking to other misfolded SOD1 proteins, forming small, highly toxic clumps called oligomers.
16:32It's basically a race, a race between the cell's garbage disposal trying to clear the misfolded proteins, and the proteins physically clumping together. You nailed it. And unfortunately, the clumping process eventually outpaces the cleanup.
16:45Once those small oligomers form, they can become too bulky for the produceum to degrade, and they continue to accumulate into the massive aggregates characteristic of end stage ALS. So a low abundance score in the human cell assay is a direct proxy for that initial dangerous structural instability.
17:02Yes. It proves the variant has the exact biophysical prerequisite needed to expose its sticky core and initiate that aggregation cascade over a patient's lifetime. Wow. That perfectly bridges the gap between a single degraded cell in a lab dish and decades of progressive neurological decline in a human.
17:20The instability is the spark, and the aggregation is the fire. beautifully put. And this isn't just theoretical modeling either. The researchers connected their lab data back to the actual clinical records of living human patients.
17:33They found a profound genotype phenotype connection. What did that connection look like? The correlation was deeply compelling. They analyze clinical databases containing the age of onset and overall disease duration for patients carrying these specific SOD one mutations.
17:49And the variants that score the lowest on the abundance map, the most highly unstable proteins that degraded the fastest in the lab correlated directly with an earlier age of ALS onset, and a significantly shorter disease duration in humans.
18:03That is wild. The biophysical instability measured in a single human embryonic kidney cell, successfully predicted the aggressiveness of a neurodegenerative disease in a human patient. It really is wild.
18:14And that predictive power fundamentally changes clinical practice. If a patient receives genetic testing tomorrow, and an entirely undocumented novel SOD one misense mutation is discovered, the neurologist doesn't have to commission a research lab to study it.
18:28Right. They can just pull up this published abundance map, locate the specific variant and immediately see its stability score. Exactly. And in the current era of precision medicine. An accurate and rapid genetic diagnosis dictates the entire treatment strategy.
18:41There are now newly approved, highly targeted anti-sense oligonucleotide therapies, such as Topherson, designed explicitly to treat ALS caused by SOD1 mutations. And those drugs work by halting the production of the toxic mutated protein, right?
18:58Yes. But insurance protocols and clinical guidelines generally require definitive proof of a pathogenic SOD one mutation before a patient can actually access the therapy. By systematically reclassifying these variants of uncertain significance.
19:12This proactive functional map actively removes the diagnostic barrier. It unlocks access to life altering disease modifying therapies for patients who would otherwise have no options. That is so inspiring.
19:22But um, we should probably look at the limitations of this massive undertaking. For instance, the VMP sec essay. Does gluing a massive, glowing, green fluorescent tag onto a delicate enzyme change how the protein behaves inside the cell?
19:37I think it's a 27 kilodalton tag on a 16 kilodalton enzyme, right? You're completely right. And fusing a tag that is significantly larger than the target protein itself drastically alters the overall biophysics and folding kinetics.
19:50It is the primary inherent limitation of the VMP sec methodology. Yeah, it has to cause some interference. It does. The tag can introduce spatial interference, potentially destabilizing a variant that might otherwise be stable, or, ironically, the tag itself might act as a stabilizing anchor for a variant that really should have degraded.
20:08The researchers absolutely acknowledge this boundary. However, the extraordinarily high correlation between the assay's final abundance scores, the known pathogenic status of established variants, and the clinical severity in human patients.
20:21Well, it proves that despite the bulky GFP tag, the system's successfully captured the core biological signal of instability. That makes sense. And there are also biological limitations with the yeast model used for the enzymatic activity assay, aren't there?
20:34Yeast cells are incredibly useful models, but they definitely aren't human motor neurons. Very true. Yeast possess a different suite of molecular chaperones, which are proteins designed to help other proteins fold correctly.
20:46Specifically, yeast rely heavily on a potent chaperone protein called HSP 104, which does not exist in human cells. Oh, interesting. So the yeast might actually be helping the human protein too much. Exactly.
20:59It's entirely possible that this unique yeast chaperone artificially assisted some of the mutated human SOD one variants in folding correctly within the heated yeast assay, which could potentially mask their true instability and toxicity.
21:13Furthermore, neither the yeast activity assay nor the human cell abundance assay is capable of measuring the complex prion-like spread of misfolded SOD1 aggregates jumping from one motor neuron to another.
21:24And that spread is a major driver of disease progression in a living human central nervous system. Still, even with those limitations, the scope of this is staggering. By proactively mapping the functional impact of nearly every possible SOD1 misense mutation.
21:40Scientists have really engineered a powerful biological tool that distinguishes pathogenic ALS variants with unprecedented accuracy. They really have. This exhaustive abundance map decodes the intricate structural vulnerabilities of the SOD1 protein, providing the immediate evidence needed to resolve variants of uncertain significance and accelerating access to targeted lifesaving therapies.
22:03It's a game changer for neurology. It truly is, which leaves me wondering, what does this massive shift from reactive testing to proactive mapping mean for the thousands of other complex disease causing genes currently hiding in our genomes and the 1000000s of patients still waiting for a definitive diagnosis?
22:20It's huge question. The potential here is limitless. This 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.
22:32If 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.
22:47Thanks for listening and join us next time as we explore more science-based by Base. Under white light, all night glow, we trade one letter. Watch it, show a protein tries to hold its shape, some changes fade, some changes break.
23:14Is it falling out of folding time? Or losing spark at the active line. Two shadows on the same disease. One map to help us finally see. We map the quiet, break in the chain. What disappears, what stays in flame?
23:31I guess work signals on the screen. Turning unknown into something seen. hold on. There's a pattern in the noise. A clear call. A steadier voice. Some sing too fast, degrade it gone, some keep their form, but miss the bond medals won't settle.
23:55Oops, miss a line, a dimer slips off the design. Heat maps like constellations, bright stability versus function in the night. Let's score that ways. What evidence means. So maybe finds a place between the quiet, breaking the chain.
24:13What disappears, what stays in flame? From east to cells, from coat to care, with the fog that lingered there. Hold on when the data turns to light. The next right patient. Guess the right fight.
