2026-04-19-gastric-gastroesophageal-cancer-dynamics

Date: 2026-04-19 Primary topic: single-cell-spatial-omics Cross-referenced from: single-cell-spatial-omics Word count: ~11,842

Auto-generated captions

This transcript is Vimeo auto-generated English. Expect misheard names and technical terms. Use the ripgrep/search approach for exact quotes.

Good afternoon. Welcome to this Advances in Organized Research sessions. So the session is about gastric and gastroesophageal cancer dynamics, including pre-cancer evolution, metastasis, microenvironment, and therapeutic breakthrough. This is my disclosure. So gastric and gastroesophageal and esophageal cancer are very important cancers because there are more than a million case newly diagnosed each year and associated with very high mortality. The TCGA has published a very nice paper about the molecular subtypes along the whole upper gastrointestinal tract, from the esophagus to the gastroesophageal junction, down to the stomach. And up in the esophagus is mostly esophageal squamous cell carcinoma, which is a kind of very different tumor, whereas down in the gastroesophageal junction, you begin to see the predominance of the CIN subtypes of adenocarcinoma. When we progressively move up into the upper gastric body and antrum, the diversity of the cancer become increased, where there are some very gastric-specific subtypes like the EBV-associated gastric cancer, the genomically stable, or with substantial overlap with the diffuse histology types of cancer, and the MSI cancers, coexist together with the CIN subtype. So previously, treatment of gastric cancer or gastroesophageal junction cancer has been really limited to chemotherapy, mostly based on 5-FU and platinum. But in the recent year, there has been increasingly molecular biomarker, like the MSI subtype, being respond to PD-1 inhibitions. The PRD1 high tumor can be treated by anti-PD-1 as well. And then the recent emergence of the claudin 18.2 as a marker for antibody against it. This is a lineage marker for stomach tight junction cells. And then there is very exciting breakthrough in the HER2-positive gastric cancer, especially those with PD-1 expression. And we are going to have speaker talking a lot about this, which is really exciting development. And there are second-line and third-line treatment, but the efficacy is really limited, and we still need a breakthrough in this area. In parallel, there are technological evolution, including organoid models, new animal models, and multiomic profiling, as well as the single-cell and spatial transcriptomic that allowed us to develop deep understanding, both in the pre-cancer evolution and also in the tumor microenvironment. And with this background, today, we are going to have three talks. Myself talking about the pre-cancer evolution of the upper gastrointestinal tract. We have Dr. Jaffa Ajani from MD Anderson Cancer Center talking about molecular heterogeneity and also advances in HER2-directed therapy. We also have Dr. Yelena Janjigian from Memorial Sloan Kettering to talk about resectable gastroesophageal cancer, time to rewrite the standard. So I'm going to start with the first talk, and we will have Q&A section after each of the talk, and also a general discussion for all three of the speaker. So you can either ask question right at the end or leave it to the end of the general discussion. So my laboratory is a molecular pathology laboratory, and we are interested in the pre-cancer evolution as well as the molecular classification of gastric cancer. More than 10 years ago, we have done a very comprehensive profiling of gastric cancer and identified these four molecular subtypes, each having their own driver mutation, in which now we are seeing emerging targeted therapy towards each of these molecular subgroups. So in parallel, we have built organoid model trying to understand all these molecular subtypes. And we have now a bank that represent virtually all the molecular subtype, including some of those that are previously underrepresented in cancer cell lines like the ALK fusion model, the RhoA mutation model, as well as the EBV and some diffuse gastric cancer models. In parallel, there is evolvement of many different modification to the original organoid cocktail that adapt to different organs in which you can actually grow organoid from cancers from these various organs, epithelial organ especially. And in parallel, the exciting development is now because of this, you can actually grow pre-cancer models covering all these different organs, and I'm going to mainly cover the intestinal metaplasia model in this talk. SoIntestinal metaplasia is the conversion of gastric cell into intestinalized cells, and it is usually a consequence of chronic inflammation, and it has a wide prevalence, especially in Asian population. Histologically, it has been classified into the complete and incomplete type, and the incomplete type is said to be associated with a higher risk of gastric cancer conversion. However, study of this is limited because there is no representative sensitive 2D cell line model, and the animal model, especially in mice, only partially resemble human metaplasia. And mice mainly displace the so-called spam instead of true intestinal metaplasia. In collaboration with Mike Stratton in Sanger Institute, we have done a deep whole genome sequencing by microdissecting individual crypts in individual gland in the stomach. And through study of more than 200 whole genome sequence of individual glands, we observe the number of SNVs increasing with age in both non-cancer and cancer patients. So this is probably the range of normal, what you would expect. By the age of 80, you will have around 2,000 SNV mutation in each of the gland in your stomach. But in some of the gastric cancer patient, we see outliers with elevated somatic mutation, and these correlate with chronic inflammation, denoted in red, and also intestinal metaplasia. And indeed, if you look at intestinal metaplasia, for the glands that has it, it has around two-fold elevation of the SNV and three-fold elevation of the indel rate. So this is example of what we have dissected, a cluster of intestinal metaplasia glands and then some gastric glands. And when you trace their evolutionary branch, you can see that all these metaplastic glands actually comes from a common ancestor, suggesting that there is actually clonal expansion from a common ancestor, and the asterisk all along in these trees would denote intestinal metaplasia. So this is another example of a bigger piece of mucosa in which you can see that each of this red gland denote a metaplastic gland. You can see families of it around it, but we only dissect one to represent them. You can see they are completely independent in terms of their lineage, showing the heterogeneity, and the field effect is dramatic. And if you imagine you are a gastroenterologist, you can biopsy this area in which you get this IM glands cluster, and this area is a totally different cluster. And only in this one cm area of mucosa, you already get, I would suppose tens or twenties of different metaplastic clones, each with their own different driver or different mutations. So it is a very complex disease because of the heterogeneity. So this phylogenetic tree shows each patient in which the glands that we have microdissected, and you can see the predominant mutation, this clock line mutation of SBS1 and 5. But you can see in the metaplastic glands denoted in asterisk, you can see the blue mutation signature coming off, which is the SBS18, suggest to be linked to ROS injury. And then interestingly, this is the indel signature again, in the metaplastic gland, you see a change in ratio between the ID2 and the ID1 signature. An ID2 signature is more enriched in tumor, especially MSI tumor. So you can see in the early metaplastic process, you already get this progressive shift in mutational signature towards a more like a cancer. So we have studied extensively the biobank of organoid that we have collected in which we grow the organoid. And then with the transcriptome, you can see this is the metaplastic organoid. This is the reference gastric and colon organoid, and we characterize them by the colon and the stomach signature score. And you can see that for the metaplastic organoids, some of them actually retain the gastric signature but begin to express the colon signature, whereas other progressively get lost in their stomach signature. So I think this spectrum probably represent the complete or incomplete transition in the metaplastic process. And this organoid model, you can see they, like the glands that we dissect, they have elevated somatic mutation rates, they have copy number alterations, and these