 Good afternoon. I will be talking about something completely different today. This slide is for computational biologists in the audience who usually look at raw digital data. This is what the raw analog data look like when you put it into your computer. If you then slice this tumor, this is what you get. You get something that is very complex. It has a number of different cell types that you can see here. Some of these are tumor cells. Some of these are stromal cells that have lots of fat, blood vessels, lots of collagen, lots of extracellular matrix. All this complexity goes into your analysis. I remember when we were first deciding what kind of tumor should we put in, we decided for varying cancer that it should be at least 80% of tumor content of cancer malignant cell content in this tumor. But it was actually difficult to find tumors with 80% tumor cell content, and that should tell us something that tells us that the majority of tumors actually have other stuff in there. This other stuff is what I will be talking about, and hopefully I will convince you that this other stuff is also very important, and we have been neglecting it to some extent in the TCGA analysis. But maybe it requires the cancer-associated analysis. This is a section through a tumor, and you can see the cancer cells here in pink. You can see also that it is surrounded by some sort of fibrotic stuff. This is the stroma that is usually present, and the question is, what is stroma really doing in tumors? Some people believe it's actually preventing from cancer progression. It's kind of encapsulating the tumor. Others would say, well, stroma actually infiltrates the tumors and, in fact, helps the progression and invasion of tumors. In fact, tumor cells can glide on top of these pink sheets, like on highways, and progress and metastasize more efficiently. One important distinction between cancer cells and stroma is that cancer cells are highly proliferative, and stroma is not. So what you see here is Ki67 is a marker of proliferation, and you can see the cancer cells are highly proliferative, but you can also see some nuclei here in this white space, and they are not stained with the marker because they are not proliferative. This is the stroma. So the cancer cells are proliferative, and this is also what is typically targeted with chemotherapy because most chemotherapies target proliferating cells. One other distinction is that, as we know, cancer cells are very genomically, genetically unstable, and this is a bit of a problem with all the chemotherapies that we are applying because of their flexibility and genomic instability. They quickly re-adapt and start using different pathways. This is not the case with stroma. Stroma is slow proliferating. It is not directly targeted by chemotherapy, although indirectly often crumbles with the tumors. It's genetically stable, and what is important, one thing about being genetically stable, then actually we may use it to our advantage when it's targeted because you may expect that genetically stable cells are not capable of re-defining themselves as quickly as the cancer cells, and as I will try to convince you, they're just as important as genetic alterations in tumor progression. One thing that we noticed is that enhanced remodeling occurs in stroma during tumor progression, and there has been, in the past five to ten years, a study of the study has shown, if you compare cancers and you're trying to find on the expression level what are the genes that predict poor survival, most likely you will end up with genes that are not genes that are expressed in cancer cells, but genes that are expressed in the stroma. And why is that? Well, that tells us that there is something in the stroma that is very important and that predicts poor survival or maybe actually contributes to that poor survival. One gene that I'm specifically interested in is collagen 11a1. This is a very rare type of collagen that we found was expressed very highly in tumor. Here you can see tumor is blue, and then you can see these brown things. That's an expression of collagen 11a1 in in situ hybridization, and you can see it's kind of dispersed throughout the tumor, but when you look at the peritumoral stroma, which is also a stroma, you can see that there's very little of it, and then once you go more than one millimeter away from the tumor, you don't see any collagen 11a1. Because collagen 11a1 is not expressed in the normal stroma, it doesn't matter how much stroma you have in the tumor. What matters is that there's specific stroma subtype that expresses this collagen, and it's usually inside the tumor, but not all tumor cells, and we still don't know what is it with that stroma cell type that expresses collagen 11a1. An interesting thing about collagen 11a1 is that if you look across different cancer types and you compare normal tissue to their corresponding cancer, you will find that in most cancers, you see much higher levels of collagen 11a1 than in their corresponding normal tissues, and this is not very common, even if you look, so this is not because there's more stroma in cancer than in normal tissues. If you take a very common stroma marker, such as actually even activated stroma marker, smooth muscle acting, you don't see that, you see kind of comparable expression in normal tissues because this acting is expressed in normal tissues and in the stroma, but with collagen 11a1, you see this huge disparity, and if you look at individual cancer types, such as breast cancer, and