Single-cell multimodal sequencing of 120 AML cases reveals four immature-cell transcriptional clusters and splits NPM1-mutated AML into two immune-evasion classes with different responses to allogeneic hematopoietic stem cell transplantation.
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 today we're diving into a problem that is, um, both clinically devastating and frankly genetically baffling.
0:16It really is. We're talking about the unpredictable failure of aggressive cancer treatment. Specifically, acute myloid leukemia or AML. I want you to imagine a scenario, because this plays out far too often in hematology wards.
0:29You have 2 patients. They're diagnosed with AML, and genetically, they look like twins. Identical on paper. Exactly. Both have the same major driver mutation, this highly prevalent MPM one mutation, which, you know, should give them a pretty favorable chance.
0:43Right. And so they both get the same intense toxic chemotherapy, the same aggressive follow-up. Often a hematopoietic stem cell transplant or HSCT. And yet, patient A, they achieve years of disease free survival, a real success story.
0:57But patient B relapses. Yeah. Violently within months and succumbs to the disease. And the core genomic readout, the one that guided this entire incredibly expensive and toxic treatment plan is identical.
1:11That is the ultimate post. How. How can one, you know, tiny molecular difference. Something completely invisible to our standard tools dictate a life or death outcome. Especially when the core mutation is the same.
1:22The startling answer, which we're going to get into of this deep dive, is that our standard diagnosis has been looking at the cancer through a fog. A fog created by cellular noise, yes. And it's been missing these crucial stable subtypes that demand radically different treatments.
1:38So that's our mission today. To understand how scientists managed to cut through that fog using just unprecedented single cell detail. And for that, we have to celebrate the work of Henrik Lilgebjorn, though as Fioretos and their colleagues at Lund University, Institute Roche, and the Teralinska Institute.
1:55They've really advanced our understanding of what they call the AML cellular state space. And it's dire clinical consequences. Okay, so let's unpack this. AML is already a formidable foe. It's highly diverse.
2:06And its lethalities driven by these rare treatment resistant cells. We call them leukemia stem cells, or LSCs. That's the core of the clinical gravity here right? No, absolutely. Despite decades of progress, and I mean, the use of really intense chemotherapy.
2:21The overall five-year survival for AML is shockingly low. How lower we're talking? Just 24%. And for older patients, it can be less than 10%. The whole goal of this intense treatment is to eradicate those LSCs.
2:35And when that fails, relapse is nearly guaranteed. So we've relied on modern genoics, you know, the foundational work from groups like TCGA to categorize AML. The NPM one mutation is a huge part of that story, defining about 30% of all cases.
2:49It's a distinct subset, yes. But if we had all this sequencing power, why did we hit what you called the genomic wall? Why didn't bold sequencing show us these differences sooner? It's a classic case of statistics just completely obscuring biology.
3:03Bulk sequencing averages the gene expression across, you know, 10s of 1000s of cells in a sample. Like throwing every piece of candy in a jar into a blender. That's a great analogy. You can tell the overall flavor profile, sure, but you can't tell which individual piece was the, you know, the sour, deadly one.
3:19You lose all resolution. Exactly. The sources confirm that the samples we get from patients are never pure cancer. They're these heterogeneous mixtures. With lots of other cells mixed in. Highly variable amounts of mature non-leukemic cells, platelets, fibroblasts, normal white blood cells, you name it.
3:39And these normal cells, they have their own pretty loud transcriptional profiles. Precisely. The malignant immature LSCs are often rare, or at least they're highly diluted. So when you average the gene expression, the signal from the actual malignant clone is completely distorted.
3:56So if there are a lot of immune cells in the sample by chance. Then the bulk readout screams, high immune activity, and it totally masks the fact that the cancer cells themselves are silently shutting down their own immune signaling.
4:08So the key problem they had to solve was this. To understand the cancer's biology. They had to somehow ignore all that noise and focus only on the malignant, immature cells. That's it. And that's where the methodology became the real breakthrough.
4:22All right, let's talk tools. How did they actually achieve this surgical separation of the cancer cells from the noise? You can't just use a microscope anymore. No, they leveraged single cell RNA sequencing or CRNA sec.
4:35Which lets you measure gene expression one cell at a time. One cell at a time, but they went even further. They integrated multiple single cell approaches. That sounds incredibly rigorous. What was the scope here?
4:45Well, they analyzed an integrated cohort of 120 AML cases with standard methods 1st just for context. But the real innovation was applying single cell multimodal sequencing to a subset of 38 AML samples and 8 normal bone marrow sample.