are the key drivers that we found. They don't have the very advanced cancer driver like P53 mutation, but ERIT1A mutation is pretty common. And the most dramatic change is actually in the epigenetics. So we can see in the metaplastic organoid, hypermethylation and downregulated gene and demethylation and upregulated gene. When we map this to the genes that related to the gastric or intestinal signature, actually, we found that some of the gastric signature genes denoted in black balls here arehypermethylated and downregulated, whereas these red genes are denoting the colon signature are demethylated and hyperexpressed. So the whole process seems to be highly epigenetically related. So the interesting genes, claudin 18, there is intense interest because the claudin 18.2 is now a target therapy. You can see they are expressed in the membrane of the antrum or body organoid, but in the metaplastic organoid, you can see progressive reduction in its expression. And probably this could constitute one of the escape mechanism to the claudin 18.2 targeted therapy, where as you can see induction of the metaplastic, the intestinal signature gene. So we perform single-cell analysis on the organoid culture, and you can see if we plot these individual organoid in the x-y axis, the colon and stomach score, you can see the organoid with a hybrid colon and gastric stomach score are in many of the organoid models. And here there is a normal gastric cells contamination, which are the gastric cells that has the gastric score but very low colon score. But you can see the heterogeneity in the spectrum. Some cells has more advanced intestinal differentiation, but other cells are more hybrid, and the lighter color is denote the cells after Wnt withdrawal when we try to induce differentiation. You can see an elevation of their score. So with this heterogeneity, then we pick individual clonal organoid and grow them and look at the spectrum of differentiation, and here is the proportion of hybrid versus more advanced intestinal cell in each clonal organoid. This is the reference gastric organoid. So you can see individual clone, each could have different proportion of differentiation into the hybrid or more advanced intestinal cell. So suggesting that actually these epigenetic deregulation constitute a spectrum. Each cell is different in terms of its epigenetic setting that whether they display more advanced intestinal or a more hybrid phenotype. So when we look at the different cell cluster, actually the cycling isthmus stem cells of the body organoid versus metaplastic organoid, you can see the key difference is the transcription factor of intestinal metaplasia are being expressed. But when we move progressively down, we can see this very interesting group we call pit-enterocyte, in which especially after Wnt withdrawal, when they differentiate, these organoid express the pit cell marker as well as enterocyte markers. So these constitute the hybrid cells that we are seeing. But we are also seeing other population like the cycling tuft cells, which is quite specific in the stomach, and then also the secretory progenitor expressed in the neuroendocrine marker. So with these, we can construct a 3D lineage trajectory of the normal gastric organoid versus the metaplastic organoid. So here you can see the cycling isthmus stem cell denoted in light green can goes to differentiate into the gastric mucus cell or through the cycling tuft cell into the secretory progenitor. And at the opposite end of the spectrum, these are the gastric chief stem cells. Whereas in metaplastic organoid, you can see they are more complex. So from the cycling isthmus stem cells, they can give rise to the Lgr5 stem cell and then the brown secretory progenitor. But on this side, it was the pit-enterocyte and the enterocyte that I will show you a little bit more in details. So this is the lineage trajectory of a gastric body organoid, and this is the cycling isthmus stem cells, which is very high in STMN1, which is we feel a very good marker. You can see you can stain them in the isthmus region, and they correlate with the high Ki67 labeling index. But from here you can see that the gastric organoid would differentiate into the pit mucus cell characterized by MUC5AC, TFF2 and claudin 18. So these are the gastric marker. But here is the interesting part, the metaplastic organoid. So for the cycling isthmus stem cells, so they still express STMN1, but they now express the stem cell, the IM markers. And then when they differentiate, you can see they can give rise to the Lgr5 stem cells, which also express the intestinal stem marker. But on the other hand, they give rise to these two different arm. One is more hybrid. You can see expression of enterocyte marker as well as gastric pit cell marker, which we denote the hybrid cells. Whereas the other arm show more advanced intestinal differentiation where you can get the more advanced intestinal marker, but reduction or loss in the gastric marker. So, in fact, it translate very well histologically for this hybrid arm to the so-called incomplete metaplasia, in which you can see theGastric gland expressing both Fabp, its intestinal marker, and Mu5ac. And the cells look transparent, with goblet cell in between, so characteristic of so-called incomplete IM. Whereas this arm corresponds to the more advanced enterocyte differentiation, in which the cell do not express any gastric pit marker at all, but mainly intestinal markers. So we find that this hybrid cell is very interesting. We will discuss it later. So here we also try to look at where is the origin of the metaplastic cells. So in the same patient, we pool the body organoid and the metaplastic organoid together to perform a lineage trajectory analysis. And you can see that the STM1 cycling isthmus stem cell seems to be common between the metaplastic and the body organoid. But the metaplastic organoids that differentiate into the enterocyte on this side and then the Lgr5 stem cell on this side. Whereas the body organoid differentiate into the gastric pit cells. So here, correspondingly immunohistochemically, you can see that this is the isthmus region with Cdx1 expression only on the upper part, whereas the lower part is not. And then you can see the Stmm1 expression here, the upper part with the metaplasia, whereas the lower part retain the gastric phenotype of Musix and Progestrine. So suggesting that actually IM start in the isthmus stem cells and move progressively upward. So when we examine the lineage plasticity of these hybrid cells, here we use the single cell and delineate their different proportion of cell types signature. And we find that the metaplastic cell, apart from expressing both gastric and intestinal marker, actually retain other, for example, epithelial non-GI signature and even some mesenchymal signature. So they remain high in the diversity, whereas the advanced intestinal organoid are more mature and unidirectional into the intestinal fate, whereas the gastric organoid unidirectional into the gastric fate. Interestingly, this arrests in the hybrid state, apart from display higher lineage plasticity, also display a fetal signature reversion. So you can see in this arm of metaplastic cell, they express this fetal signature, the fetal gut signature, including Tacstd2, which is also called Trop2, which is a very interesting target now with the ADC targeting these fetal markers. But beyond this, it also express other fetal marker. So it looks like this metaplastic process are actually undergoing fetal reversion with the lineage plasticity. So lastly, we examine whether our IM organoid cell cluster or body organoid cell cluster are more similar to the normal or IM gastric tissue or gastric cancer tissue. So you can see this is normal gastric tissue. So they preferentially map to the body organoid tree, whereas the IM gastric tissue preferentially map into the IM organoid tree. But interestingly, this is gastric cancer tissue. You can see they cell map to body organoid tree, but they map to a lot of the metaplastic organoid tree where individual nodes denote different cell cluster, as highlighted in this pie chart. So you can see the gastric tissue, cancer tissue actually show the higher similarity to the IM cycling isthmus stem cell, as well as many different cell cluster in the metaplastic process. And a minority actually looks more similar to colon, and only a very small subset resemble the body organoid. And you can see in the metaplasia and gastric cancer tissue, these cells also showed a elevated fetal and the Yap stemness score. So, we then examined the chromosomal instability status. You can see some of these organoid display trisomy of chromosome 20, and as well as in the normal body organoid. But if you look at the distribution, they are quite widely dispersed. For example, here you can see a chromosome 20 gain in one of the gland and a chromosome 13 gain in another gland. So this is another extreme example in a 64-year-old stomach cancer patient. 