Rna - Seq, exome - Seq uncover drug targets, biomarkers using ion torrent™ technology

Feeling a little bit excluded - I didnít get the musical introduction. And Rob, thank you for the promotion, by the way - Iím actually an Assistant Professor, but I appreciate it very much. So I want to thank you guys, and I want to thank the Life Technologies and the organizers for letting me talk to you today on our work using next generation sequencing - in particular, Ion Proton™ Sequencing, using RNA-Seq™ and Exome-Seq™ in translational oncology research. So every time I give a talk to a varied audience like this, I always like to start off with a very short introduction to translational cancer genomics. I think a lot of us in the cancer world are so cancer centric we forget that there's a whole genomics world out there. So I'm going to give a little quick, few intro slides. And I'm going to talk about our work, particularly in sequencing in triple negative breast cancer, which is our particular specialty; a little bit about how we're applying sequencing to clinical research studies; and then a little bit on some new analysis tools and applications. One thing I like to always show, and the reason why we do genomic sequencing in cancer and probably no surprise to many of the people in this room - is that cancer is a disease of the DNA. We know that the causation of cancer happens at the genomic level. So if we look - this is a great figure by Michael Stratton, published in Nature a few years back. And if we start off here in the fertilized egg, which we were all at one point but you may not remember. We went on, as the cells divided, to accumulate mutations in our genomes. Some of these mutations were accumulated via intrinsic mutation processes, meaning our DNA replication machinery is fantastic at replicating our DNA but itís not perfect, and mutations are introduced to the genome via these replication processes. We also get exposed to a variety of environmental lifestyle exposures - smoking, sunlight, different carcinogens, and these also introduce mutations in the DNA. But many of these mutations, as delineated by these circles, are what we call passenger mutations - meaning they get introduced into the genome, but most likely have no effect on function. But unfortunately, we do begin to develop these star mutations, what we call driver mutations. These are mutations in tumor suppressors and oncogenes that eventually cause the malignant transformation of normal cells, such that normal cells that we have present in adulthood become early clonal expansions, benign tumors, early invasive, late invasive cancer. And obviously, what weíre most concerned about is chemotherapy resistant disease. We also know that these changes at the DNA level are translated into changes at the RNA level. And here is just some example data of a principle components analysis of 10 normal samples and 10 tumor samples. And for those of you who are familiar with principle components analysis, it is just a three-dimensional representation of how alike or dis-alike samples are, based on their global gene expression - such that if samples are more closely tied together, theyíre more alike; if theyíre more separated on the 3-D space, theyíre more dis-alike. And what we know from a variety of sequencing efforts, that when we compare a tumor to normal, we have a disregulation of an expression of the thousands of protein coding genes in the transformation of normal cells to tumor cells. Thankfully, our government got involved and realized that these changes in the DNA and the RNA are very important to understanding the encyclopedia of what cancer is and what it comprises. So as many of you are already familiar with, the National Cancer Institute initiated the Cancer Genome Atlas, which is an initiative to sequence 10,000 tumors and 24 tumor types by 2015, using a variety of modalities like DNA sequencing, RNA sequencing, micro-RNA, SNP protein arrays, and methylation assays on tumor normal pairs. The main effort of this - the main goal of this effort is to create a comprehensive public catalog of all significantly recurrent genomic alterations. And I bring up the TCGA here, as I was going to show you some data later of how we've leveraged the combination of Ion sequencing with TCGA data to do discovery research. So why is this important? I think it's very cool that we could talk about technologies that can interrogate exomes and genomes, and RNA sequencing. But I think the important part is that we take this to the human level. What we know is that this combination of DNA mutations and RNA disregulation is simply lethal. Patients die from this disease due to these mutations and these RNA disregulation. Another thing that I think is very sobering is that the treatment for cancer in the last 30 years has remained largely unchanged. When you look at current FDA approved first line therapy for a variety of disease types, and also in the adjuvant setting, the most common chemotherapeutics are paclotaxel, cisplatin, and cyclophosphamide. But