 Thank you. Good morning. My name is Rachel Liao and I'm a graduate student in the laboratory of Matthew Meyerson at the Dana-Farber Cancer Institute and at the Broad Institute and I'm very excited today to tell you a story which I believe is a success story of use of the TCGA data and analysis that we all do and actually translating it to clinical insights that are affecting patient care and improving patient care Even even today as we speak Last year, I believe Peter Harriman was here and told you about the where the TCGA squamous cell carcinoma of the lung project was At that time and I believe he told you that at the prior to the TCGA study the squamous cell carcinoma of the lung was a Disease really without treatment options. It's the second most common subtype of lung non-small cell carcinoma the most common of which is adenocarcinoma Which has much larger percentage of the of the population of people who have non-small cell lung cancer But because in the last 10 years adenocarcinoma has undergone a lot of study There are many targeted therapies and advances that have been made for patient care such that many patients with this disease undergo genotyping and are able to be put on to targeted therapies that will actually target the Genomic events that are underlying their disease proliferation While at the same time a squamous cell carcinoma has had few targets and no targeted therapies that have been elucidated And so at every patient that was put on that was diagnosed with squamous cell carcinoma long was put on to the same chemotherapy radiation treatment regardless of the underlying genetic components now this TCGA Preliminary analysis or the first the first pass of the analysis was published. I believe most of you are aware Very recently a couple of months ago, and I'm going to tell you about some of the observations We've made in that data set that have actually influenced patient care for the better now the FGFR family Is a receptor tires in kinase family, which has been implicated in other cancers and in other diseases The events that exist in the TCGA lung squamous cell carcinoma data set are about 10 percent focal amplification of FGFR One which has been observed before and about 8 percent mutation Frequency across the four family members the four receptor family members that includes about 3 percent in FGFR 2 and 3 percent in FGFR 3 and Those are the two genes on which I'll be focusing this morning Now I think it's important to make the point that just observing a mutation does not necessarily imply that it's functionally relevant and And so a lot of the functional work that I did was in fact demonstrating the Potential for some of these events to transform and to potentially contribute to uncontrolled proliferation and Hopefully also therefore be targets that we can target in the clinic now I think it's also important in this audience in particular to make the note that this is not neither of these genes are significant They're not significantly mutated in this data set by a large margin And we believe that a lot of the reason for that is because the mutation frequency and lung squamous cell carcinoma as you've heard before is extremely high and so the the actual objective number of mutations that we observe is relatively low compared to that Frequency, but because of our prior knowledge of these genes being Implicated in cancer in the past you heard from John this morning that FGFR 3s are important in bladder cancer I believe you'll hear later this morning that FGFR 2s are Relatively commonly observed in endometrial cancer We believe that this was a family of receptors that That may very well be an important target in a subset of patients with lung squamous cell carcinoma Now these are the events that that I studied in FGFR 2 and FGFR 3 you can see them across the the length of the protein From the end terminal on the left to the C terminal on the right and you can see that the Ig like domains in orange Have mutations that cluster basically between the second and third Ig domain Here in FGFR 3 and a little bit more into the F the Ig 3 domain in FGFR 2. These are extracellular The structured domains in these proteins they facilitate ligand binding and dimerization of the proteins to activate the proteins And then you can see these two mutations at the residue K660 Both fall in the kinase domain have actually been described previously as being activating mutations in the kinase domain You can also see that there are other mutations that are outside of structured domains And then this one mutation which when mapped to the protein structure does not fall Close to a part of the kinase domain that would that would suggest that it was involved in the kinase function And I believe that that is again not entirely surprising I'll show you data that suggests that these might be passenger mutations And I think because mutation frequency is so high We would expect to see some passenger mutations even in these genes which are Implicated in the proliferation of the disease in other patients So this is a plot that demonstrates that essentially there are not repeatedly Cooccurring events with any of these FGFR mutations except for TP P53 mutations You can see that each row is as a lung cancer sample in the TCGA data set And then each column is the event so on the left you can see here there are amplifications and deletions that are common in this Excuse me the squamous cell carcinoma Dataset and then on the right are mutations which black are indicating miscense and and yellow or orange are sort of the Nonsense frame shift splice site mutations that we presume to be Loss of function and what you can see here is that the FGFR 2 and FGFR 3 mutations are indicated P53 which occurs at about 80% frequency in this data set occurs at about 80% frequency in the number of samples that are mutated With FGFR mutations then the rest of them are just a smattering of other events that Happen to co-occur in those samples and again because the mutation frequency is so high That's not surprising that we would see many other events in these samples But nothing that specifically is implicated as another potential driver in in a large majority of cases So the first experiment that I wanted to do was determine whether in my hands I could demonstrate that any or many of these mutations were in fact Able to drive proliferation were what we would call transforming in in a transgenic model system And so what I did was introduce each individual mutation into a cell line that's immortalized But it's not transformed and cannot proliferate uncontrollably on its own to see if the simple addition of this Mutation of this genetic mutation would be able to overcome the cells inhibitions toward uncontrolled growth and actually just drive its proliferation uncontrollably so what you can see here is graphs that indicate the essentially the number of colonies that were able to grow in suspended media basically in an auger media that that do not any more require the Sort of the contact that cells that are adherent require in plastic in order to grow so you can see that several of these mutations Form colonies at far greater rates above the wild type Which is indicated about the third from the right or the second from the right in each of these in each of these plots You'll see that I have two different isoforms of FGFR two mutations represented in these experiments. That's because there Many fewer studies that are done on the isoform differences in these mutations in particular in FGFR two Whereas in FGFR three I used only one isoform because the two mutations that you see here that are Transforming have been well studied in bladder cancer in both isoforms And you can see