75% of the gastric gland already acquire an extra chromosome. So the orange denote one haplotype and the blue denote the other haplotype. So you can see these are independent events. So sometimes it's the paternal, sometimes it's the maternal haplotype that get the gain. And it is very widespread. When we date the occurrence of this chromosomal gain, it actually dates back to when the patient was very young, and in this case, around 12 years old. So these are the different cases with the trisomy, and you can see they all happen in a pretty early stage.So these organoids, or these glands, are wild type for P53. So you can actually see onset of chromosome instability in a P53 wild type background. So coming back to the organoid model, we find that the change in the metaplastic organoid become independent of FGF10, and they have reduced FGFR2 expression. And when we examine the organoid with chromosome 20 gain, we found that they actually can survive without Matrigel for longer. So, because of this observation, we start to withdraw the Matrigel, and we found that in a few metaplastic organoid, we actually get growth that can actually grow independent of the Matrigel. And when we examine them, we call this spheroid. When we examine them, they actually get even a higher copy gain in chromosome 20 in some of the case. But all of them do have chromosome 20 gain. And we did single-cell analysis, and again, showing the enrichment of this subpopulation with the CIN extra gain of chromosome when we put them in Matrigel, so the whole population is constitute those with chromosome 20 as well as other chromosome gain. And this is another two cases, as example. So, when we look at this metaplastic organoid that can survive without Matrigel, we found that they have increased EMT markers and stemness, et cetera, and they also display tumor-like phenotype by the transcriptome. So, in summary, whole genome sequencing of microdissected metaplastic glands revealed increased mutational load, marked heterogeneity, and parallel evolution of individual clones, as well as frequent CIN in the absence of P53 mutation. And the organoid biobank can actually capture these diverse array of metaplastic gland harboring heterogeneous genetic and epigenetic perturbations. Single-cell RNA-seq revealed IM organoids are spanning a spectrum from hybrid to advanced intestinal differentiation, recapitulating the incomplete to complete metaplasia spectrum in tissue. And they show impaired differentiation potential, higher cell lineage plasticity, as well as fetal reversion. We have also shown the putative origin of IM from SDMN cycling isthmus stem cell and gastric cancer cells in tissue show high similarity to metaplastic cycling isthmus stem cell and other metaplastic cell populations. Finally, we identify a link between chromosome 20 gain and cell matrix independency in metaplastic organoid, potentially working in concert as one of the earliest tumor-initiating event in gastric cancer development. So, in parallel, there is also a very exciting evolvement of the understanding of the metaplastic process through spatial and in vivo transcriptomic and also single-cell analysis. For example, this study from the Vanderbilt group show that there is, in parallel to the metaplasia, there is the PDGF-alpha subset of fibroblasts being expanded. And co-culture of these metaplastic fibroblasts can actually promote dysplastic progression. And the Singapore Gastric Cancer Consortium has also published a lot of multi-omics data, including here suggesting that the intestinal stem cell is highly linked to early gastric cancer and also mutational signature of SBS17 in IM and mutation count, ARID1A mutation, and high clonal hemopoiesis associated with increased risk of IM progressions. So I'm going a little bit out of time now. So, we start off with pathology. We understand the tumor by their morphology, molecular, underlying genomics. Now we have a living biobank model of all these organoid. And progressively now we can even capture the pre-neoplastic organoid. So with these, I think, there is hope for therapeutic intervention, especially in the pre-cancer area. And there has been two approach previously undertaking, like using an anti-HER2/neu mythic agent or using a MYC inhibitor for targeting pre-cancer. So hopefully with all these, we will see a prevention of gastric cancer becoming a good possibility. So I would like to thanks my collaborators, especially Helen Yen, who actually co-lead this study with me, and Sarah and Yintong are the postdoc working on the metaplastic project, and Mike Stratton, for the collaboration. So I'm probably running out of time for questions. So, maybe I will leave it to the last general discussion, and I would like to invite the second speaker, Dr. Ajani. Good afternoon, everyone. I want to thank Dr. Leung and ACR for inviting me.Can you put my slides on? The second talk. Yeah. Thank you. So I'd like to start with examples of other tumor types than GEA to demonstrate the advancement that has already occurred in the defining metastatic molecularly. So this is breast cancer, and here, in the metastatic process. Can you see the pointer? How do you get the pointer? This is the mouse itself is a pointer. Yeah. The mouse. Okay. All right. I'm just going to talk you through this. So on the right side of the slide, you see the preferential mutations that occur in metastatic breast cancer, particularly ESR1. And in the middle, you see the copy number variations that have been defined in metastatic process. At the bottom right is the various pathways have been well defined, and they can be targeted. In prostate cancer, we heard from Dr. Sawyer today. On the left side, you have the primary tumor with a variety of alterations. But on the right side is the androgen receptor and P53. And androgen receptor is amplified in almost 60% to 70% of prostate cancer in certain organs. So as far as GEA is concerned, we really don't have this kind of definition of metastases. So what we do in the clinic is, patient comes in with an endoscopic specimen, or we collect them, and we do profiling biomarkers, NGS, all this, and then we treat the metastases. And I'm not suggesting we biopsy all the metastases, but we haven't really defined them very well, and this may become important in the future. So in the interest of time, I was not going to review the literature by other investigators. They really contribute a lot to our understanding. But if you look at the whole picture, it's very limited. So in 2020, we published this paper on peritoneal metastasis. We've been working on it for almost 15 years. And there you can see on the right side, right lower side, that if you have peritoneal metastasis and you look at location of the primary, whether signet ring cells are present, whether it's a grade two or grade three, whether it is intestinal type, diffuse type, there's a lot of heterogeneity. Different genes are expressed. And we didn't have primary tumor at the time, so we took the TCGA data, which is entirely from primary tumors. And then when we look at the comparison between TCGA primary tumors from other patients and then our patients, there are two genes that stood out. One is the TAF1 and RPL5, and these two genes have already been defined by the pan-cancer as driver of these tumors. So we were able to then collect specimen from the primary tumor, and here we have 41 patients, and some healthy donors. I won't go through all of this because there's a lot of data in this paper, but we have normal tissue, chronic gastritis, intestinal metaplasia, primary tumor, liver mets, peritoneal mets, and ovarian mets from the same patient. And we are looking at the immune cell and stromal cell. We're not looking at epithelial cell in this particular paper. And you can see throughout that if you look at the primary tumor versus metastasis, stage of the tumor and various features, there are a lot of differences. So this is one of the examples of how diverse the stromal component, the TME component, is from the same patient in different locations. We were also able to define these three ecotypes, I'm sorry, six ecotypes, and their cellular compositions as well as their prognostic value, and then also whether you can add immunotherapy and get an advantage. So these are also widely distributed from the same patient's tumor in different locations. And then I think Dr. Lung showed this slide where we identified SDK, Syndecan2, in the