you ask which ranked the genes that are most differentially expressed between normal breast and invasive ductal carcinoma, collagen 11a1 will be ranked as the most differentially expressed genes. In colorectal cancer, it's ranked as number three, but it gets even better, so it's not just normal and cancer, but if you look at during cancer progression, and this is an example of ductal carcinoma in situ and invasive ductal carcinoma, the difference is that ductal carcinoma in situ has a base membrane around and it's still enveloping and it has much better prognosis. Invasive ductal carcinoma, the tumor cells have already penetrated through the base membrane and this has worse prognosis. If you compare these two entities and look for genes that are differentially expressed, as I have done here with four different databases that I could find that compare these two types. Again, collagen 11a1 is one of the most differentially expressed genes during this type of progression of breast cancer. This is an example in ovarian cancer. You see that levels of collagen 11a1 are higher as you go from stages. These are different staging systems in ovarian cancer. You have stage one is the lowest stage confined to the ovary, to stage four, which is a widely metastatic disease outside the peritoneal cavity and collagen 11a1 levels rise through metastatic progression or stage progression. And even better, it actually is the highest levels of collagen 11a1 are actually during recurrence. So you see very little, this is in situ hybridization where we compared in matched patients, in this case, 10 samples. Now we have done it with 48 samples and the result is the same, that you have very little collagen 11a1 in the stroma of the primary tumor. You see a little bit of it in metastasis in the same patient. But then in the recurrences, you see a lot of collagen 11a1. So as I said, it is expressed in different cancer types. So this is now where I utilize TCGA data and I asked what genes are co-expressed with collagen 11a1. And I did that for bladder cancer, so I ranked the co-expressed genes, but their sperm and score correlation. Then I did the same for breast and for colorectal. And I did it for 12 different cancer types, all of which are adenocarcinomas. I was interested only in epithelial cancers. And you can see that some of these genes in different cancer types are actually the same genes. And I noticed that it was similar genes over and over in different cancer types. So then in all these 12 different cancers, we ranked the genes across 12 cancers by their priority score, basically. And you see that the closest across cancers is collagen 11a1, then collagen 5n1, Faberblast activating protein, and so on. The list is about, we selected a cutoff, and the list is about 190 genes or so. Then we asked, what are the top 10% genes that in bladder cancer, for example, have the top 10 priority score? And we put them in pink, then we ask the same for breast cancer, then we ask the same for cervical cancer, and so on. And when you do that sort of analysis, you find that, I don't know if you can see, but the pink, you can see most of the pink is up in the top. And that means that what is 10% most highly associated genes with collagen 11a1 in bladder cancer, it's similar to ovarian cancer, it's similar to breast cancer. So that means that these genes are very homogeneous. It's a very tight correlated set of genes. And this goes across different cancer types. We're talking about 12 different cancers with very different genetic alterations. When we did this with anti-express genes, which were not very highly correlated anyway, we didn't see this sort of association. So what is the signature now? It's a very interesting, tightly correlated signature. What does it represent? It's there in tumors. Is it a specific cell type that it represents? So as I said, tumors have many different cell types. We looked for this signature in different cell types, and we actually cannot find it in the normal cell type, as we can say, except cartilage, maybe. But really, so we don't think it's a specific cell type. We actually think it is more a process that exists in tumors and that is exacerbated during cancer progression. And that process we think is myofibroblast activation and I will not go into how we got to that. So what is myofibroblast activation? Well, fibroblast, resident fibroblast in tumors, are activated by TGF-beta growth factor. And they become myofibroblast, which are more motile, more stiff. And if that is true, we actually put this signature and asked what are the upstream, most likely upstream regulators of the signature? And TGF-beta1 actually comes up as the most likely, so that makes sense. TGF-beta1 activates these myofibroblasts. And one thing that is interesting about the fibroblast is that they produce collagen here in blue, which is kind of relaxed collagen fibers. But myofibroblast tend to produce this straight, stiff collagen fibers. And that's actually important for activation and kind of feed forward a loop of making them even more stiff and more stiff. And I won't go into the biology of that. But here, just to see if whether we can really see it in real tumors, you can see here collagen 11A1 is in brown. And you can see one of these boxes, number one. You can see the trichromes staining, which stains basically collagen and muscle. And then you can see number two, where collagen 11A is not expressed. You can see nice blue and wavy. But this is where collagen 11A1 is expressed. It's nice