the unfortunate reality is that these drugs were discovered in the 1950s and 60s, yet remain mainstay therapy to today. Also another sobering fact is that while drugs that specifically target cancer cells are under development, and - but under development, but targeted therapies that drastically extend survival will have been few and far between. That's not to say that we haven't had blockbusters like Herceptin, and Gleevec and Crizotinibs of the world. But when we think about many of the more common tumor types, targeted therapies are still quite under development, and we need new targeted therapies to drastically extend the survival. So what do I think are some of the major questions in cancer research? Obviously, what are the mutated genes in our pathways that are critical for the development of common cancers? What are the key drug targets for the development of novel therapeutics? What molecular changes cause that chemo-resistant disease that I mentioned to you? Are there better biomarkers for early detection of cancer? Are there better biomarkers for predicting prognosis and treatment outcomes? And then I think lastly, the question for this group is can cancer sequencing help answer these questions? So what I'd like to show you here very quickly is what we do as our clinical translational research pipeline at Indiana University. Many people like to embellish their labs at the end of the talk. I like to embellish them in the middle, because they're the work horses, just the phenomenal group that I work with. The majority of our studies are in triple negative breast cancer. We apply Ion Proton™ Next Generation Sequencing, which is led by my Genomics Director, Brad Hancock. We take that data and use high end bioinformatics. IU is very lucky to actually now host the fastest owned, university owned super computer in the world, Big Red II, which is a Cray system. We do a ton of our analyses on these supercomputers and this is led by Rutuja Atale, who is my Bioinformatics Director. We then take that data and take it into the lab, both in vitro and in vivo to do biomarker and drug discovery efforts. This is led by Jeff Solzak, my Lab Director, whoís actually in the audience today. So this is only a slide that I would actually show at an Ion meeting. Last year at Ion World 2012, it seemed very popular by many groups to kind of show off the names of their Protons™ and their themes. And I think last year we saw some Star Wars Protons™, variety themes. So when we got our Protons™ we thought it would be really, really important, that we need to come up with a good theme with our machines. Now, Iím from Indianapolis, but I was born and raised in Chicago. So is my Lab Director. And we thought the best theme should probably be a winning team. And we thought about it, it became very clear that our theme should be the 1985 Chicago Bears. So I'm very excited to introduce our lab Protons™. This is Proton™ Ditka. And Proton Singletary. And here's the little picture of our machine, and also the Torrent Suite™. A lot of questions - a bit question I get every time I present this is, 'Why not Walter Payton?' We didn't want him to get mixed up with a certain Payton Manning, who used to be a Colt. So what I'm going to do now is talk to you a little bit about triple negative breast cancer and how we've applied Next Generation Sequencing to study this disease. Some of you may be familiar with the disease, but those who are not, I'm going to give you a quick intro. When we have a breast cancer patient come into the clinic, every breast cancer patient is going to have a biopsy taken of their primary tumor. That biopsy is sent to the pathology laboratory to have a variety of stains done to it. About 75% of breast cancers are going to overexpress the estrogen receptor and progesterone receptor. And we know patients who overexpress the ER-PR, in addition to chemotherapy, are going to receive drugs that target the estrogens receptor - drugs like tamoxifen, fulvestrant, and aromatase inhibitors. Another 15% or so are going to be overexpressing or over-amplifying the HER2 gene, which is a cell surface receptor of tyrosine kinase. And in addition to chemotherapy, these patients receive drugs that target HER2, drugs like the antibody Herceptin, or the small molecule Lapatinib, and some more recent drugs that came on the market, T-DM1 and pertuzumab. But about another 15% or so of breast cancer patients are what we call triple negative. These are patients who are negative for the ER, negative for the PR, and negative for the HER2 gene. And these patients, in addition to chemotherapy, there are no targeted therapies. What this is translated is actually, unfortunately, to a poor overall survival for these patients. So this is a great publication in 2007 showing the overall survival Kaplan-Meier curve of all other breast cancer versus triple negative breast cancers. And as you can see, triple negative breast cancer diagnosis entails a much poorer prognosis. This makes up, again, 10 to 15% of all breast cancer, with