CAF cells that are highly prognostic in primary tumor as well as in metastatic tumor. So at the lower left side, we have the in vivo model where you take CAF cells and tumor cells and co-co-inject them in mice, and half of the mice did not get CAF with syndecan-2. And on the right side of the lower left side, you will see where CAF cells were expressing syndecan-2, and the tumor grew very rapidly. So this is just one example of how CAF can promote tumor progression. Then more recently, we published another paper in Gastroenterology. Here we have 20 patients, stage four gastric cancer with the primary tumor available, liver metastasis available, and peritoneal. And in the second or in the middle, you see the major cell types. Here we included epithelial cells, so cancer cells are included. And then on the right side, if you look at different components such as blood, adjacent normal, primary tumor, ascites, and liver metastasis, you can see the compositions. They are quite different. And this is from the same patient. This is another example of one patient where you have the circus plot predominantly looking at various pairs. One pair I want to mention is the macrophages type two and CD8 positive cells, and you can see considerable differences. Then the genius of Pawel Mazur, who is one of our collaborator, first established this model of CAR T cells in mice using the mesenchyme as a target for CAR T. But then he also identified GPX4 elevation in tumor cells. So if you look at the mouse panels, with the CAR T, you get some advantage. The IKE is the GPX4 inhibitor. But on the right side of the panel, when you inhibit GPX4 in tumor cells and then add CAR T cells, you get dramatic improvement. So today we heard from Dr. June that if you inhibit CAF cells and then do CAR T cell, you get an advantage. Here we are targeting tumor cells in two different ways. So in spite of all this heterogeneity, we've been able to make some advances in the clinic. And you have to wonder, how does that happen? Okay, so these are the individuals who have helped with this. On the left side is my team, and this would not be possible without Ruping, who's the first author of a couple of papers I showed you, and Lingua, who is the bioinformatics leader. So, let me just go on to talk about HER2. I wasn't going to plan to explain everything about HER2 because it can take a lot of time. So let me talk about the new drug, which is a bispecific antibody, and that is anidatamab. It's intentionally an asymmetric bispecific antibody that tends to conform to trans configuration. In trans configuration, it can engage two different receptors. So one arm will engage one receptor, another arm will engage another receptor. And you can see the transmission electron microscopy on the right side. So anidatamab can engage two receptors at the same time, and if you think about thousands of these bispecifics engaging HER2 receptors at the same time, you will get cluster formation. And cluster formation has very specific consequences that I'm going to mention. So this is just a trajectory of what can happen when you inject anidatamab. You engage HER2 receptor. There is little bit of clustering, and then on the right side you have cap formation, and then finally you have hexamerization of the molecule. This is very important for complement-dependent cytotoxicity. So this is an example of cap formation. So you have negative control on the left, and the second lane is anidatamab, and you can see those silver color thick caps that are being formed on the surface of the cell. And then that doesn't happen with trastuzumab, doesn't happen with pertuzumab, and doesn't happen with the combination of trastuzumab, pertuzumab. So you get this kind of cluster where the Y portion of the bispecific is engaged to the HER2And the FC portion is sticking out. And so there's a high clustering of the FC receptors, and this will invite C1q, which is a key protein for complement-dependent cytotoxicity. So this is another two different methods to show that zanidatamab induces high level of cap formation or cluster formation that other HER2-directed therapies do not. So you've got control on the left, two lanes with two different methods. Zanidatamab, you can visualize the caps, and then trastuzumab or pertuzumab and the combination. So when you have this cluster and hexamerization, the CQ1 protein really needs to engage six FC receptors in order to invite the C3 protein component to enact the cytotoxic cell toxicity. So as opposed to other HER2-directed treatments, zanidatamab can induce ADCC. It can induce ADCP, which is the macrophages are engaged. And finally, complement-dependent cytotoxicity. So there are a number of advantages for this bispecific antibody that others do not produce. And as a single agent, this is hard to see, but the paper is listed at the top, so you'll be able to get to it. But as a single agent in the fourth, fifth, and sixth line, zanidatamab produced a number of responses that were also lasting for a long time. This is something we didn't see before availability of zanidatamab, because trastuzumab cannot do this, and chemotherapy cannot do this. Then the phase two, which combined zanidatamab to chemotherapy, also produced dramatic clinical responses, and some of them were very long-lasting. It produced diarrhea when you combine with platinum and fluoropyrimidine. This led to the trial Horizon GEA-01, which had three arms. Arm A is the control arm. Arm B is the zanidatamab with chemotherapy. So similar to control, except zani instead of trastuzumab. And Arm C was zanidatamab, tislelizumab, and chemotherapy. The primary endpoints of PFS as well as OS. And this is the PFS endpoint, which the comparison is between trastuzumab chemotherapy versus zanidatamab chemotherapy. And you can see on the right upper side that there is a four-month advantage in PFS, which is also quite remarkable. Hazard ratio of 0.65, very strong P value. And you can see throughout the journey of the patients on these curves, there is a considerable separation. When we look at the other experimental arm to the control arm, in this case, zanidatamab, tislelizumab, and chemotherapy, again the median PFS is similar with or without tislelizumab, but all the numbers at different landmarks are better than you get without tislelizumab. So the area between the two curves is greater with tislelizumab than you get without tislelizumab. So in terms of overall survival, without tislelizumab, zani plus chemo versus trastuzumab chemo, you have the advantage towards overall survival, but this is not statistically significant now, but we expect it to become significant with further follow-up. But when you compare zanidatamab, tislelizumab, and chemotherapy to chemotherapy plus trastuzumab, it is statistically significant, 0.72 hazard ratio, seven-month difference in overall survival. And at every landmark, there is an advantage. And hopefully, there will be a better tail to the curve that we normally see with immunotherapy. So in summary, I think we have made considerable progress in the arena of HER2 targeting, especially in the frontline setting. So, the second column where the numbers are either blue or black. So blue is the experimental arm. So you saw 10, 12 years ago, the experimental armThey had the median survival of 13.8, but also look at the hazard ratio, 0.74. And then the black numbers are the control. So if you just run down in the second column to the black number, you can see even the control is getting much better. Our patients are doing a lot better in the clinic, even with the standard of care treatment. But the blue also continues to increase. Same thing for PFS, et cetera. So this is my last slide. We can say that dinutuximab, with its shortcoming of diarrhea, and we have to be very vigilant about its management, should be able to replace trastuzumab. And perhaps zani, Tisley, and chemo will become the standard of care. And I didn't show you the data, but this would be irrespective of PD-L1 expression. So with that, I want to thank everybody for being here today. Thank you. So there is five minutes for question, if you have questions now. Yeah. Can you go to the mic? Very nice talk. First question to Dr. Leung regarding the-- Do I pronounce right? I'm not so sure for you. But I think you both or only Dr. Johnny's right now, the question. Yeah. Because for me, leave to the end. Oh, okay. Regarding Dr. Johnny, I'm wondering the trial for HER2, the new antibody, do they have a biomarker-guided selection patient or all comers? No, these are patients who are, by definition, are HER2 positive. So they have to have three-plus protein expression, or they can have two-plus protein expression, but they have to have increased