5:03Okay, multimodal. What does that mean in this context? It means looking at multiple layers of information from the exact same cell all at once. So they use the standard CRNA sec for the RNA expression.
5:14Transcriptional readout. Right. And crucially, they also use something called CADT sec for protein markers. Ah, okay, so skaty TSEC isn't just a fancy acronym. It's critical because it confirms the cell type based on the physical proteins on its surface.
5:29Exactly. While the RNA sec tells you what that cell is doing internally. It's like a 2 factor authentication for cell identity. That's right. And this twin approach, let them identify with really high confidence, which cells were part of that AML immature population, in the end, they isolated and examine over 90,000 immature AML cells.
5:4990,000. That is a massive sample size for single cell work. It is. But wait a second. How did they know for sure that the cells they called AML immature weren't just very, very immature normal cells? How do you prove you've actually found the malignant clone?
6:04That is a fantastic and critical question. And this is where their validation was so elegant. They developed a custom assay they call Cesti RNA Mutsk. Z RNA Muxex. Yeah, this let them check the gene expression and confirm the presence of the known AML driver mutations, like NPM1, all inside those single cells.
6:21So it's the quality control step. Proves you weren't just looking at normal sales by accident. Exactly. It validated their entire approach, and the result was definitive. Practically 100% of the cells they classified as AML immature harbored the known AML mutations.
6:37Wow. They had done it. They had purified the biological signal and could now study the true undistorted expression profile of malignant cells. The payoff for all that rigor is right here. This discovery just fundamentally changes how we think about treatment.
6:53So what happened when they looked at that clean signal from the MPM one cases? The 1st major finding was really a confirmation of their premise, bulk sequencing had felled them. By analyzing only the gene expression of those isolated immature AML cells, the transcriptional clusters they found aligned far better with the expected biology.
7:12The noise had been hiding the differences. And what did that clean data reveal about the MPM one cases specifically? It's successfully subdivided NPM one mutated AML into 2 completely distinct and importantly stable classes.
7:26We'll call them class one in class two. Stable is a key word there. It's critical. They observed no switching between diagnosis and relapse. This suggests these are deeply intrinsic features of the cancer clone.
7:38So describe the 2 classes for us from a cellular perspective. Right. So NPM one class I was associated with a very homogeneous cell composition. The sample was just dominated by pure cancer. A median of 86% of the cells were those highly immature blasts.
7:54And crucially, these immature cells showed a really significant transcriptional feature. The down regulation of MHC class 2 genes. MHC Class 2. Just as a guide for our listeners, these are the molecules that are essential for the immune system to work.
8:11specifically for antigen presentation. They're the alarm system. They tell T cells, hey, the cell is infected or cancerous, come and eliminate it. So down regulations suggests class I is basically trying to become invisible to the immune system.
8:22That's its primary strategy. Then you have NPM one class two. This class was totally different. It was heterogeneous, a mixed bag of cells with a much higher proportion of differentiated, non-malignant cells.
8:33So fewer cancer cells overall. A lot fewer. Only a median of 31% were immature blasts. So class one is a pure cancer population trying to hide. And class 2 is this mixed population. This might seem academic, but the clinical implication is just staggering.
8:50How did this affect treatment? This is the massive clinical takeaway. This new classification directly dictated the treatment response to hematopoietic stem cell transplantation, or HSCT. And it held up across all the major validation data sets.
9:04TCGA, BDML, ClintSec, all of them. And for anyone listening, HSCT is not a trivial procedure, we're talking massive doses of chemo, wiping out the patient's immune system. And then infusing donor immune cells.
9:17It is toxic, high cost, and life-threatening in itself. You don't put a patient through that unless you're confident it's going to work. And this classification just shattered that confidence. It did. The results were black and white.
9:27NPM one class I patients gained a substantial survival advantage from HSCT. They showed excellent long-term survival rates. In class two. Dismal survival. NPM one class 2 patients showed virtually no significant benefit from HSCT.
9:40So the classification essentially predicted, whether the most aggressive treatment we have would be life-saving or a feudal, high-risk endeavor. Exactly. An unbelievable distinction that was completely hidden beneath the same NPM one mutation.
9:54It really reveals the failure of bulk genomics to give us true prognostic value. So what about the why? What are the mechanisms here? It's driven by two very distinct mechanisms of immune failure. Okay, we know Class one is trying to hide by silencing its MHT class 2 expression.