a disproportionately higher mortality. This is a disease that tends, or has a higher proportion of pre-menopausal women and women of African descent. Interesting also that the majority of BRCA1 breast cancers are triple negative - meaning if you are a person who is a germline carrier of a BRCA1 mutation, if you develop breast cancer almost the vast majority are triple negative, in addition to being at high risk for ovarian disease. As I mentioned, a paucity of effective targeted therapies. And I think as scientists we always have to ask ourselves, 'Where can we look to better define this disease, but also to find new therapeutic targets.' So the basis of my talk is actually based on some previous data we had generated using triple negative breast cancer and also micro-dissected normal breast tissues. So in this study we took 20 micro-dissected normal breast tissues from healthy volunteers from our Susan G. Komen for the Cure Tissue Bank; 10 triple negative breast cancers; and then combined the data with Cancer Genome Atlas data of 84 triple negative breast cancers, and 10 adjacent to tumor normal analysis, and a combined RNA sequencing data analysis; and then analyzed the data using a variety of internal software from our Center for Computational Biology, and also from Life Technologies, Partek, and Ingenuity. One powerful part of our study was actually our normal control. The Susan G. Komen for the Cure Tissue Bank at IU is home to 1500 breast, normal breast, frozen samples with coupled DNA serum, full medical history, and it's just an amazing, amazing resource that women with no disease come and actually donate a portion, a 16-gauge core biopsy of the normal breast to provide for research. So when we were comparing the RNA sequence from our triple negative breast cancers, we wanted to compare it to an enriched population of normal ductal epithelium in the normal breast. As is well known, the majority of the normal breast is made up of stroma, fibrous tissue, adipose tissue. But the origin of breast cancer are actually the ductal epithelium, the milk ducts. And so we went through quite a bit of time to laboriously micro-dissect the milk ducts. What's also, I think a big shocker to a lot of people is that while we've had a huge amount of investment in cancer research, in genomics, millions of dollars, including myself, if you search the literature, what's known about the normal is very little. We know a lot about cancer but we actually donít know a lot about the normal tissue. And unfortunately again, I don't have time to present it today, but in two papers that will coming out very shortly, we detail, using RNA sequencing, the biology of the normal breast. What we saw when we went into principle components analysis was very interesting. Here is the PCA of the data set. Here, the blue, are those micro-dissected normal tissues. The green are those adjacent to tumor normal tissues that are being used as the control for TCGA. And here in purple are the TCGA triple negs and the red are the triple negatives from Indiana University. And what you can see, as one would expect, the wide swatch in difference in the transcriptomes, differentiating normal tissue from triple negative breast cancer. Again, about the number of genes that differentiate these two groups, roughly 3,000 disregulated protein coding genes differentiating normal from triple negative breast cancer. Now, one question I get quite a bit is, 'Well, Milan, you know, why is that exciting? We've been doing microarrays on breast cancer since 2000.' Actually, one of the first microarray studies were done in breast cancer. And when you look at those studies, what they did was they performed a gene expression microarray on breast cancers and then were able to separate breast cancers into five, into actually four major intrinsic sub-types - Luminal A and Luminal B, which are the ER positive disease; HER2, again which is akin to the HER2; Basal-like, which is primarily triple negative; and then this normal breast group. But when you look into many of the details of these previous studies, the vast majority of normal breast tissues used in these studies are actually what we call reductional mammoplasties. These are breast reductions, which are actually comprised mostly of adipose tissues. So when you look at the genes that underlie the normal control, what you have is actually genes involved in adipose. So when the comparisons were done to find new, novel therapeutic targets, many cases are tumor versus tumor comparisons - it would compare one sub-type of breast cancer to another sub-type of breast cancer, or maybe a benign disease to identify a target. And obviously, I think people in this room can appreciate sort of the problematic issues with doing a tumor versus tumor comparison. You have this increased likelihood of passenger findings, and you're not really sure if those differences are actually due to the tumor, or just the use of your control. But one can postulate that when you compare a tumor to a normal sample, by definition all