number of ERBB2 copies. Okay. That mean they have a biomarker-guided, either amplification or overexpression HER2, right? That's right. And this is not highly common situation. It's about 10% to 12% of, I would say, US patients have HER2 positivity. Most patients don't. Okay. Thank you. Okay, so is there any other question for now? So if not, maybe we have the next speaker. Thank you, Dr. Ajani. So our next speaker- ... is Dr. Elena Jamcijan. Well, thank you for the opportunity, and it's always my favorite topic, discussion of how we can change practice for our patients. I am a medical oncologist, and my focus has really been on bringing new treatments to the patients, but also being able to learn from every single patient from our translational program. And I'll showcase some of the work that my group has done and give you a little roadmap. Today is the advantage of we still have the whole weekend of the meeting, so I'll give you a sense of other abstracts that you may want to focus on. So of course, we always have to bring it back to the patient. And increasingly, we are seeing this disease in younger and younger patient population. It's not infrequently that I'm seeing patients younger than my age and often sometimes even closer to the age of my children. So this is a typical patient now from my clinic, a 37-year-old man who presented with months of swallowing difficulty and weight loss. In these patients, we see delayed diagnosis because of course, the practitioners in the community are not expecting these patients to have cancer. Eventually, he had a imaging when we usually get FDG PET scan. Our disease is very much PET avid, so it lends itself well for functional imaging. Again, addressing some of the questions that were raised about ability to detect heterogeneity. And we also did an EGD, and of course, there's bulky tumor. And as you could see, it's very visible. And typical biomarkers we do in the clinic now routinely are protein-based expression, loss of expression of mismatch repair protein, gain of HER2 overexpression or amplification by Fish, PD-L1 overexpression, and claudin, as well as Epstein-Barr virus, which can also cause this disease. So in this patient, this was an MMR proficient tumor, CPS relatively low, one claudin positive, and again, overwhelmingly, at least in our clinic, these are p53 mutant tumors with very aggressive types of TP53 mutation that do lead to chromosomal instability. So what we're here today is to talk about curative disease. What you heard from Jafar is metastatic patients first-line setting. We're talking about possibility of cure and cure with patients that they can-- As you know, young people want to just go back pretending they've never had cancer. I mean, all of us want to do that, but young people in particular. So how do we cure more patients but also minimize the fallout of our treatments? So surgery alone, at least in the Western population, really results to minimal likelihood of cure because this is a metastatic disease at the get-go systemically. High risk of micrometastasis, even if you can't see them on imaging.And perioperative chemotherapy improved survival by a bit, but not sufficiently. Historically, this was a very fractionated field. In Asia, people would get surgery first and then have chemo. In the United States, we did some chemo perioperatively, but again, with very limited efficacy. So the big breakthrough in the last year was that we're now able to unify our approach, and even doctors in South Korea and Japan, Taiwan, and China are agreeing that doing chemotherapy and chemo immunotherapy first is the best option, and this is because of a large phase three study, Matterhorn study, that changed practice showing that combination of durvalumab with three-drug chemotherapy, FLOT, improved survival. The other key feature of the study, in addition to saying people all over the world can get this regimen, and we can move the field forward in a way where the comparator arm is standardized across different regions. We also showed that if you have a major pathologic response or a complete pathologic response, this will translate to survival improvement. In other words, you can have a quick result and a litmus test that could give you a sense early, instead of doing a large phase three study, whether or not the treatment that you're studying in early phases, phase two perhaps, single center study, you can get a sense of whether or not you're going to have achieved a meaningful outcome because these tumors are bulky, and if you're able to eradicate the tumor, meaning pathologic complete response, which is we see 19% of the patients with durva plus FLOT, or even a major pathologic response with less than 10% residual tumor cells, you're very likely to have a cure. So it's a very satisfying thing for us to do in the clinic after the patient is there to be able to tell them the likelihood of cure is very high and you should not worry. The other important factor that we did, because there were other studies trying to study this in adjuvant setting, we demonstrated that neoadjuvant treatment is very important. After the diagnosis, everyone has to calm down, stop, and actually not do surgery, because that was a very counterintuitive thing to do for the surgeons, certainly even for the patients. Once the patient is diagnosed and if they're healthy, they just want the tumor out in the bucket somewhere. But that's not the right thing to do. We were doing it all wrong. Taking the cancer out is a bad move because first you don't have the tumor in place to know that what you're doing is successful, right? You're only treating a hypothetical MRD sort of setting. But also you're removing the education, the educator, the tumor from the patient and the amount of antitumor immune response, and diversity of that response is really impaired if you're taking the cancer out. So everyone just needs to sort of take a breath, send the patient to the medical oncologist to be able to improve the chances of cure. So let's go back to our patients. So we know then we can impact improved survival in some patients, and this is actually the patient that I was showing you. I got his permission to share. He's very proud of his outcome. He had a complete response. So he had FLOT, durva, went on to have a YPT zero and zero tumor, and unfortunately, we took his esophagus and GE junction out anyway, so he had an Ivor-Lewis esophagectomy and went through a horrible divorce through the whole thing, through the treatment, and now he comes back four years later, cured. The only thing he complains about is having to sleep on a wedge and also the dating scene in New York. So it's very tough, apparently. But I think in this patient, if we knew he was going to be a complete response, perhaps we could have left him with his organs in place. I think he would want that. So the next frontier is to figure out who are the complete responders that we can leave and deescalate and do organ preservation. Perhaps we don't need to develop grade two neuropathy and not be able to text because of all the chemo we're giving them, and on whom we can escalate the therapy. So, you will see a great presentation on Tuesday during the plenary on the updates of the INFINITY study. I'm not going to spoil everything, but I did want to showcase the work that was done by Gano group, which was very important, showing that the total neoadjuvant immune blockade with dual immune checkpoint blockade using tramidurva can spare patients, and you can do organ preservation in early-stage gastric cancer. In fact, again, going back to my whole point of doing all the work while the tumor is in place is critical because we actually did a study at MSK where we said, "Okay, most MSI tumors are cured with surgery alone. So if they already had their surgery and they did not get neoadjuvant immunotherapy, and if they're MRD positive," which happens in about 15% of patients, in other words, after surgery, they're ctDNA positive, "let's give them immunotherapy and let's see how they do." Well, we did. We gave them pembro, and we did well. We cleared everyone's ctDNA. But as you see here from the curve on the right, 38% of patients still recurred despite ctDNA clearance at six months. So again, showing that once you're MRD positive, and if even in MSI tumors in adjuvant setting, you're treating metastatic disease. So you really need to focus on these tumors while they're still in place and risk stratify better. And I think in MSI tumors in particular, dual immune checkpoint blockade may be sufficient.So what are we doing? You just heard nice focus on zanidatamab data. While KEYNOTE-811, of course, changed practice in metastatic setting several years ago. So we're taking