the two mutations that were in unstructured domains did not form colonies above Wild type indicating that they are not in fact potential drivers at least from this experiment or are not potential drivers of this disease So you can see that there are actually four mutations out of the six in FGFR two which were which we're transforming this assay There are also four Events three of which were at this s249c event in FGFR three so four out of six Events that we observed in each gene were transforming in this assay and the next thing I asked Was whether I could actually reverse this transformation block this transformation by the addition of FGFR inhibitors I'm showing you those data here So this is the same experiment as the one on the previous slide in which cells expressing these individual mutations are seeded into an auger solution which precludes they're growing on on plastic essentially requires them to have some transformed potential in order to proliferate and So I only use the mutations in this experiment which were in fact Transforming in the previous experiment and I seeded them into increasing concentrations of drug such that We would be able to determine whether the addition of drug could block the transformation phenotype that we saw and that's exactly what happened You can see over here the different Concentrations of these two different drugs that were that were used in this experiment I did this experiment with a few more drugs But these are two of the the drugs that I had used and you can see in the case of ap24534 also known as Panatinib which is from Ariad You can see that there's loss of colony formation across many of the mutations by Between 10 and 100 nanomolars of drug which is quite quite low and in the case of BGJ 398 Which is actually a specific FGFR inhibitor from Novartis You can see the loss of colony formation in all the cells expressing mutations is at 10 nanomolars or lower While in the case of an EGFR transforming mutation You're still I'm still seeing colony formation at 10 micromolars of drug. So this was very encouraging to see and also was confirmed by Western blotting that I did to demonstrate that the loss of transformation that I saw in the colony formation assay that I just Showed you also correlates with loss of phosphorylation biochemically. So you can see here that FGFR 2 is expressed in the top two Images and then FGFR 3 expressing cells are in this bottom image. You can see that FOSO FGFR decreases very significantly at at 10 or 30 or 100 nanomolars of drug and this is just in the presence of Panatinib and FOSO FRS 2 which is also a downstream target of FGFR phosphorylation and Regulates a lot of the downstream signaling effects of the FGFR family members is also inhibited Demonstrating that this drug is in fact targeting the pathway correctly and is also Doing what we expect that it's doing in the previous experiments So I have one one slide on the next set of studies that I did which was a dependency study I introduced these mutations into a cell line in which I could actually generate dependency on this pathway Through the introduction of these mutations so that I could actually study The dependency on the pathway in the context of a panel of different inhibitors and many of you who saw my poster yesterday Saw that I actually have eight inhibitors in this panel I'm happy to send you a download of that poster if you're interested in those data and Essentially what you see here is that cells that are dependent on FGFR events are extremely sensitive I see 50 values are plotted down here for an An easier time of looking at them cells that are expressing these FGFR events are extremely sensitive to FGFR inhibitors while cells expressing an EGFR event depending on EGFR event or Parental cells are not sensitive by several orders of magnitude Now we have demonstrated to you a couple of cell based models that are certainly not They're not by themselves clinically relevant. They're not disease models per se Their cell cell models, but we we also have a clinical case That's an interesting case that demonstrates a lot of these that these data are hopefully on the right track today To modeling patient disease. This is the case of a head and neck squamous cell carcinoma patient Which of course is not a lung squamous cell carcinoma, but as you heard yesterday from Neil Hayes The HPV negative head and neck cancers Genomically are extremely similar to lung squamous cell carcinoma and this patient was found to have an FGFR 2 Mutation which has been observed previously in bladder cancer, which is known to be activating and this patient was put on to penatin I'm sorry was put on to pizopenib, which is a multi kinase inhibitor, which is approved for soft tissue sarcoma treatment But which is known to have FGFR event Inhibitory properties and this patient after two weeks on pizopenib went from having this lung metastasis, which is very large on his right Or I'm sorry this this neck metastasis on his right neck was Very much reduced and responded very Effectively to this treatment and we believe that this is because of the FGFR Targeted therapy that he underwent So just in conclusion as I told you these FGFR events that we've observed though not significant in the actual computational analysis were in fact quite significant biologically and that we've been able to model Transformation potential and also dependency and sensitivity to targeted therapy and a clinical success has confirmed that this may in fact be a very good target to pursue for squamous cell lung carcinoma and in fact a Clinical trial at the Dana-Farber has just been approved with penatinib For lung squamous cell carcinoma patients that present with FGFR events and we're hoping will be able to Extend that to a head and neck carcinoma patients Also, so it's been a very real clinical translation of the TCGA data that we've all been working on Analyzing to the clinic and to benefiting patients in a real way and so with that I'd like to end Thank you. Thank very much Matthew Myerson and also Peter Hammerman who have been very supportive in this project And I'm happy to take any questions. Thank you very much That's a really interesting study questions Lou It's exciting, but there's a challenge then you have a diversity of mutations and some of them are passengers So you get a new patient the clinic shows up with a mutation. Do you treat them or not? I? Think that's a very good question and not being a clinician I won't speak to how to make the decisions to treat patients But I believe that with more study will be able to identify More events that are known to be drivers with algorithms that enable to Perhaps infer the the function of different mutations I think you're right that it's not going to be an exact science of which mutations will respond and which won't but but you are a Biologist rapidly make that mutant put it into a model right and show it's transforming. It's true. Yes, we can Last quick question quick answer Rachel, what about the amplification events which are much more common than mutations any sense as to whether or not they're drivers from any of your data So I think that not from these data, but from other studies that I and others have done We do believe that a subset of event of tumors that present with focal FGF R1 Amplifications are in fact being driven by FGF R1 signaling events It's unclear exactly which tumors those are from just Patients presenting in the clinic and it's certainly clear that not all patients will respond to FGF R therapy So I believe that that is in fact another avenue down Which we can really make some strides in understanding patient response and and potential therapeutic options for these patients. Yeah