differences have the potential to be important. So why is that important? So I mentioned to you in my introduction that there are no FDA approved targeted therapies for triple negative breast cancer. But it's not because people haven't tried. So what I'm showing you here is a conglomerate of seven randomized Phase II clinical trials testing targeted agents in triple negative breast cancer. What I'm going to give you is the rationale for testing these targets, based on tumor versus tumor data; what the Next Gen transcriptome tumor versus normal data demonstrates; and then also give you the clinical trial outcome. Two inhibitors that were popular early on, cetuximab and gefitinib, which targeted the cell surface receptor tyrosine kinase EGFR, based on prior data, the EGFR is over-expressed in triple negative breast cancer. Interestingly, when you look at normal breast tissue, both by Next Gen, QPCR and CLIA IHC, EGFR is very well expressed in the normal breast. Actually, the amount of variability you see in a normal breast is about the amount of variability you see in triple negative breast tumors. And all two, I think three now, clinical trials have been negative. We look at the target C-Kit, again another important oncogene, based on data - I'm sorry, using the drugs imatinib and dasatinib, which target the CKId, based on prior data that C-Kit is over-expressed in triple negative breast cancer. When you look at the normal breast, you don't find the over-express. You actually find, actually C-Kit to be down-regulated in triple negative breast cancer significantly. The normal breast expresses high amounts of the oncogene C-Kit. And actually, two clinical trials, both presented at San Antonio, were negative. Then we've got the newer PARP inhibitors, drugs like Iniparib, olaparib, and rucaparib. Based on the data that PARP is over-expressed in triple negative; also based on a synthetic lethality hypothesis, and especially with the BRCA carriers in triple negative disease. We find this gene to be overexpressed in four-fold, and currently it's the only clinical trials that show any positivity in triple negative. Admittedly, itís been a little bit sporadic. It seems like a lot of the benefit has been restricted to BRCA carriers, and also there's been some question into the potency of the earlier drugs. But some of the newer drugs are now in clinical trial and we're very excited to see the results. But it also begs the question - if these are not the targets, then really what is, and what can we use - can we use the data we have from RNA sequencing to identify those targets. One of the targets that we see, to no surprise, is high over-expression of the PI3K, AKT pathway. I think a lot of the people in this room are familiar with this pathway. It's a pathway involved in tumor cell survival and proliferation. And everywhere in this pathway analysis where we see a red gene, we see high over-expression of this component of the PI3K, AKT pathway. This led us to do a study where we took a large panel of triple negative breast cancer cell lines, and then tested eight different PI3K, AKT or dual PI3K, mTOR inhibitors that are either in Phase II or Phase III clinical trial. What was interesting was that we saw a really great sensitivity of triple negative breast cancer cell lines to these inhibitors. I think one thing that's popular to do is that people will show the representative figure. I'm actually showing you the most resistant cell line in all the figures I'm going to show you now. So this is the most triple neg, was MDA-MB231, and actually we saw a very nice sensitivity to these inhibitors, using the PI3K, AKT inhibitors in these triple negative breast cancer cell lines. But what I think is more important in cancer research is we have to get away from the idea that if we hit one target we're going to cure the disease. It doesn't happen that way. We know many times, when you hit one target, the disease becomes resistant. So is there - can we use RNA sequencing to identify a rational drug combination. And I'm going to show you an example of one that we're working on actively. When we went back to the data through a variety of sleuthing - I don't have time to give details in - but we saw a very high expression in the Wnt pathway, in particular Wnt involved in epithelial development, as again the normal breast duct, there's an epithelial cell, so we're very excited to see this. What was further interesting is that we were able to find a Wnt inhibitor in the literature, as an inhibitor both Wnt/B-Catenin and PPAR-alpha and gamma. What was more further interesting is that when we go back to that original resistant cell line, and we did this across those cell lines, that you see amazing synergy when you take two - both drugs together. So here is that PI3K inhibitor, given that log dose. Here is the Wnt inhibitor, FH535 - given at log dose concentration. But when you put the drugs together, you see a massive synergy. And actually, we saw a low nanomolar] IC-50s when you put the drugs together across cell