the KEYNOTE-811 regimen, which is the pembrolizumab, trastuzumab, FLOT regimen, to early-stage disease in neoadjuvant setting. And the key points of this study is that we're giving even more treatment neoadjuvantly while the tumor is in place. So instead of doing only four cycles of chemoimmunotherapy and HER2 therapy, we're doing six, and that's been tolerable. And the protracted course of neoadjuvant therapy allows a maximum time for tumor response and again to mount the anti-tumor immune response. And again, similar to other approaches now, we're allowing non-operative management. Patients expect it. In fact, if you do not have a non-operative management in your consent form, it makes it a difficult consent form to sign with the patient. They want to at least have an option. And of course, longitudinal sample collections is critical in this study. So what we already saw, and initially we had a doublet regimen, CAPOX, compared, and then we amended the study. So this is not randomized by any means, but just showing you here to give you a sense of what we saw. Excitingly, we already seeing 83% complete clinical response. So small numbers, early data, but FLOT pembrolizumab is very doable and it's doable with six cycles per op. And CAPOX, again, underperformed. So we won't have time to talk about it, but the combination with taxane does something to tumor microenvironment that makes the responses deeper and more likely to have complete response. So what we're trying to generate now is finish this data set. It's going to be important for translational work, but also develop non-operative management strategies to stratify to know upfront which patients will be non-responders. The good news is out of our cohort of patients, everyone who had a complete response, there were only three people who elected to have surgery. They also had pathologic complete response. So at least using our modern tools of endoscopies, PET scans, serial biopsies, and ctDNA, we're good at saying that a clinical complete response translates to pathologic complete response. So now, of course, the portfolio is expanding. AstraZeneca has an important study looking at other biomarker selected strategies using claudin ADCs, HER2 ADCs, and of course, combination of bispecifics all in perioperative setting, building on the Matterhorn strategy. And we're doing an institutional small study, a single institution study, kind of Matterhorn accelerated, looking at PD-1/CTLA-4 bispecific with catumaxomab. So what worries me though is, of course, not the over-treatment, because that's important, but also the patients who are not responding, right? We talk about de-escalation, but what do we do for the majority of patients who will not be achieving complete response or major pathologic response? Who are those patients? How can we help them? Well, so we know that the big bad elephant in the room, and this is data we published from CheckMate 649, which was a large Phase 3 study, showing that majority of chromosomally unstable tumors, which is the cancers that we see in the West predominantly, don't benefit from immune checkpoint blockade in first-line setting. The hazard ratio is 0.92. The dogma is all to focus on PD-L1, and people will fight to death about what cutoff of CPS one to four or CPS less than 3.5 we should use. But really the elephant in the room is that it's something beyond PD-L1, and PD-L1 is just another sort of easy thing that we can assess. But chromosomal instability drives immune evasion in many of our patients that we're seeing in the clinic. So this is plug number one for one of the talented bioinformaticians and scientists that work with my lab. So Jerry Zhu is presenting this abstract today looking that, of course, everything is a spectrum. Nothing is as simple as just calling someone chromosomally unstable tumor. And we identified six reproducible CIN architecture features that define aneuploidy, focal amplification, and subclonal evolution. And really the purpose of this exercise was to use the facet analysis from close to 2,000 cases that we already did at MSK, but also the huge rich database of the CheckMate 649 whole exome data that BMS shared with our group to put in a framework, a unified clinical molecular framework to standardize how do we define CIN assessment to enable clinical patient stratification in the future, both in first line, second line, but also in early-stage tumors. The other important feature we saw is that in patients-- So we've been using chemoimmunotherapy in early stage disease well before the approval based on Matterhorn. So we amassed quite a group of patients mostly in standard practice who got chemoimmunotherapy. Now we're seeing some of the recurrences. So Dr. Chak and my group looked at genomic instability and predicted recurrence after perioperative immunotherapy in gastroesophageal cancer. So check out the poster. This small snippet of the data does not give it justice, but there is a story on, again, giving chemoimmunotherapy while the tumor is in place still, and even inducing a second immune response after recurrence, how CIN tumors affect that, and how the timing and duration of treatment may be important. So this is an important factor for us to consider, particularly since, of course, in the age of Matterhorn, we have to decide what to do with these patients when they recurSo what we're also seeing is that even if you induce-- So once you've been immune, sort of exposed, and if you're trying to induce a second response, sustained treatment induces increase in interferon gamma, but it does not-- So you're seeing what you would think would be adequate response, but it does not translate to clinical benefit. And this is from a phase two study that Sam Citrin in my group is running. Again, come check it out. This will be April 20th at 2:00 PM. So why is that? Well, what we're seeing, and I know Sam Baccom and Ashley Loughney gave a great talk yesterday, but chronic tumor cGAS associated with immunovasion and poor survival, and it's important, particularly in our cancers. This is data from breast cancer showing that if you have acute cGAS activation, it's potentially a good thing and can induce immune response. But what we often do, and what happens in the setting of CIN tumors, it burns out to chronic activation, which actually causes suppression of STING, and as a result, immunovasion. And this data from breast cancer patients showing that if you have, as you see here on the bottom red survival curve, cGAS high and low STING, it's a really bad thing for patient survival and outcomes. And now we're seeing the same in gastroesophageal cancer. Again, we see chronic cGAS high tumors, and we're seeing STING low. And what we're also seeing is there's a tumor stromal uncoupling between cGAS-STING axis, where we're seeing different expression. And I think what earlier discussion about heterogeneity and importance of tumor microenvironment in the stroma. In fact, again, going back to are we really looking at the right biomarkers, cGAS-STING activation appears to be, at least in our limited cohort that we looked at so far, a stronger predictor of poor prognosis in patients treated with immunotherapy than chromosomal instability itself. So considering it's potentially a druggable target, it's really important for us to continue to study this and characterize it. We're all together, sort of international group of researchers focused in the global engine for target discovery. MSK is just epicenter of it because we happen to both to have a big translational program, but also run phase three studies. And it's great to be able to collaborate with several industry partners who are then able to share with us raw data from some of these studies and use it for target discovery using newer AI tools within silica medicine, but also global collaborations. So we meet twice a year. It's both in person and virtual, and if you're interested, please feel free to ping me, and you'll be invited to the group. Our next retreat is coming up May 4th. And thanks to ACR and Stand Up to Cancer, many of these efforts are funded through various mechanisms, and these are some of the collaborators we pulled in over the years because this disease is definitely in the United States, a relatively orphan disease, where we don't have big foundations throwing millions of dollars at it. And so to have some key aspects of this funded through ACR and Stand Up to Cancer, Turkoos, Gemini, Debbie's Dream Foundation has been amazing. So in conclusion, what we're seeing here is a shift in paradigm where we're able to push the envelope and cure more patients. Because I think historically, this focus on palliation and just prolonging survival, this nihilistic approach has been