 Good morning. First of all, I want to thank the organizer for giving me the opportunity to present this story today. This story started on the latest phases of the TCGA breast cancer project and it aimed at exploring and trying to explain the diversity observed in a subset of breast tumors and I'm talking about luminal tumors and in particular luminal A. As most of you know, breast cancer is an incredibly heterogeneous disease. It can be actually thought as a collection of distinct diseases. And these diseases are mostly characterized by the status of hormone receptors ER and PR and growth factor receptor HER2. As you can see from the cartoon here, the largest fraction of breast tumors are positive for the estrogen and progesterone receptor and negative for HER2. And this definition largely identifies what we call luminal breast cancer. Luminal tumors have been further stratified using mRNA signature into luminal A and luminal B. With luminal A being less aggressive, generally associated with a better prognosis and overall the most frequently occurring breast cancer subtype in the population. However, the great diversity and abundance of genomic data that has been recently become available allowed us to do more in-depth specific analysis of this subtype and identifying a diversity, a great diversity yet to be explained. So what we did was integrating multiple data sets from different studies, accounting for roughly 1500 luminal tumors, 1,000 of which are luminal A. And from this data set we have different data type, but for almost all of them we have copy number alterations, aromatic mutation, and mRNA expression. So the first thing that we do was we went comparing the well-established mRNA-driven breast cancer subtype against copy number-driven cluster identified in different data set and showing here the meta-brick data set and the TCGA one. And what was quite interesting is that while for Basel and HER2 enriched breast tumors there is an almost perfect one-to-one correspondence between these clusters, both luminal A's and luminal B are associated with multiple and distinct copy number cluster suggesting an intrinsic heterogeneity in terms of copy number alteration. And a similar picture emerged when we went looking at somatic mutation. Indeed again HER2 and Basel are characterized by the highest number of mutation per sample on average indicated here by the blue bars, while luminal B and especially luminal A I would say are characterized by the highest number of genes mutated more frequently than expected by chance and these are indicated here by the red bars. So this tells us that these subtypes have a great heterogeneity at the molecular level, but this heterogeneity is not only molecular, indeed luminal A is quite heterogeneous also in clinics, indeed while it has the highest median overall survival, it also has the most valuable one you can see here from the box plot. And it has also been showed that the risk of late mortality persists in this group even after 10 years of initial diagnosing and it is actually greater in these subtypes than in the others in the long term. So given this overall preliminary observation we decided to focus specifically on this subtype of breast cancer, the luminal A breast cancer as it looks like the most heterogeneous both molecularly and clinically and we wanted to start exploring this diversity to address some fundamental question like can we link the variability observed in clinical outcome with the underlying molecular diversity. And we started to address this question by looking at copy number alteration. We went back to the TCGA dataset and we cluster the luminal A tumors according to the copy number levels across the whole genome for this patient. And as you can see just by eyeballing this heat map, the diversity is pretty striking. We found five major subgroups, some of which are characterized by low level or no copy number alteration. While other as like this one in red here are characterized by an abnormally high level of copy number alteration and especially focal deletion and amplification. So we took the centroid of this cluster and we wanted to see if we could validate these results in additional dataset. And first we validate that in the Metabrick dataset where we found similar clusters in similar proportions and especially in this dataset we had extensive follow up so we could do more reliable survival analysis and we saw that the copy number high or high genomic instable cluster is associated with the worst prognosis. And I'm showing here that Kaplan-Mayer specifically for this cluster in the Metabrick dataset here and we went validating this finding on a third dataset where we found a similar proportion around 10% of luminal A that have this genomic features and are associated again with the worst prognosis. So we were pretty excited because we started linking this molecular heterogeneity with the clinical variability. However we asked ourselves if we could pinpoint specific genomic features that are enriched in this subtype. So I went comparing the copy number high luminal A tumors against all the other luminal A's and what we found is that these tumors are highly enriched for P53 mutation, MIG focal amplification, 20-Q gain, 5-Q loss and are depleted for P53CA mutation that otherwise characterized most of luminal A tumors. We then looked at genes that are