and straight, as we would expect. We also showed, if we all express collagen 11A1, we make the cells stiffer. Now, it's a very good gene to target because it's expressed specifically intratumorally, but not outside of tumors. But there's one little problem, and that is that it's also expressed in the cartilage and a little bit in your eye. We also show that you can target it with SHRNA, and the tumors are smaller and not as metastatic. So it would be a good target, but how do you distinguish it from cartilage? One way to distinguish it, and this is why it's such a great target, it's actually not the same one as in cartilage. So cartilage, collagen type 11 consists of three chains, three polypeptides, collagen 11A1, our friend, collagen 11A2, and collagen 2A1. Now, when we look at tumors, only collagen 11A1 is highly expressed. But the other two chains are not. That means that in tumors, collagen 11A1 must be homotrimerizing with itself, or heterotrimerizing with something else, but not with this. Because clearly, these are not upregulated. And actually what we find in this signature is that other collagen 5A1, 5A2, 3A1, and several others are highly upregulated in this signature. And what is common to those collagen? All of them have very slow processing of their N-termini. And this is a part of collagen processing that collagen needs to go through. And because of the slow processing, these N-termini kind of hang out of the fibers, and that prevents attrition of additional fibers. So you end up with these nice thin fibers as you saw. So we don't know why these fibers have to be thin, and what mechanically why is this important, and how this affects tumor progression at the moment. But what I've shown you is that collagen 11A1 is upregulated in normal tissues and levels increase during progression. The signature that we have identified is present in multiple solid cancers, regardless of cancer origin. So that means the resident's trauma is the same, or goes through the same process of fibroblast to myofibroblast activation. And it is enriched during innovation. We think it's activated by TGF beta signaling. And fibers are stiff, stretched out. But what is most important is that we think that this is also targetable. Because it does have a unique composition of collagen fibers in myofibroblast. So we think that this may be a good opportunity for therapeutic targeting. And I would like to thank my colleagues and my lab members and my sources of funding. Thank you. So, very good question, I don't know how you suspected. Yes, we have myofibroblast in scar tissue. So we did look in scar tissue wound healing in fibrosis. So it is slightly expressed in kiloids, not so much in scar tissue. And not during wound healing, not during fibrosis. To some extent you see a slight expression, but this is nowhere near what you see in cancer. So we think that unlike smooth muscle acting, which you will see or FAP, which you will see upregulated in all of these benign non-malignant processes, collagen 11A1 is not. It's tumor specific. Very good talk. I just wanted to make a comment that we studied the protein expression data in pan-cancer analysis. I presented the results last year of the symposium. And we have found what we call a new subtype, a reactive subtype, where we have seen stromal markers are elevated in the reactive subtype. This was first discovered in breast cancer. And we saw that the prognosis was very good for those patients with this particular subtype. Then we expanded it to pan-cancer and we found that for other cancers, the prognosis was poor. So the prognosis is cancer dependent, it's context dependent. One of the markers of the reactive subtypes was collagen 6. We don't have collagen 11 in the proteomic database, but I saw that collagen 6 was highly correlated with collagen 11A. So I think one of the things that we should keep an open mind about is that this may have a good prognosis and better outcome. Or it may have worse prognosis depending on the disease type. That is correct. I will just comment that for the signature that we have been studying, we found that in all of these cancers, it actually had worse prognosis. It wasn't a great predictor of worse prognosis, but it's still predicted worse prognosis. Except for lymphoma, B cell diffused lymphoma, which has been published several years ago, where this exact signature actually shows better prognosis. But we think it's because lymphoma, it's actually the cells that are present there are not replaced by cancer. So we think because this may be resident cells in the lymphoma. So I don't know for your signature, but surely we will be finding stromal signatures that are associated with better survival. And not all stroma is the same. I think this is very important to realize that what we are looking here is a small subset of stroma. This is very intriguing data. I have a question with respect to your trimeric hypothesis. I think to be more convincing, have you looked at GTX data to actually show that the differences that you see across those three strands are non-different across and for GTX data? But I think that would be sort of more... Right. GTX data doesn't have cartilage as far as I remember. So that's one problem that these genes express in normal tissues only in cartilage. So I couldn't really find them. Thank you. So our next speaker is Dr. Winyin Zong from Pfizer, who's going to tell us about integrated analysis of TCGA and how it identified targets and patient populations for antibody drug conjugates.