centered in our disease. But now we're seeing even in metastatic, but certainly in early-stage disease, people are being cured. And what drives these cures is early immune priming, generating deep responses that will enable organ preservation. What we need to really understand is the tumor biology, including the spectrum of chromosomal instability, innate immune signaling, and that will shape immune response and interrogate all the biomarkers dynamically as we're really understanding how these tumors adjust and change while on treatment. And only working together in collaboration can we do this. We have an amazing team at MSK. I see patients in a day, 30 patients in a day. So really it takes a village to cure these patients. Really dedicated symptom management nursing care is a big part of this. And so thank you for attending the session, and I bought us four more minutes, so we have more time. So I have a few questions from the floor as well. Maybe I would like to invite Dr. Jenny to come up. We can have a group Q&A. So first, I have a question from the internet audience for Dr. Yelena. In view of current guidelines, do you offer to diffuse TPS less than one and or node negative? Yes. The answer is? Yes. Okay. So the second question. In view of the negative results for KN-525 and innovation studies, what would you assume make the pembro trastuzumab combo work? Well, KEYNOTE-585 used a doublet, and we are seeing combination therapies with FLOT-based regimen. I think if we wanted to talk about things why 585 was negative, we would need another 20 minutes. But in short, it's the combination of trastuzumab, pembrolizumab together with FLOT that is what's driving these responses because if you remember, our initial Phase 2 data of CAPOX, pembro/trast showed 100% disease control rate, and even in the large Phase 3 KEYNOTE-811 study, the complete response rate was 20%, so in the metastatic setting. So really, it's not a difference between, I think we need to be very clear. In this case, anti-PD-1 therapy, anti-PD-L1 therapy, these agents are equivalent. It's really not because of pembro versus durva, why Matterhorn was different. It's the change in the backbone, but also in a clinical trial design. It was much more simple, and I would say elegant design. So there's questions or- Christoffer has opinions about that . Yes. So I personally don't believe that addition of taxane would change the outcome like that in context of immunotherapy, because there are many trials from China now without a taxane showing benefit. So I think going forward, we will have to focus on what you mentioned, is the more underlying biomarkers in addition to a PD-L1. Philip Dennis from Novo North- I would just clarify that, of course, I think there's a different in biology. We're dealing with G junction tumors, and I think there may be some difference in that. But please go ahead. Philip Dennis from Novo Bridge. I have a question about claudin 18.2 expression. For Dr. Leong first, in your metaplastic organoids, claudin 18.2 was decreased. So is cell surface 18.2 really only evident when cells are fully transformed and when truly malignancy is present? So I think claudin 18 actually is a lineage marker. I don't think it's a neoplastic marker. So is that- Well, its subcellular localization changes, right? That's the dogma, that it's a tight junction- Mm-hmm ... protein that's accessible in the cell surface- Mm-hmm ... in pancreatic BTC and gastric cancers, right? Mm-hmm. But I'm curious when in the process of transformation that takes place. I didn't know if you have data to that question. Yeah, so my belief is because it's a gastric lineage marker and during the metaplastic process, what we are seeing is in some metaplasia when especially it become very advanced intestinal, the gastric lineage markers start to reduce. So this is what we're talking about in the metaplastic process. So I'm not sure whether in the cancer it could undergo the same types of... Because, a lot of gastric cancer actually has inherent probably metaplasia as the stem cells. So, it depends on the state. If it's a hybrid, it may still retain the claudin 18, but it may switch to more intestinal phenotype. I suppose that could be a form of resistance. Is locally advanced, resectable, is the prevalence of 18.2 the same as in metastatic? Has anyone looked at that? I think it depends on which antibody. I find that the claudin expression is really variable depending on who's looking at the, and what cutoffs you use. But in our experience so far, it's very similar. I had a question for you. Mm-hmm. So, as you know, the young onset is an epidemic for our disease as well. How do they develop? What type of metaplasia are you seeing in young patients with this disease? Have you looked at- You mean in gastric cancer? Yeah. I think it's the same metaplasia. Mm-hmm. And I also find it, the complete and incomplete classification, I think it's always a spectrum. And if you look at the heterogeneity we see in the microdissected individual gland, so if you look at it carefully, often you will see both complete and incomplete in the same patient. So it's really which one eventually gained enough driver mutation to get transformed. And it looks like based on our data, the one that are more hybrid, that probably are more likely to progress because of this plasticity and fetal reversion. Yeah. But they can, I think if you look at the clonal organoid, they show different spectrum, right? So I think to start with, the cell that turn neoplastic probably may have certain epigenetic point to start with. But whether they can still deviate afterwards, I'm not sure. Yeah. Well, a very nice presentation. I have three questions. First is, if you only observe a single tumor in a patient but you have so many mutations in the gastric mucosa, why is it that we see only one tumor? So what is the difference?The second question is, if we see so many heterogeneities, and metastases, and primary tumor, how do we handle that in diagnostics? Because all we have is the primary biopsy. The third question is, we do have sexual dimorphism. I haven't heard anything about how you treat women and men separately, how you incorporate this information about sexual dimorphism in treatment and diagnostics of gastric cancer. Thank you. So for the first question about the heterogeneity, I think it's kind of a few effect. So I think the difficulty is, if you look at the heterogeneity, basically, there are many, many glands, each acquiring their mutation, the chromosome change independently. But the chance of any one of them progress is pretty low. So that's why at the end of the day, you end up mainly with one dominant cancer. But I think in the early cancer evolution, you may have many different parallel evolution clones, but eventually, maybe one out-compete all the others. So that's the answer to point number one. So that's why by the time we see the gastric cancer, it usually originate from one clone. The sec- Fitness. It's fitness, right? Fitness. Yes. Yeah, exactly. Many mutation in the gastric mucosa, you say you have shown that. Yeah. One eventually will develop into cancer. Yeah. What happens to the remainder? I think it's not so easy to developing cancer, so they may remain as metaplasia, they may become dysplastic, but they may be still in the field in the adjacent area. Yeah. So how we cope with diagnostics? Maybe I can take that one. Yeah. I think what we really need to cope with is, and it's a circular argument, but right now the diagnostics are actually quite good, it's the therapeutics that are lacking. So the truth is, yes, there is some heterogeneity, but there's not that much of a difference between the primary and the met. The p53 mutation is the same. I would say 70% or 80% of genomic alterations, if you sequence both, are very similar. The difference comes in maybe some levels of expression of HER2 or PD-L1 and so forth, but they probably aren't that important to begin with, right? So I think what we need to catch up on is actual better therapeutics, and the diagnostics then will lead. Functional imaging, as Jeffrey said, and it's logical, you can't biopsy all the mets anyway. So I think what we've developed is functional imaging, so using radio tracers as a way to really just get a sense from the patient what level of the burden of the biomarker they have. And then you could prior strata... You think the tumor is HER2 positive, but then you go in and only 60% or 70% of the primary tumor in the met express, and we see that with HER2 PAT. 80% of patients have heterogeneity. We know that. But if you don't have an alternative to target, you will still use HER2. You'd need to develop other treatments before, and I think it's fine to do it in