differentially expressed between these two subgroups and we found that copy number high luminal A's over expresses several regulators of mitosis including both our arachinesis A and B, PLK1, CDK1 and several cyclins and they down-regulate genes involved in the endoper pathway like P53 binding protein 1, MLH3 and several polymerases. So all these features clearly point to a high proliferation, high genomic instability and well explain and match the phenotype that we observed in terms of overall survival. At this point we wanted to see if we could have a similar comprehensive picture for the other luminal A subtype. We went to look in first at some of the mutations and we saw that the most frequently occurring mutation in luminal A tumors and I'm referring here to P53CA and GATA3 in particular are enriched in tumors that have low level of copy number alteration and in particular GATA3 is highly recurrent in tumors that present the 1-Q gain, 16-Q loss pattern that characterized roughly 30% of luminal A tumors and vision enrichment is even stronger when we look specifically at hotspot mutation in P53CA and GATA3. We also saw that there is a stronger association between MAFRIK1 mutation and wide arm event on chromosome 8 and chromosome 16 that characterize the cluster here in the middle that we call the chromosome 8 associated clusters. As you may have noticed several of these genes participate in similar signaling cascade so the obvious question was which pathway are most deregulated in luminal A tumors and in the luminal A subtypes. We did a pathway analysis using MIMO, MIMO is a method that look for micro pathways characterized by genomic alteration that occurs in a mutually exclusive way. I indicated here the link, the method is available at our website for downloading use. The first set of results of MIMO nicely confirm what we already observed in the bigger breast cancer data set which are multiple genomic alteration targeting AKT, MAP kinase and ROS-ERC signaling. What we didn't find then and we could find now by focusing specifically on a single subtype was a set of multiple rare but mutually exclusive alteration affecting components of core oppressor complexes and core and smart. You can see here several mutations this time to be inactivating mutation and generally correlate with down regulation of a core respective genes. But why are core oppressor complexes relevant or important in ER positive luminal tumors? Well in these tumors proliferation is mostly driven by the estrogen receptor and ER activity can be blocked therapeutically using an estrogen antagonist called tamoxifen which recruits core oppressor complexes to block ER transcriptional program. However it has been shown in breast cancer cell that loss of core oppressor complexes may actually turn tamoxifen into a promoter of ER driven proliferation by recruiting coactivator complexes instead. So in this scenario what we have now is a set of genomic alteration which are directly linked to loss of core oppressor complexes and may therefore predict lack of response to endocrine therapy. So in conclusion we dissected the genomics of luminal tumors in multiple data set to explain the molecular and clinical heterogeneity. We identified five major subtypes actually of luminal tumors which are characterized by distinct copy number alteration and semi-reduced mutation and in particular we found an atypical luminal A subtype characterized by hygienomic instability, p53 mutation, oral kinase subregulation and associated with the poor prognosis. Furthermore we show you that luminal A hallmark mutation specifically PIC-3CA anger free are prevalent in tumors characterized by low level of copy number alteration and we also identified by pathway analysis multiple rare but mutually exclusive alteration associated with loss of core oppressor complexes and this alteration as I just told you are important as they may predict lack of response to tamoxifen treatment. And with this I'd like to conclude and thanks to all the people that contributed and are still contributing to this project and in particular Chris, Nikki and Raylene at CBIO and Chuck and Katie at UNC and of course the whole TCGA breast cancer analysis working group for all the analysis done during the breast cancer project. Thank you. Any questions? I of course always have one. I wonder if we could look at those core oppressor complexes in the in the Wash U data from their paper where they had endocrine responsiveness. Did we look in there? Did we see any of those? I didn't look at the response I look at the mutations and yeah in that I don't have a selection. I know those were rare mutations. In the fingerprint I showed only mutation according the TCGA data set but we found mutation in these components for short encore one. I think one mutation in encore two and GPS two also in their data set. So it's something which is what one thing that I didn't say is that this mutation occurred almost exclusively in luminal A2 more so highly enriched in this subtype. This is Angel from Harvard. Do you see any deletions in the ER receptor? Deletion? Yes. A recurrent deletion now and this is characterized by wide arm events like wide deletion typically occurring in 16 Q if I remember correctly 8P. But you didn't see any focal deletions in that region? Deletions of ER itself. I don't think so. Thank you Giovanni. So our next speaker