lines. So - but the question is, well, why are these drugs so synergistic? Even when we hold one drug at a constant dose of 10nM, which is a very small amount, and then give very small amounts of the Wnt inhibitor. So we went to Ion Proton™ sequencing to answer that question. What we then did was we took that triple negative breast cancer cell line and then dosed the cells with those PI3K, AKT inhibitors, and then did Ion Proton™ RNA sequencing before and after. And what we noticed, in blue, is the pre-treatment and the red is the post-treatment, that when you treated with that PI3K, AKT inhibitor, you actually induced the expression of the targets of the Wnt inhibitor. So it's very akin to what we've known in cancer research for quite a while, which is what I like to call the wack-a-mole affect. You have one target, you hit the mole on one, the second one pops up. When if you try to hit both targets, you begin to see a highly synergistic effect. And weíve been very, actually very excited about this Wnt pathway induced PI3-K - sorry, excuse me - PI3-K induced Wnt expression, and we're actually right now testing this combination in patient derived mouse xenograft models that are being characterized, fully genome characterized by Ion exome and RNA sequencing to run 'mini' clinical trials, in which we're able to take a variety of mice implanted with triple negative breast tumors that have been genomically characterized, so that we can not only test their efficacy, but also have the genomic data to test differential response. So knowing in vivo prior to going to a human clinical trial, whatís the predictors of which patients will respond and which will not. I'd like to also, as I mentioned, my talk is about translational clinical research, I wanted to talk a little bit about how we've also implemented Next Gen sequencing into clinical research studies. So the data I've shown you before is trying to find new drugs. But the question is, can we use sequencing to use current drugs better. Out of the variety of clinical trials that we are sequencing, I really wanted to present this one. This is a personal favorite of mine. This is a randomized Phase II called BRE09-146. This is a Phase II clinical trial being led by IU, but has finished enrolling patients at 17 different sites. This is a very, very innovative clinical trial. What this does is it takes patients who receive neo adjuvant chemotherapy, meaning chemotherapy prior to surgery. And when these patients who receive up front chemotherapy go to surgery, thereís two responses. One is called a pathologic complete response, meaning the patient goes to surgery, you take out the tissue where the tumor was, you go under the microscope and thereís nothing there. Those patients do really, really well. But thereís about 60% or so patients that after the chemotherapy they go to surgery, they take out the tissue, and they go under the microscope and thereís still tumor present. And in particular, when they have what we call significant residual disease, meaning tumor about two centimeters or more, these patients do very, very poorly. Unfortunately, standard of care for this population, for patients who have a significant residual disease after up front chemotherapy, is actually - is watch and wait, or physician choice. There's actually no standard therapy for this population. So this trial is very innovative for two reasons. One, it actually randomizes patients. Two, it's chemotherapy - in this case, this cisplatin for four cycles, or cisplatin plus Rucaparib, a PARP inhibitor for four cycles, followed by maintenance therapy for 24 weeks. Again, 135 patients; just finished completing enrollment in early June, with primary end points of one year and two year disease free survival. What is really cool and what's very interesting for us that we're able to get specimens from this trial. From this entire trial, our labs received pre-treatment biopsies, prior neo adjuvant chemotherapy, specimens at surgery, also normal blood and plasma to do the liquid biopsies that you heard Dr. Topol talk about earlier today - all to do correlative genomic sciences. All this work is now being sequenced on the Ion Proton™ Sequencer, using a comprehensive genomic approach of whole exome sequencing for germline and somatic mutations; RNA sequencing for disregulated gene expression; and then also plasma DNA sequencing to detect circulating tumor DNA. This data is being generated and analyzed for presentation to ASCO. But we've learned a lot during the way - and I actually went too fast here. So our correlative science goals for this study is to one, determine the mutational express changes in the tumor that mediate resistance after neo adjuvant chemotherapy. So what causes the patient to have resistant disease in the first place. Determine gene expression signatures and mutated pathways to predict response to the combination arm. Detect germline variations as biomarkers for treatment response. I actually really appreciate the comments made earlier on the importance of