parallel, but the field of therapeutics has really lacked in this disease. Thank you very much. Yep. This side. Yes. Hi, my name is Chandra Gul. I'm a general surgery resident and a postdoc in surgery at Stanford. Thank you for an excellent talk. This is a question for Dr. Chen Zhijian. Hopefully, I'm saying your name right. Yes. You spoke about the neoadjuvant immuno-chemotherapy and how that's becoming the gold standard for gastric cancer. I wonder if you have any perspective on the extent of lymphadenectomy that we currently do as a surgeon. I know a lot of the Eastern perspective historically is doing pretty aggressive lymphadenectomy, and that's what used to be considered adequate. Now with immunotherapy, are we thinking about leaving some behind for adequate therapy or response? Curious to hear your thoughts. Yeah. So looking at the pancreas experience with Vinod Balachandran and the adjuvant vaccines, I think your point is well taken. Sometimes these lymph nodes are important. In fact, doing a splenectomy and taking out more lymph nodes was associated with less of immune response. For our disease, it's slightly different because we're doing everything neoadjuvantly, and I think where the field is moving forward to doing more and more while the tumor and the lymph nodes are in place, and using surgeries almost as a staging procedure where you have to take a lot of lymph nodes out, so you know sort of what it-- Because first of all, you're worried that there's residual disease, but also to adequately stage. So I don't see the field-- In fact, right now, I think that what I struggle most is as I'm selecting centers or designing the next phase three, is quality control. Because with the surgeons, it's just you and the patient, and maybe some nurses that report to you and would never tell you, "Well, your surgery was not good enough." Right? So, that's what we struggle with. And also the fact that if you do have a good immune response, sometimes it looks like the lymph nodes are growing, but in fact, they're all residual response of therapy. So we have not deescalated surgery. I think maybe that's going to be in the next phase. But I see it's either all or nothing. You either do the surgery or you don't. But right now, the D2 lymph node dissection is the same. It's a modified version where they leave the spleen behind. Thank you. I have a couple of questions, Dr. Judkins. The first one regarding you are showing the STING pathway, STING, cGAS pathway, the increase the STING pathway can associate the poor survival. But STING actually innate immune can increase the interferon gamma or actually anti-tumor. I don't know what's behind the rationale for this poorly differentiation association with this high STING pathway as a first pass. I know. I knew that it was going to be very controversial because you could approach it from two different factors. And I suspect that in each individual tumor type, but also within each even gastroesophageal cancer, depending on where in the trajectory of the CIN tumor you look, STING can be a good thing or a bad thing. So in our cohort now, what we're seeing is that we are seeing low STING expression. STING pathway through interferon gamma can up-regulation PD-1. I'm not sure in this patient you gave additional PD-1 or PD-L1 treatment. Do they have a better response or not? So the cohort we have are all IO-treated patients, so that's why we know what immunotherapy response or duration of treatment benefit was. And what we are seeing is these were pre-treatment samples. Okay. So these were not one sort of post-progression samples, post immunotherapy. Another question regarding to Dr. Johnny Bo, for you because you treat the patient. I'm wondering because diffuse type of intestinal have different etiology and molecular biology and the genetic difference, do you treat the patient different or similar? Based on what? So I think pure diffuse type of tumor patients, they do very badly generally. But we start with the biomarkers, the four biomarkers that Yelena showed, and then you can do NGS. But consistently, the diffuse type will do poorly. Maybe with claudin, they will do a little bit better. But we don't treat based on phenotype. We're treating based on biomarkers and other profiling. Okay. Another question also both to yours regarding the immune therapy. I know PD-1, PD-L1 mostly is tested in the clinical. However, do you ever treat a patient and use TIGIT? You mentioned the TIGIT plus PD-1. Are there other LAG-3 or other? Do you have any better clinical response than PD-1, PD-L1? Yeah. So the TIGIT story, unfortunately, is negative as many other tumor types. So there was a large Phase III study that will be presented at ESMO GI in Munich that was negative despite very promising Phase II study. And LAG-3 data is also negative, unfortunately. So the only hope is with PD-1 CTLA-4 because of this Catenilimab FC silent bispecific. But it's too soon to say because the Nivo/Ipi data, the CheckMate 649 data looked promising. But because there was no chemo, because of the toxicity concern, there was a crossover in the survival curves. So it's too soon. The new bispecifics are going to probably produce good results. So one anti-PD-1, anti-PD-L1, anti-VEGF, or EGFR HER3 with ADC, they are showing some good progress. But a lot of work needs to be done. Okay. Yeah. One question for Dr. Leung on the cancer evolution in this field. I'm curious if we can detect what's going on during the pre-cancer to cancer, the transition. Yes. Any genomic bottlenecks during the transition pre-cancer to cancer? You show the mutation cluster, you use the mutation and the epigenomic and the copy number, and do you have a chance to check the timing of this event? And the ultimate question is, can we detect, as you say, the diagnosis is good. Is it possible to detect cancer in pre-cancer stage if, say, we have the precursor? That's my question. Thank you. So I think there is an intense interest in predicting which metaplasia will progress to cancer, right? I think the problem is for the metaplasia, it's quite common, but only a small percent progress. So you need to find ways to predict the one that will progress. So based on our organoid model, we feel that those metaplasia with chromosome 20 gain is probably biologically, at least from our model, seems to have more neoplastic phenotype. So in parallel, I also cite some study, for example, from Singapore groups. So they actually have other risk progression marker like mutation count, even clonal hematopoiesis, and ARID1A mutation. So I think this is still work in progress. But I think with more data and more biology, I think for certain we should be able to find better markers. Thank you. Yes. The last question? Yeah. Yeah, I have one question. Thank you for the nice talking. The question for Dr. Elena. In new adjuvant in MSI-H patients, you showed your clinical trials that combined dual IO with chemotherapy, the effect is better. But have you ever compared single IO with chemotherapy and dual IO with chemotherapy in this kind of patients? Because in many clinical trials, single IO with chemotherapy has already pretty good effect. That's a great question, and I will cover some of this in the discussion. It's actually not my data, it's Alberto's data, Infinity, and Sara Leonardi is here too. So they're presenting the durva/tremi. But that's a very good question. I think because of the heterogeneity in MSI cancers, we know that anti PD-1, like dostorlimab data, right, in gastroesophageal cancer was not as good as it was in rectal. So at a minimum, chemoimmunotherapy is needed. But I think CTLA-4 is a very attractive option because we know from CheckMate 649, looking at, at least numerically, we never compared head to head, but FOLFOX nivo versus nivo ipi, it looked better in terms of, again, small numbers, hard to extrapolate. But we know that if you respond to ipi, you never progress, right? So we know that. And also FOLFOX is not easy. And in Alberto's study, doing one dose of tremi is enough. So instead of being on six months of FOLFOX nivo or a flott durva, certainly, it's very appealing. So I think the way that we approach it is that we usually start with dual immune checkpoint blockade, but we watch the patient very carefully, and if it looks like radiographically or clinically they're progressing or not responding, we then pivot to either flott durva or FOLFOX durva or something like that. Thank you. Mm-hmm. But come to the session on Tuesday. There'll be a lot more on that. Okay. So thank you very much for your attendance. And I also want to thank the speakers. Thank you. We should take a picture. Yeah. We should take a picture together. Thank you for the invitation. Yeah. Enjoyed it.