10:10They avoid elimination by becoming invisible. Right. But what about class two? The ones with the dismal survival. Why did the donor T cells in the transplant fail so completely against them? Because class 2 AML cells demonstrated an intrinsic resistance to the donor T cells.
10:27They weren't hiding. They're actively fighting back. Disabling the T cell. Yes. In vitro testing showed class 2 cells had a significantly reduced sensitivity to these alogenic T cells. After being co-cultured with T cells, 55% of the class 2 cancer cells remained.
10:42And for class one. Only 23% remain. That's a huge difference. It is. This intrinsic ability to resist the T cell attack is the molecular reason why HSET fails for class 2 patients. The donor immune systems simply can't kill cancer.
10:57So if class one is invisible, Class 2 is bulletproof. Is this resistance driven by the classic immune checkpoint pathways we always hear about, like PD1. Not exactly. The overall checkpoint profiles were similar, but the class 2 subtype showed specific upregulation of alternative immune checkpoint lightens on the AML cells.
11:16Things like CD 86 and Vista. They also saw a higher expression of inhibitor receptors like TM3 and Tigit on the T cells that were present in the class 2 tumors. So Vista, TM3, TJ. These are major players in T cell exhaustion.
11:32So instead of one simple immune break. Class 2 is using the subtle multi-prong attack to just wear out the donor tea cells. That sums it up perfectly. It's a fundamental difference in immune evasion strategy, and that's what determines the treatment outcome.
11:43So this single cell classification is clearly superior to just relying on commutations, which clinicians used to use to try and predict prognosis. Oh much better. Things like co-occurring DNMT3A or FLT3ITD mutations.
11:57They're often associated with class 2, but they aren't a perfect substitute, are they? They're not. The research confirmed it. While those commutations do suggest a worse prognosis, they only occurred in less than half of the confirmed class 2 cases.
12:11So the pure transcriptional state of the immature cell, its immune strategy is what really matters. That's the primary driver of patient outcome. Not just the commutation status. Now, I know the researchers face some limitations when they tried to use existing bulk data to validate all this.
12:27Can you tell us about that? Because it brings us right back to the genomic wall. It's a crucial point. They found that about a 3rd of the NPM one cases and those big validation sets couldn't be reliably classified using the bulk RNA set data.
12:40Why not? The reason was simple. Those samples likely had too few AML immature cells, probably less than 33% of the cells in the sample were actually leukemic. So when the malignant signal is that diluted?
12:51The cellular noise completely swamps the data. Even with this revolutionary signature derived from single cell data, the bulk sample is just too messy. So for the most challenging low malignancy samples, single cell sequencing is an absolute necessity.
13:06It is, but and this is the exciting part. The implication isn't that we need a massive single cell platform for every patient right away. The research successfully identified a custom minimal 30 gene expression assay that could be developed for rapid clinical use.
13:21A 30 gene panel. That's something that could be easily integrated into clinical practice. Exactly. It could be used to identify these 2 subtypes immediately upon diagnosis. Which would allow clinicians to skip a potentially futile, highly toxic HSCT for class 2 patients and confidently recommend it for classified patients.
13:40That is truly personalized medicine. The clinical utility is just immense. It moves us past simple genomic definition toward a functional, immune driven prognosis. So let's bring this all home. What does this all mean for us?
13:52Well, the study provides definitive evidence that the cellular complexity of AML, which was completely invisible to our best bulk sequencing methods, was masking critical, actionable biological features.
14:05And by focusing exclusively on the single cell profiles of those immature leukemic cells. They define 2 stable NPM1 subtypes. With opposite responses to our most intense AML treatment. And this distinction, it's all driven by different immune evasion strategies.
14:21You have the class I invisibility strategy down regulating MHC2. Versus the class 2 bulletproof strategy of intrinsic T cell resistance through these other checkpoint pathways. It provides a clear, actionable guide for personalized AML treatment that just did not exist before.
14:36And this raises a much bigger question for the entire field of oncology, doesn't it? It does. What does this mean for other genetically defined cancers, you know, certain lymphomas or solid tumors that still show these really unpredictable treatment outcomes.
14:48Could we similarly find other hidden, immune driven subtypes waiting to be uncovered by the single cell lens. And could that completely change our standard treatment algorithms for other diseases? I think it's very possible.
14:59That is definitely something to think about and watch out for in the future. This episode was based on an open access article under the CZBY 4.0 license. You can find a direct link to the paper and the license in our episode description.
15:13If 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 just heard about.
15:28Thanks for listening, and join us next time as we explore more science, base by base.