pharmacogenomics, and we're going to play a large role in analyzing this data for pharmacogenomics. And then also to assess circulating tumor DNA as an accurate biomarker for relapse. As one of the lessons that we learned along the way in clinical trial sequencing, and I think that some in this room could appreciate, is that samples from clinical trials are diverse and challenging. As many of you know, these clinical trials are multi-site, could be 10, 20, 100 sites, many of them internationally, from either a tertiary medical center, or maybe from a small private practice from Honeyburg, Indiana. You get what you get. The samples you get are usually very diverse in quality, and also tumor cellularity. And I think many people in this room can appreciate samples like this where you have plenty of tumor; samples like this where you have a ton of lymphocytic infiltration in your micro-dissection; and samples like this where you have very little tumor. But youíre expected, and we have to work with these and try to make them - be successful with all these - with these sorts of samples. So one of the things that we tested early on with Ion Proton™ sequencing was the feasibility of low input paraffin sequencing. We spent quite a bit of time knowing that much of our sequencing was going to be done on paraffin - especially with those low input amounts - 100 nanogram, 50 nanogram inputs. And here, we were very lucky to have an FFPE sample - this is just some example data that we generated, where we tested a thousand nanogram RNA library versus a 15 nanogram RNA library. And we were very happy to see that with the Ion Total RNA-Seq™ kit and Ion Proton™ sequencing, we actually get a good correlation of gene expression values when you compare the two libraries. Albeit it, we haven't done a full test where weíve taken samples that say, that are seven years old and been sectioned, you know, a four micron section. Many of this - many of our - we are very lucky. At least, with many of our samples, we will cut off a fresh block, albeit, though still paraffin and still degraded - when you look at the RIN score, still very low. So we're very happy to see a good correlation in our amount of inputs. Another thing that - very recently been working with, with Matt Dyer and Mike Lelivelt, and Carl- Carl Dowds is to also beta test the Ion Reporter™ 4.0 software, which I believe will be released next month. One of the really neat things about the new Ion Reporter™ 4.0 is its ability not only to do SNPs and indels, but also do copy number variation. That same sample I mentioned to you that we tested the high input, low input RNA-Seq, we actually performed tumor and normal exome sequencing also on that same sample, and then tested the pipeline. And what was very cool was the pipeline was able to identify a copy number variation in our tumor sample. This is an example of one of the - I think, either the top hit or one of the top hits of a gene called SYBU. Itís a syntaxin gene. And I apologize - if you were to mouse over the landscape, you would see that there was roughly 400 or so mapped reads to the tumor sample, but only about a hundred or so mapped reads to the normal sample - giving data to support a true potential amplification event. Now, me personally, in all my years of cancer research, I had never heard of the SYBU gene. This was new to me when I saw this, and I'm actually kind of glad to show an example that isn't your typical P53, P10 or PI3K. But what was cool was when I ran this gene through the Memorial Sloan-Kettering c-Bio platform, which is a very neat bio-informatics platform to do Pan-cancer, TCG analysis. Everywhere you see in red is Reporter amplification of this gene across these 25 cancer types. And we actually saw amplification of the SYBU gene across many of the cancer types, in particular bladder, breast, and ovarian cancer - again, giving some more credence to evidence of a true amplification event. So with that, I'm just really lucky to work with a great group of people. All the work you've seen is just not possible without an awesome team, my laboratory, Jeff Solzak, Rutuja Atale, and Brad Hancock, along with my clinical collaborators, Drs. Schneider, Sledge, Clare, Loehrer, Ivan, Nephew, Liu and Badve - the wonderful people at the Susan G. Komen For the Cure Tissue Bank. And I really, really want to highlight just some great collaborations with Life Technologies. I'd like to even say, the unsung heroes who've helped make a lot of our sequencing possible, particularly Eric Hilligoss, Matt Hickenbotham and Jane Tueche, who've made a lot of the work we did possible. And then also many collaborators - today and also in the past who we've worked with in the bioinformatics - Fiona and Darryl and Mark, and more recently, Matt Dyer, Carl Dowds and Mike Lelivelt. So with that, I - oh, last thing - if you're interested in what we do, my lab maintains a Facebook page, so feel free to go to our Facebook page and see a little bit more about what we're doing. So with that, I thank you for your attention.